\

r
McGraw-HIII Series In Water Resources
1, and Envlronmental Englneerlng SURFACE
Consulting Editor
George TchoballOglous
WATER-QUALITV
Bailey and Ollis: Bioehenrical EngineerillR Fundanrenlals
MODELING
Bouwer: Gml/lldWOler HYllmlogy
Canter: Ellvil"lJl/l//l'lIlal ¡nrpael Assessnrenl
Chanle": EllI'ÍI1J1//II"IIwl PmlecliOlI
ChUIJra: Sur/aa lValer-Qualily Modelillg
Chuw. Maldment. and Mays: Applied Hydmlogy
Du,'is and Corn\\ell: l/llroduclion 10 Ellvirolllllelllal Engilleerillg Steven C. Chapra
Eckenrelder: 11It/lISlrial Waler Po/lulion COlllrol Ulliver.<iIY nI C%radn al Hall/der
J,IIGn'~a, Buekin~ham, and Evans: Hawmous Wasle Managenrenl
Unslcy. Franzini. Freyberg. and Tehohanoglous: Waler ReJO/trr:es alld Engilleering
Md;hee: Waler Supply and Sewerage
l\Iays and Thng: Hydro.~ysle/lls EIIgilleerillg and Mallagel//l'III
I
I\Ictealf & Elldy, Ine.: Waslewaler Engineerillg: Co/lecliOlI "lid PUlllpillg of Waslewaler
I
Metealf & Eddy, Ine.: Waslewaler Engilleering: Trealnrelll, DÜposal, Reuse
Pcavy. Rowe, and Tehobanoglous: Ellviro/llllelltal EIIgilleerillg
Sawycr and McCarty: ChenrÜ/ry for Envirollnrenlal EIIgilleering I
Tchohanoglolls. Theisen, and Eliassen: Solid Wasles: Engilleering Prillciples and
Mti/lagelllerll IHlle.~ J

Tchobanoglolls. Theisen, and Vigil: hllegraled Solid Wa.51Managenrelll: EIIgineering

" 1.1
Pri"ci,,'es alld MmltlRelllelll IUlles
!

THE McGRAW-HILL COMPANIES, INC.

New York SI. Lo,,;s San Francisco Auckland Bogotá Caracas
f Lisbon London Madrid Mexico Milan Monlreal New Dclhi
I San Juan Singapore Sydney Tokyo Toronlo
I
I

'. ~
 "

,.,
McGraw-Hill
A Division of'rheMcGraw.HiUo,mpanies
'iZ
To Cynlhia

SURFACEWATER-QUALlTVMODELlNG

Copyright @ 1997 by Ihe MeGraw-Hill Companies, Ine. AII rights reserved. Printed in Ihe United
States of Ameriea. Exeept as permiued under Ihe United Slales Copyright Aet of 1976, no par! of
Ihis publieation may be reprodueed or dis'ributed in any forro or by any means, or stored in a data
base or retrieval system, wilhout the pr'o; writlen permission of the publisher.
This book is printed on aeid-free paper.

1 2 3 4 5 6 7 8 9 O DOC DOC 9 O ~18 7 6

ISBN 0-07-011364-5
,

\
I
-::

This book was set in TImes Roman by Publiea/ioll Servia.I', lile.

The edi/ors were B. l. Clark, Dal'id A. Dams/ra, alld lames W. Bradley.
The produe/ion supervisor was Denise L Puryear.
The eover was designed by Karen K. Quigley.
Projeet supervision was done by Publieation Selvice>', Ine.
R. R. Donnelley & Sons CompallY wa,' primer and billder.

Chapra, Steven C.
Surfaee water-quality modeling I Steven C. Chapra.
p. cm.
IncIudes bibliographieal referenees and index.
ISBN 0-07 -O11364-5-ISBN 0-07-843306-1
l. Waler quality-Malhematieal models. l. Title.
TD365.C48 1997
628.1'61 '015118-de20 96- 15461
I
)
http://www.mhcollege.com

flJ
 About the Author CONTENTS
_".,..w' _

Preface
Steve Chapra teaches in the Civil, Environmental. and Architectural Engineering xvii
Dcpartment at lhe University of Colorado. His other books inelude Numerical Me/h- PART I
Completely Mixed Systems
ods for EII!?i//eersand ¡II/mdl/c/ioll/O Compu/b'Rfor ElIgilleers.
Dr. Chapra received cngineering degrees from Manhattan College and the
University of Michigan. Before joining the faculty at Colorado, he worked for the YÍ.ECTURE 1
Environmental Protection Agency and the National Oceanic and Atmospheric Ad- 1.1
3
ministration and was an Associate Professor at Texas A&M. His general research in- Engineers and Water Quality 4
1.2 Fundamental Quantities
telests focuson surface water-quality modeling and advanced computer applications 6
in environmental engineering. His research has been used in a number of decision- 1.3 Mathematical Models 10
1.4
I making contexts, ineluding the 1978Oreat Lakes Water Quality Agreement.
He is active in several professional societies and has taught over 40 workshops
on water-quality modeling in the United States, Mexico, Europe, and South Amer-
1.5
Historical Development of Water-QualityModels
Overview of This Book
14
19
Problems
ica. He has received a number of awards for his scholarly contributions, ineluding 20
the 1993 RudolphHering Medal (ASCE) and the 1987Meriam-Wiley Distinguished LECTURE 2 ReactlonKlnetlcs
Author Award (American Society for Engineering Education). He has also been rec- 24
2.1 Reaction Fundamentals
ognizcd as the outstanding teacher among the engineering faculties at both Texas 24
A&M (1986 Tenneco Award) and the University of Colorado (1992 Hutchinson 2.2 Analysis of Rate Data
29
Award). 2.3 Stoichiometry
38
Dr. Chapra was originally drawn to environmental engineering because of his 2.4 Temperature Effects
40
love of the outdoors.He is an avid fly fisherman and hiker. His primary professional Problems
goal is to apply engineering, mathematics, and computing to maintain a high-quality 42
cnvironment in a wisc and cost-efficient fashion. VLECTURE 3
MassBalance,Steady-StateSolutlon,and ResponseTime 47
3.1
f ,; Mass BaJance for a Well-Mixed Lake 47
3.2 S!.eady-StateSolutions
52
3.3
Temporal Aspects of Pollutant Reduction 57
Problems
62
-(LECTURE4 ParticularSolutlons
65
4.1
Impulse Loading (Spill) 66
4.2
Step Loading (New Continuous Source) 68
4.3
Linear ("Ramp") Loading 70
4.4
Exponential Loading 71
4.5 Sinusoidal Loading 73
4.6
The Total Solution: Linearity and Time Shifts 76
4.7
Fourier Series (Advanced Topic) 80
Problems
83
I LECTURE5 Feedforward Systems of Reactors
86
5.1
MassBalanceandSteady-State 86
5.2 Time Variable
91

tI vii

-... ...1
 ---
'~..~_._~--".

viii CONTENTS
CONTENTS ix

5.3 Feedforward Reactions 95 10.4 Tracer Studies 186

Problems 99
10.5 Estuary Number 189
Problems 190
{ECTURE 6 Feedback Systemsof Reaetors 101
6.1 Steady-State for Two Reactors 101 LECTURE 11
Control-Volume Approaeh: Steady-State Solutlons 192
6.2 Solving Large Systems of Reactors 103 Il.I Control- Volumc Approach 192
6.3 Steady-State System Rcsponse Matrix 107 11.2 Boundary Conditions
194
6.4 Time-Variable Response for Two Reactors 111 1l.3 Steady-State Solution 195
6.5 Reactions with Feedback 113 IIA System Response Matrix 197
Problems 117 11.5
Centered-Difference Approach J98
11.6
Numerical Dispersion, Positivity, and Segmcnt Size 201
120 11.7
( L~)computer Methods:Well-Mlxed
Reaetors Segmentation Around Point Sources 207
7.1 Euler'sMethod 121 11.8 ;
Two- and Three-Dimensional Systems 208
7.2 Heun's Method 124 Problems 209
7.3 Runge-Kutta Methods 126
7.4 Systems of Equations 128 LECTURE12 Simple Tlme-Varlable Solutlons 212
Problems 131 12.1 An Explicit Algorithm 212
12.2 Stability
214
12.3 The Control- Volume Approach
135 215 I
PART 11 Incompletely Mixed Systems
- 12.4 Numerical Dispersion
Problems
216
221
t,
V LECTURE 8 Dlffuslon 137 I
,.8
LECTURE13 Advaneed Tlme-VarlableSolutlons
8.1 Advection and Diffusion 137 223 \[

8.2 Experiment 138 13.1 Implicit Approa( 223 t

141 13.2 The MacCormack Method 229
8.3 Fick's First Law
13.3 Summary
1r
8.4 Embaymcnl Mndel 143 230
149 Problems 232
8.5 AddilioB.11 Tr,mspon Mechanisms
Problems 153
PART IJI Water-Quality Environments I
156 2JJ
V LECTURE
9 DlstrlbutedSystems(Steady-State)
9.1 Ideal Reactors 156 II
9.2 Application of the PFR Model to Streams 164 LECTURE14 Rlvers and Streams
:: ::¡
9.3 Application of the MFR Model to Estuaries 168 14.1 River Types . 1'\
Problems 171 14.2 Stream Hydrogeometry
I 2.i.

173
, 14.3 Low-Flow Analysis
243
ILECTURE
10 Dlstrlbuted Systema(TIme-Varlable)
173
I 14.4 Dispersion and Mixing 245
10.1 Plug Aow : 14.5 Flow, Depth, and Velocity
10.2 Random (or "Dnmkard's") Walk 177 14.6
Routing and Water Quality (Advanced Topic)
247
250
1
10.3 Spill Models 180
, f
Problems 257

I
f J.
~-~
 x CONTENTS
CONTENTS xi
. LECTURE15 Estuarles 260 19.6
BOD Loadings, Concentrations, and Rates 357
15.1 Estuary Transport 260 19.7
Henry's Law and the Ideal Gas Law 360
15.2 Net Estuarine Flow 262 19.8
Dissolved Oxygen Saturation 361
15.3 Estuary Dispersion Coefficient 263 Problems
270 365
1<;.4 Vertical Stratification
Prohlems 272 VLECTURE20
11. r Gas Transfer and Oxygen Reaeratlon 367
20.1 Gas Transfer Theories
276 369
LECTURE 16 lakes and Impoundments 20.2 Oxygen Reaeration
276 376
16.1 Standing Waters 20.3 Reaeration Formulas
278 377
16.2 Lake Morphometry 20.4
Measurement of Reaeration with Tracers 384
16.3 Water Balance 282 Problems
287 386
16.4 Near-Shore Models (Advanced Topic)
., Problems 293 V' LECTURE
21 Streeter-Phelps: Polnt Sources
389
21.1 Experiment
295 389
0ECTURE 17 Sedlments 21.2
Point-Source Strceter-Phelps Equation 391
17.1 Sediment Transport Overview 295 21.3
297 De/kit Balance at the Discharge Point 391
17.2 Suspended Solids 21.4 Multiple Point Sources
302 393
17.3 The BoUomSediments 21.5
Analysis of the Streeter-Phelps Model 396
17.4 Simple Solids Budgets 304 21.6 Calibration
17.5 Bottom Sediments as a Distributed System 307
312
.
21.7 Anaerobic Condition
398
399
17.6 Resuspension (Advanced Topic) 21.8
Estuary Streeter-Phelps 401
Problems 315 Problems
403

r._'.. LECTURE18 The "Modellng" Envlronment 317 LECTURE 22

Streeter-Phelps: D/strlbuted Sources
317 405
18.1 The Water-Quality-Modeling Process 22.1
Parameterization of Distributed Sources 405
18.2 Modcl Scnsitivity 327 22.2 No-Flow Sources
335 407
18.3 Assessing Model Performance 22.3 Diffuse Sources with Flow
339 4/0
18.4 Segmentation and Model Resolution Problems
341 417
Problems
LECTURE 23 N/trogen
419
23.1
PART IV SSOlved Oxygen and Pathogens 345 Nitrogen and Water Quality 419
23.2 Nitrification ,
421
23.3 Nitrogenous BOD Model
424
23.4 Modeling Nitrification
V LECTURE 19 SOD and Oxygen Saturatlon 347
23.5
426
19.1 The Organic ProductionIDecompositionCyc1e 347 Nitrification and Organic Decomposition 428
348 23.6
19.2 The Dissolved Oxygen Sag Nitrate and Ammonia Toxicity 430
351 Prohlems
19.3 Experiment 432
19.4 Biochemical Oxygen Demand 353
LECTURE 24
19.5 BOD Model for a Stream 355 I PhotosyntheslslResplratlon 433
24.1 Fundamentals
433

'..;,JJ,...
... ./
 xiv CONTENTS I CONTENTS xv
629
34.4 NutrientIFood-Chain Interactions
Problems
631 I PART VII Toxlcs
693

633 LECTURE40
LECTURE35 NutrlentIFood-ChalnModellng Introductlon to Toxlc-Substance Modellng 695
633 40.1 The Toxics Problem
35.1 Spatial Segmentation and Physics 695
634 40.2 Solid-Liquid Partitioning
35.2 Kinetic Segmentation 697
637 40.3 Toxics Model for a CSTR
35.3 Simulation of the Seasonal CycIe 700
641 40.4 Toxics Model for a CSTR with Sediments
35.4 Future Directions 705
Problems 642 40.5 Summary 713
Problems 713
644
LECTURE 36 Eutrophlcatlon In Flowlng Waters
644 LECTURE41
( 36.1 Stream PhytoplanktonlNutrient Interactions Mass-Transfer Mechanlsms: Sorptlon and VOlatlllzatln 715
649
36.2 Modeling Eutrophication with QUAL2E 41.1 Sorption 715
658 41.2 Volalilization
36.3 Fixed Plants in Streams 727
663
Problems l 41.3 Toxicant-Loading Concepl 732
Problems 737
665
PART VI Chemlstry I LECTURE 42
Reactlon Mechanlsms: Photolysls, Hydrolysls,
and Blodegradatlon 739
42.1 Photolysis
667 739
LECTURE 37 EqulllbrlumChemlstry 42.2
667 Second-Order Relationships 751
37.1 Chemical Units and Conversions 42.3 Biotransformation
669 751
37.2 Chemical Equilibria and the Law of Mass Action 42.4 Hydrolysis
670 753
37.3 lonic Strength. Conductivity, and Activity 42.5 Other Processes
672 755
37.4 pH and the lonization of Water Problems
673 756
37.5 Equilibrium Calculalions
676
Problems
LECTURE 43 Radlonuclldes and Metals 757
677 43.1 Inorganic Toxicants 757
LECTURE 38 CoupllngEqulllbrlum ChemlstryandMassBalance 43.2 RadionucIides
677 758
38.1 Local Equilibrium 43.3 Melals
680 761
38.2 Local Equilibria and Chemical Reactions
682 Problems 768
Problems

LECTURE 44
683 ToxlcantModellngIn FlowlngWaters 769
LECTURE39 pH Modellng
683 44.1 Analytical Solutions 769
39.1 Fast Reaclions: Inorganic Carbon Chemistry
686 44.2 Numerical Solutions 778
39.2 Slow Reactions: Gas Transfer and Plants
689 44.3 Nonpoint Sources 779
39.3 Modeling pH in Natural Waters
Problems 691 Problems 782

...
xvi CONTENTS
784 Preface t
LECTURE
45 ToxlcantIFood-ChalnInteractlons
45.1 Direct Uptake (Bioconcentration)
785
788
- -- --- ..
45.2 Food-Chain Model (Bioaccumulation)
790
45.3 Parameter Estimation This is an exciting time to be involved in water-quality modeling. Today increased
794
45.4 Integration with Mass Balance national and international concern for the envirconment,coupled with tighlly limiled
795 national and local budgets, are generaling increased inlerest in ratioDal,economical
45.5 Sediments and Food Webs (Advanced Topic)
797 approaches for waler-quality management. At the same time, hardware and software
Problems
advances are making compuler-modeling frameworks more comprehensive and cas-
798 ier to use. ....
Appendlxes
798 On the positive side these trends could result in models conlributing more d-
A Conversion Factors
801 fectively to improved waler-quality management. On Ihe down side widespread and
B OxygenSolubility 802
C Water Properties easy use of modelscould lead to Iheir being applied without insight as "black boxes."
803 The main thesis of this book is that models must be applied with insight and wilh
D Chemical Elements
805 regard to Iheir underlying assumptions. A variety of features in the book maximile
E Numerical Methods Primer
817
F Bessel Functions such a "elear box" approach to water-quality modeling, ineluding:
820
G Error Function and Complement
. Lectureformato This design was adopted for two primary reasons. The first, and
821 obvious, reason is to provide a formal that facilitates the use of Ihe text for a
References
course in water-quality modeling. The format could be especially tlseful to some-
Acknowledgments 834 one teaching the course for the first time. The second reason relales to the way
in which students assimilate information. In my teaching I've found that students
Index 835 like to receive new malerial in manageable units. By limiting Ihe exposition to
short lecture-chapters, the student is provided a well-prescribed body of material
that can be mastered at one or two sittings. In addilion the format is useful ror
those attempting 10study the material oUlsidethe elassroom.
. Theory and applications. Most formulations and techniques presenled in the book
are accompanied by an explanation of their origin andJor theoretical basis. AI-
though the book points toward numerical, computer-oriented applications, whcr-
ever possible, strong use is made of analytical solutions lo elucidate the underlying
behavior of the compuler algorithms. In addition exlensive worked examples are
employed lo relate the theory to applicalions and to iIIustrale Ihe mechanics and

.
subtleties of the compulations.
Strong computer orientation. Details of algorithms and numerical methods under-
[
Iying the models are provided in the lext and in appendixes.
In keeping with the lecture format,l've adopted Ihe active voice and first-person
singular in an attempt to speak directly lo the studenlS as I do in my lectures. In
addition, though I avoid Ihe editorial "we," 1cmploy Ihe interactive "we" cXlensiwly
to reflect Ihe active participation I try lo stimulale in my leclures. Althou!'h I mi¡:lil
risk alienating more tradilionally minded readers, 1feellhat Ihe risk b worth laking
in order to betler connect with students.
The book is organized into seven major parts. The firsl two cover modeling
fundamentals, ineluding background material on mathematics, numerical methods,
reaction kinetics, diffusion, etc. The review of fundamentals at the beginning of the
course serves two purposes. First, it provides a common foundalion of mathematical
j and general modeling knowledge upon which the rest of the course can be built. Thus
II
xvii

f
 xviii Prcface
Preface xix
il allows Ihe studenls to approach the remainder of the malerial in a more efficient
manner and with a common perspective based on malhematical intuition. Second, it 1 wOllld like lo acknowledge thoughtful reviews of Ihis manuscript and
serves to orient the science and engineering students from disciplines outside civil discussion on modeling over the years with my colleagues: in particular, Marty
and chemical engineering lo new ideas such as mass balance, chemical kinetics, and Auer (Michigan Tech) and Joe De Pinto (SUNY Buffalo) made many suggestions
diffusion. and used the draft manuscript for their classes. Other colleagues 1 would like to
Parl 11Iis designed lo inlroduce Ihe student lo some of Ihe environments com- acknowledge are BiII Batchelor and Jim Bonner (Texas A&M); Greg Boardman
monly encounlered in waler-qualily l11odeling: streal11s, estuaries. lakes, and sedi- (VPI); Dave Clough, Marc Edwards, Ken Strzepek, and Joe Ryan (Colorado);
ments. MOSIof the material deals with physical aspecls of these environments. In Tom Cole (WES), John Connolly (Hydroqual); Dave Dilks and Paul Fleedman
addition the finallecture involves issues Ihal are commonly encounlered in apply- (LimnoTech); Steve Effler (Upstate Freshwater Institule); Linda Abriola, WaJtWe-
il11!models. This description of the "modeling environment" includes information ber, Nick Kalapodes, Jeremy Semrau, and Sleve Wrighl (Michigan); Glenn Miller
011model calibration. verificalion, and sensilivity analysis. (Nevada-Reno);Wu-SengLung (Virginia);Chris Uchrin(Rutgers);Andy Stod-
The following four parts deal wilh major waler-qualily-modeling problems: oxy- dard, Sayedul Choudhury, and Leslie Shoemaker (TetraTech); Jory Oppenhcimer
gcnlbacteria. eUlrophicalion/temperature, chemistry/pH. and toxics. and Dale Anderson (Entranco); Mike McCormick and Dave Schwab (NOAA); Kent
It should be noled that although the laller sections of the book are more problem- Thomton (FTN Associates); Carl Chen (Systech); Bob Broshears, Diane McKnight,
[, ~ oriented, Ihey are not devoid of theoretical development. However, such malerial is and Ken Bencala (USGS); Terry Fulp and Bruce Williams (Bureau of RecIama-
illlroduced in a "jusl-in-time" fashion when the student would be more motivated by tion); and Hira Biswas and Gerry LaVeck (EPA). Willy Lick and Zenitha Chroneer
a particular problem contex!. For exal11ple gas-transfer theory is introduced in the lec- (UCSB) and Bill Wood (Purdue) provided information and guidance on the topic of
lures on oxygen modeling and predator-prey kinetics in the lecture on nutrientJfood- sediment resuspension. I would also like to thank B. 1. Clark, Jim Bradley, David
chain modeling. Damstra, and Meredith Hart of McGraw-HiII for expediting this project. Obviously,
In summary Ihis book has been wrillen primarily for students. They are the indi- while 1 am grateful to the above and acknowledge their insights and assistance, 1
viduals who will delermine whether 01'not water-quality models will be used more alone am responsible for any errors you might discover in this volume.
effectively in the future. If Ihis book helps educate them properly, the cause of ratio- Last, but not least, l' d like to express my gratitude to my family, Cynthia, Chris-
nal waler-qualily management will be well served. tian, and Jeff, for their support and love. Their encouragement, as well as their tol-
erance of my long and somewhat eccentric working hours, made this book possible.
ACKNOWLEDGMENTS Steve Chapra
rhis book is Ihe result of leaching cIasses in water-quality modeling over the past
12 yearsat TexasA&MUniversityand the Universityof Colorado.So, firstand
roremosl, I would like to Ihank my studenls for their challenging questions: in par-
ticular, hmes Marlin, Kyung-sub Kim. Rob Runkel. Jean Boyer, Morgan Byars,
Chris Church, Marcos Brandao, and Laura Ziemelis. Several studenls in my model-
iI1gclass in Ihe spring of 1995gave me lots of suggeslions and detecled many errors
and typos in Ihe firsl draft. These are Chrissy Hawley, Jane Delling, Craig Snyder,
I)ave Wiberg. Arnaud Dumonl, Palrick Fitzgerald, Hsiao-Wen Chen, WiII David-
son, and Blair Hanna. My Michigan sludenls during Ihe winler of 1996,Ernie Hahn,
Murat Ulasir, and Nurul Amin. also provided insighl and many suggestions.
A number of colleagues have taught me a great deal about water-quality model-
ing. These include my good friends and teachers Drs. Ray Canale (Michigan); Ken
Reckhow (Duke); and Bob Thomann. Donald O'Connor, Dom Di Toro (Manhal-
lan College); and Rick Winfield (Bell Labs). I would also like to thank my friends
in other counlries for inlroducing me lo water-quality contexts and modeling prob-
lems outside Ihe United Stales. The1:einclude Enrique Cazares (Mexico); Monica
and Rubem Porlo, Ben Braga, Mario Thadeu, Rosa and Max Hermann (Brazil);
Kit Rulherford (New Zealand); Koji Amano (lapan); Hugh Dobson (Canada); Cedo
Maximovich (Serbia); Sijin "Tom" Lee and Dong-II Seo (Korea); Spiro Grazhdani
and Sokrat Dhima (Albania); and Pavel Petrovic (Slovakia).

, '.-
 Cologne PART I
In Kmn, a town of 1/Ionksand bones,
And pavements fanged witll murderous stones,
And rags, and lIags, and lIideous wenche~', Completely Mixed Systems
1 cOlmtedtwo-and-seventy stenches.
All well defined, and separate stinks!
Ye nymphs that reign o'er sewers and sinks,
Tlle river Rlline, it is well known,
Doth wash your city of Cologne;
But tell me, nymphs, What power divine
Shall henceforth wash the river Rhine?
-Samuel Taylor Coleridge (1772-1834)

1, Part 1 is devoted to introducing you to modcling fundamentals. To keep the expla-

\ nation uncomplicated the focus is on continuously stirred tank reactors (or eSTRs).
Lecture 1 provides an overview of water-quality modeling and an introduction to fun-
1I damental quantities and units. Lectllre 2 focuses on reaction kinetics. In particular I
11 present methods for estimating reaction rates.
11 The next two lectures deal with fundamental analytical solution techniques for
a eSTRo Lecture 3 shows how a mass balance can be developed for such a reactor.
It also iIlustrates how a steady-state solution is derived and interpreted. This lecture
presents the simplest time-variable solution-the general solution-and illustrates
I the concepts of an eigenvalue and response time as metrics of a system's tempor...1
In response characteristics. Lertllre 4 explores time variable solutions for a eSTR sub.
ject to a number of idealized loading functions. These inelude step, impulse, linear,
exponential, and sinusoidalloadings.
In Lecs. 5 and 6 I begin to show how more complicated systems can be ~imulaletl
~ using groups of eSTRs. Lectllre 5 deals wilh reactors connected in series. Lecture
I 6 is devoted to reaClors with feedback and introduces matrix algebra to expedite
the analysis of multireactor systems. In particular 1 present the steady-state system
response matrix as a means to analyze and interpret such systems.
To this point all the material has dealt with elosed-form, analytical solutions.
Lecture 7 is devoted to numerical methods for solving modeling problems with the
computer. Emphasis in this lecture is placed on three techniques for simulating
~ the dynamics of individual and linked eSTRs: the Euler, Heun, and fourth-order
1 Runge-Kuua methods.
I
I
 LECTURE 1
,)
~ntroduction

( ~

1 provicle an itltrodÍlcti~l1. tb Wlttet-qÍl~it

"LÍCt'.¡ÍutOYEltVltWi
'térár.evi~\\r oftbe historical origios oí w~~t ~h~:WlISte1
efdIs9U§~OtlOf majot variables aíid uni~is~tésentéd,
. ., . a,Mt ~yáwater-quálity,tttRdelM~.~scus§,:~€
éapplied. FináUy 1reviewf1teh!srorlcal déV,

,¡~:~uíilltXJitod~ifng atid outtine the temaindér.6f.tbis setotJéPt~'?t)

One day a king arrived at a town inhabited by the sightless. Along with his troops, he
brought a large elephant. Several of the inhabitants visited !he king's encampment and
tried to gather information by touching some par! of the beast. When they returned, their
fellow citizens gathered around to learn about its form and shape. The man who felt the
trunk said: "An elephant is long. thick, and flexible, like a snake." One man who had
touched the tail countered: "No, it is thín like a rope." Anolher who had felt its leg said:
"It is solid and firm, like a tree trunk." Finally, an individual who had climbed on its back
remarked: "You're all wrong, an elephant is rough and texlured like a brush." Each had
been misled by feeling one part out of many. AII imagined something incorrect because
they did not perceive the whole elephant.
Paraphrased from
Tales of the Dervishes
Shah (1970)

Now, aside from the faet that elephants tend to generate large quantities of solid
waste, this parable might seem to have IiUlerelevanee to water-quality modeling. In
faet it has great pertinenee to the task before us. As will be cIear from the following
leeture, models were originally developed as problem-solving tools. However, be-
yond their utility in remediating pollution problems, they have a broader funetion;
that is, they provide us with a means to visualize "the big pieture." In essenee the

-,
...
,.~--

4 PARTI Complelely Mixed Systems ...

IH'fURE I Inrroduclioll 5
mathematical model offers a quantitative framework lo integrale the diverse physi-
Crltical concentration e (ML-3)
cal, chemical, and biologicalinformation that constitute complex environmental sys-
tems. Beyond solving a particular pollution problem, models provide a vehicle for
an enhanced understanding of how the environment works as a unit. Consequently .
they can be of great value in both research and management contexls. I wanted lo
start with the elephant parable lo underscore this point. You'lI gel a lot more OUIof
this book if you try to keep "the whole elephant" in mind as you proceed. Waste removed FIGURE 1.1
An uroan water-wastewattJr
system. A water treatmenl
1.1 ENGINEERS AND WATER QUALlTV plant (WTP) pllrili es river water

Engineers, particularly civil engineers, have been Ihe principal developers of the I CITY I lor human consurnption. A
wastewater troatment plant
field of water-quality modeling. This might seem odd considering thal civil engi- (WWTP) removes pollutanls
neers are most commonly associated wilh areas such as structural design and trans- Irom sewage to protecllhe
receiving walur.
portation. In fact, rather than being environmentally orienled, civil engineers are
sometimes stereolyped as hard-hatted individuals who toss drink containers out the
window of their pickup trucks while simuhaneously Irying to hit any small mammal began to be constructed. However,it was soon recognized that trealmcnl could ntnge
that crosses their path. from simple sedimentalion to costly physical/chemical treatment. In (he exlreme
Although such stereolypesare gross oversimplificalions(for example many civil case the latter might actually result in an efftuenl thal was more pristine th,tn Ihe
engineers do not drive pickups), one would more likely expect such scienlists as receiving water but at an exorbitant cosl. Clearly bOlhextremes wen.'unacceptahle.
chemists, biolegists, and ecologists to be at the forefront of this area. Further, even Consequently some design goal had lObe eSlablished Ihat would prolcCIthe envi-
within engineering, otherdisciplines such as chcmical engineering might seem more ronment adequately bUIeconomically.
naturally attuned to the environment. The reason for the slrong involvement of civil Thus it was decided Ihal waste Irealment should be based on producing an efflu-
engineers lies within the hislorical problem contexts that first drew attention to en- ent Ihat induced an acceplable level of water qllalily in the receiving waler. Howcvcr,
vironmental problems. lo determine Ihis proper level of Ireatmenl, it was necessary to predio.:!waler Ifualily
Originally there wereonly two types of engineers: military and civilian (orcivil). as a function of waste loading. As depicted in Fig. 1.1, a linkage had lo be eSlab-
As the label suggests, military engineers were concemed with the lechnologyofwar; lished between Ihe waste loading Wand the resulting conn'/llralion e in Ihe waler
Ihal is, they buih fortificalions,naval vessels, and weapons. AII other technological body. Therefore civil enginee hegan lo develop malhematical modcls for that pur-
pose.
concerns, such as roads and h"lIlllg, generally fell within the civil realm.
With the onset of the induslrial revolution. new technologies began lo develop Today water-quality lIJa/lagement has moved well beyond Ihe IIrban poiul-
and specialized engineering skills became necessary. Hence disciplines such as source problem lo encompass many other types of pollution. In addilion to waSlc-
electrical, mechanical, and chemical engineering matured and broke off from civil water, we now deal with olher poinl sources such as industrial wasles as 1\1.'11
engineering. By the early twentieth century. civil engineers were primarily respon- as nonpoint inputs such as agricultural runolI However, as depicled in Fig. 1.2, a
sible for the building of structures, Iransportation systems, and major public works
projects such as dams and aqueducts.
Stimulated by the awa..;lIcss Ihal water-borne palhogens were one ,,1tiIl ¡llune DeslredconcentraUon~
causes of disease, civil engineers began to design urban water and wastewater sys-
tems (Fig. 1.1) in Ihe late nineteenth century. Consequently civil engineers became
involved in Ihe development of water treatment plants, distribution networks. and
wastewaler collection systems. The design of these projects was fairly straighlfor-
ward, because the goals were so well defined. The goal was to deliver an adequale
ConcentraUon c
-
" Ves
quantity of potable water to Ihe urban poplllace and to safely carry olTits wasles.
In contrast, the question of what to do with the wasle was a more ambiguous
propositiol1.Al first,municipalities discharged their raw sewage directly lo receiving IContro';¡
waters. It was immedialely observed Ihat such action would ultimately transform FIGURE 1.2
rivers. lakes. and estuaries into large sewers. Hence wastewaler lrealment plants
The water-qualily managemenl process.
:¡
J
 6 PARTI Complelely Mixed Syslems LECTUREI Introduction 7

water-qualily model slill provides lhe essential link to predict concentration as a TABLE 1.1
I ~ functJon of loadings. SI (International System
of Vnits) prefixes
commonly used In
1.2 FUNDAMENTAL QUANTITIES water-quaIlty modeIlngt
Prefix Symbol Value
1n the foregoing discussion I introduced the notions of concentration and loading. Be-
fore proceeding furlher, I would like lo take a closer look at Ihese quanlities and how kilo- k 10-'
Ihey are represenled numerically. I also define some other fundamental quantities heelo- h 102
deci- d 10-'
used in waler-qualily modeling.
eenli- e 10-2
milli- m 10-3
miero- ¡J. 10-6
1.2.1 Mass and Concentration nano- n 10-'

In water-quality modeling, the amount of pollulant in a system is represented by · A eumple'e lisl ur prefixes is in.
ils ~ Such a properly is formally referred lo as an extensive property. Other cluded in App. A.

.\ examples arc heal and volume. AII these characlerislics are addilive.
In contrasl, a quanlily thal is normalized lo a measure of syslem size is re-
rerred lo as an intensive property. Examples include lemperalure, density, and
mg W L g g _
L ffiJ 103mg - m3
prcssure. For our problem conlexl, lhe intensive propcrly is mass concentration de-
fined as o(~ Thus Ihe unit mglL is idenlical to g/m3.
11I This situalion is further complicaled because, for Ihe dilute aqueous solulions
(1.1)
e = V common in mosl surface walers, concentralion is sometimes expressed on a mass
basis. This conversion is predicaled on the facllhallhe density of water is approxi-
where m = mass (M)t and V = volume (L3). The utilily of concentration lies in malely I:quallo I g/cm3:t
Ihe ract Ihat, as with all inlensive quantities, it represents the "slrength" rather than
g g m3 g
Ihe "quanlily" or Ihe pollulion. As such il is preferable lo use concentration as an
indicator of impacl on the environmenl.
m3 = m3 x (1 g/cm3) 106 cm3 = 106 g = I ppm
where ppm slands for "parts per million." Olher identities are summarized in Ta-
:\ An example from everyday life is userul in explaining the difference. One. two,
or Ihree lumps of sugar can sweelen your coffee in varying degrees. depending on
ble 1.2.
(hc size of your mug. Thc number of lumps is analogous to mass. as the sweetness
is lo conccnlralion. As an "organism" you are usually more con cerned ahout Ihe TABLE 1.2
swcetness than the number of lumps. Some
~ water-quaIlty
mI___variables along wlth typlcal unlts
"i
Concenlration is conventionally expressed in metric un¡ts. The mass in Eq. 1.1 Variables Vnits
is expresscd in the fundamenlal unil of grams combined with the prefixes in Table
1. 1. Thus Total dissolved solids, salinily gV' {:~ kgm-' {:~ ppl
Oxygen, BOD. nilrogen
I X 103 mg = I g = 1 X 10-3 kg mg V' {:~ g m-' {:~ ppm
Phosphorus. chlorophyll a, 10xics ¡J.g V' {:~ mg m-' {:~ ppb
The volume unit is not as straightforward because il is typically expressed in
Toxics
one of two ways: liters or cubic meters. Depending on the choice of the volume unit. ng V' {:~ ¡J.g m-3 {:~ pplr
lhis can lead to some confusion. because equivalent representations can result. For
cxample
EXAMPLE 1.1. MASS AND CONCENTRATION. If 2 x 10-6 lb of salt is inlro-
duced inlo l m' of dislilled water. what is the resulting concentration in ppb?
']nilially] define unils in terms nfthe fundamental dimensions: L = length. M = mass, and T = time.
Laler I will introdllee speeific melr;e unils. · Following Ihis introduelory discuss;on, unils of measure will appear in exponenlial form Ihroughout
Ihe book. Thus mglL is shown as mg L '. g/m' as g m-", and Silon.

-.

" .~ .~ - .. . .
 ......-. "
~---

~ PARTI Completely Mixed Systems

LI'<TUI<I
1 Inlrod,.c¡iul1 <)
Solution: Applying Eq. 1.1, along with the conversion factor for pound to gram from
App. A (1 lb = 453.6 g), yields delermined by measuring Ihe coneenlralion along wilh Ihe vohlllletrÍc t10w rate of

)_
_ 2 x 10-6 453.6 g 1.3(/) in the (;onduil.
e - lffi3 ( ¡¡;- -.
9072
x
10-4
gm
-3
¡I.
waler Q (L J T r). The l1lass loauing
I
rate is Iheli CUmpule!! as (hg.

Converting to the desired units, I W = Qc (1.3)

103mg ppb Volumetric flow rateo For steauy lIow Ihe t10w rale is orten ,;aI...:ulaleu wirh Ihc
e = 9.072 x 1O-4g m-3 contb",ity eqllation (Fig. 1.3b)
( g mg m- 3 ) = 0.9072 ppb Q= VA("
( 1.4)

As in Example 1.1, it is conventionallO use units that express Ihe resulting con- where conuuil
oflhe (L2). of water inlhe eonuuil (1.T- 1)anu A,. = cross-sccliollal arca
V = veloeity
centration with a magnitude between 0.1 and 999. Thus waler-quality variables are
typically expressed in particular units, as outlined in Table 1.2.
/ Mass flux rateo The lerl1ljillX is useu lo designate Ihe rale ofl1lovemclIl 01'all I
eXlensive quanlily like mass or heat nonllalized lo area. For examplc the mass flux ..,
rate Ihrough a eonuuit can be calculated as
1.2.2 Rates
11/
J=-=_ W
Properties that are normalized lo time are commonly referred lo as rates. We now t A(" A("
(1.5)
discuss severa) rates that are key to understanding water-quality modeling (Fig. 1.3).
By subslituling Eqs. 1.3 and 1.4 inlo Eg. 1.5, mass flux can also be Cxprcsscd ill
lerms of veloeily and eoncclllralion by (Fig. l.3c)
Mass loading rateo As in Fig. 1.1, waste discharges are Iypically represented
by the mass loading rate W. If the mass m of pollutant is determined over a time J = Ve
( Lb)
period t, then the loading rate can be simply computed as
EXAMPLE 1.2. LOADING AND FLUX. A pond having conslalll volullle alld no
m
W = -t (1.2) outlet has a surface area A., of 104 1112and a meml depth H of 2 111.h inilj¡llIy 11.1"a
concentralion ofO.8ppm.Twodayslatera measurelllentindicalestharlhe concclllralion
Many of these loadings enter receiving waters as point sOllrces;that is, through has risen lo 1.5 ppm. (a) \Vhat was Ihe lIIass loading rale during Ihis lillle'! (M Ir yuu
conduits such as pipes or channels. For such cases the loading rale is actually hypothesize
eSlimate tha! Ihal
the lIux only possible source of Ihis pollutanl was frolll Ihe atlllusphere,
Ihe occulTed. ,)
Solution:

(a) The volul11e of Ihe syslelll C',II be calclll¡¡leu as

V = A,H = 1041112(2111)= 2x 104m.\

The lI1ass of pollulant at Ihe inilial time (t = O) can be compuled as

Q u 11I = Ve =2 X 104 11'-\(O.~ g 111-.1)= 1.6 X lO' g

(a) Loadlng (W = Qc) (b) Flow (Q = VA,)
and all rale
loading = 2 isd is 3.0 X 104 g. Therefore Ihe inaease in m¡¡ss is 1.4 X 101 g alld I',e m,L.S

W= 1.4 x2uIO~ g = 0.7 x 10~gu

(b) The lIux of pollutanl can be COlllputed as

( J= 0.7xI04gd-1 = 0.7 g(m2d¡-I

u
(e) Flux (J = Vc)

FIGURE1.3 Sometimes the concepl of rate can be confused wilh quantily (in our case mass
Three fundamental rates used extensively in water-quality modeling. or concenlralion).
this happens. Box 1.1provides one example of Ihe problems thal can oecur when
 10 "ART' Completely Mixed Systems
""
BOX 1.1. Rate versus auantlty o, Great Lakes Water
In the early \980s, droughts in the western Uniled States prompted the government lo
seek out additional sources of water for this region. At one point the nolion of building
a ,,¡peline from the Laurentian Great Lakes to the wesl was actually enlerlained.
At faee value the plan seemed to ha'Je merit beeause the Great Lakes eontain about
21.71 x 1()12nI" ofwaler. This volume corresponds lo roughly 2()% oflhe world's liquid,
fresh, surfacc water! Consequenl,ly it seemed to make "common" sense that the Greal
Lakes had plenty to spare.
At I¡rsl. arguments revolved amund whl'lher the pumping costs would be pro-
hihitive (ies seriously uphill lo gel to ami across Ihe Rockies). However. there was a
!llore fundamental reason why the scheme was lIawed. This involved Ihe distinction
hclween the quantily of water and its lIow rateo
Although 20% of the world's water might be held in the system. the volume is
not a val id "1easure of how much water could be spared. The actual generation rate is
p rellected by ¡he lIow rate of waler out of the syslem. Inlerestingly the lakes that were
larthest west (Michigan and Superior) have small outllows (together only about I()O x
IO~ m" yr'I). It is only when you trave! east to the most downstreamlake (ünlario)
Ihal the outllow inlo the SI. Lawrence River reaches a respeclable 212 x 1()'¡ m' yr-1 .
Furlher. lhis lIow helps lO clean Ihe lakes by lIushing pollutanls. Thus. seen from this
perspeclive. Ihe Great Lakes did nol seel11 so attractive ami the schel11e died.
In essence lhe proponents 01' Ihe plan confused Ihe meanings of size and rale.
Although Ihe Great Lakes ha ve lols of waler (as relleeted by their size). they eannot
spare much of Iheir oulllow or "harvesl" rate of waler. Furlher. if a pipeline had been
buill and waler pumped al a rale in excess of Ihe average outllow. Ihe Greal Lakes
would have rapidly become known as lhe "Greal Mud Flals."
So whal generallessons does this slory otfer7 rirst. as Einslein said.
Common sense is whalmakes youlhink Ihe world is Ilal.
tJ, For some individuals a "rate" is not a "coml11on sense" concepl. Therefore,
oUllined above. irralional judgmenls can resull. Beyond Ihal the distinction belween
'¡ze and rale will he critieal to understanding lhe modcls descrihed in lhe remainder of
Ihis le xl. ¡'ve included Ihis story lo hclp you appreciale Ihe dislinclion helween lhese
.
I
two types of l11easures.
l.
1.3 MATHEMATICALMODELS
N'JWthat some fundamental quantities have been defined, we can discuss mathe-
matical models. According to Ihe American Herilage Dictionary (1987), a model is
a sITIallobject usually huilt to scale, that represenls anolher often larger object. Thus
tl10delsIypically represenl a simplilied version of reality thal is amenable to testing.
-t In our case we do nol build physical models, bul we use mathematics to repre-
sent realily. We can Iherefore define a matllematical model as an idealized formu-
lation Ihat represenls the response of a physical syslem lo eXlerna)stimuli. Thus, in
l'~ Ihe conlext o[ Fig. 1.1, a mathemalical model was needed lo compute quality (the
"" response) in the receiving waler (the syslem) as a function of Ireatmcnl planl effluent
- -- .- ~
LECTURE' Introduction 11
(the stimuli). Such a model can be represented generally as
e = f(W; physics, chemistry, biology) (1.7)
According to Eq. 1.7the cause-effect relationship between loading and concentration
x { depends
waler. on the physical, chemical, and biological characteristics of the receiving
The remainder of Ihese leclures will deal wilh refinements of Eq. 1.7.One very
simple slep in Ihis direclion is lo employ a linear relationship lo formulale Eq. 1.7 in
mathemalicallerms as
1
e =-W ( 1.8)
a
where a = an assimilalion faclor (LJT-I) Ihat represents Ihe physics, chemistry,
and biology of Ihe receiving waler. Equalion 1.8 is called "linear" because e and W
are direclly proportionallo each olher. Consequently if W is doubled, e is doubled.
Similarly if W is halved, e is halved.
1,3.1 Modellmplementations
Equalion 1.8 can be implemenled in several ways:
l. Sill/ulafion mode. As expressed in Eq. 1.8, the model is used to simulale sys-
tem response (concenlralion) as a funclion of a stimulus (loading) and system
characlerislics (the assimilation factor).
¡ . 2. De.~ignII/odeI (assill/ilafive capacify). The model can be rearranged lo yield
W = ac ( 1.9)
as
This implemenlation is refened lo as a "design" mode because it provides infor-
malion Ihal can be directly used for engineering design of Ihe system. It is for-
mally refcrred lo as an "assimilalive capacily" compulalion because il provides
an eSlimateof Ihe loading required to meet a desirable eoncentralion level or stan-
dard. Thus it forms Ihe basis for wastewater treatment plant designoIt should also
c1arify why a is called an "assimilation factor."
3. Desi1:/I/Ilode 11 (l'/Il'iro/llllellfalII/odijicafio/l). A second design implementa-
tion is
W
a= (1.10)
e
) In this case Ihe environment itselfbecomes the focus ofthe remedial effor!. Equa-
..¡ 11ion 1.10is formulated to delermine how,for a given loading rale, the environment
. might be modified to achieve the prescribed standard. This type of application is
needed when affordable treatment (that is, reduction in W) is not adequale to
meel waler-quality slandards. Examples of environmenlal modifications would
[
be dredging of boltom sediments, artificial aeration, and now augmentation.
12 PART1 Completcly Mixed Syslems I.ECTUKEI Inlroduclion 13

EXAMI'LE 1.3. ASSIMILATION FACTOR. Lake Onlario in Ihe early 1970s had a e
total phosphorus loading of approximately 10,500 mta (metric tonnes per annum, where
a metric tonne equals 1000 kg) and an in-Iake concentration of 21 JLg L -1 (Chapra and
Sonzogni 1979). In 1973 the state of New York and the province 01' Ontario ordered a '/':~!
reduction of detergent phosphate contento This action reduced loadings to 8000 mta. --- a
(a) Compute the assimilation factor for Lake Ontario. FIGURE 1.4
(b) What in-Iake concentration would resu\t from the detergent phosphate reduction Empiricalwater-qualitymodelsuse data Irommanywater
action'! bodies to statistically estimate the cause-ellect relationship
(e) Ifthe water-quality objective is to bring in-Iake levels down to 10 JLg L -1, how w between loading and concentration.
much additionalload reduction is needed?

Solution: much of c1assiealengineering is based on Newton's laws, in particular his seeond

(a) The assimilation factor can be calculated as law: F = ma. In addition the great eonservution laws are commonly employed as
organizing principies for much engineering work.
a = ~ = 10,500 mta = 500~
e 21 JLgL--1 JLgL-1 Although empirical approaches have proved valuable in certain water-quality
contexts such as lake eUlrophication (see Reekhow and Chapra 1983 alld Lec. 29
(b) Using Eq. 1.8, in-Iake leveis from the phosphorus reduction can be calcu-
lated as for a review), they have some fundamenlal limilations. Consequently I will adopt
w 8000mta primarily a mechanistie approaeh for most of the remainder of Ihese leelurcs.
e= = 16¡LgL-1 Mechanistic water-quality models are based on the conservation 01 mass; that
a 500~
JLgL-1 i is, within a finite volume of water, mass is neilher created nor destroyed. lo quantita-
tive terms the principIe is expressed as a mass-balance equation that accounls for all
(e) Usiog Eq. 1.9, transfers of maUeraeross the system's boundaries aod alllraosformations oeeurring
mta JLg wilhin the system. For a linite period of time this can be expressed as
W = l/e = 500-10- = 5000 mta
JLgL-1 L Accumulation= loadings :t transport :t reactions (1.11)
Therefore an additional 3000 mta would have to be removed.
Figure 1.5 depicts mass conservation for two hypothetieal substanees that tlow
through and react within a volume of waler. The movement of malter through the
volume, along wilh waler ftow, is termed transporto In addition to this lIow, mass is
Regardless of the mode of implementalion, Ihe effeetiveness of Ihe model is con-
gained or lost by transformations or reactions (lf Ihe substances within the volume.
tingenl on an accurate characterization of the assimilation factor. As in Example 1.3,
Reactions eilher add mass by ehanging anolher constituenl inlo the substanee being
data provides one way to estimate the factor. A primary goal ofthe following lectures
modeled or remove mass by transforming the substance into another constituent, as
wi\l be to delineate other ways in which this factor can be determined. To place these
in Fig. 1.5, where X n:aels lo I(mll Y. Finally lhe subslanee can bc inL'lcascdby
etforts into proper perspective, we will brietly review Ihe major organizing principIe
Ih .; used to do Ihis-the conservation of mass. externalloadings.

o
1.3.2 Conservation of Mass and the Mass Balance A
-o "
Traditionally, two approaches have been employed to estimate Ihe assimilation /' ~ "
factor:
. Empirical model.\' are based on an inductive or data-based approach. This was
.
SUb8lance X ~ G

\
\\
the approachemployedin Example1.3for a singlelake. Moreoften it mvolves Transport
obtaining values of W and e from large numbers of systems that are similar to
Tren.por!
In
: I I Reactlons
,
,
out
FIGURE 1.5
th.: receiving water in question. Regression techniques can then be employed to A schematic representation
SIilistically estimate the assimilation factor (Fig. 1.4). L,;;l - - "- 6'l{,-
01 the loading, transport, and
. !I¡echanistic models are based on a deductive or theoretical approach. This in-
volves the use of theoretical relalionships or organizing principIes. For example
~ '- -" ~~ boUndary
S'\S - , ~.. translormation 01two substances
moving through and reactiog
within a volume 01water.
 14 I'ARTI Complelely Mixed Syslems
LECTUREI Inlroduclion 15

~
( By combining all the above factors in equation form, the mass balance repre-
1 sents a bookkeeping exercise for the particular constituent being modeled. Ir, for the 1925-1960 (Streeter-Phelps)
period of!he calculation, !he sources are greater !han the sinks, !he mass of the sub- ProbIems:untrealed and prIme'Yelfluent
Pollutanls:BOOIDO
stance within !he system increases. Ir the sinks are greater than !he sources, the mass . Sysl"",s: slreamslesluartes (10)
Kin6tics: linear, feed.forward
decreases. Ir sources are in balance with sinks, the mass remains at a constant level SoIulions: analy1icat
and the system is said to be at steady-state or dynamic equilibrium. The mathemati-
cal ex pression of mass conservation, therefore, provides a framework for calculating
the res pon se of a body 01'water to external influences.
Since the system in Fig. 1.5 ineludes two substances, separate mass balances 1960-1970 (computerlzstlon)
should be written for X and Y. Each should inelude mathematical terms to account Problems: primary and secondary effluenl
Pollutanls: BOOIDO
fol' the transport of the substances into and out of the system. In addition the balance Syslems: esluarieslslreams(1 0120)
Kinetics: linear, feed-forward
for X should inelude a term to reflect the loss of X to Y by reaction. Likewise, the Solulions: anatytical and numerical
equation for Y should inelude the same term but with a positive sign to reflect the
gain of mass by Y due to the same process. Finally the balance for X should inelude
a term for the mass gained by loading.
For situations where more than two substanc¿s interact, additional equations 1970-1977 (blology)
could be written. Similarly an investigator might be interested in the leve\s of sub- Probfems: eulrophicatlon
PoI/ulsnlS:nutrianls
stances at various locations within the volume. The system can then be divided into Systems: lakeslestuarieslstreams
(10120130)
subvolumes for which separa te mass-balance equations would be developed. Addi- Kinetics: nonlinear, feedback
So/utions: numerical
tional transport terms could be ineluded to account for the mass transfer between !he
subvolumes. This mathematical division of space and maUer into compartments-
termed segmelltatioll-is fundamental to the application of mass conservation to
water-quality problems.
1977-present (toxlcs)
The remainder of the book wi\l be devoted to using mass balance to develop
Problems: lox~
various expressions of Eq. 1.11. In the course of doing this we wil\ describe how Pollulanls: organics. metals
I l transport and reactions can be formulated mathematically and how segmentation Syslems: sedimenl~water InteracUonsl
food-chain interactlons
figures in the process. To place these efforts into proper perspective we will now
ItP briefly review the historical evolution of water-quality modeling.
(Iakeslestuarieslstreams)
Klnetics: linear, equilibrium
SoIutions: numerfcal and analytical

FIGURE 1.6

,. 1.4 HISTORICAL DEVELOPMENT OF WATER-QUALlTV MODELS Four periods in the developmenl 01 waler-quality modeling.

'o'..
Water-quality modeling has evolved appreciably since its innovation in the early In the I 960s digital computers became wide\y available. This led to major ad-
-:0-'-
years of the twentieth century. As depicted in Fig. 1.6, this evolution can be broken vances in both the mode\s and the ways in which they could be applied. The first
down into four major phases. These phases relate both to societal concerns and to modeling advances involved numerical expressions of the analytical frameworks
the computational capabilities that were available during each of the periods. (e.g., Thomann 1963). Oxygen was still the focus, but the computer allowed analysts

~
Most ofthe early modeling work focusedon the urban waste\oad allocation prob-
lem (Fig. 1.1). The seminal work in the field was the model developed by Streeter
and Phelps (1925) on the Ohio River. This and subsequent investigations provided
..
a means to evaluate dissolved oxygen levels in streams and estuaries (e.g., Velz
1938. 1947; O'Connor 1960, 1967). In addition, bacteria models were also deve\-
oped (O'Connor 1962).
Becauseof the nonavailability of computers, model solutions were cJosedformo
This meant that applications were usually limited to linear kinetics, simple geome-
lries. and sleady-state receiving waters. Thus the scope of the problems that could
r~. b(' addrcsscd was constraincd by the available compulational tools.
\
to address more complicated system geometries, kinetics, and time-variable simu-
lations. In particular the models were extended to two-dimensional systems such as
wide estuaries and bays.
The sixties also brought changes in the ways in which the models were applied.
In particular the computer allowed a more comprehensive approach to water-quality
problems. Rather than focusing on local effects of single point sources, one could
view the drainage basin as a system. Tools developed originally in the fie\d of op-
erations research were coupled with the models to generate cost-effective treatment
alternatives (e.g., Thomann and Sobal 1964, Deininger 1965, Ravelle et al. 1967).
Although the focus was still on point sources, the computer allowed a more holistic
perspective to be adopted.

"
 16 PARTI Completely Mixed Systems
In the 1970s anolher shift occurred. Societal awareness moved beyond dis-
solved oxygen and urban point sources lo a more general concern for the envi-
ronment. An ecological movement was born and, in some qUaJ1ers,environmental
1 \ remediation became an end in itsel!'.
, The principalw.lh:r-qualityproblemaddressedduringthisperiodwaseutroph-
ication: As a consequence, modelers broadened their own scope to inelude more
mechanistic representations 01'biological processes. Capitalizing on oceanographic
research (e.g., Riley 1946, Steele 1962), environmental engineers developed eIab-
orate nutrientlFood-chain models (Chen 1970; Chen and Orlob 1975; Di Toro ~t al.
1971, Canale et al. 1974, 1976). Because 01'the existing computational capabilities,
feedback and nonlinear kinetics could be employed in these frameworks.
It should be noted that during Ihis period, major work proceeded in bringing the
urban point-source problem under control. In fact, most municipalities in the Uniled
Slales inslaIled secondary Irealmenl 01'their effluenls. Aside from ameliorating Ihe
dissolved oxygen problem in many locales, 1'01' areas where poinl-source control was
insufficienl, Ihis had the ancillary etTecl01'shifting attention toward nonpoinl sources
~' 01'oxygen-demanding wasles. Because such sources are also prime contributors 01'
¡ nutrients, Ihe emphasis on eutrophication reinforced concern over nonpoint inputs.
Al face value Ihe environmenlal awareness of Ihe early seventies should have
led to an increased reliance on Ihe syslems approach to waler-quality management.
Unfortunately Ihis was not Ihe case, 1'01'three primary reasons. First, because eu-
. ),trophicalion deals with seasonal planl growlh, it is a more dynamic problem Ihan
(urban poinl-source conlrol. Although systems analysis methods could be devised to
oplimize such dynamic problems. they are much more complicated and computa-
tionaIly inlensive Ihan 1'01'Ihe linear, steady-slate, poinl-source problem. Second, Ihe
) environmental movement foslered an almosphere 01'urgency regarding eleanup. A
'(mentalily 01'remedialion "al any cosl" led to concepls such as "zero discharge" being
I articulaled as a national goal. Third, Ihe economy was booming during Ihis periodo
I Therefore Ihe economic feasibilily 01'such a stralegy was nol seriously questioned.
As a consequence Ihe idea of balancing cosls and benefils lo devise an economi-
cal solulion waned. Legalities supplanled sound engineering as the basis 1'01'most
poIlulion conlrol stralegies. Although progress was made during this period, Ihe un-
realistic goals were never achieved.
The mosl recenl slage of model development evolved in the wake 01'Ihe energy
crisis of the mid-seventies. Togelher with increased deficit spending, the energy cri-
sis broughl the poIlution conlrol effort back to economic realily. Unfortunately Ihe
initial response amounted lo an overreaction lo the excesses 01'Ihe early sevenlies.
i '
Now, ralher Ihan an "at any cost" strategy. the public and theír represenlalives had
lo be "sold" on Ihe efficacy 01'environmental remediation. Consequently attenlion
IUmed lo problems such as loxic subSlances (and lo a lesser exlenl acid rain) Ihal,
although they certainly represented a major threallo human and ecosysh:m health,
[ could also be markeled effeclively in Ihe polítical arena.
l) The major modeling advance in this period h~s been lo recognize Ihe promi-
.., nent role of solid matter in the transport ~nd fale 01'loxicanls (e.g., Thomann and
Di Toro 1983, Chapra and Reckhow 1983, O'Connor, (988). In particular !he as-
socialion of toxicanls with settling and resuspending pmieles represenls a major
I LECTUREI IlItroductiolJ 17
I
( mechanism conlroIling Iheir transport and fale in natural waters. F'lrther, smitll 01'-
!
~~
f ) ganic partieles, such as phytuplankton and detrilus, can be ingested itndpassed alllng
¡ lo higher organisms IThomann 1981). SlIch food-chain inleracliuJls have led Ihe
modelers to view mil.. 's urganic carbon cycle as more Ihan an end in ilself. R,lIlIer,
Ihe food chain is viewed as a conveyer and concentrator 01'contaminants.
Today anolher shifl is occurring in Ihe development and application 01'waler-
qualily modeling. As was Ihe case in the lale sixties and early seventies, there is
( a strong (and growing) recognition Ihal environmenlal prolection is crilical lo Ihe
mainlenance 01'a high qualily 01'life. However, added lo Ihis awan.:n~ssare taur
faclors thal should make Ihe coming decade different from Ihe pas\:
~ l. Economicpressuresare, if anylhing,moreseverethan duringthe lalesevenlies.
· 1 Thus incenlives tal' cost-effeClivesolutions are slronger Ihan ever. This notiulI is
reinforced by Ihe fact that in the Uniled States the least expensive poinl-sollrcc
Irealment oplions have already been implemenled. As depicted ill Fig. 1.7,Ireitl-
menl currenlly deals wilh Ihe sleepest part 01'Ihe cosl curve 1'01'poinl sources.
Further, reductions 01'nonpoinl 01'diffllse sources are lypicaIly more expensive
than poinl-source controls. Today better models are needed to avoid the scvere
J {
economic penalties associated with faulty decisions.
I Developing countries around Ihe world are beginning lo recognilc Ihat envi-
'# ~ 2. ronmenlal proleclion musl be coupled wilh economic development. For Ihese
$IMGD
11$_
.
'-""' '-""'
0/0removal
o I1load I1load 100
secondary lertlary
FIGURE 1.7
Capilal conslruction costs versus degree 01Irealment 101'municipal
wastewaler trealment.Nole Ihat most decisions relaling lo lertiary
wasle Irealmenl presenlly deal with high-percent removals.
Consequently a laully decision carries a much higher economic
penalty loday Ihan in earlier years when primary and secondary
wasle Irealment were dominant.
 18 PARTI Completely Mixed Systems
countries, cost-effective, model-based control strategies could provide a means
to control pollution and sustain a high quality oflife while maintaining economic
growth.
3. Computer hardware and software have undergone a revolution over the past
decade Ihat rivals the initial advances made during the I960s. In particular,
gnlphical IJserinterfaces and decision support systems are being developed that
facilitale tI\(' generation and visualization of model output. Further, hardware
advancesaJe removingeomputationalconstraintsthat limitedthe scopeof ear-
lier models. Today two- and three-dimensional models with highly mechanistic
kineties can be simulated at a reasonable cost.
4. Finally signifieant research advances have occurred in the reeent past. (n par-
ticular, l11eehanisticeharaeterizations of sediment-water interactions and hydro-
dynamics have advaneed to the point that they can be effectively integrated into
wUler-quality-modelingframeworks. Aside from Ihe scientific advances involved
,,) in devclnpi'l,!!.Ihese meehanisms. their subsequenl inlegration into usable frame-
"'"rks is bcing madc possiblc by the advanees in eompulcr lechnology.
In summary the evolulion of water-qualily modeling over the lasl 70 years has
rcsulted in a unified Iheoretical framework Ihat encompasses bOlheonventional and
loxic pollulanls. In addition a variety of computer codes are available lo imple-
l11elllthe tllenry. Finally, although inl"requentlyused, there are a number of sys-
tCIIISanalysis teehniques that could be linked with the simulation models to provide
cnsl-cl"I"eetivcengineering solulions. Together wilh strong societal eoneern for the
environmenl, Ihese faelors have provided aclimate for a new management-oriented.
eOlllputer-aidedphase in water-quality modeling.
As airead)' stated in the prefaec. there is a possible down side to this new phase
\ 01"water-quality modeling; that is, widespread and easy use of mode1seould lead lo
,1' their being applied without insight as "blaek boxes." The main thesis of Ihis book
is thal models musl be applied wilh insight and wilh regard lo their underlying as-
sumplions. amI Jhave gone lo greal lenglhs lo maximize such a "clear box" approaeh
to walcr-quality modcling. Now 1'11describe how the book is organized to a!tain this
(,bjeelive.
-.--
BOX 1.2. Readlng the Llterature
A'ide fwm providing some historical perspective, I had an ulterior motive in writing
this sectioll. Ir you really want to become an expert water-quality modeler. I would
urge you lo read the articles quoted in the above paragraphs (as well as many 01' the
olher references throughout the remainder 01' the book). Although you can learn a lot
4~ rrom lex tbooks. space constraints prevent an author from pursuing any particular topic
:4' in great depth. The published literature is a gold mine that presents knowledge in a
detailed and comprehensive fashion. allows you to appreciate your heritage. and keeps
YOllat the clllting edge. Thus l hope that you recognize that this book is merely a starting
b L___
I
i'oint-the
literature.
lip 01'an iceberg represented by the wealth 01'insights found in the broader
LECTUREI Introduction 19
1.5 OVERVIEW
OF THISBOOK
The leclures are divided into seven parts Ihal roughly follow the historical overview
from Ihe previous section. Parts I and II concentrate on modeling fundamenlals.
Part I is devoted to modeling well-mixed systems. This material ineludes an
overview of analylical and computer-oriented solulion lechniques as well as an
introduclion lo reaction kinetics. The analytical approaches emphasize Ihe lin-
ear models thal formed Ihe basis of early witér-quality modeling. In addilion the
compuler-orienled methods provide Ihe quanlitative basis for extending the early
approaches lo address more complex systems. Part l/ uses a similar approach but
for incompletely mixed syslems. .It also inelu~esan overview of diffusi.on:
Part ll/ provides an introduetion lo the el1yiroiimen!sIhat waler-quality mod-
elers usually address: slreams, estuaries, lakes. and sediments. ( provide general
background information on Ihese systems wilh special emphasis on how their trans-
port regimes are quantified. In addilion ( also devole a lecture to the "modeling"
environment. dealing with issues such as model calibralion and verification.
Part IV looks al Ihe firsl problems addressed QYIhe pioneers of water-quality
mod!li.u.g~dissOIY.edQxygel1and ~a<ctcerlaJhe malerial on dissolved oxygen' starts
WIih background on the problem. This is followed by a delailed discussion of the
Streeler-Phelps model. NexI more moder.!U!sI>~tsof Ihe...erobl~msuc~a~nitrifica-
tion I>lanle[fefts, and s~di"!.ent...Q!ygendemand a..r.ecQvered.This is followed by
compuler-orienled models, including a leclure on the QUAL2E software package.
Part V addresses the nexl major phase of water-qualily modeling: eutrophica-
lion. After an overview of the problem and some simple approaches, the remainder
of Ihe seclion f~cuses on season~ll!.utrientJf2Qd-chainl11o<lI':J8.
Because of the impor-
lance of Ihermal stratificalion lo this Iype of model, a couple of lectures are devoted
to heal budgets and temperalure mode1ing.This is followed by a delailed elaboration
of phyloplankton and food-chain simulation.
The following parts focus on the major advances that have occurred in the
1980s. f...a,! VI presenls approaches to simulate I!le chemi§!ry of nalural walers.
Thel~p-grt vil addresses the lopic of loxicsubstanc.e..modeli.og. -
It should be nOledIhal eaeh lecture ineludes delailed examples and is followed
by a set of problems. These problems have been selected lo lest your underslanding
of the material and lo eXlendthe malerial to areas nol covered in the text. I strongly
urge you lo work all Ihese problems. There is an old proverb Ihat slates:
I hear. . . and IJorget
I see. . . and I remember
f do. . . alld I ullderstand.
Nowhere is Ihis saying more relevant than in the area of modeling. AII 100often.
noviee modelers believe Ihat Ihey "know" how lo model because they understand
and can wrile [he underlying equations. Although this is obviously a neeessary pre-
requisile. meehanies must be lempered and refined by Ihe experienee of combining
the model with data. It is Ihis experiential side Ihal is difficult lo convey in a
text. 1 have attempled lo provide a slart in this direelion through the problem
sets. Togelher wilh addilional praetice and a strong scientific understanding of Ihe
20 PARTI Complele1y Mixed Syslems LECTlJREI Inlrodllclilln 21

system and the waler-quality problem, they should prove useful in your efforts to 1 Waste source:
gain proficicncy in modeling. Q =10 MGD (106 gal d-l)
Rlver: =
e 2 mg L-1
A,. =100 ft2
U =1 fps
pROBLEMS e =200 ppb
1.1. How much mass (in g) is eontained in 2.5 L 01'eSluarine waler Ihal has a salt coneen-
tralion 01'about 8.5 ppt? -----, ,
1.2. Each individual in a cily 01' 100,000 people contribules about 650 Lcapila-I d-I 01' ,,
wastewater and 135 g eapita-1 d-I ofbioehemical oxygen demand (BOD). ,,
(a) Determine the flow rate (m3s-l) and the mass loading rate ofBOD (mta) generated
by such a populalion. Instantaneous mlxlng:
(b) Determine the BOD coneentration 01'Ihe wastewater (mg L -1). Q=?cms
c=?ppm FIGURE P1.5
1.3. You are studying a 3-km stretch 01'stream that is about 35 m wide. A gaging station on
the stream pro vides you with an estimate Ihal the average flow rale during your study
was 3 cubie melers per second (cms). You toss a I10al ¡nto Ihe stream and observe Ihat
it takes aboul 2 hr to traverse Ihe slretch. Calculate the average velocity (mps), cross- (a) Calculale Ihe eoneenlration (mg L -1) for Ihe mixlure.
sectional area (m2), and depth (m) 1'01'the slretch. To make these estimales, assume Ihal (b) Delermine the mass in each volume and in Ihe final mixture. Express your result
the strelch can be idealized as a rectangular channel (Fig. PI.3). in grams.
1.7.
You require 4 m3 s -1 01'waler wilh a sah eontenl 111'0.1 g L -1 for irrigalion purposes.
You have Iwo reservoirs from which you (;an draw waler (Fig. PI.7). Rcservoir A has
I a concentmlion 01'500 PPl1l, whereas rcservoir B has 50 ppm. Whm lIow rale I1IUSIhe
H I pumped from each reservoir lo meellhe objeclive?
~ ~"

-
-- -
l.
FIGURE P1.3
A stream stretch idealized as a rectangular channel: L = length, B = width, and
H = mean depth.

1.4. In the early I 970s Lake Michignn had a lotal phosphorus loading 01' 6950 mta and an
in-Iake coneentralion 111'¡¡ ¡.LgL 1 (Chapra and Sonzogni 1979).
(a) Determine the lake's assimilation faclor (km3 yr-I).
(b) What loading rale would be required to bring in-Iake le veis down to approximately
5¡.LgL-1?
(e) Express the results 01' (b) as a percenl rcuuclion, where

n1
d . Wpre""nt - Wru'ure
-10re ucl10n = Wpresenl (100%)
1.5. FIGURE P1.7
A waste source enters a river as depieted in Fig. PI.5.
(n) What is Ihe resulting flow rale in m3 S-I (cms)?
(b) 11'instantaneous mixing occurs, what is Ihe resulting coneentration in ppm?
1.8. You add 10 I1IL01'a glucose Solulion lo a 300-mL bottle and then fillup Ihe remainder 01'
1.6. You mix two volumes 01'water having Ihe following charaelerislics:
Ihe bottle wilh dislilled waler. 11'Ihe glueose solulion has a concenlralion 01' 100 mg L -1,
(a) Whal is Ihe concenlralion in Ihe filled bottle?
Volume1 Volume2 (b) How many grams 01'glucose are in Ihe bottle?
Volume I gal 2L
Concenlralion 250 ppb 1.9. As depicled in Fig. PI.9, many lakes in temperale regions are Ihermally slnllified,
2000 mg m-3
eonsisling 01'an upper layer (epilimnion) and a lower layer (hypolimnion).
 22 PARTI Completely Mixed Systems LECTURE1 lntroduction 23

T 1 km2 and the average depth of the hypolimnion is 5 m, calculate the sediment ftux of
phosphorus needed to cause the observation. Express your results in mg m-2 d-I. In
Epllimnlon your calculation assume that the thermocline forms an impermeable barrier between the
...- - - -- - - -... upper and lower layers. (Note: Although the thermocline greatly diminishes exchange,
_ Thermocllne some transport actually occurs. In later lectures, I show how this can be incorporated.)

1.13. The Boulder. Colorado. wastewater treatment plant enters Boulder Creek just above a
Hypolimnlon z/. USGS (U.S. Geological Survey) now gaging station.

----Á-;----- 1

z Boulder Creek
e
(a) Lake (b) Temperature
A
FIGURE P1.9 11

Onondaga Lake in SyraclIse. New York. had the following characteristics at the end of
I WWTP I
a recenl 'lInllller: ---SSOm-
",f
FIGURE P1.13
Uissolvedoxygcn
Volume concentration
(m") (mgL ') Al 8:00 A.M. on December 29, 1994, condllctivities of 170, 820, and 639 JLmho cm-I
Epilimnion 12 x lO' 8.3 were measured at A, B, and C. respeclively. (Note: Conductivity provides an estimate
Hypolimnion 9 x lO' 1.0 of the lotal dissolveo solids in a solution by measuring its capability to carry an elec-
trical curren!.) If the ftow at the gaging station was 0.494 cms, estimate the ftows for
the Irealment plant and the creek.
Compute the oxygen concentralion following asevere storm that mixes the lake from
1.14. Sediment Iraps are small collecting devices that are suspended in the water column to
topto bottom.
measure the downward ftux of settling solids.
1.10. You musl measure the now in a small brook. Unforlunately Ihe channel is so irregular
ano shallow that you cannot measllre either the velocily or the cross-sectional arca ao-
Flaat
ir equalely. You thereli¡re feeo a conservative tracer with a concentration of 100 mg L-1
inlo the hrook at a constanl rate 01' I L min l. Note that the tracer ooes not occur
.f" naturally in the system. You then go oownstream ano measure a concentration of

rBT
~ 5.5 lI'g L l. Whal is lhe origillalllow mle illlhe creek ill III~ S-I?
1.11. When solids enter a iake 01' impounoment. a portion settles ano collecls in an area
T
called the depositiO/I ZO/Ie. 'fhe rale 01' accumulation is often expresseo as a flux lhal
is calleo the lake's depo.~itioll rateo This rate is calculaleo as Ihe mass deposiled per
T
unit lime nonnalized lo the oeposition zone area. For example approximalely 5 million
lonnes of sediment is oepositeo on lhe boltom of Lake Ontario annllally. 'fhe area 01'
T T
the deposition zone is about 10.000 km2.
(a) Calculate the deposilion rale in g m-2 yr-I .
T T liJJj
,.'''' Trap FIGURE P1.14
(b) If the concentration of suspended solios in the waler is 2.5 mg L -1, oelermine Ihe
~ seltling velocity from the water to the seoimenls, assuming Ihal no resuspension
1i occurS.
Suppose that you suspend a rectangular Irap (1 m x 1 m) at the bottom of a layer of
(e) Suppose that the sediment itself has a porosity of 0.90 (that is. 90% water by vol- water. After 10 d. you remove the trap and determine that 20 g of organic carbon has
1~. lime) and Ihat Ihe inoividual sediment particles have a density 01' 2.5 g em-~. De-
termine the sedimenl burial velO\:ily (that is. the velocily al which Ihe boltom is
collecled on ils surface.
(a) Delermine the oownward nux of organic carbono
:.," being filled in). (b) If the concenlralion of organic carbon in the water layer is I mgC L -1, determine
~ the downwaro velocily of organic carbono
1.12. During \\ ,Irm periods. so me lakes hecl1me thennally malified (see Fig. PI .9). You mea-
tO surc 1I11alphosplll11'11sconccnlralion inlhe hypolimllioll and I¡nd thal the concenlralion (e) If the surface area at Ihe bottom of the layer is 105 m2, calculale how many kilo-
grams of carbon are Iransported across the area over a I-month periodo
'. inL'l'eases fwm 20 ¡.t)!.L 110 100 ¡.t)!.L I o\er a I-monlh periodo 11'the holtom arca is
"

"
 I (
IHTlIRE 2 Rcact¡on Kinctic, 25
i
LECTURE 2
--~--- - r.
""--.,,,-

Reaction Kinetics 88&&

1=0 1=1 1=2 1=3

Time
Q
o 2 3
Concentratlon I <. <, <2 <3

<
Q
.
. . FIGURE 2.1 ',-
A simple experiment to collect rate data lor a
pollutant in a natural water.

l~~~gt $()~ememods for characteri~iqg ~acq()ns iRn~" 2.1.1 Reaction Types

~~YetID'graphica1and cbmputer ~etho~s are'~~Qip~~ llr,
e(ánq'rafe ()f ¡qeaction'! In IlddiUQnJ review' ~~ ffaptiQnj" Heterog~neous reactions involve more than one phase, with Ihe rcaclion usually
", ." . . ., "'t

-
a,Y!tthe eff~t of tegtpefllture.
... 08111......___
As described in Fig. 1.5, a number of things can happen to a pollutant once it enters
a water body. Some of these relate to transportoFor example iu,;1U!.
be tran~lated and
I dispersed by currents within the system. In Ilddition the poUutant can exit the ~)~~
~ \;?yvolatilization, by sedimentation, or by transport along with outftowing wa!er. AII
~ 1 these mechanisms affect the pollutant without altering its chemical composition. In
( contrast the pollutant might be transformed into other compounds via chemical and
( ) biochemical reactions. In this lecture we focus on such reactions.
Suppose that you want to perform an experiment to determine how a pollutant
reacts after it enters a natural water. A simple approach would be to introduce some of
the pollutant into a series of bottles filled with the water. A stirrer could be included
in each bottle to keep the contents well mixed. Such vessels are commonly referred
"¡ to as batch reactors. By measuring concentration in each bottle over time, you would
\ develop data for time and concentration (Fig. 2.1).
The purpose of this lecture is to explore how such data can be employed to
characterize the reactions that affect the pollutant. That is, we will investigate how
to quantitatively summarize (model) the reaction.
2.1 REACTIONFUNDAMENTALS
Before discussing how reactions can be quantified, we must first develop some gen-
~
eral definitions and nomenclature.
~
occurring at the surfaces belween phases. In contrast a homogeneous reaction in-
volves a single phase (Ihat is, liquid, gas, or solid). Because Ihey are Ihe most {un-
damental type of reaction employed in water-quality moaelil]&J!l~ lecture focuses
on homogeneous reactions that take place in the liquid phase. ---
A reversible reaction can proceed in either direction, depending on the relativc
concentrations of the reactants and the products:
aA + bB~cC + dO (2,1)
where the lowercase letters represent stoichiometric coefticients and the uppercase
letters designate the reacting compounds. Such reactions tend to approach an eyui-
librium state where the forward and backward reactions are in balance. They are the
basisforthe area knownas equilibriumchemistry.Wewillreturn to ¡hesetypesof
reactions when we address the topic of pH later in the book.
Although reversible reactions are important in water-quality modeling, ml,re
emphasis has been placed on irreversible reactions. These proceed in a single di-
rection and continue until the reactants are exhausted. For these cases, we are dealing
with the determination of the rate of disappearance of one or more of the substances
that is taking part in the reaction. For example for the irreversible reaction
aA + bB - cC + dO (2.2)
we might be interested in determining the rate at which substance A disappears.
A common example of an irreversible reaction is the decomposition of organic
matter, which can be represented generally by

C6Hu06 + 602 - 6C02 + 6H20 (2.3)

24
 26 PARTI Completely Mixed Systems LECTURE2 ReaClion Kinelics 27

.
I
" where C6H1206is glucose, which can be taken as a simple representalion of organic
I matter. When sewage is discharged into a receiving water, a reaction of Ihis type
c

takes place. The organic matter in the sewage is oxidized by bacteria to form car-
Co
bo~ioxide and water. Although photosynlhesis (that is, plant growth) represent§ a
reverse reaction that produces organic matter and oxygen, 1t does not usually .Qc~r v
C
in the same vicinity as the decomposition. In addition because decomposition an~
pllOtosynthesisare relafivcly slow, they would not come to equilibrium on the tim~ I
, scales of interest in most water-quality problems. Therefore the decomposition is
usuallycharacterize~as a one-wayprocess. -

¡ 2. 1.2 Reaction Kinetics t

FIGURE 2.2 FIGURE 2.3
The kinetics or rate of such reactions can be expressed quantitatively by the law Plot 01 concentration versus time lor a zero- Plol 01 concentralion versus lime lor a
of mass action, which states that the rate is proportional to the c,9l!fentrationoLthe order reaction. lirsl-order reaction.
IJ reactants. This rate can berepresentcdgenerally as
VZero-order. For Ihe zero-order model (n = O),the equation to integrate is
dCA (2.4)
dt = -kf(CA. cn ) dc
1
dt = -k (2.8)
This relationship is called arate law. lt §pecifies that the rate of reaction is
dependent on the product of a temperature-de~dent constant k'and a functio!!.2f where k has units of M L-) T-I. If C = Co at t = 0, then this equationcan be
thecoñceñfniIíoñs
ofthereactanlsf(CA.Cn....). ---- integrated by separation of variables lo yield
- lfieTunclional relatlonshlp f(eA. Cn....) is almost always determined experi- C = Co- kt (2.9)
mentally. A common general form is
As denoted by Ihis equation and the graph in Fig. 2.2, this modeI specifies a constant
dCA a (3
k (2.5) rate of depletion per unit time. Thus, if a plot of concentralion versus time yields a
dt = - CACn straight line, we can infer that the reaction is zero-order.
i) The powers to which the concentrations are raised are referred to as the reaction
urder. In Eq. 2.5 the reaction is a order with respect to reactant A and f3 ordcr with V First-order. For the first-order model the equation to integrate is
respect to reactant B. The overall order of the reaction is L
I
dc
- = -kc
1 dt (2.10)
n=a+f3 (2.6)

The overall order of the reaction, or the order with respect to any individual where k has units of T-I (see Box 2.1). If c = Coat t = 0, thenthisequationcanbe
integrated by separation of variables to yield
component, does not have to be an integer. However, several of the most important
reactions used in water-quality modeling exhibit integer orders. h. c - InCo = - kt (2.11)
;t In Ibis lecture we focus on a single reartan!. For this case Eq. 2.5 is often sim- Taking the exponential of both sides gives
plified as
c = coe-kt (2.12)
dc = -kcn (2.7)
dt ( As denoted by this equation, this model specifies an exponential depletion; that is,
I \ the concentration halves per unit time. Thus, as in Fig. 2.3, the concentration curve
where r: = the concentration of the single reactant and 11= the order. I asymptotically approaches zero with time.

v 2.1.3 Zero-, First-, and Second-Order Reactions :/ BOX 2.1. The "Meaning" o, a Flrst-Order Rate Constant
11,
~~ Although there are an inlinite nUlllher 01' ways to characterize reactions, Eq. 2.7 with You may have noticed that the unils of the reaction rate d'!:pend..ill1
J!1c;u)rdero( thue-
11 = O. 1, and 2 is lIlost coml11only employed in natural waters.
aClion,F~ ~e!()-on1cx reaclion the rate an~Jts W1ltSareeasy t()inte~p!.et,lfsQmeone
 I

28 PARTI Complelely Mixed Syslems

t I.EcrURE 2 Rea,:tion Kinelics 29

stales Ihal a zero-order decay reaclinn has a rale 01' 0.2 mg I,..-I di,. it simply means 1, This relationship can be subslituled into Eq. 2.11 to give
Ihal Ihe subslance is disappearing al a rale 01'0.2 mg L -1 every day.
In contrasl a firsl-oroer rale 01' 0.1 yr-I is nol as slraighlforward. Whal does il I loge -Iogco = -k'l (2.1-'1)
"¡¡¡ean?" A way 10 gain insighl is provided by Ihe Maclaurin series approximalion 01'
Ihe exponenlial funclion: where k' = a "base-lO" rate Ihat is rclaled to thc base-e rate by

e'x
x2
= I - x + - - - + . ..
x3 k' k
, J 2! 3! = 2.3025 (2 15) ,
,
If Ihe series is lruncaled afler Ihe firsl-order lerm, il is Taking the inverse logarilhm 01'Eq. 2.14 yields

1 e-X =:1 - x
¡I
,
r e: = cOlO-k" (2.16)
¡
As depicled in Fig. 82.1, we see Ihal Ihe firsl-onkr approximalinn describes Ihe This equation yields identical predictions to Eq. 2.12.
rale of decrease well 1'01'small values 01'x. Below x = 0.5 the discrcpancy is less Ihan
20%. At higher values the approximaiion- and Ihe Irue vahie diverge.
Although most first-orderrates are wriltcn in terms ofbase-e, somc are exprcssed
in base-lO. Therefore il is importanl to understand which base is being used. Mis-
interpretation would lead to using arate that was incorrect by a factor 01'2.3025
(Eq.2.15).
Approximatlon

Second-order. For the second-order model thc equation to evalllate is

i--- , de: ,
dI = -kL'- (2.17)

where k has units 01'L3 M-J T-J. 11'e: = eo at I = O, Ihen this c'lualion Callbc
integrated by separation 01'variables to yield
o FIGURE 82.1 I
O 0_5 2 Plol 01 Ihe exponenlial lunclíon -e:o + kl (2.11S)
e
x alon9 wilh Ihe li rsl-order Maclaurin-
series approximalion.
Therefore if the reaction is second-ordcr, a plol of l/e versus I should yield a straight
!ine. Equation 2.18 can also be expressed in lerms 01'concentration as a function 01'
This leads us to Ihe following interprelation 01'Ihe "meaning" 01'the firsl-order rale time by inverting it to give
constan!. If its magnilude is less Ihan 0.5, il can loosely be inlerpreled as Ihe fraclion_
'n
1
01'!he poIlutanl Ihal is losl per unil lime. Thus arate 01'0.1 yr-I means Ihal 0.1 01' 10% e = Cn 1 + ke:nl (.! 1'1)
is losl in ayear. If Ihe magnitude nI' Ihe rale is higher Ihan 0.5, a change of Ihe unils
can be used to inlerprel i!. For example a rale 01' 6 d-I clearly cannoll!e inlerpreled as
meaning !hal 600% goes awafpcr day. However, by converting il lo an hourly rale, Thus, as was the case 1'01'Ihc first-order reaction, thc conccntration apprúaches J'ero
in a curved, asymptotic fashion.
FinaIly it should be obvious thal a paltern is emerging Ihat can be employt:d 10
k = 6d'I(2~~r)= 0.25hr'l model higher order rates. Thal is, tUl positive integer values 01'1/,where n -F 1,
we can slate Ihal 25% goes away per hour.
e:,,-I
I_1 + (1/- I)kl (:!.20)
e"-
o
01' solving 1'01' e,
The decay rale used in Eq. 2.12 is caIled a "base-e" rale, because it is used in
conjunclion with the exponenlial funclion to define the depletion 01'concentralion e:= "0 .-.!
(221 i
with time. It should be noted Ihat any base can be employed to describe the same [1 + (1/ - I)h'o-J/III(I/-IJ
trend. For example it should be recognized thal the base-e 01'Naperian logarithm is
related to the base-lOor common logarithm by
2.2 ANALYSIS OF RATEDATA
In x Inx I

10gx = In 10 = 2.3025 (2.13) There are a variety 01'ways to analyze batch-reactordata 01'the type shown in
Fig.2.1. In the presentsection,we wiIIreviewseveraImelhods.Althoughwe wilI
 30 "ARTI Completely Mixed Systems LECfURE2 Reaction Kinetics 31

12 2.5
use Eq. 2.7 as the basis for ilIustrating these teehniques, many of the general ideas
apply to other rate models. .
e
In e 1.5

2.2.1 The Integral Method

10 IS 20 O S 10 IS 20
fhe integral mrthod eonsists of guessing 11and integrating Eq. 2.7 to obtain a fune- I I
li"n. dI). 01aphicalmcthods are then cmployed to determine whether the model fits
(a)
the data adequalely. (b)
The gmphical approaches are based on linearized versions of the underlying
lI1odds. Por Ihe I.ero-orderreaclion, merely plotting e versus t should yield a straight
line (Eq. 2.9). For the first-order reaction, Eq. 2.11 suggests a semi-Iog plo\. These
O..

O.,
} ----
.
.!.
and the other commonly applied models are summarized in Table 2.1. e 0.2

I TAUI,E 2.1 O ,
. lO IS 20
Sumruary (lf the plotting strategy used for applying the integral method to
irn"'crsible. 11l1imolecular reactions FIGURE 2.4
Plots to evaluale whether the reaction is (a) zero-
(e)
Rate Dependent Independent order, (b) first-order, or (e) second-order.
O"ler units (y) (x) Intercept Slope
---
Zc", (1/ = Ü) M(LT) I (' / Cn -k which can be substituted into Eq. 2.16,
['¡rsl (1/ = 1) T-' 111(' / In ('o -k
1/(' I k e = 11.8(10)-0.0422/
Sccolld(1/ = 21 L(MT) , 1/('0
Gcnc",1 (/1 .,¡, 1) (L M')" 'T ' e' " / elo " (11- I)k
The equivalence 01' the two expressions can be illuslraled by computing e at the
----.--.- . same value 01'time.

EXAMPI.E 2.1. INTEGRAL METHOD. Employ the integral method to determine e = 11.8e-00972(5) = 7.26
whethcr the following data is zero-. lirst-. or second-order: e = 1I.8(10)-O.()422(5) = 7.26
... Thus they yield the same resul\.
I (d) o 3 5 10 15 20

«mg L ,) 12 10.7 9 7.1 4.6 2.5 1.8

Ir any 01'Ihese modcls seem to hold, evaluatc k and Co.

2.2.2 The Differential Method
Solution: Figure 2.4 shows plots to evalllale the order 01'Ihe reaclion. Each ¡neludes the The differentialmethodappliesa logarithmictransformto Eq.2.7to give
data alollg with a tlcsl-fit lillc developed with linear regrcssion. Clearly the plot 01' In (' (
de
vcrsw / 'l1osl elosely approximatcs a slraight line. The best-fitline for this case is
In (' = 2.47 - 0.0972/ (r2 = 0.995)
1 ( )
log - dt = log k + IIlog e (2.22)

TherelOlc Ihe estimates 01'the two model parameters are Therefore if the general model (Eq. 2.7) holds, a plot ofthe log(-de/dI) versus log e
k = O.0972d I
should yield a straight line with a slope of 11and an ¡ntercept of log k.
The differential approaeh has the advantage that it automatically provides an
Co = e2.47 = 11.8 mg L-'
estimate of the order. It has the disadvantage that it hinges on obtaining a numer-
Thus Ihe It'sulting model is ieal estimate of the derivative. This can be done in several ways. One of the most
common is based on numerical differentiation.
e = 11. 8e 0.11972'

1110IllOdelcould also be expressed to the base 10 by using Eq. 2.15 to calculate Numerical difTerentiation. Numerical differentiation uses finite-difference
,,'1 k' = 0.0972 = 0.0422 approximations to estimate derivatives (Chapra and Canale 1988). For example a
2.3025
32 PAIH1 Completdy Mixed Systems L.ECnJRE 2 Rcaclioll Kinetics 3.\

EXAMPLE 2.2. DlFFERENTlAL METltOI>. Use Ihe differenlialmelhud tu (;\'al

e
uate the order and the cunstant for the data from Exmnple 2.1. Use equal-area dit'tercn-
tiation to smooth the derivalive estimales.

Solution: The data from Example 2.1 can be differenliated numeriLllly to yield Ihe
Ci+l estimates in Table 2.2. The derivative estimales can be graphed as a bar cllart (Fig. 2.7).
Then a smooth curve can be drawn that best approximates the area under the histogram.
( In other words try tu balance out the histugram areas above and beluw Ihc drawn curve.
) Then the derivative estimates al the data puints can be read directly froll1the cnrve.
These are listed in the last column nfTable 2.2. Figure 2.!; shows a plOl of the log nf Ihe
C¡_I ~--_...... True derivatlve: negative derivalive versus the log nf cnncenlration. The best-tit line for lhis case is

de

1;_1 ti ti+1
FIGURE 2.5
Numerical dilferentiation.
( )
Ing - dt = - 1.049+ 1.0621nge (,.2 = 0.992)

TA8U~ 2.2 )
centered difference can be employed (Fig. 2.5): Data analysis to determine
derivative estimates I"rom time I
series 01"concentration I
dc; = /1c = C;+I - (';-1 (2.23)
1

dI - /11 1;+1-1;-1 Á... de

-Á/ {It
Although this is certainly a valid approximation, numerical differentiation is an t e
inherently unstable operation-that is, it amplities errors. As depicted in Fig. 2.6, be- (d) (mgL ,) (mgL 'd')
cause the tinite differences (Eq. 2.23) are subtractive, random positive and negative O 12.0 1.25
errors in the data are additive. As described in the following example, a technique 1.3
known as equal-area ditTerentiationcan be used to moderatc this problem (Fogler 10.7 1.1
1986). O.X5
3 9.0 O.'!
O.'!)
5 7.1 0.72 I
e e 0.50
10 4.6 045
0..12
15 2.5 0.27
IJ.l.I
20 u¡ 0.15
o o
o o
(a) (e)

FIGURE 2.6
lIIustration 01 how even small
data errors are amplified -~
by numerical dilferenliation.
(a) Data with no error;
(b) the resulting numerical
1 0.5

dilferentiation; (e) data modilied o

~I/o
~If
o
slightly; (d) the resulting
dilferentiation manilesting
o 5 10 15 20
FIGURE 2.7
.
~.
o o increased variability (reprinted t Equal-area differentiation.
(b) (d) Irom Chapra and Canale 1988).
 34 PARTl Completely Mixed Systems LECruRE2 Reaction Kinetics 35
..
n-I
logc 1
.; , Co

~
!
o

-0.25
o
1= --1
kc;;-I(n - l) [( c ) ] (2.26)

If Eq. 2.25 is combined with 2.26, the result is

1 -0.5
2n-1 - 1 1
-0.75
. 150= k(1I - 1) c;;-I (2.27)
FIGURE 2.8
-1
Plot ollog(-dc/dt) versus log c. Taking the logarithm of this equation provides a linear relationship,
2n-1 - l
Therefore (he eslimales of Ihe model paramelers are
11 = 1.062
log 150= log k(11- l) + (1 - n) log Co (2.28)

k = 1O-1.()49= 0.089d-1 Thus a plot of the log of the half-life versus the log of the initial concentration will
Thus Ihe differenlial approach suggesls Ihal a firsl-order model is a valid approximalion. yield a straighl line with a slope of I - 11(providing, of course, that Eq. 2.7 holds).
The estimate of 11can Ihen be used in conjunction with the intercept to evaluate k.
It should be noled that the choice of a half-life is arbitrary. In fact we could
have picked any other response time 1"" where cPis the percent reduction. For this
2.2.3 The Method 01 Initial Rates general case, Eq. 2.27 becomes

There are caseswhere reactions occur in which complications arise over time. For 1 = [100/(100 - cP)]n-1 - 1 ____
(2.29)
( example a significant reverse reaction mighl occur. Further some reactions are very '" k(II-1) céi-I
, slow and the time required for the complete experiment might be prohibitive. For
such cases the method of initial rates uses data from the beginning stages of the
experiment to dctermine the rate constant and order. 2.2.5 The Method 01 Excess
In Ihis melhod a series of experiments is carried oul at different initial concen-
Irations eo. Por each experiment, the initial rate -dco/dl is determined by differ- When a reaction involves many reaclants, it is often possible to add excess quan-
'1 entiating Ihe data and extrapolating to zero time. For Ihe case where the rate law tities of al! but oneof the reactants.In suchcases the reactionwill d.ere.ndsolely
lullows Eq. 2.7, the differential method [that is, a plot of log( -dco/dl) versus log eo] on the single scarce reactant. For example several decomposition reactions forToxic
l'an be used to estímate k and 11.How this is accomplished can be il!ustrated by taking .. Ssubstances(suchas biodegradationand hydrolysis)can sometimesbe represented
1 by the reaction
1helogarilhm 01Ihe negative of Eq. 2.7:
dCo A +B products (2.30)
(- di )
~
log = log k + n log Co (2.24)
where A = the toxic compound and B = another quantity (such as bacteria or hy-
Thus Ihe slopeprovidesan estimateof the order,whereasthe interceptprovidesan drogen ion) that participates in !he reaction. The fol!owing simple rate expression is
often employed to model the reaction:
estimateof the logarithmof Ihe rate.
( dCa
) di = -kCaCb (2.31)
"2.2.4 The Method 01 Hal'-Uves
where Ca and Cb = concentrations of Ihe two reactants. If the initial concentration
, The half-life of a reaction is Ihe time it takes for Ihe concentration to drop to one-half
of B (ChO) is much greater than A (coO), the ensuing reaction can have a measurable
I of its initial value. In other words
effect on A whereas B will be affected minimal!y. Consequently the reaction can be
C (Iso) = 0.5co (2.25) reformulated as
1
where '50 = half-life. Again we use Eq. 2.7 as our rate law mode!. Ir c = Coat dCa
, = O,Eq.2.7canbe integratedto give dI -(kCbO)ca = -kb2Ca (2.32)
 36 PARTI Complelely Mixed Syslems u,nllRE 1 Reac!ioll Killetics 37

.~
where kb2 = kCbU = a pseudo-first-order reaction rateoThe other techniques de- G -T H
I
scribed in the previous sections can then be employed 10 evaluate Ihe coefficients.
0.091528
1,044425
2.2.6 Numerical and Other Methods 5 '"
6 I k1 k2 '3 k4 cp ,cp-cmY'2
Aside from the foregoing approaches, there are computer-oriented methods for eval- 7 o -1.22653 -1.16114 -1.16462 -1.10248 12 12
uating rate data. The integral/least-squares metllod offers the benefits of both the 6 1 .1.10261 -1.04409 -1.04719 -0.99157 10.83658 1;)7 0,018653
integral and differential approaches in a single method. In this approach, values
are assumed for the parameters (11and k) and Eq. 2.7 is solved for c(t). However,
9
10
2.
3
-0.991691

-0.89235
-0.93929
-0.8454'
-0.94206
.0.84768
-0.69225
.0.80325
9.790448
8.849344 O.022ti9í
I
11 4 -0.80334 .0.76127 .0.76347 .0,72346 8.002317
rather than by caIculus, the solution is obtained numerically. The solution consists
12 5 -0.72354 -0.68582 -068779 -06519' 7.239604 1\ o 01948\1
of a table of predicted concentrations corresponding to the measured values. The 6
13 -0.65198 -0.61814 -0.61989 -0.5677 6.552494
sum of the squares of the residuals between the measured and predicted concen- ,. 7 -0.58776 -0.55739 -0.55895 -0.53005 5.933207
trations can be caIculaled. The assumed values 01'11and k are then adjusled until a 15 6 -053011 -0.50283 -0.50424 -0.47828 5.374791

minimum or least-squares condition is reached. This can be done by trial-and-error. 16 9 -0.47833 -0.45383 -0.45508 -0.43175 4.871037
However, modern software lools such as spreadsheets include non linear oplimization 17 '0 -0.4318 -040978 -0.4109 -0.38993 4.416389 4,ij 0033713
19 11 -0.38997 .0.37016 -0,37117 .035231 4,005877
algorithms Ihat provide an automated way lo accomplish Ihe same goal.
19 12 .0.35234 -0.33453 -o 33543 -0,31846 3.635053
The final parameter values represenl the 11and k thal correspond lo a besl-fit of
20 13 -0,31849 .0.30246 .0.30326 .0.28798 3,299934
the data. Thus the technique has the advantage of the integral technique in Ihe sense 21 ,. -0,28801 -0.27357 -0.2743 -0.26054 2,996949
that it is nol overly sensitive to data errors. Further it has Ihe benefit 01' Ihe differential 22 15 -0.26056 '()24756 -0,24821 '()23581 2.7229 25 0,049664
approach in Ihat no a priori assumption 01' reaclion order is required. 23 .6 -0.23583 -0.22411 -0.22469 .().21352 2.474917
17 -0.21354 -0.20297 .0.20349 .019341 2.250426

EXAMPLE 2.3. INTEGRAL LEAST-SQUARES METHOD. Use Ihe integral 18 -0.19343 .0.16389 .0.18436 .0.17527 2.047117
leasl-square melhod 10 analyze Ihe dala from Example 2.1. Use a spreadsheet to perform 26 ,. .0.17529 .0.16668 .0,16711 -0.1589 1862914

Ihe calculalion. I 27 20 .0.15891 .015115 -0,15153 .0.14412 1.695953 lt1 0.0108211

Solution: The sollltion 10Ihis problem is shown in Fig. 2.9. The Excel spreadsheet was SS!I 015500:'
us.:d to perfonn Ihe complllalion. Similar calculalions can be implemenled wilh other
popular packages such as Quattro Pro and Lotus 123. FIGURE 2.9
Inilial guesses for Ihe reaclion rale and order are ente red inlo cells B3 and B4, re- The application 01 the integralleast-squares method to determine the ordel and
speclively, and Ihe time slep for Ihe numerical calculalion is Iyped into cell B5. For Ihis rate coefficient 01 reaction data. This application was perlormed with Ih" Excel
case a eolllmn 01'calclllation times is enlered inlo colllmn A slarling at O (cell A 7) and spreadsheet.
ending al 20 (cell A27). The kl Ihrough k4 coefticienls 01' tllt: fourlh-ord.:r RK melhod
(see Lec. 7 for a descriplion of Ihis method) are Ihen calculaled in Ihe block B7..E27.
I These are Ihen used lo delermine Ihe predicled coneenlrations (the e" values) in eolllmn
F. The measured values (cm) are enlered in column G adjaeenl lo Ihe corresponding pre- 10
dicled values. These are Ihen used in conjunclion wilh Ihe predieted values to compule
Ihe squared residual in column H. These values are sllmmed in cell H29.
\,.,
1'_1,
\ Al Ihis poi nI each 01' the spreadsheets delermines the besl lil in a slighlly differenl e 5
way. Al Ihe lime oflhis book's publicalion, Ihe following menu seleclions would be made
on Excel (v. 5.0), QlIattro Pro (v. 4.5) and 123 for Windows (v. 4.0):
-~

( Exeelor 123: 1(001) s(olver) QI': 1(001) o(plimizer) 10 20

FIGURE 2.10
Plot 01 lit generated wilh the integr¡¡I/least-
t squares approach.
) Once you have accessed Ihe solver or oplimizer, you are prompled for a largel or so-
IUlion cell (H29), queried whelher you wanl 10 maximize or minimize Ihe largel cell
(minimize), and prompled for Ihe cells Ihal are to be varied (B3..B4). You then aClivale
-) 1 Ihe algorithm [s(olve) or g(o»), and Ihe results are as in Fig. 2.9. As shown, Ihe values in
cells 83..B4 minimize Ihe sum oflhe squares oflhe residuals (SSR = 0.155) belween Ihe There are a variety 01'olher approaches for analyzing rate data beyond Ihe ones
predicled and measured dala. NOle how Ihese coefficienl values differ from Examples des('ribed in this lecture. In addition Ihere are olher reactions aside from Eq. 2.7
2.1 and 2.2. A plol of the tit along wilh Ihe data is shown in Fig. 2.10. Ihat are used in water-qualily modeling. The reader should ,'onsult the excellent
 38 "ARTI Completely Mixed Systems LECTURE
2 Reaction Kinetics 39

presentations of Fogler (1986) and Grady and Lim (1980) for additional information. ~ Thus 100g-glucose m-3 corresponds to 40 g m-3 of organic carbon, or 40 gC m-3.
1will be reviewing some additional rate laws in later sections of this texL Methods ~ Such conversions are often expressed as stoichiometric ratios. For example the
for evaluating their rate constants will be reviewed in a "just-in-time" fashion as mass of earbon per mass of glueose can be expressed as
they are needed.
arg -_ 6 moles C x 12 gC/mole C -_ O4
. gC g-g1ueose
_1
(2.36)
180 I g-g ueose
2.3 STOICI'¡IOMETRV
where tleg = stoichiomelric ratio of carbon to glueose. This ratio can be used to
111the previolls leeture 1 inlroduced Ihe 110lionof mass cot1centration as a means to formulate Eq. 2.35 alternalively as
qmllllify the strength of a single chcmical eompound in water. Now thal we are deal-
= acgc¡: = 0.4-g-ggC g-g(ucose
il1~ with Icactiol1s, several compoul1ds llIay reaet lo form olher eom~unds. There-
fore we mighl \Vant to delermine "how much" of a reaclant or product is consumed
01' erealed "S Ihe reaclion proeeeds. The answer lO this questinn resides in the .~toi-
Cc -
1ucose 100 m.
( = 40gCm .1
) -3
. (2.37)

cf¡imnetry, 01 1111mberof moles, laking part in a rcaetion. where the subseripts e and g designale carbon and glucose, respeetively.
For exal1lple Ihe decomposition 01'oxidation of sugar is represented by (recall Aside from ealculating how much individual elemenl is eonlained in a molecule,
I~q. 2.1) sloichiometric conversions are often used lo delermine how mueh of a reactanl or
C(,H 1206 + (¡02 --+ (¡C02 + (¡H20 (2.33) prodllct is eonsumed or prodllced hy a reaetion. For example how much oxygen
would be eonsumed if 40 gC m .1of glucose reacled according lo Eq. 2.33? First, we
rlJis equat;on specifies Ihat 6 moles of oxygen will reaet with I mole of glucose to can calculate the mass of oxygen consumed per mass of glucose carbon decomposed,
fonn 6 moles of carbon dioxide and 6 moles of water. In later leetures we directly use
molar concentrations when we mathemalically manipulate such equations lo solve
roe = 6"tnnlnc-"""
moles O2 vx 32 gO/mole
~~J I_'"O2 = 267
. gO gC-I (2.38)
chemical equilibrium problems. For Ihe lime being, as outlined in the previous lec-
IlIn~. we mllsl be able to inlerpret Eq. 2.33 from a mass-concentralion perspective.
Firsl, kt's undersland how the glucose in Eq. 2.33 would be expressed in mas s where roe = mass of oxygen consumed per carbon decomposed. This ratio can be
used to delermine
lInits. This is lIslIally done in two ways. The mosl direct way is to express Ihe concen-
Iralion 011the oasis of Ihe whole moleeule. For example we might say that a beaker
eontained 100 g m .1as glucose. This is often abbreviated as 100 g-glucose m-3.
/ The number nf moles of glucose in this solulion can be determined wilh the gram 2.67:~ (40;;)= 106.67 gO m-3 (2.39)
111(\lecu(;\I'\Vcight of glucose. The gram moleculal weight can be calculaled as
Thus if 40begC
m-3 will m-3 of glucose (or 100 g-glucose m-3) is decomposed, 106.67 gO
consumed.
Number Mass of
of moles une mole

6xe = 6 x 12g = 72g EXAMPLE 2.4. STOICHIOMETRIC RATIOS. Aside fmm Ihe deeomposiliol1 of
12 xII = 12 x Ig = 12g organie carbun compounds such as glucose, olher reacliuns consume oxygen in nalural
6xO = (, x t6g = 96g
1 ~I waters. One. such pmcess, called nitrijico.tion, in~olves Ihe conv.e~sion. of ammonium
- (NH4.) lo IItlrale (N03 -). Allhough we wtlllearn m Lec. 23 that tt s a httle more com-
Oram molecular weight = IROg plicaled, the nilrifieation reaclion can be represenled by

NH4+ + 202 -> 2H+ + H20 + N03-

This resull can be used to compute the molar coneenlration.
I mole Suppose
lo you are reactíon
the firsl-order lold Ihat a beaker conlains 12 g-ammonium m-3 is nilrified aceording
= 0.556 mole m-3
100g-glueose
m3 (
180 g-glueose ) (2.34)

dna = -k,,111l
An allernative is lo express Ihe eoncenlration in terms of the mass of one of the di
componenl~of glucose. Beeause it is an organic earbon compound, glucose eould be
expres~cd ¡ISg m-3 of carbonoFor example where /la = ammonium concenlralion and k" = firsl-order rale conslant for nilrificalion.
(o) Converllhe concentralion to gN m-3. (b) Determine how much oxygen is consumed if
6 moles C x 12 gC/mole C = 40 C m-3
loog-glucose
m3 ( 180 g-glueose ) g
(2.35) Ihe nilrification reaclion goes lo complelion. (e) Calculale Ihe rale ofoxygen consumption
allhe start of the proeess (k" = 0.1 d-I).
4.
'.

40 PART1 Complelely Mixed Syslems

l.ECTURE2 Reaclion KinelÍCs 41

Solution: (a) TABLE2.3

gNH4+
12(jj3 (x1
Ix 14gN
)
14 + 4 xl gNH4+ - 9.33 gN m
2 x 32
_

_1
-3
-Some typical values of (}used in water-
quality modeling
(J Q.o Reaction
(b) ron = I x 14 = 4.57gOgN 1.024' 1.27 Oxygen reaeration
Therefore 4.57 gO are taken up for every gN thal is nilrified. For our example. 1.047' 1.58 BOD decomposition
1.066 1.89 Phytoplankton growth

)_ -3
gN{ gO 1.08 2.16
Sediment oxygen demand (SOD)
9.33m3{57 gN - 42.67gO m
(e) At Ihe onset of the experimento Ihe ammonium concentration will be al 9.33 gN
m-3. Using the oxygen-Io-nitrogen ratio. the initial rate of oxygen consumption can be k(T2) = OT,-T1 (2.4 J)
ca1culated as k(T1)

)í\9.33 gN
do gO -I -3 -1 where the temperature is expressed in 0C.
dt = -ronknn. = -4.57 gN 0.1 d ( m3 )= -4.264 gO m d In water-quality modeling. many reactions are reported at 20°C (see Probo2. 16).
Therefore. Eq. 2.43 is usually expressed as
k(T) = k(20)OT-20 (2.44)
2.4 TEMPERATURE EFFECTS
( Table 2.3 summarizes some commonly used values for O.Figure 2.11 iIIustrates
The rates of most reactions in natural waters increase with temperature. A general ) the functional dependency on temperature across the range commonly encountered
\ rule of thumb is Ihat the rate will approximately double for a temperature rise of ) in natural waters.
1 10°C. The temperature dependence of biologically mediated reactions is offen ex-
pressed as the quantity QIO.which is defined as the ratio
A more rigorous quantification of the temperature dependence is provided by
the Arrhen;us equation, k(20)
-E (2.45)
QIO = k(lO)
k(Tu) = AeRTu (2.40)
Substituting Eq. 2.44 yields
where A = a preexponential or frequency factor
E = activation energy (J mole-I)
QIO = 010 (2.46)

R = the gas constant (8.314 J mole-I K-1) Note that Eq. 2.46 can be used to compute that a Qw of 2 (recall the heuristic at lhe
Tu = absolute temperature (K) beginning of this section) is equivalent to a O of 2°.1 = 1.072. Thus a O = I.072
corresponds to a doubling of the rate for a temperature rise from 10 to 200C.
Equation 2.40 is often used to compare the reaction rate constant at two different
temperatures. This can be done by expressing the ratio of the rates, as in
a= 1.1
k(Tu2) E(Tu,-T"IJ 4
- = e RT",Tul (2.41)
k(Tu.>
--. a = 108
=>
Equation 2.41 can be simplified by realizing that: M
'-' a = l.06ó
. Becat!'.. temperatllres in most water bodies vary over a rather narrow range (273 2 a= 1.047
to 31..>,,1, the pruducl of Tul and Tu2 is relatively constan\.
f."
'-'
..:c a= 1.024
. The ditlerence in temperature (Ta2 - Ta.>is identical whether an absolute or a -- --- a= I
centigrade scale is used.
o .f:,.......................
Consequently the following can be defined as a constant: o 10 20 30
T(°C)
1 (} == eRT~Tu1 (2.42)
FIGURE 2.11
and Eq. 2.41 can be reexpressed as
The effect 01 temperature on reaction rate ler various values 01 /J.
 42 PART1 Completely Mixed Systems LECTURE2 Reaction Kinetics 43

EXAMPLE 2.5. EVALUATION OF TEMPERATURE DEPENDENCY OF RE- T("C) 4 8 12 16 20 24 28

ACTlONS. A laboratory pro vides you with the following results for a reaction: k (d-') 0.120 0.135 0.170 0.200 0.250 0.310 0.360
TI = 4°C k¡=0.12d-1
T2 = 16°C k2 = 0.20 d-I Use this data to estimate O and k at 20°C

(a) Ev¡¡llIale O for chis reaction. 2.5. An article in a lil1lnologicalt jOllrnal reports a QIO for a phytoplankton growth rate of
(b) Dclcnnine Ihc lalc al 20°C. 1.9. If Ihe growth rale is reported as 1.6 d-I at 20°C, what is the rate at 300C?
Solutloo: (o) To evaluate this information, we can take the logarithm of Eq. 2.43 and 2.6. You set up a series of 300-mL bottles and add 10 mL of a glucose solution to each.
raise the result to a power of lOto give
Note that the glucose solution has a concentration of 100 mgC L -l. To each bottle you
'og k(T2)-log k(T,) add a small quantity (that is, with an insignificant amount of carbon compared to the
O = 10 T2-T, glucose) of bacteria. You fill the remainder of their volumes up with water. Finally you
seal each bottle and incubate them at 20°C. At various times you open one of the bottles
Substitllling the data gives
and measure its oxygen conten\. The following data results:
.lugO.12-lngO.20
O = 10 4-16 = 1.0435
>
20 30 40 50 60 70
(b) Equ¡¡tion 2.43 can then be used to compute
eI (d)
(mg02 L-') I 10
O 8.4
2 6.5
5 4.4
10 2.3 1.6 1.3 1.2 1.1 1.1
k(20) = 0.20 x 1.04352()-\6 = 0.237 d-1

(o) Develop a conceptual model for what is taking place inside the bottle.
Finally il should be noled Ihat Ihere are some reactions Ihat do not follow the (b) Using the information from this lecture, atlempt to estimate the decay rate for the
Arrhenius equation. For example certain biologically mediated reactions shut down glucose.
at very high and very low lemperatures. The formulations used in such situations are
inlroduced in later lectures. 2.7. In the fall of 1972 Larsen et al. (1979) measured the following concentrations of total
I phosphorus in Shagawa Lake, Minnesota:

I
PROBLEMS Day mg m-J Day mg m-J Day mg m-J
I
r 250 97 270 72 290 62
,
I 2.1. You perform a series of batch experimencs and come up with the following data:

4 6 8 10
254
264
90
86
275
280
51
57
295
300
55
46

Ic\J.LgL
(hr) ') I O10.5 25.1 3.1 2.8 2.2 1.9
It is known that the primary reason for the reduction in concentration during this period
Determine the order (11) and the rate (k) of the underlying reaction. was the settling of particulate phosphorus. If the lake is assumed to act as a batch reactor
and settling is assumed to follow a first-order process, determine the removal rate of
2.2. Derive a graphical approaeh to determine whether a reaction is third-order. total phosphorus for the lake. If the lake's mean depth is 5.5 m, calculate the settling
velocity for total phosphorus.
2.3. To study the photodegradation of aqueous bromine, we dissolved a small quantity of
I¡quid bromine in water, pla('ed it in a clear jar, and exposed it to sunligh\. The following 2.8. Population dynamics is important in predicting how human development of a water-
data \Vere obtained:
shed might inftuence water quality. One of the simplest models incorporates the as-
sumption that the rate of change of the population p is proportional to the existing
I (min) 10 20 30 40 50 60 population at any time 1:

L" (ppm) 3.52 2.48 1.75 1.23 0.87 0.61

dp
di = Gp (P2.8)
Determine whether the reaetion is zero-, first-, or seeond-order and estimate the reaction
rateo

2.4. At a later date the laboratory informs you that they have a more complete data set than t Limnology is the slUdy of lakes. The terminology is derived from the Greek word for lake: limllo.,.
the t\Vo mcasurel1lenls lIsed in Exal1lple 2.5:

Ai
44 PARTI Complelely Mixed Syslems 'HTl'lH' 2 Reaelion Kinelies 15

where G = the growth rate (yr-I). Suppose that eensus data provides the following Time (min) o 20 50 65 150
trend in population 01'a small town over a 20-yr period: NH.OCN (mole L.') 0.381 0.264 0.180 0.151 (1.086

Determine Ihe order and rate 01'the reaelion.

;--1
P 100
1970 212
1975 448
1980 949
1985 2009
1990
2.15. You perform a baleh experiment and develop the following data:

If the model (Eq. P2.S) holds, estimate G and the populalion in 1995.

2.9. The world took about 300 years to grow from about 0.5 bíllion to 4 billion people.
Assumillg first-order growlh, determine the growth rale. Estimale the population over
;-j
e 10.0
O 8.5
2 7.5
4 6.7
6 6.2
8 5.8
10

the next eentury if this rate eontinues.

You know from experience that the reaclion should be following a third-order reaelioo
2.10. Many lakes in temperate regions are thermally stratified in the summer, eonsisting 01'an Use this information and Ihe integral melhod lo delermine a vallle 1'01'the reaclion rale.
upper layer (epilimnion) and a lower layer (hypolimnion). In general the surface layer 2.16.
has dissolved oxygen eoneentration near saluration. If il is produetive (that is, has high SlIppose that the lemperature dependenee 01'a reaetion rale is based on irs vallle at 25"C
plant growth), settling plant matter can collect in Ihe hypolimnion. The decomposition (note that this is Ihe convenlion in areas sllch as chemical engineering). For examplc
of this matter can then lead 10 severe oxygen depletion in the bottom waters. When k(T) = 0.1(1.06/,-25
turnover oeeurs in the fall (that is, vertical mixing due lo deereasing temperature and
Reexpress Ihis relalionship based on Ihe rale 's value al 20°C.
increasing winds), the mixing ofthe two layers can result in the lake's having an oxygen
eoneentration well below saturalíon. The following data were eollecled 1'01'Onondaga
2.17. The following dala 1'01'concentrations and limes were developed 1'01'a series 01' balch
Lake in Syraeuse, New York: experiments having differenl inilial conditions:

Date Sep. 30 Oel. 3 Ocl.6 Oel. 9 Oel.12 Oel. 15 Oel. 18 <;>el.

21 . ('

Oxygen conc. 4.6 6.3 7.3 8.0 8.4 8.7 8.9 9.0 O 1.00 2.00 5.(){) 10.00
(mgL-I) I 1 0.95 1.87 4.48 8.59
2 0.91 1.74 4.04 7.46
Iflhe saturation eoncentration is 9.2, use Ihis data to evalllate a tirst-order reaeration rale
1'01'the syslem (units 01' d-I). Assume Ihal the lake aets as an open batch reactor; thal Assuming tha! Eq. 2.7 holds, use the method 01' inilial rales lo determine Ihe order alJ(l
is, ignore intlows and OllttlowS 01'oxygen except gas transfer across the lake's surfaee. rate of Ihe reaetion.
Also, express the rate as a transfer velocily (unils 01'm d-I). Note thal Onondaga Lake
has a surfaee area 01' 11.7 km2 and a mean depth 01' 12 m. 2.18. Assuming Ihal Eq. 2.7 holds, use the method 01'half-lives to delermine Ihe reaclion 01'-
del' and ratc by evalualing Ihe following half-lives and inilial eoncentra!ions devel!>!,,,,1
2.11. A reaction has a QIO of 2.2.11' Ihe reaclion rate at 25°C is 0.853 wk-', whal is the from a series 01' balch experiments:
reaction rate at 15°C?

2.12. A comrnonly used anesthetic is absorbed by hllman body organs at a rale proportionallo
its eoncentralion in Ihe bloodstream. Assume that a palient requires 10 mg 01'the anes- ;:-j1251O
1",16 I1 7 5
thetie per kg 01'body weight to maintain an aeceplable level 01'anesthesia 1'01'surgery.
Compute how many mg must be administered to a 50-kg palient lo mainlain a proper
level 1'01'a 2.5-hr operation. Assume that the aneslhetic can be inlroduced into Ihe pa- 2.19. Assuming Ihat Eq. 2.7 holds, use the inlegralleast-squares method 1" d,,,"lIIine lb"
lien!'s bloodstream as a pulse input and that it deeays at a rale of 0.2% per minule. reaction order and rale by evaluating Ihe following data collecled from a balch expcri-
ment:
2.13. Estimate Ihe age ofthe fossil remains 01'a skelelon with 2.5% of its original earbon-14
e ienL. Nole Ihat earbon-14 has a half-life of 5730 yr.

2.14. In 1828 Friedrich Wohler diseovered that the inorganie salt ammonium eyanale
(NH4OCN) can be con verted into the organic eompound urea (NH2CONH2), as in ;-j
e O
10 2
7.5 4
5.8 6
4.6 8
3.8' 10
3.1

NH40CN (aq) -> NH2CONH2 (aq)

~
2.20. The eoncentration 01' inorganie phosphorus in natural waters is uSllally expressed as
The proof that this reaction oeeurred marked Ihe beginning of modern organie and bio-
ehemistry. An investigator has reported the following data 1'01'an experiment initially phosphorus (P). However, it is somelimes expressed as phosphale (P04). When reading
a scientifie artic1e, you see that an estuary has an inorganic phosphorus eoncentration 01'
eonlaining apure solution of ammonium eyanate: ,<
,.
t
46 PARTI Completely Mixed Systems

LECTURE 3
10 mg m-3. As is sometimes the case no guidance is given regarding how the concen-
tration is expressed. How does the concentration change if it is actually mgP04 m-37
By what factor would you be off?
2.21. A more complete representation 01'the decomposition reaction is provided by Mass Balance, Steady-State Solution,
Cuw.1h".OlwNu.P, + 10702 + 14H+ -> I06C02 + 16NH4+ + HP042- + I08H20 and Response Time
In contrasl to the simplified version in Eq. 2.3. Ihis reaclion renecls that organic matter
conlains Ihe nulrienls nitrogen (N) and phosphorus (P). On the basis 01'this equation,
given that 10 gC m 301'organic nmttcr is decomposcd. calculate
(a) thc stoichiometric ralio for the amount 01'oxygen consumcd pcr carbon dccom-
poscd, "",. (gO gC 1)
(b) the amounl 01'oxygen consumed (gO 10-3)
(e) the amount 01'ammonium released (expressed as mgN m-3)

LECI1JRE OVERVIEW: 1introduce the primary organiiing ~rincipl: ófwa~r:J¡f[~~'

modeling: the mass balance. It is used to deri:vea water-qua1i~ tt).gtletJÓt:
Simple case of a continuously stirred tank reactor, 0t CS~. tdevetiJ
terptet a $teady-state solUtioh and present the notions of trahsf~t futic"
residence times. i also derive the generaÍ time-variáble solUUah..of.t
balatlce.lt1odel and calculate a response time to qúantlfy tetnporál ch.
ticsoffuerecoveryofsuchsystems. '

Now that we have reviewed some fundamental concepts, let's tie them together and
actually develop a water-quality mode!. Then we will solve the model to answer the
two most commonly posed questions in water-quality modeling: If we institute a
treatment program,

· How much will the water body improve?

· How long will it take for the improvement to occur?

3.1 MASS BALANCE FOR A WELL-MIXED LAKE

A completely mixed system, or continuously stirred tank reactor (CSTR), is among

the simplest systems that can be used to model a natural water body. It is appropri-
1 at~lo!: 1!.!ece.ivjngwater in wl1ichthe contents are sufficiently welf mixed as 10 I?,e
uniformly distributed. Such a characterization is often used to model naturallakes
and some impoundment;. -
A hypothetical completely mixed system is depicted in Fig. 3.1. Note that 1
have inc1uded a number of sources and sinks that are typically encountered when

47

"
 I.ECTUKE3 Mass Balance, Slcaoy-Slale Solution, ano Response Time 4')
48 PAKTI Complelely Mixed Syslems
~('
Accumulalion = V-~t (3.6)
Loadlng

Finally ~t can be made very small and Eq. 3.6 reduces lo

FIGURE 3.1
r de
, ,1 Reactlon
, ,
A mass balance lor a well-mixed lake. The
arrows represenl Ihe major sources and
sinks 01 Ihe pollulant. The dashed arrow
i Accumulation = V -ct , (3.71

... lor Ihe reaclion sink is meanl lo dislinguish Thus mass accumulates as concentralion increases with time (positive de/dI) ami
il Irom Ihe olher sources and sinks, which diminishes as concenlralion decreases wilh time (negative dc/dl). For the steady-
Settllng are Iransport mechanisms.
( ~ state case, mass remains constant (de/dt = O).Note that the units of accumulation
(as with all other terms in the balance) are mass per time (M T-1).
modeling water quality. For a finite time period the mass balance for the system can
be expressed as Loading. Mass enters a lake from a variety of sources and in a numbcr of differ
Accumulation = loading - outftow - reaction - settling (3.1) ent ways. For example mass carried by treatment plant efftuents and tributary streams
enters a lake at a point on its periphery. In contrast atmospheric sources, such as pre-
Thus there is a single source that contributes matter (Ioading) and three sinks cipitation and dry fallout, are introduced in a distributed fashion íICrossthe air-wílter
that deplete matter (outftow,reaction, and settling) from the system. Note that other ( interface at the lake's surface. Whereas the position and mílnner of entry of load-
sources and sinks could have been included. For example volatilization losses (that I ings would have fundamental importance for incompleteIy mixed water bodies such
is, transfer of the pollutant from the water to the atmosphere) could exit across the 1 ~eams and estuaries, it is unimportant for our completely mixed systcm. This is
lake's surface. However, for simplicity, we limit ourselves to the sources and sinks because, by definition, all inputs are instantaneously distributed throughollt the vol-
depicted in Fig. 3.1. Sume. Thus, for the present case, we lump allloadings into a single term, as in
Although Eq. 3.1 has descriptive value, it cannot be used to predict water qual- )
ity. For this we must express each term as a function of measurable variables and
Loading = W(J) (3.11)

parameters. where W(t) = rate of mílssloading (M T-1 ) and (t) signities thíltloading is a function
~tim~ -
Accumulation. Accumulation represents the change of mass M in the system It should be noted that in íI later part of this lecture we formulate loading in a
over time t: slightly different fashion than in Eq. 3.8. Rather than as íIsingle value W(t), we will
~M (3.2) represent it as the product (recall Eq. 1.3)
Accumulation = -¡;¡
Loading = QCin(t) (3.9)
Mass is related to concentration by (Eq. 1.1) where Q = volumetric now rate of all water sources entering the system (L3 TI)
M (3.3) and Cin(t) = average inftow concentration of these sources (M L-3). No\<:lbat we
C =-
V <;have assumed that ftow is constant and that all the temporal variations in loading
f 1are the resuhof temporalvariíltionsin the inllowconcentration.Alsorecognizethat
where V = volume 01'system (L3). Equation 3.3 can be solved for average inftow concentration can be reIated to loading by equating Eqs. 3.8 and 3.9
M = Vc (3.4) and solving for

which can be substituted into Eq. 3.2 to give (';,,(t) = W(t) (3.10)
. ~Vc Q
l
AccumuaUon= -¡;¡ (3.5)
OutUow. In our simple system (Fig. 3.1) mass is carried from Ihe system hy
an outftow stream. The rate of mass transport can be quantitied as the product of
11n the presentcasewe
allowsus assumethat
to bringtheterm the lake's
V outsidethe volumeis constant.tThis assumption
difference: the volumetric ftow rate Q and the outftow concentration ('''UI(M L-3). But, becallse
~of our well-mixed assumption, the outflow concentration by detinition equals the
I l in-Iake concentration COUI= c, and the outftow sink can be represented by
~ ' Although Ihis is a good assumption in many cases, il is nOI always true. For example many lakes and
= QC
t- 1Ihese
impoundments
can exhibitin significanl
the weslem changes
Uniled
Slates
in
how to model such syslems in Lec. 16.
Iheir are used for
volumes overpower production
relalively and waler
short time periods.supply. Some of
We willleam \ Outflow (3.11)
 50 PART I Completely Mixed Systems LECTURE3 Mass Balance, Steady-State Solution, and Response Time 51

Reaction. Although there are many different ways to formulate reactions that
purge pollutants from natural waters, the most common by far is a first-order repre-
sentation(recallEq.2.10) - -
R.eaction = kM (3.12)
,, where k
I ==a first-order reaction coefficient (T-1). Thus a linear proportionality is
í assumed bctween the rate at which the pollutant is purged and the mass ofpoTIu-
tant that is present. El\uation 3.12 can be expressed in terms of concentration by
substituting Eq. 3.4 into Eq. 3.12 to yield
Rcaelion = kVe (3.13)
Settling. Settling losses can be formulated as a flux of mass across the surface
arca of the sediment-water interface (Fig. 3.2). Thus by multiplying the flux times
arca, a term for seuling in the mass balance can be developed as
" Sculing = \'A.,c
Before proceeding to solutions for Eq. 3.16 we should introduce some nomencla-
~ ture. Concentration e and time t are the depelldent and the ;ndependent variables,
- respectively, because the model is designed to predict concentration as a function
~ of time. The loading term W(t) is referred to as the model'sforc;ng fUlIstion 00-
i cause it re~ the way in which Ihe exter!!al world influences or "foTces"the
~'§ystem. Finally the quantities V, Q, k, v, and As are referred to as parameters or
coefficients. Specification of these parameters will allow us to apply our model to
particular lakes and pollulants.
~BOX 3.1. Parameterlzation
As described in this section, settling losses can be parameterized as the product of
settling velocity, surface area, and concentration vA..c. However, as in Eq. 3.15, the
settling mechanism can also be "parameterized" as a first-order rateo This is done by
(3.14) multiplying the seltling-velocity version by HIH and collecting terms to yield

I where v = apparent seuling velocity (L T-1) and A., = surface area of the sediments H v
(L2). The seuíing velocity is called "apparent" because it represents the net effect vA..c¡¡ = ¡¡(AH)c = ksVc
()f the various processes that act to deliver pollutant to the lake's sediments. For
example some of the pollutant may be in dissolved form and hence not subject to where k.. = a settling first-order rate constant (T-I) that is equal to
seJtling. For such cases a "real" seuling velocity cannot be used to represent the net v
cffect of this mechanism. ks = ¡¡
Because volume is equal to the product of mean depth H and lake surface area
A.f, Eq. 3.14 can also be formulated in a fashion similar to the first-order reaction, This alternative formulation has been commonly employed in water-quality mod-
eling.
as in
Now the question arises, is either way superior? From a strictly mathematical
Seuling = kJVc (3.15) \standpoint they are identical. However, because it is more fundamental, the settling
where k.. = a first-order seuling rate conslant = vlH. Notice that the ratio vlH has f, ¡velocity parameterization is superior. By fundamental I mea n that it more directly rep-
,resents the process being modeled. That is, each term in vA..e represents a characteristic
Ihe same units (T-1) as the reaction rate k. The validity of this representation is of the process that can be measured independently. In contrast the k.. tenn confounds
contingellt on the assumption thl1tthe l11ke'ssurface area and the sediment area are two¡ndependentprop~rties:settlingand deplh. _. __o
equal. The format of Eq. 3.14 is preferable to Eq. 3.1S because the forrner more (' Why is Ihis a problem? Firsl, Ihe k.. version is syslem specific (because il implicitly
faithfully captures the mechanistic nature of settling, that is, as a mass transfer across )includes a syslem-specific property, mean deplh) and hence is awkward.!o extl!lpoJ~!e
a surface (see Box 3.1). lto other systems.lf we measure a ks in a particular system, we could use it only in other
systems of the same deplh. Thus, to eXlrapolale to a system wilh a differenl depth, we
Total balance. The terms can now be combined into the following mass balance would have to revert lo the settling velocity formal anyway. Second, what if depth is
changing? For this case Ihe use of k.. clearly breaks down.
fOl a well-mixed lake:
Ii ~ Now, where might confounding parameters be advanlageous? For one thing,
within a malhematical calculalion for a particular system, we often find it useful to
Vde W(t) - Qc - kVe - vA.fe (3.16)
I 1
dt S collect lerms for mathemalicalconvenience.Second.ilis often of use to collect lerms
I l so that processes can be compared in commensurate units. For example Ihe relative
magnitudes of settling and a reaclion could be assessed by comparing vI H versus k.
Sedlment-water
J = ve Finally Ihere are some instances where we might confound several paramelers because
j one or more do not vary between syslems and/or Ihey are difficult to measure.
Interface'\ \ Throughout Ihe remainder of Ihis book Ihe issue of proper paramelerization will
AJ FIGURE 3.2 continuously arise as we attempl to quanlify the processes observed in natural walers.
When il ariseswe willdiscussils furlhernuances.
Settling losses lormulated as a flux 01 mass
across the sediment-water interface.

- .' - .~"
I --'--.

52 PART 1 Complelely Mixed Syslems

LEnlJRE.J Mass Balance, Sh:aJ)- "Slate Solulion. allo Response Time 53
3.2 STEADY-STATE SOLUTIONS
(b) The surfaee area 01'Ihe lake is lIeeded lo calclllall' Ihe atmospheríc loadillg
If tbe system is subject to a constant loading W 1'01'a sufficient time, it will attain
a dynamic equilibrium condition called a steady-state. In matbematical terms tbis A.
.
=~
H
= 50,000
2
= 2500011I2
'
S means tbat accumulation is zero (tbat is, dc/dt = O).For tbis case Eq. 3.16 can be
The atmospherie load is Ihell eompuled as
1 solved for
W Warm",phcrc
= J A.. =
c= (3.17) 0.6(25.000) = 15.()(X) g d I
Q + kV + vAs
The load from Ihe inflow slream can be ealculaled as
01'using tbe format of Eq. 1.8,
Winllow= 7500(10) = 75,Q()() g d
1
e = -W (3.18) Therefore Ihe 10lalloading is
a
wbere tbe assimilation factor is detined as W = Wfacrory+ Wa1m",phcl~
+ Winlluw= 50. (XX)+ 15.000 + 75.000 = 140,000 g o

a = Q + kV + vAs (3.19) and Ihe eoneenlralion can be delermilled as (Eq. 3.18)

1 1
Tbe steady-state solution provides our tirst iIIustration of tbe benetits of tbe c = - W = -140 ()()O= 597 mg L
a 23.454' .
mecbanistic approacb. Tbat is, it bas suceessfully yielded a formula tbat detines tbe
assimilation factor in terms 01'measurable variables tbat reflect tbe system 's pbysics, (c) The loss due lo lIushing Ihrough Ihe Ollllel can be compuled as
cbemistry, and biology.
Qf..= 75(X)(5.97) = 44.769gd-1
EXAMPLE 3.1. MASS BALANCE. A lake has Ihe following eharaelerislies: and Ihe loss due lo reaclion as
Volume = 50.000 m3
Mean deplh = 2 m kVc = 0.319(50.000)5.97 = 95.231 gd'l
Inflow = outflow = 75()()m3 d-I
Temperatllre = 25°C These resulls along wilh Ihe loading can be displayed as in Fig. 3.3.

The lake reeeives Ihe input 01'a pollulanl from Ihree sourees: a faetory discharge 01'
50 kg d-I. a flux from Ihe atmosphere 01'0.6 g m-2 d-I, and the inflow stream thal has
Tbe representation in Fig. 3.3 can now be related back to the parable of lhe
a eoneenlration 01' 10 mg L -l. 11'the pollulant deeays al the rate 01'0.25 d-I at 20°C
(8 = 1.05), blind men and the elepbant. Eaeb arrow, representing a souree 01'sink l' ,..hanislII,
is analogous lo Ibe individual parls of Ihe elephant. 11is only wben ti" y "fe licd
(a) Compute the assimilation factor.
(b) Determine Ihe steady-state eoneenlralion.
Factory
results on a plot. W -
(L') Calclllale Ihe mass per time 1'01'eaeh lerm in the mass balance and display your Inflow
loading
75 kg d -1
loading
50 kg d -1
(35.7%)
Atmospherlc
loadlng
15 kg d -1
(10.7%)
Outflow
Solution: (a) The deeay rate must first be correeled 1'01'lemperalure (Eq. 2.44): (53.6%) 44.8 kg d -1
(32%)
k = 0.25 X 1.&525-20
= 0.319d-1
FIGURE 3a ~.
Then the assimilation faclor can be ca1culated as A mass balance 101'the well-

a = Q + kV = 7500+ 0.319(50.000)= 23.454m.! d-I I mixed lake lram Example

I
- 3.1. The arrows represen!
the major sourctls and sinks
NOlice how Ihe units look like flow (that is, volume per time). This is because the same 01 the pollutant. The mass-
mass units are used in the numerator and the denominator and they cancel, as in
--... transler rates have also
I

Reaction been included along with

g d-I 95.2 kg d -1 the percent 01 total mass
g m-3 -+ m3d-I (68%) in!low accounted 101'by
each termo
 54 I'ARTI Complelely Mixed Systems
togclher by lhe mass balance lhal we can assess lheir combined effecl. Thus lhe
model provides an inlegraled view of lhe syslem.
~ BOX 3.2. "Stream-of Consclousness" Versus "Cartoon" Modellng
"Slrearn,,!" consciuusllcss" is a psychulogicul termo coined by Ihe psychologist William
James, Ihal churaclerizes individual cunscious expcrience as a succession of states con-
stanlly moving onward in time. The idea has been transferred to literalure in the form
('f "slrearn.of-cunsciousness" wriling. In ils finest expression an individual's interior
l11unulugll',' is used lo reveal chaructcl und cummenl on life. Al its worsl il mnounts lo
a self-indlllgenl mind dump.
Unfortunalely many creative exercises can be approached in the latter fashion.
For example compuler progrums are uften written without prior thoughl. Individllals
sit duwn at a computer console and jllsl begin Iyping. Invariably the final result (as
well as tl1<'IIllimate time inveslmenl) sllffers fmm Ihe haphazmd approach and lack of
design
Mathemalicul models can ulso be developed in a stream-of-consciousness fashion.
There is oftenlhe lendency to start wriling muss balunces withoul adequale forelhoughl.
As expecled, Ihe results are often incorrect or incOlnplete. In the best case a correct
model rcslllts only ufter muny lime-consllming revisions.
Some simple steps cun be applicd to avoid such pilfalls:
J. Drllll' (/ dillllram. Por the simple well-mixed models described up lo now, this merely
consists of skelching Ihe majur sources and sinks of the pollutant being modeled. AI-
Ihuugh Ihis might seem triviul, th~ act of drawing forces you to delineate the mech-
¡rnisrns governing polllllant dynamics. In later lectures, as we deal with multiple
polllllunls in segmented systems. diugrams will become essentiul. Dr. Bob Broshears
ofthe U.S. Geological Survey calls this "curtoon modeling." Although this terminol-
ogy mighl sound lIippant, il is not meant tu be. Expt:rienced modelers recognize that
a well.thullght-out schematic is critical to keeping track of all the variables and pro-
cl'~S~S in a cumplieatcd model.
.. Wril!'!!.qllllliolls. After a schematic is developed, il can be transluted inlo model equa.
~s. For Ihe simple case discussed so far, eacb arrow represents a term in the mass
bulance. In later lectures, thcre will be many variables (boxes) connected by many
pmcesses (arrows). Thus the schematic pro vides a gllide for ensuring that the math-
ematical characlerizution is complete.
. Oblllil/ (/ sO/Ulioll. This can be accomplished exactly (algebra or calcllllls) or approx-
imalely (numerical methods). For more complicated syslems. computers are neces-
.
sí!.!Y
'. Check re.m/ls. This lasl step is sometimes neglected by the novice modeler. Too
lUany people lrust model OUlput if il "Iooks reasonable." Unfortunulely this ten.dency
inrreuses when computers are involved. And if Ihe reslllts are displayed in high-
resollltion graphics in multiple colors, certain individuals lose any vestige of skepti-
c¡~m. Therefore, whether checking a homework sollltion or a large professional code,
suflicient lesting is reqllired to ensllre thut the model is producing correct results. Be-
yond obviolls und easily recognizuble bloopers (e.g., a negalive concentralion), the
simplest starting point is lo check thut muss is conserved. Beyond thal, more com-
plicated lesls are rcqllired. I willlollch un Ihese when I review mudel development
in Lec. IR.
l
- - - ---"- -
LECTURE.1 Mass Balance, Sleady-Stale Solution, and Response Time 55
3.2.1 Transfer Functlons and Resldence Time
Aside from lhe assimilalion faclor, lhere are a variely of olher ways lo summarize
Ihe abilily of a sleady-slale syslem lo assimilale pollulanls.
Transfer function. An alternalive way lo formulale Eq. 3. I7 is based on ex-
pressing Ihe loading in Ihe formal of Eq. 3.9. For Ihe sleady-slale case Ihis is
W = QCin (3.20)
Equation 3.20 can be SUbsliluledinlo Eq. 3.17, and bolh lhe numeralor and denom-
inalor of Ihe result can be divided by Cinlo yield
C
c:;;;- = {3 (3.21)
where (3 = Ihe lransfer funclion
{3 Q
Q + kV + vA, (3.22)
2 Equalion 3.22is calIed atransfer fimction t because irspecifies how lhe syslem inpul
)
(as represenled by Cin)is lransformed or "Iransferred" lo an OUlpUl (as represenled by
c). Examinalion of Eq. 3.22 provides insighl inlo how lhe model "works." If {3« 1,
lhen Ihe lake's removal mechanisms will act lo great/y reduce lhe level of polIutanl
1 in Ihe lake; lhal is, such a lake has greal assimilalive capacily. Conversely if {3 -+ 1,
Ihen lhe lake's removal mechanisms (Ihe denominalor) are weak relalive lo ils supply
mechanism (lhe numeralor). For such cases Ihe pollutanl level will approach lhal of
Ihe inflow. In olher words Ihe lake's assimilalive capacily is minima!.
Thus lhe lake's assimilalive capacity can be evalualed by lhe dimensionless
number (3. Inspeclion of Eq. 3.22 indicales thal for lhe simple model in Fig. 3.1,
assimilalion increases for large values of reaclion rale, setlling velocilY,volume,
and area. NOleIhal flow which appears in both lhe numeralor and lhe denomina-
lor acls lo bolh ¡ncrease and decrease assimilalion. It increases assimilation as ir
refleels flushing of polIulanllhrough lhe lake's out/el. It deereases assimilalion as
) it reflecls deJivcry of pollulanllhrough lhe lake's inflow.
( Residence time. The residence lime TEof a subslanee E represenls the mean
i I amount of lime lhal a moJecule or particle of E would slay or "reside" in a system. 11
) is defined for a slcady-slale, conslanl-volume syslem as (Stumm and Morg~8 J)
E
, TE = ¡dEldt~ (3.23)
where E = quanliry of E in lhe volume (eilher M or M L -3) and IdEldtl:!: = absolule
1 value of eilher Ihe sources or lhe sinks (eilher M T-I or M L- 3 T-I).
One of lhe simpler applicalions of Eq. 3.23 is Ihe delerminalion of the residence
I lime of waler in a lake. Since Ihe densily of waler is by definilion approximalely
I I g cm3, lhe quanlily of waler in a lake is equivalenl lo ils voJume. In a similar
( 'The lerll1"tmnsfer fllnction" has a relalcd blllll10re sophislicaled definilion in linear syslell1s analysis.
 56 ~AHTI Complelely Mixed Syslems
IH'TLJI!E.1 Mass Balanl:c, Stcady-Slalc Solulion, and Response I'lIlle 57

scnse the "sink" of water from a lake is measured by the magnitude 01'its outflow
(assuming that evaporation = precipitation). Subslituting Ihese vallles into Eq. 3.23 BOX 3.3. Estimating Reaction Kinetics with a Steady-State CSTR
yields the water residence time
V Grady and Lilll I 19XO)dcsl:rihe a Illelhod 1<.1'evalualing real:tion kincli¡;s usi'lg cxperi-
(3.24) ments l:ondlll:ted wilh CSTRs. In Ihis apprnadl. whi<:h they ¡;all the algebrrl/!' mellwd,
TI\' =Q Ihe real:tors are run lo a sle:Jdy-state. Allhis point the Illass hahul¡;e can he wrilten as
This relationship is useful for understanding the general notion of residence time QC¡n - Qc - rV = O (3.27)
since it has a straightforward I'hysical interpretation-it is the amount of time that
would be required for the olltllow 10replace the quantity of water in the lake. Thus where r = r;lIe 01'l:nnsllmplion ,,1'Ihe rea¡;(ant 1M L .¡ TI). Ir all the nther ljnanlilies
are measllred the ha lance ¡;an be sol ved rol' the l:nnsllmptinn rale
il is a measure of the lake's Ilushing rateoIf the volume is large and the flow is small
I
the lake has a long residence time; that is, it is a slow flusher.Conversely lakes with r = Q('III-Q(' -(¡'in - (') (J.2X¡
V Tu
short residence times (high flow and small volume) are referred to as fasl flushers.
Equation 3.23 can also be used to compule a "pollulant rcsidence time." For Ir we aSSllllle that Eq. 2.7 hnlds.
example for the system in Fig. 3.1, the sinks can be represented on a mass basis by r ::::: ke"
(.\.21))

The parameters k and 1/ l::Jn be determined hy taking the naturallngarilhlll.

Idd~ I:!: = Qc + kVc + vAsc (3.25) In r = In k + 1/ In (' (.UD)
This equation, along with Eq. 3.4, can be substituted inlo Eq. 3.23 to give
Thlls if a plnt nI' In r versus In ¡; yields a stmighlline, Eq. 2.7 hnlds, and the slope and
V inler¡;ept can he IIsed to cakulale k ;md 1/. An exereise on Ihe appmadl is p :sented in
Probo 3.5 al lhe end 01'this lecture.
T,. = Q kV + vAs
+ (3.26)
Note that Eqs. 3.24 and 3.26 are similar in form wilh the exception that the pol-
lutant residence time is affected by reactions and senling in addition lo the outllow.
EXAMPLE 3.2. TRANSFER FUNCTION AND RESIDENCE TIMES. For Ihe 3.3 TEMPORAL ASPECTS OF POLLUTANT REDUCTION
lake in Example 3.1, determine Ihe (a) inflow concenlralion, (b) lransfer fllnclion, (e)
To this point we have f{¡cused on sleady-stalc sollltions. These provide an estima~c
water residence lime, and (el)polllllanl residen¡;e time. ., 01'the average waler qualilYIha! will rcsull il' loadings are hcld conslanl liu' a suflj-
Solution: (a) The inflow concentralion is compuled as ícicnlly long limc periodoIn addilion lo slcady-slale prediclions, walcr-qualily man-
W 140,000 agers are also inleresled in Ihe lemporal response 01'nalural walers.
1867 L -I Suppose Ihat a system is al steady-slale. Al a specific lime a \1;"". removal
C¡n ~ Q = 7500- = . mg
projecl is implemented. As depicled in Fig. 3.4, Iwo iI1lerrclaledqllcslions arise:
(b) The Imnsfer ¡;oeffkient can now be delermined as

(3 = -e = -
C¡n
Q
Q + kV
= 0.32 ·
· Ilow long will illake for improved waler qllalily lo occlIr?
What will the "shape" 01' Ihe recovery look like?

To determine Ihe corrccl trajeclory, let's slart wil/¡ Ihe mass-balance modd
Thus Ihe removal proces:;es ael lo creale a lake COn¡;enlralion Ihat is 32% 01' the inflow (Eq.3.16)
concenlralion.
(c) The residence lime cau he ¡;akuhued as tic
V- = W(¡)- Q c-kVe-I'A e 0.31,
V dI .\
50,O<X) = 6.67 d
Tu' = Q 7500
Bcfore solving Ihis cqllalion, we can divide il by volulllc lo yicld
(d) The pollulant residence lime is de W(t) Q l'
V 50,000 dI - - -e - ke - -e (3.32)
V V H
T,. = Q+kV = 7500+0.319(50,000) = 2.13d
Collecling lerms gives
~ ( Because of Ihe addition of the decay lerm, Ihe residence lime 01'a pollulant is about de
'1 one-third Ihe water residence time. (3.3.\)
dI + Áe = WU>
V
 LECTURE 3
58 PARTI Completely Mixed Systems Mass Balance, Steady-State Solution, and Response Time 59

8
'. W(I)

o (a) 2
e
1
0.5 e-X
O
c-:: ~ ~ ;-~:~ ~~;; ::: ::,>-- FIGURE 3.4
O
0.693
2

(a) A waste load reduction al.ong 1.386 FIGURE 3.5

x
o (b)
with (b) lour possible recovery
scenarios lor concentration.
The exponential lunction.
\

where \

Q v (3.34) Co
A=-+k+-
V H

in which Ais cal!ed an eigenvalue (lhal is, a characlerislic value). C

I If alllhe paramelers(Q.V.k. v.H) are conslanl,Eq.3.33is a nonhomogeneous,
!linear, firsl-order, ordinary differential equalion. Its solulion consisls of Iwo parts,
\ FIGURE 3.6
C = cl! + cp (3.35)
The lemporal response 01 our well-mixed
lake model lollowing the lerminalion 01 all
where cl! = general solution for Ihe case W(t) = Oand cp = parlicular solulion for loadings al t = O.
~ specific forms of W(t).
Because Ihe general solution corre_sponds I<tlhe case where lhe loading is ler-
minaled, It is id~at fur in.Ye~ligaling a syslem's recovery time. .~ described next il Thus we can now inlerprel Eq. 3.36. As in Fig. 3.6 Ihe negative value of!he ar-
will also provide us wilh insight inlo the shape 01 Ihe recovery. gumenl means Ihat (he concenlralion dccreases and asymplolicalIy approaches zero.
Further Ihe rale of decrease is dictated by the magnitude of the eigenvalue A.If Ais
)large, lhe lake'S concenlratlonWllTdecrease rapirlly.If Ais sma~ lhe lake's response
v 3.3.1 The General Solution !will be slow.

If C = Cnal t = O,Eq.3.33wilhW(t) = Ocanbe solvedby lhe separationof vari- t EXAMPLE 3,3. GENERAL SOLUTION, In Example 3.1 we determined the steady-
ables (recal!solutionof Eq.2.10): slate concentration for a lake having the following characteristics:
}'--
C = coe-M J
Thus we have arrived al an equation lhal describes how Ihe lake's concenlralion
(3.36) Volume = 50,000m3
Mean depth = 2 m
Temperature = 25°C
Wasteloading= 140,000g d-I
Inflow = outflow = 7500 m3 d-I Decay rate = 0.319 d-I
changes as a function of lime following lhe lerminalion of wasle loading.
The behavior of Eq. 3.36 is clearly diclaled by lhe exponenlial funclion. As Ir the initial concentration is equal to the steady-state leve! (5.97 mg L -1), determine the
general solution.
in Fig. 3.5, for lhe case where Ihe argumenl of Ihe funclion (thal is, lhe value lo
which e is raised: x) is zero, Ihe exponenlial funclion's value is unily. T~~ Solution: The eigenvalue can be computed as
lhe argumenl is po~live, Ihe [unqiQ!LmCrease~~ i'!..an acceleraled fashion; lhal is, Q 7500 _1
it doubles ils value al sel inlervals of x (= 0.693). Ín conlrasl if lhe argumenl is A = V + k = 50.000 + 0.319 = 0.469d
negalivc, (he f!lndion aSYlllplolicallylIecreases lowaró zcro ¡'y halving allhe same
sel inlervals. Tlllis the general solulion is

...
'-'

60 PARTI Complelely Mixed Syslems

LEGURF.J Mass Balance, Steady Slate Solution, and Response Time Id

which can be displayed graphically

e = 5.97e-OA69'

as
-
TABLE3.1
Response times
Response time 154 103.1 "5 ,''' los '..
6 Formula 0.693/ A l/A 1.39/A 2,3/A J/A 4.61.\

:::- 4
~ Taking the naturallogarithm and solving for 150gives
01)
a 0.693
~ 2
150 = --¡- (3..19)
2 4 6 8 10 \ Thus we can see that the 0.693 we observed previously (recall discussion of Fig. 3.5)
¡
is actually the naturallogarithm of 2. Note thal the quantity 150is also commonly
t(d) FIGURE E3.3 referred 10as a half-life (recall Seco2.2.4).
NOle Ihal by 1 = 5 d Ihe concenlralion is reduced 10 less Ihan 10% of ils original value. The above derivation can be generalized to compule an arbitrary rcsponse time
by Ihe formula
\ By 1 = 10 d, for all intenls amI purposes, il has reached zero. I 100
1", = -In- (340)
Á 100-4> .
An interesting property of the general solution is that even though the loading where 1", = 4>%response time. For example if we are inlerested in detenuining hulV
is reduced to zero, the concentration will never reach zem. This introduces an ele-
long it takes to reach 95% of its ultimate recovered level, we could compule
menl of alllbiguity into the analysis. We now attempllo resolve this ambiguity by
I 100 3
t I ¡introducing the concept of response time. 195= Á In 100_ 95 = Á (3.'1I)
Table 3.1 and Fig. 3.7 show other response times. As would be cxpccted, Ihe
3.3.2 Response Time higher the percentagc of recovery, the longer the response time.
Although Iheparameter group Á clearly dictates Ihe lake's temporal response char-
acteristics, it has shortcomings for communicating with decision makers. First, it
, has the counterintuitive property that as it gets large the time for the lake to respond
gets smalI. Second, as mentioned in the prcvious section, its interpretation is c10uded =
by Ihe fact that, from a strictly mathematical perspeclive, the underlying c1eansing
process never reaches completion. Try telling a polilician thal a c1eanup would the-
~ 0.5
oretically take forever! They tend to react very unfavorably to asymptotic solutions
that extend oeyond the next election.
Both these shortcomings can be rectified by using the general solution to derive
l' a new parameter group. Called Ihe respoIIse lime, this parameter group represents
4
FIGURE 3.7
1the time il takes for the lake to complete a fixed percentage of its recovery. Thus the Iso 175 190 195
problem of ambiguity is remedied by decid:ng "how much" of the recovery isjudged If).. A plot 01 Ihe general solution showing values
01 several response times.
as being "enough." For example we might assume that if the lake has experienced
95% of its recovery we would be satisfied that, for all practical purposes, the reme-
dial measure is successful. EXAMPLE 3.4. n¡':SPONSE TIMK Detcnlllne Ihe 75%, 90%, 95'/1, "",1 -}l)'X. n>
In terms ofEq. 3.36 a 50% response time means that the concentration is lowered sponse limes for Ihe lake in Example 3.3.
to 50% of its initial value, or Solution: The 75% response lime can be compuled as

0.50co = coe-Also (3.37)

175 = 1.39
where 150 = 50% response time (T). Dividing by the exponential and 0.50co yields
0.469 = 2.96d

eAlso = 2 (3.38) In a similar fashion we can compule 190 = 3.9 d, 195 = 6.4 d, and 199= 9.8 d.
 62 PART1 Complelely Mixed Syslems LECTURE3 Mass Balance. Sleady-Stale Solulion, and Response Time 63

As in Example 3.4 there is a certain amounl of subjectivity involved in deciding A subdivision housing 1000 people will discharge raw sewage into this syslem. Each
( "how much" of Ihe recovery is judged as being "enough." In general I recommend using individual contri bu les about 150 gal capita-I d-I of wastewaler and 0.25 lb capila-I
1 eilher 190or 195.They are neither too lenient nor 100 stringent and conform lo whal most d-I of biochemical oxygen demand (BOO).
I individuals would deem an acceplable level of recovery. /(a) Oelermine Ihe BOO concenlration of Ihe waslewater in mg L -l.
/(b) If Ihe BOO decays al a rale of 0.1 d-I and setlles at arate of 0.1 m d-I, calcu-
lale Ihe assimilalion factor for Ihe pond prior lo building the subdivision. Which
of the purging mechanisms are mosl effeclive? Lisl them in decreasing order of
effectiveness.
BOX 3.4. The Rule o, 72 (e) CalcuJale Ihe transfer function faclor after building the subdivision.
Ad) Oetermine the steady-state concenlralion for Ihe lake wilh and without the sllbdi-
Before Ihe days of handheld calculalors and computers, bankers and financiers needed vision.
a quick way lOevaluate Iheir investmenls. To do this Ihey developed a heuristic thal is
called tlle rnle of 72. According lo the rule. the time required lo double your money 3.2. A lake with a single intlow stream has the following characleristics:
can be estimaled as Mean depth = 5 m
Surface area = I1 x 106 m2
. " 72
0 oubl mg tIme"" ;--
mleresl rale (%)
Residence lime = 4.6 yr

An industrial plant presently discharges malathion (W = 2000 X 106 g yr-I) to the

ror example if you invested some money al an annual interest rate of 6%, il would
lake. In addition the intlowing stream also contains maíathion (ejn = 15 mg L -1). Note
double in approximalely 12 years. The same formula can also be employed to assess
thal the volumelric rate of inflow and outtlow are equal. Assuming that a firsl-order
how the value of your money decreases due lo inflalion. For example if Ihe infla-
decay reaclion can be used lo characterize malalhion decay (k = 0.1 yr-I),
tion rale is 3%. Ihe money hidden in your matlress would lose half its value in 24
~(a) Wrile a mass-balance equation for malalhion for this system.
years. 'ib) If Ihe lake is at sleady-stale, compute the in-Iake malalhion concentralion.
This formula is derived from the concept of the half-life. In facl a more accurale
(e) 11'Ihe lake is al sleady-slale, whal industrial planlload'ií(g rale musl be maintained
represenlalion, based on Eq. 3.41, mighl be called "Ihe rule of 69.3,"
lo lower Ihe lake's concenlralion lo ~O ppm? Express your result as a percenl re-
duclion.
" 69.3
Oou blmgtIme = . ,~, (d) Evaluale each of Ihe following engineering oplions lo delermine which is Ihe mosl
effeclive for lowering Ihe sleady-state concentration:

The reasonlhal a numeralor of 72 was chosen is thal it is more easily divided by (i) Reduce Ihe present loading rate of the industrial plant by building a waste
whole number interest rales. For example trealmenl facilily Ihal will remove 50% of Ihe malalhion from the planl's ef-
tluen!.

72/1 = 72 yr 72/5 "" 14 yr 72/9 = 8 yr (ii) Oouble Ihe lake's deplh by dredging.
72/2 = 36 yr 72/6 = 12 yr 72/10 "" 7 yr (iii) OOllble Ihe lake's olllnow rale Q by diverting malalhion-free waler from a
72/3 = 24 yr 72/7 "" 10yr nearby unpolluled slream inlo Ihe lake.
72/4 = 18 yr 72/8 = 9 yr (e) Whal olher faclors would need lo be considered (aside from lowering concenlra-
lion) when making a decision in (d) in Ihe "real world"?
Thus you can quickly figure Ihe lime required to double or halve your money. (f) Oelermine the 95% response limes for each of Ihe oplions in part (d).
Aside from providing you wilh a handy means for evalualing your inveslmenls,
3.3. Recall from Example 1.3 Ihal in Ihe early 1970s Lake Onlario had a 10lal phosphorus
we have included Ihis discussion lo illuslrate how first-order processes and compound
loading of approximalely 10,500 mta (melric lons per annum) and an in-Iake concen-
inleresl are based on similar malhemalics.
Iration of 21 J.LgL -1 (Chapra and Sonzogni 1979). It is known Ihallhe only losses of
10lal phosphorus for Ihe lake are setlling and flushing Ihrough Ihe lake's outle!. Assume
.... Ihe outflow rate is 212 km3 yr-I and the sediment area is 10,500 km2.
(a) Calculate the intlow concentration for the system. Assume that water intlow equals
olltnow.
PROBLEMS
(b) Use a mass balance to estimate the apparent setlling velocity for total phosphorus
3.1. A pond wilh a single innow slream has Ihe following characterislics: in this lake.

Mean deplh =3m 3.4. A lake has the following characteristics:

Surface area = 2 x 10' m2 Volume = I x 106 m3
Residence time = 2 weeks Surface Area = 1 x 1()' m2
I
Innow BOO concenlralion = 4 mg L -1
Water residcnce lime = 0.75 yr
I
I

L
64 PARTI Completely Mixed Systems
--~--l
A soluble pesticide is input to the lake at arate of 10 X IOb mg yr-t. The in-Iake LECTURE 4
concentration is 0.8 ¡.¡.gL-l.
(o) Determine the inflow concentration (assume inflow =
outflow).
(b) Determine the transfer function. " Particular Solutions
, (e) Ir the only removal mechanism (other than flushing) is volatilization, compute the
flux of the pesticide out the lake's surface and to the atmosphere.
(d) Express the result of par! (e) as a volatilization velocity.

3.5. Arate experiment is performed for a I-L CSTR. The inflowing concentration is held
steady at a level of 100 mg L -1 . The flow is varied and the resulting outflow concen-
trations are measured:

Q (L hr-I)
c(rngL-I)
I 0.1
23
0.2
31
0.4
41
0.8
52
1.6
64

Use the algebraic method to determine the rate and order of the reaction.

3.6. Derive Eq. 3.40.

3.7. A pond with a single inflow stream has the following characteristics:

Mean depth = 3 m
Surface area = 2 X 105 m2
Residence time = 2 wk
LEC11JRE OVERVIEW; 1 develop particular solutions for specific forms 01 lhe load-
A subdivision will discharge raw sewage ioto this system. Ir BOD decays at arate of
0.1 d-J and settles at arate of 0.1 m d-I, calculate the 75%, 90%, and 95% response
jflg function: impulse, step, linear, exponential, and sinusoidalloadillgs. Then
J place emphasis Onsummarizing each solution's behavior in terms of a few
times for the pond. simple "shape parameters."
3.8. Determine a half-life for a batch reactor with a second-order decay reaction.

3.9. Compute the first-order reaction rates for the following substances:
(o) cesium-\37 (half-life = 30 yr) Up to this point we have derived steady-state and general solutions tÚr the mass
(b) iodine-131 (half-life = 8 d) balance equation for a CSTR with first-order kinetics,
(e) tritium (half-life = 12.26 yr)
dc + Al' = W(t) (4.1)
3.10. A lake (volume = 10 X IOb m3, water residence time = 2 months) is located adjacent dI V
to a railway line that carries considerable traffic of chemicals. You are hired as a con-
sultant to provide insight into potential spills into the lake. If the lake is assumed to be where A is an eigenvalue that, for the present model, is equal to
completely mixed, a spill would be distributed instantaneously throughout the volume. Q v
Therefore the resulting concentration would be Co = mlV, where m is the mass of pol-
A =V + k + H (4.2)
lutant that is spilled. Thereafter the lake's response would follow the general solution.
(o) Develop a plot of '75,195, and 199versus pollutant half life. Use logarithmic scales N..Qwlet's tum lo particular solutions.l'hese relate to cIosed-tQrm s~Julions rol'
wbere you believe they would be helpful. specific forms of W(t). As might be expected, solutions a.re avail¡¡h~-2.!!!YJQr id~J-
(b) Inelude a short "user's manual" for the plot to provide managers with guidance for
ized forms of the loading function. Some (jf the more useful forms are sUI!!.,marized_
its use and interpretation. in Fig. 4.1 (a through e).
Although these forms might appear too idealized to be realistic, Ihe)' have great
utility in water-quality modeling for three reasons:

e,· First, there are a surprisingly large number of problem contexts where they provide
i an adequate approximation of loading trends.
· Second, models are frequently used to predict future impacts. For these cases we
~( would not know the exact way in which future loads would vary. Consequently
idealizations become extremely usefu\.

65
 66 I'ARTI Complelcly Mixed Syslems LECTURE4 Particular Solutions 67

W(I) W(I)
W(,)

(a) Pulse (b) Slep

o
(8)
"'(11 e
W(tI m
V

(e) Linear (d) Exponen\lal FIGURE 4.2

o "5
(b) Plol 01 (a) loading and (b) response lor
\ 1
impulse loading.
\V!I)
function can be visualized as an infinilely thin spike centered at f = Oand having
unít area. It has the following properties (Fig 4.2a):

~ 8(t) = O t ;é O (4.3)

FIGURE 4.1
(e) Slnusoldal (t) Real
and !
I l_O)8(t) dt
O) = 1 (4.4)

The impulse load of mass to a water body can be represented in terms of !he
Loading lunclions W(t) versus lime t. delta function as

. Third. idealized loading functions allow us lo gain a better underslanding of how

a model works. This is because realistic loading functions (see Fig. 4.lfl often
W(t) = m8(t)
where m = quantity of pollutant mass discharged during a spil1(M) and the delta
make it difficult to discern model behavior. In contrast idealized solutions pro vide function has units of T-I . Substjtuting jnto Eq. 4.1 givcs
an uncomplicaled view.
) de m8(t)
In the following sections 1present Ihe analytical solutions for each of these load- ? dt + Ae = V- (4.5)
ing functions. Aside from the mathematical represenlation. 1will iIIustrate that each The particular solution for this case is
le,
solution has characteristic shape parameters that summarize its behavior. These pa-
rameters are akin to the response times that allowed us to neatly summarize the m -Al
-e (4.6)
general solution in the previous lectllre. In stressing these shape parameterS. my ul- ~ e = V
terior motive is to help you build up mathematical insight regarding these simple
solutions. It is hoped. by focusing on the underlying patterns, that you will develbp C, This solution indicates that the spill is instantaneously distributed throughout
a more intuitive sense of how natural waters respond to loadings. I the lake's volume. resulting in an initial concentration of mlV. Thereafter!he result
is identical to the general solution; that is. the concentration decreases exponentially
at arate dictated by the magnitude of A(Fig. 4.2b).
4.1 IMPULSELOADING(SPILL) Thus we can see !hat two shape parameters summarize the impulse response.
The initial concentration defines the height of the response:
The most fundamental form of W(f) is the impulse loading representing the discharge
of waste over a relatively short time periodoThe accidental spill of a contaminant
to a watcrbody would he of this type. Mathematically the Virac delta/unction (or ~ eo =V (4.7)
impulse/unction) 8(t) has been developed to represent such phenomena. The delta and Ihe résponse lime defines its temporal extenl. If a 95% response is arbitrarily
 --- -----

68 PAR:rI Complelely Mixed Syslems "'.,,,,, , P"rt¡,"I~S"I"'¡"~

chosen, this is () and Ihe response lime defines ils lemporal cxtcnl. lf a 95% response is arbilrarily
3 chosen, Eq. 4.8 holds.
(4.8)
t95 = A EXAMPI.E 4.1. STEP I.OADlNG. Al lime zero, a sewage Irealmenl planl began lo
discharge 10MGD 01'waslewaler wilh a concentralion 01'200mg L -1 lo a small delention
basin (volume = 20 x 104 m3). 11'Ihe sewage decays al a rale 01'0.1 d-I, compule the
4.2 STEP LOADlNG (NEW CONTINUOUS SOURCE) concentration in Ihe syslem during Ihe firsl 2 wk 01'operalion. Also determine Ihe shape
paramelers lo assess Ihe ultimale effecl 01'Ihe plan!.
(, Irat t = Othe lake's loadingis changedto a new constantlevel, then the forcing
/1 function is called a step input. Such behavior is given mathematical expression by Solutloo: The ftow musl be converted 10Ihe proper unils:
. the step junt:tion. This is essentially an "on-off' funclion lhat has a jump discon- 1 m3s-1 86.400 S
tinuity at t = O.Forthe presentcase, whereour loadingjumps fromzero to W al
t = O(Fig. 4.3a), it can be represented as
10MGDu u._. .__ ( ¡¡-- )_- 37.854m3d _(
W(t) = O t<O (4.9)
The eigenvalue can be delermined as

W(t) =W t 2: O A = 37.854
20 x 104 + 0.1 = 0.28927 d-I
where W = the new constant level of loading (M T-1). The pa11icularsolution for
this case is Therefore the conccnlration can be computed wilh Eq. 4.10 as

e = ..!..(l
AV
- e-Al) (4.10) ._ W
(" - AV(I
_ -Al =
e )
200(37.1154)(1_ e-U.2M'I27,)
x 104)
0.28927(20
= 131(1_ e'U2M~27')

As depicted in Fig. 4.3b this solution slarts at zero and then asymptotically con- The results 1'01'Ihe lirsl 2 wk are
verges on a new steady-state concentration. At t = 00,Eq. 4.10 becomes equivalent
to the steady-slate solution (Eq. 3.17), since lhe exponential term becomes very small I (d) o 2 4 6 8 10 12 14
as lime progresses. c(mgL-1) O 57.48 89.72 W7.79 117.92 123.61 126.79 m.5!!
Thus we can see that again two parameters summarize the step response. The
steady-state concentration defines the height of the response:
These values can also be displayed graphically as
W ,I
(4.11)
e = AV "

160
W(t) -----------.
-
~ 80
120

C)
o E
-40
(a) ~
c=
el
: ;.-.:.-.:.-.:.-.:;.:;.-.(:
- W
'-.: ~ 5 10
t(d) FIGURE E4.1

FIGURE 4.3 The shape paramelers 1'01'Ihis case are an ultimale sleady-slale concenlralion 01'
o t9S Plot 01 (8) loading and (b) response lor step 131 rng L -1, 01' which 95% will be auained in 3/0.28927 ... 10.4 d.
(b) loading.
 70 PARTI Complelely Mixed Syslems LECTURE
4 Particular Solulions 71

y 4.3 LINEAR ("RAMP") LOADING (1)

Tbe simplesl representation of a trend is a straight line with a nonzero slope. Waste
inputs can often be represented in this way, as in
W(t) = :'::{M (4.12)
where {3, = rate of change or slope of the Irend (M T-2). Notice Ihat Ihe trend can
be either positive or negalive. The particular solution for Ihis case is o

e (a)
e = :':: ~~ (Át - I + e-Al) (4.13)
Ce
In waler-quality applications, Ihe positively increasing case (Fig. 4.4a) is more
frequently used. Sometimes loadings thal are proportional to population growth can
be adequately approximated in this way. F~_ases-, aiter_!!.nin¡ti.!!!§tart-ulUim~
,

Ir
(which call be defined by a 95% response time equal to 3/ A), the solution becomes }
o{3¡ I
I e = XV t (- A) (4.14)
o te
(b)
FIGURE 4.5
Plol 01 (a) loading and (b) response
exponenlially decaying loading.
lor
Thus, as illustrated in Fig. 4.4b, Ihe solution eventually increases at a constant
slope of
{3¡ (4.15) v 4.4 EXPONENTIAL LOADING
{3¡ = AV
Another slandard way lo characterize loading Irends is as an exponential function
wilh response lagging the input by (Fig. 4.5a),
I
(4.16) W(t) = W,éP,/ (4.17)
t¡ =A
where W, = a parameter that denotes the vaJue at t = O (M T-1) and (3, specifies
the rate of growlh (posilive) or decay (negalive) of the loading (T-1).
W(t) -J-. - 3 ~ Such a funclion characterizes waste sources Ihat grow or decay in proportion lo
'JC- - I ítheir own magnitude. The particular solulion for this case is
J t¡.
e = V(AW, -
:'::(3,) (e:!:P'/ e-Al) (4.18)

This solution has many applications in water-quality modeling. The positive~-

~ ~ential has beenused~mo~elJoadi~~ that aredependenton E.~pulation
~Ih
o (see Example 4.4). The decaying exponential is more common!~ ":.n~ounter~ !le-
(a) í ~se il applies to series of first-orderdecayiñg'proce1ise~(see Example 4.2 for such
C . a case).Figure4.5b showsIhe resultfor the decayingexponentialload.Nolehow
Ihe solulion rises lo a peak and then begins to decline exponenlially.
"[hus we see that Ihere are two shape parameters: Ihe mag~and Ihe limeof
the peak. 1.h~i~e~si!Y.J!etermi~4 by differentiating.Eq. 4J.8 and ~etting Ihe
result equaJJo zero. After some algebraic manipul.ati.on,Ihe time of Ihepeak can be
rlelermined as " - - ~ ."

- FIGURE 4.4
= I~~~~) (4.19)
o Plol 01 (a) loading and (b) response lor
1 te

(b) a linearly increasing loading. wherethe subscripte signifiesIhalIhis is Ihelimeof the "critical"concenlralion.
The magnitude can be delermined by merely subSliluting Eq. 4.19 into Eq. 4.18. An
... .-

72 PAKT1 Complelely Mixed Syslems !.E<TUKE4 Particular Soluliuns 73

alternative approach can be developed by recognizing that the peak concentration or SUbsliluling Ihe parameler values,

corresponds to Eq. 4.1, with the loading equal to Eq. 4.17 (with negative exponent) Co ~ 20(e-O.I, - e-II.2,)
and de/dI = O. Substituting the critical time into Eq. 4.1 with de/dI = Ogives
The plol of Ihis solulion, along wilh Ihe solUlion for ('A, is shown in Fig. E4.2.
We
- -e -P,I,. (4.20) The shape paramelers for Ihis case can be compuled as
", Ce- AV 0.1
0.1 (20) 0.1 0.2-0.1
Further, Eq. 4.19 can be substituted into Eq. 4.20 and the result manipulated to give
p,
which will occur al
c,. = - 0.2
(--0.2) _1
= 5 mgL

= We {3ey.-¡¡,
Ce AV A) ( (4.21)
In (O. 110.2)
t,. = 0.1 -0.2 = 6.93d
-0 EXAMPLE 4.2. EXPONENTIAI. FORCING FUNCTION. The following series of Tbese values are consislenl wilb tbe graph.
tirsl-oroer reactions takes place in a balch reaclOr:
t, tI
A ,B-
Mass balance equations for Ihese reactions can be wrinen as 4.5 SINUSOIDAL LOADING
dCA
-dt = -k1cA The simplest periodic input is the sinusoidal function (Fig. 4.6(/), which can be rep-
dco
resented mathematically as
and I -dt = k,CA - k2Co
W(I) =W+ W" sin(wI - O) (4.22)
Suppose that an experiment is conducted where CAO= 20 ano ('Oll= ° mgL-l. If kl = where W = mea n loading (M T-1)
0.1 and k2 = 0.2 d-I, compute Ihe concentration of reactant B as a function of time. W" = amplitude 01'the loading (M r I)
Also, determine its shape parameters. 8 = phase shift (radians)
Solution: The concenlration 01'reactanl A can be delermined by inlegraling the first W = angularfrequency01'the oscillation(radiansT-1), whichcan be de-
differential equalion lo give finedas
CA = cAlle-kl'
This result can be subSliluted inlo the second differential equalion lo yield
o
(
)
dco
-" + k2CO = k ICAlle -t
I
,
dt
Tbus Ihe mass balance is now in the form of a lirsl-order differcnlial equation wilh an \
exponential forcing function. Tbe solulion is
(
I Co
k,CAII
= (k2 _ kl) e( -kl' - e -tI' )
20
(a)
e I«P . O
,-..
:s 10
~
e
......
~ e FIGURE4.6
I I I I
Plot01(a) loadingand (b) resp( o)lor
10 20 30
the sinusoidalloadinglunctiol.. , Ihat
t(d) FIGUREE4.2 a constantinput is also shown 111 Itus
(b) iIIustration.
 ,/
74 PARTI Complelely Mixed Syslems LECTURE
4 Parlicular Solulions 75

21T
C = C",inA(W) sin[wt - 4>(w)] (4.30)
W = Tp (4.23)
where A(w) = an amplitude attenuation coefficient,
where T p = period of the oscillation (T). Note Ihat the simple frequency of the
A
oscillation can be computed as
I A(w) = ~ (4.31)
,.¡A2+ w2
(4.24)
f = Tp Depending on Ihe frequency of the oscillation (as reftecled by the value (lf w). the
where f has units of cycIes T-1 . solution's amplilude will be attenualed (as reOecled by A) and shifted (as reflected
Beforc proceeding to the particular solution, lecognize that Eq. 4.22 ¡tseIf has by 4». The implications of Ihis solution are explored in the following example.
four shape parameters. These are w (how frequently the sine wave oscillates), W (the
mean valuc for the sine wave), W" (how high the oscillation swings vertically), and /EXAMPLE 4.3. SINUSOIDALFORCING FUNCTION. A complelelymixedlake
receives a conservalive subslance. Waler inf10wand oulf1oware equal, and Ihe inf10w
8 (how much the function is shifted horizontally relative to the standard sine wave). concenlralion varíes sinusoidally as
I Thus the parameters provide a lot of Oexibility for characterizing a sinusoidalload. ~
Now the particular solution for this case is Cin = Cin + Ca.in sin(wt)

~(1 - e-M )+ V ~A2+w2 sin[wt - 8 - 4>(w)] when~\ Cin = average inf10w concenlralion
I

c = "v "in = amplitude of Ihe inf10w concenlralion

I
w = angular frequency (= 2'TT/T,,). in which TI' = period of Ihe oscillalion
f
I
~
- V A2+ w2 sin[-8 - 4>(w)]e-M (4.25)
~
If th lake has a volume of 2.5 X 106 mJ and inf10w = oulf1ow = 9 X 106 mJ yr-I ,
determine .
'IS sensilivily lo Ihe sinusoidal componenl of Ihe loading if Ihe period of Ihe
oscillalion i. (a) 10 yr. (b) 1 yr. or (e) 0.1 yr.
where 4>(w)= an additional phase shift (radians) that is a function of the frequency,
~
I

Solution: T e eigenvalue can be compuled as

I
4>(w)= tan-I (¡) (4.26) ( Q 9 X 106
Á = V = 2.5 x 106 = 3.6yr-1
I
Inspection of Eq. 4.25 leads us to recognize thatthe first part of the solution
approaches the steady-state level, and the last part of the solution will die away For a period of 10 yr. w = 2'TT/1O= 0.628 yr-I and

: ~
\ given sufficient time (which again. since it depends on an exponential decay, can be
parameterized by a response time). Thus, after this initial start-up period, the solution
I will be
f/J(0.628) = tan-I (03~~8)
3.6
= 0.1727 radian (2TT1~a~~ans) = 0.275 yr (100 d)

and A(0.628) = = 0.985

"v + V ~
c= ~ A2+ w2 sin[wt - 8 - 4>(w)] (4.27)
J3.62 + 0.6282

Therefore Ihe solulion is almosl idenlicallo Ihe forcing funclion. The amplilude is di-
,, Now inspection of this result leads to the interesting (but obvious, if you think minished only 1.5% and the phase shift is a mere 100 d (compared lO Ihe period of 10
about it) conclusion that the concentration of the solution will also be a sinusoid with yr). However, as shown by Ihe following lable, Ihis correspondence breaks down as Ihe
the same frequency as the loading. Thus the concentration really has only two shape frequency of Ihe forcing funclion increases:
parameters that change as the model parameters change. These are the amplitude of
the concentration response, Period / q,(w)
W" (yr) (cycles yr-I) A(w) (d) (q,(w)/T,I x 100%
c" = ~ (4.28) 10 0.\ 0.985 100 2.8%
V ,.¡A2 + w2
I \ 0.573 6\ \6.7%
and the phase shift defined by Eq. 4.26. 0.1 10 0.057 8.8 24.1%
An especially interesting case occurs when 8 = 0, inftow = oUlftow,and the
Ioading is defined as
As Ihe frequency increases, Ihe solulion exhibils Iwo effecls. First, Ihe amplilude di-
W(t) = QC".insin(wt) (4.29) minishes and approaches zero. Second, as indicaled by Ihe normalized phase shift, Ihe
solulion increasingly lags the forcing funclion. In essence both effecls ref1ecl the fact
where C".in = amplitude of the inOow concentration. In this case the particular solu-
Ihal the syslem is too sluggish (as manifesled by ils eigenvalue. Á) lo "keep up" wilh the
tion becomes (after the transients ha ve died out) forcing funclion.
 76 PARTI Completely Mixed Systems e
W(/)
ni
It should be noted that there is a formal way of presenting the informatiun deve10ped v
in this example. That ¡s, both the amplitude and phase characteristics can be plotted ver-
sus frequency in what are called Bode diagrams. These are conventionally constructed
by plotting either the amplitude or the phase characteristic versus the angular frequency
(w). As shown in Fig. E4.3, we have chosen to plot the characteristics versus the periodo
o o
We have also plotted the phase shift in degrees. In the present context we believe that t.s
both modifications make the plots somewhat easier to interpret. (a) Pulse
_ W
W(/) e
Amplltude (8)
reductlon . :.-.:.- :.-.: .-.: :.: :.: :(-~:: ~-~ - - --
0.5

O
(b)
Phase 8hlft 90 o o
(degrees) t.s
(b) Step
O
0.01 0.1 1 10 100 W(/) e 1
r
Perlod of forclng functlon (yr) FIGURE E4.3

Although the phase shift certainly provides a measure of how much the solution
is shifted temporally, an alternative shape parameter that represents the same effect
is the time ofthe peak relative to time zero. This quantity can be computed by taking o t O
the derivative of Eq. 4.30 and setting it equal to zero to yield
(e) Linear
W(/) e
\ t::;: = (1TI2~+ cp
(4.32)
Ce
This formal is useful because we are often concerned with maximum concentrations
in water-quality l11odeling.

o I O
~ 4.6 THE TOTAL SOLUTION: LlNEARITV AND TIME SHIFTS te
(el) Exponential
Figure 4.7 summarizes the analytical solutions we have developed. Although they
( are usefui in their own right, they can also be employed in tandem to assess the W(/)
¡ impact of severalloading trends simultaneously. That is, if we derive a number of
particular solutions cp;, and a; are any constants, the total solution can be derived by
simple addition of the general and particular solutions,
n

(~ C = cg + La¡cp; (4.33)
;=1
I
~ This is possible only because the underlying model is linear. (e) Sinusoldal
In addition we can add loadings that are initiated at different times by merely FIGURE 4.7
4. shifting the solutions. If you have a function c(t) and want it shifted a units of time
A summary 01 the solulions developed lo Ihis point lor a substance reacling
into the future, you would merely subtract a from the argument, c(t - a). Forexample wilh lirsl-order kinelics. NOle Ihal Ihe shape paramelers are shown on Ihe
I suppose that we have two impulse loads, one that occurs at time zero and another at concenlralion response plols.

77
 LECTURE
4 Particular Solulions 79
78 PARTI Completely Mixed Systems

10
time a. The resulting concentrations can be computed using Eq. 4.6, as in
Road salt
mI -Al O$t<a (4.34) 8
( e = Ve
."
" I mI
e -_ -e -Aa m2 -A(/-a) t ~ a (4.35)
)e -
-A(I-a)
I
_ ~ 7...
( V
+ -e
--V- y
O>
§.
6

where the subscripts I and 2 denote the solutions corresponding to the impulse load <.> 4

\ at t = O and t = a. respectively. A more complicated problem is iIIustrated in the

following example.
2
M Base

EXAMPLE 4.4. LINEAR AND EXPONENTIAL LOADlNGS. O'Connor and

\ Mueller (1970) used linear and exponential forcing funclions to characterize lhe loadings
I O
J 1)of a conservative subslance, chloride. to Lake Michigan. For example they characlerized 1900 1920 1940 1960 1980
I ch'oride loadings due lo road salt by lhe linear model 1
W(t) = O t < 1939
FIGURE E4.4
W(t) = \3.2 X 109(1- 1930) 1930 s 1 s 1960

\ where W(t) has unils of g yr-I. They used an exponenlial mode' lo characterize other
I sources 01' salt thal were correlated with populalion growth in the basin (for example

\ municipal and induslrial sourccs),

W(I) =O 1< 1900_
BOX 4.1. "Back-of-the-Envelope" Calculatlons
A subslanlial portion of water-qualily modeling ineludes large, elaborate calcula-
W(I) = 229 x 10geO,0I5(,-I900)1900 s 1 s 1960
lions involving digital computers. Allhough such compulalions can be eXlremely
Finally lhey considercd lhat lhe lakc had a background chloride concentralion of ~ userul, mosl practilioners perform a significant amount of what is called "back-of-the-
3mgL-1. 1 envelope" modeling.
According lo O'Connor and Mueller, Lake Michigan had the following average As the name implies, lhis alternalive approach involves skelching out a solution
characteristics for the period from 1900 lo '960: oulflow = 49.1 x 109 lE yr-I and on a piece of papel' no bigger lhan lhe back of an envelope. Although this mighl seem
volume = 4880 X LO"m3. Calculale l!te chloride concenlralion in Lake Michigán r..óiñ a lad "Iow-Iech," a 101of w;eful modeling gels aecomplished in this manner.
1900 lhrough 1960. The pasl lwo leclures have been orienled loward such calculalions. In 1.ec. 3, 1

{ Soludon: Because ch'oride is a conservalive subslance. lhe eigenvalue is simply lhe ¿ presented Ihe steady-stale and general solutions for a well-mixed walerbody to provide
\ quick eslimales of lhe "how much?" and lhe "how long?" of load reduclions. In lhis
( reciprocal of lhe residence time:
49.1 X 109 leclure lhe shape paramelers provide a rapid means lo eXlfllcllhe essential fea tu res of
(
1 A= Q
,
-V 4880 X 109 = 0.01 yr-I ( concenlralion response lo a variely of loading shapes. In all cases a pencil, a pad. and a
\ pocket calculator are all the computalional firepower required lo oblain rapid eslimales.
Therefore lhe solulion is In some sludies such "quick-and-dirly" estimales mighl be all that's required (01'
,I From 1900 lo 1930: affordable) to address a water-quality management queslion. Furlher, even when large

~: 229 X 109 ~) compulalions are inevilable, back-of-lhe-envelope calculalions can provide invaluable
e = .3 + 4880 X 109(0.01 + 0.015) (e+O,OI5(,-I900)
- e-O.OI(I-1900) insighls and checks.
I Finally a back-of-lhe-envelope approach can somelimes be very effeclive for com-
'~
From 1930 lo 1960: municaling results to decision makers. In the hands of a masler modeler, the exercise
can amounl lo an elegant distillalion of lhe essential elemenls of a problem solution.
\ 229 X 109
(' = 3 + 4880 x 10"(0.01 + 0.015)
(e+OOI5(1-I900)
- e-o,Ol(t 19O())) As such it can serve as a vehiele for expressing complex model resulls in a elear and
easily underslandable fashion.
13.2 x lO" -0.01(1-1930) Much of lhe remainder of the text points toward large-scale numerical calcula-
lions. However. wherever possible. I will allempl to provide concise analyticalmeth-
- (0.()J)24880
x 10,)[1- e - 0.01(1- 1930)]
ods lo facililate your own back-of-lhe-envelope calculations. In particular many of lhe
Nole Ilml inilial condilions are nol included explicilly because Ihey are reflecled in lhe end-of-lecllIre problcms muslnecessnrily be kepl brief and poinled. As such Ihey can
conslanl hackgl'llllllll k\'l'1 01' -' mg L l. The results. along with data. are depictcd in pruvide gnnd praclke :lI1dexpcriel1ce in develllping quick. bmad estimates lu cumple-
Fig. E4.4. 1menl ynur cnmputer mndeling.
--
 80 ~ART. Completely Mixed Systems
IHTIJKI' 4 Particular Soluliolls HI
. .
4.7 FOURIER SERIES (ADVANCED TOPIC) Another example relates to the design of a detention basin. Suppos.: lhat during
the summer a eompany discharges a small l10w of a highly eoncentrated contaminant
Aside from the sinll"".,1 presented in Seco 4.5, there are other periodic inpllts into a tributary of a larger river. The company diseharges from 6:00 P.M.unlíl 6:00 A '1
(Fig. 4.8). Such functions can be represented by a Fourier series Measurements at the mouth of the tributary indicate that a half-sine wave provides .iIJ
..~ adequate approximation of the eoneentralion time series over Ihe period

W(t) = ao+ ¿k=1[akcos(kwt) + bk sin(kwt)] (4.36)

" Cin(l) = c" sin(wI) o :S I :S TI'
2

) where w = 27TITpis referred to as thefundamentalfrequency, and its constant Cin(t)= O TI' < I :S TI'
2-
I multiples 2wo, 3wo, etc., are cal1ed harmonics. The coefficients can be evaluated by
l fTI' where Cin(t) = time series of the inflow coneentration at Ihe tributary mouth (¡.tg Lo")

! ao = Tp Jo W(t) dt (4.37) and Ca = amplitude of the half-sinusoid (¡.tg L-1). Nole that time zero is assumed lo be
at 6:00 P.M.

11 is proposed that the company build a detention basin at Ihe tributary mouth lo
2 (1' for k = 1,2. .. . (4.38) moderate the effeel of the diseharge on the river. The flow in lhe tributary (including Ihe
(j ak = TI' Jo W(t)cos(kwt)dt diseharge) can be adequalely eharaeterized by a constant level of Q = Ix 10('rrr!d-I. In
. addition the detention basin's volume is Ixl 06 m'\ and Ihe amplitude of the coneentration
and 2 fT" for k = 1. 2, .. . (4.39) , is 10 ¡.tg L -l. If no losses oeeur, determine the long-term coneentration response using
\ bk = Tp Jo W(t) sin(kwt) dt the well-mixed-Iake model (Eq. 4.1) along Wilh a Fourier-series approximatiun of Ihe
half-sinusoidalloading up lo Ihe seeond harmonie.
The Fourier series represents the loading as a summation of sinusoids. Then this
Solution: The mass balance for Ihe detenlion basin can be wriuen as
summation can be substituted into Eq. 4.1. The individual solutions are then obtained
(for example using Eq. 4.27 for the sinusoidal terms) and the results summed to dc
determine the concentration response. The process is elaborated on in the fol1owing dI + Ac = Aein(t)
example. where the eigenvalue for Ihis example is Q/\ l. The eoeflieienls for Ihe Founer-series
approximation of the half-sinusoid can be e\.. <oled,as in
EXAMPLE 4.5. HALF-SINUSOID FORCING FUNCTION. The half-sinusoid
(Fig. 4.8a) can be employed in a number of water-quality-modeling eontexts. For ex-

~ ample it has been used sueeessfully to simulate the affeet of diurnal variations of plant
I photosynthesis on stream dissolved-oxygen levels.
(/0 = e"-I
o
I
i J/2
sin(21TI)dl = C,,-
1T
I

2 (1/2
(/J = c"l Jo sin(21TI)cos(21TI)dl = O
WW
2 (1/2 1
bl = C" 1 Jo eos(21T1)sin(21TI)dI = e" ()
2
2 (112 2
(/2 = e"l Jo sin(21T1)cos(41T1)dl= col - 31T ( )
2 1/2

W(t)
(a) Half-slnusold
b2 = e,,-
I o isin(21T1)sin(41T1)dl
=O

Therefore Ihe loading funclion can be approximaled by (wilh e" = 10)

10. 20
Cin(/) = -71" + 5SIl1(21Tt)
- _371"eos(41Tt)
NOle that the last term can be represented as a sine with a phase shift of 71"12,

FIGURE 4.8
Two commonly used periodic inputs. sin (471"1+ Í) = eos(471"t)
(b) Pulse traln (a) Half-sinusoid; (b) pulse train.
 LECTURF.
4 Particular Solutions 83
82 PARTI Completely Mixed Systems

Therefore the loading function becomes

BOX 4.2. Donald O'Connor and Closed-Form Solutlons
11"
. (4
e¡n(t) = Ti +
10
5 2
sm( 11"1)- 311"sm
20.
\ 11"1+ 2" ) You may be wondering
several other parts
why in this age of computers
of this book) to analytical
I've devoted
or "closed-form" solutions.
this lecture (and
In fact prior to
This approximation is displayed in Fig. 4.9a. Notice that although it is not perfect (there S 1960 such solutions provided the only means to address complex mathematical prob-
are evcl1 sume slighlly negative valuer.), the series provides a reasonable approximation ) lems (recall Fig. 1.6).
of Ihe half-sinusoid. Way back then there were two primary reasons for developing such solutions. On
To evaluale concentration, the approximation is subslituted inlo the mass balance ( the one hand certain mathematical manipulations can be expressly designed to generate
to give ... 1) answers in an efficient manner. An example was the use of "equivalent resistances" to
simplify circuit analysis in electrical engineering. In water-quality modeling, Iineari7.a- (' I ..
de
dI + ÁC =
10,\ .
--;¡,- + 5'\sm(211"1) - ).;sm
20,\ . (
,411"1+ 2" )
11"
tion was often employed to make otherwise impossible or difficult problems tractable. )
Today because computing power is so accessible, some of these manipulations have
become less utilized and more of historical interest.
The rcsponse can be determined as
The second reason for developing analytical solutions is to provide insight. Thus
.
_
11"
\ )
(" =
10
11"
+
5'\.
J,\2 + (211")2
sm(211"1
- O.) -
20A.
311"J,\2 + (411")2
(
sm 41r1+ _2
- 02 ? closed-form calculus can be employed to neatly summarize and iIIuminate the essential
behavior (lf the systems being modeled. A classic example is the use of dimensionless
numbers to characterize systems involving nuid dynamics. Such manipulations are far
where from outmoded and remain highly relevant in all areas of engineering.
In the field of water-quality modeling Prof. Donald O'Connor of Manhattan Col-
1211"
01 = tan I~T ) = 1.413 radians (= 5.5hr)
lege has been the main innovator and proponent of such insight-oriented closed-form
solutions. Although he certainly had predecessors (notably Harold W. Streeter, Gordon
Fair, and Clarence J. Velz), Dr. O'Connor is generally considered the father of water-
quality modeling. Starting with his doctoral work on stream reaeralion in the 1950s
02 = tan-I(4~)= 1.491 radians(= 5.7hr) (O'Connor and Dobbins 1958), he has published (and continues to publish) a series of
The results, displayed in Fig. 4.9b, indicate that the detention pond has a definite effect significant papers over the past four decades Ihat have advanced and guided the devel-
on the concentration discharged to the river. Whereas the innow concentration swings . opment of our field. Although all involve generaling quantilative results, their hallmark
isthe fundamental insights they provide.
between O and 10 ¡Lg L-I, the pond concentration moves between aboul 2.4 and 4.1
¡Lg L -l. Further, the peak occurs at about 4:00 A.M.rather than at midnight. Beyond acknowledging Dr. O'Connor's contributions, I also wanted to emphasize
the importance of closed-form solutions al this junclure. They are especially critical
nowadays because you can implementjust about any waler-quality-modeling calcula-
c1o(/) t: tion on Ihe computer. As will be stressed repealedly throughoul Ihis text, use of com-
10 puler models as "black boxes" can lead to blunders and erroneous interprelalion of

o
. \ results. Insight-orienled analytical solutions are one imporlant and effeclive way to
~ build up modeling inluilion and use computer-generaled results wisely.
, ...',-
~...,
..;,' ,-.'-. ~...,
'.;" ,
1st harmonJe 2ndharmonle
FourJer approxJmatlon PROBLEMS
c(t) 4
4.1. Use calculus lo derive Eq. 4.6.
.1 4.2. Use calculus to derive Eq. 4.18.
o
. "" ., , \lo
, .
, ""
'-
., 4
4.3. Use calculus lo derive Eq. 4.27 for the case 8 = O, W = O,and the transients have died
3 out. Hilll: if you use Laplace Iransforms, recognize that the following partial fraclion
/-{
1st harmonle 2nd harmonle expansion applies:
(WIV)w As+B D
FIGURE 4.9 -+-
(.f2+ w2)(s + '\) .f2 + w2 s + ,\
Loading and response for a detention basin. (a) Inflow
concentration approximated by a half-sinusoid; (b) the
4.4. A spill of 5 kg of a soluble peslicide takes place into a well-mixed pond. The pesticide
resulting concentration in the basin.
is subjecI lo volalilization that can be characlerized by the following tirst-order nux:

\t
 -_._-

IHTUI(E 4 Particular SOIUI¡ ,i5

H4 PARTI Completely Mix.ed Systems

J = vye 4.11. The following data describe Ihe seasonal cyele 01' so!tlr energy incident on a snmll
mountain lake:
where v,. = a volatilization mass-transfer coefficient 01'0.01 m d-I. Other parameters
for the lake are Monlh F M A M A s () N D
Surface area = 0.1 X 106 m2 Solar (cal cm.2 di) 73 157 278 4()6 505 549 527 443 322 1\14 95 51
Mean depth = 5 m
Outftow = I X IQ6m) d-I
Fit the following funclion lo Ihis dala (thal is, eslimate Ihe paramelerS):
(a) Predict the concentration in the lake as a function 01' time.
(b) Determine the 95% response time for the system. J(t) = J + Jam!,sin(wl - 8)
(e) Calculate the time required for the concentration to be reduced to 0.1 J.LgL -l. Assume a 360-d year.
4.5. A well-mix.ed lake has a steady-state concentration 01'5J.Lg L -1 01'total phosphorus. At 4.12. A well-mix.ed lake is subjecllo Ihe following pulse-lrainloading:
the beginning 01' 1994 it received an additionalloading 01'500 kg yr- I from a fertilizer
processing plant. The lake has the following characteristics:
lnftow =
= outftow 5 X lOS m) yr-I ,
Volume = 4 X 107 m)
Surface area = 5 X 106 m2

11'total phosphorus settles at arate 01' 8 m yr-I , compute the concentration

from 1994 through 2010.
in the lake ~" DJln un n
o 0.5 1.5 2 2.5 3 3.5 4 4.5 5 I (yr)

4.6. For the lake described in Probo 4.5 suppose that. rather than a fertilizer plant, a small FIGURE P4.12
subdivision will be located on the lake in 1997. The population 01'this subdivision can
be described by an ex.ponential model:
The concenlralion response 01' a CSTR lo such a loading can be conslrucled using Ihe
p = 200é'" step and the general solulions. Determine Ihe response 01'a small pond lu such a loading
from I = O lo 5 yr. Note that the initial condition is e = O at I = O. The pond has
where Gp = first-order population growth rate = 0.2 yr-I. 11'each persongenerates
a residence lime 01' 2 yr, a volume 01' 100,000 mJ, and a pollulant reaction rate 01'
0.5 kg yr- I 01'total phosphorus, calculate the concentration in the lake from 1997 k = 0.25 yr-I.
through 2010.
4.7. Determine the concentration in the lake described in Probs. 4.5 and 4.6 if both the 4.13. Repeat Probo 4.12, bUI use the Fourier-series approach lo ubtain your solulÍon. Nole
thal the Fourier series for the pulse wave shown in Fig. P4.12 is
fertilizer plant and the subdivision are allowed to discharge total phosphorus. Perform
your computations from 1994 through 20 IO. 40 ' l. l.
2
4.8. A lake has the following characteristics: W(t) = 10 + 7T [ Slll( 7T1)+ 3 Slll«(m/) + 5 Slll( 10m) + ... ]

Inftow = outftow = 20 x 106m) yr-I Use three terms for your solution and compare il with Ihal obtaioed in Probo 4.12.
Mean depth = 10 m 4.14. A radioaclive sllbstance decays by firsl-order kinelics and forms a dauglller sllbslance
Surface area = lO X IQ6 m2 that also decays in a first-order fashiun:
A canning plant discharges a pollutant to the system that decays at arate 01' 1.05 yr- l. (, k2
(' 1 ' ('2 +
Recause 01' the seasonal nature 01' the product being canned, peak waste discharges
occur in the fall (October 1) and minimum levels in the spring (April 1). The di s- You perform an ex.periment in which you place so me 01' Ihe first compound in a batel!
charge has a mean level 01' 30 X 106 g yr-I, with a range between the ex.tremes 01' reactor. The following data is collecled over lime:
30 X 1<1 g yr-I. Ifthe initial condition is e = () at I = O,
(a) Compute the concentration in the system from I = O to 10 yr.
(b) Al'ter sufficient time the concentration will reach a dynamic steady-stale. At this
I'" (d)
(uCi L ') I O
10 20
6.3 40
4.0 60
2.5 80
1.6 IO()
LO
poiot, on what day 01' the year will the in-Iake concentration be at a max.imum
value?
In addition the peak concentration 01' Ihe second compuund occurs at 35 d. Use this
4.9. Repeat Ex.ample 4.5 for the case where the contaminant reacts at a first-order rate 01' information to determine Ihe half-lives 01'the compounds.
0.1 d -1 . Interpret how the reaction affects the mean, amplitude, and lag 01'the response.

4.10. Derive Eq. 4.32.

 LEcrUREs Feedforward Systems of Reaclors 87
LECTURE 5
.. - -- - - ---
Feedforward Systems of Reactors

(8) Feedforward reaetora

FIGURE5.1
Two general ways In whlch
reactors can be connected.
(b) Feedback reaetora The dashed arrows in (b)
represenl feedback.

We will keep the following derivalion simple by IilOiting it to two lakes in series
(Fig. 5.2). Mass balances for these reactors can be written as
dCI
V, di = W1 Q'2CI - - k,V'C, (5.1)
dC2
1develop steady-state and time-variáble sólutiol1sfor cou" V2-
\. LECI'URI OVERVIEW:
pled reactors without feedback and apply these solutions to a series bf takes. ln
dt
= W2 + QI2C, - .
Q2JC2 - k2V2C2 (5.2)

addition 1extend the analysis to a series of first-order reactions withirt a single : ( Note that both lakes are assulOed to have conslant volulOes. For silOplicity the inflow
reactor. ", I into the first reactor has not been shown explicitly.
\
~

The terlO QI2CI represenling Ihe outtlow of lOass frolO Ihe first lake appears in
S
¡ both balances.
2 depends Thus,
00 the because
solution lOassflows
for lake l. frolOlake I lo lake 2, the solution for lake
To Ihis poinl we have exarnined Ihe sleady-state and tiDle-variable solutions for a
single eSTRo We can now attempt to model more complicated systems by combining , At sleady-slale Ihe lerlOscan be collected and expressed as
reactors.
S; a'ICI = W, (5.3)
Figure 5.1 shows two general ways in which completely mixed reactors can be
connected: feedforward (Fig. 5.1a) and feedback (Fig. 5.1b). The difference between where
-a2lc, + a22c2 = W2 (5.4)
Ihe two types of configurations is thal in the feedforward setup, water never flows
\. through the same reactor twice. This greatly simplifies both steady-state and time- al! = QI2 + kl VI
(5.5)
variable solutions. \ a21 = Q'2 (5.6)
Aside from providing a model for systems of reactors, the feedforward frame-
a22 = Q23 + k2V2
(5.7)
work is also useful for characterizing natural water systems. For example il can be
employed lo model a chain of lakes or a slream. Furthermore, it can be used to sim-
ulate serial reactions.

Q2jC2
5.1 MASS BALANCE AND STEADY-STATE

The reactors in series model applies lo a variety of problem contexts in environmental

engineering. AlI10ng Ihe lI10st straighlforward is a series of lakes connected by short
ri verso FIGURE 5.2
Lakes in series.
86
 -- - -_. ---

88 PARTI Complelely Mixed Syslems

LECfUHI:j Feedl"orward Syslems 01"Rea<:lors !S9

We can now see Ihe advantage of Ihis soft of system. Because the reactors are in
series, we can solve Ihe equalions in series. That is, we can solve Ihe first equation = 596.13 + 38.86 = 634.99 ¡.tgL

for (', and Ihen substitute Ihe result into the second equation, which can be solved
r ;lISEq. 5.3 can be solved for
CJ=
Q34
WJ
+ vAJ
+_ Qnl'2
QJ4 + l'A3
l l
cl = -WI (5.8) I x 10~ + 1.0x IOh(634.99)
al! Q'2 + klV, W1 I x 106+ (10 x 1x 1()6) 1x 106+ (10 x I x 1(6)
This result in lurn can be subsliluted inlo Eq. 5.4, which can be solved for = 148.64¡.tgL-I
C2 = W2 + a2lc.
a22
= l
Q23+k2V2
W2 + QI2 l
Q23+k2V2Q'2+kIVI
W, (5.9)
(b) The delerminalion of how mu<:h 01"the Ihird reaClor's <:oncentralion b due lo Ih<:
loading lo Ihe second reaclor can be eSlablished by inspecling Ihe solulion fiJr C2 above.
We see cIearly Ihal Ihe concenlralion of lake 2 depends on the loadings lo bOlh ilself As can be seen, 596.13 ¡.tg L -1 01"C2 is due lo direcl loadings (thal is, ",2), whereas
and the loading lo lake 1. Becunse Ihe underlying model is linear, the effects of Ihe 38.86 is due lo Ihe loadings lo reaclor l. Therel"ore the elTeel 01"Ihe second 1eactor on Ihe
Ihird reaclor can be calculaled as
( Iwo loadings are independen!. As in Ihe following example. Ihe approach can be
11easily extended lo addilional reaclors.
EXAMPLE 5.1. LAKES IN SERIES. Suppose Ihal Ihree lakes connecled in series cJ(due lo loading lo reaclor 2) = 1.0 x IOh(596.13)
I x 1()6 + (10 x 1 x J()6)
= 54.19,..g 1._1
have the following characleristics:

5.1.1 Cascade Model

2An inleresling inSlance of reaclors in series is Ihe case where Ihe reaclors' size amI
Iftow are identical. This is somelimes referred lo as a cascade of CSTRs (Fig. 5.3).
For Ihis case Ihe equalions are simplified,
1 2 3
Q
3 Ct
Vo\ume, 106mJ 2 4 Q+ kVl'o (5.10)
Mean deplh, m 3 7 3
Surfaee area, \ ()6 m2 0.667 0.57\ 1.000 Q Q Q.
C2
2000 4000 1000 Q + kVc1 (5.11)
Loading, kg Yíl Q+kVQ+kV(O
Flow, 100 mJ Yíl 1.0 1.0 1.0 1.0

Q
"
lf Ihe pollulanl seules al a rale of 10m yr-I ,
(a) Calculale Ihe steady-stale eoncenlration in each 01"Ihe reaelors.
C" =
(Q + kV ) Co (5.12)
(b) Delermine how mueh of Ihe eoncentration in Ihe Ihird reaclor is due lo lhe loading
to the second reaClor. Thus we can see Ihal each reaclor altenuales Ihc concenlralion by a constan! fraclioll
because Ihe parenlhelical lerm will always be \ess Ihan one.
Solutioll: (a) The concentralion for Ihe reaelors can be delermined by

CI =
WI
QI2+vAI
2 X 109 _ = 260.76 ¡.tg 1 -1

C2=
W2
Q23+ vAz
+-
QI2CI
Q23+ vAz
4 x 109 1.0 x IOh(260.76)
~...~ 1 2 3 n-2 n-1 n
= + FIGURE 5.3
1.0 x \()6 + (10 x 0.57\ x \()6) 1.0 x \()6 + (10 x 0.57\ x \()6)
Cascada 01 CSTRs.
 90 PART1 Complelely Mixed Syslems LECTURE5 Feedforward Syslems of Rcactors 91

EXAMPLE 5.2. CASCADE MODEL OF AN ELONGATED TANK. Use Ihe cas- Therefore for Ihe single-segmenl approximalion,
cade model lo simulale the steady-state distribution of concentration in an elongated
lank.
c(50) = Looo~~Iooo)] / = 0.3333mgL-1
whereQ = VA = 100(10)= 1000.Similarlyforn = 2,

~ ~ c(25) = [10001~500)] I = 0.5 mg L-I

l' l' l. I I " ," I I .~
1000 2
I I . I ! I I I . I t I I {";¡
1.._ _ L .!.._ _ ! l..- _ J 1..- _ J _ !.. _ _ L !..._ _ L _ !.. _ _ L.!.. .~..
I I ... 1 1 _ I I I .... and c(75) = [1000 + 2(500) ] I = 0.25 mg L-I
--- The olher cases can be computed in similar fashion. The reslllts, as summarized in
óx
et Fig. 5.4, are inleresling. As more (and smaller) reaclors arc used, Ihe sollllion approaches

'1
O
, -L
, x 1) a paltern Ihal looks jusl like an exponential decay. We will explore this observation in
7 more delail when we invesligale models of elongated reaclors in Lec. 9.

The tank has cross-sectional area A,. = 10 m2, lenglh L = 100 m, velocity V = 100 5.2 TIME VARIABLE
m hr-I, and lirsl-order reaction rale k = 2 hr-I. The inflow concentralion is 1 mg L-l.
Use 11 = 1, 2, 4, and 8 CSTRs lo approximate Ihe lank. Plot the results.
Equations 5.1 and 5.2. with zero loadings. can be wriUen as

Solution: As in Eq. 5.12, Ihc lIlodel for such a systclIl is -dCI

dI = -Al)cl (5./3)
"
Q
C" =
(Q + kV )Cn and dC2
-dI = A21CI - A22C2 (5./4)
where
e n=1 QI2
0.5 AII =-
VI + k, (5.15)
o QI2
20 40 60 80 100 A21=- (5./6)
1o V2
e n=2
0.5 Q23
A22= - + k2 (5./7)
o V2
o 20 40 60 80 100
1 Again. as was Ihe case for sleady-slate. Ihe equalions can be solved in sequence.
e If CI = CIO and C2 = C20 all = O. Ihe general solulion for Eq. 5.13 can be devel-
0.5
n=4 oped as
o =+===1 r C - C e-AII'
o 20 40 60 80 100 I I 10- (5./8)

e I This result can be subsliluled into Eq. 5.14. which can Ihen be solved for
0.5 n=8
(.' C2 = + (e-AII1
o FIGURE 5.4
Approximalion 01 an
I
c20e-A221 A21CIO
A22 -Al)
_ e-A22'
) (5.19)
o 20 40 60 80 100
elongaled lank by a
X series 01 n eSTRs. Note also that Ihis result is identical to Ihe solulion for a single lake wilh an expo-
nentialloading (Eq. 4.18).
,------.

LECTURE5 Feedforward Syslems 01' Readors 93

92 PART1 Complelely Mixed Syslems

Di Toro (1972) furlher showed Ihallhis form requires 2" 1 evalualions 01' Ihe el
Now suppose Ihat Ihere are a Ihird and a fourth lake in series. For Ihese cases
function (Eq. 5.22). Since Ihere are only 1/ independenl values oflhis funclion. Eq. 5.26
solutions are (O'Connor and Mueller 1970, Di Toro 1972)
is inef1icienl. He Ihen offered a more efficienl version based on a parliul-fraclion ex-
pansion. The general form 01'Ihis faclored version is
C3 = C30e-A.n' +
A33 A22 - (
A32C20 e-A22' - e-A)]' ) ,,-1 11
('1(1. Aj,iJ
(',,(l. AII A,,_I.,,_I.A".,,) = fI ¿;-1 " (5.271

+ A32A21 CIO
e-AII' - e-An' _ e-A2¡'- e-A))' (5.20)
j=1
Aj+J.j
n (A),) - A1,' )

A22 - AII ( A33 - A\I A33 - A22 )

j~ l(j~;)

and C4 = C40e-A44' + A43C30

Á44 A33 - (e-A))' - e-A...') EXAMPLE 5.3. TEMPORAL RESPONSE OF LAKES IN SERIES. During Ihe
lale 1950s and early 1960s, nuclear weapons lesling inlroduced large quanlities 01' ra-
dioaclive subslances inlo Ihe almosphere. As Fig. E5.3-1 shows, Ihis resulted in a falloul
e-A22'- e-A...' _ e-A))' - e-A«' flux 01' Ihese substances lo Ihe surface 01'Ihe earlh.
+ A43A32C20
A33 - A22 ( Á44 - A22 A44- A33 ) 00

tñ~ 25
e-AII'- e-A..' _ e-A", - e-A..' :i !.. 20
+ A43A32A21CIO
(A22 - AII)(A33 - A\I) ( A44- A\I A44- A33 ) o....
)( I 15
::JE
lO:._
_ A43A32A21CI0 e-A22'- e-A«,
- e-An' - e-A..' (521)
_ () 10
60> FIGURE E5.3-1
(A22 - A\I )(A33 - A22) ( A44- A22 A44- A33 ) .
:fg 5
o
Falloul Ilux 01 90Sr lo Ihe Greal Lakes (Irom
Lerman 1972) along wilh Ihe impulse load
Inspection of Eqs. 5.18 Ihrough 5.21 suggesls Ihal a pattern is emerging. 1975 used lo approximate Ihe input.
O'Connor and Mueller (1970) first nOled Ihis pattern and applied il 10 simulale
Ahhough Ihe falloul has conlinued beyond Ihe 1960s, Ihe pronounced peak in 1963
I lakes in series. Di Toro (1972) developed the insighl further by recognizing Ihal Ihe allows idealizalion 01' Ihe resuhing load as an impulse funclion,
7 pattern could be expressed as a recurrence relalionship (Box 5.1). W(t) = l"AJi(l- 1963)
where 1" = 70 X 10-9 Ci m-2 (Ci denoles Ihe radioaclivily unil Ihe curie)
A, = lake surface area (m2)
BOX 5.1. Efflclent Schemes to Compute Concentrations lor Serial Systems c'i(l- 1963) = unil impulse funclion located al IIJ())
Inspeclion 01' Eqs. 5.1810 5.21 suggesls a pallern Ihal Di Toro (1972) expressed as a Predicllhe rcsponse urlhe Ore;11Llkes lo Ihis l1ux¡rlhe hulr-lire 01''H'Sris ;Ippru"imatcly
recurrence relalion. For example for Ihe case where only Ihe (irsl lake has an inilial 28.8 yr (k = 0.0241 yr 1J.
condition ('10, Ihe equalions can be reformulaled as
Solution: The Oreat Lakes can be represenled as a series o" reaclurs:
CI(I. ÁII) = clOe-AlI1 (5.22)
Á2\
C2(1.ÁII' Á22) = \
[
22 - Á 11 C\(l. A\I) - CI(I. A22)1 (5.23)
Á32
]
(')(1. AII' A22. AJJ) =, 1
JJ - A22 ('2(1. AII' A22) - ('2(1. A\I. AJ.1) (5.24)

- cJ(l. AII' A22.A44)I

A4J
C4(1.AII' A22.AJJ. Á44) = '.44- AJJ ICJ(I, AII' A22.AJJ)

The sequence can be generalized fUrlher for Ihe case 01'Ihe I/Ih reaclor,
(5.25)
Jrr~-l!r-l!r-
11-1 n l!r Lake
Huran Lake
Erle Lake
Ontarlo
('\(1. Aj,i) (3) (4) (5)
n (5.26) Lake Michlgan
Cn(l.AII An-I.n-a. An.n) = nAj+l.j¿
j= I ;-1
n
j= IIj..i)
( A),) -A 1,1) (2)
FIGURE E5.3-2
 94 PARTI Completely Mixed Systems LECTUREs Feedforward Systems of Reactors 95

4
Both Lakes Superior and Michigan are "headwater" lakes. Their outftows feed into Lake
Superior Michlgan
Huron, which discharges to Lake Erie. Lake Ontario is the last lake in the chain. The
parameters for the system are summarized as

Huron Erie Ontario

Parameter Vnits Superior Michigan
146 85 59 19 86
Mean depth m
106m2 82.100 57.750 59.750 25.212 18.960
Surfacearea ~o
12.()()() 4,900 3.500 468 1.634 '"
Vnlnme 10. m3 I
67 36 161 182 212 E4
Olllnnw 10" m" yr I Huron
Ü
O>
I
52
Thc initial concentration in each lake in 1963 can be computed by e 2
i.... (E5.3.1) ~
Co = H ~
eel)
o
where H = mean dcpth (m). The results are 50
u 1980 FIGURE 5.5
4 Response 01 Ihe Greal Lakes
Onlarío
Vnits Superior Michigan Huron Erie Ontario lo an impulse loadlng 01 90Sr
0.814 in 1963. Dala (.) lrom Lerman
0.479 0.824 1.186 3.684
Cn lO'. Ci m-3 (1972), Alberts and Wahlgren
(1981), and Inlernational Joinl
Commission (1979). The
dashed line represenls Ihe
Equations 5.18 through 5.21 can then be applied to compute the responses of each response 01 Ihe lake lo lIs own
lake, and the total solution arrived at by summing the individual components. As in loading excluding Ihe effecl 01
1960
Eq. 5.18 the model for Lake Superior is upslream lakes.

·
('1 = 0.47ge-11.029681
~ For a slowly decaying contaminant like 9OSr, the upstream Great Lakes have a sig-
and for Lake Michigan is
nificant elrect on the downstream lakes. The effect on Lake Ontario is so pronounced
C2 = 0.824e-o.O.Jl4S1 that its peak concentration does not occur in 1963 but lags 2 lO 3 yr due to upstream
effecIs.
Equation 5.18 is also used to predict how Lake Huron purges itself 01'its initial con-
centration. However, to compute the total response Eq. 5.19 is also employed to calculate
the cffect 01'Lakes Superior and Michigan on Huron's concentration:

CJ
= 1 186 -0.07011
. e
+ 0.01914(0.479)
0.0701_ 0.02968 e ( -O.02%Rr _e .-O.0701t
)
5.3 FEEDFORWARD REACTIONS

Althoughwe are emphasizing reaclors in series, the models developed in this leclure
0.01029(0.824) -O.OJI4S1 -l1070It

+ 0.0701_ 0.03145(e - e
) 11-
can jusI as easily be applied lo a series of reaclions occurring wilhin a single reaclor.
For example if Ihe following series of reaclions
(
The concentrations for the other lakes can be determined in a similar fashion. The
) A -> B -> e -> D -> ... (5.28)
results along with data are displayed in Fig. 5.5. The simulation duplicates the general
trend of the data, with the exception that the computation decreases somewhat faster follows firsl-order kinelics, Ihe mathematics described in the previous section are
than the data. This is due, in part, to the use of an impulse forcing function to idealize directly applicable.
the continuous loading function. Although environmental reactions do proceed as in Eq. 5.28, they do nol involve
This analysis results in two conclusions:
an infinite chain of reactions. For example the following sequence is an interesting
. If two lakes receive an cqual impulse flux of a pollutant, their response is inversely
and relevant case:
proportional to their depth (Eq. E5.3.1). This is the reason why shallow Lake Erie's k h
initial concentration is about 4 times higher than for the other lakes. A~B~C (5.29)

- -. .-- -.- ._p

 LECruRE 5 Fecdforward Systems of Reactors 97
96 PARTI Completely Mixed Systems

This situation differs from those in the previous section in that the compound B
C does not undergo further reaction. In effect it represents a dead end. Such serial
{ reactions appear in a number of water-quality contexts, the most noteworthy example
t being the nitrification of ammonium to form nitrite and ni¡rate.
If the transformations are assumed to be first-order, the mass balances can be
written as
dCa = kabCa
dt
_ (5.30)

dCb = k"bC" - kbcCb (5.31)

dt

dcc (5.32) FIGURE 5.7

-;¡¡ = kbcCb Concentrations 01 a series 01
three reactants mapped on
If Ca = CaO,Cb = Cc = Oat t = O,the solution is
(5.33) A e a triangular space (redrawn
lrom Fogler 1986).
Ca = caoe-kUbl

(5.34)
of our three reactants, A, B, anl! C. If we set up our experiment with the initial
Cb = k"b
kabCaO(e-kb<1
- kbc - e-kUbl) conditions, Ca = C"O,Cb = Cc = O,the initial state would begin at vertex A. Then
the reaction would move up and to the right as first B and then C are created as time
(5.35) progressed. After a certain time, B would pass its peak and begin to decrcase. Thus
Cc = CaO- caOe-kUbl
- kab
kabCaO(e-kb'"
- kbc - e-kUbl)
the curve would move down toward the lower right vertex as the vessel approached
An example simulation using these equations is displayed in Fig. 5.6. Notice apure solution of C. Now the height of the curve, which indicates how much B
that the three species must sum to the initial condition of reactant A. Also observe is formed, depends on the relative values of k"b and kbc. If kbc « k"b, significant
howthe solutionforreactantC (Eq.5.35)is merelyCc = CaO - Ca - Cb. amounts of B will be formed as the second reaction acts as a bottle neck. Conversely
Now an interesting question relates to the middle reactan!. Since it is being if kbc » k"b, negligible amounts of B will be formed.
An alternative way to look at the same phenomena, but in the more familiar
gained and lost simultaneously, when will its concentration be significant? Con-
versely could there be cases where the reaction from B to C would be so much faster temporal context, is provided by a dimensionless analysis. In such analysr.. dimen-
than from A to B that B would never reach significantly high levels and hence could Ssionless groups are developed so that system parameters are incorpOfaicd IlIto the
be ignored? 1/variables. Jf done properly this generalizes the solution behavior as a function of a
The situation is depicted in the triangular or phase diagram in Fig. 5.7. Such few dimensionless parameter groups.
( reprcsentations are commonly u:.el!in fields like chemical engineering to map sys- For the present case this can be done by defining
1 tems that manifest three states. In our case each vertex represents apure solution t* = kabt (5.36)

50 *
Ca -_ C"o
C"
C¡;
Cb

= C"o
c.c _- Cl.:
C"o
(5.37)
40
k* = kb,. (5.38)
30 kab
e
20 Equations 5.36 and 5.37 can be solved for nondimensionless time and concen-
trations and the results substituted into Eqs. 5.30 to 5.32 to yield
10
dc~
(5.39)
FIGURE 5.6 dt* = -c:
2 3 4 Concentrations 01 a series
01 reactants where the last
d * ..
t .
is a dead end."
r:'I
~dt' = c:- k cb (5.40)
 LECTURE5 Feedforward Syslems of Reaclors 99
98 PARTI Complelely Mixed Syslems

lo omitting B from the analysis. The graph shows thal results for e are quite cIose
for the two cases, indicating Ihat for all intents and purposes we would be justified
0.8 in omitting B.
0.6 FIGURE 5.8
c*2
0.4 Plot 01dimensionless PROBLEMS
concentration versus
0.2 dimensionless time lor the
second 01 three reactants. The 5.1. A spill of 5 kg of a soluble peslicide lakes place in Ihe firsl of Iwo lakes in series. The pes-
diflerenl lines represent licide is subjecl lo volalilizalion that can be characterized by Ihe first-order flux J = v.c,
2 4 6 8 10 diflerent values 01 the where v. = a volalilizalion mass-Iransfer coefficienl of 0.01 m d-I. Olher paramelers
for the lakes are
t* dimensionless reaction rate.

LakeI Lake2
de;.
dt* =
k
* *
Cb
(5.41) I Surface area (m') O.lxlO" 0.2 x lO"
Mean deplh (m) 5 3
Oulf1ow(mJ yr-I) I x lO" I x (()6
The ¡n¡lialconditions are now c;' = 1, c¡' = c; = 0, and Ihe solution is
(5.42)
c~ = e-f
(a) Predict the concenlralion in bolh lakes as a funclion of lime. Presenl your results as
C
*
b
= -l -l k' (e-k',' -e -,. ) (5.43) aplot.
(b) The second lake is used as a waler supply reservoir. Therefore it is crilical lo de-
lermine the time required for Ihe second lake lo reach its maximum concentration.
C
*
e
= I-e _,'
-
- l -l k*(e
-k','
-e
-l'
) (5.44)
Derive an analytical formulalion lo make Ihis determination. Verify your resulls by
comparing it lo the p}OIfrom part (a).
Now. aside from appearing much simpler (compare wilh Eqs. 5.33 to 5.35), the 5.2.
In 1970 lotal phosphorus loadings lo Ihe Greal Lakes were (Chapra and Sonzogni 1979)
power 01 Ihe dimensionless approach is illuslraled in Fig. 5.8. This diagram, which
shows c¡' for various values of k*, illuslrales cIearly that for k* > 5 the concenlration Units Superior Michigan Huron Erie Ontario
of the second reaclanl never rcaches significantly high levels. In other words if the 4000
W lonnes yr- r 6950 4575 18,150 6650
second rcaction is 5 (or more) times fastcr Ihan the first, the second reaclant can be
effectively ignored. Thus the analysis provides a way lo determine when an analysis
could be simplified by ignoring a state variable. Thal ¡s, it provides an objective basis
Using the paramelers from Example 5.3 and assuming Ihal lolal phosphorus settles al a
t"ordeciding when a box (stale variable) becomes an arrow (rale). rate of approximalely 16 m yr-I ,
The conclusion of our present analysis is underscored by Fig. 5.9, which shows (a) Calculale Ihe sleady-slate concenlralion for each lake.
lhe dimcnsionless results for lhe cases where k* = 5 and k* = 'Xi.The latter amounls (b) Delermine how much of each lake's concenlralion is due lo upslream lakes.
5.3. Suppose Ihal 5000 Ci of 137Cs(1so ;;; 30 yr) was spilled into Lake Huron due to a nuclear
accident. Calculate the response of Lakes Huron, Erie, and Ontario lOIhe spill.
~100
el 5.4. Duplicale Example 5.2, bul do nol limit yourself lo n :S 8. Plol your results on semi-
X logarilhmic paper. As n increases, Ihe plot should approach a straight line. Interprel Ihe
~ 80 slope that results.
.2
~e" 60 5.5. You are given the task of building detention ponds to remove settleable solids from a
"e small stream prior lo discharge inlo a lake. The ponds musl be 2 m deep, the lIow in
o" the stream is 20 cfs, Ihe solids settle al arate of 0.2 m d-I, and you must achieve a
lO
lO sleady-slale removal of 60% of Ihe solids.
.!! FIGURE 5.9 ..¡..
e (a) Delermine Ihe size of a single CSTR lo achieve Ihe desired removal.
o Dimensionless concentrations
.¡;
e (b) Delermine Ihe sizes of a pair of idenlical CSTRs in series needed lo achieve the
versus dimensionless time lor a desired removal.
~
i5
2 3 4 5
series 01 reactants where the last
(e) Which seems lo be Ihe besl oplion? Why?
t* is a "dead end."

-....
100 PARTI Completely Mixed Systems

LECTURE 6
5.6. Develop a computer program to implement Di Toro's eflicient scheme for implementing
feedforward first-order systems. Test your framework by using it to duplicate the results
- ...
of Example 5.3.

5.7. Everyone's heard about the five Great Lakes described in this lecture. However, not Feedback Systems 01 Reactors
everyone knows that there is a sixth lake in the system, Lake SI. Clair. that is located
on the river connecting Lakes Huron and Erie. By the standards 01'most other lakes. it's
morethanadequate(V = 6.6km3,As = 1114km2.H = 5.9m,Q = 170.5km3yr-I).
However, in the context of the Great Lakes, it's at best a "Very Good Lake." ~,

(a) Determine the responses of Lakes Huron, SI. Clair, and Erie to an impulse flux of
strontium (as described in Example 5.3) to Lake Huron. That is, compute the re-
sponse as if the fallout fell only on Huron.
(b) Use the dimensionless analysis described at the end of the lecture to justify omitting
Lake SI. Clair from the Great Lakes.

¡~. 0rF~YlEW¡'deyel!)p stea~y-~~te and time-varia~'e solutions fpr cou-

~'. "~iict()ff wit1tf~dback. For the steady-state case, matrices are offered as
\'nlC}ap~19~oncisely represent systems of coupled reactors. l introduce the ma-
tP~jpverse as a meaus fa sort out !he interactions between the reactors. Por tfte
~e-virlable ca!¡e,fue general solution is derived and an ejgenvalue approach
pseOto~ahrinsighnnto the system's dynamics. In addition to systems ofre-
¡¡c~rSthe t~ture also deals with'modeling coupled reactions within a single
~l!9t~r,

In the previous lecture we developed models for reactors in series. Now let's add
feedback ftows to these systems. Although this greatly ¡ncreases the range of ap-
plication of such models, it also complicates their solution. In fact, for all but the
simplest systems (that is, two or three reactors), it means that computers are es-
sential.

6.1 STEADV-STATE FOR TWO REACTORS

Mass balances for two CSTRs with feedback (Fig. 6.1) can be written as
dCI
Vt di = W1 + QOICO - QI2C¡ - k,V1CI+ Q21C2 (6.1)

dC2
V2 di = W2 + QI2CI - Q21c2 - Q23C2- k2V2C2 (6.2)

At steady-state the terms can be collected and expressed as

101
 LECTURE
6 Feedback Systems of Reactors 103
102 P¡\RT1 Completely Mixed SystemS
Algebraic manipulation can also be employed to express these results as
I I
CI W, + W2 (6.13)
all - (a2IaI2/a22) a21- (alla22/aI2)
Quc, I I
QO\CO C2 = al2 - W, + W2 (6.14)

- -- QUCl
(alla22/112d a22 - (a2IaI2/all)

Thus we see thal the concentralions are expressed in the format of Eq. 1.8.
However, lhese solulions are considera31y more complicaled lhan Eq. 1.8. For Eqs.
~ 6.13 and 6.14 the assimilalion factors are now composites of a number of parameter
k1 V lCl '1groups. But due to lhe linearity of the original model (Eqs. 6.3 and 6.4), these results
k.V.c\ can provide greal insight into how the system operates. In particular notice how each
\
equation consists of two independent parts-one that is solely dependent on the firsl
FIGURE 6.1 reactor's loading, W" and the otherdependent on the loading to the second. W2. This
A system of reactors. The dashed arrow represents feedback.
( fact,whichis a productof the model'slinearity,allowsus to separatelyevaluatethe
(6.3)
\ ~ impactof lhe individualloadings.
Cl1\C, +Cl12c2 = WI (6.4) Note lhat for more than Ihree equations Cramer's rule becomes impractical be-
lI21cI + lI22c2 = W2 cause, as the number of equalions ¡ncreases. lhe determinants become too time-
¿ consuming to evaluate by hand (or by computers if they're calculated by expanding
where
a1\ = QI2 + kl VI (6.S) / minors!). Consequently. as described next, more efficient methods are used.
al2 = -Q2\ (6.6)
(/21 = -Q12 (6.7) 6.2 SOlVINGlARGE SVSTEMSOF REACTORS
a22 = Q2\ + Q23+ k2V2 (6.8)
The extension of the mass balance to more than two reactors is straightforward. For
and the loading to the I¡rst reactor must be modified to inelude the inflow input example for three coupled reactors with feedback, a set of three equations with three
WI _ WI + QUICU (6.9) unknowns would be generaled,
allCI + al2C2+ al3c3 = WI (6.IS)
Thus we have two equalions (6.3 and 6.4) with two unknowns. There are a
variely of ways to oblain solutions. One convenient way, which is appropriate for a21CI+ a22c2+ a23c3 = W2 (6.16)
1 smatlllumbers of equaliollS, is C,ame,'s ,,,le. This rule slates that each unknown aJI CI + a32c2+ aJ3cJ = W3 (6.17)
11 in a syi;tem of linear algebraic equations may be expressed as a fraction of two
determinanls with lhe denominalor as the system determinant. For example for a where a's = constants that reflecl the system parameters (that is, Q, V, k, etc.)
two equation system the denominator would be W's = conslanlloadings
a1\ a12 c's = unknown concenlralions
D = a22 \
= lI1\a22 - a12a21 (6.10)
\ (/21
l, Note Ihal the unknowns are all raised lO the first power and are multiplied by a
i constan!. Equations with these characteristics are called linear algebraic equations.
The Ilumerator is obtained from D by replacing the column of coefficients of the
!n lhe previous section we used Cramer's rule to solve two equations with two
unknnwn in question by the constants W. For example CIwould be computed as
WI a12 unknowns. For more Ihan two equations, this technique becomes computationally
\ W2 (/22\ a22WI - a\2W2 (6.\1) burdensome. Therefore in such cases, computers and numerical methods must be
CI = D = a1\a22 - a12a21 used. Several l1umerical melhods are available for this purpose (see App. E and
Chapra and Canale 1988 for details).
Aside from the mechanics of oblaining solulions, the representalion and manip-
Similrlrly the second unknown could be determined as
(/1\ WI ulation of large numbers of equations is diflicult using simple algebra. For this reason
a1\W2 - a21WI (6.12) the next sectiol1is devoted to matrix algebra.
\ Cl21 W2 \
C2 = o-- al1a22 - a12a21
 LEC.'TURE
6 Feedback Systems of Reactors 105
104 PARTI Completely Mixed Systems

Mal algebra follows rules in the same fashion as algebraic manipulations 01'
6.2.1 Matrix Algebra simple \¡¡;I,lbles. In the present context the two most important involve matrix mul-
A matrix consists of a rectangular array of elements represented by a single symbo1. tiplication and matrix inversion.
For example
a 11 al2 an Matrix multiplication. The prodllct of twa matrices is represented as [C] =
(6.18) [A][B], where the elements of [C] are defined as
[Al = a21 an a23
11
[ a31 a32 a33]
Cij = ¿k=1 Uikbkj (6.22)
where [A] is the shorthand notation for the entire matrix and aij designates an indi-
vidual element of the matrix. where 11 = the column dimension of [A] and the row dimension 01'[B]. That is, the
A horizontal set of elements is caIled a row and a vertical set is called a colllmn. t
Cij element is obtained by adding the product of individual elements from the ith
The first subscript i always designates the number of the row in which the eIement row of the first matrix, in this case [A], and thejth column of the second matrix lB].
lies. The second subscript j designates the column. For example element a23is in Box 6.1 iIIustrates a simple way to visualize matrix multiplication.
row two and column three.
A matrix can be characterized by its dimensions. It has 11rows and m columns
and is said to have a dimension of 11by m (or 11X m). Such a matrix is commonly
refen'ed to as an n by m matrix. BOX 6.1. A Simple Method lor Multlplylng Two Matrices
Matrices with column dimension m = l such as Although Eq. 6.22 is well-suited for implementation on a computer, it is not the simplest
means for visualizing the mechanics of multiplying two malrices. What follows gives
(6.19) more tangible expression 10the operalion.
(8) ~ {~} Suppose Ihal we want lo multiply [X) by [Y] lo yield [ZI:

are caIled column vectors. Note that for simplicity the second subscript of each el e-
" 'nt is dropped. Also it should be mentioned that it is often desirable to distinguish [ZI = [Xlly] = r~ 1][; ~]
Ilumn vector trom other I,"pes of matrices. Consequently we use special brackets,
A simple way lo visualize Ihe compulation of [LI is lo raise [Y], as in
}, to endose the values.
Matrices such as the one in Eq. 6.18, where 11= m, are caIled sqllare matrices. 1i'
Square matrices are particularly important when solving sets of simultaneous linear
equations. For such systems the number 01'equations (corresponding to the number [~ ~] +- [YI
of rows) and the number of unknowns (corresponding to the number of columns)
must be equal for a unique solution to be possible. Consequently square matrices of
coefficients are encountered when dealing with such systems.
There are a number of special types of square matrices. In the present context IXI~[~ :][ '! ]+-IZI
the most important is the i.l, ¡¡tity matrix. For this case aIl elements are zero except Now Ihe answer 12] can be computed in Ihe space vacaled by 1YI. This fOrlllal has
for the diagonal that consists 01'ones. For the three by three case, . ulilily because il aligns the appropriate rows and columns Ihal are lo be multiplied. For
1 O O example according lo Eq. 6.22 Ihe elemenl <11is oblained by multiplying Ihe t¡rst row
[1] = O l O (6.20) of IXI by Ihe firsl column of [Y]. This amounts to adding .he producI of XII and .\'11to
the product of XI2 and )'21, as in
[O O l]

rr=[;
This matrix has properties similar to unity. lust as a X l = a in simplealgebra,so
also does
(6.21) ;]
[A]x [1] = [A]

[~ ~] ~ [3 x S + I x 7 = 22]
t A simple mnemonic is provided by the horizontal "1 .ws" of a theater and the vertical "columns" of a
Thus <11is equal to 22. Elemenl <21can be compuled in a similar fashion, as in
temple.

"lo
:\
 LECfURE 6 Feedback Systems of Reactors 107
106 PART1 Completely Mixed Systems
Thus the multiplication of a matrix by its inverse is analogous to division, in the
sense that a number divided by itself is equal to l. That is, multiptication of a matrix
(; ~1 by its inverse leads to the identity matrix.
i The inverse of a two-dimensional square matrix can be ca\culated simply, as in
022 -012
[AJ-1 = I 011 ]
(6.28)
011022 - °12021 [ -021
[~ :] _ ['X'+;~7' 8']
Similar formulas for higher dimensional matrices are much more compticated. Con-
The computation can be continued in this way, following the alignment of the rows sequently computer algorithms are usually used. One simple method involves stan-
and colllmns. to yield the result
22 29
dard computer solution algorithms such as Gauss elimination. For such cases each
columnj of the matrix inverse can be determined by using a unit vector (with a I in
[ZI = 82 84 the jth row and Oelsewhere) as the forcing-function vector (that is, the right-hand-
( 28 81 side constants). Additional details can be found in App. E.
N.)te how this simple method makes il clear why it is impossible lo multiply two Once the inverse is obtained, a formal way to obtain a solution is to multiply
matrices if Ihe number of columns of Ihe first matrix does not equal the number of rows each side of Eq. 6.23 by the inverse of LAj,
in the second matrix. Also note how it demonstrales that the order of multiplication [A][Arl{e} = [Arl{W} (6.29)
matlers (that is. malrix multiplication is not commutative).
Because [AJ-1LA]equals the identity matrix, the equation becomes

It should now be clear that matrices provide a succinct notation for represent- {C} = [Arl{W} (6.30)

ing simultaneous linear algebraic equations. For example Eqs. 6.15 to 6.17 can be Therefore if we multiply the inverse of the coefficient matrix [Arl by the matrix
expressed concisely as (6.23) of constants {W}, we obtain the solution for the unknowns {C}. This is another ex-
[AJ{C} = {W} ample of how the inverse plays a role in matrix algebra similar to division. That is,
where the matrix (A] contains the coefficients c.,just as e = (llo)W represents the steady-state solution for a single CSTR, Eq. 6.30
\ represents the solution for a system of CSTRs.
011 012 013 It should be noted that the matrix inverse is not a very efficient means to solve
(6.24)
l.'
(A] = 021 022 02) a system of equations. Thus other approaches, such as the etimination and iterative
[ 031 032 033] methods described in App. E, are used in numerical algorithms. However, as de-
~ scribed next, the matrix inverse has great value in the engineering analysis of such
the vector {C} contains the unknowns Jsystems.
(6.25)

tC} . U:} 6.3 STEADY-STATE SYSTEM RESPONSE MATRIX

and thc vector {W} contains the right-hand-side constants or forcing functions, Now that we have learned a titile matrix algebra, we can explore its imptications
for steady-state solutions of coupled reactors with first-order kinetics. As described
(6.26) in the previous section, the matrix inverse can be used to obtain the solulion for the
1
{W) · {S:} sleady-state case. In addition the terms in Eq. 6.30 have a definite physical interpre-
tation. For example the elements of {e} are the response of the syslem as reftected
At this point you should apply the rule for matrix multiplication to convince yourself by the concentrations of Ihe reactors. The right-hand-side vector {W} contains the
that &Is. 6.23 and 6.15 to 6.17 are equivalent. values of the system's stimuli or forcing functions-the loadings. Finally the ma-
trix inverse LA] 1contains Ihe paramelers that express how the parts of the system
Matrix inversion. Although multiplication is possible, matrix division is not a are coupled and purge themselves of the pollutant. Consequently Eq. 6.30 mighl be
defineu operation. However, if a matrix (A] is square, there is usually another matrix reexpressed as
[A]-I called the inverse of LA]for which {Response} = [interaclionsJ{stimuli} (6.31)
(6.27)
[AllAl-1 = [Arl[A] = [1]
 108 PART. Completely Mixed Systems LU-ru~E" Feedback Syslems nI' Rea.:lnls 1(1)

In addition nolice Ihal ElI. 6.30 is Ihc multidimcl1siol1al mal1ifcslaliol1of ElI. l.!!. K17 x 10" O 0.5 X 10" CI 2 X lO"
This can be seen by applying Ihe definilion of malrix multipliealion lo Eq. 6.30. This -1.5 X 10" 7.21 X ID" O c, = 4 x 10~
[, O -1.5 x 10"
gives 11.5 X ID/>J{ C3} { I x lO')}
- (-I) (-I)
C. - a 11 W I + a 12 W2 + a (-nl\' (6.32) The matrix lIl~erse can be determined as
l' I
(-I) W ( ¡, W 1.23 X 10-7 1.11x IO-~ 5.33 x 10~
c2 = a21(-I) W I + a22 2 + U23 3 (6.33)
2.55 X lO-K1.39x 10 7 I.IIXIO~
(-I) (-I) (-I)
W3 [ 3.33 x IO-~ 1.81X 10K 8.71 x 10 K]
C3 = a31 Wt + a32 W2 + U33 (6.34)
whichcan be multiplied by {W} to give
where aj-:-l)denolesIhe elemenlin Ihe ilh row and Ihejlh columnof Ihe malrix 255
inverse.thus wefindIhalthe invertedmalrixitself,asidefromprovidinga solution, CI
C2 = 608
~ has exlremely useful properties. Thal is, eaeh of its elements represenls Ihe response { C]} { 166}
7 of a single par! of Ihe syslem lo a unil slimulus of any other part of Ihe system.
(b) The effect of a load lo segment two on the eoncenlralion of segmenl three can be
EXAMPLE 6.1. STEADY-STATESOLUTION FOR LAKES WITH FEED- based on a~21) = 1.81 x lO-K ¡.tg L -l/mg yr'l. The effecI of Ihe 4 x I O~ mg yr-I load
BACK. kecall Ihat in Example 5.1 we compuled Ihe steady-stale distribulion of a to segment Iwo on Ihe coneentralion of segmenl three can be calculaled as
pollutanl in Ihree lakes conneeled in series. The same syslem is shown below, wilh Ihe
exceplion thal we have recycled a fraction of Ihe ftow (a) from Ihe Ihird lake back to Ihe cddue lo loadings to reactor 2) = 1.81 x IO-K(4 x 10~) = 72.5 ¡.tg L I
firsl lake. (c) For the case where a = O, Ihe Ihree silllultaneous equalions can be expressed in
matrix form as

7.67 X 10" O O CI 2 x 10~

-1 x 10" 6.71 x 10" O C2 = 4XIO~
[ O -1 x 10" 11.0 X 10"] { C3} { 1 x 10~}
The matrix inverse can be determined as
I I

1.30 X 10-7 O O
I 2 3 1.94 X lO-K 1.49 x lO 7 O
Volume. 106m3 2 4 3 [ 1.77 X 10-9 1.35 X lO-K 9.09 x 10 K]
Mean depth, m 3 7 3
0.667 0.571 1.000 Notice how the superdiagonallerms go lo zero when feedback is omiued.
Surface area, I()6m2
Loading, kg yrl 2000 4000 1000 FIGURE E6.1
Before proeeeding lo lime-variable solulions, we should commenl a hit more
(a) If Q = 1 X 10" m] yr-I. a = 0.5, and Ihe pollulanl seules al a rale of 10 m yr-I, on Ihe slrueture 01'Ihe malrix inverse. As depicled in Fig. 6.2, regions 01'Ihe malrix
calculate Ihe concentrallUII 11Icach of the reaclors.
(b) Use Ihe matrix invcr." lo delermine how much of Ihe concenlration in the Ihird
reaclor is due lo Ihe loading lo the second reactor.
Superdlagonal
(c) Delermine the malrix inverse for Ihe case where a = O. (effecl 01 downslream Ioadíngs
on upslream segments)
Solution: (a) The sleady-stale mass balances for Ihe Ihree reactors can be writ-
len as
O = W. - (Q + aQ)c1 - vAlc) + aQc3 Diagonal
(effect 01 <¡írec! loading)
O = W2+ (Q + aQ)c1 - (Q + aQ)c2 - vA2c2
O = W3 + (Q + aQ)c2 - (Q + aQ)c] - vA]c] Subdlagonal
(effecl01upstreamloadings
Subslitutingthe parametervalues,the Ihreesimultaneous equalions can be expressed in on downslreamsegmenls) FIGURE 6.2
matrix form as The elements01Ihe malrix Inv,!rse[A) 1
have a specilic physical inlerprelation.
 110 PARTI Completely Mixed Systems LEcruRE 6 Feedback Systems of Reactors 111

in verse lAr I have a speeifie physical interpretation. Diagonal terms speeify the re- 6.4 TIME-VARIABLE RESPONSE FOR TWO REACTORS
sponse of the segments to direet loadings. The superdiagonal terms reReet the effeet
of downstream segments on upstream segments. The subdiagonal terms reReet the Equations 6.1 and 6.2 (wi!h no loads) ean be written as
effeet of upstream segments on downstream segments. Thus, as in part (c) of the pre- dCI
vious example, a feedforward series system would have zero superdiagonal terms. (~ Jt = -a"cl + al2c2 (6.35)

~
dt = a21CI - a22C2
, dC2 (6.36)
BOX 6.2. Input-Output Modeling and the Delaware Estuary Study where
The sleaúy-state system response matrix úescribeú in this lecture is an example of
input-o/lt/1lI1 modeUng. This approach was originally úevelopeú by W. W. Leontiff, al! QI2 + kl (6.37)
the winner of the 1973 Nobel Prize for Economics. Leontiff derived linear systems of VI
equations \'ery similar to the ones in this lecture. However, rather than linking pollutant
Q21
loaúings with concentrations. he deviseú linear models that linked economic inputs al2 (6.38)
J J
VI
I (e.g.. prodllction of goods by sectors 01'lhe economy) with outPlltS (lhe consumption of
1 the gnods by olher sectors). QI2
lhe Delaware Estuary was the sile 01' the first application of this approach in the a21 (6.39)
V2
water-qualily domain. Located near Philadelphia, the Delaware Esluary is a critical wa-
ter resource in lhe heavily populateú Eastern United States. In the I960s it was the site - Q23 + QI2
+k (6.40)
01' lhe Delaware Esluary Comprehensive Sludy (FWPCA 1966), one of the first appli- a22 - V2 2
calions 01' computer-orienled water-l\lIalilY modeling and systems analysis (Fig. 1.6).
Despile having occurreú over 30 yr ago. it remains one of lhe most comprehensive and If CI = CIOand C2 = C20,then thegeneralsolutioneanbe developedas
innovative studies of its type (see Thomann 1972 for a nice sumlllary).
Al1Iong the sludy's many contributions. the most relevant to waler-quality model-
\ CI = cIJe-AI/ + clse-A.,/ (6.41)
ing was lhe developmenl of lhe conlrol-volume approach described in this lecture. This ¿ C2 = c2Je-AI/ + c2,e-A,/
frmncwork (Thomann 1963) represented lhe doctoral research ofthe slUúy's Technical
(6.42)
Director. Bob Tbomann. now a professor al Manhattan College. Along with the model, where the A'S are eigenvalues that are defined as
Dr. Tholllann developed his own expression of lhe input-output approach in the form
or Ihe sleady-state system response matrix. l AJ _ (al! + a22):t J(al! + a22)2 - 4(al!a22 - a12a21)
In Ihe present leclllres 1 have emphasi7.ed bow the matrix proviúes a concise way I ~ - 2 (6.43)
to link loadings with response. It should be noted lhat Dr. Tbomann and his leam (in-
and the eoefficients are
clocling Matt Sobal and Dave Marks. now professors at SUNY Stony Brook and at
M.I.T.. respectively) extenúed it well beyond lhis application. In parlicular they lIsed
(AJ - a22)CIO - a12C20
the computer lO develop treatmenl strategies Ihal produced acceptable water quality CIJ (6.44)
at a minimum cosl (Thomann and SobaI1964). Specifically Ihey employeú a systems AJ - As
analysis approach known as linear programming. designed to optimize an objective
a 12C20- (As - a22)cIO
¡ function subject to constraints. For the Delaware application, the objective function (6.45)
represented treatment cost as a function of treatment leve!. Among other factors the CL, = AJ - As
conslraints ref1ected lhat the concentrations in the estuary should be at or below stan-
dards. The system response matrix was lIsed to connect the loadings with the resulting -a2ICIO + (AJ - a")C20
C2J (6.46)
concentratiuns in lhe estuary. Linear programming Ihen provideú a means to optimize AJ - As
lhe oojective function (that is. minimize cos!) while salisfying Ihe constraints (that is,
making sure the resulting concentralions met standards). = -(As - all)C20 + a21cIO
(647)
As already menlioned in our hislorical overview (Lec. 1). such cost-effective ap- C2.f AJ - As .
proaches ha ve never fuund wiúespread use. Today, as economic pressures build, a
reviv:d may be at hand. The cause-effect mudels and systems analysis approaches de-
As we saw for the steady-state solution, the time-variable ease is much more
vcl0l'eú 30 yr ago un the Delaware provide an elegant example uf how it can be done. eomplieated than for a single eompletely mixed lake or for reaetors in series. Aside
from the eomplexity of the eoefficients, the major differenee is that the reeovery of
í eaeh segment now depends on t':'.~_expon!ntial~s.
 ...

l.ECTURE6 Feedback Syslems of Reaclors 113

1\2 PART1 Completely Mixed Systems

1500
The total recovery depends on the relative magnitudes of the eigenvalues. Note
that Al will always be greater than As. Consequently Al and Asare often referred
to as the "fast" and the "slow" eigenvalues, respectively. This nomenclature derives 1000
froro the rate at which they approach zero as time progresses. As in the following e
example. there are cases where the fast eigenvalue is much larger than the slow 500
eigenvalue.
o
EXAMPLE 6.2. LAKESWITH FEEDBACK(TIME-VARIABLE).Thefollowing o 1 2 3 4
two lakes inc1ude feedback of a fraction of the flow (a) from the second lake back to the
firstlake: t FIGUREE6.2-2

Although the foregoing analysis can be extended to many coupled reactors,

Q+aQ ~ most analyses of such systems are implemenled numerically. We will review such
1 techniquesin the next lecture.Despilethis faetIheanalytical,two-reactorcase has
general utility in water-quality analysis. This is because several importanl problem
contexts can be modeled as two eoupled reaclors (Fig. 6.3). For example when we
aQ l~model toxie substanees, a lake and its underlying sedimenls can be eharaclerized
( in this way (Fig.6.3c). In suchcases the sedimentstypicallyrespond much more
2 ~slowly than the water. When loads are reduced. the lake's recovery will be retarded
Volume, 1()6 m} 0.2 10 ~ as sediment toxics seep back into the water. The eigenvalue approach outlined above
4 20 Ican be useful in analyzing such cases.
Mean deplh, m
Surface area, \()6 m2 0.05 0.5
2000 4000
Loading, kg yrl 6.5 REACTIONS WITH FEEDBACK
FIGURE E6.2-1
Aside from coupled reactors, the models developed in this lecture can be applied
lo reversible reactions occurring within a single reactor. For example, as previously
Notice !hat for this case the second lake is much larger than the first.
(a) If Q = I x 1()6 m3 yr-I, a = 0.5, and the pollutant settles at arate of 10 m yr-I,
calculate the steady-state concenlration in each of the reactors.
(b) Using the concentrations calculated in (a) as initial conditions, determine the re-
sponse of each lake if their loadings are temlinated at t = O.

Solution: (a) Using the same approach as in Example 6.1, the steady-stale concentra-

ti'M'" ~... reoc".,," b<"::~<d[

~~;~5J (
(b) The temporal response can be determined by substituling the parameter values into
Hypo
Eqs. 6.41 and 6.42 to yield
CI = 981.24e-lUlII' + 243.25e-0611
C2 = -15.67e-IO.041 + 913.63e-061' (a) (b) (e)
The reslllts are plotted in Fig. E6.2- 2. As depicted, the concentration of the first reactor FIGURE 6.3
initially drops precipitously because of the dominance of its fast eigenvalue. After this Three syslems Ihat can be characlerized as a pair 01 coupled reaclors. (a) The
initial drop its recovery slows down as its response becomes tied to the more sluggish epilimnion (surface) and hypolimnion (bottom) 01 a slralilied lake; (b) a lake with
recovery o" !he second reactor. a major embaymenl; (e) a lake underlain by a sediment layer.

'1J'f!'
.-........
 LECTURE
6 Feedback Systems of Reactors 115
114 PARTI Completely Mixed Systems
\
6
dcveloped in Lec. 2, a simple reversible reaction can be represented as
k"b C<IO
A~B (6.48) C¡,
k"" 4

where k,,/.and k¡,,,= the forward and backward rates of reaction, respectively. We e
now cxaminc how such rcactions can be modeled within both cIosed (batch) and 2 r-
I
,opcn (CSTR) readors. I Ca
CIJO I
I
o ,
) o 4 8 FIGURE 6.4
6.5.1 Closed Systems (Batch Reactors) 195
The progress 01 a reversible reaction
If the reactions are assumed to be lirst-order and to take place in a batch reactor, the
t within a batch reactor.

roIlowing balances can be wriuen:

dc" Thus, because we are at equilibrium, the concentrations of A and B are at fixed
(6.49)
- k"¡,c,, + k¡,,,Cb fractions of the total.
dt
Next let us examine the time-variable solution. Although these equations can be
dCI, (6.50) sol ved using the methods described earlier in this lec tu re, an alterna ti ve approach
dt = k"¡,c,, - klwc¡, involves solving Eq. 6.52 for Cband substituting the result into Eq. 6.49 to give
í
I The steady-state solution of these equations can be obtained by setting the dCa
(6.58)
¡ derivatives to zero and solving either equation for dt = -kabCa + kba(C - ca)
\ c¡, = ka¡' = K (6.51) or coIlecting terms,
c" k¡,,, dc"
(6.59)
\
I where K = an equilibrium constan\. Thus we arrive at the familiar result that, at dt + (ka¡, + kba)Ca = kbac
equilibrillln, the ratio of the products to the reactants of a chemical reaction wiIl be Thus the equation is a first-order ODE with a constant forcing function (recaIl Seco
constan\.
4.2). If C = c"o at 1 = O,it can be solved for
To determine the magnitudes of c" and C¡"we can define
(6.52)
C = c" + c¡, Ca = caOe-(k,ék",,)1+ c,,(I - e-(kab+k",,)I) (6.60)
where C = total mass of substances A and B. Now because the system is cIosed, this where ca = the ultimate steady-state concentration of Ca,defined by Eq. 6.54.
quantity is a constan\. Thus Eq. 6.51 can be solved for c¡,and the result substiluled The concentration of Cbcan be deterrnined by substituting Eq. 6.60 into 6.52.
into Eq. 6.52 to give After some manipulation the result is
(6.53)
c = Ca + KCa Cb = CbOe-(kab+k",,)1
+ cb(l - e-(kab+k",,)I) (6.61)
which can be solved for
(6.54) where ChO= the initial condition for B and Cb = the ultimate steady-state concen-
Ca = Fac tration defined by Eq. 6.56.
where the overbar is included to make it cIear that this is an equilibrium solution, As depicted in Fig. 6.4, the solution asymptoticaIly approaches the steady-state
and Fa = the fraction of the total mass that is represented by species A. The fraction condition. The eigenvalue kab + kb" can be used to estimate that the reaction will
is related to the equilibrium coefficient by have a 195of 3/(kab + kb,,).
I l The foregoing analysis has significance beyond merely obtaining solutions. In
I F,,=- (6.55)
I l +K particular the eigenvalue provides insight into how the solutions would be obtained
mathematicaIly. If we are interested in time scales less than 195,the differential equa-
I
This result can then be substituted back into Eq. 6.52, which can be solved for tions would be solved directly to yield concentrations as a function of time. This is
I C¡, = F¡,c (6.56) sometimes caIled a kinelic approach, because we are interested in the time-variable
I
I or dynamic variations of the constituents. However, if we are not concerned with
where F¡, = the fraction of the total mass that is represented by species B, shorter time scales, the steady-state solutions (Eqs. 6.55 and 6.56) would be per-
K (6.57) fectly adequate. This is sometimes caIled an equilibrium approach, because we are
F¡,= I - F" = I + K interested in the equilibrium states of the reactants rather than in how they got there.
116 I'AIUI Complelely Mixed Syslems I.ECTlJRE (¡ Feedhack Syslems 01" Rea<:lors 117

Outflow
PROBLEMS
Inflow
6.1. Three malrices are detined as

k kb
A~B [YJ = [~ ~] [ZI = [~ ~]
kb.
[X] = [~ I~]
FIGURE 6.5
A mass balance lar a caupled reactian Perform all possible multiplicalions that can be computed between pairs of these ma-
trices.
taking place in a eSTA.

6.2. In 1970 Ihe total phosphorus loadings to the Greal Lakes were (Chapra and Sonzogni
1979)
6.5.2 Open Systems (CSTRs)
Units Superior Michigan Huron Erie Ontarío
Now let's explore how the foregoing analysis pertains to an 0(><:11
system. As in Fig. 4000
W lonnes yr-I 6950 4575 18.150 6650
6.5 suppose that the reversible reaetion occurs in a eSTRo Both A and B are carried
in and out of the reactor by flow. In addition assume that B is also subject to an
additional reaetion that removes it from the system. Mass balances can be written as (a) Iftolal phosphorus settles at arate 01"approximately 16 m yr-I, calculale Ihe steady-
state concenlration 1'01'each lake using the malrix inverse approach.
dCa Q Q (b) Use the malrix inverse lo determine how much 01'Lake Ontario's concelllration is
(6.62)
di = VCa.in- VCa - k"bCa + kbaCb
..&1
due to the loading to Lakc Huron.
(e) Use the matrix inverse lo determine how much Lakc Ontario's cOllcenlration will

l dCb
di
Q Q
= VCb,in - VCb + k"bC" - kb"Cb - kbCb
(6.63 ) change if Lake Erie's loading is reduced
halved.
by 25% and Lake Huron's loading is

These equations can be solved using the approaches described in earlier parts 6.3. Use Laplace transforms to solve Eqs. 6.35 and 6.36 1'01'Eqs. 6.41 and 6.42.
of this lecture. However, we will look at a spedal casco This relates to situations
6.4. As depicted in Fig. P6.4, a lake and its bottom segments can be modeled as two CSTRs
in which the coupling reaction rates kab alld kba are much faster than the input- wilh feedback.
output 01'purging rates Q/V and kb. For example if kab and k"" were on the order of
1 hr-I and the purging rates were I yr-I, the coupling reactions would always be Inflow Load Outflow
at a local equilibrium on an annual 01'even a daily time scale. In this case Di Toro
(1976) has shown that Eqs. 6.62 alld 6.63 l:an be added logelher to yield
Lake
dc Q Q (6.64)
- = -Cin - -c - kbCb
di V V
(
where c = total concentration = Ca + Cb and Cin = total inflow concentration =
Ca,in + Cb,in'
Notice that the terms representing the fasl reactions were canceled by the addi-
tion of the equations. This is justified by the fact that if the reactions are at equilib- FIGURE P6.4
rium, the terms kabcaand kbaCbare equal and would cancel out.
Now at this point the mass balance, although simpler, cannot be solved because The following parameters are given:
it is a single equation with two unknowns, C and Cb.However, the second can be Inftow = outftow = 20 x 106 m3 yr-I Lakc area = sediment area = 2.5 x 106 m2
eliminated by substituting Eq. 6.56 to give Lake volume = 150 x 106 m3 Sediment volumc = 100 X 1041111

de Q Q Settling velocily = 10 m yr-I Resuspension velocity = 1 mm yr-I

di = VCin - VC - kbFbC (6.65) Burial velocity = 2 mm yr-I

Thus the original system of differential equations has been replaced by a sin-
gle mass balance Ibat can be solved for the total mass in the system as a function
of time. Then at every time step, Eqs. 6.54 and 6.56 can be used to calculate the
concentrations of each of the two interacting species.
(a) If the lake receives a constant inftow concentration 01'100 I'g L -1 and the pollu-
-l. tant does not react, but does settle, compute the steady-state concentration in the
sediments and the water. Use the matrix inverse to determine the concentrations.
(b) If a sediment concentration 01' 100,000 I'g L -1 must be maintained, use the matrix
I inverse to come up with the necessary inftow concentration.
118 PART1 Complelely Mixed Syslems
LEClURE6 Feedback Systems of Reaclors 119

", (c) Suppose Ihal a spill of 20 kg of a conlaminanl occurs in Ihe lake. Calculale !he
concenlralion in Ihe lake and Ihe sedimenls as a funclion óf lime in Ihe following
years.

6.5. Suppose Ihal a spill of 5 kg of a peslicide lakes place in Ihe lake described in Probo
1
a:
6.4. Assume Ihal 50% of Ihe peslicide associales wilh so lid maUer and seules bUI
docs nol volalilize (Ihal is, il is nol losl lo Ihe almosphere). The olher 50% is solu-
j
CJ
ble Hlld suhjecl to volatilizalion (hulnol seulingl Ihal can be characterized hy Ihe firSI- ~.:-
order flux ~~
~..o
J = v,.Fdc C>>)~('

where "v = a volalilization mass-Iransfercoefficienl ofO.OI m d-I and Fd = Ihe frac-

tioll of peslicide in dissolvcd form (in Ihis case 0.5). Calculale Ihe response of Ihe 1;
>
s:vslem lo Ihe spill. When does Ihe peak concenlralion occur in Ihe sedimenls? Nole a: NIVllo
o
...
Ihal Ihe lake is peslicide-frce prior lo Ihe spill.
S
6.6. SUI'I'0se Ihal Ihe rollowing firsl-c>rder reaclions lake place in a balch reaclor: 8
kab
A _ B

\ ei-
k""
k,. klH;
k., k'b
Lekl
H8V.SU
FIGURE P6.6
FIGURE P6.8
Tha Colorado Rivar systam.
Thc following paramclers and inilial condilions apply:

k"" = 0.1 d-I k,>o= 0.2d-1

k,., = 0.4 d I k,." = 0.5 d 1 TABLE P6.8
k,,,. = 0.3 s I kw = 0.6 S-I
Characteristics or impoundments on the Colorado
C,~I= IOmgL-1 cw = 20mgL-1 River system
('dI = 70 mg L-1
Outnow Volume
(a) Calculale Ihe sleady-slale concenlralions of Ihe Ihree reaclanls. Impoundment Chlorlde loadlng
(10' 10.1yr-l) (10' mJ) (10' g yr '1)
(bl Calculale Ihe concenlralions or Ihe Ihree rcaclanls ror 10 d. Taylor Park 226 602 900
Blue Mesa 853
6.7. Repeal Probo 6.6 for a CSTR wilh a residence lime or I d. Nole Ihal Ihe inflow concen- 4.519 4.000
Morrow Point 914 851
lralions or Ihe Ihree reactanls are Ihe same as Ihe inilial condilions in Probo 6.6. 400
Fonlenelle 1.429 2./09 5.700
6.8. Figure P6.8 shows a simplified schemalic or Ihe Colorado River syslem. and charac- Aaming Gorge /.518 19,581 20,000
lerislics or Ihe impoundmenls are summarized in Table P6.8. Navajo 1.250 9.791
(a) Solve ror Ihe chloride concenlration in each of Ihe reservoirs.
~
'.
Powell 13.422 150.625
5.000
714.500
Mead 12.252
(1.» Determine Ihe sleady-slale syslem response malrix. 180.750 300.000
Mohave /2.377
(e) Use Ihe result or par! (b) lo evaluale Ihe new concenlralion in Lake Havasu if Ihe 12.050 102.000
Havasu 11.797 4,142 30.000
loading lo Lake Powell is halved.
(d) Suppose Ihal global warming reduces all oUlflows by 20%. Recalculale Ihe con-
cenlralions ror Ihis scenario.
(l') Calculale Ihe errecl on Lake Powell ir Ilonne (1000 kg) of a conservalive conlam-
inanl is spillcd inlo Flaming Gorge.
(n Repeal parl (1') hui for a conlalninant wilh a half-lire or 2 yr.
 LECTURE 7
LECTUKE
7 COl11puter Methous: Well-Mixeu
-¡
Reaclors 121 '
_'''F~r~ ,,,,, ,,,..,,,. ~ " '_;.h~ ...........
Al! the above problems can be addresscd by using the computer and numerical
methods. For these reasons you mllst now begin to learn how compllters are lIsed
to solve differential equations.
Computer Methods:
Well-Mixed Reactors /

V 7.1 EULER'S METHOD

Euler's method is the simplest nllmerical method for solving ordinary differential
equations.
mixed lakeOne way to introduce Euler's method is to derive it to solve the completely
model,
de W(t)
dt =V - Al' (7.1)
where

A=g+k+ (7.2)
V H r
The fundamental approach for solving a mathematical problem with the com-
puter is to reformulate the problem so that it can be solved by arithmetic operations.
J..ECftJRE OVEJlVlEW¡ . showhow computerscan be used to obtainsolutionsfor OuronestumblingblockforsolvingEq.7.1in thiswayis Ihederivativetermde/dt.
i~~h'jduat ~¡tgtofS and sY!ltpmsof reactors. Three numericaJ approacpes {Ire I However, as we have already shown in Lec. 2, difference approximations can be
~escfibe(1:the'}Mer, fJeun, 1U\4fourth-order ~unge-KlItta meU1o~s. 1 used to express derivatives in arithmetic terms. For example using a forward differ-
.. ence, we can approximate the first derivative 01'c with respect to t by

To this point we have used ealculus to develop solutions for a number of idealized
dc¡ = ó.c C¡+I - c¡
dt - ó.t t¡+t - t¡
(7.3)
cases. Although these solutions are extremely useful for obtaining a fundamental
understanding of pollutant fate in natural waters, they are somewhat limited in their where c¡ and c¡+ 1 = concentrations at a present and a fUlure lime t¡ and t¡+1, rcspec-
ability to characterize real problems. There are four reasons why the analytieal ap- tively. SlIbstituling Eq. 7.3 into Eq. 7.1 yields
proaches are limited:
(, . Nonidealized loading jUllctions. To attain closed-form solutions, idealized load-
ing functions must be used. For example the loading must be adequately repre-
('¡,'I - C¡
t¡+1 - t¡
W(t)_
V Ac¡ (7.4)
which can be sol ved for
sented by functions such as the impulse, step, linear, exponential, or sinusoidal
forms described in Lec. 4. Although realloadings may sometimes be rcpresented
'..
in this way, they are more often arbitrary, with no apparent undcrlying pattern (7.5)
(rccall Fig. 4.1f). ('¡+I = c¡ + [W~t) - AC¡]([¡+I- ti)
\, . Variable parameters. To this point we have assumed that all the model parameters
(that is, Q, V, k, v, etc.) are constan!. In faet, they may vary temporally. Notice that the term in square brackets is the differential eqllalion itself (Eq.
( . MlIltiple-segment systems. As you have learned in the past two ehapters, systems
7.1), which provides a means to compute the rate of change or slope 01'c. Thlls
the differential equatiolJ has been transformed into an algebraic equation Ihat can
) of more than two segments require computers for their effieient solution.
.
Nonlinear killetics. To this point we have emphasized first-order kineties. This be used to determine the concentration at t¡+1lIsing the slope and a previous value
means thar we limited our studies to linear algebraic and differential equations. of c. If you are given an ¡nitial value for concentration at some time ti, you can
Although first-order reaetions are important, there are a variety of water-quality easily compute concentration at a later time ti+l. This new value of c at ti+1can in
~
problems that require nonlinear reactions. In most of these cases analytical solu- turn be used to extend the computation to [¡+2, and so on. Thus at any time along
the way,
tions cannot be obtained.

120 ) Newvalue = old value+ (slope)(step)

~ (7.6)
 122 rART1 Complelely Mixed Syslems lEC'TURE7 Computer Methods: Well-Mixed Reactors 123

e .,.I Predlcted The calculation is continued in a similar fashion lo obtain additional values, The results,

,
,
'~
,
I

True
} Error along wilh Ihe analylical solulion, are
_ e (mg L-') e (mgL-')
-.
t(yr) Numerlcal Analytlcal t(yr) Numerlcal Analytlcal
c¡ O 15.00 15.00 6 133.21 127.20
I 59.75 52.75 7 136.59 131.82
dc¡ 2 88.84 79.37 8 138.78 135.08
lit 3 107.74 98.12 9 140.21 137.38
4 120.03 111.33 10 141.14 139.00
5 128.02 120.64 00 142.86 142.86
h
FIGURE 7.1
The numerical solution is plolted in Fig. 7.2 along with Ihe analylical result. It can
Graphical depiction 01 Euler's
ti ti+1 t method, be seen thal Ihe numerical melhod accuralely caplures Ihe major fealures of Ihe exacl
SolUlioll. However. because we have used slraighl-lille segmenls lo approximale a con-
linuously curving funclion, Ihere is some discrepancy belween Ihe two results. One way
This approaeh can be represented generally as lo minimize such discrepancies is to use a smaller step size. For example applying Eq.
('i+1 = C¡ + ¡(ti, ('i)II (7.7) S 7.5 al 0.5-yr inlervals results in a smaller error, as the slraighl-line segmenls track closer
~ lO Ihe lrue Solulion. Using hand calculalions Ihe effor! associaled wilh using smaller and
where ¡(ti, ('i) = dCildt = the value of the differential equation evaluated at ti and smaller slep sizes would make such numerical solulions impraclical. However, with Ihe
Ci and h = step size (= ti+1 - ti). This formula is referred to as Enler's (or the aid of Ihe compuler large numbers of compulalions can be performed easily. Thus you
I fJoillt-slope) metllOd (Fig. 7.1). can accuralely model the concenlralion wilhoul having lo solve Ihe differenlial equalion
exaclly.
EXAMPLE 7.1. EULER'S METHOD. A well-mixed lake has Ihe following charac-
lerislics:
v = 106m3 1 The Euler method is a first-order approach. Among other things this means that
Q = I()~ m3 yr-I
:=5m k = 0.2 yr 1 it would yield perfeet results if the underlying solution of the differential equation
were a linear polynomial or straight line. The accuracy of Euler's method can be
\' = 0.25 m yr-I
improved by using a smaller time step. Other methods are also available that at-
Al t = O il receives a slep loading of 50 x 106 g yr- ~nd has an inilial concenlralion tain better aceuraey by improving the slope estimate used for the extrapolation (see
of 15 mg L l. Use ElIler's melhod lo simlllale Ihe concenlralion from t = O lo 20 yr us- Chapra and Canale 1988 for a review). The following sections describe some of these
ing a lime slep of I yr. Coml'iII'c Ihe resulls with the analytical sollllion methods.
W
C = C(leAl + -(
AV
1 - e Al)

Solutlon: First, the eigenvalue can be compuled as 'r :~ 150

_. ".._. _. oO.
Euler's
Sleady-state
""_.._. _.._. .

/
~-
.
'fl'''J/y \ = ~ 02 0.25 = 035 -1 method

~ 1\ 106+ . + 5 . yr
(t¡ = O) Ihe cOllcentration is 15 mg
100
Al Ihe start of the compulation L -1 and the loading e
is 50 X lit. g yr-I. Using Ihis informalion ami the parameter values. Eq. 7.5 can be
employed to compule concelltnltion at tj+ 1: 50
50 x lit' _

c(1) = 15 + [ -~ ]
- 0.35(15) 1.0 = 59.75mg L 1
o
FIGURE 7.2
Comparison 01
For the Ilcxl inlerval (from t = I lo 2 yr) the computalion is repeated. with the result o 2 4 6 8 10 analytical and Euler-
melhod solutions lor
the well-mixed lake
c(2) = 59.75+[501~J(~06 -O.35(59.75J]UJ = 8R.8375mgL-1 t with a step loading.
 I.HTUKE 7 Computer MClhods: WclI-Mixcd Rcactors 125
124 PARTI Completely Mixed Systems

V7.2 HEUN'S METHOD ¡{C;tl

--;¡¡- = f (/ ¡ iI ,('~)t I )
(7.10)
A fundamental source of error in Euler's method is that the derivative at the begin-
ning of the interval is assumed to apply across the entire interval. One method to Thus the two slopes (Eqs. 7.8 and 7.10) ean be combined to obtain an average slope
for the interval
improve the estimate of the slope involves the determination of derivatives across
~ the
are interval--one
then averaged at
tothe initial
an point and another
estimateatofthe
theend point.
forThe two derivatives
obtain improved slope
approach, called Heun's melhod, is depicted graphically in Fig 7.3.
entire interval. This ;
(dc df = f(I;.c;)+ f(I¡+I,C:)+I)
2 (7.11)

Recall that in Euler's method the slope at the beginning of an interval This average slope is then used to extrapolate linearly from c; to c¡+ t,
dc¡
lit = f(I;. c;) (7.8) C¡+I = c¡ +
f(I¡, c¡) + f(l;+I.
2
c?+ I )
,
1 (7.12)
is used to extrapolate linearly to ci+1: The Heun method is called a prediclor-correclor melhod. As derived above it
can be expressed coneisely as
C?+I = Ci + f(ti. c¡)h (7.9)
For the standard Euler method we would stop at this point. However. in Heun's
Predic/or: c?d = c¡ + f(/¡,Ci)h (7.13)

method the c?+I calculated in Eq. 7.9 is not the final answer but an intermediate f(li, Ci) + f(li+l. c?+t)
Corrector: Ci+1 = C¡ + _ h (7.14)
prediction. This is why we have distinguished it with a superscript O.Equation 7.9 :,
provides an estimate of ci+1 that allows [he calculation of an estimated slope at the D
Note that because Eq. 7.14 has Ci+) on both sides of the equal sign, it can be sol ved
end of the interval:
iteratively to refine the result. That is, an old estimate ean be used repeatedly to pro-
vide an improved estimate of c¡+l. It should be understood that this iterative proeess
SIOpe2 = f(t'+1 ,..+,) will not necessarily converge on the true answer but will converge on an estimate
with a finite truneation error. However, this truncation error will be smaller than for
e cruder approaches like Euler's method. .
As with other iterative melhods, a termination criterion for convergenee of the
corrector is provided by

/ cj - cj-I
% error = i+1 i ¡+I (100%) (7.15)
I c¡+1 I

h
where ct:.' and c{+I are the results from the prior and the present iteralion of Ihe
corrector, respectively.
t/ t/+I t
(a) EXAMPLE 7.2. THE HEUN METHOD. Use the Heun method lo solve the same
.. problem as in Example 7.1. For Ihe present application do nol ilerale the corrector.

e Solution: Allhe start 01' the computation (1; = O) Ihe concentralion is 15 mg l. I and
the loading is 50 x 10b g yr-I . Using this information and the parameler valucs. bl. 7.1
can be used lo compute a slope eSlimate at 1;:

feO, 15) = 50 - 0.35(15) = 44.75

which can Ihen be used lo cakulalc the concentration at Ihe end 01' Ihe interval

c(l) = 15+(44.75)1.0 = 59.75mgL-1

FIGURE 7.3 This value can in 10m be employed to estimate a slope allhe end 01' the interval
1/ li+1 t Graphical depiction 01 Heun's method.
(b) (a) Predictor; (b) corrector. f(l,59.75) = 50 - 0.35(59.75) = 29.0875
 126 PART1 Completely Mixed Systems ~ lECTURE7 Computer Methods: Well-Mixed Reactors 127

c
Then the two slopes can be input into the corrector to calculate the final result k2 ¡
c(l) = 15 + !(44.75 + 29.0875)1 = 51.91875 mg L-1
, ..~
~
which is much closer to the true value than was obtained with Euler's method. The .
1
:..:...~,
I
calculation is continued in a similar fashion to obtain additional values. The results,
along with the analytical solutioll, are
kl
!

0, .
," di
~::.:
I
c(mgL") ... .,",
. '. ". .'...
c(mg L ') .. -.-'.
Numerical Analytical 1 (yr) Numerical Analytical
I (yr)
O 15.nO 15.00 6 126.30 127.20 ~ ~4
--.
1 51.92 52.75 7 131.08 131.82 FIGURE 7.4
2 78.18 79.37 8 134.48 135.08
t¡ t¡+112 ti+1 t Graphical depiclion 01 the lourth-
3 96.85 98.12 9 136.90 137.38 order RK method.
4 110.14 111.33 10 138.62 139.00
5 119.58 120.64 :x; 142.86 142.86
'\
i k4 = f(t¡ + h. e¡ + hk3) (7.21)
where the functions are merely the original differential equation evaluated at specific
values of t and e. That is,
I¡ The Ileun method is a second-order approach. Among other things this means
that il would yield perfect resulls if the underlying solution of the differential equa- de
tion \Velea quadratic polynomial. Thus it is superior to the Euler approach. However, f(t, e) = dt (t, e) (7.22)
the faet that two derivative evaluations must be made for each time step means that
a complItational price is paid for the gain in accuracy. ~ The fourth-order RK method is similar to the Heun approach in that multiple
estimates of the slope are developed to come up with an improved average slope for
the interval. As depicted in Figure 7.4, each of the k's represents a slope. Equation
v0.3 RlINGE-KUTTA METHODS 7.17 then represents a weighted average of these to arrive at the improved slope.
)
The Runge-Kutta (or RK) methods are a family of numerical methods that are used i EXAMPLE 7.3. THE FOURTH-ORDER RK METHOD. Use Ihe classical fourth-
extensively in waler-quality modeling. The RK methods all have the general form order RK melhod lOsolve Ihe same problem as in Example 7.1.
(7.16) -,
..
c¡+I=c¡+cjJh
Solution: At Ihe starl of Ihe compulation (t¡ = O) Ihe concenlration is 15 mg L -1 and
Ihe loading is 50 x 106 g yr-I. Using Ihis informalion and Ihe parameler values. Eqs.
=
~
whele l' a slope estimale (fonnally called an incremelltfullction). Comparison 7.17 lo 7.21 can be used compute concentralion al li+l:
of Eq. 7.16 wilh Eq. 7.7 indicates that Euler's method is actually a first-order RK Ii iAJ k"'
method with cjJ= f(t¡. c¡). In addition Ihe Heun method (without corrector iteration) ~ kl = 1(Q,j2) = 50 - 0.35 15) = 44.750 V - 1"--
is a second-order RK algorithm (Chapra and Canale 1988).
The mosl comll1onlyused RK ll1ethodis the classical fourth-order method that
~ k2 = 1[0+ 1(1),
1t
15+ !(1)44.750]
/
= 1(0.5.37.375) = 50-0.35(37.375) = 36.919
l I J

has Ihe rorll1 kJ = 1[0 + !(I). 15+ !(1)36.919]= 1(0.5.33.459) = 50 - 0.35(33.459)= 38.289
c¡+ 1 = c¡ + I~(kl + 2k2 + 2k3 + k4)]h (7.17) k4 = f[0 + 1.15 + 1(38.289)] = 1(0.5,53.289) = 50 - 0.35(53.289) = 31.349

where -+ (~ ) c(l) = 15 + [~(44.750 + 2(36.919 + 38.289) + 31.349)] (1) = 52.75 mg L-1

kl = f(t¡, c¡) (7.18) For Ihe next interval (t = 1lo 2 yr) Ihe compulalion is repealed, wilh Ihe result
(7.19) c(2) = 52.75 + [W1.537 + 2(26.018 + 26.984) + 22.092)](1) = 79.36mgL-1
k2 = .r(t¡ + ~h.c¡ + ~hkl)
<. The calculation is conlinued in a similar fashion to obtain addilional values. The resulls,
(7.20)
k.1 = .r(t¡ + ~h. ('¡ + ~hk2) along with lhe analytical solUlioll, are
..

..
 »..----
~
r~~
128 I'ARTI Compleh.:l)' Mixed Syslems IH I¡;Rr 7 COlllpUh.:r MClhods: WclI-Mixcd I{caclors 129 i.~

c(mgL
Numerkal
1)
Numerical
c(mg 1. ') Subsliluling Ihe parameler values results in ,"
I (yr) Analytical I (yr) Analytical e/CI
e/I
-2cl
O 15.00 15.00 6 127.19 127.20
1 52.75 52.75 7 131.82 131.82 e/C2
8 135.08 I!
2 79.36 79.37 135.08 dt = 1.667cI - 1.(¡()7c~
3 98.11 98.12 9 137.38 137.38
4 111.32 11\.33 10 138.99 139.00 If c) = 10 ILg L -1 and /"2 = O, Ihen Ihe::ge::nclal ,ululiun C:1I1be deve::(oped as
5 120.63 120.64 '" 142.86 142.86 CI = lOe-21

Thus, for this example, Ihe lIul1Ierical melhod very accuralely follows Ihe analylical so-
C2 = 1.667(
. u_ 10)"
_(e - - 1"
16671
) ~ ,1
f

JUlion. Discrepancies are limiled 10 a unil quanlily in Ihe second decimal place.
Now we can proeeed wilh Eule::r's melhod. Firsl. Ihe differenlial equations are em-
ployed to calculale Ihe slopes all = O, 1
e/('I
7.4 SYSTEMS OF EQUATIONS -(O)
e/I = -2(10) = -20

AlJlhe methods described previously can easily be adapled lo simlllale syslems of

ditTerenlialeqllations of Ihe form (~/(O) = 1.667(10) - 1.667(0) = 16.661
del These values ean Ihen be used lo extrapolale OUIlo 1 = O.I yr.
/I(CI. e2, c,,) (7.23)
dI
('1(0.1) = 10-20(0.1) = 8ILgL I
dC2
!2(cI. C2 c,,) (7.24) ')}
dI /"2(0.1) = 0+ 16.667(0.1) = 1.667 ILg L
"'
dc" For Ihe nexl il1lerv:11 (t = 0.1 lo 0.2 yr) Ihe COlllpul:llion is repealed. Firsl, Ihe slopes an:
(7.25) determined all = O.1 yr:
dt = f,.(el. C2 , e,,)
The solution of such a syslem requires Ihat 11inilial conditions be known at Ihe
e/CI
--(0.1)
dI
= -2(8) = -16 li
.-~
starting point. Then Eqs. 7.23 to 7.25 can be employed to develop slope estimales for
-
d('2
each of the unknowns. These slope estimates are used to predict vallles of concen- ( ,1 (0.1) = 1.667(8)- 1.667(1.667)= 10.556 (
tration at a new time. The process can then be repealed to project 0111another lime
slep. These values ean then be used lo extrapolale oullo I = 0.2 yr,
:='1
EXAMPLE 7.4. SPIU. FOR TWO LAKES IN SERIES. A spill of 5 kg of a soluble
cl(O.2) = 8-16(0.1) = 6.4lLgL I

pesticide lakes place in Ihe first 01'two lakes in series. Nole Ihal bolh lakes are eomplelely ('2(0.2) = 1.667 + 10.556(0.1) = 2.722 ILg L
mixed. The peslicide is subjecl to no losses excepl for flushing. Paramelers for Ihe lakes
are The calculation is eontinued in a similar fashion lo oblain addilional values. Thc
results, along wilh Ihe analytieal ~olulion, are
I.ake I Lake 2
Vnlumc(m') 0.5 x )()" 0.6 x 10" C, (,..1:1. ') C, (,..1:1. ,)
OUlllow (m' yr ') 1 X lO" 1 X lO" I (yrl Numerical Analylical Numerical Analylical
X 0.0 IO.(J() 10.00 O.!X) O.(X)
Predict the eoneenlralion ofboth lakes as a funelion oftime using Euler's melhod. (:on1- /" 0.1 8.00 8.11) 1.67 1.39
pare Ihese results wilh Ihe analytieal solulion. Presenl your resulls graphlcally. e 0.2
0.3
6.40
5.12
6.70
5.49
2.72
3.35
2.31
2.89
Solution: Firsl, we ean develop Ihe mass balances for the syslem: 0.4 4.10 4.49 3.63 3.20
e/CI e//"2 0.5 3.28 3.68 3.71 3.34
-dI = -AII/"I -dI = A21CI- A2,C2
. 0.6 2.62 3.01 3.64 333
where 0.7 2.10 2.47 3.47 3.24
0.8 1.68 2.m 3.24 3.09
QI2 QI2 0.9 1.14
AII A21 A22 = Q2.1 I.h5 2.98 2.89
VI V2 V2 1.0 1.07 1.35 2.71 2.68
-
f
 130 PART1 Completely Mixed Systems
LECTURE7 Computer Melhods: Well-Mixed Reactors 131

The results are plotted in Fig. 7.5, along with the analytical solution. As was the case PROBLEMS
for the single equation (Example 7.1), one way to minimize the discrepancies is to use a
smaller step size.
\ 7.1. Population-growth dynamics is importanl in a variety of engineering planning studies.
One of the simplest models of such growth ¡ncorporales the assumption that the rate of
change of the population p is proportionallo the exisling population at any time t:
The Heun and the fourth-order Runge-Kutta methods can also be used to simu-
late systems of ODEs. However, as illustrated by the following example, care should dp = Gp
dt (P7.1)
be taken lo sequence the calculation of the slopes prior to computing new concen-
trations.
where G = a growth rate (yr-I). This model makes intuitive sense because the greater
EXAMPLE 7.5. SYSTEMS OF OI)ES WITH THE FOURTH-ORDER RK the poplllation, the greater the number of potential parenls.

METHOD. Use the classical fourth-order RK method to solve the same problem At lime / = O an island has a population of 10,000 people. If G = 0.075 yr-I, use
as in Example 7.4. Euler's method to predict the population al/ = 20 yr using a slep size of 0.5 yr. Plot P
versus / on standard and semi-Iog graph paper. Delermine the slope of the line on the
Solution: The important featllre of this computation is that all the k's must be com- semi-Iog plot. Discuss your results.
1';-1 plltcd for Ihe enlire syslem of ODEs before compllting Ihe next set or concentrations.
For example 7.2. Although the model in Probo 7.1 works adequalely when population growth is unlim-
I
ited, il breaks down when factors such as food shorlages, pollulion, and lack of space
kr¡1 = -20 kb,1 = -18 kJ.1 = -18.2 k4.1 = -16.36
inhibit growth. In such cases the growlh rate itself can be thought of as being inversely
kJ¡.2= 16.667 k~2 = 13.611 kJ.2 = 14.032 k4.2 = 11.295 proportional to population. One model of this relationship is
These can Ihen be used to compute the two increment functions G = G'(Pmax- p) (P7.2)
4>1= H-20+2(-18-18.2)-16.36) = -18.127 where G' = a population-dependent growth rate [(people-yr)-I] and Pmax = the max-
4>2 = W6.667 + 2(13.611 + 14.032)+ 11.295] = 13.875 imum sustainable population. Thus when population is small (p « Pmax), the growth
rate will be at a high, constant rate of G' Pmax. For such cases growth is unlimited and
These can be used to extrapolate out in time,
Eq. P7.2 is essentially identical to Eq. P7 .1. However, as population grows (that is. as P
cl(O.I) = 10-18.127(0.1) = 8.19 mg L-1 approaches Pmax), G decreases until at l' = Pmax it is zero. Thus the model predicts that
when Ihe populalion reaches the maximum sustainable level, growth is nonexistent, and
C2(0. \) = 0+ 13.875(0. \) = 1.39 mg L -1 the system is al a steady-state. Substituling Eq. P7.2 inlo Eq. P7.1 yields
~ The calculalion is continued in a similar fashion to obtain additional values. The reslllts
dp '
match the analylical solulion to two decimal places of accuracy. Thus, for this example, G
d/ = (Pmax- p)p
the numerical method very accllrately follows the analytieal solution.
For the same island stlldicd in Probo 7.1 use Euler's method to predict the population
at / = 20 yr using a step size of 0.5 yr. Use values of G' = 10-5 (people-yr)-I and
10 Pmax = 20,000 people. At time / = OIhe island has a population of 10,000 people. Plot
e P versus / and interpret the shape of the curve.
8
7.3. Recall that in Probo 4.8 we studied a lake having the following characteristics:
6 Inllow = outllow = 20 x 106 mJ
Mean depth = 10 m
4 Surface area = 10 X 106 102

2 A canning plant discharges a pollutant to the system that decays at arate of 1.05 yr-I .
In Probo 4.8 you were asked to approximate the seasonalloading from a cannery as a
o sinusoidal input. The following measurements pro vide a better estimate of the loading
pattern over the course of ayear.
o 0.2 0.4 0.6 0.8 1
t Month J F M A M J J A S O N D
FIGURE 7.5 Load (mta) 29 26 1I O O 9 23 43 44 64 53 50
Comparison 01 analytical and Euler-method solutions lor a system 01
well-mixed lakes. Ifthe initial condition is c = O at / = O,

__o
~ ~-
 I
I

132 PART1 Complelely Mixed Syslems I

IHTlJKE 7 Compuler MelhlJus; Wcll-Mixeu Reaclurs 133
i
(a) Use Ihe numerical melhou 01'your choice lo compute Ihe concentnllÍon in Ihe system 7.6. A spill 01' 5 kg 01' a soluble pesliciue takes place in the lirsl nI' Iwn lakes in series. The I
from I = O lo 10 yr. peslieide is subjecI 10 volalilizalion Ihal can be characlerized by f¡rsl-order decay mIes
(b) Afler sufficienl lime Ihe concenlralion will approach a dynamic sleauy-slale. Al Ihis
point, on whal day 01'Ihe year will the in-Iake concentralion be al a maximum value?
kl = 0.002 d-I and kl = O.(X)333 d l. Olher paramelcrs '()r Ihe fakes are
.
I

7.4. A small pond has Ihe following characlerislics: Lake 1 Lake 2

Surface area = 0.2 x 106 m2 t~ Surface area (m2) 0.1 x 10" 0.2 x 10"
Mean deplh (m) 5 3
Mean depth = 5 m OuUlow (mJ yr-I) Ix 106 I x 10"
OUlftow = 1 X 106 m3 d-I

The lemperalure in Ihe pond varies diumally as

Using Ihe fourth-order RK melhod,
Time Midnighl 2:00 4:00 6:00 8:00 10:00 (a) PredicI Ihe coneenlralion in bolh lakes as a funclion 01' lime; present your resulls as
apio\.
Temp. (.C) 21 20 17 16 18 21 (b) Delermine Ihe lime requircd 1'01'Ihe second lakc 10rcaeh ils maximum eoncenlralion.
Time Noon 2;00 4;00 6;00 8;00 10:00
7.7. Asidc from Ihe Heun melhod, Ihere is anolher seeond-order approaeh 1'01'solving ordi-
Temp. (.e¡ 25 27 28 26 23 21 nary differenlial equalions ealled Ihe midpoinl 01'improved polygon me/llOd. It can be
represenled malhemalieally by
Delermine Ihe response 01' Ihe syslem lo a 2-kg spill 01'a pollulanl Ihal decays al a rale "
k = 2 d-I. Calculale the response if the spill lakes place (a) al midnighl 01'(b) al noon. c¡t ~= c¡ + f(l¡, (")2
Plol the responses on Ihe same graph. NOle Ihal Ihe reaclion has a temperalure depen-
dence characterized by fJ = 1.08. C¡+I = c¡ + f(lit ~,c¡+ ~)h

7.5. Lake Washington, a beauliful
general characleristics:
Volume = 2.9 x 109 m3
Mean deplh = 33 m
Outflow = 1.25 X J09 m3 yr-I
lake localed in Seanle, Washington, has Ihe following Thus Ihe firsl equalion uses Euler's melhod lo lIlake a prcdiclion 1'01'C at Ihe midpoinl 01'
!he inlerval. This value is Ihen used lo generale a centered slope eSlimale Ihat is applied
lo predicl Ihe value al Ihe end 01' Ihe inlerval wilh Ihe second equation. Employ Ihis
approach lo solve Ihe ODE deseribed in Example 7.2.

In Ihe 1950s and 1960s il began lo deleriorale because 01' increased loadings 01' Ihe nu-
Irient phosphorus. As a consequenee, in Ihe lale 1960s, sewage inpuIs were greally cur-
lailed. The loading patlem from 1930 Ihrough Ihe lale 1970s can be summarized as in
Ihe following lable:

Year Load (mta) Year I.oad (mta) Year Load (mta)

1930 40 1961 137.4 1972 103.4
1940 40 1962 148.5 1973 42.9
1941 55 1963 156.5 1974 58.5
1949 55 1964 204.2 1975 99.3
1950 84.8 1965 142.8 1976 42.9
1951 81 1966 124.8 1977 60.3
1956 81 1967 54.3 1978 48.6
1957 93.2 1968 59.1 1979 60.5
1958 104.3 1969 48.2 2: 1980 60.5
1959 115.3 1970 59.0 \
1960 126.4 1971 53.8

TOlal phosphorus settles at a rale 01'abouI 12 m yr-I .

(a) Use the Heun melhod lo compute the lake's response from 1930lhrough 1990. Note:
Set Ihe ¡nitial eondition in 1930 al 17.4 ¡J.g L -l.
(b) Compare your results wilh Euler's method.
(e) Compare your results with !he fourth-order RK melhod.
 PART 11

IncompletelyMixed Systems
I i

..

Part II is designed to introduce you to the fundamentals of modeling incompletely

mixed systems. Lecture 8 provides an introduction to transport with emphasis on
the important topic of diffusion. This process allows CSTRs to be applied to sys-
tems composed of segments linked by open boundaries. The case of an embayment
connected with a large lake is used as an iIIustration.
The next two lectures are devoted to closed-form solutions for idealized elon-
~ gated reactors.These are the basis for analytical solutions for streams (plug-flow re-:
\ actor) and estuary (mixed-flow reactor) water-quality models. Lecture 9 derives the!
mass balances and steady-state solutions for these systems. Lec/Ure 10 deals withc'
time-variable solutions. f
The remaining three lectures in this part focus on computer-oriented methods
for incompletely mixed systems. First, a general steady-state method, the control-
volume approach, is developed in Lecture 11. In addilion to presenting the method,
this lecture also explores the two constraints on the steady-state approach, model
positivity, and numerical dispersion. The steady-state approach is extended to time-
variable simulations in Lecture 12. The issues of accuracy and stability are intro-
duced and discussed. Finally several advanced time-variable methods are described
in Lecture 13.

135

...
LECTURE 8
- - -- - . ..
Diffusion

LEC11J!IE OVERVIEW: To extend our fraroework to incoroptetely mixed systems,

¡introduce the mechanism of diffusion, because of its role in the transfer of
l1)ass across open boundanes. Then 1 develop mass balances for a simple in-
coropletj:ty mixed case: a 'ake with an embayment. Both steady-state and time- I
vllril)ble liolutions are derived for this system. , ¡

Recall that in Lec. 1, I stated that two major processes dil:taled how pollulanls be-
haved in natural waters: kinelics and transporl. Ilhen devoted Lec. 2 to kinctics,
outlining dilferept1JYJ.>Csof reactions and how they can be measurcd.
In the ensuing lectures we treated Ihe other process-mass transpon-in a sim-
ple fashion. Because we were dealing almost exelusively with CSTRs, w~ had to
contend only with flows between reactors. Such one-way mass transpurt Ihmugh
rivers or pipes could be handled as a simple product 01'flow and concentralion.
In this lecture I extend our theoretical framework to encompass incomplclcly
mixed systems. These inelude systcms such as rivers, estuaries, embaymenls, and
coastal zones that are not chemically homogeneous. Hence we now musl .k..lwith
the issue of Ihe movemenLof mass within syslems, that is, across open boundaries.
\
8.1 ADVECTION AND DIFFUSION
Numerous types of water motion transpon matter within natural waters. Wind en-
ergy and gravity impart motion to Ihe water thal leads to mass transporto In the
present context within-system molion can be divided into two general categories:
advection and diffusion.

137
 --
138 PART11 Incomplelely Mixed Syslems LECI1JRE8 Diffusion 139

. . 8...
........ ..
Time
Time
. . ..e._.
e. ... t =O
.- QI .'9, - . .. .'-; .....
.':. . . .,, .' ...
........
.......
u .- .:,. '::::';::.
C)
ea
o-
C/)
.- al
o-
(J)
. .'
... .:.: .. 8.::;:.. . FIGURE 8.1
The transport
~
.. . . .....1
8...
. l. . .
/1 .- 01a dye patch
8. ... ~I t=l1t
in space and ... . ,I . . .
time via (a)
.- advectionand .
(a) Advection (b) Diffusion (b) diffusion.
. . .... ,. .
. . .. '. . . .
Advectioll results from ftow that is unidirectional and does not change the iden-
... .. .. It . . .... t=2l1t
,
tity of the substance being transported. As in Figure 8.la advection moves matter ,.
t from one position in space to another. Simple examples of transport primarily of this
type are the now of water through a lake's outlet and the downstream transport due
to ftow in a river or estuary. .. ..."
.. l.. . . .
Diffu.~iollrefers to the movement of mass due to random water motion or mix- .. :. .. . . . I , = 00

ing. Such transport causes the dye patch depicted in Figure 8.1b to sp~c:;adout and
... . ...~..
. .. ,. ........
i dilute over time with negligible net movement of its eenter of mass. On a micro- (1) (2)
FIGURE 8.2
The diffusion 01 mass between two vqlumes.
, scopic scale lIIoleclllardiffusioll results from the random Brownian motion of water
I molecules. A similar kind of random motion occurs on a larger scale due to eddie.s
I
and is calledtllrblllelltdiffllSioll.Both havea tendencyto miniQ1!ze gradients(that of the experiment (t = O) the partition is gently removed so that all the particles
¡s, differences in concentration) by moving mass from regions of high to low con- remain on the left side. If we return to the tank after a period of time (1 = dI), severaI
centration. \ ~ ~ ' c> ,,-1 J J \jY' ~ ~- of the particles will have wandered to the right side of the tank. Later (t = 2dl) more
The breakdown of the motion into the two idealized forms of advection and dif- would migrate to the right. Finally after a sufficiently long period, the eoncentrations
i
-} for the two sides would be equaI.
I
fllsion is innuenced by the sc~6f the phenomena being modeled. For example the
motion of water in an estuary might be WE.:.ived aS-E.rimarily advective on a sho!!. ~ The speed oé the process wiII be di<?tated by the strength of the ~ixing.!~rcc::~~
I time scale, since the rising or falling 113e causes water to move unidirectionally into Ir the tank were covered, Brownian motion would impart enough energy to the par-
i Or out of the estuary. If the modeling problem concerns the effeet of baeteria!QQ.!..- - tieles to move so me to the right, albeit at a slow rateo The imposition of mechanical
lution fro~r!.-tern:!...stormwater overflow epiSode, día~terization of the trans- mixing, say, by blowing air randomly aeross the tank's surface, would aecelerate the
port [lSadvection would be necessa:ry. Oñá lóngert1me scale, however, the tides process. In any event, as time progressed, more particles would show up on the right
I
would move water back andforth 1ñaéyclÍcaCíaSiíion and the lllotiQn J!ligbtprimar- unti! eventually they would be uniformly distributed throughout the tank.
, ily bt<.gassified as diffusive. In many cases transport-is considered a combination of ( As we already know, such movement of mass due to random water motion or
the two ~es, wfihthe emph~eflt!ent-un!!Jej;care Qf!fjeJ!!:0blem.- --" ~mixing is referred to as dijJusion. We can quantify this mechanism by developing
"' Water motioñwithin a i13tural waiér is a complicated process"ánd, as with any a mathematical model for the experiment from Fig. 8.2. Ir the left and right sides of
idealization, the above concepts are intended as a brief introduction to some of its the tank are designated I and 2, respectively, a mass balance for the left side can be
wri tten as
I essential features. In the following sections we focus our attention on the simplest
I dCI ,
I example of an incompkiely_mlx~~_syslem-=-aliiKe wjih_a single_tm1baYJl1~I!!:After
I using tnis .coittéXt to introduce diffuslon, asubsequent section will be devoted to
VI dt = D (C2 - CI) (8.1)
some other introductory information regarding transport. where VI = volume of the left side
I
CI and C2 = concentrations of the partieles in the left and r¡ght sides, respectively
D' = diffusive ftow (m3 yr-I)
8.2 EXPERIMENT
r Thus diffusion is modeled as an equal, two-way ftow connecting the volumes
Figure 8.2 illustrates an experiment that demonstrates what I mean by diffusive mass Fig. 8.3).
transport. A tank is divided in half by a removable partition. A number of tiny, neu-
Notice that three factors contribute to the diffusive transport between the two
trally huoyant particles are introduced into the left half of the tank. At the beginning
sides of the tank. First, the mixing ftow D' reflects the intensity of the mixing. Thus
I
 ~

140 PAR'T11 Incomplelely MiKed Syslems I.ECIlJKEx DilTusion 141

D' This result can Ihen be substiluted back ¡nlo Eq. ES.l.l to yicld
-'-------' 4/)'
cl CJ
o---
C2
FIGURE 8.3
C2
CIO
=T (1 - e
__o/
v )
D' A two-way Ilow model 01the diffusion 01mass. Thus, as depicled below, Ihe sides asymplolically approach a concentration 01'clO/2
with eigenvalues of 4D'/V. Therefore, Ihe lime lo gel 95% complele miKing is 3 V/(4D').
if the tank were subject to only weak mixing such as that due to Brownian motion,
D' ~oul~~~~!. If, however, it were subjected to "igorQ"-~ ¡>~'y~icalmixing, D'
would be large. 1 V,i.J...el 1
Second, the mass transport is directly proportional to the interface area. Thus "'

? if the interface area were doubled, we would expect twice as many particJes to be ... ~
transported. This effect would also be reftected by the magnitude of D'. 2
Finally diffusive transport is proportional to the difference in concentration be- -,
een the two sides.ThiS concentration difference, 01'grat!ie~t, inftuences both the
e tw
/ direction
. . ~nd the
. magnitud~
. of the transporto In the case of direction, if CI > C2, t95 FIGURE E8.1
theñmass woullbe expected to move from the left to the right. Such motion occurs
because more particJes are available to move from the left than would move from '>
the right. Consequently net motion is from left to right. Equation 8.1 represents this
motion by yielding a negative value; that is, as a loss of mass from the left side. 8.3 FICK'S FIRST LAW
Conversely if C2> CI>then mass would move from the right to the left.
In the case of magnitude, a large difference would mean that proportionatély Now that we have developed an intuitive and mathematical feel for diffusion, we
more particJes are transported. If no difference existed (CI = C2), then we would will explore how scientists and engineers have formally characterized the process.
expect that leftward motion would balance rightward motion, giving a net transport In 1855 the physiologist Adolf Fick proposed the following model of diffusion:
~
of zero. Such zero net transport is reftected by the fact that Eq. 8.1 would yield a dc
zero value. Jx = -D dx (8.2)

EXAMPLE 8.1. DIFFUSION OF MASS BETWEEN TWO VOLUMES. Model

= massftux in the x direction (M L-2 T-I) and D = a diffusioncoef-
whereJ x
the time required for the eKperiment shown in Fig. 8.2 lo go 10 95% of complelion.
ficient (L2 T-I). This model, which is called Fick's first law, specifies that mass
ftux is proportional to the gradient (that is, the derivative 01'rate of change) of con-
Solution: Recalllhat in Seco 6.5.1 we simulaled Ihe lemporal progression of IWOreac- centration. As dep!'cted in Fig. 8.4 the negative sign is incJuqeq to ensure that mass
tants undergoing a reversible reaction in a balch reaclor. A similar approach can be used . ~ux pr~~th~oITect di!.ection. Thus it is analogous to other constitutive laws
here. First, recognize thal if Ihe volumes of Ihe IWOsides of Ihe lank are identical, Ihe such as Fourier's law tor heat conduction 01'Ohm's law for electrical conduction.
balance for Ihe left side can be eKpressed as
.~';t For example Fourier's law states that heat ftows from regions of high to low temper-
V dCI ature. In a similar way Fick's law states that mass ftows from regions of high to low
--
2 dt
= D(C2 - c¡)
I
concentration.
where V = 10lal volume of Ihe lank. ~ Thediffusion coefficient D is a parameter used to quantify the rate of the diffu-
Further, again because the Iwo sides have equal volume, Ihe sum of Ihe concenlra- §iye process.As in our discussion of the experiment in the previous section, if the
lions in each side must equallhe initial concenlralion in the left side, tank were covered, the diffusion coefficient would be low to reftect the small rate of
(ES. 1.1) diffusion. If mechanical mixing were present it would be large.
CIO = CI + C2
Fick's law can now be used to model the situation depicted in Fig. 8.2. To do
This equation can be sol ved for C2 and liluted into Ihe mass balance 10give this we can write a mass balance for the left side of the reactor:

which can be sol ved for

dCI
di" + ~
D'
("1 =
2D'
VCIO

where
VId(
dCI
= -JAc
Al' = cross-~('ctional area of the interface between the two sides (m2) and
(S.3)

J = ftux between tb "olumes.

.4D', CIO _4D'
A finite-differencc approximation can be used to estimate the derivative at the
CI = clOe v + T (1 - e v I ) interface between the two halves,
 142 PART11 Incompletely Mixed Systems
" LECfURE
8 Diffusion 143
I
e e
-- molecular
!he - case the parameter is defined a~
Dlrectlon of Dlrectlon of D
mass flux mass flux
. .. Vd = e (8.7)
~
whereVd is called a diffusion mass-transfer coefficient (L T-I).
Further !he bulk diffusion coefficient D' or E' is freQ.!!ently used for mathemati-
de
~venience. Reca"LlhiL!h!~~~e.leri;i¡;;-;C~.;§l ~<!nfotJnds'i)~ih
l~earea
de <o
dx
- X
dx >0

X
-
and mixmg
tfiebulr --Téngih with the diffusLon. cQefl1.!:j~m..F.Q[ ~}!.amlli.eJQ!t!:lrbu!~f!.t diffusion,
coeffiCiel1t
~

E' = EAc
(a) (b) e (8.8)
FIGURE 8.4 is often used for convenience in model formulations.
GraDhical depicliDn DI the effect DI the concentration gradient on the mass flux.
~
Because mass moves "downhill" Irom high to low concentrations. the Ilow in (a) is
I Irom lell lo right in the positive x direction. However. due to the orientation 01cartesian 8.4 EMBAYMENT MODEL
cooldinales, Ihe slope is negative lor this case. Thus a negative gradient leads to a
positive flux. This is the origin 01the negative sign in Fick's first law. The reverse case
is depicled in (b), where the positive gradient leads to a negative mass flow Irom right
:~ ~ depicted in Fig. 8.5 a wc:!!-mixedbl!l c2.!lIIe~t~to a \¡Ir~ke i~!l simpleillus-
to lelt. tration oI an incomplete]y illixed system (Chapra 1979). This is particularly in-
structive because
described it is directly
in the previous comparable with'1t~ completely mixed lake model
lectures.
dc = c2 - cl
dx - e
(8.4) Mass balances for the main lake and the b~ can be written as

where f = mixing lenglh (L), which is the length over which Ihe diffusive mixing
lakes phlce. Equations 8.2 and 8.4 can be subslituted into Eq. 8.3 to give
dCI
V'-
dc
[0
dt =WI-QIC¡-kl . ICI Q2c2+ ""
~:~~
I

. E(C2-CI) (8.9)

~
I dCI DA,. I

¡ V, di = {(C2 - C,) (8.5) and V2 dt2 = W2 - Q2C2 F(c, - C2)

- , (8.10)

COlllparison of Eq. 8.5 with Eq. 8.1 indicates that we have now defined the where the subscripls I and 2 refer to the mai~~,-apd the bay, respectively.
diffusion flow in terms of more fundamental parameters,
8.4.1 Estlmatlon of Dlffuslon
D' = DAc (8.6)
e
The bay/lakemodelcanbe usedto estimatethe bulk diffusioncoefficienlforcases
According to Fick's law, diffusJQn fI()wis made up of th~ components. The where there is a gradient of a conservative substance (that is. one with k = O)
diffu~ciéñt D reflects the vigor of the mixing process.The area Af, accounts
for lhe fact tha" ¡he mass transport should be directly proportional to the size of the
inteitacem:ross which the l'I1ixing occurs. Finally the mixing length edefines the
~ across which the mixing takes placeJ _
, I -J1.!'0I!ffusi.2n coefficient D (L 2 r 1) serves as the fundamental parameter to
, I
I
qu~nti.!i' the diffusion process. It should be mentioned that alternative parameteriza-
tion\; and nomenclature are somelimes used and can lead to confusion. Maln
,
l'
For example a distinction belween molecular diffusion and turbulent mixing is lake
! often made. Even though they take the same form malhematically, we use different (1)
nomenclature to designate them: D for molecular and E for turbulent diffusion.
Aside from nomenclature. parameters are often clustered for convenience. l!L
cases where e~tim~~.n ~f ..!..hemixing le.n&!his. difficul1.Qr impossible.J.J~ ]~IDU!L
FIGURE 8.5
is oncn combincd wilh - Ihc- dTITusloncoefticicnl
- - lo devclop
--' a sin~le
. parameler. For A lake/embayment system.
 --.a ~

144 I'ART11 Incompletely Mixed Systems LECTURE~ Dilfusioll 145

between the main lake and the bayoThis is done by writing a mass balance for the or in more conventional unils,
conservative substance, as in 10" cm2 Yr o

V2di
dS2
= W2 - Q2S2
I

+ E (SI - S2) (8.11)

E = 1.48 X 109m2 yr-I x ---
( m2 365o 86.400 s )
= 4.7 x lOs cm2 S-I

where S designates the concentration of the conservative substance (M L -3). At

steady-state, Eq. 8.11 can be sol ved for 8.4.2 Steady-State Solutlon
E' = W2 - Q2S2 (8.12) At steady-statethe concentrationin the bay and the main lake can be determined
S2 - sI ;f using methods described in Lec. 6.

EXAMPLE 8.2. USING NATURAL TRACERS TO ESTIMATE DlFFUSION. For EXAMPLE 8.3. STEADY.STATE TOTAL P BUDGET FOR LAKE HURON
Saginaw Bay the bulk diffusion coefticient can be estimated by using the gradient of the AND SAGINAW BAY. Information for Ihe Lake Huron/Saginaw Bay system is sum-
conservative substance chloride. Estimate the oiffusion coefticient ano the mass-transfer marized in Table 8.1. The sedimentalion loss of tolal phosphorus can Ix: parameterized
coefficient for the process. by an apparenl seuling velocity v of approximalely 16 m yr-I (Chapra 1977). Use
Ihe models described in Lec. 6 10 determine (a) the inllow cOl1cenlmtion ano (b) the
Solulloo: Substituting values from Table 8.1 inlo Eq. 8.12 yielos steady-state concentration.

E ' = 0.353 X 1012 [7 X !tI

15.2 _ 5.4
- 2)] = 25 2 109
. X
3
m yr
-1 Solutioo: (a) ". l/late the inftow concentration for Ihe main lake, we musl oeler-
mine its inftow 1;1 I his is done by correcting its outftow rate to account for Ihe f10w
Note that the rale of oiffusion is considerably greater than the bay's outftow rate of 7 x from Saginaw Bay:
109 m3 yr-I.
The cross-seclional area belween Saginaw Bay and Lake Huron is approximately
Ql.in = QI - Q2 = 161 X 109 - 7X 109 = 154 X 109 m3 yr-I

0.17 X 1(j> m2. The mixing lenglh is approximately 10 km. Therefore the mass-transfer The other values from Table 8.1 can then be used to compute
coefficient can be compuled as
4.05 x 10'2 _
E' 25.2 X 109 CI.in = 154x 109 = 26.3 ¡.LgL 1
1.48 x lOs m yr-I
Vd = -- = 0.17 x 1()6
1.42 X 1012 _
and the diffusion coefticielll as C2.in = 7 X 109 = 202.9 ¡.LgL 1

E = Vde = 1.48 x IOs(lO x 103) = 1.48 X 109 m2 yr-I

TABLE8.1
Parameters for tbe Sagill. l\aylLake Huroo system
1.42 X10'
Parameter Symbol Value Units

SagillulV Bay:
Volurne V2 8 10. rn]
H2 5.81 rn
Depth 0.62 X 1012
Surface area A2 1,376 10" rn2
7 (11.4%)
oulnow Q2 10. rn] yr I
Chloride concentration S2 15.2 g rn-]
Chloride loading W.,2 0.353 1012g yr" 1
Phosphorus loading W.,,2 1.42 1012rng yr I
4.13 X 1012
Lt,ke HUYOIl:
(75.6%)
Volurne V, 3507 10. rn]
H¡ 60.3 rn " Saglnaw Bay Lake Huron
Deplh
Surface area Al 58,t94 10" rn2
Outnow QI 161 10. rn] yr-I FIGURE 8.6
Chloride concenlrdtion SI 5.4 grn-J A total phosphorus budget lor Lake Huron and Saginaw Bay computed in
Phosphorus loading w.pl 4.05 1012 rng yr" 1 Example 8.2. Note that the percentages are based on the totalloading to
the system.
 LECTURE
8 Diffusion 147
146 PART
11 Incompletely Mixed Systems

Thus the inftow concentration for Saginaw Bay is extremely high. CI = -0.037e-6.717t + 4.476e-O.3141t
C2 = 26.18e-6.717/ + 2.18e-o.3141'
(b) The steady-state solution is (Eqs. 6.13 and 6.14)
1 1 ~\i.
Values 01' time can be substituted into these equations and the results displayed
CI = \.102 x 1012Wpl + 1.857x 1012W,.2= 3.671+0.768 = 4.44J.LgL-1 in Fig. 8.7. In addition a semi-Iogarithmic plot 01' the response is depicted in Fig. 8.8.
The response 01' Lake Huron is dominaled by its slow eigenvalue. This implies that its
I I -
response is only slightly inftuenced by the response 01' Saginaw Bay. For Saginaw Bay
C2 = 2.373 X \012 WI'I + 5.345 x \010 W,'2 = 1.705 + 26.658 = 28.36 J.LgL I
the fast eigenvalue dominates initially. Thus the bay manifests a substantial improvement
Figure 8.6, which shows the magnitude 01' the individual terms 01' the steady-state in less than 1 yr. However, after 1 yr the slow eigenvalue retards Saginaw Bay's recovery
as its fate becomes dominated by the sluggish recovery 01'Lake Huron.
budget, indicates the importance 01' the sedimentation mechanism. Note that about 87%
01'the total phosphorus load ultimately finds its way into the bottom sediments. In addi-
tion note that diffusion represents a significant mass-transfer mechanism. It accounts for 100
approximately 42% 01'the total phosphorus that leaves Saginaw Bay.
II Saglnaw Bay (total)
I
~ I
I
!
10 LakeHuron
8.4.3 lime-Variable Solution (totaUslow)
The time-variable concentrations in the bay and the main lake can be determined
e
I
using the methods described in Lec. 6.

EXAMPLE 8.4. TIME. VARIABLE SOLUTION FOR LAKE HURON AND

1:1: \\ Saglnaw Bay (slow)
SAGINAW BAY. Determine the temporal response 01'the Saginaw BaylLake Huron Saglnaw Bay ('asl)
system following termination 01'loads. Assume lhat the system is initially at the steady- 0.1
state concentrations determined in Example 8.3. o 4 8
Solution: The eigenvalues can be computed as (Eq. 6.43)
t (yr)
Al _ J _ _ 6.777 -1
( 1 0.1693) -0.314Iyr
+ FIGURE 8.8
A., -3.5451_
General solulions lor Lake Huron and Saginaw Bay lollowing
These, along with other values from Table 8.\, can be used to determine the general Ihe terminalion olloadings 01 tolal phosphorus. In this case Ihe
sol'ltions responses are displayed on a semi-Iogarilhmic plollo make the
eigenvalues obvious.

4.5- 30
BOX 8.1. IThe Mlxlng Lengt~(Con'uslon about Dlffuslon)
3.5
\ Estuaries were the first syslems where the diffusion mechanism was incorporated into
::' 2.5
t\ /
Slow 20
\ water-quality modeling. When the estuary-modeling theory was applied to other in-
completely mixed systems, some confusion resulted regarding the proper choice 01'the
.:,¡ mixing length.
t \ I . Total
The typical estuary modelillg situation is depicted in Fig. B8.la. As shown, such
'<:;' 1.5
Total systems were usually divided into segments. Because many segments w~r~ e~p.!!?~q...
the gradientsof pollutant betweenthe. sel\'!I~!l!S we!~-Us~al1Y..$lILa.U:consequently the
0.5 Fast s_ FIGURE 8.7
General solutions lor
"Segiiiei1i léiigtfi served as the proper choice for the mixing length. The validity 01' this
O (a) Lake Huronand ) approach can be affirmed by recalling that the purpose 01' the length is to provide an
8 . accurate finite-difference estimate 01'the derivative in Fick's law (recall Eq. 8.4).
-0.5 4 8 O 4 (b) Saginaw Bay
lollowingthe In subsequent years, when modelers began to simulate bayllake systems, the
t (yr) t (yr) terminalion olloadings situation was usually quite different. As depicled in Fig. B8.lb the bay and the lake
(a) Lake Huron (b) Saginaw Bay ollolal phosphorus. are treated as CSTRs separated by a constriction. Thus each has a relalively constant
 I.ECTlJREx Diffusion 149
148 PART11 lncompletely Mixcd Systems

e ] \,Lt..-I [ ~
t t +-

FIGURE 8.9
Complicated incompletely mixed systems such as estuaries can be
represented as a series 01 coupled completely mixed systems.
(a) One-dlmenslonal estuary x

8.5 ADDITIONAL TRANSPORT MECHANISMS

To this point we have provided a brief introduction to the diffusion mechanism. Now
. we would like to present some additional information about transportoFirst, we dis-
I ,
, I cuss the magnitudes of diffusion coefficients encountered in natural waters.
I I
I I
I I

ti
,I
---: " ~ 11
8.5.1 Turbulent Diffusion
e "
I
,
I
, Just as mass is transported via random molecular motion, matter is mixed by largcr
, scale eddies or whirls in natural waters. If long enough observational time and space
I ,:1:1'
I
I
scales are employed, this motion can be viewed as random and treated mathemati-
I
I \ cally as a diffusion process. Although equations of the kind developed for molecular
motion can be used to characterize turbulent mass transport, two important differ-
ences should be noted.
(b) Bay/maln lake x c: First, because the eddies are much larger than the random motion of molecules,
'í mixing by turbulent diffusion is much greater than by molecular diffusion. As shown
FIGURE B8.1
in Fig. 8.10, turbulent diffusion coefficients are several orders of magnitude greater
'11 than those on a molecular scale. Note also that horizontal diffusion is generally much
concentration, and the gradient or transition occurs in the narrow boundary region greater than vertical diffusion. Likewise, effective diffusion through porous media
where diffusive mixing is minimized. Consequently the mixing length is best repre- such as bottom sediments is less than molecular diffusion in free solution because,
sented by the length of the boundary region rather Ihan by the average length of the among other things, the solute must move around particles.
adjacent sections. Second, in contrast to molecular diffusion where the motion is assumed to
When we use either the bulk diffusion E' or the mass Iransfer coefficients Vd.
the issue of mixing length is a moot poi nI because Ihe mixing length is part of the
coefficient. However, when the analysis is based on the diffusion coefficient E, the
1 be ¡denticalin scale,turbulenceis composedof a wi1crarge of ed~y siz:s. Thus it
choice of the length is critica!. A related example wil\ occur later in this text when we r _ Estuarles
model vertical temperature profiles in stratified lakes. LONGITUDINAL DISPERSION
_ Streams
... I . Oceao
In summary the incorporation of an embayment into our completely mixed
model has certainly added to its complexity. If the lake had two bays, the analyti-
cal solution would be even more unwieldy (three equations with three unknowns).
-
MOLECULAR

DIFFUSION
HORIZONTAL TURBULENT DIFFUSION
Lakes

Consequently more complicated incompletcly-mixed systems are almost invariably DIFFUSION IN POROUS .. .. . Surfacelayer
solved numerically with a computer (see Lees. 11 through 13 for details). . ... VERTICALTURBULENT DIFFUS,dN
MEDIA (SEDIMENTS) .. .. Deeplayel
Although the modeling of most incompletely-mixed systems involves comput-
erized solution techniques, the conceptual basis for these models is straightforward.
As in Fig. 8.9, an estuary is often treated as a series of coupled, well-mixed reactors.
10-8 1O-S 1Q""4 1~ 100
1()2 í 1()4
108 108

cm2.-1
The derivation of the mass balances for these reactors is then handled in a similar
fashion to the lake/embayment system addressed here. The equations are then solved FIGURE 8.10
simultaneously using numerical methods. Typical ranges 01the diffusion coefficient in natural waters and sediments.
 LECTURE
H Diffusion 151
150 PART11 Incnmp\ctely Mixed Systel11s

Length Scale (km)

1 10 102 103 104 ',.
10-2 10-1

--
(a) Dlffuslon
109

I 108 Radial..l.
CJ
el)
en
..
.
HYPOllmlllulI. 'l

-
I
- 41 ) f1" (b) Dlsperslon

g 106
'0::J
-----
't
J FIGURE8.12
Contrast between dillusion and dispersion. 80th tend
,/
:t:: .., ( to "spread out" pollutants. Dillusion is due to random
11°l
i5 105 \ --- motion 01 the water in time, whereas dispersion is due
i to dillerential movement 01 the water in space.
\ "
1104 - :J't\8 ¿f \ 1\, 1J' or shear. The net effect of these differences in mean velocity is to spread or mix the
Lo d~.a!Q.n,g ~I:léiipe's axis~ Iniérestiñgly it can be shown('táylor 1953, Fischer et al.
::J
.... 1979) that, coupled wÍth ranoom radial movement due to diffusion, this dispersion
-
tü 103
e
FIGURE 8.11
Relalionship 01 horizontal
(givel!.enouglitimc) cañ be represeñte~-as a Fickian..diffusion process.
2 dillusivityand scale length in In the environment, turbu\ent diffusion an~ dispersion cari, individually or in
.¡: 102 the ocean and Lake Ontario. ( concert, cause mixing of a substance. For example in rivers and eswaries, dispersion
O
:::I: Unes deline an envelope around '-¡ usually predominates because of the strong shears developed by the large mean flows
10
104 105 106 107 108 109
the oceanic data with a slope 01
4/3 (Okubo 1971, with additional
) (du~ t9 gravity now and ti des) and the constraining channel. Dispersion can also be
103 Irf of some importance in highly advective river-run reservoirs. Additionally, for all
Length Scale (cm) data Irom Murthy 1976).
cIasses of water, dispersion is often important on smaller time and space scales. In
particular, shears can be developed near system boundaries such as the shoreline and
would be expected that the resulting transport is scale-dependent. It has been shown
the bouom.
thatthe turhulentdiffusioncoefficientE varieswiththe ~ powerof the scaleof the For lakes and wide bodies of water such as bays, diffusion tends to predominate.
phenomenon (Richardson 1926). This relationship is supported by observations in For these cases the wind is the primary agent imparting random motion to the water.
hoth oceans and lakes (Fig. 8.11). In particular, for longer term simulations, wind mixing acts as a random process and
The scale dependence of eddy diffusivity can have great practical consequence can be adequately represented as turbulent diffusion.
for modeling pollutant transport in natural waters. For cases where a pollutant is In the remainder of this book we will use the term "dispe~siol.!':wJ1e~ealing
discharged rapidly in a small area (that is, a spill), the diffusion of the resulting with narrow, nowing bodies of water such as one-dimensional streams and estuaries.
cloud would accelerate as it grew. This is caused by larger and larger eddies playing For lakes, bays. and vertical transport we will usually employ "diffusion."
a role in the spilrs mixing as it spread. When the spill grew bigger than the largest
eddy, a constant diffusion coenicient could be used to approximate further mixing.
Models to account for this sort of phenomenon are described elsewhere (Fischer et 8.5.3 Conductlon/Convectlon
al. 1979). Finally there are some additional mechanisms and terminology that are used in
water-quality modeling and bear mention. Conduction and convection are two pro-
I 8.5.2 Dlsperslon cesses that originate from heat transfer and aerodynamics that are roughly analogous
I to diffusion and advection.
To this point we have focused on diffusion due to the random motion of water (Fig. ConductiOl' refers to the transfer of heat by molecular activity from one sub-
I 8.12a). Dispersion is a related p~<; that als05~':'~~ pollutfi~t~_to,spre~~.~ stance to another or through a substance. Because conduction is similar to molecular
in contrast to random water motlOn In time, disper~iolt i~ t e result of velocity dif-
diffusion of mass, individuals will often use the two terms interchangeably. Some
ferences'in space:Porex;;;TIple ~<;uppoSéthal dye were introduced into water nowTng
\
may even take the liberty of extending the term "conduction" to refer to turbulent
through a pipe (Fig. 8.12b). In such cases ~molecule of dye near the wall of the pipe
would move more slowly t~an a mo.!:.<:ulenear the center dueto a velQcity gradient

.-- -
152 PART11 Ineomplelely Mixed Syslems LECTUREM Diffusion 153

diffusion. In lhis text we will use it only when we eover surfaee heat transfer in PROBLEMS
Lec. 30.
Convection, which generally refers to the mOlionsin a fluid that result in the 8.1. A wdl-mixed lidal bay i~ ,. ..¡¡cctt:d wilh Ihe ocean by a challnel.
transpon and mixing of the fluid's propenies, takes two forms. Free convect;on
refers to venical atmospheric motions due to the buoyancy of heated 01'cooled fluido
For example in meteorology, the rising of heated surface air and sinking of cooler Lenath
air aloft is called "free conveclion." Oftentimes the term "convection" is assumed
to imply free convection [for example, in meteorology (Ahrens 1988)]. In contrast
forced convection is due to external forces. An example is the lateral movement of Ocean
heat 01'mass due to the wind. Tnus forced convection is akin to advection. In the
Wldth
present context, we will use the term convection only when we are discussing free f
convection in relation to thermal stratification. Chennel
In summary water molion in natural waters is a complicated process and, as with Bay FIGURE P8.1
any idealization, the above concepts are intended as a brief introduction to some
of its essential features and terminology. In-depth discussions of the subtlelies of The bay and Ihe channel have Ihe following charaelerislies:
mixing (Fischer et al. 1979) and lake hydrodynamics (Hutchinson 1957, Monimer
1974, Boyce 1974) are presented elsewhere. In addition a more detailed discus- lIay Channel
sion of transpon processes in a number of natural waters is included in Lecs. 14 1000 100
Length (m)
thrcugh 17. Width(m) 500 10
Deplh(l1I) 5 2

BOX 8.2. Modellng Cocktall.Party Etlquette AII = O d, 100 kg 01' a conservalive dye is inslanlaneously released and dislribuled
Imagine you're al a cocklail pany with a group 01'environmenlal engineers. As might across Ihe bayo The following measuremenls 01' dye concentralion are made over Ihe
ensuing period:
be expected Ihey're making smalllalk aboul pollulion. In a loud voice you offhandedly
mention Ihal you are sludying a lake wilh a horizonlal eddy diffusion coefficienl 01'
aboullO-6 cm2 s-l. Suddenly Ihe room grows unnalurally quiet. One oflhe engineers I (d) 20 40 óO !lO 100 1.'11
rolls her eyes, while two glance al each olher and snicker. Olhers avoid eye conlacl, as c(ppb) 32 29 23 20 17 16
one by one Ihey wander off. Before you know il you're standing alone. Upon reftec-
tion your heart plummels. You recognize Ihal you have committed an environmenlal-
modelingfaux paso Sweallrickles down Ihe small 01'your baek and your face bums as (a) Whal is Ihe ditTusion coeflicienl for exclmnge belween the bay alld Ihe ocean'l
you recall (Figs. 8.10 and 8.11) Ihal your slated diffusion coeflicient is at leasl 6 to 8
Express your result in m2 d-I .
orders 01'magnitude below a credible value. You slink loward Ihe door, berefl, a pariah. (b) Whal mass loading 01' a reaclive pollulant (k = 0.01 di) could be illpullo Ihis
No one says good-bye. syslem under steady-slate eondilions if Ihe allowable eoncenlralion in (he syslem
Clearly one mark 01' Ihe experienced modeler is Ihe eapabilily lo recognize rea- were I ppm? Express your result in g d-I. Note Ihal bolh Ihe conservalive dye and
sonable values 1'01'common model paramelers and variables. Aside from prevenling Ihe reaclive pollulanl have negligible concentralion in the oeean. Also Ihe residence
social embarrassmenl, such knowledge can be invaluable in professional conlexls. lime in the channel is negligible compared lo the residence lime 01' Ihe bayo
For example you could be ealled upon as an expert lo crilique someone else's 8.2, A well-mixed lidal bay i~ adjaeenlto Ihe ocean as shown in Fig. P8.2. Thc 1.1) has a
model. Recognizing odd paramelers and variables is an importanl tool in such exercises. diameler 01' I km and a deplh 01'5 m. The short channel conneeling Ihe Iwo walcrbodies
In addition, when developing your own models, you will oflen be unable 1" eSlimate is 50 m long, 10 m wide, and 2 m deep. The Iransfer belween the bay and Ihe ocean
:111model paramelers direetly. In such cases educaled guesses must be made. Finally
somelimes a seemingly odd parameler eSlimale serves as a valuable clue in Ihe model-
developmenl process. For example it could mean Ihallhe model is missing an importanl
mechanism. Altemalively il could mean Ihallhe syslem being sludied has atypical char-
aClerislics. Such evidence could result in Ihe idenlificalion 01'research and dala gaps.
In all these cases the ability lo idenlify Iypical ranges 01'paramelers and variables Ocean
is an invaluable asset 1'01'Ihe expert modeler. Throughoullhe remainder 01' Ihis lext I
will endeavor to present such ranges to help you build Ihis capabilily.
FIGURE P8.2

I
 LECJ'URE8 Diffusion 155
154 PART11 Incompletely Mixed Systems

QR (103 m3 cJ-1)
has been characterized by a turbulent diffusion coefficient of I<P m2 d-I. Determine
the bay's 95% response time for a spill of a substance that settles at arate of 0.1 m d-I.
1000
8.3. Green Bay is a major embayment connected with Lake Michigan. The Green BaylLake
Michigan system can be modeled as three reactors,

;.....
CR (J.1g L-I) Bay ;

Lake ;...
200
Huran

Lower Upper
2 I(d)
Green Bay FIGURE P8.4

8.5. A welI-mixed circular bay is connected to a large lake. The lake is so big relative to
the bay that for alI intents and purposes it can be assumed to be an "ocean"; that is,
the bay has negligible impact on its state. At steady-state the bay and the lake have
chloride concentrations of 30 and 7 mg L -', respectiveIy. The bay has the folIowing
characteristics:
Lake Michigan
36
Radius = 0.25 km
5.4 10.8
Flow. km' yr -, 70 Inftow = outftow to main lake = 5 x 106m3yr-I
20 30
Butk diffusion. km' yr-'
4846 Inftow chloride concentration = 70 mg L-1
7.5 55.4
Volume, km'
53537 Mean depth = 5 m
953 3260
Surrace area, km'
(2nO 200 5500 Cross-sectional area of interface with main late = 500 m2
Phosphorus loading, mla Mixing length of transition zone between the bay and lake = O.I km
¡;al
F!GURE P8.3 (a) Determine the diffusion coefficient to parameterize turbulent exchange between
the two water bodies. Express your result in cm2 S-l.
(b) How much loading of a highly concentrated waste (that is, its now is negligible)
If phosphorus settles at arate of approximately 12 m yr-I , use the matrix inverse ap- could be added to the bay while maintaining a concentration of 10 mg L -I? Note
proach to compute the steady-state concentrations. Assume that Lake Huron has a con- that the waste decays according to first-order kinetics at I yr-I. Assume that the
stant concentration of 4.4 J.LgL -l. main lake will have zero concentration of this polIutant. Express your result in kg
yr-I.
8.4. A bay is subject to a runoff event of a contaminant (see Fig. P8.4) that volatilizes at (e) What is the bay's 95% response time for the waste in days?
arate of 1 m d-I. Assume that innow =
outftow during the runoff event and that the
main body of water has negligible concentration of this contaminant. The bay has the
IÚllowing characteristics:

Volume = 10 x 106 mJ
Mean depth = 3 m
Normal outnow = O
Interface diffusion coefficient = I x lOs m2 d-I
Interface cross-sectional area = 1500 m2
~>
Interface mixing length = 100 m J:.

(a) Detcrminc the pcak conccntratioll in the bayo

(b) Whcn will the bay conccnlration be lowered to 5% of Ihe peak value?

.-
 ..

u,nURE ~ Distrihuted Systems (Stcady-Slatc) 157

LECTURE 9
- lin

Distributed Systems -~
-
e
(Steady-State)
o L x
FIGURE 9.1
An elongated rectangular reactor.

where AV = volumeof the element(L3) = Ac Ax

Ac = cross-sectionalareaof the reactor(L2) = BH
B = channel width (L)
H = depth (L)
lin and = flux of mass in and out of the element due to transpon (M L-2 T--I)
loul
reaction = gain or loss of mass within the element due to reaction (M TI)

9.1.1 Plug-I:.!0"!f!!actor (PFRJ

y~ TheDlu~-flow
react'l.risonein~hicha~~ect!ondo~Jlt.es..AsinFig.9.2, this means
that a "plug" of conservative dye introduceóáfüñe end will remain intact as it passes
through the reactor. In other words~isch~r~~dln.Jl1e~'luen~
To this point I have focused on completely mixed reactors or CSTRs.In Lec. 8, as thev enter the.reaClli!L
I illustrated how incompletely mixed systems can be modeled by breaking them F~~w reactor.,t!t~~illtQ.the ele~entJs defined as
into a coupled set of CSTRs (recall Fig. 8.9). Such characterizations are formally
referred to as lumped-parameter systems, so called because each reactor has its own " !in = t!f (9.2)

set of parameters. Thus even though the system is continuous, we approximate it as

"lumps."
In contrast it is possible to model such systems while maintaining their contin-
,
where U
---
= velocity (L T-I)
Taylor-seriesexpansion,
-
r
= Q/Ato.The
- -
ac
-
flux out is estimated by a hrst-order

uous nature. Such characterizations are formally referred to as distributed-parameter

systems. In this lecture I will introduce you to such systems. To do this I focus on
F(
jJoul = U e + axAX ) (9.3)

..
two idealized reactors: plug-flow and mixed-flow reactors. I also show how these
models can be used to simulate steady-state distributions of pollutants in streams
and estuaries. - - -. -..
9.1 IDEAL REACTORS e
B{)thplug-flow and mixed-flow reactors are elongated rectangular.basins..as d~pi~ted
i~9.-l. They aie_a~s.tLñ.jed
tobe 'weIH1}i~d laterallyJl) and vertically (z). Thus
we are concerned only with variations in the long!tudinal dimension (x).
.. - ... - -.,_._-

As in Fig. 9.1, a mass balance is taken for a differential element ofiength tu, o L x
\ . ac flc..\JF.E: 9.2
AVat = linAc - lou¡Ac :!: reaction (9.1) A plug-flow reactor.

156
 158 PART11 Incompletely Mixed Systems LECTURE
9 Distributed Systems (Steady-St.ate) 159

Finally, assuming a first-order decay reaction, - t

Therefore this equation can be used to compute
those from Example 5.2 are shown below:
the distribution. The results along wit~

[Reaction =- k~Ve J (9.4)

'/ where the bar over the Cdenotes the average value for the element. Substituting the
I individual terms into Eq. 9.1 yields 0.8
ac ac
AV-
- (
aI = VArc - VAc c + -ax Ax - k AVe ) (9.5)
e
0.6

0.4
Combining terms gives
ac ac (9.6) 0.2
AV- = -VAc- Ax - k AVe
al ax
20
Dividingby AV = Ac Axandtakingthe limit(AxJ -+ O)yields 40 60 80 100
. x
',Bc( ac W5;.
! '--
I al = -u ax" - k~ c:leC(,;,IM""'"' ,..,. (9.7) FIGURE E9.1-2

Qt steady-state Eq. 9.7 becom~ ~ tIv.\'() .;¡ d ')

_
As the number of reactors becomes larger, the cascade model seems to be approach-
ing the analytical solution. In fact it can be proved that the analytical solution is the lim-
\ IO = - V dc - kc
.sJ.x
) (9.8) it of the cascade approximation.

which, if c = Coat x = O,can be solved for

\ -1., -- (9.9)
c = coe u 9.1.2 Comparison 01 the ~ a'1d the PFR
EXAMPLE 9.1. PLUG-FLOW REACTOR. In Example5.2 we used a cascade Perf~rmanceof the CSTRand the plug-ftowreactorcan nowbe compared.A nice
m()c1elto simulate the steady-state distribution of concentration in an eIongated tank. l way to do this is via the residence time required to achieve a certain leve! of per-
~ formance efficiency. For example recall that the steady-state solution for the CSTR
wjth fi(st-ordeueaction can be represented aS.
Ac
Q
c = cinQ + kV (9.10)

'1
-
e 6.x
The performance efficiency can be defined in terms of the transfer function (Eq.
1~ ~
~.. ~
x
_c _- Q
13-- (9.11)
L - Cin - Q + kV
o L
Dividing the numerator and the denominator by Q gives
FIGURE E9.1-1 I
(9.12)
13= I + krw
The tank has cross-sectionalarea A,. = 10 m2, length L = 100 m, velocity U =
100 m hr-I, and a first-order reaction rate k = 2 hr-I. The innow concentration is I mg where Tw water residence time = V/Q. This result can be solved for residence
L -1. Use the plug-flow model to compute the steady-state concentration distribution for time,
the tank. Display the concentrations along with the results for the 11 = 4 and 8 CSTR
approximations from Example 5.2. I 1-13
(9.13)
Solution: The plug-flow model is
Tw = kT
e = 1('-I~r' = le-O.02x In a similar fashion the efficiency of the PFR can be defined as

.....
 160 PART11 Incomp1elely Mixed Systems L~CTUK~~ Distributed Syslems (Sleady-Statc! 161

15 For the mixed-tlow reactor, the flux in is defined as

ae
J.
11I
= Ve: - E -ax I
(9 17)
.
~ 10
where E = lurbulenl ditfusion. Note Ihat Ihe second lerm is Fick's tirsl law (Eq.
8.2).The flux out is
~ ' ae ae a ae:
~
-------------------- J001 ( ) [
= V e + axt1x - E ax + ax ax t1x ()] (9.18)

o 0.2 0.4 0.6 0.8 FIGURE 9.3 Substitutingthe individuallerms into Eq. 9.1 yields
Ralio 01eSTR lo PFR residence
Efflclency times versus efficiency.
ae ae
I
t1Val = VA"e - VA,. e + axt1x ( )
(.)._ e
=-=eu -Lx .\. ae ae a ae _
I
I ein
(9.14)
- EA,.-
ax + EA" [ uX + ax
-;- - -ax t1x - k t1Ve:
] () (9.19)

I
(!I Recognizing thallhe residence lime of Ihe PFR is equalto L/ V , Combiningtermsgives
I J3 = e-kTw (9.15) ae ae ~e
I t1V-al = -VAc-;-t1x + EAc2t1x - k t1ve (9.20)
I This result can also be sol ved for residence time, uX ax
:ti
I
Dividingby t1V = A,.t1x andtakingthe limit(t1x -+ O)yields
II Tw = ~ In (~) (9.16)
{ ae ae a2e
(9.21)
The two types of reactors can be compared by plotting the catio of the two res- 1 al = -V ax + Eax2 - ke
--
I
I idence times versus efficiency. The result is shown in Fig. 9.3. This plot shows that
I the CSTR always requires a higher residence time lo achieve a desire4 efficiency.
or at steady-state, -
I This leads to the general conclusion that for first-order reactions the PFR is more de ¡(le
I ~ efficientthan the CSTR for removing pollutants. l\-o = -V-dx + E- - ke:
\
(9.22)
I dx2
I
The general solution can be obtained in a variely of ways. A simple approach is
9.1.3 Mlxed-Flow Reactor (MFR)
to assume thatthe Solulion has the form

The mixed-flow reactor is one in which bOlh advection and diffusionldispersion are e = eAx (1).23)
importan\. As in Fig. 9.4, this means that a plug of conservative dye introduced at
This solution can be substituted into Eq 9.22 to arrive at Ihe characterislic equalion
one end will "spread out" as it passesthrough the reactor.
EA2-VA-k=0 \,J.24)

(
\
- ~
~~

~ ~
which can be solved for

( Al V rill V ~
(9.25)
) A2 = 2E ( I:!: Vi + lf2 )= 2E(l :!: ,,1 + 417)

1
o:

- - -
o L JC
?:
where 17= kE/U2.
Therefore the general solution is

where F and G =
e
constants of inlegration.
= FeA,x + GeA2X (9.26)

The constants of integration can be evaluated by boundary conditions. For Ihe

FIGURE 9.4 tank in Fig. 9.1 a boundary condition can be developed by taking a mass balance at
A mlxed-flow reactor.
'" the inlet,

- ....--..
 162 PART11 Incompletely Mixed Systems LECTURE
9 Distributed Systems (Steady-State) 163

dc They can be displayed as

(9.27)
QCin = Qc(O) - EAc dx (O)

Dividing by Ac and substituting Eq. 9.26 gives 1

(V - EA¡)F + (V - EA2)G = VC;n (9.28)
0.8
The seeond boundary eondition relates to the faet that no diffusion of mass is
assumed lo oceur Ihrough the outlet. Therefore no gradient should exist at Ihe end of 0.6
Ihe lank. e
0.4
dc (L) = O (9.29)
dx
0.2
Using Eq. 9.26,
o
(9.30)
(A1eA'L)F + (A2eA2L)G = O o 20 40 60 80 100
These two boundary eonditions are eommonly ealled Danckwerts boundary condi-
tions, after the ehemical engineer P. V. Danekwerts, who proposed them originally
t
(Danekwerts 1953). FIGURE E9.2
Equalions 9.28 and 9.30 now represent a system of two equations with two un-

¡ --
knowns. They can be solved for l]ms al zero diffusion. the model converges on the.YEUs diffusioll is inSffiased
it approaches Ihe eSTRo
Vc. A2eA2L
F= In (9.31)
(V - EA,)A2eA2L- (V - EA2)AleÁIL

and G = VC;nAleA1L (9.32) BOX 9.1. The Peclet Number

(V - EAÜA,eA,L- (V - EA¡)A2eA2L
A dimensionless analysis of Eq. 9.22 can be used to gain insight into the results of
These eonslants can be substitutedinto Eq. 9.26, which can then be used to ealculate
Example 9.2. To do'tlJis, the fo\lowing dimensionless parameter groups can be defined:
eoneenlration along the tank's lenglh.
.
x = -x
L (9.33)
EXAMPLE 9.2. MIXED-FLOW REACTOR. For the same system as in Example
9.1, compule concenlration using the mixed-flow model with diffusion coefficients of c . =- c
E = 2000 and 10,000 m2 hr-I. Plot your results along with the PFR and CSTR models Cin (9.34)
for the same lank.
These can be solved for x and c and the result substiluted into Eq. 9.22 to give
Solution: The fo\lowing results can be generated: U dc' E d2c'
O = - L dx' + U dX'2 - kc' (9.35)
E = O(PFR) E = 2000 E = 10,000 E = oo(CSTR) le Mulliplying Ihis equalion by L/U yields
Á, -0.02
'X> 0.0653
-0.0153
0.02
-O.O!
0.0
0.0
dc' E d2c'
O=--+ kL
c. (9.36)
A2 dx' UL dX.2 U
F o 5.66 x 10-5 0.0126 0.1667
G I 0.7656 0.5063 0.1667 or dc' 1 d2c' .
(9.37)
1.0000 0.7656 0.5189 0.3333 {'h fmtlL 0= - dx' + P, dx'2 - Dac
x=O (¡
x = 20 0.6703 0.5638 0.4333 0.3333 1./1/\-:' where P, is ca\led !he Peclet number,
x = 40 0.4493 0.4157 0.3674 0.3333
x = 60 0.3012 0.3083 0.3197 0.3333 ( P _J-U _ rate of adveclive transport (9.38)
x = !lO 0.2019 0.2354 0.2899 0.3333
0.3333
\ ,- E - rale of diffusive/dispersive Iransport
x = IIX) 0.1353 0.2044 0.2794
and D" is ca\lcd ,he Damkohler number,

I
\.-
 164 PART11 Incomplelely Mixed Syslems
I.ECTURE9 Dislrihuled Syslems (Sleady-Slale) 165

D .= -kL
U
=. rale 01'consumplion
rale 01' advecllve
by decay
Iranspon
(9.39) Q.,c.,

Thus if P, is high (> 10), Ihe syslem approaches a PFR because Ihe firsl-derivalive ! !
lerm dominales Ihe second-derivalive lerm in Eq. 9.37. Conversely for low P,« 0.1),
Q,cr ---1- --L. Qco
1 longitudinal mixing dominales and Ihe syslem becomes more like a CSTR. For inler- ! I
I mediale values Ihe MFR model is required. I FIGURE 9.6
I I a Mass balance lar a point source
o r
discharged into a plug-lIow system.
i A Iv::JvV

9.2 APPLlCATIONOF THE PFR MODEL TO STREAMS

W + Qrcr
Just as the CSTR is the fundamental model for a lake. the PFR is the fundamental
Co = Qw + Qr (9.43)
( rnm!eUor a-stre3m.1IL thIT]eC1Ionwe emphasize ho~a plug-fIOwmodel can be
, a,pplied to analyze two types of sources: point and distributed loadings (Fig. 9.5). Now, in some cases, the waste source has a relatively small flow ami a high
concentration, whereas the river has a relatively high flow and a small concenlration.
In other words,
9.2.1 Polnt Source Qr » Qw (lJ.44)
The following analysis focuses on the case where a point-source loading is injected (9.45)
and Cr«cw \
into a channel having constant characteristics. For this case the solution would follow
Eq.9.9. For these cases, the following approximation holds:
The initial concentration Cocan be computed by taking a mass balance at the W
injection poiot. If it is assumed that complete mixing takes place in the lateral and Co =- (9.46)
Q
venical dimensions, the situation is as depicted in Fig. 9.6. A flow balance yields
Q = Qw + Qr (9.40) EXAMPLE 9.3. POINT SO URCE FOR A PLUG-FLOW SYSTEM. A poinl
/ source is discharged 10a river having Ihe following characlerislics: Q, = 12 x 1(¡6mJ
A mass balance can then be developed, d-I, c, = 1 mg L-I, Q.. = 0.5 X 106 m3 d-I, and e.. = 400 mg L-I.
O = Qwcw+ Qrcr- (Qw + Qr)CO (9.41) (a) Delermine Ihe inilial concenlralion assuming complele mi¡¡ing venically anJ laler-

that can be solved for

I ally. Assess whelher Eq. 9.46 is a good appro¡¡imalion for Ihis case.
( (b) Calculate Ihe concenlralion 01'Ihe pollulanl for 8 km below Ihe injeclion pojn!. NOle
QwCw + Qrcr J Ihatlhe slream has a cross-seclional area 01' 2000 m2 and Ihe pollulant reaCls wilh firsl-
Co = (9.42) I order decay (k = 0.8 d-I).
Qw + Qr
or (because Qwcw = W) Solulloo: (a) Firsl, Eq. 9.42 can be used lo compute Ihe inilial concentralion,

C(J = 0.5 ~ ~06(~~>?) + 12 x 106(1) = 16.96 mg L I

-- Equalion 9.46 can also he used lo make Ihe same eSlimale, wilh Ihe resuh

C(J = 0.5 0.5 x 106(400)

x 1()6+ 12 x 1()6
= 16 m L I
g
(8) Polnt source
which represents a 5.7% error. This mighl be considered 100 I:lr¡;<.:an error and Ihe ap-
pro¡¡imalion deemed inadequale.

-- (b) The velocily in Ihe slream can be compuled as

U =R
Ac
= 12.5 X 106 = 6250 m d-I
(b) Nonpolnt 01'dlstrlbuted source ~nnn

RGURE 9.5 Using Eq. 9.9.

Poin! and nonpoin!sourceslar one-dimensionalsystems.
c = 16.96e-6~~x
 166 "ART11 Incomplelely Mixed Syslems LECfURE9 Distribuled SYSlems (Steady-State) 167

Subsliluling values 01'x gives Solution: The evalualion from Example 9.3 holds for the first 8 km and provides the
boundary condition for the next reach. For km 8 to km 16 the velocity must be recalcu-
o 1.6 3.2 4.8 6.4 8 laled because 01' Ihe increased area,
x (km)
16.96 13.82 11.26 9.18 7.48 6.09
c(mgL-')
J u = 12.5 x 1()6 = 4167 m d-I
3000
Thlls Ihe pollutant is decaying exponentially as il is being transported downslream. Equalion 9.48 can Ihen be used to calculate the concenlration in the second 8-km strelch,

e = 6.0ge -Mfx-ROOO) + ~.~ (1 - e -Mfx-IIOOO»)

9.2.2 Distributed Source Substituling values 01'x gives

A distributed source is one lhal enlers a waler body in a diffuse manner.t In the x (km) 8 9.6 11.2 12.8 14.4 16
conlexl of a channel, íi would enler along its lengifi.The simples! type is the uniform
distribuled source, which is constant spatially (Fig. 9.7). c(mgL-') 6.09 9.44 11.90 13.71 15.05 16.03
TI1\'dislrihuled source can he incorporaled inlo the steady-slate mass balance,
de Finally Eq. 9.9 can Ihen be used lo calculale Ihe concenlralion in Ihe third 8-km strelch,
O = - U- - ke + Sd (9.47)
dx
e = 16.03e-Mfx-'6.000)
If e = en al x = O.then Substituling values 01'x gives
-1-, Sd _1-x
e = eoe U' + k I - e u ( ) (9.48)
x(km)
c(mgL-')
16
16.03
17.6
11.79
19.2
8.67
20.8 22.6 24
6.38 4.69 3.45

S¿{M L-3T-I) ~he entire solution can be displayed graphically as

" I I 1111
Polnt Dlffuse

o x
-
FIGURE 9.7
Uniformdistribuled source. .-.
i
..;¡
OIJ
EXAMPLE 9.4. DlSTRIBUTED SO URCE INTO A PLUG-FLOW SYSTEM. e
'-'

Suppose Ihat for the same problem setting as in Example 9.3, a uniform diffuse source ~ 5
loao with no now conlribulion begins 8 km downstream from Ihe poi ni source. The dif-
fuse source has a value 01' 15 g m-3 d-' alld continues for an additional8 km. Thereafter
allloadings lerminate. Determine Ihe concentration for the stretch 24 km below Ihe point -4 o 4 8 12 16 20 24
sOllrce. Note that the area ¡ncreases to 3000 m2 at the 8-km poin!.
x (km)
FIGURE E9.4
'Note thatthe dislrihuted sources described hcre contribute mass but no now. In Lec. 22 we will model
distribotcd sources that contribute both mass and now.

r---M
168 PARTU Incomplelely Mixed Syslems LECTlJKE'¡ Dislribuled Syslems (Steady-Slatc) 169

9.3 APPLlCATION OF THE MFR MODEL TO ESTUARIES dCI dco

W + VAc¡(O) - EA-
,
ex
(O) - VAC2(O)+ EA- -(O)
dx
=O (9.53)
The mixed-ftow syslem is Ihe fundamenlal model for one-dimensional eSluaries. In
Ihis seclion I emphasize how Ihe mixed-ftow model can be applied lo analyze poinl
and dislribuled loadings. Substiluling Eqs. 9.51 and 9.52 yields

W + VAco - EAA1co - VAco + EAA2co = O (9.54)

9.3.1 Point Source which can be sol ved for
The folIowing analysis focuses on Ihe case where a poinl-source loading is injecled W I
inlOa channel having conslanl characlerislics. The impacI of dispersion on Ihe ini- (9.55)
lial concenlralion al a poinl source can be compuled by laking a mass balance al co = Q J I + 41/
Ihe injection poin!. If il is assumed Ihal complele mixing lakes place in Ihe laleral Therefore Ihe final solullon is
and vertical dimensions, Ihe silualion is as depicled in Fig. 9.8. For simplicily we
assume Ihal Ihe loading does nOIadd significanl ftow al Ihe mixing poinI (Ihal is, W U(
c= x:sO (9.56)
relalive lo Ihe ftow in Ihe eSluary). Problem 9.8 explores Ihe case where Ihe wasle Q JI + 41/ e2E 1+fi+41j)x
ftow signilicantly elevales Ihe eSluary ftow.
We use Ihe same approach as was used lo solve Eq. 9.22, and Ihe general solulion W U(
for Ihis case is C= x~O (9.57)
Q JI + 41/ e2f: 1- jf+:¡~)\
c = FeAlx + GeA2X (9.26)

where Ihe A's are NOleIhal when E approaches zero, Ihis modcl convcrges onlhc plug-Ilow mode!
(see Probo9.9). Also observe Ihal Eq. 9.55 is in Ihe formal of Eq. 1.8. Thal is, il
(9.25) relales a crilical concenlralion Colo Ihe loading W. Thus as illuslraled in Ihe folIow-
~~ = 2~ (1 ~ J1+4;1) ing example, il provides a means lo perform an assimilalive capacily cakulalion for
The conslanls of inlegralion can be evalualed by boundary condilions. For Ihe presenl an eSluary.
case Ihese are
EXAI\IPI.E 9.5. POINT SOURCE TO A MIXED-I<'LOW SYSTEM. A poinl
c=O @ x = -00 (9.49)
source is discharged inlo an estuary having the following characlcrislics:
c = O @x=oo (9.50)

Applying Ihese boundary condilions reduces Ihe number of unknowns lo one, Value Units

Dispcrsiou cocflidcl1l xo x tO" m2d I

Cl = coeA,x for x :S O (9.51) Flow 12X tO' m' di
for x > O Widlh 0.5 m
C2 = coeA2X (9.52) ti m
Dcpth
The mass balance depicled in Fig. 9.8 can now be used lo evaluale co. For Ihe
case where Ihe wasle load conlribules negligible ftow, Ihe mass balance is
The pollulanl selllcs (\' = I 111di) amI rCilclsviii firsl-ordcr kinctics (k = 0.2 di).
Whal mass loading could be inpul to Ihis syslem under steady-stale conditi,ms if Ihe
w allowable conccnlralion allhe omfall is 10 ppl11? Express your results in kg yr l. A~slIme
complele laleral and vertical mixing al the outfilll.

! Solulion: The velocily can be calculated,

t" T
t FIGURE9.8 u = 12X IOJ
I Mass balance lor a poinl source
0.5x8 = 3000md-1
I discharged inlo a mixed-Ilow
O x syslem. Equation 9.55 can be used to compule
 170 I'ART" Incomplelely Mixed Syslems I.ECTURE 9 Dislribuled Syslems (Steady-State) 171

~ \J;" J1+4;i + 1
S,I JL(I- JI+4'1 a JL I_
W = Qco 1+_Hr
(
4 k+ ~.;;

e = -k ( 2 J1+4;i
1 + 417 )(1 - e2E
) (
)e2E
JI+4'1"x-a¡
, x ~ a (9.60)
U2
As depicted in Fig. 9.9 the solution approaches the plug-flow case when E is
~ As E ¡ncreases.~nliQO ~\'enLilaffy!!P..m:()arn~~
a..¡;eIl~shaped.distrihution
4(0.2 + ~ )80 x 106 centered on Jhe.l1li!I.no¡nt of J.heAWribu~<.lJQilding.
= 12 x IOJ(IO) = 425.206gd-1
I + (3000)2

or converting lo the proper units.

PROBLEMS
. g d-I
= 425206 365 d ~ J = 155200
. kg yr-1
W
( yr 10 g ) 9.1. For the mixed-ftow model nOlice that c(O) ::; Cin.That is. the concentration immediately
inside the reactor will be less than or equal to the inftow concentration (Fig. E9.2). When
~
will they be equal? Use mathematical and physical arguments to establish why they are
different.
9.3.2 Distributed Source
9.2. Determine the Peclel numbers for the four cases represented in Example 9.2.
T~oma.!!n~nd Muelier tI9.t!.npr~~ent the foliowing solution for the case of a dis-
tributed IOl!dQfa first-order decaying substance intoan a~y~<;.tive/dispersivesystem 9.3. At the end of Box 9.1 we specified bounds for the Peclet number. Verify these bounds by
determining concentration profiles for the tank from Example 9.2 using Peclet numbers
(seeFig.9.9): -- of O.I and 10 (by changing E). Display your results graphically along with the CSTR
and PFR solutions.
J1+4;i- 1 1 _ e -#;(1+ J'+4'1)tI
e = Sd
k ( 2 J1+4;i )( )e~H'+ JI+4'1)x
9.4. Determine the steady-stale distribution of a pollutant (k = O.I d-I) for x = Oto 32 km
in the following system. Note that the concentration in the river immediately upstream
X :5 O (9.58)
of the distributed load is 5 mg L -l. How far downstream will the system return to the

- I - )1 + 417- 1 e2E
Sd JL(I+ JI+4'1)t.t-tl¡
level of 5 mg L -1 ?
c=
k [ 2 J1+4;i S4= 10 9 m-3d-1 W = 1000 kg d-1

- )1 + 417+ 1e2E
~(I- JI+4'1)X O:5x:5a (9.59) ~ -.illJ]]] !
2 J1+4;i ]

e
- Sd(M L-3T-') 12.2:[
Q¡

;;- 60
I

Sd1k .5.
-t!." 20
L
O 8 16
x (km) FIGURE P9.4

9.5. A point source is discharged into an estuary having the following characteristics:

Value Unlts
o a x !(},
Dispersion coefficienl crn2S-I
Flow 5 x IIJ4 rnJd-1
FIGURE 9.9 Widlh 100 m
Uniform distribtJted sottrce in.amiJ!~-f!9j'V .system. Seve~al diflerent levels Depth 2 m
.é.fdispersi().Qa[e..displaye<!along wilh the plug-flow solution.

'
172 PART11 Incompletely Mixed Systems

The pollutant settles (v = 0.11 m d -1). What mass loading could be input to this system LECTURE 10
under steady-state conditions if the allowable concentration at the outfall is 10 ppb? Ex- - - - - -- - -- - -
press your results in kg d-I . Assume complete lateral and vertical mixing at the outfall.

9.6. A point source is discharged into an estuary having the following characteristics:
S pistributed Systems
Value Units
1 (Time-Variable)
Dispersion coefficienl \()6 m2 d-I
Flow 5 x 10" m3d-J
Widlh 200 m
Deplh 2 m

The pollutant decays at arate ofO.2 d-I at T = 20°C with QIO = 1.7. The estuary has
an ambient temperature of 27.5°C.
(a) Determine the reaction rate for the ambient temperature.
(b) What mass loading could be input to this system under steady-state conditions if
the allowable concentration at the outfall is 20 ppb? Express your results in kg d-I .
Assume complete lateral and vertical mixing at the outfall.
9.7. A stream receives a point and a diffuse source as shown in Fig. P9.7:

;:(.Wro~ Q}1UIVJlIW: J COQtinpe Qur discussimJ of distrib"ted systems by ~W~ying

2S mg L-I d-I 10 kg d-I ~i@i~ tefJ1P?J1'J charac~ristjp~of pJug.noWI1l1d mixed-ftQW$yste~s,wi~fQ()UI!
-- HHH . .¡ e
. . ppfuejp~t!"1tapeous
~... \f¡¡eful!or
discbarge Í1\tC)11<)Qe"dime!1sion¡ij chllhnel.~uch'ltJodels
mode1i!1g~pills an~ for fracer studies in Sfream~!114estuarie¡¡.
A B r
¡" 1.0..........................--..------- 1nowdescribesomemodelsforsimulatingtime-variablechangesin distributedsys-
!
QI
0.5 tems. As in the previous lecture, the discussion is limited to one-dimensional systems
with constant characteristics and to the instantaneous discharge of a substance into
x (km)
such channels.
The resulting models are extremely usefui in environmental engineering. In par-
18 .-.0000 00.................
g ticular they can be used to simulate an accidental spill in a stream or an estuary. In
12
addition they can be used to analyze tracer studies, that is, for those cases where we
~"
10 25 J5
deliberately introduce a tracer (such as a dye) into a water body to estimate some
x (km) of its characteristics (for example velocity, dispersion, reaction rate).
FIGURE P9.7
10.1 PLUG FLOW
The boundary condition (at x = O) is e =
10 mg L -l. If the pollutant decays at arate
of 0.2 d-I. determine the steady-state concentrations at points A, B. and C. The time-variable mass balance for the plug-ftow system can be wrilklo trigo
lO.!)
9.8. Rederive Eqs. 9.56 and 9.57 for the case where the loading contributes a significant ftow
to the estuary. ac ac
-at = -u- ax - kc (10.1)
9.9. Rederive Eq. 9.57 so that it converges to the plug-ftow solution at low dispersion. Note
that one way to do this utilizes the following altemative formula for determining the As derivedin Box 10.1,if a spill causesa concentrationCoat t = x = O,thesolu-
roots of a quadratic (Chapra and Canale 1988): tionis
-e
x-- ~ (9.6\) c = coe-kl fort = x/U
b:!:. ..¡b2 - 4ac c=O (10.2)
otherwise

173
 LECTURE10 Distribuled Systems (Time-Variable) 175
174 PART11 Incomplelely Mixed Syslems

e de
dt
= iJc
iJt
+ iJx
ae dx
dt
(10.7)

where dx/dt is the slope of the curve AB in !he x-t plane. The \erm dcldt on !he left-
hand-side represents the rate of change as measured by a moving observer. The term
acliJt represents the change of t at a fixed position. The term (dxldt)(iJclax) represenls
the change due to the observer's moving into a region of possibly different e.
x FIGURE 10.1 Now suppose Ihat the curve is chosen so thal ils slope is equal lo V. That is,
Depiction 01 Ihe movemenl 01 dye in
space and time lor a plug-flow system. dx = V (10.8)
dt
The whole plugs show the movement 01
a conservative substance. The shaded In olher words suppose that the observer is moving at the same rate as the constant
portions show a substance that reacts velocity V. Ifthis is tme Eq. 10.7 becomes
with lirst-order kinetics.
de
dt
= iJe
iJt
+ ViJe
iJx
(10.9)

Further, the velocity establishes a direct relationship between time and space, Thus Ihe righl-hand side of Eq. 10.9 ts equal lo the left-hand side of Eq. 10.5. Conse-
x quently
(10.3)
t =Ü de = -ke (10.10)
dt
That is, the solution can also be written as This is a nice result. In essence we have converted !he original partial differential
k equation into a pair of ordinary difTerential equations (10.8 and 10.10). The fon.~er
e = coe-Vx for x = U t (10.4)
(10.8) represent.~a curve in the x-t plane !hat is called !he eharacteristic eurve. In the
c=O otherwise present case it is a straight line that has a slope of V. If we assume thal t = Oal x = O
(!hal is, Ihese are Ihe coordinales where Ihe observer starts to move downstream with
Ihe velocily), Eq. 10.8 can be integraled lo yield

BOX 10.1. Characterlstlcs of the Plug-Flow Equatlon

x = Vt (10.11)

The time-variable mass balance for the plug-ftow system can be written as
Thus Ihe velocily defines a linear relationship between space and time. That is, at time
t!he observer will have moved a distance Vt downstream.
_
iJc
iJt
+ V-
iJe
iJx
= -ke (10.5) Thl: concentration seen by the observer is Ihen oblained by integrating Eq. 10.10
along this curve. For examplc if e = eo at t = O,

Now suppose Ihat we wanl to find a solulion lo Ihis equalion along an arbilrary curve e= eoe-~' (10.12)
in Ihex-t plane(Fig.B10.1).The change of e, de, from poinls A lo B can be written as
Consequently, as depicted in Fig. 10.1, the solulion to the original partial difTerential
iJe iJc
de = -dt + -dx (10.6) equation is an exponential decay that occurs along the line defined by Eq. 10.11.
iJt iJx

Dividing Ihis equalion by dt gives

EXAMPLE 10.1. SPILL INTO A PLUG.FLOW SYSTEM. Fivekg of a conserva-
tive pollutant is spilled into a stream over a period of about 5 mino The stream has !he
/ following characteristics: ftow = 2 m3 sol and cross-sectional area = 10m2.Determine
B
!he concentration and the extent of the spill and how long it takes to reach a water intake
located6.48km downstream.
80lutloo: The spill concentration can be estimated as follows:

. = 5 kg 1000 g 1440 min

. -I
A WsplIl
5 mm k g d ( )= . 144 X 1M
v' g d

----
x FIGURE B10.1 Q = 2 m3 sol (86,~OO s) = 0.1728 X 106 m3 d-I

-..-
 LECruRE 10 Oistributed Systems (Time-Variable) 177
176 PAKI'11 Incompletely Mixed Systems

_ W,pill_ 1.44x IQ6 _ 8 33 -3 An altemative perspective exists for a moving observer who travels downstream
e- Q - 0.1728X 1()6- . gm at the same velocity as the spill. As depicted in Fig. 10.2 this observer would watch
the concentration decay exponentially as if the process were taking place in a batch
The extent of!he spill in this stretch is calculated as (note, U = 2/\0 = 0.2 m s-1) reactor.
The interrelationship between space and time also applies to steady-stale solu-
X,pill= 5 min(0.2 m S-I) 6~
mm
s
( )= 60 m pons. For example recall that the steady-state solution for a continuous load into a
plug-ftow system is
The front of !he spill will reach !he water imake at -.!x
6480m hr c = coe u (10.13)
t= u . (3600s )=9hr Now we should recognize that this equation could also be written as
from the start of the spill. Five minutes later it will have passed beyond the intake. c = coe-kl (10.14)
where I represents travel time below the point source. The inl mgeability of
The preceding example took the perspective of an obseiver at a fixed location space and time will recur in our subsequent analyses ofplug-ftow sY~It:msthroughout
the remainder of this book.
x downstream from the spill. This observer will see nothing until t = xlU. At this '
instant the spill will pass the observation point. For the case where the spill is de- Clearly the foregoing diSCUSsion is a highly idealized representation of stream

..
caying, the concentration will be reduced as specified by the exponential decay in transportoIn even IfuLmost
.
. highly.
. advective syste~p
.. . rii~p"rsi.Qllwilloccur.
~.1O.2. Before showing how the dispersíve prOcesscañbe- modeled, 1'11first introduce the
~ random walk model of the diffusion/dispersion process.

~ ~ ~ ~
........ ~ -.r-
Movlng
observer
10.2 RANDOM (OR "DRUNKARD'S") WALK
~ _ r The random or "drunkard's" walk is a term used to describe the typc of random
Flxed motionfound in many diffusion processes. The name "drunkard's walk" stems from
SplllO
..O.nd'.0 t observer the similarity between this type of motion and the random stumbling around that a
drunkard might exhibit.
Suppose that a population of particles is confined to motion along a one-
dimensional line (Fig. 10.3). Assume that each particle has an equal likelihood
l!J-l!J-l!J-l!J-l!J-l!J-
,=0 ,=t.I ,=Zt.I ,=3t.I '=4t.I ,=St.I
of moving a small distance dx to either the left or the right over a time interval dI.
Al I = O all particles are groupt:d at x = Oand are allowedto take one random
x=o x=Ut.I x=ZUt.I x=3Ut.I x=4Ut.I x=SUt.I step in either direction. After dI has elapsed, approximately one-half of the panicles
would have stepped to the right (dx) and the remainder would have stepped left
e (-dx). After another time interval (that is, after 2 dt had elasped) approximately
one-fourth would be at -2 dX, one-fourth at 2 dx, and one-half would have stepped
back to the origino
With additional time the particles would spread out (Fig. 10.3).Note "tat the
distribution of the population is not uniform but has a higher density at the (1. ¡ I il ¡¡nd
diminishes at the end. This is due to the fact that a particle would have tI! ,'..:ute
many successive moves in a single direction to reach the extremes. For example in
Figure 10.3, after 4 dI, a particle would have to execute 4 successive right steps to
reach 4 dx. Because there is aSO-50 chance of moving left or right in each time
interval, it is more likely that a particle would stumble around in the vicinity of the
originoThe net outcome is that the random walk of the individual particles results
x=Ut in a spreading bell-shaped distribution of the population. In addition note that this
F\GURE10.2 spreading tendency amounts to a general movement of panicles from high to low
concentrations.
Twa perspectives lar viewtng temporal and spatial changes In a plug-flow System.

" _......-
 178 PAKT11 Incompletely Mixed Systems U!C1tIRE10 Distribuled Systems ('Í'ime- Varilll:lre)t;§

W41 41 i
I,J

I I I I , I I I I I I t= 18;' [\ ').1
-56x -36X -6X O 6X 36x 56x ," ,1 ,.

& JI t=O FIGURE 10.4

]
.c
-3 -2 -1 o 2 3
Representation 01the random walk by
a normal or "bell-shaped" distribution.
() 1I Jl 1=61
.11
(M L -2). t Thus the distribution ofthe population of particles is described by a series
1ií
al of bell-shaped curves that spread out over time symmetrically about the origin (Fig.
"U n Il JI t=26t 10.4).
, I 'E It should be noted that Eqs. 10.15 to 10.17 apply only if the number of time
steps and spatial intervals are much greater than i. In other words the observation
'O 6 11 JI q t=36t
f & time t and the observation space x must be much greater than the magnitudes .DIthe
I +:1 duration t!.tand size t!.xof the individual steps. Because the random walk is an ánalog
111
()
-, 11 11 11 - t=44t of the diffusion process, these conditions will have relevance to the application of
diffusion models in the real world. That is, the process holds only when the random
FIGURE 10.3 motions, whether molecular or due to eddies, have smaller time and space scales
Graphlc representa ti on 01 a random walk. At time t = O, all than do the phenomena being modeled.
particles are grouped at the origin (x = O). During each time
step 4t. hall the particles at each location move left and hall
move right. The result is that over time the partlcles spread
out In a bell-shaped pattern. BOX 10.2. Mathematlcs o, the Random Walk

The random walk can be formulated mathematically by realizing that it can be rep-
... resented as a binomial distribution. The binomial or BemoulU's distribution results
, The random-walk process can be expressed mathematically as (see Box 10.2 for
details) from our tendency to place observations in one or the other of two mutually exclusive
categories. Ir p is the probability that an event will occur and q = 1 - P is the prob-
1 _ x2 ability that it will not, the probability that the event will happen exactly x times in n
p(x, t) = c;:;:e 4Di (10.15) trials is given by
2..¡ 1fDt

where p(x, t) = the probability that a particle will be at x arter an elapsed time t and
D = a diffusion coefficient, defined as
p(x, n) = {
\n -
n
x )pxqn-x

where the parenthetical operation represents the number of possible ways or combina-
t!.x2 tions by which the event can occur x times in n trials, where
D=- (10.16)
2t!.t ( n n!

If the populationis groupedat theoriginat timerero, thenumberof individuals

\n - x ) = x!(n - x)!
a:Jd n! = n(n - I)(n - 2)...1.
at position x at a subsequent time t would be proportional to the probabilit)j.ot an The randomwalk is this sor! of processbecausethe particle is limited to two
individualparticle's beingat x. Thus Eq. 10.15can be expressed in terms óf mass modesof motion:left or right. Thusthe binomial distributioncanbeusedto determine
and cOl1centration, the probability that, after n,time steps, the particle will be n. spatial intervals from th~
originoTo do this the total numberof stepswould be divided into n, to the right and
mp :. x2 (10.17)
c(x, t) = c;:;:e 4Di
2..¡ 1fDt

'Note Ihat mp is formally called aplane source since il enlers!he syslem al aplane suñace--the cross-
where m p total mass of the particles normalized to the cross-sectional area sectional area.
 ...

180 PART11 Incompletely Mixed Systems LECfURE10 Distributed Systems (Time-Variable) (SI

n, - n, to the left. To wind up at n., the difference between the right and left steps,
n, - (n, - n,), w()"ld have to equal n.. Therefore the number of right steps would have
to be n, -
= ('" I 11,)/2and left steps would have to be (n, n,) = (n, - n.)/2.The
probability that after n, time steps the particle would be n. spatial intervals from the
origin can be represented as (remember there is an equallikelihood of moving left or
right; that is, p = q = 0.5)
FIGURE 10.5

p(~..n,) = (n.
1 n,
; n,)G r~n. (~r;nx
n,
/ '-3.30(90%)-
'-3.90(95%)
5.20(99%)
A standardized normal distribution
showing the probability (expressed
as percent) encompassed by various
multiples 01 the standard deviation. For
or
() -3 -2 -1 o 2 3
example 3.9/1 encompasses
area under the curve.
95% 01 the

p(n.,n,)= 2 (n,~n.)!(nl-~n.)!
As the number of intervals becomes very large, the binomial distribution ap- This relationship is sometimes referred to as Fiek's seeond law. The solution for the
proaches the normal distribution. If we define continuous variables x = n. Ax and case where the substance is initially concentrated at x = Ois
t = n, At, the probability that the particle would be at distance x at time t can be
formulated as a normal distribution with a mean of zero and a variance of t(Ax)2/At, m _xl
v
as in (see Pielou 1969) e(x.t) = 2 vbe
1TEt
4Ei (10.19)
1 _L
p(x. t) = 2 vc;:::
1rDt
e 4D, ~ which is identical to the solution based on the random walk (Eq. 10.17). Thus the
! solution is a bell-shaped curve with mean at zero and a variance of 2,Et.
where p(x, t) = the probability that a particle wi1l be at x after an elapsed time t ~nd Because the variance is a measure of spread, this equation can be used as the
D = a diffusion coefficient, defined in the limit (as Ax and At become small) as
basis of a simple engineering computation to assess the effect of dispersion on a
AX2 pollutant spill. For example if a conservative substance were discharged in a lump
D=-
2At sum to a water body, its tendency to spread outward from its center of mass could be
represented by the standard deviation

u = J2Ei (10.20)
10.3 SPILL MODELS or multiples of the standard deviation as depicted in Fig. 10.5. For example 95% and
Now fhatwe have reviewedthe fundamentals of the random-walkmodel,we can 99% spreads can be roughly approximated by 4u and Su, respectively.

\ integrate dispersioninto modelsof instantaneousdischargesinto uniform,one- EXAMPLE 10.2. CONSERVATIVE SPILL IN A CHANNEL WITH NO I~LOW.
) dimensionalchannels.First,we investigatethecase wherethe spilloccursinstanta- A barge releases a large quantity of a highly persistent contaminant in the center of a
( neously. Then we present a model for a continuous input. canal that is not ftowing. If the dispersion coefficient is approximately lOSm2 d-I, how
far wi1l the contaminant spread in 1 d? In 2 d? Assume that a 95% band adequately
approximates the extent of the spi1l.
10.3.1 Instantaneous or "Impulse" Spllls
Solution: Using Eq. 10.20 and the multiple of the standard deviation that encompasses
for ~~iscase a significant input occurs over a very short time period at a point in 95% of the distribution (Fig. 10.5) yields
space. Many spills can be approximated in this fashion. In the following paragraphs x(l d) = 3.9j2(1OS)1 = 1744m
we build [he solution by adding mechanisms on a term-by-term basis. First, we ex-
.
amine the situation where dispersion or turbulent diffusion is the only mechanism. and x(2 d) = 3.9 j2(1OS)2 = 2466 m

(P'jffusionldispersio". As developed in Lec. 9, a mass balance for a substance

that disperses in a one-dimensional channel can be written as (Eq. 9.21 with k = Dispersionladvection. Now we can add advection to the model. This relation-
U = O) . ship is sometimes referred to as the adveetion-diffusion (or advection-dispersion)
equation
de = Ed2e ~ de _uae + Ea2e
dI dX2 (I~18) dX dX2 (10.21)
\ dt
 182 PART11 Incompletely Mixed Systems LECruRE 10 Distribuled Systems (Time-Variable) 183
/ .M: e, same, !he peak progressively decays due lo dispersion. In addition !he spill spreads as
e
po ) timeprogresses.Theseresultscan be summarizedas

. 1=3hr t=6hr \1=9h~)

. ,
Extentofspill(m)' 181 25~ 314
x Peakconcentration(mgL-I) 4.29 (3.03) 2.48 ../
'Defined 3.. distance encompassing 95% or!he mass.

FIGURE 10.6
Gfupers~on/adv;tiol!!dec~ Finally a first-order reaction can be added to the
The movemenl 01 conservative dye in model,
space and lime lor a mixed-flow syslem. I ac k a2~
f *'" I~ -at = -u-ax + E- - kc
ax2 _.-1
(10.23)
The solution for the case where the substance is initially concentrated at x = Ois
The solution for the ca~e where the substance is initially concentrated at x = Ois
m _(x-Ut)2 AI)'z::..
c(x, t) = ---1!-.-e 4EI ( 10.22)
2 j-TTEt c(x,t) = 2..¡mbe -
(x:¡Ut)2
EI-kl (10.24)

-
Note Ihat in comparison with Eq. 10.19, ~ effect o~~e~ion
{ dispersion -
solution intact downstream at velocity U (Fig. 10.6).

EXAMPLE 10.3. SPILL INTO A MIXED-FLOW SYSTEM. Evaluate !he spill

from Example 10.1. but ¡nelude the effect of dispersion. Assume !hat a dispersion co-
is lo "move" the
7rEt
In comparison with Eq. 10.22, !he effect of decay reduces to the area under !he bell-
shaped curve as it moves downstream (Fig. 10.7).

Fixed versus global observer. Note that Eq. 10.24is a function of two indepen-
efficient of 0.1 m2 S-I holds for the entire stretch. Also assume that the spill occurs dent variables x and l. Thus it can be viewed from two perspectives. As in Example
instantaneously as aplane source. 10.3 we can compute the spatial distribution at a fixed time. Thus we can develop
Solullon: The spill can be expressed as aplane source, as in the giobal perspective manifested by !he bell-shaped curves shown in Fig. 10.7.
~} 5 X 103 g '-2 Conversely we can compule the temporal distribution at a fixed point in space.
IIIp= 10m2 =5OOgm This latter case relates to a static observer. For this case the view can be skewed
because !he bell-shaped curve continues to spread out as it is being observed (Fig.
The spill concentration in the stretch can then be estimated by 10.8).
500 (x-o.2I)2
c(x. 1) = e - 4(0.1)/
2 ¡?r(0.1 )1
Results can be calculated and are displayed in Fig. EI0.3. These results are 10.3.2 Contlnuous Spllls
very different from Example 10.1. Although the travel time to the water intake is the
Forsome t.!.aceEstu~ies, as well as ~olTle_s"piIIs,the inputjumps to a constant leve!.
I As depicted in Fig. ,10.9, two ídéalized cases can be modeled. First, !he concentration

1=0
t=3 ./'
-
~?
4 4.L 1=0
1=6
~ 1=9 "'a¡
1=3
'" '" a¡ 2 2
,...
.s .s 1=6
FIGURE10.7
1=9 The effecl 01decay on Ihe spill
mode!. The Ihin Une~are Ihe
I ' o O. same as calculaled in Example
~I O 2 4 6 8 01 246 8 10.3. The heavy Unes are tor
Ihe same case bul with k =
x (km) FIGURE E10.3 x (km) 2d '.
 ----

lH4 ~Akl" lncompletely Mixed Systems lECfURE 10 Distribuled Syslems (Time-Variable) 185

150
1=1
x ..-. ----------------
100
~ o 5
I I
10
I I I I
15
~~
k=O

1=2 '-'
El
50
~
x k = 0.0005 S-I
I I I I I I I
-1 o 5 10 15

1=3 -4 o 4 8 12
x
-1 t (hr)
~ o 5 10 15
FIGURE 10.10
1=4 A simulation 01 a "breakthrough" curva. A step increase to a
x concentration 01 100 mg L -, is initiatad at x = O at t = O. Shown
.., I I I I I
is the change in concentration at a sampling point 2000 m downstream

" o t 5
10 15
lor a conservativa and a nonconservative raleasa. This exampla used
U = 0.1 m s., and E = 5 m2 .s.'.
Ir

t x=3 AGURE 10.8

.where I
di" -,
r f Although tha distribution 01a pollutant spill r = Jl + 417 ( 10.26)
is ball-shapad in spaca. a lixed observer
would .see" a skewed shape in time kE
because the bell-shaped curve continues and (10.27)
-s o 5 10 15 to spread out as it is being observed. 17 = U2

The error function complement, erfc, is equal lo one minus the error funclion:
increase can be m~intaine<ffor an intinite duration (Fig 1O.9a).For this case the 1 - erf. Also, erf( - x)
= -erf(x). The error funclion is simply the evaluation of the
following detinile integral,
"
solution of Eq. 10.23 with constant coefficients can be expressed as (O'Loughlin
and I!owmer 1975)
\ erf(b) = ~ J: e-132d{3 (10.28)

~ c(x, t) = ~ [e~(l-r¡erfc (x; %r)+ e~:(l+r¡erfc(x2+)~r)] (10.25) where {3 = a dummy variable. We have inc\uded selected values for the error func-
(
tion in App. G. 1t should also be noted that the error function is available on stan-
e dard software libraries [e.g., the International Math and Statistics Library (IMSL),
Numerical Recipes (Press et al. 1992, etc.), and as a function on many software
packages (e.g., Excel, Mathematica, etc.)].
A simulation using Eq. 10.25 is shown in Fig. 10.10. Such curves are referred
p to as "breakthrough" curves. They are used extensively in both surface and ground-
~ (a) i water
. problem contexts.
The second idealized application applies to cases in which the step input termi-
nates after a tinite time (Fig. 1O.9b).For this situation the solution is presented in two
FIGURE 10.9
~'Q
!'lo .\.
t,¡j ., . ponUnuous inputs are characterized in two
parts. For t < T, Eq. 10.25 holds. Thereafter the following formula applie\(Runkel
1~96):
~b) ways: (!') inlirite and (b) lin~e durations.
 ."186 PART11 Incompletely Mixed Systems LECTURa10 Distributed Systems (Time-Variable) 187
'¡-<

x =100 x = 2000m e
150

,-.,
~t)I)
'-'= 50 k =0.0005 S-I 1
~

RGURE 10.12
Concentration data sampled
\
/'
\
\
-4 o 't 4

t (hr)
8 12

to ti tI t3 t~ t
at a point In space to
characterize the dlstribution01
a tracer.
FIGURE 10.11
A simulation 01 a dye release or spill 01 linite duration-a Mass
-- "breakthrough" curve. A step increase to a concentration 01 1M =-Qc(tn - ¡). (10.31)
100 mg L-1 is initiated at x = O at t = O and lasts lor T = 2
hr. Shown is the distribution at x = O (dashed line) along with
\¡
curves at x = 100 and 2000 m lor a conservative (bold line) Travel Time
and a nonconservative release (light line). This example used n-I
f
I U = 0.1 m s-t and E = 5 m2 S-l. ¿(C;t; + Ci+lti+I)(ti+1 - ti)
011, t= ;=0 " (1Ó.32)
n-I
I
co uX(I-n -" x-utr -" X-U(t-T)r ¿(c; + Ci+t>(ti+1 - ti)
C( x, t ) = - e2E ellc - ellC ;=0
2l [ ( 2 jii ) ( 2 JE(t - T) )~
uX(I+n -" X + Utr -" X + U(t - T)r 10 29
temporal Varlan.ce
+ e2E [ ( 2 vEt )
ellC r;::: - ( 2 JE(t - T) )~}
ellC ( . ) n-I
A simulation using Eq. 10.25 is shown in Fig. 10.11. NQtice how !I)~ SQI\.ltiolL ¿JC¡t; + Ci+lt;+I)(t;+1 - ti)
s7
1
= ;=0
n-I
- (ríl
11 (10.33)
approaches a bell-~haped cur:.e as it moves downstre~m..;..-
¿(C;
;=0
+ C;+t>(ti+1- ti)
.
~
10.4 TRACER STUDIES
If data are available from two stations, located at XI and X2,the mean velocity
Aside from accidental spills, the models determined above have utility when com- can be estimated by
pounds are deliberatelYdischarged a s in a trace r Study. In such_casesthe distribution X2 - XI
d~nSlream from the injection point can be used to determine key characteristics
.
al( .. U = ~
t2 - tt
(10.34)
t such as the velocity,. the dispersion coefficient, and the decay rateo
.
.

To do this It isnecessary to estímate some quantities from cQncentratio~'ta.

.
The velocity estimate can, in tum, be use<Í to calctilate the dispersion coefficient by
I Such data is usually measured at discrete points in time (Fig. 10.12). . (Fis~~er 1968)
I For such cases the following formulas can be used: U2 s2 2)
'( E = ( 12 - S,t (J 0.35)
-+-- 2(t2-tt>
Mean Concentration n-I
2)c; + C;+I)(I;+I - ti) EXAMPLE 10.4. EVALUATION OF A TRACER STUDY. A tracer study is con-
_c= ;=0 ( 10.30) ducted in a stream with a flow of3 x lOSm3 d-I and a width of 45 m. At t = 0,5 kg of
2(tn - tO) !kuJ v
 -.----

188 PART11 Incompletely Mixed Systems LECfURE10 Dislribuled Systems (Time-Variable) 189

a conservative substance, lithium, is instantaneously injected al x = o. Concentrations where the travel times and the masses are calculated according to Eqs. 10.32 and
are measured at Iwo downstream slations:
10.31, respectively. We elaborate more on the use of tracers to estimale reaction
x = 1km rates when we discuss the measurement of reaeration laler in the lext (Lec. 20).

, (min) 30 40 50 60 70 80 90 100 110 120

O 100 580 840 560 230 70 15 3 O
10.5 ESTUARV NUMBER
Lilhium (,..g L-I)

TI¡e relative importance of advection and dispersion can be assessed by a dimen-

x = 8 km sionless analysis of Eq. 10.23:
ac ac a2c
, (min) 370 400 430 460 490 520 550 580 610 -at = - U-ax + E - - kc ( 10.23)
ax2
Lilhium (,..g L -1) O 10 80 250 280 140 35 5 O
Three dimensionless parameter groups can be defined:
Determine (a) the velocity (m d-I) and (b) the dispersion coefficient (cm2 S-I). c =
c . (10.37)
Solutlon: (o) The velocity can be evaluated by determining Ihe travel time between co
the two sampling points, kx
x'
- - 2.982,600 _62 2 . U (10.38)
ti - 47,960 - . mm
t' = kt (10.39)
- - 23.133,000 _ 4819 .
t8 - 48,000 - . mm These equations can be solved for c, x, and t, respectively, and the results substilUted
into Eq. 10.23 to yield
Therefore the velocity can be computed as

_ 8 - 1km 1000m _ . _1 1440min _ _1 ac. a2c' ac' .

(10.40)
U - 481.9_ 62.2min ki1I - 16.67mmm( d )
- 24,014md ( ) where11is calledthe estuarynumber,
at' = 11ax.2 - ax' - c
(b) The dispersion coefficient can be estimaled by calculating the lime variances of Ihe
lithium concentration distributions, kE
11= U2 (10.41 )
2 1.92 x 108 2 . 2
S'I = 47,960 - 62.2 = 137 mm
From Eq. 10.41 it should be cIear that Ihe magnitude of 11determines whclher
2 1.12xlOIO 2 .2 the second derivative lerm in the differential equalion is significant relalive lo the
S'2 = ._ ___ - 481.9 = 1043mm
other terms. Thus the following guidelines can be developed
Then Eq. 10.35 can be used lo compute
Suggested ranges
E = (16.67 m min-I)2(1043 min2 - 137 min2) 1/»1 Diffusion predominales
2(481.9 min - 62.2 min) 1/>10
1/""1 Adveclionldiffusion important 0.1 < 1/ < 10
-_ 300m2 mm
. -1 -- 10" l:m2 min _- 50,019 cm2-1 1/«1
s Advection predominates 1/ < 0.1
m2 60 s
( )

EXAMPLE 10.5. ESTUARY NUMBER. Evaluale the eSluary number for the slream
Dye studies can also be used to determine first-order reaction rates. For this case from Example 10.4 for a nonconservative tracer with a half-life of 1 d.
the mass under the concentration-time curve is determined at two positions. The rate Soludon: The parameters must be expressed in common units,
can then be estimated br

k= ~lnM1 (lO.36) E = 50,000 cm2 S-I (IO,~ cm2 86.~00 s) = 432~000

t2 - t, M2 m2 d-I
U = 24.000 m d-I
190 PART11 Incompletely Mixed Systems
LECI1JRE10 Distributed Systems ('lime-Variable) 191

0.693 = 0.693 d-I e

k= Id
.,./
The estuary number can be computed as
s--
= 0.693(432.000) = 5 2 x 10-4
11 (24.000)2
which clearly indicates a highly advective situation.
.
I -A
Channel FIGURE P10.3

PROBLEMS (a) Determine Ihe concentration time series at poinls A, B, and C. Plot your results and
express the concentration in ppb.
10.1. A one-dimensional estuary has constant dimensions and flow:
(b) How much algicide can be applied if the concentration in !he river is never to
exceed 10 ppb?
Width = 1000 ft
Depth = 10 ft 10.4. Al ( = O. 10g of a conservative algicide is introduced into a trout halchery pond having
Flow = 500 cfs
V = 2500 m3. Flow in the channel and the river is plug flow and !he hatchery pond is
Dispersion coefficient = Ixl 06 m2 d-I complelely mixed. The dimensions of the channel are B = 2 m, H = 1 m, and L =
0.5 km. Flow in the river upstream oflhe diversion is 20.000 m3 <1-1, and 2000 m3 d-I
(a) Calculate the estuary number for a herbicide that is subject to first-order decay
is diverted Ihrough the hatchery. 80lh the river and the pond flow inlo a weU-mixed
(0.05 d-I) and volatilization (0.3 m d-I).
lake having V = 10.000 m3. Determine the time when the maximum concentradon
(b) Suppose that 10 kg ofthe herbicide is spilled at mile point 2. Plot the concentration occurs in the lake.
distribution as a function of time at a water intake at mile point 10.

10.2. A tracer study is conducted in a stream with a flow of 3.7 x lOS m3 d-I and a width
of 60 m. At ( = O. 50 kg of lithium is instantaneously
concentration is measured at two downstream stations:
injected at x = O. Lithium -
x = 1km

(min) 60 80 100 120 140 160 180 200 220 240 260

Lithium(/LgL 1) o 2 24 78 108 89 52 23 9 3 o
Channel
FIGURE P10.4

x = 5 km 10.5. Suppose thal a contaminant is injecled into a sediment at a depth of 10 cm. Assuming
thal the diffusion coefficient is 10-6 cm2 S-I,
(mi") 550 600 650 700 750 800 850 900 950 (a) How long will it take the contaminant to reach the sediment water interface? Define
the front as being the point that encompasses 99% of the contaminan!.
Lithium (/Lg L -1) O 7 26 47 43 23 8 2 O (b) Determine the flux inlo the overlying water al this time.

Determine Water
(a) velocity. m d-I
/
t
10cm
(h) depth. m
(e) dispersion coefficient. cm2 S-I ~

10.3. At I = O. 10 g of a conservative algicide is introduced into a trout hatchery pond hav-

!J~!"t~:"
ing V = 2500 m3. Flow in the channel and the river is plug flow and the hatchery \

pond is completely mixed. The dimensions of the channel are B = 2 m. H = 1 m, Sedlh1ént8

and L = 0.5 km. Flow in Ihe river upstream of the diversion is 20,000 m3 d-I. and
2(}(K)m3 d 1 is divertcd through the hatchery.
FIGURE P10.5
 ~ ~ , ~

LECTURE 11 Control- Volume Approach: Sleady-State SolUliuns 193

LECTURE 11 ¡
-. ,~~ ~ n __ ~. ~~, ,~"-' '~'~"'--

Control-Volume Approach:
Steady-State Solutions
I
I /
L-l- J
I
l.
I FIGURE 11.1
I Mass balance around a control volume.
I

where double-subscripted terms refer to interfaces between volumes. For example

the term Q¡-J.j refers to the flow between volumes i - I and i.
Note that a problem with Eq. 11.1 relates to the terms, Ci-I,i and c¡.¡+I-that
is, the concentrations at the interfaces between the segment and its upstream and
downstream neighbors. Although it makes physical sense, inc1usion of these con-
centrations introduces two additional unknowns to the equation. To eliminate these
unknowns we must reexpress these interface concentrations in terms of the model
Sunknowns. One approach for doing this is referred to as a backward or upstream
1
difference. Such terminology relates to the fact that the interface concentration is
talcen to be approximated by the concentration of the "upstream" segment. Using

'\ "'"0=.pprooch.w,
W¡ + Q¡_I,¡C¡_I- Q¡,¡+IC¡+ E:-U(Ci-1 - Ci)+ E¡,¡+I(C¡+I- Ci) -
,," oxpre" Eq. 11.1"
k¡V¡c¡
(11.2)
Thus the model is now totally formulated in terms of the unknown concentrations.
In the previous lectures you have learned how ca1culus can be used to obtain ana- Collecting terms yields
Iytical solutions for distributed systems. Now 1 describe a method for modeling such
-(Q¡-I,¡ + E¡-u)c¡-I + (Q¡.i+1+ E:-u + E:' I t k¡V¡)c¡ - (Ef ,ki+1 = W¡
systems with the computer. 1 talce a straightforward approach by developing some (11.3)
simple computations to illustrate the method. Then 1 generalize the approach and
explore its limitations.
or ai,¡-IC¡-I + a¡.ic¡ + a¡.i+lc¡+1 = W¡ ( 11.4)

where
11.1 CONTROL-VOLUME APPROACH
a¡.¡_1 = -Qi-I.i - E:-I.i (11.5)
As depicted in Fig. IU the approach is based on dividing the water body into finite
segments or "control volumes." The segments O and n +I represent boundary seg- ai,¡ = Q¡,i+1 + E:_I.¡ + E:,i+1 + k¡V¡ (11.6)
ptents. Therefore there are n unknowns that need to be determined: CI, C2, Cn.
Consequently n simultaneous equations must be developed to obtain a solution. a¡,¡+1 = -E¡,i+1 (11.7)
To do this a steady-state mass balance can be written for volume ..s
If Eq. IIA is written for volumes I through n in Fig. 11.1,there are n equations
0= W¡ t Q¡_I,¡C¡_I,¡ - Q¡,i+IC¡,i+1 + E¡_J.j(c¡-1 - c¡) + E¡,i+I(Ci+1 - c¡) - k¡V¡c¡ with n + 2 unknowns (co through Cn+I)'Therefore boundary conditions are needed
(11.1) to eliminate two of these unknowns.

192
194 PART11 Incompletely Mixed Syslems LEcrURE 11 Control- Volume Approach: Steady-State Solutions 195

11.2 BOUNDARY CONDITIONS Notice how we have merely omitted the dispersion tenn foe the entry interface. When
we collect tenns the final equation looks like
There are two fundamental types of boundary conditions used in water-quality
modeling: WI = al.ICI + al,2C2 (11.13)
where
. Dirichlet boundary conditions specify the concentration at the boundary.
. Neumann boundary conditions specify the derivative of concentration at the WI +- WI - al,oco (11.14)
boundary. al,o = -QO.I (11.15)
In addition there are occasions when a combination of these might be appropriate. a1.l = QI.2 + Ei.2+ kl VI (11.16)
In this seclion, because they are the most commonly used, 1describe how Dirichlet al.2 = -Ei.2 (11.17)
conditions are treated. Some examples of Neumann conditions wil\ be explored in Thus the loading term is modified to inelude the inflow concentration, and the a's
the end-of-chapter problems and in later lectures. are adjusted to omit dispersion across the inflow interface.
Two types of Oirichlet conditions are il\ustrated by Fig 11.2. In Fig. 11.20 the A similar derivation at the outlet yields
boundaries are open, as they would be in a channel. For such cases Eq. 11.4 can
be applied directly. For example for volume 1, the equation is Wn = an.n-ICn-1 + an.nCn (11.18)
where
WI = a1.lcI+ al.2c2 ( 11.8)
an,n-! = -(Qn-I,n + E~-I.n) (11.19)
where the loading is modified to reflect the fact that it ineludes the effect of inflow
from volume O, an.n = Qn.n+! + E~_I,n + kn Vn (11.20)

WI +- WI - al,oco (11.9) Notice that because of the lack of dispersion, the loading term is not affected by
Cn+l. Similarly the diagonal term an,1I does not inelude the term E~,n+I because no
Similarly the balance for volume n reflects the fact that Cn+1is a known, dispersion occurs across the outlet interface.

Wn = an.n-ICn-1 + an.nCn ( 11.10)

where the loading is modified to inelude the effect of volume n + 1, 11.3 STEADY-STATE SOLUTION
WII +- Wn - an.n+ICn+1 (11.11)
The complete set of equations to be solved is now specified as
For both cases the coefficients (the a's) are as defined in Eqs. 11.5 to 11.7. a1.lcI + al,2c2 = WI (11.21)
When mass enters and lea ves the tank via plug flow, the situation is more inter-
esting. For the case of the inflow volume (i = 1). the original mass balance becomes
a2,1 CI + a2,2C2 + a2,3c3 = W2 ( 11.22)
a3,2C2 + a3.3C3
+ a3,4c4= W3 (11.23)
O = WI + QO.ICO - QUCI + E;.2(C2 - ct> - kl VICI (11.12)

--;.. ~ an-I,n-2Cn-2 + an-l.n-ICn-1 + an-I,nCn = Wn-I (11.24)

I _
= Wn
~

an,n-ICn-1 + an,nCn (1l.25)

(s) Open boundarles

which can also be expressed in matrix fonn as
[A]{c} = {W} (11.26)
/ where
1
~
--, ~ .-
1....-
all al2 O O
a21 a22 a23 O O
(b) Pipe boundaries O a32 a33 a)4 O O
[A] = I.
FIGURE 11.2
Two types 01 Dirichlet (Iixed concentralion) boundary conditions lor one- O O
I (11.27)

an-I,n-2 an-I,n-I an-I.n

dimensional systems. (a) Open advective/dispersive; (b) pipe advective O O an,lI-1 an,n
boundaries.
 --- --.--.-

196 PART11 Incomplelely Mixed Syslems

I.ECTURE11 Conlrol-Volllmc Approach: Sleady-Slale Sollllions 197

Cl Wl e 1
C2 W2 0.8
C3 W3 0.6
{c} = {W} = (11.28)
0.4
Cn-I Wn-I 0.2
Cn Wn o
o 20 40 60 80 100
Note that, as described previously in Lec. 6, a variety of methods are available x FIGURE E11.1-2
Cor solving such equations. An application is provided by the foIlowing example.

EXAMPLE tU. APPLYING THE CONTROL-VOLUME APPROACH TO A
MIXED.FLOW REACTOR. Recalllhat in Example 5.2 we used a cascade model
to simulate the steady-state distribution of concentration in an elongated tank. In this ex-
ample use the saf!le syslem but add dispersive Iranspon to the framework. as indicaled
in Fig. Ell.1-1. .
FIGURE E11.1-1
The tank has cross-sectional area A,. = 10 m2, length L = 100 m, velocity U =
100 m hr-I, first-order reaction rate k = 2 hr-I, and E = 2000 m2hr-I. The inftow
J. 1 differences.
concentration is 1 mg L-l. Use n = 5 and display the concentrations along with the
analytical solution computed in Example 9.2.
Solutloo: The following matrix representation can be developed for the syslem:
The numerical results generally follow Ihe Irend oflhe exact solution. If a liner seg-
mentalion were imposed (Ihal is. a smaller .:1x). the numerical oulcome would approach
the analytical case.
Despite the fact that the results ofthe foregoing example seem acceplable, dosel'
inspection ofFig. EIl.J-2 indicates that the discrepaneies seem systemalie. That is,
at the head end of the tank the numerical results are lower than the analytieal solu-
tion. At the tail end the numerical results are higher. In addition the discrepancy is
most pronounced where the gradient is steepest. at the head of the tank. Sueh errors
might oecur if we had used too high a dispersion coefficient for the numerieal cal-
eulation. In faet, in Seco 11.6, I wiIl show that there is a "numerieal" dispersion that
aecompanies computations of the sort il\ustrated here, that is. those that use baekward
11.4 SYSTEM RESPONSE MATRIX

The solution for Example 11.1 can be obtained in a variety of ways. For example both
2400 -1000 O O O CI 1(00
iterative amI elimination methods are available for aceomplishing this task (App. E).
- 2000 3400 - 1000 O O C2 O In addition the matrix inverse could be used to genera te the solution
O - 2000 3400 -1000 O C3 = O
O O - 2000 3400 -1000 {c} = [A¡-I{W} (11.29)
C4 O
O O O - 2000 2400 {Cs) { O ) Although this is not the most efficient means for obtaining a solution. aside from
providing the solution veetor, the matrix inverse allows the computation of the effeet
which can be solved for the concentrations. These results. along with Ihe calculations of loading to a particular segment on any other segment. As previously deseribed in
from Example 9.2, are summarized in th~ following table and figure: Seco 6.3, the element of the matrix inverse a¡) represents the change in eoncentration
of segment i due to a unit loading change to segment j. This property is explored in
Dlstance Analytical Numerical further detail in the following example.

O 0.76563
EXAMPLE 1l.2. MATRIX INVERSE ANAI.YSIS OF A MIXED-FLOW REAC-
10 0.65700 0.60919
30 0.48401 0.46206 TOR. Use the matrix inverse lo evaluale Ihe following loading scenarios for the mixed-
ftow reactor from Example 11.1:
50 0.35753 0.3526I
70 0.26760 0.27476 (a) Determine how much the inftow concenlralion must be reduced to reduce the outftow
90 0.21319 0.22897 concentration to O. 1 mg L -l.
100 0.20441 (b) Calculate the concentration distribulion if 2000 g hr-I is injected in the middle of
!he reactor and 1000 g hr-I is injecled at!he end. Set the inftow concentration to zero.
 198 PART11 Incompletely Mixed Systems
lECrURE 11 Control- Volume Approach: Steady-State Solutions 199

Solutlon: (a) The matrix ¡nverse for !he system can be calculated as
interface concentration as the average concentration of the two adjacent segments.
0.0006092 0.0002310 0.0000882 0.0000343 0.0000143 For equal-sized volumes, the resulting balance is
0.0004621 0.0005545 0.0002116 0.0000824 0.0000343
Q. .
C¡_I+C¡
0.0003526 0.0004231 0.0005430 0.0002116 0.0000882
0.0002748 0.0003297 0.0004231 0.0005545 0.0002310
0.0002290 0.0002748 0.0003526 0.0004621 0.0006092
O=w
I + ,-1., ( 2 )_ Q ..
1.1+I (
C¡+Ci+I
2 )
+ E/_1.¡(c¡_1 - c¡) + E!.i+1(Ci+I - c¡) - k¡V¡c¡ ( 11.30)
The load reduction needed to lower the oudet concentration to O. 1 mg L -1 can be calcula-
Collecling lerms yields
ted as

Cs = a5"IIWI = a5".'Qco O¡.¡_IC¡_I + oUC¡ + o¡.¡+ICi+1 = W¡ ( 11.31)

which can be solved for where

Co =~ = 0.1 gm-J = 04367 m-3 Q¡-U '

a5"11
Q 0.0002290[g m-3(g hr-I )-II( I000 m3 hr-I) . g
O¡.¡_I = -~ - E¡_I.¡ (11.32)
Thcrefore the inflow concentration must be reduced,
ni . 1 - 0.4367 1 % 56 3 o¡.¡ = E/_u + E!.i+I + k¡V¡ (11.33)
I d
I "/0re uctlOn = 1 (00) = . %
I ,!
(b) The appropriate elements can be used to evaluate the effect of the two loadings on
O¡,/+I = -E!.1.'+ 1 + ~
Q¡.i+1
(11.34)
I the entire reactor, as in

I CI = 2000(0.0000882)+ 1000(0.0000143) = 0.1764+0.0143 = 0.1907mgL-1 E~AMPLE

but 11.3. CENTERED.DlFFERENCE
now use centered differences. MODEL. Repeat Example 11.1,
C2 = 2000(0.0002116) + 1000(0.0000343) = 0.4232 + 0.0343 = 0.4575 mg L-1 "-
CJ = 2000(0.0005430) + 1000(0.0000882) = 1.0860 + 0.0882 = 1.1742 mg L -1 Solution: U;¡ln~ centered differences we can develop the following matrix repre-
I
I C4 = 2000(0.0004231) + 1000(0.0002310) = 0.8462+ 0.2310 = 1.0772mg L-1 sentation for the system. Note that the inflow and outflow segments must be treated
Cs = 2000(0.0003526) + 1000(0.0006092) = 0.7052 + 0.6092 = 1.3144 mg L-1 differently from what was done in Eqs. 11.13and 11.18 lo account for the cenlered
differences (see Prob. 11.1),
I These results are displayed below:
1900 -500 O O O CI IOOO
-1500 2400 -500 O O C2 O
e 1.6 2000 t 1000 t O
O - 1500
O
2400
-1500
-500 O
2400 -500 C4
C3 = O
O
[ O O O -1500 1900J{ Cs}
1.2 { O }
which can be solved for the concentrations. These results, along with the calculations
0.8 . from Example 9.2, are summarized in the following table and in Fig. EII.3:
0.4 ....... j{'1000 Distance
o Analytical Numerlcal
o 20 40 60 80 100 O 0.76563
x FIGUREE11.2 10 0.65700 0.65338
.../ 30 0.48401 0.48284
50 0.35753 0.35748
70 0.26760 0.26741
90 0.21319 0.21111
100 0.20441
11.5 CENTERED-DIFFERENCE APPROACH

Note that the backward difference presented in the previous sections is but one way
The numerical results follow the trend of the exact solution much better than in Exam-
to represent the advective term in finite formo Another altemative is to specify the pie 11.1.

~ ...
 -. .-----

l.ECTURE11 Control-Volume Approach: Steady-State Solutions 201

200 PART11 Incompletely Mixed Systems

C 1 tu; _1 :. tu;
0.8
0.6 ~ ;-t ¡- ~ lB
0.4
0.2
O
100 FIGURE B11.1
O 20 40 60 80 ''¡--C;_I,; Graphical depiction 01
x FIGURE E11.3
how Ihe centered-
dillerence approach
amounts lo linear
inlerpolalion between
adjacent segment
In Example 11.3 we observe that the centered difference does not exhibit the concenlrations.
"numerical dispersion" that afflicted the backward difference in Example 11.1.
Therefore, at this point, you might be wondering why we ever used the backward-
difference approach in the first place. In brief, although the centered-difference
method yields superior results for the present example, there are other cases where
it yie1dsphysically unrealistic results. Por such cases the backward difference pro-
vides an alternative. The next section deals with these issues. 11.6 NUMERICAL DISPERSION, POSITIVITV, AND SEGMENT SIZE
In addition note that the centered and backward differences are actually two
In the pre~ections we developed a straightforward numerical approach for
specific instances of a more general weighting scheme for handling the advective
term in our model. This approach is outlined in Box 11.1. modeling sleady-stat~llutant levels in dislributed systems. In Example 11.1
we discovered a discrepancy between the numerical and analytical results for a
backward-difference approach. Based on this experience, we take a more system-
atic look at two fundamental questions. Pirst, does the control-volume approach
BO~ 11.1. A Welghted-Dlfterence Formulatlon always yield meaningful results? Second, when it works, how accurate are the
Now that we have developed the backward and centered approximations, we can gen- ~ calculations? We first look at the accuracy issue.
eralize them into a general weighted-difference approach. To do this we can formulate
the concentnttion al the interface between segments by a weighted average,
Cj,k = aj,kCj + (Jj,kCk 11.6.1 Numerical Disperslon
where Now let us look at the problem we encountered in Example 11,I when we developed
t:..Xk a numerical approximation to the mass balance equation
a'k=-
J, t:..Xj+t:..Xk
de d2e
t:..x' O = - U- + E - - ke (11.35)
and f.l.J'k
t' ,
= 1 - a J'k't:..Xj
= --1.
+ t:..Xk
dx dx2

These can be substituted into Eq. 11,1 to give Recall that using a backward-difference approximalion seemed to cause the com-
putation 10exhibit 100much dispersion. To gain insighl into whal is happening, we
O = W¡ + Q¡_I,¡(a¡_I,¡c¡_1 + (J¡-I,¡C¡) - Q¡,i+l(a¡,i+lci + (Ji.i+IC¡+I)
can recognize that for constant-parameler systems, the conlrol-volume approach is
+ Eí_I,¡(C¡-1 - c¡) + Eí,i+I(C¡+1 - c¡) - k¡V¡c¡ identical to a finite-difference approximation of Ihe underlying partial differential
The a's and {J's serve two purposes, First, as depicted in Fig. 811.1, they provide equation (see Box 11,2 for details).
a way to apply the centered-difference approximation for unequally spaced segments. The finite-difference perspective allows the use of Taylor-series expansions lo
In addition lo accounting for unequal segment sizes, they also provide a means for characlerize Ihe errors of the finite-difference approximations, Por example we can
{
shifting belween backward and centered spatial approximations. Note that for a = 1 expand concentration in a backward Taylor series,
¡t!'!d fJ = O the equation becomes the backward-difference representation. In contrast de d2e /lx2
if q = fJ = 0.5 (or for unequai-sized segments using the weights calculated above), = e; - -/lx + --
dx dx2 2!
- ... (11.36)
Ihe resull amounts lo a centered difference.
~ e¡_1

_.' ......-
202 PART11 Incomplelely Mixed Syslems LECTURE11 Control- Volume Approach: Steady-State Solutions 203

If the series is truncated after the second-derivative term, it can be solved for a forrnulation can also be developed that is the equivalent of Eq. 11.30. Problem 1í.4
deals with this case.
de = e¡ - e¡_1 !!.xd2e 1137
dx - !!.X + 2 dx2 ( . )
The divideddifferenceis substituledfor the firstderivativein Eq. 11.35,and the EXAMPLE 11.4. CORRECTING FOR NUMERICAL DISPERSION. Repeat Ex-
resultis ample 11.1, but calculate the numen cal dispersion. Then reduce the dispersion coeffi-
cient by this amount and see what happens.
0= Ed2e _ U de _ !!.xd2e _ ke (11.38)
Solution: Equation 11.40 can be used to calculate numerical dispersion as
dx2 (dx 2 dx2)
Collectingtermsyields En = 202m (100 mhr-I) = looom2hr-1

_ Udedx
(
O = E + !!.XU d2e
2 )dx2
- ke (11.39)
Therefore we can reduce the dispersion coefficient from 2000 to 1000 m2 hr-I
done, the following matrix representation can be developed for the system:
. If this is
Thus we see that the real dispersion is enhanced by the quantity
1900 -500 O O O 1000
CI
!!.X
( 11.40) -1500 2400 -500 O O OC2
E" = TU

~ O -1500 2400 -500 O C3 = O

whereE" is callednumeriealdispersion. O O -1500 2400 -500 C4 O
O O O -1500 1900 { Cs} { O }
Note that this system is identical lo Ihe one derived using the centered-difference
BOX 11.2. The Flnlte-Dlfference Approach approach in Example 11.3. Consequenlly Ihe compuled results will be identical with the
The finite-difference method offers a more conventional (though not necessarily supe- cenlered-difference characlerizalion. Thus correcling the dispersion coefficienls or using
centered differences yield the same oulcome!
rior) approach to deriving computer solutions for distributed systems. In this method
we substilute finile-difference approximations for the derivatives in our fundamental
equation
dc d2c On the basis of the foregoing example, you might now be wondering why we
( O= - V - + E- - kc did not just use a centered difference to start with. The answer lies in an additional
) dx dx2 problem-solulion positivity.
For example the first derivative can be approximated with a backward difference,

dc :: c¡ - C¡_I
e/x - I1x 11.6.2 Posltlvlty
The second derivative can be approximated with a centered difference, It can be shown that as a system becomes relatively more advective, solutions can
C;+1 - c¡ c¡ - C¡_I go negative (Hall and Porsching 1990). The following example iIIustcates this phe-
e/2c:: ~-~ C¡+I - 2c¡ + C¡_I nomenon.
elx2- I1x I1x2

Suhstituting these into the mass balance yields EXAMPLE 11.5. NEGATIVE SOLUTIONS FOR HIGHLY ADVECTIVE SYS.
TEMS. Repeat Example 11.3,but increase the velocity in the tank to 400 mhr-I. Also
O = EC¡+I - I1x2
2c¡ + C¡_I _ VC¡ - C¡_I - kc.
I1x I
set the inflow concentration to zero and add a loading of 4000 g hr-I to the middle of the
_./'
tank. As in Example 11.3 use a centered-difference approximation.
Collecting terms and multiplying by Vi = A..I1x gives Solution: The following matrix representation can be developed for the system:
-(Q + E')c¡_1 + (Q + 2E' + kV¡)c¡ - E'ci+1 = O 3400 1000 O O O CI O
which is ofthe same form as Eq. 11.3.Thus for constant-parameter syslems, the control- - 3000 2400 1000 O O C2 O
volume and finite-difference approaches yield identical formulalions. O - 3000 2400 1000 O C3 = 4000
By substituting the centered-difference approximation for the first derivative, O O - 3000 2400 1000 C4 O
dc :: ('¡+I - c¡ O O O - 3000 3400 { Cs} { O }
,Ix - 2 I1x

-~-'-
 -- -:o1't-----

204 I'ART11 Incompletely Mixed Systems Lr.CTURE11 COlllrol-Volume Approach: Steady-State SolutiuJlS 205

which can be solved for the concentrations. These results, along with caIculations from TABLE 11.1
an analytical solution, are summarized in the following table and figure: Constraints due to numerical dispersion and positivity
ror solving the advection-dispersion-reaction equation
Distance Analyücal Numerical with backward and centered finite-difference schemes
O 0.00003 Numerical dispersion Positivity
10 0.00026 0.08477
0.01584 -0.28823 Backward difference
30 E = 0.5UÓx Óx < 00
50 0.95346 0.94608 (a = 1; 13 = O)
70 0.86483 0.86470 Centered difference E
90 0.78665 0.76297 E=O
(a = 0.5; 13 = 0.5) Óx < 0.5U
100 0.76479
-

1.2
E
I1x < - ( 11.43)
e 0.5U
0.8 Thus if centered differences are used, the spatial discretization must be below the
level specified by this equation to maintain positivity. Note that this constraint is
0.4 often formulated as

0.0
S Pe = utu < 2 (11.44)
I '- I I E

-0.4
I . 80 100
)where P e is the ceII Peclet or Reynolds number (recall Box 9.1).

11.6.3 What It AIIMeans

FIGURE E11.5
At this point let's regroup and tey to make sense out of what we've just learn.:d. We
now know that we are dealing with two constraints:
Although the results downstream from the load are adequate, the simulation above is
c1early in error. In fact it yiclds the physically unrealistic outcome of negative concen- En = U I1x(a - 0.5) numerical dispersion ( 11.45)
trations.

Notice that the matrix in Example 11.5 differs from the others in this lecture.
~~ I1x <
E
positivity

Note that these are now expressed in terms of the weighted-difference fOffnulation
(11.46)

In aIl the other matrices the off-diagonal terms have a different sign fcom the terms
00 the main diagonal. That is, the main diagonal terms are positive, whereas the
off-diagonal terms are negative. In the previous example the superdiagonal terms
are positive. It can be shown (Hall and Porsching 1990) that this is the root cause
of the negative results. To quantify this constraint, we can rephrase it as specifying
that the superdiagonal term must be negative to obtain positive results. RecaII the
general formulation for the superdiagonal term (that is, the constants multiplying
Ci+l),
-E' + 0.5Q < O
of Box 11.1 (compare Eqs. 11.45 and 11.46 with Eqs. 11.40 and 11.43).
The situation is summarized in Table 11.1. If we use centered differenccs we
do not have numerical dispersion. However, for systems where advection is high
relative to dispersion, we are forced to use small segment sizes to obtain physical-
Iy realistic solutions. In contrast if we use backward differences, our solutions are
always positive, but they are contaminated with numerical dispersion.
What's the bottom line? Close inspection of Eqs. 11.45 and 11.46show that the
numerical dispersion and positivity constraints are structuralIy identical. To under-
(11.41) stand what this means, recognize that regardless of the difference scheme our goal
should be

.
Substituting E' = EAcll1x and Q

EAc
= AcU yields Ep = Em + En (11.47)

(11.42) / "NU~" dispersion

- I1x + 0.5UAc< O Real "phYSiCal"/
dispersion The numberyou inducedby difference
which can be solved for use in your scheme
"model"
 LECTURE11 Control- Volume Approach: Sleady-Stale Solulions 207
206 PART 11 Incompletely Mixed Systems

That is, we would like the model output to match the actual dispersion that occurs
J) in theSubstitute
physical world.
Eq. 11.45 into 11.47 and rearrange.
) Em = Ep - U6.x(a - 0.5)
and substitute this result into Eq. 11.46to give
(1 6. Ep - U6.x(a - 0.5)
J x < U(l - a)
The right side of Ihis equation can be manipulated algebraica\ly to give
A Ep
L.\Xc < 0.5U
Thus the a 's drop out and we see thal both the numerical dispersion and the positivity
constraints amount to a single constraint on the segment size. Consequently we have
used Ihe nomenclature 6.xc to connote thal Eq. 11.50defines a critical segment size.
In essence if backward differences are used, this criticallength induces a nu-
/ merical dispersion thal is exactly equivalent to the physical dispersion. Thus we
could set the model dispersion Ernequal to zero and obtain an accurale simulation
because the backward difference would induce Ihe proper dispersion numericaIly.
Conversely this length also represents the poinI al which we could set the model
dispersion to the level of the physical dispersion and obtain positive concentrations
with a centered-difference scheme.
Now what is the significance of these results in real streams and estuaries? Fig-
ure 11.3 shows the critical length versus the width of a coIlection of streams and
éstuaries. This plot indicates that because they are predominantly dispersive sys-
tems, Ihe estuaries have criticallenglhs of more than I km. This is also troe for the
v
c: larger rivers. However. for the sma\ler, more highly advective streams. Ihe lengths
, I faIl below 0.1 km. This means lhat many more segments would be required to ade-
~ In summary, for eSluaries a centered difference can usuaIly be used along with
~ í measured dispersion coefficients. If backward differences are used, the model dis-
persion should be reduced to account for numerical dispersion. For streams. smaller
( 11.48) ~ segment lengths are usua\ly required to ensure positive solutions and to avoid excess
~ ~ numerical dispersion.
One additional practical point should be noted. Although smaIler segment
(11.49)
lengths are required to exactly match the physical dispersion in streams, lhe fact is
( that dispersion is not that important in such systems relative to advection. Therefore.
[ particularly for steady-state solutions where gradients are not especia\ly sbarp, the
conslraint embodied in Eq. 11.50 can often be relaxed wilhout introducing major
(11.50) inaccuracies into the solution. Test calculations against analytical solutions can of-
ten be performed to verify whether tbis is valid. However. it sbould be stressed tbat
'" time-variable solutions for problems such as spills can be very sensitive to numerical
~ dispersion. We'\l explore tbis issue in Lecs. 12 and 13.
11.7 SEGMENTATIONAROUNDPOINTSOURCES
Finer segmentation is typica\ly used to better capture tbe sbarper gradients around
point sources. The problem is iIIustrated in Fig. 11.4. Wben a finer scbeme is used
(Fig. II.4a), the error amounts to less tban 5%. In contrast the coarse segmenta-
tion scheme in Fig. II.4b results in a 30% underestimation of tbe concentralion al
the entry point of tbe point source. MueIler (1976, as described in Thomann and
Mue\ler 1987) has developed some guidelines to sbow how segment sizes can be
related to acceptable errors.
Based on a backward-difference scheme. be developed a set of formulas lo com-
pute segment size needed to acbieve a desired error level in the peak concentralÍon.
\ For advective systems (1/ < 1),

, 1quately simulale such systems. \ U 1+1/

l 6.xo = k (l _ e)2 - 1/- I (11.51)

. ..
[ ]
1000 .
. ..
, .. .....
100
20T_ Error;:5%

><"
..
<1 "11
.. ...--
I /
0.1d
. .
r 111"" , " """ . ,...."
0.01 10.000 100.000 -20 -10 O 10 20 30 -20 -10 O 10 20 30
10 100 1000
Widlh(m) (~ (~
FIGURE 11.4
Comparison 01 numerical and analytical solulions lor a poinl source ¡nlo an esluary.
FIGURE 11.3 J The line segmenlalion scheme in (a) yields much beller resolulion Ihan Ihe coarse
Plol 01 crilical segmenl lenglh versus widlh lor a number 01 rivers and
estuaries. The original dala are laken Irom Hydroscience (1971) and \version in (b).
Fischer el al. (1979).

~M ... ~ po _ __ :: -------..---
 208 PART11 Incomplelely Mixed Syslems lI:L-ruRE 11 Conlrol- Volumc Approach: Sleady-Slalc SOIUliof)s 209

3 4 Out

FIGt 11.6

FIGURE 11.5
Atw0 n. ..msionalvertical segmentation 01 the lar-right
segment Irom Fig. 11.5.
An elongated lake with a side bay is a
simple example 01 a two-dimensional
system.
}ayeras depicted in Fig. 11.6. The founh mass balance would be modified to give
where U = velocity O = Q3.4(.'3 - Q4.ou1C4 + Ej.4(C3 - ('~) - kV4C4 + E4.S(cs - (.'4) (11.57)
k = reactionrate
11 = estuary number = kE/U2 and a fifth mass balance could be written for the lower layer,
e = the desired fractional relative error
Lj O= E4.S(C4 - cs) - kVscs (11.58)
For diffusive systems (11> 1), a comparable formula is
PROBLEMS
T+(T7i]) _ I - ~
(1 - e)2 jii
~XO = k (11.52) 11.1. Develop equalions for Ihe enlry (i = 1) and exil (i = 11) segmenls of a reaclor us-
ing cenlered finile-difference approxim¡¡linf), for Ihe derivalives in Ihe adveclion-
These formulas provide useful staning points for choosing segment sizes around dispersion-reaclion equalion. Assume Ihal IlIl' lIow enlers and leaves Ihe reaclor via
point sources. Additional information regarding their derivation can be found in pipes. Your equalions should be presenled in Ihe formal of Eqs. 11.13 lo 11.20.
~ Thomann and Mueller (1987). 11.2. (a) Perform Ihe same compulalions as in Example 11.1, bUI use a finer segmenlalion
(11 = 10).

(b) Compule percenl relalive errors for your results and for Ihe coarser case (11 = 5)
11.8 TWO- ANDTHREE-DIMENSIONALSYSTEMS as compared wilh Ihe analylical results. Discuss your results.
The extension of the methods described in this lecture to two- and three-dimensional 11.3. Repeal pan (a) of Example 11.2 using a segmenlalion scheme of 11 = 10. Discuss
systems is straightforward. In most cases backward-difference approximations of how and why Ihe finer segmenlalion affecls Ihe reduclion of Ihe inflow concenlralion
lo meellhe 0.1 mg L -1 goal.
the advective terms can be used because any numerical dispersion that is introduced I
\ ¡s usuaIly dwarfed by physical dispersion. However, centered differences can also
be simply applied for cases where numerical dispersion could have a detrimental
effect on solution accuracy.
The simplicity of the applOach can be iIIustrated for the four-segment system -1IDJJ]
s¡= 10 9 m-3ct"1 W = 1000 kg ct"1

!
shown in Fig. 11.5. Using backward differences, we can write steady-state mass
balances as
2.25
0= QO.)Cin- Quc) + E;,3(C3- c) - kV1c) (11.53)
Oi"
O= E2.3(C3 - C2) - kV2C2 (11.54) E 1
.2-
~
O = Quc) - Q3.4C3+ EÍ.3(CI- C3)+ E2.3(C2- C3)
+ Ej.4(C4- C3)- kV3C3 (11.55)
O = Q3.4C3- Q4.outC4
+ Ej.4(C3- C4)- kV4C4 (11.56) g60
Thus we have generated four equations that can be solved for the four unknown ~"20
concentrations.
o 8 16 x (km)
Three-dimensional problems can be handled in a similar fashion. For example
suppOse!hat !he fourth segment were verticaIly stratified into an upper and a lower FIGUREP11.5
 lECTURE11 Control- Volume Approach: Steady-State Solutions 211
210 PART11 Incompletely Mixed Systems

(d) Verify your result by simulating the system with the backward-difference numer-
11.4. Derive Ibe centered-difference version of Ibe one-dimensional, constant-coefficient, ical scheme.
advective-dispersive-reaction equation using the finite-difference approach outlined
in Box 11.2. 11.8. The lake in Fig. P 11.8 receives pollutant inputs Ibat settle at arate of 20 m yr-I. Using
-' the data from Table P 11.8,
11.5. (a) Use a backward-difference scheme (6.x = 0.5 km) to determine Ibe steady-state
distribution of a pollutant (k = 0.1 d-I) for x = Oto 32 km in the system shown
in Fig. PII.5. Note: The concentration in the river immediately upstream of the ...
distributed load is 5 mg L -1 . .....
(b) Estimate the numerical dispersion for the calculation.
(e) Compare your results with the analytical solution from Probo9.4.
+~-
11.6. A one-dimensional estuary has the following characteristics: 2
Value Units
11 lOS cm2 S-I
Dispersion coefficient FIGURE P11.8
Flow 5 x 10" m3 d-I
1:
Widlh 100 m
Deplh
2 m
(a) Calculate the system 's response to the loadings.
(b) Determine Ibe improvement that would occur in segment 3 if the loading to seg-
ment 2 were reduced by 50%.
The pollutant seules (v = 0.11 m d-I). A mass loading of2 kg d-I is discharged into
this estuary. Note that the boundary conditions are TABLE Pll.8
en =O
A, % V W t
dC18-19 =O Segment (km2) (m) (kmJ) (103 kg yr-I) (km)
dx
1 280 4.3 1.2 500 16.7
Use centered differences and Ihe segmentation scheme in Fig. PI \.6 to model the 2 280 4.3 1.2 2000 16.7
dislribution of the pollulant in the estuary. Compare your results with the analytical 3 280 4.3 1.2 300 16.7
solution. Q E A, Width E' Co
~ ;j'; Intedace (kmJ yr-') (km2yr-') (km2) (km) (km3yr-I) (pg L-1)
~j 'L " Inlel-I 170 10
,~ I 1.\ ~' 1-2 170 1500 0.072 16.7 6.5
2-3 10 1500 0.072 16.7 6.5
'I'j\
1. 1 ,;~!¡' 1t 2-oullel 180
, !:I. W'
l 1¡l
..
,:1,1
1" .
ir,
0.5 km 1km 2km 4km
11.9. A mixed-ftow reactor is 10 m long, 2 m wide, and I m deep and has the following
11i~ i,
i, .1 FIGURE P11.6 hydraulic characteristics: E = 2 m2 hr-I and U = I m hr-I. It is subject to a con-
'
~' ;'\i, stant inftow concentralion ofachemical, ei. = 100 mg L -1, Ibat decays by first-order

t kinetics k = 0.2 hr-I . Determine Ibe steady-state distribution of the chemicaI along

I \
11.7. A tidal river has the following characteristics:
the length of the reactor.
~'~
.

.
.
. .. '~! E = I X 106 m2 d-I
. '11 ~
.

U = 250md-1
/

Ac = 1000 m2

A poinl source of 12 X 106 g of a first-order decaying substance (k = 0.075 d-I) is

introduced into Ihis system.
(a) Delermine the concentration at Ihe mixing poin\.
(b) Calculate the estuary number.
(e) Use the appropriate formula based on the result of part (b) (either Eq. 11.51 or
11.52) lo delermine the segment length needed lo oblain a numerical estimate of
Ihe concenlralion allhc mixing poinllhat is approximalely 5% low (8 = 0.05).
. ._h. ...

LECTURE12 Simple Time-Variable SOIUliolls 213

LECTURE 12
t

Ir Simple Time-VariableSolutions tó /
Óx
(,+1 . i.r + 1
-.-
. ;-1,( . i,t . ;+ 1,'
1-1 . i.r-1

;-1 ;+1
x
FIGURE 12.1
A computalional grid used lo characlerize Ihe spalial and
temporal dimensions.

A simple approach for discretizing Eq. 12.1 for such a grid involves using a
centered spalial derivalive and a forward temporal derivative. That is,
aee ef+I - et
() -at ==' dt 1 (12.2)

and a2e
_= et1+1- 2et1 + et1-1
ax2 - dx2 (12.3)
~
Now let's tum to time-variable computer solutions for distributed systems. Aside Notice the notation change in that superscripts are used to denote time. This is done
from describing numerical methods, I emphasize the issues of stability and accu- so that a second subscript could be used to designate a second spatial dimension in
racy. SÚlbilily connotes that errors are not amplifieu by the solution scheme. Hence the event that the approach is expanded to cases thal are two-dimensional in space.
an unstable solution is one where errors ultimately "swamp" the tme solution. Substituting lhese differences into Eq. 12.1 gives
Accuracy deals with how well the solution scheme conforms to the true solution of ef+I - et et - 2et + et
the underlying differential equalion.
1 1 =E 1+1 1 ,-) (12.4)
dt dx2

12.1 AN EXPLlCITALGORITHM
xO Grid polnt Involved in time difference

Grld polnt Involved In second-derlvative

space dlfference
To simplify lhe analysis the time-variable approach is derived for the mass balance,
ae a2e
t( -at = E-ax2 (12.1)
1+1

~!hus we neglect advection and reaction and assume that the dispersion coefficient
liS constant.
A numerical solution for this equation can be developed by substituting finite-
difference approximations for the derivatives. In contrast to the previous lectures on FIGURE 12.2
steady-state solutions, we must now discretize in time as well as space. Figure 12.1 A computational molecule or stencil for
shows a computational grid for one-dimensional spatial cases. ; -1 the simple explicil method using FTCS
;+1 difference approximations.
212
 214 PART11 Incompletely Mixed Systems LECI1JRE12 Simple Time-Variable Solutions 215

( This approach is sometimes referred to as aforward-tímelcentered-space (FTCS) that we are dissatisfied with our spatiaI resolution arid rloubte the number oi st\iiUál
)difference equation. It can be further simplified by collecting terms and solving for
the concentration at the advanced time step,
~ segments; that is, we cut /:1xin haIf. According to Eq. 12.10 we must quatter the
time step to maintain stability. Thus to perforrn comparable calculations, the l1um.
e( - 2c( + e( ber of time steps must be increased by a factor of 4. Furtherrnore the computátion
ef+1 = e~+ E i+1 ¡ ¡-I /:1t ( 125 ) for each of these time steps will take twice as long because halving &x doubtes
I I /:1x2 .
~ the number of segments for which equations must be written. Consequently for the
Inspection of this equation reveals that it is actually a form of Euler's method. one-dimensional case, halving /:1xresults in an 8-fold increase in the number of cal-
That ¡s. it is in the format
culations. This is a serious computational penalty for a 2-fold increase in spatial
resolution.
New = old + slope(step) (12.6)
For two- and three-dimensional problems, the stability constraint is even more
! This lI1ethodis referred to as an explicit metllOd, so named because each difference severe. For example for a two-dimensional problem, halving the spatial grid re-
\ equationhas oneunknown(theconcentrationat the next timestep)thatcan be iso- sults in a 16-fold increase in computations. Although the simple explicit method
latcd 011the left side of the equal signo A computational molecule for this explicit is severely limited by the stability constraint, it is still widely used for water-quality
method is depicted in Fig. 12.2, showing the nodes that constitute the spatial and modeling because it is extremely easy to programoThis advantage is particularly
temporal approximations. important when dealing with nonlinear kinetics (we elaborate on this aspect in the
II
Wc should also recognize that substituting finite differences for the right-hand next lecture).
I1 side 01'Eq. 12.1 effectively transforms the partíal differential equation into a system
lit" of ordinary differential equations. Thus the use of Euler's method to integrate the
systelll of ODEs, which results from substituting Eq. 12.2 for the left-hand side, 12.3 THE CONTROL-VOLUME APPROACH
l' reprt"scntsthe most straightforward way to obtain solutions.
I ' 1'11now show how the general control-volume approach developed in the previous
lecture for steady-state solutions can be extended to simulate time-variable results.
12.2 STABILlTV Using a forward difference for time and a weighted difference for space, we can
write a mass balance for segment i as
To investigate the stability of Eq. 12.5, we first collect terms to give
(+1 - e~
c¡ =
(, d+'= Acf_1 + (1 - 2A)cf+ Acf+, (12.7) V¡ /:1t
I

wf + Q¡-U(U¡-,,¡cf_, + 13i-J,¡cf) - Q¡,i+1 (U¡,¡+tcf + 13¡,i+lcf+l)

r I
I
I where
')

I
I E/:1t + Ef-u (cf-I - cf)+ Ef,i+1(cf+1- cf) - k¡v¡cf (12.11)
! A = /:1x2 (12.8) Dividing both si des by V¡ and rearranging yields
i
I
I , and ,\ is formally called the diffllSioll Illlmber. f+ I _ ( /:1t (
I
Ir Now we will examine the stability of Eq. 12.7. Stability connotes that errors are c¡ - c¡ + V¡ [W ¡ +
Q
¡-I,¡ (
u¡_I,¡c¡_1
(a (
+ I-'¡-I,¡C¡ )- Q¡,i+l (u¡,¡+IC¡ (a (
:+-I-'¡,i+1 ci+ 1)
II not amplified as the computation progresses. It can be shown that Eq. 12.7 will yield
II (
stablc results if the diagonal term is positive; that is, 1 - 2A > O. In other words,
+Ef-I,¡ (cf-I - cf) + Ef,i+' (cf+1 - cf) - k¡V¡Cf] (12.12)
I

I
¡I A< ! (12.9)
We collect terrns to give
In addition it should be noted that setting A < ! could result in a solution in which er-
I
( rors do not grow but oscillate. Setting A < ! ensures that the solution will not oscil- ,
C(+I Wf
V¡ l:1t -
/:1t
V¡ (-Q¡-I,¡U¡-U
,
- E¡_u)c¡_1
(

0\ lale,and A = ~ tendslOminimizetruncalionerror(Carnahanet al. 1969)../

I
A1though satisfaction of Eq. 12.9 guaranlees slable solutions, it also places a
I slrnng limilalion on the explicil method. This can be seen by equating Eqs. 12.8 and
12.9 and solving for
+ [1 - ~: (-Q¡-u13i-J,¡ + Q¡,i+IU¡,i+1+ Ef-u + E:'i+1 + k¡V¡)]cf
I /:1t , e
I l/:1x2
- (
V¡ Q¡,i+113u+I - )
E¡,¡+I ci+1 (12.13)
j /:1t< lE (12.10)
I ( l1ecause it allows liS to see how /:1tchanges as we vary the spatial step size Depending on the value of u, this equation can represent either a forward-timelback-
! I /:1x,Ihis equalion offers a IIkc pcrspective onlhe constraint's significance. Suppose ward-space (FTBS) for u = 1or an FTCSforu = 0.5.
I
i
'1,
 .

216 PART11 Incomph:lely Mixed Syslems LECTURE12 Simple Time-Variable Solulions 217

As with tbe simple model described in the first section, solutions for this equation
will be stable if the diagonal term is posilive. For Ihe case where all parameters are BOX 12.1. Error Analysls lo Estlmate Numerlcal Dlsperslon
constant this is
For simplicily we analyze Ihe lruncation error of the adveclion equation
dt
(12.14) ac = _uac
l - V( -Qf3 + Qa + 2E' + kV) > O al ax (12.20)
which can be sol ved for
Recall from Lec. Illhat a backward Taylor-series expansion can be lruncaled lo yield
(recall Eq. 11.37)

l dt<
Q(a -
V
13)+ 2E' + kV
(12.15)
ac
_=
ax
cf - C(_I
1.-+
Ax
Ax a2c
T ax2 (12.21)
or dividing the numerator and the denominator by the cross-sectional area,
In similar fashion a forward series can be wriuen in lime and lruncaled lo give

\ ..
~t <
U dx(a -
(dx)2
13)+ 2E + k(dx)2
(12.16) ac C(+I- e!
- ... --1
al Al
Al a2c
J. - - -
2 al2
(12.22)

Careful jnspection of either Eq. 12.15 or 12.16 indicates that, beyond the fact Equations 12.21 and 12.22 can be combined wilh 12.20 lo yield
that the matrix diagonal elements must be positive, the stability criterion actually
r\ G ac = _Uac + UAx a2c _ Al a2c (12.23)
I has a physical meaning. That is, it says that the time step cannot exceed the pollutant al ax 2 ax2 2 al2
~ residence time (recall Seco 3.2.1) for the segment. This can be seen c1early for the
The second-order lime derivative can be converted lo a space derivalive by diffel""u-
~ case involving pure advection with a backward difference (a = 1, k = E = O).For
ating Eq. 12.20 wilh respecllo lime,
this case the stability criterion becomes
a2c a2c
dt< -QV (12.17)
al2 = -u alax (12.24)
Next differenliate Eq. 12.20 with respecllo space,
which is the cell residence time. Further we can also multiply the top and bottom of a2c a2c
-aJ.a-l = -U-ax2 (1~.~5)
the equation by the cross-sectional area to yield
Therefore Eqs. 12.24 and 12.25 can be combined lo give
dt<dX (12.18)
~ ~ U _al2 = u2_
a2c a2c
(12.26)
ax2
This result indicates that another interpretation of stability is that the time step cannot
which can be substituled imo Eq. 12.23 lo yield
\
h.: so large tbat the water could move more than the segment length dx for a given
'.city U. In other words the stability criterion specifies that if the computation ac = _uac
I'gcts ahead of itself' it will "blow up."
Note that Eq. 12.18 is formally referred to as the Courant condition. It is of-
al ( AX U _ U2AI
2 2 )
a2c
ax2 ax (12.2/)

ThusIhe numericaldispersionis
~ ten presented as 'Y < 1, where 'Y = Udt! dx is called the advection number or the
I Courant number. E = AxU _ U2AI (12.2M)
"2 2

12.4 NUMERICAL DISPERSION

Close inspection of Ihis formula indicates that there are two independent dfc\:b on
In Lec. 11 we used a Taylor-series expansion to estimate the numerical dispersion numerical dispersion: spatial and temporal discretization effects,
created by approximating advection with backward spatial differences. Now we ex-
ten~ this analysis to determine the effect of forward temporal differences. It can be L U2dt
shown that tbe numerica! dispersion can be represented as (see Box 12.1 for details) { En = Udx(a - 0.5)- ---z- (12.29)
I
Spatial
,
Temporal
(12.19) discretization discretizallon
1 En = Udx[(a - 0.5) - ~:;]
 218 PART11 Incompletely Mixed Systems LECTURE
12 Simple Time-Variable Solutions 219

( An interesting aspect of this formula is that the time step has a negative effect Now let us tum to the FfCS approach, which exhibits no spatial numerical dis-
persion. For this case Eq. 12.29reduces to
(( upon the numerical dispersion. In other words as the time step is increased, the for-
mula states that the numerical dispersion should decrease. In the extreme it might U2t:.t
lead you to believe that at a sufficiently high value of the time step a negative disper- En = ---y- (12.31)
sion could result. Since this is clearly physically unrealistic (it means that gradients
would sharpen up, a clear violation of the second law of thermodynamics !), we must Thus a negative numerical dispersion is created. In this case an additional dis-
closely examine the formula to truly understand its implications. persion equivalent to the absolute value of the quantity ca1culatedin Eq. 12.31would
To do this let's first determine the time step that would yield zero numerical be added to the actual physical dispersion used in the mode!. This correction is il-
dispersion. This can be done for the case of a backward difference (ex = 1) by set-
lustrated in the following example.
ting Eq. 12.29 to zero and solving for
EXAMPLE 12.1. NUMERICAL SIMULATION OF A SPILL. A river has !he fol-
t:u lowing characteristics:
t:.t= U (12.30)
Depth = 1 m Flow = 144,000m3hr-I
( Comparison of this result with Eq. 12.18 indicates that this time step conforms Width= 60m Dispersion= 150,000m2 hr-I
-- \ to the point at which the stability condition would be violated. Thus because of the ~ Velocity = 2400 m hr-I Cross-sectional area = 60 m2

stability constraint, we cannot use a larger time step and hence it is impossible to A spiII of 5 kg of a conservative contaminant occurs about 0.5 km below a city.
create a negative dispersion. Use both (a) analytical and (b) centered-differeJ)ce numerical models to simulate the
Now the second implication of Eq. 12.29 is that the total numerical dispersion distribution 0.2, 0.4, and 0.6 hr folIowing the spilI.
increases as the time step is decreased. Why would this occur? A graphical depic- Solution: (a) Equation 10.22 can be used-t<rgenerate the analytical soluti~n
tion of the situation for the FfBS difference is shown in Fig. 12.3. In Fig. 12.3a an
initial concentration is set in a segment that is t:.x wide. This concentration is ad- 5 X 103/60 _ (x-500-24001)2
o,
o
c(x, 1) = e 4(150.000)1
vected downstream over a time step equal to the critical value ca1culated in Eq. 2 jl7(150,OOO)1
12.30. Thus as shown, the entire mass in the segment is displaced downstream with
Values for time and space are substituted into this formula and used to generate the solid
no mixing. As in Eq. 12.29 this occurs because the positive spatial numerical dis- linesin Fig. 12.4.As expected,the spiIIis transported2400m hr-I x 0.6 hr = 1440m
persion is canceled by the negative temporal dispersion. downstream and has spread
In Fig. 12.3b the numerical experiment is repeated but using a time step that X99 = 5.2j2(150,OOO)0.6
= 2206m
is one-half the critical value. In this case half the mass wi\1be moved downstream
while half remains in the original segment. Thus a net numerical dispersion is cre- over the 0.6-hr periodo

ated since only part of the spatial dispersion is diminished by the negative temporal
dispersion. In the limit (t:.t _ O) the temporal component wi\1 approach zero and
\ the total dispersion wi\1approach the spatial component. 0.2 , (a)
1= 0.4
't
0.1
t
FIGURE 12.4
-- (a) Analytical (solid
::, O Una) and numarical

'lL
~ x
,.0 'b=.- x
FIGURE 12.3 ,-
01
E
O 1000 2000 3000 simulations (horizontal
dashes) 01 el spill in
a river obtained with
Depiction 01 the numericá(
dispersion created by the lorward-
timeibackward-spaca (FTBS)
0.2
0.1
t (b)
a At = 0.01 hr. (b)
Unstabla results with

11l '~
a Al = 0.0225 hr.
differance approximations. In (a) Note thal dashed
the calculation uses the critical
Unes are used
time step basad on stabiUty. For O
"..IU x
this case no numerical dispersion
lor the numerical
results lo make
26x x occurs. In (b) a smallar time
O 1000 2000 3000
the instability more
step is usad. For this case the x(m) evident.
(a) M =!ulU (b) M=6x/(2U) concentration is dispersed.
220 PAR,.11 Incompletely Mixed Systems IH'TlIlIE 12 Simple Timc-Variable Solutions 221

(b) The first step in developing a numeri¡;al solution is to determine the critical segment advection. In such a ease Ihere mighl nol be sufficienl real dispersion to compensale
length (Eq. 11.50), for the numerical dispersion.
150.000 = 125 m As in Ihe previous example, Ihe firsl slep in answering Ihis queslion is lo calcu-
~ Ax,. < 0.5(2400) late the critical segment length (Eq. 11.50).
To be on the safe side. we adopt a Ax = 100m. 2EI'
Next we determine the time step relJuired to maintain stability using Eq. 12.16. ~xc = U (12.32)
(100)2 . Next the slabilily criterion must be met. For the eentered difference with no reae-
Al < 2(150,000) = 0.033 hr (= 2 mm) tion this is (recall Eq. 12.16)
A
...1<-(~x)2
Again to be on the safe side. we use a time step ofO.OI hr. (12.33)
The numerical dispersion can be determined with Eq. 12.31. 2EI'
Now substitule Eq. 12.32 into Eq. 12.33 to give
E = - (2400fO.01 = -28800
n 2 . m2 hr-I
(2E /U)2 2E
This means that a value of 150,000 + 28.800 = 178.800 m2 hr-I will be used in the
~I < P
2EI'
= -E.
U2
(1234)
.
numerical mode!.
Now before performing Ihe simulalion, we musl address Ihe issue of an inilial
Finally subslitute Ihis time step into the numerical dispersion estimate (Eq. 12.31),
condition. The initial condition for the analytical solulion is an impulse function. Thal U2 (2E / U2)
is, the spill is assumed lo occur instantaneously in space and lime. Because our control- En = - 2/' = -El' (12.35)
volume approach breaks lime and space ¡nlo discrele chunks, it is impossible lo exaclly
represenl such a function in our control-volume approach. An approximation involves Consequently for this case the stability poinl represents a boundary where a nega-
distributing Ihe mass across a single segmenl al the start of the computation. For Ihe tive numerical dispersion exactly balances the physical dispersion. Thus because
present case Ihis means Ihal the following concentration would be established in Ihe moving beyond this point with a larger ~I wiIIyield an unstable solution, a physically
segment where Ihe spill occurred: unrealislic negative dispersion cannOIbe generaled.
m 5000 _1 In summary we have successfully developed a simple approach for simulat-
c(500) = B AxH = 60(100) I = 0.833 mg L ing lhe time-variable distribution of a pollutant in a distributed system. Because the
A value of zero would be used as an initial condition for Ihe olher cells.
scheme is so straightforward, it has utility for water-quality applications. However,
Using this approach we depict the results of Ihe model run in Fig. 12.4a, along with
beeause of its stability constraint, more refined techniques have been developed.
~ The next lecture deals with some of these methods.
the analytical solution. As manifested by a small reduclion in Ihe peak and small exlen-
sion of Ihe lails, the numerical values seem to exhibit slightly more dispersion Ihan for
the analytical case. However, Ihe overall agreemenl is adequate. The numerical results PROBLEMS
generally follow the trend of the exacl solulion. If finer segmenlalion were imposed (that
is, smaller Al and Ax), Ihe numerical oulcome would approach the analytical case. 12.1. Suppose Ihat for Example 12.1 you redueed Ihe spatial seale lo Ax = 25 m. Ca1culate
However, it should be noled that the lotal number of ca1culations required to obtain Ihe time step required to (a) maintain slabilily, (b) avoid oscillalions, and (e) minimize
the solution amounls lo
trunealion errors. For (e) determine exaelly how many ea1culalions would be required
n = XIOI.I1101.1 =
Ax Al
3000 ~
100 0.01
= 1800 lo perform Ihe same ca1culation as in Ihe example (Ihat is, from I = O lo 0.6 hr).
12.2. Boulder Creek has the following eharaeteristies:
Ifthe space discrelizalion were refined, Ihe calculalions would increase in Ihe rapid fash-
ion specified by Eq. 12.10. . Parameter Value
Figure 12.4b shows Ihe results for the case Al = 0.0225hr. As can be seen, the Width 12m
solution is beginning lo become unslable. If larger sleps were used, the ea1culalionwould Depth 0.3 m
begin to oseillale even more violently. Thus the explicit melhod's slabilily eonstrains our Flow 1.7mJ S-I
ability to lengthen the time slep. Dispersion coefficient 4.5 m2S-I

~ow one more question needs to be addressed: For the centered-difference case Ifyou are evalualing a pollutanl Ihat deeays wilh firsl-order kinelics (k = I d-I),
is if possible to generate a negative dispersion? Such might be the case for a highly (a) Caleulate Ihe eSluary number. Inlerprel your result-is the syslem adveelive, dis-
persive, or adveetive/dispersive?
advective system; that is, one where the physical dispersion was low relative to
 222 PART11 Incompletely Mixed Systems

LECTURE 13
(b) For a steady-state ca\culation, determine the critical segment length.
(e) If you use this segment length, delermine Ihe lime slep needed lo jusl maintain a
stable solution with a cenlered-difference, explicil (that is, Euler) solulion.
(ti) What numerical dispersion would be manifested for par! (e)?
(e) Ca\culale how a 2-kg spill of a conservalive subslance would move through a 20-
Advanced Time-Variable Solutions
km stretch of Ihe syslem.

12.3. An estuary has Ihe following characleristics:

Deplh =4m FIow = 100,000 m3 hr-I

Velocity= 250m hr-1 Cross-seclionalarea = 400 m2
Widlh = 100m Dispersion= 1,000,000m2hr-I

A spill of 10 kg of a reactive conlaminant (k = 0.1 d-I) occurs. Use both (a) an an-
alylical and (b) a cenlered-difference numerical model to simulale Ihe distribulion of
the contaminanl in the esluary.
12.4. A mixed-flow reactor is 10 m long. 2 m wide, and I m deep and has Ihe following
hydraulic characterislics: E = 2 m2 hr-I and U = I m hr-I . It initially conlains pure
water. Al t = O il is subjecllo a slep loading of a chemical, CiD= 100 mg L-I, thal
decays by firsl-order kinelics k = 0.2 hr-I . Delermine the distribution of the chemical
along the length of the reactor as a funclion of lime using Ihe explicil method and
centered spatial differences.

12.5. Suppose that 20 kg of a conservative toxic compound were spilled in segment I of the
lake described in Probo 11.8. Use the explicit approach with backward spatial differ-
ences lo ca\culate the time-variable response of the system to the spill.
)
\

The simple explicit finite-difference formulation described in the last lecture has
problems related to stability and numerical dispersion. The two types of algorithms
described herein, implicit and predictor-corrector methods, attempt to obviate these
deficiencies.

13.1 IMPLlCIT APPROACHES

One source of the difficulties with the methods described to this point is that, as de-
picted in Fig. 13.1, they exclude information that has a bearing on the solution. That
is, the value at the advanced time step is affected only by the values of the segment
and its two adjacent neighbors at the previous step. Clearly it is also impacted by
the concentrations at other segments in both the spatial and temporal domains.
Implicit methods overcome this difficulty at the expense of somewhat more com-
plicated algorithms. In this section 1 describe two types of implicit approaches that
are commonly used.

223
r" " -- y-......

224 PART11 Incompletely Mixed Systems LECTUKEIJ Advanced Time-Variable SOIUliol1s 225

e! X Grld point involved In time difference

/' O Grld polnt Involved In space difference

-- - 0- -- -« -- -0---
,, ,, '.
-- -?- - - ~ - - -y--- "':'

\
I I I I
I I I I

- - - ~_ - - + - - - I --+---';'---<:>
,. ,, '' e+ 1
, ,
: '
---0--- -- -0----o
. " boundary
,,

- - - "" : ---( /'-~'

I
;-1 ; +1 ; -1 ;+1
x
Inltial
condltlons
~ (a) Expllcit (FTCS)
FIGURE 13.2
(b) Impliclt (BTCS)

FIGURE 13.1 Computational molecules demonstrating the fundamental diflerences

Representation of the eflect of other nodes on the finite-diflerence between (a) explicit and (b) implicit methods The computational module in
approximation at node (i.e) using an explicit finite-diflerence (b) is sometimes callad a "backward-time/centered-space" (BTCS) scheme.
scheme. The shaded nodes have an influence on (i.e), whereas
the unshaded nodes. which in reality aflect (i,e), are excluded.

We now illustrate how the scheme is set up for the control-volume approach.
Using a weighted difference for the advective terms,
i 13.1.1 Simple Impllclt or "Backward- Time" Algorithm
Cf+l - cf _ f+1
The fundamental difference hetween explicit and implicit approximations is de-
picted in Fig. 13.2. For th. "hcil form, we approximate the spatial derivative at
V¡ ~t - W¡ + Q¡-I,¡ (a¡-l.ic¡_t (+1 + 13¡-I,¡c¡(+1 )
time level t (Fig. 13.2a). i,ecall that when we substituted this approximation into f+ 1 '
(a¡,¡+lC¡ + 13 i.i+lc¡+l(+ 1 + E¡-I.i .(+ I ti 1
- Qi.i+1 ) (C¡_I - c¡ )
~
the P 'al differential equation, we obtained a difference equation with a single un-
known. us we could solve "explicitly" for this unknown.
In imph ~methods, the spatial derivative is approximated at an advanced time
leve!. For exampie.the first and second derivatives could be approximated by cen-
'
+ Ei,i+1 (C¡+I - c¡
(+1 (+1
)- k ¡ V¡c¡(+1 (13.3)

tered differences at time t + l by (Fig. 13.2b) Collecting terms gives

-j)c c~+I
=,+1 - c(+
,-11 ( Wf+IM _ ~t
)C¡-t.t+1
I

j)x - 2~x
(13.1)
c¡ + ~ - - (
V¡ Q¡-I,¡a¡-I.i + E¡-I.i
~t
j)2c _ cf¡¡ - 2cf+1 + cf~:
j)x2 = ~x2
(13.2) ( I I (+
+ [ 1+ V¡ -Qi-I,¡13¡-I.¡ + Q¡.¡+la¡.¡+1+ E¡_I,¡+ E¡,¡+I+ k¡V, ] c¡ )
1

~t
Note that because the spatial derivatives are evaluated at the advanced time step, ( I (+ 1
- V¡ -Q¡,¡+I13i.i+1 + E¡,¡+I c¡+l ) ,
(1 d)
( this approach is sometimes referred to as a backward-time/centered-space (BTCS)
I implicit scheme. The relationship can be simplified further for the centered-difference approximation
\ When this relationship is substituted into the original PDE, the resulting dif- (a = {3 = 0.5) and constant parameters,
f¡:rence equation contains several unknowns. Thus solutions cannot be obtained by W~+I~t U~t E~t f+1 2E~t
c~+ --L- f+1
l, simple algei>raicrearrangement as was done in the explicit approach. lnstead the , V¡ = [- 2~x - (~x)2 ] c¡_1 + [ l + (~x)2 + k~t ] c¡
\ entifesystem of equations must be solved simultaneously. This is possible because, U~t E~t
a}Ol1gwith the boun<tary conditions, the implicit formulations result in a set of lin- (+1
earalgebraic equations with the same number of unknowns. The method therefore (13.5)
+ [ 2~x - (~x)2 ] c¡+1
reduces to the solution of a set of simultaneous equations at each point in time.
 226 PART11 Incompletely Mixed Systems LECI1JRE 13 Advanced TIme-Variable Solutions 227

0.2 T (a) The results of applying the BTCS approach are shown in Fig. 13,3a. Observe how¡
although there is some inaccuracy, the numerical solution follows the general trend of
0.1 the analytical solution and is much better Ihan the unstable eJtplicit lechniqüe from
Fig. 12.4b.
--
::. o
CII o 1000 2000 3000
.5- Although Ihe simple implicit method is uncondilionally slable, there is an ac-
... FIGURE 13.3 curacy limil lo Ihe use of large lime sleps. Consequently it is not that much more
(a) Analytical (sol id !ine) efficient Ihan Ihe explicit approaches. Where il does shine is for steady-state prob-
0.21 (b) and numerical simulations lems.
0.1
(horizontal dashes) 01 a spill Recall from Lec. 12 that malrix solulion methods can be used to oblain steady-
in a river oblained wilh Ihe
slate solulions. An allemalive approach would be lo mn a time-variable solution
o eTCS implicil approach wilh a ~
M = 0.025 hr. (b) The same
until it reached sleady-slale. In such cases, because inaccurale inlermediate results
o 1000 2000 3000
calculalion is shown. bul wilh are not an issue, implicil methods allow you lo use larger time steps and hence
x(m) Ihe Crank-Nicolson algorilhm. can generale steady-slale results in an efficient manner. In fact the simple implicil
~ approach outlined here is particularly well-suited for this task since it converges
This equalion applies lo all bul Ihe firsl and Ihe lasl inlerior nodes, which must ~ quickly(for exampleil is superiorlo Ihe Crank-Nicolsonmelhodin this regard).
be modified to reftect Ihe boundary condilions. These are handled in a fashion sim- \ Consequently water-qualily models such as QUAL2E (Brown and l!.~well(987)
ilar lo Ihe sleady-slate case (recall Seco 11.2). When Eq. 13.5 is written for all the . O attain sleady-stale soIUlioñs.- ~
~ use this algoritl1

~
elemenls, the resulting set of n linear algebraic equations has n unknowns. In ad-
dilion for Ihe one-dimensional problems described here, the method has Ihe added
13.1.2 Crank-Nlc
( bonus Ihat the syslem is Iridiagonal. Thus we can utilize Ihe efficient solulion al-
~ --Ison -
gorithms Ihal are available for Iridiagonal syslems. In the lasl leclure we used an explicil scheme that was forward-time/ceittered-
\
Using an analysis similar to Ihe one in Box 12.1, it can be shown that the nu-
space. In Ihe previous seclion we used an altemalive algorithm thal was backward-
merical dispersion for Eq. 13.5 is
time/cenlered-space. Both approaches have the disadvantage that the spatial and
U!:!.t temporal derivatives are out of synch-whereas Ihe space differences are centered,
En = U!:!.x[ (o: - 0.5) + 2!:!.x
] (13.6) the time differences are biased.
~ The Crank-Nicolson method remedies this situation by using a centered-time/
Consequenlly Ihis equalion again shows that, when we use centered spalial dif- ~ centered-space strategy (Fig. 13.4).In essence, estimates ofthe spalial derivative are
ferences, Ihere is no spalial numerical dispersion. However, because the lime deriva- made at both Ihe present and Ihe fulure lime step. These estimates are then averaged
live is biased (that is, backward), a lemporal numerical dispersion is induced, lo obtain a spalial estimate Ihat corresponds to the midpoint of the time slep.
1'11now ilJustrate how Ihe scheme is sel up. Using constant parameters and cen-
~ E _ U2!:!.t (13.7)
~ n - ---"2 tered spalial differences,

Nole Ihal Ihis result is opposite in sign lo the simple explicit result (compare with Eq. c~ - c~ c~+1- C~+I
- c~ /+1 /-1 + /+1 ,-1
12.31). Therefore an addilional dispersion equivalenl to the quantity calculated in
C~+I
/ / - _U 2 !:!.X 2 !:!.x
1ft - 2
Eq. 13.7 would be sublracled from the actual physical dispersion used in Ihe model.
EXAMPLE 13.1. SIMPLE IMPLlCIT SIMULATION. Apply the backw~imel
~ ( )
,~ centered-space algorithm to the spill simulation described in Example 12.1. Use a tem- Cf+I-2Cf+Cf (+1- 2c~+1+ cf~1
poral step size of 0.025 hr for the calculation.
ci+!, c¡(+ c~+I
,
~~
Solu!ion: The numerical dispcrsion can be determined

E = (2400)20.025 = 72 000
2 h -1
with Eq. 13.7,
+E

[
(!:!.x)2
+2 (!:!.x)2 - k
]
2
( )
" 2 ,m r
I (13.8)
This means that a value of 150,000 - 72,000 = 78.000 m2 hr-I will be used in the Collecting lerms gives
numerical mode\.
 ~

-
228 PART11 Incompleldy Mixed Systems
I.ECfURE13 Advanced Time-Variable Solutions 229

xO Grld polnt Involved In time dlfferenee 13.2 THE MACCORMACKMETHOD

Grld polnt Involved In seeond-derlvatlve
spaee dlfferenee ~The MacCormack (1969) method is a predictor-corrector approach that is closely
related to the Heun method described previously in Seco 7.2. In its most convention-
o
~

Grld polnt Involved In flrst-derlvatlve al form it uses the stencil shown in Fig. 13.5. Thus it uses the following equation
spaee dlfference
lo make an initial prediction of the slope at the base point:

--t;:J..
:-
J. FIGURE 13.4 c~ - c~ c~ - 2c~+ c~
~J."~~.. .., ~.~
".... (+1 A compulalional molecule or slencil s . = -U 1+1 , + E ,+1 , ,-1 - kc~ (13.12)
........ lor Ihe Crank-Nicolson melhod uslng l., dx (dx)2 ,
~....-........ .....
.". 1+1 a cenlered difference approximalion
....... lor the advective termo The dashed Thisslopeis then usedin conjunctionwith Euler's methodto finishthe prcdictor
lines are meanl to show Ihat Ihe spatial
calculation,
,'1':.1.
derivalive estimales al the present (t) (
and lulure (t + 1) limes are averaged to I1 c[+1 = cf + SI.;dt (13.13)
-<i) // ~T:r-- obtain the estimate at the intermediate
1-1 1+1 time (t+ H. NOlice that this is the first time we have used afonvard spatial difference (that is,
in the advection term).
Then the predictor is used to make a slope estimate at the end of the interval,
-(A + -y)cf::1+ 2(1 + A+ kd1\ 1_(A- y)cftl
= (A + y}ct-I - A - kd/)cf C~+I - C~+I C~+I - 2C~+1+ cf+1
+ 2(1 + (A - y)c[+1 (13.9)
1 s . = -u ' ,-1 + E ,+1 , ,-1 - kc~+1 ( 1314
where 2., dx (dx)2 , . )
Edl
i A = (dx)2 (13.10)
Observe that now we use a backward difference for the advection termo
Finally an average of the two slopes is used to obtain the finalcorrector equaliolJ,
Udl S I . + S2 .
y = 2 dx
st
c[+1 = cf + ( .' .')2 dI (13.15)

l
This equation applies lo all but the first and Ihe last interior nodes, WhiCh
be modified to reftect the boundary conditions. These are handled in a fashion si i- The MacCormack method has the advantage that it is more stable than the FTCS
lar to the steady-state case (recall Seco11.2).When Eq. 13.9 is written for all th el- approach described in Lec. 12. In addition it is not plagued by numerical disper-
ements, the resulting set of n linear algebraic equations has n unknowns. In ad ition 1 asion. Consequently, as described in the next example, it yields a good prediction at
reasonable computational price.
the metbod again has the added bonus that the system is tridiagonal. Thus we can
utilize the efficient solution algorithms that are available for tridiagonal systems.
EXAMPLE 13.2. CRANK-NICOLSON SIMULATION. Apply the Crank-Nicolson X Grld polnt Involved In lime dlfferenee
algorithm to the spill simulation described in Example 12.1. Use a temporal step size of O Grld polnt Involved In seeond-derlvallve
0.025 hr for the calculation. spaee dlfferenee

Solutloo: Because we are using centered differences in both space and time, numerical
dispersion is not an issue for Crank-Nicolson. Consequently the actual dispersion can
o Grld polnt Involved In flrst-derlvallve
spaee dlfferenee
be used. Using the other parameters from Example 12.1, the results of applying the
Crank-Nicolson approach are shown in Fig. 13.3b. Observe how, although there is some
inaccuracy, the numerical solution follows the general trend of the analytical solution ---t;::~
J. Jo(,
¡;. '~
.. .
1'+ 1
and is superior to the simple implicit approach outlined in Example 13.1. In addition
the results are much better than the unstable explicit technique frem Fig. 12.4b.
Although the Crank-Nicolson method is efficient for linear problems, it becomes
less convenient when used for the nontinear models that are sometimes encountered in .. . J "":1.
¡.J..'
~ water-quatity modeting. For such cases the system of equations contains nonlinear terms
¡r~ FIGURE 13.5
aIJ~cannol be directly solved using matrix algebra approaches. Thus, as described next, "'rr-
j al! ¡¡4vanCedexplicit method can be used. ;-1 ;+1 A computational molecule or slencil lor
Ihe MacCormack method.
 LECI1JRE13 Advanced Time-Variable¡ Solutions 231
230 P¡\RT11 Incomplelely Mixed Syslems

TABLE 13.1
EXAMPLE 13.3. MACCORMACK SIMULATION. Apply the MacCormack algo-
Comparlson of numerlcal methods for solvlng one-dlmenslonal water-quality models
rithm to the spill simulation described in Example 12.1. Use a temporal slep size of
0.025 hr for Ihe ca1culalion. Programmlng
Solution: As was Ihe case wilh Crank-Nicolson, since we are using cenlered differ-
Numerlcal
Method Dlsperslon Stablllty Linear Nonllnear
ellces in bolh space alld lime. numerical dispersion is nol an issue for the MacCormack
melhod. Conscquenlly Ihe aclual dispersioll can be used. Using Ihe olher parameters Explicit:
from Example 12.1. Ihe resulls are shown in Fig. 13.6. Observe how, although there is fTBS
some inaccuracy, the numerical solution follows Ihe general Irend of the analylical so- (backward-spaceEuler) Condilional Easy Easy
!Ux- !U21
IUlioll. 11is comparable in perforrmmce lo Ihe Crank-Nicolson melhod and is superior
fTCS
lo Ihe simple implicil approach oUllined in Example 13.1. In addition the results are
(cenlered-space Euler) -!U21 Condilional Easy Easy
mllch hctler Ihan Ihe unslable exp1icil lechniquc fmm Fig. 12.4/1.
MacConnack (Heun) None Condilional Easy/moderale Easy/moderale
Implicit:
0.2
I .......
1=0.2
BTCS (simple implicil) !U.1x + !U2.1t Unconditional Moderale Difficull
::.. FIGURE 13.6 Crank-Nicolson None Unconditional Moderale Difficult
C1 0.1 Analylical (sol id line) and
g numerical simulations
'"
o
(horizontal dashes) 01a .7 temporal. 80lh types of numerical dispersion are problematic because they v~
3000
spill in a riveroblainedwith I as a function of step size (bolh I1x and 111)and velocity. 8ecause these (particu-
o 1000 2000
x(m)
the MacCormackapproach
with a Ilt = 0.025hr.
\ larly spatial step and velocity) are not usually constants, this means that a variable
1 numerical dispersion is introduced. 8eyond such accuracy considerations, the prin-
~ cipal shortcoming of the Euler-based approaches relates to stability. In short they
ard dif- '7 extract a harsh execulion-time penalty for increases in spatial resolution.
It should be noled Ihal variations ofMacCormack's melhod use a bac 8y centering its time-derivative approximation, the MacCormack approach re-
( ference for Ihe predictor and a forward difference for the corrector. Further, for more moves all second-order temporal numerical dispersion. In addition, although it is
( conditionally stable, its stability constraint is more liberal than for the simple explic-
\diffusive systems, centcred differences can be employed.
¡ The MacCormack method is a good, general numerical approach fo water-
I quality ca\culations.Firsl, becauseit is a centered-time/centered-space a ,proach,
~ )ear
it methods.
models This
makescombined with itsalternative.
it an allractive ease of programming for both linear and nonlin-

it has good accuracy. Second, because the melhod is explicit. it is relatively easy
( lo prngram. In particular. since the MacCormack method does not involve simulta-
\ neous equations, it handles nonlinear problems in a straightforward
The down side of the approach is thal it is not unconditionally stable. Although
its stability constraint is more liberal than for the simpleexplicit methods (FT8S
or F}'CS), there is a time-step limit beyond which the solulion wi\l blow up. These
and ulher trade-offs among the methods are described next.
13.3 SUMMARV
Tablc 13.1 summarizes all the methods we have described in the previous tWolec-
tures.
Implicit methods. Two implicit methods are commonly used. The simple im-
fashion. ( plicit approach writes all spatial derivatives (the transport terms) at the advanced
( time step. Thus the resulting equations are implicit and solutions must be obtained
by solving simultaneous equations. Although the implicit melhod is uncondition-
~ ally stable, it requires small time steps to obtain accurale solulions for highly ad-
( vective systems. In addition it generales time-step-dependent numerical dispersion.
Although these represent serious deficiencies when generating time-variable solu-
tions, they becorne less important when steady-state solutions are desired. Thus
it is applied with a large time step for such cases, because inaccurate interme-
diate results are not a detriment. In addition the numerical dispersion drops oul
as the steady-state condition is reached. Finally second-order approaches such as

Explicit methods. The primary advantages of all these approaches is that they
are rt"lativelyeasy to code for both linear and nonlinear problems. The simplest ver-
sions are manifestations of Euler's method. Two treatments of the spatial advection
r derivative are usually used: backward (f<1'8S) and centered (FTCS). The former
\ has hoth spatial and temporal numerical dispersion, whereas the laller has only
1
\ Crank-Nicolson are available that do not manifest time-step-dependent numerical
dispersion. However, they are limited to linear models.
Finally it should be noted that all the techniques in Table 13.1 exhibit third-
\ order truncalion errors. These tend lo manifest themselves by inducing oscillations
\ and skewness into rnodel simulations. These shortcomings lend to be significant in
)the simulalion of sharp fronts. Although Ihey have not traditionally been emphasized
 ...
---

232 PARr11 Incompletely Mixed Systems

PART 111
L in water-quality modeling, such fronts should receive future attention in problems
I involving spills and transient flows. Hoffman (1992) discusses the issue. --- - - -...............

¡OBLEMS Water-Quality Environments

13.1. AII cstuary has !he following characteristics:
Depth = 4 m Aow = 100,000m3hr-I
Velocity = 250 m hr-I Cross-sectional area = 400 m2
Width = 100 m Dispersion = 1,000,000 m2 hr I

A spill of 10 kg of a reactive contaminant (k = 0.1 d-I) occurs. Use the (a) simple
implicit, (b) Crank-Nk(,lson, and (e) MacCormack methods to simulate the distribution
of the contaminant in the estuary.

13.2. A mixed-now reactor is 10 m long, 2 m wide, and I m deep and has the following
hydraulic characteristics: E = 2 m2 hr-I and U = 1 m hr-I. It initially contains pure
water. At t = O it is subject to a step loading of a chemical, ein = 100 mg L-1, that
decays by lirst-order kinetics k = 0.2 hr-I. Determine the distribution of the chemi-
cal along the length ofthe reactor as a function oftime using centered spatial differences
andthe
(a) simple implicit method
Pan J// is designed to introduce you to some ofthe environments commonly encoun-
(b) Crank-Nicolson method tered in water-quality modeling. Most of the l1lall:rialdeals with physical aspects of
(e) MacCormack method these environments. Just as Lec. 2 served as an introduction to reactions, this part
provides information and modeling techniques that have relev¡mceto the other major
13.3. Calculate how a 2-kg spill of a conservative substance would move through a 20-km facet of water-quality modeling-transport.
stretch of Boulder Creek as described in Probo 12.2 using the
The first three lectures describe the three major elasses of natural waters encoun-
(a) simple implicit method
tered in water-quality modeling: rivers, estuaries, and lakes. Lecture 14 deals with
(b) Crank-Nicolson method
(e) MacCormack method characterizing flows and mixing in rivers and streams. This ineludes using data to es-
timate model parameters and an introduction to how the hydraulics of such systems
are simulated. Lecture 15 performs a similar function for estuaries and tidal rivers.
Emphasis is on characterizing mixing due to tidal motion. Lecture 16 introduces as-
pects of lake and reservoir physics. Aside from describing lake water budgcts, this
lecture ineludes additional models to assess the distribution of pollutants in the near-
shore zone. This lauer material can also bc applied to the other watcr bodics (that
is, wide rivers and estuaries) as well as to pollutant discharge in the oceanic coastal
environment.
Lecture 17 describes an additional environmcnt that is common lo mm,1water-
quality modeling comexts-sediments. Information is provided on both suspended
and bouom sediments.
Finally a very different type of environment is explored in Lecture IB-the
"modeling environment." This lecture deals with issucs that arc commonly cncoun-
tered in applying models, and ineludes information on model calibration, verilica-
tion, and sensitivity analysis.

'"
 LECTURE 14

Riversand Streams

As described in Lec. 1, water-quality modeting originally focused on river and

stream pollution. Because they provided drinking water and navigation, rivers were
the sites of many large cities and their associated waste discharges. In addition
to bearing the brunt of urban point-source pollution, rivers also provide water for
other uses such as agriculture, industry, and power supply. Each of these uses can
potentially affect water quality.
Beyond their utility, eIean rivers and streams are beautiful. Anyone who has
ever meditated at the foot of a waterfall or strolled along the banks of an unpolluted
creek can appreciate their aesthetic value and their contribution to the quatity of tife.
Consequently rivers and streams are still the site of great environmental con-
cero. This lecture is designed to provide you with background informadon on these
systems. In particular we focus on aspects of river transport that havea bearing on
water-quality modeting.

14.1 RIVER TYPES

Rivers can be eIassiljed in many ways. In the present context we are primarily con-
cerned with their hydrogeometric properties. That is, to model rivers, we must know
their hydrautic (f1ow,velocity, dispersion) and geometric (depth, width, slope) pa-
rameters. Table 14.1 ineludes a size range of rivers and streams along with some of
their hydrogeometric parameters.

235

----.-
'.

236 PART11I Water-Quality Environments l.ECTURE14 Rivers and Streams 237

TABLE 14.1 100,000

Hydrogeometric parameters for a range of rivers ordered by flow (Fischer et al.
1979) 1,000
-
Mean 10
depth Width Velocity Flow Dispersion 0.1
River (m) (m) Slope (mps) (mJ S-I) (10' cml S-I)
Missouri 2.70 200 0.00021 1.55 837.0 150.00 0.001
Sabine 3.40 116 0.00013 0.61 254.6 4<;.30 Hlcl 3o-Jan l-Jun 1-0cl
WindylBig Horn 1.63 64 0.00135 1.22 144.1 10.10 1979 1980
Yadkin 3.10 71 0.00044 0.60 140.1 18.50 (a) Hudson River al Green Island, New York
Clinch, Tennessee 1.68 53 0.00054 0.70 74.5 3.83 (walershed area = 20,950kml)
JohnDay 1.53 30 0.00239 0.92 41.8 3.95
Nooksack 0.76 64 0.00979 0.67 32.6 3.50 100,000
Coachella Canal, California 1.56 24 - 0.71 26.6 0.96
Bayou Anacoco 0.93 32 0.00050 0.37 10.9 3.60 1,000
Cinch, Virginia 0.58 36 - 0.21 4.4 0.81
Powell, Tennessee 0.85 34 - 0.15 4.3 0.95 10
Copper, Virginia 0.56 17 0.00130 0.32 3.6 1.51
Comite 0.43 16 0.00059 0.37 2.5 1.40 0.1

.-.
0.001
l-ocl 3o-Jan l-Jun l-ocl
Aside from size, Ihe other importanl aspecl of slream physics is Ihe lemporal I 1978 1979
ti)
distribution of flow. This lime series is quanlified by a plol of flow rale versus lime
!>fyear that is calIed the annual hydrograph. Figure 14.1 shows four lypeS.
The firsl (Fig. 14.1a) is for a stream in a lemperate cJimale with aperennial flow
-
C')

E
== 100,000
(b) Mili Creek near Belville, Texas
(walershed area = 974 kml)

regime. This type of flow paltern is typical of many of Ihe streams in the easlern and O
midwestern Uniled Slates, where waler-qualily modeling was firsl developed. NOle ¡¡: 1,000
that most of Ihe peak mnoff occurs during the spring when heavy precipilation and 10
some snowmelt occur. More important, from Ihe standpoint of early waler-qualily
modeling, is the occurrence of a relative1y slable low-flow period during the warm 0.1
( summer mOl1ths.Such warrn, low-flow periods are usualIy Ihe most crilical from a 0.001
1 waler-quality perspective. Thus Ihe low-flow period provided a stable design conlext 1-0ct 3o-Jan l-Jun 1-0cl
Ifor early wasteload alIocalion sludies. 1978 1979
Figure 14.1b shows a hydrograph for a perennial slream, bUIin a more humid
(e) Easl River al Altmonl, Colorado
climate typical of the soulheaslern Uniled Slates. The spikes are due to rain Slorms, (watershed area =748 kml)
whereas the slowly varying underlying trend, calIed basejlow, is due lo ground-
water contribulions. For Ihis case Ihe difference between summer and spring is not 100,000
as pronounced because of the more even dislribulion of precipilalion over the year
1,000
and the absence of snowmelt. Although Ihere seems lo be a low-flowsummer period,
it is more punctualed by storrn events. 10
Figure 14.1c shows a hydrograph for a river wilh a significanl amounl of
0.1
snowmelt. The bulk of Ihe peak flow occurs because of heavy snowmelt in the
early summer. The release of water from baseflow and residual slowly melting snow 0.001 FIGURE 14.1
results in a sleadier distribution over the remainder of Ihe year. l-ocl 3o-Jan l-Jun l-ocl Annual straamllow hydrographs
finally the hydrograph in Fig. 14.ld is an example of an ephemerallJow regime 1978 lor (a) parannial tamparata rivar,
1979
for a stream in an arid cJimate. For this case there are significanl periods when the (b) parannial humid rivar, (e)
riyer is dry. The periods of 1J0ware due to large storms over Ihe basin. (ti) Frlo Rlver near Uvalde, Texas snowfed river, and (d) aphameral
(walershed area = 1712 kml) river.
 238 I'ART11I Water-Quality Environments LECTURE14 Rivers and Slreams 239

Human inftuence can further modify hydrographs in several significanl ways:

Impowlding. The regulalion of slreams by dams lends lo moderale lhe seasonal
swings in ftow.This also lends lo shift lhe liming of minima and maxima. For
example whereas lhe unregulaled river mighl have peak ftows in lhe spring,
the peaks could be caplured by impoundmenls and then released laler in lhe (a) Stream cross sectlon FIGURE 14.2
summer for agricultural use. (a) A stream cross section along
Urhanization and channelization. Urban drainages lend lo have more of lheir with (b) a transect showing depth
x
land area covered by impervious surfaces such as pavemenl and parking and velocity measurements
101S.Consequently runoff becomes more pronounced and concenlraled be- needed to calculate mean depth,
cause liule rainfall will be absorbed by soil and vegelalion. Further, slorm flow, and other hydrogeometric
sewers, which are conslrucled lO convey lhe waler off lhe slreels rapidly, parameters. Note that lhe velocity
measurements (x) are taken at
result in speedy delivery lo receiving walers. In addilion many slreams 60% depth 101'shallow water and
have been channelized so lhal lhey would lransmil high ftows more effi- (b) Depth and veloclty measurements
at 20% and 80% 101'deep points.
cienlly and prevenl flooding. AlIlhese faclors conlribule to making spikes
in lhe annual hydrograph more pronounced. Moreover ftow during lhe dry
season is usually lowered as groundwaler sources become less predom- ., (14.1)
inan!. Ac = I: Z(x) dx
Human water lI.fe.Many rivers are used eXlensively for irrigalion. During lhe
growing season lhis can lead lo high lemporal and spalial variabilily as farm- I H = Ac
B (14.2)
ers lake and discharge waler. Slreams lhat are used for hydropower 01'cooling
waler also exhibil fluclualionsdue lo associaled waler use. In conlrasllO olher where Ac = cross-seclional area (m2)
uses, power-associaled wilhdrawals and releases lend lo exhibil a more pe- x = distance measured across lhe stream (m)
riodic nalure. This is so because human power needs lend lo manifesllhem- z(x) = depth measured allocatíon x (m)
selves on prediclable daily, weekly, and annual cycles. H = mean depth (m)
B = slream widlh
In slImmary slreams exhibil a greal deal of flow variabilily. For some syslems
r, At each deplh measuremenl a corresponding velocily rneasurernent can be ta-
lhe cIassical pauern of low, warm summer ftow slill provides a sound design context
for evalualing pollulion conlrol, particularly poinl sources. However, olher syslems, ken. There are several ways lo do lhis (Gupta 1989). The rnost cornrnon are:
because of lheir highly lransíent nalures, human effecls, and lypes of pollutíon prob- l. For deeper waters, deplh > 0.61 m (2 ft), the velocities taken at 20% and 80% of
lems (rOl'example, nonpoinl runoff), require a relhinking of lhis perspeclive.ln mosl the lotal deplh are averaged.
of lhe remainder of this leclure we deal wilh issues relevanllo steady-ftow model-
2. For shallower water, deplh < 0.61 m, a single velocity rneasurernenl is laken al
ing. In lhe final sectíon we address lhe issue of lime-variable flows. 60% of lhe depth.
It can be shown (Box 14.1) that these choices have a rnalhernatical basis.
14.2 STREAM HVDROGEOMETRV
A slream's hydrogeometry consisls of ilShydrologic characlerislics (velocily, flow,
dispersion) and its geomelry (deplh, widlh, cross-seclional area, slope). Two ap- BOX 14.1. Mathematlcal Basls o, the Two-Polnt and Slx-Tenths-DepthMethods
proaches are available for delermining lhese paramelers: al a poinl 01'for a re~ch. A typical distribution of velocity with depth is depictedin Fig. 814.1-1. Dickinson
(1967) determined that the distribution of velocity in streams could be lit by the simple
power law
14.2.1 Point Estimates l/m
z~ - z
As lhe name implies, a poillt estimate is evalualed al a specific slream location.
Typically, measurements of deplh and waler velocily are laken al a transecl across
( )
V(z) = Vo -
Z~-lo
(14.3)

lhe slream (Fig. 14.2).The deplh dala are used lo estímate lhe mean deplh and cross- where V(z) = velocity at depth z
seclional area. This is done via integratíon, Vo = known velocity at a depth lo from the surface
 -

240 PART11I Water-Qualily Environmenls I.ECTURE14 Rivers and Sl/eams 241

u -1.0 o
Z=o
-e.6 - 0.57736 0.2zb

0.0
=%0 0.5zb

0.6 0.57736 0.8zb

Z=Zb
1.0
Zb
Z FIGURE 814.1-1
(o) Gauss quadrature (b) Two-point method FIGURE 814.1-2
Zb = deplh of the bouom
m = a constanl that varíes from 6 to lO, depending on Ihe Reynolds number,
and is approximately 7 (Daily and Harleman 1966)

This equation can be integrated lo delermine the average velocilY, Once the average velocity for each point on the transect is derived, V(x), the
lIm
values can be integrated to arrive at the mean fIow
- 1
i (- - )
Zb
1 Zb Zb Z l/m m Zb
U = -
1
ZbO
U(z)dz = -
ZbO
Uo
Zb-Zo
dz = -uo
m+1 (-
Zb-Zo ) (14.4)

i
.., Q = r U(x)z(x)dx
(14.8)

Now suppose Ihat the mean velocily occurs at a distance Z from Ihe bouom. Seuing
Uo = U and Zb - Z = Z in Eq. 14.3 and equating Ihe result with Eq. 14.4 gives Finally the mean velocity can be calculated with the continuity equation

u---
l/m
Z .J'>{ Q
Uo-
( )
Zb-Zo
l/m_
- m U
m+lozb-Zo ( Zb
) (14.5)
I Ac
(14.9)

which can be sol ved for The approach is illustrated in the following example.
m
m
Z=
( ) m+ 1 Zb
(14.6) EXAMPLE 14.1. POINT ESTIMATION.
stream cross section:
The following data were collc.;tcd for a

for values of m belween 6 and 10, Ihis equation can be used lo delermine that
Z ==0.4 Zb. Thus a single measllrcment taken at aboul 60% of the total depth should x (m) O 4 8 12 16 20
provide a good approximm,. "f IhC'average velocity over the entire deplh. z(m) O 0.4 I 1.5 0.2 O
A similar analysis can ¡,. ¡...'It,umed to determine a best Iwo-point approach. For U(X)O.2(mps) O 0.2 0.3 O
this case the mean is obtained by laking the average of the values at 20% and 80% of U(X)06 (mps) O 0.05 0.07 O
the total depth (Gupta 1989). U(X)08 (mps) O 0.12 0.2 O
Interestingly, independent support for the two-point approach comes from the nu- U(x) (mps) O 0.05 0.16 0.25 0.07 O
merícal integration method called Gauss quadrature. If the integration inlerval ranges U(x)z (m2 S-I) O 0.02 0.t6 0.375 0.014 O
from -1 to 1, it can be shown (Chapra and Canale 1988) Ihat an optimal integral es-
limate can be determined "Y .idding Ihe values of the integraled variable at 1/ fj and
11J3. Thus an average value for the interval could be computed as Use Ihis data to determine the (a) area, (b) mean deplh, (e) ftow,and (ti) velo¡;ily.

- _ el I(x) dx _ 1 -1 Solutioo: (a) Because the sampling points are equally spaced, Simpson's rulc~ iCliapra
and Canale 1988) can be used to implement Eq. 14.1. The !
rule can be appl" .1111Ihe
1--1_(_1) -0.51 ( 13) +0.51 ( 13) (14.7) i
firsl two iOlervals and Ihe rule to the last Ihree.
Now by a simple change (lf variable, il can be recognized Ihat -1/ 13 on a domain
from -1 to 1 corresponds tI>.l value of approximately 0.22 Zbon a domain from ° to 12.617m2
Ac = (8_0)0+4X6°.4+ 1 +(20-8)1 +3(1.5;0.2)+0
Zb.Similarly 1/13 correspomh to 0.78 Zb(see Fig. BI4.1-2). Hence the results support
the strategy of averaging values taken at 20 and 80% ofthe total depth (Fig. BI4.1-2). (b) Equation 14.2 can theo be used to calculate the meao deplh,
A nice result of the Gauss-quadrature approach is that it is more general than the
foregoing derivatioos. That is, it does not depend on having a power function of Ihe
form ofEq. 14.3. Consequently it should hold regardless ofthe velocily distribulion. /1 = 122~17= 0.6308m
 242 PART11I Water-Quality Environments t.ECTURE14 Rivers and Streams 243

(c) The now can be evaluated by using Simpson's roles to integrate Eq. 14.8 numeri- which ean then be used to detennine the mean depth
cally,

Q = (8 _0)0 + 4 x 0.02 + 0.16 +(20 _ 8)0.16 + 3(0.375 + 0.014) + O

,,
1
Ac H
=B (14.12)

6 8 EXAMPLE 14.2. REACH ESTIMATION OF VELOCITY AND MEAN DEPTH.

I
= 2.3105 Suppose thal the point estimale calculated in Example 14.1 is allhe downstream end
i m~ s I
of a 2-km reach with a mean width of 22 m. Recall that the point estimate of ftow was
(d) The mean velocity can then be calculateu as (Eq. 14.9) 2.3105103 S-l. You perform a dye study and determine that it takes 3.2 hr for the dye
I
to traverse the 2 km. Use the reach approach to determine the velocity, cross-seclional
area, and mean deplh for the reach.
¡ _ 2.3105 _ 103 1 -1
i ¡
U - 12.617 - O. o. 80S Solution: The mean velocity can be calculated as (Eq. 14.10)
I
I I
The above procedure can be repeated for a number of Iransects along a stream.
t ?
U = 2 km .J..!!... 1000m = 0.1736m sol
(
3.2 hr 3600s km )
1

¡
I
The dala can then be plolled versus slream lenglh and the average segment param- The velocity and ftow rate can then be used to determine the average cross-sectional area
elers delerll1ined(Fig. 14.3). according lo Eq. 14.11,
¡

2.3105 2
Ac = 0.1736 = 13.3 m
14.2.2 Reach Estimates
which can then be used to determine the mean depth (Eq. 14.12),
A reac¡' approaeh provides an alternative to point estimales. This method is usuaIly
) predicaled on Ihe generaIly valid assumption thal the str.¡:am width is less variable
1 than its depth. If this is lrue a stream reach is identified that has a relatively constant H = Ig;3 = 0.605m
\ widlh. After measuring this width Ihe flow al the end of the reaeh is determined by
a pointestímate.In additiongagingstalionsmay alsoprovidethe sameestimate.
NeXIa lravcllime is delermined by injecling a lracer sueh as a dye at the head In general the reaeh approach is particularly effective for shaIlow systems. This
) end and liming how long it takes lo traverse Ihe reach. The mean \iélOcityfor the .,. is because the width is usuaIly much less variable than the depth for such streams.
1 In contrast the transect method is more commonly applied to deeper systems.
1 reach can then be calculaled as .

U=~t (14.10)
14.3 LOW-FLOW ANALYSIS

I wherex = reaehlenglhandt = lraveltime. As mentioned in this lecture's introduction, traditional water-quality modeling has
The velocity and flow rale can then be lIsed to delermine the average cross- used the steady,low-ftow summer period as its design condition. Therefore the ques-
seclional area tion arises: How do you establish the low-ftow condition?
~ The minimum 7-d flow that would be expected to occur every 10yr is generaIly
( Al' =g
U
( 14.11) ( accepted as the slandard design ftow.Called the 7QIO, this value can be detennined
,.. \ for the entire year, by month, or for a season. For the present discussion we limit
ourselves to a hYPolheticalsummer period consisting of the months of luir through
.. . September.
The first step in computing the 7QIO is to obtain a long-tenn flowreeord for the

/l
-r---r \l j ~ location being modeled. For our case we would examine the data for each year to
~ detennine the smallest ftow that occurs for seven consecutive days in the summer

b FIGURE 14.3 , periodo

~---:---:---:- Dala Irom individual Iransecls Next we would tabulate the n ftows in ascending order and assign them a rank
. . __8 . L_
------- can be plotted versus dislance
lo develop average values 01
m. The cumulative probability of occurrence is given by
m
x paramelers lar reaches. p = N+ 1 (14.13)
 ¡ ;

244 PART11I Waler-Qualily Environmenls LECTIJKE14 Rivers and Streams 245

The recurrence interval is then defined as 14.4 DISPERSION AND MIXING

1
T =- (1~1~ In rivers, we examine two mixing regimes. First. for one-dimensional modcls, we are
P concerned with mixing in the direction of tlow or longitudinal mixing. This process
is parameterized by a dispersion coefficient. In addition we are also interested in the
Probability paper can be used to pick off the 7QIO as depicted in Fig. 14.4 and
mixing across the slream or lateral mixing. In this case we would like to rtssess our
described in the following example.
assumption that point sources are instantaneously mixed in this dimension. Therefore
EXAMPLE 14.3. CALCULATIONOF 7QI0. The following7-d low ftowswere !he goal here is to quantify the longitudinal ftow length required to aUain lateral
compiJed for a river:
mixing.

1971 1.72 1976 4.23 1981 4.48 1986 5.39

1972 3.03 1977 4.11 1982 3.03 1987 3.00 14.4.1 Longltudlnal Dlsperslon
1973 2.76 1978 1.92 1983 2.84 1988 2.50
1974 1.65 1979 2.14 1984 3.66 1989 2.47 Severalformulasareavailabletoestimatethelongitudinaldispersioncoeflicientfor
1975 2.00 1980 1.48 1985 1.87 1990 3.07 streamsandrivers.Fischeret al. (1979)havedevelopedthe following:

Use this data 10 determine the 7Q 10. E = 0.011 ~2B2 (14.15)

J-1 HU'
So.utlon: The data can be ordered and labulated along wilh Ihe cumulalive probabilily
of occunence and the recunence interval: where E has units of (m2 s-I)
U = velocity (m s-I) .
Rank Flow
Recurrence
Interval Rank Flow
Recurrence
Interval
B = width(m)
Probabillty Probability H = mean depth (m)
1 1.48 4.76 21.00 11 2.114 52.38 1.91 U' = shear velocity (m S-I), which is related lo more fundamental
2 1.65 9.52 10.50 12 3.00 57.14 1.75 characteristics by
3 1.72 14.29 7.00 13 3.03 61.90 1.62
4 1.87 19.05 5.25 14 3.03 66.67 1.50
5 1.92 23.81 4.20 15 3.07 71.43 \.40 U' = JgHS (14.16)
6 2.00 28.57 3.50 16 3.66 76.19 ql
7 2.14 33.33 3.00 17 4.11 80.95 1.24
I I
8 2.47 38.10 2.63 18 4.23 85.71 1.11
where g = acceleration due to gravity (m S-2) and S = channel slope (dimension-
less).
I I 9
10
2.50
2.76
42.86
47.62
2.33
2.10
19
20
4.48
5.39
90.48
9524
1.11
1.05 Table 14.1 contains measured values of longitudinal dispersion for several
I I
streams and rivers. Figure 14.5 shows a comparison between these measurements
This data can be ploued on normal probabilily paper. As in Fig. 14.4 a value of
I I aboul 1.66 cms conesponds to the 7Q 1O.
FIGURE 14.5
I
6
. 10,000
Comparison 01 measurements
(Table 14.1) with predictions

....
using Eqs. 14.15 (.) and 14.17
(.). The solid line represents a
_ 4 ~ per/ect correlation between the
¡"
Q/¡
- ........ i predictions and the measurements,
2 ..... FIGURE 14.4
Q. whereas the dashed lines
correspond to a lactor 01 5 above
and below the 1:1 correlation. Note
Frequency curve corres- that some additional data beyond
!I ponding to the 7-d minimum """
10 25 50 75 90 99 10,000 Table 14.1, Irom some laboratory
Ilow lor Ihe river Irom !lumes and canals (Fischer et al.
Probablllty Example 14.3. Measured 1979), have been includad on this
plol.
 246 PART11I Water-Quality Environments l.ECI'URE14 Rivers and Streams 247

and values predicted with Eq. 14.15. As can be seen the predictions are within a u = 0.3 mps
factor of 5 of the data. B = 15m
H = 0.4 m
McQuivey and Keefer (1974) proposed the altemative formula f
\'I '. - Q (14.17)
S = 0.004
Q = 1.8cms
,I E - 0.05937S8
Use this data to determine the longitudinal dispersion coefficient and the length required
¡ to attain complete lateral mixing.
I where Q = mean flow (m3 s-I). They studied rivers with flows ranging from 35
to 33,000 cfs. They limit this formulation to systems with Froude numbers (F = Solution: Substituting Ihe values inlo Eqs. 14.16 and 14.15 yields
I
UI Jiii) less than 0.5. We have also included predictions from this model on Fig.
14.5. As with the Fischer formula (Eq. 14.15) the predictions are approximate to a U' = J9.8(0.4)0.004 = 0.125ms-1
range of plus or minus half an order of magnitude.
E = 0011 (0.W(15)2 = 4 45 2 -1
. 0.4(0.125) . m s
The McQuivey and Keefer formula (Eq. 14.17) can also be applied to calculate
14.4.2 Lateral Mlxlng
Lateral mixing of point sources is the second facet of mixing relevant to one- E = 0.059370.~~15) = 1.78 m2 SOI
dimensional stream water-quality modeling. Fischer et al. (1979) have developed
the following formula to estimate the lateral or transverse dispersion coefficient for Thus for this case the estimates differ by about a factor of 2.5.
a stream: The mixing length can be determined as (Eqs. 14.18 and 14.19)

(14.18) E/al = 0.6(0.4)0.125 = 0.03 m2 S-I

, E/al = 0.6HU.
152
where H = mean depth (m) and U. = shear velocity (mps). This value can then LnI = 0.4(0.3) 0.03 = 898 m
be used to compute the length required to attain complete lateral mixing. For a side or with Eq. 14.21 as
discharge, the formula is
(15)2
(1 Lm = 0.4U E
/al
_82 ,
, (14.19) LIn = 8.52(0.3) 0.4 = 1438m
Conseqllently these reslllts indieate that it would require about I km of distance
For a discharge in the center of the channel, to reach complete mixing. In actuality a USGS gaging station is located just below the
outfall. A weir at this point constricts the flow severely and cffcctivcly results in complete
82 mixing at that poin!.
(14.20)
Lm = O.IU E/al

Yotsukura (1968) has proposed the following altemative formula for a side dis-
charge: 14.5 FLOW, DEPTH, AND VELOCITY
82 Suppose that you have a stretch of river that has constant dimensions and bottom
(14.21)
Lm = 8.52U¡¡ I slope. Ir a steady flow rate Q enters the upstream end for a sufficiently long period
\.. ) ~ of time, the water wil\ reach a steady, uniform-flow condition. That is. the flow wil\
where Lm = mixing length (m) be constant in lime and space.
U = velocity (mps)
Now what aboul the depth and velocity of the water? They wil\ also be constant
8 = width (m) and are relaled to flow by the continuity equation
H = depth (m)
1 Q = UAc (14.22)
EXAMPLE 14.4. LONGITUDlNAL DlSPERSION AND LATERAL MIXING. where A,. = cross-sectional area and U = mean velocity.
The following data applies to Boulder Creek, Colorado, immediately below the Boulder Equalion 14.22 represents a dilemma. It says Ihat the flow can manifest it-
waslcwalcr trcalmcnl planl dischargc:
self in two ways: as deplh and as velocity. Clearly additional information is needed
 248 PA".¡ \\.II,,-Quality Environments IH'OUIW 14 River. and Str.:ams 249

TABLE 14.2 TABLE 14.3

Average values and ranges or exponents The Manning roughness
in hydrogeometric correlations coefficient ror various open
Correlatlon Exponent Value Range
Material
---
channel surraces (Chow 1959)
n
Velocity-flow b 0.45 0.3-0.7
Depth-flow {J 0.4 0.1-0.6 Artificial channels:
Width-flow / 0.15 0.05-0.25 Concrete 0,012
Gravel bouom with sides:
Concrete 0.020
Monared stone 0.023
lo delennine bOlhquantilies. As described nexl Ihis infonnation is handled in two Riprdp 0.033
ways: discharge coefficients and Manning's equation. Natural stream channels:
Clean. straight 0.030
Clean, winding 0.040
Weeds and pools. winding 0.050
14.5.1 Dlscharge Coefflclents 0.100
Heavy brush, limber
Power equationscan be used to relate mean velocilY,deplh, and widlh lo ftow
(Leopoldan~ Maddoc~!~~3),
u = aQ!> (14.23) The Manning fonnula can also be substituted into the conlinuity equation to ,:ornpute
ftow,
H = aQP (14.24)
j B=cQf (14.25) .¡. I Q = C" AcR2I3S~/2 (14.27)
) n
where H = mean depth and a, b, a, f3, c,J are empirical constants that are deler- lf ftow is given and Ihe area and hydraulic radius can be expressed as fllnctions
, mined from stage-discharge rating curves (log-Iog plots between the parameters). of depth, Eq. 14.27 has one unknown-{jepth. Consequently il can be sol ved as a
) Because these parameters are interrelated, the coefficienls are not tOlallyinde- roots problem. Depth can then be used to determine Ihe cross-sectional area. Finally
pendent. For example because Q = UAc and Ac = BH, the sum of the exponents ~ deplh. can be used in conjunction with ftow lo ca1culate velocily with Ihe cuntinllity
should be equal to unilY,b + f3 + f = l. Table 14.2 outlines values and ranges for 1 equatton.
the exponents. To illuslrate this method, suppose thal the geomelry can be representell as a
trapezoid (Fig. 14.6) with constant dimensions and bottom slope. '

Because we are dealing wilh a Irapezoid, Ihe cross-sectional area a..d Ihe hy-
1...0.2 The Mannlng Equatlon draulic radius can be expressed as a function of depth,
The Manning equation, which is derived from a momentum balance for Ihe channel, Ac = (Bu + sy)y (14.28)
provides a means to relate velocity to channel characleristics, P = Bu+ 2y H+I \ 14,29)

u = Co R213st/2 (14.26)
~ n · R = Ac = (Bu + sy)y ( I UD)
where Co = a constant (= 1.0 for metric and 1.486 for English units)
P Bu+ 2y g-:¡:¡
I n = Manning's roll!!hness coefli..i,~nt
(Table14.3)
l R = channel'shy I l.;1..radius (m or ft) = AclP
t ) P = wetted perimeter (m or ft)
I s. = slope of the channel's energy grade linet (dimensionless) FIGURE 14.6
A cross section 01 a Irapezoidal channel
showing the paramelers needed lo uniquely
INote!hat sincc we have assumed that flow is steady and the cross section is constant, the energy slopc deline Ihe geomelry: So = bottom widlh,
So s = side slope.
is cquallo the channel slopc.
1 7

"
~'o

. c'
250 I'ART111 Water-Quality Environmcnts LECTURE14 Rivers and Streams 251

Equalions 14.28 and 14.30 can be subsliluled into Eq. 14.28 to yield "'.
on a frequent basis during the growing season. Another relevant example might be
I [(80 + sy)y
2I3S,
¡
5/3
1/2 (14.31 )
f the tailwater below a power-supply reservoir. In such cases flows often vary on an
, hourly basis due to changes in demand for electricity. Transport could be an ex-
Q = -;; ~ tremely significant determinant of the water quality for such systems.
(
Bu + 2)' ../s2 + I )
If Q is given, Eq. 14.31 is a non linear equalion with one unknown, y. Thus this
14.6.1 Routlng Water
equation can be reforlllulaled as a mols problem,

·
5/3

f( 1') = ~_'
I( 8 u + sy ))' 1

2/3 SI/2 - Q (14.32)

There is a spectrum of models that are available to simulate dynamic water move-
( ment for streams. AII are based on the Sto Venant equations. These result from tak-
. Il ~
(8u + 2)' ../s2 + I ) )lateral
ing mass and momentum balances on a one-dimensional channel. For the case where
inflow,wind shear, and eddy losses are omitted, the results are
The root (that ¡s, the value of)' that makes this equation zero) is the reach depth.
Contilluity equation (mass balance):
EXAMPLE 14.5. THE MANNING EQUATION. A ehannel has the following ehar-
l,i
aeteristies: aQ + aAc = O (14.33)
ax at
Flow = 6.25 ems
Channel slope = 0.0002 Momentum equation (momentum balance):
BoUom width = 10m
=2 ay
Side slope
~ .!. aQ + .!. ~ Q2 (14.34)
Roughness = 0.035 Ac al Acax (Ac) + gax - g(So - SI) = O
Determine the depth. cross-sectional area, and veloeity for the channel.

Solution: Substituting values into Eq. 14.32 yields As depicted in Fig. 14.7 the momentum balance can be simplified by dropping
terms. In this section, we develop a solution for the simplest form, the kinematic
f(y) = 0.4041 [( 1O + 2y)yf/3 - 6.25 wave. For this case the pressure and acceleration force terms are dropped and the
(10 + 4.472y)2I3 following pair of differential equations result:
whieh can be solved for y = 1.24 m. This result can be substituted into Eq. 14.28 to
} aQ
ax
+ aAc
al
= O (14.35)
compute \

Ac = [10 + 2(1.24)11.24 = 15.5 m2 1) So = SI (14.36)

which. along with the now rate, can be substituted into the continuity equation to deter-
mine

Q - 6.25 = 0.403 m S-I

U = Ac - 15.5
~a Q+~~ Q2
+ ~_
Ac at Ac a x ( Ac ) g ax g (So- SI) =o
( Computer packages, such as EPA's stream model QUAL2E, use the Manning
equation lo model steady f10win streams. To do this the stream is broken into reaches Local Convectlve Pressure Gravlty Frlctlon
acceleratlon acceleratlon force force force
) that consisl of constant hydrogeometric properties. Then Manning's equation is u~ed
( to determine the velocity and depth for each reach. This information is then uséd in 1+-- Klnematlc
wave
Imass balances for individual pollutants.
¡. Dlffuslon wave

1- Dynamlc wave
14.6 ROUTING AND WATER QUALlTY (ADVANCED TOPIC)
FIGURE 14.7
Although the Manning approach is adequate for determining velocities and depths
The SI. Venan! equation lor momen!um can be simplilied by dropping
for streams with steady f1ows,Ihere are other cases where flow changes temporally. !erms as shown (redrawn Irom Chow et al. 1988).
For exmnple in agricultural regions. farmers withdraw and release irrigation water
f
.! ..... "

252 PART11I Waler-Qualily Environmenls

L.I:l"rUKE14 Rivers and Slremns 253

The continuity equation merely expresses the idea that if the area is increasing
EXAMPLE 14.1,. KINEMATIC WAVK Supposc Ihal wc have a reclangularchannel
temporally at a point in space, then the flow rate passing this point at that time must Ihal has Ihe following characteristics
, Jecreasing spatially to account for the water that is stored in the increasing cross
Flow = 2.5cms
.,In.The momentum equation states that the accelerating effect of slope is exactIy
ed by the retarding effect of friction. Channel slope = 0.004
Bottom widlh 15 m
JW Manning's equation can be incorporated into the scbeme by substituting
Roughness = 0.07
So = SI and R = AclP,
1 A513 For simplicity we assume Ihat the width is much greater than Ihe dep!h, and therefore

\ Q = _n -LS1f2
p213 o
(14.37) Eq. 14.40 holds. At x = O the flow increases and then decreases
release. The excess flow Q, can be approximatedby Ihehalf-sinusoid,
due to a pnwer-plant

which can be sol ved for Q, = 2.5 sin wl O S t s 0.25 d

Q. = O t > 0.25 d
Ac = aQfJ (14.38)
where w = 2'IT(0.5 d)
where f3 = ~ and
a = n p2I3 ,15
(14.39)
6
(a)
{J ( }S;j .....
(1)
If the channel has uniform, ideal geometry Eq. 14.39can be simplified further. E
U
4
For example for a rectangular channel that is much wider than it is deep, it becomes "-
CI 2
a = nB213 ,15
(14.40)
~ ( }S;j o
o
Thus for this simple case, a is a constan\.
Next Eq. 14.38can be differentiated with respect to time,
0.8 T(b)
oAc ..... 0.6
(14.41)
al ot
= af3QfJ-1oQ
(1)
Q.
E 0.4
"-
which can be substituted into Eq. 14.35 to yield
0.2

i ~
oQ
ox +af3Q/HOQ
ot =O (14.42) o
o
l Thus tbese manipulations bave generated a single differential equation with one un-
(known, Q, that is the equivalent of the original two equations (14.35 and 14.36). 0.6 T(C)
A numerical solution can be generated by substituting forward-timelbackward- 0.6
space differences into Eq. 14.42 to give ..... FIGURE 14.8
Simulation 01 a wave
Qt - Q Qt+I - Q
t t .§. 0.4
ti: moving through a
1 ; ;-1 + af3(Q~)fJ-I;; (14.43) channe!. The boxed
.:1x I .:1t 0.2
numbers represent
wlUch can be sol ved explicitly for the amount 01time (d)
O I 1 I I I I I I I lollowing the onset 01 the
1 (Qnl-fJ O 20 40 60 80 wave. Also the dashed
Qt+1 = Qt + Qt_ _ Q~l.:1t ) (14.44)
~1 I I {[ af3.:1x
] (I 1 I f x (km)
lines ¡ndieate the peak
values.
 252 'ater-Quality Environments LECTURE14 Rivers and Streams 25~
tJ.
'"
~the steady-state dep!h and velocity prior to the power-plant release.
v _ 2.51-0.06
he explicit numerieal method to prediet how !he depth and !he velocity e - 3.14(0.6)= 0.766ms-1
.ower-plant ftow propagates down the ehannel.
Ve _
- _.51-0.6
.._ __= 1.01ms-I
",olullon: (a) Substítuting values into Eq. 14.40 yields
315
213
a =
[0.07(15)
jO.OO4 ]=
Next Eq. 14.38 ean be solved for the sleady-state area,
3.14 Now suppose that we approximate the wave as a half-sinusoid. The average velocity ean
be ealculated as

- _ {OO.25[O.766
+ (1.01 - 0.766)sin(wt») dt _ 92 _1
Ve - 0.25 - O. m s
A,. = 3.14(2.5)°.6 = 5.44 m2
Inspeetion of Fig. 14.8 indieates that the peak ftow is at 6500 and 71,500 m at t =
whíeh can he used to ealculate depth and velocity: 0.2 and 1.0 d. respeetively. These values ean be used to determine an average eelerity of
J -1
= 2 .5 m' s
(~86.400s)= 094. ms
-1 = (71,500 - 6500) m = 81 250 d-I -1
= O46 V
V
5.44 m2 . ms e (1.0-0.2)d . m
whíeh eorresponds c10sely to the eomputed average. Also note that !he water velocity
5.44 m2 = 0.3627 m
H = l5ñI ranges between about 0.46 and 0.6 m S-I (Fig. 14.8). Thus the eelerity is approximately

(b) Using time and spaee seales of 0.007 d and 1 km, respeetively, we ean use Eq.
I i 1.67 times larger than the water velocity.

14.44 tn generate the solution shown in Fig. 14.8. The solution in Fig. 14.8 exhibits
some numerieal dispersíon. Thís would not be unexpected beeause of our use of the
forward-timelbaekward-spaee seheme. More refined numerieal methods ean be used to
J11itígate this effee!.
~ The foregoing example shows that the wave and water velocities are different.
Their relative magnitudes can be evaluated for the simple case of a wide, shallow
rectangular channe\. For such a channe1, the hydraulic radius can be approximated
by the mean depth

Before using the foregoing mode1to simulate pollutant transport, I want to point (
) R_ Ae _ By _ By _
=--y (14.48)
out an important feature of the kinematic wave mode\. To do this recall that the I P 2y + B B
advection masg.balance equation for a conservative pollutant can be written as
Then the Manning equation can be used to represent the velocity and ftow as
ac = _Vac 1
at ax (14.45) V = _y2l3SIn.
n (14.49)

'.vc can rearrange Eq. 14.42 so that it is in the same format,

l
aQ and Q = _ByS/3SIn. (14.50)
- VeaQ (14.46) n
at ax
Next Eq. 14.50 can be differentiated fo'give
whcrc
aQ = ~ !By2l3SIn.
Ve = -
QI-P
a{3
= -
aQ
aAe
(14.47)
l

¡ ay (
3 n ) (14.51)

Recognizethat for a rectangularchannel,

, Comparisonindicatesthatjust as pollutantmass is propagateddowna channe1at a
velocityV (Eq. 14.45),a ftowwavemovesat a velocityVe (Eq. 14.46).The ve10dty aQ laQ
(14.52)
(,
VI is formallytcrmedthe kinematicwave's celerity. aAe B ay
Thus the celerity is equal to
E:XAMJ'LE 14.7. CELERITY. Determine the celeríty for Example 14.6 and use Fig.
14.8 to VCIil y that the wave moves at this velocity. 1,
Ve = iV (14.53)
Solution: In Example 14.6 the ftow in the channel varies between 2.5 and 5 ems. Thus
the celerity IVould also vary. Substituling values into Eq. 14.47 for each bound yields
In other words the wave velocity propagates approximately 1.6 times faster than
the water velocity. This result confirms the ca1culations in Example 14.7. Although
 -

lECfURE 14 Rivers and Slreams 257

256 PART111 Water-Quality Environments

(a)
( this resull was dcrived for a simple case, the fact that the kinematic wave celerity is 6
Ifaster than the water velocity holds for other geometries. -en 4
14.6.2 Routing PoJlutants
Now we will investigate how the varying f10wscan be used to model water quality.
-E
u
CI 2

A mass balance for a conservative substance with time-varying f10wcan be wriUen O

as O
a(Acc) a(Qc) (14.54)
at ax

A forward-timelbackward-space scheme can be used to approximate the derivatives

- 200 T (b)

(Acc)f+' - (Acc)f _ _ (Qc)f - (Qc)f-,

/:1t - /:1x (14.55) -DI 100
E

To simplify we can multiply both sides by 6.x to give o I

o 20 40
(Vc)f+1- (Vc)f = (Q )~ _(Q ) e (14.56)
60 80

6.t c ,-1 CI
x (km)

Now it can be recognized that the term (V c)f+I can be evaluated as FIGURE 14.9
Simulation 01 (a) a hydrograph and (b)
a dilution wava or "dilutograph" moving
(Vc)f+' = [Vf +(Qf-. - Qf) 6.t]cf+1 (14.57) through a channa!.

This result eanthen be substituted into Eq. 14.56. Because there is only one term at
the advaneed time point e + 1, the resulting equation can be solved explieitly for
The foregoing example demonstrates that poIlution and water waves travel al
t+ I Vfcf+ (Qf Ic[_1 - QfcD 6.t
c¡ = e e f\ (14.58)
V¡ + (Q¡_I - Q¡I6.t f different velocities in one-dimensional streams. For the kinematie wave sludied hcrc,
¡ the poIlutograph lags the hydrograph. A practical implication relates lo water w¡th-

( This equation has a simple interpretation. The numerator represents an Euler- 1drawals duringwave events. Ignorance of the lag might lead a water user to begin
method predietion of the mass in segment i at the next time step. The denominator J.. '; withdrawals
after the passage of the erest of the water wave, only lo Ilnd that the
t ) is an Euler predietion of its volume. The ratio therefore provides an estimate of the ) maximumpoIlution level would oeeur some time later.
It should be noted that Ihe foregoing description of river hydraulics is a ver)'
1 new coneentration.
simplified representation. It was intended to introduce you to how sueh computations
EXAMPLE 14.8. WATER.QUALlTY WITH THE KINEMATIC WA VE MODEL. are performed and some of the issues involved in their implementation. The expan-
SUPI'<lse that the syslem from Example ~4.7 inilially has a sleady concenlration of 100 Sionof the water-quality modeling paradigm to cxplieitly include hydrodynamics is
mg L -1 of a cunscrvalive pollulanl. If Ihe sinusoidal fIow carries no concenlration, eurrently an important area of research. You should consult other sourees sueh as
determine the Iransienl concenlration of Ihe pollulanl in Ihe channel. 1 textson hydrology(Ponee1989, Chow et al. 1988, Bras 1990) or open-channel f1,)W
Solution: The results of applying Eq. 14.58 are displayed in Fig. 14.9. We have included j
(Chaudry 1993) for additional information on routing.
the fIow for comparative purposes. Notice Ihat the concenlration shows an inverse paltern
to the fIow. Thus a dilution "wave" moves through Ihe river.
Although the shapes are similar, notice that Ihe concenlration wave moves slower PROBLEMS
than the fIow wave. In fact the forroer moves at about 60% Ihe velocity of Ihe (alter. This
is due to the difference between Ihe celerity of the wave and the velocity of Ihe waler.
"'.1. The data in Table P14.1 were colIected on Boulder Creek, Colorado, on December
23, 1994. The slope at Ihe sampling location is approximately 0.004. Use Ihe dala
Consequently the evolution of the concentration changes proceeds at a slower rate Ihan
~l those Ibat influence the hydrodynamics. along with appropriate formulas lo delermine (a) cross-seclional area, (b) mean depth,
 258 PART11I Water-Quality Environments
LEcrURE 14 Rivers alld Streams 259

x
-
TABLE P14.1

H V60 V20 V8()

14.5. The data in Table P14.5 were collected on Boulder Creek on December
the kinematic wave model to simulate this data. Assume the following values for the
29, 1994. Use

(ft) (fl) (fps) (fps) (fps) creek's hydrogeometric parameters: B = 12 m, n = 0.04, S = 0.0042, and E = 7 X
1()4 cm2 S-l. Note that your simulation results for KP 5.5 might differ from the data.
O O
3 1.3 0.3
TABLE
6
9
1.25
0.85
0.35
0.48
---- --P14.5
12 1.2 0.46 KP= 0.5 KP = 5.5
15 1.85 0.58 Time Q. Time Ql Time Ql
18 1.85 0.66 (hr) (cms) CJ (hr) (cms) Cl (hr) (cms) Cl
21 1.75 0.68
5.08 0.544 522 5.40 0.716 450 12.75 LQ16 110.,.
24 1.9 0.67
6.88 0.479 597 7.00 0.544 500 13.00 0.994 692
27 2.25 0.4 0.68
7.87 0.479 639 7.50 0.532 500 13.25 0.929 703
30 2.4 0.4 0.62
8.00 0.500 7.75 0.524 508 13.50 0.917 710 f.o t\-t
1. 33 2.1 0.29 0.5
8.42 0.544 665 8.00 0.507 514 13.75 0.906 709
36 1.5 0.37
J9 1 0.25 8.95 0.642 693 8.25 0.490 518 14.00 0.906 707
42
45
1.3
1.2
(1.09
0.13
9.42
10.00
0.726
0.819
'ill!!
696
".. 8.50 0.481 522 14.25 0.883 700
8.75 0.470 529 14.50 0.872 694
48 1 0.14 10.57 0.923 672 9.00 0.453 536 14.75 0.861 684
51 0.8 0.08 10.70 0.942 9.25 0.453 543 15.00 0.841 678
54 O O 1'.....
11.15 (!.22.9 626 9.50 0.470 550 15.25 0.818 674 .. f'I\'l\.
11.67
12.12
0.923
0.887
ill . 9.75 0.481 557 15.50 0.818 678
10.00 0.507 566 15.75 0.818 685
(e) fiow, (d) mea n velocily, (e) longitud in al dispersion coefficient. and (j) distance 12.37 0.853 706 10.25 0.524 575 16.00 0.830 699
needed to auain lateral mixing. 12.75 0.836 712 10.50 0.544 583 16.25 0.818 711
14.2. Suppose a stream channel has the following characteristics: 13.43 0.819 717 10.75 0.617 593 16.50 0.810 720
BoUom width = 10 m 14.00 0.787 723 11.00 0.696 598 16.75 0.810 727
14.67 0.787 723 11.25 0.716 608 17.00 0.798 733
Side sJope = J.5 14.80 0.787 723 11.50 0.779 618 17.25 0.798 736
Channel slope = 0.004 15.23 0.772 723 11.75 0.818 621 17.50 0.798 738
Roughness = 0.07
16.17 0.756 722 12.00 0.872 628
Develop log-Iog plots and power equatíons of depth and velocity versus lIow based on 16.80 0.726 720 12.25 0.929 642
the Manning equation. Use a range of flows from I to 20 cms. 17.23 0.726 12.50 0.982 656
1".3. The following data represent 7-d minimum flows in cms:
14.6. A rectangular channel has the following characteristics:
1977 0.023 1982 0.311 1987 0.091 1992 0.142
1978 Flow = 20 cms Channelslope = 0.0005
0.105 1983 1.331 1988 0.127 1993 0.340
1979 0.024 1984 0.051 1989
Width = 30 m Roughness = 0.025
0.089 1994 0.651
1980 0.187 1985 0.065 1990 0.453 1995 0.017 Determine the depth, cross-sectional area, and velocity for the channel.
19X1 0.010 1986 0.062 1991 0.849
14.7. Repeat Probo 14.6, but for a trapezoidal channel with a bouom width = 30 m and a
side slope = 1.5.
Determine the 7QI0 for this stream.

I tiA. An industrial plant discharges at the side of a stream having the following characteris-
tícs: width = 30 m, depth = 0.9 m, velocity = 0.9 mps, and slope = 0.0002.
(a) Determine the length of stream needed for complete mixing.
(h) If Manning's equation holds and the channel is roughly rectangular, what is the
stream 's roughlless?
(e) Delerllline the longitudinal dispersion coefficient.
 LECfURE15 Estuaries 261

LECTU RE 15

~s~uaries
Rlver
- . ----
Tldal
~
{~¡
---t-
rlver Estuary
/.
.~.t.

Ocean
(a) Bay

-
!.
20

g
... 10

Situated at the interface between rivers and the sea, estuaries are the site of tremen-
dous human development and biotic diversity. From ahuman perspective the role
of estuaries as shipping ports made them the locales of some of the world's great-
esl cilies. From a biolic perspective, estuaries receive targe quantities of nutrients
from their inftowing rivels As a consequence the typical estuarine ecosystem teems
with life.
This leclure is designed to provide you with background information on these
systems. In particular we focus on aspects of estuarine transport that have a bearing
on water-quality modeling.
15.1 ESTUARV TRANSPORT
An estuary is the region where a free-ftowing river meets the ocean. As depicted in
Fig. 15.1 it can be treated as a number of zones based on the interactions of advec-
tion, dispersion, and salinity. At one boundary is the river, which is devoid of sea
salt and in whieh tidal mol,,'1I is negligibte. Next is the tidal river, where the effect
of tides begins to affect transport, but the water is stil1 relatively fresh. This is fol-
lowed by the estuary itself. This region is marked by fulI tidal reversals. The water
here is calIed brackish, meaning that it is a mixture of salt and fresh waters. Often-
times the estuary will widen into a bayoThe bay has a higher salt conct:ntration due to
(b)
FIGURE 15.1
(a) A schematic 01the various zones in an estuarine system. (b) Corresponding salinity
concentration.
its proximity to the ocean. In addition it is dislinguished from the estuary in tha! il
is wider. Thus il must usualIy be modeled as a Iwo-dimensional horii'u, 1syslem.
Also it can somelimes be deep enough thal it must be lrealed as a thrcl !lllIensimml
body. In the bay, the freshwaler ftow is dwarfed by the aetion of lides and winds.
FinalIy the estuarine system ends al the ocean.
There are several aspects of estuarine transport Ihat have direet relevan(;e lO
water-quality modeling. Foremost is the effeet of tidal action. Due to the allraeliun
of both Ihe moon and the sun, tidat motion causes waler lo move in and oul of Ihe
eSluary in a periodic fashion. Thus the ftow, veloeilY,and deplh change ciynamically.
Depending on the seale of lhe probtem being addressed, the tidal 11llIi;"n C¡,IIh.:
perceived as being eilher advective or dispersive. For short-scale pru¡.. ,liSsuch as
the discharge of highly reactive substances or spills, the motion would h.. pcrceiveJ
primarily as adveetion. On a longer time scale, however, the tides would l1Iovewater
back and forth in a cyc1ical fashion and the motion might be c1assified as dispersive.
In this lecture we limit ourselves primarily to the long-term perspective. Thus
we focus on the steady-state eondition averaged over a number of tidat cycIes. The
shorter seale or "real time" aspects of estuarine transport are beyond lhe scope
of this lext. The reader can eonsult other referenees for additional information on

260
262 PART11I Water-Quality Environmenls
LEC11JRE
15 Estuaries 263

15.3 ESTUARY DISPERSION COEFFICIENT

estuarine hydrodynamics (for example Dailey and Harleman 1972; Officer 1976,
1983; Fischer el al. 1979; Lung 1994; elc.). The natural salt gradient can be employed to estimate the dispersion coefficient in
an estuary. The following sections show how this can be done. In the first section the
estuary is assumed to have constant hydrogeometric characteristics. Although this is
15.2 NET ESTUARINE FLOW
not usually the case, the analysis provides insight into the method. This is followed
Net estuarine f10w refcrs to Ihe advcclive movemcnl of waler oul of Ihe estuary over by a more general, finite-difference approach.
a lidal cycle 01'a given number of lidal cycles. For cases where Ihe upstream river
flow dwarfs other waler sources. Ihe nel f10wwould correspond lo the river f1ow.
More comlllonly, where poinl inflows ami diffuse inflows such as surface runoff and 15.3.1 Constant-Parameter Analysis
groundwater are significanl, a water balance musl be developed. A constant-parameterestuary can be representedas in Fig. 15.3.The steady-state
Where Ihe advective flow is large. Ihe tidal flows can sometimes be analyzed massbalancefor thiscase is
lo determine the net f1ow.As depicted in Fig. 15.2 the tidal f10w varialions can be
id(.~tlized as a pair of half-sinusoids. o = _Uds + Ed2s (15.5)
dx dx2
. 7T(f - O)
O :S I :S O + T,. ( 15.1)
Q = q,. Slll [ T,. ]
Using the same approach as was used to solve Eq. 9.26, we obtain the solution for
this case as
7T(t - O- T ) e = FeÁ1x + GeÁ2X (15.6)
Q = qf sin 7T+ f O + T,. :S I :S O + T,. + T¡ (15.2)
[ ] T¡ where the A's are
whcre thc subscriptse and.fdenoleebbamif1ood.
respectively.An averageflowcan U
A, =- E (15.7)
be developed by inlegrating Eqs. 15.1 and 15.2 and dividing by Ihe lidal period,
. 7T(I - O) H+7',+ 7'} . 7T(f - O - Te) A2 = O (15.8)

Q" =
r~7'.
qeSIl1
[ T,. ]
dI +
fH+7',
q¡sll1
[ T¡ ]
dI
The co~stants of integration can be evaluated by the boundary conditions.
T,. + T¡ s = So @x = O (15.9)
(15.3) s=O @x = -00 (15.10)
'''hich can be evalualed lOyield Applyingtheseboundaryconditionsleadsto the solution
2q T - qJT J '

Q" = _7T " f

Te + T¡
(15.4) !Lx
s = soef: for x :S O (15.11)
Although Eq. 15.4can be useful, it should be employed with caution for systems s = So for x > O (15.12)
whcre Ihe nct l10w is small. For sllch cases Ihe numeralor would consist of the dif-
Equation 15.11 provides a model for evaluating the dispersion coefficient. Tak-
lerence belween two very similar, large numbers. Suchdifferencesare notoriously
~cllsilive lo dala error. ing its naturallogarithm yields
In-
s u
(15.13)
So EX
Q I Ebb(+)

E ~

U -x
rr
FIGURE 15.2 O
Simple sinusoidal approximation FIGURE 15.3
01 Ihe tidal flow can be used to Representation 01 a constant dimension estuary as a mixed-
Flood (-) eslimate the net estuarine flow. Ilow system with an ocean boundary condition.
264 PART11I Water-Quality Environments !.IJC'rlJlWI~ Esluarics ~65

Tberefore ifEq. 15.11 holds, a plot ofthe naturallog of sIso versus distance should ~Xk
( 15.15)
yield a straight line with a slope of U/E. If we know U, we can use this slope to aj,k = ~Xj + ~Xk
estimateE. ~X'
and {3J'k = 1 - a J'k = A JAXk (15.16)
, 'Xj +
t:XAMPLE 15.1. USING THE SALINITY GRADlENT TO ESTIMATE DlS-
fERSION. A one-dimensional estuary has constant dimensions and ftow: width = If salinities and flows are known, and this equation is written for an n-segmcnt
300 m, depth = 3 m, and ftow = 15 cms. You measure the following chloride concen- ~tuary, there will be n equations and n + 1 unknown dispersion coefficients. Con-
trations aIong its length (nole Ihal kilometer points increase toward the ocean): sequently one of the coefficients must be specified. In many cases the most natural
choice is to extend the analysis u(::he estuary to the point where dispersion becomes
Kilometer point o 3 6 9 12 15 18 21 negligible. Thus the dispersion across the first interface is set to zero and the equa-
Chloride (ppt) 0.3 0.5 0.8 1.4 2.2 3.7 6.0 10.0 tions can be solved.
To simplify the analysis we assume that we are dealing with equal-sized seg-
Determine the dispersion coefficient for the estuary in cm2 S-l. ments; that is, a = (3 = 0.5. The steady-state salinity mass balance for the tirst
segment can be written as
Solution: The velocity for the system can be computed &S
0= Qo.I(0.5so+ 0.5s¡) - Q1,2(0.5s1+ 0.5S2)+ EÍ,2(s2- s,) (15.17)

U = ¡= 3~3) = 0.016667mS.I (86.~00s) = 1440md-I which can be solved for

A semi-Iog plot of sIso yields
E' - QO,I(SO + sd - QI.2(SI + S2) ( 1518)
1.2- 2(sl - S2)
x (km) Subsequent values can then be computed by solving Eq. 15.14 for
-25 -20 -15 -10 -5
, Q¡_I,I(S¡_I + Si) - Q¡,¡+I(S¡+ s¡+d + 2E¡_1¡(S¡_I - Si)
E¡,¡+I = M '(15.19)
-1
-2 '-'"2 EXAMPLE 15.2. USING THE SALINITY GRADlENT TO ESTIMATE nts-
PERSION. Thomann and Mueller (1987) present the following data from Ihe Wicomico
-3 .5 Estuary, a small tributary of Chesapeake Bay. Use this data lo delermine the dispcrsion
..... coefficients for the system.
FIGURE E15.1

Dlstance
'fhe slope can be converted to 1.667 X 10-4 m -1 and then used to estimate the dispersion (km) Segment Inteñace Area (mz) Flow (m) S-I) SaIlnItYIPI't)
coefficient 1.61 Bay 11.3
O 7.bay 2601 2.83 10.1
E =~ = 1440md-1 104cm2 ~
= Ix 106cm2s-1 -1.61 7 9.0
slope 1.667x 10-4m-I ( m2 86,400s ) -3.22 6.7 2322 2.72 8.0
-4.83 6 7.0
-6.44 5.6 1672 2.66 6.1
-8.05 5 5.2
15.3.2 Flnite-Difference Analysis -9.66 4.5 1393 2.60 4.35
- 11.26 4 3.5
Although estuary reaches can have constant geometry and hydrology, it is more -12.87 3,4 1208 2.55 2.75
-14.48 3 2.0
likely that they will vary. To extend the previous analysis to such cases, we can -16.09 2,3 929 2.41 1.5
develop a finite-difference model of salinity, -17.70 2 1.0
O = Q¡-I.¡(a¡-l,is¡-I + (3¡-I.¡S¡)- Q¡.¡+I(a¡,¡+IS¡+ (3¡,i+ISi+I) -19.31 1.2 836 2.35 0.6
-20.92 1 0.2
+ E¡_I,¡(S¡-I - Si) + ELi+I(Si+1- Si) (1514) - 22.53 0,1 650 2.29 0.2
-24.14 O 0.2
where
266 PART11I Water-Quality Environments LECTURB
15 Estuaries 267
,
Solutlon: Equation 15.18can be employed to compute the bulk dispersion coefficient -..:.... ..f-- -¡-, -..:....
for interface (1,2), O ¡ . ,2 3 ¡ 5
--t- . . -- -r-
E' - 2.29(0.2+ 0.2)- 2.35(0.2+ 1) _ 1 19 3 -1 ~---'r...t .
1.2- 2(0.2 _ 1) - . m s 4
which can be used to calculate a dispersion coefficient of FIGURE 15.5
An overview 01 a simple two-
3.22 km 1000 m lO,ooocm2 dimensional estuary.
E = E,Ax = 119 3 -1 = 458x Hr 2-1
1.2 Al' . m s ( 836 1112 km m2 ) . cm s
The foregoing analysis can be extended to two- and three-dimensional systems
Equation 15.19 can be employed to compute the bulk dispersion coefficient for interface using matrix techniques. An example of a simple two-dimensional system is shown
(2.3),
in Fig. 15.5. If dispersion across the first interface is zero, then mass balances of a
conservative substance can be written as
E' = 2.35(0.2+ 1)- 2.41(1 + 2)+ 2(1.189)(0.2- 1) = ] 157 3 _1
2.3 2(1 _ 2) ., m s
0= QO.I(UO.ISO + 130.1'~¡)- QI.2(UI.2SI+ I3I.2S2)+ E[.2(S2- SI) (15.20)
which can be used to calculate a dispersion coefficient of
O = QI.2(UI.2SI + I3I.2S2)- Q2,3(U2,3S2+ 132.3S»+ Q2.4(U2.4S4 + 132.4S2)
3.22 km 1000 0110.000 cm2 = 109 1,,5
E =] 157 3 -1 2-1 + E1.2(SI- S2) + E~.3(S3- S2) + E~.4(S4- S2) (15.21)
2.J . . m s ( 929 m2 km 012 ) . X \J cm s
O= Q2.3(U, 3'~2 + 132.3S3)- Q3.5(U3,5'~3
+ 133.5S5)+ E2.3(S2- S3) + E3.5(.~5- '~3)
The remainder of the values can be calculated and tabulated. as below. They can also be (15.22)
displayed graphically (Fig. 15.4).
O = W4 + E~.4('~2- S4) - Q2.4(U2.4'~4
+ 132,4S2) (15.23)
Distance E' E These are four equations with four unknowns. However, in contrast to previous si-
(km) Interrace (m3 S-I) (cmz S'I) multaneous equations (e.g.. Lec. 6), the unknowns are not concentrations but bulk
° 7.8 12.29 1.52x 10-' dispersion coefficients. ColIecting terms gives
3.2 6.7 10.65 1.48 x 10-' IdSj{E'} = {Qs} ( 15.24)
6.4 5.6 8.76 1.69x lOS
9.7 4.5 6.38 1.48x lOS where [dS] = a matrix of concentration differences,
12.9 3.4 4.37 1.16 x lOS
16.1 2.3 3.16 1.09 x 10' S2 - SI O O
19.3 1,2 1.19 0.46 x 10' S3 - S2 O
(15.25)
-(S3 - S2) S5 - S3 -(S2:- S4)
dS = [ -(S2f SI) O O S2 - S4 ]

and {Qs} = a vector of advection terms and sources,

16 E -Qo, I(uo. ISO + 130.ISI) + QI,2(UI.2'~1 + I3I.2S2)
(104 em2 S-I)
12 {Q,} = -QI.2(UI.2SI + I3I.2S2) + Q2,3(U2,3S2 + 132,3S3) - Q2.4(U2.4'~2 + 132.4S4)
. -Q2.3(U2.3S2 + 132.3S3) + Q3.5(U3,5S3 + 133.5S5)
8 E',Q
(ems) { - W4 + Q2.4(U2.4S2 + 132,4S4) }
4 ( 15.26)
and {E'} = a vector of unknown dispersion coefficients,
-'-5 -20 -15 -10 -5

x (km)
Ei'2

FIGURE 15.4 {E'} = E~,3 (15.27)

Plot 01 flow, bulk dispersion, and dispersion versus distance E3.5
as calculated in Example 15.2.
{ E~.4}
268 PAlO'111 w"lcr-Quality Environments LECfURE15 Estuaries 269

- ---- -
Because both Eqs. 15.20 and 15.23 have only one unknown, this particular example
can be solved simply with algebraic manipulation. However, for more complicated
cases, compuler methods provide a more general means for determining the disper-
sion cn, ffiCll,';!~.
-- - - -- ~

HWS Mean LWS

Notice thal scgment 4 has a source term, which can be eilher a tributary (Qs) or a
e ~ ~
waste loading OV).In essence all upstream terminal segments must have such source " "
termS"for the successful implementation of this estimation technique. For example ,, .., ,"
I , FIGURE 15.6
I .' I ' Samples around a
if segment 4 had no source term its mass balance would be I ' I ',
I
, ' ,, point discharge 01 a
O = E2.4(S2 - S4) (15.28)
,
, ,, substance Ihat decays
, , with first-order kinetics.
which specifies that there is no gradient; that is, S2 = S4. If such were the cast(, . , 80th high- and low-
segments2 and4 shouldbe consolidatedintoa singlewell-mixedcell. .
The foregoing approach should be applied with caution because, as with all
"
------ waler slacks are shown
along with a mean
o x concentration pro/íle,
data-based approaches, its results are dependent on the quality of the underlying
data. further, because the ca1culation is based on gradients (that is, differences),
Jh~ r~sults are particularly sensitive to data errors (recall Fig. 2.6). However, if
these reservations are considered, it can sometimes provide order-of-magnitude es- The plot indicates that, given the same lenglh scale, mixing is more intense in
ti¡nates of dispersion. Several problems at the end of the leclure are devoted to the streams and estuaries than in lakes and the ocean. This should nut b~ surprisil1g
fiPpr~ac~. since streams and estuaries are subject lo the same wind forces that are the primary
cause of horizontal turbulence in lakes and the ocean. However, addilional forces
supplement the wind and lead to higher total mixing. In streams, gravity-induced
15.3.3
1 Slack-l1de
, Sampllng . flow leads to shear that causes dispersion. In estuaries the tidal molion creates the
The presence of tides frequently makes estuarine sampling a case of "hitting a mov- same SOrtof enhancement.
ing target." The issue is especially relevant to the steady-state dispersion approach Because dispersion and diffusion are different processes, we have mixed "llP-
H~e<lin the previous sections. For the steady-state approach, we are trying to define pies and oranges" in developing Fig. 15.7. However, it is useful in cUlltrastil1gthe
al}average condition for a system that is continuously changing. 1Dagnitudesof horizontal mixing for the three types of waters.
~Iack-tide sampling provides one way to address this dilemma. The term slack
tfde refers to the ROintat which the tides reverse direction,that is, when the flow
~Qesto zero. I~ s0!11esystems this state moves progressively up or down the estu- E
1ifY:''ft1usjf the sampling teaijl has a fast enough boat (approximately 40 km hr-I),
t~ey can collect samples at the slack tide. (cm2 S-1) 109
Because there is both a high-water (HWS) and a low-water (LWS) slack !ide,
such surveys would yield two "snapshots" of water quality: one biased up the estuary ~~
.al!
107
(JiWS) and the otherdown (LWS). Therefore an average would have to be deve10ped
(Fig. 15.6) to be consistent with the long-term approach described in the previous j!g
section. E~
~
(/)- 'ti
~ 1 105

15.3.4 Comparlson with Streams, Lakes, and the Ocean

Recall that in Lec. 8, I presented a diffusion diagram (Fig. 8.11) thal related. Ihe Key:
diffusion coefficient in lakes and the open ocean to length scales. At this point it
1()3
. Estuary
D Stream
would be interesting to relate these values to the dispersion coefficients in streams
FIGURE 15.7
{\!1destuaries. 10
Okubo's diffusion diagram (recal!
.. fjgure 15.7 reprodllces Fig. 8. J.\ \>utadds dispersion coefficients from a number 10 102 1()3 1()4 105 106 107 Fig. 8.11) along with dispersion
of streams and t(st\1~es~ The channel widtf1is taken as the length scale for these values Irom a number 01 streams
waters. . Width (m) and estuaries.
270 PART11I Water-Quality Environments LECTURE15 ESluaries 271
I
1.03

-'?
BOX 15.1. Freshwater Bays

Freshwater systems such as lakes and impoundments can have embayments that act 1.02
like estuaries. We have already addressed the case where such systems can be charac-
terízed as a single well-mixed system when we discussed diffusion in Lec. 8. However,
as depicted in Fig. B 15.1, some are elongated and hence must be treated as a distributed
~ystem.
a.
-5
el
1.01

1.00

0.99
O 10 20 30
7
~
o 3 ¡ 4 T (OC)
+ ~ FIGURE 15.8
Density as a lunction 01temperatura and salinity.

Rlver Elongated bay Lake FIGURE B15.1 To understand this regime, the following equations of state quantify the relation-
ship of density to salinity and temperature (Millero and Poisson 1981):
Because such systems are often fed by polluted rivers, gradients of conservative
sllbstances often occur. As with saltwater systems, such gradients can be used to esti-
P = Po + AS + BS312+ CS2 (15.32)
mate turbulent diffusion coefficients. However, in contrast lo estuaries where the con-
where P = density (g L-1), S = salinity (ppt), and
centration increases loward \he sea, a polluled bay's concentralion usually decreases
to\\'ard the main body of Ihe lake. A = 8.24493 X 10-1 - 4.0899 X 1O-3T + 7.6438 X lO-sT2. ,
Mass balances of a conservalive substance can be written as - 8.2467 X 1O-7T3 + 5.3875 X 1O-9T4 (15.33)
o = QO,I.fO- QI.2(aI.2SI + {31.2S2) + Ei.2(s2 - SI) (15.29) B = -5.72466 X 10-3 + 1.0227 X 1O-4T - 1.6546 X 1O-6T2 (15.34)
e ,.;,4.8314 X 10-4 (15.35)
O = W2 + QI.2(a1.2SI + {31.2S2)- Q2.3(a2.3S2+ {32.3S3)

+ Ei.2(sl - S2)+ E2.3(S3

- S2) (15.30) in which T = temperature eC) and Po = density of fresh water,
O = Q2.3(a2.3S2+ {32.3S3) - Q3.4(a3.4S3
+ {33.4S4) p" = 999.842594 + 6.793952 X 1O-2T - 9.095290 X 1O-3T2
+ E2.3(S2- S3)+ E3,4(S4- S3) (15.31) + 1.001685 X 1O-4T3 - 1.120083 X 1O-6T4 + 6.536332 X 1O-9Ts
Now it can be seen that because the system is fed by a river, it can usually be (15.36)
assumed Ihal diffusion across Ihe interface (0,1) is negligible. Therefore the first bal-
As displayed in Fig. 15.8, density increaSeS with salinity and decreaseS with
ance can be solved for its only unknown, Ei.2' Thís result can then be substituted into
temperature. Note that a peak density occurs at 4°C for the freshwater case.
the second equation, which can be solved for E2.3, and so on. In addition for wider
and deeper bays, two- and three-dimensional approaches can be applied as des<;ribed Because of its high salinity, Sea water is denser than fresh water. Further, in
previously at the end of Seco 15.3.2. many cases, ocean water is often den Ser beeaUSeit is frequently colder than river

,/' -- \1

-_ - - -
--
15.4 VERTICALSTRATIFICATION
Freshwater
The bulk of this lecture has been predicated on the assumption that Ihe estuary is Innow t t t _______.
welJ-mixed laterally and vertically. However, as mentioned in our discussion of Fig.
15.1, wide estuaries can exhibit lateral gradients and deep bays can exhibit vertical
stralification.
__
_ _ - -
I
I
"
/~---T T r--- Saltwater
Inflow

Il must also be noted that bays are not the only place where vertical stratifica-
tion takes place In addition a more gencrally occurring vertical transport regime is FIGURE 15.9
eSlabli"hed due to the interaction of salinity and heat. Two-dimansionalverticaltransport in a chemicallystratili ed estuary.
272 PART11I Water-Quality Environments LEc.TURE15 Estuaries :.73

E You measure the following chloride concentrations along its length (note tha! mile
points increase toward the ocean):

Mile point O 2 4 6 8 \O 12 14

Chloride (ppt) 0.3 0.5 0.8 1.4 2.2 3.7 6.0 10.0

Determine the dispersion coefficient for the estuary in cm2 s-1 .

Tldal mlxlng 15.2. A tidal estuary has the following characteristics:

Depth = 5 m
Freshwater mlxlng Cross-sectional area = 100 m2
Flow = 0.3 m3 S-I
Head Mouth
FIGURE 15.10 x (miles from mouth) -30 -24 -18 -12 -6 O
Factors contributing to the tidally averaged dispersion coefficient in a Chloride (ppt) 0.4 0.7 1.4 2.7 5.2 10
chemically stratified estuary.
A pollutant that settles at arate of 0.2 m d-I is to be discharged at kilometer poilll lb.
water. As a consequence, as shown in Fjg. 15.9, denser sea water can foem a salt (a) What is the dispersion coefficient?
wedge at the bottom of the estuary. Less dense freshwater river water ftows over (b) What is the estuary number?
the top of the wedge. This seaward transport is supplemented by salt water moving (e) What loading can be discharged iflhe maximum allowable concentration is I ppm?
upward. Thus the surface layer is brackish.
15.3. Determine the net estuarine ftow for the following case:
The net result is that longitudinal mixing is enhanced in the region of the wedge.
Start of ebb = 0700 End of ebb = 2000 Peak ebb ftow = 2.5 cms
As depicted in Fig. 15.10, estuarine longitudinal dispersion is made up of three com-
Start of ftood = 2000 End of ftood = 3100 Peak ftood ftow = 2.1 cms
ponents. In the freshwater region the system is dominated by riverine dispersion due
to the shear ftow. In the estuarine section this mechanism is enhanced by the tidal 15.4. Thomann and Mueller (1987) report the following salinity data for the Back River, all
oscillations. And in the area of the wedge the effect of salinity leads to a region of estuary of Chesapeake Bay:
maximum mixing.
Aside fcom dispersion, estuarine hydraulics have additional effects on water Distance from mouth (km) O 5.5 8.6
yuality through the impact of ftow and mixing on suspended solids. This is due to the Salinity (ppt) 3.9 2.5 0.32
fact that there is a point in the estuary called the dead zone where the downstream
river ftow and the tidal motion tend to cancel. This convergence zone is the site of Ifthe net nontidal velocity is 3 cm S-I > determine the dispersion coefficient in cm2 s
enhanced sedimentation.
The solids themselves are carried into the estuary with the freshwater ftow and 15.5. Suppose that the following data are available for a vertically stratified estuary (Fig.
PI5.5):
are also generated by plant growth. These partic1es settle to the bottom saline layer
where they are transported back upstream to the dead zone. This concentration of QOI =4 X 106 m3ts Q32 = 2.5 X 106 m3ts EÓI= O SI = 8.5 ppt
partic1es is important in most water-quality contexts. For example many toxic com- S2 = 14.5ppt = 15 ppt
S3 So = Oppt
pounds preferentially associate with solid matter. Thus the "dead zone" is also a
"deposition zone" where large sediment concentrations of contaminants and nutri- (a) Determine the bulk dispersion coefficients, E;2' Eí3> and E23. Assume thal Ihe
ents can accumulate. .
latter two are equal and use backward differences for the advective lerms.
(b) Suppose that the inftowing river has a concentration of 5 mg L -1 of a consCIva.
tive pollutant. Assuming that segment 3 contains negligible concentration 01'the
substance, determine the concentration in segments I and 2.
rROBLEMS
}S... J\ Qne-dimensionalestuaryhas constantdimensionsand ftow: 15.6. The following data are available for the system depicted in Fig. PI5.5:

Width = 1000 ft QOI= 1 x lQ6m3d-1 So = Oppt = 30ppt

55
pepth = 10ft QI2 = 1x lQ6m3d-1 SI = 1.74ppt W4 = O
flow = 500 cfs
 LECTURE15 Estuaries 275
274 PART11I Water-Quality Environments ~

15,8. 1\vo tributaries enter an elongated impoundment along its length (Fig. PI5.8). Flow
QB = 1.5x 1ü6mJ d-I .~2= 10.47ppt and chloride concentration for these tributaries and the inflow river are
Q.15 = 1.5 X 106 m3 d-I S3 = 22.33 ppt
Q24 = 0.5 X 106 m3 d-I S4 = 7.48 ppt Chloride Phosphorus
Flow concentration concentration
Determine the bulk dispersion coefficients for the interfaces in this system.
(m3 yr-I) (mg L -1) (ltg L -1)
Inflow 10 x leY' 150 30
'SI.. Tribulary A 7.5 X 106 400 100
, ' Tribulary B 5 x leY' 100 50
O: I ... 1 ~

: t--t
2
~ I
~
3 The resulting steady-state chloride concentrations are shown at the bottom of the plot.

Rlver Estuary Ocean FIGURE P15.5 7 8 9

1".7. A vertically stratified water-supply reservoir has a small embaymenl (Fig. P 15.7). The '.

water utility typically fluoridates the water at a constanl level with the result that all
water in the lake has a constant nuoride concentration of I mg L -l. To study the lake's
..... 300
mixing characteristics, you talk the ulility into abruptly stopping nuorida!ion. In the 'i
...1
ensuing weeks you measure the following concentrations as the lake purges itself of . . .
fluoride:
1:1I
.s 200 . .
Times. d O 20 40 60 80 100 120 140 160 GI
"D
1: . . . .
I 0.49 0.26 0.15 0.10 0.08 0.05 0.04 0.04 o 1
Epilimnion
I 0.98 0.92 0.85 0.78 0.73 0.65 0.60 0.54 (.) 100
Hypolimnion
0.27 0.20 0.16 o 500 1000 1500 2000 2500
I 0.93 0.77 0.61 0.47 0.38
Bay
Dlstance(km)
The parts of !hc lake have the following volumes:
FIGURE P15.8
Vr = 3 X 107 m1
V" = 8 X lO" mI The following segmentand interfacedataare available:
V" = 3 X lO" mJ
Length Wldth Depth Chloride Area
The inf10w = outnow =
I x ({t, m3 d l. all of which f10ws through the epilimnion.
Segment (m) (m) (m) (mgL 1) Interface (m%)
Use this informalion to dcterminc Ihc bulk diffusion coeflicicnls across the thermocline
and betwccn the bay and the epilimnion. 1 500 250 1 150 1-2 250
2 500 250 I 15] 2-3 250
3 500 250 I 157 3-4 250
Inllow Outflow 4 500 300 2 200 4-5 600
5 500 300 2 253 5-6 600
6 5110 300 2 245 6-7 600
Eplllmnlon
Bay 7 500 ItXJO 5 222 7-8 5000
8 500 ({JO() 5 222 8-9 5000
9 500 ({JO() 5 222

G
(a) Employ Ihis data to eslimate !he longitudinal turbulcnt diffusion coefficients for
the interfaces in this system.
(b) Determine the total phosphorus concentration profile if TP settles at a mte of
FIGURE P15.7 IOmyr-l.
 lEl rURE 16 Lakes and Impoundments 277

LECTURE 16
OriginoBy thisI meanwhetherthe waterbodyis natural(a lake)orartiticial(al!
impoundment or a reservoir). Although there is a tremendous amount of vario
ability within these two categorics, there are some generalizations that typify
Lakes and Impoundments each in a general sense. In particular impoundments often have controJled
outftow. In contrast naturallakes are uncontrolled. In addition, as descrihed
next, they usually have different shapes.
Shape. Artiticial impoundments are almost always created by damming a r¡ver.
. Consequently they tend to be elongated or dendritic since they consisl uf
drowned river valleys. In contrast natUrallakes tend to be less elongated and
more circular. Of course there are also some dendritic and elongated natlllal
lakes and circular impoundments, but the reverse situation is more likely.
Size. The two aspects of size that have a strong bearing on water quality are
residence time and depth. In general,lakes are divided into short (Tw< I yr)
and long (Tw> 1 yr) residence time systems. Further they are c1assitied as
shallow (H < 7 m) and deep (H > 7 m). The latter c1assitication is signiticanl
because deep lakes are often subject to thermal stratitication during cerlain
times of the year.

_ _ M .. - ..-

\:Lr
(a) Completely mlxed (b) Deep, vertlcally
reactor segmented
'n contrast to ftowing waters, lakes and impoundments were not emphasized in the
~arly years of water-quality modeling. This is because, with the exception of large
navigable systems like the Great Lakes, they have not historically been the major
focus of urban development.
Starting in the 1970s, however, it was recognized that natural and man-made
lakes are equally, if not more, important than estuaries and rivers from a recreational
standpoint. Further their use for water supply, hydropower, and ftood control also
contribute to their significance.
This lecture is designed to provide you with background information on these (e) Dendrltlc, laterally (d) Elongated,
segmented
systems. After a brief overview of lake types, 1describe how lake morphometry and longltudlnally segmented
hydrology can be quantified. Then 1 describe how pollutants advect and diffuse in
the near-shore zone.

11. .1 ST4NDING
'. WATERS
'fbj§ '~tUf~ focuses oq standing waters. Such water bodies range from small deten- (e) Rlver-run reservolr,
longltudlnally and vertlcally
"00 ~~as&Q~lJge $Yste~s li~e~c Oreat Lakes and Lake Baikal. Lake scientists (or 1 segment"d
IiJlJIIQWI"'f)d"l~sjfy Jakes in O1any ways (see Hutchinson f957 and Wetzel 1975
(!;Irr~vi~ws.).. ;11}the standpoint of water-quality modeling there are three major FIGURE 16.1
f~1ltID"eipf .*e~ tb~t Rlate te>QPfefforts to sip1Ulatettteir ~sport and fate:
Typical segmentation schemes usad lor lakes and impoundments.
276
278 PART11I Waler-Quality Environrnents

As we will see throughout the remainder of this text, these attributes have an
impact on how lakes and impoundments are modeled. For the time being the pri-
mary implication relates to model segmentation. As depicted in Fig. 16.la, CSTRs
are often used to model standing waters. However, such systems are also segmented
vertically (Fig. 16.1b),laterally (Fig. 16.lc), and longitudinally (Fig. 16.1d). Further,
as i11the deep river-run reservoir in Fig. 16.le, the system can be divided in several ~
;¡ :1 ::t
::.: -~-~-~--
~ ~
11
~(1)

dimensions.ln Lec. 18 we revisit the topic ofmodel segmentation in a more general E

¡..:¡ >-
sensc. 1\1\1\1\ .c
iii
~ ~ ~ ~ ..( e .c
11I.
E§
16.2 LAKE MORPHOMETRV ., e~
e ti tIf :z:'" :i
The lirst step in characterizing any lake or impoundment is to determine its geometry
N
~~
(1)-
.. c:
11I'-
or, as it is fonl1ally called, morphometry. To do this the lake's bathymetry must be 0111
.5 '1:
mapped. Batllymetry refers lo a topographic map showing depth contour lines (Hg. .5 (1)
16.2). eE
(I)::J
_1:
The analyst can then determine how much area is encompassed by each depth (1)>-
'O.c
cOlltour. This can be done using a mechanical device called a planimeter. Ir a '081
planimeter is unavailable, an altemative is to superimpose a grid over the bathy- ~E::J
.-
metric map. The areas can then be estimated by summing the areas of the grid cells :!!g
falling within each contour. 8CD
~o ~.5
As depicted in Fig. 16.3 the resulting areas and corresponding depths represent a Q¡e

\
tabular function of the dependence of area on depth, A(z). The volume for the system E~
OCD
can be detennined by integration. For example the volume from the surface (z = O) ~-I , ,
~~
0-
down to a particular depth (z = H) can be calculatcd as Ea¡
°6 ~
fll . i! i
~'O
V(H) = Jo A(z)dz (16.1) = ~ ~ :z:'"' :z:'"
N m5i
!!!'2
In addition the volume betwecn two deplhs can be evaluated. For example Ihe vol- 1Hi
lImc hctween Iwo adjacent depths can be represented generally as 11I"3
lli.' ~-
CD'fiI
-1:
V;.i+1 =
flIi
A(z)dz (16.2)
=
~CD
..oS
11I(1)
-1;;
Clearly, hecause we are dealing wilh tabular data, numerical methods must be
BIQ
11(1)
uscd to evaluatc Ihe integrals. This is most commonly accomplished by applying the
(')01;;
cci:3m
'-1'!(I)
liieF
a::a..
o' ::ICDI}
~FE

FIGURE 16.2
A bathyrnetric rnap 01 Lake Onlario (courtesy 01 Mike McCorrnick,
GLERUNOAA).
279
280 PAR111I Water-Quality Environments
I.ECCURE lb Lakes and Impoundmenls 21S1

.. rule to Eq. 16.2, TAIJLE 16.1

trapezOIdal _ A(H¡) + A(H¡+d (H¡+I _ H,). Summary of bathymctric data and morphometric
V;.;+I - [ 2 ]
(16.3)
-
calculatlons for the Tolt Reservoir, Washington
Cumulative
In this way the volume can be determined for each layer starting at the surface Depth Area Volume volume
and working down to the bottom. The individuallayer volumes can be accumulaled Index (m) (106m2) (106m3) (106m3)
to determine the total volume down to a particular depth, " O O 5.180 29.73 O

Vi+1= t
J=O[ A(Hj) +2A(Hj+d ](Hj+1 - Hj)
(16.4)
1
2
3
4
6.10
9.14
12.19
15.24
4.573
4.249
4.047
3.683
13.44
12.64
11.78
10.67
29.73
43.17
55.81
67.59
where ; + 1 = depth at which the volume is to be determined. 5 18.29 3.318 16.90 78.26
6 24.38 2.226 11.22 95.16
Not only can volume be computed from areas, but area can be determined from 7 30.48 1.457 7.401 106.4
volumes. 8ased on the inverse relationship between differentiation and integration, 8 36.57 0.971 4.255 113.8
9 42.67 0.425 1.789 118.0
A(H) = dV(H) (16.5) 10 48.77 0.162 0.493 119.8
d1. 11 54.86 O O 120.3
Thus the rate of change of the cumulative volume at a point provides an estimate of
the area.
Now Eq. 16.6 can be used lo calculale Ihe area at inlerface 1,
Again numerical methods are required. A common approach is to use a centered
divided difference, Al = dVI == V2 - Vo = _(43.17
x 106 -0)m3 = 4.72 x 106m2
dz Z2- lo (9.14 - 0.00) m
A- = dV; = V;+I - V;_I (16.6)
I dz - 1.;+1- 1.;-1 Observe the discrepancy belween Ihis result and Ihe lroe area (4.573 x 106) in Table 16.1.
The disparily is due in part to the inadequacy 01' Eq. 16.6. The alIemalive formulalÍ,,,.
However, this approach is ftawed for unequalIy spaced depths. In addition it cannot (Eq. 16.7) provides a beller eSlimale,
be used to evaluate the area of the top-most segment. A formula that takes care of
dV(6.1O) ==0+29.73x 106 2(6.10)-0-9.14
both shortcomings is (Chapra and Canale 1988) dz (6.10 - 0)(6.10 - 9.14)

~(Z) = dV(z) == V; I 2z - Z; - Z;+I + V; 2z - Z;-I - Zi+1 4317106 2(6.10)-0-6.10 =4572xI06 2

dz (Z;-I - z;)(Z;-1 - zi+d (z; - Z¡-I)(Z; - Zi+I) + . X (9.14-0)(9.14-6.10)' m
In addilion, Eq. 16.7 can be used lo determine Ihe 10psurface area,
+ V;+I (Zi+1 2z- -1.;-d(1.i+1
Z;-I - Z;
- 1.;).
(16.7)
dV(O) ==0+29.73 X 106 2(0)-0-9.14
dz (6.10 - 0)(6.10 - 9.14)
EXAMPLE 16.1. BATHYMETRIC AREA AND VOLUME CALCULATIONS.
Bathymetric maps ofthe Tuh Reservoir (a waler-supply impoundment for Seallle, Wash-
6 -
2(0) O- 6. 10 6 2
~43.17x 10 (9.14-0)(9.14-6.10) = 5.176x 10 m
ington) can be used to determine the areas and deplhs in the second and Ihird columns
of Table 16.1. Delermine the volume if Ihe reservo ir level is at z = O. whichalsocompare>well wilhthe measuredarea(5.IHOx 106).
Solution: Equation 16.3 can be used lo compute Ihe volume of Ihe lop segmenl,
Allhe end of the previous example we saw how Eq. 16.7 provides superior esli-
mates ofthe afea based on the derivative ofthe volume. However, becallse numerical
VO.l= e.180 X 106; 4.573 x 106)(6.10 - 0.00) = 29.73 X 106m3 differentiation tends to amplify data errors, it is not recommended thal area be de-
Similarcalculationscan be performedforthe othersegments(Tabie16.1)and the results termined in this fashion. Rather the procedure outlined in Fig. 16.3 (bathymetry -+
substitute4iotoEq. 16.4to computethe totalvolume, arca -+ volume) is the proper approach.
It should be noted that the results showed in Table 16.1 are often presented in the
VII = :2)9.73 X 1~6+ 13.44x 106+ ... + 0.493 X 106 = 120.3 X lOSm3 graphical format of area-(jepth (Fig. 16.4a) and volume-depth (Fig. IbA/.) plots. As
The cumulative volume is listed in the last column of Table 16.1. described above, the area-deplh version is more fundamental, with volumes being
generated through inlegration of the area-depth curve.
 ..'

2HZ I'ART11I Waler-Qualily Environmenls LECTURE16 Lakes and Impoundmenls 283

For mosl cases it is much easier to measure inflow and outflow than the olher
6a)
-- 4
quantities in Ihe balance. Thus for many water-qualily modeling applications, !he
ea'" other terms are neglected. This simplification is based on the assumption !hat pre-
cp E
.. ... cipilation balances evaporation and groundwater flow is negligible.
c(o 2
=- Although this simplification is often invoked for expediency, it should be
checked for validity. Unforlunalely although precipitation measurements can some-
times be oblained (usually contingent on access lo a nearby wealher station), direct
20 40 60
measuremenls of evaporalion and groundwater are much more difficult to obtain.
The following example illustrates how an inaccurate water budget can impact mod-
cp::;,
120 t (b) eling calculations.
EE 80
.2... EXAMPLE 16.2. EFFECT OF WATER BUDGET ON QUALlTY MODEL-
00
>-.......
40 ING. A lake has Ihe following characlerislics:
Volume = I X 107 m3
20 40 60 River inflow = I X 106 m3d-I
Riveroulflow= 0.8 X 106 m3 d-I
Depth (m)
Suppose Ihal a firsl-order decaying (0.2 yr-I) dissolved pollulanl is discharged lo Ibis
FIGURE 16.4
system al a constanlrale of mass loadingof 1 X 107g yr-I. CaIculalethe lake con-
(a) Area-depth and (b) volume-depth curves lor the Tol!
centration for Iwo cases, if the discrepancy belween inflow and outflow is due lo (a) a
Reservoir, Washington. ·
groundwaler loss or (b) an evaporalion loss.
16.3 WATERBALANCE Solutlon: (a) For Ihis case the loss will carry Ihe pollutanl inlo Ihe groundwater. There-
fore the total oUlflowis equal to 1 X 106 m3 d-I and the concentration can be calculaled
A waler balance for a well-mixed lake can be wrilten as as

S = -dV
dt
= Qin - QOUI + G + PAs - EA,
.
(16.8) e =~
Q+kV
= I X JQ7
',~I~.nl"H.n7'\
= 5 m Vi
g
(b) For Ihis case Ihe loss will not carry Ihe pollutant inlo the almosphere. Therefore Ihe
where S = slorage (m3d-l) IOlaloutflow is equal lo 0.8 X 106 m3 d I and the concenlration can be caIculated as
V = volume (m3)
1 X 107 _1
t = lime (d) e = __ . _, _. .. . _.. = 5.556 mg L
Qin = inflow (m3d-l)
QOUI = outflow (m3 d -1 ) Therefore a discrepancy of about 10% results, depending on your assumption regarding
G = groundwaler flow (m3 d-I) the discrepancy belween the river inflow and outflow.
P = precipilalion (m d-I)
E = evaporalion (m d-I)
A.. = surface area (m2) Of course if either evaporation or groundwater flow can be deemed unimpor-
tant, the situation is greatly simplified. For example if the lake overlies imperme-
Equalion 16.8 can be applied in Iwo general fashions: steady-slale and time-
able bedrock, the assumption could be made that groundwater is unimportant. In
variable. The following sections treat both Ihese applicalions. / such inslances if inflow, outflow, and precipitation are known, Eq. 16.9 can be used
to estimale the evaporation rateo
16.3.1 Steady-State
IlllUany instances lakes and impoundments (especially large ones) do not experience 16.3.2 Evaporation
drastic volume changes for the time periods addressed by waler-quality models. For Although evaporalion can be eslimaled by difference, Ihere are many syslems where
~llchcases Eq. 16.8 simplifies lo this is impossible. In such cases direct measuremenls and model equalions provide
O= Qin - Q"UI + G + PA.,- EAs (16.9) altematives.
 lRTURE 16 Lakes and IlIIpoundmefils 285
284 PART11I Waler-Qualily Environmenls

Compule Ihe evaporalion ftow.

The most comlllon direct method involves deploying a pan of water. The pan
can be either Iloatcd on the water surface or positioned on land close to the water Solution: The required paramelers can be delermined as

body. The amount of water lost over time is monitored daily (with corrections for L, = 597.3- 0.57(25) = 583.05cal g-I
precipitation) and is expressed as the pan evaporation rate Ep(cm d-I). It is then f(Uw) = 19.0 + 0.95(2)2 = 22.8
adjusted by a correction factor called the pan coefficient kp to extrapolate the result 17.27(25)
to the natural water. It can then be used to compute the quantity of Ilow lost via es = 4.596e237.3+(2S)= 23.84
evaporation, as in 17.27(20)
Qe = O.OlkpEpAs (16.10) eair = 4.596e237.3+(20)
= 17.59

where Qe = evaporative water Ilow (m3 d-I), As = surface area (m2), kp averages These values can be substituted ¡filOEq. 16.1210 give Ihe fIow rate,
about 0.70 for the United States with a range from 0.64 to 0.81, and the 0.01 is
includedto convertcm to m. Q, = 0.0122.8(~~~~5~1~3.84) 1 x 106 = 2440.8 m3 d-I
There are alsosomeequationsavailabletocalculateevaporationbasedon mete-
Note that the posilive sign connoles Ihal the fIow is losl from Ihe lake.
orologicaland lake conditions.Forexamplethe energy Iluxdue to evaporationcan
be computedas
He = f(Uw)(es - eai,) (16.11)
It should be noted that there are many other predictive approaches lor calculat-
where f(Uw) = a function rellecting the effect of wind on evaporation and es and ing evaporation. Any good book on hydrology (e.g., Chow et a\. 1988, Ponce ]989,
eai, = the vapor pressure corresponding to the water and the dew-point tempera- Bras 1990) will contain an extensive review on the subject. In addition wc pro-
fures(mmI-J~),respectively.The heat transfercan then be convertedto a waterIlow vide additional information on evaporation in our discussion of thcrmal mouding
rate by in Lec. 30.

Qe = 0.01 f(Uw~(es -
'epw
ea;,)As (16.12)
16.3.3 Tlme-Varlable
where Le = the latent heat of vaporization (cal g-I) and pw = water density
(g cm-3). The 0.01 is included so Ihal Ilow is in m3 s-l. The latent heat and the For simplicity we assume that precipitation and evaporation are appro.\ill1alelyequaJ
wind function can be computed wilh . and that groundwater flow is negligible. For this case Eq. 16.8 reduces to
Le = 597.3- 0.57T (16.13) dV
di = Qin - QOUI (16.16)
~
and (Brady,Graves,andGeyer 1969)
If the initial volume and the inflow time series are known, the solulion or Eq.
f(l/w) = 19.0+ 0.95U; (16.14) 16.16 depends on characterizing the outflow. Of course if the outl1ow is regulatcd
where T = temperature en
and Uw = the wind speed measured in m s-I at a or measured, the solution is simple. In many cases, however, it is not given. For these
situations the relationship between outflow and volume must be detennined. This is
height of 7 m above the water surface. The vapor pressures can be delermined by
the fOl1l1ula(Raudkivi 1979) done by establishing a functional relationship between outflow anu hcad (volume)
17.27T in the reservoir.
e = 4.596e237.3+T (16.15) For certain spillway structures, equations are available to establish the relalion-
where the surface water and the dew-poinl temperature are used to generate es and ship. These formulas are usually of the general form
éai" respectively. QOUI = CLHu (16.17)
These formulas can then be used to estimate the Ilow lost via evaporation. This
where C and a = coefficients
value, along with inllow, outllow, and precipitation flows, can then be subslituted
L = length of the spillway
.into ~q. 16.9 to estimate the groundwaler flow. H = the total head or surface-water elevation
t:JAMfLE ¡f).3. EVAPORAT¡ON CALCULATION. A lake has the following For cases where an equation is not given, relationships can be established by mea-
"hD,11lcteristics:
surements of flow and elevation. In either case Eq. 16.16 can be written as
&urface arel! =;¡ t x IQ6 1J!3i'
WlmlfP.m ~. .7OJg\
Water t~JTlperature= 7~~C:::
. dV
di = Qiu\ Qout(H) (16.18)
Dew-point temperalUre = 20"C
286 I'ART11I Water-Quality Environments
LECTURE16 Lakes and Impoundments 287

The process is further complicated by the fact that the relationship of depth and
volume must be established. In other words Now that we have learned how to route water through a pond, we can model
pollutant transport and fate for such a system. To do this we can write a general
dV = A(H)dH (16.19) mass balance for a pollutant that reacts with first-order kinetics as
Thus an area-stage relationship A(H) must be established (recall Fig. 16.4a). If dM M
tlti.. i~done Eq. 16.19 can be substituted into Eq. 16.16 to give dt = Qi,,(t)Cin(t) - Qout(H)y - kM (16.21)
Acculnulation Innow Outftow Reaction
dH = Qi,,(t) - QOllt(H) (16.20)
dt A(H) Observe how this equation has been written in terms of mass rather than concentra-
T"is approach is sometimes called the level-pool routing technique. tion because the volume is changing. Thus to soIve this problem, we must integrate
I.;XAMPLF 16.4. LEVEL-POOL RESERVOIR FLOW ROlJTING. A small de-
Eqs. 16.20 and 16.21 simultaneously to calculate mass and volume at each time step.
Then, if we would like to determine the pond's concentration, we merely use
tcntion pond has a surface area 01'2ha and vertical sides. The discharge-head relationship M
Ilas heen rneasured:
C = y (16.22)
EIevation Outllow
(m) (m"s ')
0.0 0.0 EXAMPLE 16.5. LEVEL-POOL RESERVOIR POLLUTANT ROUTING. Sup-
0.5 0.0
pose that 1'01'the detenlion pond from Example 16.4, Ihe inllow has a constant con-
1.0 0.0 central ion 01' 100 mg L-1 or a pollutanl thal seltles at arate or 1 m d-J. Calculate the
1.5 1.7 concentralion and Ihe rate or mass oulllow. Assume that e = Oat t = O.
2.0 5.0
2.5 9.0 c 120 T (a) Inflow
3.0 14.0 o
:; 80
3.5 20.0 ::
4.0 26.0 Cc> 40 Outflow
fj
c-E
o
NOlice that no Olltllow occurs when the water level is below 1 m. This volume is called
(J o
dead storage. A storm creates the inllow hydrograph in Fig. 16.5. Use the level-pool
o 40 80 120 160 200
1200
lllOdello route Ihis lIow thl'ough the detention basin. Assurne thal inilially the pond has
a depth 01' I In.

Sollltion: The results of applying the level-pool rouling technique are displayed in
--Oí::.-Ec 800

Fig. 16.5. Notice how the pond tends to diminish and spread the lIow. The peak di s- la c> 400
:¡¡- Outflow
charge is diminished from 10 cms at t = 50 min rol' the inflow to approximately 4.5 cms
al about t = 78rnin rol' the oulllow. o FIGURE 16.6
40 80 120 160 200 Inflow and outflow (a) concen-
12 tralion and (b) mass 101'a
Time (mln)
small detention pond.

, 8
::;-
Solllllon: The results 01' applying the level-pool l'Outing technique are displayed in Fig.
: 16.6. Notice how Ihe concentration increases rapidly as the inllow carries the pollutant
.§. 4 inlo Ihe system. After Ihe peak lIow has passed, the concenlralion diminishes slowly as
seltling becomes the primary removal mechanism. In conlrast the mass outflow rate more
closely rollows Ihe oulflow hydrograph (Fig. 16.5), which rol' this case is determined
mainly by Ihe outllow rateo
40 80 120 160 200
t (min)
FIGURE 16.5
16.4 NEAR-SHORE MODELS (ADVANCED TOPIC)
Inflow and outflow hydrographs lor a small detenlion pond.
The outflow hydrograph was calculated with the level-pool Since many pollutants enter lakes (and other waterbodies for that matter) at their
routing method.
peripheries, another important water-qllality problem is the distribution of contam-
inants in the vicinity 01'a waste discharge 01'a river. SlIch areas are important in a
 '!j

288 PART11I Water-Quality Environments

LECfURE16 Lakes and Impoundments 289

y
d2e 1 de k
d,2 + -¡. d, - "He = O (16.26)
the solution of which is

{k;2 (k;2
-..
'" ./' . e = Bloy E + CKoy E (16.27)
x x };I' where B and C are constant!. of integration and lo and Ko are modified Bessel func-
, FIGURE 16.7
, Cartesian (x, y) and radial (r)
.~ tions of the first and second kind, respectively. By invoking the l'ollowing bOlllldary
conditions:
coordinates used lor coastal zone
models.
e(,o) = ea ( 16.28)
and c(oo)= O ( 16.29)
management context since human use and perception of a lake are typically intense O'Connor solved Eq. 16.27l'or
in the coastal or near-shore zone, where beaches and other recreational areas are
Jpc~ted. W
c: KOYE
After an inHial period 01'mixing due to the turbulence of the discharge jet (see
(1a.30)
fisc~er et al. 1979 for a discussion of modeling jets and plumes), the concentration
of a near-shore input is dependent on the transport processes in the lake and the
ea
~
KOYE
pollut~nt's reaction characteristics. For a vertically well-mixed, constant-depth layer
suc~ asthe epilimnion of a stratified lake, the distribution of a substance reacting with O'Connor's first boundary condition (Eq. 16.28) sets the concentration equul to
firshorder decay is represented by a constant at a distance '0 from the origino This distance can be thought of as the
periphery of a mixing zone. Although this forrnulation has the disadvantage oEnot
_
ae ae ae a2e a2e
=
a. t -U x-
ax - UY-ay + EX-
ax 2 + EY2
ay
- ke (16.23) being directly related to the waste source at, = O,it is useful for circumventing the
problem that the solution (Eq. 16.30)actuallyapproachesinfinityas , approaches
where the x and y axes are defined to be parallel and Ferpendicular to the shoreline, zero. This can be seen from an altemative solution to Eq. 16.26 that uses a boundary
condition at , = O,
respectively (Fig. 16.7). The following sections present solutions to Eq. 16.23.

~teady-state case in an intinite fluid (no advection). In situations where ad- e = ~Ko
1THE
{k;2
YE (16.31)
vective currents are negligible, the steady-state version of Eq. 16.23 is (assuming
thal diffusion is equal in all directions) where H = the water depth. This solution goes to infinity as, approaches zero ~inú:,
a2e a2e
mathematically, the waste source emanates from a point (or more accurately a Jine
since the source enters over the depth of the layer) of infinitely small thickness at
E (ax2 + ay2 ) - ke = O (16.24)
the shoreline. Di Toro (1972) has presented an analysis of the problem along w!th
a tabulatedsolution that is finite at the origino While the modeler should conslllt
O'Connor (1962) transforrned and solved Eq. 16.24 for polar coordinates, asdepicted
in Fig. 16.7. If, is the radial axis, Eq. 16.24 becomes Di Toro's paper, Eqs. 16.30 and 16.31 provide adequate approximations in many
cases and will be used in subsequent examples.
a2e 1 ae 1 a2e k
- + -- + -- - -c = O (16.25)
ar2 r ar ,2 ao2 E EXAMPLE 16.6. RADIAL UACTERIAL MODEL. O'Connor í I"(,~) used 1'1.
16.30 to solve for bacterial distribution in the vicinity of Indiana Harbor in Lake Midu-
If we assume that e is constant for a given " then ac/ao and a2c1a02 are zero gano The radius of the mixing zone was taken to be 45.7 m, which is one-half the
~nd Eq. 16.25 reduces to a Bessel equation t of order zero, approximate width of the harbor outlet. A dilfusion coefficient of 2.6 x 106 m2 d-I
(3 x lOScm2 S-I) was used to characterize turbulent mixing. 1\vo decay rates, 0.5 and
3.0 d-I, were used to estimate bounds for bacterial die-off under summer temperalUre
conditions.
'Differential equations of a particular fono (e.g., Eq. 16.26) are called Bessel equalions. The solutions
for tbese equations are called Bessel functions. These functions are tabulated in numerous matbemat- Solution: The results of applying O'Connor's model (Fig. 16.8) indicate general agrec-
ical reference volumes and bandbooks. Additionally many computer systems carry easy-Io-use library
ment between model results and 3-month averages (June, July, and August) of b¡¡cIerial
functions to compute tbeir value. conccntration.
 LECTURE16 Lakes and Impoundments 291
290 I'ART11I Water-Quality Environments

Steady-state case in a bounded fluid. The preceding solutions are appropriate

for situations where the reaction is rapid andlor the transport weak andlor the lake
wide enough that the opposite shoreline of the lake has no effect on the solution.
For example Eqs. 16.31 and 16.33 would be valid for the analysis of discharge of
coliform bacteria into one of the Great Lakes since the bacteria would die off within
a few kilometers of shore and such lakes are on the order of 100 km wide. However,
for narrow lakes or for substances that react slowly or are conservative. the effect
of the oppositeshorelinemust be considered.In that case Eq. 16.33 is expressed
directly in cartesian coordinates as
2
w U,x k V
c(x, y) = -e
7THE
lE Ko (x2 + y2) - + ---3-
~ [
E 2E )
] ( (16.35)

Then for a lake of width Y, the solution is obtained by the iterative fonnula (Boyee
and Hamblin 1975),
oc

e = c(x, y) + 2)c(x, y + 2nY) + c(x. y - 2nY)] (16.36)

n=1
In this solution the contining effeet of the opposite shoreline is aeeounted for by the
intinite series of additions in the seeond half of Eq. 16.36. In praetice only a small
FIGURE 16.8 number(== 2 or 3) are neededto eharacterizethe effee\.
Protiles 01 bacterial concentration (normalized to
concentration at the edge 01 the mixing zone) EXAMPLE 16.7. BACTERIAL MOOEL WITH AOVECTION ANO DIFFU-
versus distance r (km) Irom the edge the mixing
SION. A municipality discharges sewage at arate of 4 X 104 m3 d-I at a point on
zone in the vicinity 01 Indiana Harbor, Lake
a lake's shoreline. The wastewater is laden with coliform bacteria (30 X 1<1'number
Michigan, as originally computed by O'Connor
per 100 mL) that die at arate of 1.0 d-I. Ca1culate eoneentration pro fiJes in the lake's
(1962). The lines represent model calculations
based on two estimates 01 the bacterial die-off vertically well-mixed surfaee layer (H = 20 m) for the following cases:
rate. The data are 3-month averages lor June (a) The lake is infinitely long and wide and horizontal diffusion (Ex = Ey = 5 X
through August. 1<1'm2 d-I) is the sole transport meehanism.
(b) The same eonditions exist as in (a), but with a eurrent of 0.5 X 1ü4 m d-I moving
from east to west.
Steady-state case in an in'inite fluid (advection along the shoreline). AI- (e) The same eonditions exist as in (b), but with a finite width of 4 km.
thIJugh the preceding is appropriate where turbulcnt mixing is the primary trans-
Solution: (a) The waste load W is ea1culated as the product of flow and coneentration
pOli mechanism. some lakes havc persistent unidirectional currents that can advect
of the diseharge, as in
1'f)lIutantsalong the shoreline. For this case Eq. 16.23 hecomes, at steady-state,
(Jl' a2e (J2e W = (4 X 104 m) d-I )(30 X 106 number per 100 mL)
U,~ =E ~ +- - kc (16.32)
. (Jx ((lx2 (Jy2 ) Equation 16.31 can then be used to determine eoneentrations in the lake,

where Ux = current velocity along the shoreline. Hoyce and Hamblin (1975) have
plt'~:entedthe following solution: e = (4 X 1~~)~30 ~_!06) KoJ 5 ;2106
2
W u,:' k VI The resultsof this eomputation for a seetion of the lake are displayed Fig. 16.9a.
l' = -c2f,
7THE
Kn r - + -
[
E 2E ( )] (1633)
. (b) In this caseEq. 16.33 is used:
2
I (4 X 104)(30 X 106) (O.5XI04), I (0.5 X 1Q4)X
e = e 2(5XUr)Ko r
whcre 1T(20)5 X 106 5 X 106 + [ 2(5 X 106) ]}
,. = Jx2 + v2 ( 16.34) {

Notlce that for lJ \ = O.Eq. 16.33 reduces to Eq. 16.3t. The resultsare shownin Figure 16.9b.
292 PART11I Water-Quality Environments
LECTURE16 Lakes and Impoundmenls 293
- - -- -
1(8)-- - - - - -N-
I 1km t
,
I
f(b) - -
IAdvectlon
As in Example 16.8, Fig. 8.11 pro vides a first estimate of the expected rallgl: of
I I
the horizontal eddy diffusion coefficient for such long-term computations. Howcver,
as with any default coefficient, addit. "11 measurements, such l'llibration with
I
V1~~ I
conservative substances, are ad\'i. \oyce and Hamblin';, "nration of ¡heir
modello chloridecontoursis a goodexampleof such an approach.
1 2~ 1 I

i/i 3\
I

:~~:
3~ I I
I
I
/ PROBLEMS

16.1. The following depth-area dala is available for Lake Youngs inlhe State ofWashinglon:

Deplh(m) o 6 14 28 31
FIGURE 16.9 Area (012) 2,830,000 2.405.500 1,188,600 424,500 O
Countours 01 colilorm bacteria (number per 100 ml) as a result 01 a waste discharge
al a point on a Jake's shoreline lor three cases: (a) unbounded Iluid with diffusion, (b)
unbounded Iluld with' diffuslon and advection, and (c) bounded fluld with diffusion and (a) Determine Ihe volume of Ihe lake when it is filled.
advection. (b) Ifthe Jake is vertically stratified inlo two layers at a deplh of 10 m, delermine Ihe
volume of the surface (epilimnion) and bouom (hypolimnion) layers.
(e) In this case Eqs. 16.35 and 16.36 are used with n = 2. The results are displayed in 16.2. The following polynomial defines lhe area of Cascade Reservoir, Idaho:
Fig.16.9c.
A(z) = 0.1675333(z
- 1453)4- ?446(z- 1453)3
+ Il5.5717(z - 1453)2- 101.05(z- 1453)
EXAMPLE 16.8. BACTERIAL MODEL FOR LAKE ERIE. Boyce and Hamblin
(1975) developed Eq. 16.35 and used it to simulate steady-state chloride profiles in the where A(z) = surface area (lQ4 m2) and z = deplh (m). NOle Ihat z = 1453 m is ¡he
central basin of Lake Erie (Fig. 16.10). They assumed that the controlling dim¿nsion bottom of Ihe reservoir and that depth increases upward.
of the lake (that is, its narrowest width) defines the largest energy-bearing eddy and (a) Delermine the volume ofthe lake when it is filled (z = 1473 m).
specifies the point at which the acceleration of the diffusion process comes to a halt. (b) If the lake is vertically stratified into two layers al a depth of 10 m from Ihe bot-
Thus for long-term computations, Fig. 8.11 with scale length spcdfied by the controlling tom (z = 1463 m), determine the volume of the surface (epilimnion) and bOllom
(hypolimnion) layers.
width provides a first estimate of the order of magnitude of the diffusion coefficient. For
example for central Lake Erie, the minimum width is approximately 80 km. Figure 8.11 (e) Develop a polynomial to predict volume as a function of depth.
indicates that the diffll~ion coefficienl for this scale length ranges from approximately 16.3. The following data have been reporled for Shagawa Lake, Minnesota:
5 X lOSto 8 X 1Q6cm2 ~ ¡:igure 16.lOb shows Boyce and Hamblin's (1975) simulation
using Eq. 16.35 and values of E = 1.5 X IQ6 cm2 s-I and n = 2. Deplh(m) o 246 8 10 12
Volume(%) 100 68 41 19,5 4.45 O

If the total volume is 53 x 106 m3, use this information to determine the area-deplh
curve for Ihe lake.

16.4. You are studying a small pond during Ihe sllmmer. "has Ihe following ,haraclcrbll. s:
/' Volume = 1 x lOS013 Surfacearea = I X 105m2
/' QlIlflow = 3 x 103 m3 d-I PrecipilallUn = 0.1 cm d-I
/' Inflow = 1 x 103 m3 d-I

(b)
FIGURE 16.10
Chloride model 01 lake Erie Irom
Boyce and Hamblin (1975). (a)
Chloride contours in ¡.1Mlor lake
Erie during 1970; (b) model resuhs
'1g Eq. 16.35.
You measure a pan evaporalion rate ofO.25 cm d-I. Assuming a pan cocflicienl oro.?,
determine the flow due to groundwaler.
16.5. Compule lhe evaporation flow for the lake in Probo 16.4 if
Wind speed = I mps
Water temperalure = 30°C
Dew-point lempcralllre = 25°C
294 PART11I Water-Quality Environments '1.

LECTURE 17
16.6. A conical mining pit is shown in Fig. PI6.6. The pit is located in an area where pre-
cipitation is negligible. The ratio or the radius to !he depth is 20: l. What is its ultimate
depth ir it is being filled by a constant groundwater inflow at arate or 1 x 1Q4m3 d-I?
Note that it is subject to the rollowing meteorological conditions: Sediments
Wind speed = 1mps
Water temperature = 28°C
Dew-point tcmperature = 22°C

- = ===--=
~ 7
~~: ' .
~/ FIGUREP16.6
LECTURE OVERVIEW: Following a general discussion ot sed1ment trarts'portl1l1cl
suspended solids, we present an overview of the mechan1sm of settUng. Thik
iÍ1c1udesa review of both observed values and the Stokes' 1Itwmóde!. Afler
16.7 Route the following hydrograph and pollutograph through the system in Examples 16.4
and 16.5. Note that the pollutant is conservative. providing background on the bottom sediments, 1develop a simple budget for
a well-mixed take underlain by a weU-mixed surface sediment tayer. This is
Time (min) o 20 40 60 80 100 120 140 followed by a description of models that characterize the sediments as a ills-
tributed system, that is, as a conlÍnuum. The lectUreends with an introduttion
Inl10w(m3S-I) O 4 10 10 4 2 1 O
0102010 O O O O to the topic of sediment resuspension.
Conrcntration(mg L 1)

)r..8. A lirst-order decaying pollutant (k = 0.1 d-I) is discharged at a continuous rate into For most or its history water-quality modeling has been concerned with how sol-
a large lake. Ir Ihe radius of the mixing zone is 100 m and advection is negligible, utes moved through aquatic environments. Consequently adequate characterization
determine the radius at which the concentration drops to 5% of its initial value. Note of hydrodynamics and water budgets have been an essential component of our fteld.
Ihat the dirrusion coefficient is 105 cm2 S-l. Over the past 20 yr there has been increased attention directed to how water pol-
\(,.'.1. 1\ municipality discharges sewage at a rale of 1 x 10-5m3 d-I at a point on a lake's lutants interact with solid matter. The interest originalIy focused on the simulation
shoreline. The wastewater contains coliform bacteria (200 x 106 number per 100 mL) of the settling or organic partic1es produced by eutrophication and their subsequent
that die al arate of 0.5 d-t. Calculate concentration profiles in the lake's vertically decomposition in the bottom sediments. Today modeling ofboth suspended and bot-
well-mixed surface layer (H = 10 m) for the following cases: 10msediments has been heightened by the recognition that many toxic substances
la) The lake is infinitely long and wide and horizontal diffusion (Ex = E)' = 106m2d-t) associate with solid matter. FinalIy in some systems, solids are considered a polIu-
is the sole transport mechanism. tant in their own right. For example the reproduction of some endangered fish can
(b) The same conditions exist as in (a), but with a current of IQ4 m d-I moving rrom
be affected by sediment deposition on spawning beds.
east to west.
Thus bottom sediments and suspended solids represent an "environment" that
(e) The same conditions exist as in (b), but with a finite width or 5 km.
must be understood to adequately simulate the transport and fate of many polIutants.
This lecture provides background information on this environment.

17.1 SEDIMENT TRANSPORT OVERVIEW

Oncesuspended solids are introduced into a natural water they are transported and
transformed by a number of different mechanisms. A portion of the organic solids

295
296 PARTIIJ Water-Quality Environments ~
LECTURE17 Sediments 297

will be lost by decomposition. The residual organic particles along with inorganic
solids are subject to a number of transport processes.
'fhe particles will be carried laterally by water currents. At the same time they
will settle differentially, depending on their size and density. Although a portion
will remain perrnanently on the bottom, solids can be reintroduced into the water by
~ur~ulence.~uch resuspension tends to occur due to strong currents and in shallower
are~ due to wind mixing.
Although the situation is very complex, some general pattems have been ob-
served. In particular, fine-grained sediments will tend to collect in low-energy areas.
'fhis has sóme interesting rarnifications for the water environments we studied in the
previous hn::e lectures.
for streams (Fig. 17.1a), deposition zones tend to forrn at low-energy regions
such as pools and the inside ofbends in meandering rivers. In estuaries (Fig. 17.lb),
a null zone is forrned where inflowing river flow and tidal action tend to cancel. This
convergence region is typified by higher particle concentration in the water and large
sediment accumulation.
¡I\ {akes (Fig. l7.k), wind- and current-induced lurbulence leads to the accu-
mulation of coarser solid matter in shallow water and finer particles at depth. Tbis
proce~§, r~ferred ~Qas focusing, means that a fine-grained-solids deposition zone
wiUbe fO!111~d at ~e center of a lake.
!fnpoundments exhibit two effects (Fig. 17.1d). First, a delta can form at the
point where the turbulent river enters the much lower energy impoundment. Second,
very pne sediments can be carried past the delta by density currents that hug the
!>ow>m.Suc~ advective transport can eventually lead to large accumulation of fine-
~raiq~d solids at the pool above the dam.
Finally all these deposition patterns can be modified by the ef!¡;¡;tsof humans
and extreme natural events. For example sediment beds are typically found in the
vicinity of waste outfalls. Further, floods and high winds can scour and redistribute
bottom sedimenls.
The level of modeling detail required to simulate lhe patterns depicted in Fig.
17.1 is beyond the scope of this book. In the following sections we cover introduc-
tpry inforrnation and some simple models related to both suspended and bottom
/ solids that has direct relevance to water-quality modeling.
17.2 SUSPENDED SOLIDS
Let's start our discussion with suspended solids. The following section provides in-
~~
forrnation on suspended solids properties such as typical concentration, size, and
pensity. Then I introduce Stokes' law as an idealized model of the settling mech-
anism.
17.2.1 Suspended Sollds Properties
Tbe suspended solids concentration of natural waters is reported on a dry-weight
basis. Values range from below 1 mg L-) for extremely clean waters to over 100
mg L-) for highly turbid systems. Levels for sewage and for storm-water overflows
can be even higher. Table 17.l1ists some typical values.
Although suspended solids are expressed on a dry-weight basis (that is, dry
weight of solids per volume of water), their dynamics require a more in-depth char-
acterization of their composition. One aspect of this characterization relates to thcir
origino

- - -
Nul!zone TABLE 17.1
-
_l. .1 ey.:
Suspended solids concentrations encountered

- -:t -~----------
~
Flne sedlments
4--""--
in natural waters and sewage (from Di Toro
el al. 1971, O'Connor 1988c, Lung 1994,
Thomann and MueIler 1987)
(b) Suspended solids
System (mgL-I)
Great Lakes:
Superior/Hunm 0.5
Saginaw Bay 8.0
Western Lake Erie 20.0
Flint River, Michigan 8-12
Rapid Creek. South Dakota 158
Clinton River, Michigan 10-120
Hudson River, NY 10-60
fe) (el) POlomac Estuary 5-30
Ja.' E\tuary, Virginia 10-50
FIGURE 17.1 ! Sa~I..,n"l\to-San Joaquin Delta, 50-175
r~ItElm~of deposition 01 line sediments in natural waters. (a) Top view 01 river; (b) side California
view 01estuary; (e) top view 01lake; and (d) side view 01impoundment. Raw sewage 300
298 PART11I Water-Quality Environments

Solids in natural waters have two primary origills: thc drainage basin and the
photosynthesis process. Particles from these two sources are formally referred to
as aUochl/ronous and auloe/tl/ronous solids, respectively. They are also informally
duhhcd "brown" and "green" solids, respectively, based on their general color. ...
The Iwo Iypes of solids diffcr in several ways: ....

Orgal/jc carf¡ol/ cOl//el//. Autochthonous solids are organic matter and hence
have high organic carbon contenl. In contrast allochthonous solids usually E
E G>
¡....
originate from soil erosion. Therefore allochthonous solids have much lower i...
organic carbon contcnt than fresh photosYllthetically-derived solids because 01
Z
allochthl1nous solids (1) include inorganic solids from weathering of rocks
that are low in organic carbonoand (2) their organic content has been reduced
I 11I
1...... I
1,
:2 "e
by decomposition. Further, thc make up of thc organie carbon differs. Freshly 'O '"
produced plants are eomposed of a variety of organic carbon compounds. Af-
11I UI I ¡""
8-
ter they die and begin decomposition. the more easily decomposable portion 'ti
e §
is lost lirsl. Thus with time. the fraction remaining becomes more refraclory GI
Q, fi 2 !,
....
(that ¡s, resistant to decomposition). Consequently the organic carboll in the 11I
:1 IIJ .!:!
allochthonous solids is usually less reactive. UJ ]1
.c ..
G>
Del/si/y. The density is the mass of a particle normalized to its volume. Thus
I e01 U 01
E E i:01 't:
densily has the same units as cOllcentralion. For example the density of wa-
;-1
:1
:z: Q. Ü
,.,
¡.... E
ter is approxirnately I g cm" J. Autochthonous solids tend to be much less
11I 01 01
dense than allochthl1nousparticles because inl1rganicminerals lend to be more I .. ü
-8
;¡o
mq¡
E (i)
'O ,.,
dense than organic carbono In addition fresh organic matter has a high water ()
content. Consequently, as listed in Table 17.2, organic matter such as bacteria SG> 8.
..
E
01'plant rnatter has a dcnsity that is very close to that of water. ..
:1
S;~e.A general classification scheme is shown in Fig. 17.2. AlIochthonous solids o
't:
tcnd to cover a wide spcctrum of sizes. Although organic parlicles also range :: :11 I
2
in sizc. thc principie Iypc of aUlochthonous solids (bacteria and floating uni- I :> ....
cellular planls callcd phytoplankton) tend to reside at thc lower end of the
I ..
rangc. However, it should be noled that both inorganic amI organic particles ¡gl I
.2 :1 ..G>
can agglomcrate to form larger assernblages called flocs and colonies. Such ,¡ .2
aggrcgation will have an effect on the particle's settling characteristics.
I GI
"i5,
::-r Eo
1:.
.... ui

Finally it should be noted that chernical precipitation can enhance settling. In

'3
¡;t
'O
' =: g e
pall ¡cular. hard-waler systerns cxpericncc calcium-carbonate 01'ca/cile precipita/ion .!I
iñ-
.... c:
TAULE 17.2
e .5
o 11I
;; Q)
Dcnsitics or water and u ea C\I:§
'! c: .t::
particulatc mattcr al CII !ECII ,.. '"
e al U al ... a.
ala. ~a.
9!
.c w-
a: o
Density CJ~ rn~ O
Suhstance ::) 11I
(~cm .1) CJ
~ :ti
-- iLi:ñ
Wuler
Orgunic malla
Wct-wl'ighl hasis 1.02-1.1
Dry-wcight husis 1.27
SiliccolIs mincrals 2.65
(¡urncl sunds 4

299
300 PART11I Water-Quality Environments LECrtlRE 17 Sediments 301

when plant productivity reduces carbon dioxide and raises pH. Such events some- EXAMPLE 17.1. STOKES' LAW. Determine the settling velocities for (a) phyto-
times produce so much ca1ciumcarbonate that they are commonly called "whitings" plankton (Ps = 1.027 g cm-3) and (b) silt (P., = 2.65 g cm-3) for the cases 01' diam-
due to the resulting milky appearance of the water. Organic compounds can sorb to eters 01' 10 and 20 ¡Lm. Assume that all particles are perfect spheres (a = 1.0 and
d = diameter).
these partic1es and phytoplankton can serve as a nuc1eus for calcium carbonate for-
mation. Hence as the partic1es settle, they carry organic matter with them. Wetzel Solution: Stokes' law (Eq. 17.2) can be used to ca1culate the settling velocity for the
(1975) provides a discussion of the mechanism. 1O-¡Lm phytoplankton,

/.. Vs = 0.033634(1.027 - 1)102 = 0.09 m d-I

17.2.2 Settllng and Stokes' Law
The other cases can be computed and tabulated
The settling velocity of a partic1e can be estimated using Stokes' law,
d = 10 ¡Lm d = 20 ¡Lm
- ex-g ps
_ - PW 2
Vs
18
-
( J.L )d
(17.1)
Phyloplanklon (P. = 1.027 g cm-3) 0.09
5.55
0.36
22.20 l
where Vs = settling velocity (cm S-I) Silt (P. = 2.65 g cm-3)

ex = a dimensionlessformfactorreflectingtheeffectofthe partic1e'sshape
on settling velocity (for a sphere it is 1.0) r
Nule huw increasing diameler and density lead to much larger scttling velucitics.
g = acceleration due to gravity (= 981 cm s-2) I
Ps and Pw = densities of the partic1e and water, respectively (g cm-3)
J.L = dynamic viscosity (g cm-I S-I)
d = an effective partic1e diameter (cm)
In general, partic1es in natural waters have complex shapes (which leads to val-
ues of ex lower than 1). Thus the results for both phytoplankton and other partic1es
Thomann and Mueller (1987) have reexpressed Stokes' law in a convenient form, would be lower than for Example 17.1. As summarized in Table 17.3 measured val-
Vs = 0.033634ex(ps - pw)d2 (17.2) ues for phytoplankton and organic solids average about 0.25 m d-I , with a range of
0.1 to 1 m d-I (Table 17.3). For systems with ca1cite precipitation, the average value
where Vsis in m d-I, the densities are in g cm-3, d is in J.Lm,and water viscosity is can rise to leveis on the order of 0.5 m d-I .
assumed to have a constant value ofO.014 g cm-I S-l.
According to Stokes' law the settling velocity is linearly dependent on partic1e
density and quadratically dependent on diameter. Figure 17.3 shows values for a TABLE 17.3
range of diameters and densities. These dependencies are i1\ustratedin the following Settling velocities of particles found in natural waters
example. (WetzeI1975, Burns and Rosa 1980)
100 Diameter Settling velocity
Particle type (!Lm) (md-I) l.
80
I
;:- Phyloplanklon:
2 0.08(0.24)'
ü:¡'
O 60 Cye/otella meneghiniana
Thalass'ios;ra nana 4.3-5.2 0.1-0.28 \
GiS 8.4 I
>- 40 Scenedesmus quadr;cauda
25
0.27(0.89)
O.2( 1.411)
A.rter;onella formosa
Thalass;os;ra rotula 19-34 0.39-2.1
20 50 1.9(6.11)
Coscinod;scus lineallls I
Me/os;ra agass;z;; 54.8 0.67( 1.87)
O 84
Rh;zoso/en;a robusta 1.1(4.7)
O 20 40 60 80 100 1-10 0.2
Particulale organic carbon
10-64 1.5
Dlameter (mm) >64 2.3
Clay 2-4 0.3-1
FIGUftE 17.3 Silt 10-20 3-30
Plot of settling velocity versus diameter lor various levels 01
particle density. This figure assumes that the particles are 'Parenthelical numbers are for Ihe Slalionary phase (see Lec. 32 for an ex-
perfect spheres (a = 1.0 ane! d = diameter). planation of the different pbases of microbial growth).
302 PART11I Water-Quality Environments LF.crURB17 Sediments 303

In practice few models use Stokes' law to directly delermine settling velocities volume is solid (Fig. l7.4b). Such systems are referred to as porous media. They
of organic matter. For one thing Stokes' law is predicated on the assumption that requirethedefinitionofseveralnewparameters. .
tlow is laminar, whereas tlow in most natural systems is turbulent. Further, living The porosity refers to the volume of the sediments that is in the liquid phase
particles such as phytoplankton can become buoyant due lo internal gas vacuoles. and is interconnected (Engelhardt 1977). Strictly speaking this exc1udes isolated
In addilion Iheir settling velocities may vary with their physiological state. For ex- pore space Ihat is considered part of the solid phase. However, because such isolated
ample as in Table 17.3, Ihey may exhibil differing settling velocities, depending on pores are ra~ely found in fine-grained sediments (Bemer 1980), the porosity fjJ is
whether or nol Ihey are aClively growing. Finally the formation of tlocs and precipi- operalionally defined as Ihe fraclion of the lotal volume that is in the liquid phase,
lales complicates the direct application of Stokes' law.
Vt
The settling velocity also depends on the model segmentation (Lec. 18). Coarse,
one-dimensional vertical models (e.g., Ihose Ihat treat Ihe water column as two lay- fjJ = V2 (17.3)
ers) typically use lower values than two- and three-dimensional characterizations. where Vt = volume ofthe liquid part ofthe sediment layer (m3) and V2 = total vol-
This has been attributed lo Ihe fact Ihat the one-dimensional approaches do nol ade- ume of the sediment layer (m3). Note that Ihe subscript 2 is used for the sediments
q\lalely characterize hydrodynamic phenomena such as upwelling and entrainment because we wil\ shortly use Ihis nomenc1ature when we model sediment-water in-
thal tend to reduce the effect of settling (Scavia and Bennett 1980). Thus a lower teractions (look ahead lo Fig. 17.5).
sellling velocity is used to compensale for the omission of Ihe mechanisms. The fraction of Ihe sedimenl Ihat is in the solid phase immediately follows as
Because of all the above faclors. the settling velocity is usually determined by
direcl measurement or calibration in most water-quality models. Nevertheless V
Slokes' law provides a useful theoretical reference, particularly in assessing the I - fjJ = V~ (17.4)
relalive effects of density, diameter, and shape on particle settling. In addition it
could become more directly useful in the future as both Ihe hydrodynamic charac-
where Vp = volume of the solid or particulate phase of the sediment (m3).
terization are refined and water-qualily modeling becomes increasingly involved
wilh problems involving inorganic solids.
17.3.2 Denslty and Sedlment Sollds Concentratlon
Another quanlity that is used in modeling porous media is the density, which can be
17.3 THE BOTTOM SEDlMENTS represenled in terms of more fundamental parameters as
Some of the suspended solids are eventually deposited and become part of a wa- M2
tcr body's bottom sediments. This seclion deals with how this environment can be p = Vp (17.5)
qtHlntified.
where p = density (g m-3) and M2 = mass ofthe solid phase in the sediments (g).
We can now use Ihe above quantities to define a number of parameters that are
17.3.1 Poroslty needed to model sedimenl-water inleractions. First, recall that the suspended solids
concentration is Ihe critical melric of the solids content of the water. A "suspended
In all the models described in the previous lectures, we dealt exclusively with the solids" concenlration in the sediments, m2, can be defined as
waler column. Because natural waters are typically dilute (Fig. 17Aa), it made
sense to represent concentration by normalizing the quantity of a chemical or solid M2
(17.6)
lo Ihe water's volume. We now tum lo a system, the bottom sediments, where the m2 = V2
silualion is nol quite as cIear cut. This is because a significant fraction of Ihe sediment
Equation 17.5 can be solved for

FIGURE17.4 .
/ M2 = pVp (17.7)
The conlrast between (a) dilute

D
:::::::::::::::::::::::::::::::::::::::::; natural water and (b) bottom Equation 17A can be solved for
,.........................................
::::::::::::::::::::::::::::::::::::::::::
sediments. The signilicant
':.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.: amount 01 solid matter present Vp (17.8)
::::::::::::::::::::::::;:;::::::::::::::: In the latter case must be I V2 = 1_ fjJ
accounted lor when developlng
(8) Dllute aqueous (b) Porous media mass balances lor sediment- Equations 17.7 and 17.8 can be substituted into Eq. 17.6 to yield
suspenslon water systems.
m2 = (1 - fjJ)p (17.9)
 I FlTIIRE 17 Sediments 305
304 PAR1'11I Water-Quality Environments

Thus we have reexpressed the sediment solids concentration in tcrms of pa- wherev, = resuspension velocity (m yr-I) and vb = burial velocity (m yr-'). Equa-
rameters that are conventionally used to measure porous media, We now use this tion 17.9 can be used to express sediment suspended solids 11'12 in terms of sedimcnt
porosity and density. In addition the subscript of m( can be dropped. At steady-state,
expression to develop a solids budget for a sediment-water system, In the follow-
the resulting solid balance equations are
ing discussions,it will be usefulto recognizethat the termsO - cP)prepresentthe
"suspended solids" concentration of the bouom sediments, o = Qmin- Qm - vsAsm+ v,AsO - cP)p (17.14)
and O = vsAsm - v,AsO - cP)p - VbAsO - cP)p (17.15)
17.4 ~IMPLE SOLIDS BUDGETS ~xt Eq. 17.15can be solvedfor
Now tbat we know something about suspended and bouom sediments, we can de-
~m (17.16)
velop a solids modeL For simplicity the model will be developed for allochthonous O - cP)p= v, + Vb
sotids in a well-mixed lake. As in Fig. 17.5 two cases will be examined. In the first
a ope'-way loss to the sediments is used. Then we couple the sediments and water which can be substituted into Eq. 17.14 and the result solved for
by ad4ing sediment resuspension. 11'1= Qmin
(17.17)
for fig. 17.5a the following mass balance can be wriuen for the water: Q + vsAsO - F,)
dm
where F, is a resuspension factor that is defined as
Vd( = Qmin - Qm - v.,Asm (17,10)
v,
where Vs = seuling velocity (m yr-t) and As = area of the sediment-water interface (17.18)
F, = v, + Vb
(ro2). At ste¡ldy-state~.
.. ", '.,
17.10 can be solved for
'

Note that Eq. 17.17is very similar to the well-mixed-Iakemodelsdescribed

Qmin
11'1=- (17.11) previously. The effect of adding the sediment layer is isolated in the dimensiouless
Q + vsAs
parameter group F,. This group represents the balance between the resuspension
. Now we can add a sediment layer to the model (Fig. 17.5b). Mass balances for rate a.w the total rate at which the sediment purges itself of solids (that is. both
so~idsin the waterand the sedimentlayercan be wriuenas burial and resuspension). Thus if burial dominates resuspension (Vb » v,). F, .-
Oand Eq. 17.17reducesto a well-mixedmodelwith no sediments. In contrast, if
dm,
V, (jf = Qmin - Qm, - vsAsmJ + v,Asm2 (17.12)
resuspensiondominatesburial(vb« v,).F, - 1andEq.17.17reducestom =
In other words when resuspension is relatively dominant. the water concentration
mln'

dm2 (17.13) will approach the inftow concentration because everything that seUles is immediately
¡ptd V2(jf = vsAsm, - v,Asm2 - VbAsm2 resuspended.
The above solutions are in the simulation mode. where all the parameters are

Loadlng Outflow Loadlng Outflow

. .. known. Altbough the solidsmodelcan be used in this way.it is moreconventional
for the model to be employed to estimate some of the parameters. In the next section
we show how this can be done.

17.4.1 Parameter Estimation

The parameters in the model are P. cP.m, min, Q, As, Vs.v,. and Vb.For the steady-
Resuspenslon Settllng
Settllng state case. Eqs. 17.14 and 17.15 represent a pair of simultaneous algebraic equations.
Thus given seven of the parameters, they generally provide us with a means of esti-
matingtbe remainingtwo. Although an algorithm can be developed for this general
problem.wetake a differenttack.Wewill try to assesswhichof the parametersare
~eastlikely to be available and tben show how they can be estimated with the modeL
1\. ;t Je) ~~ !tecllment-v.rettr
'ln'~r~p.J!on'"
(b) Sedlment-water
'.QJ!Jr~gt!on Of tbe nine param et .we assume tbat p and cP are known. 1'ypical values for
f/ . ")l," , fine-grainedsedimept are p= 2.4-2.7 x 106g m-3 and cP = 0.8-0.95. We also
-rIPn ~l11fti~
URI! 17.Q:\)h;
01a.~qll~s
I ,\ t1 ;j ~ l'
bud9E!1 fpr a. Vfell-mixed lake (a) without and (b) with sediment ílssumetbat
. tbe flow'
'1
,liE ,

d area, Q and As, are given.

'm~back. J
306 PART11I Water-Quality Environments LEC'I'UfU¡
17 Sedimen~ 1b7

Therefore we are left witlt five unknown pararneters: nt, min, vs, v" and V/1. it is recognized that they represent a simplificaüort ot the dyb~lt~ 6f iloliclsJrr~h
Now among these, one is extremely difficult to measure: v,. tbis is the focos of our
systems. In particular it is understoód that sediment resuspen!lionis Mí« §leady.§f!m
parameter estimation. There are two situations that usuaIly occur. process. Rather, it occurs episodically-usúatIy due.to high-winl:tevtnts lntakM 11M
First, there is the case where m and min have been measured. In addition the high-current events in rivers. In the last section of tbÍs lectUte we lhtrorltice a md~~l
settling velocity Vs may have been measured directly or estimated from literature that simulates resuspension episodes in lakes.
values. For example a value of 2.5 m d-I represents a typical value for allochthon- 41'
ous fine-grained particles (O'Connor 1988). Equations 17.14 and 17.15 can be
ac\ded to give 17.5 BOTTOM SEDIMENTS AS A DISTRIBUTED SYSTEM
o = Qmin- Qm - vbAs(l - 4»p (17.19) In the previous section we characterized the bottom sediments as a si ,. .ei.Atl
which can be used to estimate Vb, though such lumped models have proved useful in areas such as toxicant modé 'ng,
sediments can also be characterized as distributed systems. As depicted in Fi~. i 7.6
Q min-m
the simplest such approach views the bottom sediments as a one-dlmensionát coh-
Vb = As (1 - 4»p (17.20) tinuum in the vertical. '.
Three processes are shown in Fig. 17.6. The substance being modeled is sl1bj~d
_ Second, the burial velocity is sometimes measured direc~. This is often accom-
plished using sediment-dating techniques (see Box 17.1) to simple first-order decay. We also assume that it diffuses within the Pore water.
Regardless of whether it is measured directly or computed, once Vb has been Finally because solid matter tains down from lhe overlying water, substarices in (he
approximated the resuspension velocity can then be estimated by solving the steady- sediment are buried. Thus allhough a layer of sediment does not move physically,
state version ofEq. 17.15 for its distance from the sediment-water interface increases with time as matter accti-
mulates on the bottom. In essence the s'edimen!-water interface is aHvecting ü~'W&fd.
m _ Vb (17.21) From our modeling perspective, however, it is convenient to conceptualize the pro-
- v, = v"(l_ 4»p cess as if the interface were static and the sediments were advecting downward.
For the case of a dissolved contaminant the three mechanisms can be cothbined
EXAMPLE 17.2. SOLIDS BUDGET. Thomann and Di Toro (1983) presented tbe
foIlowing data related to the solids budget for Lake Ontarlo:
into the following mass balance: .,., . \
ac ac a2c
Volume = 1666 X 109m3 Suspended solids concentration = 0.5 mg L-I - = -Vb- + 4>D- - kc (17.22)
at az az2
Solidsloading = 4.46 X 1012g yr-I Flow = 212X 109 m3 yr-I
Area = 19.485X 106 m2
where e = concentration of a dissolved contaminant (mg L-1)
They assumed that the solids settle at arate of 2.5 m d-I (912.5 m yr-I) and that tbe Vb = the burialvelocity(m yr-I)
sediments have p = 2.4 g cm-3 and 4> = 0.9. Determine the burlal and resuspension D = an effective diffusion coefficientthrough the sediment pore waterS
velocities by a mass balance approach. (m2yr-I)
Solution: First, an inflow concentration can be determined as We have assumedconstaptpararnetersfor Eq. 17.22.It shouldbe nottd,thiU
4.46 X 1012 _ this is not necessarily true. With respect to transport, sedirnentsare subject to
mi. = 212x 109 = 21 mgL I
Next Eq. 17.20 can be used to calculate
Settllng Dlffuslon
IWater I
Vb = 212 X 109 21 - 0.5
19.485 X 106 (1 - 0.9)2.4 X 106 = 0.000929 m yr
_1
= 0.929 mm yr
-1
/ --- %=0
This result can be substituted into Eq. 17.21 to determine
eo
üi
:J
:1:
'6
.. »-
-
---
v, = 912.5 (1 _ 0.9~~4 X 106 - 0.000929 = 0.000972 m yr-I = 0.972 mm yr-I

I1 !
; Q
=: I Sedlments I
FIGURE 17.6
The simple budgets described above have been used in conjunction with contam-
inant balances to model toxic substance dynarnics in lakes (see Lec. 40). However, Z1 I
oI!!
Q.
- Schematic 01 a sediment
viewed as a vertical distributed
system.
308 PART11I Water-Quality Environments LECfURE17 Sedimenls 309

compaction as the weight of overlying sediments presses down on deeper layers. In Thus diffusion overwhelms advection. To calculate the distribution in the sedimellts,
the simplest sense such a process means that both the velocity and the ~orosity vary the eigenvalue can be determined as
with depth. Although in most water-quality contexts the effect is assumed to be neg-
ligible, modifications ofEq. 17.22 are available to account for the phenomenon (Rob-
bins and Edgington 1975, Chapra and Reckhow 1983). Á = ..0',002 [1- JI +4(492)] = -1.528m-1
Equation 17.22 is of the same form as Eq. 10.23 for a mixed-ftow reactor or This value can be substituted inlo Eq. 17.25,
an estuary. In fact we are treating the sediments as if they were an estuary on its ,/
e = lOe -1.528z
side. Consequently all the previous analytical and numerical solutions developed
in this text for such systems are applicable to the sediments. which can be used to calclll:.te the profile displayed in Fig. EI7.3.
A 95% penetration depth can be computed as
3
17.5.1 Steady-State Dlstrlbutions X95 = 1.528 = 1.96m
Suppose that the pore water at the sediment-water interface is held at a constant lev-
Thus as depicted in Ihe graph, the pollutanl penetrates about 200 cm inlo the sediment,.
el eo for a sufficiently long time so that the sediments come to a steady-state. For
this case, Eq. 17.22 becomes
de d2e e (mg L-I)
O = -Vbdz + cpDdz2 - ke (17.23) 5 10
with boundary conditions
e(O.1) = eo
( 17.24) 50
e(oo,1) = O
Using the techniques described in Lec. 9, the solution is 100
e = eoeÁz (17.25)
where
150 II ¡'
Á
Vb
= 2cpD(
I _ l + 4kcpD
v2
(17.26) ¡(cm)
200 .....
b ) FIGURE E17.3

EXAMPLE 17.3. STEADY-STATE SEDIMENT DISTRIBUTI0N. A contaminant

has a pore water concentration of 10 J.LgL-1 al the sedimc:nl-water interface. If it has a
half-lifeof 10 yr, how far will it penetrate into the sediments if.pD = 0.9x 10-5 cm2 sol
and Vb = 2 mm yr-)? 17.5.2 Time-Variable Distributions
Solution: First, let's ca1culate the reaction rate and express the other paramelers in com-
Althoughanalyticalsolutions are available for a variety of idealized cases (Carslaw
and Jaeger 1959), numerical melhods provide a more general approach. Equation
mensura~;~J' 17.22 can be integrated using any of the finite-difference methods described in Lecs.
k ~~ = 0.0693 yr-I Vb = 2 mm yr-I (Iex: mm) = 0.002 m yr-I 12 and 13. For problems where adveclion is negligible, the simple explicit FrCS
approach (Lec. 12) can be used to express it as
.pD = 0.9 x 10-5 cm2 S.I ~ 86.400s365d = 0.02838 m2 yr-I e~ - 2d + el
( I()4cm2 d yr ) el+1
1
= el,~+ -I-.D 1+1
tl~,
1 ,-) - kcl tlt ( 1727
. )
( )
Before ca1culating the concentration, we first determine the estuary number (recall Seco
10.5) to ascertain which mechanism dominates, advection or diffusion, Boundary conditions can be incorporated by developing finite-difference equa-
tions for the top and bottom segments. For the top segment (that is, the one at liJe
= k.pD= 0.0693(0.02838)
= 492 sediment-water interface), the simplest case is the Oirichlet boundary condition. For
1/ v~ 0.0022 this case the difference equation would be
310 I'ART11I Water-Quality Environments LECTURE17 Sediments 311

et - 2e' + el sediment oxygen dcmand in Lec. 25, we will deal with multiple reactants in a sedi-
et+ I = e' + "'D 2 I o - kc' !:J.t (17.28) ment system.
1 I ('f' ~Z2 1)
Further, many contaminants sorb onto particulate matter. Consequently the dif-
where the suhscripl Odcsignates the conccntration of the ovcrlying water. fusion mechanism can be reduced by this effect. In some cases, such as the radionu-
For the botlom segmcnt (subscript = 1/)a "at gradient can be assumed. This clide lead-21O, sorption is so strong that diffusion can become negligible (see Box
/II'~:II1S
thc net transport across the lower boundary is zero, 17.1). We will describe this sorption effect in detail when we discuss the modeling
of toxicsubstancesin the sedimentsin Part VII.
el c( -
(.r+1 = el + q,D 1/-1 1/ - kel ~1 (17.29)
"-,, ( tlz2 ")
BOX 17.1. Sedlment Datlng wlth Lead-210
":XAMI'LE 17.4. TIME-VARIABLE SEDlMENT DISTRIBUTION, Use the
~ame systCJJ1
as Example 17.3. Assume that at I = O the sediments are devoid 01'
Lead-210or 2IUPbis a naturally occurring radionuclide(an unstable isotope) that is de-
livered to the sedimenl-walerinterfaceat a constantflux rateoBecauseit sorbsstrongly
Ihe conlanlinant and Ihal the boundary concentralion is immediately brought up to the
andhasa long (but not too long) half-life 01'19.4yr, it is aneffective tool for quantifying
level 01' lO 111gL l. Use Ihe simple explicit approach to compute sediment proliles
the rate 01'sedimentationfor lakesandother standing waters.
dnring Ihe bloild-up in the sediments.
For a substanceIhat associatesexclusively wilh solid matter, Eq. 17.23 can be
Solutioll: rirst, let's delermine the lime slep lo achieve stabilily. Using a spatial scale expressedas
01 2.5 cm. 1'.11.12.16can be employed to calculate dI'
O = -v,,- - kc (17.30)
dz
~ 0.252
I = _(
'" ().O"'X1X
_. ) + O()
. 691. «(.) 2 5)-' = 1.1yr where V = amount 01' contaminant per mass 01' solids. When the contaminant is
measured on a mass basis Ihe units 01'v mighl be something like ILg g -l. For a radionu-
rherefore we will use a limc slep 01'0.25 yr. The resulting calculation is shown in Fig. dide like 2IUPb, where the amounl is measured in curies (see Lec. 43 for additional de-
E 17.4. As <'an be secn the sedimenl prolile takes over a decade lo build up to the sleady- tails on Ihis and other units 01'radioactivilY), the units would be Ci g-I. Aside from the
slale previ'''lsly compuled in Exumple 17.3. change in the expression for concentration, notice that the diffusion term has dropped
out 01'the model. Thus because the substance is not free lo move in the pore water,
the model becomes plug flow.
e (III~ 1.-1) 11'the concenlration al the sedimenl-water interface is 1'0and the burial velocity
o 5 10 and decay rale are constanls, Eq. 17.30 can be solved for
O .L
v = voe ¡¡¡;- (17.31)

50 According lo this equation the concentration should decrease exponenlially with depth
into the sediments. Consequently if the model holds, a semi-Iog plot 01' concenlration
versus depth shollld yield a slraight line with an intercept 01'1'0and a slope 01' -klv¡,.
100 Because Ihe half-life 01' 210Pb is known. the slope can be used to evaluate the burial
velocity. Problem 17.5 al the end 01'Ihe leclUre deals with Ihis calculation.
Aside from eSlimaling the burial velocity, the 210Pbtechniques can be used to date
150 sediments. Once the burial velocity and the decay rate are known, each stratum can be
assigned a time horizon below the sediment-waler interface by recognizing that time
and depth are related by \'" = xII. Problem 17.5also deals withthis estimate.
200
Finally it should be menlioned that other approaches are available to date sedi-
ments and hence eslimate the burial rateo A group 01' techniques are based on identi-
250 fying the remains 01'aqualic and terrestrial organisms. Thus the approach is similar to
archaeological dating basedon fossil remains.
z (cm) For example pollen in a watershed can change abruptly due to agricultural clear-
FIGURE E17.4 ing 01'land. disease, or the introduction 01'exolic plants. For example farming in many
parts 01'North America was accompanied by Ihe slldden rise 01'Ambrosia. or common
ragweed. Ir Ihe time 01'agriculturalization can be pinpointed, the sediment strata where
the pollen tirst shows up is associated with Ihat time. Wetzel (1975) provides a nice
In this :;edion wc have limilcd ollrsclvcs to a single. dissolved compound. It discussioll 01' this and other methods.
sltollld hc noled thalmany prohlems in\'olvc several compollnds. Whcn wc disclIss
 ..--

312 PART11I Water-Quality Environments

LECTURE17 Sediments 313
17.6 RESUSPENSION (ADVANCED TOPIC) ,- Fetch -,
As depicted in Fig. 17.7 the resuspension of sediments is a function of several factors. Wind
The process begins with the energy delivered to the water surface via the wind.
The amount of energy is dependent on both the wind velocity and fetch. The latter
refers to the length of exposed water surface in the direction that the wind blows. ,~
This wind energy creates waves. In general the greater the wind velocity and /
fetch, the greater the height and period of the resulting waves. Under the surface
Ute water qloves in circular eddies, the energy of which dissipates with depth. The
orbital velocity ofthese gyres (Iarge eddies) exerts a shear stress at the bottom. Scour
or resuspension of the sediments can then occur, depending on the magnitude of the
stress and the type of bottom sediments.
Engineersand scientistshave developeda variety of formulas to quantify the FIGURE 17.7
proceSs.The following development uses the approach outlined by Kang et al. (1982) A simple model 01 how wind induces wave action that leads to ~J
JOsimulate wave action and bottom shear stresses in Lake Erie. sediment resuspension.

Wave height, period, and length. Ijima and Tang (1966) have developed for- where L = waveleugth (m) and Lo = deep-water wavelength (m), which can he
mulas to predict wave height and period for shallow waters. These equations are estimated by
T2
dependent on mean depth, wind velocity, and fetch, ..
Lo = ;; (17.3ó)

gH
0.75
tanh
(
0.0125 fJ;
r.42
H 0.75
(17.32) Orbital velocity. The orbital velocity created by the wave can be determined by

g:; ~ 0.28"0'+53 (U,)] { Ionh[0.53(¡¡; j ]

}
- 7TH.
U = T; sinh(27TH/L)
100 .
(11.37)

where U = orbital velocity (cm s-I). The hyperbolic sine can be computed as "
0.25 eX - e-X
H 0.375
( ) 0.375
0.077 fJ; > (17.33)
sinh(x) = 2 (17.38)

::~ ~ l.2 Ionh[0.833 (~) ] ~oh ¡ Ionh[0.833 ¡¡¡() ]

Shear stress. For shallow lakes where currents are generally small,
stress can be approximaled by
the shcar

T = 0.003U2
where H. = significant wave height (m) ( 17.39)
T. = significant wave period (s) where T = shearstress (dyne cm-2).
U = wind velocity (m s-I)
H = mean depth (m) Bottom scour.
F = fetch (m) be calculated with The amount of mass of sediments scoured from Ihe bottom can

Note that the hyperbolic tangent is available as an intrinsic function in Fortran as

e =O T S; Te
well as on other software such as some spreadsheets. It can be also be computed ao 3 (17.40)
with the exponential function e = 2"(T - Te) T> Te
Id
eX - e -x
tanh(x) = + e-X
(17.34) 1':
where e = mass of sediments seoured (g m-2)
eX ao = 0.008
The wavelength can aIso be determined by solving the following equation iter-
Id =7
ative'y, Te = criticalshearstress(dynecm-2)
(17.35) t;r
L = LotanhC~H) Suspended solids concentration and entrainment rateo Experiments have
shown that the enlrainmenl process lakes on Ihe order of an hOllral a conslanl shear
314 PART 11I Waler-Qualily Environmenls LECTURE17 Sediments 315

stress. The entrainment rate E(g m-2 hr-I) is constant over this period and consid- Substituting!he parameters for!he present problem into Eq. 17.35 ~ives,
ered negligible thereafter,
e
t s I hr L = 22.81025tanhel.:16)
E=Thf (17.41) Using Lo as the inilial guess for L, we can solve the equation iteratively for
E=O t> I hr
Iteratlon L % error
I"hcresulting suspended solids concentration can be estimated by
O 22.::10
e = IOOOO~ (17.42) I 20.081 13.59
. H 2 20.897 3.91
where e = suspendedsolids concentration(mg L- 1). Further, the results can be 3 20.660 1.14
4 20.730 0.336
expressed as a resuspension velocity by 5 20.710 0.0983
E 6 20.716 0.0278
Vr=- (17.43) 7 20.714 0.00843
(1 - f/J)p
Inspection of Eqs. 17.39 and 17.40 indicates that below a critical shear stress This result can be substituted into Eq. 17.37 to calculate the orbital velocity
no resuspension will occur. Above the criticallevel, however, resuspension is depen-
dent on the orbital velocity raised to the sixth power. Thus scour will be particularly - 1T(1.03) 100 _1
critical during extreme storm events. U = 3:82 sinh[21T(5)/20.714] = 39.15cms
The orbitalvelocitycan then be employedto delerminethe shearstress,
EXAMPLE 17.5. CALCULATlON OF SEDlMENT RESUSPENSION. A uni-
formwind blowsover a rectangularbasin as shownin Fig. EI7.5: 'T= 0.003172= 0.003(39.15)2= 4.6 dynecm-2

-
U ,
~-----------------------
which is used to calculate the sediment erosion rate

~
_ 8 X 10-3(4.6 _ 1)3 = 0.0076 g cm-2
- 49
-
F This result can be normalized lo depth to yield a concentration in the water,
FIGURE E17.5

e = lO,oooO.~76= 15.22mgL-1
The crilical shear stress is equal lo 1 dyne cm-2. If the wind velocily and Ihe deplh
are equal lo 20 mps and 5 m, respeclively, delermine Ihe sediment resuspension al a It can alsobe expressedas a resuspensionvelocity,
puinl F = 10 km down Ihe basin.
80lutloo: First, Eqs. 17.32 and 17.33 can be used lo delermine the significanl wave _ 0.0076g cm-2 hr-I _ _1
height and period, Vr - (1 _ 0.9)2.5g cm-3 - 0.0304cm hr

Hs = 202
9.80.283""h [053 (9.:; S)"'] ,,"" 0.0125 (9.8x,.:,
lanh 0.53 9.8 0X000
{
5 \075
[ ( 202)
r ] )
_
- 1.03~/
/
PROBLEMS
17.1. Use Stokes' law lo calculate the settling velocilies for the phytoplankton and the in-
organic solids (c1ay and silt) in Table 17.3. Assume that!he form factor is 1. Develop
a plot of the measured velocities versus your calculated values. Plot both !he grow!h
and stationary values for the phytoplankton. Use differenl symbols on !he plot to dis-
T = 21T(2W criminale among inorganic solids. growlh-phase phytoplankton, and stationary-phase
I.2I'"h[0.833(9.:; Sr']"'h 0.077(9.8x;.ooo)'" phytoplankton. Discuss the plol.

{ ) 17.2. Repeat Example 17.1, but use Ihe area ofthe deposition zone (10,000 kJn2) ratber than
~ 3.82, ",h [0.833(9.:; S)""'] the lake ~urface area as the actual interface across which the vertical transfers take
place.
316 PART11I W..lcí-Quality ~nvironments

17.3. A lake is fed by two rivers as shown in Fig. PI7.3. The inflowing rivers deliver solids
LECTURE 18
at rates of 0.1 x 109 and 0.2 x 109 g yr-I, respeclively, and Ihe oullel earries 0.05 x ---
109 g yr-I. The lake's deposition area is 106m2.
(a) If the sediment
loeity.
has p = 2.5 X 106 g m-3 and tP = 0.9, delernllll! Ihe burial ve- .,.' The "Modeling" Environment
(b) Sediment traps (area = 250 em2)are deployedin the lake.These yield an annual
.-.'
average aeeumulation rate of 20 g yr-I of solid matter. In addition the average
suspended sotids eoneentration in the water is measured as 5 mg L -l. Detennine
the settling veloeity.
(e) Estimate the resuspension velocity.

FIGURE P17.3

17.4. A lake has a steady-state eoneentration of 100 fJ.gL -1 of a dissolved substanee that
has a half-tife of 28 yr. Assuming that the sediments are also at a steady-state and that ..~~ j)~~~¡,] 4sYPf~dIe b~}~9f the JeC;tI1re tJteYfU'iQUS
steps that cQn- ~
~~~.~~frj~tiQP~
the burial veloeily is negtigible, determine the llIoleeular diffusion eoeffieient based on
the following sediment data (p

Deplh(cm)
= 2.5 X 106 g m-3 and tP

o 50
= 0.9):
100 150 200
. . ".1.4WQ~hptprocess. ![PIS is foUowoo

,
9J!11~
,
.'

,
. r.~gP9:
,
',~.eqIp'l~rrd fgr
,
I\DPn~hif pf.:--ptodet Yme.
_.
"
, ,

sP'tf~'llqd
nClP4
flIUi1'y 1Cotwo
' ,,
of tr.c~~.Jqn~s
mqdel-qeveJopm?nt
e t e<:tqre WJtb tl\sks:~e.nsiHyi!),'
"

,
, ,
'

,
,
', ,
I1~
,
'
, ,
~netic sCllles.
, ,
"

"
,
~H~lY~J§
. !IDovetVieW Qf the IlJterre.

100 60 40 20 15
011___
Concenlration(¡.tgL-1)

Because their primary intent is to provide Ihe reader with an organizcd body of
17.5. Determine the buri¡¡1 locity and sediment deposition flux for Lake Michigan based knowledge, many textbooks give Ihe impression that the world is a "''I'Vneat place.
on the data below for llOPb (assume that p = 2.4 X 106 g m-3 and tP = 0.85). Also if In fact the "real world"is fraughtwithambiguitiesanddoesnot USII;,;\ yidd to our
the sediment eore were taken in 1975. determine Ihe date at which the layer at 3 cm effons in the same "black-and-white" fashion as in the end-of-chap¡'.'1problems.
was deposited. Water-quality modeling seems to be especially prone to such ovcr~ill1plificalion
beca use, before you can even develop realistic models, you must first assimilate a
Deplh (cm) o 0.5 1.5 2.5 3.5 4.5 5.5
considerable body of technical information dealing with theory, mathcmatics. alld
Concenlralion (lO-12 Ci g-I) 8.1 7.2 5.4 3.6 1.5 0.6 0.5 computers. As a consequence the neophyte can come away from the lopk thinkillg
that ies aIl about writing mass balances and solving differential eqllallolJs.
( In reality the mathematical aspects are but one facet (albeil an IIlIp..r~antolle) 01'
17.6. Suppose that 5 Ci of a dissolved radionuc1ide (half-tife = 20 yr) is spilled into a 5-m-
deep lake that has a residenee time of 10 yr and a surfaee area of 2 x 106 m2. Caleu-
V ' how models are developed and applied to sol' water-quality-modeling problems. It
1 is the task of this lecture to provide a larger plcture 01'the modeling process.
late the prolHe in the sediments over the first 10 yr following the spill. Assume that thr
;j /
burial veloeity is negli¡!ill ...! Ihe difh ¡\'n coefficient is 5 x 10-6 cm2 SI.

17.7. Repeat Exampk ¡
suspended sotid~ concclIlrallOn
, ¡rom O to 20 m s -l. Plol Ihe resulting 18.1 THE WATER-QUALITV-MODELlNG PROCESS
versus wind velocily. Diseuss your resuhs.
A water-quality model is merely one linkage in a larger management process (rccall
Fig. 1.2). FUr1herthe watcr-quality model itself does not arise out 01'Ihin air. Rather
it results from a multistep model-developmenl ¡¡ndapplication proccss (Fig. I g.I).
 318 I'ART11I Water-Quality Environments LECruRE 18 The "Modeling" Environment 319

This lecture is designed to give you a glimpse of the larger picture by providing MANAGEMENT OBJECTIVES,
OPT1ONS ANO CONSTRAINTS
background information on the model-development process. In so doing 1 introduce
tasks such as model calibration and confirmation that must be implemented prior to PREEXISTlNG OATA
management decisions. I also cover some techniques such as sensitivity analysis that tPREUMINARY
DATA COUECT1ON)
ari~e in model application.

18.1.1 Problem Specification: Getting Started

Before a single equation is written. as depicted in Fig. 18.1, the modeling process THEORY
slarls with a problem specijicati()// phase. In other words the waler-quality engineer ~---..
must be provided with a clear delineation of the objectives of the customer. This can
be an individual or a decision-making entity such as a corporation, a municipality,
or a regulatory agency. Oftentimes, because he or she usually has broad expertise re-
lated lo water quality, the modeler is included in this process. This is particularly tme
when the decisi('n-making entity requires additional technical expertise to clarify its
ullimate goab.
Note th¡.t t\Voprimary information sources feed this phase. The first are man-
agement objectives, control options, and constraints, which might include physical :.-
constraints as well as legal, regulatory, and economic information.
The secolld source is data related to the physics, chemistry, and biology of the
waler body amI its drainage basin. Oftentimes such information is sparse or nonex- .. IDATACOUECT10N I

iSlcnl. In such cases some preliminary premodeling data collection could be in order.
When this phase is complete, the modeler should have a clear idea of the problem ;-01-- --- - - - - - - - - -- -- -- DATA COUECT1ON

objcctives and the accompanying water-quality variables required to assess whether

Ihe objectives are attainable. In addition, based on the available data, she or he should
have a preliminary idea of how the system presently functions. In particular the
modeler should understand the temporal, spatial, and kinetic resolution that will be
compatible wilh the problem needs. 1'11elaborate on the topic of model resolution at
the end of this lecture.
Finally il should be noted that modelers often provirle simple back-of-the- Actual
enve10pe calculations during this phase (recall Box 4.1). These are usually helpful water quaUIy .. IDATACOUECTlON I
in bounding the problem and providing rough estimates of system behavior.
(a) Process (b) Informatlon
18.1.2 Model Selection FIGURE 18.1
(a) The water-quality-modeling process along with (b) the necessary information needed for
Once you know what you're dealing with, the next step is to obtain a mode\. In some its effective implementation.
I instances this can be accomplished by choosing an existing software package. This
is obviously Ihe preferred route for two reasons. First, someone else has done ~ for the entire range of pollutants encountered in natural waters. In addition you may
work for you. Second, there are several models that have been widely used and, as a be confronting a new problem that has never been modeled before. In such cases you
consequence. have a certain cachet in legal and regulatory contexts. Thus they have must develop your own model from scratch. This consists of two phases: theoretical
crcdihility in Ihe eyes of judges and other decision makers.
development, and numerical specification and validation.
UnfortunalcJy there are many waler-quality problems for which canned pack-
ages are inadcquate or unavailable. These renect both kinetics (for instance, they Theoretical development. First, you must provide a theoretical development
don'l ~imulalt' Ihe right pollutants) and the time-space scale considerations. For ex- that specifies the required variables and parameters along with the associated con-
ample current interest in nonpoint source pollution has prompted the need for water- tinuity equations. The continuity eqnations are most frequently written for mass
qua lity models with built-in hydrodynamics. Unfortunately these are not available
 320 PAR1'11I Water-Quality Environments I.E(TlIHE IX The "Modeling" Ellvln>nment 321

andJor energy. In cases where hydrodynamics are simulated, momentum balances I model that gives inadequate predictions 1'01'the problem being addrc~sed. Con se-
are also ineluded. .
~ quentlyadditionalinformationmustbe cI)llled tobringthe modeluptotherequired
level.
One of the most important aspects of the theoretical design phase is the issue
of model complexity. As depicted in Fig. 18.2, there are trade-offs between model AII this is relaled to the fact that you shuuld always strive to deveh,p Ihe simples!
complexity, uncertainty, and information. The straight line indicates the underlying . model that is consistent with the data and Ihe problem requirements. This dicturn is
assumption that, if we have an unlimited budget, a more complex model wiII be ~ sometimesreferredto as applyingOckham'srazort to the model.The idea was also
more reliable. In essence as we add complexity to the model (that is, more equations
with more parameters), we assume that there are sufficient funds to perform the nec-
f expressed by Einstein. To paraphrase his words, "every model should be as simple
as possible, but no simpler."
essary field and laboratory studies to specify the additional parameters. In fact this
assumption itself may not be true, because there may be limits to our ability to math- Numerical specification and validation. After we develop the theore¡ical
/ ematically characterize the complexity of nature (sort of a "Heisenberg uncertainty
specifications, the equations are implemented on the computer. This phase cnn-
principie" of ecology). However, even though we may not be a!tk lOtotally charac- sists of several substeps, ineluding algorithm design (that is, specification ol' data
terizea ,. ..turalwatersystem,wegenerallyfunctionunderthe nOlionthat if wehave structure and numerical solution techniques), coding in a computer language, de-
more and better infonnation, our models will be more reliable predictors.
bugging, testing, and documentation. The numerical melhods are chosen to generale
The real rub comes from the fact that we almost never have a large enough sufficiently stable and accurate solutions 1'01'Ihe minimum computational effort.
1 budget to even come elose to approaching that limit. Rather we are usually in the
Before proceeding we should elaborate a bit on the testing phase of program
j position that we must make due with a limited sampling and lab budget. In such development. One of the most common mistakes made by novice programmers is lo
cases there are two extreme outcomes: at one extreme a very simple model wiII be assume that if the model yields "reasonable" results, it is adequalely tested. Thus il
so unrealistic that it will not yield reliable predictions; at the other, a very complex
results are positive (1'01'variables such as concentralion, which shollldn't go lIega-
model will be so detailed that it outpaces available data. Consequently the second tive), relatively smooth, and have expected magnitudes, Ihe compulation is deemcd
model can be equally unreliable because of the uncertainty of the parameters. As successful.
\ in Fig. 18.2 there is an intermediate point where Ihe model is consistent with the In fact such criteria are no guarantee that Ihe results are correct. For example It is
) available level of information.
relatively simple to violate the conservation 01'mass and still obtain smooth, posiu ve
At this point a third dimension must be interjected: the reliability required to
values that are in the right ballpark. Consequently the model should be subjeclcd
solve the problem. Clearly you might be at the optimum for your data but have a I) to a battery of tests to ensure that it is mathematicalIy correct. This is sometimes
') referred to as numerical validation (Oreskes et al. 1994). The following are some
specific suggestions for establishing that the model cakulation is valid:
Cost=unlimlled $ () · Mass balances. The model should be checkcd to ensure Ihat mass is h:danced. 1''':11
is, over the period 01'the caIculation the change 01'accumulation Sh;'Li.JcqU¡.: .;e
integral (that is, the summation) 01'the sources minus the sinks 01'mass. Sl'lIIt:
water-quality models do this as a part 01'their algorithms. For example Ihe Corps
ofEngineers two-dimensional reservoir model, CE-QUAL-W2 (0,1.' :md Buchak
~
:o Required rellability
1995), ineludes a mass-balance check as an automatic part of each rlln.
.!! -- -- -:.: -~ -". :-..:-.:: -~ -- - - - - - - - -. - - - - - - - - --- - - -. -. - - - -.
1) · Simplijied sollltiom'. The model output can be compared with simplilied CilS';s
'ii ",...( : 't-, for which closed-form solutions are possible. For example time-variable models
a:
".,.: ,: ":, can be run to steady-slate. The results can then be checked againsl c1osed-fnrm
'/ : : : , steady-state solutions. Similarly analytical orclosed-form solutions al'';sometillll';
: : : t available 1'01'idealized geometries and linear reaclion kinetics. The model can lit.:
'l
: i i
... Casi = $ + ó$
......... evaluated to see how well it matches sllch solutions.
/ \ · Rallge of cO/lditiom'. Although the foregoing suggeslions can help lo identify lIIa-
Casi =$ I jor model flaws, they will not ensure that Ihe model operates under a wide mnge
of conditions. Therefore the model should be tested 1'01'robustness by applying it to
Complexity
AGURE 18.2
'Named after .he monk William of OdhulI1.
The trade-off between modal reliability and complexity.
 122 PAR! 111 Walcr-Qualily Environmcnts LEITUREIR The "Modeling" Environment 323

different types of systems under differing initial conditions, boundary conditions, .0._ ¡
and loading scenarios. tlon --- ~.~¡;.¡..yiics
¡
:oBoundary
d con
dltlons i¡
. Graplzica/ resul1s. Although it does not address specific problems per se, a graph- loadlngs ¡
Data conec
ical interface is invaluable in assessing model performance and identifying model :¡"nltlal.~()"
an dltlons;
=.:.:.::.:.:......
l1aws. For example unless negative results occur, mild instabilities (that is, nu-
1 MODEL!
IIIcrically generated oscillations) are extremely difficult to identify by perusing Calculated
state
nUlllcrical output in the form oftables. However, they are readily visible as graph- Fleld and lab studl~ loKlnetlc parameters
Llterature values variables

· ical output.
Benchmarkillg. Finally benchmarking or "beta testing" involves having a wide
number of users implement the model. Not only does this process result in iden-
Callbrate
(that la. adJ.."
e"

tilication of errors, but it also provides guidance on how to modify the model so parameters
Error = (cm - C,
lhal it more adequately meets the user's needs.
Measured
state
variables
18,1.3 Preliminary Application cm
I T~
The next step in the process is to perform some preliminary simulations for the sys-
Callbratlon complete Data collectlon
tem being modeled. Although necessary data will often be lacking, the exercise is
tremendousiy useful for identifying data deficiencies and theoretical gaps. In partic- FIGURE 18.3
ular it can provide an important context for designing the field and laboratory studies Schematic diagram 01the model calibration process.
reqllired to fill the gaps.
Aside from identifying obvious information needs, this phase can also be useful These inelude
in identifying which model parameters are most important. When working under a
funiJ.edbudget, it is important to reco&!1i~ wh~ch mod~arameterShave the greatest
. Forcing functions and physical parameters:
( Boundary conditions and loads
im~ct on model Qredictions.lfthis can be done, more ofthe available resources can
) Initial conditions
be directed toward determining these parameters.
One way to identify the important parameters is with a model sensitivity analy- IPhysics
sisoIn its simplest form a sensitivity analysis consists of merely varying each of the . Calibration parameters:
parameters by a set percent and observing how the predictions vary. A later section '1 Kinetics
in lhis lecture provides details on how this can be done.
Notice how we have separated the kinetic parametersfrom the rest of the in-
formation. In essence we are assuming that sufficient information is availabie to
18.1.4 Calibration adequately define the other parameters. Thus before beginning the c~librationpr9-

: The next step in the process is to calibrate or "tune" the model to fit a data set. This
, consists ~
~'lrYing th~mo~elparameters to ol>.!a~n_llll.9Ptimala~reement between
1 TI1emod¡:1ca.!c~lations and the data ~~t.
- The calibration data set shouid be as similar as possible to the design condition
for the problem being studied. For example if the model is being used to analyze a
wastewater clischarge into a stream, the design condition wouid usually be for tile
summer low 110w(that is, summer 7QIO as described previously in Lec. 14). Conse-
quently a data set collected during the summer under 10w-l1ow conditions would be
most appropriate, whereas calibrating the model for a data set collected during the
spring runoff would not be as effective.
After the data set is collected, there will be many parameters that can be varied.
However, rather than arbitrarily trying to vary all of these, there is a systematic way
lo attain the hest fit (Fig. 18.3). To IInderstand how this is done, it shouid first be
!ecognized that thcre are several typcs of information that must be fed into the model.
~ cess, we must measure the syst~m-'~.!!>.ads,~oundaryconditions, and initial condi- _
-tiOfiswTllisufñeielll predslon that they ar~not considered to be a significant source
,¡of lincertainty. Similarly an accurate characterization of the system's geometry and
~ hyoraullcs must be developed. Once this is done, the physical parameters should not
1 be varied.
When this is accomplished, the focus of the calibration becomes the adjustment
I of the kinetic parameters. This process also should be approached in a systematic
fashion. For example if certain kinetic parameters have been directly measured in
an adequate fashion, they should be fixed.'Further some parameters can be estimated
I from the measured values of the state v;"iables. Finally the remaining parameters
can be varied within reasonable ranges until a best lit is attained. That is, adjustments
\ are made until the simulation matches the data in some optimal sense.
There are §everal approaches for _makingthe adjustme_nts.At one extreme, tun-
ing can proceed in a trial-hy-errorfashion. At the other, automated techniques such as

324 PART11I Water-Quality Environments
\ least-squares regression can be imp1emented. Both are il1ustraledin an example later
1 in this lecture.
I
\Ihey can be dislurbing lo Ihe original modelers, such discrepancies can be eXlrclllcly
Iuseful sincc Ihey often provide clucs lo missing mechanisms and informalion. Thus
18.1.5 Confirmationand Robustness
~ Once you have developed the calibrated model, al1you real1yknow is that the model
Ifits a single data setoBut befure the model can be used with confidence to make man-
agement predictions, it must be confirmed, To do this the calibrated model should
I be run for a new data set (or ideal1y, several data sets), with the physical parame-
{ ters and the forcing functions changed to reflect the new conditions, In contrast the
I kinetic coefficients must now be kept fixed at the values derived during the original
( calibration. If the new model simulations match the new data, the model has been
confirmed asan 'étTectiveprediction tool for Ihe range of conditions defined by tht;
original calibration and the confirmation data sets. If there is no match, the model
( can often be analyzed to determine possible reasons for the discrepancy(s). This may
Ilead to additional mechanism characterization and model refinement.
Even when the model is deemed adequately confirmed, Ihe simulation wil\ not
exactly match the confirmation data setoThus some investigators fine tune the model
after confirmation so that il optimal1y matches bOlhthe calibration and confirmation
data.
It should be noted that in the past, water-quality modelers referred to this phase
as "verification." This jargon implied thal once the model successful1y simulated an
independent data set, il represented an establishment of"truth"; that is, it represenled
an accurate representation of physical reality. In fact this can never real1y be proved
~ absolutely. At best al1 that can be concluded is thal our lesting has nol proved the
~ modelwrong(Oreskeset al. 1994).
This then brings us lo Ihe notion of model robustness. If we can never total1y
verify a model, we are len with the quality or rigor of its confirmation. As put by
( Oreskes et al. (1994), "the greater the number and diversily of confirming obser-
vations, the more probable it is that Ihe conceplualization embodied in Ihe model
'
is nol flawed." Such a model is said lo be robusto Therefore the aClual goal oflhe
l confirmation process should be to establish Ihe model's robuslness.
f
18.1.6 Management Applications
Many modelingstudies result in remedial actions. For example wastc treatment
\ plants might be built or upgraded, or environmental modifications such as aeration or
( dredging could be implemented. The effecliveness of Ihese actions can be evaluated
~by modifying the model paramelers and forcing functions and running (he model to
) predict the effects of the changes on the state variables.
18.1.7 Post-Audit
( After implementation of remedial actions, a check can be made to learo whether
! the model predictions were valido In a very few cases such model post-audits have
IH'TlIKE IX The "Modeling" Envirunment 325
shown excel1enl agreemenl belween Ihe prediclions and Ihe resulting waler quality.
However, in olher inslances, such is nOIIhe case and differences occur. Although
Ihey provide a furlher way for model developers lo improve Ihe robuslness of thcir
frameworks.
BOX 18.1. Great Lakes Cladophora Model
Over the past 40 years the Great Lakes have been threatened by a number of waler-
quality problems. One of the most visible is the growth of nuisance filamentous algae
such as CladopllOra glomerata on rocks, breakwalls, buoys, and other structures inter-
secting the water line. Profuse growth 01' such organisms takes place in areas enriched
with nutrients. The problem is particularly vexing because it occurs near shore, whele
people come intu the most frequent contact with the waler body.
In the late l 970s and early 1980s Professors Ray Canale (University 01'Michigan)
and Marty Auer (Michigan Technological University) conducted a major observational
and modeling slUdy of these organisms. Ahhough they mapped Cllldoplwra distribu,
tion in all the Great Lakes, their modeling study focused on the Harbor Beach area uf
Lake Huron.
Aside frum representing a landmark in the modeling of large aqualic plants, tllls
study was also significanl from Ihe perspective of model developmenl. This was due
lo two faclors.
Firsl, after the model equations were developed, a Iradilional model calibration
was not performed. Rather Canal e and Auer directly measured all the model rales in
the laboratory or in the field. For example lo determine the planl pholosynthesis and
respiration cales for the model, they collaboraled with Dr. Jim Graham 01' the Univer-
sity of Wisconsin to grow the organism under a conlrolled environmenl in a ccoss-
gradient room called a BIOTRON (Graham el al. 1982). This facilily allows plants lo
be grown and measurements to be carried oul sillluhaneously over a malrix of lighl and
temperature conditions. The resulling measurelllenls were Ihen used lo develop modcl
parameters in a fashion similar to the kinetic analyses previously described in Lec. 2.
By measuring all rales independently, Canale ami Auer eliminaled calibralion param-
elers from the resllhing model. Thus for beller or worse, it yielded predicliuns with n"
tuning.
Second, Canal e and Auer then used the model to simulate their system in a "beforl~
and after" sense. In other words they tirst employed the model to sillllllalc polluted
conditions in 1979 and unpolluted condilions in 1980. The pollulion reduction was due
to the installalion of phosphorus removal by Harbor Beach between their Iwo sampling
periods. In both cases the model paramcters were set al the vallles dctermined in the
laboratory and ficld mcasurements. Thus only the loading and physical parallll~lcrs such
as light and temperature were changed to renecl the differences belween the 2 years.
The results of the model simulations are shown in Fig. B 18.1. The model did a
good job in caplUring the improvement in soluble reaclive phosphorus levels in the
water. Although the response uf the Cllldophora standing crup (biomass) does nut ap-
pear as accurate, it mllst be considered Ihat the biomass measuremenls are mllch less
precise than for a chemical measuremcnt like phosphorus. If Ihis is takcn into consid-
eration, Ihe model generally capturcd Ihe impruvemenl in biomass, parlicul:lrly in .he
vicinity of Ihe outfall (dislance = 0.0), where Ihe problem was grealcsl.
326 PAR!"111 Waler-Qualily Environmenls \. LECTURE18 The "Modeling" Environment 327
}

60 In summary Canale and Auer's Lake Huron Cladop¡'ora study is unique in cou-
.::..
~'" pling direct mechanism measurement with modeling. Professor Auer has edited a spe-
~.=, dal issue of the Joumal of Great Ltlkes Researc¡' (Auer 1982) devoted to !he study.
e"
..~
:ii.c
Aside from representing a seminal contribution in the area of algal growth modeling,
"Q.
-" it provides an interesting case study related to the model development and application
oo . . issues described in this lecture.
1I).c
Q.
.{).5 O 0.5 1.5

Q.- 18.2 MODEL SENSITIVITV

e7
u E
~i Before modeling,il is advisabl~!Q\Illd.erstao..d!he.&,eneralbehavior of a water-quality
'6~
!t
(1)-
moaeI. One way lo do tlií.s Is called a sellsitivity alialysis. There are a variet)' of
ways lo imptemenl such an analysis. Two of the mosl common are simple parameter
.{).5 O 0.5 1.5 2 perturbalion and firsl-order sensitivily analysis. In addition Monle Carlo approaches
provide a more general approach. We describe each in the following sections.
Dlstance Irom nutrient source (km)

(a) 1979 (before treatment) 18.2.1 Parameter Perturbation

For simplicity we illustrale both parameter perturbation and first-order sensitivily
.::.. 60 analysis for a simple mass-balance equalion for a well-mixedlake,
.2:",
t>.=, 40 ( dc
.. 1 V-d = QCin-Qc-kVc (18.1)
.. 2
-o t
....c
"Q.
°8
1I).c
Q.
20
8.U'T . . . which, al sleady-stale, can be solved for
Q
o P
I

1
I

I
'{).5 o 0.5 1.5
C = Q + kVCin l ~ .1 (}8.2)

300 Recognize Ihat according lo Ihis formula c is a funclion of each of the model
Q.¡;-
o, parameters and forcing funclions; Ihal is, e = f(Q, k, V.Cin)'Thus one way of visu-
1; E 200

..
alizing Ihe dependence of Ihe solulion on one of the palamelers (for example k) is
'6
i
r+"- . . by developing a plol of c versusk (Fig. 18.4). í _-" .

!t
(1)-
As Ihe name implies, parameter perturbatioll consiSís of ~rying each of the
1 I model paramelers (for example raised and lowered a fixed percent) while holding
.{).5 O 0.5 1.5
, all otherterms conslan!. The corresponding varialions of Ihe stale variables reflect
Dlstance Irom nutrlent source (km) ~ Ihe sensilivily of the solulion lo Ihe varied parameler. Thus if we desiredlo quantify
the sensilivily of Eq. 18.2 lo a parameler such as the reaction rale k, we would merely
(b) 1980 (after treatment) substitute k - I:1k and k + I:1k inlo Eq. 18.2 to ca1culate c(k - I:1k)and c(k + I:1k).
The error of Ihe prediclion could Ihen be ca1culaled as
FIGURE 818.1
Plols 01 SRP and biomass versus dislance Irom I:1c= c(k + I:1k) - c(k - 1:1k) (18.3)
Ihe outfall al Harbor Beach, lake Huron (redrawn 2
Irom Canal e and Auer 1982). The plols represenl
Ihe condilion belore and alter phosphorus reduction A graphical depiction of Ihe approach is shown in Fig. 18.4a.
technology was inslalled al Ihe Harbor Beach
waslewaler trealmenl plan!.
18.2.2 Flrst-Order Sensltlvity Analysis
An alternalive lechnique Ihal yields similar information is based on a jirst-order
sellsitivity allalysis. This approach uses Ihe derivative of Ihe funclion with respect
328 PAKT11I Water-Quality Environments LEnURE IX The "Modeling" Environment 329

to the parameter as an estimate of the sensitivity. One way to derive it is to employ Equation 18.5 can be subtracted from 18.4 and the result solved for
first-order Taylor-series expansions of the model (Eq. 18.2) around the value of the
parameter. For example forward and backward expansions can be written as
Al.' = l.'(k+ Ak) -2 c(k - Ak) = al.'(k)
ak
Ak ( I¡S,h)
ac(k)
c(k + Ak)= c(k) + a¡¡-Ak (18.4) Note that for this formula the sign of the derivative indicates whether a po~itive
variation of the parameter results in a positive or negative variation of the prediction.
ac(k) As depicted in Fig. 18.4b, because the slope of the straight-line approximatioll is
c(k - Ak) = c(k) - a¡¡-Ak (18.5) negative,an increasein k resu\tsin a decreasein l.'.
A nice refinement on both the perturbation and first-order sensitivity analyses
e ~ is to express the resu\ts as condition numbers (Chapra and Canale 1988). For the
J first-order sensitivity analysis such a number can be derived by dividing both sides
ofEq. 18.6 by c. Then the right-hand-side can be multiplied by k/k and the Icslllt
e-Ae solved for
,/
2Ac .......--..
e+Ae 0 ---....-.- Al.'
l.'
= CNk Ak
k
( ¡¡S.?)

o where CNk is the condition number for Ihe parameter k,

o k-Ale k+AIe k kac
CNk = -- (1 ~.H)
~2A1e/ l.'ak
(a) Parameter perturbatlons
e Thus Eq. 18.7 indicates that the condition number provides a transfer functioll lo
propagate the relative error of the para meter into the relative error of the prediction.
Note that although the derivation is for the first-order sensitivity analysis, it call he
used for the perturbation analysis. However, for this case a discrete form is used for
the derivative,
/e-Ac ............
2Ac
-- e+Ac 0 ..........
k Ac
(I¡S~)
CNk=cAk
o
o k-Ale k + Ale k EXAMPLE 18.1. SENSITIVITY ANALYSIS. As depicted below, two chclIlll'..1
~2A1e/ species react within a lake. Mass balances for the two reactants can be writtclI as

(b) Flrst-order analysls

e Loadlng Outflow

---..................--..
-. ~...-. ~----... ~ .Reactlon
e. .
le2

o
o k Settllng FIGURE E18.1

FIGURE 18.4 V-dCI = QCin - - kl2 VCI

Graphical depictions ,
l I
QCI + k21 VC2

01 three methods
dC2
lor assessing model
(e) Monte Carlo analysls sensitivity. V-;¡¡ = -QC2 + kl2VCI - k21VC2- 1',A,C2
.BO I'A~r 111 Water-Quality Environments LECTUREIR The "Modeling" Environmenl 331

where V = volume 01' lake = 50,000 m3 These reSUllS lead lo Ihe following general conc1usions:
Q = inllow = oUlllow= 50,000m3 d
A, = surface area 01' lake sedimenls =
I
16,667 m2
· The lransfer rale kl2 has Ihe grealesl
sile impacl on Ihe Iwo slale variables.
effecl
Whereas
on bolh variables. However,
CI is inversely dependent
it has an oppo-
on Ihe rale,
kl2 = first-order conversion rate 01' CI lo C2 = 1104 d
k21 = firsl-order conversion rate 01' C2 lo CI = 0.5 lo 0.7 d' 1
'.. = settling velocity 01' C2 = O lo 1 m d 1
· C2 is direclly dependenl on Ihe rale.
The impacl 01' k21 is opposite lo Ihal 01' k12. Thal is, a rise in k21 causes CI lo rise and

The ranges 1'01'Ihe parameters connole literature values. The term Cin represenls Ihe con-
centration ofthe infJowing slream (mg L 1). which during Ihe period ofsludy has a mean
\'alue 01'approximately 10 mg L l. Use Ihis informalionlo estimate Ihe sensilivily 01'Ihe
l1Iodello Ihe three paramelers k12. kzl. and 1'...Employ firsl-order sensilivity
express your results as a lable 01' condition numbcrs.
Solulion: Firsl. we can develop the sleady-slate solulion 1'01'each 01' the unknowns
· C2 lo fall. However, Ihe magnilude 01'Ihe impact is less Ihan 1'01'k12.
The settling rale causes bOlh concenlralions lo fall, bUI has aboul 4 limes more impacl
analysis and
· on C2 Ihan on CI.
The effccl 01' Ihe paramelers on Cr is much less Ihan on Ihe individual consliluenls
because flushing is Ihe primary inpul-oulPUI mechanism (see Probo 18.1). 01' alllhe
paramelers, Ihe 10lal concentralion is mosl sensilive lo Ihe only one Ihal direclly affects
inpul-output: 1'...
as

k~ + kqk21 + kqk.
CI = 2 Cin
kq + kqk21 + kqk.. + kqkl2 + kI2k.. A problem with the foregoing example is that it did not consider the uncertainty
of the parameter estimates. By using constanl perlurbations we ascribe the same
kqkl2
C2 = ., Cin uncertainly lo cacho However, based on [he ranges provided in lhe problem, they
k-¡, + kqk21 + k"k.. + kqkl2 + kI2k..
exhibil quile differenl uncertainlies. For example k21is known wilh a much greater
where kq = QIV precision than eilher kl2 or lhe settling velocity, One way of rectifying this problem
k.. = I'..IH involves propagaling lhe range through eilher the perturbation or firsl-order proce-
11 = mean deplh = VIA. dures. This can done simply by using Eq. 18.7 to propagate the relative error in each
Because we have no basis for anylhing better (allhis poinI), we willlake Ihe average parameler inlo the resulting error in the prediclion.
value 01'cach parameler as Ihe mosl likely value. For example 1'01'kl2 we would use
(1 + 4)/2 = 2.5 d-I. This value, along wilh Ihe average values 1'01'Ihe olher parame- EXAMPLE 18.2. UNCERTAINTY ANALYSIS. For Ihe case from Example 18.1,
Icrs (k21 = 0.6 d 1 and k. = 0.5/3 = 0.1667 d-I) can be used lo compule our mean delermine the expecled uncertainlies 01'Ihe concenlralions due lo Ihe uncertainlies 01'Ihe
prediclions. paramelcrs.

CI = 12+ 1(0.6) + 1(0.1667) 10 = 3.772m L-I Solution: The relalive error in Ihe parameler kl2 can be calculaled as
(2 + 1(0.6)+ 1(0.1667)+ 1(2.5)+ 2.5(0.1667) g
~k12 =~ = 0.6
. - 1(2.5) 10 - 5138 Vi kl2 2(2.5)
q - 12+ 1(0.6)+ 1(0.1667)+ 1(2.5)+2.5(0.1667) mg
which can be propagaled Ihrough Ihe condilion number lo eSlimale Ihe expecled uncer-
\Ve can also delermine Ihe tOlal concentralion as CT = CI + C2 = 9. 110mg L-l. lainly in c"
Now wc can pcrform a firsl-order scnsilivity analysis. To do Ihis Ihe parlial deriva-
I¡ves must be analyzed. For example 1'01'Ihe effecl 01' kl2 on C¡, ~CI
- = -0.623(0.6)= -0.374
CI
dCI - (k~ + kqk21 + kqk. )(kq + k,) The olher uncertainlies can be evalualed and labulaled:
iik = 2 Cin
12 (q + k"k21 + kqk, + kqkl2 + kI2k,) n '7' ku kZI k,
which along wilh Ihe olher values can be subsliluled inlo Eq. 18.8 lo give CI -0.374 0.035 -0.030
51 0.226 -0.021 -0.125
-k (k + k ) C2
CN. = 12 q .\ = -O 623/ CT -0.022 0.002 -0.086
1.1'2 l.~ + kqk21 + kqk., + k,¡kl2 + kI2k., .
The olher ('ondilion numbers can be evaluated and labulaled: These results indicale Ihal:

ku k2J k, · The uncerlainly in kl2 is a large conlributing factor

values 01' CI and C2. However, because il merely shifts mass belween the two forms,
to Ihe uncertainty in the relative

('1 -0.623 0.212 -0.030

C2
Cr
0.)77
-O.OJ7
- O.12R
0.(1)
-0.125
-0.086 · Ihe uncertainly does not slrongly impact Ihe 10lal quanlity, CT.
The uncerlainty in k21 has 'ittle impacl on the uncertainty 01' the componenls 01' Ihe
tOlal concenlralion.
 332 PART11I Water-Quality Environments IHTlIKE 1M The "Mudeling" Environmel1l 333

· The settling parameter

significant contribution
has a larger impacl un the uncertainty 01' C2. It alsu makes a
to the uncertainty 01'the tOlal prediction.
Random numbers are then used lo generale a series 01'parameler estimales tha! fol-
low the distribulion. Each 01'Ihese estimates is used lo compule model simulalions.
On the basis 01' this simple analysis, it can be concluded that efrort should be ex- The oulcome is a dislribulion 01'concentralions (Fig. 18.4(").
pended in refining the estimates 01' kl2 and t,. Figure 18.5a shows a few ideal probability dislribulions thal are commonl) IIsed
to describe parameler variabilily in waler-qualily modeling. The probabilily densilY
functions or p.d.f. (Fig. 18.5u) represenl plols 01'frequency 01'occurrence f(x) versus
18.2.3 Monte Carlo Analysis the parameler value x. Thus Ihe areas under Ihe probabilily densilY funcliol1Sare
equal lo 1, connoting thal Ihey encompass Ihe total probabilily 01'occurrence.
t As the name implies, Monle Carlo analysis is related lo games 01'chanceo Ralher than The uniform distribution (Fig. 18.5.1) assumes that there is an equallikelihuoJ
~ prescribing a range for Ihe paramelers, the analysis charaeterizes their distributions. \ 01'occurrence belween Iwo bounds. This distribution is often used when all that is
,known is a range for the parameler. In contrast the normal distribution assull1es
F(x) thal Ihe parameter follows a symmetrical bell-shaped curve. Thus there are more
j{x) 1 occurrences al a mosl likely value in the cenler and diminishing occurrences at the
extremes (Fig. 18.5.2). Because 01'its flexibility, a triangular distribution has mi litY
0.5 in describing a range 01'shapes. The symmetrical version shown in Fig. 18.5.3 ean
be used to approximate the normal distribution. However, because the pea\.. "oes
X x not have tu be centered betwecn the extremes, it can also be emplo}cd lo rcprcscnt
o o dislributions that are off-centered or "skewcd."
o 2 3 o 2
The cumulalive dislribulion funclion or c.d.f. reprcsents the integral 01'Ihe p.d.f.,
(1) Uniform

F(x)
Unlform 1
j{x) F(x)
1 random number
generator
(0-1)
0.5 0.5

X x
o o o
o o 2 3
o 2 3
(2) Normal
x

.f(x)
j{x) F(x)
1

0.5

x
o o
o 2 3 o 3 o
(3) Triangular o 2 3
x
(a) (b)
FIGURE 18.6
FIGURE 18.5 Graphical depiction 01 how the c.d.l. is used in conjunction with random
(a) Probability density and (b) cumulative distribution lunctions lor three distributions commonly numbers to generate values 01 model coeftiients in Monte Carlo
used to characterize uncertainty 01 water-quality modeling parameters: (1) unilorm. (2) normal, simulation.
and (3) triangular distributions.
 334 PART 111 Water-Quality Environments LECTURE18 The "Modeling" Environment 335

or collecting terms,
F(x) = roe f(x)dx (\8.\0) o = -0.2222x2 + 1.7776x - 3.3553
This integral specifies the probability that the parameter wil\ be less than x. The two which yields the root 3.0515. (The quadratic formula also gives another root, 4.9485.
distributions are also related inversely by differentiation, which is disregarded because it's outside the parameter range, 1 to 4.) This value can
then be substituted into the model (recall Example 18.1) along with the mean values for

f(x) = d~~X)
(18.11) =
the other parameters to give CI = 3.317 and C2 5.729. The process can be repeated
for other random numbers. The resuhing distributions are displayed below for the case
In olher words Ihe frequency of occurrence is equal to the rate of change of the c.d.f. where 400 random outcomes were generat'.:d. Notice that ahhough the mean values are
similar to those from Example 18.1. the distributions are skewed.
wilh respect to the rale of change of the parameter.
As depicted in Fig. 18.6 the c.d.f. provides the key tool for implementing Monte
Carlo analysis. A random uniformly distributed number in the range O to 1 can be
generated with the computer. This value is then propagated through the c.d.f. to . >-
60 t /1... 60

generate a value of the parameter x. By repeating the process, a series of parameters U

e 40 + I I \ 40
GI
will be generalcd that together are distributed according the underlying p.d.f. These
CT
valucs are thell used in a water-quality model to simulate distributions of concen- 20 + I I \. 20
I!!
tWlion. IL
\
o I I 1 '1 '1 I '1 I o
EXAMPLE 18.3. MONTE CARLO ANALYSIS. For the case from Example 18.1. o 2 4 6 o 2 4 6
determine the expected uncertainties of the concenlrations due lo the uncertainty of the
parameter k12. Assume that the parameter follows the triangular distribution CI C2

¡(x) = 0.4444(x - 1) I s x < 2.5 FIGURE E18.3-2

¡(x) = -0.4444(x - 4) 2.5 s x < 4

These can be integrated to develop equations for the c.d.f.,

F(x) = 0.2222(x - 1)2 1 s x s 2.5 18.3 ASSESSING MODEL PERFORMANCE

F(x) = -0.4444(0.5x2 - 4x + 5.75) 2.5 s x s 4 As described previously in Seco 18.1.4 the aim of calibration is to adjust the param-
eters so that model solulions fit observations in an optimal fashion. Now we must
Both functions are displayed below:
( define what we mean by a "best fit." There are two general approaches for assess-
( ing the quality of a calibration: subjective and objective. Subjective assessment is
0.8 based on a visual comparison of the simulation with the data. This usually consists
0.8 of individual time series plots for each of the state variables. The modeler adjusts
0.6
parameters until he or 5he deems that the agreement between the model output and

.f(x) 0.4 F(x) data"Iooks" adequate. Q.N~cA~\d.J...

In contrast objective approaches hinge on developing some quantitative mea-
sure of the quality of the fit (usually some measure of error). Once such a measure
0.2
\ is adopted, the parameters are adjusted until a best (usuaIly a minimum) value is
o o ! attained.
o o 3.05 4 There are a number of measures that can be developed to assess a tit. In !he
FIGURE E18.3-1 present discussion we wil\ focus on minimizing the sum of !he squares of the resid-
'> uals, "
Ranc\om numbers between O and I can be generated. For example a value of 0.8 é:\ Sr = L(cp.; - cm.;)2 (18.12)
might b(' generated and substituted into the e.d.r. to give ;=1

0.8 = -0.4444(0.5x2 - 4x + 5.75) where Cp,; = ¡th model prediction of concentration and Cm.; = ¡th measured concen-
tration.
 LEerURE 18 The "Mudeling" Environmcnt .117
336 PART111 Water-Quality Environments

15
The sum of the sguares provides a total vulue that should be u minimum whcn
the predietions and measurements are in agreement. Thus it provides u single score
that can be monitored as you adjust parameters.
It should be noted that the parameter adjustment proeess for the objeetive ap-
-
...
~
10 (a) """"''''''',
"
Total ....
.......
proaeh can also be implemented in two fashions. First, as with the subjeetive method, C)
- x" C _ x x x
it can be approaehed in a triul-and-error fashion. However, along with a visual as- I ): ',IS .
sessment, the best-fit measure (Eg. 18.12) provides an additional means of judging .5.5 .
the effeets of parameter adjustment. Seeond, numerieal optimization methods can be
~
used to adjust parameters in an automatie fashion. 80th are illustrated in the follow- C2 o
o I
ing example.
o 2 3 4 5
EXAMPLE 18.4. MODEL CALIBRATI0N. For the same system studied in Exam-
pIe 18.1. the following data were collected during a 5-d sampling survey: 15

t
(d)
CI
(mgL-1)
Cl
(mgL 1)
Cr
(mgL 1) -
...
~
10
(b)
X
.x .'_ x
0.0
0.5
1.0
7.0
7.3
6.6
4.0
3.5
4.7
11.0
10.8
11.3
C) ... ., . ~.. · '..> x ""0(___., ,
1.5 8.9 4.0 12.9 .5.5 o o --...,¡ .
2.0 6.6 5.0 11.7 ~ o .
2.5 6.2 3.1 9.3 o
7.6 o
3.0 3.3 4.3
4.7 2.2 6.8
o I I I I
3.5
4.0 3.75 3.9 7.7 o 2 3 4 5
4.5 6.1 1.3 7.4
5.0 6.0 4.0 10.0
Mean 6.03 3.64 9.67 15

The concentration of the innowing

sented by the sine function,
stream ein during the period of study can be repre- -
í
..J
10
(e)

.
x ~~

C) ., .
,."'" '''. ,.. '"
'. >(" '. '. " ~
, ./~oo" o'
----.... .
Cin = 10 + 5sin(~I) .5.5 o
~ o o o
In addition in situ experiments were performed at two times during the study to estimate
kl2 directly. These yielded eSlimates of 1.05 d-I and 1.55 d-I for that parameter. Use this o
information (and any other tcdmiques at your disposal) \O estimate the three parameters
o
k12. k210 and v,. o 2 3 4 5

Solution: As in Example 18.1, in the absence of additional information, the average

values for the parameters are the most likely: kl2 = 2.5 d-I. k21 = 0.6 d-I, and v, =
0.511Id-I. Using these parameters along with the other information results in the simu-
lation shown in Fig. 18.70. Notice that the model does an adequate job of matching the
total concentration. However, it underpredicts CI and overpredicts C2.The differences
between the data and model are large as renected by the total sum of the squares (calcu-
Jated based on both CI and C2),which is 97.69.
The fact that the simulation misses CI low and C2high suggests that kl2 might
be a good candidate for adjustment (recall the analysis at the end of Example 18.\).
Thus we can repcat the simulation using the average value of the direct measurement of
k12, (1.05 + 1.55)/2 = 1.3 d-I. This value along with the average values for the other
t(d)
FIGURE 18.7
Calibration. (a) The original lit using average values lor the parameters; (b) improved
lit alter using the measured value lar k'2; (e) linal calibration.
parameters can be employed to compute another prediction. As in Fig. 18.7b this adjust-
ment brings a great improvement (SSR reduces to 18.35).
Next we can try to improve our estimate of v,. A rough estimate of this parameter
can be obtained by recognizing that the two original differential equations can be added
to yield
:BH I'A~T111 W¡¡tcr-Quality Environrnents LECTURE18 The "Modeling" Environment 339

18.4 SEGMENTATION AND MODEL RESOLUTION

V-dc]' = Qc' - Q C1' - \. A C2
III 111 ., ,~
The final concepts that must be explored before developing mechanistic water-
This clJualiol1 descrihes the input-olltpul relalionships that govern how the lotal quality models are segmentation and model resolution. As stated previously, seg-
mass varies ternporally. Now allhough it descrihes transienl changes. inspection 01' the mentatiOll is the process of dividing space and matter into increments. In space
data indicates thal although the two dependenl variables c]' and C2 oscillate, they ap- a water body can be divided into volumes for which mass-balance equations are
proximately rclurn to Ihcir initial condilions over the study periodo Thus this equalion written. Within each volume, matter, in lurn, may be divided into different chemical
c¡m be solved at sleady-slate for
and biological forms for which separate equations would again be written. Thus, if
QZ'¡n - Qc]' a waler body were divided into 11segmenls in space and matter were divided into In
\'l = A.\C2 substances, 111X 11mass-balance equations would be used to define the system.
Aside from space and matter, there is a temporal aspect to segmentation related
where the overbars indicale Ihat the coneenlrations represent average values for the si m- to the fact that the mass balance defines changes in Ihe water body overafinite period
ulation periodo These ean he developed frorn Ihe dala and used lo calculate 01 time. Allhough Ihis concept will become clearer in the subsequent development
of actual models, the key point is Ihat, jusl as a model describes additional spatial
50,OOO(10.67 - 9.67) = 0.82 m d-I
\'.. = 16.667(3.64) and material detail by using more segments, the model also has a temporal focus
that can be increased by using a shorter "finite period" or time step for the mass-
Notice how Ihe average inf10w and total eoncentrations are very close to each other. balancecomputation.Forexamplethe water body may be undergoinga significant
This implies that f1ushing is Ihe dominant purging mechanisll1 and Ihal seuJing traps a yet gradual change due to longer term phenomena such as increases in nutrient loads
relatively small fraction 01'the total inf1uent in the lake. It also suggests that the estimate (Fig. 18.8a). At the same time, as in Fig. 18.8b, phytoplanktont undergo seasonal
01' the settling velocity is probably not very certain. However, since the estimate lies cycles due lo, among other things, ftuctuating light and temperalure leveIs over the
wilhin the range 01'expected values, we can use it to de;elop a new simulation. As mighl
course of the year. By adjusting the time step of the computation, we can design the
be expected the new value has very liule effeet on the results.
model to resolve part or all of Ihis variability. In a sense this can be thought of as a
Thus we are left with one rernaining parameter. k21, Ihat must be specified. Look-
ing at Fig. 18.7b, we see that C2 is still being overpredicted. Thus an increase in k21
segmentation in time.
seems in order. This adjustment can now he made by either trial-and-error or by using The degree to which space, time, and matter are segmented is called model
an optimizer. The result is a value 01' 0.7 d-I. Substituting this parameter estimate into resolution. This concept is analogous to photography, where Ihe camera's lens is
Ihe mode) yields the final calibralion in Fig. 18.7c. The final total sum 01' the squares is
14.95. 'Microscopic floating plants.

Now as a linal exercise we can repeat the calibralion using an optimizer to minimize
lhe sum 01' Ihe squares numerically (recall Example 2.3). The result is kl2 = 1.24 d-I, 20
.
k21 = 0.74 d-I , and v.. = 0.93 111d-I with a total sum 01' Ihe squares 01' 14.23. COlise-
quently for this case the automaled ca)ibration yields results that are close to the manual
version. However. also notice that the estimate for k21is outside the expected range 01'0.5
lo n.7 noted in Example 18.1. Such behavior can occur in autornated calibration schemes
IIsing unconstrained parameter estimates. - 10

In the previous example we carne up with similar results using two different
approaches. The first was a slep-by-step, Irial-and-error approach that used some
-el
:1-
q¡
direct measuremenls of certain parameters. The second was a fully automated least- >.
.c
sqnares approach. Although Ihey yielded ahnost identical parameter estimates for c. 18 FIGURE 18.8
the present case, in olher instances Ihe Iwo approaches could yield very differént 2O Plot 01chlorophyll a concentra-
parameter estimates. Such would be the case when some of the key state variables :E tion ( g L-') versus time. (a)
were highly uncerlain. O Depiction 01 the long-term
trend (straight Hne) along with
Al face value it mighl seem thal Ihe objective method is superior. However, 10
the underlying seasonal cycle
because the firsl approach uses additional information, it is preferable. Further the (wavy Hne); (b) depiction 01
suhjective, visual approach also has value since il employs Ihe modeler's intuilion in the seasonal cycle (wavy
the process. Modern inleractive compuler graphics coupled with optimization soft- Hne) along with the underlying
2J FMAMJ JASOND long-term trend (straight Hne).
warecouldprovidea meanslo integrateboth approachesin futuremodelinterfaces.
 ,a

340 PART11I Waler-Qualily Environmenls LECfURE 18 The "Modeling" Environmem 341

adjusted to bring different parts of the field of view into focus. At times the fore-
ground is important; at other times distant details might be of interest.i
In the present context there are two basic ways in which the "focus" or resolu-
tioTlof a model has significance for water-quality analysis. First, the fine-scale phe-
nOIl...nonmay have a direct, causative influence on predictions made on the coarser
scale. This is primarily a function ofthe substance's properties and the physical char-
acteris~ics of the system. For example certain pollutants (e.g., enteric bacteria) die
rapidly upon entering a water body. Therefore they typically are at high levels near
a sewage discharge and then decrease rapidly, so that in the open waters of lakes and
estuaries they are often at "background" levels. Thus a model of near-shore bacte-
rial pollution would require relatively fine spatial and temporal segmentation around
sewage outfalls. In contrast conservative or slowly reacting substances often can be
modeled with coarser schemes that treat the whole water body as a single well-mixed
segment.
The physical characteristics of the system also can dictate the required level of
segmentation. For example dendritic lakes have numerous embayments that make
treatment of tbe system as a single segment unrealistic. In addition problems such
as eutrophicationt that are strongly influenced by thermal stratification may !leed
multiple vertical segments for adequate characterization.
The second way in which model resolution has significance is the influence of
the problem context on the choice of scales. For example although phytoplankwn
go through growth cyc1es on seasonal time scales and exhibit patchiness on small
space scales, a water-quality planner might not have the funds to develop models to
simulate such short-term variability. This would often occur when large numbers of
smalllakes were being evaluated. Altematively the issue of concem to the plar,ner
may be such that spatial and temporal aggregation results in a negligible loss of
relevant planning information. In these cases the manager might opt for a co¡m,er
scale model. In other situations, as with bacterial contamination of beaches, a finer
scale approach would be necessary to handle the problem efkctively.
The temporal, spatial, and kinetic scales of a problem often are interrelated. As
depicted in Fig. 18.9, fast kinetic processes such as jet mixing of thermal efflucnts
or bacterialdie-off tend to manifestthemselveson local (that is, small)time and
space scales. In contrast problems determincd by slow reactions such as the decay
of persistent contaminants are more likely to be important on a whole-system, long-
term basis.

t Profuse growlh of plants due 10overfertilization.

Horizontal spaee seale (km)
10-> 10-> 10"' 10 103 103 104 PROBLEMS
10'
i
Century 18.1. Develop condilion numbers for Example 18.1 for flow and inflow concentralion.
ITn"

104 Pen ¡slenl lo.les 18.2. The bay/main lake system depicted in Fig. P18.2 has the following characterislics:
Nul ents Decade VI = 2 X \07 m3 V2 = 5 X \06 m3

-
-'C
Q)
fti
103

103 Sollds
Year
Season
As.1 = 2 x \OÓm2 As,2 = 1 x 10ó m2

The outflow from the main lake is 1 x 107 m3 yr-I and the flow from the bay to Ihe
main lake is 0.2 x \07 m3 yr-I. The main lake has an inflow concentration of 10 f.Lg
L -1 of a substance Ihat reacts with first-order kinetics (half-life = 1 "l). An induslrial
(,) Month
u, plant is lo be localed on the bay with a flow ofO.1 x 107mJ yr-I alld a l:oncentralion
Q) of 100 f.Lg L -l. The inlerface between lhe main lake and the bay has a cross-seclinnal
10 Becterte Wgek
E Thermellels
area of 250 m2 and a mixing length of 200 m.
(a) Use Okubo's diffusion diagram (Fig. H.II) to estimate the range 01'tllrbulclll dir-
¡: fusion coefficients expected for such a system.Usea valueof 500 m for lhe length
1 Day

10"'
Hour

10"2
10 102 103 104 10' 10" 107 h~
Horizontal spaee seale (m)
1 )
FIGURE 18.9
Approximate time and space scales o, water-quality problerns.
~I\ FIGURE
P18.2
342 PART 111 Walcr-Quality Environmcnts LECTUREIR The "Modeling" Environment 343

scale. In addition to !he range, estimate a most likely value as the geometric mean The concentration of the innowing stream. Cin.during the period of study can be rep-
of the minimum and maximum values obtained from !he plot. The geometric mean resented by the sine function,
can be cornputed as
GM = JEminEmax
Cjn = 15 + IOsin(~t)
where Emin and Emax are the minimum and maximum diffusion coefficients from
Okubo's plot. Note that the geometric mean is the appropriate statistic for central Unfortunately none of the parameters were estimated directly. Use the observations to
tendency when the pararneter is distributed log normally (that is, the logarithms of estimate the three parameters k12. k21, and v...
the values follow a normal distribution).
18.5. A general representation ofthe triangle function is depicted in Fig. PI8.5.
(b) Use the mean diffusion coefficient along with the other data to estimate the ex-
pected concentration in the main lake and in the bayo
(e) Use a first-order sensitivity analysis to estimate the uncertainties in these predic-
tions. f(x}
(d) Use a triangular distribution to approximate the log-normal distribution of the dif- flXpoak)
fusion coefficient and perform a Monte Carlo analysis to compute the expected
distribution of the bay's concentration.

18.3. A total phosphorus mass balance for a lake can be solved at steady-state for
W
p = Q + vAs x
where p = total phosphorus concentration (mg m-3)
Xmln xpoak Xm81
W = total phosphorus loading (mg yr-I)
Q = outftow (m3 yr-I) FIGURE P18.5
v = an apparent settling velocity (m yr-I)
As = lake-surface area
(a) Determine f(Xpeak) sothat the areaunderthe function is equalto I.
(a) Apply this model to calculate the total phosphorus concentration of Lake Ontario (b) Develop the equations describing f(x).
in the early 1970s (Q = 212 x 109 m3 yr-I, V = 1634 X 109 m3, As = 19.000 X (c) Develop the equations describing F(x).
106 m2, W = 10,000 mta). Assurne that the settling velocity is 12.4 m yr-I.
(b) Supposeyou haveadditionaldataavailableindicating that the seuling velocity can
range from about 4 to 40 m yr-I. Use this informatiol1 and a first-order sensitivi,y
analysis to translate this range into an uncertainty estimate for the total phosphorus
prediction.

18.4. The same kinetic reactions as in Example 18.1 occur in a pond having the following
characteristics: V = 20,000 m3, Q = 50,000 m3 d-I, and A.. = 50,000 m2. The fol-
lowing data were collccted during a IO-d sampling survey:

t CI Cl
(d) (mg L -1) (mg L -1)
O 1.7 7.3
I 3.5 6.9 /
2 2.5 10.5
3 3.9 10.2
4 2.6 13.6
5 4.6 12.8
6 2.8 13.6
7 3.3 11.0
8 1.3 11.9
9 2.9 7.2
10 0.7 8.8
PART IV
-
Dissolved Oxygen
/
and pathogens

Part IV is designed to introduce you to our first water-quality-modeling prohlem

context: dissolved oxygen. In addition it provides information on another c1assic
water-quality problem: bacterial pollution.
The first two lectures describe the fundamental building blocks that form ¡he
foundation of oxygen modeling. Lecture /9 deals with biochemical oxygen demand
or BOD. In addition the subject of oxygen saturation is covered. Lecture 20 is de-
voted to the general topic of gas transfer, with special emphasis on oxygen reaer-
ation.
After this review of fundamentals, the next two lectures present the Streeter-
Phelps oxygen model. Lecture 2/ describes the point-source version, whereas Lec-
ture 22 deals with diffuse sources.
The succeeding lectures outline refinements that have been made to the origmal
Streeter-Phelps framework. These cover the inc1usion of nitrification (Lecture 23),
plant photosynthesis and respiration (Lecture 24), and sediment oxygen demand or
SOD (Lecture 25).
Next, computer aspects of oxygen modeling are discussed. The first part of Lec-
ture 26 deals with general computer approaches, with particular empha~is on e~llI-
ary models. Then Ihe remainder of the lecture is devoted to a description 01'o>.ygell
simulation with the QUAL2E software package.
Finally the subject of pathogen modeling is explored in Lectl/r(' 27. Thi>;l."
ture focuses on c1assicfirst-ord"1 rrameworks designed lo simulate the lranspoll ;110'
I
fate of enteric bacteria in natural waters and sediments. In addition new modeb !.Ir
protozoa are explored.

.\-t5
 LECTURE 19

BOD and Oxygen Saturation

In Lec. 1, I stated that engineers originally became involved in water-qua!ity mod-

eling to assess the impact of sewage on receiving waters. We wi\l now study the
m~at were deve~ to ~olve_tbis..problem.To place these models in coi1text,
.1 l' U first describe the cycJe of organic production and decomj2Qsitinntba.U>ccursin
theolos~e.r.e-:-
--- --- - --
:ti

19.1 THEORGANICPRODUCTIONlDECOMPOSITION
CYCLE
\ As depictedin Fig. 19.1the biospherecan be viewedas a cyc1eof !ifeand death.
\ Powered by the sun, a,!!otI:Qphict Q.rganisi11s(prilTImIY-p1ants)_coDYertslmpleinot-
~./
ganic nutrients int~ more complex or&.a.!1i<;
mole¡::..!!les.
In photosynthesis, sotaren-
ergy is ~oredJls chemical energy in the organic molecules. in additiort oxygen-rn
,liberated and cacbon dioxide is consumed.
The organic matter then serves as an energy source for heterotrophic organisms
(ba~teria and animals) in the rev'efse processes of respiration and decomposition.

tThe tenn autotrophic refers to organisms like plants that do not depend on other organisms f()rnubiUpn.
In contrast he!erotrophic organisms consist of animals and most bacteria that subsist on organic matter.

347
" ~

34~ PARTIV Dissolved Oxygen and Pathogens

LECTURE 19 BOO and Oxygen Saturatíon 349

"I~~ enem
- RIVER

rJ Saturatlon
- --T-~ -'--j_.~":"~ ------ -----------.--.----- ------.
/ ... ('
) , r..,
/1 l \),\,.. .
01
.! - -
O
Q t- R';;;o;;ry ZOñe
(plants) ,
r "
,J ") ~ r
Critical concentration
~ \., '~ll
/
(decompositlon =reaeration) .i4-t '~ \" ,\, }',,\ Dlstance

Ir ,1\1 \ .\."}
Lowenergy
.
Decomposition I
Reaeration ·
f'GUR~ 1~.1 domlnates dominates
Tlw o~tur~1
cycle 01 organic production and decomposition. FIGURE 19.2
The dissolved oxygen "sag" that occurs below sewage discharges into streams.
TPese retuffi the organic matter to the simpler inorganic state. During breakdown,
oxygen ͧ con~umed 3!1dcarbon dioxide is liberated.
T~e cyc1ecan be represented in chemical terms by the following simple expres-
siop:
photosynthesis
6C02 + 6H20 , . C6HI206 + 602
respiration (19.1)
CaJbon Waler Sugar Oxygen
dioxide

According to this reversible reaction, carbon dioxide and water ar~useA t~ s'yn!h~ize
organic watter (the sugar glucose) and to create oxygen in the forward photosyntheslS
reaction. Conversely tlte organic matter is broken down and oxygen is consumed1ñ
!he reverse respiration and decomposition reactions.
The cbemistry of the productionldecomposition cyc1e is far more complicated
tban Eq. 19.1. For example many different organic compounds are create!i and bro-
ken down in tbe process. In addition other elements beyond carbon, bydrogen, and
oxygep are involved. In later lectures 1'11present a more complete representation.
However,Eq. 19.1providesa startingpointforour effortsto quantifythe process.

1,~.2 THE DI~SOLV!=D OXYGEN SAG

Now that we have the big picture, let's link the life!death cyc1ewitb the environment
il1a sffeam t>e!owa wastewatef discharge (Fig. 19.2). If the stream is originally un-
polluted, dissolved oxygen levels ab9y~ !he disctlarge Willbe near saturation. 1ñe
introduction of the untreated sewage will elevate the levels of both dissolved and
350 PARTIV Dissolved Oxygen and Pathogens \:1
LECTURE
19 BOD and Oxygen Saturation 351

'sms
J) consistingof ciliates,rotifers,and crustaceans.The diversityof higher ~
decreasesdrasticall in the degradationand ac . ositionzones ow the .

I dischar.&e.At the same time the total number Qf organisms increases (Fig. .
..
I Rtvet I
- ímd e). As recovery erisues, diese trends are reversed.
AII the chemical interactions form the characteristic dissolved oxygen "sag"
Degra- Active shown in Fig. 19.2. The key featpre of the sag is the critical or minimum concentra-
Cleanwater datlondecomposltlon Recovery
tion. The location and magnitude of this critical concentration depends on a number
&.s- 16 of factors, including the size Qfthe loading, the.stream's ftow and morphometry, wa-
'ti a. E
GI U J ter temperature, etc. The ~oal of the models formulated below will be to simulate the
(ñ'tl"-' O sag as a function of these factors.

(b)
19.3 EXPERIMENT
¡as
GI
f
a:cr The first step in modeling the DO sag will be to characterize the strength of the
wastewater. To do this we will focus on the respirationldecomposition portion of
O the life/death cycle. In terms ofthe general chemical representation from Eq. 19.1,
this is
Sewage bacteria
I \ no.mL-1 respiration

r
C6H 1206 + 602 ~ 6C02 + 6H20 (19.2)
200 Collform RoUfera and
e no. mL-1 Now let's imagine a closed-blltch experiment to investigate how this reaction
:1 100 crusl8ceanl
!.O 300 no. mL-1 (x25,OOO) affi:cts a simple environment (recall Lec. 2). Suppose that you placed a quantity of
O sugar into a bottle of water with an initial oxygen content of Oo.Youalso add a small
amouitt of bacteria and stopper the bottle. Assuming that decomposition proceeds as
a first-order reaction, a mass balance for the glu~Q.se_canl:>ew1Ítten~L
dg
(19.3)
¿ V dt = -k. Vg
O where g = glucose concentration (mg-glucose L -1) and kl = the decomposition rate
in the bottle(d-I). If the initiallevelof glucoseis go,thisequation_c;.fbe solvedfor
~ 3
(e)
(g = goe-kll J c: '.!: '" e (19.4)
!i
.8')(
1.5
Aquallc Insecls
(x 25) Next a mass balance can be written for oxygen,

E"-' Vdo _ (19.5)

~
z O dt - -rogkl Vg
-4 O 4 8 12
where o = oxygen concentration (mgO L-1) and '''8 = the.stoichiometricratiQ ol
Travel time (d) oxygen~0!1su!!l~<!tQg!!lC9Se
decomposed(ptgOmg-glucose-I).Equation 19.4can
be substitutedintoEq. 19.5to give
FIGURE 19.3 _ t
The changes in the biota below a sewage treatment plant elOent (redrawn from d -~ , (19.6)
Bartsch and Ingram 1977). Vd; = -'"gkl Vgoe-kll
If the initiallevel of oxygenis 00,this equationcan be solvedfor

o = 00 - '''l1go(l - e-kl') (19.7)

 lUTl RE I~ BOD and O\~g~n S .LlIIl1I1 :':'->
352 PARTIV Dissolved Oxygen and Pathogens

12
According to this equation, the bottlc will originally have a db ..Ived oxygen
level of oo. Thereafter the oxygen will decrease exponentially and asymptotically - Oxygen level in bottle
~
approach a level of

o --+ 00 - roggo
(19.8)
DI
§.
e
.2
~e
8
/ O ".1..1.."
consumed durlng
glucose decomposltlon
Before proceeding further, let's work an example.
e8 4
EXAMPLE 19.1. OXYGEN DEPLETION IN A CLOSED BATCH SYSTEM. You 8
ó
place 2 mg of glueose in a 250-mL bottle. After adding a small quantity of bacteria, you ci
fill the remainder of the volume with water and stopper the bottle. The ¡nitial eoneentra- o
tion of oxygen is 10 mg L-l. If glueose deeomposes at arate of o. I d-I, determine the O 10 20 30 40 50
oxygen eoneentration as a funetion of time in this closed batch system. r
Time (d)
Solution: First, we must determine Ihe initial eoneentration of glucose,
FIGURE E19.1
"\
I()OOmL
go = ~
250 mL (--c- )=!! mg L I
'-~
Next we must determine the ralio of mass of oxygen eonsumed per mass of glueose
b
deeomposed. Using the sloichiomelry of Eq. 19.2 we can ealculale 19.4 BIOCHEMICAL OXVGEN DEMAND
The experiment outlined at the end of the last section was intemled to show how
6(32)
rog = 6 X 12 + I x 12 + 6 x 16 = 1.0667 mgO mg-glueose-I the decomposition process could be modeled for a simple batch system. At face
value a similar approach could be used to mo.delhQw s~age wO.ltld.aJ!ectoxy'gen
Therefore lotal deeomposilion of the glueose would consume Ihe following amount of levels. However, such an approach would be probleJl1atic becau1i.eas menTI~ed
oxygen:
previously, sewage is not composed of ~imple.sugar. Thus to rigorOl!~!),1ípplyIhe
approach, we would have to characterize the concentratio.ns of the myriad .Q!game
roggo = 1.0667 x 8 = 8.5333 mg L-.1 compounds in each sewage sample. We would fUr!her need to.determine th~i-
chiometry of the decomposition for each reaction, Fipally CdI~ho(ihi:..~ompounds
Consequenlly the oxygen level in Ihe boule will ultimately approaeh (Eq. 19.8) could decompose at a different rateoQbviously such a r!gQrousapproac.h W01~ be
impractical.
o -+ 10 - 8.5333 = 1.4667 mg L -1
In the early days of modeling, the constraints on such an approach were even
more severe because of the limitations of the technology for characterizing orgallic
Equalion 19.7 can then be used lo eompule the oxygen level as a funetion of time. compounds. As a consequence the first water-quality analysts took an empirical
Some results. along with Ihe glueose levels (expressed as oxygen equivalents) computed
approach and simply disregarded the composition of the sewage. As in our sim-
with Eq. 19.4 are displayed below:
( ple experiment from Example 19.1, the analysts introduced some scwage inlo a
" batch reactor and merely measured how much oxygcn was consume1l.]]lercs.uliln/t
Time Glucose Oxygen quantity was dubbed biochemical oxyge1l demand 01'BOD.
(d) (mgOL-') (mgO L 1) (, In terms of our simple model we can now define a new variablt,\/. (mgO 1. 1)

\
8.5333 10.0000
) .,
O that~the amount 0.!2xidizable.organi~~maining in the bOIl!.~es.:;ed as
4 5.7201 7.1867 -4 &.~~~enequivalents. A mass balance for L for the batch system can be written as
8 3.8343 5.3009 I
/
12
16
20
2.5702
1.7229
1.1549
4.0369 ·

2.6215
3.1895 \
sr v dt = -kl VL l:j
If the initiallevel is Lo,this equationcan be solvedfor
(19.9)

A graph of the results can also be generated: . ~e-kll) (19.10)

354 PARTIV Dissolved Oxygen and Pathogens
LECTURE 19 BOD andOxygen Saturation 355

Note that the oxygen consumed during the decomposition process can be de- Lo = roggo (19.16)
fined as

y=Lo-L Thus by disregarding the exact composition of the organic matter, we avoid the ne-
(19.11)
cessity of characterizing the organic matter and its stoichiometry relative to oxygen.
or substituting Eq. 19.10, Finally it should be noted that although the exact composition of the organic
'1.:::.D\b
(y = Lo(l ~ 'f)1-lfNW' 1" .( (19.12)
matter in wastewater is not characterized, the organic carbon content can be mea-
sured directly. In these cases Eg. 19.16 can be reformulated to estimate the BOO
where y = BOO (mgO L -1). We can now see that the value Lo can be defined as on the basis of organic carbon content, as in
either the initial concentration of oxidizable organic matter (expressed in oxygen Lo = roeCoeg (19.17)
units) or as the ultimate BOO. This notion is reinforced by Fig. 19.4, which shows
bolh Eq. 19.10 and 19.12. where Corg = organiccarbonconcentrationof the wastewater(mgCL-1) androe =
Next a mass balance can be writtcn fo~~n, ratio of mass of oxygen consumed per mass of carbon assimilated (mgO mgC-I).
;---~.- '---Il.~ Again, Eq. 19.2 provides a basis for estimating the ratio
Vdodt = -kl VLoe-k" (19.13) _ 6~~ _ _1
(19.18)
roe - 6(12) - 2.67 mgO mgC
If the initiallevel of oxygen is 00, this equation can be solved for
Fina!ly it should be noted that_along with_lI!e decompositionJlÍ-Cilf...QQnaceou!i
't- G<W\'V\()
fo = 00 -
(19.14) Lo(l - e-kl')} I matter!a.!!...l!d~i!!?naI2x~gendemandil;_exerted~Ee to the oxidation oe ammimia.to
~', ."'\v IV '\j l!iU:alpiR thp.process called.nitrificatia. The oxygen demand nllp.In nitrificatioo
According to this equation the bottle will originally have a dissolved oxygen i~ sometimes referred to as nitrogf!nous BOD or N1JQD to distinguish, it from fue
level of oo. Thereafter the oxygen will decrease exponentially and asymptotically carbonaceous BOO or CBOO described above. In Lec. 23 we will describe nitrifica-
approach a level of tion and NBOO in detail.
o~oo-Lo .¡ (19.15)
Notice that this development is identical to the glucose experiment. In fact the J 19.5 'BOD,MODELFORA STREAM
glucose experiment could be modeled in terms of BOO by substituting
~ Now..we_c.a1l.take..ow:
model and apply it to a..streambelow a s~ageJ!:~J>lahi.
Whenwe do this we mustnow cQl!side.rtl1¡y,jñ:8aititíOnto decomposition, the BOO
10
c~ also be removed_Qv_s.edimentatiQI1. J'here(ore the m_ass_bal!!~efor a constiiñt-
fIow,constant-g~()roetry
channel.can..be.
written -
_ 8
.. ~ .. .. .. - .. .. .. .. .. .. . .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. --aL aL , ,
-',
í < _ = -u- - krL ) (19.19)
o , at ax
~
ge 6 where kr = total removal raté (d-I), which is composed ofboth decomposition and
o settling,
!e 4
Lo ~ = kd + W (19.20)
3 Oxidizabie organlc matter where kd = decomposition rate in the stream (d-I) and ks = settlingremoval rate
e 2
8
o
/ /
/
(d-I). Recognize that the decomposition rate represents the same type of process as
the k. from the bottle experiment (recall Eqs. 19.3 and 19.9). WeJ1a'leusedJl(Hff~r-
ent subscript in Eq. 19.2QJ<>_highlight that decomposition in-a naturat~nvironment
O 10 20 30 40 50 sucl1as a river wiI1g~!l~rally b..e:different from thaliluLhoWe. In addition note that
Time (d) the settlingxat~iSJelated tomoreJuJ1d~ental par~ters by
FIGURE 19.4 Vs
ks =- (19.21)
The value 4, can be .defioed.1!Seither the initial concentration 01oxidizable H
orgal1iC.lI!ID!~t or élSthe ultimate BOD.::- - -
wher~~ settlingvelocity(m d-I) andH = water depth (ni).
356 PAJaIV Oissolved Oxygen and Pathogens LECTURE19 BOO and Oxygen Saturation 357

At steady-state, Eq. 19.19 becomes kr

InL = In Lo - UX (19.25)
\~dL
~~ = -u dx- -:---
krL) (19.22) Consequently if the simple model hol.!. a plot of In L versus x/u (lhat is, travel
time) should yield a straight linewith a slope of kr.
If c!>fDplete
p1ixingis assum~ .~. t!!e-LocatioJl.of thedischarge, an initialconcentra- Figure 19.5 shows a typical pattem thal might be observed in a stream receiving
qo'l} clffioe cafculated as fue ftow-weighted averag~ (recall Eq. 9.42) of the Joadjpg . untreated sewage. NQ!e how the rate is higher immediately below the discharge.
(~u~~~fÍptw~ ¡m~ the BOD iñ the river upstreaOl of th,e discharge (subscript r), t~ . ~ve!al factors underlie such a patteJ:11~ In particular tbeJ1.!g!1erfate~!I'r

Qw+ Qr
'
!
/
.. caused by t~e fast degradation of readily decomposable organics and the seltling 01'
.. ..: .
i¡ .
sewag e partículates. - --- - -
~ ~+¡¡;-J - (19.23) !'

Using thisvalueas an initialcokaldulI,wccan solveEq. 19.22for

\ --
19.6 BOD LOADINGS,CONCENTRATIONS,AND RATES
. L = Loe-ü~k, J . (19.24) Jkfore proceedingto olheraspectsof 00 modeling,let's reviewsomeof Iheparam-
'fhus tbe BOOis reducedby decompositionandsettlingas it is carrieddownstream. eters that relaleto biocheinicaloxygeodemando
AsidefrommoddingthedistributionofBOObelowa poi~-ftow
str~amwith_conl'!an1.Panímeters,:pq~td~~ a frameworkforestimatiñgilie
19.6.1 BODs (5-pay BOD)
removalrate in such s~ste~.To do thíS,~~en..atu.~llogarithm
can be taken,
As in Table 19.1, <ypical values forJhe: BOO boule decay rate range from 0.05 lo
0.5 d- I, with a geometric mean' of about O.15 d-::.1.This information can be usedlo
estimate a 95% response time for the bottle test as 195 = 3/0.15 = 20 d. Because
.. s!lch a long measuremenl period is unacceptable, water-quality analysts early 00
Rlver adopted a 5-day BOO test.
Although shortening the incubation time to 5 d makes the test practical, we must
then have a means to extrapolate the 5-d result to the ultimate BOO leve!. This is
usually done by performing a long-term BOO to estímate the decay rateoUthe first-
.~ , order decomposition model holds, Eq. 19.12 can be used to compuh':
I k41 > k421
, Lo Y5 (19.26)
~o = 1 - e-kl(5)
.s
where Y5 = 5-day BOO. Table 19.1 ineludes typical values foe the ralio of 5-day lo
ultimate BOO.
Deep stream

19.6.2 BODLoadings and Concentrations

Table 19.2 provides typical values of flow rale and BOO for raw se\\ a~~ frono hulll
the United Slates and developing countries. In general the ftow ra!\: 1m Ihe lJllih:tI

t=xlU TAnLE 19.1

Typical values of the BOO bottle decomposition
fiGURE 19.5
rate for various levels of treatment. BODu is tbe
Plot of BOD downstream Irom a point source 01 untreated sewage ultimate BOD. Values bere are for CBOO
Into a plug-Ilow river having constant hydrogeometric characteristics.
11'1
the ¡nltialstrelch. high BOD removal rates will occur due to settling Treatment ~. 80D5180D.
~ncl Ill$t c:1~!j1positiol'l01easily degradable organic matter. Further
downstream thl}lower removal rates willoccur as tlle more relractory Untreated' 0.35 (0.20 .(1..)0) 0.83'
organk: matter degra~es at a slower rateo Primary 0.20 (0.10-0.30) 0.63
Activated sludge 0.075 (0.05--{).1O) 0.31
358 PART IV Dissolved Oxygen and Pathogens LECTURE 19 BODand OxygenSaturation 359
"
TABLE 19.2 10
Typicalloading rates for untreated domestic sewage

Per-eapita flow rate Per-eapita CBOD CBOD coneentratlon

-.
Uniled Slulcs
Dcveloping counlries
(m' eapita ' d ')
0.57 (150)'
0.19(50)'
(m' eapita-' d-')
125 (0.275)'
60 (0.132)'
(mgL-')
220
320
-
*'
...
1

0.1
I (¡"lInns c"pila I d"y 1; Ipnunds capil" I day I

0.01
SWles is highcr because higher water usc typically accompanies a higher standard 1 10 FIGURE 19.7
0.1
(11"living. Per capita generation rate 01"BOO is also higher because of garbage In-stream decomposition rale
dí~posals and other accouterments of a developed economy like that of the United H(m) versus deplh (Bowie et al. 1985).
Slalcs. The average concentration of the devdopi!!g countries is generally higher
bccause the lower water use in these coun\riesout~ighs the higher per capita BOO
Clll1lributionfor the United States. As might be expected. the bolll¡;:rate can rarely be direc1ly applied to rivers
\ \ because the bottle environment is not a good represeotaliQJ!of the r¡ver. In fa~
\1 - ~

Iy i~ de~.JIlow rivers~ld !he t}YQc..onv~..ln mo_~t other rivt:.f.JjJenvlronmental

19.6.3 BOD Removal Rates f~ClO-rste.!!4to mak~JeJJ1.QYal.bi&.herthan forOth~ ~ Ihe "primary causes of this
1 .increase...a{e~e.t~Ln&.
and be(l ~ff~c.t~,
Thc bottle BOO decomposition rate provides a first estimate of the removal rate in
J1¡lluralwaters. As listed in Table 19.1. Iherates depend on Ihe.degreeof treatment v Settling. Settling effects relate to the fact that for sewage witl} a significant
01'Ihe sewage priQr 10discha!Re. Raw sewage is a mixture Q! compounds rangi.Dg fraction of Qrganicsolids. the total removal rate in streams is a combinalion of seltling
fr(lm easily dec()I1lP'Q.,~(\!?le
sugars to refraclory substal!~«s that take 10ng,<rtQ~reak_ and decomposition (~ecall Eqs. 19.20 and 19.21).
d(lwll. Because wa"~t.!;.
Ire<\tmcJ!t 1~11~~!Q!ieJeclivcly re.!!1QY~tllc former. BOO bottle Vs'
mIes tend to be lower for treated sewa~e. ~ - kr = k" + H (19.27)

, ,0.0' í -'t "7 \()

r' l' ('

{\ ¡I"( ( ",,-\ Usingsome typicalseltlingvclocities.wecan 1>~e
(fig. 19.6)lb.a.~the settling etfect
3T I J can be particularly significant for~~raw sewage in shallow streams (that Is. < 1 m).
') ,\ ._~
r r
I 8ed effects. AII other Ihings equal. aJtach~t.lbac~eriagenerally are more effec-
\. . (
--
n (:'-1
tive decomposers than free-floating bacteria. Boltom decomposition can be param-
2+: V
- ,
L eterized as a mass-transfer flux of BOO. Thus in a way similar to seltling, bol~m
use the effect
decomposilion becomes more pronounced in shallower syste ms beca
r
....
I
"'C becomes more significant relative to the volumc::tricdecomposition in Ihe water...
. .
.

..:;c...
~ This trend. which is displayed in Fig. 19.7, has been fit by the equations (Hy-
droscience 1971)
-0.434
H
.. , /v..=O.Smd-1
.~-.-.-._.-.....-._..._._--------- ()
k" = 0.3 '8 O :S H :S 8 ft (19.28)

o vs= 0.1m d-I k" = 0.3 H > 8 ft

O 2 4 6 Thus up lo about 2,4 !TI(!!.ft)t!1e rate decreases wilh.del?!h. A\2Qy~2.4 m the rate
H(m) approacj1~ a.~Q.J1stantyalu~ihat i§ t)'pic.al.Qfhou1e.rates.
Ftnally. BOO decomposilion rates can be extrapolated lo other temperatures us-
FIGURE 19.6
ing Eq. 2.44 (k = k200T-20). with (1 == 1.047.
Plot 01 total removal rate versus stream depth lor BOD that is 50% in settleable
In summary 1:50Dremovalrates tend to increase with tel11(>~ratllre and lend
lorm. A range 01 settling velocities is depicted. Note that a decomposition rate 01
0.35 d ' is used. lo be higher immediately downstream I"rompoint sources. The latter effect is more
 360 PARTIV Dissolved Oxygen and Pathogens
}bCTURE 19 BOD and Oxygen Saturation 361
pronounced for untreated wastewater. In addition enhanced settling and hed effeGt& TABLE 19.3 /
~eans_tl!!t! shallower s)'stems typically exJ:¡il>ithigher BOD removal rates t!!.an_ Henry's constants for some gases commonly encountered
deeperwa~~ - in water-quality modeling (modified from Kavanaugh and
Trussell 1980)

19.7 HENRV'S LAW AND THE IDEAL GAS LAW Henry's constant (20"C)

Compound Formula (Dimensionless) (atm m3 mole-I)

If a beaker of gas-free distilled water is opened to the atmosphere, gaseous com- /

poun.ds such as oxygen, carbon dioxide, and nitrogen cross the air-water interface Methane C"" 64.4 1.55 x 100
(Uldenter into solution (Fig. 19.8). The process will continue until an equilibrium is
,,1; Oxygen O¡ 32.2 7.74 x 10-1
Nitrogen N2 28.4 6.84 x 10-1
e~ta~lished between the partial pressure of the gas in the atmosphere and the COncen- Carbon dioxide CO2 1.13 2.72 x 10-2
tration in the aqueous phase. This equilibrium is quantified by Henry's law, which Hydrogen sulfide H¡S 0.386 9.27 x 10-3
fan be representedas . Sulfurdioxide S02 0.0284 6.84 x \11-'
Arnrnonia NH3 0.000569 1.37x lO-s
- p
He. -- C
(19.29)
.-----------
wtIere He = Henry'sconstant(atmm3mole-I) H' = He = Cg (19.31)
p = partial pressure (atm) e RTa CI
C = water concentration (mole m-3)
where Cgand cl are the gas and liquid concentrations, respectively (mole m-3).
Henry's constants for some gases commonly encountered in water-quality modeling Notice that Henry's law specifies that at equilibrium the ratio of the gaseous to
are summarized in Table 19.3. the water concentration will be maintained at a constant value. The water concen-
Equation 19.29 can also be represented in dimensionless form by invoking the tration for a particular gaseous level is referred to as the saturation concentration.
ideal gas law In the following example we determine the saturation concentration for oxygen.
EXAMPLE 19.2. HENRY'S LAW AND OXYGEN SATURATION. Determine Ihe
C = R~a (19.30) saturation concentration of oxygen in water at 20°C. NOlethat clean, dry air near sea level
where R = universal gas constant [8.206 X 10-5 atm m3 (K mole)-J] and Ta = is composed of approximately 20.95% oxygen by volume.
absolute temperature (K). Thus the ideal gas law provides a means to express the Solution: Assuming that Dalton's law holds, the partial pressure of oxygen can be COIII-
partial pressure in concentration units of moles m-3. Substituting Eq. 19.30 into puted as
/ '2'f~ and rearranging yields the dimensionless Henry's constant p = 0.2095(1 alm) = 0.2095 alm
This value can then be substituted inlo Eq. 19.29 along wilh the value 01'Henry's conslanl
~ ~ ~ from Table 19.3 lo yield
p 0.2095 -3

C]CJI!}
H:c,<c, H:c,< c, H:c, =c,
or in mass unils,

.. e
V e =

= 0.2707
He =

mole m-3
0.774 = 0.2707 molem

(
32 g-oxygen
mole )= 8.66mgL-1

(a)

FI~UR¡: 19.,
(b) (e)

Tt111closed system in (a) is undersaturated with oxygen. When

it Is opened.to the atmosphere (b), oxygen comes into solution
until ~n equilibrium (e) is re¡¡ched. Henry's law provides the
mea.ns to 9uantily this equilibrium condition.
\ 19.8 DISSOLVED OXYGEN SATURATION

,As_calculatedin the previous example, the saturation concentration Qfoxygen in a

natural water is on the order of 10 mg L- l. In general several environmental fac-_
tors can affect this value. From the perspective of water-quality OJ.Q..deling,
important of these are:
the most
362 PART IV Dissolved Oxygen and Pathogens LECTURE19 BOD and Oxygen Saturation 363

TCF) 19.8.2 Salinity Effect

The followingequationcan be used to establishthe dependenceof saturationon
16 salinity(APHA 1992):
( 1.0754 X 101 2.1407 X 103

- 12 Ino
where 0.,.,
= In 0,[ - S (1.7674X \0-2 -
= saturation conce,ntration Qf disso,lved
Tu
oxygen
+ T~
in saltwater
) (19.34)
at 1 at!!!. (mg
L'I) and S = salinity (g L-1 = parts per thousand, ppt. sometimes given as %0).
-..
CI
E
Q
8 Salinitycanbe relatedtochlorideconcentrationby the followingapproximation:
S = 1.80655 x Chlor (19.35)
4 where Chlor = chloride concentration (ppt). The higher the salinity, the less oxygen
can be held by water(Fig. 19.9).
o EXAMPLE19.3.OXYGENSATURATlONFOR AN ESTUARY.Determine
o 10 20 30 40 the saluralion for an esluary with a lelllperalure 01'200e and a salinity 01'25 ppl.

TCC) Solution: Equalion 19.32 can be used lo compute

FIGURE 19.9
I
no,!
= _11...934411 +
1.575701 x I(}~_ 6.642308 x \07
293.15 293.152
Relationship 01 oxygen saturation in water to temperature and salinity.

1.243800 x \010 _ 8.621949

X \011 = 2207
+ 293.153 293.154 .
. Telllperature c,
. Salinity This corresponds lo a freshwaler saluralion value 01'
· Parlial pressure variations due to elevation ¡./"
o.,! = e2.2117 = 9.092mg L-1
Several empirically derived equations have been developed to predict how these
Equalion 19.34 can be used lo correel for salinity.
faclols influence saturation. These are reviewed in the following sections.
, 1.0754 x 101 2.1407 X \03

19.8.1 Temperature Effect

In0" = 2.207 - 25 1.7674x 10- -
( 293.15 + 293.152 ) = 2.060

which corresponds to a saltwaler salUralion value 01'

The following equation canbeused to establish the dependence of oxygen saturation

on telllperature (APHA 1992): 0,. = e2.()(,o= 7.846 mg L-1
6.642308 X 107 Thus Ihe saltwater value is aboul 86% 01'Ihe freshwater value.
In 0.'1 = -139.34411 + 1.575701x \05 _ T2
u
1.243800x \OtO 8.621949 X 1011
+ (19.32)
TJ T4u
19.8.3 Pressure Effect
wherc 0.,[ = saturáfiOn concentration ofdissolve.Q.QB'gen in fresh water at 1 atm The following equation can be used to establish the dependence of saturation on
(lIIg L 1)and Tu == absolutetefnperature\K). Rememberthat
pressure (APHA 1992):
T" = T + 273.15 (19.33)
where T = temperature (OC).According,Jo this eq\.lation, saturation decreases with
0,1' = O.d [1
(1 - ~)(I - fJ[1) ( 19.36)
increasing temperature.As displayedin Fig. 19.9, freshwater concentration ranges (1 - [11\'1')(\- fJ)
from about 14.6 mgL-t at O°C to 7.6 mg L -1 at 30°C.
 364 PAKfIV Dissolved Oxygen and Pathogens LECfURE 19 BOD and Oxygen Saturation 365

PROBLEMS
TCF)
40 50 60 70 80 90 100 19.1. Use separalion ofvariables to solve Eq. 19.6 for Eq. 19.7.
16 I I 1 I I I 1
19.2. A tanker lruck careens off the road and dumps 30,000 L of glucose syrup into a smaU

-
mountain lake. The concentration of the syrup is 100 g-glucose L-l.
12 (a) Compute the grams of CBOD spiIJed.
(b) Determine the lake's saturation concentrationof oxygen (T = _ JOoC;clev
1" 11,000 fl.).
el 41 19.3. A lake with a bay has the following characteristics:
!...
Lake Bay
~ -
1
\ Mean depth
Surface area
810
1.6 x 1()6102
3m
0.4 x 1()6102
Inflow
o
o 10 20 30 40~
) Inflow BOD concenlratíon
50,000 m] d-I
O mg L-'
4800 m] d-I
57 mg L -1

TCC) A subdivision housing 1000 people is planned that wiIJ discharge raw sewage into
FIGURE 19.10 the bay. Each individual contributes about 0.568 m3 capita-I d-I of wastewater and
~elationship o, oxygen saturation in water to temperature and elevation above 113.4 g capita-I d-I of carbonaceous biochemical oxygen demando
i :;ealevel. - (a) The bay inftow has a chloride concentrations of 50 mg L - t. The lake and the bay
/ have chloride concentrations of 5 and 10 mg L -l. respectively. Determine the bulk
diffusion coefficient between the lake and the bayo
where p = atmospheric pressure (atm) (b) If the BOD decays at arate of 0.1 d-I and seules at arate of O I '01d-I, deter-
(Jsp = saturationconcentrationof dissolvedoxygenat p (mg L-1) r mine the steady-state BOD concentration of the lake and bay in IHg L--I. Make
this determination with and without the subdivision.
Osl = saturation concentration of dissolved oxygen at I atm (mg L-1)
pwv = partial pressure of water vapor (atm) 19.4. The foUowing data have been collected for dissolved and total BOD below a point
source of untreated sewage into a stream. Use this data to estimate the BOD removal
fwv c::~ be calculated by rates (kr. k.. and kd) for the river. The velocity and depth are 6600 m d-I and 2 m,
respectively.
1 = II 8571 _ 3840.70 _ 216,961 (19.37)
npwv. ru r¡
x (km) O 5 10 15 20 25 30 35
'[he para meter (J can be computed as Dissolved(mg L-1) 20.0 17.0 14.7 12.9 11.6 10.5 9.6 1\.9
TOlal(mgL-I) 40.0 29.4 22.5 17.8 14.6 12.3 10.7 9.6
/ (J = 0.000975- 1.42~x 1O-5r + 6.436 x 1O-8r2 (19.38) x (km)
Dissolved(mg L-')
40
8.3
45
7.8
50
7.3
60
6.6
70
6.0
80
5.5
90
5.1
100
4.7
N ¡tice that thj~ formula is wriuen in terms of temperature in degrees Celsius rather TOlal (mg L-') 8.7 8.0 7.5 6.7 6.0 5.5 5.1 O
Ihan Kelvin.
Zison et al. (1978) have developed a handy approximation based on elevation,
19.5. Determine the saluration concentrations al 20°C for (a) nitrugen gas (¡l:>.
j % by volume
Osp = Osl[l - 0.1l48 x elev(km)] (19.39) in the atmosphere) and (b) carhon dioxide (0.0314%).

or in English units, _
19.6. A ftow of 2 cms with a 5-d BOD of 10 mg L-1 is discharged from an activated-sludge
,,-,,\. treatment plant to a stream with a ftowl){5 cms and zero BOD. Stream characteristics
Osp = Ostp ..,. 0.OOOO3~ x elev(ft)] (19.40) are kr.20= 0.2 d-I , cross-sectional area = 25 m2, and T = 28°G,.
(a) What is the. óJnct'l1Ir:ltionofBOD at the mixing point? )
(b) How far 1,,-, ! le dliuent will the stream BOD concent tion faU to 5% of ils
whereelev = elevation above sea level -\s displayed in Fig. 19.10 this relationship original vah
indicatesthat as pressuredecreasesat hlgherelevations,the saturationdrops. r
366 PARTIV Dissolved Oxygen and Palhogens

19.7. The folIowing temperalures, salinities, and IIxygen concelllralions are measured in an LECTURE 20
esluary:

Distance from ocean, km

Temperature, °C
30
25
20
22
10
18
Gas Transfer and Oxygen Reaeration
Salinity, ppt 5 10 20
Dissolved oxygen 5 6.5 7.5

Calculate Ihe percenl saturalion of oxygen at Ihe Ihree localions.

19.8. Whal is Ihe oxygen saluralion concenlralion of a saline lake (moslly sodium chloride)
Ihal is localed al an elevalion of 1 km, and has a salinity of 10 ppt, and is al a lempera-
lure of 25°C?

--r.>:
- --
LECrtJRE OVERVlEW: t describe two theories of gas trlUlsfet: tlIe tWo.flI~ni,bdel:
and the surface renewal model, and show howUteycan be ltsedfo Í110aét ~ty~ ' ~
gen ttansfet lh natural waters. In addition t review fofmuias to I'h~alH relÍé¡'a~ I
tlbn lb streams.

Suppose we fill an open bottle wilh oxygen-free dislilled waler. We know froro e
previQus leclure thal, given sufficient lime, .almoliphericoxygen wil\ enler until Ihe
solulion reaches Ihe saluralion level defined by Henry's law. Similarly if we had a
bottle órwater that was supersaturaled, over lime oxygen would leave the solulion
uñlil ihesatiffiítíon value was reached.
The key queslion is, "How long will il lake?" In olher words we would like lo
assess Ihe rale of the process. Now let's imagine an experiment 10quantitatively
answer Ihis queslion.
We have our open bottle filled with oxygen-free distilled water. As depicted in
Fig. 20.1, we place a mixing device in the bottle. T.his device keeps the volume
well-mixed except in Ihe bottle's narrow neck, where molecular diffusion govel1JS
Iransport.
To model Ihis syslem, assume Ihal Ihe waler al Ihe air-water inlerface is al Ihe
saluration concenlralion. Under Ihis assumplion, a mass balance for the bottle can
be written as

r Vdo = DA o,.-' o
~ -d.J---A
(20.1)

where D = molecular diffusion coefficienl of oxygen in water (m2 d-I)

A = cross-seclional area of Ihe bOltIeneck (m2)

367
 LECfURE20 Gas Transfer and Oxygen Reaeration 369
368 PARTIV Dissolved Oxygen and Pathogens

which can be used to determined the foJlowing values as a function of time:

d
o,
Time o 80 160 240 320 400
Oxygen 0,00 4.01 6.25 7.50 8,21 8.60
~.l~.~...........

A graph 01' the results can also be generaled:

o 10
~a_~ _ _. ~_. __.
Saturation
_._.................
FIGURE 20.1
An open well-mixed bottJe with a narrow 8
neck in which molecular diffusion governs --
transporto 6
!J
al
o. = oxygen saturation concentration (mg L-1) E
-- 4
o = oxygeri concentration in the botde (mg L -1) <:>

I1H = length of the neck (m) 2 )

The modelcan alsobe expressedas
do (20.2) 100 200 300 400
V dt = KLA(o. - o)
Time (d)
wpere KL = oxygenmass-transfervelocity(m d-I), whichis equalto FIGURE E20.1.1
D
KL= I1H - (20.3)
Over a long period of time the concentration approaches the saturation value. The
Dividing both sides ofEq:-20.2 by the volume and rearranging yields time can be quantified by delennining a 95% response time,
do 3
19S= _ = 412 d
dI + kao = kao. (20.4)
where ka = reaeration rate (d-I), which is equal to KLAIV. Together with the initial . Thus according to the model, it would lake 0\ '\ I yr lo reach 95% of saturation.
conditionthat o = Oat t = O,Eq. 20.4 can be solvedfor
o = o.(l - e-kal) (20.5)
In the foregoing example we calculated that it would take over I yr for a boule
of water to reoxygenate. Although the situation in nature is not as slow, gas transfer
¡i:XAMPLE 20.1. OXYGEN TRANSFER FOR A BOITLE. You fiJl the 3OO-mL in natural waters involves many of the principies we used to model the bottle.
~ottle ~ho,!\,nin Fig. 20.1 with oxygen-free water. Calculate the oxygen concentration as
a function oftimeif D = 2.09 x )O-s cm2 S-I, d = 2 cm, and ~H = 2.6 cm. Assume
that the system is at a temperature of 20°C and the saturation concentration is 9.1 mg L-l. 20.1 GAS TRANSFERTHEORIES
Solutlon: First, we must detennine the mass-transfer velocity,
We will now describe two theories that are widely used to describe gOl,,'allsfer in 1I,lt- !
~ = 0.006945 m~-
KL = 2.09x "1&'_-
)O-s cm2 S-I
( 100cm
86,400s
d ) /
ural waters. Although both are used in stn':lffiS, estuaries, and lakes, 111.'slagnanl 111111
theory is more widely used in standing \,aters such as lakes, wherea~ tlle surturc- (

renewal model is more commonly used in flowing waters such as streams. \

Next we can calculate the reaeration coefficient, \

0j' . ·=
. k, 0.006945m300
d-I[1T(0.0l)2]m2
mL ( )=
1Q6mL
m3
0.007273d-I
~0.1.1 Whltmal'!'s Two-Fllm Theory
::1
.
'1'h"
.

P~ete~
f .~ ~'Jt ¡'. , '.
,,1m\!I¡§\lbstitured intoEa 205" ,
. , .,
tt
1 A simple mo<JeJof 8as e~change is provi~e~ \:!yWhitman's two-filmor two-
.,;¡ . '.'T~J' o , ' ,f"'t1i .
'J ~ j'~ "W l' '1 ( ..,..j
iS! , " f"" 9.1Ú - ~-o.lXI7ihí) fysistancemodel(Whitman1923,~wis and)Vhitman1924).
:no PARTIV Dissolved Oxygen and Pathogens LECI1JRE20 Gas Transfer and Oxygen Reaeration 371
,j,

I
The transfer eoefficients can be related to more fundamental parameters by
Turbulent BULK
GAS
Pg
KI =-DI (20.8)
ZI

and Kg _-- Dg (20.9)

Zg
I AIR-WATER
laminar where D, = liquid moleeular diffusion eoefficient (m2 d-I)
= INTERFACE
Dg = gas moleeular diffusion eoefficient (m2 d-I)
Z, = thickness of the liquid film (m)
Zg = thickness of the gas film (m)
c,
j A key assumplion 01'the two-film theory is that an equilibrium exists at the air-
t
Turbulent
water interface. In other words Henry's law (reeaIl Eq. 19.29) holds:
BULK p¡ = Hec¡ (20.10)
I
LlaUID
Equation 20.10 can be substituted into Eq. 20.6, whieh can be solved for
FIGURE 20.2
Schemalic reprasenlalion 01 Whilman's Iwo-lilm Iheory 01 gas Iransler. Liquid (20.11)
and gas concenlralion allhe inlerface are assumed lo be al an equilibrium as p¡ ~ He(~ + CI)
delined by Henry's law. Gradienls in Ihe lilms conlrollhe rale 01 gas Iransler
belween Ihe bulk fluids. Equation 20.7 can be solved for

p¡ = pg - -RTaJg (20.12)
As dcpicled in Fig. 20.2, the bulk 01'main body oflhe gaseous and (iquid phases Kg
ale assumed lo be turbulently well-mixed and homogeneous. The Iwo-film theory as-
Equations 20.11 and 20.12 can be equated and solved for flux,
SlllllCSIhal a subslance moving belween the phases encounters maximum resistance
inlwo laminar boundary layers where mass transfer is via molecular diffusion. The
ma'iS Iransfer Ihrough Ihe individual tilms would be a function 01'a mass-transfer (20.13)
J = v,,(~: - CI)
velqcity and Ihe gradienl belween the concenlrations at the inlerface and in the bulk
I1nid.For eXílll1plelransfer through the liquid film can be represented by where v" = net transfer velocily aeross the air-water interface (m d-I), whieh can
be compuled by
J, = K,(c¡- c,) (20.6)
l 1 RTa
wlll're J, = mass l1uxfrom Ihe bulk liquid lo Ihe interface (mole m 2 d-I) - =- +- (20.14)
VV K, HeKg
K, = mass-transfer velocity in Ihe liquid laminar layer (m di)
( , :llId c, = liquid eoncentrations al Ihe air-water interface and in Ihe bulk liquid, Equation 20.13 now provides a means lo compule mass transfer as a funetion of
rcspeelively (mole m .1). the gradient belween the bulk levels in the gaseous and the liquid phases. In addition
it yields a net transfer velocily (Eq. 20.14) that is a function of environmental ehar-
Similarly lransfer Ihrough Ihe g<lseousfilm can be represented by aeleristics K, and Kfi and Ihe gas-speeifie parameter He. Note that Eq. 20.14 can be
Kg inverted to ealculate the mass-lransfer velocity direetly,
(20.7) ~
J~ = RT" (Pg - p¡) H,
Vv = K, (20.15)
whne J~ = mass flux from the inlerface to the bulk gas (mole m-2 d-I). He + RTa(K,IKgj
Kg = mass-transfer velocily in Ihe gaseous laminar layer (m di)
I',! and p¡ = the gas pressures in Ihe bulk gas and allhe air-water interface, respec- Notice that I have modified the nomenclature slightly by using a "v" rather than
lively (atm) a "[(" for the nel transfer veloeity. I did Ihis lo make the eoefficient's nomenclature
eonsistent with its units-thal is, a veloeity. The subseript v is intended to signify
t.J,,'ice thal for both Eqs. 20.6 and 20.7, a posilive flux represents a gain lo the water. that the eoeffieient is a volatilization mass-transfer veloeity.
372 PARTIV Dissolved Oxygen and Pathogens LECI'URE20 Gas Transfer and Oxygen Reaeralion 373

Llndane NH3 Aldrln

\ \
\ SO, NaphthaleneH,S

~
el) Dleldrln

" l\D~ \\~/:~D,N,CH'

~ roo%
g'
::
.5 50%
B
I :
:&
:j'
..,

~
,,
1:-!Y
f
:/
IJ
".

e
R =R. + R2
1
-=-+- 1 RT. FIGURE 20.3 S 1:
~'
Llquld controlled
v. K¡ H,Kg The two-lilm theory 01 gas transler ti)
.¡¡¡
is analogous to the lormulation lor el) 0%
(a) Two reslstors (b) Two fllms twq resistors in series in an electrical a: I I I I
In series
11'1!$8r1es circuit. ~ 1~7 101 10~ 104 10~ 10~ 1~1 1 10

He (atm m3 gmol)
Also observe that Eq. 20.14 seems to be analogous to the formulation used to
Soluble - - Insoluble
determine the effect of two resistors in parallel in an electrical circuit: FIGURE 20.4
Percent resislance to gas transler in the liquid phase as a luncUon 01 H.. the
1 1 1
- = - +- (20.16) Henry's constant lor lakes. Values 01 H. lor some environmentally important
R RI R2 gases and toxic substances are indicaled (modilied lrom Mackay 1977).

Although Eq. 20.14 might superficially be in this format, it can be recognized that
the resistance in each film is actually the reciprocal of its mass-transfer velocity. 20.1.2 Surfac,e Renewal Model
Consequently Eq. 20.14 is actually analogous to the formulation used to determine
the effect of two resistors in series in an electrical circuit (Fig. 20.3). We now turn to a model that takes a different approach to gas Iransfer from Ihe I
As in Eq. 20.15 the total resistance to gas transfer is a function of the individ- two-film theory described in the previous seclion. Rather than as a slagnanl film, I
ual resistances in the liquid and the gaseous boundary layers. The liquid, the gas, or the system is conceptualized as consisting of parcels of water Ihat are hrought lo Ihe
both layers can be the controlling or limiting factor depending on the values of the surface for a period of time. While at the surface, exchange takes place. Then the
three coefficients K¡, Kg, and He. This can be quantified by using Eq. 20.15 to de- parcels are moved away from the surface and mixed with the bulk liquid (Fig. 20.5).
velop (Mackay 1977) Higbie (1935) suggested that when the liquid and gas are firsl hrought into
conlact, the liquid film will be at the concenlration of the bulk liquidoThus, prior
R¡- - He (20.17) to the situation envisioned by the two-film theory (Fig. 20.2), the dissolved gas
He + RTa(K¡IKg)
must penetrate Ihe film. Hence it was dubbed the penetration theory. The evohlliull
where R¡ = ratio of the liquid-layerresistanceto the total resistance.For lakes, of this penetralion is depicted by the succession of dashed lines in Fig. 20.6. lf the
Kg varies from approximately 100 to 12,000 m d-I and K¡ from 0.1 to 10 m d-I Flux
(Liss 1975, Emerson 1975). The ratio of K¡ to Kg generally ranges from 0.001 to
0.01, with the higher values in smalllakes due primarily to lower Kg because of
sheltering from wind. A plot of R¡ versus He (Fig. 20.4) indicates where the liquid,
gas, or both films govem transport for contaminants of differing solubility. In gen-
eral the higher the Henry's constant, the more the control shifts to the liquid film.
A¡so note mat sQ1allerlakes tend to be more gas-film controlled than larger lakes.
1~ p¡entioned previously the two-film theory usually represents a good approx-
iltlation for standing ~aters such as lakes. Next we tum to another theory, one that
e LU)/
e¡¡:tendsthe two-film theory to systems such as streaQ1Sthat have strong advective
FIGURE 20.5
ft!>w..
Depiction 01 surface renewal model 01 gas exchange.
374 P¡\HTIV Oissolved Oxygcn and Palhogens LECTURE20 Gas Transfer and Oxygen Reaeration 375

c(O. t) = c.I boundary condition at air.water interface (20.20)

Turbulent BULK c(oo.r) = c¡ bouom boundary condition
GAS
where D¡ = (iquid diffusion coefficient
c, = concentration at the air-water interface

Laminar
¡ P,
\7 AIR-WATER
CI = concentration in Ihe bulk water
Applying these conditions, we can solve Eq. 20.19 for
INTERFACE
e(z. t) = (c,-CI)erfc ~
(2.,¡D¡r )
(20.21)

C¡ "~:::::::::::::.:,:.:~,:.> = where erfc is the error function complement, equal to ( - erf. where erf is the error

Increaslng

~~~
~,'
,
~
"'...,.,\ '.
..'
?0----
function (recall Seco 10.3.2 and App. G),

erfc/J = -
2

fio 1
01> 2
e-~ dg (20.22)
Turbulent
: BULK The flux across Ihe air-water interface can be computed by applying Fick's first
I
: LlQUID law at the inlerface (: = O),
FIGURE 20.6
J(O. r) = - D¡-8c(0.
az
r)
(20.23)
The temporal evolution 01 Ihe liquid li 1m immedialely alter il is broughl inlo
conlacl wilh Ihe gas.
and the average flux is detennined by

prot.:essis nol inlerrupled. Ihe Whilman Iwo-lilm condilion (the solid line) will be at- J=- Id' J(O. r) dr
(20.24)
ta¡ned. ('
As described in Box 20.1, lhe penelralion Iheory can be used lo eslimate the flux where (' = average contact lime of Ihe fluid parcel al Ihe interface. Equalion 20.21
01'gas across Ihe air-waler inlerfaee as can be differentiated and substiluled inlo Eqs. 20.23 and 20.24 and solved for

rn; I

J = V ;'F(c, - c,) (20.18) J = {§;

-(c,
7T(' .
- el! (20.25)

wh"le D, = liquid diffusion eoefficienl

c., = concenlralion allhe air-waler inlerface
(', = cllncenlralion in lhe bulk waler One of Higbie's underlying assul11ptionswas lhat all packels of water have lhe
(' = average conlacllime of Ihe fluid parcel allhe inlerface same contacl time at lhe interface. Danckwerls (1951) modified lhe approach by
assuming Ihat the fluid elements reach and leave the interface randomly. That is, their
'j his eqllalion is 01'lillle vallle in itself hecause lhe average contact lime al the in- exposure is described by a stalislical distribution. This approach, which is called the
lerface is í!ifficult lo measure. However, Eq. 20.18 yields Ihe valuable insight that if sur/ace rellewal tlleory, was used to derive
111('p('nelration theory holds, the mass-Iransfer velocily is proportionallo the square
WlIt"f Ihe gas's molecular diffusivily. J = JD,r(c., - c,) (20.26)
where" = I¡quid surface renewal rale, which has units 01' T-I .
The surface renewallheory can also be applied to the gaseous side of the inter-
BOX 20.1. Derlvation o, Penetratlon Theory face by assuming that packets 01'gas are brought into contact wilh Ihe air-water in-
Surpose Ihal a pareel of water moves lo Ihc air-waler interface (Fig. 20.5). The pareel terface in a random fashion. The transfer velocities for the liquid and gaseous phases
can thus be written as
can be idealized as a one-dimensional semi-infinile medium described by Ihe equation
a('
- = /),-
(12('
(2().19)
K, = /r,D, (20.27)
a( ,1,,2
and K~ = ,'rgD~ (20.28)
s1Ihjcel lo Ihe inilial and hOUlu.Jarycondilions
These relationships can be subslituled inlo eilher Eqs. 20.14 01'20.15 to eSlimale a
c(~. O) = (', initial eondilion
lotal transfer velocil)' 1'01'the inlerface.
376 PARTIV Oissolved Oxygen and Pathogens LECfURE20 Gas Transfer and Oxygen Reaerati()n 377

We can now see that a major difference between two-film and the surface re- where ka = reaeration rate, which is equivalent to
newal theories relates to how the liquid and gas-film exchange velocities are formu-
lated. In particular for the two-film theory, the velocities are proportional to D (Eqs. ka = K¡
H
(20.35)
20.8 and 20.9), whereas for the surface renewal theory they are proportional to the
square root of D (Eqs. 20.27 and 20.28). Regardless of how the oxygen transfer rate is parameterized, Eq. 20.32 or 20.34
We retum to the topic of gas transfer "i¡¡.:n we cover toxic substances later in provides insight into how the mechanism of oxygen reaeration operates. The direc-
this textoAt that time, I provide additional information on Henry's constant and ex- tion and magnitude of the mass transfer depends partially on the difference between
change coefficients for organic toxicants. For the time being let's naITOWour focus the saturation value and the actual value in the water. If the water is undersatu-
and concentrate on the problem at hand: oxygen transfer. rated (o < os), then transfer wil\ be positive (a gain) as oxygen moves fcom the
atmosphere into the water to try to bring the water back to the equilibrium state of
saturation. Conversely if the water is supersaturated (o > os), then transfer will be
20.2 OXYGEN REAERATION negative (a loss) as oxygen is purged from the system.
At this point we have a general equation for the flux of any gas (Eq. 20.13), Oxygen reaeration cates can be extrapolated to other temperatures by
ka.T = ka,200T-20 (20.36)
J = vv(;;' -c¡) (20.29) where O = 1.024.
Now let's apply it to oxygen reaeration. Because of its high Henry's constant (= 0.8
atm 11)3mole-I), oxygen is overwhelmingly liquid-film controlled. Consequently
20.3 REAERATIONFORMULAS
Vv = K¡and Eq.20.29becomes
Many investigators have developed formulas for predicting reaeration in streams
and rivers. Comprehensive reviews can be found elsewhere (Bowie et al. 1985). In
J = K¡ (~: - o) (20.30)
this section we describe some of the more commonly used formulas that have been
where o = oxygen concentration in the water. Further, because oxygen is so abun- developed for natural waters.
dant in the atmosphere,the partialpressureis constantand therefore
J = K¡(os - o) (20.31 )
20.3.1 Rivers and Streams
where Os = saturationconcentrationof oxygen. Numerous formulas have been proposed to model stream reaeration. Among these,
Next the mole flux can be converted to a mass flux, and the ¡¡quid concentrations
three are very commonly used: the O'Connor-Dobbins, Churchil\, andOwens-Gibbs
can be reexpressed in mass rather than mole units by multiplying both sides of Eq. formulas.
20.31' by the molecular weight of oxygen (32 g mole-1). The equation can also be
~n~formed from a flux to units of mass per time by multiplying it by the surface area O'Connor-Dobbins. In Seco20.1.2 we developed the surface renewal mnde!.
of ftte liquid exposed to the atmosphere. Thus for a well-mixed open batch reactor, For oxygen this model can be used to formulate the transfer velocity as
a mass balance for oxygen can be written as
do K¡ = Jr¡D¡ (2(U7)
V dt = K¡As<os - o) (20.32) O'Connor and Dobbins (1956) hypothesized that the surface rencwal cate cuol!!
where As = surfacearea of the waterbody. be approximated by the ratio of the average stream velocity to depth,
Finally there are many cases (especially streams and rivers) where the transfer U
velocity is expressed as a first-order rate. In cases where the air-water interface is rl = ¡¡ f1 /~ ,1 )

not constricted (as was not the case for the bottle from Fig. 20.1), the volume is
This hypothesis was backed up by experimental measurements. SUbsliLulingttus
V = AsH (20.33) value into Eq. 20.37 yields

whereH = meandepth.If this is true Eq. 20.32canbe expressedas (20,39)

[
K¡ = JD~U
do
(20.34)
V dt = kaV(os - o) The relationship is usually expressed as a reaeration rate,
J7!! PARTIV Dissolved Oxygen and Pathogens LECTURE20 Gas Transfer and Oxygen Reaeration 379

UO.S TABLE 20.1

k" = jD;H 1.5 (20.40) Ranges of depth and velocily used to develop the
O'Connor-Dobblns, Churchlll, and Owens-Glbbs
The diffusivity 01'oxygen in natural waters is approximately 2.09 X 10-5 em2 S-l. formulas for stream reaeration
Therefore the O 'COIlnor-Dobbills formula ean be expressed as
Parameter O'Connor-Dobhins Churchill Owens-Gibbs
Me/ríe: ElIglish:
Dcpth. m 0.30-9.14 0.61-3.35 0.12-0.73
ft 1-30 2-11 0.4-2.4
U°.5 UO.S
k.,,-- 12.9- (20.41) Vclocity. I11pS 0.15-0.49 0.55-1.52 tJ.03-0.55
k" = 3.93 fiT3 H1.5 fps 0.5-1.6 1.8-5 0.1-1.8

Unils: k,,(d '), U(lllpS), H(m) Units: k,,(d-I), U(fps), H(ft)

Churchill. Churchill et al. (1962) used a more empirical approach than that Notice that the O'Connor-Dobbins formula generally gives lower values than \he
uscd by O"Connor and Dobbins. They exploited Ihe faet that the water leaving some Churchill and Owens-Gibbs formulas. One possible explanation is that the slower,
nI' the reservoirs in the Tennessee River valley were undersaturated with oxygen. deeper channels for which O'Connor-Dobbins performs best are more idealized (i.e.,
Thc)' therefore Illeasured oxygen levels in the slrelehes below these dams and eal- more like a flume) than faster. shallower streams where drop structures and riffles
culaled associatcd reaeralion rates. They then correlaled their results with depth and may enhanee reaeration.
vclocily lo oblain
MI'/rie: Ellglí.l'h: Vclocity (fps)
0.1 0.2 0.30.4 0.60.8 1 2 3 4 56
U U . . . r . .
(20.42) 15 1,. ' . I ./ -->.....> 50
k" = 5.026 HI.67 k" = 11.6 HI.67 -
10
8 j lA :
Same unils as Eq. 20.41.
6 __________ ().I I ----- J/ .A- 20
5
()wens and Gibbs. Owens el al. (1964) also used an empirical approach, but 4
r- ...............

Ihcy induced oxygen depletion by adding sulllte lo several streams in Great Britain. 10
Tlo<:vcOlllbinedIheir results wilh the data frolll the Tennessee River and lit the fol- 8
2 -1 ----- ----- I
1,,\,ing formulas: ./16
5
Me/rie: EIIgli.\'¡': 4
()cplh(l1I) I _______ 3 Dcplh(fI)
0.8
k,,--_e .,.32-UO.67
HI.X~
k ,,--- 2 1.6-
UO.67
!II.X~
(20.43)
0.61
0.5
- 5" _ -->_ 5Y'/,--,u- o......-::;::
I 2

0.4
0,3 I
Same units as Eq. 20.41.
w
Compurison among formulas. As summarized in Table 20.1, Ihe O'Connor- 0.5
0.4
Dnhhins, Churchill, and Owens-Gibbs formulas were developed for differenl types O.I
r.- O..o¡;;-l 0.3
01 slleams. Covar (1976) fOllndthat they could be used jointly lo prcdict reaeration O.Og
ror ranges 01'dcpths and velocity eombinations (Zison et al. 1978). Aecording to Fig. 0.0.\ (W6 0.1 0.2 0.4 0.60.8 I
20.7. O'Connor-Dobhins has Ihe widest applicabilily being appropriate for moderate Vdocity (1111'S)
lo deep streams with moderale to low velocilies. The Churehill formula applies for
silllilar dcpths but ror faster strcams. Finally the Owens-Gibbs relationship is used FIGURE 20.7
101'.hallowcr systcms. Reaeration rate (d ') versus velocity and depth (Covar 1976 and Zison et al.
1978).
380 PARTIV Dissolved Oxygen and Pathogens LECTURE20 Gas Transfer and Oxygen Reaeration 381

Other formulas. There are many olher reaeralion equalions bcyond Ihe O'Con- 20... .} Standing Waters and Estuaries
nor-Dobbins, Churchill, and Owens-Gibbs formulas. Bowie el al. (l9~5) provide an
eXlensive compilation of many formulas along wilh references 10major criliques and
For standing waters. such as lakes, impoundmenls,and wide eSluaries,wind be-
comesthe predominantfactorin causingreaeralion.
inlercomparisons that have been performed.
In addition. along with allowing the user to specify reaeration values directly,
Lakes. The oxygen-transfer coefficient itself can be estimated as a func!ion
software packages such as EPA's QUAL2E model also provide the option to automat-
ically compute reaeration rates according to formulas. I will present Ihese formulas of wind speed by a number of formulas. Some. such as the following relalionship
when I describe the QUAL2E model in Lec. 26. developed by Broecker et al. (1978), indicate a linear dependence,
K¡ = 0.864Uw (20.45)
where K¡ = oxygen mass-transfer coefficient (m d-I) and Uw = wind speed mea-
20.3.2 Waterfalls and Dams sured 10 m above Ihe water surface (m s-I).
Oxygen transfer in streams can be significanlly influenced by the presence of wa- Others use various wind dependencies to altempt lo characterize the dijj~rellt
terfalls and dams. Butts and Evans (1981) have reviewed efforts to characterize this turbulence regimes Ihat resuh al the air-water interface as wind velocity increases.
transfer and have suggested the following formula: For example the following is a widely used formula of Ihis Iype (Banks 1975, Banks
and Herrera 1977):
r = 1 + 0.38abH(1 - O.IIH)(1 + 0.046T) (20.44)
K¡ = O.728U~5 - O.317Uw + O.0372U;;. (2046)

where r = ratioof the .'.

ÍI above and he10wthe dam Thus at high wind velocities, the relationship becomes dominaled by the second-
H = difference 111w¡¡(Cfe1evation (m) order term, as shown in Fig. 20.8.
T = watertemperatureeC) As wilh stream reaeration formulas. lake oxygen-transfer formulas have bolh
a and b = coefficients that correct for water quality and dam type empirical and theoretical bases. For example Wanninkhof el al. (1991) useJ gas
tracer experiments in lakes to develop the following formula:
Values of a and b are summarized in Table 20.2. 0.5
S
K¡ = ( )
O.108u~646~ (20.47)
TABLE 20.2
where Se = Schmidt number, which for oxygen in water is approximately 500. If
Coefficieot values for use io Eq.
this value is adopted the Wanninkhof formula reduces to K¡ = O.0986U ~64 .
20.44 to predict tbe effect oe dams 00
stream reaeratioo
10
Water-quality coefficient
Polluted state a
Gross 0.65
1.0
Moderale
Slighl 1.6 t
Clean 1.8 ! 5
Dam-type coefficient i:<"'¡

Dam type b
Flal broad-creSled regular Slep 0.70
Aal broad-cresled irregular slep 0.80
Aat broad-cresled vertical face 0.60 ." ~.,-..-
Aal broad-cresled slmighl-slope face 0.75 o
Aal broad-cresled curved face 0.45 O 5 10
Round broad-crested curved face 0.75
Sharp-cresled straight-slope face 1.00 !i Uw(m S.I)
'!1
Sharp-crested vertical face 0.80
Sluicegates 0.05 FIGURE 20.8
Comparison 01wind-dependent reaeralion formulas.
3H2 I'ARTIV Dissolved Oxygen ano Palhogens LECTURE20 Gas Transfer and Oxygen Reaeration 383

Finally O'Connor (1983) has developed a Iheoretically based sel of formulas to Saturatlon
10 . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _'.., _, _ _ _,.oo.. oo, _ _ _ _.
compule transfer for low-solubility gases as a function of wind. His scheme can be
applied to oxygen (Fig. 20.8). Calculated wlth
1I should be noled Ihal there are many more formulas'for calculaling oxygen
Iransfer as a funclion of wind. Many of Ihese are summarized in general references -- dallywlnds ~
such as Bowie el al. (1985). As evident from Fig. 20.8 these formulas yield a wide ~ .
01
range of prediclions. Consequenlly il is advisable lo oblain system-specific measure- 5
.§.
menls lo check Ihe validily of Ihe formulas before using Ihem in model calculations. O
As was the case for Wanninkhof et al. (1991) this can be done wilh artificial trac- Q

ers (see Seco 20.4). In addition natural oxygen-depletion events can sometimes be
exploiled to oblain direct measurements (Box 20.2).
o
O 10 20 30
BOX20.2. Dlrect Measurement o, Reaeratlon In Lakes
Elapsed time slnce overturn (d)
l\1any lakes in lemperale regions are Ihermally slratilied in Ihe summer. consisling 01'an FIGURE 820.2
IIpper layer (epilimnion) and a lower layer (hypolimnion). In general the surface layer
ha~ dissolveo oxygen concenlralion near saluralion. 11'il is productive (thal is, has high
phnt growlh). setlling plant maUer can collecl in Ihe hypolimnion. The decomposilion
01 Ihis maUer can Ihen lead to severe oxygen deplelion in Ihe bouom walers. When
lurnover (thal is, vertical mixing due lo dropping temperature and increasing winds)
Estuarles. Because estuary gas transfer can be affected by both water and wind
m;curs inlhe fall, Ihe mixing 01' Ihe Iwo layers somelimes results in Ihe lake's having
velocity, efforts lo determine reaeration in esluaries combines elements of current
an oxygen concenlralion well below saluralion.
In certai n cases Ihe lake can be assumed to act as an open balch reaclor: Ihat is, we
and wind-driven approaches.
can ignore inflows and oUlflows 01'oxygen, excepl gas Iransfer across Ihe lake's surface. The waler velocily effecls are Iypically compuled with the O'Connor-Dobbins
11any addilional sources and sinks 01' oxygen (such as sedimenl oxygen demand) are formula (Eq. 20.40),
nt"gligible, a mass halance for oxygen can be wriuen for Ihe lake inlhe period following
!!verturn as ka = JD¡Uo (20.50)
fj3i2
do
v- = k,,(o,,-o) (20.4R) where Uo = mean tidal velocity over a complele tidal cycle,
dI
The wind effects can be computed with any of the formulas developed for stand-
11'Ihe saluralion value is constanl over Ihe ensuing period, this equalion can be sol ved ing waters in Ihe previous paragraphs. For example Eq. 20.46 can be expressed as a
11\1(with o = o; al I = O) reaeration rale, as in
o = (/¡e-k", + 0,,(1 - e-k",) (20.49)
ka = 0.728U~5 - O.317U,,:-+O.0372U; (20.51)
H
rhus if Ihe oxygen concentralions are measured as a funclion 01' lime, Ihis model pro-
vides a means lo eSlimale Ihe reaeralion rale. Thomann and Fitzpatrick (1982) have combined the two approaches for estuar-
Geloa el al. (1996) applied such an approach lo Onondaga Lake in Syracuse, New ies affecled by both tidal velocities and wind,
y,¡rk. Figure 820.2 shows oxygen concenlralions Ihal occurred in Ihe lake following
fall overturn in 1990, along wilh a curve lil wilh Eq. 20.49 using a reaeralion rale 01' k = 3 93 jfJ'; 0.728U~5- 0.317Uw+ 0.0372U; 205 2)
approximalely 0.055 di. In addilion Ihe plol also shows a simulalion using a wind- a . H3/2 + .. ( .
dt:'pendenl reaeralion rale. The superior lil oblained wilh variable winds iIIustrates how
importanl accurale wind eslimales are for simulaling gas lransfer in lakes.
The approach 01' Gelda el al. (1996) is particularly appealing because il ís nonob-
Irw;ive; Ihal is, il does not depend on Ihe introouclion 01' lracers ano dyes lo Ihe envi-
20.3.4 Extrapolatlng Reaeratlon to ?ther Gases
"'IImenl. It is also atlractive hecause il oirectly measures oxygen concenlralion on a As slaled earlier, we are going to return to Ihe topic of gas transfer when we model
whole-Iake basis. Wherever possible such oirecl measuremenls are preferable lo indi- toxic organics later in this text. However, beyond toxics there are a few other com-
Iel I formula eslimation.
mon gases that are of interest in environmenlal engineering.
"l..

384 PARTIV Dissolved Oxygen and Pathogens

l.ECfURE20 Gas Transfer and Oxygen ReaeralH>n 385

The most important of these are carbon dioxide and ammonia gas. The former is
Unfortunately, because the other factors are difficult to measure accurately, es-
important in pH caJculations, whereas the latter relates to the problem of ammonia timates obtained in this fashion are usually highly uncertain. However, as in
toxicity.
Churchill's studies on the Tennessee River (recall discussion of Eq. 20.42), there
Mackay and Yeun (1983) have provided a way to extrapolate from commonly
are certain problem settings where the appro¡¡ch works nicely. Box 20.2 outlines
studied gases (such as oxygen and water vapor) to these other gases. For example such a case that sometimes occurs in lakes.
the liquid-film exchange coefficient for a gas can be determined by
0.5 · Deoxygenation with sodium sulfite. As in Owens' studies of British streams, oxy-
gen levels can be artificially lowered by adding sodium sulfite to the stream. This
KI = KI.02 ~ (20.53) method would be particularly auractive for relatively clean systems, where other
( DI.02 )
where KI and DI = exchange coefficient and diffusivity, respectively, and the sub-
script 02 designates the values for oxygen. Similarly the gas-film exchange coeffi-
· effects would be negligible.
Diurnal oxygen swings, In some streams, plant growth can induce diurnal swings
in oxygen leve\. Chapra and Di Toro (1991) have iIIustrated how such data can be
cient can be scaled to that of water vapor by used to obtain reaeration estimates. I will describe this approach when 1discuss
0.67 the impact of photosynthesis on oxygen in Lec. 24.
(20.54) The fourth method for measuring reaeration in the field takes a decidedly dif-
Kg = Kg.H20( DDg
g,H20)
ferent tack. Rather than oxygen, a different volatile substance is injected iuto the
where it has been suggested (Milis et al. 1982) that the gas-film coefficient for water system. Such substances are chosen because (1) they volatilize in an analogous fash-
can be approximated by ion to oxygen, (2) they do not react, and (3) their concentrations can be measured at
a relatively rcasonable cost. Most commonly, radioactive (e.g., krypton-85), hydro-
Kg,H20 = 168Uw (20.55)
carbon (e.g., ethylene, propane, methyl chloride, etc.), and inorganic tracers (e.g.,
where Kg,H20 has units ofmd-I and Uw = wind speed (m s-I). sulfur hexaftuoride) are used. These tracers are usually discharged together with
Schwarzenbach et al. (1993) have correlated diffusion coefficients with molec- a conservative, nonvolatile tracer (tritium, lithium) to determine dispersion (recal!
ular weight. For a temperature of 25°C, the resulting equations are Seco 10.4). r
I
2.7 x 10-4 Either continuous or pulse experiments are normally conducted. For the contin-
,1
DI = MO.71 (20.56) uous case the tracers are injected at a constant rate until steady concentration levels
are attained at two downstream locations. The first-order gas-transfer rate can then
1.55 be estimated by Eq. 10.36,
and (20.57)
Dg = MO.65 1 MJ
k = =--=- In - (20.60)
Finally some investigators have combined relationships such as Eqs. 20.53 t2 - ti M2
to 20.56 to directly calculate the exchange coefficients as a function of molecular where the subscripts I and 2 represent the upstream and the downstream locations,
weight. Using this approach Milis et al. (1982) have come up with the 1's are the average travel times to the two locations, and the M's are the masses
0.25
32 of the tracer. Because the experiment is continuous, the masses should be equal
KI = KI.02 () M
0.25
(20.58) to the flow times concentration at each point. Therefore the equation can be ex-
pressed as (assuming constant ftow)
18
and
Kg = Kg,1I20 () M
(20.59)
k
I
= =--=-In -
t2 - tJ C2
CI
(20.61)

20.4 MEASUREMENT OF REAERATION WITH TRACERS
Aside from formulas, reaeration can be measured directly in the field. Four methods
are commonly used. The first three consist of techniques that back-calculate reaera-
,ion based on a mass balance model and field measurements of oxygen. These are:
: Ste~dy-state oxygen balance. If all the other factors governing an oxygen sag (that
¡s, deoxygenation rate, sediment oxygen demand, etc.) can be determined inde-
A similar approach can be used for the pulse experiment, except that the masses in
Eq. 20.60would be determined by integration as described previously in Lec. 10.
Once the first-order gas-transfer rate is estimated, the result must be extrapolated
to oxygen. One way to do this would be by using the empirically derived correla-
tions such as Eqs. 20.58. Although this can be done, the developers of tracer methods
have directly estimated the scaling as represented by
ka = Rk (20,62)

pendently, the only unknown governing the sag will be the reaeration coefficient. where R = scale factor to relate the tracer exchange rate to reaeration (Table 20.3).
3116 PARTIV Dissolved Oxygen and Pathogens LECTURE20 Gas Transfer and Oxygen Reaeration 387

TABLE 20.3 . 20.4. A flat broad-crested regular step dam with a drop of2 m is situated on a grossly polluted
Factors for scaling gas-tracer exchange rates to stretch of river below a sewage outfall plant. The river is located at an elevation of 2
oxygenreaeration rates
Tracer R
--
Reference
km. Determine the oxygen concentration below the dam if the water upstream has a
concentration of 2 mg L -1 and a temperature of 26°e.

20.S. Derive a relationship of the form of Eq. 20.58 by combining Eq. 20.56 with Eq. 20.53.
Elhylcnc 1.15 Ralhhun el al. (1978)
Pn)p¡1I1c \.3\1 Rathhun el al. (1978) 20.6. A dye study yields the following data for time and concentration of ethylene:
Melhyl chloride 1.4 Wilcox (1984t/. b)
Station I (6 km)
Sulfur hexal1uoride 1.38 Canale el al. (1995)
Kryplon 1.2 Tsivoglou and Wallace (1972)
t(min) O 10 20 30 40 50 60 70 80 90 100

e (p.g L-') O 9 69 81 78 74 71 80 80 80 O

PROBLEMS
Stalion 2 (13.5 km)
20.1. A tankerIruck careensoff the road and dumps 30,000 L of glucose syrup into a small
mounlain lake. The concenlration of the syrup is 100 g-glucose L -l. The lake has the t(min) 9 9.5 10 10.5 11 11.5 12 12.5
following characteristics following the spill: residence time
in the period immediately
c(p.g VI) 0.0 0.0 1.4 2,9 3.4 4,6 3.4 3.0
= 30 d. depth = 5 m. arca = 5 x 1ü4m2.altitude = 11.000 ft. wind speed = 2.235 m
S-l. and temperature = 10°e. Note that the lake is assumed to be completely mixed, t(min) 13 13.5 14 14.5 15 15.5
and has 7.ero BOD ami is at saturation prior lo the spill. Also recognize that the inflow
e (p.g L-1) 2.9 2.3 2.1 1.1 1.1 0.6
river is at saturation.

(a) Compute the grams ofCBOD spilled.

An accompanying tracer study using rhodamine dye and lithium has yielded estimates
(b) Compute how the CBOD and oxygen in the lake change after the spill.
of 0.5 d and 8.3 x 104 cm2 S-I for trave] time and dispersion, respectively. The river
(e) Determine the time of the worst oxygen level in the system.
has the following eharacteristics: Q = 3.7 cms, B = 46 m, and T = 21°C.
20.2. A lake in the United States has a surface area of 5 x 10~m2, a mean depth of 5 m, (a) Estimate the reaeration and compare your results with appropriate reaeration for-
and a residence time 01' 1 wk. How r.lrge a community could discharge to the system mulas.
during Ihe summer (wind specd = 0.89 mps. temperature = 30°C, and elevation = (b) Use Eq. 10.24 to compute the continuous distribution of ethylene at the second
1 km) if the BOD decay rate is 0.1 d 1 and Ihe desired oxygen level is 6 mg L -17 station. Plot the data on Ihe same graph for comparison.
Assume that the sewage has zero dissolved oxygen concentration and does not seltle.
20.7. You continuously discharge sulfur hexanuoride into a stream having constant hydro-
Also express your result as an equivalenl innow concentration,
geometric characteristics. You measure concentrations of 400 and 150 pptr at locations
20.3. Surpose that the 300-mL holtle described in Example 20.1 had an open lop as iIIus- 0.5 and 4 km downstream from the injection point, respectively. Use this data to esti-
tmled in Fig P20.3. Repcat Ihe examplc 1'01' Ihis case. As with Ihe boltle neck. assume mate the reaeration rate if the velocity overthe stretchis 0.2 m s 1.
that Iransfcr Ihrough Ihe thin liquid film takes place by molecu!:lrdiffusion.
20.8, The following data are measured for a polluted lake with a mean depth of 12 m fol-
lowing overturn:
r = 3.115 cm

-L Time (d) O 4 8 12 16 20 24 28 32
611=1mili
---¡--- I "-- DO (mg L-I) 5 6.4 6.8 7.8 8 8.5 85 8.5 8.8
I ....---

Determine the reaeration rate and the oxygen mass-transfer coefficient if the saturation
11 ': 10.26 cm during the sampling period was 9 mg L-l.

20,9. A stream has a velocity of 0.4 mps, a depth of 0.3 m, and a temperature of 23° e.
Estimate (a) the reaeration rate and (b) the comparable rate for carbon dioxide.

20.10, A stream with a rectangular channel has Ihe following characlerislics: S = 0.00 1, B =
20 m, 11= 0.03, Q = 1 cms, and T = 100e. Determine the reaeration rate following
the introduction of a new point source Ihat will discharge an additional 0.5 cms (T =
FIGURE P20.3
25°C) to the channel.
--

388 PARTIV Dissolved Oxygen and Palhogens

LECTURE 21
20.11. The epilimnion of a lake (H = 7 m) has a lemperalure of 16°C and is subjecllo a - -- -.--------- -~
conslant wind nI"2 mps. (a) Estimale the oxygen reaeration rale in d-I. (b) Estimate
the mass tran~fcr velocily for ammonia gas in m d -l. Assume Eq. 20.46 holds.
20.12. A wide river (B = 200 m) has a flow of 800 cms and a mean deplh of 2.7 m. It is Streeter-Phelps: Point Sources
subject lo a wind of 1.5 mps and its mean lemperature is 25°C.
(a) Determine its reaeration rate in d-I. ..
(b) Delermine a mass-Iransfer rale for ammonia in m d-I.
Assume Eq. 20.47 holds.

J4¡cr0rg¡:QVE1!YI~W; 1 4etiv~ tbe classic Streetef.r~eJps equation fOr a single

~ ~b1J

.
~~~eqfcarP.op~c~()PS
~~ljb~te~; tie~tJShOW
'

. . .
iJ)uslrate'i1QW ~1lfCmUst
. .
.
~ t~n
..
.
"'

,. when
.
. . ,... .
.
'
'

.. makir¡g
. .
, an.. o;¡;ygcm
.. . . .
. . .
:
~QP anq provide some ~uidan.c~ on how it can jJe
,, . Jt9W t Canb e useq10P0 dermulti p
}.e point~ourees. an.4
,

.. . l>m!ltlce
. . . .
.

fOra
'

.
.

. . . po1Dt
. .

~ql,ln.~ t !pe6 sfiqw h()}Y,the¡pode! 1Saffected by anaerobic conditions. FinaUy

1~l!ntt ¡!J~ pJ'!~.ft()W fJw~eJ wit}} pie fJlixed~fto,!" V~fsiQl) \hat is appropriate

-
ÍIl sf~~nl!¡'$qF!i ~ esffiartt:s. w~ere !jjspersion iss!~mpClplt.
'R

\ the Slreeler-Phelps model lies logelhenhe Jwopri~!lry_mechani:;ms gúvt:rni~is~

soJved oxygen in a slr~am receiving sewage:~omposilioll.J!f orga..njcmatter amI
oxyg,enreaeralion.As such il provides.an analYlicalfral1leworkfor predicling Ihe ef-
fecI ofbotÍ! póínt and nonpoinl sources of organic waslewaler on slream and t:sluary
dissQlyedQ.~ygen.In Ihis leclure I f(}cus0-"'poinl sources. --
21.1 EXPERlMENT

Now Ihal we have explored ways lo characlerize gas Iransfer, lel us rt:lurn lo the
experimenl Ihal we developed in Lec. 19. Recall Ihal we pl¡¡!;.e<ia.qualll!!y
~~
ic matter in a sealed bottle and Ihen simulaled how oxyge!!.w.a.§.
<)epltele<!,.
!'-I0\.Y
we
can ¡:>erformIhe same simulalion, bUIwilh Ihe bottle open lo Ihe almosphere.
The mass balances for BOD and dissolved oxygen can be wriuen as
, dL
(21.1 )
/V dt = -kdVL
389
390 PARTIV Dissolved Oxygen and Pathogens I.ECTURE21 Streeter-Phelps: Poi"t Sources .391
J .-- /' J,
do r
and
V dt = -kdVL + kaV(os- o)

Now before proceeding with the solution, we malee a transformation that sim-
plifies the oxygen balance. To do this we introduce a new variable,
(21.2)

-
\
D = Os - o J (21.3)
o x
where D is caIled the dissolved oxygen dejicit. Equation 21.3 can be differentiated FIGURE 21.2
In give A point source discharged into a stream reach 01 constant geometry and
hydrology.
dD do
(21.4)
di = -di
21.2 POINT-SOURCE STREETER-PHELPS EQUATION I
.J
Equation~21.3 and 21.4 can be substituted into Eq. 21.2 to give
Now we can move the foregoing theory into a natural water. SpecificaIly we will
dD model a stream with a single pOlnt source of BOD. As depicted in Fig. 21.2 the
(21.5)
V-;¡¡ = kdVL - kaVD reachis at steady-state and is characterized by plug ftow with constanthy.illology and
geometry. This is the simplest manifestation of the classic Streeter-PheJp~ mode\.
'1hus the use of deficit simplifies the differential equation. -Mass balancescanbe written as JL-
f1-~ 4
I 1n.~
Al _ / t I

If L = Loand D = Oat t = O,then Eqs. 21.1 and 21.5 can be solved for
At - o. f
- ¡ O = - U dL
dx- k.L
. (21.8)
L = Loe-kd' (21.6) ¿
--- dD. 1,
\ ~.f

t' 1"
,Ij \... -~
(21.9)
= if
alld D kdLo
ka-kd (e-kd' _ e-k", ) (21.7)
and _ O= -U dx + tL - kaD f/
wherekr =¡kd+ ks. . I¡..,trl r~
As depicted in Fig. 21.1, opening the lid means that oxygen first decreases, but it 11
Ir L = Lo and D = Do at t = O.then these equations can be solved for '"
Ihen recovers as reaeratiorueplenishes the oxygen. - 1..,<'-c-i
¡-L k,
\
.
L _ = Loe-Vx
k"
'
kiLo k, k"
(21.10)

D = Doe--.'u l --x o-x

DO 1>. ~ 'i'
., and
~ J
+-
ka - kr (e u - e u
) (21.11)

~ k
~
These equatioJ1S constitute tt1t;ocJassic "Streeter-Phelps" .fl1odeIaon4 are.!h~ equatioDs
BOO DeflCII. ,.
~ LL D~
behind the oxygen "1'.ag" we discussed in Fig. 19.2.
Aside from being written for a plug~ftowsystem, these equation~ diffeJ frol11.
(a) ~21.6 alld 21,7 in two important ways. First, only part of the removal has an
effect on the defici~ Second, th~ ~ficithas an initial value.

cfr km?" 21.3 DEFICIT BALANCE AT THE DISCHARGE POINT

~
(b)
FIGURE 21.1
LL D~.~ We introduced the dissolved oxygen deficit to stmplify our mathematical modelo Un-
fortunately it also complicates the"deiermination of the boul)dary concentration at the
mIxing pointbecause most waste intlows exhibit a significantly different tempera-
turefrom the-rece-Ivrng water. COilsegüently iheir ox~en~~t~~i()ns wJH differ and,
How (a) closed and (b) open systems are affected by BOD decomposition. hence, a simple deficit balance will be erruneous., This is best iIIustrated by example.

- -----.-.---
- , -.--.

392 PARTIV Dissolved Oxygen and Pathogens I.ECTUHE21 Slreeler-Phelps: Poinl Sources 31U

EXAMPLE 21.1. OXYGEN BALANCE AT A DISCHARGE POINT. A poinl 21.4 MULTIPLE POINT SOURCES
source and a receiving slrearn at sea level have Ihe following characteristics:
Now that we have an understanding 01'how lo perform mass balances al dischargc
Polnt points, we can investigate how multiple point sources can be simulated. In general
Values soun:e Rlver the approach consists of treating the river as a series of uniform reaches linked by
'>.
Flow(m3S-I) 0.463 5.787 boundary conditions. The compulation slarts at Ihe farthest upstream point where
Temperature("C) 28 20 bgundary concentralions are specified. Then the model equations (Eqs. 21.10 and
00 (mgL-I) 2 7.5 I 21.11) are used to calculate conceñtrátions in the downstre_am Jirecí!Qñ.
27 '9.692- r_ 0<'
00 saluration(mg L-1) The calculation is conlinued until a boundary is encounler~<L.Two types 01'
00 deficit (mg L-1) 5.827 1.592 r ...,.... /"\!Jt;..
boundaries occur. The first represents the case where.thesystem paramelers changc.
¡l~/. For example the bottom slope might change, whichwouldlea~to a cha'!,gejn~
Perform mass balances for temperature and oxygen assuming complete mixing. channel's velocity, deplh, reaeration rate, etc. I.nsucJ.Lcasestl:!ecou<;entrationsal
the end of the upstream reach directly serve as the initial concenlrations for the mod--
Solutloo: If we assume that Ihe density and heal capacity of water are relatively con-
stant, a heat balance for temperature can be developed in a similar fashion to a mass el equationswith revised parameters. Second, for sltuations where a point source
balance, ~ } enters the system, mass balances are used lo e.stabli~bthe starting concenlrations 1'01'
Ihe new rcach. The approach is described in Ihe following example.
T, = 5.787(20) + 0.463(28) J;o590(:
o 5.787+ 0.463 . \
EXAMPLE 21.2. MULTIPLE SOURCES. Figure 21.3 shows a river Ihal receives ¡¡
A mass balance for oxygen can be calculaled simply as
sewage trealment planl efIJuenl al kilomeler poi nI '100 (KP 100) and a Iribulary inIJo\\!
t') _ 5.787(7.5) + 0.463(2) _ 7 093 L-I al KP 60. NOle Ihal the channel is lrapezoidal wilh Ihe characlerislics shown. Thc d.:-
00 - 5.787+ 0.463 -. mg oxygenation rate for CBOD is equal to 0.5 d-I al 20°C. For 20 km downstream ¡'rom Ih.:
trealment planl, Ihere is a CBOD setlling removal rale 01'0.25 d-I.
The saturationvalue for 20.59~g L-l. Thereforethe deficitat the mixing Assuming Ihat Ihe O'Connor-Dobbins rcaeralion formula holds and Ihal Ihe strealll
pointcan be delerminedas Do :. '~7.093 = 1.894mg L-l. This is the correct is at sea level, compute Ihe concentralion 01' dissolved oxygen in Ihe system. To sim-
value.
plify Ihe calculalion, we have independently delermined Ihe heal balance~ In addilion
Now lel us see whal happens when we Iry to balance the deficil directly:
we have computed Ihe syslem's hydrogeomelric paramelers and reaclion kinelics. This
informalion is included in Ihe following lables:
D = 5.787(1.592)+ 0.463(5.827)= I 906 L-I 1. -1 ~ J~ "J, . .P.f - l'
o 5.787+ 0.463 . mg
Parameter VoUs KP > 100 KPl KP < 60
which represents a discrepancy of 1.894 - 1.906 = -0.012 mg L-I. Why does this
erroroccur?It resultsfromthe factthat lemperatureandoxygensalurationare relatedin Deplh m 1.19 1.24 1.41
a nonlinearfashionby Eq. 19.32. <.J' (ft) (3.90) (4.07) (4.62)
c-J Area m2 14.71 15.5 18.05
@ Flow 103 sol 5.787 6.250 7.407
103d-1 500,O()() 540,O()() 640,000
It should also be mentioned that the type of errors illustrated in the previous 1\ (cfs) (204) (221) (262)
section can also be incurred for systems where saturation changes longitudinally due Lf) Velocity lOS I 0.393 0.403 0.410
IOdl 33.955 34.819 35.424
to temperature, elevation, or salinity variations. Examples inelude systems with high
(fps) (1.29) (1.32) (1.35)
elevation changes (upland streams) or estuaries. For such systems, de/icit balances
at junctions can introduce slight discrepancies if not handled correctly. Several end-
of-chapter problems illustrate the errors.
fortunately the discrepancy incurred by using de/icit is usually not great be- \ Parameter

rrC)
KP > 100

20
KP 100-80 KPSO-6O KP < 60
~
19.72
cause tf1erelationship of oxygen to temperature, elevaiion, and salinity is not highly '<
o. (mg L-1) (f092 9.1.Ú
noqHnear over the ranges commonly encountered (recall Figs. 19.9 and 19.10). 1.902
k. (d-I) 1.494
Thi:Jjfpre the errors incurred ar¡: usually not significant. In addition, now that com- k,WI) 0.50 0.494
putéij ¡¡reIlbiquitous, dissolved oxygen (rather than deficit) is simulated directly, kJ (d-I) ( 0.50 0.494
-
IU\~!be problems associated with deficit balances are becoming a mool point. , () " ( o v
394 PARTIV Oissolved Oxygen and Palhogens
) LEC11JRE
21 Slreeler-Phelps: Poinl Sources 395

1'\'
_ 0.764 2Qooo
L = 16.667e 34.819 = 1O.75mgL-1
L, =2-mgL-I' / L 0\
'->
o, = 7.5 mg L-I
A, = 5.787 cms ~ ~ '(VP ..,¡
~ ,. In Ihe nexl reach Ihe BOO does nol settle. bUI conlinues lo decay. Al KP 60 Ihe value of
BOO will have droppedlo )
T, = 20.C L.r )
"
""'1 v '11
Q514:

Lw = 20Qmg L-I J
L = 1O.75e -34.,,19(40.000-20.000) = 8 mg1.-1
0w = 2 mg L-I ~)( :x Now another mass balance musl be laken lo accounl for Ihe-inf1owing Iribulary,
Ow = 0.463 cm¡J ~.o '"~ , = 540,000(8)+ 100,000(5)= 7 53 L-I
Tw =(28.~ ( ~ . mg
~ O
L() 640,000

which then decays downslream. Thus by KP O.Ihe value of BOO will have droppedlo
0(;
0.0002 ~
Channel slope
Slde slope 2 fJI L = 7.53e -~(HX),OOO-40.()(X)) = 3.26 mg L-I
Bottom wldlh. m 10
For dissolved oxygen a mass balance is developed at KP 100.
Roughness 0.035
L,=5mgL-1
o, = 9mgL-1 J 00
_ 40,000(2) + 500,000(7.5)
- 540,000 -.
7093
mg
L- J _
... / a, = 1.157cms
T, = 15.C
" which represenls an in¡lial deficil of

Do = 8.987- 7.093= 1.894mg L-1

The deficil in Ihe nexl downslream reach (KP 100 lo KP 80) can be compuled wilh Ihe
Slreeler-Phelps formula

D = 1. 894e -3'4~I~' + 0.514(16.667)

0.764 - 1.842 (
e -3~~129x - e -~~R~91
)
Fór example, al KP 80. Ihe deficil would be
,
= 1894 -JWJ~20.(0) 0.514(16.667)
Channel slope
Slde slope
0.00018
2
D
. e + 0.764 _ 1.842 ( e
-;'¡~1~20.000_
e
-~~R~920'()()()
)= 302
. mg
L-I

Bottom wldlh. m 10 which corresponds lo an oxygen concenlralion of

Roughness 0.035
0= 8.987- 3.02 = 5.97 mg L-1
Okm
.. The defkil in Ihe nexl downslream reach (KP 80 lo KP 60) can be compuled as
= 5 97 _e
FIGURE 21.3
A slrelch 01 slream wilh Iwo poinl sources and changing hydrogeomelric characlerislics.
D
. e
-3'4~'-20.IXJO) 0.514(10.75)
+ 0.514 _ 1.842 e ( -1'¡~~('-20.()()()) -~~~I~(X-20.,XX))
)
Al KP 60 (thal is, 40 km downslream from Ihe poinl source), Ihe delicil would be
Solution: Firsl, the concentration of carbonaceous BOO can be calculaled for each
strelch. A COOD mass balance is dcvcloped al KP 100. D = 5.97 e }'¡~129(40.fX)()-
20.(0))

, = 40.000(200) + 500.000(2) =/ 16 667 L -J

0.514(10.75) _
L() 540.000 ~). mg + 0.514 _ 1.842 (e
-}'¡~129(40.10)-20.()(X))
e
-~~~I~(40.ooo-20.000)
)
which thendecaysdownslreamaccordinglo = 2.70 mg L-1
0.514+0.25
L = 16.667e" J4.KI9--' which corresponds lo an oxygen concenlralion of
Al KP 80 (that ¡s. 20 km downslream from Ihe point source) Ihe value of BOO will have o = 8.987- 2.70 = 6.29mg L-1
droppcd lo
I (1 .¡. \
J I t ~ 'tI" JI
~
-]
396 PARTIV Dissolved Oxygen and Pathogens IHTIIKE 21 Slrcclcr-l'hclps: Pninl Snurccs 391

20 The criticallravel time can be dctcrmincd by dilTcrcnlialing El.!.21.12, sClling

the resultequal to zero,andsolvingfor
!J
{ [1 _
15 kll
Do(kll - kr)
Q te = ~In (21.13)
E kll - kr kr k"Lo ]}
The critical deficit can be determined as
10
............. Saturatlon
k,
- k,,-k,
{ [_
kll
De = k"Lo 1 Do(kll - kr) (21.14)
5 DO kll kr k"Lo ]}
These results indicate that Ihe presence of an initial deficil causes Ihe crilical
deficit to be larger and farther downstream from the discharge. To gain further in-
100 sight, lel us now simplify the analysis by assuming Ihal Do is zero. For this case Eqs.
-20 O 20 40 60 80 21.13and 21.14reduce lo
Dlstance (km) I k"
(21.15)
FIGURE E21.2 t,. = kll _ kr In kr
k,
Now another mass balance is taken to account for the inftowing tributary, k"Lo _ - k,,-k,

+ 100,000(5) = 2.43 rng L -I

and De = _
k" (
,rk
kll

) (21.16)
Lo = 540,000(6.29)
,..n nnn
Now some additional general characteristics emerge. First, notice that the cril-
This value, along with the mass balance for CBOD, serves as Ihe houndary condition for ical time depends only on the removal and the reaeralion rates. To understand this
the next reach. The final results are summarized in Fig. E21.2 and the following table: relationship. note the plot of te versus kr for various values of k" in Fig. 21.4. This
leads to the general concIusion that an increase of either one or both of these param-
eDOD Dellcit DO
eters means that the critical point on the sag will move cIoser to the source.
KP (mg L-I) (mg L-1) (mg L-1)
Second. inspection of Eq. 21.16 shows Ihat the magnitude of the critical deticit
-10 2.000 1.592 1.500 is linearly related to k" and Lo. This makes sense, because each of these p... .uneters
O 16.661 1.894 1.093
10 13.384 2.814 6.113 contributes linearly to Ihe amount of BOO Ihat is available for decomposilion.
20 10.148 3.022 5.965
30 9.214 2.918 5.991 20
40 1.532 2.433 6.110
50 6.553 2.391 6.152
60 5.100 2.260 6.883
10 4.959 2.085 1.059
80 4.314 1.891 1.252
90 3.153 1.696 1.441
100 3.265 1.509 1.635

21.5 ANALYSIS OFTHE STREETER-PHELPS MODEL

In this section we scrutinize the simple Streeter-Phelps model to try to gain some
0.01 0.1 10 100
insight into its behavior. This is extremely important for model calibration. For
simplicity the following analysis is written in terms of travel time. Recall that the kr (d-l)
basic equation is
FIGURE 21.4
D = Doe-k., + kdLo e-k" - e-k.'
( ) (21.12) A plol 01te versus k, lor various values 01k. (boxed numbers).
ka - kr
398 PARTIV Dissolved Oxygen and Pathogens LEcruRE 21 Slreeter-Phelps: Point Sources 399

21.6 CALlBRATION k,
0.2775 -o.277S-kr
6643 = kJ(lI) 1 _ 1.173(0.2775 - kr)
The relationships developed in the previous section can be extremely useful in model . 0.2775, { kr [ k,,( 11) ]}
calibration for the case where the critical time and deficit are known. For example if
the initial BOD and reaeration are predetermined. Eqs 21.15 and 21.16 can be used These Iwo non linear equations can be sol ved simultaneously for kr and kd = 1.159
to calculate values for kcl and kr. and 0.97 O-l. These values, along wilh the olher model paramelers, can then be used in
conjunclion with Ihe Slreeler-Phelps modello compule the tit shown in Fig. E21.3.
EXAMPLE 21.3. PARAMETER ESTIMATION. A point source and a receiving
stream al an allilude of 1524 m (5000 ft) above sea level have the following charac- 10
lerislics:
... - -.....
Polnt ......
Values source Rlver
~
Aow (mJ s.') 0.7078 3.398 DI
5
Temperature ("C) 22.7 14.6 .§.
BOD(mgL-') 40 5 O
Q
DO(mgL ') 2 8

For 32 km below Ihe oischarge point, Ihe stream has a relalively constant width of 61 m o
ano a mcan oeplh of 1.83 m. The following DO measurements are available: O 10 20 30
Distance (km) FIGURE E21.3
I>ownstream Downstream
distance 1>0 distance 1>0
(km) (mgL 1) (km) (mgL 1)
1.61 2.9 16.09 4.8 21.7 ANAEROBIC CONDITION
3.22 1.7 19.31 5.6
4.83 1.5 22.53 6.2 It is possible Ihal Ihe magnitudc of a BOD discharge is so greal that a stream wil\
6.44 1.7 25.75 6.7 be devoid of oxygen.. For ~<¡u<j1 ~asesthe $Ireeler-Phelps model must be modified.
8.05 2.1 28.97 7.0 Gundelach and Caslillo (197m llave developed anice analysisfor such cases.
9.66 2.7 32.19 7.3
For simplicily we again express Iht;;.equalions in terms ol travel tíme. In addition
12.87 3.8
we assume Ihal k,t = kr (that is. there are no senling losses). Por th-íssituation the
Slreeter-Phelpsmodelis
Delermine Ihe lolal removal rale (kr) and the deoxygenation rate (kJ) from Ihe data, L = Loe-k,/, (21.17)
assuming Ihallhe O'Connor-Dobbins l'om1Ula aoequalely predicls reaeration.
Solution: Mass balances allhe outfall can be used to delermine: Lo = 11 mg L -1.00 = and D = Doe-k",+ kclLo (e-k", - ek.,) (21.18)
k" - kcl
6.97 mg L-I, and To = 16°C. The lemperalure can be used in conjunclion with Eqs.
19.32 and 19.39 to compule o... = 8.143 mg VI. This value, in lurn. can be employed The point where the syslem goes anaerobic can be dClerminedby solving Eq.
lo calculale Do = 1.173 mg L -l. Finally Ihe O'Connor-Dobbins formula along wilh a 21.18 for t with D = o.,: This must be done numericaIly. In other words. the solutloÍ1
lemperalure correction can be useo lo calculale ka = 0.2775 d-I.' amounts lo finding the smaIlest root of
The DO oala can be used to delermine oeticil downslream from Ihe discharge. From
this intnrmation it can be eslimalcd Ihallhe crilical deticil 01'6.643 mg L -1 occurs al a
lravellime of 1.517 d. Subsliluling values into Eq. 21.13 gives f(t) = Doe-k", + k:~~cI(e-k't'- c",) - o. (21.19)
0.2775 ~hen Ihis {JOinl is .rcachcd, ()x~_gen depleli~n can!19-lQ!1g~ l1rocecd at Ihe rate of
1.517 = I In 1_ 1.173(0.2775- kr)
0.2775 - kr { kr [ k,,(lI) ]} k"L. Ralher it wiIl be limited ~y Ihe rale at which oxygen passes across th~ air-waler
inlerface via reaeration. Thal is,-
and inlo Eq. 21.14, .

~ !
~ ~~~ \..:." ~ _..~: ~ -'¡.=.' ..,

-r
400 PARTIV Oissolved Oxygen and Palhogens LECfURE21 Slreeler-Phelps: Poinl Sources 40 I

20
dL 't
, ,

d~ = -k~Os\ (21.20) 15
--
Thus once the oxygen is totally depleted, the reaction l:>c:<cQmes
initial condition that L = L¡ at t
zero-order. Using the
= ti, we can calculate the BOO in the stretch from ~CII 10 ~ Anoxic 1
t¡ to (1 as .s
...¡
L = L¡ - kaos(t - ti) (2\.21) 5
Consequently as with all zero-order reactions, the BOO is reduced linear/y.
Finally we must determine where the anaerobic zone ends. At this pÓlnt, o
O 5 10 15 20 25 30 35
kaos = kdLI (2\.22)
(a)
-

-
Combining Eqs. 21.21 and 21.22 yields
- -lkdL¡
ti = t¡ + k51 _ kaos ,;
-=-kaos
(2 \.23)
8 Saturation

--
/ 4
EXAMPLE 21.4. ANAEROBIC CONDlTIONS. Repeal Example 21.3, bul double CII
Ihe BOO concentralion of Ihe trealmenl planl. Thal is, increase Ihe poinl source of BOO .s
Q
Hlgh
BOD
to 80 mg L -l. AIso assume Ihal no settling losses occur (Ihal is, k, = kJ = 0.97 d-I). FIGURE 21.5
O
All olher values are assumed 10be Ihe same as for Example 21.3. Plots 01 (a) BOO
30 35
Solution: The in¡lial CBOO concenlralion musl be reca1culaled lo reflecI Ihe higher and (b) dissolved
x (km)
concenlralion oflhe waslewaler. The result is 17.93 mg L-l. This value, alo:1gwilh Ihe
olher paramelers, can be subsliluled inlo Eq. 21.18 lo give
-4 ~ Uncorrected
tor anoxla
oxygen down-
stream lram a
poinl source where
1.
,"
lit'
_o.97 _o.2775
1I¡.
= I . 173e_0.2775 0.97(17.93) oxygen levels drop ,.
D 3182x +
0.2775- 0.97 ( )
e 3182x- e 3182x (b) lo zara. I Ii'
í~,
",
When we set D = 0" Ihis equalion can be used to eSlimale Ihal the oxygen con- !.j"
~, '.
centralion will reach zero al a Iravel time of 0.589 d, which corresponds to a dislance of
x .~ 1.87 km. At Ihis point the BOO will have decomposed 10a level of 10.13 mg L-l. ¡
21.8 ESTUARV STREETER-PHELPS
This serves as Ihe value of L; lo be used in conjunction wilh Eqs. 21.21 and 2\.23. ¡
The lenglh of Ihe anoxic slrelch can be delermined via Eq. 21.23: For a dispersive system such as an estuary, the Streeter-Phelps equation can be writ-
ten as
= 0589 ~0.97(10.13)-0.2775(8.143) = 404d ':1
¡¡
If . + 0.97 0.2775(8.143) .
o = Ed2L
dx2 _ UdL
dx - k,L (21.24)
which corresponds 10 x = 12.85 km. The BOO al Ihis poinl can be calculaled wilh Eq.
21.21,
d2D dD
and
L = 10.13 - 0.2775(8.143)\4.04 - 0.589) = 2.33 mg L-I O = E dx2 - U dx + k,¡L - kaD (21.25 )

This Ihen provides a boundary condilion lo ca1culale Ihe remainder of Ihe concenlra- The solution for BOO is
lions using Ihe nonnal Slreeler-Phelps model. The final results for bolh BOO and oxygen
are shown in Fig. 21.5. NOlice how for BOO, after an inilial sleep exponenlial drop, Ihe L = Loej"x x::sO (21.26)
degradalion slows 10 a linear decrease in Ihe anoxic region. Afler oxygen moves above
zero, the more rapid exponenlial decrease relurns. '
L = Loeh,x x2:0 (21.27)
For DO, along wilh Ihe simulalion corrected for anoxia, we also display Ihe original and for oxygen deficit is
resull for the low-BOO load as well as a high-BOO-J!!ad simulalion where we make no
correclion for anoxia (Ihal ¡s, we allow Ihe model lo go negalive). Note how Ihe correclion
causes Ihe zone of low oxygen lo expand because of Ihe depressed decomposilion thal
D = ~ W ej"x _ ej,ux x::sO (21.28 )
ka - kr Q ( ar aa )
occurs.
 402 PARTIV Dissolved Oxygen and Pathogens lECTURE 21 Streeter-Phelps: Point Sources 403

D = ~ka - W eh'X _ ehaX

kr Q ( ar aa )
X~O (21.29)
Xc =. (- . )
In ar ~2a
aa 12r
(21.35)
where 12r 12a

w
(21.30) and De =~ W eh'Xc _ eh"xc (21.36)
Lo= arQ k" - kr Q ( ar aa )
4krE Observe that, as with the plug-flow case, the critical deticit is linearly related to
ar = J 1 + U2 (21.31)
the loading and the deoxygenation rateoIn addition the location is solely dependent
on the removal and reaeration rates. As with the plug-flow model, higher removal

aa = J 4kaE
1 + U2
(21.32)
and reaeration rates tend to move the critical deticit cIoser to the outfalI.

jlr u PROBLEMS
(21.33)
hr 2E(l :!::ar)
21.1. A tanker truck careens off the road and dumps 30,000 L of glucose syrup into a small
ha U mounlainlake.The concentration ofthe syrupis 100g-glucoseL-l. Thelalcehasthefol-
(21.34) lowing characteristics in the period immediately following the spill: residence time =
ha = 2E(l :!::aa)
30 d, depth = 5 m, area = 5 X 1Q4102,altitude = 11,000 n, wind speed = 2.235
m s-1, and temperature = 10°C. Note that the lake is assumed to be completely mixed,
The solution is displayed in Fig. 21.6, along with the plug-flow case. Notice how
and has zero BOD and is at saturation prior to the spill. Compute the oxygen level in
dispersion spreads out the sag. the lake during the period following the spill.
In á fashion similar to the derivation of Eqs. 21.15 and 21.16, critica! distances
and deticits can be determined, as in 21,2. Some organic matter is added to a closed reactor so that its initial concentration is 4
mg L -1 of TOe. As the organic matter decays, you know (from a previous study) that

-
it will deoxygenate the water at arate of 0.2 d-I at a temperature of 20°C.
(a) If the present experiment is conducted at 15°C, predict !he concentration of BOD
versus time in the reactor. Plot your results.
(b) Determine the dissolved oxygen and the deficit versus time within the doseó re-
actor. Plot your results.
12
(e) Open the reactor to .he atmosphere and repeat parts (a) and (b). Note that your
Saturatlon laboratory is located at an altitude of 2 km and thereis a gentle breezeof 1 mps
blowing over the reactor during the experiment.

-~ 21.3. You make the following measurements below a sewage plant outfall into a stream at
sea level:

al Traveltime (d) o 2 4 6 8 10 12 16 20
§. 4
Q CBOD (mg L-I) 50 38.9 30.0 23.0 17.7 13.9 11.4 6.7 3.7
DO(mgL-l) 10.0 5.3 2.7 1.6 1.3 1.6 2.2 4.0 5.7
o

" '!jI'
1:',
-10 o 10 20 If the water temperature is 10°C, determine the reaeration and deoxygenation rates.

Dlstance from polnt source 21.4, A fresh stream having the characteristics QI = 7 cms, SI = 1 ppt, 01 = 12 mg L -1,

FIGURE 21.6
and TI = IDoC discharges into a saline stream (Q, = 2 cms, S, = 15 ppt, o, =
3 mg L-1, and T, = 18°C).Determinethe oxygenand the oxygendeficit assum-
Plot 01 dissolved oxygen lor a point source into an estuary. The solution ing complete mixing at the confluence of the two streams. Note that the elevation is
lor the plug-flow case is shown lor comparison. 1.5km.

r
... -.-- --- ..,-- --- -

404 PARTIV Oissolved Oxygen and Pathogens

LECTURE 22
..21.5. A waste source (Qw = l cms, Lw = 25 mg L-I, o" = 2 mg L-I, Tw = 25°C)dis- .~',~-

charges into a stream (Q, = 10 cms, L, = 2 mg L-1, O, = 10 mg L-1, and T, =

15°C¡. Downsln:am, the velocity is 0.3 mps and the depth is 0.3 m. Ca1culate the pro-
files 01"both BOO and oxygen downstream, assuming that the stream is located at 5500 Street~r-Phelps: Distributed Sources
ft elevation and that the BOO decays at arate of 1 d-I. Determine the value and the
location of the maximum deficit.

21.6. Determine the profiles ofBOO and dissolved oxygen for the foIlowing sea-Ievel stream:

Q. = 1cms
T. =2S'c Q,= 1cms
Q,=4cms L. =SOmgL-I T, =20'C
T,=lS'C J
L,=SmgL-1-r-
0.=2mgL-1
~
L, =10mgL-1
o, =SmgL-1
0,=7mgL-1

H(m) 3 S 7
U (mps) 0.5 0.2 0.4

o 30 km 60km

FIGURE P21.6

.1 LE~ 0VERVlEW:1 focus on tbe effect of distril>uted sources of BOD an<J

~ deficit Qn the Streeter-Phe)ps model. The initia) sections describe cases in
r wbic\1
. . tbe distributed sources contribute polIutant mass but ~o not add sig-
nificant ~uantities of ftow. 1follow tbis witb a description of modds of flow-
cQntributmg sources.
L_ ~

To thisjuncture I have focused on point sources. Now let's look at nonpoinl or diffuse
sources. As the name implies, these inputs enter Ihe system in a diffuse manner. For
one-dimensional streams and estuaries this means Ihallhe loadings enter the system
along its length.
In this lecture 1focus on two types of diffuse sources: those that do nol contrib-
ule ftow and those that do. Before doing that I'iI briefty describe how the rates of
diffuse source contributions are parameterized.

22.1 PARAMETERIZATION OF DISTRk TED SOURCES

Recall that in Lec. 9, I showed how a steady-stale mass balance for a JislributeJ
source (Fig. 22.la) could be written for a plug-ftow system with uniform hydrology
~pm~ ~

de
O = - - - ke + Sd (22.1)
dt
where Sd = rate of the distributed source (g m-3 d-1) and t travel time for a
plug-ftow system. If e = O at I = O,then
405
 406 I'ARTIV Dissolved Oxygen and Pathogens
r
1/'
i
I
(/
~

LECfURE22 Streeter-Phelps: Distributed Sources 407

Sd(M L-J T-')

r ,-,. ,-,. ,-,. ,~r '-1'",-,. r,-,., ID
paramelerize such inpuIs wouid be per lenglh of Slream, S:í (M L-1 T-1). Such a
II ,. " 11" ,. ,. I I I I
""""""""" rale could be converted lo Ihe proper volumelric unils for Eq. 22.1 by
"L S:í
Sd = Sd- = - (22.3)
I
o . x=Vt V Al'

(a) where L = 10lallenglh

V = volume
A,. = cross-seclional area of Ihe reach being loaded

l~ oUUu nuun l'

FIGURE 22.1
Thus normalizalion lo slream cross-secl;¡>nalarea would converl Ihe per-Ienglh rale
inlo Ihe proper volumelric unils for Ihe mass balance.
As in Fig. 22.2bolher dislribuled loads enler as an areal source across Ihe bol-
V . ______ (a) Uniform distributed source and (b) resulting
res pon se if the substance decays with first-order
10m 01'Ihe lop surface areas. Some examples incl'!.~~~dilTlenl and almospheric
sources and O¡¡y&<:n.Jluxe~ f!0ll!.r()QI~dplanls: The besl way lo parameleriie such
(b) kinetics. mpuiswould be as a fluxS;, (M L-2 r 1).Such areal unils could be converted lo the
proper units for Ihe mass balance by
Sd
(1 kl) ,A., Sd
c=T -e (22.2) Sd = SrV = H (22.4)
Thlls the SOllltiollstarls al zero and asymplolically approaches a level of Sd/k as
where H = stream depth.
lravellime ¡ncreases (Fig. 22.1b). Finally, as in Fig. 22.2c, some distribuled loads enter in a volumelric fashion.
Now althoughthe proper units for the rate are M L- 3 T-I, we should note An example is Ihe oxygen produced by floating planl§ (phyloplanklon). Because Ihe
lhal such loads enler in Ihree commoll ways. The manner in which the loadings besrwayiO paraitierellZesÚellinpuIs is on a volumelric basis, Ihey would nalurally
be in Ihe proper formal Sd (M L-3 T-I).
clller the system influence how they are parameterized. As in Fig. 22.2a, some dis-
tributed loads enter as a line source along the side of the slream. Examples mighl
hc loads cO~lIIectedwilh bank erosion 01'wilh leaching from landfills.The best way to 22.2 NO-FLOWSOURCES
J'
Let's now foclls on diffuse sources Ihat deliver mass bul nol waler lo rivers and
s;' 1M 1,-' T-'I slreams. Although Ihese are primarily used lo characlerize oxygen, as described firsl,
Ihey can also be applied lo BOO.

[_(-~ , ,..'-:.L..KY
(a) Une load
t 22.2.1 BOD

The primary examples of distributed sources Ihal inlroduce BOO inlo a syslem
s; ,1\1 1,-' T-') wilhoul adding significanl f10ware inleractions with porous media. For example a
polluted bottom sediment can contain very high concentration of dissoived organic
matter in its pore water. If the pore-water BOO is much greater than the slream
dJ~-~_~.~j~~W~b BOO. the diffusion can be characlerized as a zero-order source of mass that does not
(b) Arealload
contribute water. For such cases Ihe mass balance can be written in a fashion similar
to Eq. 22.1,
dL
0= --
Sd(M L-.,.,~I) d t - krL + SL
~
(22.5)

-,.,- -. ,--- where S'- = rateof Ihe800 distributedsource(g m-3 d-I) andkr = rate of BOO
FIGURE 22.2 removal (d-I). Nole the use of kr rather than kd. This is done for completeness,
~.,~.:8 .. e._.,-..'0
r~ o 01J . __.
Ways in which distributed loads enter one-dimensional although in realily a sediment or leaching source would be in dissolved form and not
(e) Volumetrlc load water bodies.
subject lo settling losses. The Solulion,if L = Oal t = O,would be

"""'-.-.-
\.-:'~. ~ ~'''''::'-~-~~ ,;",:.. ~---

408 PARTIV Dissolved Oxygen and Pathogens LECfUKE22 Streeter-Phelps: Distributed Sources 409

(a) D= k"SL (22.10)

k,ka

.......

~ 22.2.2 Dissolved Oxygen

01 L
E
..... The po-flow DO sources and sinks have traditionally been the most commonly em-
~
p¡Oyeddistributed loads. It has been used to simulate the effect of plants and sedi-
ment oxygen demandoThe mass balance can be written as
dD S'
O = - di - kaD - P + R + ¡; (22.11)
Sag where P and R = volumetric rates of plant photosynthesis and respiration, respec-
Saturatlon
.-......-....-.......................... tively (g m-3 d-I)
.......
SB = arcal rate ofsediment oxygen demand (g m-2 d-I)
~ Step Totalresponse D H = depth (m)
01
! The solution, if D = Oat t = O,would be
~

o,-D D= -P+ R + (SBIH) (l - e-ka') (22.12)

ka

Travel time Thus the deficit increases in a step-response fashion and asymptotically approaches
a steady-state levelD, where the net deficit gain is balanced by reaeration losses
FIGURE 22.3 (Fig.22.4),
Plot of the (a) BOO and (b) oxygen responses due to a
distributed BOO loading. D = - P + R + (S;iH) (22.13)
ka
L = SL(l
k,
- e-k,,) (22.6)
22.2.3 Total Streeter-Phelps Model
Thus the BOD increases in a step-response fashion and asymptotically approaches
a steady-state level (Fig. 22.3a) L, where the BOD gain is balanced by removal, We have now developed formulations for handling both point and nonpoint sources
of BOD and oxygen. These can be combined into a comprehensive Streeter-Phelps
- SL
model to handle both types of sources. For a stream stretch with a point source at
L = k, (22.7) its upstream end and diffuse sources along its length, the resulting equations for BOD
A mass balance for oxygen delkit can be written as

O = --(dD
- k D + -kd SL (1 - e-k'' ) (22.8) i
dt , a k' Saturatlon t
¡.
Thus the BOD would act as a forcing function contribution to the deficit. For the --" 'f --,.....
case D = O at t = O this balance can be sol ved for .......
D
~
DI
D = kdSL(l _ e-ka') _ kdSL (e-k" - e-ka') (22.9) !
k,ka k,(ka - k,)
~
This is an interesting solution since it contains both a step and a sag response. Thus o,-D FtGURE 22.4
aSdepicted in Fig. 22.3b. the initia! response is s-shaped due to the sag termoThere-
after the delkit increases in a step-response fashion and asymptotically approaches ,'"
¡ Plot 01 the oxygen response
due to a distributed oxygen
t "
Travel time loading.
a steady-state level,
 410 PARTIV Dissolved Oxygen and Pathogens LEcruRE 22 Streeter-Phelps: Distributed Sources 411

and oxygen deficit are ,~.I e FIGURE 22.5

Plot 01 conéentration versus distance
L = Loe-k" + SI.,(1 - e-k,,) (22.14) Noflow." where nonpoint sources add flow and
kr a dissolved conservative pollutant
to a stream. For this case the in-
Point Oistributed
stream concentration asymptotically
approaches the concentration 01 the
""
diffuse source. In addition we also
D = Doe-k" + k:d_~ r (e-k" - e-k",) + -p + Rk"+ (Sj/H)O - e-k",)
show how the concentration would
Point Point Oistribuled increase exponentially lor the case
deficit BOO deficit where the diffuse flow contributes
x mass but not flow.

+ kdSL(I
krk"
_ e-k",)_ kr(k"
kdSL
- kr)
(e-k"_ e-k",) (22.15)
o=- d(Qc)
dx + dQ
dx cd - kAcc (22.17)
Oistributed BOO
O'Connor(1976)developedboth steady-stateand time-variablesolutionsfor
These equations now pro vide a framework for simulating the steady-state effects of
dissolved solids (k = O). In his analysis O'Connor idealized flow increases as an
multiple-point and diffuse sources in a fashion similar to Seco 21.4.
exponential funclion,
Q = Qoeq'x (22.18)
22.3 DIFFUSE SOURCES WITH FLOW
where Qo = flow al x = O(m] d-I) and q' = exponenlial rate of flow increase
To this point we have investigated the response of a stream to diffuse sources that (m-I). Equalion 22.18 can be substiluled into Eq.22.17 to yield
do not contribute significant flow. Although this has been the standard approach in dc , ,
traditional stream-oxygen modeling, recent concem over nonpoint-source pollution -dx + q c = q Cd (22.19)
has directed attention to distributed sources that contribute flow.
In this section 1'11 first present closed-form solutions for some idealized steady-
By applying the boundary condition c = Coat x = O,O'Connor obtained the so-
lution
J state cases to demonstrate the behavior of pollutants affected by such sources. Then
1'11describe a numerical approach that has more general applicability. c = coe -q' x + Cel(1 - e -q' X) (22.20)
As in Fig. 22.5 this model predicts that stream concentration wil\ asymptotically ap-
proach the nonpoinl concenlralion. Thus, because both mass and flow are introduced
22.3.1 Analytical Solutions at the same rale, the concentration of the conservative substance slabilizes at the
A mass balance for a diffuse source that contributes both flow and mass can be writ- nonpoinl concentration.
ten as (Thomann 1972,Thomann and Mueller 1987) This situalion can be conlrasted wilh Ihe case where the pollutant mass enters in
a( A.c ) a(Qc) aQ an exponenlially increasing fashion, but without an associaled flow increase. For this
~ + - = -c - kA c (2216 ) case, which is also depicted in Fig. 22.5, the conservative coneentralion increases
a, ax ax el e .
exponentially beeause Ihe new mass is not being diluted with new flow.
where Ac = cross-sectional area (m2) Now Ihat we have an idea how eonservative diffuse sources affect streams, we
c = concentration (mg L -1) can turn lo noneonservative substanees. Unfortunately the addition of a reaetion term
t = time (s) eomplicales the analysis beeause we musl now eonsider how the diffuse flowaffeets
Q= flow rate (m] s-l) the stream's hydrogeometrie parameters. In particular we must aeeoUnt for the faet
x = distance (m) that flow additions will impaet both veloeity and eross-sectional area (Fig. 22.6).
Cd = concentration of the diffuse source (mg L-1)
k = first-orderdecayrate (s-l)
(Note that we are using seconds for our time unit to be consistent with subsequent
parts of this section. In actuality day or hour units are more commonly used.) At
steady-state, Eq.22.16becomes
- -- -
FIGURE 22.6
Distributed sources 01 Ilow cause
water in a channel to bolh move
tasler and run deeper.
~ ~ ' __ -1 ". ~ .." ~~..:. '- .-" ~ ~ --

~
412 PARI IV Dissúlved Oxygen and Pathogens LECfURE22 Slreeler-Phelps: Dislribuled Sources 413
f
Early attempts to model diffuse sources did not account for these impacts. In C
fIo
most cases it was assumed that either the area or the velocity remained constant
while the other was allowed to vary (for example Thomann and Mueller 1987). In Constantarea(¡J =0)

fact as we have already seen in Lec. 14, both the vdOi.:ityand area should increase
as diffuse flow is added.
To incorporate such effects into our model we employ a linear function to rep- ~
,
=0.6)
resent the flow increases,
FIGURE 22.7
Q = Qo + qx (22.21) Co
Plot 01 concentration versus
where q = a constant parameterizing the rate of the linear increase (m2 d-I). Sub- Constantveloclty(¡J =1) distance 101'three cases: constanl
stituting Eq. 22.21 into Eq. 22.17 gives area, area and velocity varying
according to Manning's equation,
dc x and constant velocity.
O = -(Qo + qx) dx - (q + kAJc + qCd (22.22)
Because of the decay the ultimate resuh will always be less than the diffuse concen-
I have chosen the linear characterization (in place of the exponential model of tration. However, the solution does not go to zero. In faet it approacht'~ 111;. constant
Eq. 22.18) for two reasons. First, although inflowcan exhibit an exponential increase fraction 01'the diffuse concentration that represents a balance betwecn thc dclivery
over large distances, the increáse can just as well be characterized by a linear rela- (v) and the decay (k) ratcs.
I
tionship. This would particularly be the case over shorter distances that are typical Clearly the correct result for realislic cases lies between these two eXln:mes.To I
when dealing with nonconservative pollutants. Second, the numerical approach de- quantify this intermediate result we assume that the Manning equation provides an
scribed in the next section assumes linearly increasing flow. adequate representation 01'the lOolOentumbalance for steady-tlow in the ehannel, l J
As a prelude to incorporating realistic hydraulics, we first develop two so- C
lutions to Eq. 22.22 that correspond to extreme cases. The first assumes that the Q = -.!!.AcR2I3S;12
n (22.27) tJ~1
cross-sectional area is constant and all distributed flow manifests itself in increased
velocity. For this case the solution is where Co = a constant (= 1.0 for metric and 1.486 for English units) !t~~:t
n = Manning's roughness coefficient "
1,'1'1\11
k+V k+V !<111'11
Uo v V Uo v R = channel's hydraulic radius (10 or ft) = AJ P
C
(
= Co
Uo + vx ) + Cd-
v+k
[
1-
( + vx
Uo ) ]
(22.23) P = wetted perimeter (10 or ft)
Se = slope of the channel's energy grade line (dimensionless)
¡ ..

'::'~'. ~\
'
. .
where For steady flow the last quantity is approximated by the channel slope So' ~ "1' .
Assuming metric units, we can solve Eq. 22.27 for ~'1i
Q
(22.24)
v = A,-o Al' = exQfJ (22.28) /1:
"14
in which AcO= initial area (m2). where{3 = ~ and,forthecase of a wide,shallowrectangularchannel, ,I.J;
At the other extreme the velocity is assumed to be constant and all distributed '
I/B- /1.
3/5

flow manifests itself in increased cross-sectional area. For this case the steady-state
result is ex = (¡s;,) 1.:2.29)

_.Lx V _-1..< Uo whc!c B = channel width. Thus for this simple case, because exis a conSlanl, a
C = coe Uo + Cd-k (I - e Uo' -Uo + vx
(22.25) simple relationship between flow and area is established.
[ )] Note also that the parameter {3specifies the relationship between area and ve-
Both solutions are shown in Fig. 22.7. As can be seen, the constant velocity locity.The extreme cases of constant area and constant velocity correspond to values n
result eventually goes to zero. In contrast the constant-area solution asymptotically of {3ofO and 1, respectively.
approaches a constant concentration We apply the method of characteristics, and the original partial differential equa-
v tion (22.16) can be simplified by expressing it as a coupled pair of ordinary differ-
(22.26) ential equations,
c = Cd v + kAcO
414 I'ARTIV Dissolvcd Oxygen and Pathogcns I.ElTURE22 Streeter-Phelps: Dislributed Sources 415

e Unfortunately a cIosed-form solution 1'01'Eq. 22.33 cannot be obtained. How-

ever, it can be easily integrated numerically. The results are displayed in Fig. 22.7.
As can be seen the solution lies between the constant area and velocity cases.
NOlicethat the solution evenlually approaches zero. Por lhe conservalive case
(k = O) the concenlralion wOllld evenlually approach the diffuse source concentra-
tion as the rate 01'delivery 01'lhe diffllse source would evenlually balance the re-
duction in concentration by dilution. In contrast the presence 01'lhe increasing area
in the reaclion lerm means lhallhe denominalor will continuously increase. Thus 1'01'
lhe nonconservative case. the concentralion would evenlually approach zero.
x
22.3.2 Numerical Method

The analylical approach descrihed prcviollsly has limiled applicabilily. Conse-

-.
.)1Characlerlstic I quenlly a compllter-oricnted nllmerical melhod is helter slliled 1'01'more general
appl ical ions.
The following dcscriplion is similar lo Ihe approach employed in Ihe QUAL2E
model. h is based on dividing Ihe river into reaches wilh uniform geomelry. Each
t ..... reach is fllrlher divided inlo equal-Ienglh complltalional elemenls. These are Ihe
same as Ihe conlrol vollimes descrihed previously in Lec. 11.
FIGURE 22.8 Once Ihis scgmcnlalion scheme is adopled, the Iransporl through Ihe syslem
Graphical depiction 01 the method 01 characleristics'lor a plug-Ilow system where musl be eSlablished. Becallse 01'ollr inleresl in diffuse sOllrceswe use the approach
flow ¡ncreases with dislance. depicted in Fig. 22.9. For Ihis silualion a 10lal diffllse now Q,I is eSlablished COI'
a reach. This now is Ihen divided eqllally al110ngIhe segments. For an II-segment
dx reach, each clemenl would receive Q'¡/IInow. Consequenlly such a scheme is Ihe
u (22.30) discrele fonn 01'Ihe linearly increasing load described in Ihe previous seclion (recall
di
Eq. 22.21). [n facl if Ihe lenglh 01'each segmenl is ~x. Ihen Q,Iand q are relaled by
dc q q (22.31)
-=-kc+-cl--c
di Al' A,. Q,. = Q,¡ (22.35)
11 = q ~x
As depicled in Fig. 22.R,Ihe I¡rslequation describes Ihe solulion's characteristic
where Q,. = increll1enlalnow for each clell1enl(m' di).
lrajcctory. thal is, how time and space are rclatcd via velocity. The second equation
shows how the polllltanl changcs lemporally as ilmoves along Ihe cllaracteristic. In The transporl calclllation is Ihen performed on an elemenl-by-element basis
olher words the lirsl eqllalion represenls Ihe posilion in space and lime 01'an observer slarling al the elcll1entfartheslllpslream. Firsl, simple now balances are cOl11pllled
moving allhe slream's velocilY,and Ihe second equalion indicales Ihe concenlralion Q, Q, Q, Q, Q, Q, Q,
changes Ihal wOllldbe perceived hy Ihc moving observer as a fllnclion 01'lime.
SlIbSlilulingEqs. 22.21 and 22.28 inlo Eqs. 22.30 and 22.31 yields
dx I Q"
(22.32) I I
di ~(Qo + q.d' (3)
T
--dx
dc
- q(c,1 - c) Q" + Qd
ti/ = a(Qo+qx) (3 - kc (22.33)

Applying appropriale inilial condilions (all = 0, x = 0, andQ = Qo),we can

solve Eq. 22.32 for Q"
"(3
FIGURE 22.9
1 Plot 01 a linite-difference

.1 \ (~rq I f Q[~
) ~ Qo1 ;¡
(22.34)
.r
scheme nccounting 101'a
uniformly distributed load.
 "

LECruRE22 Slreeter-Phelps: Distributed Sources 417

416 PARTIV Dissolved Oxygen and Pathogells

for the cell. For example for the first element, Real streams exhibit dispersion (Sec. 14.4.1). Consequently, given a physical
dispersion, Eq.22.42 provides an upper bound on the spatial step,
QI = Qo + Qe (22.36)
2E
Áx =- (22.43)
where Qt = outflow from element l. U
Next, for a wide rectangular channel, the flow can be used to determine the
element volume in a similar fashion to the derivation of Eq. 22.28, If this value is used the backward difference will generate a numerical dispersion
equal to the real dispersion. Unfortunately, because velocity is changing, this means
V = ÁxAc = ÁxaQ{3 (22.37) that each cell would require a different Áx to attain the desired result. However,
Once the flow and volume are determined, the mass balance for a pollutant can because advection dominates dispersion in steady-state streams, using a constant
be developed using backward differences. For example balances for BOO and oxy- Áx will not lead to significant error.
gen can be written as Finally real systems never exhibit uniform hydrogeometric properties for the
distances used in the examples in this lecture. Thus the algorithm outlined above
0= Q¡_IL¡_I - Q¡L¡ + QeLd,¡- k,.¡V¡L¡ (22.38) would usually be implemented in a piecewise fashion. This would be done by divid-
- Q .o. + Q 0 d ' - Si,. ing the system into reaches with constant properties. The solutions for the individual
0= QJ'_I O'_ k d YL. + k Y ( o . - o. ) + p. - R. - ~
I I " e ,1 ,1 " ti," S,I 1 I 1 H; reachescouldthenbe linkedby mass balancesat Iheirboundaries.
(22.39)

Because we are using backward differences, the balances can be solved in sequence, PROBLEMS

22,1. Asmalluncontaminated slream IJows past a landlill for 10 km. At Ihe end 01'the stretch, J'
L¡ = Q¡_IL¡_I + QeLd.¡ (22.40) the stream has a concentration 01'2 ¡.tg L -1 01'a contaminant with a known half-life 01'
Q¡ + k,.¡V¡
'
b ''
5 d. Ifthe stream's velocity is 0.1 m S-I, determine the rate ofthe distributed source in
Q,'_IO¡_I + QeOd¡,. - kd ¡V¡L¡+ ka"¡V¡o.¡ + p¡ - R¡ - (S b ,./H¡) '

O¡ = ---'--- (22.41) mg m(streamlength)-I d-I.

jlf~' . ..
.
Q¡ + ka,¡V¡
22,2. Determine the oxygen concentration for a lO-km stretch 01'stream IJowing over a sludge
'~¡II
The results of applying this model are displayed in Fig. 22.10. For this case we bed that exerts a sediment oxygen demand 01'5 g m-2 d-I. Note that the stream has
omitted photosynthesis, respiration, and benthic (sediment) oxygen demandoIn ad- an oxygen saturatiun of 8 mg L -1, a velocity 01'0.2 m S-I, a width of 50 m, a slope of di!
¡~I
dition we used boundary conditions of Lo = Oand 00 = 10mg L- 1. The distributed 0.0002, and a roughness coefficient 01'0.03. The stream's cross section can be approx-
imated by' a rectangular channel.
BOO and oxygen loads had concentrations of 80 mg L-1 and Omg L-1 , respective-
ly. As can be seen the BOO solution conforms c10selyto the analytical result ob- 22.3. An excess of 2 mg L -1 of oxygen develops in a r¡ver at the end 01'a 20-km section
:'1:,.
tained previously. The resulting oxygen profile has a similar shape to the c1assic00 with a high concentration 01'attached plants. Determine the net photosynthesis rate of
sag. However, because it is the result of distributed sources, the sag is much more these plants in g m-2 d-I. The river has the following characteristics: U = 0.2 m d-I,
drawn out. H = 0.5 m, and ka = 2 d-I. Assume that the river is at saturation when it starts to :1

It should be noted that this solution technique will manifest a level of numerical flow over the plants.

dispersion that is approximately equal to (Eq. 11.40)

Áx Lo = 80 mg 1.-'
Do = O mg 1.-'
En =2 U (22.42)
SOD=2~ : P-R=lgm"'d-':-
, ,
, ,
12 I i i
::
i.
............................................. .!?~~.u!:~~.i!?,!..... O 20 40 60
x(km)
8 BOD U (mps) 0.1 0.15 0.1
!,
al
.s 4 DO
~ H(m)
k,(d-')
k.(d-')
0.8
0.2
0.1
1
0.1
0.1
1
0.1
0.1
FIGURE 22.10 k.(d-') 1 1.2 1.2
Plot 01 concentration versus 0, (mg 1.-') 10 9 8
O distance lor BOD and DO lor FIGURE P22.4
x distributed sources with Ilow.
 .",. t,," " *, rifo",.. 41' 4IJ;L ~ ' ",; ~ ',.0 ~ ~,~: ~
~I.Wa
-~
/' .. ,; .~~~ ~'"... y' .r;" ~'".J 40' o, ~ f" ~~ ~~~::

/ /. ' , '/ , f '.' ./ rl ~ <(;; A~.~"'; ,.,J'ffJl: ¡el;. "'...r4,: V.tV'J~ ""; '1J"'.t1Z...t" ~ 'be
i,.. d. I,~,,~'1 '~'_." ,';/', ":::"<.')ft'l'''.. ~" lI'. , .. '.;"h' ~.-:. .' <!'t~ ,. '';:'''': '..~ crr¿ ..~¡r--~ Nitrogen
1'1 'JI'.'nl.I:.-'j ')f :!:';,!¡ t 'Jf d.,: o'.f':\:.h

B = 20 m S = 0.001 n = 0.035 q = 0.00006 m2 5-1 Ld = O Od = 0,

The government is interested in evalualing the sustainable populalion for Ihe area. This
quanlily is lo bc dclincd as Ihe number of people who can populalc Ihe slretch if Ihe
minimum acceptable oxygen concentralion is 2 mg L -l. Assume Ihe basin is localed
al sea level and Ihal the lemperature (20°C) is unaffected by the addilional tlow and
loadings generated by Ihe populalion. Employ Ihe O'Connor-Dobbins formula as your
reaeration model, and assume Ihat all Ihe BOD and tlow generated by Ihe populalion is
input lo Ihe stream. Furlher assume Ihal Ihe wastetlow added by Ihe populace has zero
dissolved oxygen.
.l
22.6. A slream tlows pasl a landlill which leaches water al arate ofO.OOI m2 S-1 wilh car-
bonaceous BOD and DO concenlralions of 50,()()() mg L -1 and O mg L -1, respeclively.
The landlill runs for 5 km of slrcam Icngth. Determine the BOD and DO of the stream
for 20 km downslrcam frrm thc sourcc.
(a) Assuming that the leachale tlow does not significanlly affecI the slream deplh and
velocily.
(/J) Assuming Ihal Ihe leachate tlow does significantly afrecl Ihe slream deplh and
velocily. LEcruREOvERVIEW: 1describeandthensimulatetheeffectofnitrogén~:S~i1tI1~1
The slream has a reclangular channel. Assume\the following values for olher pertinenl dissolved oxygen. After an overview of nitrogen and watetqüa1itY.c1p~ltf~
paramelers: \ a description of the process whereby nHrogen affects dissolved dxy~e1i;illí}i¿
o, = 10mg L 00 = 10 mg L-I LO = OmgL-1 T = 20°C fication. Tben 1 present two frameworks that have beeti used to sitttlÍlate.lli~
k,¡ = O.I d-I R = 20 m S = 0.0005 11= 0.03 phenomenon: nitrogenous BOD and nitrHication models. Finally Í prov!deit
brief introduction to ammonia toxicity.
\
At the beginning of this book I presented the parable of the "Blind Men and the
Elephant." This story has great relevance to the impact of nitrogen on water quality
in that no less than four specific, but interconnected, water-quality pro9tems are as-
sociated with this nutrient. Although this lecture focuses on dissolved oxygen, 1'11
start by presenting an overview of nitrogen as a pollutant.

23.1 . NITROGEN AND WATER QUALlTV

Figure 23.1 depicts the nitrogen cyde in natural waters. As can be seen the cy-
de affects the water's oxygen level. In addition several other water-quality prob-
lems occur. As described next these problems can be grouped into two categories.
In the first group are nitrification/denitrification and eutrophication. For these prob-
lems nitrogen serves as the cause of the problem rather than as a problem in itself. In
the second group are nitrate ollution and ammonia toxicity. In these cases nitrogen
species are the actual pollutants. Althoug I we can IVI e t e problems in this way, as
described next, all the problems are interconnected.

419

. .- "
~~- ~~:... \ '.'_~~~ --:. ~-:~ ...~_ ---i+- __ ~ ~

420 PARTIV Dissolved Oxygen and Pathogens

IH.TURE 2.1 Nilrogen 421
Ga.'lIm Llquld.'lIm
controlled
-Alr

Water
volatlllzatlon
controlled
volatlllzatlon
Na.. sewage discharge into a shallow stream. For such cases suspended solids will be
low and ammonia will be high. Consequenlly profuse plant growth can occur. In
the late aftemoun such plant photosynlhcsis can deplete carbon dioxide from the
,

water, and for weakly buffered waters it can induce a large increase in pH. The
late aftemoon is also coincidentally the time of day when temperatures are high-
est. Consequently high pH and lemperalure can result in elevaled levcls of un-
Denltrlflcatlon ionized ammonia. ~
1-- -- 1
1 1
I 1 In summary Ihe nitrogen problem is interconnected and multifaceled. First,
I 1 ammonia can cause oxygen depletion via nitrification. If Ihis occurs, one of Ihe
Ammonlflcatlon I Nltrlflcatlon
byproducts is nitrate, which itself is a pollulant. Furlher, depending un lempera-
Na
Nltrlflcatlon .1 N02"I.. _ _ _ ~I NO;-I
Denltrlflcatlon
ture and pH, Ihe ammonia can manifest itself in an un-ionized form that is toxic
to aquatic organisms. Finally bOlhammonia and nitrate are essenlial nutrienls for
Net productlon photosynthesis. Thus Ihey can stimulate excessive planl growth, which constitutes a
Nltrogen tlxlltlon water-quality problem in its own right, and exacerbate some of Ihe other I1roblems.
With this overview as background the remainder of Ihis leeture focu~es on ni-
FIGURE 23.1 trificalion and, to a lesser extent. nitrate pollution and ammonia toxicitlln laler
The nilrogen cycle in nalural walers. The dashed arrows indicale Ihal Ihe denilrilicalionreac. lectllres 1'11describe eutrophication. Although this approach runs the rjsk of mis-
lions lake place under anaerobic condilions. Nole Ihal. allhough il is nol depicled on Ihis
diagram, Ihe nel production 01organic N can have an impacl on a Watli' body's oxygen level. sing the "whole elephant," I have done this because of our presenl emp9asis on dis- I
solved oxygen and because nitrification was the first cuntext in which rbodelers ad-
tp
~ Nitrificationldenitrification. As depicted in Fig. 23.1, ammonia due lo direct load- dressed the
sections, younitrogen cycle inthe
will appreciate natural waters. J hupe that, afler covering ~I relevant
big picture. ,.
ings and lo the decomposition of organic nitrogen is oxidized in a Iwo-slep process
lo form nitrite (N02 -) and nitrate (NO) -). The process consumes oxygen and,
Ihus, can seriously deplete the water body's oxygen levels. Ir conditions go anaer- 23.2 NITRIFICATION
obic the nitrate can be reduced to nitrlli:...andthe nitrite convened to f~e nitro~en ~
~y denitrificatlon. gecause free nitrogen is in gaseous form, this mechanism can In addition to carbonaceous BOD, nitrogen compounds in wastewater also have an -I.r
tIJ
result in the loss of nitrogen to the atmosphere. In addition free nitrogen can be impact on a river's oxygen resources (Fig. 23.2). Sewage nitrogen can be broadly
utilized by cenain nitrogen-fixing algae and bacteria. broken down into organic nitrogen compounds (for example proteins, urea, etc.) and
. Eutrophication. Aside from its other characteristics, nitrogen serves as an essen- ammonia. With time the organic nitrogen compounds are hydrolyzed to crear.' addi- : "ti'
n J.
! nutrient for plant growth. Thus it acts 'as a fenilizer that can overstimulate tional ammonia. AUlolrophicbacleria then assimilate Ihe ammonia and crealc 1I¡lrite
plant growth in \he process called eutrophication. This excess growth can impact (N02 -) and nitrate (NO) -).
waler quality directly (e.g., unsightly scums, clogging of the water course, etc.) or The conversion of ammonia lo nitrate is collectively ealled nitrijicatitm. It can ,]
.~
indirectly by exacerbating other problems (e.g., oxygen, ammonia toxicity, etc.). be represented by a series of reaclions (Gaudy and Gaudy 1980).In the first, bacteria
. Nitrate pollution. As depicted in Fig. 23.1, the ultimate result of the nitrification of the genus Nitrosomolllls conven ammonium ion (NI-Lt+) lu nilrite,
process is nitrate. In sufficiently high concentrations, nitrate in drinking water can
NH~+ + 1.502 --+ 2H+ + H20 + N02- (23.1)
have serious and occasionally fatal effects in infants. The problem can become
especially critical in agricultural regions where nonpoint sources of nitrale from
fenilization supplement high leveis due to'nitrification from point sources. FIGURE 23.2
. Ammonia toxicity. Ammonia exists in two forms in natural waters: ammonium
Sewage Sewage The decomposillu, I 01
ion (NH4+) and ammonia gas (NH). Whereas the former is innocuous at the lev- n¡trogen compounds
els encountered in most natural waters, the un-ionized form is toxic to fish. The below a wastewaler
equilibrium relationship between the two forms is govemed primarily by pH. At trealmenl planl discharge.
Models have been
high pH (and to a lesser extent at high temperatures), ammonia exists principally
developed lo simulate
in the toxic, un-ionized formoFor example at moderate temperatures ("" 20°C) Oxygen Oxygen this process basad on
and pH leveis above 9, upward of 20% of the total ammonia will be in the un- ~--y ' a series 01 first-order
Nltrlflcatlon reactions.
ionized formoSuch conditions can occur in the recovery zone below a secondary
422 PARTIV Dissolved Oxygen and Pathogens
I:f, LECrURE23 Nitrogen 423

In the second, bacteria 01'the genus NitmbClcter convert nilrite to nitrate,

N02 - + 0.502 - N03 - (23.2)
The growth rates 01'the two bacteria are generalIy slower than for heterotrophic ~')..., f;fh'i;.'~
hacteria (d I as contrasted with hr I rates). In addition the conversion from nitrite
to nitrate is I"aster Ihan the conversion 01"ammonium lo nitrile. This renects the fact Recovery
lhal about 3 times Ihe amounl 01"suhslrale is consumed in the second step lo derive
the same amount 01"energy. (a) Cerbon
Thc oxygen consul11ed in the two stages can be computed as I
20

roa = 14
1.5(3~ = 3.43 gO gN (23.3) :s
10 Saturatlon
r(Ji = O.5(3~
14 = 1.14 gO gN (23.4)
O DO
wherc "Otlami ro; represent the amounl 01'oxygen consumed due to the nitrilJcation
01";unmonium and nitrile. respeclivcly. The entire process can be represented as
""tI = "Otl+ "tI; = 4.57 gO gN I (23.5) 10
Saturatlon
where ro" =
amount 01"oxygen consumed per unit mass 01'nilrogen oxidized in the
total process 01' nilrifJcation. It should be noted that some 01' the ammonium will í
5
he used I"orhacterial celI production. As a conscqucnce Gaudy and Gaudy (1980)
E NO,
suggest that lhe oxygen demand aclllalIy approal:hes 4.2 g 01"oxygen per g 01'nitro- NO,
gen oxidized. O
'.-....-.....................-..-
The occurrence 01'nitrification depends on additional I"actors besides the pres-
ence 01"mnmonium. The most important 01"these col"actors are
10
l. The presence 01"adequale numhcrs 01"nitril"ying bacteria. This depends on both the Saturatlon
., - ~-
level 01"trcatment amllhe stream type. 11'Ihe sewage has been subjecI to secondary í
5
treatmenl. lhere is thc possibility lhal some nitrifJcalion will have begun before E
discharge. As I"orslream lype. shalIowcr streams wilh rocky boltoms provide a
suhslralc upon which hacteria can grow more efl"ectivcly.
O
2. Alkaline pH levels (oplimalIy at ahoul 8). This tends lo nculralize the acid thal is +-
produced. -4 O 4 8 12
3. Sufficienj oxygen (grcaler than I lo 2 mg L-1). Travel time (d)

The nel effect 01"lhese cofachlrs is lo inhibit nitrification immediately down-
slream I"rompoint-source effluenK In this region oxygen and pH will be depressed.
In addilion I"ordeeper streams, lhe establishment 01'sufficient numbers of bacteria
is hindered by their relatively slow growth rate and the lack 01"substrate. Thus ni-
trilication and its accol11panyingoxygen demand would generally be exerted farther
downstream than the oxygen demand due to oxidation of organic carbono
As with the decol11positionof carbonaceous malter, nitrilication wilI cause a dis-
solved oxygen sag in the stream. The entire process, as depicted in Fig. 23.3, shows
the colIective effect 01"organic carhon and nitrogen decol11position on a stream's oxy-
gen resources. In Fig. 23.30 lhc carhonaccous BOO is elevalcd at Ihe poi nI where the
sewage enters the stream. It lhen decreascs as the CBOO is decomposed and settles.
The effect of decomposition in tandem with reaeration results in an oxygen sag.
In Fig. 23.3h nitrogen also induces a sag. However, beca use 01'the sequential na-
ture of the nitrificalion reactions. thc sag's minimum occurs I"arther downstream than
FIGURE 23.3
The trends 01 carbon, nitrogen, and oxygen below a wastewater
treatment plant discharge into a river.
for the carbon induced depletion in Fig. 23.3a. In addition the cofactors described
earlier can cause the nitrification sag to lag even farther downstream.
Figure 23.3c shows the total depletion that results from the combined effect of
carbon and nitrogen. Note that the resulting 00 profile is worse than the individual
sags, and the critical 00 level occurs at a distance somewhere between the carbon
and the nilrogen cases.
Early altempts to simulate the impact 01'nitrification used a nitrogenous BOO
(NBOO) to characterize the process. In recent years a more mechanistic approach
that attempted to model organic nitrogen, ammonia, and n¡trate explicitly has been
developed. Both are described next.
" ----- -.-,.- ,

424 PAIITIV Dissolved Oxygen and Pathogens LECTURE23 Nilmgen 425

"

23.3 NIT~OGENOUS BOD MODEL 14

'.,
figure 23.4 shows an oxygen demand curve for sewage that has a significant amount
of organic carbon and nitrogen. Note that a second BOO demand is exened due to
nitritication. Tbis led early modelers to treat the demand as a nitrogenous BOO or
-
!.
10 _................. Saturatlon

01 6 ",
NBOO. Thus, using a direct analogy to C.aOD, mass balances could be written and .5.
solved for NJJOO and oxygen deficit. ¡JI
Thc NBOp could be measured directly from the BOO bottle measurement. ~ l'.,
.2 16 20 , '"
In addition i~could be measured indirectly through knowledge of the nitrification
process. First, it was recognized that the total amount of oxidizable nitrogen was FIGURE 23.5
-6 NBDD simulation.
equal to the sum of organic and ammonia nitrogen. In fact there is an analytical mea-
surement calIed total Kjeldahl nitrogen (TKN) that represents this quantity. Funher,
according to Eq. 23.5, the oxidation would consume 4.57 g of oxygen per g ofTKN.
Thus NBOO could be estimated by Figure 23.5 shows the results of a simulation where Ihe initial condition i \ 50Q¡(,
organic N and 50% ammonium. Oxygen satllration is set at 10 mg L-1, Ihe N OD
= 4.57 TKN
LN (23.6)
decay rale is set al 0.25 d-I , and Ihe reaeralion rale is sel al 0.5 d-I . Ahhollg il is
where LN = NBOO (mg L- t) and TKN = total Kjeldahl nitrogen (mgN L-1). physical!y lInrealistic, we have Jet Ihe oxygen go negalive. Notice Ihal Ihe maxi um ~
For a plug-flow system, mass balances for NBOO and oxygen deficit can be deficil is - 3 mg L-1 and occurs al a Irave! lime of 2.5 d from the oulfal!.
written in a similar fashion as for CBOO. For a point source, if LN = LNOand D = Although Ihe NBOO framework provided a firsl way lo in~rp~rale nilrification
Do at t = O,then these equations can be solved for inlo Ihe Streeter-Phelps framework, it had some serious shorlcomings. In panicular
k. it could nol adequalely simulale the time lags ci6served in me 'BCD1iotue measw=e:-
LN = LNoe-UX (23.7) ments and in natural systems. This can be betler u!lderstooEby scruliñizing ¡he aS:

an~
--k.x
D = Doe u +
k nL o
-ka - kn (
e
k.
--x
U - e
k.
--x
U
) (23.8) the aClualnitrification process:' -
sumplions,that are implicit in Ihe adoplion of NBOO afLa!ladeQ!lalesurrogateTO¡:-
-

where kn = rate ofNBOO oxidation (d-I). Thus an oxygen sag due to nitrification 1. Ir we lump the oxygen demands of organicnitrogen and ammonium inlo a sin-
could be simulated. Because the mass balances are linear, the results for CBOO and gle quanlity (NBOO), the conversion rate of organic N to amrnonillnrN'\s é(-
NJJOO could be added to derive the total deficit due to the combined effect of carbon fectively ignorcd. .Thus when organic niirogen iSábundiml, a time I'!.&,w!.!![!ie
erroneously omitted. ~
and nilfogen oxidation.
Values of kn generalIy range from 0.1 to 0.5 d-I for deeper waters. For shalIower 2. Ir we use a single, first-order NBOO decomposition rale, Ihe two-stcp lransfor- t
streams, values of greater than 1 d-1 are often encountered. mation of ammonium lo nitrite lo nltfaié ISIgriOre~..2ill!íougiitiie:.relaIJvc
rá§iity
of the second step (nilrile to nilrale) mitigates this.shoflcoming, anotbe! lime fag- ~
is omitted.' 111).,
3. Fii!iílIy,and probably mosl imp.,<>rtan.hJhe
NBOD perspeclive does nol clirf>"'~ lij
address the other cofaclors(presence of bacteria, pH, and oxygen) Ihat are re- .~,
quTredror niiñficatioñ tO'occur. - ~~!¡

-~ NBOD
Al! three of Ihese shorlcomings mean Ihal tlirect use of NBOO would result
-' . --'-. - ~
in simulaled oxygen depletion océurring too cIose to Ihe eftluent. Analysls could
~t
~I
':
DI
!. c1fcumventIhis problem in ¡woconlrived"ways.First, ifiey coulOiiSeslower rales for
o
o
VI j....
NBOO
'-- decomposition.
- -
Second, Ibey could anificially add dead time lóihcir mode\$.
. _o, 'o. '
BOlh remedies are inadequate: the first because 1\ involves a distonion Of rea~
. ~_ '
11
CBOD FIGURE 23.4 and the second because, although il might adequately represenf the obser'leijJag,
. . .

Oxygen demand curva for
sewage that has a significant
amount of organic carbon and
TIme nitrogen.
under present condilions, it cannot ad'dresshow Ihelag might change once Ireatmenl _
is implemented. As a consequence ¡he NBOO approach has been discar<fedand, ~t;
as described next, a more mcchanistic characterization of Ihe nitrification process
developed.
 -
LECTURE
23 Nilrogen 427
426 PART IV Dissolved Oxygen and Pathogens

23.4 MODELlNG NITRIFICATION

!.. 12
If we assume first-order kinetics, the nitrification process can be written as a series zC) (a)
of first-order reactions,
dNo
Te I = -k"oNo
(23.9) -E
ti)
Q)
8

dN"
'u
(23.10) Q)
'Tel = ko"No - k"¡N,, Q.
ti) 4

dN¡
e
(23.11) Q)
T( I = k"¡N,, - k¡"N¡ C)
o...
:t::
dN" = k. N. (23.12) Z 16 20
dI 11I'

whcrc the suhscripts 0, 11,;, and 11denote organic, ammonium. nitrite. and nitrate,
respcctivcly. An oxygcn dcficit balance can he wrillen as
dD
-
( ,I + ro¡k¡"N¡ - k"D
= r,,,,k,,¡N,, (23.13) -~ 12

8
(b)
................................................................................

Because Eqs. 23.9 to 23.12are sequenlial. Ihey can be solved analylically using
Ihe techniques elabomted previously in Secs. 5.2 and 5.3 for feedforward reactors
and reaclions. For the case where No = N,~)and N" = N"o al I = O,the so\utions
-
O
e
C)
E
4
1'01'Ihe nilrogen spccies are
(23.14)
N" = N(~II' k",' o
20
(e ko,,'_ e-k",,) (23.15)
N" = N "01' k,,,t + k""N,A)
k-~ k
1Il 00 .4

N¡
k"¡N,,u
-(1'
k tk
'" - e
t
",) Time
k¡" - k,,¡
FIGURE 23.6
k .k N, ) l k,,,' - e' km'
_ e-k",t _ e-k'''t (23.16) Nilrificalion simulalion.
+lIIoao\__
tui - koa (e ki" - k"tI k¡"- k,,¡ )
aboulO mg L 1amI has moved lo a lillle over 4 d Iravcllime bclow the source. Thus
N" = N,~)+ N"u - N(~)I' k""t- N"ue-k"" the use of sequenlial reactions has spread out and delayed Ihe impacl on oxygen.

_ k,,¡
ko"N(~)
- ko" (e k,,,' _ e k,,,t)_ k¡"
k"¡N,,u
- k,,¡ (e-k",t - e-k,,,,) 23.4.1 Nitrificatlon Inhibitlon
(23.17) Although Ihe explicil simulalion of nilrificalion is more realislic Ihan using NBOD.
_ k:,;k""Nou e k:"t- e km' _ e k,,,t- e-k'''t
k,,¡ -k"" ( k;" - ko" k¡" - k,,¡ ) it slill has some deliciencies. In parlic~u:j[doeR-nOI cJll1'iiill:rJbclinútingcofaclor~.
One altempl lo address Ihis sh(J!1col1tingÍl1Volvf'sthe inhibitiou_oLthe.nitrificalj()n
The dcficil can be determined in a similar fashion (see Probo23.3). In addition the
reacliOlld!!.e-IQ <J~pre.ssed
uxygen Ie.vcls.The following factor is multiplied by each
difTerelltialequalions can be inlegraled numerically.
of Ihe nitrificalion mIes k", amI k;,,_(Brown and BarnweIl1987):
Figure 23.6 shows Ihe results 01'a simulalion where Ihe inilial condition is 50%
.(";,,. = 1 - e - k"l/,.o (23.18)
organir N antl 5()"h.ammonium. Oxygen saluration is set al 10 mg L-1 and all rales
arc 0.25 ti 1. wilh (hc cxccptions 01'k¡" (al 0.75 ti 1) ami rcaeralion (al 0.50 di). where k,,¡t, = lirsl.onJer nilrilicalion inhihilion coefficicnt ("" 0.6 L mg 1). As plot-
In contrast wilh Fig. 2:\.5 themaximumtleficilhas beenreducedfrom- 3 mg L 1to led in Fig. 23.7 Ihe factor is close lo I for dissolved oxygen concenlrations grealer
 -

LECTU~E23 Nitrogen 42~

428 PAR'!'IV Di~~ulved Oxygen and Pathogens
representation for decomposition (Eq. 19.2):
1.2 respira(iuo
C6HI206 + 602 · 6C02 + 6H20 (23.19)

0.8 Now that we are beginning to deal wilh nutrients, a more detailed representa-
tion is needed. Such a representation was provided by Redfield et al. (1963), whu
Initr added phosphate and ammonium to develop a more complete stoichiometry for or-
0.4
~anic maUer,
(CIH201)106(NH3)16(H3P04) = CI06H263011ON16PI (23.20)

4 8
10 Using this formula a more comprehensive chemical representation of respiration
2
can be expressed as (Stumm and Morgan 1981)
O (mg L-1)
respir.atiun
FIGURE 23.7 CI06H26301l0NI6P. + 10702+ 14H+
Nitrification inhibition factor. 106C02 + 16NH.¡+ + HPO/- + 108H20
(23.21)
Note Ihat this formula provides a similar result foeIhe oxygen demand per unil mass
than about 3 mg L -l. At lower levels the factor approaches a linear relationship,
of organic carbon as was oblained previously with Eq. 23.19 (recall Eq. 19.18),
which at zero oxygen goes to total inhibition. Thus as oxygen approaches zero, ni-
trification shuts down. 107(32) _1
Figure 23.8 slÍOws the results from Fig. 23.6, along with the same simulation but roc = 106(12) = 2.69 gO gC (23.22)
with the inclusion of the inhibition factor. Note that the latter was determined with a
Although ammonia can originate from a variety of sources (for example fertil-
numerical solution (fourth-order RK). For this case the oxygen bottoms out .at abQut izers), let's limit ourselves to that which originates purely from the decomposition
2 mg L -l. In addition it spreads out the lowest DO and prolongs the recovery. of organic matter. According to Eq. 23.21 Ihe stoichiometric yield of nitcogen from
i the decomposition of a gram of organic carbon should be
i
16(14)
O 176 N C -I
,
23.5 NITRIFICATION AND ORGANIC DECOMPOSITION
am, = 106(12) =. g g (23.23)
In addition to the oxygen yield, it is also interesting to relate nitrification back to the
l.! This result can now be linked wilh Eq. 23.5 lo compule an oxygen demand generaled
organic matter being decomposed. Recall that we have previously used the sugar
,1 per gram of carbon decomposed due to nitrificalion,
glucose, C6H1206, to represent organic matter, and we used the following simple
11
,I rollallC = 4.57 gO gN-1 (0.176 gN gCI) = 0.804 gO gCI (23.24)
Thus the oxygef! cQQgJmed in nitrification is abOUI30% of !heOxy~~ru:Q!lli.\!l!!ed in
14 carDonaceousoxidation of pure organic matter. -
-~Finanylet's look at wastewater.The nitrogen concentration of untreated waste-
10 .1-.- __ __-- -- _-- ___ -- -- --- __ _-- _ ___-- -- ____ ____ __-- -- _- - - - - -~~!~!~~I~I)_- water in the United Stales is approximalely 44 mgN L-l. which breaks down into
..... 17.5 and 26.5 mgN L-1 of organic and ammonia nilrogen. respeclively (Metcal!'and
!J 6 Eddy 1991, Thomann and Mueller 1987). Multiplying 44 mgN L-1 by 4.57 mgO
al With mgN-1 yields an NBOD eSlimale of aboul 200 mgO L-l. Thus Ihe NBOD of raw
E
-- Inhlbitlon sewageshould be roughlyequivalent lo itsCBOD(= 220mgO L-1 as in Table 19.2).
I 2
o Now if sewage were purc organic matter (Eq. 23.20), we would expect thal
I
~ 8 12 16 20 NBOD would be about 30% of CBOD. Several faClorscould explain the discrep-
,
I !u I
j
-~ f Wlthout Tlm, ancy. For example not all the organic matter might be decomposable under the con-
! I! Inhl!2ltlon ditions of the BOD test. Further, the sewage might be enriched in nitcogen from
,\l' -6
~
sources other than organic wastes. Regardless these simple calculations suggest that
1,
(a) NBOD can be significant and comparable to CBOD, and (b) wastewater is not
I -" fI~~f1~23.~
,

simply composed of readily decomposable organic matter.

''\NitI\llQi~9D:,l~,I~tioJ\ wtt~ IJ1hi~iti9nat low di~olve~ oxygen levels.
Hr~ ; }

1', f: TI" .,

'1l~ r
 \ 430 PARTIV Oissolved Oxygen and Pathogens
LECl'URE23 Nitrogen 431

23.6 NITRATE AND AMMONIA TOXICITY . 2729.92 ..

As described at the beginning of this lecture, forros of nitrogen are considered water pK = 0.09018 + ---:¡;- (23.28)
pollutants in their own right. Both nitrate and ammonia fall into this category. where T is in kelvins and pK = -log1o(K)..
To gain insight into the partitioning, we can solve Eq. 23.27 for the ammonium
ion and substitute the result into Eq. 23.25, which in turo can be solved for (see
I
23.6.1 Nitrate Pollution Prob.23.4)
¡ The presence of high levels of nitrate in drinking water appears to be the cause of [NH)] = F,,[NH)]r (23.29)
I
methemoglobinemia or "blue babies." This disease primarily affects infants « 6
I months) but may impact children up to age 6. Levels above 10 mgN L-1 of nitrate
where F" = fraction of total ammonia in un-ionized formo
I are believed to lead to the condition (Salvato 1982). 1
I Because nitrate is the terminal point of the aerobic nitrogen cycle and because F" = I + ([W]/ K) (23.30)
, I
raw sewagc contains about 40 mg L-1 of total nitrogen, the potential exists for nitrate
to become a problem below point sources without nitrogen removal. Further, the Thus we can now see that the fraction un-ionized is a function of the system's
I
presence of nitrogen in agricultural fertilizers means that nonpoint sources could pH and temperature. Although Eq. 23.29 is written in molar terms, it applies directly
I compound the problem. to massconcentrationsexpressedas nitrogensincebothammoniaandammonium
I
I
The Ilitrificationmodels described above provide a starting point for the analy- have the same masses of nitrogen per mole.
II sis of nitrate pollution. However, more advanced frameworks should also consider
other sources and sinks of nitrate beyond nitrification. For example plants use ni- EXAMPLE 23.1. CALCULATION OF UN-IONIZED AMMONIA. Oetennine the
trate as a nutrient source and give off nitrogen when they decay. Further, the process concentralion of un-ionized ammonia al a poinl in a slream Ihal has a total ammonia
of denitrilication can act as a sink of nitrate under anoxic conditions. This process coneenlration of 1 mgN L-1, a pH of 8.5, and a lemperature of 30°C.
I
can occur in the water when oxygen levels drop to zero or near-zero levels. How- Solution: The lemperalure can be used lo determine Ihe equilibrium coefficienl,
ever, even when the water contains oxygen, denitrilication can occur in anoxic sedi- 2729.92
ments and biofilms. We describe such processes when we discuss sediment oxygen pK = 0.09018 + 303.15 = 9.095
demand in Lec. 25.
K = lO-pK= 10-9095= 8.03 X 10-10

The hydrogen ion eoneenlralion can be delermined from Ihe pH,

23.6.2 Ammonia Toxicity
rW J = lO-pH = 10-85 = 3.16 X 10-9
Ammonianitrogencanexistin twostatesin naturalwaters:ammoniumion(NH4+)
and un-ionized ammonia (NH), These values can Ihen be SUbsliluted into Eq. 23.30,

[NH)]r = [NH4+] + [NH3] (23.25) F = 1 = O202

" 1+ [(3.16x 10-9)/(8.03
x IO-IO)J .
where the brackets connote molar concentrations and [NH3]r = total ammonia. At which can be used lo compute
sufficiently high levels (on the order of 0.01 to 0.1 mgN L-1), the un-ionized form
can be toxic to lish. Hence the term "ammonia toxicity" is associated with this effect. NH3 = 0.202( 1) = 0.202 mgN L -1
The amounts of the two forms can be represented by the equilibrium dissociation
reaction
" Figure 23.9 summanzes values for percent un-ionized ammonia for a range of
NH4+ ;::=NH3 + H+ (23.26) temperatures and pH's. As can be seen, systems with high pH and temperature have
the highest percentages.
wherc the ratio of the reactants to the products is specified by an equilibrium coeffi-
cient As was the case with nitrate pollution, the nitrification model developed in Sec,
23.4 provides a starting point for modeling the un-ionized ammonia problem. In the
K = [NH3][W] (23.27) simplest sense the nitrification model provides a means to simulate total ammonia
~
[NH4+] as a function of loads and transport. Then Eq. 23.29 provides a means to determine
The equilibrium coefficient for the reaction is relatcd to temperature by (Emerson et how much of the total ammonia is in un-ionized formo Although this is a simplified
al. 1975) view ofthe phenomenon, it provides a means to develop first-cut predictions to meet
un-ionized ammonia standards in natural waters.
 -'-"._~.~"""'c

432 PARTIV Dissolved Oxygen and Pathogens

LECTURE 24
tII
e
100
-- - - . -- - - - - -....
2
E 1(1
Photosynthesis/Respiration
l'
Jg /
'T
e
:1

~GI 0.1
~
0.01 FIGURE 23.9
o 10 20 30 40 Percent un-ionized ammonia
versus temperature lor various
Temperature rC) levels 01pH.

PROBLEMS
h

~." A point source has the folIo~ing characteristics: ¡"~;~VP!Yt~w; J fQQ{J§ QQbowPRotO$ynthe$is
anqrespiratiqq
rates¡}fj'mea.

! J
Ii ..¡u ~1f~iUl»!I~q9J!lWI1f\Wta~\y~r~,
!VD = 6mgL-1 N. = 6mgL-1 "" Firsh rp~~W~!ntrQ9\!9~orym~}~¡ia1
Qw = 2 cms k.. = 0.05 d-I
~ai = 0.q75 d-I kin = 0.2 d-I [~'~'u»WJ~
~~". .,
é!a~~rJ!
i4~~jtPplpQ~sr~~. Jb~j,~~,¡m<f@J\~ il!gNt Th~DJ 1escfiP~$r~~ 'WP.{!>~b~~
Ii~lU~n9q mt~ S:Jr!t9tlW1~~rew~01 nu~1fi
..'
'

If the receiving water has a ftow of 5 cms and negligible nitrogen, compute the profiles
ofthe nitrogen species below the plant. Note that n = 0.03,S = 0.001,and B = 20
Qt~ ~.I¡NUJ ID9ae~ m~JPQd,
, , ,
"

~
,

- ,j - '

m. Disregard temperature.
23.2. A waste treatment plant discharges into a stream so that at the mixing point the concen-
tration of organic nitrogen and ammonium are 5 mgN L -1 and I mgN L-1, respectively,
and the deficit is 2 mgO L-l. Downstream the river has the folIowing characteristics:
U = 0.2 m d-I, H = 0.5 m, Os= \Omg L-1, and k. = 2 d-I.
(o) Compute the oxygen sag in the stretch given the folIowing decay rates: k""
0.25 d-I, k.i = 0.2 d-I, and kin = 0.6d-I.
(b) Ifthe pH and temperature in the stretch are 8 and 15°C, respectively, compute the
maximum leve! of un-ionized ammonia that occurs.
23.3. Develop a closed-form solution for deficit as a function oftime by integrating Eq. 23.13
with D = Do at t = O.
23.4. Perform the algebraic manipulations described in Seco 23.6.2 to solve for Ihe fraction
un-ionized (Eq. 23.30) as a function of hydrogen ion concentralion and the equilibrium
coefficient. In addition solve for a fraction ionized, F¡, where F. + F¡ = l.
23.5. A smalI detention pond (depth = I m, area = I x lOS m2, and residence time = 2
wk) receives inftow concentrations of 5 and 2 mgN L -1 of organic and ammonium
nitrogen, respectively. Determine the steady-state levels of alI species of nitrogen if
k.. = 0.1 d-I, k.¡ = 0.075d-t, and kin = 0.25d-I.
.1
,\
The photosynthesis and respiration of plants can add and deplete significant quanti-
ties of oxygen from natural waters. As described in the previous lecture, such effects
can be incorporated into the Streeter-Phelps framework as a zero-order distributed
source. For such cases the resulting deficit can be calculated as
= R-P
D = -(1 - e-k",) (24.1)
ka
Such a zero-order characterization was deemed adequate when water-quality mod-
els were primarily used to assess the impact of treating raw sewage. However, as
treatment became more refined and concern for nonpoint sources grew, such charac-
terizations have been discarded.
In a later part of this book I will describe a mechanistic theory that predicts
plant growth and death (and hence its etfect on oxygen) as a function of nutrients,
light, and other environmental factors. For the time being, we focus on quantitat-
ively estimating photosynthesis and respiralion rates and their impact on oxygen.
24.1 FUNDAMENTALS
Deforedescribing measurement methods, we must cover some fundamental infor-
mation. First, 1'11provide a brief description of plant activity in streams. Then I will
describe measurement units and parameterization of photosynthesis and respiration

433
 434 PARTIV Dissolved Oxygen and Pathogens LECTURE
24 PhotosynthesisIRespiration 435

rates. Finally 1'11show how light, the plincipal drtermilling factor of photosynthe- of organic matter in the environment,
sis, is quantified.
6C02 + 6H20~C6H1206 + 602 (24.2)
We then used this representation to determine how much oxygen would be consumed
24.1.1 Stream Plant Activity in the decomposition of a unit mass of organie carbon,
In Lec. 19, I described the environment in a stream below an untreated sewage out-
fall. Recall that for the stretch immediately below the discharge. biological activity 6(32) = 2.67 gO gCI (24.3)
roe = 6( 12)
is primarily heterotrophic. That is, it is dominated by organisms sueh as bacteria that
obtain their energy by consuming organie matter, and in the proeess. deplete oxygen. In the present context we can now see that just as this factor allows us to char-
Plant growth in this lOne is suppressed beeause of a number of factors, including acterize the breakdown of organic matter, it also allows us to determine how much
light extinetion due to turbidity. oxygen is produced when a quantity of organic carbon is created via photosynthesis.
Further downstream. as the stream begins to recover, levels of nutrients such as This one-to-one correspondence between organic matter and oxygen brings up
nitrogen and phosphorus will be high. In addition most particulate solids will have the point that photosynthesis and respiration rates can be expressed in either oxygen
settled and the waterclears. Consequently autotrophic plant growth usually becomes 01'carbon units. The chuice is usually dictated by the probJem context. Early oxy-
dominant. gen modelers adopted the former, whereas those interested in the later problem of
Because photosynthesis creates oxygen and respiration depletes oxygen, the eutrophication tended to choose carbonolf a subscript e is used to designate the car-
plants will have an impaet on the stream's oxygen resources. Beeause photosynthesis bon version, Ihe two expressions can be simply related by using the stoichiometric
is light-dependent, this effeet can have both seasonal and diurnalmanifestations. conversion of Eq. 24.3. Thus, for photosynthesis,
From a seasonal perspective, photosynthesis will tend to dominate during the
P = roePe (24.4)
growing season, whereas respiration and plant decomposition wil\ prevail during
the nongrowing periodoThus from a mean-daily perspeetive, plant activity will tend The second issue that must be addressed relates to how the rates should be nor-
to ameliorate 01'exaeerbate the oxygen problem depending on the time of year. malized. This depends on whether the plants are free-floating 01'fixed. In the former
In the growing season, swings in oxygen can be indueed by diurnallight varia- case a volumetric expression, M L- 3 T-I, is proper, whereas in the latter an areal
tions. Thus stream oxygen levels could be supersaturated during the afternoon and approach, M L- 2 T-1, is preferred. As described in the previous lecture, the two
,everely depleted just before dawn. quantities can be related by depth. Thus for photosynthesis,
i Finally, depending on the depth. two types of plants would tend to grow. For
P'
deeper rivers, floating plants called phytoplankton would dominate. For shallower P= - (24.5)
slreams. where lighl can reach the lower depths, bottom plants would tend to make H
up most of the plant biumass. Thcsc ean indude moted and attaehed macrophytes.
where the prime denotes the areal rateo Note that the above conversions and para m-
;ISwell as mierol1oralgrowth ealled periphyton. eterizations also apply to respiration.
Fixed plants tend tu have a greater impact on stream oxygen resourees than
Iloating plants due to two factors:
l. Because they are usually situated in shallower water, fixed plants will tend to 24.1.3 Light and Photosynthesis
have a greater effeet on oxygen levels. Thal is. for an equal growth 01'respiralion
In most water-quality models the rate of photosynthesis is assumed to be directly
rate, the impaet on a shallower system will be greater. proportional to light energy,
2. Because they are fixed in spaee. they tend to be more eoneentrated longitudinally.
P(t) ex:1(/) (24.6)
In addition, as described next, whether the plants are floating 01'attaehed will
have a bearing on how their rates are parameterized. . where 1(1) = available light (langleys d-I), where Ilangley (ly) is I cal cm-2.
One aspect of this relationship that has relevance to the present discussion is
the light variation over the diurnal cycle. This variation can be idealized by a half-
24.1.2 Units and Parameterization sinusoid function. If we assume that Eq. 24.6 holds, the photosynthesis rate can then
al so be represented by a half-sinusoid (Fig. 24.1),
Recallthat at the beginningof our deseriptionof oxygenmodeling,we introduced
the followingsimplechemicalrepresentationof the produetiunand decomposition P(/) = PIIIsin[w(t - Ir)] 1,.< f < /.,
(24.7)
P(/) =O otherwise

!
¡
 ..,
..~,"
r ~ ,. .
#),
.
' .

...
.

LECfURE24 Pholosynthesis/Respinuioll 437

436 PARTIV Dissolved Oxygep and Pathogens

Thus the average daylight rate is 63.7% of the maximum rate regardless of the photo-
P(t)
periodo The average daily rate is about 32% of the maximum for a l2-hr photoperiod,
with a range of fcom 20 to 40% for S-hr and 16-hr pholoperiods.
P,.
EXAMPLE 24.1. PHOTOSYNTHESIS RATES AND STOICHIOMETRY. r~
O'Connor and Di Toro (1970) present Ihe following data for the Grand River, Michigan, ,~~

p. //on AugusI 30, 1960: ~I

~,
Depth (m) 0.58
P Sunrise (hr) 0700
Pholoperiod 0.5417( 13 hr)
Solar noon (hr) 1330
\ f \ Avemge daily pholosynlhesis rate (mg-oxygen L-, d-') 17.3 J
I I ' t
O tr tn t, Tp i:j
J
t.
Y

fI'p
Determine time of sunset, and the maximum and Ihe average daylighl
rates. Express the average daily rate on an areal basis in carbon unils.
pholosynlhesis
.
Solution: The lime of sunsel can be calculaled from Eq. 24.9 as il,
J'
, ,/
FIGURE 24.1 Ji
The variation in photosynthesis over the diurnal cycle due to a 1s = 21. - 1r = 2(13.5) - 7 = 20
hal'-sinusoid representation 01 light intensity.
Thus Ihe sun seIS al 2000 or 8:00 P.M.
The maximum rale can be calculaled from Eq. 24.10, TU. !
.~ '~i'\

wtlere Pm = maximum rate (g m-3 d-I) , \I~&,

-,' w = angular frequency [= 1r/(fT p)] 'TT
'1
Pm=17.3
t~ = timeof sunrise(d) [ 2(0.5417) ] = 50.17gm-3d-l i..s:
t. = time of sunset (d) 1
and Ihe average daylighl rale from Eq. 24.11,
1 = fractionof day subjectto sunlight(!hephotoperiod)
l' p = daily period (= 1 d, 24 hr, etc., depending on the units of time) ( Po = 50.17~
'TT
= 31.94gm-3d-1
tJote !hat the pbotoperiod is related to other parameters by
t -t Finally Ihe average dailv rale can be con verted lo aceal and cacbon unils by using
1= T p -(24.8)
Eqs. 24.4 and 24.5 lo compule

P ' = !.. H = 17.3 gO m-3 d-I O58 = 3 76 C -2 d -I

i
and solar noon can be calculated as ['1' t
e roe 2.67 gO gC-1 . m . g m
tr + t. (24.9) r:t~¡
tn =2 t ft f
i~
In general, values of the average daily areal pcoduction rate range fmll. :Inuut '
FinalIy in addition to its maximum value, we are often interested in the average f,'i ¡
0.3 to 3 gO m-2 d-I formoderatelypcoductivestreams.Highlypcoductivc.~ ~I.:ms '1: ~J
value of Eq. 24.7. Two such averages can be developed. The average daily rate can
can range fcom 3 to 20 gO m-2 d-I. In the following section we explore sOllleof
\Je calculated by integration as the ways in which such values are oblained.
r\, l
(Tp
J, P(t) dt (24.10)
i.j ,
:J
p=-L- = Pm21 ~t'l
Tp 1r 24.2 MEASUREMENT METHODS ..
~rJ l'
Now that you have some feel for the quantities involved in photosynthesis and respi- '~¡I "
The avera~e 4aylight r~te Cal}be deterplined by ration, 1 will describe techniques tbat are used to measure lhese rates. We first look at ',~,
(p estimates that are made in situ- dircctly in the environment. Then we examine two
J, P(t)dt indirect appcoaches: the first uses biomass estimates in conjunction with a theoretical
Pa = o = Pm-2 (24.11)
model of phytoplankton gcowth; the second uses diurnal oxygen data.
ITp 1r
 438 I'ARTIV Oissolved Oxygen and Palhogens LECTURE
24 PholosynthesislRespiration 439
¡;¡

24.2.1 Llght-Bottle/Dark-Bottle Technique EXAMPLE 24.2. LlGHT BOTT~FJDARK BOTTLE. A light-bottleldark-bottle

experiment conducled over an 8-hr period starting at 9 A.M. yields the following data:
The first approach, calIed the light-bottle/dark-bottle technique, is very similar in
spirit to the BOO bottle measurement described in Lec. 19.That is, a microcosm or o¡¡ = 7mgVl o/¡ = 12mgL-1 OJí = 7mgL-1 0J¡ = 6.5mgL-1
bottle environment is used to simulate the chemical environment in the water body. Nole Ihal sunrise and sunset occur al 6:00 A.M. and 6:00 P.M., respectively. The filtered
The method consists of placing a sample of a natural water in two bottles. One BOO of the waler is 2 mg L -1 and the BOO deoxygenalion rate is 0.05 d -l. Use this
is a clear or "Iight" bottle, where both photosynthesis and respiration can occur.The dala lo compule the slream's pholosynlhesis and respiration rate.
other. which does not allow light penetration and is hence calIed "dark," should ex- Solution: Equalions 24.12 and 24.14 can be used lo compule
hibit only respiration. After determining the oxygen concentration in each bottle, we
cap them and place them in the water. Sume time later we open the bottles and de- n.. .
P nel = (12 - 7) mg L-1 = 15 g m-3 d-I
termine the oxygen concentrations. The difference between the oxygen levels can
then be used to compute the photosynthesis and respiration rates, as described next. and Rtm
. = (7-6.5)mgL-1
" I = 15
. gm -3d-1
Suppose that the experiment is conducted over a time 1,and the initial and final
concentrations in the light bottle are designated 0li and o/f' Because we are dealing These values can then be substiluled into Eq. 24.13 to give
with a reaction thal is assumed to be zero-order, the nel pholosynthesis rate can be Pb = 15 + 1.5 = 16.5 g m-3 d-I
determined simply by
The planl respiration can be determined as
Poel = 01{ -I °li (24.12) Rb = 1.5 - 0.05(2) = 1.4 g m-3 d-I
Next the maximum photosynthesis rate can be delermined by integrating Eq. 24.7
where Poel is defined as between the beginning and Ihe end of the test, tI and t2, and setting the result equal to
-
Pb(t2 ti),
Poel = Pb - RCIII (24.13)

where the subscript b designates that this is the "bottle" rate, and the respiration
rale is subscripled with "cm" to signify that it is a "community" rate which not only
r Pmsin [¡;,. (t - tr)] dt = Pb(t2 - ti)

reflects planl respiration but also other oxygen-consuming reactions such as bacterial The left side can be inlegrated and Ihe resulting expression solved for
respiration.
The respiralion in the dark bottle (in other words, the community respiration) Pm = Pb(t2 - ti) ¡T,. = 20.66g m-3 d-I
I "
'1
" can be delermined in a similar fashion. If the initial and final concentrations in the
1T{cos[¡;,. (tI - tr)] - cos[¡;,. (t2 - tr)]}
dark bottle are designaled 0Jí and oJf and we assume that the incubation time is the
same, Finally Eq. 24.10 can be used to delermine the average daily rate,

Re", - -
_ OJí - OJf
(24.14) P = p}¡1T = 20.662(0.5)
1T
= 6.6gm-3d-1
I
This value can then be used in conjunction with Eq. 24.13 to calculate Pb.
The community respiration can be correcled to compute plant respiration by There are various ways in which the light-bottle/dark-bottle approach can be
determining the filtered BOO, Lf' that will be exerted over the incubation time. implemented. Por shallower systems, where bottom plants dominate, light and dark
This value can be used to calculate
sediment chambers must be used. As depicted in Hg. 24.20, probes and automated
recorders are typically used to generate continuous oxygen readings within the cham-
Rb = Rc",- -LfI (24.15)/ bers. Por deep systems, it is necessary to deploy arrays of bottles (Pig. 24.2b) to de-
termine the variation in photosynthesis due to light extinction with depth. Integration
where the correclion can be measured directly or calculated from the bottle decom- is used to translate the resulting information into a single depth-averaged value.
position rate kl as Lfll = kILf;. where Lfí is Ihe initial filtered BOO in the bottle.
The results are average rates for the bOltleover the period of the study. However,
the photosynthesis rate varies accurding lo the half-sinusoid function (Eq. 24.7). 24.2.2 Biomass Estlmates
Thus further manipulation is required to determine maximum and average photo-
synthesis rates. The folluwing example iIIustrales how Ihis is done.
As 1statedat Ihe beginningof this leclure,a theoryof phytoplanktongrowthwill
be developedin a later lecture.Por the time being,1 presentsomeresults of this

._~, ~~_.",_. ."-_.,. --T--''''' ;0., .___

""1,., . ~....._.
--.----

LEC!UKE24 Photosynthesis/Respiration 441

440 PARTIV Dissolved Oxygen and Pathogens

where 1" = average daylight light intensity (ly di)

1, = light intensity at which plant growth is optimal (range = 250 to 500 Iy
d-I)
ke = a lighl extinction coefficient (m-I)
Note that extinction can be relaled lo other paramcters, such as Secchl-di..1;deplh
(m), by formulas. For example
1.8
,/' (24.20)
ke = SD

Equation 24.17 generales a number ranging between Oand 1, which reflects the I

effect of light on plant growth. Thal is, a valuc lIf O indicates total suppression of
productivity. If it is set to 1, Eq. 24.16 represents the oxygen production at optim'll
light levels, I

.
P, = r""Omax 1.066T-20a (24.21)
(a) Benthlc chambers (b) Vertical array
where the subscript "s" denotes "saturating" light intensity. Thomann and Mueller I
FIGURE 24.2 (1987) have nOled that, for T = 20°C, r"" = 0.125 g mg-I, and Om"x= 2 d -1,
Variations 01 the in situ lighl-bottle/dark-bottle melhod. The benthic chambers Eq. 24.20 reduces lo
In (8) are usad to measure rates lor bottom planls. The vertical array 01 I
bottles In (b) 15 employed lo measure Ihe effecl 01 phytoplanklon in deeper P, = 0.25a (24.22)
walers.
which is an oft-ciled heurislic or "rule of thumb."
theory thal can be employed to estimale pholosynthesis from planl biomass measure- The planl respiralion rale can be estimated by
ments.
If nutrients are not limiting, which is often the case below sewage treatment
R = r""k",1.08T-20a (24:23)

plant outfalls, the daily average plant photosynlhesis rate can be computed by where k,a = respiration rate of the plants (range = 0.05 to 0.25 d-I). Again.
p = roaOmaxl.066T-204>/a (24.16) ThomannandMueller(1987)pointoutthal anothercommonlyusedruleof Ihumb,
wl1ere roa = oxygen generated per unit mass of plant biomass produced (range = R = 0.025a (24.24)
0.1100.3 gO mg-Chla-I)
Omax= maximumplant growthrale for optimallighl conditionsand excess corresponds to Eq. 24.23, with r"" = 0.25 g mg-I and T = 20°C.
nutrients(range= 1.5to 3.0d-I)
EXAMPLE 24.3. BIOMASS ESTIMATE OF PHOTOSYNTHESIS AND RESPI-
T = watertemperatureeC)
a = concentration ofplanl biomass (mg-Chla m-3) RATlON. A stream has a phytoplankton population of about 20 g dry weight m-2. The
dry weight is about 40% carbono Employ this biomass estimate to detennille the aver-
4>/ = attenuation of growth due to light age daily areal production and respiration rates on a dear day in Seplember. Note Ihal
The light auenuation factor for a layer of deplh H is calculated as (complete Ihe average temperalure is 16°C. the photoperiod is about 12 hr, and the mean daily
derivations of the following relationships are presented in Lec. 33) solar radial¡on is 485 Iy d -1 at this time of year. The slream is aboul I 111deep alld
has an extinction coefficient 01'approximalely 0.5 m -'. Use values of /., = 300 Iy d -, ,
-/.. -a, -ao )
'I'/=-e2.718/ ( -e (24.17) r",. = 0.15 g mg-', and GOl.. = 2 do'.
keH
Solution: The mean daylighl light intensily ¡;all be calculaled as

------ where
lo -k,H
al =-e
ls
(24.18) 1" = 485 ~~ = 970 Iy d-'

and ao =-
la
ls
(24.19) The parameter values can be substituted inlo Eq. 24.20 to compute the lighl auenuation I
c/JI = 2.~~;~~)5) (e'1.9S4 - e-3.233) = 0.276 ~r
 442 PAR,. IV DissolveJ Oxygen and Pathogens
LECfURE24 PhotosynthesislRespiration 443

Thc biomass can be converled to organic carboll by 0.4(20) = 8 gC m-2. Because 30

we are dealing with carbon lInils we mllst use ,.,...(ElI. 24.3) in placeof "oain Eq. 24.I6. prlmary productlon
..
P' = 2.67g0gC 1(2d 1)I.06616-10(O.276)8gCm 1 = 9.1 gOm-2d-1 -b 20 ~ ~Ion
:::. 10
C)
24.2.3 The Delta Method E O
Over 40 years ago. OdulII ( 195ó) sllggesled Ihal pholosynthelic produclion could be -10 6 12 18
eSlimated using a mass-balance nllldel in conjunction wilh diurnal oxygen measure-
menls. Over the years a number 01'investigators (rol' eX:lmple O'Connor and Di Toro
1970; '-Iornberger and Kelly 1972; Erúmann 19790, 197%; Schurr and Ruchli 1975,
-
:::.
4 --.:

--
1977. etc.) have extended amI relined Ihe idea. Recenlly Chapra ami Di Toro (1993) C) 2
ha ve presenled a graphical expression 01' Odum's idea Ihat is easy to implement. In E
addilion they ha ve shown how Ihe scheme can be extended lo pro vide an eslimale
01' slream reaeration. Their approach. called Ihe delta met/wd, is the subject 01' Ihe '(3 O
fÚllowing paragraphs. ;:
Q)
The basic approach is appealing hecause it is hased on direel observations and e
hence is superior lo the biomass approach descrihed in Ihe previous seclion. Further -2 solar noon
it is rclalivcly inexpensive and can be applied lo very shallow syslems. Thus it also
has advantages over Ihe light-/dark-hotlle lechnique.
As deserihed hy O'Connor ami Di lbro ( 1970). a mas s balance for dissolved
-
:::.
12
saturation
oxygcn delieil in a slream can he writlen as 10
,7D
--:--
dI =
,7D
U -:--
rI.r - "al) -1- R - P(I) (24.25)
-E
O
C)

8
e
t where D = oxygen deficil (Illg L 1) 6
1 = lillle (d)
U = slream velocily (m di)
O 6 12 18
.r = distance (111)
k" = reaerat ion rale (d 1) t (hr)
R = respiration (Illg Lid 1) FIGURE 24.3
PU) = primary produclion (l11g Lid 1) (a) Planl primary produclion and respiration (mg L-, d-'); (b) dissolved
oxygen deficit (mg L '); and (e) dissolved oxygen concentralion versus time
For silualions where the plants are unitimnly distrihuted liJr a sufficienlly long dis- (hr). Nole that (a) shows the half-sinusoidal photosynthesis function along
lance (> 3Ulk,,). de/icil does nol vary spatially (,7D!t1.r ==O). and Eq. 24.25 silllpli- with the two-term Fourier series approximation (redrawn with permission from
)ies lo ASCE).

tlD
-
,
(1 -+k"f) = R - 1'(1) (24.26) where
41TIf
The planl production rate can be approxilllaled by a half-sinusoid (Eq. 24.7 and
Fig. 24.3a). Notethat li)r the presentderivation.we have assullledIhal time zero
bll = COS(Il1Tf) (1TIf)2 _ (21T1l)2 (24.28)

eorresponds to sunrise. O'Connor and Di Toro ( 1970) proposed that Eq. 24.7 can be The two-term approximalion is superimposed on Fig. 24.30.
exprcssed as a Fourier series. The endogenous respiralion R is assumed lo be constan!. This is the traditional
i
supposition of previous models [e.g., Odum (1956)]. In light of presently existing
)
I
P(I) = PII/

[
2./;' -1- i" I
/1"cos 2~11
r 7/, (1 - ITE ~

J
2 )_ }
(24.27)
informalion (Welzel 1983), it is considered a sound first approximation.
Assuming thal Ihe effecls of initial conditions have decayed to zero, the solution
to 24.26 wilh 24.27 as Ihe forcing function. can be delermined as

L
-*,--'-..~ ; , ;:¡;c,.... ¡ .~"...._ --..-

444 PARTIV Dissolved Oxygen and Pathogens lECI'UREN Photosynthesis/Respiration 445

poral variation of P{t). Finally the lime lag cP between solar noon and Ihe time of
D(t) = 15- Pm i
{ a=1
ba cos
27rn
Jk~ + (21TnITp)2 [ Tp
t
( _ fTp
2 )
_tan-'
21Tn ~
( kaTp )~ }
minimum deficit is strictly a functjon of the reaeration rateoThese observalions can
be used as the basis of a simple method for estimating reaeration, primary production,
(24.29) and respiration in a sequential fashion.
where 15 = average daily deficit (mg L-1),
Reaeration. Because the solution's phase shift is dictated solely by reaeration,
- R-P there must be a uniquefunctionalrelationshiplinkingk" and cP. Unfortunalelyan
D=- (24.30)
ka analytical expression for this relationship is difficult to obtain because it involves
the solution of a transcendental equation. However, a numerical solution is straight-
Examples of the solution are shown in Fig. 24.4. To ease interpretation of this forward. I
figure, 1 have plotted oxygen concentration above saturation rather than deficit. In The numerical solution is shown in Fig. 24.5, where the reaeration rate is ex- I

addition the respiration is set equal to zero. pressed as a function of the time lag cP. Figure 24.5 shows that lower ka's are asso- i
At large values of the reaeration coefficient ka, the solution does not diverge ciated with longer time lags. It also suggests that for time lags greater than about 3 I
much from saturation. AIso the shape of the solution is similar to that of the forcing hr, the reaeration prediclion is particularly sensitive to the photoperiod.
function P(t), with the maximum oxygen concentration (minimum deficit) occur- Note thal Fig. 24.5 is extremely sensitive to dala errors for cases with short 1
ring shortly after solar noon. However, as ka decreases, the solution moves farther time lags (that is, streams with high reaeration). Consequently uncertainty in the I
trom saturation. In addition the shape departs from that of the forcing function and time lag can translate jnto a sizable uncertainty in the reaeration estimate. For such I
becomes more sinusoidal. The maximum concentration occurs later in the aftemoon, cases other methods might be required to obtain a sufficiently accurate estimate of
and a c1earlydefined minimum occurs just after dawn. In effect the small ka retards reaeration. For example the rate can also be determined by direct measurement or
the rate of reaeration transfer, allowing the effect of the previous day to carry over by using predictive equations (Sec. 20.3).
jnto the present day.This effect accounts for the minimum dissolved oxygen concen-
tration (maximum deficit) occurring slightly after dawn. The decay toward equilib- Primary production. As with the reaeration estimate, an analytical expression
rium is slow during the evenings and is not completed at sunrise. Thus the dissolved for ~ as a function of primary production is difficult to obtain. However, a numerical
oxygen is stiU above saturatjon as dawn arrives, and a slight decay occurs until the solution can be easily developed.
source rises to a value sufficient to start the solution increasing again.
Inspection of Eq. 24.29 and Fig. 24.4 leads to some general¡zations regard- 100
ing the mode\. As summarized in Fig. 24.3b, it can be conc1udedtImt the solution
has three fundamental characteristics: mean vertical displacement, vertical range, 50
and horizontal displacement. The mean vertical displacement of the entire curve 15
is govemed by the interplay of reaeration, primary production, and respiration as 20
specified by Eq. 24.30. The vertical range ~ is dictated by reaeration and the tem- 10

5
12
DO I -------------- 1.- . .....t
I '1 ..:2 2
above 8 1
saturation
(mg L-1) 4 0.5

O 0.2

O 6 12 18 24 0.1
O 1 2 3 4 5 6
t (hr) tf>(hr)

FIGURE 24.4 FIGURE 24.5

Dissolved oxygen concentration above saturation versus time as compú'ted by Eq. Plot 01 k. (d-') versus </>(hr) lor range 01 photoperiods (hr). Nole Ihal
24.28; Pm = 25 mg L-, d-', f = 13 hr, R = O mg L-, d-'. Boxed nUT~~rs indicate we have expressed photoperiod in hours lor Ihis plol (redrawn with
diflerent values 01 the reaeralion rate (redrawn wilh permission lrom ASCE). a permission Irom ASCE).

j
441> I'AHr IV lJissolvcd OxygclI alld l'alhogclIs LECTURE24 PhotosynthesislRespiration 447

0.8 TABLE 24.1

0.7
Data for Grand River, Michigan (38 km below the
Lansing sewage treatment plant) on August.10, 1960
0.6 as reported by O'Connor and Di Toro (1970).

0.5 .Quantlty Symbol Value Unlts

6 (a) Measured
¡; 0.4
Depth H 0.58 (1.9) m (ft)
0.3 Flow Q 8.35 (295) m3 s 1(efs)
Water lempcrature T.. 25.9 .C
0.2 Oygen saturation o, 8.13 mgL I
Sunrise 1, 0700 hr
0.1
Photopcriod J 0.54(13) d (hr)
Solarnoon l. 1330 hr
0.2 0.5 2 5 10 20 50 100 (b) Caleulated
k" (d") Reaeration t k" 5.5 d-'
Primary prmluclion
FIGURE 24.6 Maimum 50
P.. mgL 1d-I
Plot 01 ÁlP versus k. (d ') lor range 01 photoperiods (hr). Note that Average P 17.3 mgL 1d-I
we have expressed photoperiod in hours lor this plot (redrawn with Rcspiration R 16 mgL 'd-I
permission 110mASCE).
t Computed with Ihe OTonnor and Dobb;ns (1958) fnrmula.

As shown in Fig. 24.6, the numerical solution indicatef that the ratío dI P is least-squares regression can be used to lit a sinusoid to the data (Chapra and Canale
dependent on the photoperiod and the reaeration rateoAt low reaeration levels (that 1988).The resultingestimatesare75 = 0.,- (} = 8.128- 8.4 = -0.272 rng L -',.:l =
¡s, k" < 1.0 di) the mtio dI P becomes solely dependent on the photoperiod. In 8.4 mg L -', and 4> = 3.93 hr.
contrast, for slreams with high reaeration the photoperiod is less importan!. since the Using Fig. 24.5 wilh f = 13 hr, the value of 4> = 3.93 hr can be employed to
ratío becomes dictated primarily by reaeration. e~timate ka = 2.7 d-I. The reaeration eslimate can then be employed in eonjunetion /
It should be noted that ifl and k" are known, this plot can be used in two ways. Ir with Fig. 24.6 lo delermine I1IP = 0.47. This value can, in turn, be used to compute

diurnal oxygen data are available. Fig. 24.6 can be used to estimate prirnary produc-
tíon. Conversely for a production rate calculated by other rneans (light/dark bottle, 16
cte.), the graph can be employed to estímate d.

Respiration. After obtaining the reaeration and primary production estimates, 12

we can determine the respiration rate by rearranging Eq. 24.30 to give
R= P + k,,15 (24.31) O>
g B
O
EXAMPLE 24.4. THE DELTA METHOD. Table 24.1 summarizes O'Connor and e
Di Toro's (1970) results for a slation on the Grand River, Michigan. Ihal is marked by 4
profuse growth of aqualie planls. The slalion is also sufficiently far downstream from
the Lansing sewage treatmenl planl that Ihe assumption 01' negligible spatial gradients
(iWliJx =o O) holds. Use Ihis dala to implement the delta method.
O
Solution: The diurnal oxygen dala for Ihis case are shown in Fig. 24.7. Visual inspee- O 5 10 15 20 25
tion of the plot suggests a mea n oxygen level of about 8 mg L -1 wilh an amplitude of t(hr)
approximately 4 mg L -l. In addition, Ihe Irend seems to peak between 5:00 P.M. and
6:00 P.M., which amollnls lo a time lag 01'about 4 hr. FIGURE 24.7
Plot 01 dissolved oxygen (mg L ') lor Grand River, on Augusl10, 1960;
Although a visual inspection can be used to estimate the trend parameters. regres-
superimposed curve was generated using Eq. 24.29 with parameter values
sioll techniqllcs provide an altcrnative thal is somewhat les s sllbjcclive. For example estimated in Example 24.4 (redrawn wilh permission Irom ASCE).
(- observalions;_m__1¡lIedcurve).
-~ .1

I
J.
448 PARTIV Dissolved Oxygen and Pathogens LECTURE24 Photosynthesis/Respiration 449 \

p = 17.9 mg L -1 d-I. Finally the primary production and the reaeration rate can be data to compute the stream's photosynthesis (average and maximum) and respiration
rateo
substituted into Eq. 24.31 to yield R = 17.2 mg L -1 d-I.
Our estirilates compare favorably with those obtained by O'Connor and Di Toro (Ta- 24.3. The epilimnion of a lake has a phytoplankton population of 5 ¡.Lg-chlorophyll L -l. The
ble 24.1) using their more complicated model. The reaeration rate manifests the greatest carbon to chlorophyll ratio of the plankton is about 40: l. Use this biomass estimate
discrepancy: 2.7 versus 5.5 d-I. It should be noted that O'Connor and Di Toro used a to determine the average daily areal production and respiration rates on a clear day in
formula (O'Connor and Dobbins 1958) to obtain their rateo However, it can also be seen September. Note that the average temperature is 22°C, the photoperiod is about 12 hr,
that if the phase lag had been estimated as 3.5 hr, we would have obtained a reaeration and the mean daily solar radiation is 500 Iy d-I at this time of year. The epilimnion is
rate of about 5 d-I from Fig. 24.5. about 7 m deep and has a Secchi-disk depth of 3.6 m. Use values of ls = 350 Iy d'l,
roa= 0.15gmg'l,andGmax = 2d-l.
The delta method has three deficiencies: one specific to systems with high reaer- 24.4. Temperature and dissolved oxygen were measured at a station on Boulder Creek (about
8.5 miles below the Boulder WWTP) on September 21 and 22,1987:
ation and two of a more general nature. As shown by Chapra and Di Toro (1991) the
reaeration computation is particularly sensitive at low time lags. Thus there is con- T DO T DO T DO
siderable uncertainty associated with the resulting reaeration estimate. Consequent- Time (oe¡ (mgL-I) Time ("C) (mg L 1) Time ("C) (mgL'I)
ly predictive equations or direct measurements might be necessary for such cases. 11.0 2000 18.5 6.7
0400 12.4 6.0 1200 17.2
A more general difficulty pertains to the serial nature of the parameter evalua- 0500 12.6 6.2 1300 19.2 11.2 2100 17.4 6.8
tions; that is, the caIculation of each parameter depends Olithe previous estimates. 0600 12.6 6.2 1400 20.8 11.2 2200 16.9 5.4
This could rcpresent a problem in any application. However, it could be especially 0700 12.6 7.4 1500 21.5 10.6 2300 16.2 5.1
significant in highly aerated systems where the most uncertain estimate, reaeration, 0800 12,8 8.2 1600 21.6 9.2 2400 15.6 5.4
0900 12.4 8.8 1700 21.3 8.8 2500 14.9 5.2
is the first to be computed. Consequently error in k" would propagate to P and then
1000 12.4 10.8 1800 21.0 8.2 2600 14.0 5.3
R. An alternative approach might be to use nonlinear regression to approximate all 1100 15.6 11.0 1900 19.2 6.9 2700 14.0 5.4
parameters simultaneously.
A further limitation of the aPl1roachrelates to our assumption that photosynthesis
can be adequately represent~.¡ci a half-sinusoid. This assumption precludes appli- The elevation at the station is approximately 5500 feet above sea level. The stream's
cation of the delta method to periods (overcast days) or systems (canyon streams) velocity above!he station was about 0.8 fps and the depth was about 0.75 fl. Use the
delta me!hod to determine the reaeration rate and the maximum and the mean daily
that are shaded. lf such effects can be quantified, some of the other more compli-
cated expressions of the Odum framework can be employed. For example Schurr productivity and respiration at the station. Express !he productivity and respiration
in g m-2 d-I. Note that the photoperiod is approximately 13 hr and solar noon is at
and Ruchti (1975) proposed a way to incorporate cloud cover into the framework. 1:00P.M.
Despite these deficienci.:s. because of its ease of application, the delta method
has utility for water-quality 1I10deling.For cases where physical or economic lim- 24.S. A s\ream has a reaeration rate of I d-I, DO saturation of 10 mg L -1, an average daily
itations preclude adequate measurement of plant activity and reaeration, the delta photosynthesis rate P of 6 mg L -1 d-I, and a respiration rate R of 5 mg L -1 d-I . If the
method provides a quick and inexpensive means to quantify these processes. For sit- photoperiod is 10 hr and dawn is at 7:00 A.M., determine the value and the time of the
uations where acceptable sampling can be conducted, it offers a simple, independent maximum dissolved oxygen concentration.
cj1eck. 24.6. A series of light and dark bottles are suspended in the 8-m-thick epilimnioll of a lake.
The resulting average daily rates of photosynthesis and respiration were calculated as
-,--
PROBlEMS Deplh (m) O 2 4 6 8

Surface area (m2) 1 x 10" 0.95 x 10" 0.9 x 10" 0.83 x 10" 0.7 x 10"
24.1. A stream has the following characteristics: depth = 0.8 m, sunrise = 0600, noon =
Pholosynthesis (g m..3 d-') 1 6 10 8 5
1300, and maximum photosynthesis rate = 10 gC m-2 d-I. Determine the average (()
Respiration (g m'3 do') 0.5 1 1 4
daylight and daily photosynthesis rates in gO m-3 d-I .

24.2. A light-bottle/dark-bottle experiment conducted over a 1O-hr period staning at 8 A.M.

yields the following data:

Oli = 8mgL-1 Ol! = 14mgL-1 Od; = 8mgL-1 Od! = 5.5mgL-1

Note Ihat sunrise and sunset occur at 5:00 A.M. and 7:00 P.M., respectively. The filtered
u~ lb. d,U .. d""~;~ ,Ire'plhm,',,', Ph('" "d re,p""'" m",.
BOO of the water is 4 mg L -1 and the BOD deoxygenation rate is 0.05 d-I. Use this

LECTURE 25
) r
Sediment Oxygen Demand
LECTURE OVERVIEW: t review typical SOD values observed in rtatlÍral. \Vatel'$
and then ()utline mechanisti.c approaches for modeUng thé phenóme"ort.l
first derive a simple model that 1s consiste"t with the classic Sb-eéteH?he1ps
parádigm. Then 1 describe arnore tealistic framework thataccodtit~lor phe-
nornena such as gas forrnation and hence provides a superior approach tbt
sirnulating SoD in natural Waters.
.~
Sedimenl oxygen demand or SOD is due to the oxidation of organic matter in bot-
10msedimenls. These benlhic deposilSor "sJudge beds" derive from several sources.
Waslewaler particulales. as well as olher allochlhonous particulales (leaf litter and
eroded organic-rich soils), can result in sedimenls with high organic conten!.ln addi-
(ion,!n highly produclive environmenls such as eUlrophiclakes, eSluaries,and rivers,
pholosynlhelically produced planl maller can settle and accumulale on Ihe bottom.
Regardless of Ihe source, oxidation of lhe accumulated organic maller will result in
a sediment oxygen demando
As menlioned previously in Lec. 22. Ihe firsl attempls lo model SOD used a
l.ero-order or constant source lerm. Such a characterization was deemed adequate
when waler-qualily models were primarily uséd to assess Ihe impact of treating
raw sewage. However, as Irealment becomes more refined and concern for nonpoint
sources grows, such characterizations become increasingly inadequale. This is due
to the facl that Ihere is no satisfaclory way to decide how the SOD changes following
lreatmen!.
The two mosl common approaches were lo (1) leave the SOD unchanged or (2)
assume linearily and lower SOD in direcl proporlion to the load reduction. Such ar-
hilrary melhods were usually jllslified under the assumplion that a truly mechanistic
lInderslandingof the processwas nolnecessarywhenmaking crudeassessmentsof
lhe effecls of primary and secondary Irealmen!. Although this assumption is probably
450
"
LECTURE
25 Sedimenl Oxygen Demand 451
1$-1
FIGURE 25.1
The "missing link" between
BOD seUling losses and SOD
in classical Streeter-Phelps
theory .
queslionable in ilself, il becomes even more lenllous as models are used lo evaluale
advanced wasle treatmen!.
Such shortcomings arise because Ihe zero-order approach treats SOD as a model
input rather than as a calcuJated variable. This problem is iIIustrated in Fig. 25.1.
Recall that in our earlier discussion of the Streeter-Phelps model, we included a pa-
rameter, ks, to simulate that particulate sewage can be removed by settling. Later,
we expanded Ihe modello inelude Ihe zero-order SOD to simulate the oxygen de-
mand due to the decomposition of the settled sewage in the bottom sediments. As
iIIustrated in Fig. 25.1. there is a missing piece that was left out of the process. That
is, the use of the zero-order SOD neglects the mechanism whereby the sediment
organic matter is converted inlo oxygen demando
The pre nt lecture is designed to show how this missing link can be modeled.
To do this. 1 ill first review dala in order lo delineale the levels that occur in natural
~
waters andthe major factors that intluence SOD. Then I will construct a siJr.ple
"Streeter-Phelps" SOD model which. althollgh it is unrealistic, provides a context for
our subsequent discussions. Finally I describe a more recently developed framework
for calculaling SOD in a mechanistic fashion.
25.1 OBSERVATIONS
SOD is typically measured in three ways: two approaches based on modeling ob-
served oxygen levels and one based on direct measuremen!. In the first model-based
approach, a DO model is developed for the water body and all rates except the SOD
are determined. The SOD can then be estimated by adjusting the SOD rate until the
model predictions match the observed DO levels. Although this calibration method
was widespread in the early years of modeling, it is tlawed because it assumes that
all other model parameters (reaeration, deoxygenation rates, etc.) are known with
confidence. Because such is rarely the case. values obtained wilh this method are no
better Ihan order-of-magnilude estimales.
The second model-based approach is expressly designed for stratified lakes.
For such systems the deeper walers (or hypolimnion) can be idealized as a eIosed
system, and an areal hypolimnetic oxygen demand can be detennined as
 .~ .J~",-"", ,
~ .., '"w.;..: ~.. __,____., ,)'
~
I

452 PARTIV Dissolved Oxygen and Pathogens LEcrURE 25 Sediment Oxygen Demand 453

AHOD = -Hh
02 - 01
(25.1)
The effect of temperature on SOD can be represented by
t2 - ti (25.2)
SÍ¡ = S;l.208T-20
where AHOD = areal hypolimnetic oxygen demand (gO m-2 d-I) wherc SÓ.20 = areal SOD rate at 20°C and 8 = temperature coefficient. Zison et al.
01 and D2 = hypolimnetic oxygen levels (mg L-1) measured at two times, ti
(1978) have reported a range of 1.04 to 1.13 for 8. A value of 1.065 is commonly
and t2 (d), during the stratified period
employed. In addition the SOD for temperatures below 10°C typically declines faster
#h = mean thickness of the hypolimnion (m)
!J¡an indicated by Eq. 25.2. In the range from O to 5°C, it approaches zero.
If it is assumed that the primary cause of oxygen depletion in this layer is d::com-
position of organic matter in or at the surface of the bottom sediments, the oxygen
~epletion over the summer-stratified period provides an estimate of the SOD. That BOX 25.1. ZOD or "Zebra Mussel Oxygen Demand"
iS,we assume that SÍ¡ Si AHOD.
This approach has a number of deficiencies that relate to its underlying assump- You may havenOliced that along with typical environmental settings. Table 25.1 con-
tions. First, it does not distinguish between decomposition in the water column and tains the organism Splwerolim.\'. These organisms are attached, fi' 'liS higher bac-
teria that are communly and erroneously called "sewage fung'h l) pical'" grow
in me sediments. Whereas the former is not usually as important as the laUer, in
in channels dominated by raw sewage, and as noted, their respiralion resuhs in a high
deeper systems the assumption might be violated. Second, although strong thermal oxygen demando
gradients diminish mass transfer between the surface and bouom water layers, some Other OOttom organisms can have a similar effec!. Among the most notorious
transport does occur. It should be noted that corrections can be made to account for are the freshwaler mollusks commonly called zebra mussels (Dreissena pol)tII/Jrpha).
mennocline transport (Chapra and Reckhow 1983). However, since such corrections Originally native to the Caspian SeaIBlack Sea region, they were probably introduced
J¡ave rarely been made, most literature AHOD values usually represent underesti- inlo the Great Lakes in the late 1980s when a ship from Europe released its ballast
mates of me actual depletion. water inlo Lake Sto Clair (Ludyanskiy et al. 1993, Mackie 1991). This freshwaler mol-
The final estimation memod involves direct measurement. This is done by en- lusk attaches to lake and river bottoms, channels, and pipe walls. 1'0 date, the biggest
closing me sediments and some overlying water in a chamber and measuring me problem connected with the organism is clogging at drinking and cooling water intake
dissolved oxygen concentration in me water as a function of time. This approach is pipes. However, recent evidence suggests thal they can also have a significant effect
used in bom me laboratory and the field. As with all microcosm measurements (e.g., on water quality.
the BOD Oflight-bottle/dark-bottle tests), this test has the drawback that encJosing One such case occurs in the Seneca River, N.Y., a tributary of Lake Omario. In
1993 the Upstate Freshwater .titute (U.F.I.) in Syracuse, N.Y., conducted a study of
the water in a chamber makes the system different from me natural environment.
1'ypicaI values of SOD are listed in Table 25.1. In general, values from about 1
to 10 g m-2 d-I are considered indicative of enriched sediments. Also notice that
~
a 16-km section of the r¡ver th t was infested with the mussels. Steve Effter, Clifford
Siegfried, and Ray Canale meas red (Effter and Siegfried 1994) and modeled (Canale
and Efller 1995) the section, and their results document the impact of the or)' lism.
the or~anism Sphaerolitus is incJuded in the table to indicate that benthic organisms They measured an average mussel density oyer the stretchof 6000 im.l1viduals
other than bacteria can process organic matter from me overlying water and exert an m-2. They then independently measured an organism respirationrate 01'0.9 mgO
SOD (see Box 25.1). individual-I d-I. The product of these numbers yielded an oxygen demand, or as
dubbed by Prof. Canale, a "ZOD," of 5.4 gO m-2 d-I. Together with other measured
TABLE25.1 parameters for the section, they used a simple oxygen model of the form of Eq. 22.12
Sedlment oxygendemand values (from Thomann 1972and Rast and to determine that oxygen leveis should drop to aOOul 3.9 mg L-1 at the end of the
Lee 1978) stretch. As depicled in Fig. B25.I, this prediction conforms nicely to measurements.
Beyond that the model predicts that 80% ofthe oxygen demand is due to zebra mussel
S~.2t (g m-1 d-I) respiralion.
Range The zebra mussels also haye a dramatic effect on the nutrients in the syslem. The
Bottom Iypeand locallon Averagevalue
mussels in the reach filter water at arate of 60 m3 S-l. This organism "lIow ralc" is
Sphaerolilus(10 g-dryWIm-2) 7 2 times the lIow rate in the rivcr! As they filter, they take in and consume floating I
Municipalsewagesludge: phytoplanklon. As a consequence, chlorophyll ti levels in the stretch drop from 45 to I
Oulfallyicinily 4 2-10 I
5 J.l.gL-l. At the same time, the mussels excrele and egest soluble phosphorus and am-
Downstreamof oUlfall."aged" 1.5 1-2
Estuarinemud 1.5 1-2 monium,leading to increases of 48 J.l.gP L -1 and 0.4 mgN L -1, respectively. Although J
0.2-1 the phosphorus increase is consislem with model results, the ammonium observations
$aqdyboltOm 0.5
Mineralsoils 0.07 0.05-1 are less thanpredicted.However,this discrepancy can be attributed to nitrification.
Arealliypoliinnelic oxygen demand (AHOD)-Iakes 0.06-2 In summary, aside from its other nuisance impacts, the zebra mussel can also di-
rectly impact water chemistry anl! quality through their metaOOlism and processing o"
 454 "AHTIV Dissolvctl Oxygcn mili Palhogcns LECTURE~ Sedimenl Oxygen Demand 455
1

10 waters. Baity (1938) concluded that this was the case for oxygen levels greater than
2 mg L-l. Several investigators have tried to represent the dependence by a saturat-
ing relationship,
~ .~.~~~~~~i.<?~ .~~.~...
, ,
= -km + Os8
1:1I O
5 S8(0) (25.3)
É.
oQ Model _Data
where S~(o) = oxygen-dependent SOD
O = oxygen concentration of the overlying water (mg L-1)
o t---+- I I I I I k.w = half-saturation value for the dependence (mg L -1)
O 4 8 12 16
Lam et al. (1984) have suggestcd a '.'alue for k.1Oof 1.4 mg L -l. Thomann and
I .r (km) Mueller (1987) fil the data of Fillos and Molof (1972) with a value of k.1O=
I FIGURE 825.1 0.7 mg L -l. This lalter result corresponds to independence at oxygen concentrations
.1 grealer Ihan about 3 mg L- I.
nrganic mattcr. The more gcncrallheme 01"such biological impacls will be claborated
on in grcal dclail whcn wc descrihe cutrophicalion inlhe ncxt parl 01"Ihis book.
25.2 A "NAIVE" STREETER-PHELPS SOD MODEL
'1
Initial efforts lo develop a mechanistic represenlation of SOD took the logical step
Asidc fmm lemperalurc. Ihe Iwo olher factors Ihat afTeclSOD are the organic of simulating Ihe sedimenl compartment in a similar fashion lo the water column.
l'ontcnl oflhe sedimenls and Ihe oxygcn concenlralion of Ihe overlying walers. Baily The segmenlation scheme for such a model is depicted in Fig. 25.3. Notice that the
( 11)38)and Fair el al. ( 11)41)pmvidcd Ihe (irslevidence of how Ihe organic conlenl of subscripls w and 2 will be used for waler and sedimenls, respectively. In addition the
Ihe sedimenls affecled SOD. Their work suggesls a square-rool relalionship helween water concentralions will not be simulaled as state variables. Rather they will serve
the SOD and the sedimenl volatile solids. A similar elTeclis provided hy Fig. 25.2(/, as boundary conditions for the sediments. Finally we will assume that the surface
which correlates SOD with sedilllenl CODo area of the water is equal to the surface area of the sediments and that the tempera-
AHOD also seems lo he rclated lo lake total phosphorlls concenlration hy a tures of Ihe two segments are identical.
square-rool dependency (Fig. 25.2h).Because lakes wilh high phosphorus tend lo We can now develop sediment CBOD and NBOD mass balances using first-
have high productivily. such syslems would be expected to have elevaled sedimenl order kinetics. For CBOD the equation would be
organic-carhon conlen\. Thus Fig. 25.2h seems lo imply Ihe same square-root rela-
tionship helween SUD amI sediment carhon. In summary alllhe dala displayed in (25.4)
Fig. 25.2 lead lo Ihe general conclusion Ihat there is less SOD per organie carbon as
"'2 d:r2 = k.,V".Lw- kd2V2L2
the sediments hecome more enriched.
Oxygen is the other factor that allects son. Clearly if water oxygen concenlra-
tion goes to zero the SOD will cease. Converscly ahove a eertain level. it is usually
assumed Iha! the SOD is independent of the oxygen conccnlralion in the overlying

FIGURE 25.2
8, (a) 2 , (b) . Two seIs 01 observations
~ ~ Ihal indicale a square-
í
0'0 o~ rool relalionship belween
01 4 O N 1 SOD and organic conlenl
WE :z:' 01 sedimenls. (a) SOD
S
«E versus surlace sedimenl
S
COD (Gardiner el al. 1984);
o f- ,. -+ I---t- o 1"--+ -+ +-t---+ t (b) AHOD versus total FIGURE 25.3
O 100 200 O 100 200 300 phosphorus concentration
A "Slreeler-Phelps' model 101'SOD. Nole Ihat Ihe waler
(Rasl and Lee 1978. Chapra concenlralions 01 BOD and delicil are considered as
COD (mgO L-1) TP (pgP L-1)
and Canale 1991). boundary condilions. ralher than slale variables.

k
 ~'.'.,J.'''''',.,w."-'a}'''~. '..~,~. '.~ ..... ~,

456 PARTIV Oissolved Oxygen and Pathogens LECfURE25 Sediment Oxygen Oemand 457

where kd2 = decomposition rate in the sediments (d-I). At steady-state t!1is balance TABLE 25.2
can be solved for SOD in g m' Z d-I caJculated with the "naive"
Streeter-Phelps SOD model
k,Hw Lw (25.5)
L2 = kd2H2
Particulale CDOD
Sellling veJocily(m d -')
where H = layer thickness (m). (mg L-1) 0.1 0.2 0.5
A similarbalanceforNBOOcanbe wriuenas
0.5 0.065 0.13 0.325
d~n ;/ 1 0.13 0.26 0.65
(25.6)
V2T = anoronk,VwLw - kn2V2~ 5 0.65 1.3 3.25
10 1.3 2.6 6.5
wh~re ano = stoichiometric yield of nitrogen from the decomposition of seuling 50 6.5 13 32.5
BOO _ _ __ ¡,t;.- 100 13 26 65
ron = oxygen demand to nitrogen ratio due to nitrification = 4.57 gBOO ~
gN-1 shows a matrix that applies this result with a range of organic seuling veloci-
kn2 = nitrification rate in the sediments (d-I) ties and particulate BOOs that are Iypically encountered in natural walers. Thus our
Note that, according to the stoichiometric relationships from Eq. 23.21, the nitrogen
simple framework would lead us to concJude thal SOOs in nalural walers should
range between aboul 0.05 to 65 g m-2 d-I. The upper end of Ihis range is well be-
yield can be ca1culated as
( j'
yond Ihe typical values in Table 25.1. This overprediclion occurs because Eq. 25.12
_ 16(14) _ _1 specifies a linear relationship between SOO and the downward flux of organics.
ano- 107(32)- 0.0654gN gBOO (25.7) This contradicts Ihe square-rool relationships depicted in Fig. 25.2. Thus we can
At steady-state, Eq. 25.6 can be solved for concJude that our "naive" Streeler-Phelps SOO model is not capable of capturing
the diminished oxygen demand for highly productive systems. At best it provides
ksHw
~n = 0.3-k H Lw (25.8) an upper bound on the process. Now we turo to a more sophisticated framework that
n2 2 is capable of simulating the square-root dependency.
where we have substituted and combined the values for ano and ron. Thus we see
that according to our simple model, the seuling particulate CBOO carriesan addi-
tional 30% of oxygen demand due to nitrification. 25.3 AEROBIC AND ANAEROBIC SEDIMENT DIAGENESIS
finally the oxygen deficit balance can be written as
In a landmark paper, Oi Toro et al. (1990) developed a model of the SOD procesr
dD2 , that mechanistically arrives at the same square-root relationship exhibited by Fig.
k
- sBAs
V2tit = d2V2L2+ kn2V2L2n (25.9) 25.2. Before plunging ioto its mathematics, we would first like to provide a more
realistic description of how organic carbon decomposition (or diagenesis) leads to
where As = area of the sediment water interface (m2). At steady-state thls balance SOO.
can be solved for
,,,¡ As depicted in Fig. 25.4 the sediments are divided into an aerobic (oxidizing)
S8 = l.3ksHwLw (25.10) and an anaerobic (reducing) zone. Within these zones both organic carbon and ni-
lll:an further be simplified by realizing that trogen undergo transformations that ultimalely create SOO. We will descrihe Ihero
separalely.
lc' = ksHwLw = vsLpw (25.11)
where lc' = downward flux of organic carbon expressed as oxygen equivalents
(gO m-2 d-1)
Vs = seuling velocity of particulate organics (m d-I)
Lpw = particulate BOO concentration in the water (mg L-1)
Carbono Particulate organic matter (POM) is delivered to the sedimenls by set-
tling. Wilhin the anaerobic sedimenls, .Iheorganic carbon decomposes to yield dis-
solved methane,
l
CH20 --> !C02 + !CH4 (25.13)
Substituting this re1ationshipinto Eq. 25.10 gives
S8 = l.3lc' (25.12) where CH20 is a simplified representation of organic matter. The methane diffuses
I I
upward to the aerobic zone where it is oxidizcd, .
Consequently the model yields the rather simple result that the steady-state SOO
should l>eequal to about 130% of the downward flux of ultimate BOO. Table 25.2 !CH4 + O2 --> !C02 + H20 (25.14)
 458 PARTIV Dissolved Oxygen and Pathogens LECI1JRE2S Sediment Oxygen Demand 459

an NSOD. Ammonium that is not converted to nitrate diffuses back into the water
- by diffusion.
The oxidation of ammonium in the aerobic layer can be represented by
e
E NH3 + 202 HN03 + H20 (25.15)
::1 ~
'O
u
.. This reaction consumes 4.57 gO gN-1 , which together with the nitrogen yield from
S
organic matter decomposition (0.0654 gN gO-I) leads to the nitrogenous oxygen
~ demand caIculated previously in Eq. 25.8.
Now, in contrast to the simple model described in the previous section, nitrilica-
tion is not the last step in the process. In fact much of the nitrate is denitrified to form
Aerobic zOQe nitrogen gas. If it is assumed that the carbon source for denitrilication is methane,
.. this reaction can be represented as
~
.!!
e Anaerobic zone iCH4 + HN03 ~ ~C02+ !N2+ ~H20 (25.16)
l' ~
~
The entire process can be compressed into a single reaction by combining Eqs. 25.14
to 25.16 to give
CD
>
ti
el: Productlono.
NH4 and CH. I ~ NH3 + ~02 !N2 + ~H20 (25.i7)
( I ~

Thus the <}xidation of the nitrogenous matter yields

FIGURE 25.4
An overview 01 the mechanistic SOD Iramework (redrawn Irom Di Toro et al.
1990).
1 silmption would be much lower than the 4.57 gO gN-1 usually expected from the
In the process an SOD is generated. Any residual methane that is not oxi.dizedin
the aerobic layer is diffused back into the water where additional oxidation can take
place.
A problem with modeling this scheme is that we are dealing with several chem-
¡cal species: CH20, CH4, and O2. Thus according to Eq. 25.13, l gC of organic
matter would yield 0.5 gC of methane in the anaerobic sediments. Then, according
to Eq. 25.14, the 0.5 gC of methane would consume 2.67 gO in the aerobic zone.
These conversions can be avoided by expressing all the species as oxygen equiv-
alents, that is, as the oxygen required for complete oxidation. Di Toro uses the
nomenclature 0* to distinguish oxygen equivalents from the actual oxygen con-
centration O2. If this is done, l gO* of organic matter yields 1 gO* of methane that
consumes 1 gO* of oxygen. Thus the need for stoichiometric conversions is unne~
essary. It should be noted that the idea of oxygen equivalents should not be new
to you. In fact we have already used them in the previous section when we used
CBOD and NBOD to represent the oxygen equivalents of organic carbon and ni-
trogen.
Nltrogen. The two-zone scheme depicted in Fig. 25.4 also affects nitrogen. In
the anaerobic zone, ammonium is generated via ammonification of organic N. Thus
a gradient is created that feeds ammonium to the aerobic layer. In the aerobic zone,
I nitrification of the ammonium serves as an oxygen-demanding reaction that exerts
r~n = 0.7~¡32) = 1.714 gO gWI (25.18)
where r~n = oxygen demand for nitrilication, corrected for denitrilication. "'
This is an extremely important result since it indicates that the oxygen con-
nitrilication reaction. This reduction occurs because methane is consumed in the
denitrilication process. Consequently Eq. 25.12 should be adjusted to
SÍ1 = 1.llJe- (25.19)
Thus, rather than the 30% additional SOD due to nitrification (Eq. 25.12), the inclu-
sion of denitrilication lowers the increase to about 11%.
In summary the foregoing description does not lead to serious modiflcations
to the simple Streeter-Phelps approaches described in the previous section. How-
ever, it does suggest that denitrilication mitigates the NSOD effect. Next we see that
modeling the vertical distribution of reactants in the sediments along with gas feit-
mation will have a much more significant impact on our calculations.
25.4 SOD MODELlNG (ANALYTICAL)
Di Toro and his colleagues used the framework described in the previous section
to explain the square-root relationship of SOD to sediment organic carDon contént.
Their theory is depicted in Fig. 25.5. In essence when the organic carboll content of
the sediment gets high, the amount of methane produced in the anaerobic sediments
can exceed its solubility and bubbles formoThese bubbles advect upward due to
their buoyancy and consequently represent a loss of organic carbon tiJat does not

L___ _ - -
.J...~_:ld'/:1Ito '. , ,- _o.,

460 PAR1'IV Dissolved Oxygen and Pathogens LECfURE25 Sediment Oxygen Demand 461

J,.=Sc!l J...,=S(H J,.= Sc!l J.., =s, 1

e e
o o
Water
CH4 column NH4

terob!c layer
Lo
sediment
_
Oxldatlon
Bubbles

/ L,

~- -Jmuum~ J~~~.~s,on H

_ Bubbles
Dlffuslon H , "sJ¡l"
2D
...;~..
c-~'
,- 2D

~:~: ~yer H1 z
l .----------.. -----.------------- (a) (b)
z
FIGURE 25.7
(a) Carbon dlagenesls (b) Nitrogen dlagenesls
Vertical profile when methane in Ihe pore walers (a) does not exceed
FIGURE 25.5 and (b) does exceed saturation.
A schematic diagram 01 Di Toro's SOD model (redrawn Irom Di Toro et al. 1990).
interstitial waters can be written as I
exert a sediment oxygen demando This loss is one component of the observed square- d2c
root relationship between SOD and sediment organic carbon content. O = D dz2 + Se (25.20)
The following sections are intended to describe Di Toro's approach. Because the where D = diffusion coefficient of methane in the sediment interstitial waters I
framework is somewhat complicated, I describe it in several parts. It is hoped that
fftis will help you to understand it more thoroughly and, in so doing, appreciate its
(m2d-l) k
e = concentration of methane as measured in oxygen equivalents
elegance.
¡.¡
'
(mgO L-1) \
Z = depth inlo the sediments (m)
~5.4.1 A Simple Methane Balance Se = a constant source of methane (mgO L- i d-I)
Rather than presenting the complete framework, we can develop a simplified version We .can integrate this equation for two cases. In Ihe firsl case we assume that 1;
to gain insight into Di Toro's approach. To do this we initially concentrate on carbono the methane never exceeds saturation. Thus all methane is in dissolved form and no
Then in a later section we broaden the framework to inelude nitrogen. bubbles formoFor this case the following boundary conditions can be assumed:
The active sediments can be idealized as a one-dimensional vertical system with
e(O) = O (25.21)
a thickness H (Fig. 25.6). A mass balance for dissolved methane in the sediment
de
e - O (25.22)
dz z=H
l

11: O The firstcondition specifies that the methane concentration ofthe 0\\'11\ing wa-
ter is zero. The second stipulates that transport across the bottom of the active layer
I is zero.
1 Using these conditions, we can solve Eq. 25.20 for
SeR z SeZ2
11 e = -¡) - 2D (25.23)
:¡¡
IH:' .' . ,'" '''.' '"
As displayed in Fig. 25.7a, the solution is a parabola that is zero at the surface and
1\~ j". J".!.;!lI.ii'.l'..m;1
increases to a maximum concentration at the bottom of the active sediments,
.
¡;..'..' mriF'1;}.t...;!.j~,,:.,I..,..'". '~:~~ ~~~~ nal scheme lor charac-
'Ít~I,Ü: ' y . ¡SI 'tP~ I.!'! I tE¡riZing
. the 'vertical
. distribulion 01 c(H) = SeH2 (25.24)
.
~ . ,;. 1
¡; "tI' , ,r " constituenlsin an activesedimentlayer: 2D
 ~;
-
, 462 PARTIV Dissolved Oxygen and Pathogens loofUl(US"~MltI1!ltt.O\t~gel\
~ftlitlft1 _~

Differentiatloh 6f Eq. 25.23 lind substitultort HitoFick'~ first law shows that thl! Twofutthértriodi6c!\UórlAcDIlbétnadeby~AtUtl(\Jt ffialumratéol'~
flux of oxygen equivalents back to the Water column is productioncan be relatedtothe incomingcarbortftdx1>9:1' "
dc ~
S = Jt;- J~.~':
lou! = Ddz = ScH (25.25) C H \ "J.
As would be expected, the flux out corresponds to all the methane production in the (where both Sc and lc' are measured in oxygen equivalentsj, and llie diffuMort<:8-,
active sediment layer. Since all the organic carbon in the sediment is assumed to be efficient can be expressed as a mass-transfer coefficient,
due to settling of particulates from the water column, this means that the flux out is D " l
equal to the flux in. Consequently if the methane never exceeds saturation, the lin- KD =H (25.3:»
ear model described in the previous section would be valid.
Next we look at a second case where the organic carbon levels are high enough If we assume that all the methane is completely oxidized in a small aerobic tá~~r
that methane saturation is exceeded and, hence, bubbles formo For this case, along at the sediment-water interface, we can then express the maximum carbohaceous
with the surface condition (Eq. 25.21), the following bottom boundary conditions SOD as
hold: CSODmax = J2KDCslC' (2SJ6)
c(L,) = c, (25.26) j' t
dc EXAMPLE 25.1. MAXIMUM CSOD. Use the model developed above to,detef!lilne
the downward flux of organic carbon that corresponds to the point at which Iñetháne
dz l z= L,
=O (25.27)
becomes saturated for a 10-cm-thick active sediment lOen develop a plot of CSODmax
versus downward flux for the same sediment. Noté that the diffusion mass-trailsfeÍ' có-
The first condition specifies that there is a depth L, at which the methane concentra-
efficient for methane is 0.00139 m d-I, and e, ==100 mgO L-l.
tion reaches saturation, C,. The second condition stipulates that net transport across
this depth is zero. Solutloo: First, we can combineEqs. 25.24 [wlth e(H) = e,], 25.34,aild 25j~ !Uid
'h
solvefor
Employing these conditions, we can solve Eq. 25.20 for ~
~
il
1
',
,,
,, Sc 2 Je- = 2KDe, = 2(0.00139)(100) = 0.278 gO m-2 d-I
((
;J c = c..- 2D(L..- z) z :S L, (25.28a) Thus, for carbon fluxes lower than this value, the linear SO)) mude. holdt Tl!é
c = c.. z> L, (25.28b) square-root relationship of Eq. 25.36 can be used to determine the CSOD for higlier
values. The results are displayed below. .
I\s displayed in Fig. 25.7b, the solution ic¡again parabolic. However, now it reaches
saluration at a level L, that is above the bottom of the active layer. This depth can
he determined by substituting c = Oand z = O into Eq. 25.280, which can be solved
ror
:::- 10 I .

t
. '

(25.29)
L.. = J2DC,
Se
O 5
t A SOD =J c.1
In addition the flux at the sediment-water interface can be determined as S
Q
lou! = ScL, (25.30) O
1/)
Because L, < H, the SOD for this case would be less than for Eq. 25.25 because ..Y
o
some of the organic carbon is lost as methane bubbles. The gas flux loss can also bé'
O 20 40 60
quantified as
19as = Sc(H - L,) (25.31)
J co (gO m-2 (1-1)

FIGURE E25.1
Finally more useful forms of the flux equations can be developed by substitut-
ing Eq. 25.29 into Eqs. 25.30 and 25.31 to yield
lnn! = J2Dc..Sc (25.32) The foregoing analysis is very important because it pro vides an explanadon for
and the square-root dependence of SOD on organic carbono That is, it indicates that the
19as = ScH - J2Dc,Sc (25.33)

¡,
 " .~

LECTIIKE25 Sediment Oxygen Demand 465 ,\

464 PARTIV Oissolved Oxygen and Pathogens
I
I
o Concentratlon (mgO L-1) ,
Aeroblc zone (1) o 100 200 o 5 10
L
1!' °1 o !
I 'u 'u'UU" L I
1 i' '. ke=o FIGURE25.9 I

l
t Anaeroblc zone (2) -- CH4 "./ Proliles01(a) methaneand

T
H FIGURE 25.8
t
e!.
5
~
0.05
K ...
'. '..
Lo (b) methane and oxygen
'Ior TheDi Toro's line
dashed SODshows
model.
the
limiting case where there is no
Segmentation scheme lor an active 10.L I 0.1 .L "' oxidation in the aerobic zone I
sediment layer composed 01 aerobic (redrawn Irom Di Toro et al. I
% and anaerobic zones. (a) (b) 1991). r
t

square-root relationship is caused by carbon losses due to methane bubble formation

at high organic carbon levels. -D dC¡ = -D dC2 flux continuity at anaerobic-aerobic
dz z=4:"
1
dz z=4"
1
interface (25.41)
c2(Ls) = Cs saturation coneentration at saturation
~!$'1'~ P.xldatlon In the Aeroblc Zone
depth (25.42)
In the previous section we assumed that all the dissolved methane would be com-
pletely oxidized in a thin aerobic layer at the sediment surface. Depending on the A numerical solution for the active sediment is depicted in Fig. 25.9a. This graph
re'atiye ptagnitude of the oxidation reaction and diffusive transport, this might not shows two cases-with (solid) and without (dashed) oxidation-indicating that the
J>etfUe. for example if the oxidation reaction were much slower than the transport, inc1usion of oxidation has negligible effeet on the profile. However, a blow-up of
rnucl1of the methane would merely pass through the aerobic layer. If such were the aerobic layer in Fig. 25.9b shows that oxidation has a major impact within the
$,e cas~ th~ resulting SOD would be less than computed by Eq. 25.36. aerobic layer. Further, the oxygen profile for the aerobic layer, shown in Fig. 25.9b, '1
.
To quantify this effect, methane oxidation can be added to our model. To do this indicates that (1) the aerobic layer is much smaller (on the order of l mm) than ei- I ~
we c~Q\yrite mass balances for each zone in Fig. 25.8, ther the active layer (10 cm) or the undersaturated layer (2 cm), and (2) Ihe oxygen
d2c¡ profile is approximately linear. i
(25.37) Although the profiles shown in Fig. 25.9 are instructive, we are preselltly more
O = D dz2 - kecl
interested in the impact of oxidation on the SOD. To quantify this impact, Di Toro I
d2c2 (25.38) et al. (1990) have also developed an analytical solution for the same system. Their
and O = D- + Se solution can be differentiated and substituted into Fick's first law to compute a for- I
dz2
mula for the flux of dissolved methane at the sediment-water interface,
where Ct and C2 = dissolved methane concentrations in the aerobic and anaerobic
zones, respectively, and ke = decomposition rate of dissolved methane in the aero- lOUI= j2KDCslc' sech(AeLo) (25.43 )
bic layer. Note that we have omitted the source term (Se) from Eq. 25.37. Although where
¡
this assumption is not necessary to obtain a solution, it greatly simplifies the analysis.
Tpe orqission of the source term can be justified by the fact that, as we will discover, ~ (2<;.44) ¡
I

.h!f Ihickness of the aerobic layer will be much smaller than that of the anaerobic Ae = '-1D
\~yer. ~onsequently the methane production in this layer would be dwarfed by the and the hyperbolic secant is computed as
I
, (
I

, ~~nilude of the rnethanefluxffom\he anaerobiczone.

These 4Uf~feotial equations are subject to four boundary cO!1ditions: 2 I
(25.45)
sech(x) = eX + e-x
r'l! ~,(P) = O" ~ro metltaneat the sedimem-water
mterface (25.39)
Jn addition they computed the flux of methane gas,

::~ ,11," '!CI1L~)

<i )JI.II " ¡Il,
== f2(i~} ,t~¡¡concen~at~oncontinuityat
,,' Jt anaeroblc mterface
aerobic- (25.40) 19as= le' - j2KDCsle. (25.46)
 466 PART IV Dissolved Oxygen and Pathogens LECTURE2.5 Sediment Oxygen Demand 467

where

'i1
:c
g
11I
I
0.5
Ke =
M De
2
ke---!!.

where we have used the subscript c to distinguish the methane diffusion coefficient
(25.52)

....
(De) from that for oxygen (Do).
FinallyEq. 25.51can be substituledintoEq.25.47 to yieldthe finalmodelof
)1 FIGURE 25.10 CSOD,
2 4 6 A plot showing the behavior 01 1 -
X sech(x) versus x.
(25.53)
and the CSOD,
CSOD = J2KDC..le { 1 - sech[ Ke C~~D]}

CSOD = J2KDC..le[1 - sech(AeLo2)] (25.47) Note \hat for low carbon flux (le' < 2KDC.),the square root terro is replaced by le',

Now inspection of Eq. 25.47 indicates that the incorporation of oxidation mod- (25.54)
¡Iies the former solution (Eq. 25.36) by the term I - sech(x). As in Fig. 25.10, this CSOD = le { I- sech[ KC C~~D]}
term starts at zero and approaches I as x becomes large. In our case Eq. 25.47 will
approach Eq. 25.36 when the product of Ae and Lo2 becomes large. The former Because CSOD is on both sides of these equations, this equation must be solved
makes sense because Ae is directly proportional to ke and inversely proportional numerically as a roots problem,
to D. Thus we would expect Eq. 25.47 to approach maximum levels as the oxida-
tion rate increases or the diffusion coefficient decreases. Similarly the laller is rea- (25.55)
sonable because more oxidation would take place in a larger aerobic zone. f(CSOD) = le { I - sech[ Ke C~~D]} - CSOD= O
This numerical approach is described in the following example.
25.4.3 CSOD EXAMPLE 25.2. CSOD. Develop a plot of CSOD versus downward carbon flux for
\he same sediment as in Example 25.1. Also, compute the thickness of the aerobic layer
The final step in the analysis involves the determination of the thickness of the aer-
for this case. Assume that the oxygen concentration in the overlying water is 0(0) =
obic layer, Lo2. This can be accomplished by recognizing that \he total SOD is the 4 mg L -1 and that the reaction
coefficient is Kc = 0.575 m d-I . The diffusion coefficient
gradient of the pore-water oxygen concentration at the sediment-water interface. Us- for oxygen is approximately 2.1 X 10-5 cm2 S-l.
ing Fick's first law,
Solution: As an example supposc that Ihe downward flux is 10 gO m-2 d-I. This value
do
along wilh Ihe olher model paramelers can be substituted inlo Eq. 25.55 to give
SOD = - Do dz z=o l
(25.48)

where D" = diffusion coefficient for oxygen in water. The derivative can be approx-
imated by a first-order !inite-divided difference (that is, by a straight line), f(CSOD) = 10 [1 - sech (0.575 CS~D )] - CSOD

do The root of this equation can be delermined in a number of ways. In this example we use
== 0(0) - 0(Lo2) = _ 0(0) (25.49) the modified secant method,
dz ,=0
I O- Lo2 Lo2
Equation 25.49 can be substituted into Eq. 25.48, which can then be sol ved for €CSODf(CSOD)
CSOD = CSOD - f(CSOD + €CSOD) f(CSOD)
0(0)
(25.50)
Lo2 = D"SOD where € = a small perturbalion fraction. An excellent initial guess is provided by Eq.
25.36. For the present case this is
This result, along with Eq. 25.44, can be used to reformulate the argument of the
hyperbolic secant term in Eq. 25.47 as, CSODmax = j2(0.()()139)IOO(IO) = 1.6673

0(0) This value can be substituted into the formula for the modified secant method. Ir we use
AcLo2 = KC SOD (25.51) a value of € = 0.1, the resulting iterations are
 ...

468 PARTIV Dissolved Oxygen and Pathogens l.ECTURE25 Sediment Oxygen Demand 469

Iteration CSOD Approximate error (%) where l

O
I
1.6673
1.1557 44.27 [Df (25.57) ¡
2 1.1928 3.1 KN = V kND;.
3 1.1926 0.015 t
where kN = oxidation rate ofammonium to nitrogen gas and DI/ = diffusion co-
efficient for ammonium ion in water. Note that again for low carbon flux (lc' < I
The thickness of the aerobic layer can be ca1culated as 2KDCs),the square-root term is replaced by lc.,

' I
4 g m-3 m 10mm86,400s _ {
0(0)
[ [
0(0)
~ = 2.1 X 10-5 cm2 S-I
(1.1926gm-2d-1100cmcm d )- 0.608mm SOD = lc. 1 - sech KCSOD ]} + rOl/al/olc. { l - sech KN SOD ]}
(25.58)
I
i
Inspection 01'Eq. 25.56 shows that the NSOD is mathematicalIy similar to the
Results for other levels of downward flux are displayed below along with the maxi-
CSOD. The differences relate to the conversion of the carbon flux into nitrogenous
k'
mum CSOD values determined previously in Example 25.1. N¡¡tice how the inclusion of
oxidation reduces !he CSOD because some dissolved methane ditTuses to the overlying oxygen demand (r~l!al/o) and the absenceofthe square-root term becauseorthe lack i
of gas formation in the anaerobic layer. However, as with CSOD, the hyperbolic
I
waters before it can be oxidized.
secant term accounts for both the diffusion/reaction competition in the aerobic zone I
and the shutdown of SOD at low oxygen leveIs. As was the case for Eqs. 25.53
10 and 25.54, because SOD appears on both sides of Eqs. 25.56 and 25.58, they must I
be solved numerically as a roots problem. I
t
..,
ir SOD-~
EXAMPLE 25.3. TOTAL SOD. Develop a plot of total SOD versus downward carbon
I
I
E lIux for the same sediment as was studied in Examples 25.1 and 25.2. Assume thallhe
I
O 5 !
S I~~~~l-- parameters have the following values: KN = 0.897 m d--I, r;", = 1.714 gO gN-1, and
Q ano = 0.0654 gN gO-I.
O I
R ICSODI Solution: As wilh Example 25.2, suppose Ihat the downward flux is 10 gO m-2 d-I.
O This value along with the olher model paramelers can be subslituled inlo ElI. 25.56 i
O 20 40 60 .1 to give

J C' (90 m-2 ct-1) FIGURE E25.2

10
~
~

~
~

25.4.4 Nltrogenand Total SOD i

'ti
..,
ISOD=le.1 /150Dmul
Using a similar analysis as applied to carbon, Di Toro et al. (1990) also evaluated
!he effect of nitrification on SOD. As depicted in Fig. 25.5b nitrogen differs from E
5 ~
~
~
~
~

0(0) =10
O ~
~

carbon in that the ammonium formed from organic nitrogen does not become satu- S ~
~

/
rated and form bubbles as was the case for the methane from organic carbon.. Q ~

O /
~
0(01'- 4
This difference is accounted for by the folIowing equation for the total SOD due U) /.
/
~

to both carbon and nitrogen,

O

SOD = j2KDCslC.{ 1 - sech[ KC

l __
'v
;66]}+ r~nanolc
J l
{ 1 - sech[ KN
v
;66]} J
O 20 40 60

NSOD
lc. (gO m-2 d-1)
. CSOD
(25.56) FIGURE E25.3
;l 470 PARTIV Dissolved Oxygen and Pathogens LECl'URE25.' Sedimertt.Oxygeít Demaftd 471
"..
for estuarine systems. However; in this section we I¡mit our discussion to the ffesh-,

f(SOD) = J2(0.OOI39)IOO(IO>(t - sech(0.575 S~D)] water case. At the end of the discussion we wil\ describe how estuarine systems
differ from freshwater environments.!.
In this section we first describe Ii simplified sob model that toctises on i1thttt~-
+ 1.714(0.0654)(10) [1 - seCh(0.897S~D)] - SOD nium creation and oxidation. Later we wil\ incorporate carbon and other elements
into the framework.
The root of this equation can be determined with the modified secant method, with an
initial guess provided by As in Fig. 25.11, we can represent the active sediments as two wetl-mixed I~y_
\ ers. Notice that in contrast to the analytical model, we are representing the downward
CSODmax = j2Kvc"JC + r;..annJC'
flux of organic matter in carbon units (rather than oxygen equivalents).
This value can be substituted into the formula for the modified secant method. The results A mass balance can be written for organic carbon in the lower sediment layer as
along with SODs for other flux rates are displayed in Fig. E25.3. In addition the plot dC2
shows results when the oxygen in the water is raised to 10 mg L -1. As expected this
leads to higher SODs.
V2dt = leAs - kc2V2C2 (25.59)

where C2 = concentration of organic carbon (gC m-3)

le = fluxof organiccarbonsettlingfromthe overlyingwater(gC m-2 d-I)
Before proceeding, let's regroup and consider what the analytical model tells us As = surface area of the sediment-water interface (m2)
about SOD:
kc2 = first-orderdiagenesisrateof organiccarbon(d-I)
l. Ir there is oxygen in the overlying water, oxidation of methane and ammonium Now because the surface area between all the model segments is equal to As and
takes place in a very thin aerobic layer at the sediment-water interface. V2 = A.,H2,wecandividebothsidesof the balanceby As to yield
2. In the thin aerobic layer a competition exists between the oxidation reaction and
diffusion. The l - sech terms in Eq. 25.58 account for this competition as well dC2
as the suppression of SOD when oxygen in the overlying water gets low. H2dt = le - kc2H2C2 (25.60)
3. Methane can exceed its saturation concentration in the anaerobic sediments. Con-
Thus we have expressed the mass balance in terms of flux (M L-2 T-I) rather
sequently bubbles can be lost from the system, as reflected by the square-root than rate of mass transfer (M T-I). Similar flux balances can be written for ammo-
. term in Eq. 25.58. Because ammonium does not form a gaseous phase at the
pHs encountered in sediments, the relationship of NSOD to downward oxygen-
equivalent flux in Eq. 25.58 is linear.
nium as
dn2
H2dt = allckc2H2C2+ vdll12(nl- n2) (25.61)
By providing a means to calculate SOD as a function of externalloadings, the
and dnl
Di Toro framework is a major advance over prior schemes. Predictions should be (25.62)
HI dt = Vdlll2(n2 - n.) - Vdllwlnl- klllHtnt
more realistic than those obtained with the two alternatives used by earlier mod-
clers: fixed SOD or a linear decrease in proportion to load reductions. The former
is overly conservative whereas the latter is overly optimistic. By allowing a more
physically realistic middle course, the Di Toro approach should provide a basis for
le
better management decisions. Most importantly it allows SOD to be computed in-
lernally in a mechanistic fashion rather than in a prescribed or empirical manner.
To date Di Toro's framework has been applied to the Chesapeake Bay (Cerco and
Cole 1993, Di Toro and Fitzpatrick 1993). In the coming years numerical versions Aeroblc layer (1)
of the approach wi\l undoubtedly be incorporated into mos! water-quality models tó dlffuslon
provide a more rigorous characterization of SOD.

25.5 NUMERICALSOD MODEL

~~ ammonlflcatlon
Anaeroblc layer (2)

Although the analytical solution is elegant and instructive, numerical solution tech-
njques provide greater flexibility and broader application. The following models FIGURE 25.11
are similar in spirit to the approach outlined by Di Toro and Fitzpatrick (1993) Schematic 01 a control-volume approach lor modeling NSOD.
'---' ~ _~. ~ __~_. ._ _h

I
472 PARTIV Dissolved Oxygen and Pathogens LECTLJRE
2S Sedimenl Oxygen Demand 473

I -
where n = ammoniumconcentrationexpressedas nitrogen(gNm-3) where Do = oxygen diffusion coefficient (m2 d-I).Now it can be recognized that
Vdnl2= diffusivemass-transfercoefficientbetweenthe twolayers(m d-I) the SOD is also equivalent /0 the rale at which ammonium is being oxidized,
vdnwl = diffusivemass-transfercoefficientbetweenthesurfacelayerand the
SOD = r~l/kl/lHlnl (25.70)
water(md-I)
knl = nitrification rate of ammonium in the surface layer (d-I) where r~n = oxygen demand for nitrification/denitrification (according lo Eq. 25.1 g,
NOlicethat !3q. 25.62 assumes that the ammonium concentration of the overlying 1.714 gO gN-I). Assuming that Hwl = H.. Eqs. 25.67 and 25.69 can bc substituted
water is negligible. Although this will not necessarily be true, we have adopted this into Eq. 25.70 and the result manipulated to give
assumption to be consistent with the analytical framework. Also recognize that the 1
mass-transfer coefficients can be related to more fundamental parameters by NSOD = r~l/ancle- (25.71)
D1/NSOO'
Vdnl2 = -HDn Vdnwl = -HDnwl (25.63) 1 + D~o~knl I
12 I
where we have changed SOD to NSOD to acknowledge that we are presently limit-
where Dn = diffusioncoe/lI cnt for ammoniumin water(m2d-I) and HI2 and ing ourselves to ammonium oxidation. This equation can be solved numerically for l.
HWI = mixing lengths for the two diffusive transfer processes (m). NSOD. This result can then be used in conjunction with Eq. 25.69 to solve for the I
As described later these balances can be solved numerically to simulate the thickness of the aerobic layer. I
concentra/ions as a function of time. Before doing this it is instructive to first solve Thus Eq. 25.71 is a simple finite-difference representation of the analytical ap- i
them at s/eady-state. Along with insight into how the model functions, such solu- proach which for NSOD is the second part of Eq. 25.56, I
tions aIlow direct comparison between our lumped, numerical characterization and
Di Toro's distributed, analytical solution.
After setting Eqs. 25.60 to 25.62 to steady-state, Eq. 25.60 can be solved for I
= ,;.",,1c
Je
C2 = k H
c2 2
_ (25.64)
NSOD [1 - =h(
EXAMPLE 25.4. COMPARISON OF LUMPEO ANO DISTRIBUTEO NSOD.
k..~: N;~D) ] "'.72)

which in tum can be substitu/ed into Eq. 25.61, Develop a plol 01'NSOD versus downward carbon flux for the same sediment as was
studied in Examples 25.1 through 25.3 using both the distributed and lumped NSOD
o = ancle + Vdnl2(nl- n2) (25.65) models. Assume Ihal the parameters have Ihe following values: Ow = 10 mg L-1,
Next Eqs. 25.65 and 25.62 can be added to eliminate n2, KN = 0.897 m d-I, r;.. = 1.714 gO gN-I, Do = 1.73 X 10-4 m2 d-I, D" = 8.47 x
lO-s m2 d-I, and al/<'= 0.176 gN gC-1.
O = aneJe - knlHlnl - Vdlllvlnl (25.66)
Solullon: First, we must determine Ihe nitrification rate from Eq. 25.57,
This equation can then be solved for
aneJe k,,1= K~~;o = 0.8972.~.~! x !~-.s_ = 2282d-I
ni = knlHI + Vdnwl
(25.67)
Next the parameter values can be subslituled ioto Eq. 25.72 to oblain the <lbuibuted
whichcan thenbe substitutedbackintoEq.25.62togive, solution. For example for the case where lc' = 10 gO m-2 d-I (which corresponds to
le = 3.75gCm-2d-I),
Vdnl2 + knlHI + Vdn",1
n2 = Vdnl2(knlHI + Vdnwl)
an<' J e (25.68)
NSOD = 1.714(0.176)3.75 1 - sech J2282(1.73 lO-s NSOD
8.47xx 10-4)2 10
Therefore Eqs. 25.64, 25.67, and 25.68 provide steady-state solutions for all the { [ ]}
variables as a function of the organic carbon flux. which can be sol ved ileratively 1',)1'NSOD 1.1 JO gO m-2 d-I.
Oxygen demand can be integrated into Ihe framework by recognizing that the In a similar fashion the lumped case can be set up for Eq. 25.71,
oxygen concentration profile through the surface layer is very cIose to being linear.
Thus the SOD at the sediment-water interface can be nicely approximated by the NSOD = 1.714(0.176)3.75 1
following simple expression of Fick's first law (recall Eq. 25.50), 1 + c. ~_ ---,_ .._.._._____
Do
SOD = Hlow (25.69) which can be solved for 1.113 gO m-2 d-I. Thus the approaches yield similar results.
Additional solutions are displayed in the plot shown in Figure E25.4:
 474 PARTIV Dissolved Oxygen and Pathogens LECI'URE25 Sediment Oxygen Demand 47S

10
ASOD;'h..1
~
Aeroblc layer (1)
1
o 5
.!!! Anaeroblc layer (2)
Q
O
IJ) FIGURE 25.12
Anaeroblc layer (3) Schematic 01 a control-volume
o approach lor modeling CSOD in
O 20 40 60 Anaeroblc layer (4) Ireshwater systems with melhane
lormation.
le<(gO m-2d-') FIGURE E25.4

relationship manifested by the analytical solution does nol occur. This problem is
compoundedbecausethe saturationis usuallyexceededat loworganiccarbonflux.
Thus not only does the CSOD hil a ceiling, but it does so at an unrealistically low
25.6 OTHER SOD MODELlNG ISSUES (ADVANCED TOPIC) leve!.
At the limeofthis book'spublicalion,SOOmodelingis an activeandrapidlyevoIv- A remedy for Ihis problem is depicled in Fig. 25.12. By dividing!be anaerobic
ing research area. Consequently Ihere are a number of relevant issues that are cur- zone inlo layers, Ihe saluration phenomena will come inlo play in stages rather Ihan
rently unresolved and that were nol included in the prior sections. Because these at a single instant. In olher words, the lowest layer will saturate first at which point
topics are highly relevant to how SOD will be computed in the coming years, they its level becomes fixed. Thereafter, the middle and then the top layerwill satutate.
are the focus of this section. Once alllayers saturate, the CSOD wil\ be fixed. However, the sequential saturation
up to that pointwillfollowthe analyticalsolutionnicely.In additionit extendsthe
va!idityof the approximationto higherlevelsthan wouldbe the case for the single
25.6.1 Methane Bubble Formation layer.
Although the foregoing section shows how numerical, finite-difference methods can
be developed, it did not address the issue of methane bubble formation in the anaero- 25.6.2 Time-Variable SOD Calculatlons
bic sedimenls. The simplest approach is merely to model methane in a fashion similar
to ammonium, as in To this point we have focust:d on steady-state computations. For most water-quality
problems, time-variable solutions would be of equal, if nol greater, importance.
dC2 The most straightforward approach would be to merely integrate the equations
(25.73)
H2d( = le - kc2H2C2 numerically using either explicit or implicit numerical algorithms as described pre-
viously in Lees. 7 and II through 13. For example Eqs. 25.60 to 25.62 could be
dm2 integrated to delermine time-variable NSOD.
(25.74)
H2-;¡¡- = kc2H2C2 + vdnrdml - m2) Although this is a viable option, a problem arises in that the thin surface layer
requires much finer temporal resolution (Ihat is, a much smaller time step) lo attain
dml
(25.75) stable solutions than does the thicker anaerobic layer. In numerical methods this is
HI-;¡¡- = Vdnrl2(m2- mi) - vdmwlml - kmlH,ml referred to as a "stiff" solution.

Once the deep aerobic layer methane exceeds saturation, it can be kept fixed at the This apparent dilemma leads to a beneficia! insight. As recognized by Di Toro
saturation leve!.This means that Eq. 25.75 becomes and Fitzpatrick (1993), because it is much thinner, the surface layer will always be
at a steady-state relative to the time frame needed to resolve variations in the thicker
dml anaerobic layer.
(25.76)
HI-;¡¡- = VdmI2(C"
- mi) - Vdmwlml- kmlHlm, Consequently Eq. 25.62 can be solved at steady-state for
Vdnl2
At this point this equation can be directly solved for the methane concentration and ni = RI2n2 = Vdnl2 + Vdn...1 + knlHI
n2 (25.77)
SOD in a fashion similar to the derivalion of Eq. 25.71.
Although Ihis approach may seem appealing, il has one major flaw. That is, where R 12 = ralio of the ammonium concenlration in the aerobic (1) to the anaerobic
once the saluration is exceeded, the CSOD becomes constan!. Thus the square root layer (2). This value can then besubstituted into Eq. 25.61 to give
1,

_.- --- _.-

...r~ --" ~ ---L

476 PARTIV Dissolved Oxygen and Palhogens LECfURE25 SedimcnI Oxygen Demand 477

dll2 the water. Thus for deeper waters with relatively low flows (for example, deeper
I (25.78) lakes and estuaries), such a boundary layer can be significan!.'Consequently it could
H2{ft =; anck('2H2C2 + vd"dRI2 - 1)112
serve as a compartment where additionalll' ¡dation could take place.
Thus instead of integrating three differential equations (one of which is less stable The foregoing finile-difference modcl:. can be extended to account for such a
Ihan the other two), we integrate the two more robust ODEs (Eqs. 25.60 and 25.78) layer. For example mass balances can be wriuen for ammonium as
and determine the third unknown, 111,algebraically with Eq. 25.77. Note that this
dll2
trick applies equally well to methane and CSOD as well as to other elements in- H2{ft = a"ckc2H2C2 + Vd,,12(1I1 - 112)
(25.79)
volved in sediment ca1culations.
dlll
(25.80)
HI{ft = Vd,,12(1I2 - lId + Vd"OI(IIO -111) - k"IHlfll
25.6.3 Water Boundary Layer
dllO
In Example 25.4 a very high nitrification rate is needed so that the model as originally (25.81 )
HO{ft = Vd"OI (111 - 110) + "d"uO(lIw - 110) - k"oHolIO
fonnulated yields realistic NSOD values. A similar conclusion can be also made for
CSOD. In general, biologically mediated reactions such as nitrification and methane Notice how we have now included the ammonium concentralion in the overlying
oxidation generally proceed at rates on Ihe order of 0.0 I to 5 d-I rather that the water,IIw.
magnitude of 1000 to 10,000 d I used in E>.ample25.4. At steady-slale Ihese equations can be solved simultaneously (along with Eq.
One reason for the high rates is that sediments have much higher bacterial 25.60) for the organic carbon and the ammonium concenlrations. The values 01'am-
biomass than the water. A second explanation lies in Di Toro's assumption that all the monium for the upper Iwo layers can then be subslituled inlo Ihe following relation-
decomposition takes plact: itl a Ihin aerobic sediment layer. Although this assumption ship for SOD:
was certainly valid from Di Toro's perspective and does nut negate his conclusion, it
must now be reexamined as we begin to integrate the mechanistic SOD mode' into
SOD = r~,,(k"IHIIII + k"oHoIIO) (25.82)
large water-quality-modeling frameworks. . Thisformulacanbe combinedwilh Eq.25.69inorderto developan ileralivemelhod
As depicted in Fig. 25.13, along with the sediment layers, a laminar boundary for determining the SOD.
layer can exist in the water. The size of the two layt:rs would be determined by Although the foregoing analysis provides an indicalion uf the importance 01'the
different processes. The aerobic sediment layer thickness would be dictated by the waterboundarylayer,it suffersin that the Ihicknessof Ihe boundarylaYl'rlIIuslbe
balance between oxygen consumption and diffusion as expressed by Eqs. 25.69 given.A superiorapproachwould linkthe transferin Ihislayerlo the flO\\velocity
and 25.70. In contrast the water boundary layer is an artifact 01'the turbulence in in the overlyingwater.Recenleffortsare beingmade to developsuchn:lalionships
(Nakamuraand Stefan 1994).

°w
le 25.6.4 Nitrate and Phosphate
Becauseof our present emphasis on oxygen, alllhe other compounds disclIssed to
uTmuu

Water boundary this poinl (ammonium, methane) directly impacl SOD. However, beyond 11.,'\,'com-
Ho nltrlflcatlon pounds Ihere are a variely of olhers Ihat are important and are signilicanU) aflccled
layer (O)
by oxygen.

nltrlflcatlon Aerobic Nitrate. Nilrate is an important plant nUlricnl that is released dll.il\)! nilrifi-
H.
. ...u ~~y_~~_P)-------_...-_.-. cation. To Ihis point we have assumed Ihat it is 10lally rcduced 10 nitlll'" , ~.~asby
denilrificalion. Although Ihis is a reasonable firsl assumpll that \\, di f(lr
oo,

u f many systems, a more refined viewpoinl WOIIi, ,:unsider 11. IIly pan gen-
Hz Anaerobic erated nitrate flux would be denitrified and par! n:leased lo Ih.: overlyin¡; .. .ller. Di
layer (2) Toro and Fitzpatrick (1993) show how Ihis can be done by slightly modllying Ihe
SOD model described previously.
First, nitrate is modeled explicitly. For the Iwo-Iayer model, balances would be
wriuen as
FIGURE 25.13
An NSOD model including a water laminar boundary layer.
 47R I'AUTIV Dissolved Oxygen and Pathogens LECTURE
25 Sediment Oxygen Demand 479

H d;2
"
k H
. where cp = sediment porosity (== 0.8 lo 0.95)
(25.83)
2di = Vd,,12 ( '1 - '2) - ;2 2/2 p = sediment density (== 2.5 X 106 g m-3)
Ktlp = phosphorus sorption coefficient (m3 g-I)
and HI ~: = vd"di2 - ;1) - VtI"wl(iO
- ;1) + k"IHlnl - kilHI;1 (25.84)
The sorption coefficient is the key to the phosphorus model. At high values
where; = nilrale concenlralion (mgN L -1) and k; = denilrification rate (d-I). For (==0.01 to 0.001 m3 g-I), the dissolved concentration will be low and diffusion
the steady-stale case Ihese equations can be solved for ;1 and ;2. These values, in in the pore water will be low. Conversely at low values of K"p (==0.00005 to
tum, can be used lo compute the flux of nitrogen gas that would be created and lost 0.0005 m3 g-I), pore water concentration and diffusion become high.
from the sediments, Aside from sorption, three additional features ofEqs. 25.87 and 25.88 bear men-
lN = kilH1;1 + ki2H2;2 (25.85) tion. First, notice that I have inc1uded terms for partic1e mixing. Such transport
is due to burrowing and other activities of benthic organisms. Second, I have in-
This value is then used to correct the carbon flux to acknowledge that methane is uti- c1uded an advective term to account for the movement of matter away from the
lized during denitritication. The final model for conditions below methane saturation sediment-water interface due to the accumulation of settling solids in the sediments.
would look like
Finally I should mention that whereas we have used the steady-state case for all
1 1 previous constituents, it is not meaningful for a substance like phosphorus that is
SOD = r,,('(le - ac"lN)
DcSOD
2 + ronancle
DnSOD
~ (25.86) not continuously subject to a constant diffusion mechanism. Thus, unless the over-
1+ 2 2 1+ ~ Iying water is always aerobic (in which case the phosphorus Oux would proba-
D"owk mi Doowknl
bly be negligible anyway), phosphorus in the anaerobic layer must be simulated
where aCII = methaneloss due to denitritication= 1.07gC gN-1. Also,since we dynamically.
are directly discounting the CSOD, the true nitrification oxygen demand rnn =
4.57 gO gN-1 is used for the NSOD.
25.6.5 Estuarles
Phosphate. Phosphorus is another important plant nutrient that is greatly af-
fected by sediment oxygen levels. However, in contrast to nitrate ils accurate mod- Because of the dominant role of sulfur, estuaries tend to exhibit a different carbon
eling involves some major modifications. In particular the dynamics of phosphorus chemistry from freshwater systems like lakes and streams. In essence the SOD is
are significantly dictaled by its association with particulate matter in the sediments. tied to the reduction of sulfate (S04) lo produce sulfide (H2S). The key reaction is
A number of investigators starting with Mortimer (1941, 1942) have theorized (Bames and Goldberg 1976)
Ihat phosphorus in the aerobic surface layer is sorbed to precipitated iron hydrox-
¡des. Such attachment then impedes its transfer back into Ihe water. When oxygcn 2CH20 + H2S04 - 2C02 + H2S + 2H20 (25.90)
levels drop, the iron hydroxides are dissolved as they become reduced. This I~adsto
an associated release of dissolved phosphorus into the sediment pore waters where In the anaerobic zone, sulfirleis produced. A portion of it reacts with iron to form
particulate iron sulfide [FeS(s)]. The remainder diffuses into the aerobic zone where
it is free to diffuse into the overlying water.
These effects can be integrated into the SOD framework by writing balances for a part is oxidized. In addition partic1emixing can move some of the particulate iron
sulfide into the aerobic zone where it can be oxidized to form femc oxyhydroxide
phosphale that account for Ihe sorption mechanism,
[Fe203(S)], again consuming oxygen in the process.
dP2 Thus the model would be quite similar in format to the balances for phosphorus,
H2d( = a,)(.kc2H2C2 - Ftl2P2) + Vtlpl2(FplPI
+ Vtlp12(Ftl1PI - F,,2PÜ with the exception that
+ Vb(PI - P2) (25.87)
. Sulfideis modeledin placeof phosphorus.
dpl · Sulfide undergoes oxidation in the surface layer.
Hld( = Vtlp12(Ftl2P2 - FtlIPI) + Vtlp12(Fp2P2 - Fp2PI) · Precipitation reactions would not depend on oxygen in the way that phosphate
+ VtI",wl (FtlwP2 - Ftll PI) (25.88) sorption does.

where FtI and F" are the fractions associated with dissolved and particulate forms in Di Toro and Fitzpatrick (1993) provide a very thorough description of this sulfide
the two layers. Later in this book when we discuss toxic substances, we will describe model along with all the other elements of SOD modeling discussed in this lecture.
how Ihey are derived. For the time being they can be calculated by Finally, it should be noted that sulfur plays an important role in many freshwater
sedimenls.ln the fUlure, sulfurdynamics must be incorporated into SOD frameworks
F" = 1
4>+ KtI,,(I - 4»p
F" = 4>+Ktlp(l - 4»p
K,I".(I- 4»p
(25.89) developed for such systems.

1,
I
480 PARTIV Dissolved Oxygen and Pathogens LECTURE
25 Sediment Oxygen Demand 481

PROBLEMS 25.3. Repeat Example 25.2, but for a downward flux 01' 20 gO 01-2 d-I and a water oxygen
concentration 01'3 O1g L -1 .
25.1. A small pond located at 10,000 ft has the following characteristics:
25.4. A lake has a particulate organic carbon concentralion 01'5 O1gL -1 that settles at arate ot
Residence time = l wk
0.25 m d-I. Determine the steady-stale SOD and the thickness 01'the aerobic layer for
Area =l x lOS 012
the system if the overlying waler has an oxygcn concenlration 01'4 mg L -l. The acti ve
Mean depth =2m sedi01ent layer is 10 cm thick and assume that all the nilrate produced by nitrificalion
The pond has an inftow CBOD concentration 01'50 mg L -l. The CBOD does not settle is denitrified to nilrogen gas. Assume thallhere is no waler boundary layer.
but decays at arate 01'0.05 d-I at 200C. The temperature ofthe pond is 15°C. The wind (a) Use the analytical solulion to oblain your answer.
speed over the pond is 2 m S-l. The inftow has a temperature 01' IO.C and a deficit 01' (b) Use a numerical approach wilh the anaerobic sediments divided inlo four segments.
3 mg L-I. Hall' 01' the sediments exert no SOD whereas the other hall' exert an SOD 25.5. Repeat Probo 25.4, bUI add a l-c01-thick boundary (ayer in the water. Assume Ihal
01'0.2 g 01-2 d-I at 20.C. Compute the following (don't forget temperature correction diffusion and reaclion in bOlh Ihe sedi01enl aerobic and waler boundary layers are Ihe
factors): same.

(a) The CBOD concentration 01' the system

(b) The oxygen saturation concentration for the pond (use oftable in App. B is accept-
able)
(e) The reaeration rate for the pond at 15.C (units 01'd-I)
(d) The oxygen concentration 01'the pond

25.2. The epilimnion 01'a lake has a particulate organic carbon concentration 01' I mg L -1 and
adissolved oxygenconcentration 01'9 mg L -l. Thecarbon settles at amte 01'0.05 md-I
and is oxidized in the hypolimnion at arate 01'0.05 d-I. The bulk diffusion coefficient
across the thermocline is 0.05 m d-I and the hypolimnion thickness is 5 m. The situa-
tionis depictedin the followingdiagram: .

Eplllmnlon P8rtlculate Oxygen

org8nlc e

Thermocllne

se"lIng dlffualon dlffualon

!typollmnlon

SOD

Sedimenta FIGURE P25.2

(a) Compute the steady-state concentration 01'carbonaceous BOD and dissolved oxy-
gen in the hypolimnion. Assume that the epilimnion le veIs are constant and that
once the particulate matter settles to the bottom it does not exert an SOD.
(b) Por the case described in (a) compute the SOD that would be exerted by the par-
ticulate matter that settles from the hypolimnion to the sediments using the naive
Streeter-Phelps SOD model described in Seco 25.2. Assume that the particulate
matter is co01pletely oxidized in !he sediments. Express the resulting SOD in units
ofg 01-2 d-I.
 lECfURE 26 Compuler Methods 483

LECTURE 26 We showed how each element of [A]-1 provides the change in concentration in ele-
- -- -- -- - ment i due to a unit load change to segmentj.
We now il\ustrate how this idea can be expanded to BOD/oxygen modeling.
Computer Methods To do Ihis we limit our analysis to a one-dimensional syslem with carbonaceous
BaO and oxygen. The approach can be easily extended to multidimensional sys-
lems, NBOD, or linear represenlations of nitrification. A steady-state mass balance
for eBOD in segment i can be wrillen as
0= Wu + Q¡-I.i(CI¡-I,¡L¡-I+ f3¡-I.iL¡) - Q(CI¡,¡+IL¡
+ f3¡,i+ILi+I)
J
+ E:_I./(L¡_I - L¡) + E:,¡+I(L¡+I- L¡) - kr¡V¡L¡ (26.2)
Writing this equation for an n-segment syslem with appropriate upstream and
downstream boundary conditions gives
[A]{L} = {Wd (26.3)
with a solution

{L} = [Ar'{Wd (26.4)

Now the coefficienls of [A] bear scmtiny. They are

---
a/'/_I = -Cl¡_I,¡Q¡_I,¡ - E:_I,¡ (26.5)
LEctuitkOvEaVJEW:J reviewsomecomputef-orientedirlethót1$lór~drviriAt5t~- Transport
gen balances with the computer. First 1 develop thé $teiidy-atlitésystem
response m!\tOx for the Streeter-Phelps framework. ThenJ. próvlde.ltn. InttO- a¡.; CI//+IQ//+I - f3l-1¡Q¡-1 ¡ + E¡'_ I,' + E,! ,+I +
'
kr¡V¡
\,. " , 'J :..y..:.. (26.6)
t
duct/on to the QUAL2E ItIbdtH. After somé historiCáI bllCkgroutíá, de&etlbe Tran~port Kinetics
its underlying strUcture. Therl 1 illustrate 118useb)'lIpplyhl~ itto,.sHnúbité
oxy.gen levels in a strefint, with a focus oil oxygen d~pleUort due id iidb Ilrtd a¡,i+1 = f3/,/+IQ¡,i+1 - E:,i+1
(26.7)
SOD. Transport

AII the terms marked "transport" would be identical regardless of the poIlutant. Thus
the LA] matrix can be divided into two parts, as in
Although we have louched on some computer methods, lo Ihis poinl our discus-
sion of oxygen modeling has focused on analylical. closed-form solulions. Now we [A] = [T] + [krV] (26.8)
broaden our perspective lo show how numerical methods allow us to implement such where [T] is a "transport" matrix identical to malrix [A] but conlaining only the
caIculalions wilh Ihe computer. First I describe a malrix approach Ihat has great util- transport terms, and [krV] is a square diagonal matrix containing the terms kr¡V¡on
ily for developing sleady-slale solutions for linear DO models. Then I provide an the diagonal and Oelsewhere.
inlroduclion lo EPA's QUAL2E software package. Using a similar approach, we can write mass balances for deficit (see Thomann
and MuelIer 1987). However, because such an approach can lead to errors for sys-
tems where saturalion is variable (recall our discussion in Lec. 21), we write the
26.1 STEADY-STATE SYSTEM RESPONSE MATRIX
mass balance for oxygen,
In Lees. 6 and 11 we developed the steady-state-system response matrix as a meanS 0= Wo¡+ Q¡_I,¡(CI¡_I./O¡_I
+ f3¡-I,/0¡) - Q¡,i+I(CI/,/+IO¡
+ f3/,/+IOi+I)
lo summarize the interaclions of loadings and responses for linear water-quality mod-
els. Recall that the steady-slale mass balances for such systems could be expressed + E:_I,/(o¡_I - Di) + E:,¡+I(Oi+1- Di) - kd¡V¡L¡
as a set of linear algebraic equations. These equalions could then be solved as + ka¡V¡(os¡- Di) + p¡V¡ - R¡V¡ - S~As¡ (26.9)
{C} = [Arl{W} (26.1) Writing this equalion for an n-segment system with appropriate upstream and
where {e} = veclor of unknown concenlrations downstream boundary conditions gives
{W} = veclorof loadings [B]{o} = {W,,} + {PV} - {RV} - {SSA,,}+ {kaVO,,}- {kdVL} (26.10)
lAI I = malrix inverse or Ihe steady-slale-syslem response malrix
482
~'

4~4
"'-:

PARTIV
~ '-----

Dissolved Oxygen and Pathogens

--- ~

'..
.:
.!
.

..
..
.. ..
~
.

LECfURE26 Computer Methods 485 ~

where (b) The minimum DO is 2.84 mg L-1. Thus. to bring this value up to 5 mg L-1, we
must increase the DO by do = 5 - 2.84 = 2.16 mg L-l. This value can be translated
[B] = [T] + [kaVJ (26.\ 1)
into a load change by using the appropriate value fcom the matrix inverse,

J}y c~n~ctin~t~nns,wt?can~rite the resulting system of equations as I.64E-08 2.66E-Q9 3.25E-1O 3.53E-II 3.6IE-12
[B]{o} = {W~} - [kdV]{L} (26.12)
1.33E-08 1.80E-08 2.84E-Q9 3.43E-1O 3.7IE-II
w~ere {W~}is a matrix containing al1the external oxygen sources and sinks. 8.12E-Q9 1.42E-08 1.81E-08 2.85E-Q9 3.44E-1O
4.42E-Q9 8.58E-Q9 1.42E-08 1.81E-08 2.85E-09
{W~} = {Wo} + {PV} - {RV} - {SBA} + {kaVos} /
(26.13)
Extemal DireCI PhotosynIhesis Respiralion SOD Reaeration "¡f [C¡-I = 12.25E-09
I.IIE-Q9 4.64E-Q9
2.36E-Q9 8.60E-Q9
4.65E-09 1.42E-08 1.81E-08
8.60E-Q9 1.42E-08
sources loading gain loss loss gain 5.28E-1O 1.16E-Q9 2.36E-Q9 4.65E-Q9 8.60E-Q9
Multiplying both sides by the matrix inverse of [B] yields 2.47E-1O 5.5IE-1O 1.l6E-09 2.36E-09 4.65E-09
{o} = [Brt{W~} - [Brt[kdV)[Ar'{Wd (26.14)
1.l4E-10 2.57E-1O 5.5IE-1O 1.16E-09 2.36E-09
5.00E-Il 1.I4E-1O 2.47E-1O 5.28E-1O 1.11E-Q9 (
or {o} = [Brl{W~} - [Crl{Wi} . (26.\5)
3.54E-13 3.38E-14 3.i7E-15 2.92E-16 2.56E-17
where[Cr I is a systemresponsematrixthat relatesoxygenconcentrationto BOO 3.78E-12 3.70E !1 3.53E-14 3.29E-15 2.92E-16
toading, 3.72E-- 11 3.78E \70E-13 3.53E-14 3.17E-15
[Crl = [Brl[kdV][Arl (26.16) 3.44E-lO 3.72E-ll 3.78E-12 3.70E-13 3.38E-14
2.85E-09 3.44E-1O 3.72E-ll 3.78E-12 3.54E-13
t:~AMPLE 26.1. MATRIX APPROACH FOR OXYGEN. A one-dimensional es- 1.81E-08 2.85E-Q9 3.44E-1O 3.71E-ll 3.6iE-12
IUIIryhl\s !he following characteristics: 1.42E-08 1.81E-08 2.85E-09 3.43E-IO 3.53E-ll
Flow = 1 X 101m3 d-I BOD decay = 0.2 d-I 8.60E-Q9 1.42E-08 1.81E-08 2.84E-Q9 3.25E-1O
Width = 1500ql Reaeration= 0.25d-I 4.64E-Q9 8.58E-Q9 1.42E-08 1.80E-08 2.66E-09
Depth = ~m Saturation= 8 mg L-1 2.'. 09 4.42E-Q9 Q 12E-Q9 1.33E-08 I.64E-08
Dispersion = 1 X 107 m2 d-I
'.he estuary is 100 km long. The boundary conditions at both !he upstream and down- The needed value is in the fourth row anu the fourth column: c4} = 1.81 x
,.:am ends are L = O and o = os. Loadings of BOD and oxygen of 300.000 and lO-s (mg L -1 )/(g d-I). (Remember, the column location is dictated by the positiol1
tQQ,ooo kg 4-1, respectively, enter the estuary at KP 35. Centered differences Were of the loading and the row by the location of the resulting concentration.) This value can
used to approximate space. be used to compute
(á) Calculatethedistributionof BODand oxygenin !heestuaryusingsegmentsizesof
kg
10km.
(b~ Determine the BOD loading reduction needed to raise the minireum oxygen con-
dWL = 2.16mgL-1
1.81 x lO-smgL-1
gd-I
(iooog) = 119.337kgd-1
, ~lItraqon in the estuary to 5 mg L-l.
Solution: (a) Using the parameter values, we can write and solve the mass balance Thus the BOD load must be reduced by 40% to meet the oxygen goal.
equations. The results are displayed in Fig. E26.1.

10 The matrix approach is predicated on having a linear modeI. Although this is

10 ....
true for the elassic Streeter-Phelps framework, more recent oxygen models inelude
~al 'i

E
.....
Q
O
5
-
..1

E
O
Q
al 5
nonlinear tenns that preelude the matrix approach. Consequently it might seem that
the method outlined in the present section is obsolete. There are three reasons why
this isn't necessarily true:
IR \1 o 1. Even though nonlinearities appear in models, some of the new mechanisms are
( " ~ 50 7$ 1~ O 25 50 75 100
not highly nonlinear. Consequently some mechanisms are amenable to lineariza-
Qlstance (kIP) Dlstance (km) tion via a Taylor-series expansion. For such cases the foregoing approach could
\.
be employed without incurring major errors.
fiGURE E26.,
 486 PARTIV Dissolved Oxygen and Pathogens LEC'lURB26 Computet Methods, 481
¡
2. Although the Streeter-Phelps model n1ayseem passé, it is still useful where pn-
11
mary and secondary treatment are to be evaluated. Such management contexts
still occur in developing countries.
3. The general approach outlined here is useful for other problem contexts beyond
oxygen. For example many of the modeling frameworks used to assess toxic
substances are linear. Hence the matrix solution techniques are possible and
useful.

26.2 THE QUAL2EMODEL

e
The QUAL2E software package is presently the most widely used computer model
for simulating stream-water quality. It is capable of simulating up to 15 water-
quality constituents (Table 26.1) in dendritic streams that are well-mixed laterally
and vertically. Among its many capabilities it allows for multiple waste discharges,
withdrawals, tributary ftows, and incremental (that is, distributed) inftows and
outftows.
QUAL2E has its roots in the QUAL-I model developed by F. D. Masch and As-
sociates and the Texas Water Development Board (1970). In 1972 Water Resources
Engineers, Inc. (now Camp, Dresser and McKee), under contract with the U.S. En-
e
vironmental Protection Agency, modified and extended QUAL-I to create the first
version of QUAL-II. Over the ensuing years the model was upgraded several times
I (for example Roesner et al. 1981a, 198Ib). The present version (Brown and Bam-
!
l'

I
I
'!
"
well 1987) is known as the "enhanced QUAL-Il model" or QUAL2E for short. It
is currently maintained by the EPA's Center for Water Quality Modeling in Athens,
Georgia.
It should be noted that QUAL2E was originally developed using punch cards
as its input media. As the software moved to time-sharing systems and then per-
e 25
sonal computers, the input files to run the model maintained the data structure of
! the punch-card input. Recently a user-friendly interface for entering the input file
and viewing the results of QUAL2E simulations has been developed and is being (b)
(8)
distributed (Lahlou et al. 1995).This interface should further enhance the utility and
widespread application of this model. Although the interface employs easy-to-use, FIGURE 26.1
spreadsheet-like input. it still requires precisely the same information as for the flat- (a) A river basin and (b) a QUAL2E representation as reaches and elements.
file version. Thus this lecture (as well as later descriptions ofQUAL2E applications
in Lec. 36) uses the ftat-fi1eformal. So whether you use the original version or the 26.2.1 Spatlal Discretlzation and Model Overvlew
new interface, our input files provide a concise representation of all the necessary
information for performing model simulations with QUAL2E. As depicted in Fig. 26.1, QUAL2E treats a river as a collection of reaches, each
j'
having homogeneous hydrogeometric properties. Each reach, in tum, is divide<!Into
a series of equal-Iength computational elements or control vo1umes.
TABLE 26.1
A mass balance is used to keep track of the water-quality constituents \isted in
Tbe 15 constituents tbat can be simulated by QUAL2E Table 26.1. This balance can be written generally as
Dissolved oxygen Ammonia as N Coliform bacteria
Biochemical oxygen demand Nitrite as N Arbitrary nonconservalive constituent
Temperature Nitrate as N Conservative constituent I Vac = a(AcEac)
ax dx _ a(AcUc) dx + Vdc + s
Conservative constituent 11
at ax ax dt
Algae as chlorophyll a Organic phosphorus as P
Organic nitrogen as N Dissolved phosphorus as P Conservative constituent 11I Accumulation ,Dispersion Advection I Kinetics Externalsourceslsinks
y
Transport (26.17)
... -.....

488 PARTIV Dissolved Oxygen and Pathogens LKTURE 26 Computer Methods 489
¡ti
where V = volume
e = constituentconcentration
FIGURE 26.2 ~
-1c= elementcross-sectionalarea
,
E = longitudinal dispersion coefficient A trapezoidal channel showing the three
x = distance parameters needed to uniquely delie
I
the geometry. 80 = bottom width; 5,
U = average velocity Bo and So¿= side slopes.
s = external sources (positive) or sinks (negative) of the constituent ,.

In ~he foUowjng sections we elaborate on the three terms on the right-hand side of 1
Q = -A (26.22)
this balance. nC' R2I3SIf2
where R = channel'shydraulicradius(m) andS. = slope of the channel's en-
ergy grade line (dimensionless). Note that since we have assumed that flow is
26.2.2
.
Transport steady and cross sections are constant, the energy slope is equal to the channel
As in Eq. 26.17, transpon consists of two components: advection and dispersion. The slope. QUAL2E assumes that the channel has a trapezoidal cross section (Fig. 1
fooner specifies the movement of the constituent with water as it flows downstream. 26.2), so the cross-sectional area and hydraulic radius can be expressed as a func-
The latter relates to the spreading of the constituent that occurs primarily due to tion of depth. If Q is given, this means that Eq. 26.22 is a nonlinea. equation
shear. that can be solved for depth numericaIly.Oepth can then be employedto de-
termine the area, which can be used to compute velocity with Eq. 26.21 (recall
Advectlon. 'ft1e QUAL2E model assumes steady, nonuniform flow. The term Example14.5).
steadyJfowmeans that flow does not vary temporally. The term IIonuniform jIow
connotes that it varies spatially. For such a characterization a flow balance for ele- Dispersion. The QUAL2Emodeluses the foIlowingrelationshipto compute
ment i can be written as dispersion as a function of the channel's characteristics (Fischer et al. 1979),
E = 3.I1KnUH5/b (26.23)
Qi-I :t Qx.i - Qi =O (26.18)
where E = longitudinal dispersion coefficient (m2s-I)
where f2i-1 = flowfromtheupstreamelement n = channel's roughness coefficient (dimensionless)
Qi = outflow from the element U = meanvelocity(mps)
Qx.i = lateral flow into (positive) or out of (negative) the element H = mean depth (m)
K = a dispersion parameter (dimensionless)
Once the flowbalance is established, it is necessary to determine the other hydro-
geometric characteristics for each element. In particular the resulting water velocity, defined as
depth, and cross-sectional area must be established. The relationship of the element's E
other hydrogeometric characteristics to its flow rate is handled in two ways (recall K = HU. (26.24)
H:c. 14):
where U. = shear velocity (m s-I). Equations 26.23 and 26.24 represent circular
J. power equations can be use<! to relate mean velocity and depth to flow, reasoning. Thus they differ from Eqs. 14.15 and 14.17, which calculate dispersion
purely as a function of channel parameters. However, once K is established it pro-
U = ad' (26.19)
vides a formula to compute dispersion as a function of nonuniform flow conditions.
and H =- aQfJ (26.20) It is in this way that K is used in QUAL2E.

where H = mean depth and a, b, a, and ~ are empirical constants that are deter-
mined ñom stage-discharge rating curves. Once velocity has been determined, 26.2.3 Klnetlcs
the cross-sectional area can be calculated from the continuity equation
To keepthis lecture simple we limit the discussion to two constituents: carbona-
Q
ceous BOO (CBOO) and dissolved oxygen. The kinetics for these constituents are
Ac = U (26.21)
\ displayed in Fig. 26.3 and can be represented mathematically by
2.. 'fhe Manning equation p'royides a means to relate channel characteristics and dL
flQw.In metfic units, dt = -KIL - K3L (26.25)
 490 PARTIV Oissolved Oxygen and Palhogens I.ECI'URI!26 Computer Methods 491
11I
tABLE 26.2
Reaeratlon formulas In QUAL2E.Note tbat some o; tbese formulas (2, i, and 4)
dilTerslightly rrom tbe versions in Lec. 20 due to roundlng errors
Optlon Aulhor(s) Kz (d-' al 2O"C) Unlts
Sedlment oxygen
, \ 0, demend(SOD) I User-specified value
T~r 2 Ü (m s.,)
Churchill el al. (1962) 503 UO...
. - HI.613 H(m)
¡
3 Ü (m S"I)
, (lt O'Connor and Dobbins 395 U'l5
! (1958) . H '.3 H(m)
i
I 7}7T 4 Un'7
i Owens el al. (1964) Ü (m SO')
j IlJj, 5.34 H'.85 H(m)
i
I
i " .llz
-

lo
y
I :r 5 Thackston and Krenkel
(1966) 24.9
(1 + ./F)u'
H
where F is Ihe Fraude number,
F (dimensionless)
u' (m SO')
H(m)
I I F=-
u'
./iii
Ü (m S-I)

I (JJ- and u' is Ihe shear velocilY,

I
I . Unfi
-IX,1' u = jHS,g = HI.67
I
I <ttl' 6 Langbien and Durum
(1967)
U
5.13 HI.33
H(m)
Ü (m S"I)

I
0, ')",,- 7 User-specified power a(t Q (cms)
funclion
FIGURE 26.3
QUAL2E kinetics. Note that the highlighted constituents and processes 8 e (m-')
r Tsivoglou and Wallace e !1H
I are described and modeled in this lecture. We will exlend our (1972); '/ !1H (m)

I discussion lo Ihe remainder 01 the diagram in Lec. 36. Tsivoglou and Neal
(1976)
where !1H is change in waler-surface elevalion '/ (d)
in Ihe elemen!, '/ is Ihe flow lime in Ihe elemenl,
and e is a flow-dependenl escape coefficient: -
do K4
H e = 0.36 for 0.028 s Q s 0.28 cms
,
I and dI = K2(0., - o) - KI L - (26.26)
e = 0.177 for 0.708 s Q s 85 cms
I where L = carbonaceous BOD (mg L -1)
I KI = BOD decomposition rate (d-I)
I K3
o
=
=
BOD settling rate (d-I)
dissolved oxygen concentration (mg L -1) 26.2.4 Numerlcal Algorlthm
K2 = reaeration rate (d-I)
1; Now that we have an understanding of the major components of QUAL2E, we can

I
I o.,
K4
=
=
dissolved oxygen saturation concentration (mg L -1)
sediment oxygen demand (g m-2 d-I)
discuss how the model obtains solutions numerically. To do this, Eq. 26.17 canbe
divided by volume and written as
Note that all the rates (the K's) are corrected for temperature by
K = K200T-20 (26.27)
ae ax
a(AcEae) a(AcUe) s
where K = rate at temperature T (26.28)
1I K 20 = mte at 20°C
at = Acax - Acax + re + p + V
O = a temperature correction factor Observe that we have divided the kinetics into two separate terms,
AII the rates in Eqs. 26.25 and 26.26 can be entered directly to QUAL2E. How-
I de
ever, the reaeration rate can also be internally calculated using eight different for- (26.29)
mulas. These formulas are summarized in Table26.2. dI = re + p

1)
 _. "'!"_:-......-. ...

492 PARTIV Dissolved Oxygen and Pathogens JETURE 26 Computer Methods 493

Flow
(26.36)
g¡ = -l(AxE)¡ V¡Il~xJ

D 1-1
· I ·
1+ 1
[ FIGURE 26.4
QUAL2E elemenls.
Z¡ =
n
c¡
s¡ Ilt
+ -v;- +
A
p¡ ut
These equations fonn a tridiagonal system that can be sol ved efficiently for con-
(26.37)

centration as a function oftime. Note that the external sources and sinks (the s¡ terms)
. are treated as constants in this fonnulation. As noted previously, some of these are
The first tenn on the right denoles those reactions that are linearly dependent on
-' nonlinear functions of other constituent concentrations. Thus the QUAL2E solution
concentration, and the second tenn denotes internal constituent sources and sinks (for
algorithm handles nonlinear tenns by treating them as constant contributions to the
example, benthic sources, nutrient loss from algal growth, etc.). Although some of
the latter are constants, olhers are nonlinear functions of constituent concentrations. forcing function that are updated at each time step.
Two fonns 01'the solution are implemented in QUAL2E:
A general representation of the QUAL2E element scheme is shown in Hg. 26.4.
Equation 26.28 can be written for element i by using a backward difference, as in Steady-state. The model is run until it reaches a steady-state. This is the con-
ventional implemenlation mode for QUAL2E. Thus the time-variable algo-
_ AcEac rithm is a means to an end-the steady-state result (recall our discussion 01'
ac¡
at
( ) V¡
+ AcEac
ax ¡-J ( ax )¡ + (AcUC)¡-1- (AcUc)¡
V¡
Ihe backward-time implicit approach in Lec. 13).
Time-variable. The model can also be run in a normal time-variable mode. At

Acc In Out In Out present this type of implementation is limited to diurnal simulations.
Dispersion Advection
Now Ihat we have covered the basics ofthe QUAL2E software package, 1'11pro-
s¡ vide an overview of how it is implemented. However, before doing this l' 11introduce
+ r¡c¡ + p¡ +
V¡ (26.30) a problem context where we will apply the modeI.
First-order Internal External
reactions sources/sinks sourceslsinks

Then backward differences can be used to approximate the remaining spatial deriva- 26.2.5 QUAl2E Application
tives, Figure 26.5 shows a river that receives a sewage treatment plant efftuent at kilo-
meter point 100 (KP 100) and a tributary inflow at KP 60. Note that the channel
ac¡ (AcE)(Ci+t - c¡) (AcE)(c¡-J - c¡) Q¡-Jc¡-J - Q¡c¡ s¡
at is trapezoidal with the characteristics shown. The deoxygenation rate for CHOO is
V¡Ilx¡ + V¡Ilx¡ + V¡ + r¡c¡+ p¡ + V¡ equal to 0.5 d-I at 20°C. For 20 km downstream from Ihe treatment plant there is a
(26.31) CHOO settling removal rate ofO.25 d-I. In addition there is a sediment oxygen de-
finalIy a backward difference can be applied in time to yield mand for this reach of 5 gm-2 d-I. Assume that the O'Connor-Oobbins reaeration
fonnula holds and that the stream is at sea leveI. Values for the channel's geome-
c~+1 - c~ A E .. cH1 - c~+I A E . . c~+1- d+' )
, , = ( e ) '.'+l( 1+1 , ) + ( e ) '-.'1 ( ,-J , try and hydraulics are listed in Table 26.3. Kinetic parameters and temperatures are
~ ~Il~ ~Il~ summarized in Table 26.4.
(26.32)
Q . cHI - Q .cH1 s. TABLE 26.3
+ ,-1 ¡-I ' ¡ +. HI + . + -.!...
V¡ r,c¡ p, V¡ Hydrogeometric parameters
~ualion 26.32 can now be reexpressed by colIecting terms to yield a linear system Parameter Unlts KP > 100 KP1 KP < 60
e¡c¡~l + /;c¡+ I + g¡c¡:11 = Z¡ (26.33) Depth rn 1.19 1.24 1.41
(ft) (3.90) (4.07) (4.62)
wherr Area rn2 14.71 15.5 18.05
Flow rn) sol 5.787 6.250 7.407
, Ilt Q¡-J Ilt rn) d-' 500,000 540,000 640,000
(26.34)
f?¡,7; - [ (AcE)¡-t V¡ Ilx¡ + -v;- ] (cfs) (204) (221) (262)
Velocity rns-' 0.393 0.403 0.410
.,..~ Ilt Q¡Ilt rnd-I 33,955 34,819 35,424
/; = ) + [(AcE)¡-I .A .
+ (AcE)¡] V,uX" + - V. - r¡Ilt (26.35) (fps) (1.29) (1.32) (1.35)
.
 494 PARTIV Dissolved Oxygen and Pathogens LECTURa 26 Conlputet Methods 495
~-<o
.. '.'

.. The lirst step irtusirtgQtJAL2E1sto develo(j the spatfalsegm~rttÍiu.olÍ~rr~bié

for the system being morleled. This involves dividirig the' system iriio reache§ de
L, " 2 mg L-I
c, " 7.5 mg L-I constant hydrogeometric characteristics. These reaches, in tum, consist of e<juát..
a,
T,
"
"
5.787 cms
20'C
lengthcomputationalelements. . .
A segrnentation scherne used for the present case is depicterl in Fig. 26.6. Ob-
L.. " 200 mg L-I
serve that we have divided the systern into six reaches that we will call:
c.. " 2 mg L-' 1. MS-HEAD
a.. " 0.463 cms
TI' " 28'C
2. MSIOO-MS080
3. MS080-MS060
4. MS060-MS040
I
Channel slope 0.0002 5. MS040-MS020
Slde slope 2 6. MS020-MSOOO
Bottom wldth, m 10 I
I
.1 Roughness 0.035 The names for each reach are at the user's discretion and rnerely serve to iden-
L,,,5mgL-1 I
'-" c,,,9mgL-1 tify each reach. We have chosen to use the abbreviation "MS" to designate !hat
, ," a, .. 1.157 cms we are sirnulating a "main stern" of a river with no tributaries modeled explicitly.
T, .. 15'C

7
Also notice that we have used distance in kilometers to identify the extent of each
~
reach. However, we have also reversed the sense of the distances. That is, we have
rneasured distance upstream from the downstream end of the system, rather tluin ,
downstrearn as in the original problem staternent. We have made this modificatiori

/ to rnake our scheme consistent with the way in which distances must be entered into
QUAL2E.
Next, inspection of Fig. 26.6 shows that our scherne consists of 2-ktn elements.
I

Channel slope
1_
0.00018
2
The elements that rnake up the reach rnust be nurnbered in order from the headwater
to the rnost downstream point in the systern. In addition the type of each elernent
rnust be designated. There are seven element types:
I

i Slde slope
Bottom wldlh, m 10 1. Headwater elernent
, Roughness 0.035 2. Standard elernent
-------. Okm 3. Elernent just upstream from a junction
---------....... 4. JunclÍon elernent
5. Last elernent in system
FIGURE 26.5
A slream receiving BOO loadings from a poinl source and a Iribulary.
6. Input element
7. Withdrawal element
In the present exarnple we use only four of these types. The lirst (1) and last (51)
elernents are type I and 5, respectively. Elernents 2 and 22 are type 6 because they
both receive point inflows. The rernainder of the elernents are the standard type 2.
TABLE 26.4 Once the systern segmentation is delined, we can create a data lile to run
Temperatures and kinetlc coefficients QUAL2E. The lile that conforms to the present problern is shown in Fig. 26.7. Note
that the data in IbisfiJernustbe typedexactlyas shown(rninuslinesand shading,
Parameter KP > 100 KP 100-30 KP80-60 KP<60
of course). This is because QUAL2E is wriUenin FORTRAN 77.Thus each line of
T("C) 20 20.59 20.59 19.72 input corresponds to an 80-column input-card format that derives frorn the punched
0, (mgL-') 9.092 8.987 8.987 9.143 cards used in early FORTRAN prograrns.
k"W') 1.902 1.842 1.842 1.494 It should be noted that at the time of this book's printing, a user-friendly interface
k,W') 0.500 0.767 0.514 0.494
k,/ (d-") 0.500 0.514 0.514 0.494 for entering the input lile and viewing the results becarne available (Lahlou et al.
k. (d-') 0.000 0.254 0.000 0.000 1995). The interface requires precisely the sarne information as in Fig. 26.7. Thus,
SOD(gm-2d-') 0.000 5.175 0.000 0.000 whether you employ the original version or the new interface, Fig. 26.7 can serve as
a guide for perforrning the QUAL2E sirnulation outlined in this lecture.
 ...
--_.-

496 PARTIV Dissolved Oxygen and Palhogens

,-..tu.... Data.,....
o 1 3 ] . 5 fi 7 8
USfS67890U34567S90123fS678901:3)nfi1890UJ4561890123456789012JC5fi7890123f56'1890
Reach name Reach no. Element no. Element type TITLEOl EXERCISE 1, OUAL-2EU WORKSHOP: SlMPt,E CBOD/SOD
TITLE02 Steve Chdpra, May 18. 1994
TITLEO) NO CONSERVATIVE: MIN~RAL 1
TITLE04 NO CONSERVATIVE MINERAL 11
TITLEOS NO CONSERVATIVE MINERAL 11 I
M5-I1EAD 1 1 1 TITLE06 NO Tt:HPERATURE
T1TLE01 VES 8Hx.:Ht:HICAL OXYGEN DEMAND
2 6 TITLE08 NO AI.t:AE AS CHL-A IN [}GIL
3 2 TITLE09 NO PHOSPtlORUS CYCLE AS P IN MGIl. Ti t le dat.
TITLEI0 10FtGANIC-P; OISSOLVt:D-P}
4 2 TI1'LE11 NO NITROGENCYCLEAS N IN M<.>1l.
5 2 TITLEI2
TITLE]J VES
(ORGANIC-N¡ AMMONtA-N; NITRI1.t.:-N;'
DISSOINED OXYGEN IN MG/L
NI'fRATE-NI

2 TITLE14 NO FECAL COLIFORH IN NO./lOO ML

MS10G-MS080 2 I7
6
2 TITLE) 5 00 ARBITRARY NON-CONSE~VATIVE
ENDTITLE
8 2 NO LIST DATA INPUT progr_ control
9 2 NO WRITE OPTIONAL SUI1MARY
NO FLOW AUGMENTATlON
(dat.. t~ U

10 2 STEADY STA1'E
TRAPEZClIDAL CHANNELS
1 2 NO PlI:tNT I.cD/SOLAR DATA
12 2 NO PLOT DO AND BOO
FIXEO DNSTM (,.'ONC t VES; 1J o. !>p-Ul:r BOl:. O.2S
13 2 INPU1' METRIC 1. OUTl'UT MF.Ti<1\ 1.
14 2 NUMBER OF RI::ACIIES .. NUMBER OF "I¡NI, 1t,I~S O.

15 2
NUM OF HEADWATERS 1. NUMBER OF PO Htf LOADS 2.
TIME STEP IHOIJRSJ O. LNTH. COMI'. ELEMENT I KMI 2.
MS080-MS060 3 2 MAXIMUH ROlrl'E TIME (HRSI lO. TIME tNe. f'OH R1'T2 IBHSI
116 LATI'fUDE OF BASIN (CEe) OO. LONGITUDE Of' BASIN (DEGI OO.
17 2 STANDARD MERIDIAN IDEG) OO. OAY OF YEAR START TIME o.
18 2 EVAP. COEF., IAE) 0.0000000 EVAP. cm:F., (8t:J . 0000000
ELEV. OF BASIN IMI:.'TERS) O. PUST AT'rENUATI0N COEt.. 0.00
19 2 ENOATA1
20 2 ~
INDATAl'
-- -.-- - 21 2 STREAM REACH 1. RCH; MS-HEAD PR" 102.0 1'0 100.0 "acb ld.atif1catioa
22 6 STREAM REACH
STREAM REACH
2.
3.
RCU..
RC'H;
MSIOO-MS080
MS080-M5060
YR"
PR"
100.0
80.0
1'0
1'0
80.0
60.0
and rlv.r
NUa/lr.l1088t.r data
23 2 STHEAH REACH 4. RCU; MS060-MS040 t'ROM 60.0 1'0 40.0 (dat.a t.ype, 2.
24 2 STREAH REACH 5. RCH; MS040-M5020 FROH
f'ROH
40.0
20.0
1'0
1'0
20.0
0.0
STREAM REACH b. RCR: MS020-MSOOO
25 2 ENDATA2
MS06G-MS040 4 126 2 iiffiATi3'
FLAG FIELD RCH.. l. 1. l. CC8pUt..tion.l
27 2 f'LAG FIELO RCU.. 2. 10. 6.2.2.2.2.2.2.2.2.2. al_at.
28 2 Ft.AG FIELO RCH.. ). 10. 2.2.2.2.2.2.2.2.2.2. "lag ti.ld d.ta
2 FLAG FIELO RCU.. .. 10. 6.2.2.2.2.2.2.2.2.2. (data t~ .,
29 FLAG f'lEI.D RCH.. 5. 10. 2.2.2.2.2.2.2.2.2.2.
30 2 FI.AG FIEI.O RCU.. 6. 10. 2.2.2.2.2.2.2.2.2.5.
ENDA"!'A.
31 2 HYORAULICS RCII.. ). 0.00 2.0 2.0 10. .0002 .ti}';> Hydraulic8 data
32 :2 HYORAULICS RCH: 2. 0.00 2.0 2.0 10. .0002 .035 (data type S)
2 HYDRAULICS RCII; 3. 0.00 2.0 2.0 10. .0002 .035
33 HYDRAULICS RCU.. 4. v.OO 2.0 2.0 10. .000111 .0]5
34 2 HYORAUL.lCS RCH: 5. 0.00 2.0 2.0 10. .00018 .0)5
HYDRAUI,lCS RCH" b. 0.00 2.0 2.0 10. .00018 .035
35 2 ENOATA5

r
MS04G-MS020 5 136 2 ENDATASA
37 2 REACTCOEF RCH.. l. 0.00 0.000 0.000 l. 0.000 0.0000 0.0000 BODaD4 DO
38 2 REACTCOEF RCH.. 2. 0.50 0.250 5.000 3. 0.000 o. 0000 0.0000 r.actloa
REACTCOEF RCH.. 3. 0.50 0.000 0.000 3- 0.000 0.0000 0.0000 aat. coa8tant8
39 2 REACTCOEFRCH.. .. 0.50 0.000
0.000
0.000 3- 0.000 0.0000 0.0000 Idata type6)
40 2 REACTCOEFRCII.. ,. 0.50 0.000 3. 0.000 0.0000 0.0000
REACTcot.:t. RCII.. .. 0.50 0.000 0.000 3- 0.000 0.0000 0.0000
41 2
42 :2 ENDATMA
!:NDATA6S
43 2 INlTlAL COND-1RCH.. 1. 22.00 8.11 0.0 0.00 0.00 0.00 0.000 0.0 ¡alUd condlUor...l
44 2 INlTlAL COND-l RCII.. 2. 20.'59 8.11 0.0 0.00 0.00 0.00 0.000 0.0 148t.. type '. I
INn'IAL COtm-l RCH.. 3- 20.59 8.11 0.0 0.00 0.00 0.00 0.000 0.0
45 2 lNITIAL CONI)-1 RCH", .. 19.12 8.11 0.0 0.00 0.00 0.00 0.000 0.0
2 INITIAL COND-! RCH.. ,. 19.12 8.11 0.0 0.00 0.00 0.00 0.000 0.0
MS02G-MSOOO 6 146
47 2
lNt'("tAL COND-) RCH"
NDATA1
.. 19."12 8.11 0.0 0.00 0.00 0.00 0.000 0.0

48 2 F'.NDATA7A
INCR INFLOW.l RCH; l. 0.000 00.00 0.0 0.0 0.0 0.0 0.0 0.0 O. Iocr...nt.al intlow-l
49 2 INCR INFt.OW-l RCII: 2. 0.000 00.00 0.0 0.0 0.0 0.0 0.0 0.0 O. (data type 8)
50 2 INCR INFLOW-I RCII: 3- 0.000 00.00 0.0 0.0 0.0 0.0 0.0 0.0 o.

51 5 INCR INFI.QW-I RCH... .. 0.000 00.00 0.0 0.0 !:l.O 0.0 0.0 0.0 o.
INCR tNFLQW-) RCH; ,. 0.000 00.00 0.0 0.0 0.0 0.0 0.0 0.0 O.
L... INCR INFI.QW-) ReH'" .. 0.000 00.00 0.0 0.0 0.0 eJ.O 0.0 0.0 O.
EN(}ATAa
FIGURE 26.6 ENnATASA
ENDA'l'A9 f)l
(J
The QUAL2E segmenlation scheme conforming lo Ihe river HEADWTR. 1 HDW.. I UPSTREAM 5.7tf10 20.0 7.50 2.U 00.0 0.0 0.0 H.adwat.r 8Oluc..-l
ENOATAI0 n "\ r') I (dat.a type 10)
shown in Fig. 26.5. ENDATA10A
POINTLD-l PTLc 1. MSO 0.00 0.463 28.0 2.00 200.0 0.0 0.0 0.0 point load-l
POIN'rLD-I PTL.. 2. MS60 0.00 1.157 15.0 9.00 '.0 0.0 0.0 0.0 ldata type, 11)
ENDATAll
¡m¡.¡:;:n-¡
BNDA'fA12
f;ND.\TAl1
TAUA

;11- FIGURE 26.7

QUAL2E inpul file.
497
 LECl'URE26 Computer Methods 499
,\ 498 PARTIV Dissolved Oxygen and Pathogens

Title data. These specify identification information for the mn and set the con- 20
stituents that are to be simulated. Notice that we have typed "YES" for BOD and
dissolved oxygen. 16

Program control (data type 1). These consist of two parts. The first defines 12
:.J
Ihe program control options. The second sets the characteristics of the stream sys- 00
E
tem configuration as well as some of Ihe geographicallmeteorological conditions for 8
lIIodcling tcmperature. Most of the information is self-explanatory. If dala is not nec-
essary it is omitted or set to zero. For example latitude is not needed because we are 4 I I

not simlllating temperature. The "MAXIMUM ROUTE TIME (HRS)" entry bears Oxygen
additional cxplanation. When performing a steady-state computation, this entry sets o I . . 100
-20 20 60
the maximum number of iterations of the numerical method. This way, if the solution
Distance (km)
does not converge it will be halted. I have found that a value of 30.0 works weIl for
thc applications I have developed with QUAL2E. However, for other systems some FIGURE 26.8
adjustment of this parameter might be necessary. Comparison 01 QUAl2E output with analytical solution.

Reach identification and river mile/kilometer data (data type 2). The cards as either the reach or element numbers. Ratber the headwaters are numbcred con-
in this group identify the stream reach system by name and river milelkilometer. The secutively (starting at 1) from the farthest upstream headwater.
latter is done by listing the stream reaches from the most upstream point to the most
downstream point in the system. Observe that the river milelkilometer data must be Polnt load-l (data type 11). This card group, one card per point source or
in descending order. withdrawal, defines the percent treatment, inflow or withdrawal, temperature, dis-
Computational elements flag field data (data type 4). This group of cards solved oxygen, BOD, and conservative mineral concentrations of each point load or
withdrawal. Note that the point load numbers are not the same as either the reach or
identifies the type of each complltational element in each reach.
element numbers. Rather the point sources or withdrawals are numbered consecu-
Hydraulics data (data type 5). Because we specified "TRAPEZOIDAL tively'(starting at 1) from the most upstream to the most downstream.
CHANNELS" in card 5 of data type 1, we use the Manning fonnula to deter-
r' n I ')1") Sb,
"? -
mine the hydrogeometric characteristics of each reach. Consequently these cards Model output. The output for this mn consists of > ~.-o '/ -;
inelude the parameters necessary for the Manning coefficient calculation (that is,
í channel slope. side slopc, roughness, etc.).
. Hydraulics summary ,
. Reaction coefficient summary
800 and DO reaction rate constants (data type 6). This group of cards in-
eludcs reach information on the BOD decay rate coefficient and settling rate, sedi-
lIIentoxygen demand, and the method 01'computing the reaeration coefficient.
.
. Water-quality variables
Dissolved oxygen data V <!,) ",
~ " ~

A plot of tbe QUAL2E output for CBOD and oxygen is shown in Fig. 26,8. The
QUAL2E results generalIy follow the analytical solution.
Initial conditions-I (data type 7). This card group, one card per reach, estab- However, notice that there is a discrepancy at KP 100. This is due to the finite-
lishes the initial values of the system for temperatllre, dissolved oxygen. BOD, and
difference approximation used in QUAL2E. Let us look at CBOD to understand
the three conservative minerals. Initial conditions for temperature must be specified
what is happening. RecaIl that in the analytical solution, a mass balance is made at
whether it is simulated or notoFor the present case, when it is not simulated. the ini-
the mixing point, "... 1:'
tial conditions are used to determine the temperature correction factors for the rate
constants. Other values can be set at zero for steady-state applications. QrLr + QwLw - (Qr + Qw)Lo = O

Incremental inflow-l (data type 8). Even though we willnot simulate incre- which can be used to calculate (Fig. 26.90)
mental inllows. these cards must be ineluded. As in our example, alI the values can
be set to zero. L. _
- 40,000(200)540,000
)
+ 500,000(2) 16.667 mg L-1

Headwater sources-l (data type 10). This card group, one card per headwa- which the~ decays downstream according to
ter, defines the lIow,temperatllre, dissolved oxygen, nOD, and conservative mineral
_0.514+0,25.1..
concentrations 01'the headwater. Note that the headwater numbers are not the same L = 16.667e 34.819
500 PARTIV Oissolved Oxygen and Pathogens LECTURE26 Computer Methods 50 I

~
-8- -
4 mg L -1 must be maintained in all segments. Assume that segments O and 3 con-
tain negligible BOO and are at saturated le veis of 00 (o, = 10 mg L -1). In addition
segment 2 has an SOO 01'0.1 g m-2 d -1 and a surface area 01' 10.000 m2.

\l
,
1 '
(8) Analytlcal mass balance (b) Numerlcal mass balance o +----
a. polnt Input

~
at polnt Input
FIGURE 26.9
Comparison 01 mixing schemes
mass balances.
at point sources lor (a) analytical and (b) QUAl2E
":1"
I!

Rlver
: t--f
2 -- ,,
,
,
Ocean
3

"
Thus at 1 km the value of BOD will have dropped to
Estuary FIGURE P26.1 '1

_~767 1000
L = 16.667e 34.819 = 16.304 mg L-1 26.2. A waste source (Qov = l cms, Lw = 25 mg L -1, 0011= 2 mg L -l. T", = 25°C) dis.
charges into a stream (Qr = 10 cms, Lr = 2 mg L -1, Or = 10 mg L-1, Tr = 15°C).
In contrast, for QUAL2E, the loading is introduced into a well-mixed element Oownstream the stream flows through a rectangular channel with roughness = 0.03,
(Fig. 26.9b). Therefore a mass balance must now inc1udethe sources and sinks for channel slope = 0.0005, bottom width = 20 m, and a side slope 01'3. Use QUAL2E to
t yolume, calculate the profiles of both BOO and oxygen downstream assuming the BOO decays
at arate of l d-I. Determine the value and the location of the maximum ti.. Ikit.
QrLr + QwLw- (Qr + Qw)Lo - (K. + K3)VL = O 26.3. Use QUAL2E to determine the profiles of BOO and dissolved oxygen for Ihe following
sea-Ievel stream:
which can be used to calculate

l..(} = 40,000(200)+
~..."'nnrr. . .."" /" 500,000(2),
,...".. ".... = 15.98mgL-I Qw= I cms Q,= I cms
Tw= 2S'C T,= 20'C
Qr= 4cms Lw= SOmgL-I L, = 10mgL-1
Thus there is a discrepancy because of the finite discretization at the mixing point.
As discussed in more general terms in Seco11.7,care should be taken to ensure that q Tr = IS'C
Lr=SmgL-1
J °w= 2 mg L-I !
, ¡ o,= S mg L-I
the element size is small enough that such errors do not have a significant impact on 0r = 7mgL-1
the application of the model for decision making.
In summary the QUAL2E model provides a convenient tool to implement oxy- H(m) 3 S 7
gen balances in stream networks. We will revisit the software in later leclures when U (mps) 0.5 0.2 0.4
we tum to the eutrophication problem in Part V.
o 30 km 60 km

FIGURE P26.3
PROBLEMS
26.1. Suppose that the following data are available for a vertically stratified estuary (Fig. 26.4. Figure P26.4 shows a river that receives a sewage treatment plant effluenl ;tt KP O and
f'26.1): a withdrawal at KP 70. At a distance of 150 km, the stream enters a much larger river.
QOI =4x 106 m3S-I Note that the channel is trapezoidal with the characteristics shown. The deoxygenation
Q32 = 2.5 X 106 Q21 = 2.5 X 106 QI3 = 6.5 x lQÓ
rate for CBOO is equal to l d-I at 20°C. For 20 km downstream from the treatment
EÓI =Om3 S-I E32 = 2.5 X 106 E21 = 1.2 X 106 Eí3 = 2.5 X 106 plant, there is a CBOO settling removal rate of 2 d-I. In addition there is a sediment
oxygen demand for Ibis reach of 4 g m-2 d-I. Assume that the Churchill reaeration
Determine the steady-state system response matrices for BOO and oxygen if the formula holds, that the stream is at sea level, and that the dispersion constant is zero.
deoxygenation and reaeration rates are 0.1 and 0.2 d-I, respectively. Use the matri- Use QUAL2E to simulate the levels of CBOO and oxygen for this case.
j:es to determine how much BOO loading can be discharged to segment l if a level of
 502 PARTIV Dissolved Oxygen and Palhogens

LECTURE 27

Pathogens

Slde slope 2
Boltom wldlh, m 10
Roughness
Cho_'._ - "",1 0.035

'~
60 km

Channel slope 0.0008

Slde slope 2
1 Boltom wldlh, m 10
Roughness 0.033

The initial impetus for water-quality conlrol arose because of concerns over water-
borne diseases. In developed countries, much has been done to diminish the threat
FIGURE P26.4 posed by disease-carrying organisms in natural waters. However, bacteria, protozoa,
A slream receiving SOD loading Irom a poinl source and a loss 01 waler
and viruses still pose problems related lo disease transmission and interference with
Ihrough a wilhdrawal. uses of water for recreation. In addition most undeveloped and developing countries
still experience great problems related to pathogens.

27.1 PATHOGENS
Contaminated water is responsible for Ihe spread of many contagious diseases. The
primary agents of these diseases are calIedpathogens. These are disease-producing
organisms that grow and multiply within the host. Some pathogens enter the human
body through the skin. More commonly they are ingested along with drinking water.
Pathogens can be divided into categories. The most common groups associated
with water polIution are summarized in Table 27.1.
1, Although Ihe organisms listed are the cause of waterborne diseases, their con-
cenlrations are often very difficult to measure. Consequenlly, as described next, in-
dicalor organisms are used lo Irack and manage palhogen problems.
r
503
el
'-----
~-_...

504 PARTIV Dissolved Oxygen and Pathogens LECfURE27 Pathognl' 505

j'
TABLE 27.1
preferable to TC. They are measured by running the standard total coliform test I
Some waterborne pathogenic organisms at an elevated temperature (44.5°C). As a general rule-of-thumb the FC is about I
Category Descrlption Specles and groups 20% of TC (Kenner 1978). However, there is a wide spread in the ratio.
Bacleria
3. Fecal streptococci (FS). These inelude several varietics of streptococci that origi- I
Microscopic, unicellular organisms Ihal Víbrio cholerae I
lack a fully-detined nueleus and conlain Salmonella
nate from humans (Streptococcusfaecalis) as well as from domesticated animals
no chlorophyll Shigella such as caule (Streptococcus bovis) and horses (Streptococcus equinus).
Legionella
Although the TC measurement has traditionally been the most widely used in- I
Viruses A large group of submicroscopic (10 Hepatitis A dicator of contamination, its use is problematic because of the presence of nonfecal
lo 25 nm) infectious agenls. They Enleroviruses
Polioviruses
coliform bacteria. Consequently emphasis is shifting more to fecal coliforms and -
are composed of a protei n sheath
surrounding a nueleic acid core ando Echoviruses fecal streptococci.
thus. conlain all Ihe infonnalion Coxsackieviruses Further, the ratio of FC to FS (FCIFS) has been used to determine whether con-
required for Iheir own reproduclion. Rotaviruses tamination is due to human or animal sourceS. In general an FC/FS > 4 is often
However. Ihcy require a hosl in which taken to indicate human contamination whereas FCIFS < I is interpreted as orig-
to live
inating from other warm-blooded animals. However, the ratio should be used with
Protozoa Unicellular animals thal reproduce by Gianlia llllllbia care because of differential die-off ofFC and FS. Thus, as the distance from a sewage
tission Entll/lloeba histol)'lica outfaIl increases, the ratío can change and simplístic interpretation of the ratio could
Cryplosporidium
be misleading.
Naegleria fowleri
Helminlhs Inteslinal worrns and wormlike parasites Nematodes
(intestinal worms) Schistoso/lla haelllatobium
27.2.2 Concentrations
Algae Large group of nonvascular plants. Anabaella jlos-aquae .
Cenain species produce loxins Ihal Microc)'slis aeruginosa Per capita generation rates of indicator organisms are summarized in Table 27.2.
if consumed in large quantilies may be Aphanizomenon jlos-aquae Loading concentrations depend on the water use. For example in the United States,
harrnful
where per capita water use is high, the concentration of raw sewage is on the order
of 20 x 106TC/IOO mL. In contrast for a country like Brazil, where water con-
sumption is lower, concentrations of 200 x 106have been measured.
27.2 INDICATOR ORGANISMS Diffuse sources can also be contaminated. For example the geometric mean 01'
.~'J,
concentrations for urban runoff is about 0.3 x 106TC/IOOmL. Because the source
Because individual pathogens are usually difficult and/or expensive to measure di- of suchpollutionis animalwaste(rodents,dogs,cats,etc.),the FCIFS ratiotendsto
rectly, eIassical water-quality management and modeling has focused on the levels be low « 0.7) for such sources. In contrast combined sewer overflows (CSOs) carry
of indicator organisms. These are groups of organisms that are convenient to meas- both human wastewater and storm water. Consequently their FCIFS ratío is higher
ure and that are abundant in human and animal waste. If they are present it is as- (> 4). In addition, because of the presence of wastewater, their concentratioll tends
sumed that pathogens may also be present. to be higher(geometricmean == 6 x 106TC/lOOmL].

TABLE 27.2
27.2.1 Types
Per capita generation rates of intestinal bacteria for
There are three major types of indicator bacteria: warm-blooded animals (Metcalf and Eddy 1991)
---
1. Total coliform (TC). A large group of anaerobic, gram-negative, nonspore- TC FC FS
forming, rod-shaped bacteria that ferment lactose with gas formation within Animal (10. nlllllber capila I di) FC/FS
4S hr at 35°C. They exist in both polluted and unpolluted soils and occur in the
Human 100,000--400.000 2,000 450 4.4
feces of warm-blooded animals. Escherichia coli (or E. coli) and Aerobacter Chicken 240 620 0.4
aerogenes are common members of the group that occur in organisms and soils, Cow 5,400 31,000 0.2
respectively. Duck 1].000 18,000 0.6
2. Fecal coliform (FC). A subset of TC that come from the intestines of warm- Pig 8,900 230,000 0.04
Sheep 18,000 43,000 0.4
blooded animals. Thus, because they do not inelude soil organisms, they are
Turkey 130 1,300 0.1

I
 506 PARTIV Oissolved Oxygen and Pathogens LECl'URE27 Path()gens 507

TABLE 27.3 27.3.2 Light

Concentration standards for total and fecal coliform
The bacterialloss due to the effectof light can be representedas (Thomannand
Use TC (No. I 100 mL) FC (No. I 100 mL) Mueller1987)
Drinking waler O O kIJ¡ = al (27.4)
Shellfish 70 14
Fishing 1000-5000 100--1000 where = light decay rate (d-I)
kIJ¡
Conlaet reereation 1000-5000 100--1000 a = a proportionality constant
1 = average light energy (ly hr-I)
Finally Table 27.3 summarizes concentration standards for several water uses. Using data from Gameson and Gould (1974), Thomann and Mueller (1987) con-
Recognize that these values are much smaller than the typical concentrations of eluded that a is approximately unity.
sewage. For example the least stringent standard in the table (1000 to 5000 num- As depicted in Fig. 27.1, light extinction for a well-mixed layer of water can be
ber/loo mL for fishing and contact n~creation)is about 4 orders of magnitude less modeled as an exponential decay represented by the Beer-Lambert law
than the average United States sewage concentration of 20 X 106.
I(z) = loe-k,z (27.5)
where I(z) = light energy (ly hr-I)
27.3 BACTERIALLOSS RATE lo = surface light energy (ly hr-I)
The los s rate for total coliform bacteria can be represented as ke = an extinction coefficient (m-I)
z = depth (m)
kí, = kIJI + k¡,¡ + kb., (27.1)
Note that the extinction coefficient is a function of the quantity of particulate matter
where kí, = totalloss rate (d-I) and the color of the water. It can be related to Secchi-disk depth (m) by
kIJI = base mortality rate (d-I)
kIJ¡ = loss rate due to solar radiation (d-I)
kIJs = settling loss rate (d-I)
k
e
=~
SD
(27.6)

Note that we have marked the loss rate with a prime because it ineludes a transport
or to suspended solids, m (mg L -1), by (Di Toro et al. 1981)
mechanism: settling. Later in this lecture, when we model sediment-water interac-
lions, we will separate that effect from the actual mortality. ke = 0.55m (27.7)

The average light for the layer can then be estimated by integrating over the
27.3.1 Natural Mortality and Salinity depth, as in
The followingequationcan be usedto calculatea basemortalityrate for totalcol- - Jo11loe-k,z dz =_ lo (1 _ e-k,H) (27.8)
iforms(Mancini1978,ThomannandMueller1987): I = H keH
kIJI = (0.8 + O.006 P..)I.07T-ZO (27.2)
where P" = percent sea water. Thus this formulation assumes a freshwater loss
rate of 0.8 d--I. This freshwater loss is supplemented by a saltwater loss that is lin- Sunllghl
early dependent on salinity. Consequently the totalloss rate ranges from 0.8 d-I for J 10
fresh water to lA d-I for salt water. The totalloss is then modified to account for ~
temperature. Recall from Lec. 2 that a value of 1.07 denotes a strong temperature
I dependence (that is, a doubling of the rate for a 10°C temperature rise).
I H
If sea water is assumed to have a salinity of 30 to 35 ppt, Eq. 27.2 can also be
written as
, ------------ FIGURE 27.1
kIJI = (0.8 + 0.02S)I.07T-zO (27:3)
,t z
The exponential extinction 01 light tor a well-
where S = salinity (ppt or g L-1). mixed layer.
-- '-~ -_..-

508 PARTIV Dissolved Ox.ygen and Pathogens LECTURE27 Palhogens 509

Equation 27.8 can be substituted into Eq. 27.4 to give If the settling velocity of the particles is v, (rn d-I), the loss due to settling can then
be represented as
(27.9) V,
kbi = :'~(I - e-k,H) (27.18)
kbs = Fl'fj

27.3.3 Settling
Settling losses depend on how rnany of the organisrns are attached to particles. To
27.3.4 Total Loss Rate
model this process we rnust first discrirninate between the quantity of bactena that The totalloss rate can now be forrnedby substituting Eqs. 27.3,27.9, and 27.18 into
are free-ftoating and attached, 27.110 give
N = Nw + Np (27.10) I T-20 aJo -k H V,
kb = (0.8 + 0.02S) 1.07 + keHO - e ' ) + Fl'fj (27.19)
where Nw = concentration of bacteria that are free-ftoating (nurnber/lOOrnL) and "
Natural Lighl Senling
N p = concentration of bacteria that are attached to particles (number/l 00 rnL). mortalily
The quantity of bacteria on the particles is often expressed as a rnass-specific
concentration, r (nurnber/g). Thus the volurne-specificconcentration on the particles EXAMPLE 27.1. BACTERIA MODEL FOR A WEI.L-MIXED PONt)o A com
can be expressed as munity of 10,000 people is being developed alongside a stream. The development's
Np = 1O-4rm (27.11) wastewater has a percapita flow rate ofO.5 m3 capita-I d-I, acoliform general ion rate of
IXIOII number capita -1 d -1, and a suspended solids conlribution of 100 g capita-I d-I ,
whére m = suspended solids concentration (rng L -1) and 10-4 is included to convert A sample ofthe raw wastewater is centrifuged and it is determined that K,t = 0.05 m3 g-I .
volurne to 100 mL. It is proposed that the sewage be passed through a small artificial pond prior to
The tendency of the bacteria to attach to particles can also be represented by a discharge to the stream. The pond has the following dimensions: V = 2.4 X 104 11I3,
linear partition coefficient, H = 4 m, and As = 6 X 103 m2.
Determine how much the pond will reduce both the suspended solids and the col-
(27.12) iform levels during cal m periods. Assume that the solids settle at arate of 0.4 rn d-I and
Kd = 10-4 r
that there is no resuspension of solids when the weather is calm. AIso assume that the
pond's temperature is 25°C and the mean daily solar radiation is 350 Iy d-I.
where Kd = a partition coeflicicnt (rn3 g-I j ,10 4 is again included to convert
volume to 100 mL. Solution: First, we can determine the flow that will be contributed by tbe development,
If the rate at which the bacteria adsorb ami desOlu frorn the particles is fast, a IO,ooocapitaxO,5m3capita-1 d-I = 5OOOm3d-1
local equilibrium is assurned to occur. Equations 27.11 and 27.12 can be substituted
into Eq. 27.10 to give which means that Ihe pond's residence time will be 24.000/5000 = 4.8 d.
The inflow's and the pond's suspended solids can be caiculated as
N = Nw + KdmNw (27.13)
which can be solved for
.. _ lO,ooocapita(loogcapita-1 d-I) _ 200 L-I
min - 5000m3d-I - mg
Nw = FIVN (27.14)
m = 5000 ~:(6000) 200 = (0.67)200= 135mg L-I
where FIV= fraction of the bacteria that are free ftoating,
~. Thus the pond ren,..."~ 33% of the solids.
1 The ex.tincti'JiI ,,,d'ficient can be determincd as ke = 0.55( 135) = 74..\ I!! 1, and
F =- (27.15)
IV 1 + Kdm the fraction of the bacteria associated with the solids as

0,05(135)
Equation 27.14 can be substituted into Eq. 27.10 and solved for
Fp = I + 0.05(135) = 0.871
Np = FpN (27.16)
This value, along with the other parameters, can be substituted into Eq. 27.19 to calculate
where Fp = fraction of the bacteria that are attached,
+ O870~
oí k' = (O8) 10725-20 + 350/24 ( 1 - e-74.3(4)
b .. 74.3(4) . 4
Kdm
(27.17)
Fp = 1 + Kdm = 1.122 + 0.049 + 0.087 = 1.258 d-I
510 PARTIV Dissolved Ox ygen and Pathogens LEcrURE27 Pathogens ~11

The inflow and the pond bacteria concentrations can then be calculated as where the subscripts 1 and 2 designate the lake and sediments, respectively, and vr
and Vb = resuspension and burial velocities, respectively.
10-4 m3
At steady-state, Eq. 27.23 can be solved for
N
In
= 10,000 capita(1 x 1011 number capita-I
5OOOm3d-1
d-I)
( IOOmL ) N2 = R21NI (27.24)
= 20 x 106 number/lOO mL
where R21 = ratio of sediment to water concentration,
5000 6 6
N = __ u u 20 x 10 = 2.84 x 10 number/lOOmL
Fpv.. (27.25)
R21 =
v, + v" + k"2H2
Thlls allhollgh ¡tbout 85% of the bacteria would be removed, the loading to the river
wOllld still be vcry largc. This can b~ substituted ¡nto Eq. 27.22, which can be sol ved for

27.4 SEDIMENT-WATER INTERACTIONS

/ N
I
= q..
q, + Fl'v"+ kblHI - VrR21
N
In
(27.26)

where q, = Q/A.,.
In the foregoing analysis, bacteria were lost to the sediments via settling. For shalIow
systems the bacteria can be reintroduced into the surface water via storm events. To EXAMPLE 27.2. BACTERIA MODEL FOR A WELL.MIXED POND AND
gain insight into this mechanism, we must first reformulate Eq. 27.19 so that set- SEDlMENTS. For the pond described in Example 27.1, (a) calculate the concentra-
tling is separated from mortality, tion in the sediments dllring calm periods (that is, no resuspension). Assume that the
bacteria are confined to a thin surface sediment layer that is 5 mm thick. Also assume
Vs
k'" = k" + Fpli
(27.20) that buriallosses are negligible, the temperature of the surface layer is 15°C, and negli-
gible light reaches the sediment-water interface. (b) Determine the water concentration
where k" = mortality rate, of bacteria immediately after a storm that resuspends the top 40% of the layer (2 mm).

Sl,Ilution: (a) Bacterial death in the sediment layer can be estimated with Eq. 27.21,
(27.21)
k" = (0.8 + 0.02S)I.07T-2o + :'~(l - e-k,H) kb = (0.8)1.0715-20 +O= 0.57 d-I
Once this is donewe can writemass balancesfor the waterand sedimentsas The ratio of sediment to water concentration can be calculated as
(see Fig. 27.2)
R 0.87(0.4) 22
dN1 21 = O + O + 0.57(0.005) = 1
(27.22)
VI dt = QNin - QNI - Fl'v..A.,NI - kb VI NI + ~'rAsN2
This ratio, along with NI = 2.84 x 106number/lOOmL from Example27.1, can be
dN2 substitutedinto Eq. 27.24 to give
(27.23)
V2dt = Ff,v"A..NI - k"2V2N2- vrA"N2 - v"A.,N2
N2 = 122(2.84 X 106) = 347 X 106 number/lOO mL

Inflow Outflow Thlls the sediments have a much higher concentration of bacteria than the water.
(b) The storm loading to the water can be idealized as an impulse loading. The quantily
" of bacteria added lo the water can be computed as

V2N2 = 30 X 106 mL(347 x )06 number/lOO mL)O.4 = 4.17 x 1013

Lake
which can be normalized to the pond's volume to determine the resulting increasein
concentration,
Settllng Resuspenslon
Sediments V2N2 4.17x 1013number IOOmL = 0.174x )(j6number/lOOmL
ANI1= --v;- = 24,000x 106mL (100mL)
FIGURE 27.2
lake and underlying sediment layer.
 -.-.--

512 PARTIV Dissolved Oxygen and Palhogens IHTIIIU: 27 Pathog':lIs 513

TABLE 27.4
27.5 PROTOZOANS: GIARDIA AND CRYPTOSPORIDIUM
As reflected by this chapter, bacteria have been Ihe focus 01'most pasl efforts to
Export coefficients
--
for Cryptosporidillm
, ""
(IIanscn and Ongcrth 1991)

Orainagc arca Export coemcient

manage pathogens. However, as oUllinedin Table 27.1, Ihere are olher organisms that Basin Land use (ha) (oocysts ha- I d. ')
can create waterborne health problems. In particular IWOparasitic protozoa, Giardia
Cedar Prolected; mountainous f()rest land 31.077 7.72 x 10'
and Cryptosporidium, have received greal attenlion as drinking-water palhogens.
Snoqualmie Unproteclcd; mountainous forcslland; 97.115 1.04 x JO'
Giardia was first found by Leeuwenhoek in 1681, who discovered il in his own . (upper) heavy recrealional use
stools (Schmidt and Roberls 1977).At presenl, Gillrdillis Ihe most common disease- Snoqualmie Unprotecled; dairy farming 59,(J46 1.20 x 10'"
inducing intestinal parasite in the United Slales (Smilh and Wolfe 1980). When in- (Iower) 5.41 X 10"1
gested the organism causes giardiasis, a disease marked by a number of symploms
t Entire s'udy periodo
inc1udingdiarrhea, nausea, cramps, dehydralion, and headaches. 1 Durillg high runotT.
Cryptosporidium was first discovered around Ihe turn of the cenlury. Species
names are associated wilh the host organismo For example Cryptosporidium par- values were oblained in Arizona, which had raw values 01' 1732 and trealed values
vum is the name given 10the species found in the domestic mouse. This is also the of 489 oocysts L-l. Madore et al. (1987) reported higher values 01'5180 oocysts
major species responsible for iIIness in humans. Currently there is a dispuled number L-1 (850 to 13,7(0) and 1063oocysls L-1 (4103960) for raw and Irealcd efHuenls,
of dislinct species becausc an organism from one animal host can infecl another, respectively.
inc1udinghumans. In general the disease caused by the organism, cryptosporidiosis, Hansen and Ongerth (1991) sludied two walershcds in Washinglon Ihal were I
has similar symptoms to giardiasis and is se1f.limiting in individuals with healthy
immune systems. However, for individuals with impaired immune systems (such as
nOIimpacted by sewage. As in Table 27.4, they found a large difference in Cryp- II
tosporidillm export coefficients between the prolecleo Cedar River ano the un
those infected wilh the AIOS virus) no cure is currenlly available. protected Snoqualmie River. In aodilion, wilhin the Snoqualmie watershed Ihey Il.
Both Giardia and Cryptosporidium exist in Ihe environment in resting stages detected significant increases due to land use and seasonal effects. I
I
called cysts and oocysts, respective1y.When ingested, these can halch in the gas-
trointestinal tract to produce their growing slages, which can multiply and lead to
the associated iIInesses. Bolh cysts and oocysts range from spherical to ovoid in 27.5.2 Concentrations in Natural Waters
shape. LeChevallier and Norton (1995) report thal Gillrdill has an 8.6-¡.c.mwidth
LeChevallier et al. (1991) have measured raw drinking-water levels of surface water
(6.6 to 11.9) and a 12.3-¡.c.mlenglh (8.6 10 16.5). These dimensions are equivalent
to an effective diameter of about 10 ¡.c.m(7.5 to 14). Clyptosporidit/l1loocysts are treatment plants in the ~nited Stales and Canada. The geometric mean of detectable
smaller, with effective diameters on the order of 5 ¡.c.m(3 to 7). BOlhhave specific (69 out of 85 samples),Giardia was 2.77 cysts L-1, with levels ranging from 0.04
to 66. If the nondetectables are assigned the value of 0.04 at the lower end of the
gravities in the range from 1.05 10 l. l.
range, the geometric mean would be lowereolo about 1cyst L-l. This \Sconsislenl
with a value of about 0.9 cyst L-1 reported by Rose in Western U.S. streams ano
lakes.
27.5.1 Loadings
LeChevallier et al. (1991) also reported a geometric mean of detectable (74 out
In general it is believed Ihal elevaled Giardia leveis are moslly due to Ihe inlroduc- of 85 samples) c,yptosporidillm as 2.7 oocysts L-), with levels ranging from 0.07
tion of sewage effluents, whereas Cryptosporidium is more c10selyassociated with to 484. If the nondetectables are assigned the value of 0.07 at the lower end of the
nonpoint sources (LeChevallier et al. 1991). However, both can originate from either range, the geometric mean would be lowered to abollt 1.7 oocyst L-) . They suggest
point or nonpoint sources. that their results compare favorably with a range of 0.002 to 112oocysts L-) reponed
Sykora et al. (1991) reported that annual average geometric means of Gil/r- by other investigators.
dia in raw sewage were about 1500 cysts L-1 (with a range from about 650 to In summary,environmentalconcentrationsfor both Cryptosporidiumand
3000). Their data also indicated a significant seasonallrend with valucs from OCIO- Giardia are on the order of 1 cyst or oocysl L-), ranging from about 0.05 for
ber through January being higher (,,=2500 cysts L-1) than for the other monthsof Ihe pristine waters to 100 for polIuted systems. In general the Cryptosporidillm levels
year (,,= 1000 cysts L-1). They also suggest Ihat secondary Ireatmenl by activated tend to be somewhat higher. LeChavallier states that Cryptosporidium levels are
sludge is superior (90 to 100% removal) to trickling filters (40 to 60%). Gassmann typicalIy about 1.5 times as high as for Giardia.
and Schwartzbrod (1991) reported a range from 800 10 14,000cysts L-1 in sewage. Before proceeding I should note that cysts and oocysts can become nonviable in
Again. higher values were detected in the winter and early spring months. tbe environment. Thus only a fraction of Ihe total concentration may lead to infection.
Rose(1988) reportedaveragevaluesof Cryptosporidiumin raw and treated In general, lower viability tends to occur at high temperatures (Wickramanayake
sewage in tbe westem United States as 28.4 and 17 oocysts L-l. respectively. High. et al. 1985, deRegnier 1989),
I 514 PARTIV Oissolved Oxygen and Pathogens .L ~,I t..BdtnUí!t7 ~athogell§, 3B
27.5.3 Drlnklng-waier Treatmenl and Acceptable ~Isk Levels bewriHeüCoUhesUf.Fllc!
cyst and the Cryptosp'ori~iI¡tn oocyst.Massbaláíte~s'~hri
The EPA (1989) has ruled that an acceptable risk for giardiasis is 10-4, that is, one and bottom layers of a thermaUy stratified lak~ M . tI¡. 1/" ~'. "
-, M

infection per year per 10,000 people. To achieve such a risk it has been calculated h dce. A .
A b, , t:
that potable water should not contain more than 7 X 10-6 cyst L-1 for Giardia Ye"d( = Qc¡o- \,!Ce - V.,e ,Ce + J:,,(Ch - Ce) tiH1j
on the basis of the geometric mean for 1 year. To achieve such a risk for Cryp-
lo.rporidillm,it has been calculated that potable water should not contain more than
3 x 10- 5 cyst L- I on the basis of the geometric mean for one year (Rose et al. 1991, Vh d~h = v..eA,ce - V..hA,Ch + E:(ce - Ch)
(27.28)
Rcgli et al. 1991, LeChevallier and Norton 1995).
Figure 27.3, which summarizes all the information compiled above, indicates where the subscriptse andh representepilimnion(surface)andhypolimnion(bot-
that significant levels of treatment will be necessary to meet these objectives. Pres- tom), respectively, V = volume (m3), t = time (d), C = concentration (cysts L-I),
ent conventional filtration technology attains about 99 to 99.9% removal (2-log to A, = area of the thermocline separating the upper and lower layers (m2), and
3-log reductions). Thus raw drinking waters subject to sewage and nonpoint runoff E; = bulk diffusion coefficient for turbulent mixing across the thennocline (tn3
would require additional treatmenl. The problem is furthercomplicated because Gia- d-I). Note that different settling velocities are used for the surface and bottom lay-
rdia can be treated effectively with chloriné whereas Cryplosporidillm cannol. Thus, ers. This is done to reflect that settling should be higher in the surface layer because of
to mitigate the high cost of treatment, water-quality management may be needed in the lower viscosity of warmer surface water.
order to improve the quality of raw water supplies. Models of the sort described next
lIlay be useful in guiding such managemenl. EXAMPLE 27.3. GIARDIA AND CRYPTOSPORIDIUM SIMULATIONS. A wa-
ter supply reservoir has surface and bottom volumes of V. = 3 X 107 and Vh = 1 X
107 m3. Simulate the lake's response given the following seasonally varying parameters:
27.5.4 Giardia/Cryptosporidium Model
A simple;Uodel for Giardia and Cryplosporidium can be developed for a stratified ,.;;;
Unstratlfted RunotT Stratlfted Unstratlftecl
lake. In th~ following discussion we will use the term cyst to refer both to the Giardia Parameter (t s 120 d) (120 < t siSO d) (120 < t s 150d) (150 < t s 360 d)
T. 4 6 20 6
T. 4 6 8 6
1/1 E', 00 00 1.2 x i1J4 00
.. x tOS
!ea Q 3 15x tOS 3xtOS 3 x tOS
C...in 10 50 10 10
Cn.in 10 50 10 10
1 x 106, UJ ;;; In
Secondary v".-Cryp'o. 0.08 0.08 0.12 0.08
effluent v,..-Cryp'o. 0.08 0.08 0.08 0.08
1x104+tl J\ v".-Giardia 0.3 0.3 0.475 0.3
v".-Giardia 0.3 0.3 0.3 0.3
100+ l. - / \ "ir
Notethat the ,",Uliogvelocitiesare ba.sed00 Stokes' law.
1
Solution: Equations 27.27 and 27.28 can be integrated to give the time series displayed
0.01 in Fig. 27.4. In general the responses of the two organisms are similar in that both (1)
exhibit heightened levels during runoff and (2) have higher hypolimnion ooncentrations
1 x 1D-4 J. during summer stratification. They differ in that Cryptosporidium exhibits somewhat
!..................-Lll ................. ''''t ti 1 rlsk leveV higher concentrations. In addition the impact of spring runoff is more prolonged for Cryp-
1 x 10-6
I ...,- ....... tosporidium. Both these results are due to Cryptosporidium s lower settlin,g velocity.
log Removal

(a) Giardia (b) Cryptosporidium The previous example points toward some possible management strategies for
(cysts L-1) (oocysts L-1) meeting drinking-water standards for protozoan pollution. First, it suggests that dur-
FIGURE 27.3 ing stratified periods, water should be drawn from epilimnetic waters that are subject
Summary 01 Giardia and Crytosporidium levels in waste, natural, and drinking waters. to both flushing and higher settling velocities due to higher temperatures. Second,
516 PARTIV Dissolved Oxygen and Palhogens LECTURE27 PalhogclIs 5 17

it suggests that treatment might be managed from a seasonal perspective. That is, wasle discharge has a ftow of 100,000 m3 d-I and negligible coliform bacleria. There
increased treatment might be required during and after high ronoff events. are negligible suspended solids in the slream (Ihe eXlinction coefficient is 0.5 m-I).
The foregoing analysis should be tempered because other drinking-water re- (a) Assuming inslantaneous mixing, compule lhe lemperalure allhe discharge point.
quirements such as toxic disinfection byproducts (e.g., trihalomethane formation due (b) Assuming inslanlaneous mixing, compule Ihe coliform concenlralion allhe dis-
charge point.
to chlorination) and taste and odor problems must be concurrently managed along
(e) Compule Ihe velocily in Ihe slream below Ihe mixing poin!. Nole Ihallhe slream
with pathogen problems. In some cases these other problems might be minimized
is 20 m wide and 0.5 m deep.
by adopting similar strategies as for pathogens. However, different and contradic- (d) Compule Ihe base mortalily rale (kbl) of Ihe bacleria in Ihe slream below Ihe mix-
tory strategies might just as likely be required, e.g., taking water from the lake's ing poin!.
hypolimnion to obtain cooler water to minimize trihalomethane formation. (e) Compule Ihe lighl decay rale (kbi) of lhe bacleria in Ihe slream below Ihe mixing
In summary the management of pathogenic protozoans is complicated by a num- poin!.
ber of factors. In particular they are difficult to measure (especially their viability) (f) Compule Ihe concenlralion of the baCleria at a beach localed 10 km downslream
and the concentration range from sources to finished water spans many orders of from Ihe discharge.
magnitude. However, sufficient information has been compiled to allow the devel-
27.2. A poinl source of bacleria enlers an eSluary channel having Ihe following charac-
j
opment of simple, order-of-magnitude models of their leveis in natural waters. teristics:
Value Units
(a)Giardia
Dispersioll coefficient 10" m2d-I I
120T Fiow 5x1O' mJd I

!. 80 I Inflow Width
Deplh
200
2
m
m
I
1

! 40 The eSluary has an ambienllemperature of 27. 5°C. 11al so has a mea n daily lighl Ílllcn-
sily of 200 cal cm-2 d-I and an eXlinclion coefficienl of 0.2 m-l. Allhe poinl of Ihe
inpul the estuary is 50% sea waler. If Ihe bacleria do not sellle:
O (a) Delermine Ihe bacterial die-off rale.
O 120 240 360 (b) Whal mass loading could be ¡npullo Ihis syslem under steady-stale condiliolls if
Ihe allowable concentralion allhe oUlfall is 1000 number/mL? Express your resulls
in number d-I. Assume complele lateral and vertical mixing allhe oulfall.
(b) CryplO.
120 27.3. A lake with a residence lime of 2 monlhs (inftow = outflow) has a depth of 7 m alld a
surface area of 5 x lOS m2. The average surface solar radialion is 200 Iy d-I and Ihe
!. Inflow lemperalure is 22°C. The lake has a suspended solids concentralion of 2 mg L -1 Ihal
!>- 801 I I Hypo sellle al a rale ofO.3 m d-I. It also has an eXlinClion coefficient ofO.4 m-l. Noh: thallhe

8 bacleria associale wilh Ihe solids wilh a linear partilion coefficienl of 0.005 013 g-I . If
O Ihe 10lal concenlralion of bacteria in Ihe syslem cannol exceed 1000/100 mL, delerm ine
40 the acceplable loading rale for the lake.

27.4. For the same lake as in Probo 27.3, compule Ihe loading needed lo maintain a sedimenl
O concenlralion of 100,000/100 mL. Assume Ihallhe sedimenllayer is 10 cm Ihick, Ihal
O 120 240 360 resuspension and burial are negligible, and thallight has a negligible effecl on baclerial
Day FIGURE 27.4 die-off in the sedimenls.

27.5. ESlimale the mean and range of sellling velocities for bolh Giartlia and Cryptosporitl-
iUIIIusing Slokes' law. along wilh Ihe size alld densily inform¡uion from Seco 27.5. Base
rROBLEMS your estimale on the density and viscosily 01' fresh waler al a temperalure of 20uC.
17.1. A stream receives waslewaler from a cily wilh a populalion of 100,000 people. The per 17.6. (a) Calculate the rate of Giare/ialhal could be loaded lo a smalllake (residence lime
1 capilaloadingrale of totalcolifonnbacteriaforIhecityis 1.5x 1011numbercapila-I =I monlh, volume = 10 x 106 mJ, and mea n deplh =2 m), if the only loss
d-I and Ihe per capita ftow rate is 0.5 mJ capila-I d-I . The wastewaler and the stream mechanisms are flushing and sellling (v, = 0.4 m d-I) and Ihe desired concen-
(above Ihe discharge) have temperatures of 25 and 15°C, respectively. After mixing Iralion is 0.1 cysl L-l. Express your result as an inflow concenlralion.
!8kes place; assume that Ihe temperature in Ihe stream remains conslant. The peak (b) How much addilionallrealmenl (express as a log removal) would be required lo
solar radiation is 650 cal cm-1 d-I and the photoperiod is 13 hr. The stream above the meellhe 10-4 risk level for drinking waler?
 '"'

pART V

~uttbph lentit>t1
áhdTerripetatUre

We now tum from decomposition processes to a more complete representatlon of

the life/death cyele in the aquatic environment. The problem of eutrophication will
serve as the focus for the following set of lectures. In addition we will also cover the
rudiments ofheat budget models that are so important in thermally stratified systems.
Lecture 28 provides an introduction to the eutrophication problem along with
information on some of the nutrients that stimulate plant growth in natural waters.
Lecture 29 describes the phosphorus loading concept that was developed to provide
simple models of eutrophication in lakes.
The succeeding two lectures deal with temperature modeling because of tbe
great impact of thermal stratification on eutrophication. Lecture 30 describes how
heat enters and leaves the surface of a water body due to atmospheric interactions.
Then Lecture 31 illustrates howheat budgets and thermal stratification are simulated.
Next the interactions between nutrients and organisms are explored. Lecture
32 provides an introduction to modeling microbial growth kinetics. Tl1is material
relates to algae as well as to other organisms such as bacteria. Then the microbial
growth model is applied to algae (Lecture 33) with special emphasis placed on the
role of light in photosynthesis. Lecture 34 deals with predator-prey interactions be-
tween organisms.
Finally a total nutrient/food-chain framework is developed and applied to lalces
(Lecture 35) and streams (Lecture 36). The latter lecture ineludes inf()rmation on
how the QUAL2E model can be used to simulate both heat balances and stream
eutrophication. .

519
' ~ ...A..-_..

LECTURE 28
- - - - - -_.-
Tt1eEutrophication Problem
an~ Nutrients

J..E9TVREOYJ!RVlEW: , provide backgroundinforma~ion on the eutropJ¡jcatio~

prpblem. M'ter sn 9yervi~w of the ~ymptoms of eutrop~ica~ion, , disc!Jss ~t1e
PriWIU}' puments U,1~Jsti"uIJsJe the rrocess. , sIso descriJ>e plant stoicmometry
.

m.d th~"~*rn~en:pt¡OSpbpfUS
. . ratio.

Several limes ayear I fertilize my lawn and garden so I can grow green grass and 11
10lsof fresh vegelables. In a similar way Ihe addition of nutrienls to a natural water I

slimulales plant growth. In small quanlities Ihis can be a good thing. For example
because of Ihe higher human population of ils drainage basin, Lake Michigan re-
ceives more nulrienls Ihan Lake Superior. Consequenlly Lake Michigan has more
planl growth, which ullimalely allows il lo support more game fish.
However, as with mosl everylhing in life, there can always be 100much of a good
Ihing. When lakes, streams, and estuaries are overfertilized, Ihe resulting excessive
plant growth can become a serious water-qualily problem.
This "overfertilizalion" phenomenon is generally referred lo as eutropllication.
This lerminology was originally coined to describe Ihe natural aging process
whereby a lake is transformed from a lake lo a !rtí\fsR'ro'a ~'éJ8'8"w.Thb process
can take Ihousands of years lo occur naturally. However, Ihe process can 1M:greatly
quickened by excess nUlrients from human aClivities. This accelerated process is
sometimes called cultural elltrop¡';catioll.
Waler bodies are often classilied as lo Iheir Irophic Slate.The generallerms are
() .' ..1
. Oligolrophic(poorlynourished)

..
. Mesolrophic (moderately nourished)
Eutrophic (well-nourished)
Hypereulrophic (ovemourished)
Although such terms were originally developed for and are mosl commonly applied
lo lakes, Ihey are also appropriate descriplors of slreams and eSluaries.

521

- - ,~-.~.
...~
 522 PARTv Eutrophication and Temperature LECTURE
28 The Eutrophication Problem and Nulrlehls 123

28.1 THE EUTROPHICATION PROBLEM phosphorus, serves to remove phosphorus from tbe water to tbe bottom sediments.
In general, eutrophication can have a number of deleterious effects on water bodies. For cases where the water in contact with the sediments contains oxygen, such
These inelude: sediment phosphorus becomes chemically trapped.

· Quantity. The profuse growth of fln"tiJlg p"'n's clp.creases water elarity,

species form unsightly scums. Further, certain floating "Ian's c"n clog filters at
and some Although phosphorus is naturally scarce, many human activities result in pHos-
phorus discharge to natural waters. Human and animal wastes both contain substan-
~~ 'rp.:I'mentplants, and oyerJ!.rowthof rooted plants can hinder navigation :lncl- tial amounts of phosphorus. In the recent past the former has been supplemented by

· recrea,ion hy clogginJ!.waterway.s..
Cltemistry. Plant growth and respiration can affect the system's water chemistry.
Most notably, ox en and carbon dioxide levels are directl im acted b lan
~c!!.vity.O~XJ!.enhas implications re ated to the survival of omanisms sllch :lSfisb.
In particular the bottom waters of thermally stratified systems can become totally
devoid of oxygen due to the decomposition of dead plants. ..c.adwulioxide-eatt
detergent phosphorus. In addition nonpoint sources from agricultural and urban land
both conlribute excess phosphorus. Part of the enhancement of diffuse sources is
due to fertilizers and other phosphorus-containing chemicals associated with human
land use. Moreover, human uses lead to soil erosion, which also enhances phosphorus
transport into waters.
Phosphorus in natural waters can be subdivided in several ways. One scheme,

·_impact
Biology.
pH.
Eutrophication can alter the species compÓsition of an ecosystem. Na-
tive biota may be displaced as the environment becomes more productive. Certaln
which stems from conventional measurement techniques and modeling necessity, is
(Fig. 28.1)

· Soluble reactive plto.~pltorus(SRP). Also called orthophosphate or soluble inor-

sp.eciesof algae cause taste and odor roblems in drinkin water. Further, certain ganic P, this is the form that is readily available to plants. It consists of the species
blue-green algae can be toxic when consume y animals. Many of these problems H2P04 -, HP042- , and P043-.
I become prominent as the water body becomes more eutrophic. · Particulate organic P. This form mainly consists of living plants, animals, and
Now that we have a feeling for the types of problems caused by eutrophication, bacteria as well as organic detrilus.
I

I we must figure out how it works. The first step in this process involves the inorganic · Nonparticulate organic P. These are dissolved or colloidal organic compounds
i nutrients that serve as the raw materials from which plant biomass is synthesized. containing phosphorus. Their primary origin is the decomposition of particulate

I 28.2 NUTRIENTS
· organic P.
Particulate inorganic P. This category consists of phosphate minerals (e.g., ap-
atite phosphorus), sorbed orthophosphate (e.g., on elays), and phosphate com-
I plexed with solid matter (e.g., calcium carbonate precipitates or iron hydroxides).
Inorganic nutrients provide chemical building blocks for life in aquatic systems. · Nonparticulate inorganic. This group ineludes condensed phosphates such as
I
I Some are required in large quantities for cell development and hence are caBed those found in detergenls.
I macronuliients. These are carbon, oxygen, nitrogen, phosphorus, sulfur, silicii, and
iron. Smaller quantities of micronutrients, such as manganese, copper, and zinc, are
also necessary. Water-quality modeling has focused on four macronutrients: phos-
I
phorus, nitrogen, carbon, and silica.
ORGANIC

28.2.1 Phosphorus
Phosphorus is essential to all life. Among other functions, it has a critical role in t
genetic systems and in the storage and transfer of cell energy. INORGANIC
From a water-quality perspective, phosphorus is important because it is usually
in short supply relative to the other macronutrients. This scarcity is due to three
primary factors:
.~ UNAVAILABLE
l. It is nol abundant in the earth's crusl. Further, the phosphate minerals that do exist
are nol very soluble. NONPARTICULATE 4. PARTlCULATE

2. It does not exist in a gaseous formoThus. in contrast to carbon and nitrogen, there FIGURE 28.1
is no gaseous almospheric source. Forms 01 phosphorus lound in natural waters. The principal
J. Finally phosphate lends lo sorb slrongly to fine-grained partieles. The settling lorms involved in Ihe production/decomposition life cycle are
shown in bold.
(Ir ,hese parlicles. along wilh sedimenlalion (Ir organic partieles containing
~ ,.. ~.. - - ~

LECfURE28 The Eutrophication Problem and Nutrients 525

524 PARTv Eutrophication and Temperature

Atmolpheric ,lI,lmospheric Iosses

Phosphorus is measured in several ways. Soluble reactive phosphorus (SRP) is nltrogen al high pH
. .
measured by adding ammonium molybdate, which forms a colored colnplex with . A:
the phosphate. Organic phosphorus is not rletected by the test unless it is first hy-
~ -----..-...-.----- .:~
drolyzed, that is, converted to orthophospll<tte by digestion with heat and a strong
acid. In addition filtering can be used to divide between particulate and nonpartic- I :

ulate components. Finally application of the SRP method to a digested, unfiltered
sample plOvides a means to measure all the phosphorus in a sample. As we will see
later in this lecture and the next, such a total phosphorus (TP) measurement has
been used widely to quantify eutrophication.
As mentioned above, the partitioning in Fig. 28.1 is based on available mea-
surement techniques and modeling necessity. The distinction between particulate
and nonparticulate forms is made so that the former can selectively be removed by
settling. The division of available phosphorus from the other speciesis made because
it is the only form that is directly available for plant growth. It should be understood
that the other forms are not absolutely "unavailable." Rather, they must first be con-
verted to SRP bcfore they can be consumed by plants.
It also should be noted that Fig. 28.1 is not the last word regarding phospho-
rus segmentation. In fact from a strictly scientific basis, each of the compartments
could be broken down into finer detail. Conversely from the perspective of water-
quality modeling, some of the compartments are often consolidated whereas others
are broken down into finer segments. For example the distinction between inorganic
and organic unavailable forms is not usually made. That is, they are lumped into
unavailable particulate P and unavailable nonparticulate groupings. An example of
more refinement is that living particulate organic P is often distinguished from non-
living forms. In other words, groups of organisms (such as phytoplankton, zooplank-
ton, etc.) are modeled separately. In such casesthe amount of phosphorus contained
in these groups would have to be subtracted from the particulate P compartment. We
retum to this topic when we build models in a later lecture.
28.2.2 Nitrogen
I havealreadydescribedthe nitrogen'cyele in Lec. 23. Recall that the primary
forms are
.
. Free nitrogen (N2)
Ammonium (NH4 +)/ammonia (NH3)
. Nitrite(N02- )/nitrate(N03-)
. Organic nitrogen
As depicted in Fig. 28.2, the organic nitrogen can be broken down further into par-
ticulate and dissolved components.
Some of the major processes governingthe dynamics of these groups are
. Ammonia and nitrate assimilation. This ineludes the uptake of inorganic nitro-
gen by phytoplankton. Although phytoplankton utilize both ammonia and nitrate,
their preference for the former has been demonstrated (Harvey 1955, Walsh and
Dugdale 1972, Bates 1976).
ORGANIC
t
INORGANIC
Free
nitrogen AVAILABLE UNAVAILABLE
NONPARTICULATE PARTlCULATE
FIGURE 28.2
Forms 01 nitrogen lound in natural waters. The principal lorms ¡nvolved in the
production/decomposition lile cycle are shown in bold.
. Ammollification. This is the transformation of organic nitrogen to ammonia. This
is a complicated process involving several mechanisms, ineluding bacterial de-
composition, zooplankton excretion, and direct autolysis after cell death.
. Nitrification. 'Ibis is the oxidation of ammonia to nitrite and nitrite to nitrate via
the action of a kct group of aerobic bacteria. This process utilizes oxygen and
typically is represented by first-order reactions. The fact that the transformation
of nitrite to nitrate is relatively fast supports the use of a single functional group to
encompass both nitrate and nitrite in some nutrient/food-chain models. For these
cases the nitrification process would be represented kinetically by the oxidation of
ammonia.
. Denitrification. Under anaerobic condition~-.-for example in the se,lillll'nts <lJId
the anoxic hypolimnia of some lakes-nitrate can serve as an electron acceptor
for certain bacteria. Nitrite is formed as an intermediate, with the principal end
product being free nitrogen.
.
Nitrogenfixation. A number of organisms can fix elemental nitrogen. An impor-
tant group from the standpoint of nutrient/food-chain modeling is blue-green al-
gae possessing heterocysts. These organisms are important because, for lakes with
high phosphorus loadings, phytoplankton growth can depress nitrogenlevels to the
point where nonfixing algae will become nitrogen-limiled. The ability ofthe blue-
green algae to utilize free nitrogen gives Ihem a competitive advanlage in such
situations. The resulting dominance by the blue-greens has implications lo water
quality since many species have objectionable characteristics; for example Ihey
form ftoating scums.
Although nitrogen isjust as necessary for life as phosphorus, they differ in three
ways:

t
526 PART v Eutrophication and Temperature LECTURE28 The Eutrophication Problem and Nutrients 527

· Nitrogell ha.f a gas phase. Further, as mentioned above, certain blue-green algae
are capable of fixing free nitrogen. This gives them a competitive advantage in
Atmoopheric
carbon dloxlde

situations where other forros of nitrogen are in short supply. This state of affairs
can sometimes occur when advanced treatment ineludes nitrogen removal. In such

· cases blue-green algae can become dominant.

/lIorgallic lorms olllitrogell do lIot sorb as strongly to particulate matter as does
phosphorus. Consequently, although particulate forms of nitrogen are carried to
the sediments by settling, they are more easily introduced back into the water. ORGANIC
In addition inorganic forms of nitrogen (particularIy nitrate) are more mobile in

· groundwater.
Dellitrification represents a purgillg mechanism that does not occur lor phospho-
rus. Because it occurs only in the absence of oxygen, denitrification is insignifi-
cant for many surface waters. However, for productive systems where denitrifica-
tion can occur in anoxic sediments, a deficiency of nitrogen can be created.
As with phosphorus, nitrogen discharges to natural water result from human
activities. Human and animal wastes both contain substantial amounts of nitrogen.
In addition nonpoint sources from agricultural and urban land both contribute ex-
cess nitrogen. As mentioned above, because forros such as nitrate do not associate
strongly with solid matter, they can be easily transmitted to surface waters along
with groundwater flow.
Most of the aforementioned characteristics mean that phosphorus has usually
been identified as the primary controllable nutrient governing the eutrophication pro-
cess in fresh waters. However, productive estuaries can tend to be nitrogen limited.
28.2.3 Carbon
Carboncan playthreerolesin water-qualitymodeling:
· Nutrient. In the same way as phosphorus and nitrogen, carbon can be thought of
as a nutrient. However, in most cases modelers usually assume that carbon c:mnot
limit algal growth, although some models (Chen 1970, Chen and OrIob 1975) do
allow for potentiallimitation. Studies dealing with carbon-limited algal cultures
raise questions regarding the forro of carbon needed for photosynthesis by differ-
ent algae (Goldman et al. 1974, King and Novak 1974). It also has been suggested
that the relative abilities of green and blue-green algae to use various forms of in-
organic carbon could partially explain the succession from greens to blue-greens
in enriched natural waters (King 1972, Shapiro 1973). These and other develop-
ments (see references in Goldman et al. 1974) indicate that control of primary
production by carbon limitation could be important in certain systems.
· Bioma.fS. Because it usually constitutes a large component of organic compounds,
· carbonis oftenusedas a measureof biomass.
Pollutant. Finally carbon is an important factor in other pollution problems be-
yond eutrophication. First, as already discussed in great detail in Part IV, the de-
---{
r=l INORGANIC
AVAILABLE ~ _1-+ UNAVAILABLE
T
NONPARTICULATE 1.""'" PARTtCULATE
FIGURE 28.3
Forms 01carbonlound in naturalwaters. The principalforms
¡nvolvedin the production/decomposition
lile cycle are shownin
bold.
the generation, transport, and fate of organic carbono Finally naturally produced
organic carbon can, itself, be transformed into a toxic compound. For example
chlorine can react with organic compounds to forro toxic trihalomethanes.
Figure 28.3 is an effort to represent the carbon cyele in natural waters. Note
that the dissolved inorganic carbon compartment actually consists of several species:
carbon dioxide (C02), bicarbonate (HC03 -), and carbonate (C03 -).
28.2.4 Silicon
Although it might be considered a minor nutrient, silicon has significance in the
dynamics of phytoplankton because of its importance as a major structural element
in the cells of an important phytoplankton group-the diatoms. These organisms use
dissolved reactive silicon [mainly as Si(OH)4] to build afrustule or "glass wall" that
surrounds the cell. Silicon in frustules is not available to other diatoms, and ambient
concentrations of available silicon can therefore become low enough to limit further
growth of these algae.
In most modeling work to date, silicon has not been simulated. Where it has been
ineluded, silicon has been treated as one compartment, such as dissolved inorganic
silicon (Lehman et al. 1975), or two compartments, such as available silicon and
unavailable silicon (Scavia 1980).

composition of organic carbon can greatly affect a system's oxygen concentration. 28.3 PLANT STOICHIOMETRV
Second, it is known that many toxicants preferentially associate with organic mat-
ter. Thus the dynamics of toxics in the environment is intimately connected with Aside from nutrients, the other key part of the eutrophication process is the food
chain. As depicted previously in Fig. 19.1,the exchange between the two components
. .. .~.IL.ÍII ,ó\5'" __.~ ~ \ _,..__
_._-- --_o

528 PARTv Eutrophication and Temperature lECTURE28 The Eutrophication Problem and Nutrients 529

represents a cycJe. That is, production converts inorganic nutrients into or~anic mat-
ter whereas decomposition reverses the process.
An important factor in the process is the stoichiometric composition of organic
matter. AJthough the composition varies, the dry-weightt composition can be ide-
alized as in the folIowing detailed representation of the photosynthesis/respiration
process (Stumm and Morgan 1981):* N
Oth er .
7 20;.0 P

w
,/ 0;.o
106C02+ 16NfLt+ + HPol- + 1081'hO;::CHJ6H2630IloNI6PI
+ 10702+ 14H+
51.8%, ,
..

"Algae" ~ . :C
(28.1) ~ .
...
,..
,
40"""

This formula can be used to determine the mass ratios of carbon to nitrogen to
phosptlorus:
FIGURE 28.4
C : N : P
Pie diagram showing the percentages 01 nutrients and water constituting
106x 12: 16 x 14: 1 x 31 (28.2) the average phytoplankton biomass.
1272 : 224 : 31
EXAMPLE 28.1. PHYTOPLANKTON STOICHIOMETRY. Suppose Ihat a lake
It is also known that plant protoplasm is about 1% phosphorus on a dry-weight basis. has a volume of 1 x 106 m3 and a phytoplanklon concentration of L -) of chloro-~
Therefore we can normalize the ratios to the mass of phosphorus and express the phyll a. If the carbon-to-chlorophyll ratio is 21 uf(,hL~flgC-1 and all the other stoi-
chiometry follows Fig. 28.4: (a) Reexpress the phytoplankton concentralion as organie
results as percentages of dry w~i~~t, carbono (b) Iflhe phytoplankton are decomposing at arate ofO.1 d-I, what is the result-
C : N : P
(28.3) ing rate of oxygen demand in g m -3 d-I? (c) What is the rate of release of nitrogen and
40%: 7.2%: 1% phosphorus in g d-I?

Solution:
Th4s a gram dry weight of org~atter contain; approxim~tely 10J!1Eof phos-
phorus, 72 mg of nitrogen and 400_mg of carbono FinalIy it should be noted that (a) The phytoplankton concentration as organic carbon can be ca1culated as
dry-weight biomass has a density of about 1.27 g cm-3 and wet-weight biomass is
gC
abOuf90% water. Figure 28.4 summarizes the fundamental information regarding
ceII stoichiometry. Additional information, particularly regarding the variability of
JOmgChla
m3 (
25 mgChla )= OAO 'C m-3
g
tbese numbers, can be found in references such as Bowie et al. (1985). (b) The decomposition of a gram of organic c bon ulilizes 2.67 of ox 'el (recall
Although the foregoing information provides a way to break down biomass
Eq. I . us e amo n o oxygen consumed can be ea cu aled as .r. '>........
into its individual components, additional information is required because phyto- ,'QL-. '"Jf::."7
gO o.l gC -3-1
plankton are often measured in units other.than dry weight. In cases where organic
carbon is used, the foregoing stoichiometric information can be employed directly. ()
r""kdc = 2.67gC d 0.40m3 = 0.1068gO m d
More commonly the phytoplankton are measured as chlorophylI a. In general the yo -;: ,~
(c) The phosphorusgeneralionratecan be calculatedby O ,'<'P.:;.C..:"\\T::l
chlorophylI-to-carbon ratio ranges from 10 to 50 J,LgChlmgC-I. The lower value
~ I X 106 m3 mgChla I gP
is usualIy typical of welI-iIIuminated waters such as oligotrophic systems. For such
systems less chlorophylI is required because of the high solar radiation. In contrast,
a ,kdVa
P<
= J mgP
mgChlu () d (
10 m3 )( 1()()()mgP )
= 1000
g
Pd-I

waters with less light, such as eutrophic and turbid systems, would tend to have The nitrogen-Io-ehlorophyll ralio can be compuled as
phytoplankton with a higher chlorophylI content. 1...\\\0 -t, ¿ ..--
f'l\':::l~
mgP mgN mgN '(~.:b
a"" = (
1mgChla 7.2 mgP )= 7.2 mgChla
which can then be employed to determine the nilrogen generation rale,
'Dry weigbt means Ihe weight of the organie malter after it has been dehydrated.
'This fOflllula holds when ammolllum is Ihe source of inorganie nitcogen. for the ease where nitrate is
the source. ¡/If reaetion is modified. a..kdVa = 7.2m:~~a(o(nl x Jü6m3(lOmg~:la)C~g~gN)= nOOgNd-1
106C~ + t6NO¡ - + HPOl- + 122H20 + 18W :;:::::C'IX>H26JO"oN,.P, + 13802
 LECTURE28 The Eutrophication Problerh and NutrleittS ~:H
}I 530 PARTv Eutrophication and Temperature

I
tAOLE 28.1
Amounts of nltrogen and phosphorus In untreated domestic
sewage In the Unlted States. Numbers In itarentheses
represent ranges (from Metcalf and Eddy 1991, Thomalíri
and Mueller 1987)

Per-caplta
Concentratlon loadlng rate
Nutrient (mg L-1) (g caplta-I d-I)!
FIGURE 28.5 Nitrogen 40 23
Pie diagram showing the (20-85)
percentages 01 nutrients on a Organic N 15 8.5
dry-weight basis lor diatoms (8-35)
with diflerent silicon content Free ammonia 25 14.2
SI SI in their .glass. cell walls or (12-50)
20% 50% Irustules.
Phosphorus (with detergents) 8 4.5
'(4-15)
Diatoms differ frorn otber forms of pbytoplankton in tbat silicon makes up a Organic P 3 1.7
large fraction of their biomass. The percentage of ury weight ranges from 20 to 50% (1-5)
Inorganic P 5 2.8
depending on the structure of tbeir frustule. It is usually assumed that the remainder
(3-10)
of tbeir dry-weight biomass follows the proportions in Eq. 28.4. Figure 28.5 shows
how the dry-weight percentages would decrease as more of tbe total dry weight is Phosphorus (without detergents) 4 2.3
dedicated to silica.
t 8ased 00 a per-capila flow cate orO.57 mJ capita-' d-I (150 gal capita-I dO').

28.4 NITROGEN AND PHOSPHORUS TABLE 28.2

Because tbey are the primary controllable nutrients, nitrogen and phosphorus have Nitrogen and phosphorus export rates (kg ha-I yr-I) generated
lJeen the focus of most efforts to control eutrophication. from various nonpolnt sources In the United States (numbers in
~
parentheses represent ranges)
Nutrlent Forest Agricultural Vrban Atmospherlc
28.4.1 Nitrogen and Phosphorus Sources 3 5 5 24
Nitrogen
Point sources of nitrogen and phosphorus are summarized in Table 28. l. In general ( 1.3-10.2) (0.5-50) (1-20)
the use of phosphate detergents has had a great impact on the amount of phosphorus Phosphorus 0.4 0.5 1 j
in wastewater. (0.01-0.9) (0.1-5) (0.1-10) (0.05-5)
Some typical nonpoint sources are summarized in Table 28.2. Notice that urban
and agricultural use greatly increase the export ofboth nitrogen and phosphorus from
Ibe land. 7.2. Hence an N:P ratio in tbe water tbat is less than 7.2 suggests that nitrogen,is
limiting. Conversely, higher levels imply that pbospborus wilt limit plant growth.
.'" The rationale bebind the ratio is derived in the following example.
28.4.2 N:P Ratlo
* EXAMPLE 28.2. N:P RATlOS. Suppose Ihat a batch reactor has an initial concen-
We now know that as plants grow they take up inorganic nutrients from tbe water in tration of algae of I J-tgChl L -l. If the plants are growing according lo first-order Id-
proportion to their stoicbiornetry. In eutrophication management it is often important netics (kg = I d-I), determinehow bothplant and nutrientconcentrationsevolvefor
to identify wbicb of the several nutrients used for planl nutrition ultimately controls initial nutnent leveIS of (a) no = 100J-tg~L-1 and Po = 10J-t.E!' L-1, and (b) ~=-
7~-1 and Po = 36 p.g'P L l.
the level of plants in the water body. A first cut at identifying this "limiting nutrient"
is to compare tbe levels of the nutrienls in the waler with the ceil stoichiometry. Solution: Mass balances for algae, phosphorus, and nitrogen can be written as
This is most commonly done for nilrogen and phosphorus. A rough rule of da dp dn
thumb for assessing whicb nutrienl is lirniting relates to the nitrogen-to-phosphorus dt = kga dt = -apakga dt = -a.akga
ratio. Recall tbat the ratio of nitrogcn lo phosphofUSin biomass is approximately
. ~~t.;..""".""~F'~ _~ ",\., \. .... .-.

532 PARTv Eutrophication and Ternperature LECTURF 28 The Eutrophication Problern and Nutrienls 533

(a) Similarly estuaries tend to be l1. nt in nitrogen and hence are usually nitrogen
limited. In contrast those systems subject to pbosphorus removal and llonpoint-
100

- 80
60
oo. phosphorus
limited
source input are generally phospborus limited.
Although Table 28.3 provides general patterns, individual natural waters must
be assessed on a case-by-case basis. Consequently mathematical models have been
$.. 1& developed to simulate the eutrophication process.
:~ fAl.....
t:J
PROBLEMS
(b) 28.1. A sewage treatment pond has the following characteristics:
100

- 80
60
oo. nitrogen
limited
Residence time = 3 wk
Area = I x 10:;m2
Mean depth = 2m
$.. 40
FIGURE 28.6 The inflow concentrations of nilcogen and phosphorus are 50 and 5 rng L -1, respec-
20 . fAl .....
~.uu~A Algal growth in a batch reactor starting with tively. These inputs result in a level 01'200 mgChla m-3 of phytoplankton in the rondo
initial N:P ratios that lead to (a) phosphorUs The phytoplankton are removed by settling at arate of 0.5 m d-I. AIso, Ihe nilcogen is
2 3 limitation (initial N:P = 10) and (b) nitrogen removed by denilrification at a rale of 0.05 d-I . Compule the following:
Time limitation (initial N:P = 2). (a) The steady-state nitrogen and phosphorus concentrations in the pond
(b) Based on the nitrogen-to-phosphorus ratio, which is the lirniting nutrient?
which can be sol ved for , ~
Note that Ihe chlorophyll-to-phosphorus ratio in Ihe phytoplankton is l.
a = aoekgt p = po - apaao(ekgt - 1) n = no - aa"ao(ekgt- 1) 28.2. A stretch ofa river(U = 1 m S-I) has a uniform, nel photosynthesis rate of 10 gO m-2
(a) The solutions for the first case are depicted in Fig. 28.6a. Because it is in relatively d-I. The boundary conditions 01'inorganic nitrogen and phosphorus at the head end 01'
sbort supply (N:P = 10 > 7.2), the pbospborus runs out first at approxirnately t = 2.5 d. the stretch are 20 mgN L -1 and 4 mgP L -1, respectively. If plant activity is the only l
At tbis point tbe algae can no longer grow and excess nitrogen rernains in tbe water.
(b) The second case, where nitrogen is in short supply (N:P = 2 < 7.2), is depicted in
significant source or sink 01' nutrients, determine and plot the N:P ratio for the stretch
until one of the nutrients runs out. ~I
Fig. 28.6b. For tbis case tbe nitrogen runs out first. Again, at tbis point, tbe algae can
28.3. Right al'ter tbermal stralification is established, a lake has available phosphorus and
'f~'.
~
no 10nger grow. However, now excess pbosphorus rernains in the water.
available nitrogen (mostly nitrate) concentrations 01' 10 ¡tgP L -1 and 100 ¡tgN L -l. 1"
If the chlocophyll-to-carbon ratio is 25 ¡tgChl mg-I: (a) How rnuch biomass could
potentially be produced by photosynthesis in ¡tgChl L -I? (b) How much carbon in
As displayed in Table 28.3, sewage is generally enriched in.phosphorus. There-
rngC L -I? (e) How much oxygen would be required in mgO L -1 to nitrify Ihe organic
fore water bodies dominated by wastewater effluents tend to be nitrógen limited. nitrogen produced by photosynlhesis?

28.4. A town of 20,000 people discharges raw sewage into a river with a now of 1 crns. If
the river has 10 ¡tgP L-1 and 100 ¡tgN L -1 prior lo Ihe addition 01' the sewage, what
TABLE 28.3
is the N:P ratio allhe mixing point? What is the limiting nutrient above and below the
lIf:f ratlos for point, nonpoint, and marine waters (data from Thomann and Mueller discharge?
1987, pmernik 1977, and Goldman et al. 1973)

Source type TNrrpt INIIJ» Limlting nutrient

Rawsewage 4 3.6 Nitrogen

Activated sludge 3.4 4.4 Nitrogen
Activaled sludge plus nitrification 3.7 4.4 Nitrogen
27.0 22.0 Phosphorus
Activa!ed sludge plus phosphoRls removal
Aclivll!ed sludge P'IIS niarogen removal
0.4 0.1 Niarogen
3.0 2.0 Nilrogen
Activated sludge pll!s niarogen and phosphoRls removal
Nonpoint sources
28 25 PhosphoRls
Marine waters 2 Nilrogen

tTNffP = tOlal nitrogenllotal phosphorus; 'INnp = inorganic nitrogenlinorganic phorphorus.

 LECfURE29 Phosphorus Loading Concept 535
LECTURE 29 10
- -
dangerous
i "eutrophlc"
Phosphorus Loadlng Concept ¡
Q.
'. .9..
...¡

0.1
1 10 100 1000 FIGURE 29.1
H(m) Vollenweider's (1968) loading
plo!.

As displayed in Fíg. 29.1, the plot provides an easy-to-use model to perform

LEcrtJRE OVERVIEW:Jfir~tdescribe thephosp~oru:
an easy-to-usemethod
¡oJ¡fg-h16~~~i&í~a
fot obtaining order-of-niagnihicleesthna1eS oflnkt\)a.
i to attain a desired trophic state for a lake of a particular mean depth.
r trophlcation. th~91 devetop a simple mass.balá1tcefran1eWorkto~notftUi~..
I
I
iliIr predictions ártdj!lustrate .how the plots and the "'.IISSbtU!!bc~,sái'éft\l~ted.
Next t review sortie 01 the empiri!:al correhttions tbili fiá\í~ Séeit.J¿\tis~~fÓpl'é-
~ict eUb'ophicatiol1sy~p.t0?ts (chlorophyU a.liypdli!M~U.b j6);y~~Hd~tnlíiltlf "
etc.) as a fúhction óflpa~iil~s an~ in-Iake cofil:~itttáüó~~!ti!l~!xl $mU&~e r.~~~
iUre,\\Iithtwo tnechalliStic.mod~isbased ohthe phospRót .. ,
á sediment-watermode' fot total phosphot'Us ánd á. wáter!cdtuinn
differentiates between different forms of phosphonis.
'fhe phosphorus loading concept is based on the premise that phosphorus is the pri-
niarf,controlfiióTe limiting nlltrient of lake and resé!vOir eutrophication. A number of
simple empliTcal models ha ve been developed to predict lake eutrophication on the
basis of this premise. The earliest and most fundamental were developed by Richard
---
Vollenweidcr. As a consequence they are often termed "Vollenweider plots/, - TI<"
both simulatíon and wasteload allocation calculations for lakes. For a simulation
calculation the modeler would use loading and depth data to predict trophic state. For
a wasteload allocation calculation the modeler would determine the loading required
In a subsequent paper Vollenweider (1975) added a second determining factor
11 into the loading plot framework. He recognized that not only depth but also residencl?
time had an impact on e!;'troe~!cat!oll'~nessence fíe observed-that faiter ftushing
r .esseemedTo-be iess susceptible to eutrophication than lakes with long residence
times. He incorporated this effect into the plot by adding the inverse residence time
to the abscissa.
As in Fig. 29.2, he again plotted lakes on this space and superimposed lines.
However, along with straight lines. he suggested that curves represented a superior
fit of the data. Again, as with the earlier version (Fig. 29.1). the plot can be used for
both simulation and wasteload allocatíon predictions.
It should be noted that one problem with using mean depth and residence time
is that they are not independen!. In fact, the abscissa, H!Tw.can be shown to be
independent of depth.
_
-H - HQ
- _
- -HQ _- -Q =
- q, (29.1)
V H As As .

29.1 VOLLENWEIDER
LOADING
PLOTS
The first loading plot was developed by Vollenweider (1968). This model was based .
on Rawson's (1955) insight that deeper lakes are less susceptible to eutrophication'
th:m shallower systems. Vollenweider compiled arealloadings of total phosphorus
Lp (mgP m-2 yr-I) and mean depth H (m) from north temperate lakes from around
the world. He used these points to locate the lakes on a space defined by log Lp ver-
sus log H (Fig. 29.1). He then labeJed each lake as to its trophic status (oligotrophic.
mesotrophic, eutrophic). Finally he superimposed lines dividing the various cate-
gories of lakes.
where q, is called the hydraulic overflow rate (m yr-I). Note that engineers involved
in water and waste treatment have historically correlated sedimentation in treatment
reactors with the overftow rate (Reynolds 1982).
A final refinement was developed independently by Vollenweider (1976) and
Larsen and Mercier (1976). Their models corresponded to plots of tbe logarithm
of Lp versus the logarithm of qs(l + JiC). Again, curves were used to divide the
loading levels on the plo!.
Other investigators. notably Rast and Lee (1978), have applied Vollenweider's
(1976) approach to larger data bases and have extended it to predict trophic status.
We review these extensions in Seco29.3.

534

- --------..-
 .. " ..~--- -. .1....

536 PAIITv Eutrophication and Temperature LECTURE2~ Ptíosphorus Loading Concept 537

10 On the basis of such budget calculations, Vollenweider conc1udedthat the loss rate
Dangerous could be approximated by
"eutrophic" 10

í-
ks = H (29.5)
1 'h
~
>-
Chapra (1975) suggested that because the loss of phosphorus was due lo settlil1g ,
~
of particulate phosphorus, Ihe loss term should be represented as
!=>.... (,\
.......
-.:je
-
Q.
en 0.1
000

00"
o" 00

00
00
.......

Permissible

"oligotrophic"
e Vt = W - Qp - VAs~ (29.6)
where vis Ihe apparent settling velocity (m yr-I). lf Ihis is lrue, al sleady-state Eq.
29.6 can be solved for
1(..

W
0.01 P = (29.7)
, Q + vA s.
0.1 1 10 100 1000 Thus we can see Ihat Vollenweidcr's cstimatc that ks = IO/H supports Ih~ settlil1g
velocily approach. In faCIEq. 29.5 is equivalenl lo using Eqs. 29.6 and 29.7 with v =
Hh:w (m yr-1)
10 m yr-I . Data analysis by a number of individuals (e.g., Chapra 1975, Dillon al1d
FIGURE29.2 Rigler 1975, Thomann and Mueller 1987) has delermined Ihat the settling velocity
(1975)loadingplol.
VO"~rWf'li~!lr's most commonly takes on values in the range from about 5 lo 20 m yr-I . However,
values have been reported from less than 1 lo over 200 m yr-I .
\1.
The unity between loading plots and budget models can be iIIustrated (Chapra 4 f
29.2 ¡:JUDGETMODELS and Tarapchak 1976) by dividing the numerator and denominalor ofEq. 29.7 by Ihe n !
surface area to give
Early in the development of phosphorus loading models, it was recognized that sim- L ~~¡ .1
ple mass-balance models could provide the same predictions as loading plots. In (29.8)
fact Vollenweider (1976) wrote one of the first phosphorus mass-balance models for P = qs + v
a welJ-mixed lake as or L = p(qs + v) (29.9) 1.,

where V = volume (m3)

\ ví = ~ --~ --k~vcl (29.2)
TakingIhelogarithmof Eq. 29.9gives
log L = log p + log(qs + v) (29.10)
¡el
!!
It is assumed thal for phosphorus-limited syslems, lrophic stale is correlated with
p = total phosphorus concentration (mg 01-3) phosphorus concentrations. Vollenweider and others have suggested that mesotrophy
t = time (yr) is bounded by lotal phosphorus concentrations of 10 and 20 j.lgPL-1 (Table 29.1).
W = total P loading rate (mg yr-I) lf this is true, Eq. 29.10 can be IIsedto draw lines on a graph ofIog Lp versus log qs.
Q = outflow (m3 yr-I) As in Fig. 29.3, the result is a loading plot that is quite similar lo Vollenweider's
ks = a first-ordersettlinglossrate(yr-I) (1975) mode!.
At steady-state this equation can be solved for Beyond showing Ihe unity of the two approaches, the budget approach is useful
W in elucidating Ihe mechanisms Ihat underlie the shape of the loading plot. Inspecting
(29.3) Eq. 29.10 indicates thal there are two asymplotes. Al one extreme (low fIushing
p = Q + ks V
lakes; small qs) Eq. 29.10 reduces 10
Based on phosphorus budget data (that is, inputs, outputs, and concentration of phos- log L = log p + log v = constant (29.11)
phorus), the loss rate can be determined as
W - Qp
Thus as assimilalion becomes solely dependent on sedimentation, the curves fIatten
ks = = -W - - (29.4) out on the left side of Fig. 29.3. Conversely for high fIushing lakes (high qs), Eq.
Vp Vp Tw 29.10 approaches
 538 PARTV Eutrophication and Temperature LECTURE29 Phosphorus Loading C()ncept 539

TABLE 29.1 . 29.3 TROPHIC-STATE CORRELATIONS

Trophic-state cIassificationbased on total phosphorus concentration as
well as on other variables renective of eutrophication in the previous seetions we ealculated total phosphorus eoneentrátioIts and t~téf:-
preted the resulting levels as indicators of trophic status. Another dpprmí1:his tQ ~~e
Variable Oligotrophic Mesotrophic Eutrophlc phosphorus coneentration (or in some cases loadings) to predict other lrophiC-silite
< 10 10-20 > 20 variables that more directly reflect the deleterious effeets of eutrophieation.
Total phosphorus (¡.tgP L-')
Chlorophyll a (¡.tgChl" L ') <4 4-10 > 10 As in Table 29.1, these other variables provide measures of eutrophication. In
Secchi-disk depth (m) >4 2-4 <2 faet because they are more directly reflective of the adverse effects of eUlI'ophication,
Uypolimnion oxygen (% saturatíon) > 80 10-80 <10
they are preferable to total phosphorus eoncentration.

/ Total phosphorus
loadlng

....é
10gL = logp + logq, (29.12)
Cunsequently, as assimilation becomes solely dependent un flushing, the curves
approach straight lines with a slope of one. rus
Beyond the Vollenweider (1975) plot, the above theoretical development can be
cmploycd to gain insight ¡nto his 1976 model, which can be formulated as
L
p=- (29.13)
q,( 1 + jT:)
,. Total phosphorus
concentratlon
Comparing Eqs. 29.8 and 29.13 leads to the conclusion that the 1976 model has a
settling velocity
H
(29.14)
v = q.. jT: = jT: Chl
or a first-order rate uf

k -- (29.15)
,f - jT: TP
Ch,IorophyllB
ncentratlon
COI
10
SD AHOO
--
í..
,I
~
>-
I
I
1. Chl TP
a.. ~
S 0.1
~...
Slow
flushers
:
I
Fast
flushers Secchl disk Areal hypollmnetlc
- - - - -' depth oxygen demand
0.01
FIGURE 29.4
0.1 1 10 100 1000 Schematic 01 approach used by Chapra (1980) to predict trophic state variables based on phosphorus
qs(myr-l) loading model predictions. The approach consists 01a number 01submodels and correlalions Ihal lorm
a hypolhesized causal chain Ihal starts wilh lolal P concenlralion predictions based on budgel models
FIGURE 29.3 or loading plols. This concenlration is used in conjunclion wilh a series 01correlalion plols lo eslimate
Loading plot derived Irom a phosphorus budget model (Eq. symptoms 01eutrophication such as chlorophyll a concentration, Secchi-disk depth. and hypolimnetic
29.10withvc 12.4myr '). oxygen demando
540 PARTv Eutrophication and Temperature lECfURE 29 Phosphorus Loading Concept 541

One such scheme is outlined in Fig. 29.4. The approach consists of relating AII these models show an increase of chlorophyll with increasing phosphorus.
predicted total phosphorus concentrations to symptoms of eutrophicatioh such as In all cases the relationship is nonlinear. However, the Dillon and Rigler model dif-
chlorophyll a concentration, Secchi-disk depth, and hypolimnetic oxygen demando fers in that its exponent is greater than 1 (slope of log-Iog plot = 1.449),connoting
Each step is dictated by empirically derived corcelations, as described next. that more polluted lakes exhibit proportionately higher chlorophyll than less pol-
luted lakes. The other corcelations show less chlorophyll per phosphorus for more
productive systems.
29.3.1 rhosphorus-Chlorophyll Correlations In addition all the models are assumed to be appropriate only for phosphorus-
Initial attempts to extend phosphorus loading models attempted to caIculate chloro- I~íted systems. Smith and Shapiro (1981) have presented a modified corcelation
phyll a levels as a function of total P concentration. Most of these are based on a fit that attempts to account for potential nitrogen limitation,
of a log-Iog plot. Several examples are
10g(Chla) = 1.5510g(p) - b (29.19)
Dillon alld R;gler (1974): where
I
6.404 .'
10g(Chla) = 1.44910g(pv)- I.136 (29.16) (29.20)
b = I.5510g [ 0.0204(TN:TP) + 0.334 ]
Rast alld Lee (1978):
in which TN:TP = total nitrogen to phosphorus catio.
10g(Chla) = 0.7610g(p) - 0.259 (29.17)
Bartsch alld Gakstatter (1978):
29.3.2 Chlorophyll-Secchi-Disk Depth Correlations
log(Chla) = O.80710g(p)- 0.194 (29.18)
AttemptstorelateSecchi-diskdepthtochlorophylllevelshaveagainusuallystarted
where Chla = chlorophyll a concentration (J.Lg L -1) with log-Iog plots. One such gcaph is shown in Fig. 29.6, and it can be described by
P = total P concentration (J.LgL -1) the equation
Pv = spring total P concentration (J.Lg L-1) (29.21)
10g(SD) = -0.473Iog(Chla) + 0.803
Figure 29.5 shows the Bartsch and Gakstatter (1978) version.

100 L 10

./
o o

o
IJ DD o "
OD~O
a:> o ""
0
ocPq¡t)
o o
00
00 0
08 SI 8 o o
D
-
'i a cPtI oy
'/
o
o
DD
o
8 "
00
%
o
"'\:)
~"
"V
o

o
o o 00ti
o
o00
00
i' ~o
O" 00"
o o
0..8
"

i
o o

e 10
6 lb
~g o o
'\ 806 OdJ&:b
",Yo ti o "
66f o 1 o
606 o o ° oligotrophic "
o" A mesotrophlc o
A Q eutrophic
o o
6 o
1 10
100 1
1 10
P (g L-1) Chla (~g L-1)
FIGURE 29.5
FIGURE 29.6
The relationship between chloraphyll a and phosphorus in some United States lakes The relationshipbetweenSecchi-diskdepthand chlorophylla (framRast and Lee 1978).
and reservoirs (from Bartsch and Gakstatter 1978).
 } LECfURE29 Phosphorus Loading COllcept ~43
542 PARTv Eutrophication and Temperature

where SD = Secchi-disk depth (m). which can be manipulated to give

When this equation is transformed into normal coordinates, it becomes 1 (29.27)
SD = 6.35Chl a-0.413 (29.22) I + ¡LCi1iQSDmax
Thus the lit traces out a hyperbolic shape that, as expected, exhibits large Secchi- where JL = alkwc and SD;"ax = 1.9lkwc. Thus the relationship starts at an ¡nitial
disk depths at low chlorophyll concentrations and small Secchi-disk depths at high Secchi-disk depth corresponding to particle-free water. Then, as chlorophyU Costers
chlorophyll concentrations. light extinction, the Secchi-disk depth is reduced to zero. Figure 29.7 iHustrates this
The shape can be related to more fundamental measurements by recognizing pattern.
that light extinction in natural waters is often described by the Beer-Lambert law
f
I = 10e-k,HJ (29.23)
29.3.3 Areal Hypolimnetlc Oxygen Demand
where I = light at depth H
lo = light at the surface Rast and Lee (1978) presented the following correlation to predict the areal hypolim-
ke = extinction coefficient of the water netic oxygen demand in lakes (Fig. 29.8),
A number of investigators have related Secchi-disk depth to light extinction. - 1.07 (29.28)
logAHOD = 0.46710g
For examplc a rough rule of thumb is that the Secchi-disk depth corresponds to the [ q,(1 +L jT;
Tw)]
depth at which about 85% of the surface light is extinguished (Sverdrup et al. 1942, where AHOD = areal hypolimnetic oxygen demand (gO m-2 d-I), or taking the
Beeton 1958). When we assume this level, ~<.l:...23 becomes antilog, .461
0.15 = e-k,SD (29.24)
Further, the extinction coefficient is often related to chlorophylllevels. One common
model is a linear proportionality,
AHOD = 0.085 I [ q,(l +L jT;
Tw) r (29.29)

Thus even Ihough the equation seems to represent a correlation with loading,
\ke = kwc + aChlaJ (29.25) it actually correlales AHOD with in-lake total P concentration (recal! Eq. 29.13).
where kwc = extinction due to water, color, and nonalgal particles (m-1) and a =
a coefficient (L JLg-1 m-I). Substituting this relationship and taking the natural .
logarithm gives
InO.15 = -(kw{o+aChla)SD (29.26)
.

.: .r
. .
20
- 1

Q~
0'1'
=E
<O
- .
~
A
~f
~
A
·
.
,..-
Q)

-
E
~
10 . ./..~
8
.8
./
0.1
1
~ . . ,0.06
10
A ·
'

100
o L
O 5 10 15 20 25
q. (1+ ..¡c::)(~g L-1)
Chla (J.1g L-')

FIGURE 29.7 FIGURE 29.8

The relationship between areal hypolimnelic oxygen demand and phosphorus loading
The relalionship between Secchi-disk deplh and chlorophyll derived
from Ihe Beer-Lambert law and light extinclion relalionships. (Rasl and Lee 1978).
544 PARTv Eutrophication and Temperature ,t
l.ECruRE 29 Phosphorus Loading Concept 545
!
Chapra and Canale (1991) realized this and reanalyzed Rast and Lee's data to deter- 29.4 SEDIMENT-WATER INTERACTIONS
mine a direct correlation with total P concentration, I
Bottomsediments have long been acknowledged as a potenlial source of phosphorus
AHOD = 0.086po.478 (29.30) to the overlying waters of lakes and impoundmenls. As such, sediment feedback I
could have a significant impact on the recovery of such syslems. This would be I
where p = mean total P concentration of the lake (/LgP L-1). particularly true in shallow lakes or Ihose with anaerobic hypolimnia.
When this equation is plotted on untransformed scales (recall Fig. 25.2b), it WehavealreadyaIludedto sedimentfeedbackof nutrientsin Seco25.6.4.Such I
traces a saturating curve that suggests that as lakes become more productive, the !Uechanisticframeworks provide one means to simulate the process. In this section I
amount of AHOD does not increase as rapidly as the phosphorus increases. This is 'wedevelop an altemative approach that is more akin to the phosphorus loading mod- II
consistent with Fair's observation (Fair et al. 1941) that scdiment oxygen demand els describedherein.That is, semiempiricalformulationsare used to simulatesed- I
increases as the square of the organic carbon content of IllIIent.In addition it iment feedback in conjunction with simple lotal phosphorus budgels for the water J
conforms to Di Toro et al. 's (1990) SOD model. Ir it is assumc :1I.lt lakes with higher and the sediments. Such a simplified approach is somctimes more consistent with J
total P have higher sediment organic content, Eq. 29.30 and Fair's observation and typical data collection programs for many water bodies. 't
Di Toro's model show some consistency. ¡
29.4.1 Sediment-Water Model I

29.3.4 Summary In this section we deveiop a simple modeling framework lo address this problem
for stratified lakes. The framework, which is expressly designed for management I
The models and correlations constituting the phosphorus loading concept have been applications, ineludes two components: a total phosphorus budget and a model of II
widely used, in part because they are so easy to apply. They also have the intrinsic hypolimnetic oxygen deficit. Each is described briefly in the following paragraphs. I
appeal of any empirical approach. That is, they directly reflect observations-"what I
you see, is what you get." I
Total phosphorus model. A sediment-water model for total phosphorus in a
Ul1fortunatelythese models and correlations also have shortcomings: lake and its underlying sediments (Fig. 29.9) can be written as III
. fJecause the plots are alllog-Iog and exhibit large scatter, the prediction errors are
substantial. In the forms presented here, these errors are not explicitly displayed. di
VI dpl =W- QPin - vsAsPI + VrA,'P2 ~ (29.31)
Thus the user might naively employ a highly uncertain prediction with unwar-
ranted confidence in its validity. Reckhow (Reckhow 1977, 1979; Reckhow and
Chapra 1979, 1983) and Walker (1977, 1980) have addressed this deficiency in
~dI~
J di' VrAsP2 -~ VbAsP2
where the subscripts 1 and 2 designale the water and the enriched surface sediment
(29.32)

detail and offered remedies. Unfortunately because the loading models and plots
layer, respectively, Vs = settling velocity of phosphorus from the water to the sedi-
are so easy to use, uncertainty is rarely connected with the resulting predictions.
ments (m yr-I), A.I'= surface area ofthe deposition zone (m2), Vr = recycle mass-
. The models are commonly developed from widely heterogeneous data bases. For
transfer coefficient from the sediments to the water (m yr-I ), and vb = a burial mass-
example lakes from different regions and different types of lakes (e.g., well-mixed
transfer coefficient from the enriched surface layer to the deep sediments (m yr-I).
lakes and eIongatcd impoundmls, some nitrogen-limited systems, etc.) are of-
ten ineluded in the same corrdation. The net result is that the prediction error
\tJ loadlng
becomes inflated by regional and lake-type variability. One remedy is to develop flushing Q <,....
customized correlations for specific lake regions or types.
. They provide little mechanistic insight into the mechanisms underlying the eu- Pl
trophication process. Further, their utility is limited to the specific applications
for which they were intended-that is, simulation and assimilation capacity esti-
mation. In contrast, mechanistic models can be extended to assess environmental
Qlodifications(e.g., dredging, reaeration, etc.) and to guide research and experi- settllng recycle
lP~n~tiOt1.
In spite of these shortcomings, empirically derived loading models often provide
useful order-of-magnitude estimates. As such they provide a quick means to "see P2 FIGURE 29.9
;f
the big picture." In other words they offer a way to discem how eutrophication in a Schematic diagram 01 a
burlal phosphorus budget model lor a
pamc\llar lake relates to how lakes generaIly behave. lake underlain by sediments.
 LECJ'URE29 Phosphorus Loading Concept 547
546 PARTv Eutrophication and Temperature

Hypolimnetic oxygen model. A zero-order model is employed here to simulate TABLE 29.2
hypolimnetic oxygen during periods when the lake is stratified, Data for Shagawa Lake (1967 to 1972)

Symbol Value Volts

AHOD Parameter
53 x 1<J6 m3
0h = o; - -¡¡;;-(l - l.,) (29.33) Volume VI
Al 9.6 x 1<J6 m2
Surface area
m
where Oh = hypolimnetic dissolved oxygen leve) (g m-3) Mean depth HI 5.5
2.2 m
o; = initial oxygen concentration at the onset of stratification (g m-3) Hypolimnion thickness Hh
4.8 x 1<J6 m2
A2
AHOD = areal hypolimnetic oxygen demand (g m-2 d-J) Deposition znne area
H2 10 cm
Surface sedimentthickness
H h = average hypolimnion thickness (m) W¡n 6692 x 1<J6 mg yr-I
Total P loading
t = time (d) Total P oulftow WOUI 4763 x 1<J6 mg yr-I
1, = time of onset of stratification (d) PI 56.3 mg m-3
Mean waterP concenlration
Mean sedimenl P concentration pz 500,000 mg m-J
Although frameworks to predict AHOD as a function of the quality of the overly- Th.. 15 .C
Hypolimnion lemperalure-summer .C
ing water are currently under development (Lec. 25), empirical approaches provide Th..' 4
Hypolimnion temperalure-winler
v. 42.2 m yr-I
a much simpler means lo estimate this quantity. For example, Eq. 29.30 can be used. Tolal P seuling velocity
8
Summer initial hypolimnelic DO DO,... mgL-1
For dimiclic lakes, an AHOD can also be exerted during'winter inverse stratifi- 8 mgL-1
Winter ¡nitial hypolimnetic DO DO,...
cation. To simulale depletion for these syslems Eq. 29.30 can be eXlrapolatedon the
basis of temperalure using
AHODw = AHOD,1.08T..-T. ./ (29.34)
TABLE 29.3
where T., is Ihe lemperature (0C) al which Ihe summer AHOD, is measured and Tw Stratificatlon data for Shagawa
is the lemperature eC) corresponding to the desired winter AHODw. Note that a Lake (1967to 1972)
similar temperalure correction factor is used to scale the recycle velocity.
Event Day

Start 01'spring mi~ed period 120

Start 01'summer stratificalinn 150
29.4.2 Appllcatlon: Shagawa lake 255
Start 01'fan mixed period
Shagawa Lake. Minnesota, is an ideal setling to iIIustrate the model. This is because Start 01'winter stralification 320
it is one of the first systems where sediment feedback was recognized as being im-
portan!. In addition it has been the subject of numerous studies and, hence, there is
a large quantity of information available for model calibration. Finally after being
polluted for mallYyears, it was the subject of significanlllutrient load reductions in
Ihe early 1970s.
W1n I -:- - - - - - - - - - - ~~ W~Jt -
Mode! calibration is based on Ihe extensive data reported by Larsen. Malueg,
and colleagues (e.g. Larsen and Malueg, 1976, 1981; Larsen et al. 1975, 1979, 1981;
f'vlalueget al., 1975; Bradbury and Waddinglon, 1973).Parameter values were deter-
mined by calibrating for the pretrealment period for which data is available (1967-
1972). As in the following example, it is assumed Ihat Ihe lake was at a steady-state
during this time frame.
EXAMPLE 29.1. CALIBRATION OF SEDlMENT-WATER TOTAL P MODEL.
Data for Shagawa lake from 1967 through 1972 are summarized in Tables 29.2 and
FIGURE E29.1-1 FIGURE E29.1-2
29.3. Use this data to calibrate the sediment-water total P mode!. Specifically, estimate
(a) the burial and (b) Ihe recycle velocilies.

Solution: 11'the system is at steady-slate. the sources and sinks 01'phosphorus should
balance. As depicted in Fig. E29.1-1. the amount 01'phosphorus buried must be equallo
the difference belween phosphorus inflow and outflow.
...

548 PARTv Eutrophication and Temperature LECTURE2~ Phosphorus Loading Concept 549

Thus Ibe burial velocity can be computed as ~\~'\).((¿ . 120

= Win - WOU1 = -
6692 X 106 4763 X 106 = 803 X 10-4 -)
Vb A2P2 4.8 X 106(500.000) . m yr ;¡ 100
E
(b) The recycle velocity can be estimated by taking a mass balance around the sediments C)
to determine how much phosphorus is recycled from the sediments to the water on an
annual basis (Fig. E29.1-2). The balance can be sol ved for Ibe amount of phosphorus
.s 80
(/)
...
:J
recycled per year,
VrA2P2 = VsA2PI - VbA2P2 = 11,410 X 106 - 1928 X 106 = 9476 X 106 mg yr-I
o
c.
(/)
60- ._
·í/'--. · · - ,\,·
o
Now, this value must be distributed over the summer and winter anoxic periods. To do c. 40 8
Ibis we must determine the AHOD rates. For the summer,

AHOD = 0.086(56.3)°.418 = 0.5905 g m-2 d-I

-
(ti
o
1- 20' ¡;'ygo0h,.. rI5mgL:'\ 4 o g
E

This value along with other parameters can be substituted into Eq. 29.33 to deter- \.., ................ ........ .... ................ ..:.. c.
mine how long after stratification the lake would go anoxic (lbat is, below approximately O O
\.5 mg L -1), O 80 160 240 320

( _ )= (Oi - oanoxic)(H!,)= (8 - 1.5)2.2 = 24 2 d FIGURE 29.10

1 Is AHOD 0.5905 . Plot 01 phosphorus and oxygen in the pretreatment period (1967-1972)
which means Ibat in the summer the lake will be anoxic for 105 - 24.2 = 80.8 d. A simi-
lar calculation (with Ibe temperature correction from Eq. 29.34) can be used to determine
Ibat Ibe lake would be anoxic for 108.5 d during the winter.
The total amount of recycled phosphorus can be set equal to the terms in the mode!
accounting for recycle in the summer and winter anoxic periods,
W,ecyde= Fa,svr1.08T-Th"A2P2 + Fa,wvr1.08T-Th.w
A2P2
where Fa.s and Fa,w = fractions ofthe year when the hypolimnion
mer and winter, respectively. This equation can then be solved for
W recyde
V = -
r A2P2(~la,s 1.08Th,,-20 + ~Ia,w 1.08Th.w-20)
= 9476 X 106
80.8 108.5 = 0.01663 m yr-I
4.6 X 106(500,( ")
[365' ( 1 08 15-20)
+ 365 (1.084-2°)
Figure 29.10 shows simulation results for a single year during the steady-state
calibration periodo Data shown are for 1972 (Larsen et al. 1979). Notice how the
phosphorus increases due to the heightened release rate when the oxygen level fal1s
below 1.5 mg L-l. A beUer fit could have been accomplished by additionalluning
of the paramelers or by al10wingthe model parameters to vary seasonal1y.However,
considering the simplicily 01'the present calibration process, Fig. 29.10 is judged to
be an adequale approximalion of the general trend of Ihe data.
To obtain some perspective on Ihe long-term dynamics of Shagawa Lake, a sim-
,ulation was performed for Ihe period from 1880 to 2000. Actual ftows were used for
~e period 1967 to 1979. An average ftow is used for al1other years.
Qecause measurements were not made prior to 1967, an idealized long-term
loading scenario was developed (Fig. 29.11). Measured loadings are used for Ihe
lor
Shagawa Lake. The data are lor 1972 (Larsen et al. 1979).
period 1967to 1979.Before 1890and after 1979,an average"natural"loading01'
1311 kg yr-I (Larsen et al. 1975) is assumed to apply.
In 1890the town of Ely was established in Ihe lake watershed. From 1890to Ihe
present, the town's population has been relatively stable. An idealized scenario is
used to characterize the town's contribution. The scenario assumes that immediately
is anoxic during sum- upon Ely's establishment, the loadings increased stepwise to the high average levels ¡.
of the late 1960s. Although the actualloading was undoubtedly different from this
idealization, in the absence of direct measurements it is considered to be an adequate
firsl approximation.
The results of the long-term simulation along with measured data are shown in
Fig. 29.11. The plot indicates that after the 1973 load reduction, tht I ,¡.<.:eXJ>eri-
enced an immediate quick response. However, by 1974 the lake's recovery slowed
] significantly due to feedback of phosphorus from Ihe sediments.
Aside from its simplicity,the use of a step increaseto characterizethe his-
torical loading scenario allows a clearer visualization uf the model's response
characleristics. As in Fig. 29.11, Ihe model shows an immediate sharp increase'
in 1890 following the step load increase. Then, as sediment feedback begins to
dominate, the remainder 01'the increase proceeds at an extremely sluw rate (aboul
0.2 mg m-3 yr-I). By the early 1960s, the lake was just beginning to approach a
steady-state. Although it had not 10tal1yreached a sleady-state allhat time, it seems
to have been close enough to make our calibration acceptable.
The simulation results and data for the recent past are depicted in Fig. 29.12. Th.:
adequacy of the simulation can be assessed by three features of the response. Firsl,
the model approximately matches the prediversiun levels of 50 to 60 mg m-3. Sec-
ond, the initial rapid drop 01'about 25 mg m-3 is close to the observed drop. Finally,
the rate of the subsequent retarded recovery is quite close in the data. Conside..ing
that the model parameters are untuned, these results are encouraging.
 550 PARTv Eutrophication and Temperature LECfURE29 Phosphorus Loading Concept 551
1>

120 10000
~
. LakeI
Outflow
¡.
;
! II ..- 8000

i?' 80 o . ¡~ L.
>.
~
E r'" ... ... oo' , ,. '''v' \
C>
§.
<J)
.¡ruuul=
,
. ( Ji
CI
CI
6000

2 t c: I \
o , '5 4000 I !
.c: 40
c. I, a1
<J) o
o
.c:
c.
t ...J
¡ IExtemalloading
r""\
I tU
(5
f-
2000
.I i
~
\¡,
¡.
.....
J o O
1880 1900 1920 1940 1960 1980 2000 1880 1900 1920 1940 1960 1980 2000
Year
Vear
l'
FIGURE 29.11
FIGURE 29.13
Long.term total phosphorus concentration lor Shagawa lake as
long.term trends 01 sediment recycle (thick line) 01 phosphorus lor
simulated by the phosphorus.oxygen model (thick line). A plot 01 Shagawa lake as simulated by the phosphorus.oxygen mode!. The
inflow concentration is superimposed (thin line) lor comparison. externalloading is superimposed (thin line) lor comparison.

Figure 29.13 is a plot of external and internal loadings as calculated by the

model. Notice how the internal sediment feedback load takes many decades to build 29.5 SIMPLEST SEASONAL APPROACH
up to the high levels observed during the 1960s. Also notice how after advanced
treatment was installed in 1973, the sediment feedback experienced a small abrupt Temperature changes have a profound effect on mass cycling within the water col-
drop followed by a very gradual decrease. It also indicates that, although !he con- umn. The presence of a strong thermocline essentiaIly divides the lake into two ver-
stant feedback model would be adequate for short-term projections (that is, less than ticallayers with markedly different characteristics (Fig. 29.14).
a decade), long-term predictions would have to account for the gradual decline in The surface layer (epilimnion) is warm and weIl-iI1uminated.Consequently al-
s~diment release. gal photosynthesis leads to transformation of dissolved nutrients into particulate 01'-
In summary the foregoing provides a preliminary framework for assessing the ganic matter. Although the thermocline greatly reduces vertical mixing. some of this
impact of sediment feedback of phosphorus on long-term lake recovery, which be- particulate matter settles and diffuses to the bottom layer (hypolimnion), where it de-
cause it depends on a few parameter values, offers a simple means to assess the composes and eventually returns to a soluble formoMixing across the thermocline
long-term impact of murient loadings on lake eutrophication. then reintroduces some of the dissolved nutrient to the surface water, where it is
again taken up by the phytoplankton. Because many contaminants in lakes are asso-
80
¡;;- ciated with particulate organic matter, this cycle has significance to their transport
E and fate.
O> This section describes an approach to simulating basic features of this cycle. The
.s 60
<J) key characteristic of the approach is that it partitions the substance being modeled
:J
C5 into two fractions. Although the model is specifically developed for phosphorus, this
.c:
a. 40 / partitioning is basic to many other substances (such as certain toxic compounds),
<J)
o
.c:
a. ---,
. and the approach could serve as a preliminary framework for analysis of the seasonal

. dynamics of these contaminants. In addition to its use in simulating mass cycling, the
D D

tij 20
(5 lake model can also be employed to simulate oxygen concentration in !he water column.
f-
D Outflow FIGURE 29.12
O Recent total phosphorus 29.5.1 Model Formulation
1964 1968 1972 1976 1980 concentration lor Shagawa
lake as simulated by the Mostof the modelsdescribedin earlier lecturesdeal with the dynamicsof a sin-
Year
phosphorus.oxygen model.
gle substance in a vertically well-mixed water body. As previously mentioned, many
'lo. ~_~-~ 0 ____

552 PARTv Eutrophication and Temperature LECfURE29 Phosphorus Loading Concept 553

'"
e
'"
e
:E
.,:J :E
.,
:J
ii: ii:

I
'\ Uptake
(
Loading Loading c:
SRP Retease NSRP O
.c
INORGAMC 1- '~ODUCTlOÑ\B I /' E
/ e e '" 'Q.
o
- - - o -
.¡¡S
:J
- - _ .¡¡S
:J -s - - - w
:1: :1: c:
- - i5 i5 O
'2

. tz
Loading
SRP
Retease
NSRP
Loading
.E
"O
~
:I:

cl '-
:I: Seltling

1 FIGURE 29.15
Schematic representation 01 the .simplest seasonal approach"

I
developed by O'Melia, Imboden, and Snodgrass.

phosphorus (NSRP).t This breakdown has an operational basis because the two
INORGANIC I\ /).::::-:----::O(\O I ORGANIC I w most common pho~phorus measurements are for SRP and total phosphorus (TP).
ti)
Q
w NSRP can therefore be estimated by the ditTerence of the two quantities as in
NSRP = TP - SRP. In the model itself a further distinction is made in that NSRP
is subject to settling losses whereas SRP is no!. This is based on the assumption

J!l
FIGURE 29.14
Idealizad representation 01 the cycle 01 production and decomposition that plays
a critical role in determining the vertical distribution 01 maller in stratified lakes.
constituents occur in various chemical andJor biological forms that are subject to
transformations due to thermal stratification of the water column. O'Melia (1972),
Imboden (1974), and Snodgrass (1974; and O'Melia 1975) have provided a model-
ing framework that makes a first attempt to simulale some of these changes for the
nutrient phosphorus. Simons and Lam (1980) have dubbed this framework the "sim-
plest seasonal approach" or SSA (Fig. 29.15). Although each of their approaches has
unique features (for which tbe reader can refer to the original publications), in gen-
eral all the models have tbe following characteristics:
1. Phosphorus is separated into two components. These conventionally refer to
soluble reactive phosphorus (SRP) and phosphorus that is not soluble reactive
I
that a significant portion of the NSRP is in particulate formoHowever, because
NSRP also ineludes nonsettleable dissolved organic phosphorus (DOP), the dis-
tinction is not precise. We return to the subject of the definition and interpretation
of the phosphorusfractionsand the generalquestionof kineticsegmentationat
the end of this section.
2. The lake is segmented spatially into well-mixed upper and lower layers úf con-
stant thicknesses.
3. The year is divided into Iwo seasons represenling a summer, slralified period
during which lurbulent exchange between Ihe layers is minimal and a winler,
circulation period when lurbulent Iransport is inlense and Ihe lake is well-mixed
vertically.
4. Linearfirsl-orderdifferenlialequationsare used to characlerizetheIransportand
kinetics representing the mass exchange belween the componenls. As schema-
tized in Fig. 29.15, Ihese exchanges inelude the following:
'Note that Imboden (1974) Ireats phosphorus as disso\ved P available for bioproduction and parliculate
P. He therefore does nol explicilly accounl for dissolved organic phosphorus. The queslion of kinetic
segmentation is explored in further detail al Ihe end of Ihis section.
554 PARTv Eutrophication and Temperature LEcruRB 29 Phosphorus Loading Concept 555

. Waste input,f. Mass loadings of SRP and NSRP fractions can be input lo bolh TABLE 29.4
Ihe epi- and the hypolimnion. Typical ranges of parameters for the slmplest seasonal approach
. Transport. Flushing of mass through Ihe lake's outlet is characterized mathe- for modellng phosphorus
Unlts
matically in a similar fashion lOa single CSTR (Ihat is, as flow times concen- Parameter Season Symbol Range'
Iralion), yet il differs in Ihal only Ihe surface layer loses mass in this way. A Summer kit., 0.\-5.0 d-I
Epilimnctic uptake rate d-'
smalllevel of vertical lurbulenl lransfer 01'diffusion belween Ihe layers is used Winter k"... 0.01-0.5
during Ihe slralitied periodoFor Ihe winter, Ihe diffusion coeflicienl is increased Summcr k", 0.01-0.\ d-'
Epilimnetic release rate d-I
Winter k", 0.003-0.07
lo Ihe poinl where for all practical purposes the lake is well-mixed. 0.003-0.07 d-'
Summer k,.h
. Settling. By detinilion, only NSRP is removed from the waler column using lIypolimneticreleaserate
Winter k,.h 0.003-0.07 di
settling velocity relalionships. Separale seuling velocilies can be inpuI for Ihe Annual \/,.,Vh 0.05-0.6 md-I
Scllling velocity
Iwo layers as well as for Ihe Iwo seasons.
. Uptake. To accounl for the facl Ihal SRP is transformed into particulale matter t Val"es are taken primarily from Imboden (1974) and Snodgrass (1974).
via phyloplanklon produclion, uplake of SRP is characlerized by a firsl-order
reaclion. This mechanism is included only in the epilimnion under Ihe assump- The mass balances for Ihe Ihree remaining pools are
lion thal light limitation makes il negligible in Ihe lower layers. In addition a
much higher uplake rale is used during Ihe summer to account for Ihe facl Ihat V. d~;.. = W".. - Qp".. + v,A,(p"./. - P"..)
produclion is grealest al Ihal lime. + kll.'V,P.". - kr,.V.p",. - v.A,p".. (29.36)
. Release. A number of mechanisms such as decomposilion, respiralion, zoo-
(29.37)
planklon grazing, elc., acl lo relurn phosphorus from Ihe NSRP lo Ihe SRP VI,d~~h = W,.1r + v,A,(p.,.. - P.,.Ir)+ kr.1rVIrP".1r
pool. A tirsl-order reaclion dependeni on Ihe concenlralim, of NSRP is used lo
approximale Ihis phenomenon.
V dplI.1r W A( )
IrT = + v, , P"..- P".h
".Ir
To Ihis poinl Ihe kinelics of mosl models in Ihis volume have been characlerized
- kr.hVIrP".1r
+ v.A,p".. - VIrA,P".h (29.38)
a~ simple firsl-order decay of a single pollulant. Wilh Ihe SSA we begin lo inlro-
duce more elaborale represenlalions of subslance inleraclions. As in Fig. 29.15, the Definilions and typical values 01'the parameters are contained in Table 29.4.
phosphorus forms are coupled by uplake and release reaclions. For example in Ihe
epilimnion, phosphorus is losl from Ihe NSRP compartmenl 01'"pool" by a release
reaclion. This loss, in lurn, represenls a gain for Ihe SRP pool. Therefore, when we 29.5.2 Applicatlon to Lake Ontarlo
wrile Ihe mass balance for a particular pool, each arrow in Fig. 29.15 represenls Parameters for Lake Ontario are summarized in Tables 29.5 and 29.6. In addition the
a 'erm in Ihe resulting differenlial equalion. For example for Ihe SRP pool in Ihe
epilimnion, heal exchange coefficienl across the thermocline is 0.0744 m d-1 during tbe summer
stratified period, and Ihe waler column is well-mixed vertically during olher limes
of Ihe year.
V dps.. - Ps..)- ku..V.ps..+ kr..V.Pn..
Ws..- Qps.. + v,A,(Ps.h
· dt Equations 29.35 to 29.38 can be integrated numerically with a method such
Accumulalion Loading- Flushing + Diffusion Uptake + Release as the fourth-order Runge-Kutta technique. The results are displayed in Fig. 29.16.
(29.35) As can be seen, the primary feature in the epilimnion is the shift of mass from the
SRP and the NSRP fraction during summer due to the large uptake rate. Because
where the subscripts n and s designate NSRP and SRP fraclions, respeclively, and of the lack of production, the hypolimnion is generally a more stable system, with
Ihe subscripts e and h designale epilimnion and hypolimnion. respectively. concentrations maintained al fairly constant levels throughoul the year.
Although the reactions in Eq. 29.35 are all first-order, they could just as eas- Although Ihe above framework captures many of the essential features of the
ily be more complicaled formulations. For example in Lec. 33, nonlinear relation- seasonal cYcle, it has several limitations. In particular. its use of constant coeffi-
ships are used to characterize phyloplankton growlh. In addition if olher substances cients and tirst-order kinelics limits its general applicability. Many of the processes
(such as additional forrosof phosphorus 01'other nutrients) were to be included in the governing substance interactions in the water column are nonlinear and dependent
model, compartmenls (each representing a differential equation) and arrows (each on faclors not accounted for in this model. For example the epilimnetic uptake rate
represenling a lerm in the differential equalion) could be included. The point is Ihal depends, arl10ngother things. on light intensity, temperature. and levels eJfbolh the
regardless of the complexity of Ihe situalion, Ihe conservalion of mass, as reftected phytoplankton and the dissolved nutrient. In addition the dependence on the nutrient
hy Ihe sel of differential equalions, is a simple bookkeeping exercise lOaccounl for is besl described hy a nonlinear relationship. Thus some efforts to refine Equalions
how, when. and where mass moves within Ihe system.
556 PARTv EutrophicalÍon and Temperalure LECTUHE
29 Phosphorus Loading Concepl 557

T\BLE 29.5 . 30
. ,f'Mmationon
- Lake Ontario in the early 1970s Total P
. ........-- I 25
I ...ameter Symbol Value Vnits
Area I 20
Surface
Thermocline
A.
A,
19.000
18.500
10" m'
106m' 15
r.--fJ
:'-0- A
\
..
. 4 o o ~ ~NSRP "',.....
Meandepth
Wholelake H 86 m
'"
10 ...
"- - -~...
-~ - :f.-ó";>O4
Epilimnion H, 15 10
Hh 71 m
g 5
4 SRP: 4
Hypolirnnion c:
Volume g ¿¡¡ ¿...f
!!!
Whole lake V 1634 lO" lO' c: O
Q) r,
Epilirnnion V, 254 lO"lO' g 30
J380 lO" m' oo
Hypolimnion Vh
U)
üulllow Q 212 10" 10'\yr I e 25
TotalP q,
o
SRP load
g. 20
Epilirnnion
Hypolirnnion
W.1,c'
WJ.h
4000
O
lO" rng yr I
lO" mg yr I
o
Q.. NSRP .
NSRP load /1_ /1
;
_ _.o.~ - ~ ~. ..:,.- _~
..-...D...............
o o .... FIGURE 29.16

~
W".r 8000 109 rng yr-I
Epilirnni'>D
~ '4 i:..:-- '" l> -.J
",/ -o";:....
l>
Data and simulation results
Hypolilllnion W",h O lO" rng yr' I . 5 o
using the simplest seasonal
approach lor total phosphorus
SRP
in Lake Ontario. (Top-
29.35 to 29.38 have focused on more mechanistic characterizations of the kinetic O epilimnion; Bottom-
interactions (lmboden and Gachter 1978). Later lectures review nutrientlfood-chain J F M A M J J A S O N D hypolimnion).
models that incorporate sufficient refinements as to constitute an alternative method.
Other ways in which the approach can be modified would be to divide the hy-
polimnion into severallayers to try to resolve vertical gradients in the bottom waters 29.5.3 KineticSegmentation
(Imboden and Gachter 1978). In addition the metalimnion can be modeled as a third There are three basic rationales underlying kinetic segmentation:
segment.
Likewise, phosphorus can be subdivided into additional components to more . First, the division of matter can be based on measurement techniques as in the case
realistically define its dynamics. For example a distinction can be made between of the SRP/NSRP scheme. Similarly a dissolved/particulate splil (Fig. 29.17b) is,
living particulate phosphorus and detrital phosphorus or between dissolved organic
and dissolved inorganic forms. Figure 29.17 depicts a number of possible kinetic
segmentation schemes for phosphorus.
. in part, based on the use of filtration to discriminate between these pools.
Second, the segmentation can have a mechanistic basis. For example the break-
down of matter into pools with similar kinetic characteristics facilitates derivation
and measurement of the input-output terms and coupling mechanisms between
TABLE 29.6 components. The division of the particulate phosphorus into phytoplankton and
Kinetic parameters for Lake Ontario detrital components (Fig. 29.1711) is illustrative of this rationale since the settling
in the early 19705 rates of these two pools are different and can be measured separately.
Parameter Season Value Vnits
. Finally the segmentation scheme can have a management basis. For example the
explicit formulation of a phytoplankton pool has informational value for the plan-
ku,e Surnrner 0.36 d-I ner trying to assess the deleterious etJects of eutrophication.
Winter 0.045 d-I
k"e Surnrner 0.068 d-I A general advantage of adding compartments is that the transport and reac-
Winler 0.005 d-I tion processes can usually be formulated in a more mechanistic manner based on
k'h Surnrner 0.005 d-I
Winter 0.005
measurements. A disadvantage is that additional effort must be expended to obtain
d-I
Ve.V. Annual 0.103 rnd-I these measurements. In addition the more "sophisticated" representations usually
require more effort to program and run on the computer, and they are moredifficult to
558 PARTv Eulrophicalion and Temperalure LECTURE29 Phosphorus Loading Concept 559

~ ~ 25 cm yr-I wilh a phosphorus conlenl 01' 40 kg km-2 yr-I. Almospheric sources 01'
10lal phosphorus are 24 kg km-2 yr-I. Ir 10lal phosphorus seules at arate 01' 12 m yr-I,
calculate Ihe pil's total phosphorus concenlration. Note thal evaporation exactly equals
direcl precipilation.
29.2. A lake (surface area = 10fi 103, mean deplh = 5 m, residence time = 2 yr) has a tOlal
~ ~ phosphorus loading 01' 2.5 X lOS mg yr-I. Nole Ihal the lake has a k..< = 0.15 m-l.
Delermine Ihe following:
(o) TOlal phosphorus innow and lake concenlralions
(b) Chlorophyll t1 concenlralion
(e) Secchi-disk deplh
(d) Loading required lo mainlain Ihe lake al Ihe border between oligotrophy and
mesolrophy
29.3, A slralified lake (surface area = 1.5x JOfi 102, Ihermocline area = 1 X IOfi 102, epi-
limnion volume = 1x 107m3, hypolimnion volume = 0.8 x 107 m3, residence time =
3 yr) has a lolal phosphorus loading 01' 7 x JOKmg yr-I. Determine ils 10lal P and
chlorophyll t1 concenlralions and ils Secchi-disk deplh. Also, calculale ils AHOD and
delermine lhe concenlralion 01' oxygen in ils hypolimnion over Ihe summer slratified
period 01'3 monlhs.
29.4, Lake Sammamish, localed in Washinglon Slale, has Ihe following characterislics:
Lake surface area = 119.8 km2 FIow = 2.03 X 108m3 yr-I
Sedimenl surface area = 13.068 km2 Active sedimenllhickness = 10 cm
Waler volume = 3.5 x 108m3 Sedimenl porosity = 0.9
Hypolimnion volume = 9.8 x 107m3 Sedimenl densily = 2.5 x 106 g 10-3
Hypolimnion lhickness = 7.5 m
The lake was heavily polluled in Ihe 1960s when the lotal P concenlration in the lake
was approximalely 33 mg m-.1. Allhis time the lake's innow concenlralion was about
100 mg m-.1. The sedimenl P concenlralion allhe same lime was aboul 0.12% P.
The lake is monomiclic wilh a slralified period from day 135 lo day 315. The
oxygen al Ihe beginning 01' Ihe slralified period is aboul 8 mg L I and the summer
hypolimnelic temperature is approximalely 10°C.
In 19691he innow concenlralion was abruplly dropped lo 65 mg 10-3. The follow-
ing dala are available I'or lake lolal P (in mg 10-3) before and after Ihe diversion:

Year 1964 1965 1966 1971 1972 1973 1974

FIGURE 29.17 Tolal P 32 35 32 27 29 32.5 25.5

Allernalive kinelic segmenlalion schemes lo model seasonal phosphorus dynamics. Ye:lr 1975 1979 1981 1982 1983 1984

inlerprel. Furlher, prediction reliability may actualIy decrease wilh addilional com- Total P 20 14 22.5 18.5 16.5 18.5
parlments, bccause Ihe least underslood pools would lypicalIy be incorporaled lasl
(recalI Fig. 18.2). Therefore Ihe decision lo expand on a basic framework such as Ir a value 01'46 m yr-I is assumed I'or lhe seuling velocily, calibrale and simulale lotal
Fig. 29.15 musl be made only after considering Ihese factors. P in Ihis lake I'rom 1960 Ihrough the year 2000. Present your resulls graphically. In
addilion invesligale Ihe response if Ihe innow concenlralion had been dropped lo 45 or
25 mg m 3 in 1969.
PROBLEMS
29.5. Use Ihe simplesl seasonal approach lo delermine Ihe concenlralions in Lake Onlario
29.1. A gravel company excavales a 405-ha gravel pil thal is 1.52 m deep. The pil is filled over Ihe annual cycle if Ihe loadings in Ihe early 1970s were doubled.
wilh groundwaler al a mle 01' 10010.1 d-I. The groundwaler is devoid 01' nulrienls. The
pil has a small drainage arca 01' 607 ha. Rainl'all on Ihis drainage area yields runoff 01'

--
 -------.--

LECTURE
30 Heat Budgets 561

LECTURE 30 Remedial measures have been proposed to reverse these effects (Oswald and Roth
19H¡\). I.¡thematicalmodels could prove useful in the evaluation of these modifi-
cations. In addition diurnal temperature variations are relevant to the modeling of
the fate of pollutants in such systems. For example water-quality problems such
Heé:Jt puctgets as ammonia toxicity are sensitive to diurnal temperature variations.
. Temperature effects on biota. some of which are threatened or endangered, are
directing attention toward heat management. Aside from the drainage-basin mod-
./i'fications noted above, reservoir releases can also have an impact on temperatures
/ in taíl waters.
For all these problems, the extensive body of theory concerning the fate of heat
provides a basis for temperature models of streams, estuaries, and lakes. In this lec-
ture l' 11describe the part of this theory that relates to surface heat exchange.
,
J
I

30.1 HEATAND TEMPERATURE

To this point we have focused on mass balances of substances in natural waters.
In particular we have used concentration to provide a measure of the intensity of a
pollutant. For a volume of water V, the concentration e is related to the mass m by
m
e = V (30.1)
Recall from our discussion in Lec. I that mass is an extensive (that is, size-
dependent) property whereas concentration is intensive (size-independent).
We ended the previous lecture with a model of a thermally stratified lake. As we saw, A similar relationship can be developed for heat,
thermal stratification can have a pronounced effect on water quality. Consequently, H
T = (30.2)
understanding heat and temperature modeling is an important facet of water-quality Pe l'V
modeling. where T = temperature
The mathematical modeling of the transport and fate of heat in natural waters H = heat
has been the subject of extensive study. Edinger et al. (1974) provide an excellent p = density
and comprehensive report of this research. Thomann and Mueller (1987) have sum-
el' = specificheat
marized the fundamental approach as it relates to water-quality modeling. .
Most of this work has been oriented toward evaluating cooling-water discharges Thus heat is the extensivequantitywhereastemperatureis intensive(Fig. 30.1).
and has dealt with systems such as cooling ponds and large rivers. Today there is
a broadening of interest in temperature modeling that goes beyond the effects of
point sources of heat such as power plants. Several problem contexts that involve MASS BALANCE HEAT BALANCE
temperature and heat are:
. Physical processes (such as thermal stratification) along with biological and chem- Mass extenslve Heat
ical tnmsformations are sensitive to temperature. Therefore to adequately charac-
terize these other problems, there may be cases where an accompanying analysis Concentratlon Intenslve Temperature
of heat would be needed.
. There is growing interest in the diurnal temperature variations of shallow, turbu-
c=y-
M
relatlonshlp T=~
lent streams that are commonly found in upland regions. Although these systems pC.V
are sometimes subjected to anthropogenic heat loads, their response to natural forc- ----
ing functions is also of interest. Physical modifications, such as channelization and FIGURE 30.1
riparian zone denudation, can have a pronounced effect on their thermal regimes. The analogy between mass and heat.

560
562 PART
v Eutrophicalion and Temperature LEcrURE30 Heat Budgets 563

TABLE 30.1 30.2 SIMPLE HEAT BALANCE

Thennal unlts
lust as a mass balance can be written for a volume of water, a heat baltJnce can
Symbol MKS CGS Engllsh also be developed. As was the case for mass, the heat balance states that for a finite
Temperarure T °CorK °C or K °F or °R volume of water over a unit time period,
Heat H joules calories Blu Accumulation = sources - sinks (30.3)
Density p kgm) gcm) Ibmn .J
Volume V m) cm.1 n' A hypothetical completely mixed system is depicted in Fig. 30.2. Por a finite
Speci tic heal C. J (kg °C) , cal (g °Cj-' Blu (Ibm°Fj-' time period, the heat balance for the system can be expressed as
Accumulation = innow - outflow :t surface heat exchange (30.4)
Note that several systems of units can be used (Table 30.1) for the quantities in In this balance there is a single inflow source that contributes heat. Although this is
Eq. 30.2. labeled "inflow" it represents both heat entering through tributary streams as well as
Observe that two parameters, density and specific heat, are included in Eq. 30.2 point discharges. A loss is included due to heat leaving the system through the lake's
to reflect the type of substance that is being described. As in the following example, outlet. Finally surface heat exchange represents the heat gained across Ihe air-water
these parameters innuence the temperature change that results from adding heat to interface due to interactions with the atmosphere. As indicated by the plus or minus
different substances. sign, this term can be either a source or a sink, depending on the state of the lake and 1-
the atmosphere.
EXAMPLE 30.1. TEMPERATURE AND HEAT. Determine how much heat needs Other sources and sinks could have been included in the heat balance. For ex-
to be added to I mJ of air, waler, brick, and iron lo induce a 1°C rise in temperature.
ample exchange of energy with the sediments can be an important terrn in very shal-
Thermal properties of these subSlances (Kreith and Bohn 1986) are listed below:
low systems. However, since sediment-water exchange is not significant for most
natural water bodies, we limit ourselves to the terrns outlined in Fig. 30.2 and
Denslty Speclftcheat
Substance (kgm-J) [J(kg °Ct' ] Eq. 30.4.
Although Eq. 30.4 has descriptive value, it cannot be used to predict water qual-
Dryair 1.164 1012
Water 998.2 4182 ity. To do this we now express each term as a function of measurable variables and
Common brick 1800 840 parameters.
Casi iron 7272 420
No'e: all valacs are for 20"C.
Accum!llation. Accumulation represents the change of heat H in Ihe system
over time t,
' !::J.H
80lutlon: For air, Eq. 30.2 can be rearranged to compule A l (30.5) 1---
ccumu atlon = "I;t

H = pCpVT = 1.164~~ (1012kg1oC)(I mJ)WC) = 11781

The amount of heal for the olher subslances can be compuled in a similar fashion. Al!
the resulls can be labulaled as

Added heat
Inflow
. Outtlow
..
Substance (J)
Dry air 1178
Waler 4.17x10"
Common brick 1.51x 10"
Casi iron 3.05 x 10"

NOlice Ihat, primarily due lo ils lower densily, much less heal is required to induce a
lemperalure rise in dry air Ihan in the olher substances. Also observe thal allhough it
is nol as dense as Ihe solids, more heal must be addcd lo Ihe waler because of its bigh FIGURE 30.2
sJ1ccitic hc:!!.
A heat balance lar a well-mlxed lake. The arrows represent the majar sources
and sinks 01 heat.
 Jt

LECTURE.JO Heal Budgets 565

564 I'A,(!" bu, ,¡ph¡cation and Temperature

EXAMPI.E 30.2. HEAT BALANCE FOR A WELI..MIXED POND. A pol1d h;ts

Substituting Eq. 30.2 into 30.5 yields the following characteristics:
Accumulation = ÁpC p VT (30.6) Volume = 50.000 m3
Surface area = 25.000 m2
Mean depth = 2 m
In the present case we assume that the density and specific heat of water are relatively Inflow = oUlflow = 7500 013d-I
constant over the range of temperatures for the lake. We also assume that the lake's
volume is constant. These assumptions allow us to bring the parameters outside the ./ The pond's inllow has a lemperature of 20°C. The net heat gain from the atmosphcre
difference. Finally when Át is made very small, Eq. 30.6 reduces to . is 250 cal cm-2 d-I. If there is no other heat exchange, calculale the steady-stale tem-
perature.
Accumulation = pCpV~~ (30.7) SoIution: First, the surface heal flux can be converted to the proper units,
1O'()()()Cm2
Thus heat accumulates as temperature increases with time (positive dTldt) and di- --~
minishes as it decreases with time (negative dTldt). For the steady-state case, heat
j = 25()~-
cm2 d
( m2 )( 0.2388 cal )= 1.047 X 107 J m-2 d-I

remains constant (dTldt = O). Note that the units of accumulation (as with all other Then, Eq. 30.12 can be solved for
terms in the balance) are heat per time (J d-I). This can be seen by inspecting the
_, A,J _ 2 25,000(1.047X 107) _ 3 _ °
units of the individual terms in Eq. 30.7,
T - Tm+ QpCp - 0+ ~~""".."n,.,,,. "',.,. - 20 + 8.. 6 - 28.36 e
dT kg J 3 OC J
pC p V di = ( )
m3 kg 0C (m ) (d ) = d (30.8)
Thus the lemperalure is incrcased 8. 36°C by Ihe surface heat lIux.

Inflow. As was the case with mass, we lump all point and nonpoint sources oí In the previous example, we treated atmospheric heat exchange as 11single entity.
In fact, as described next, it consists of a number of diffcrent mechanisms.
heat entering the lake's periphery into a single term, '

Inflow = QpCpTin(t) (30.9)

30.3 SURFACE HEATEXCHANGE
where Q = volumetric flow rate of all water sources entering the system and Tin(t) =
average inflow temperature of these sources. For the prcsent case we have assumed As depicted in Fig. 30.3, surface heat exchange can be modeled as a combinatioll
that flow is constant and that all the temporal variations in heat inputs are due to of five processes. Note that the processes can be clustered in two ways. Pirst, wc
tu,,¡)V(...1varlations in the inflow temperature. can distinguish bctween radiative and nonradiative mechanisms. Radia/ion relel s
to energy that is transmitted in the form 01'electromagnetic waves and thus does not
Outftow. In our simple system (Fig. 30.2) heat is carried from the system by an
outflow stream. The rate of heat transport can be quantified as a function of the out-
Radlatlon terma Nonradlatlon terma
flow temperature To. However, because of our well-mixed assumption, the outflow A A
temperature by definition equals the mid-Iake temperature, T o = T, andtheoutflow ( \ ( \
sink can be represcnted by
= QpCp T (30.10)
Outflow
'x/ 'x/ Alr-water
Surface heat exchange, The net heat exchange across the lake's surface can
be represented as a flux,
" " Interface
Surface heat exchange = AsJ (30.11) Solar Atmospherlc Water Conductlon Evaporatlon
ahortwave I longwave longwave and and
where As = lake's surface area (m2) and J = surface heat flux (J m-2 d-I), with a I radia~lon radlatlon I radlatlon I
convectlon
condenaatlon I
positive flux representing a heat gain. ~ ) ~ )
y y
Total balance. The terms can now be combined into the following heat balance Net absorbed radiatlon Water-dependent terma
for a well-mixed lake:
FIGURE 30.3
dT
The eomponents 01surtaee heat exehange.
VpCPdi = QpCpTin(t) - QpCpT + AsJ (30.12)
566 PARTv Eutrophication and Temperature LECl'URB30 "eat Budgets 567

depend on matter for its transmission. In contrast, processes such as conduction and can have a wide range of water content. The range of water content can be tepre-
evaporation depend on the motion of molecules. sented by the percent relative humidity Rh, which is the ratio ofthe air's actual wa-
The second way in which the mechanisms can be divided is based on whether ter content to its maximum possible level at the same temperature,
they are dependent on the temperature of the water. This division relates to whether
the process acts as a model forcing function or ineludes the dependent variable: wa- Rh = 100 eair (30.15)
esat
ler temperature. Processes such as solar radiation and atmospheric longwave radia-
tion are independent (lf the condition of the water. Hence these terms act as forcing where eair = vapor pressure of the air (mmHg) and eSal = satunition vapor pressure
functions. As labeled in Fig. 30.3, they constitute the net radiation absorbed by the (mmHg). The latter quantity is a function of temperature and can be calculated by
water. In contrast, although they are dependent on external conditions (air tempera- (modified for units from Raudkivi 1979)
ture, moisture, wind speed, etc.), processes such as evaporation are also a function 17.27T
of water temperature. eSa! = 4.596e237.3+T (30.16)
The total surface heat nux can be represented as
Figure 30.4 shows the resulting dependency of saturation on temperature.
1 = l.m + lan - (hr + le + lr) Finally the dew-point temperature is the temperature at which a mass of air
Net absorbed Water-dependent
(30.13) just becomes saturated when cooled at a constant pressure and water content. The
radiation lenns name makes sense if you think about the situation that occurs at night. At sundown,
wherel.", = nelsolarshortwaveradiation if there is no precipitation, the air willlikely have a water content below saturation.
.fan = net atmospherie longwave radiation After the sun goes down, the air usually cools unti! it reaches its lowest value just
1br = longwave back radiation from the water before sunrise. As Ihe air cools (assuming that the pressure and the moisture content
1r = conduction remain relatively constant), the saturation level decreases according to Eq. 30.16.
1r = evaporation Ir the process continues, there will come a point at which the moisture content will
equal the saturalion value. At this point the air cannot hold additional water and
In the following sections we outline how these terms are quantified. Before doing this dew forms.
we must provide some background information on two subjects that have a strong Inspection of Fig. 30.4 indieates that Ihe air vapor pressure is equivalent to the
bearing on some of the terms: the Stefan-Boltzmann law and atmospheric moisture. saturation pressure corresponding lo the dew-point temperature,
17.27Td
30.3.1 Stefan-Boltzmann Law eair = 4.596e237.3+Td (30.17)

AII objects with a temperature above absolute zero emit radiation. The higher the These concepts are important since they relale to quantifying the heat loss due
object's temperature, the shorter the wavelength of the emission and the greater the to evaporation. For example Dalton (1802) proposed that evaporation was propor-
quantity of encrgy emitted per unit of surface area. tional to the difference between the vapor pressure of the air and the saturated vapor
The Stefan-Boltzmann law stales that Ihe maximum rate of radialion emitted pressure calculated at the temperalure of the water surface Ts:
per unit area of surface can be expressed mathematieally as lr = - !(Uw)(es - eair) (30.18)
lrad = EuT: (30.14) where es = saturation vapor pressure corresponding to the surface temperature of
where Tu = absolute temperature (K)
u = the Stefan-Boltzmann constant [= 11.7 X 10-8 cal (cm2 d K4)'I] 100%
E = emissivityof the radiatingbody 50
80%
The emissivity is a correction factor to account for the fact that the body is no( a r
e, ..---- -- ~~~--
60% FIGURE 30.4
Di

l
perfect emitter of radiation. The law is named after two Austrians, Josef Stefan, who :z: The dependency of saturation
discovered it experimentally, and Ludwig Boltzmann, who derived it theoretically. E 25 40%
oS vapor pressure on temperalure.
~ e.lr Unes 01 conslant relative
~.-- 20%
humidity are displayed. Also
30.3.2 Atmospheric Moisture o shown are the vapor pressures
corresponding to the water,
Recall from Example 30.1 Ihal dry air and water have very different densities and O 10 T.20 T", 30 T, 40
air, and dew-point temperature
~pecificheats. Consequently something must be done to account for the fact that air T(<>c) Irom Example 30.3.

-.-
 " ,
--- - ----

568 I'AKTv Eutrophication and Temperature

LECfURE30 Heal Uudgets 569

Ihe water eC) and f(Uw) = a transfercoefficienllhatdependson Ihe windspeed 800

measured a fixed distance above the water surface. :¡,
'1
It is unlikely that you will be provided with a value for Ihe vapor pressure in
Ihe air above a waler body. 11is more likely thal you will be able to obtain estimates ~ 600
for aie temperature, dew-point lemperature, and/or relative humidity. As ilIustrated !
by the following example, all you require is two of these quanlities to calculale the 8 400
:;::
remaining value as well as the vapor pressure of Ihe air. ea
:a
EXAMPLE 30.3. RELATIVE HUMIDITY, DEW POINT, AND AIR TEMPERA- f! 200
...
11I
TURE. The atmosphere above a lake has an air temperature of 25°C, a relative humidity 'O
of 60%, and the suñace water temperature is 35°C. Use this information to determine C/) O
(a) the air vapor pressure and Ihe dew-point temperature and (b) whether evaporation or
condensation will occur. J F M A M J J A S O N D
Soluüon: (a) Equation 30.16 can be used to compute the saturation vapor pressure, Month
17.27(2S) FIGURE 30.5
e = 4.596e237.3+2S = 23.84 mmHg
Daily totals 01 solar radiation lor dillerent latitudes as a
Therefore Eq. 30. 15 can be used to determine the air vapor pressure, lunction 01 lime 01 year (Irom Kreider 1982).
. - Rhe, _ 60(23.84) _ 143
e.Ir - 100 - "'" - . mmHg
Then this value can be substituted into Eq. 30.17,
· Scattering ami absorption. Once sunlighl enters Ihe almosphere iI is scaltered by
dUSl, reftected by clouds, and absorbed by atmospheric gases.
17.27TJ
14.3 = 4.596e237.3+Td
· Reflection. Upon reaching Ihe water, a portion of Ihe radialion will be scatlered
by refleclion allhe surface. The percenl of radialion relurning from a surface con 1_
which can be solved for
pared lo thal which slrikes il is formally referred to as Ihe surface's aibedo. As de-
Td = 237.3 = 1670C picled in Fig. 30.7, for a ftal waler surface Ihe fraclion reftected is significant only
17.27 _ 1 . when the sun is low in the sky. In addilion Ihe condilion ofthe sky (clear/overcasl)
In (14.3/4.596) and Ihe waler surface (ftat/waves) can affecl reftection.
(b) The saturation vapor pressure corresponding to the surface temperature of the water · Shading. Some streams are located in deep canyons or are lined by lalllrees. In
can be computed as such cases Ihe resulting shading can greally decrease solar radialion.
17.27(3S)
e, = 4.596e237.3+3S = 31.93 mmHg Solar radialion is usually oblained from direcl measuremenls or equalions. 'rhe
lalter are a funclion of faclors such as lime, posilion, and cloud cover. Some of Ihe
Because e, > e.ir. Dalton's law indicates that heat (and water) will be lost from the lake
by evaporation. commonly used algorilhms are summarized in a number of references (EaglcsolI
1970, Bras 1990, Brown and Barnwell1987, elc.)
-.
*' 2000
30.3.3 Net Absorbed Radlatlon '1
E
u 1500
1\vo sources of radiation add energy to elwater body: shorlwave radiation from Ihe
sun and longwave radiation from the atmosphere. !e
o 1000
:;::
Solar shortwave radiation. The magnitude of solar radialion is dependent on 11I FIGURE 30.6
I several factors: :a 500 Diurnal Irends 01 clear-sky solar
I!
.. radiation on a horizontal surface at
· Solar a/titude. This will vary depending on the date, time of day, and location 11I
;S
O latitude 400N on the summer solstice
on the earth's surface. Figure 30.5 shows seasonal trends of solar radiation for O (June 21), the winter solstiee (Dee.
6 12 18 24
different values of latitude. Figure 30.6 depicts the daily trend for various days of 21). the vernal equinox (Mar. 21). and
the year. the autumnal equinox (Sep. 21) (Irom
TIme 01 day (hr)
I Kreider 1982).
 570 PARTv Eutrophication and Temperature
CI LECTURE
30 Heat Budgets 571

100"/. A
to mass movement of fluids. Both can occur at the air-water interface aod can be
eGI described by
1:1
Ü
e le = c¡f(Uw)(Ts - Tair) (30.21)
50"'{'
where CI = Bowen'scoefficient (~ 0.47mmHg°C-I). Thetermf(Uw) definesthe
ja; FIGURE 30.7 dependence of the Iraosfer on wind velocity over the water surface, where Uw is
Ir
Ralio 01 reflecled lo incidenl solar the wind speed measured a fixed distance above the water surface. Many relation-
0"10 radialion (albedo) as a lunclion ships exist to define the wind dependence. Bras (1990) provides a review. Edioger
30 60 01 solar altitude. Both clear and
O 90 et al. (1974) have suggested the relationship proposed by Brady, Graves, and Geyer
overcast sky conditions are shown
Solar altltude (deg) ( 1969),
(Brown and BarnweIl1987).
f(Uw) = 19.0+ 0.95U;, (30.22)

Atmospheric longwave radiation. The atmosphere itself emits longwave ra- where Ihe wind speed is measured in m s -1 at a height of7 m above the water surface.
Jiation. This gain can be represented as a modification of Ihe Slefan-Boltzmann law,
Evaporation and condensation. The heat loss due to evaporation can be rep-
resented by Dallon's law, ~-
lall = u(Tair + 273)4 (A + 0.031 je;.)(I - R¿)
Slefan-Boltzmann
(30.19) le = f(Uw)(es - eair) (30.23)
Atmospheric Renection
law attenuation
where e" = saturalion vapor pressure at the water surface and eair = vapor pressure
where u = the Stefan-Boltzmann constant = 4.9 x 10-3 J (m2 d K4)-1 or 11.7 x in the overlyingair (mmHg).
10-8 cal (cm2 d K4)-1
Tair = air lemperature eC)
A = a coefficient (0.5 lo 0.7) 30.3.5 Total Heat Budget
eair = air vapor pressure (mmHg)
The individual terms developed in the previous paragraphs can now be incorporated
RL = reflection coefficient
into Eq. 30.13,
The reflection coefficient is generally small (~ 0.03).
1 = 1m + U(Tair + 273)4 (A + 0.031 je;.)(I - R¿)
Nel solar Atmospheric longwave
30.3.4 Water-Dependent Terms
Water longwaveradiation. The backradiationfromIhewatersurfacecanalsobe - EU(T..+ 273)4 - CIf(Uw)(T,. - Tair)- f(Uw)(es - eair) (30.24)
represenledby the Slefan-Boltzmannlaw, Water Jongwave Conduction Evaporation

lbr = Eu(Ts + 273)4 (30.20) This surface heat flux can then be substituted ioto the heat balance for the well-mixed
lake (Eq. 30.12). As described next, the resulting equation then provides a basis for
where E = cmissivily of water (approximately 0.97) and T., = water surface tem- predictiog water temperature as a function ofheat loads and atmospheric cooditions.
perature. The emissivity is a correclion factor that accounts for the fact that the water
is nol a perfect emitler of radiation. /

30.4 TEMPERATURE MODELlNG

Conduction and convection. To this point all the terms have involved radia-
tion. We now tum to the last two mechanisms: convection/conduction and evapora-
tion/condensation. In bOlhcases heat lransfer is linked to matler.
Conduction is the transfer of heat from molecule to molecule when matler of
different temperatures come into contac!. As such, it is analogous lo the diffusive
Iransport previously described in Lec. 8. Convection is heat Iransfer thal occurs due
Now that we have developed a mechanistic model of surface heat exchange, the total
heat balance can be used lo predict the temperalure of a natural water. For simplic-
ity the following discussion focuses on the simple completely mixed lake introduced
at the begirlning of this lecture. 11should be noted that the approach call be easily
extended to olher bodies of water such as streams and estuaries.
 ----

572 PAkl v EUlrophicalion and Temperature LECTURE30 Heal Budgels 573

30.4.1 Steady-State This result can be substiluled along with other values lo compule Ihe conduction losses,

At steady-statea heatbalanceforthe well-mixedlakein Fig.30.2canbe writtenas -CI j'(U..)(T, - Tair) = -0.47(27.55)(T, - 25) = -12.95(T, - 25)

_ pCpQTin 4 ~ pCpQTs Evaporation: We musl firsl determine Ihe vapor pressure for the waler,
0- . + 1sn + U(Tai' + 273) (A + 0.031.,¡eai' )(1- Rd - ~ 11.271",
e, = 4.596e231.3+T,
- Eu(Ts + 273)4 - c¡f(Uw)(Ts - Tai,) - !(Uw)(es - eai')
(30.25) / This result, along wilh olher paramelers, can be subsliluled inlo Eq. 30.18,
11.21T,
EXAMPLE 30.4. STEADY-STATE HEAT BALANCE FOR A WELL-MIXED - f(V..)(e, - eair) = -27.55(4.596e2313+T, - 14.3)
LAKE. A pond has the following characlerislics:
The individuallerms can now be consolidaled inlo Ihe 10lal heal budgel,
Volume = 250.000m3
Surfacearea = 25.000m2 o = 300 + 300 + 642- 30T, - 11.35x 10 x(T,+ 27W
Inftow = outftow = 7500 m3 d-I I
17.21T,
I
The pond's inftow has a temperature of 10°C. In addition, it is subject to the following (
- 12.95(T, - 25) - 27.55 4.596e2313+T,- 14.3 )
meteorological condilions:
Net solar radiation = 300 cal cm-2 d-I
This nonlinear equalion can be solved numerically for T, = 17.3°C. This value can Ihe"
r-
be substiluted back inlo Ihe heal budget 10 assess the relative magnilude oflhe individual
Air temperature = 25°C terms. The results are shown below:
Dew-point temperature = 16.7°C
Wind speed = 3 m S-I
Relative humidity = (;()%
1,,=300 1..=642 1br=-806 1,=99 1,=-16
Calculate the steady-state temperature.
11n= 300 100' = -520 cal cm-2 a-1
Solution: Each terro in the steady-slate heat balance (Eq. 30.25) can be dealt with sep-
arately.
.. .
lnjlow:
pCpQTin _ 1(1)7500 x 106(10) _ 300 I -2 d-I
~ - 250 x 1()6 - ca cm Ts= 17.3
Atmospheric longwave: This lerro requires estimates of A and eai,. We will arbilrar-
ily choose a value of A = 0.6 as the midpoint of Ihe range suggested previously. The
vapor pressure of the air can be computed as
11.21(25)
FIGURE E30.4
eai, = 0.~4.596e231.3+25) = 14.3 mmHg
These values can be substituted along with other paramelers to compute

U(Tai, + 273)4 (A + 0.031 Je.;;)(1 - Rd 30.4.2 Time Variable

= 11.7 x lO-s(25 + 273K)4 (0.6 + 0.031 Ji43)(1 - 0.03) = 642 cal cm -2 d-I The surfaceheat exchange(Eq. 30.24)can also be incorporatedintoEq. 30.12 10
Outjlow: derivea time-variableheat budget,
_ pCpQT, _ 0.9982(1)7500X 106 T, = -30T,
A, - 250 x 1()6
dTs = 9.r + ~ + U(Tai,+ 273)4(A + 0.031Jeir)(I - Rd
Water longwave: dt V ID pCpH pCpH
-Eu(T, + 273)4 = -0.97(1 J. 70 x lO-s)(T, + 273)4 = -11.35 x lO-s(T, + 273)4

Conduction: Fiest, we musI compute the wind effect (Eq. 30.22),

_ 9.T _ Eu(Ts+ 273)4 _ c¡f(Uw)(Ts - Tai,) _ !(Uw)(es - eai,)
l'I
8,~ V s pCpH pCpH. pCpH ,
f(Vw) = 19.0 + 0.95(3)2 = 27.55 r (30.26)

E>
 574 PARTv Eutrophication and Temperature LECTURE30 Heat Budgets 575

EXAMPLE 30.5. TIME. VARIABLE IIEAT BALANCE FOR A WELL.MIXED ature occurs at about the same time of year as the peak air temperature. In contrast the
LAKE. Compute the annual heat budget for the same pond used in Example 30.4. Val- peak solar radiation occurs about a month earlier.
ues for meteorological variables have been provided (Table E30.5): Finally Fig. E30.5-3 shows the individual terms of the heat balance in ftux units.
Observe that the largest gains and losses are due to atmospheric and water radiation.
TABLE E30.S
Thermal units
-- -- 30 900
Solar Alr Dew.polnt Wlnd
radiatlon temperature temperature speed ~
Month (cal cm Zd 1) (OC) ("C) (kmhr 1) 20
~u
Jan 169 8.3 2.8 11.6
E
Fcb
Mar
274
414
9.0
13.5
3.3
4.9
11.7
16.4
f...
10 !..
Apr 552 13.9 4.0 15.6 -upu-Ilnnow I--~'---'_:" ". ea
May
Jun
651 21.8
24.7
5.3 16.6 ..
IDewpo1ntl
-~:~~:.:;~.~---
,
~
684 7.8 16.7
o I I I I I I I I I I I I I O
Jul 642 29.4 11.8 12.7
537 26.6 11.5 11.7 O 60 120 180 240 300 360
Aug
Sep 397 24.9 7.7 14 Day
Ocl 259 15.0 6.8 12.9
Nov 160 9.7 6.5 14.8 FIGURE E30.5-2
Dec 127 6.6 2.4 11.6 Simulation results.

Assume thal the volume and nows are constanl and that the innow temperature
lixed at a constanllevel

Solution:
E30.5-1):
of 10°C.

Before developing the solulion. we can plot Ihe meleorological

is

variables (Fig.
- 800

400
"t
E O
U

-
30
¡¡¡
U -400

20 )C

10 if -1200 I-800 f
O
aCk

60
ad I
120 180 240 300 360
'. Day
o
O 60 120 180 240 300 360 FIGURE E30.5-3
FIGURE E30.5-1 Components 01 the heat balance.
Day Meleorological variables.

NOlice thal we have compuled and displayed the relalive humidity along with Ihe other
meleorological variables. The peak solar radiation occurs on day 167 (mid-June) whereas
the peak air and dew-point temperatures fall about a month later on day 198 (mid-July).
High winds blow in the spring and to a lesser extent in the fal!. Finally the highest relative
humidities take place in the winter months.
Equation 30.26 can now be integrated with a method such as the fourth-order
Rungc-Kutta approach described in Lec. 7. The results are displayed in Fig. E30.5-2.
We have included the solar radiation, as well as the innow and the air and dew-point
tempcratures. on the plot for comparative purposes. Notice how the peak water temper-
PROBLEMS
30.1. A 30 x 10 x 10 cm brick is placed in a 2-m3 container along with I m3 of water.
The container is perfectly insulated and is immediately sealed after the brick is added.
If the brick, the water, and the air in the container are originally at 50, 10. and 20°C,
respectively, determine the equilibrium temperature ofthe system. How much heat will
be contained in the brick, the water, and the air before and after equilibrillm? Express
your results in kilocalories (kcal).

.-
576 PARTv Eutrophication and Temperature

LECTURE 31
30.2. A heated discharge has a flow of 7 m3 s-1 and a temperature of 40°C. A completely -- -- ~---"'- ___ ao-
mixed cooling pond is to be built. If it is assumed that J = - 200 cal (cm2 d) -1,
determine the surface area so that the pond's exit temperature is 25°C.

30.3. Determine the net atmospheric
has a temperature of 30°C.
30.4. Calculate the air temperature
llUmidity is 30%.
surface heat flux so that the lake from Example 30.2
Thermal Stratification
if the dew-point temperature is 21°C and the relative

30.5 Repeat Example 30.4, but use a wind speed of 6 m S-l.

. LECfURE OVE!!VlEW: Tdescribe the venical distribution of temperature in a lake.

I r
Th~n illustrate how venical temperature distributions cap t>eemployed to
:..-. -
q~antify., vertical transp. on in freshwater systems such ~s lakes.

A lake's vertical thermal regime has dual significance for the water-quality modeJer. I

As mentioned throughout this text, temperature has a strong influence on the rates
I
of chemical and biological reactions. However, it also has additional significance as
a tracer of transpon in the water column. In fact, heat balances are a primarytooLfor I

estimating vertical mixing rates in freshwater systems. Before descri6íñg how this
is done, I will briefly describe the seasonal temperature~ges in the water columñ
of a lake.

31.1 THERMAL REGIMES IN TEMPERATE LAKES

Hutchinson (1957) defines temperate lakes as those "with temperature above 4°C in
winter, thermal gradients large, two circulation periods in spring and late autumn."
Although other lake types can be severely polluted,t the present discussion focuses
on temperate lakes because many of the world's developed areas are in temperate

t For example lakes Ihal never mix (amictic) or mix incompletely (merolllicric) are eXlremely sensilive lo
pollUlanl inpuls. In addilion, because of growing urbanization and induslrializalion in Ihe world's lropical
regions, many nonlemperale lakes are being subjected 10severe waler-quality stress. Although some of
Ihe approaches in this seclion mighl serve as a slarting poinl for modeling these syslems, additional
p.
research is needed for nonlemperalc lakes.

577
 578 PARTv EUlrophicalion and Temperalure LEcruRE31 Thermal Stralilicalion 579

climales and, consequently, many lakes in Ihese climales have been subjecI lo pol- (8)
lulion. Thus mosl engineering models have been developed for lemperale syslems.
tó"
The Ihermal regime of lemperale lakes is primarily Ihe resull of Ihe inlerplay of
Iwo processes: (1) heal and momenlum Iransfer across Ihe lake's surface and (2) Ihe i
.2 0.9975
force of gravily aCling on densily differences wilhin Ihe lake. Depending on Ihe Q.

season of Ihe year, heal Iransfer lends to either raise or lower Ihe temperature al
Ihe lake's surface as a consequence of a number of factors, including the magnilude 0.995
of solar radialion, air lemperalure, relative humidity, wind speed, and cloud cover O 10 20 30
as described in the previous lecture. Winds blowing over the lake's surface lend to
mix Ihe slIrface waters and transfer heal and momentum down Ihrough the water
column. The extent of Ihis mixing is, in lurn, inhibited by buoyancy effects. This
relates lo the fact that the densily of waler varies over Ihe range of lemperatures
encounlered in lakes (Fig. 31.1). Therefore, denser waters accumulate at Ihe lake's ~Q,
bouom and are overlaid with lighler walers. FIGURE 31.1
... 0.0001
For example in Lake Ontario in August (Fig. 31.2), surface waters of 18°C Plots 01 (a) density (g cm-3) and (b) rale
. '1
with a densily of 0.9986 g cm-3 overlay deep waters of 4°C with a density of 01 change 01 density per rale 01 change
01 lemperature (g cm-3 °C-') versus
1.0000 g cm-3. Allhollgh these densily differences may seem small, considerable 10 20 30
temperature (OC). Note that the maximum
work must be expendedto mix the enlire column(Ihatis, lift Ihe heavier bouom T(OC) densily 01 water occurs at 4°C.
waters againsl the force of gravity lo mix them with the lighter surface walers). The
energy lo do this work comes from the wind. The resull is that buoyancy works gradients are neither large enough nor deep enough to prevent mixingof the wa!er
againsl and mitigates wind-induced turbulence. The interplay between these fac- column by major storms. AUl!e.end of spring and Ihe beginning of summer, surface
lors can be expressed quantitatively by a dimensionless parameter, Ihe Richard.wn heating increases to the point that mixing is confined lo an upper layer.In tenns of the
number, thal represenls the ralio Iof buoyancy
. lo shear forces, as in bulk Richardson number, the fluid has reached the point where the density gradiegt
r. 1'.'
(glp)(iJpliJz) 's sharp enough that even large storms do not reduce R¡ below the stability criterion.
R¡ = buoyancy (31.1) At this stage the lake is said to be stably slratified. The attainment of persistent
shear (alllaz)2
stratification ilJa lake leads to Ihe existence of three regimes: the upper (epilimnion)
where z = deplh (L), which is posilive in the downward direction and lower (hypolimnion) layers separated by a narrow region of sharp temperature
g = acceleralion due to gravity (L T-2) change-Ihe thermocline or metalimnion. t
p = density of the fluid (M L-3) During midsummer Ihe net daily heat flux al the surface is low and, although
iJplaz = gradienl of densily wilh deplh (M L-4) the thermocline deepens gradually. the density gradient between the epi- and the
iJuliJz = gradienl of horizonlal velocily with deplh (T-I) or shear hypolimnion remains strong and stable. Although transport of heat and energy across
the thermocline occurs, it is al a low level and exchange between the upper and lower
If R¡ is significantly greater Ihan a criticallevel (- 0.25),
a stable regime results.
layers is at a minimum.
11'R¡ is significantly less Ihan 0.25, connoting strong shear relalive lo slratification,
In late summer and fall, loss 01'heat due Iargely to falling air temperatures re-
Ihen shear-induced turbulence is generated.
sults in a net heat loss from the lake. As surface waters cool, they become more dense
Figure 31.2 depicIs Ihe seasonal changes in the vertical temperature distribution
than underlying epilimnetic waler. Since this is an unstable situation, strong vertical
uf Lake Onlario. Allhough Ihis lake is very deep, its thermal regime has many of Ihe
mixing called convection occ~r~.Together with increased winds during fall, the pro-
characterislic fealures of smaller lemperale lakes. Some lime after the dis~ar-
cess e~odesthe metalimnion from above, giving the impression of a sinking thermo-
ance of an)' ice coyer in sprlng, temperatures t~!~ughout Ihe ~r column d.st:..~or cline. As the lake cools further, a point is reached at which the deepening surface
wilhin a few degrees of, the maximum density of ~ater (4°C). At Ihis lemperature,
layer becomes denser than the bottom layer, and complete mixing of the column
healing or cOQlingQfthe lake's surface resulls invery small de~sity differences (Fig.
occurs. This episode is calledJall overturn.
31.lb) and, consequently, only asma" amountof wind stress is required lo keep the
waler cohiínn well-mixed. In lerms of 1~~,Richardsonnumber, buoyancy forces (Ihe
nUlnerator)are small, and Ihereforeshear forc,es(Ihe d«<nominator)of small magni-
'The lenns "thermocline" and "melalimnion" often are used inlerchangeably lo de.~ignate Ihe layer of
tude are sufficienl to keep R¡ below the.criticallevel.
sharp temperature change. Allhough Ihe melalimnion is really a layer. Ihe Ihermocline is aClually Ihe
As spring progresses, solar radialion increases, air temperalures rise. and ther- plane passing Ihrough Ihe poi nI of maximum decrease in temperalure wilh deplh.
mal stralilicalion will be eslabli.~hedin tbe near-surface waters. However, density

-
" ..", ~j

580 PARTv Eutrophication and Temperature I EerURI,.1I Thcrmal Stratilkalilln 51iI

o
T
¡
Eplllmnlon
(e)
E
4

Hypollmnlon
150 (h)
FIGURE 31.2
Depth-time diagram 01
isotherms ("C) at a mid-
..~ .
lake station in Lake Ontarío,
z
1972- 1973.
(a) (b)
The lake continues to be well-mixed and lo lose heat as temperatures drop in FIGURE 31.3
winter. In some cases the surface water cools below 4°C and an inverse stratifica- Idealization 01 summer stratilied periad in a temperate lake.
tion resulls owing to the low densily of water below 4°C (Fig. 31.la). This inverse (a) The segmenlation 01 the water column into a well-mixed
s¡ratification is strengthened if ice forms al Ihe lake's surface. epilimnion and hY¡;Oilmnion COI )Qnding to the vertical
temperature distribution in (b)
In summary Ihe seasonal changes in a lemperate lake can be idealized in both
time and space. Temporally Ihe cycle consists of two stages: a summer period of dT.
strong stratification and a nonstratified period of intense vertical mixing. Spatially VePCI'd/ = QpCI'Tin(t) - QpCI'Te:t lA,. + v,A,PCI'(T¡, - Te) (31.2)
the summer stratified period can be treated as consisting of two layers separaled
by an interface of minimal vertical mixing. These idealizations form the basis of d~ . I I

some efforts to develop engineering models of vertical heal and mass distribution in VhPCpd/ = v,A,PCp(Te - T¡,) (31.3)
1,
temperate lakes. where Ihe subscripts e and h designate the epilimnion and hypolimnion, respeclively,
T = temperatureeC), p = density (g cm-3), Cp = specific heal (cal g-I °C-I),
II I

31.2 ESTIMATION OF VERTICAL TRANSPORT Q = volumetric flowrate ofall watersources enlering Ihe syslem (g cm-3), Tin(t) ==
average inflow lemperature of Ihese sources (0C), A,. = lake's surface area (cm::!),
1-
In previous lectures (for example Lecs. 8 and 15) we showed how gradients of con- 1 = surface heat flux (cal cm-2 d-I), VI = l~rmocJine heat transfer coefficient
servative substances, such as chloride, could be used to estimate horizontal diffu- (cm d-I), and Al = thermoc1inearea (cm2).
sion. Because of the large seasonal temperature gradienls in temperate lakes, heat It should be nOledthat exchange across Ihe thermochne also can be parameter-
balances serve an analogous role in vertical models. This implies thal water motion izedas a verticaldiffusioncoefficienlE" whereEl has unilsof cm2d-1 andis relal-
has an identical effect on determining heat and mass transport in Ihe water column. ed to the heal exchange coefficient by
AlIhough Ihis analogy does not hold strictly for large particles or dense solutions, it
El = vIH, (31.4 )
is a good first approximation for most pollulants.
We can now wrile a heat budget in a similar fashion lo those developed in the where H, = Ihermocline thickness (cm). This altemalive expression is usefuJ in
previous lecture. Consider the simplest case of a temperale lake during the mid- comparing Ihe magnitude of vertical mixing with other turbulenl processes. How-
summer stratified periodoAt this lime (Fig. 31.2) the tempcratures in Ihe epi- and ever, il has Ihe disadvantage Ihal il requires specificalion uf the Ihennoc1ineIhickncss
hypolimnion are fairly uniform, with the major gradient occurring al Ihe Ihermocline. lo parameterize vertical mixing.
AlIhough the thermocline deepens during this period, we assume thal the descent is, Equation 31.3 can now be used tu eSlimale the Ihennoc1ine heal exchange CIJ-
at most, very gradual. A simple model for this case consists of two well-mixed lay- efficient. To do [his we assume Ihat during the summer stratified period, the epil-
ers of constant thick~ess separated by an interface across which diffusive transport imnion temperalure is constant (see Fig. 31.4). lf Ihis is lrue, Eq. 3J.3 can be wril-
occurs (Fig. 31.3). Therefore the metalimnion is not modeled explicitly. Ralher the len as
thermocline acts as the interface between the surface and the bottom layers. For such
a system, heat balances can be written for each of the layers as .~ dTh
dt
+ ÁhT¡, = ÁhTe (31.5 )
582 PARTv Eutrophication and Temperature LECTURE31 Thermal Stratification 583

20

15
..
-¡.----- -- 16.74'C
TABLE 31.1
Mean monlhly epilimnetic (O lo 15 m) and hypolimnetic (15 m lo bottom)
le'!lperalures for Lake Ontario (Chapra and Reckhow 1983).
o
Lo
f....
10
. .
. FIGURE 31.4
Plot 01 mean monthly epilimnetic Month
T,
(0C)
Th
rC) Monlh
T,
(0C)
Th
rC) Monlh
T.
(OC)
Th
rC)

.
and hypolimnetic temperatures lor Dec 5.47 5.18 3.9\ 3.20 Oct 11.29 5.64
5 May

.. . I ..~_..... Lake Ontario during 1967 lrom

Thomann and Segna (1980). The
Jan 3.\8 3.30 Jun 8.86 4.05 Nov 7.92 5.68
Feb 0.89 2.00 Jul 15.58 4.35 Dec 5.47 5.18
o mean epilimnion temperature lor the Mar 1.55 2.04 18.56 4.6\ Jan 3.\8 3.30
Aug
J F M A M J J A S O N D summer stratilied period is displayed Apr 2.06 2.04 Sep \7.74 5.\2
(dashed line) along with the simulated
Monlh
hypolimnetic temperature (sol id line).
These coefficients can now be employed in conjunction with Eq. 31.7 to calculate
the hypolimnetic temperature and response time. First, the eigenvalue can be computed
whcre 1'..designates thalthc epilimnion tcmperature is constant and A" is the eigen-
valuc (¡Ir thc hypolilllnion. as (E4. 31.ó)

A" = vIAl (31.6)

Á
¡,
= 10.1(l5.nOO X 1010) = 000 108 7 d-I ( = 039 7 -1 )
1380 X 1015' . yr l
V"
Then Eq. 31.7 can be used to calculate
Equation 31.5 is identical in form to the completely mixed lake model, with step Tr. = 4.2e-o.OOI0871
+ 16.75(1 - e-O.DOIOR7,)
I

input described previously in Lec. 4. If the hypolirnnion temperature atthe beginning

of the summer period is T".;. Eq. 31.5 can be solved for Values calculated with this formula are plotted in Fig. 31.4, As can be seen, Ihe lit seems
linear because the response time of the hypolimnion is so long,
T" = T".¡e-Ahl + 1'1'(1 - e-A",) (31.7) 3 3 I
Equation 31.7 can then be arranged to estimate the heat exchange coefficient across 195= Áh = 0.397 = 7.6 yr
the thermocline (Chapra 1980), I
The transfer coeflicient can also be used to compute mass transfer of substances
V" Te - TIIi across the thermocline. For example in Lake Onlario during the summer, the eon-
VI = -
Al f..
In
(
Te - T"...
'
) (31.8) centrations of soluble reactive phosphorus in the epilimnion and hypolimnion are ap-
proximately 3.1 and 8.ó ¡.LgL -1, respeetively. The mass-transfer coefficient from the
~
where 1..= time after the onset of stratification al which the hypolimnion lempera- hypolimnion to the epilimnion can be estimated as
lure T"... is lIIeasured. I
m3 365d mta

(
_1 10 2 -3

EXAMPLE 31.1. THERMOCLINE TRANSFER COEFFICIENT. Lake Ontario 10.1 cm d (15.000 x 10 cm )(8.6 - 3.1) mg m 1()6 cm3 )( yr )( 109 mg yr-I )
is strongly stratilied during the months of July, August, and September. During this time = 300 I tonnes yr-I
the thermocline is at a depth of approximately 15 m. The average epilimnetic temperature
over the period is approximately 16,74°C and the surface area of the thermocline is This estimate can be put inlo perspective by comparing it to the 12,000 metric tons per
approximately 15.000 X 10(, m2. The temperature in the hypolimnion (volume = 1380 X year (mla) of total phosphorus that enters the lake from external sourees. The faet that
109 m~) rises from 4.20°C in the beginning of July to 5,38°C al the end of September diffusive transport to the epilimnion aeross Ihe thermocline amounts to about 25% of
(Table 31.1). Use Eq. 31.8 to estimate Ihe thermocline transfer coefficient. In addition the externalloading suggests its relative importance in regulating surface produetion of
compute the thermocline diffusion coefficient if H, = 7 m. phytoplankton during the growing season.

Solulion: Equation 31.8 can be used to compute

1380x 1015cm~ 16.74-4.20 Snodgrass (1974) has summarized eslimates of the thermocline difftlsion coeffi-
V, = 15,000x 10IOcm2(90d)In 16.74-5.38
~
( )= 10.1 cm d-I
cienl El for a number of lakes. The values are positively correlated with mean deplh
(Fig. 31.5) and range over several orders of magnitude from 0.003 lo 2.4 cm2 s-) .
which in tum can be used to compute a diffusion coefficient
Note Ihal Ihe following formula can be derived frorn Ihe plot:

E, = 10.1 cm d-l(7oo cm) (86. 100 s) = 0.082cm2 S-I El = 7.07 X 1O-4H1.I505 (31.9)

where E, is in cm2 s-t and H = mean depth (m). With knowledge of the
584 PARTv Eutwvhication and Temperature LECTURE
31 Thermal Stratification 585

10 m2 86.400 s = -J
~

í BA
.
V,
= E, = 0.01 cm2s-J
H, 3m (
10,000 cm2 )( d )
0028
. 8md

U) For the IInstratilied period, a large enough coeflicient is used to simulate complete mix-
Eu ing. The magnitude of this exchange coefficient is estimated by trial and error.
--
-e Equations 31.2 and 31.3 can now be integrated with a method sllch as the fourth-
order Runge-Kutta approach described in Lec. 7. The results are displayed in Fig. E31.2.
al We have included the single-Iayer model from Example 30.5 for comparison. Observe
'0
how the single-Iayer-model temperature falls between the results for the two-Iayer casc.
~U As expected, because it is subject to meteorological inftuences, the epilimnion tempera-
e 10-1 turc is warmcr during the stratified periodo
o
.¡¡¡
30
~
'6
al
.5
"8 10-2 FIGURE 31.5 20
E Thermocline diffusion coefficient
~ (cm2 so') versus lake mea n
depth lor a number 01 temperate
f-,

lakes. Redrawn Irom Snodgrass 10 , ,-...-...........

Hypolimnion
(1974) with new data lrom Lake
.:;:......
Ontarío (Ll) and the central
basin 01 Lake Erie (O).
O
O 100 200 300 400
approximate thickness of the metalimnion, the plot or the equation can be used to Day
make a first estimate of the thermocline exehange eoeffieient for a lake.
Now that we have estimated the thermocline exehange eoefficient, it is relatively FIGURE E31.2

straightforward to combine the two-Iayer lake model with the surfaee heat exehange
model deseribed in the previous leeture. ~
EXAMPLE 31.2. TIME. VARIABLE HEAT BALANCE FOR A STRATIFIED 31.3 MULTILAYER HEAT BALANCES (ADVANCED TOPIC)
LAKE. Compute the annual heat budget for the same pond used in Examples 30.4 and
30.5. Note that the following additional information is available: Although the lumped approaeh of the previous seetion is widely IIsed, onc-
dimensional distributed approaches are also employed. The primary reason for using
V, = 175,000 m3 A, = 11.000 m2 As = 25,000 m2
sueh approaches relates to the faet that the hypolimnion is not completely mixrd.
Vh = 75,ooom3 H, = 3 m In fact it is more accurately characterized as a one-dimensional system where heat
Assume that all atmospheric interactions are limited to the epilimnion and use Eq. 31.9 transport occurs by diffusion.
to determine the thermocline transfer coefficient for the summer stratified period (April 5ueh a distributed approach may be less important for unproductive or very decp
1 through September 30). Also assume that the lake does not stratify during the winter. systems (Fig. 31.6a). For these cases, vertical gradients of constituents may not be
significant in the hypolimnion. Thus a well-mixed bottom layer serves as an ade-
Solutloo: First, we must determine the thermocline transfer coefficient. To do this the
quate approximation.
pond's mean depth can be computed as
However, for shallower, productive systems, significant hypolimnctic gradicllh
H = ~ = 250,000 = 10m can occur during stratification (Fig. 31.6b). In most cases such gradients are cre-
A, 25,000
ated by sediment release of constituents. Because the hypolimnion is govemed by
Then Eq. 31.9 can be used to compute diffusion, concentrations rise becauseof the sediment sources:Thus, higher coneen-
trations first develop near the sediment-water interface and slowly diffuse up through
E, = 7.07 X 10-4(10)1.1505= 0.01 cm2 S"I
the water column. A distributed characterization is required to simulate such gradi-
The transfer coefficient can be determined as ,,'
ents and the associated mass transporto
 586 PART v Eutrophication and Temperature LECJ1JII.E
31 Thermal Stratificatiod sA7

I
TilO,
I
SRP
I
Tilo,
I
FIGURE 31.6
Typical vertical protiles 01temperature,
oxygen, and soluble reactive
~ SUrI_
~ --
...
, phosphorus in (a) an oligotrophic he8tnux hoot nux
u ",..

o
... and (b) a eulrophic lake. For Ihe
, ...
oligotrophic syslem, sediment
sources are nol strong en~
to inducE!"'yertic~ g!adienls in IhB- m +
boltoJJ1waler. I~rast eulrophic
syslems typically have proauCfíve.::.-
sediment Ihallake up oxygen and +
reJease nuírienls. Consequenlíy,
slrong vertical gradienls usually
(a) Ollgotrophlc (b) Eutrophlc occur. .'~¡ .l t..............

Two approaches have been developed lo model vertical temperature distribu- t.............. t..............
tions. The firsl is based on a turbulent diffusion approach (Fig. 31.7a). Numerical
solulions are developed by dividing Ihe verlical dimension inlo segmenls. Heal is t.... t..............
supplied lo Ihe surface and Ihen dislribuled through deplh by diffusion in a fash-
ion similar lo Ihe dislribuled models described earlier in Ihe book (for example Ihe +-.........
t......
elongaled reaclors described in Parlll).
The second melhod, called a mixed-layer approach (Fig. 31.7b), uses a mech-
FIGURE 31.7
anical energy balance lo predicllhe Ihickness of Ihe epilimnion. This Ihick surface
Contrasl between the MO primary
layer is Ihen 1110deledas a well-mixed segment. The hypolimnion is modeled as a
! series of layers using Ihe lurbulenl diffusion approach.
(a) Turbulent dlffuslon
approach
(b) Mlxed-Iayer
approach
models used lo distribute heat

r Because of ils simplicily we focus on Ihe lurbulenl diffusion approach. A one-

vertically in lakes and reservoirs.

I dimensional heat balance can be writlen for Ihe waler column as

" I and a1' (H) =O (31.13)
a1' = !..- E(Z) a1' + 1.",(Z)A.. (31.10) al.
~;.
¡
¡ al az [ az ] pCp
The first condition specifies Ihe heat flux at the air-water interface. The second con- I
I where l' = lemperalure
z = deplh, measured dowl1wardfrom Ihe surface
dition seIs an insulated condilion al the boltom.
Also observe how a portion of the solar radialion is absorbed within a few cen-
E(z) = a verlical turbulenl diffusion coefficienl
limeters of the surface. The remainder is absorbed in an exponential fashion as it
1.",(z) = nel solar radiation distribuled by deplh
radiates down through the water column.
A.. = surface area
At this point the model formulated above merely distributes heat through the wa-
p = water density
ter column by mixing. The key to simulating stratification relates to how the vertical
C,' = water specific heal diffusion coefficient varies with depth. A variety of approaches have been proposed.
Nole that Ihe solar radialion is distribuled by depth according lo Ihe formula One of the first was that of Munk and Anderson (1948). This model assumed that
the diffusion coefficient was a function of the Richardson number,
1.,,,(z)= (1 - Fa)J""e-k,z (31.11)
where Fa = fraclion of Ihe solar radialion thal is absorbed immedialely al the sur-
face
E(z) =~ (31.14)

J"" = nel solar radialion delivered lo the surface

ke = exlinclioncoefticient where Eo = the diffusion coefficient at neutral stability
R¡ = Richardson number
The following bOllndary conditions are required at Ihe waler surface and the a = a constant
bol\om:
1.", - (1 - F")1.,,, A variety of formulations have been proposed to relate Eo to depth and wind
_ E(O) (71'
az (31.12) velocity. We will use the simplest formulation,
flC,.
588 PARTv Eutropllu.:ation and Temperature
IInUkE 31 Thermal SlratilJ, '''11 5H9

Eo = cWo (31.15) 31.2. The Central Basin of Lake Erie had the following general characteristics Inr Ihe sum-
Oler stralilied period from 1967 through 1972:
where e = an empirical constant and Wo = wind shear velocity,
Mean deplh = 17.8 m Hypolimnion Ihickness = 4 m
{!¡ (31.16)
Volume hypolimnion = 40 km3 Epilimnetic lemperature = 19.94"C
Wo = V p:. Area thermocline = 10.000 km2 Epilimnion oxygen = 9.66 mg L I
Metalimnion Ihickness = 7 m TOlal P concentration = 20 f.LgL 1
in which Pw = water density and Ts = shear stress at the air-water interface, ~. In addition, the following dala can be extracted from temperature and oxygen timc
series:
Ts = PairCd U;' (31.17)
June I September2
where Pair = air density Hypolimnion temperature eC) 7.77 12.5
CJ '" a drag coefficient = O.OOO52U~44 Hypolimnion oxygen (mg L -1) 10.72 0.60
Uw = wind speed
Aside from Eq. 3I.l, altemative formulations are available to compute the (a) Determine the thermocline diffusion coefficienl in cm2 sol and compare it with Ih~
value from Ihe Snodgrass plol.
Richardson number. For example (b) Delermine:

R¡ =?
- (glp)(aplax) (3.18)
(i) The apparent areal hypolimnelic oxygen demand in g m 2 d
wol (zs - z) (ii) The actual areal hypolimnelic oxygen demand in g 01-2 d J corrected fur
diffusion
where Zs = the water surface elevation (m). (iii) The empirical areal hypolimnelic oxygen demand in g 01-2 d 1as a funeli,lIl
AIthough the foregoing sequence of formulas may seem involved, their ultimate of lotal P concenlration (Eq. 29.30)
effect is embodied in Eq. 31.14. In essence the numerator is a manifestation of the (c) Compare and discuss Ihe results of (b).
kinetic energy delivered to the surface by the wind. Thus, as we would expect, the
31.3 UseIhe fourth-order RK melhod to compule the hypolimnelic lemperalure for a lak,:
vertical diffusion within the lake will increase for higher winds. The denominator having Ihe characleristics given in Table P31.3. Use a value of 0.13 cm2 s-I for Ih~
decreases the diffusion as the Richardson number increases. Thus, as the density thermocline diffusion coefficient for Ihe summer slratified periodo Also assume that Ihc
gradient increases or as the velocity gradient decreases, the diffusion coefficient de- lake does not strdtify during the winter. Use the same epilimnion temperatures as lO
creases and stratification can occur. Table 31.1.
The foregoing is an introduction to vertical temperature modeling in lakes. It
reviews one of many formulations that have been proposed to model this phe- TABLE P31.3
nomenon. Ford and Johnson (1986) and Henderson-Sellers (1984) can be consulted Lake characteristics
for additional information.
Parameter Value VoUs
Lake surfaee area 25.000 m2
Thermocline area 11,000 m2
pROBLEMS
Epilimnion volume 175.000 m'
31.1. A pond has the following characteristics: Hypolimnion volume 75.000 m'
Thermocline thiekness 3 m
Epilimnion volume = 150,000 m3 Thermocline area = 10.000 m2 Epilimnion Ihickness 7 m
Hypolimnion volume = 50,000 m3 Inflow = outflow = 5000 m3 d-I Hypolimnion Ihickness 6.8 m
Surfaee area = 25,000 m2 Thermocline thiekness = 3 m Inllow = outllow 274 x 10" m' yr I
The pond's inflow has a temperature of 10°C and enters and leaves Ihe hypolimnion.
Stan 01'summer slratifiealion 120 d
End 01'slralificatioll 290 d
In addition it is subject to Ihe following meteorological conditions:

Net solar radiation = 250 cal cm-2 d-I Dew-point temperalure = 15°C
Air temperature = 20°C Wind speed = 2 m sol

Calculate the steady-state temperatures of tbe epilimnion and hypolimnion.

 LEcruRB 32 MicrobeJSubstrate Modeling 591
LECTURE 32

:i
La9 + Exponenllal
growth +- Slatlonary -+- Death ..
Microbe/Substrate Modeling i
1:1I
e

I ~
I 'O
!E
,
I :1
.5-
1:1I
o
I
..J FIGURE 32.1
I Typical growth curve lor
I
...'
Time bacteriain a batch reactor.

· Lag phase. When bacteria are added to the reactor, time is often required for Ihe I

bacteria to acclimate to the new environment.

· Exponential-(or log-) growthphase.After the celIsare acclimated.becauseof
~
~-----. the abundance of substrate (food), they typically grow at a maximum rateJimited
only by their ability to process the substrate. Because they grow by binary fissióñ,
b.stra te t edes~ri.bc
t..Atl~raho\\,
! micr,obiatgrowth . '. .ffi . .~C. J
tál1~e.ffiódér&hsli the resulting increase follows the exponential growth model.
b.~et. revie"'!,0
ttEcrtJUOVERVD:W:
n
i~. of S'U
,. .lev. .
.. .
. .Jba t n:aIImI.VLtI:
.. C .e
.
d. Io.log M
.:....w.. . .
, .!Í!1~ M
.' !c
.
.¡~.
· SflifiiJ/lary
phase. Growth eventually levels off as the bacteria exhaust their food
.
.
.
bi1da~dfi~1
.
.
. . .
pres~fitmeMlchheits-~efite~J:titefohntif~lóff1!i~J!n8IJib$,
. ..
.
. .. ... . . .. ~
.
. supply. Thus the growth rate of new cells comes into balance wilh the dealh of
párameters can ~ estilb,ated
. .. . . ... . . ...... .. . ... ... . .. . . .. .. ... .... . ... .
.
f . .
fro¡n diUa. Tljeh l11UStráEe.{.~W~é~lc~ae~}¡
. .
. .. . ..
I
old celIs, and net growlh goes to zero. The leveling off may be hastened by en-
tipUttrit~~r~fU!"
tisedtb ltiodéliillc~biálgfflWl6,¡~
Men~11 ró1ititllati~ttcitl1.be vi ron mental factors such as the generation of toxic metabolic by-products by Ihe

Q
torsartdCSTRs.1ftthelattl;tt introducetbr.importanttdeii p!a W~Hotitrateo I
J.:íinaUy
nl1tOent.1show.howthe theorycan be apphedto algaegrówiligon
,. .. ¡¡limítlttg · bacteria.
Death phase. If incubation is continued, celI death will eventually outpace growth,
and a decline will set in.
The following mass balance represents a simple starting point for modeling
Fig.32.1:
A large body of theory has been developed to simulate how microorganisms such as
bacteria break down organic carbonoAlthough this theory was largely developed to
dX / /'
(32.1)
model waste treatment processes, it has great relevance to Ihe modeling of natural di = (kg - kd)X
waters. In particular it is pertinent to the algal models discussed in later lectures. where X = bacterial concentration (celIs L -1)
Beyond its connection with algal modeling, microbial kinetics has several other kfi = bacterialgrowthrate(hr-I)
applications in water-quality modeling. First, it provides an al~rnative way to model kd = bacterialdeath rate (hr-I)
bacteria from the traditional first-or~er decay approach described previously in Lec.
p. In particular it óf(ers a means to incorporate growth, and in so doing, could
Thusthe statusof the bacterialpopulationis dictatedby the balanceof growthand
death.
lead to improved pathogen models. Second, it provides a theory to characterize the
microbial-mediated decomposition mechanisms in models of bolh conventional and It is important to recognize that the growth and death rates should not neceSsarily
toxic organic pollutants. be considered as constants in this simple representation. However, if the growlh rate
is constant and the death rate is negligible, Eq. 32.1 would aptly model tbe exponen=-
tial growth phase. Similarly if a constant death rate grew much larger thaQ..gfQwth,
32.1 BACTERIAL GROWTH the model couldapproximate the death phase. Further, nmking growth dependp.ntQJl
Jhe food supply co~he model tosimulate Ihe progression from exponential
Suppose that a batch reactor initially contains a large amount of biodegradable or-
ganic carbon and a small amount of bacteria. As depicted in Fig. 32.1, the numbers
growth to the stationary I!!taseas ihe ~!lQ§!rate.i§.de.Qféied~ - -
1n !he fóllowing sections we will be concerned primarily with modeling this
01'bacteria will typically manifest four phases over time: ~Mt from exponential growih to substrate-limited growth. 1'0 do-ii1is we now dlscJ!§s
590 ho~ s~bstrate limitation of growth is modeled.
592 PARTv EUlrophicalion and Temperalure Lt:LTuHE
32 Microbe/Subslrale Modcling 593

32.2 SUBSTRATE LlMITATION OF GROWTH in Lec. 2, almost all the reactions until now have been first-order. This was a vulid
~ation b~ween the growlh rate and Ihe concentration of sub~lrate can be de- approximation for modeling the gross decomposition of sewage in oxygen mod-
scribed by the folIowing empirical model: eling.
S However,nowthatwe mustdeal withbiologicalgrowthas an issue unto itself,
kg = kg,max ~ (32.2)
the boundless nature of first-order growth begins to become inadequate. Hence the
Michaelis-Menten formulation allows us to incorporate a limit to growth (food sup-
where kg.max = maximum growth rate when food is abundant (hr-I ) ., ply) into model kinetics. The introduction of such limits is characteristic of the eu-
S = substrate concentration (mg L -) trophication models in the coming lectures and is reflective of the fact that unbridled
ks = a half-saturation constant (mg L -1) growth is checked and moderated by finite resources in nature.
This model is sometimes called the Michaelis-Menten model, after the equation
used in the study of enzyme kinetics. It is also referred to as the Monod model, after 32.2.1 Parameter Estimation
the Nobel Prize-winning microbiologi'st Jacques Monod, who was the tlrst to apply
it to model microbial growth. Recall that in Lec, 2 we presented a variety of methods to estimate rate parameters y
As depicted in Fig. 32.2, the half-saturatLon constant represents ,the substrate for simple reactions. Michaelis-Menten parameters can be evaluated in a similar i
fashion, Many of the approaches involve transforming Eq. 32.2 so that it plots as a I
concefltra!iQI} <!tWhich growth is half the maximum. Thu_s if provides the parameter
tJiaidlctatcs at what level the substrate becomesÜniiting. At lowfood levels (S « straight line. One of the earliest methods involved inverting it to give i
ks), the growth rate becomes directly proporti'onal to thefoód suppIy,
- I = --+-
k, I I
(32.5) I
kg == kg,maxS (32,3) kg kg.maxS kg.max
ks I
Thus, if the Michaelis-Menten model holds, a plot of I1kg versus l/S should yield a
In the context of Eq. 32, I the growth process becomes second-order.
straight line with an intercept of 1/kg.maxand a slope of kslkg.max.A linear regression
At high food levels (S » ks) th_egr~wth rate becomeU:2.nst~nt ~nd_indepen- can then be used to determine estimates of the coefficients. This approach is called
denLQLthe fQQdsuPlID::, I
a Lineweaver-Burk plot.
kg == kg.max (32.4) Another linearized version is obtained by multiplying Eq. 32.5 by the substrate I
concentration to give I~
In the context of Eq. 32.1 the growth process becomes first-order, As in I'ig. 32.2,
this level is approximately 5 times the half-saturation constant. Thus Eq. 32.2 ad-
justs the rate as a function of whether food is abundant or scarce. -S = -SI +- ks (32.6)
kg kg.max kg,max 1-
Before proceeding we should pause and consider how the Michaelis-Menten
formulation is so radically different from the way in which we have characterized For this case, which is called a Hanes equation, a plot of Slkg versus S should yicld
reactions to this point. Aside from the mention of zero- and second-order reactions a straight line with an intercept of ksfkg.maxand a slope of I/kg,max.
A third option is obtained by multiplying Eq. 32.2 by (ks + S)/S to give

2 .¡ :
. k
- . . . . . . ¡::'!'~~
.
kg = kg.max - ks (i) (32.7)

-'i For this case, which is called a Hofstee equation, a plot of kg versus kglS should
yield a straight line with an intercept of kg,maxand a slope of - k.,.
Finally, aside from linearization, the integralneast-squares method (recall Seco
e.1 FIGURE 92.2 2.2.6) can be used to estimate the coefficients numerically. AII the approaches are
""," Foodsupply i.-- GrowthIndependen! Michaelis-Menten or Monod
curve used to characterize
applied in the following example.
Ilmldng ¡ ollood supply the microbial growth rate EXAMPLE 32.1. PARAMETER ESTIMATION FOR THE MICHAELIS-
O. as the lunction 01 substrate
concentration. For this plot, MENTEN MODEL. The following dala were generated froro Eq. 32.2 wilh val-
O k, 10 20 Sk, 30 40 1;
S (mgL-')
kg, = 2 hr-' and k, =
5 mg L-I.
. ues of kg,max= 2 and k, = 5, The resulting values of kg were Ihen roodified by adding
and sublracling an error of 0, I froro alternale values:
 LECTURE32 MicrobelSubstrate Modeling 595,

S i, S i, S i,
4 0.7889 20 1.5000 36 1.6561
( 8 \.3308 24 1.7552 40 1.8778
12 1.3118 28 1.5970
16 1.6238 32 1.8297

- T!q
e
,. r a;
"O
Estimate the parameters of Eq. 32.2 wilh the Lineweaver-Burk,
transformalions and Ihe inlegrallleast-squares

80lutlon: The Lineweaver-Burk

approach.
Hanes, and Hofstee

approach can be implemented by plotting l/kR versus

'" O
ci '- g E
1:
l/S. As displayed in Fig. 32.3a, a straighlline can be fit to!he data with linear regression.
The best-filline is
S
1:
Q) 1 l
:;! ;:E kR = 3.031 S + 0.472
ci \- teN '" "- .!!l
a;.
cuUI which can be used to determine
.c:GI
:g .-U cu
.. 1
ci X :;! ;:E=>
g"
GIUI
.c:.!.
kR.ma. = 0.472 = 2.117
-UI
_cu
O 01 k, = 2.117(3.031) = 6.418
8 (1)S
e ci
I
N -I
e
e .. cu
01"
QjOl
The other transformation approaches can be implemented in a similar fashion.
... ... E Transformation plots and fits are displayed in Fig. 32.3b and c. The integra1J1east-squares
CU.-
:a't;' approach can be applied in a similar fashion to Example 2.3. The fit Une is shown in
a. Fig. 32.3d. The parameler estimates of all methods are summarized in Table 32.1.
01"0
.c:1:
-CU
01'
1: 01
:¡!
E- The results of the previous example provide some insights into tlte strengtii
ci
N
,. I
g and weaknesses of tbe various approaches. Of tbe four methods the Lineweave
Burk performs the poorest. This is because tbe most stable points (that is, tbose
O .
-UI
UI 01 higher levels of S) tend to cluster together near tbe origin, whereas those tbat usu
"01:
OCU Iy manifest the most variability are spread out far from the origin and hence will ha'
ci \ t"'s
N "- =I
GI
E
an inordinately large influence on tbe slope. Thus the value of ks would be expect.
to be particularly poor. '.~'
-i
:;! .2=> The Hanes transformation provides a better slope estimate by spreading out the
..ID
O .:. data along the abscissa. However, the intercept tends to be cIose to the origin and
C")-GI
. UI
e I NGIGI
I
'" '"
ci e ci g
I \ I e
e
,§ TABLE 32.1
ci :;! a:.s
Results of four methods for obtalnlng
:!!
;r '" a¡J;:::i
¡¡:GIíij'
ID parameter estlmates for the Mlchaells-
.'/ Menten model. The data were
generated wlth perturbed values based
on k,.ma. = 2 hr-I and k, = S mg L-l
Melhod k,..... k,
Lineweaver-Burk 2.117 6.418
Hanes 2.054 5.767
Hofslee 2.040 5.521
Inlegrallleasl-squares 2.048 5.551
594

------
596 p,,¡ I V EII!r.)phicationand Temperature L.Ec-ruln,.l2 MicrobclSubslrale Modeling 597

hence becomes les~reliable. The value of ks tenus to be puor because, for the Hanes the substrate pool. In contrast, uecay represents losses of bacterial biomass thal nei-
model, it depends on the intercepto ther SY1ITlíCsizenew cells nor release organic carbon (for example, respiration). Mass
The Hofstee approach rectifies both problems by spreading the data out and gen- balances for Ihe bacteria and the substrate can be wriuen,
erating a beuer interceptoHowever, because the dependent variable (kg) appears in dX S
hoth the ordinate and the abscissa, least-squares regression is not strictly applicable. t
As in Table 32.1, the result nevertheless often pro vides good results.
di = (kg.max ks +S - kd )
- k, X
(32.9)

Finally, the integraVleast-squares approach also provides decent results. In ad- dS I S

(32.10)
dition, it has the advantage that the equation is lit directly rather than via a transfor- dI = -ykg.maxks+SX+ kdX
mation.
Notice how we have used the Michaelis-Mt:nten formulation to quanlify the depen-
, dence of bacterial growth on substrate concentration.
32.3 MICROBIAL KINETICS IN A BATCH REACTOR
EXAMPLE 32.2. BATCH REACTOR. A balch reaclor has Ihe followillg charact.:r-
Michaelis-Menten kinetics can now be used to simulate the dynamics of bacteria and iSlics:
'\
substrate in a batch reactor (Fig. 32.4). For simplicity, we will use organic carhon as
Xo = 2 mgCL-1 ) So = 998 mg<::.L-1J
the substrate and as the measure of bacterial biomass. _
kg.max - 0.2 hr
_ _'
k, - 150mgCL
_~J
As in Fig. 32.4, the bacteria grow by utilizing the substrate. However, not all of kd= k, = 0.01 ~r-~ Y = 0.5 gC cells (gC subslrale)-I
the carhon that is utilized becomes new bacterial cells. A significant pan is con verted
Simulate how Ihe subslrale, bacteria, and 10lalorganic carbon change over lime in Ihis
iplo carbon dioxide and water. T~~ is quantified
--=--
by the cell yield, reactor.

y = gC~lJs (32.8) Solution: The paramelers can be Subsliluled inlo Eqs. 32.9 and 32.10 and sol ved nu-
gC substrate mericaIly wilh Ihe fourth-order Runge-Kuua melhod. The results are displayed graphi-
I
caIly in Fíg. 32.5. Notice how the bacleria firsl grow exponentiaIly because the substrale
Thus, for this example, because the cells and the substrate are measured in com-
concentralion far exceeds Ihe half-saluratíon conslan!. However, at aboul 40 days tlle
mensurate units, the cell yield represents the efficiency of the conversion of organic subslrale concenlration becomes low enough Ihat il limits growlh and Ihe bacleria lev-
carhon into cell carbono eloff.
Retuming back to Fig. 32.4, the cells are degraded by two processes: death and On Ihe basis of the yield coefficient. we would expecl Ihat Ihe original 998 mgC
decay. Death represents losses whereby the cell organic carbon is released back into L -1 of subSlrale would be con verted lo 449 mgC L -1 of bacteria. However, because \11
the decay ]oss, a peak of only about 433 mgC L -1 is reached. Thereafler the bacterial
populalion declines gradually as both decay and assimilalion slowly deplete the system's
carbon resources.
tj..east-squares regression is predicaled on me assumplion Ihallhe independen! variable (me "x" axis) is
measured precisely and Ihal all me measuremenl error is posiled in Ihe dependenl variable (see Chapra
and Canale 1988 for addilional delails regarding regression).

1000
- - . . '-:. ' . JOe
750
S, substrate", ". .
I ,...,
p'.' ~ :s
'. ) , 500
!
250

FIGURE 32.5
O Growth 01 bacteria as
Y-lkX
y . O 10 20 30 40 50 a lunction 01 substrate
concentratlon in a batch
t (br) reactor.
FIGURE 32.4
FloW diagram lor !dnetlc interactions between bacteria and
organlc carbon substrate in a batch reactor.
598 PARTv Eutrophication and Temperature

32.4 MICROBIAL KINETICS IN A CSTR

Although the batch reactor is informative, natural systems are not closed. Therefore
we now examine how the bacterial growth behaves in a continuously stirred tan k
reactor or CSTR (Fig. 32.6). Adding inllow and outllow terms lo Eqs. 32.9 and 32.10
gives

(32.11)
l:/~ = (kR...""k..~ S - k" - k, - ~)X
tlS 1 S Q
(32.12)
dt = -yk~..."" k, + SX + k"X + V(Si.. - S)

EXAMPLE 32.3. CSTR REACTOR. A CSTR has the same parameters as the batch
reactor from Example 32.2. In atltlition it has the following characteristics:
V =; 10 L
Sin = 1000 mgC L 1
So = O mgC L I
SimllJ;1Ie how Ihe sllhstrate. bacteria. and total organic carbon change over time in Ihis
reactor fil)' three residence times: (a) Tu = 20 hr. (h) Tu = 10 hr. and (e) T.. = 5 hr.

Solution: Thc parameters can be sllbstitllled into Eqs. 32.11 and 32.12 and sol ved nu-
merically with Ihc fOllrlh-order Runge-Kulla method. The results are displayed graphi-
cally in Fig. 32.7. For the longest residence time, case (a), the reslllts are qllalitatively 1000
similar to Ihe batch reactor. As the residence time is lowered. less bacteria are grown.
For the shortest case, (c), the bacteria go extinct' That is, they do not grow rast enough lí SubstratelTOC

to counterbalance the rate at which they are being nllshed.

u.. 500
! Bacteria
___o
L
'"

J-- o I I
,
o 50 100 150
l (hr)

t:'~ kdX FIGURE 32.7

Growth 01bacteria as a lunclion 01 substrate concentration in a
Bacteria eSTR lor three residence times. (a) T.. = 20 hr, (b) T.. = 10 hr, and
X ~~b~trat~ (e) T.. = 5 hr.

Now it can be recognized that the maximum possible growth rate would occur
1'-1
y"k.X
if S = Sin' Therefore Eq. 32.13 can be reformulateoaS-- - - -
Si.. 1
FIGURE 32.6 kR,max-
k .f + S in
= kd + kr + - (32.14)
'Troin
Flow diagram lor kinetic interaclions between bacteria and organic
carbon subslrale in a eSTA. where Tmin = residence time at which washout occurs. Equation 32.14 can be solved
for
-F;'f
The condilion depicled in Fig. 32.71' is refelTed to as "wash 0111." Thc. rcsidence 1jlli.. =
k..+ Sip (32.15)
~e al which it occllrs can be deterlllined by solving E4¡ 32.11 at steady-state for _ k
( R.max -
k
d -
k S
r) ~ k
~- ( d +
k )k
r ..

S Q Another interesting feature 01'the model can be'i'Úhstrated by solving the steady-
k~..."" -/.:.
' --c: = k" + k, +, - (32.13)
l' '/',) V stateversions01'Eqs. 32.11and32.12for
I

1 A\
i "-o -.2" ";~
-----

600 PARTv Eutrophication and Temperature LEcruRE.12 Microbe/Substrate Modeling 60 I

-dp = -apakg.max~ p ailt ..

a¡Jakda + -Qy (Pin - p) (32.19)
s= (kd+k,+~)k. (32.16)
dl k 'p + P

wherea andP = concentration of plants (U"Ustands for ualgae") (mgChla m .1)

kg.max
- (kd + k, + ~) and phosphorus (mgP 013),respectively
p
_Q(S. - S
- k.p = half-saturation constant for phosphorus (mgP 01-3)
and -x= V ID ) (32.17) / J - apa = ratio of phosphorus to chlorophyll a in algae (mgP mgChla-1 )
1
y kg - kd
Two important distinctions can be made between this scheme and the one rol'
microbial decomposition from Fig. 32.6:
where S and X = steady-state concentrations of the substrate and bacteria, respec-
tively, and kg is defined by Eq. 32.2 with S = S. Thus the ultimate steady-state
. Because the plants and the nutrient are measured on a different basis, stoichiomct-
substrate concentration is independent of the inftow concentration. In contrast the
steady-state bacterial concentration varies directly with the inftow concentration.
. ric conversions must be applied to transfers between the compartments.
There is no loss of mass due to degradation processes as was the case for carbono
Thus the stoichiometric conversions do nOIinvolve a loss of mass, merely a changc
EXAMPLE 32.4. WASHOUT RESIDENCE TIME AND ULTIMATE SUB- of units.
STRATE CONCENTRATION. (a) Determine Tminfor Ihe: re:aclor from Example
32.3. (b) Calculale the steady-state substrate level for Example 32.3 with Tw = 20 hr. EXAMPLE 32.S. ALGAUNUTRIENT INTERACTIONS. The epilimnion 01' a
lake:has the following characteristics allhe beginning 01'(he stralified penod:
Solution: (a) Equation 32.15 yields
150+ 1000 /al!. = 0.5 mgChlam-3 po = 9.5 mgPm-3
Tmjn = (0.2 _ 0.01 - 0.01)1000 _ (0.01 + 0.01)150 = 6.5 hr "Pin'= IOr.ngPm-3 to;:'QI:'''7l.5mgPmgChla-1
kg.m.. = 1d-I · k.
'F 2 mgP m-3
(b) Equation 32.16 gives kd '" 0.1 d-I "V i.;.= 30d

Assuming that mass transfer across the Ihermoc1ine due lo diffusion and senling is neg-
S = (0.01 + 0.01 + 0.05)1.50 = 80.77 mgC L-1
0.2 - (0.01 + 0.01 + 0.05) ligible, (a) simulale how Ihe algae and phosphorus change over lime, (b) calcillale Ih.:
which corresponds to the numerical result (Fig. 32.7a). ultimate, sleady-slate level 01'phosphorus, and (e) determine Tmjn.

Solution: (a) The paramelers can be substituled into ~qs. 32.18 and 32.19 and sol ved
numerically with Ihe fourth-order Runge-Kuna method. The results are displayed
graphically in Fig. 32.9») The algae grow to a peak 01' 6.6 mg m-3 in about 4 d.
32.5 ALGAL GROWTH ON A LlMITINGNUTRIENT
(b)
,
"" ~ r 1-
The theory of the previous sections can be easily extended to the growth of algae
as a function of a limiting nutrient. A CSTR is depicted in Hg. 32.8 for the case of S = [kd + (QIV)]k. _ d 10.1 + 1/30)2 t /' '

algae and the nutrient phosphorus. Mass balances can IJewritten.as kg.max- Ikd + (QIV)] . 1 - 10.1 + (1/31\" = 0.lQ8~m-3_
"':1 ,\)">:: (e) ~ , ~ "-
- = kg,max 1!L- ,- kd - -Q '~."

1';1>
I
f da
d
l
k
( 'p + P;..,o
y
r
.
,.~ (32.18)
Tmjn= k. + pin _ + 10 1
(kg,ma.- k,,)Pjn- k.¡(k,) - (1 ~ O.I)IQL 0)'(2) = 1.36 d
f i/ l' tI 1/
10 !
J l ~
1105
.:.
I PI:nt ~ i,1I
:
lI,.i,1I
NU~lent I
FIGURE 32.8
Flow diagram lor kinetic interactions
..

2 4 6
FIGURE 32.9
Flow diagram lor kinetic interactions between
between plants and a limiting nutrient I(d)
In a eSTRo plants and a limiting nutrient in a eSTA.
 .,.
602 PARTv Eutrophication and Temperature
'.11I
PROBLEMS
LECTURE 33

32.1. The following datawere collected for algal growth rate as a function of the inorganic
phosphorus concentration:
Plant Growth and
SRP (/LgP L -1) 0.5 2.0 3.5 5.0 6.5 8.0 9.5 11.0
k, (d ') 0.33 0.52 1.15 1.02 1.09 1.28 1.27
Nonpredatory Losses
1.55

Use (1/) the Lineweaver-Burk and (b) the integral/least:squares methods to estimate ks
and kg.m<l"

~ 32.2. Su Jse that during Ihe springtime the algae in a lake's epilimnion die at arate of
'~.02"'i.d'. her.mpcline diffusiQn is negligible, but the phytoplankton settle at arate of
<1},1 m .,-1 and the epilimnion thickness is 5 ,ÍÚ The following conditions hold:
, Uo = 0.25 mgChlu m-3 po = 20 mgP m-3 '1
vp¡" = 50 mgP m-3
~=2mgfm ~
tF,: = 1.5mgP mgChlo-1
T:"= 30d
-,-i
I
I
_.~ --- ..=-
AI.Iother paramelersare asin Example32.5. Calculate(o) Ihe time requiredto reach
~-
" equilibrium, (h) the ullimate cO!1celltralionsof phosphorusalld algae, alld (e) the
washout i"ále.
n.3. RepeatExamples32.3 and32.4, but with all Sinof 500 mgC L -1.,.1
1 =
""
- - ---

hC1'llllk OVEIIVlEW:
,l.reviewthe tactors fhátcontri,BUtét6
pliyf~'
1
tempetatur,e¡ n\1trient~.. andlight1'he,n,l shu,~~o' . ~'
..., " \..J ~1Uibe used to cofup~lém8SSrates ,StlChliSptimltj ' ',u'","
~ ~.:' _.'
genemüon, Finally 1conclude with a brief ihtrodüttionto~ehUy ~eveid
algal gtowth models employing variable carbon..fo-cWorophyllratio$,
l. -1'"'"\
1"
).
lV' {FI'+ ('/('i-.,.J. In the previous lecture we leamed how to model the dynamics of microorganisms.
(;..
At the end of our discussion we focused in on a specific group of microorgan-
\
r, isms: algae. Now we will refine our plant growth framework by developing a more
_ "\),.-1- complete mode! of an important group of algae: the free-floating microorganisms
11 (
called phytoplankton. In particular, along with nutrient limitation, we will ex-
pand the framework to inelude the effects of light and temperatur,on phytoplank-
ton growth. '

33.1 LlMITS TO PHYTOPLANKTON GROWTH

Just as we used first-order reactions as the starting point for characterizing the decay

~
of organic matter in Lec. 19, we can use a similar approach to mode! growt or a
batch system. a mass balance can be written for algae as
da
dt = kga (33.1)

wherea == concentrationof algae(mgChla m-J) and k8 = first-ordergrowthrate

(d-I). If the initial condition is a() = Oat t = O.then Eq. 33.1 can be solved for
603

---_._-
604 PARTv Eulrophication and TemperalUre LECTUREJJ Planl GI'OWlhami Nonpredalory Losses 605

a = lIOekg/ (33.2) The growth rale frlllll Eq. 33.3 can he represented by
It is known that the phytoplankton growth rate is on the order of 2 d-I. If an ini- kg(T. N. J) = kg.TcPNt/lL (33.4)
tial condition of I mg m-3 were used, the following values can be computed wilh
where kg.T = maximum growth rale (thal is, under oplimallighl and excess nUlri-
Eq. 33.2:
ents)at a particular temperatureand t/lN andcPL= attenualionfactorsfor nUlrienl
/ (¡J) o I 10 100 and light limitation, respeclively. The attenualion faclors take on values between
a (rngrn-3) 7.11 4.115x W" 7.2 x 10"" Qand 1, specifying complele limilation (O) and no limilalion (1).
/ The shapes of the individuallerms in Eq. 33.4 can be explored hy conducting
Thus in only 100 d we would have a ridiculously large quantity of plants. experiments. As displayed in Fig. 33.1, Ihe effect of temperature, nulrients, and lighl
can be determined by holding two of Ihese characterislics lixed while varying the
In nature such levels are never reached because, along with growth, there are
Ihird. Such experiments indicale Ihal in general all Ihe faclors seem 10stimulale .
a number of loss processes. Some of these are transport-related such as settling and
growth. However, light and lempcrature show a suppression of growlh al high lev-
diffusionldispersion. Others are kinetic such as respiration, excretion, and death by
els. We will now look al how Ihese effects cun be formulated mathemalicully.
predation. Further, the growth rate itself is not a simple COrlstant,but varies in re-
sponse to environmental factors such as temperature, nutrients, and ligh!. At. low
levels, and in some instances at high levels, these factors can limil growth. 33.2 TEMPERATURE
To incorporate these effects, Eq. 33.1 can more realistically be written as
da A variely of formulations have heen developed lo reprcsent the cffect nf lempcra-
(33.3) ture on plant growth (Fig 33.2). The simplesl is a linear model with some minimum
dt = kg(T, N, /)a - kda
temperature below which growth does not occur,
where kg(T, n, 1) = growth rate as a function of temperature T, nutrients N, and kg.T =O T :5 Tlllin
light 1 and kd = loss rateo Later in this lecture we learn that the loss rate is also
T- Tlllin (33.5)
not a simple constan!. However, for now, we will focus on how the growth rate is T> Tlllin
formulated. kg.T = kg.rd 7'rer - TIIIIII

where kg,T = growth rate (d -1) at lemperature T(°C)

kg . kg.rer= growth rate (d-I) allhe reference temperuture Tr.r(OC) "

(a) Experlment 1:
Tmin = temperature below which growth ceases
o()ptlmallight
'Excess nutrients A more commonly used form is the theta model previously describcd in Lec. 2,
n\ dry iemperature
kg.T = kg.20(}T-20 (33.6)
T Eppley (1972) pro posed a value of () = 1.066 based on a large number of sludies
involving muny species of phytoplanklon. Recall from Lec. 2 Ihal Ihis coefficicnt
kg value roughly corresponds lo a doubling of the rate for a 10°C rise in temperature.
(b) Experlment
oOptimallight
2:

nFixed temperature
.. 8
(20.C)
nVary nulrlenl
6
N
i 4
FIGURE 33.1
(e) Experlment 3:
k, 'k k.c,upl
The results 01 three experimen s 2
nFixed temperature to determine the phytoplankton
(20.C) growth rate as a lunctlon 01 (a) k,,zo FIGURE 33.2
nExcess nutrients
temperature, (b) a nutrient,
~
an I
O
Some 01 Ihe models used lo
nVarylight (e) light. The lines are intended O Tm1n 20 T... T"...40
characlerize Ihe effecl ollemperalure
lo Indicale the general patlern¡ T (.C) on phyloplankton growth.
I exhibiledbyIhepoinls. ,
606 PARTv Eutrophication and Temperature LECTURE33 Plant Growth and Nonpterlatory LosseA 601

Several Q!2!Jelshave been proposed to represent a temperature dependenee that TABLE 33.1
is zero at a minimum temperature, inereases to a peak growth rate at an optimal Half-saturatlon constants
for nutrlent IImltatlon of
temperature, and then deereases at higher temperatures. The simplest approaeh is a
linear representation, phytoplankton growth
Nutrlent k,
k¡¡.T = O T :S T min
Phosphorus 1-5 ¡LgP L-I
T - Tmin Nitrogen 5-20 ¡LgN L-'
T min :S T :S Topt
k¡¡.T = k¡¡.op,Topt - Tmin (33.7) Siliea(diatoms) 20-80 ¡LgSi L-I

Tmax - T
T > Topl I
k¡¡.T = k¡¡.opl T max - Topt
Vogel (1974) showed that major algal groups had differing sensitivity to temperature
Other investigators have suggested various funetions to fit sueh a shape smoothly (Fig. 33.3). Their work iIIustrated that groups such as diatoms grew at low temper-
(e.g., Shugart et al. 1974, Lehman et al. 1975. Thornton and Lessem 1978, ete.). atures. In contrast blue-green algae grew better at high temperatures. Inclusion of
For example Cerco and Cole (1994) have used the following formulation based on such effects in water-quality models is neeessary to determine their influence on
Ihe normal or bell-shaped distribution: eompetition among groups.
k - k -Kt(T-T.)2 T < T.
¡¡.T - ¡¡.ople .",. "" - Opl (33.8)
= k ¡¡.opte-K2(TnI":'n2\
k ¡¡.T T > T J/ 33.3 NUTRIENTS
lopt

where KI and K2 are parameters that determine the shape of the relationship of
glOwthto temperature below and above the optimal temperature, respeetively.
Which approach is best? Equation 33.6 is typically used when phytoplankton
are simulated as a single state variable. This usually implies that there will always
he species that grow at any particular temperature. Further it can also be used when
the simulations are performed for temperatures lower than the optimal temperature
for a particular species.
When several individual species or groups are modeled. the level of detail
provided by Eqs. 33.7 or 33.8 may beeome necessary. For example Canale and
As previously described in See. 32.2, the Michaelis-Menten equation provides the
most common approach for handling nutrient \imitation,
N
I/JN= ..- (33.9)
..N
where N = concentration of the limiting nutrient and k..N = half-saturation con-
stan!. This equation is linearly proportional to eoncentration at low nutrient levels
and approaches a constant value of one at high levels (Fig. 33.4). Values of the half-
saturation constant are summarized in Table 33.1. Note that the half-saturation con-
stant may also vary depending on the form of the nutrient that is limiting (e.g., SRP

-
í
~2
3

¡...,
~
.:c1 1JN 0.5

10 20 30 40

T (OC) ks
N
FIGURE 33.3
FIGURE 33.4
Temperature dependence 01 growth rates lor several algal
The Michaelis.Menten (or Monod) model 01 nutrient limitation.
groups (Canale and VogeI1974).

--
 ---

J
lEnURE 33 Planl Growth and Nonpredatory Losses 609
608 PARTv Eutrophication and Temperature

vs. dissolved P for phosphorus; ammonium vs. ni trate vs. total inorganic nitrogen for
nitrogen).

Multiple nutrients. There are several ways in which the nutrient limitatiOn Extemal Uptake
term can be refined. The most important relates to how multiple nutrients are han- nutrient -
dled. The most common case involves phosphorus and nitrogen. For such cases, .'
~

separate limitation terms would be developed for each nutrient,

- p
4>p
- _k'P + P (33.10) FIGURE 33.5
Internal nutrient pool phytoplankton
n Phytoplankton cal! mode!.
and 4>n = -k,n + n (33.11)
Luxury uptake and variable stoichiometry. Most water-quality models as-
where p and n are concentrations of available phosphorus and nitrogen, respectively. sume a fixed stoichiometry. However, il is known lhat phytoplankton stoichiometry
Three primary approaches are used to determine the combined effect of the nu- varies.
trients: A variety of mode1shavc becn developcd to simulatc this phcnomcnon. Thcy
l. Multiplicative. In this approach the two limitation terms are multiplied, as in are all based on lhe nOlionlhat nutrient growth limitation is a function of the phyto.
plankton's internal concentration rather than the nutrient concentration in the water.
4>N = 4>p4>n (33.12) This necessitates a separation of the mechanisms of uptake and growth (Fig. 33.5).
Thus it is assumed that the nutrients have a synergistic effect. That is, several The first attempts to model the phenomenon used the Michaelis-Menten formu-
nutrients in short supply will more severely limit growth than a single limiting . lation to model uptake,
nutrient. The multiplicative form has been criticized as being excessively low v = vmaxN (33.16)
when several nutrients are limiting. It also suffers in that, as more nutrients are km +N
included, it becomes more restrictive. where V = uptake rate
2. Minimum. At the other extreme is the case where the nutrient in shortest supply Vmax = maximum uptake rate
k,u = half-saturation rate for uptake
l
controls growth, I
4>N= min{4>p.4>n} (33.13) The growth limitation term was then computed as (Droop 1974)
This type of approach, which is similar in spirit to Liebig's law oIthe minimum, l - qo
4>N = (33.17)
is the most commonly accepted formulation. q
3. Harmonic mean. The reciprocals of the limitation terms are combined, as in where q = internal nutrient concentration, mass of nutrient per dry weight of al-
m gae, and qo = minimum internal concentration. Subsequent investigators (e.g., Auer
4>N= "'1 (33.14) 1979, Gotham and Rhee 1981, etc.) have refined this basic approach.
2:- The internal nutrient pool models have not been widely used (noteworthy ex-
j=l4>j ceptions in engineering-oriented water-quality modeling are Grenney d al. 1973,
where m = the number 01'limiting nutrients. For the case of nitrogen and phos- Bierman 1976,and Canale and Auer 1982). However, models are now bCIIIgdevel-
phorus, it would be oped to acknowledge that cell chlorophyll content varies. These modcls also includ~
2 variable nutrient content. We will describe such approaches at the end of this lecturc.
4>N= 1 1 (33.15)
-+-
4>p 4>n 33.4 LlGHT
This formulation was developed to allow some interaction among multiple limit- The effect of light on phytoplankton growth is complicated by the fact that several
ing nutrients while not being as severe as the multiplicative approach. However, factors have to be integrated to come up with the total effect. As summarized in ~.
it bas been criticized (Walker 1983) as being unrealistic for the case where one 33.6, these factors are diurnal surface-light variation, light attenuation with depth,
nutrient is limiting. for example suppose that phosphorus is limiting (near zero) and dependence of the growth rate on light.
and nitrogen is in excess (neaeone). for this case Eq. 33.15 approaches 24>prather The dependence of the growth rate on light can be quantified by experiment.
than 4>pas expected. As depicted in Fig. 33.6c, lhe growth rate peaks at an optimallight level~A variety
 ,,..1 610 PARTv Eutrophication and Temperature LECTURE33 Plant Growth and Nonpredatory tosses 61 í

where 1 = lightleveland l. = optimallight leveÍ. Note that 1. ranges from áboltt

(a) Dlurnal 100 to 400 Iy d-I. Lower values are appropriate ror low-light-adapted species,
varlatlon o. whereas high values are for those that are adapted to high light intensity.
IIght The temporal variation in light can be characterized by a half-sinusoid. For 8uth
Mklntght Sunrlse Noon Sun'" Mklnlght cases the average light over the daylight hours can be computed as (Eq. 24.11)

D 11\~ (33.20)
la=lm(~)
~
I where 1m = maximum light intensity andf = photoperiod (sunlight fraclion of day).
(b) Llght ! Thus the average daylight value is about ~ of the maximum.
attenuatlon Mlxedlayer The spatial variation of light down through the water column can be modeled
wlth depth
I
-.- - - - - - - - - - - - - - - - - - - - - - - - - - -'- - - - - -
by the Beer-Lambert law,
l(z) = loe-keZ (33.21)
where lo = solar radiation at the surface añd ke = extinction coefficient. The latter

D ".r,
can be related to more fundamental quantities by (Riley 1956)
= k~+ O.0088a
+ O.054a2l3
k' 8 ke (33.22)

(e) Dependence
o. growth rate
on IIght

FIGURE 33.6
~ 11\~
Ü
where k~ = light extinction due to factors other than phytoplankton, which can be
either measured directly or ca1culatedvia (Di Toro 1978),

where kew
k~ = kew+ O.052N + O.174D
= light extinction due to partic1e-freewater and color (m-I)
N = nonvolatile suspended solids (mg L-1)
(33.2j)

i\
"
1:
Incorporating light into the phytoplankton growth model involves
integratingthree separatelactors: (a) the temporal variation 01 solar
radiation at the water surface, (b) the attenuation 01 light as it passes
through the layer 01 water being modeled, and (e) the effect 01 the light
D = detritus (nonliving organic suspended solids) (mg L-1)
AII the preceding formulas can be applied to compute !he mean light limitation
for a weIl-mixed layer (Fig. 33.7). For example using the Steele model, we lirst
on plankton growth.
substitute Eq. 33.21 into Eq. 33.19 to give an equation for!he growth limitation at
I
depth
z, --
of models have been developed to lit light dependence. For example a Michaelis-
Menten formulation is sometimes used, lae-k,z _l.ae-1e'+ I

F(/) = .
1
k... + 1
_ (33.18)
-i!."
F(/) = -e
l.,
l., (33.24)

This function can then be integrated over depth and time to develop !he mean
value
:I where k.,¡ = a half-saturation constant for light (ly d-I). Thus, as for nutrient lim-
1 H2 I fTp lae-kez
I
itation, this relationship is linearly proportional to light intensity at low light levels
and constant at high levels. Consequently it does not capture growth attenuation at
high light intensity. As was the case for Eq. 33.6 for temperature, such a formu-
cPL= - -
f
H H, Tp o
-e
l
l.,
_lae-1e'+1
l., dtdz (33.25)

~ Evaluating this double integral results in

lation might be justilied when the phytoplankton was simulated as a single group,
In such cases it would be implicitly assumed that (1) some species of phytoplankton
~
would grow optimaIly at any particular range of light intensity and (2) increasing
light always leads to increasing growth. It would also be appropriate for a single
species when the experirrients were conducted below the optimallight intensity.
A somewhat different approach was taken by Steele (1965), whose model ac-
knowledges that growth is inhibited at high light levels, FIGURE 33.7
HCO' H¡ -------.-._------------------- A layer 01 water 01 thickness H =
1 _L+I H2 - H,. For this particular case, the
F(/) =-e l., (33.19) surface layer is deplcted (H, = O).
1"
: .III ~ -,-

612 PARTv Eutrophication and Temperature LECTURE33 Planl GroWlh and Nonpredalory L"",.:s 613

(a) Compute Ihe phytoplankton growlh rale. Assume Ihar Ihe suspended solids concen-
2.718/ (e-a, _ e-a,,) (33.26)
cPL = ---¡¡;¡¡ Iralion (olher Ihan phyloplanklon) is negligible.
(b) Delermine Ihe primary produclion rale in g m-2 d-' if Ihe chlorophylllo-carL0n
where ratio is 20 ¡.LgChla mgC-1 .
1
ao = ...!!..e-k,HI (33.27) Solution: (u) The eXlinclion C()efficien~ c,an be dctenJlined by
ls V P V'
k, = 0.3 + 0.0088(4) + OS4(4)2I3 = 0.471 m-I
la -k H (33.28)
and al = -e ,2
ls which can used in conjunction wilh Ihe Ijghl dala lo ca1culale
SOO 1
It should be noted that the light values used in all the preceding equations is ao = _e-O,47I(O) = 1667
3<>2 .
visible, photosynthetically ,!' 'Iable ligh!. This value is typically about 40 to 50% ,-
of the energy in the compkh. standard spcclrum used for such calculations as heat
306e-0.471(51 =
and ,
budgets (Lec. 30) or photolysis (Lec. 42). For example Bannister (1974) and Stefan al = SOO 0.158 I

et al. (1983) suggest values from 44 to 46%. It should be noted that almost all the
radiation outside the visible range is absorbed in the first meter below the surface which can be subsliluled inlo Eq. J3.29 lo give
(Orlob 1977). ~ lo ~
= 2 . 1066 25\-20 2.718(0,S) . 20
l
k
g .,
[
0.471(St
(e -I,667_

e
-O'15H
)]
mm
( ,
,

10,+ 2(f 2.+3

,

) ¿.
33.5 THE GROWTH-RATE MODEL 2.7S3 X 0.3838 X L¡0.6 = 0.634 d

Thus we can see Ihal the lempcralure seIs Ihe maximum rate, whereas tllt, lighl and
The complete model of phytoplankton growth can now be developed as
nulrients diminish Ihe ultimate value. Furlher, Ihe equalion indicates thal phosphorus

kg = kg.20l..066T-20 2.718/ (e-al .

- e -ao mm ~ ' -L is the limiting nutrient.

(km+n k ~+p (33.29) (b) The rate can be lranslated inlo a daily prirnary produclion rale by
[ k eH )] )
Temperature Lighl Nutrients Pr = a,."kgHa
..
.'
This formula has a variety of applications. In a later lecture we use it to calculate SUbsliluling val~es v (..
phytoplankton growth as part of a nutrientJfood-chain model. In addition it can be 0.634 t 1000L
= ImgC P.gChI1l--.!.L
employed to calculate some other quantities of interest to water-quality modelers.
For example if the quantity ol' phytoplankton is known, it can be used to compute
P
r ( )
20p.gChlu d (m S )(4 L )( IOOOmg )( m3
= 0 634"C -21
. c> m , )
the primary production,
Pr = amkgHa
33.6 NONPREDATORV LOSSES
with units ol' gC m-2 d-I, or the oxygen produced by photosynthesis,
P = rocacakga A numberof processesconlrihulelo Ihe 1055raleof phyloplanklonin Eq. 33.3. In
waler-qualily modeling, Ihree losses are cmphasized:
with units of gO m-2 d-I. Finally it can be used to gain insight into which l'actors
are limiting phytoplankton growth. AII the~eapplications are explored in the follow- · Respiratioll. This refers to Ihe process Opposilelo pholosynthesis, whcre Ihe plalll I

ing example.
· utilizes oxygen and releases carbon dioxide.
Excretion. This process has lradilionaIly l'ocused on the release of nulrients. How-
'+ EXAMPLE 33.1. PHYTOPLANKTONGROWTH RATE. The epilimnionof a
lake has tbe following parameters:
T = 20°C la = SOOIy d-I
· ever, algae can also release organic carbon as eXlracellular byproducls.
Predatory losses. Oealh of algae due lo grazing by zooplanklon.

Because the first Iwo processes are difficult lo measure separately, they have
ls = 300 Iy d-I k~ = 0.3 m-I usually been modeled as a single first-order decay. Thus Ihe death rale from Eq.
Available P concentration = 3 mg m-3 P half-saturation constant = 2 mg m-3
N half-satUralion constant = 10 mg m-3
33.3 is usuaIly expanded, as in
Available N concentration = 20 mg m-3
,~, ~/11orophylla concentration = 4 mg m-3 f = O.S kd = kru + kgz (33.30)
1:..20= 2 d-I H = Sm
 614 PARTv Eutrophication and Temperature LECTURE33 Plant Growth and Nonpredatory Losses 615

where kra = loss due to the combined effects ofrespiration and excretion (d-I), and 10
k¡¡z= grazing losses (d-I). Values for krarange between 0.01 and 0.5 d-I , with typ-
ical values on the order of 0.1 to 0.2 d-I . A theta model is usually used to COITectthe ~
respiration/excretion rate for temperature. A value of () = 1.08,connotinga strong QI 5
t"mperature effect, is conventional. ~
...

It should be noted that although they are often treated as a single process, the FIGURE 33.8
division between respiration and excretion should not be considered a trivial distinc-
Flow diagram lor kinetic inleraclions
tion. This is particularly tme as nutrient/food-chain models evolve toward a more 10 20
between planls and a limiling nulrienl in
IIccurate representation of the organic carbon cycle. In such cases, processes that (d) a eSTA.
IClldto generate carbon dioxide and liberate available nutrients (respiration) should
be separated from processes that liberate organic forms of carbon and nutrients (ex-
cretion).
33.7 VARIABLE CHLOROPHYLL MODELS (ADVANCED TOPIC)
At this point we can now integrate the growth and decay mechanisms into our
modeling framework. In an analogous fashion to Eqs. 32.18 and 32.19, we can de- As describedto this poinl, mosl waler-qualily modelshave used a conslantstoi-
vclop the folJowing mass-balance equations for a limiting nutrient and algae for a chiometry lo characlerize algal concentralion. An easy-to-measure quantity such as
C'STR: chlorophyll a was usualJy adopled as the measure of algal biomass. Transfers be-
tween phytoplankton and other model slate variables such as nutrient pools were
-da _ P _ 2.7181 -a, _ -ao Q
e ) - k ra - _
thenhandledby simple stoichiometricconversionfactors.
k sp + P kr H (e
d t - ¡¡.r- v] a (33.31)
Although this has proved to be a good first approximalion, il has long been un-
[k
derstood by p/rycologists (scientists who study algae) that celJ stoichiometry is not
dp
- = -a/1ak¡¡.r- p 2.718/ ( _a
d(
_ -a
k .'p + P krH e I - e 11
Q
)a + apakr.aa+ _v (Pin - P)
constan!.We havealreadyalludedtothis inour earlierdiscussionof luxuryuptake.
Beyond nutrients, a more fundamental issue relates to carbon and chlorophyll.
In particular,the chlorophyll-to-carbon ratiois not constant,but variesin response
(33.32)
to light levels and the physiological state of the cells. Reviews of the published lit-
Notice that we have omitted grazing losses, which will be described in more detail eraturé (e.g., Bowie et al. 1985) indicate Ihat the chlorophyll-to-carbon ratio varies
\ in Ihe folJowing lecture. between about 10 and 100 JLgChlamgC-1 .
There are two reasons why such variability could have significance in the com-
ing years:
EXAMPLE 33.2. ALGALlNUTRIENT INTERACTIONS WITH LlGHT LlMI-
¡) TATION. Perform the same calculation as in Example 32.5, but now inelude the effect . The focus of modeling is shifting from nulrient/food-chain interactions to an or-
of light limitation. Note that the epilimnion has the following characteristics at the be- ganic carbon cycle characterization. This is due to the application of water-quality
ginning of the stratified period: frameworks beyond eUlrophicationto encompass problems such as toxic pollution,
ao = 0.5 mgChla m-3 Po = 9.5 mgP m-3
sediment-water interactions, and Ihe impact of disinfection byproducts on drink-
Pin = 10 mgP m-3
a{>fl = 1.5 mgP mgChla-I k¡¡.T= I d-I
ing water.
k;p = 2 mgP m-3
km = 0.1 d-I T...= 30d-1 f = 0.5 · Water-quality models are being increasingly used to analyze cleaner systems
la = 400 Iy d-I 1, = 250 Iy d-I H=lOm than were studied in the pas!. Many water bodies are cleaner because their load-
k; = 0.1 m-I ings have been reduced through waste treatmen!. Further, new problems such
, ; as drinking-water quality are being addressed. Because heavily polluted systems
Assuming that mass transfer across the thermoeline due to diffusion and settling is neg:. are more turbid and dimly iIluminated,they tend to have a more constant
ligible, simulate how the algae and phosphorus change over time. chlorophyll-to-carbon ratio. In contrast clean water bodies typically exhibit more
variable light levels because significant light can penetrate beyond well-mixed
Solution: Using Eqs. 33.22 and 33.26 to quantify the light effect, we display the re- surface layers. The existence of deep chlorophylllayers in some estuarine systems
sults graphically in Fig. :n.8. In contrast to Fig. 32.9, two major differences are evident.
and clear lakes provides circumstantial evidence for such effects.
First, the final algae leve! is lower and the phosphorus higher for the present case. Sec-
ond, the time to steady-state is longer when light limitation is ineluded. Both are due Over the past decade a number of models have been developed to account for
to the reduction of algal growth from light limitation. Thus, because the growth rate variable clUorophyllcontent in plants. Today the models are beginning to be inte-
is smaller, the final partitioning between plants and nutrient is less extreme. Similarly grated into water-quality frameworks. The purpose of Ihis section is to ilIuslrate how
the response time is lengthened.
such integration can be done.
 ....

616 PARTv Eutrophicalion and Temperature lECJ"UKE 33 Plant Growth and Nonpredatory Losses 617

Laws and Chalup (1990) have developed such a mode!. In their approach, TABLE 33.2
growth and respiration rates, along with chlorophyll-to-carbon ratios, are simulated Parameters developed by Laws and Chalup (1990) for the application of their
as a function of light and nutrient levels. Because they focused on a marine diatom, -.-... ........... -
algal growth model to Ihe marine diatom Pavlova III/ller;
Laws and Chalup used nitrogen as their limiting nutrient. Consequently they also Parameter Symbol Value Range Units
computed a nitrogen-to-carbon ratio for phytoplankton biomass.
Rale of change of T per unil T, 0.28 0.03-0.43 gC cell 1d I
To derive their model, Laws and Chalup developed mass balances around a
change in gross photosymhe-
phytoplankton cell for four types of cell carbon: structural carbon, storage carbon, sis rale per cell
and carbon associated with both the light and dark reactions of photosynthesis. The
Basal respimljon rale T"/C 0.03 -0.JI-O.16 d-I
resulting model can be represented by the following coupled algebraic equations:
Ralio of 10lal cell C 10 nUlri- WN 6.9 6.2-7.7 gCgN 1
~.- _ K.(I - rg)(I - S/C)I _ ro enl in slruclural, lightldark
- (33.33) reaclions of C
K.I/pO + 1[1 + k.I(I,/pO)] C
QuolÍenl of nUlrienl-limiled F 0.22 0.17-0.27
rlC = (ro/C + rg~)(I - rg) (33.34) N/C ralios al relalive growlh
(
rales of O and 1
I
N/C = F + (1 - F)(I - ~/~s) (33.35) Gross pholosymhesisrale per K, 3.6 2.7-4.2 d '
WN
1

unil quamily of dark reaclion

,1
carbon
'!
I - (1 - F)(I - ~I ~s) - SIC _ ~ + ro/C RalÍo of 10lal cell C lOchloro- W'b' 17 13-34 gC gChl I
ChlalC = (1 - rg)K.
(33.36) phyll a in Ihe lighl reacljon I
WChl componenl of C

where 1 = incident irradiance (mol quanta m-2 d-I) Struclural C per lolal C SIC 0.1 -0.2-0.2
~ = algal growth rate (d-I) Value of gross rale of pho- /,.. 0.28 O.16-0.51 m2 mol quanla-'
~s = nutrient-saturated algal growth rate (d-I) losymhesis per unil lighl
r = respiration rate per cell (gC cell-I d-I) reaclion carbon per unil lighl
inlensily in Ihe limil of zero .\
The remainingmodelparametersalongwithtypicalvaluesare listedinTable33.2. irradiance
Insight into the model can be gained by substituting the values from Table 33.2 Half-saluralion irradiancc /.. 64 10--122 0101'1"'111111111 2 d I
I
into the model equations. For example we can first substitute the parameters into
Eq. 33.33 to give
1 11
1 = max-- (33.39)
(33.37) k.,;+ 1 km + 11
~s = ~max ks; +1
Consequently we now can see that growth is related lo light and nutrients in a simi-
where ~max = maximumgrowthrate at saturatedlight and nutrient levels and lar fashion to Eq. 33.4. That is, it corresponds to a maximum rate limes light ami
ti = light half-saturation constant. For the parameters in Table 33.2, ~m.x = nutrient attenuation coefficients.
. d-I and ks; = 10.7 mol quanta m-2 d-I. The latter value can be converted Next, we see that rcspiration is a simple linear function ot' growth,
,) more familiar units by multiplyingit by a conversionfactor of about 5 Iy dC = 0.042 + 0.389~ (33.4())
(mol quanta m2)-1 to give 53.5 Iy d-I. Thus the model specifies that nutrient-
saturated growth is related to light level by the Michaelis-Menten model specified In other words, beyond a small basal metabolism rate of 0.042 d-I , growth-related
earlier in this lecture (Eq. 33.18). respiration will be about 39% of the actual growth rateo
If the Michaelis-Menten model is also assumed to apply to nutrient limitation The nitrogen-to-carbon ratio is linearly relaled to the cells nutrienl cundition,
as in
(as we did previously in Eq. 33.9), the ratio of the actual to the saturated growth rate
would be specified by N/C = 32 + 1131:. (33.41)
~s
~ _ n
(33.38) where we have changed the expression ofthe ratio from Laws and Chalup's gN gC-1
~s - ksn+ n
to the mgN gC-1 that is more consistent with the units used in previous sections.
and f~e actual growth rate by the product of Eq. 33.37 and 33.38,
Thus in plants grown in nutrient-poor environments (~I ~s = O),the nitrogento
 618 PARTv Eutrophication and Temperature
.. LECTURE33 Plant Growth and Nonpredatory Losses 619

carbonratio is 32 mgN gC-I. For nutrient-richwaters(pJ11-..= 1), it approaches 60

145 mgN gC-r. Recall that in Lec. 28 we developed an average N/C ratio for
,........
plant matter of 7200 mgN/40 gC = 180 mgN gC-1 , which corresponds to the upper
40
bound of the range. Thus the model implies that phytoplankton are at the upper
bound when they are in a nutrient-rich environment, and drop to about 20% of '-'
!heir maxirnum level when nutrients are severely depleted. olu 20
Lasl hut not Icusl. Ict's look at the chlorophyll-to-carbon ratio

Chla/C = 6.378 + 45.882 11-..

E:...- 22.711- (33.42)
~>~ 11,
FIGURE 33.9
Relationship 01 chlorophylHo-
200 400
carbon ratio to light lor various
states 01 nutrient enrichment
whcre again 1 have changed the units from gChla gC-r to mgChla gC-1 . I (Iy <1")
(Jif Ji,).
This relationship indicates that nutrient enrichment tends to raise the ratio
, whereas ligh! tends to lower it. It can be simplified further by substituting Eqs.
/. =.!!!
33.37 and 33.38 to give
.IV k,H (1 - e
-k,H
) (33.48)

Chla/C = 6.378 + 45.9-

k.."+
11
11
- 44-
11
k.."+ 11
1- _k..i I+ I (33.43) where lo = surface light (ly d-I).
t Thus !he lirst lerm in Eq. 33.46 is net growth (photosynthesis minus growth-
or collecting lerms and assuming that 45.9 and 44 can both be approximated by 45, related respiration) and the second term is basal respiration. Note how the N/C ratio
11 ksi
is employed to translate the plant photosynthesis and respiration sources and sinks
(33.44 ) into sinks and sources in the nitrogen balance.
Chla/C = 6.4 + 45 k.u,+ 11 k.o; + I Finally where does Chla/C lit in? In the present framework it would be em-
Now this relationship provides additional insight into the mode!. The last term indi- ployed to translate algal carbon into algal chlorophyll for use in determining the
cates that Ihe negative effect 01'light on the ratio has a hyperbolic shape that is the extinctioncoefficientas in Eq. 33.22.It could also be usedto makethe conversion
mirror irnage of !he Michaelis-Menten curve. Tha! is, Eq. 33.44 can be expressed as in order to compare model output in carbon units with measured chlorophyll concen-
trations.
Chla/C = 6.4 + 1 - q,/J
45q,,,( (33.45)
EXAMPLE 33.3. ALGALlNUTRIENT INTERACTIONS WITH VARIABLE
The result, as displayed in Fig. 33.9, illustrates that higher chlorophylllevels should CHLOROPHYLL. A batchsystemhas the following characteristics:
occur in darker waters and when the cells have plenty of nutrients. Conversely they ao = 0.5 mgChla m-3 /lo = 700 mgN m-3
would have low chlorophyll in highly illuminated waters or when they were nutri- kR.T = 1d-I k,,>= 15mgNm-3
en! limiled.
k.t., = O.I d I Tu'= 30 d
Nowil canbe seenhowthis modelcanbe integratedintoourlargeralgalmodel- f = 0.5 ('v = 400 Iy d-I
ing framework. For the simple case 01'phytoplankton and a single limiting nutrient, H=IOm k~=O.IIII-1
IllaSSbalunces for a CSTR can be written as
Determine how Ihe algae and phosphorus change over time.
dc"
= 0.389
( -
lav
_k + 11 Q
I
( f
1.94
53
.. 5 + 1av )
11
c" - 0.042c"- -V c" (33.46)
Solution: Using Eq. 33.22, the initial extinction coefficient can be computed as
k, = 0.1 + 0.0088(0.5) + 0.054(0.5)113 = 0.138 m-I
dll lav 11 Q which can be used in conjunction with Eq. 33.48 to determine (he average irradiance in
dt
(
-(N/C) [ 0.389 1.94535
-. .¡_I JVk.." + 11 C" - 0.042c" ] + _V )
(lIio - 11) the reactor,

(33.47) 1 =
ov
400
(
0.138(10)
1 - e-O.138(JO)
) = 216 .61y d-I
where 11 = concentration of available nitrogen (mgN L-1)
C" = algal carbon (mgC L 1) The initial nutrient saturation state and growth rate can be computed as
lav = averageirradianceforthe layer(lyd-I ) /L 700
/L, = 15 + 700 = 0.979
For a constant !emporal light source, thc average irradiance would be related to depth
and light extinction by 216.6 700 _1
/L = 1.94,~, . ~ U / 15 + 700 = 1.94(0.802)0.979 = 1.523 d

---'
, ..' *~;" -" ...,a"- ~... .

620 PARTv Eutrophication and Ternperature LECTL/HE

33 Plant Growth and Nonpredalory Losses 621

Thus, as rnight be expected, the systern has very liule light or nutrien!limitation. Initial syslems. However, the algal chlorophyll rises to a peak and lhen declines. The reason
N/C and Chla/C ratios can be deterrnined with Eqs. 33.41 and 33.42, for Ihis behavior is illuslrated by Fig. 33.3c, which shows how the chlorophyll-to-carbon
ratio is affecled by lighl and nutrients. As lhe simulation progresses. light is diminished
N/C = 32 + 113(0.979) = 143 mgN gCI because 01'shading by Ihe increasing algal biomass. This leads lo an increased ChICl/C
Chla/C = 6.378 + 45.882(0.979) - 22.7(1.523) = 16.72 rngChl gC' ratio. The resulting peak is Ihen rapidly dilllinished as the ralio drops when the nutricnls
are even!ually used up.
The nitrogen-to-carbon ratío is near its upper bound because there are abundan!
nutrients. The chlorophyll '>icarbon ratio is near its lower bound because Ihe irradiance
is high. The chlorophyll-to-carbon ratio can also be employed lo calculate the initial algal
biomass as carbon,
The foregoing example is intended to ilIustrate how variable carbon can be in-
tegrated into water-quality frameworks. Jt also suggests a number of implications of
mgChla gC -.)
c. = 0.5ffi3 (16.72 mgChla ) = 0.0299 gC m
the mechanism. According to the model,
·
.

Because it varies, chlorophyll is a deccptive measure of tme biomass. This is

Now, using these initial conditions, Eqs. 33.46 and 33.47 can be integrated (with
Q/V = O to simulate the batch systern). Al each time step, all 01' the ratios are recorn- somewhat unfortunate since the chlorophyll a measurement provides such a con-
puted. The results are shown in Fig. E33.3.
The rcsults are interesting. The nitrogen and the algal carbon fall and rise as sig-
moid (that is, S-shaped) curves as would be expected from our previous rnodels 01'batch
venient and economical way to discriminate plant biomass from other particulate
mattcr in natural waters. However, if the foregoing model is an adequate descrip-
tor of reality, the caIculated ratio provides a means to translate chlorophyll into
l
calbon.
· Because of the opposing effects of nutricnts and IiglH 011the ratio, the mcchanism
n seems to sharpen up growth events. This bchavior might allow water-quality mod-
c. Chis (a) els to better simulate both temporal and spatial gradients of carbon and chlorophyll
(mg L-1) (¡tg L-1) (1I9NL-I)
20 200 1400 in natural waters. Examples of such phenomena are spring blooms, deep chloro-
phylllayers, and horizontal gradients in waste plumes.
n
10 100 700
PROBLEMS
O
o1 o 33.1. A very clear body 01'water has Ihe following condilions:
O 5 10 15 20 25
T = 10°C Available P concenlration in hypolimnion
1.=700lyd-1 = 4 rngm-)
200 (b\ Is = 250 ly d-I P half-saturation constant = 2 mg m-3
k~ = 0.2m-' f = 0.5
(same in epi and hypo) kg.21J= 2 d-I
100
Calculale Ihe phytoplankton growth rate in the hypolimnion. Note that th~ ,,'Clayer
has a Ihickness of 4 m and the hypolimnion has a thickness 01'3 m. Assume that phos-
O I I I I I phorus is the limiling nulrient.
O 5 10 15 20 25
33.2. Use (a) Eq. 33.7 and (b) Eq. 33.8 to fit the effect of temperature on the growth ofthe
o ftagellates, as in Fig. 33.3.
:¡; (e)
f! 33.3. Develop the equalion for light limitation (comparable to Eq. 33.26) for 11 ',,'haelj,,-
g
.l:. Nltrogen
Menlen form of Eq. 33.18.
(J 33.4. Repeat Example 33.2, but include a phytoplanklon seUling velocily 01'0.2 m d-I.
'O 0.5
e 33.5. Perforrn a sensitivity analysis (recall Lec. 18) on Exarnple 33.2 for the following pa-
o ramelers: ao. Po, Pin, kg.r, ksp, k,., and Is.
;:1
e 5 10 15 20 25 . 33.6. Duplicale Ihe compulalion from Exarnple 33.3, bul for a CSTR wilh a residence lime

i.. Time (d)

of 10 d. Discuss how ftushing modifies Ihe shapes of Ihe curves in Fig. E33.3.

FIGURE E33.3
 J LEcruRE 34 Predator-Prey and NutrientlFood-Chain Interactions 623

LECTURE 34 markets from 1910 to 1923. On the basis of the data, he hypothesized that lack ol
- - - --- fishing during World War I led to a higher proportion of predators. He cornmunicated
his findings to his father-in-Iaw, the noted mathematician Vito VoJterra. Over the
predator-Prey and following year, Volterra developed a number of mathematical models to simulate
the interactions among two or more species. Independently the American biologist
NutrientlFood-Chain A. 1. Lotka carne up with many of the same models.
There are a variety of these models, which are now commonly called LotluJ.
Interactions Volterra equations. In this section we describe the simplest version. To do this we
can first write a growth equation for a single prey species in an isolated environment
(that is, no predators and plenty of food). For such a case a simple first-order model
can be written as
dx
- = ax (34.1)
dt

where x = number of prey and a = a first-order growth rateo

" Next an equation can be written for a single predator y in the absence of its sole
food source x,
dy (34.2)
dt = -cy

where e = a first-order death rate.

LECTURE OVERvmw: 1now describe the predator-prey Jnte~actionstháldictate an Now the interaction between the two species should depend on both their mag-
important component of plant death. To do this 1 introduce you ti:>the ml1the- nitudes. If there are either too few predators or too few prey, the interactions would
matics of predator-prey interactions. Thell.. apply this theory to phytoptanktorl decrease. A simple way to represent this interaction is as a product xy. Thus, because
I and their key predators: zooplankton. FinaUy 1 integrate predator-prey intemc- the interaction spells death for the prey, it can be added as a loss to Eq. 34.1,
I tions with a nutrient balance to form a simple nutrieritlfood-chairi moclet.
~ 'J- 1 dx
j ..) -d t = ax - bxy
~~
(34.3)
In the previous lecture I outlined some external physical and chemical factors, such where b = a parameter that quantifies the impact of the interaction on prey mortal-
as temperature. light. and lIutrients, that limit phytoplankton growth in natural wa- ity. Conversely for the predatOJ:,the interactioo represents a gain,
Icrs. I also dcscribed losses such as respiration and excretion. Now let's turn to fac- 'd
lors that limit algal populations by causing their death. In particular we'lI focus on
the action of predators such as zooplankton that use algae as a food source. To pro-
I ;, = -e y + dxy (34.4)
vide backgmund for these phytoplankton-zooplankton interactions, 1'11firstdescribe where d quantifies the impact on predator growth.
general mathematical models that have been developed to simulate predator-prey in- The results of integrating these equations is shown in Fig. 34.1. Note that a
teractions. cyclical pattern emerges. Thus, because predator population is initially small, the

34.1 LOTKA-VOLTERRA EQUATIONS

Aside from phytoplankton and zooplankton, there are other cases involving a pair of
organisms in which one serves as the primary food source for the other. For example
in southern oceans. the shrimplike Antarctic krill serves as the principal food source
lor baleen whales. Moose provide the primary food source for wolves on isolated
FIGURE 34.1
island ecosystems such as Isle Royale in Lake Superior.
Time-domain representation 01
In 1926 the Italian biologist Humberto D' Ancona estimated the populations of numbers 01 predators and prey
prcdator ami prey species in the uppcr Adriatic Sea based on the numbers sold in fish Time lor the Lotka-Volterra mode!.

(,22
 .,~.
~- ...---

628 PART
v Eutrophication and TemperatUre LECfURE34 Predator-Prey and Nutrient/Food-Chain Interactions 629
(
Note that the efficiency factor ranges between O and 1, where O means nq assimila- by our choice 01'paramelers and (he fact that we are using a batch system. Other pamm-
etcrs, including transporto such as flushing and seUling, can alter the size and phasing
tion and 1 means total assimilation. Consequently it provides a mechanism for es-
oí' the peak s You will be provided some opp0r!unity to explore sllch sensitivity in tht
tablishing how much of the prey biomass becomes predator biomass and how much end-oí'-lectllrt problems.
is released as detritus. Similarly the carbon-to-chlorophyll ratio provides a means to
translate ingested chlorophyll into zooplankton carbono 80th effects are illustrated
in Fig. 34.5b.
Now let's look at the phytoplankton-zooplankton equations under the simplified _34.3 ZOOPLANKTON PARAMETERS
case where temperature is constant, the half-saturation effect is omitted, and the net
Grazing rates generally vary between 0.5 and 5, with values most commonly in the
phytoplankton growth rate is a constant,
range from 1to 2 m3 gC-1 d- t. The temperature correction factor is generally on the
da high side, with a value of 1.08 being commonly used. The half-saturation constalll
dt = (kg - km)a - CgZza (34.21) varies between 2 and 25, with values most commonly between 5 and 15 ¡LgChlL-l. I1
dz Grazing efficiencies commonly vary between 0.4 and 0.8.
(34.22) Because it is a composite of several factors, the zooplankton loss rate must be
dt = (acaBCgz)za - kJzz
broken down into its component parts. As with phytoplankton, a common approach
Because all the parameters are constants, these represent Lotka- Volterra equations. is to break it into nonpredatory and predatory losses,
kJz = k,.z + kgzc (34.23)
EXAMPLE 34.1. ALGAL-ZOOPLANKTON lNTERACTIONS. A batch system
containing phytoplankton and zooplankton has the following characteristics:
where k,.z = nonpredation loss rate (d-I) and kgu' = predation loss rate (d-I), in
which the subscript e connotes that the predation is carnivorous. It should also be
ao = 1 mgChla m-3 lo = 0.05 gC m-3 recognized that there are omnivorous zooplankton that eat both animals and plants.
aca = 0.04 gC mgChl-1 CgZ = 1.5 m3 gC-1 d-I The nonpredatory losses consist primarily of respiration and excretion. Because
e = 0.6 kg - k,.a = 0.3d-I these are difficult to measure separately, they are lumped together. This rate is com-
kdz = 0.1 d-I
monly between 0.01 and 0.05 d-I, with reported values ranging between 0.00 I
Calculate a few cycles oí' holh organisms. andO.1 d-I.
Solution: The results 01' integrating Eqs. 34.21 and 34.22 with Ihe fourth-order RK The grazing loss is typically handled in two fashions. Where upper leveis 01'
method are displayed in Fig. E34.1. Note that we have expressed the phytoplankton the food chain are not modeled explicitly, it is often treated as a constant with a
concentration in carhon units so that the biomass values are directly comparable. We temperature correction. Conversely when higher predators such as carnivores are
have also included the total oí' zooplankton and phytoplankton on the plot. . simulated, it is modeled in a similar fashion to Eq. 34.17. We will illustrate how
The biota peak about every 45 d. The zooplankton peak about 5 d after Ihe phyto- such "food-chain" modeling is implemented in the following lecture.
plankton reach a maximum. Note that the behavior exhibiled by this plot is influenced Summaries 01'data for all the aforementioned zooplankton grazing parameteh
can be found in Bowie et al. (1985). This publication also summarizes alternative
formulations and approaches for modeling algal-zooplankton interactions.

34.4 NUTRIENT/FOOD-CHAIN INTERACTlONS

..'.Total :...

Now we integrate a limiting nutrient into our phytoplankton-zooplankton ~l.hemc to

~ complete our fundamental picture 01'how nutrient/food-chain interactiol1 :lrc moJ-
O 0.5 eled. Following the simple kinctic scheme in Fig. 34.6, the resulting e4u"".,lI1s could
01
E be written as
da
dt = (kg - kw)(/ - CgZw (34.24)

o dz
o 40 80 dt = (ameCgz)za - k,.zz (34.25 )

Time (d) tlp

elt = (/1',,(1- e)CgZza + af".k,.zz - af",(kg - k,.,,)a (34.26)
FIGURE E34.1
 g. LEcruRE 34 Predator-Prey aod NutrieotIFood-Chaio loteractions 631
630 PARTv Eutrophication and Temperature

FIGURE 34.6
A simple represenlalion 01
Ihe inleraclions simulaled in a
nulrientllood-chain mode\.
in equivalenl carbon unils so Ihallhe values of the Ihree forms are direclly comparable.
We have also included Ihe total of Ihe three componeols 00 Ihe plo\.
The biola peak aboul every 42 d. The zooplankloo peak about 1 wk after the phyto-
plankton reach a maximum. Notice Ihal the plaots recede because of the grazing ef-
fect ralher than the nulrienl limilalion (Ihe phosphorus never falls below 6 J.LgP L-I
during Ihe simulalion). In olher cases Ihe nulrienls could be depleled firs\. One value
of nulrientlfood-chain modeling is Ihe capabilily lo simulale and discriminale belween
such effecls. As wilh Example 34.1 Ihe behavior exhibiled by Ihis plot is inftuenced by
our choice of parameters and Ihe fact Ihat we are using a balch system.

The equations now form a complelely closed syslem where nulrients are turned
into biomass by net production and recycled by respiration and death. The following
example illustrates the resulting dynamics of the three components.
This lecture has been designed to iIIuslrate how predator-prey kinetics work and
to show how they can be integrated into a more comprehensive nutrientlfood-chain
computation. For simplicity we have limited our exposition to batch systems. In the
next lecture we expand the framework and implement it in an open, natural water
body: a stratified lake.

EXAMPLE 34.2. NUTRIENT/FOOD-CHAIN INTERACTIONS. A balch system

conlaining phyloplankton, zooplanklon, and available phosphorus has the following PROBLEMS
characlerislics:
34.1. Integrate Eqs. 34.3 and 34.4 numerically using fourth-order RK and parameter values
ao = 1 mgChla m-3 Zo = 0.05 gC m-3
ofa = I,b = O.I,c = O.S,andd = 0.02andinitialconditionsofx = 20andy = 5.
aca = 0.04 gC mgChl-1 a"a = 1 mgP mgChl-1
g = 0.6 34.2. Perform a sensitivity analysis on Example 34.1 as described in Lec. 18. Vary the fol-
Cgz = 1.5 m3 gC-1 d I
kg - km = 0.3 d-I káz = 0.1 d-I lowing parameters in your analysis: net growth (kg - k,a), CgZ' g, and ktJz.

[lo = 20 J.LJiPL -1 34.3. Repeat Example 34.1, but use the following parameter values:
Calclllale a few cycles of bolh organisms and Ihe nlllrien\. Include a nlltrienllimilation ao = 2 mgChla m-3 lo = 0.03 gC m-3
effecl inlhe phyloplanklon groWlh lerm. Use a value of 2 J.LgPL -1 for Ihe half-saturation aea = 0.04 gC mgChl-1 CgZ = 2 m3 gC-1 d-'
conslan\. B = 0.5 kg - km = 0.5 d-I
Solution: The nulrienl limilalion lerm is included by modifying the growth term, kdz = 0.075 d-I
as in 34.4. Modify Example 34.1 by including a carnivorous zooplankter that eats the herbivorous
_ P
kg - kg'''-k zooplankton. Also, modify the model by incorporating a saturation lerm for herbivore
.'P + [1 grazing on algae. Along with the parameters from the example. you will require the
following additional values:
where kg = a conslanl growlh rale al saluraled nlllrienllevels. The results of substitut-
ing !his relationship into the system and integrating wilh the fourth-order RK method are Zeo = 0.02 gC m-3 Cgze = 3 m3 gC-1 d-I
displayed in Fig. E34.2. Note that we have expressed the phytoplankton and phosphorus Be = 0.6 kdzc = 0.1 d-I
ksa = 15 J.LgChl m-3

where the subscript e denotes the camivores. Also, remove the lumped respira-
Total tionlexcretionldeath term for the herbivores and model respirationlexcretion explicitly
------------------------.
(k'zh = 0.01 d-I) along with the carnivore grazing effect.
\
. .... \ .. ./r'\..
~
u 0.5
. ... 34.5. Modify Example 34.2 so that the system is a CSTR with a residence time of 20 d and
inftow concentrations of a¡a = 0.5 mgChl L -1 and [lin = 10 mgP L -l.
C>
E 34.6. Isle Royale National Park is a 21 O-square-mile archipelago composed of a single large
island and many small islands in Lake Superior. Moose aITived around 1900 and by
1930 their populalion approached 3000, thus ravaging vegetation. In 1949, wolves
o crossed an ice bridge from the Canadian Province of Ontario. Since the late 1950s
O 50 100 150
the numbers of the moose and wolves have been tracked (Allen 1973, Peterson et al.
Tlme(d) FIGURE E34.2
1984).
 -~._' ., --' --

632 PARTv EUlrophicalion and Temperalure

LECTURE 35
Year Moose Wolves
1960 700 22
Year Moose Wolves
1972 836 23
I ----- --
1961 - 22 1973 802 24
1962
1963
-
-
23
20
1974
1975
815
778
30
41
I NutrientlFood-Chain Modeling
1964 - 25 1976 641 43
1965 - 28 1977 507 33
1966 881 24 1978 543 40
1967 - 22 1979 675 42
1968 1000 22 1980 577 50
1969 1150 17 1981 570 30
1970 966 18 1982 590 13
1971 674 20 1983 811 23

(a) Inlegrale Ihe LOlka-Volterra equalions from 1960 Ihrough 2020 wilh Ihe fourth-
order RK melhod. Delermine Ihe coeflicienl values Ihal yield an oplimal lit. Com-
pare your simulalion wilh Ihe dala using a lime-series approach, and commenl on
'1
Ihe results.
I
(b) Plollhe simulalion 01'(a), bUI use a slale-space approach.
(c) Afler 1993 suppose Ihallhe wildlife managers Irap one wolf per year and Iransport
il off Ihe island. Predicl how Ihe populalions 01' bOlh wolves and moose would
evolve lo Ihe year 2020. Presenl your resulls as bOlh lime series and slale-space
plols. For Ihis case [as well as for (d)], use Ihe following coefficienls: a = 0.3,
b = O.OIIII,c = 0.2106,andd = 0.0002632. Lu-ruRE OVERVU:W:1 combine the material from the preceding two Jectures into a
(d) Suppose Ihal in 1993 some poachers sneak onlo Ihe island and kill 50% 01' Ihe complete framework to compute nument/food-chain interactions for a stratified
moose. Predicl how Ihe populalions 01'bolh wolves and moose would evolve lo Ihe
Jake,
year 2020. Presenl your results as bolh lime-series and slale-space plols. ,;./

In Ihis lecture a simple mode! 01'a two-Iayer lake is used to illustrate how the
nutrient/food-chain framework developed at the end of the previous lecture can be
applied to a nalural water. It should be emphasized thal the framework describcd
below is but one of a variety of formulations proposed to model nutrient/food-chain
dynamics in natural waters. The reader also can consult Scavia's (1979) review 01'
the subject, as well as critical artic1esby Riley (1963), Steele (1965), and Mortimer
(1975b).

35.1 SPATIAL SEGMENTATION AND PHYSICS

The present framework handles physical segmentation (two vertical layers), load
ings, and transport in the same way as the two-component mode! (SSA) described in
Seco 29.5. Mass balances rOi a substance in the epilimnion (1) and the hypolimnion
(2) can be wriuen as (Fig. 3:>.1)
dCI
VId! = W(t) - Qc, + V,A,(C2 - c¡} + SI (35.1)

dC2
and (35.2)
V2d! = v/A,(c, - C2)+ S2

633
 634 l'ARrv ElIlrophicalion and Temperatllre
LEITURE35 NutrientIFood-Chain Modeling 635
~
Intlow Outtlow
TABLE 35.1
Eplllmnlon Model state-variables
(1)
State-variabte Symbol Unlts

Dlffusion Food ehain:

1. Algae a mgChla m-3
2. Herbivorous zooplanklon Zh gCm-3
Hypollmnlon 3. Carnivorolls zooplanklon Z, gCm-3
(2) Nonliving organie earbon:
I 4. Partieulate
FIGURE 35.1 cp gC m-3
5. Dissolved C,I gC m-3
Physical segmentation scheme and transport
representation. Nutrients:
I 6. Ammonillm nilrogen ". rngN m-3
1 ' where V = volume 7. Nilrate nitrogen ni rngN m-3
e = concenlration 8. Soluble reaetive phosphorus p, rngP m-3
t ~., time
I W(t) = loading AII the equations in Secs. 35.2.1 and 35.2.2 are written for a generic layer. Thus
i Q outnow
==
they can be applied to either the epilimnion or the hypolimnion. The only cases
V, == thermoelinetransfercoefficient where the equations for the layers would differ is for the variables subject to settling
A, = thermoeline area
(algae and particulate organic carbon). For these variables the following equations
S = sources and sinks
are written for a layer that is overlain by another layer. As such they apply strictly to
the hypolimnion, which receives 'Settlinggains from the epilimnion. When applied
to the surface layer, these settling gains are merely omitted from Ihe equa~on.
35.2 KINETIC SEGMENTATION

As depicted in Fig. 35.2, the model consists of eight state-variables. These can be 35.2.1 Food Chaln
, divided into three major groups (Table 35.1). Mass-balance equatioris are written for
each of the state-variables for each of the layers. The source and sink terms for each The food chain consists of a single plant group along with two zooplankton grollps.
state variable are described in this section.
Algae. As described in Lec. 33, algae grow as a function of temperature, nutri-
ents, and solar radiation. For the hypolimnion, algae are also gained by settling from
\éAJ,V ICarnlvorous
zooplankton ~ the surface layer. Sinks inelude respiration/excretion, grazing, and settling losses:
da
V dt = kg(T.n,. p"./)Va - kra(T)Va- Cllh(T,a. Zh)Va+ vaA,au- vaA,a
(35.3)
~'FHerblvorous
zooplankton where kll(T. n" P.,./) = algal growth rate (d-I)
Dlssolved
Zh kra(T) = losses due to respiration and excretion (d-I)
organlc C
CII/¡(T. a. Z¡,) = grazing losses (d -1)
Cd
Va = phytoplankton settling velocity (m d-I)
Soluble r "!"t,
reactiveP and the subscript u designates an upper layer.
p, The growth and the grazing rates are dependent on environmental factors, as in
(recall Eqs. 33.29 and 34.18)

FIGURE 35.2 kg(T ,n" p".1) -- k 11.20.

1 066 T-20
. _ n,
2.718f e-al - e-ao mln ps
_k H ( '
Kinetic segmentation. [ . )] (km+~ k~+~
)
(35.4)

--
_
- "'AA<''''_
J~ >':~~' . ~ p '- ........-.- .,.

LECrURE35 Nutrient/Food-Chain Modeling 637

636 PARTv Eutrophication and Temperature
.. .
a T-20 35.2.3 Nutrlents
and Cg/¡(T,
a,z/¡)= _ksa + Cg/¡(} Z/¡ (35.5)
a
The nutrients are divided into inorganic nitrogen and phosphorus. The former is also
where n, = total inorganic nitrogen = na + n¡ and other parameters are defined in split into ammonium and nitrate nitrogen.
Lees. 33 and 34.
In addition the respiration rate is corrected for temperature with the theta model, Ammonium nitrogen. Ammonium i I gained due to hydrolysis 01'oIissolved
o¡ganic carbon and from food-chain respiration. lt is lost via plant uptake aud nitri-
k,a(T) = k,a.20(}T-20 (35.6) ftcation:
AIl other temperature-dependencies are corrected in this fashion. d~ . .
Vd! = an,.k/¡(T)VCd+ anak,a(T)Va + anck,/¡(T)Vz/¡+ an,.krt.(T)Vz,.
Herbivorous zooplankton. Part of the consumed algae is converted into her-
- Famanakg(T, n,. p,.. 1)VII - kn(T) Vna (35.11)
bivorous zooplankton. The herbivores are depleted by carnivore grazing and respi- i
rationlexcretion losses: where Fam = fraction 01'inorganic nitrogen that is taken from the ammonium pool
by plant uptake.
dz/¡ /''\. ,A
o.¿) a, z/¡)V~- 1 Cgc(T, z.,)V z/¡ - k,/¡(T)V z/¡
Vd t =( acar/¡Cg/¡(T, (35.7) na
(35.12)
Fam = kam+ n" I ~
where aca = the stoichiometric coefficient for the conversion of algal chlorophyll a
to zooplankton carbon (gC mgChla-1). in which k"m = a half-saturation ¡;onstant for ammonium preference, antl al/(' alld
an" = the ratios 01'nitrogen to carbon and chlorophyll. respectively.
Carnivorous zooplankton. Part of the consumed herbivorous zooplankton is
converted into carnivorous zooplankton. The carnivores are depleted by respira- Nitrate nitrogen. Nitrate is gained due to nitrification and it is lost via plan!
tionlexcretion losses ánd a lirst-order death due tó grazing by organisms higher on uptake:
the food chain (primarily fish).
(35.13)
dzc V~~¡ = kn(T)Vn" - (1- F",n)anakg(T.n,. p,.,¡)Va
V di" = 8cCg(.(T.zc)Vz/¡- krc.(T)Vzc- kdc(T)Vzc (35.8)
Soluble reactive phosphorus. SRP is gained due to hydrolysis 01'dissolved
organic carbon and from fooo/chain respiration. lt is lost via plant uptake: l'
35.2.2 Nonlivlng Organlc Carbon v' 'y I
The nonliving organic carbon is divided into particulate and dissolved fractions in
vd~s = apck/¡(T)V~ + apak,a(T)Va + apck,/¡(T)Vz/¡ + ll/}(.k,c(T)Vzc I
order to distinguish between settleable and nonsettleable forms. - ll/",kg(T.ni.P.,.¡)Va (35.14)
where apc and apa = the ratios of phosphorus to carbon and chlorophyll. respec-
Particulate. Inefficient grazing (egestion) along with carnivore death result in tively.
gains to the particulate nonliving organic carbon pool. For the hypolimnion. POC is
also gained by settling from the surface layer. Sinks inelude a first-order dissolution
, 35.3 SIMULATION OF THE SEASONAL CYCLE
reaction and settling losses:
dcp By using the kinetic interactions from Eqs. 35.3 through 35.14. we can write Eqs.
ir 35.1 and 35.2 for each of the eighl state-variables. The resulting 16 ordilldiYdiffer-
Vd! =tlI;'a(1- 8/¡)Cg/¡(T,
-it;<T)VCp + vpA,c/JU - vpA,cp -
a, z/¡)V (1 - 8c)Cgc(T,z,.)Vz/¡+ kdc(T)Vzc
(35.9) ential equations can be integratetl simultaneously using a numerical melhod such as
the fourth-order Runge-Kutta melhod.
Dissolved. DOC is gained via the first-order dissolution reaction and lost by
EXAMPLE 35.1. NUTRIENTIFOOD.CHAIN MODEL FOR LAKE ONTARIO.
hydrolysis:
Physical parameters 1'01'Lake Ontario during the early 1970s are summarized in Table

VdCd
tlt
= kp(T)Vcp - k/¡frr)VCd
'!J
(35.10) 35.2 and Fig. 35.3. Note that the hypolimnetic temperatures are simulated with a heat-
balance model (recall Lec. 31). Loads and ¡nitial conditions are summarized in Table
35.3. and model parameters are listed in Table 35.4.
 TABLE 35.2
Hydrogeometric parameters for Lake Ontario
LECTURE35 NutrientIFood-Chain Modeling 639
Parameter Symbol Value UnUs
Surfacearea A. 19.000 x lO" m2 TABLE 35.4
Thermoclinearea A, 10,000 x lO" m2 Model parameter values used for Lake Ontarlo slmulatlon in the
Epilimnion volume V, 254 x lO' m] early 1970s
Hypolimnion volume V. 1.380 x lO' m'
Thermocline thickness H, 7 m Parameter Symbol Value Unlts
Epilimnion thickness H, 17 m
II Hypolimnion thickness
Outtlow
H,. 69 m Algae:.
Growth rate k,.20 2 d-I
Q 212 x lO' m' yr..' O. 1.066
II Thermocline diffusion
Summer-stratilicd
v,
0.13
cm2 S-I
Temperature factor
Respiration rate
Temperature factor
km
Om
0.025
1.08
d-I
Wintcr-mixed 13
Start 01'summer stratification Settling velocity v. 0.2 md-'
¡ 100 d
Optimallight 1, 350 Iy d-'
¡ Time to establish stratilication 58 d 2
P half-saturation
I Onset 01'end 01'stratification
End 01'stratification
315
20
d
d
N half-saturation
k..p
k.fl! 15
¡.tgPV'
¡.tgN L-'
Background light extinction k; 0.2 m-'
)
I Herbivorous zooplankton:
Grazing rate C,. 5 L mgC-' d-'
I
I 10(Iy d-I) f Temperature factor O,. 1.08
I 20 Grazing efficiency €. 0.7
Respiration rate k,. 0.1 d-'
, Temperature factor O,. 1.08
600 Algae half-saturation k.fQ 10 ¡.tgChla L-'

O' Carnivorous zooplankton:

~ 10 400 Grazing rate C,. 5 L mgC-' d-'
0.5
¡.., Temperature factor O,. 1.08
Grazing efficiency €, 0.7
200 Respiration rate k" 0.04 d-'
Tel11peraturefactor .Orc 1.08
Oeath rate k., 0.04 d'"
o o o
Temperature factor O., 1.08
o 60 120 180 240 300 360 Herbivore half-saturation k,. 004 mgCV'

Time (d) Nonliving carbon:

Particulate settling vp 0.2 md-I
FIGURE 35.3 Oissolution rate kp 0.1 d-I
Temperature and light data lor Lake Ontario. Temperature factor Op 1.08
Hydrolysis rate k. 0.075 d-I
Temperature factor O. 1.08
TABLE 35.3
Nutrients:
Boundary (Ioading) and initial conditíons for Lake Ontario in the Nitrification rate k. 0.1 d--'
early 1970s
- - --.------- Temperature factor O. 1.08
Variable Units Ammonia preference half-saturation k.m 50 ¡.tgNL-1
Loadlng' Inltial condltlons'
Algae ¡.tgChlaL-I I I
Herbivorous zooplankton mgCL-' O 0.005 Solution: The model equalions were solved numerically for 2 yr. Some of Ihe resulls
Carnivorous zooplankton mgCL-' O 0.005 for Ihe second year are summarized in Fig. 35.4. Figure 35.40 shows results for Ihe
Particulate organic carbon mgCL-' 0.8 0.12 food chain in the epilimnion. Nole Ihal the zooplanklon are expressed in chlorophyll
Oissolved organic carbon mgCL 1 0.8 0.12 units to allow comparison among Ihe variables. The resulls indicate Ihat predalor-prey
Ammonium 15
¡.tgNL-' 15
Nitrate 220 interactions are taking place, wilh Ihe peaks of Ihe algae, herbivores, and camivores
¡.LgN1. ' 250
Solnble rcactivc phosphnrlls ¡.tgl'L' 14.3 occurring al approximately days 160, 180, and 240, respeclively.
12
Results for epilimnelic inorganic nilrogen and phosphorus are iIIustrated in Fig.
t Mulliply by oulflnw lo cnllverl In mass.lomJing ralc. 35.4b. According lo the calculalion the lake is overwhelmingly phosphorus-limited
*Samc fnr cpilimnion .mil hypolimnion.
 ---' ~ ~ -~-- -~. -~

640 PARTv Eutrophieation and Temperature LEcrulUi .t\ NUlrientlFood-Chain Modeling 641

The foregoingexampleilluslralesIhe Ihreeprimaryslrenglhsof nutrientlfood-

16 T (a) A I chainmodels:
C\I . Temporalresolutionof eutrophicationeffects.The modelsprovidepredictionsof
:c
Jh
fealures like peak chlorophylllevels and, hence, produce informalion that is ex-
O
al
:1. ] I tremely usefulto water-quality managers. This is because the public is often most
concemed with extreme events in a water body rather than in se¡tsonal or long-
. lerm average conditions. Because of its daily time step, the nutrient/food-chain
o ¡. r 1 I -<f' I -< 11-'-1 I ..h.U1 models are capable of generaling both types of information.
o 60 120 180 240 300 360 . Nutriefll and light limitation. Identifying the limiting nutrient or whether the sys-
(b) tem has significantlight limitalionis a criticalstep in controllingeutrophication.
300+ rI I By mechanisticallyincludingseveralnulrienlsand lighl,Ihenutricntlfood-chain
modelsprovidea meansto makesuchidentifications.
. Org(/IIic carbo" cycle. By providing predictions of organic carbon levels, nutri- I
entlfood-chain frameworks provide a means to as se ss both oxygen und toxic sub-

...t
100 1 slances in lakes.
r
II
o
o 60 120 180
.
240
. "7.
300 360
I 35.4 FUTURE DIRECTIONS
The framework described in this lecture can be extended and refined in a number
1.ooT
I
(e) of ways. First, it can be eXlended to other water bodies such as elongated lakes,
0.75 streams, and esluaries. Second, additional nutrients and food-chain compartmenls
can be included. For example the algae might be divided into several functional
0.50 groups (e.g., dialoms, blue-greens, greens, elc.), or new Iypes of organisms such as
bacteria or rooted plants might be inc1uded.
0.25
Over the coming decade one fundamenlal way in which Ihe nutrient/food-chain
Nonllvlng poc framework could be modified is lo shift the focus from chlorophyll to organic carbono
0.00

60 120 180 240 300 360

Because of concern over eutrophication, early nutrientlfood-chain models focused
o
on plant growth (Fig. 35.5a). For such applications, chlorophyll a served as a con-
Time (d) venienl measure of the effects (lf enhanced productivity.
FIGURE 35.4
Simulation results lor the epilimnion 01 lake Ontario as
computed with a nutrient/lood-chain model. (a) Food chain
with all components expressed as chlorophyll, (b) inorganic
nutrients, and (e) organic carbono

during the summer stratified periodoThe faet that there is exeess nitrogen and phosphorus
in the water refleets the low growth during the remainder of Ihe year.
The organic carbon eycle for the epilimnion is shown in Fig. 35.4c. This plot indi-
cates a clear difference between the productive summer months and the rest of the year.
High organic carbon generation during the stratified period has several implieations.
First, its eventual decomposition can have an impact on the oxygen content of bottom
(a) (b)
waters. This would be particularly important fur water bodies with smaller hypolimnion
(and hence smaller oxygen reserves) than for Lake Ontario. Second, as we will see in FIGURE 35.5
later lectures, the transport and fate of toxic substances can be strongly associated with (a) Traditional nutrient/lood-chain kinetics have used chlorophYII as their locus. (b) In
organic matler. the luture, such productionldecomposition models may shift toward a more detalled
characterization 01 organic carbono
 ~.'~ . j¿.>. .. Ia . - - - ._- .- --
'/.
I

¡
642 PARTv Eutrophication and Temperature LECTURE35 Nutrient/Food-Chain Modeling Ij4j

TABLE 35.5 Data from the lake has been collected and is summanzed Irt Table P35.l The Jxjiíd's liíl
The ditTering reactivity of autochthonous particles In natural ftow has available and unavailable phosphorus concentrationsof .
20 Jigi> L-~ Assti¡¡je
waters can be modeled by dividing them into "G classes" of that the lake is never limited by nitrogen. Use the same solar radiation ánd pbotdperlM
ditTering reactivity (Berner 1980, Westrich and Berner 1984, as forLakeOntario(Fig.35.3). ,. . .

Di Toro and Fitzpatrick 1993) (a) Calibrate the nutrientlfood-chain model developed in this lecture to tbIs dai¡{ ;el. "
(b) Determine the innow concentration of available P needed so that the peak epilim-
Particle type Fraction of fresh partlcles Decayrate (d-I) !iI'
netic chlorophylllevel does not exceed 10 p,g L -l.
A, (rapidlyreacting) 0.65 0.035 (e) Determine the washout rateo
A, (slowlyreacling) 0.20 0.0018
G, (refraclory) 0.15 o TABLE P35.3
t SRP. P.",.t
T. T. a. z". z.. SRP. P...
Today, water-quality managers have broadened their concerns to encompass Day eC) eC) (I'gL-I) (mgCL-I) (mgCL-I) (I'gPL-') (I'gPL-I) (I'gPL-1) (I'gPL-I)
problems other than eutrophication, such as toxic substances and disinfection 0.01 13.5 t.I 15 1
15 3 4.3 4.9 0.01
by-products. In such cases both dissolved and particulate organic carbon must 45 1 2.3 6.5
be characterized. Further, as already discussed in Lec. 33, because of variable 75 2 1.8 9.7
chlorophyll-to-carbon ratios, chlorophyll a may not even represent an adequate 0.05 4.0 1.8 9.0 1.6
105 2 2 12
2.7
135 4 2.8 8.7 0.2 0.005 2.8 3 7.5
measure ef particulate plant carbono Consequently, as depicted in Fig. 35.5b, 01'- 4 8.5 3.4
165 7 5 7 0.2 0.04 1.6
ganic carbon may become the focus of newer modeling frameworks. 0.16 0.4 4.3 12.7 2.7
195 15 5.5 6.8 0.07
Further, distinctions will have to be made between easily decomposable and 7 0.07 0.12 0.4 3.6 15.7 1.8
225 20 5.7 1.3
refractory forms of carbonoOne case where this could be important would be short- 6.6 0.07 0.105 0.49 3.3 17.6
255 20 5.9
residence-time impoundments that are dominated by allochthonous sources. Another 285 12 6 6.3
relales to the simulation of sediment-water interactions. In such cases the distinc- 315 8 6.5 6.5
345 5 6.3 4.8
lion among varying decomposing fractions is important (Berner 1980).
Di Toro and Fitzpatrick (1993) have adopted a scheme devised for sediment t Unavailablephosphorus.
diagenesis by Berner (1980) and Westrich and Berner (1984) to handle different
.~ reactivity cIasses of particulate matter. As listed in Table 35.5, fresh particulate 01'-
ganic carbon is divided into three "G fractions" corresponding to easily degradable,
slowly degradable, and nondegradable fractions. By tracking each fraction sepa-
rfllely (thal is, with separate mass balances), the differential degradation can be de-
lermined.

PROBLEMS
35.1. Increase the loading of phosphorus to Lake Ontario until the spring bloom becomes
limi¡ed by nitrogen or light. Express your result as an innow concentration and identify
which factor dictates the maximum algallevel.

35.2. As described in Lec. 18, perform a sensitivity analysis on the simulation performed in
Example 35.1.ln particular determine which kinetic coefficients have the major impatÍ
on the timing and peak of the spring phytoplankton bloom.

35.3. Oreen Lake has the following characteristics:

Epilimnion volume = 150 x 106 1113 Epilimnion depth = 10 111

Hypolimnion volume = 600 X 106 m3
Thermocline area = 10 x 106 012
Therlllocline thickness = 3 m
Time to establish stral. = 30 d
Time lo end stra!. = 30 d
Surface area = 20 X 106m2
Inflow = outflow = 150 x 106m3 yr-I
Start stratification = 150 d
Begin. of destral. = 300 d
Extinction due to water/color = 0.201-1
 ----------- ._--

LECfURE36 Eutrophication in Flowing Waters 645

LECTURE 36 3f '3imple PhytoplanktonlNlltrient Analysis

1 ¡I.II and Muelk ,\'1' ü.:veloped a simple analysis that is extremely
uselul for understand¡. "_ ,.pact of point sources on stream eutcophication.
E~tropl1icatiol1tn Flowing Waters First, the analysis addresses the question of whether nutrients will be limiting
in the immediate vicinity of a sewage outfall into a civer. Recall from Table 28.3
that, with the exception of cases where phosphorus removal is implemented, point
~6urces of sewage are generally nitcogen-limited. Further, nonpoint sources are lISU-
/ ally phosphorus-limited. For cases where the receiving civer above the point sOllrce
is dominated by nonpoint sources, Thomann and Mueller used a simple mass bal-
ance at the outfall (recall Eq. 9.42) to draw the following general conclusions:
. If efftuent ftow is les s than 2% of the total river ftow after mixing, phosphorus
J
.
'
would tend to be the limiting nutcient downstream. I
For efftuent-to-ftow ratios greater than 2%, the limiting nutrient would depend on
the treatment process. In particular, if tiule or no phosphorus removal were used,

. the stream would tend to be nitrogen-limitcd.

In all cases, if significant phosphorus removal is used, the stream would tend to
be phosphorus-limited.
f

There are two major reasons why special modeling frameworks are necessary for
ftowing waters such as streams and estuacies. First, and most obviously,their physics
are fundamentally different fcom standing waters such as lakes. In particular, hor-
Next they investigated whether nutrient limitation would occur at the outfall.
To do this they recognized that the Michaelis-Menten fonnulation for nutcient limi-
tation becomes significant only when nutcient concentrations fall below a value tha.
is 5 times the saturation constant (recall Fig. 32.2). Thus, assuming ksn = 10 to
20 p.gN L-) and ksp = I to 5 p.gP L-1, they estimated that nitcogen and phospho-
rus limitation would occur only when concentrations fell below about 0.1 mgN L -1
and 0.025 mgP L-1, respectively. Since sewage typically has nutcient conccntrations
in the mg L-1 range (untreated sewage has about 40 mgN L-1 and 10 mgP L-(), f
they concluded that point-source-dominated streams (that is, those with efftuent-to-
ftow ratios greater than about 1%) without significant nutcient removal would have
excess nutcients at the mixing point.
Even though nutrients are in excess at the mixing point, subsequcnI plan!
gcowth downstream will tend to lower nutcient concentrations downslr '11 Thus ,

izontal transport must be considered. Second, shallower ftowing waters can be the stream could eventually become nutrient-limited downstream. Two qucstioJls
dominated by fixed plants ach as macrophytes and periphyton rather than the follow from this reasoning: At what distance downstream will the stream become
frec-floating phytoplankton l1¡a1we have emphasized to this point. nutrient-limited? Which nutrient will become limiting first?
fhis lecture deals with both topics. First, I show how the phytoplankton models To address ¡hese questions, Thomann and Mueller developed the following
d,'. I ¡bedin earlierlecturescan be appliedto deeperstreamsandestuacies.Then I steady-state mass balances for a downstream stretch with constant hydrogeometric
desccibe some appcoaches for modeling fixed plants in shallower systems such as pcoperties:
streams.
da da ~
(36.1)
U dx = dt* = [ kg(T. 1) - kd - H ] a = k"ela

36.' STREAMPHYTOPLANKTONlNUTRIENT INTERACTIONS
In this section , show how the lake phytoplankton models developed in earlier sec-
pon~ can be apptied to ftowing waters. I first desccibe a simple back-of-the-envelope
patcul~tion mat is helpful in plac,ng the subsequent numerical solutions in perspec-
liY~.
where t* = travel time (d) and k"el = net gain rate of phytoplankton (d IJ, which
as shown, depends on growth, kinetic losses (respiration, excretion, death), and set-
tling. Notice that the growth rate does not depend on nutcients as per our previous 1;
discussions. Thus the folJowing analysis will apply only when n 2: 0.1 mgN L -(
and p 2: 0.025 mgP L-l.

644
 646 PARTv Eutrophication and Temperature LECTURE
36 Eutrophication in FIowing Wa~rs 647

If we assume a constant plant stoichiometry. balances for nitrogen and phospho- limitation would set in after about 8.6 d. For a stream with aJt average velocity of
rus folIow. 0.15 to 0.3 mps (0.5 to I fps). a week's travel time corresponds to about 90 to 180
dn km (60 to 115 mi).
dt* -a""kg(T. I)a (36.2) Note that Thomann and MuelIer have shown that Eqs. 36.5 and 36.6 can be used
to determine the critical trave! times exactly.
dp
dt* = -a""kg(T. I)a (36.3)
h
wereno=-I aa"kgnO (36.7)
Ohserve thal bolh kinetic losses and setlling are assumed to be terminallosses; that . = knel In (n
t"
noI - 25)
+ no knet
¡s. they do nol fced hack to Ihe nutrienl pools.
Using houndary condilions of II = lIO, 1/ = l/o.andp = Poallhe mixingpoint. (36.8)
where p = a"pkgpo
wc can sol ve these equations . _ knel
tp -
I In (p Po- 1(0)
+ Po
I knet
a = lloeklll',l. (36.4)
where t;' and t~ = travel times (d) for nitrogen and phosphorus limitation. respec-
1I,/(/k~lIO k l. tive!y.
1/ = l/o + ---'-- (1
- e "" ) (36.5) Aside from telling us where nutrient limitation begins, critical travel times also
k"rI
indicate whether nutrient load reduction wiII even have an impact on plant growth.
l/""kgl/o .. Suppose that the stream length of interest is less than both t;' and t;. The analysis
p = Po+ k".,
:-- (I - e'''''/.) (36.6) indicates that initial nutrient load reductions would not have an impact on eutroph-
ication. Of course. at a certain point, load reductions could bring the critical travel
The Ihree equations are displayed in Fig. 36.1 for a stream with a signifi- times within the stream length of interes!. The equations could provide an estimate
canl efftuent-Io-now ratio. The simulation also renects a pnsitive net phyloplank- of the leve! of treatment required for this point to be reached.
ton growlh. a signilicanlly high algal initial condition. and an efftuent with a low
N:P ratio. As expected from Eqs. 36.4 through 36.6. the nUlriérJtsdrop as the phyto-
plankton grow. 36.1,2 Stream and Estuary Phytoplankton Models
Two other features of Ihe plol bear mention. First. nitrogen becomes limiting
before phosphorus when the nilrogen level falIs below 100p,gN L -1 at a travel time Phytoplankton kinetics (Fig. 36.2) can be integrated into a general model for a one-
J dimensional stream or estuary, as in
of about 7 d downslream frolllthe source. Second, if there were more nitrogen in the
cmuent so thal N-limilalion did not occur. the plot also indicates that phosphorus aa a2a aa Va
-at = E~ax - U-ax + (k¡¡- k",)a- H a (36.9)

119 L-1
2000
L.. Simple model no
200 Jonger applles
a

nllmlllng
a 10: I a, p 1000O /
p IImltlng
5

-1000 Travel time (d)

Time when n
becomes IImltlng
FIGURE 36.2
FIGURE 36.1 A simple oxygen balance for a stream
Plo: 01 nutrient and phytoplankton concentrations downstream Irom a point source. Including phytoplankton grawth.
'fIt~~,_" .,:~~.~':!~:-.~. --.~ ~~ ..
~

648 PARTv Eutrophication and Temperature LECTURE36 EUlrophication in Flowing Waters 64~

aL
at
= Ea2L
ax2
_ uaL
ax
_ kdL- ~L
H (36.10) ~g L-1
2000
ap a2p ap
-at = E~ax - u-ax - apa(kg- kra)a (36.11)
200
ana a2na ana
-at = E-ax2 - u -ax - ana (k g - k ra)a-n k na (36.12)
a 100 n, p 1000
an¡
at = Ea2n;_
ax2 uan;
ax + knna (36.13)
..................
_
ao a20 ao o o ~nllnliting
pllmlllng ,
a t = E~ax - u-ax - kdL - rOl,k"na+ roa(kg - k,.a)a+ k,,(os - o) (36.14) o 5 10
We have developed this representation so that it is direclly comparable to the Travel time (d) Time when n
back-of-the-envelope mude! described in the previous section. Thus it is a some- becomes limitlng
what simplified version of a full-blown stream eutcophication model. For example FIGURE E36.1
it assumes that the plants can utilize only ammonium as a nitcogensuurce. Further it
ignores the organic nitrogen and phosphorus pools that are included in more complete where i = a control volume. Along with appropriate boundary conditions. these thrcc
representations. Of course these mechanisms, as well as others such as sediment- equations can Ihen be integraled numerically using the solution techniques presenled in
water transfers, can be readily added. The resulting set of equations can be solved Lees. 11 through 13. Figure E36.1 shows the resulting steady-stale proliles employing
numerically, as described in the following example. parameter values comparable to Ihose used to construct Fig. 36. J.
Notice how this model behaves in an identical fashion to Fig. 36.1. up lo Ihe poi ni
EXAMPLE 36.1. NUMERICAL STREAM PHYTOPLANKTON SIMULATION. that nitrogen becomes limiting at about t* = 7.3 d. Whereas the simple mode! breaks
Duplicate the calculations in Fig. 36.1. but use nutrient limitation explicitly. down beyond that point. the numerical model is valid because the nutrient limilation

Solution: The model can be made directly comparable to Thomann and Mueller's ap-
lerm shuts down uptake as nitrogen leveis drop. Consequently both Ihe nitrogen ami h
phosphorus leve! off rather than going negative. Further the phytoplankton begin to de-
proach by omitting BOD and oxygen and combining ammonium and nitrate in a sin-
cline since growth falls below the respiration and settling losses.
gle inorganic nitrogen pool. The resulting mass balances can be written for a plug-ftow
stream with no light limitation as
da da . Va The foregoing is intended to show how nutrientlfoud-chain kinetics can be inte-
4>p.4>"- kraJ a - Ha
dt = -u dx + {kg(T. I)mm [1 grated into a one-dimensional, advective-dispersive transport framework. Now we
dn dn . will describe how the QUAL2E framework performs a similar calculation.
dt = -u (Ix - a"a(kg(T. 1)10m[4>/"4>,,] - kraJa
dp dp . 36.2 MODELlNG EUTROPHICATION WITH QUAL2E
di = -u dx - apaI kg(T, J)mm[4>p.4>,,1-kraJa
This section shows how EPA's QUAL2E model can be used to simulate both tem-
These equations can also be expressed in a numerical control-volume approach with
backward spatial differences as
perature and nutrientlalgae dynamics in flowing waters.

da¡ . Va

V¡lit = Q¡_I,¡a¡_1- Q¡ ,'1;+ {

[ ]
kg(T. I)mm 4>p.4>" - kra - H} ; V;a¡ (36.15) 36.2.1 Temperatureand QUAL2E

(36.t6)
Recallthatthe generalmasstransportequationforQUAL2E(Eq.26.28)is
V; ~¡ = Q¡_I,¡n;_1 - Q¡.¡+ln¡ - a"al kg(T. J)min[4>p. 4>,,]- kraJ; Vía;

dp "
s YI 4: = Q¡-I.¡P¡-I- Qi,i+IP¡- apllIkg(T, 1) min[4>P' 4>,,] - kraJ; Vial
(36.17)
ae = a(AxEac)
ax dx _ a(AxUe)dx + de + .:. (36.18)
at Axax Axax dt V
'~11 650 PARTv Eutrophication and Temperature LEC11JRI!
36 Eutróphication in Aowing Waters 6S t

In a similar fashion a heat balance can be written as

~. . 1 ..a j... 00:-"",.' .0 oj;,:üt:¡

¡~
I aT =
al
~
a AxE-
ax dx _ a(AtUT)dx + ~
AxdX A.tax pCV
(36.19) ~
2Je567t90~2]«56189ÓUJ
TITLEOl

IJ'ITLE03
'tITLEOC
45t7t9.01mU1~t~J2Je..~~'01aJ.45~i8t!!.i~i8¡Q

NO
NO
MANOS-ON 3a,
Steve Chapra,
CONSERVATlVE
CONSUVATIVE
QUAL-2EU
May 18.
MINERAL
MINERAL
WQRKSHOP: TEMPERAruRE
199.

U
I

TITLtOS NO CONSERVATIVE MINERAL Itl

Notice that we have omitted the source term-dT/dl. This means that internal heat TITLE06 VES TEMPERATURE
TITLB01 VES
generation or loss (for example viscous dissipation of energy and boundary friction) TITLE08 ~
810CHEMICAL OXYGEN DÉMAND
ALOA&: AS CHL-A IN UG/L
is negligible. TITLE09
TITLE! o
NO PHOSPHORUS CYCLE AS P IN NO/L
IORGANIC-P; DISSOLVBD-P)
In addition note that, because it is usually negligible, the transfer of heat between 1'1T1&11 NO
1'tTI,112
NITROOEN CYCLE AS N IN Ha/L
(ORGANIC-N; MMONIA-NI NITRITE-NI' NITRATE-M)
the bottom sediments and the stream is ignored. Therefore the external sources and 'U1'LE!3 YlS
TITLE1. NO
DlSSOLVED OXYGEN IN MG/L
FECAL COLÍF'ORM IN NO./I00 ML
sinks of heat are purely dependent on transfer across the air-water interface, TITLB15
tNDTITL!
NO ARBITRARY NON-CONSERVAT]V&

NO LlST DATA INPUT

~¡.I
s = H.w + Hall - (Hbr + He + He) No WRtTE OPTIONALSUHMÁRY ~
NO FLOW AUGMENTATIbN
Net absorbed (36.20) STEADYSTATE
Water-dependent TRA,PE'ZOIDAL CHANNELS I
radiation terms PRINT LeD/SOLAR DATA
NO PLOT 00 AND90D
1 where H.", = net solar shortwave radiation FIXED rMSTMCONC'(YES.I). o. 5D"ULT BOD CONV J(. COEE" 0.25 ¡¡¡:¡¡¡¡
;" INPUT
MtTRlC . 1. 1.
j Hall = net atmospheric longwave radiation NuMaER
OF'REACHES . 6.
OlFrPUT
NÚMBER
METRIC
OF JUNcT10NS o.
'jJ
Hbr = longwavebackradiationfromthe water
NuHOP'HBADWATERS.
TIMESTtP (HClURS) . 1.
O.
NUMBER OE" POINT LOA.OS
LNTH. COMP. ,tLEHENT (ÍM)
2.
2.
He = conduction
:J¡,¡ MAxIHUM ROUTE TtME (HRS).. 30. TIME me. FeR RI>T2 (HM)
LATITU[)E OY BASIN IDEG) z 40. WNGlTUDE OF BASIN IDEGI= 105.
1,
He = evaporation STANDARD HERtDIAN
EVAP. COEf.. (AEI
IDEG) '"
z 0.0000062
90. DAY OF YEAR START TIME
EVAP. COEF., (BE)
180.
.0000055
ELEV. Of BASIN (HETERSI ,. 1670. DUST A'M'ENUATION COEr. 0.01
The individual terms in Eq. 36.20 are calculated in a fashion similar to the ENDAT41
INnATAlA .
scheme outlined in Lec. 30. However, it should be noted that solar radiation is cal- tNDATAlB
S1'REAM REACH 1. RCH_ MS..HEAD FROM 102.0 10 100.0 1
culated internally based on parameters such as latitude and time of year. This is a STREAM RF.ACH 2. RCII.. MSI00wMS080 FROM 100.0 10 80.0
STREAM tEACH 3. RCU.. MS080-MS060 FROM 80.0 1'0 60:0
really nice feature of the model since it obviates the need for the user to obtain such STREAN REACH .. RCH_ HS060-MSO.0 FRCM 60.0 TO eo.o ~
STREAH REACH S. RCH_ MS040-MS020 FRCM «0.0 1'0 20.0
information independently. S1'REÁM REACH 6. RCH. Hso20-HSOOO FRCM 20.0 TO 0.0
ENDATA2
Equation 36.20 can be substituted into Eq. 36.19 to yie1dthe final heat balance. ENCATA)
FLAG FtELD RCH", 1. 1. 1.
Within QUAL2E several additional modifications are made to (1) express the con- FLAG FIELO RCH_ 2. 10. 6.2.2.2.2.2.2 ;2.2.2. ,1
FLAG FlELD RCH.. ). 10. 2.2.2.2.2.2.2.2.2.2.
duction exchange as a linear function of water temperature and to (2) linearize the FLAG rULD RCH,. 4. 10. 6.2.2.2.2.2.2.2.2.2.
FLAG FlELD RCH.. S. 10. 2.2.2.2.2.2.1.2.2.2.
nonlinearheatexchangetermssuchas back radiationandevaporation.Whenthese FLAG FlELO RCH", 6. 10. 2.2.2.2.2.2.2.2.2.5.
ENDATA.
llIodificationsare made, the heat budget can be computed using QUAL2E's efficient HrDRAULICSRCH_ l. 0.00 2.0 2.0 lÓ. .0002 .035
solution algorithm. HYDRAULtCSRCH. 2. 0.00 2.0 2.0 lO. .0002 .OJ5
HYDRAULICSRCH_ 3. 0.00 2.0. 2.0 lO. .0002 .OJ5
An input fiJe has been developed to silllulate telllperature for the simple kYORAUL1CSRCH. ... 0.00 2.0 2.0 lO. .00018 .035
HYORAULICSRCH. S. 0.00 2.0 2.0 lO. .ÓOOU .OJ5
CBOO/oxygen/SOO system described in detail previously in Lec. 26. This file HYDRAULICSRCH.. 6. 0.00 2.0 2.0 10. .00018 .035
tNDATA5
is displayed in Fig. 36.3. The cards that IlIUStbe modified to model temperature are TEMP/LCD RCflr l. 1670. 0.01 0.25 25. 20. 825. ,.
ENDATASA
highlighted in Fig. 36.3 and are described next. REACT COKF RCH.. 1 . 0.00 0.000 0.000 lo 0.000 O.:0000 .0000
REAC'I'COEr RCH.. 2. 0.50 0.250 s.ooo J. 0.000 0.0000 .0000
RSÁCTCOEF RCH.. 3. 0.50 0.000 0.000 3. 0.000 Ó.OOOO .0000
REACt COEr RCJh. ... 0.50 0.000 0.000 J. 0.000 0.0000 .0000
Title data. The title inforlllation should be changed to reflect the fact that th~ REACTcaEr RCR. 5. 0.50 0.000 0.000 j. 0.000 O. Ó008 ,0000
REACTCOEF RCH. 6. 0.50 0.000 0.000 3. 0.000 o.oooó .0000
mn deals with temperature. Also, card 6 is changed to a YES so that the temperature tNDATA6
will be simulated. ENDATA6A
ENDATA6B

FIGURE 36.3
Program control (data type 1). We have chosen to llIodify the 6th card in this
QUAL2E input tile tor a temperature run.
data type so that the LCO (local eIimatological data) and the solar data is printed.
In addition a variety of new information must be added on the 14th through 18th
cards of this type. These ¡nelude the location (latitude, longitude, standard meridian,
elevation), time (the day of the year), and some parameters related to the heat budget
(evaporation and dust attenuation coefficients).
 f ..
. .~'
...

~
652 PARTv Eutrophication and Temperature LECfURE36 Eutrophication in Flowing Waters 653

I CoIunms """'11P"
R 1 ~ 3 . S . 7 8 Point Tributary
,+234~67890}~:t.567e~0123.5n890123456789012)4567e90123456789012345678901234561890
10
IN1'l'tAL CONO';"1 RCH_ 1- 22.00 8.1 0.0 0.00 0.00 0.00 0.000 0.0
t¡.\T1.!-'1 CH_ 2. 20.59 0.1 0.0 0.00 0.00 0.00 0.000 0.0
llltnA¡. .1 Re"_ 3. 20.59 8.1 0.0 0.00 0.00 0.00 0.000 0.0
TA\o 1 "CH_ .. 19.72 8.1 0.0 0.00 0.00 0.00 0.000 0.0
1 RH_ 8.1 0.0 0.00 0.00 0.00 0.000 0.0 6
n.n
i 1tCti".: i 19:72 8.1 Ó.O 0.00 0.00 0.00 o.oao 0.0
¡~f.'f.'"
l~n1~ ~~~.~~~~~~~....._....................
yP~'!!'r1"J~H 8CH- 0.000 00.00 0.0 0.0 0.0 0.0 0.0 0.0 o. 6
0.000 00.00 0.0 0.0 0.0 0.0 0.0 0.0 O.
.II!'R !NI'LO!I'I
"9~ I!'I'LO!I l . 8<;.-
RCI-t:ll: 0.000 00.00 0.0 0.0 0.0 0.0 0,0 0.0 O. O>
RCih. 0.000 00.00 0,0 0.0 0.0 0.0 0.0 0.0 O. g
~. .
NcRINFLa..;l
.. .
..8 . :INFLOtf..l
RCH>;
RCH..
0.000
0.000
00.00
00.00
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0,0
0.0
0.0
0.0
O.
O.
O
e 4
lfl'i'R '.~'LO!I-l
~T~O
PtDATA8A
;"DATA~ ~
HEADWTR~l !:tDW. UPSTRBAH 5.18'10 20.0 7.50 2.0 00.0 0.0 0.0 2
iNQAfAl0
~roUtTLD.1
.
~1'~~~.
'
FTL- 1. HSO 0.00 0.463 28.0 i;oo 200.0 0.0 0.0 0.0
I1QlIf1'LD-l m- 2; RSOO 0.00 1.157 15.0 9.00 5.0 0.0 0.0 0.0
O r
1IND~TP.1I
iND~TAÍ1. -20 O 20 40 60 60 100 I
~T~í2
~T'n Distance (km) I
itro~TA:ÚA
FIGURE 36.5
I
FIGURE 36.3 (cantinued) QUAL2E autput lar axygen with variable temperature.

A plot of the QUAL2E output for oxygen is shown in Fig. 36.5. Notice how
Temperature and local climatology data (data type 5a). We use a single card
the temperature rise causes the oxygen saluration to decrease in the downstream
here JOspecify the key m. nl'ological dala needed to simulale a heat budget. An direction.
alternative is to specify dilkr':l\l data for each reach.
Finally, as was mentioned in Lec. 26, a user-friendly interface for entering the
A plot of the QUAL2E output for temperature is shown in Fig. 36.4. The tem- ~"
input file and viewing the results is now available (Lahlou et al. 1995). The inter-
perature rise is due to the meteorology we have used for the example. AIso, observe
face requires precisely the same information as in Fig. 36.3. Thus whether you em-
the drop in temperalure induced by the tribuiary inflow.
ploy the original version or the new interface, Figs. 36.3 and 36.7 can serve as a
guide for performing the QUAL2E simulations outlined in this lecture.
30

36.2.2 Nutrients and Algae in QUAL2E

26
- We will now expand our analysis lo encompass nutrients and plants. As d,'pi,:ti"din
Fig. 36.6, QUAL2E simulates the kinetics of Ihe nutrients: nitrogen and ph'''i'h;)l1Is.
22 - In addition, it calculates how these nutrients impact plant biomass.
2: The addition of these conslituents has two effects on oxygen. Firsl. Ihe con-
f.o
16 version of ammonia to nitrate in the nitrification process uses oxygen. SCl:ond.the
nitrogen and phosphorus can induce plant growth. The resulting photosynll"is and
respiration of the plants can add and deplete oxygen from the stream. AdJilional
14
information on both these processes is provided next.
The QUAL2E kinetics for the nutrientlplant components can be wriuen as
10
-20 20 60 100
Algae (A):
Dlstance (km)
dA
FIGUA, ¡.LA pA U'A
dt H (36.21)
QUAL2Eautput lar temperature.
AccumuJalion Growth Respiration Senling
 654 PARTv Eutrophication and Temperature U!CTURB.16 EutrOphlcaHon
ItU10wlllg Watéts M~

1,
Atmospherlc Nlti'ate nUrogen (1\13):
reaeratlon
dN3
dt ¡M'h - ti - Jo)áttLA t36j~t
Accumulation Nitrification Growth

I ,
K4
. Sedlment oxygen
demand (SOD) Organlc phosphorus (p.):
I D
dPI
I
I dt
a2pA - {34P. - usp. (36.26)
I I K
CBODI
Accumulation Respiration Decay Seuling
,

L
I

\,
I:' G:513,
V
E
D

O
Inorganlc phosphorus (P2):
dP2
dt
(34P\ +
U2
-H - a2IJ-A (36.27)
I X
Accumlllation Decay Sediment Growth ¡:;
I
¡ I I I r-4t1- Y Note that the nitrogen, phosphorus, and kinetic constituents can be slfuU~
G:6132 G lated without computing oxygen and CDOD. However, if they are computed, the
E oxygen kinetics are modified to account for the effects of nitrification and plant
. I
growth/respiralion,
;,¡,.
lo Carbonaceous BOD (L):
G:,/-L G:2/-L
- dL
G:,p G:2P dt
- KIL - K3L (36.28)
Accumulation Decay Seuling

Dlssolved oxygen (O):

FIGURE 36.6
QUAl2E kinetics showing nutrienllplanl inleractions. dO K4
K2(0.1 - O) K.L H
dt
Organk nitrogen (N4): Accumlllation Reacralion Decomposition SOD

dN4
dt alpA {33N4 - U4N4
(36.22) + (a3IJ- - a4P)A - as{3.N, - a6{32N2 (36.29)
Accumulation Respiration Hydrolysis Seuling Growth - respiralion Nitrificalion

An input file has been developed lo integrate plants and nutrients into !be tem-
Arnrnonia nitrogen (N t): perat:Jre program introduced earlier as displayed in Fig. 36.7. The cards that trlusibe
dN\ U3 / modified lo model plants and nutrienls are highlighted in Fig. 36.7 and are described
dt {33N4 {3IN\ +
H
- F1a\IJ-A (36.23) nex!.
i\ccumulation Hydrolysis Nitrification Sediment Growth
Tltle data. The title informalion should be changed. Also, cards 8 through 12
Nitrite nitrogen (N2): should be changed to YESo

dN2 Global algal, nitrogen, phosphorus, and light pararneters (data type lA).
dt {3IN\ - (32N2
(36.24) These cards inelude all Ihe global paramelers for the nutrient/plant inleraclions. Sev-
Accumulation Nitrification Nitrification eral of them bear menlion:
 ...
___ .. -,,:,,:,:,"w:_ ~ ~ ~ 04-- .~. .....-.-.-

656 PARTv Eutrophication and Temperature

Dolo11..
~HH~§,"8'9H3456H90H34587890UH567nOI2H567890UH567890UH567890UH587890
.. t ~ 1 o:-- 5 6 7 8

tWf08';(JH3b, QUAL;':¡EU 'WORJ[SffQP; N/P/ALGAE

.t d;wt.pn. Ma)" ~~. 1994
CON.ltRy~nV. ~lmRAL 1
CONSItRVAT¡VB MIN\!RAL ti

~~~ ~IM~~.
81QCIIEl<lCALOXVOSM OSIW<P
m
ALGAE AS CHL-A IN UG/L
PHOSPHORUS CVCLE AS P IN MG/L
10RGANIC-P; DISSOLVED-P)
VES NITROGEN CVCLE AS N IN HalL
CORGANIC-NI 1.141401"11/1.-;4;NITRITE-N;' NITRATE-N)

:~~l~:~I:O~fto ItL
~.lrMltV OOII-COIIS..V~r¡VS

~-
'.""O';Ulol' BOD COHV 1( CQEr 0.2.
~~; OUTPUT HETRJC 1.
e. tfUH8&R OF JUNCTIOHS o.
1. JjUHS"'I '" ""m LOADa
2.
q. ~".COMP,..B~ (KM) 2.
¡o. TI~ ¡He.' FOR 1tPT2 (HRS)
""'f01'¡'I)DB OF""",M 10"'" 105.
:~: DAV O",YEAR STAIlT TIME 180.
I~, . 0.0000062 JVAP;.. COEP.. (BE) .0000055
e_BRS) , .\670. DQ$'1'A'1'Tf.NUA1'IONCOEF. . 0.01

o UPTAICE BY NH3 OXIDIMG

o PROD. BY ALGAEIHG O/MG A)
N CONTENT OY ALGAE IHa N/Ha Al ..
O/MG N)_
- 3.5 o UPTAKEBV 002 OXIDCMG OlMO N)
1." o UPTAKE BV ALGAE IHG OlMO
.085 P CONTENT OF ALGAE (MG P/HG A)
A)
'"
=
'"
1.20
2.0
.013
ALG MAX SPEC GROWTH RATE(l/DAV)",
N HALF SATURATION CONST. IMG/L)_
1.5 .61&6E RESPIRATION RATE (l/DAV)
.155 P HALF SATURATION CONSTo IHG/L)=
.. .~
0.1

LIN ALa SHADE eo (l/H-UGCHLA) :. .0088 NLIN SHADE 11/H-IUGCHA/L,u2l3).. .0540

LIGHT FUNCTION OPTION tLFOOPTI '" 3.0 LIGH'r SATURATION COEF I INT/MIN) = .2083
3.0 1,IWl"r AVERAGING FACTOR (AFACT) .94
DAILV AVERAGING OPTION(I.AVOPT)
NUMBER ~y DAYLIGHT HOURS IDLH)
ALQY GROWTH CALC OPTION(LGROPT)_
..
. lU!'~.L DAILV SOLAR RADTN tINT)
ALGAI. PREF FOR NH)-N IPREFNI
000.0
0.5
0.44 NITRlflCATION INHIBITION COEF 0.6
. .
ALG/TEMP SOLAR UD FACT(TFACT) ..

:{j

~.1\q".".-s~JUW) ",al 102.0 'l'O 100.0

02. I\CU' M'¡oO".1fIj080 '!\Ci\I 100.9 'l'O 80.0
!~: eCRf !lS08QcMS080 fRQII 80:0 T9 80.0
s, 8C., MS060'MSOSO hON 60.6 ro '0.0
~. BCII. MSoso."a020 fROM ,o.0 ro 20.0
\'I'OM 20.0 ro 0.0
t. ,W!~ ~o~o.\t.ooo

ACH~1. V 1"
'CH_ l. 10. ..2.2.~.2,~~c2.2. .2.
JU:H.i;~ 10. 2.~.2.2.2.2,'2.2. .2;
ItCH_.. lÓ. 6.~.2.2.2.2.2'2' .2.
!ICJI' ~. 10. 2'2.2.2'~.2'2'~' .~.
BCII. .. 10. 2.2.~.2.2.2~'2.2. .5.
ItCH_ 1. 0.00 2.0 lO .0002 .035
J«:I!~ a,r 0.00 2.0 10. .000. .015
!ICJI.. l. 0.00 2.0 10. .0002 .035
j!'¡ORt,V~1C'!ICJI",. 0.00 2.0 10. .00018 .035
10. .OOOl .035

5 Au¡.¡c~8C11.;ts; 0.00 2 °
R~" ft:
,..DRAUt.tC. 0,00 2_U 10. .00018 .035
TA5
I!'Bllf!¡,co _CH.í. lnO. 0.01 0.25 25. 20. 82S. 2.
IIII1lúM1,

FIGURE 36.7
QUAL2~ Input lile lor a planVnutrientrun.
 '~ 658 PART v Eutrophication and Temperature lECTURB36 Eutrophication in Flowing Waters 659
4<
'11
.
rt.,
. '1 · The model allows linear, nonlinear, or no self-shading according to the values of
Trlbutary
( .11

1 the shading coefficients (UN ALGSHADECO and NLlN SHADE).

10
1 " · Different functions can be employed to represent the effect of light on plant
growth. This is controlled by the coefficient (UGHTFUNcnONOPTION). In the
present example we use Steele's model (option 3). Note that the interpretation 8
of the LlGHTSATURATIONCOEF depends on the option.
· There are several different light averaging options (DAILYAVERAGINGOPTION). , ~~~!~~-.......-..-..-
6
· There are several different options for trading off the light and nutrient limitations C>

· (ALGY GROWTH CALC OPTION).

A preference
FOR NIB-N).
for ammonia or nitrate as a nitrogen source is allowed (ALGALPREF
.s
O
e
4
With N:p:algae

2
N and P coefficients (data type 6a). These cards inelude the reach-specific
data for the nutrients.
O
AIgaeJother coefficients (data type 6b). These cards inelude the reach- -20 O 20 d 60 80 100

specific data for the algae. In addition they are used to enter data for the other Dlstance (km)

constituents (conservative/nonconservative) that might be modeled. FIGURE 36.8

QUAL2E oulpul lor dissolved oxygen wilh and wilhoul algae and
Inltlal condltlons-2 (data type 7a). This card group, one card per reach, es- nulrienls.
tablishes the initial values of the system for the nutrients and algae.

Incrementallnflow-2 (data type 8a). Even though we will not simulate in- 36.3.1 Simple Attached Plant/Nutrlent Analysls
cremental inftows, these cards must be ineluded when plants and nutrients are sim- As for phyloplanklon, Thomann and Mueller (1987) also have presented a simple
ulated. As in our example, all the values can be set to zero. back-of-Ihe-envelope approach lo model Ihe effect of attached plants on stream quat-
,
ity. In contrasl lo ftoating plants, they assumed that the rooted plants have a constant
Headwater sources-2 (data type 10). This card group, one card per headwa- biomass over the study stretch. Thus, rather than being modeled, the plants are as-
ter, defines the boundary conditions for plants and nutrients at the upstream ends of sumed to aet as a zero-order sink of nutrients,
the system. dn
H - = -ana k,(T, I)a ' (36.30)
dt*
Polnt load-2 (data type l1a). This card group, one card per point source or
withdrawal, defines the loading values for plants and nutrients. Note that the point H -dp = -apa k,(T. J)a
'
(36.31)
load numbers are not the same as either the reach or element numbers. Rather, the d t*
point sources or withdrawals are numbered consecutively (starting at 1) from the
most upstream to the most downstream.
where a' = fixed-plant biomass in areal units (p.gChl m-2) and H = depth (m).
Usingboundaryconditionsof n = noandp = Poat themixingpoint,these
equationscanbe solvedfor
A plot of the QUAL2E output for the algae/nutrient run is shown in Fig. 36.8.
The oxygen levels are less than the previous runs because of the effect of nitrifica-
tion. n = no _ anakga' (36.32)
-¡¡-t"
apakga' . (36.33)
36.3 FIXEDPLANTS INSTREAMS p = Po- -¡¡-t
Now that we know how to simulate phytoplankton dynamics in ftowing waters, we NotethatEqs.36.32and 36.33can be solvedforthe criticaltraveltimes,
tum to the fixed plants that are particularly prevalent in shallower systems. These
range from large macrophytes to small microalgae called periplryton.
t~ = no - 100 (36.34)
am,k,a'

ff~
..
 ...
-....----

660 PARTv Eutrophication and Temperature LECTURE

.16 Eutrophication in Flowing Waters 661

n, p (Jlg L-1)

2000
a' (gChl m-2)
I "- a'
1000
'] O
'\1 /' n IImhlng ..
-,.J
f
Time when n
--Pllmlllng
(a) Floating plants (b) Fixed plants
_11m/llng
Tnw"!! time (d) -:1 FIGURE 36.10
FIGURE 36.9 Conlrasl belween how lighl is characlerized in (a) Iloallng
A simple oxygen balance lor a stream including lixed-plant growth. (inlegraled over deplh) and (b) lixed-planl (Ievel delermined
al deplh 01 planls) models.

1; = Po- 25 (36.35) was delerminedby inlegralinga growlhmodeloverlimeanddeplh(Fig.36.10(/).

apakga'
The model results are displayed in Fig. 36.9 for a stream with a significant
effluenHo-flow ratio. The simulation also reflects an effluent with a low N:P ratio.
As expected from Eqs. 36.32 through 36.33, the nutrients drop as the plants grow.
However, in contrast to Fig. 36.1, the drop is linear.
Observe that the travel time to reach nutrient limitation is shorter than for Fig.
36.1. This particular result is due to the parameter values, the initial conditions cho-
sen for the two plots, and the assumptions underlying the Thomann-Mueller models.
However, it should be recognized that, all other factors being equal, attached plants
and their impacts tend to be more localized than floating plants. That is, their bio-
mass and associated effects tend to be more concentrated immediately downstream
ftom point sources because they are not being transported downstream as they grow.
Having I'aid this, I should stress that each situation must be assessed on a case-
by-case basis. In particular the impact of light has been neglected in the foregoing
analyses. Therefore, as described next, more detailed mass balance models are re-
quired for site-specific water-quality evaluations.
36.3.2 ModeIlng Attaehed Plants (Advaneed Topie)
Attached plants differ from floating plants in some fundamental ways:
· water body.
· As the name implies, fixed plants do not advect downstream and they don't set-
He.Consequently a phytoplankton model (e.g., Eqs. 36.15 through 36.17) can be
· There is typically a maximum density for attached plants. This maximum depends
adapted to attached plants by merely removing advection and settling.
· Although some attached algae can extend up through the water column, most re-
!¡ide on or near the bottom. Thus light attenuation must be handled differently than
fQr F}1ytoplankton. Recall that for the phytoplankton, the light supplied for growth
For bOllomplants, growlh would be dependent on Ihe lighl delivered to the depth
al which the plants reside (Fig. 36.IOb). Consequenlly Ihe growlh allenualioll
factor for light could be simply determined by integrating over time for a tixed
depth, as in
cPl = LIT'. F(/)dl (36.36)
TI'
where F(/) is one 01'Ihe growth models for light. For example, if Steele's model
is used, Eq. 33.19 and the Beer-Lambert law (Eq. 33.21) can be substiluled inlo
Eq. 36.36, and Ihe resuh integrated to give
f(,e-k,.1I -!.,,~+ I
cPl= e 1, (36.37) -,
Is
The form of this function means that light will have a much more dramatic impac.;1
on bOllomplanls Ihan on noating plants. In parlicular it means Ihal water depth
will have a much grealer impacl on whether fixed planls can even grow in the
Allached planls onen require a parlicular bOllom substrate in urder to !!rIJw.Sumc
are rooled and consequenlly need a fine-gmincd, soil-type hollum. ( .llIns alladl
lo rocks. Thus jusl because Ihere is optimal lighl, nutricnts, and tcmpcnllurc, iI
parlicular planl may nol grow if suilable substrale is absent.
on light limilation and self-shading, space constraints due to a finile area of sub-
strate, and the fact that the plants are not mobile. Ahhough ftoating plants can also
reach such limits, attached planls tend to more easily approach maximum density,
particularly in nutrient-ric.;henvironments with plenty of light.
 662 PARTv Eutrophication and Temperature LECI1JRE36 Eutrophication in Flowing Waters 663

PROBLEMS
J..l9 L-1 36.1. In Sec. 36.1.1 we assume that both kinetic losses and settling are terminal; thal ¡g, Utey
2000
do not feed back to the nutrient pools. How could such assumptions be justified71f they
are erroneous, what would be the ramifications for the subsequent conclusions of the
200 analysis, if any?
36.2, The highest recorded phytoplanklon chlorophyll levels found in temperate lakes are
1000 aboul 100 ¡.Lg L -l. The maximum macrophyte den sities are about 500 S-dry weight
a 100 n,p m - 2. Because of their high content of structural carbon, macrophytes typically have low
chlorophyll-to-carbon ratios, on the order of 10 mgChl gC-1 .They have typical carbon
n IImltlng contents of about 40 gC g-dry weight-I. If the macrophytes grow in I m of water,
translate the maximum density into a volume-specific concentration of chlorophyll.
o o pllmltlng 36.3. A phytoplankton-dominated stream has the following characteristics:
o i 5
10
Upslream.'
Flow = 0.5 cms P = 50 ¡.LgPL-1 11= 1000 ¡.LgNL-1 a = SO ¡.LgChlL-I
Time when n Travel time (d)
becomes IImlllng Po;,,' source:

FIGURE E36.2 Flow = 0.75cms P = 5000¡.LgPL-1 11= 25.000¡.LgNL-1 a = O¡.LgChfL-1

Use the simplified analysis described in Sec. 36.1.1 to determine (a) the travel time
EXAMPI,E 36.2. NUMERICAL STREAM FIXED-PLANT SIMULATION. when nutrients become limiting, (b) which nutrient is limiting, and (e) what will happen
Duplicate the calculations in Example 36.1, but fix the plants by removing the transport if nitrogen removal is installed and the nitrogen efftuent is reduced to 5000 ¡.LgN L -l.
terms from the algal balance. Assume the following parameter values: kg = 0.5 d-I, k. = 0.2 d-I, n:p:a = 7.2:1:1,
Solution: The model equations from Example 36.1 can be modified, as k..p = 4 ¡.LgP L -1, and ks. = 15 ¡.LgN L -l.

:.l 36.4. Repeat Prob. 36.3, but use the control-volume approach described in Example 36.1.
-
V¡~~¡ = (kg(T. 1) min[l/>p.1/>.] kraL V¡a¡ (36.38) 36.S. Repeat Probo 36.4, but inelude nitrification ofammonium to nitrale at arate of 0.05 d-I .
Assume that the nitrate is unavailable for phytoplankton growth.
(36.39)
V¡~¡ I
= Q¡_l,ill¡-I - Qi.i+III¡- a.a kg(T. I)min[l/>p. 1/>.]- kra\¡ V¡a¡ 36.6. Using the same river as in Prob. 26.4, modify your file so that temperature is com-
a, puted. Note that the simulation is to be performed for September 10, at latitude 40° N,
longitude 111°W, and an elevation of 1000 m. The wind speed is 2 m S-I, barometric
V¡;:,¡ = Q¡-I.¡l'¡-I I
- Q¡.¡+IP¡ - aptl kg(T. I)min[l/>p. 1/>.] - k",\ ¡ V¡a¡ (36.40)
pressure is 970 millibars (mb), and the dry- and wel-bulb air temperatures are 30 and
Along with appropriate boundary conditions, these three equations can then be integrated 23°C. respectively. The dust attenuation coefficient is 0.10 and the eloudiness is 0.2.
numerically using the solution techniques presenled in Lees. 11 through 13. Figure E36.2 Use QUAL2E to simulate the levels of CBOD, oxygen, and temperature for this case.
""
shows the resulting steady-state profiles employing parameter values identicallo those
36.7. Modify the file from Prob. 36.6 so that nutrients and plants are computed. The following
used in Example 36.1.
concenlralions are available for the headwater and the point load:
Notice how this model behaves in a radically different fashion from Fig. E36.1. The
plants are at a higher level than in the phytoplankton model. Further. they are at a fairly Headwater Polnt load
conslant level until the travellime when nilrogen starts to become limiting. At this point Constltllent VnUs (r) (w)
the plants die out as losses dominate growth. The resulting shape is quite similar to the
Organic nitrogen mgNV' 0.05 10
assumed constant distribulion ofThomann and Mueller's simplified approach (compare
Anllnonia nitrogen mgNL I 0.05 15
with Fig. 36.9). 0.02 0.05
Nilrite nilrogen mgNL'
Nitrate nitrogen mgN L-' 0.1 1
Organic phosphoru mgP L-' 0.01 0.5
To summarize, different mode! formulations are required to simulate floating Inorganic phophorus mgP L-' 0.01 10
and fixed plants in flowing waters. In general, given favorable light conditions, Algae ¡.tgChl L-' 1.0 0.0
fixedplants tend to growto higherbiomasslevelsand to be concentratedcloserto
point sources than floating varieties. However, light attcnuation with depth and sub- For alllhe olher necessary coefficienls employ the same values as were used in Fig.
strate suitability are strong determinants of whether fixed varieties can thrive. Future 36.7. Use QUAL2E to simulate the levels of CBOD, oxygen,temperature,nitrogen,
stream eutrophication frameworks including both types should be developed. Such phosphorus, and algae for this case.
models could prove useful in dctermining whether either fixed or floating plants
36.8. Repeal Probo 36.4, but fix Ihe plants.
might dominategivenchanginglighl. n\ltricnt,amiternpcratureconditions.
~_ ; L...j ... , ., ._ ...._..
_.- ---

PART VI

Cf1emistry
/

I
I
t-

Part VI is designed to introduce youto the rudiments of integrating aquatic chemistry

into our water-quality-modeling framework.
Lecture 37 provides background on equilibrium chemistry. After an introdu¡;-
tion to units, the subjects of chemical equilibria and pH are reviewed. Approaches
for making chemical equilibria computations are also outlined.
In Lecture 38 I describe how equilibrium chemistry can be integrated into a
simple mass-balance framework. The important concept of a local equilibriurn is
introduced and the dissociation of ammonia is used as an example.
Finally the calculation of pH is used as a context for illustrating the basic
approach in Lecture 39. This material focuses on simple CSTRs dominated by
carbonate chemistry. CSTRs are used to show how pH is calculated in systems
subject to both plant activity and gas transfer.

665
 LECTURE 37

Equilibrium Chemistry

.¡

.1:P~~~ijf:s~M~t\1
'. '''ñé~ i sh~tIt.,_,_
,k;~l~w,tHeJ¡\t.'blffiiAs~t:tion,lItidtlttJi
'
f
m' 0.tf&"'\V:e~ b>'ll deSCH
,

ptiQn ,OfP~~~idaltYJ ,!
GhHériuefil
,, ~UitlbHt mit tah be ~irli)ldyeald~
, ', , ,
,, , , , , ,

1 f¡Y'i~íitj, " , '

,"" i' ,." .. """Io:.,~,", f

At the beginning of this book, I made the point that the transport and fate of sub-
stances in natural waters was dictated by their physics, chemistry, and biology, In
the earlier lectures the chemistry and biology were represented by very simple first-
orderkinetics.Anyonewhohasever takena coursein biologyor chemistryknows
that there's a lot more to these subjects. In the last severallectures we've finalIy be-
gun to address the realistic modeling of biology. Now we will try to do the same for
chemistry.
To do this we must first review some fundamental concepts. In particular I de-
scribe the field of equilibrium chemistry and show how calculations are mad~ in that
field.

Ij¡!\
37.1 CHEMICAL UNITS AND CONVERSIONS
To this point we have exclusively expressed the "quantity" of a substance by its
mass. An alternative, which is superior from a chemical perspective, is to use moles
or gram-molecular weight. A mole refers lo the molecular weight in grams of a com-
pound. For example a mole of glucose, C6H12Ü6,can be calculated as being

667

-----
668 PARTVI Chemislry IlnUIIE.\1 Equilibrium Chelllhlry 669

moleeular weighl
1 moleofC6HJ206 = 6CC
12g Ig
+ 12HH +600
16g
= 180g Equivalenl weighl = 11(number ü( prolons)
60 g mole ~ = 60~.eq
-¡-;;;¡ lilole 1
= 60 mg lIleq

mass per liler

The gram-molecular weight is important because each mole ineludes the same
Normalily = equivalenl weight
]() mg 1.
--1
60 mg mel(
~ = 0.5 meq L
number of molecules regardless of the compound. This quantity (6.02 X 1023)is
calledAvogadro's number. A molar solution consists of I mole dissolved in enough Therefore the solulion is 0.5 N wilh respecllo carbonale.
water to make a total volume of I L.t (e) This is also an acid-base reaclion:
Another related way to express concentration is as equivalents per liter, or nor- .
J;',,' I welg h I = lIlolecular weight 60 g mole
lIIalconcentration. A one normal solution (designated as 1N) contains one equivalent ,-,,\ulvaen! =
11(number of prolons) 2 eq mole
weight of a substance per liter of solution. Thus normality is defined as
= 30 g eq 1 = ]0 mg meq 1
l . =.
NormalIy massper liter
. (37.1)
equlvalentwelght NOrilla l.Ity = --.mass pcr liler = -
]() mg L
-- I mCl( 1. 1
el(uiv;llelll wcighl ]() mg l11e'l I
Theequivalentweightis calculatedas Therefore the solution is I N with respecllo carbonale.

Equivalent weight = molecular

n weight
(37.2)
As in the previous example, a suhstance c¡m have different equivalenl weights
whcr.: n depends on the type of substance being quantified. In natural waters it is depending on the reaction. Further, as illustcaled in parts (b) and (c), an aeid or .1
usually based on one of three eriteria: base can have more than one equivalent weight. Although this might initially seem
. Charge of an ion confusing, it is offset by the advantage of expressing concenlration in equivalent
. Number of hydronium (protons) or hydroxyl ions transferred in an acid-base re- units. This advantage is that when substances react to 1'01'111
a product, the equivalents
of the reactants will be the same as Ihe equivalents nI' the prodllcts.
action
. Number of electrons transferred in an oxidation-reduction reaction
The following example iIIustrates the approach. 37.2 CHEMICAL EQUILlBRIA AND THE LAW OF MASS ACTION

EXAMPLE 37.1. CHEMICAL CONCENTRATION. Delermine Ihe normalily of In our discussion of reaction kinetics in Lec. 2, we presenled the reaction (Eq. 2.1)
30 mg L-1 of carbonale (C032-) for the reactions: I'¡
tiA + bB;:=('C + dO (373)
(a) Ca2+ + C032- ;::CaC03(J) 1'"

(b) HC03 - ;::C032- + W

(e) C032- + 2W ;:: H2C03 where the lowercase letters represenl stoichinmetric coefticiellls . '" "I'percasc
letters designate the reacting cumpounds.
Thc kinetics 01'rate of such reactinl\s can be expressed quantilall' hy the
The molecular weight of carbonate is 60.
law 01mass action, which states that the cate is propnrtionallo the concelllration 01
Solution: (a) For this case, the reaction involves the ionizalion of carbonale: the reactants. A simple way to formlllate the forward cate rf nI' the readinl1 in Eq.
37.3 is
Equivalent weight = molecular weight
n (ion charge)
60 g mole-I
2 eq mole-I
= 30~eq = 30 mg meq-I rf = kflAl"lBlb (37.4)
where the bracketed terms are molar concentrations 01'the reactants. t The rale of Ihe
mass per liler 30 mg L-1
I meqL-1 backward reaetion can be modeled as
Normality = equivalent weight 30 mg meq-I

Therefore, the solution is 1 N with respect to carbonate.

rb = kb[C]'IDJ'/ (37.5)
(b) This is an acid-base reaction:

, As discussed in Ihe next seclion, the reaction rate for ions should be defined in terms of aClivity. How-
'lbis is in contrasl 10 a molal solulion, where t mole is dissolved in t kg of solvent. ever, concentration provides an adequate approximation when dealing wilh dilule aqueous soluliolls.
670 PARTVI Chemistry ~I LECTURE37 Equilibrium Chemistry 671

After a sufficient period of time, the backward rate wil\ equal the forward rate, TABLE 37.1
The size parameter a for varlous 1005 (Kielland 1937)
k¡[A]"[Bt = kb[CnO]d (37.6)
Ion
which can be sol ved for a (ang$troms)

H+, A13+, Fe3+, La3+, Ce3, 9

K = k¡ = [C¡C[O]d (37.7) Mg2+, Be2+ 8
kb [A]"[B]b 11' Ca2+, Zn2+, Cu2+, Sn2+, Mn2+, Fe2' 6
Ba2', Sr+, Pb2+, C032- 5
where K is the equilibrium constant of the reaction. Na', HC03-, H2PO.-, CH3COO-, SO.2-, HPO.2-, PO.3- 4
K', Ag+, NH.', OW, CI-, CIO.-, N03-, '-, HS- 3

37.3 IONIC STRENGTH, CONDUCTIVITV,AND ACTIVITV

AZ7jü
Although concentration provides a good approximation of the chemical activity of
ions in dilute solutions, it is not adequate when dealing with concentrated solutions logy¡ = - 1 +Ba¡
'jü U (37.13)
such as sal! water. For such systems the effective concentration, or activity, is di- where ai = a parameter (in angstroms) corresponding to the size of the ion (Table
minished by forces of attraction between positive and negative ions. 37.l), and
The activity of an ion, {c}, can be determined by multiplying its molar concen-
tration [c] by an activity coefficient, A = 1.82X 106(eT)-3n = 0.5 (forwaterat 25°C) (37.14)

{c}= y[c] (37.8) and B = 50.3(eT)-ln = 0.33 (for water at 25°C) (37.15)

where the activity coefficient y is less than or equal to one. where e = the dielectric constant (= 78 for water) and T is in Kelvin.
To compute the activity of a solution, the strength of the ionie interactions must
be quantitied. Lewis and Randall (1921) introduced the concept of ionic strength to EXAMPLE 37.2. ACTlVITY CORRECTlONS. Carbonic acid (H2COJ) dissociates
measure the electric field in a solution, by the following reaction,
H2C03 ~ HC03 - + H+
IL = !¿[c]¡Z¡
¡
(37.9)
The equilibrium constant for this reaction is (25°C)

where Z¡ = the charge of species i. KI = 10-6.35 = [HC03 -][W]

A number of correlations have also been developed that allow ionie strength to [H2C03]
be computed without determining all the ions in solution. For example Langelier If the tolal dissolved solids is 100 mg L-1 and the hydrogen ion concentration is 10-7.
( 1936) indkated that the ionic strength could be related to the total dissolved solids, determine the ratio of bicarbonate (HC03 -) to carbonic acid with and without activity
corrections.
IL = 2.5 X 10-5 (TOS) (37.10)
Solution: First, we can estimate the ionic strength as (Eq. 37.10)
where TOS is the concentration of total dissolved solids in mg L-l. In another at-
tempt, Russell (1976) suggested that the ionie strength could be estimated on the ¡.t = 2.5 x 10-5(100) = 2.5 x 10-3
basis of the specific conductance, Because the ionic strength is less than 5 x 10- 3, Ihe activity coefficients can be calculate<Í
as (Eq. 37.12)
IL = 1.6 X 10-5 (specific conductance) (37.11)
where specific conductance must be measured in ILmhocm -l. 10g'YH' = log'YHc03- = -0.5(1)2 h.5 X 10-3 = -0.025
Once the ionic strength is quantified, a number of equations are available for de-
termining the activity coefficient. One, which holds for solutions with ionic strengths
'YM' = 'YHC03 - = 10-0.025= 0.9441
less than approximately 5 x 10-3, is called the Debye-Hückellimiting law, These values can be substituted into the equilibrium relationship,
logy¡ = -0.5Z; jU (37.12) 10-635 = 0.944 1[HC03 -]0.9441 (10-7)
[H2C03]
An altemative equation is also available for more concentrated solutions (ionic
strength < 0.1). Called the extended Debye-Hückel approximation, it has the form, which can be solved for
 --.-

I.WWRE 37 Equilibrium Chemistry 673

672 PARTVI Chemistry

[HC03 -] 10-6.35 This is sometimes referred to as p(x) notation (Sawyer el al. 1994). For Ihe hydro-
[H2C03] = 0.94412 (10-7) = 5.011 gen ion, Ihe use of Ihis nolation results in Ihe pH (short for Ihe French puissance
d'Hydrogene, which Iranslates as Ihe "power or inlensily of hydrogen"),
Jf we had not taken activities into account, the ratio would have been (37.21 )
pH = -loglO[Wl
[HC03 -] 10-6.35
[H2C03] = 10-7 = 4.467 Because of the negative sign and the logarilhmic transformalion of [H' \, a high
pH represents a low hydrogen ion aClivilY,whereas a low pH represents a high hy-
wf¡ich amounts to about an 11% differ~nce. (Irógenion activily. Consequently Ihe pH scale reflects the intensity of Ihe acidic or
basic (or alkaline) condition of Ihe water. As in Figure 37.1, low pH connOlesacidic
and high pH connoles alkaline walers, wilh a neulral condilion at a pH of 7.0. The
pH of natural walers ranges from 1.7 for volcanic lakes to 12.0 for c10sedalkaline
37.4 pH AND THE IONIZATION OF WATER systems (Hutchinson 1957). However, the typical range is from 6.0 to 9.0.
Pure water dissociates to yield a hydrogen ion (or more properly, hydrated protons;
see Stu~ and Morgan 198\), H+, and a hydroxyl ion, OH-.
37.5 EQUILlBRIUMCALCULATIONS í
H20~H+ + OH- (37.16)
Now thal we have defined our lerms, we will explore how chemisls use Ihe concepl
I

If the reaction is at equilibrium. the concentrations of reactants and products are of equilibrium lo estimate Ihe levels of chemical species in aquatic syslems. The I
defined by basic idea is lo delineale the various species that exist in a solulion. Once this has II
been done, the nexl step is to determinehow many of these are unknowns.Then
K = [W][OW] (37.17) equations reflecting Ihe interaclions among Ihc compollnds must be specificd. Fi-
[H20] nally. if the number of equations equals Ihe number of unknowns, the equations can
Because the concentratioJlof water in dilute aqueous solutions is much greater be solved for the concentralions of the species.
than the ioos and is decreased so liUleby the ionization. it may be assumed to be at a In Ihese lectures we focus on acid-base syslems. For such syslems Ihere are three
constant leve\. Therefore the equilibrium relationship is conventionally expressed as
Kw = [W][OW] (37.18) ..
types of equations Ihat are used:
Equilibrium relalionships
Concenlration conditions or mass balance
where Kw is known as the ion, or ioniZJltion, product o/water, which at 25°C is equal . Electroneutrality conditions or charge balance
to approximately 10-14. The value of Kw can be computed at other temperatures by The aClual approach can be elaborated on by an example.
(Harne4 and Hamer 1933)
4787.3 EXAMPLE 37.3. EQUlLIBRIUM CALCULATIONS. Determine the cou.:enlra-
(37.19) tions of the species in a 3 x 10-4 aqueous solution of HCI at 25°C. Note that dissociation
pKw = ---;¡;;- + 7. I32110glOTa + 0.010365 Ta - 22.80 constant for hydrochloric acid is 103.

This relationship predicts a mildly decreasing trend ofpKw with increasing temper- Solution: First, we can list the reslIlting reactions,
ature. In other words there is more dissociation at higher temperatures. H20~W +Otr (37.21)
For puce water, equal amounts of [H+] and [OH-] are formed. and therefore \\7.23)
and HCI~W +CI-
the concentration of either is equal to apprüximately 10-7. When an acid is added
to the water. it ionizes and increases the hydrogen ion concentration. Consequently. Thus there are four species thal must be determined: HCI, CI-, H+, and 011 . Couse-
for Equation 37.18 to hGld.the hydroxyl concentration must decrease so that the quently we require four equatious.
product of the ions equals 10-14.For example if the addition of an acid to pure water The first two equations are provided by the c¡¡uilibrium couditions li)r water,
raises [H+] from 10-7 to 10-5. then [OH-] must decrease to 10-9. [W][OW] = 10-14 ,37.24)
Because it is somewhat cumbersome to work in terms of negative powers of 10,
and for hydrochloric acid,
chemists typically employ .
the negativc¡logarithm:
)

[W][Cn = 103 (37.25)

p(x) = -loglO X (37.20) [ifCI]
 LEcnJRE 31 Equilibrium Chemistry 675
11:II

Next we recognize that this is a closed system and that the amount of chloriné in
, the system must be equal to the amount added originally. This is termed a concentration
condition and can be written as
..
.,... b... ~ 'g
'O
.&:- Qj
¡¡;
Cr = [HCI]+ [Cn = 3 x 10-4 (37.26)

I
I
111110
!~~
o
:z:
- .&:-
UCD
m .
~
.S
c
.2
where Cr = total
balance for chlorine.
concentration of chlorine. Note that this relationship amounts to a mass

Finally the positive and negative charges in solution must be in balance,

I
.!!~ ~
I m a)- e [W] = [OWI+ICn (37.27)
I g¡:::-
e":
"t:I
>-
.c.
I et:.
el: ~
"t:I
This is termed an electroneutrality condition.
C These four eqllations can now be sol ved for Ihe concentrations of the species. No-
1. m ni tice that Ihey are a nonlinear system. Therefore some algebraic manipulation is in order.
o:íñ c
.2 First, sol ve Eq. 37.25 for [HCI],
I 11 ~ ~~
m'"
c
GI
I g,
&:
::lE
e-
I!! ~ [HCIJ = [W j[CI- J
103 (37.28)
I

I
11
.5
j
Ci
f
_:1.,
~ocó
- ..11
'&-
"t:I

~
>-
.c.
Next. substitllte Ihis result illto Eq. 37.26 and solve for
.z..- ~~ 'O
i -me. ..
c . [CI ) = 3 X 10-4 (37.29)
I 1íñ o¡ .2 e
¡¡;o
[W]
I ,.. :z: 6... '+' 6... L'
:z: mIS I~_
-..o ..=
-o 1+101
'1 a. =. e. ~~ i;1S !5.c
OGl Equation 37.24 can be sol ved for

t ::lE!!!.

!&:~
:s- .
Q,I!!!!!.
I
Q c.c.
0-
O¡¡;
~i
10-14
(37.30)
11
g,
&: =-
oC'! g~
[OW) = [W)
I
1 I!! J:'- Finally Eqs. 37.29 alld 37.30 can be substilllted into Eq. 37.27 to give
I U iíñ o!2. c.~
'6 ID~ &.-
&:U'" lV"GI
10-14 3 x 10-4
ü I!!'S"; :::(¡;
el: o..
0-- IW] = IH') + [W)
(37.31)
!-
-"1 .e.~
.c.c
~s 1+ 103
~r1..m o..
...o Further algebraic maniplllation yields
¡r-iCñ -miJ!
-gcq.5~
(/)!:!.> ..GIoc [W]3 + 103[WJ2- 0.3[W) - 10-11= O (37.32)
SE
g 3 ¡;;- e>E
c o Thus our manipulations have generated a cubic equation in [H' J that can be solved
i:a.!i
..J .~ ~ numerically for (H '] = 3 x 10-4, which corresponds to a pH of 3.52. The other species
.c.E

o ~...
..
b
-
lIüC'!
~'g
¡me.
"o
~(/)(/)
(')-0
w~..
GI_
can be calculated as

10-14
(37.33)
a) a:GI:a. ./' IOW) = 3 X 10-4 = 3.31 X 10-11 M
5!GI
:§: ~ ~ ¡¡:~~ 3 X 10-4
[CI-) = 3 X 10-4 = 2.9999991 X 10-4 M
(37.34)
1+-
103

(HCI! = 3 x 104 - 2.9999991 X 10-4 = 9.06 X 10-11 M (37.35)

674
 ~ ~._-

676 PART VI Chemislry

PROBLEMS LECTURE 38
-- ---
37.1.Suppose that you introduce '2 mg of nitrogen in the form of the ammonium ion into a
300-mL ftask. Immediately allcr introducing the ammonium. you IiII the remainder of
the volume with distilled water and stopper the ftask. Determine Ihe pH if the temper- Coupling Equilibrium Chemistry
ature is 25°C. Nole that ammonium dissociates according to
and Mass Balance
NH4 + ~NH3 + H+

where Ihe equilibrium coeffkienl for the reaction is related to temperature by

pK = 0.09018 + 27~.92

in which T. is in Kelvin.
{
37.2. Determine molar concentrations of all the species in a 10-4 M water solulion of ilcelic \
acid at a temperature of 25°C. Note that :1' . acid dissociates according to
CH3COOH ~ Ii + CH3COO-
where the pK for the reaction is 4.70.

37.3. Repeat Example 37.2, but use the extended Debye-HUckel approximation wilh T =
25°C.
37.4. Determine the hydrogen and hydroxyl ion concentrations of waler al a pH of 8.5 and a ~OYERVIEW: 1showhow the equilib¡ .111chemistry approach can be inte-
temperature of 30°C.
~te~ wjth *e J11as~-l>atance models that we bave used thtoughout this bOQk.
~Q 4.. .. .Sj lj~Jft>;4~.c~ th~ POtio. n ~.f,10001~Uiljb p'um \1n4 illu'str* howIt
?~
proYnt~~1!lJ\eim~
. . . . to. '§olve
.. . fast IUld. slow
. . reactions In tandem. Then
. J extend the
Ir
< .
.~I1Pfo~ch'm .
modelhlg c~emistfYIn ¡tatural systems.
.. . .
_ .....-

To Ibispoinl we have addressed Iwodifferenl modeling perspeclives Ibal are capable

of characlerizing tbe chemistry of natural waters. (Fig. 38.1). Tbroughout most of
tbe previous lectures we have used tbe mass-balance approach (Fig. 38.la). Then in
Lecture 37 we learned about equilibrium chemistry modeling (Fig. 38.1b). We now
unify tbese two perspectives. Before doing tbis 1must first present the mathemalical
idea Ibat underlies tbis unification: tbe local equilibrium.

38.1 LOCAL EQUILlBRIUM

Thelocal equilibrium idea relales to systems conlaining bolh fasl and slow reacliolls.
Suppose Ibat you are modeling Ihe situation depicled in Fig. 38.2. This beaker in-
eludes two reaclants thal are subjecllo Iwo purging mechanisms Ihal deplele lI1ass
from Ihe system: flusbing and a decay reaclioll. Observe that flushing acts 011bolh
reactants, whereas tbe decay loss affects only Ibe second. Furlher, Ihe reaclants are
coupled by two ftrst-order reaclions. Mass balances for the two conslituents can be
written as
dCI
(38.1)
V-¡¡¡ = QCI,in - QCI - kI2VC¡ + k21VC2

677
 -

678 PARTVI Chemistry LECfURE38 Coupling Equilibrium Chemistry and Mass Balance 679

If the above scenario is troe, Di Toro (1976) has kuggested that thé two maSs-

-er
" dIdc _ aA + bB ~
i cC + dD
FIGURE 38.1
Two modeling perspectives used
balance equalions (38.1 and 38.2) can be added lo yield

where W = lotalloading
dCT
Vd( =W- QCT - kVC2
of both conslituents (in which W = QCl.in +
(38.5)
QC2.in)and
- QC;n- Qc - kVc to characlerize the chemistry 01 CT = CI + C2 (38.6)
la\ Mass balance (b) Chemlcal equlllbrlum natural waters.
Notice Ihal Ihe lerms represenling the fasl reactions were canceled by the addi-
tion of Eqs. 38.1 and 38.2. This is due lo the fact Ihat if the reaclions are at equilib-
dC2 rium and Eq. 38.3 holds, the lerms kl2VCI and k21VC2are equal and would cancel
V-
dt
= QC2in. - QC2 + kl2VCI - k21VC2 - kVC2 (38.2)
out.
Now al Ihis point Ihe mass balance, although simpler, cannol be solved, be-
Given initial conditions, we can solve these equations for lhe concentration of
cause il is a single equation with two unknowns, CT and C2.To rectify this diffi-
lhe two reaclanls as a funclion of lime. Now suppose Ihal Ihe reaclions linking Ihe
culty, we recognize Ihat Eqs. 38.4 and 38.6 represent a system of two equations
reaclanls are mllch fasler Ihan Ihe olher mass Iransfers and Iransformalions. For ex-
that can be used lo compute CI and C2in terms of CT.For example Eq. 38.4 can be
ample let's say Ihal kl2 and k21were on Ihe order of I s-I and all the olher model solved for
rales were on Ihe order of Id-l. For example Ihe residence lime (VIQ) mighl be
I d and Ihe decay rale of Ihe second reaclanl mighl be 1 d-I . ]f we were simulaling C2 = KCI (38.7)
Irends on a lime scale of days, due lo lhe presence of Ihe fasl coupling reaclions, a which can be substituted inlo Eq. 38.6, which in turo can be solved for
very small lime slep would have lo be used lo mainlain slabilily.
Now let's Ihink aboul Ihe silualion a Httle more cIosely. ]f Ihe coupling reac- CI = FICT (38.8)
lions are Sllfficientlyfasl, Ihey will go lo completion much quicker Ihan Ihe olher where F I is Ihe fraction of the total concentralion thal is represented by the first
mechanisms. In facl Ihe backward and forward reaclions should rapidly reach a constiluenl,
balance. Thus, when viewed from Ihe perspeclive of Ihe olher processes, Ihe fol-
lowing equilibrium would hold: 1
(38.9)
F. = 1 + K
kl2 V CI = k21V C2 (38.3)
Equation 38.8 can be subsliluled back into Eq. 38.6, which can be solved for
Consequenlly Ihe ralio of Ihe Iwo reaclanls should be fixed al a constanl value. This
of course is the idea behind an equilibrium conslan\. For Ihe presenl example, Eq. C2 = F2CT (38.10)
38.3 can be used lo derive
where F2 is the fraclion of the total concentration Ihal is represented by the second
K = ~ = C2 (38.4) constituent,
k21 CI
K
F2=- 1 + K (38.11)
where K = an equilibrium conslanl Ihal specifies Ihe ratio of Ihe species. J
Finally Eq. 38.10 can be subslituted into Eq. 38.5 to give
dCT
V- dt =W- QCT - kVF2CT C~8.12)
Qc!,!!!
QC,.ln Thus the original system of differential equations has been replaced by a single
mass balance that can be solved for the total mass in the system as a function of
time. Because we have eIiminated the fast terms fcom the differential equation, a
C1 ~
reasonable time step can be used in the solution scheme. Then at eve¡y time step,
FIGURE 38.2 Eqs. 38.81and 38.10 can be used to calculate the concentrations of each of the two
kVc, Two reactants in a complelely mixed system. interacting species. These species are said to be at a local equilibrium.
~ ~ ""r~'~'_-.t-_ ___ l._:.. ':...~ ; ~ _.. .

680 PART VI Chemistry LECTURE38 Coupling Equilibrium Chemislry and Mass 8al;'I1"': 681
,
38.2 LOCAL EQUILlBRIA AND CHEMICAL REACTIONS Because the chemica] reaction is much faster than the other transpon and trans-
formation terms, a ]ocalequilibrium holds. Therefore the two equations can be added
In the previous section we modeled a coupled reaction that looked like to yield
k'2VC,
(38.13) dC1'
CI~C2 V- = (38.19)
k21VC2 d1 QCTin
. - QC1' - knV[NH4+] - vvA,[NH3]

Such ~ reaction is directly analogous to the chemical equations developed in the w}1ere
/'
preyious lecture. For example the dissociation of ammonium ion can be represen- C1' = [NH4 +] + [NH3] (38.20)
~edas
,:1
k,V[NH.+1 and K = [NH3][W] (38.21)
NH4 + , . NH3 + H+ (38.14) [NH4+]
k.V[NH)J(WI
As was done earlier in the lecture, Eqs. 38.20 and 38.21 can be solved simulta-
Thus the forward and backward arrows can be represented as reactions that neously for (
transfer mass forward and backward between the reactants and products. If the re-
action is at equilibrium. it means that [NH4 +] = F¡C1' where F¡ = [H+]/K (38.22)
1 + [H
I
kI V[N~ +] = kbV[NH3][W] (38.15) and [NH3] = F"c]' whereFu = 1 (38.23)
from which follows the equilibrium constant I

kl [NH3J[W] where F¡ and Fu are the ionized and un-ionized fractions, respectively.
K = kb = [NH4+] (38.16) FinaIly Eq. 38.22 and 38.23 can be substituted into Eq. 38.19 to give
dC1'
l'j'ow suppose we want to model the system depicted in Fig. 38.3. Mass balances
V dt = QC1'.in- QC1'- knVF¡C1'- vvA,FuC1' (38.24)
for IlIJImonium ion and arnmonia can be wriuen as
Thus this mass balance can be solved to determine levels of both ammonium and
yf~+] = Q[N~+]in - Q[N~+] - kIV[N~+] + kbV[NH3][W] - knV[NH4+] , ammonia in the system. However, observe that the pH is required to obtain solutions.
(38. (7) This is ilIustrated in the foIlowing example.

EXAMPLE 38.1. AMMONIA BALANCE FOR A LAKE. A lake has Ihe following
d[NH3] k + k + characteristics:
[
V~ = Q[NH3]in- Q[NH3] + I V[NH4 ] - b V NH3][H ] - vvAslNH3]
(38.18) Volume = 50.000 m3
Surface area = 25.000 m2
Mean depth = 2 m
Inflow = outflow = 7500 m3 d-I
Volatlllzatlon
Wind speed = 3 mps
v,A(NHJ]
Inflow Outflow The lake's inflow has an ammonium concentration of 6 mgN L-l. If the lake's pH is 9
and the temperature is 25°C, determine the steady-state concentrations of ammunium and
Q(NH¡]ID
..Q(NH¡] ammonia. Note that the pK for the reaction is equal to 9.25. Also, employ a volatilization
velocity ofO.1 m d-I and a nitrification rate ofO.05 d-I.
QINHJ]ID Q(NHJ]
Solution: First, we must convert the inflow concentration to moles, ¡.
18 gNH4+ moleNH4+ ~
Nltrlflcatlon
(N"¡]
k,V(NH¡]

k.flNHJ[U+]
(NHJ]
[NH +¡. = 6gN
4 In m3 ( 14 gN )( 18 gNH4+ )( 103L . )
= 04286 x lO'] M

Equations 38.22 and 38.23 can be used to compute

k.V(NH¡]
[H]+/K 10-9/10-9,25
FIGURE 38.3
F¡ = 1+ [H+]/K = 1 + 10-9/10-9,25= 0.64
Ammonium ion and ammonia in a completely mixed system. and Fu = l - 0.64 = 0.36
682 PARTVI Chemistry

LECTURE 39
At steady-state, Eq. 38.24 can be solved for

CT -- Q
CT'
Q + kIlVF¡ + v"AF" ' .10
pH Modellng
_ 7500
- 7500 + 0.05(50.000)0,64 + 0.1(25,000)0.360.4286 x 10-3 = 0.3215 x lO-3M
The individual species can be determined as
[NH4+) = 0.64(0.3215 X 10-3) = 0.2057 X 10-3 M

and [NH3) = 0.36(0.3215 X 10-3) = 0.1157 x lO-3M

These results can al so be expressed in mass units as

c¡ = 0.2057 x 1O-3M(';~~)C~3L)= 2.88 mgNL-1

_ 14gN loJL
and c" = 0.1157x 103M ( mole )(---¡n3) = 1.62 mgN L-1

PROBLEMS
38.1. Repeat Example 38.1. but develop a plot of total ammonia and its species versus pH.

38.2. Repeat Example 38.1, but perform a time-variable calculation over 3 d, where the pH
varies sinusoidally according to
pH = 8 + 1.25 COS[21T(t- 0.25») Now I apply the theory from the previous two lectures to simulate the ptt be nat-
ural waters. Because the pH of many natural waters is dominated by the carbonate
where t = time (d) and t = O at midnight.
buffering system, I begin this lecture by describing the inorganic carbon system.
38.3. A detention pond at sea level is at a temperature of 20'C. The pond has a pH of 9.5 and
ammonia concentration of 5 mg L -l. The gas-film exchange coefficient for the lalce is
200 m d-I and the Henry's constant for ammonia is 1.37 X 10-5 atm m3 mole-l. Com- 39.1 FAST REACTIONS: INORGANIC CARBON CHEMISTRY
pute the following (/lote: tables and figures are unacceptable; you must use formulas):
(a) The un-ionized ammonia concentration The tendency of natural waters to remain within a relatively narrow band af hydrogen
(b) The flux of ammonia across the air-water interface ion activity is due to the presence of buffers that resist pH changes. The buffers do
this by scavenging H+ and OH- ions. In many freshwater systems. much of the
buffering is related to the dissolved inorganic carbon species: carbon dioxide (C02),
bicarbonate ion (HC03 -), and carbonate ion (C032-).
-/ When carbon dioxide is introduced into an aqueous solution. it combines with
-
water to form carbonic acid,
C02 + H20~H2C03 (19.1,
The carbonic acid. in tum, dissociates into ionic form, as in
H2C03 ~ HC03 - + H+ (39.2)
Becalfse the equilibrium constant for the hydration of carbon clioxide (Eq.,39.1)
is so small, the proportion present as carbonic acid is !1egligible.Therefore the pro-
cesses of hydration and dissociation conventionally are treated as a single reaction,.
683
 -- ",.~"

()84 fART VI ~~e¡nistry LEcrURE 39 pH Modeling 685

kJ Alkalinity can be formulated as t

H2C03. ~ HC03 - + H+ (39.3)
4 Alk = [HC03-] + 2[C032-] + [OW] - [W] (39.12)
where
To summarize, we now have developed five equations:
[i'hC03.) = [C02] + [H2C03]== [C02] (39.4)
o
K _ [H+)[HC03-] (39.13)
Tlte equi!ibrium constant for the combined reaction is ,- [H2C03.]
,/
/
K _ [Wl[HC03-] (39.5) K _ [W)[CO/-] (39.14)
I - [H2C03.] 2- [HC03-]

where K, is called the composite acidity constant of H2C03., or the firstdissociation

Kw = [Wl[OW] (39.15)
constant of carbonic acid. The value of K, can be computed as a function of absolute cT = [H2C03.]+ [HC03-] + [C032-] (39.16) \"
femperature by (Harned and Davis 1943)
Alk = [HC03 -] + 2[C032-] + [OW] - [W] (39.17) ¡
3404.71
pK1 = ;¡;;- + O.032786Ta- 14.8435 (39.6) with five unknowns: [H2C03.], [HC03-], [CO/-], [H+], and [OW]. ~
There are a variety of ways to solve this system of equations. Stumm and Morgan
T~e t>icarbonate ion, in turn, dissociates to yield a carbonate and a hydrogen ion, (1981) have shown how Eqs. 39.13 and 39.14 can be solved for
ks
HC03 - ~C032-
kt,
+ H+ (39.7) [H2CO/] = [W)[HC03-]
KI
(39.18)

Thl'( ~quilibrium constant for this reaction is

[CO 2-] = [HC03 -]K2 (39.19)
I
3 [W]
K _ [W)[C032-] (39.8)
2- [HC03-] These results can be substituted ioto the mass balance (Eq. 39.16), which can be
solved for I
I

vvhere K2 is called the second dissociation constant of carbonic acid. The value of
[H2C03.] = FOCT (39.20)
K2 can I>ecomputed as a function of temperature by (Harned and Scholes i941) I
I
t'O .
[HC03-] = FICT (39.21) II
I
(39.9)
pK2 = 29~39 + O.02379Ta- 6.498 [C032-] = F2CT (39.22)

The inorganic carbon syslt:Jn delineated above consists of five unknowns: where Fo, F" and F2 are the fractions ofthe total inorganic carbon in carbonic acid,
H'hCQ3.], [HC03 -], [C023-], [H+], and [OW]. Therefore five simultaneous bicarbonate, and carbonate, respectively:*
eqq~~jQnsare needed to solve for these unknowns. Three of the five are provided [W]2
t>y.he equilibrium relationships defined by Eqs. 39.5, 39.8, and the dissociation of Fo = (39.23)
[H+]2 + K,[H+] + K,K2
water (Eq. 37.18).
A fourth equation is the concentration condition, K,[W]
F, = (39.24)
(39.10) [H+]2 + K.[H+] + K.K2
CT = [H2C03.] + [HC03 -] + [C032-]
K,K2
)Vhere CT is total inorganic carbono 139.25)
Finally a fifih equation results from the fact that the solution must be electrically
F2 = [i-tt'j2+ K,iH
peutral. 'fhis c~n be expressed as a charge balance or electroneutrality equation,
1''''

. CB
(
+ [W)
~I"
= [JiC03 -) + 2[C03-] + [OW] + CA (39.11) tOur definilion of alkalinily is applicable only 10syslems in which olher buffers are al negligible lev-
els. Other reactions contribute lo alkalinily ifl fresh walers, but Ihese are often ignored because of the
where CB and CA are theamounts of base and acid that have been added to the system, overriding importance of inorganic carbon in many systems. Carbonate chemistry is not adequate for
describing saltwaler systems, and other buffers must be considered.
~~~f1ve1f ~ecause ittis usually impractical to measure the amounts of acid and
INole Ihat Stumm and Morgan employ Ihe parameters ao, a" and a2 lo designale Ihe fnlctions. We have
'~ase:a new quantity, alkalinity, is defined as the acid-neutralizing capacity of the used F's lo be consiSlenl with nomenclalure used elsewhere in Ihis lex\.
system.
r- 686 PARTVI tbemistry tJritJRBj9' "tI Mbciellfil M1~
I
I

Now Eqs. 39.20 to 39.22 can be substituted, along with Eq. 39.15, into tbe charge 't .... \.tI
'\'I(!
balance to give

O = FICT + 2F2cT + [~:] - [H+] - Alk (39.26)

1
Although it might not be apparent, Eq. 39.26 is a fourth-order polynomial in [H+]. I
Therefore it can be solved for [H+]. '- _ _ _ _ _ _ _11!2f9.!!1ICLC8~'lo

EXAMPLE 39.1. pH FOR THE INORGANIC CARDON SYSTEM. Determine

the pH for a system with an alkalinity of 2 meq L -1 and a total inorganic carbon concen-
tration of 3 mM L-l. Assume tbat pKw = 14, pK1 = 6.3, and pK2 = 10.3.
Solutlon: First. the terms in Eq. 39.26 can be combined to give a fourth-order polyno-
mialin [W]:
Slow reactlons I Fast reactlons
[W]4 + (K1+ Alk)!W]3 + (KIK2 + AlkKI - Kw - Klcr)!W]2
FIGURE 39.1
+ (AlkKIK2 - KIKw - 2K,K2cr)rWl- KIK2Kw = O
A slmplified representation 01 the Inorganic carbon system
Substituting parameter values gives and Its relationship with the loed chain. The Inorganlc carbon
specles shown are carbonate (C032-). bicarbonate (HC03 -),
[W]4 + 2.001X 1O-3[W]3- 5.012X 1O-IO[W]2- 1.055X 1O-19[W) and carbon dloxlde/carbonlc acld (H2C03'). The reactlons
-2.512 X 10-31 =O between these specles are relatlvely last, with rates on the
order 01 seconds to minutes. In contrast the reactlons wlth
Thisequationcan be solved for[W] = 2.51 X 1O-7(pH = 6.6).Substitutingtbisvalue the loed chaln [photosynthesls (p) and resplratlon (r)) and
back into Eqs. 39.20 through 39.22 gives volatillzatlon (v) are typically much slower, with rates on the
order 01 days.
[H2C03'] = (2.51 X 1O-7r 3 X 10-3
. (2.51 X 1O-7r + 10-6.3(2.51 X 10-7) + 10-6.310-10.3
= 0.33304(3 X 10-3) = O.OOIM Figure 39.1 is a simple representation ofhow inorganic carbon interacts wlth the
atmosphere and tbe food chain. Carbon dioxide enters and leaves the inorganic car-
10-6.3 (2.51 X 10-7) bon system via two major pathways: atmospheric and biological exchange processes.
[HC03 -] = 3 X 10-3
Note that other heterogeneous reactions such as calcium carbonate dissolution and
(2.51 X 1O-7r + 10-6.3(2.51 X 10-7) + 10-6.310-10.3
precipitation are neglected in this simple representation. .
= 0.666562(3 X 10-3) = 0.002M As can be seen in the figure, tbe reactions are divided into fast and slow ¡nter-
10-6.310 -10.3 actions. Before developing a model, we can review these interactions.
[COl-] = 3 X 10-3
(2.51 X 1O-7r + 10-6.3(2.51 X 10-7)+ 10-6.310-10.3 Atmospherlc exchange. The atmospheric exchange can be quantifled vla
= 0.000133(3 X 10-3) = 1.33 X 1O-4M

Watm = vvAs([H2CO/]S - [H2C03'])e~3L) (39.27)

39.2 SLOW REACTIONS: GAS TRANSFER AND PLANTS wbere Watm = "Ioading"of inorganiccarbondueto atmosphericexcbahgé
(mole d-I)
The inorganic carbon buffering system is inftuenced and enhanced greatly by a Vv = volatilization transfer coefficient (m d -1 )
lIumber of beterogeneous reactions tbat occur in nature. We may recall that het- A. = surface area of tbe air-water interface (m2)
erogeneous reactions are those tbat occur between phases, as opposed to homo- [H2C03'1. = saturation concentration of carbon dioxide (mole L-1)
geneous reactions, which occur within a single phase. For the inorganic carbon
system, the heterogeneous reactions inelude atmospheric exchange of CO2, disso- Carbon dioxide is overwhelmingly liquid-film controlled. The lral1sfer coeffi-
lution/precipitation of carbonate minerals such as calcium carbonate (CaC03), and cient can be correlated with the oxygen exchange coefficient, K¡ (m d- 1), by (Milis
photosynthesis/respiration. et al. 1982)
~_~_~."'_,,-."_'L.-'-'-;:;~_":"'~_ -.- \1 "----_

688 PART
VI Chemistry LECTURE 39 pH Modding 689

- 375 The reaction can be formulaled in a manner compatible with the carbon mass
balance as
-
l¿
E
a.
a. 350 W"p = ocX<R - P)As (39.32)
e
o where W,'p = net carbon load due to the combined effects of respiration and pho-
I 325 1855 11165 1975
/ R and P
tosynthesis (mole d-I)
= areal rates of respiration and photosynthesis (g m- 2 d- I)
§ 300
ocx = a factor to convert the mass units lo moles based on the stoichiometry
of Eq. 39.31 as well as the mass units used for R and P
8..
O For example suppose Ihat R and Pare calculaled in g02 m-2 d-I, as is commonly
q 275 done for Streeter-Phelps models. The convcrsion factor could be calculated as
1750 1800 1850 1900 1950 2000 _ 6moleC02 _ 003125 moleC02 '"
Year Oca- 6 X 32 g02 - . g02 (39.33)
AGURE 39.2 where Ihe notation Ocodenotes that the conversion is belween carbon and oxygen
Recenl concenlralions 01 carbon dioxide in Ihe almosphere as recorded at the units.
Mauna Loa Observalory, HawaiL The annual cycle is superimposed on Ihe long- In summary the heterogencous reactions wilh the almosphere and the foodchain
lerm Irend (Irom Keellng el al. 1982, 1989; Neftel el al. 1985).
- --¡¡roceed on time scales of hours to days and serve to introduce and remove CO2
from the waler. In tum the carbon dioxide takes part in the fast reactions among the
°.25 inorganic carbon species as dclineated above.
32
v. = ()
M K¡ (39.28)

where M = molecular weighl, which for carbon dioxide is 44. .'

39.3 MODELlNG pH IN NATURAL WATERS
Tbe saluration concenlralion, [H2C03 *)s' is Ihe level Ihat the waler would reach Now that 1 have delineated the major reactions affecting the inorganic carbon sys-
under sleady-slale conditions for a given lemperature. This level can be estimaled tem, I'll show how Ihey can be formulated to calculale the pH of natural walers.
on the basis of Henry's law, The key to this coupling is Ihat the heterogeneous reaclions goveming the inter-
[H2C03*). = KHPCO;¡ (39.29) relationships between the inorganic carbon species and water are Iypically much
fasler than atmospheric transfer and photosynthesis/respiration. Therelore the com-
when: KH = Henry's constant [mole (L atm)-I) and PC02 = partial pressureof putational strategy (as originally devised by Di Toro 1976) is to use a dynamic, dif
carbon dioxide in the almosphere (atm). ferential equalion 10compute the effect of Ihe helerogeneous reactions on the IOlal
The value of KHcan be compuled as a funclion oftemperature by (Edmond and inorganic carbon level of Ihe syslem with a lime slep on the order of days. Because
Gieskes 1970) they are much more rapid, the homogeneous reactions can always be assumed lo be al
2385.73 equilibrium on this time scale. Thus algebraic solution techniques can be employed
pKH = -r:;- - 0.0152642Tu
+ 14.0184 (39.30) to solve for pH and Ihe individual inorganic carbon species al each time step.
Mass balances can be developed for Ihe individual species involved in the inor-
The partial pressure OfCÜ2has been increasing, largely due lo Ihecombustion of ganic carbon syslem,
fossil fuels (Fig. 39.2). Present values are approximalely 1O-3.4satm (= 355 ppm).
d[H2CO]*¡ * + -
PhotosynthesisIRespiration. In addilion to reaeralion, CO2 is introduced to V dt' = Waun+ W,'p - k3V[H2C03 ) + k4V[H ][HCO] ¡ (39.3t>
and removed from a lake via biological reactions. To incorporale these mechanisms d[HCO]-] * + _
into our inorganic carhon model, we assume that photosynthesis and respiration pro- V dt' = k3V[H2C03 ) - k4V[H ][HC03 ] - ks V[HC03
~ed via the following simple formula:
+ k6V[W][C032-) (39.35) II
phOIO
6C02 + 6H20;::::::::::=C6H1206+ 602 (39.31)
.\';
I
respir
vd[C~:2-) = ksV[HC03-] - k6V[W][C032-] (39.36) ,
690 PART VI Chemistry tECTURB39 1>HModelinR 69 t

Since the reactions between the inorganic carbon species are much faster than
the gains and losses due to atmospheric and biotic exchange, a local equilibrium [HC03_] = 10-6.3(1.1X 10-8) . 2.034 X Ib-3
(1.1 X IO-sf + 10-6.3(1.1 X lO-S) + 10-6.310-10.3
assumption can be made and Eqs. 39.34 through 39.36 can be combined to yield
= 0.001981 M
dCT
(39.37) 10-6.310-10.3
V dt = Wa1m + Wrll' [COl-) = 2.034 X 10-3
(1.1 X IO-sf + 10-6.3 (1.1 X lO-S) + 10-6.310-10.3
where CT is defined by Eq. 39.10. Equation 39.37 aIlows computation of the dynam- = 9.011 X 10-6 M
ies of total inorganic carbon as a function of biotic and atmospheric interactions.

EXAMPLE 39.2. pH FOR THE INORGANIC CARBON SYSTEM WITH BI-

OTIC AND ATMOSPIIERIC INTERACTlONS. Delermine the steady-state pH for
PROBLEMS
Ihe same system as in Example 39.1, except that the syslem is open lo the atmosphere

[pKH = 1.46, Pc~ = 10-3.45alm, and v,. = 0.5 m d-I] and is subjectlo a level of R- P 39.1. Develop the fourth-order polynomial in [W) by combining the lerms in Eq. 39.26.
of 0.5 gO m-2 d- . Nole Ihallhis means Ihat respiralion dominales pholosynthesis and
Ihal Ihe planls are adding C02 lo Ihe waler. As in Example 39.1 use an alkalinity of 39.2. Determine the pH for a system with an alkalinity of 3 meq L-1 and a I()tal inorganic
carbon concenlration of 3.5 mM L-l. Assume all other values are Ihe same as in Exam-
2 meq L 1 and assume thal pK.. = 14, pKI = 6.3, and pK2 = 10.3.
pIe 39.1.
Solution: The sleady-stale inorganic carbon budgel can be written as
39.3. Delermine the steady-state pH for Ihe same system as in Example 39.1, except Ihat
v,.([H2CO/ls - [H2C03']) + aco(R - P) = O the system is open to Ihe atmosphere (pKH = 1.46,peoz = 10-3.45 atm, and v. =
where a,o is 0.03125 mole C02/g02 as calculaled previously in Eq. 39.33. The saturation 0.5 m d-I) and is not subject lo significanl plant activity. As in Prob. 39.2, use an
concenlralion of carbon dioxide can be computed as (Eq. 39.29) alkalinity of 2 meq L-l.
(H2C03'l., = 10-1.4610-3.45= 10-4.91= 1.23 X 10-5 M 39.4. Repeat Example 39.2, but perform a time-variable calculadon over 3 d, wl1erethe plant
photosynthesis varies diumally as a half-sinusoid with a photoperiod ofO.S d and a peak
'DIese values can be subsliluled along wilh olher parameters inlo the carbon budget,
of 3 gO m-2 d-I. Note that respiration is 1.5 gO m-2 d-I and Ihe system is 3 m deep.
39.5 A wastewaler treatment plant (Q =
= 1 X 1Q6 m3 d-I; T 25°C;pH 6.1; Alk = =
0.5J (1.23x 1O-~m~le- [H2C03'J)('~3L)+ 0.03125m;leO'\~d =O 0.003 eq L-I) discharges inlo a river (Q = 5 X 1Q6 m3 d-I; T = IOOC;pH = 7.2;
which can be sol ved for Alk = 0.0012 eq L-I). Assuming instantaneous mixing, calculate Ihe (a) alkalinity,
(b) total inorganic carbon, and (c) pH at the mixing point.
IH2CO/] = 4.355 X 10-5 mole L-1

Thus, for Ihis case, the carbonic aeíd level is fixed by Ihe atmospheric and biolic
inleraclions. At Ihis point, we still have five equalions, but Cr has replaced [H2C03'] as
an unknown.
Now, because we know Ihe level of carbon dioxide, Eq. 39.20 can be sol ved for
[H2C03']
Cr = (39.38)
Fo
which can be substituled ¡nlo Eq. 39.26 to give
~
.. "
'"

0= ([H2C03']/Fo)(F1+ 2F2)+ [~:) - [W] - Alk

Thus Ihe problem reduces lo solving this equalion for [H+). This can be done numerically
for [W) = 1.1 x IO-R(pH = 7.96). Subslituting this value back inlo Eqs. 39.38, 39.21,
¡lIId 39.22 gives

(1.1 X IOR)2 + 10-63(1.1 x IO"R)+ 10 6310-103

Cr = ,4.355 X 10-5
(1.1 X lO R)
= 0.002034 M
- ,... 1. __~

PART VII

Toxics
./

J
.,

The lectures in this part provide an introduction and an overview of water-quality

modeling of toxic substances.
Lecture 40 is an introduction to the toxic substance problem. After an overview
of the problem. I develop a mass balance for a toxic substance in a well-mixed lake.
My intent is to illustrate how the modeling of conventional poJlutants and of tOKic
substances differ. In particular the role of suspended solids and solidlliquid parti-
tioning is introduced. Then I add a sediment layer underneath the lake, and develop
a solids budget and a toxic substance model for the sediment-water system. Steady-
state and time-variable solutions aI\: ¡:enerated and compared with the well-mix.ed
lake model without sediment-water interactions.
The following lectures are devoted to the mechanisms that govern the ways in
which the toxicant is transported and transformed within the environment. Leclure
4/ deals with sorption of toxicants onto particulate matter and volatilizatioll across
the air-water interface. The first part of the lecture explains the mechanislI1of sorp-
tion and delineates how it is quantified. The second part is devoted to voLllli/ation.
Leclure 42 describes reactions that act to deplete toxics from water. These indude
photolysis, hydrolysis, and biodegradation.
Lerlllre 43 describes how radionucJides and heavy metals are moddcd. LeClurt'
44 illustrates how toxic substances are modeled in streams and estuaric~. FinaJly
Lecture 45 deals with modeling the fate of toxicants within organisms an.! Ihe fond
chain.

693
 LECTURE 40

introduction to Toxic-Substance
Modeling

..'
/'

The following lecture is intended to provide an overview of toxic modeling. Our

emphasis will be on developing simple mass balances for a well-mixed lake and its
underlying sediments. In this way we can investigate the similarities and differences
between conventional and toxic-pollutant models.
"
40.1 THE TOXICS PROBLEM
To this point the models described have almost exc1usively dealt with what are called
conventional pollutants. That is, pollutants that are a natural by-product of human
:~ activities and that overstimulate the natural productionldecomposition cyc1e. Before
.!
developing toxic models it is useful to reftect on how toxics differ from conventional
wastes.

40.1.1 Contrast of Conventlonal and Toxlc Pollutlon

The major differences between conventional and toxic pollution relate lo four areas:
t. Natural versus alíen. The conventional pollution problem typically deats with
the natural cycle of organic production and decomposition (recall Fig. t9. t). The

695
~.!#_~.----
696 PARTVII Toxics
discharge of sewage adds both organic matter and inorganic nutrientf to a wa-
ter body. The decomposition of the organic matter by bacteria can result in oxy-
gen depletion. The inorganic nutrients can stimulate excess plant growth. In t>oth
cases the problem involves an overstimulation of the natural processes govem-
ing the water body. In contrast many toxics do not occur naturally. Examples are
pesticides and other synthetic organics. For such "alien" pollutants the problem
fSone of poisoning or interference with natural processes.
2. Aes,hetics versus health. Although it would be an overstatement to contend that
conventional pollution deals solely with aesthetics, a strong case can be made
that the mitigation of "visual" pollution has been a prime motivation for conven-
tional waste treatment.ln contrast the toxic-substance issue is almost totally dom-
inated by health concems. Most toxic remediation focuses on the contamination
of (a) drinking water and (b) aquatic food stuffs.
3. Few versus many. Conventional water-quality management deals with on the
order of about 10 "pollutants." In contrast there are tens of thousands of organic
chemicals that could potentially be introduced into our natural waters. Further, a
large fraction of these are synthetic and are increasing every year (Schwarzen-
bach et al. 1993). If even a fraction of these proved toxic, the sheer numbers of
potential toxicants will have a profound effect on the resulting control strategies.
f!1rtl1erit is difficult to obtain detailed information on the factors goveming their
tfansport and fate in the environment. The study of problems such as dissolved-
oxy~en depletion and eutrophication is facilitated by the fact that they involve a
fewc~emicals. In contrast toxicant modeling is complicated by the vast number
of contaminants that are involved.
4. Single species versus solid/liquid partitioning. As was done in previous lectures,
the conventional paradigm usually treats the pollutant as a single species. Con-
sequently its strength in the water body is measured by a single concentration.
In contrast the transport, fate, and ecosystem impact of toxicants is intimately
connected with how they partition or associate with solid matter in and below the
water body. Thus toxic-substance analysis must distinguish between dissolved
and particulate forms. This distinction has an impact on transport and fate in the
sense that certain mechanisms differentially impact the two forms. For exam-
pie volatilization acts only on the dissolved component. The distinction also has
importance from the standpoint of assessing ecosystem impact. For example if bi-
otic contamination is being assessed, toxicant mass normalized to biotic mass is
a better metric of contamination than toxicant mass normalized to water volume.
The four points outlined above have implications for the modeling of toxicants.
Rather than an assimilative-capacity approach (recall Lec. 1),an alternative "screen-
ing" approach is sometimes adopted. In essence the strategy is designed to determine
which toxicants pose the greatest threats in a particular environment. One way to
focus this strategy is to divide the toxicants into classes.t
'It should be noted that compound-specific approaches are often required for adequate
and prediction of (ate and transpon. Schwarzenbach et al. (1993) is a good reference
LECfURE40 Introduction to Toxic-Substance Modeling 697
40.1.2 Classification of Toxics
There are many ways that toxic substances can be broken into categories. For exam-
pIe they can be divided by their functions (e.g., pesticides, cleaning products, fuels,
etc.) or their primary industrial association (agriculture, petroleum refining, textiles,
etc.). Alternatively a classification could be based on chemical criteria.
One effort at classification, that has become commonly used, was developed
tp/categorize the so-called "priority pollutants." As the result of a court settlement
involving the EPA and a number of environmental groups, the EPApublished such a
list. Although the list has been the subject of criticism (see Keith and Telliard 1979),
it provided some structure to the toxics problem.
The list of priority pollutants consists of 129 inorganic and organic toxicants.
The inorganics consisted of heavy metals and other contaminants such as asbestos
and cyanide. As listed in Table 40.1, the organics were divided into nine categories. t I
The resulting classification scheme is a mixed bag of usage (pesticides) and some
broad chemical classes.
Although the categories in Table 40.1 could be significantly modified, improved,
and refined, we will adopt them for use in this book.ln a later section we'lI attempt to
use their key chemical properties in conjunction with mathematical models to assess
how these chemicals could potentially impact lakes.
40.2 SOLlD-LlQUID PARTITIONING
In previous lectures we have focused on conventional pollutants. In all these exam-
pies the pollutant was represented by a single concentration. Toxic substances differ
from conventional pollutants in that they are divided or "partitioned" into particulate
and dissolved forms.
The primary reason for solid-liquid partitioning is to produce a more mecha-
nistically accurate characterization of the toxicant mass balance. In particular it is
known that several key mechanisms act selectively on one or the other of the two
forms. For example volatilization (that is, loss of the contaminant from the water to
the atmosphere) acts only on the dissolved fraction. Conversely settling acts solely
on the particulate-associated fraction.
Mathematically the total contaminant concentration, c (¡.tg m-3), is separated
into the two components
(40.1)
c = Cd + cp
where Cd = dissolved component (¡.tg m-3) and cp = particulate componenl
(¡.tg m-3). These components are assumed to represent fixed fractions of the total
concentration, as in
(40.2)
Cd = FdC
and cp = Fpc (40.3)
characterization where the parameters Fd and Fp are the fractions of the total contaminant that are
for this approach. in dissolved and particulate form, respectively. These fractions are a function óf the
 698 PARTVII Toxics
TABLE 40.1
Tbe categories of the organlc
Pollutant
Peslicides:
Generally chlorinated
hydrocarbons
l Polychlorinaled biphenyls
(PeBs):
Used in eleclrical
1
capacitors and Iransform-
ers, painls, plastics,
inseclicides, olher
induslrial producls
Halogenaled aliphalics
(HAHs):
Used in fire eXlinguishers,
refrigeranls, propelIanls,
pesticides, solvenls for
oils and greases, and in
dry cleaning
Elhers:
Used mainly as solvents
for polymer plastics
Phthalate eslers:
Used chiefty in
produclion of polyvinyl
chloride and Ihennoplaslics
as plaslicizers
priorlty pollutants (CEQ
Characterlstlcs
Readily assimilaled by
aqualic animals, fal-
soluble, concenlraled
Ihrough the food chain
(biomagnified), persistenl
in soil and sedimenls
Readily assimilaled by
aquatic animal s,
fal-soluble, subjecl
lo biomagnificalion,
persislent, chemicalIy
similar lOIhe chlorinaled
hydrocarbons
Largesl single class of
"priorily polIulanls"; can
cause damage to cenlral
nervous syslem and liver,
nol very persislenl
Polenl carcinogens; aqualic
loxicily and fale nOIwelI
underslood
Common aqualic polIulant,
modcralely toxic
bul leralogenic and
mutagenic properlies
in low concenlralion;
aqualic invertebrales are
particularly sensilive lo
loxie effecls; persistenl
and can be biomagnified
LECTURE
40 lntroduction to Toxic-Substance Mooeling 699
1. TABLE40.1 (continued)
1978)
Pollutant Characterlstlcs Sources and Remarks
Sources and Remarks "
Monocyclic aromalics Cenlral nervous syslem Enler environment during
Direcl applicalion lo farm and (MAHs) (excluding depressant; can damage production and by-product
foresl lands, runoff from phenols, cresols, and liver and kidneys produclion states by direct
lawns and gardens, urban phthalales): Used in Ihe volatilization, waslewater
runoff, discharge in induslrial manufacture of olher
wastewaler. Several chlorinaled chemicals, explosives,
hydrocarbon peslicides already dyes, and pigmenls, and
reslricled by EPA: aldrin, in solvents, fungicides,
dieldrin, DDT, DDD, endrin, and herbicides
heplachlor, lindane, chlordane
Municipal and induslrial discharges
Phenols:
Large-volume induslrial
Toxicily increases with
degree of chlorinalion of
Occur naturalIy in fossil fuels;
waslewaler from coke ovens,
disposed of in dumps and compounds used Ihe phenolic molecule; oil refineries, lar dislillation
landfilIs. TOSCA ban on chiefty as chemical very low concentralions planls, herbicide and plaslic
produclion after 6/1n9 bul will inlermediales in Ihe can lainl fish ftesh and manufacluring, can all contain
persisl in sediments; reslriclions produclion of synthelie impart objectionable odor phenolie compounds
in many freshwaler fisheries as polymers, dyesluffs, and lasle lo drinking
result of PeB polIulion (e.g., pigmenls, herbicides, and waler; difficult lO
lower Hudson, upper Housalonic, peslicides remove from waler by
parts of Lake Michigan) eonvenlional Irealmenl;
carcinogenie in mice
Produced by chlorinalion of waler,
vaporizalion during use. Large- Polyeyclic aromalic Carcinogenic in animals Fossil fuels (use, spills, and
volume induslrial chemicals, hydrocarbons (PAHs): and indireclly linked produclion), incomplele
widely dispersed, bUlless Ihreal Used as dyesluffs, lo cancer in humans; combuslion of hydrocarbons
lo Ihe environmenllhan persistenl chemical inlermediales, mosl work done on air
chemicals peslicides, herbicides, polIulion; more is nceded
molor oils, and fuels on Ihe aqualic loxicily
of Ihese compounds;
nol persislenl and are
Escape during produclion and use. biodegradable, although
Allhough some are volatile, bioaccumulation can
elhers have been idenlified in occur
some natural walers
Nilrosamines: Tests on laboratory Produelion and use can occur
Wasle disposal vaporization during Used in Ihe produclion animals have shown sponlaneously in food cooking
use (in nonplaslics) Ihe nilrosamines lO be
of organic chemicals and operalions
rubber; palenls exisl on some of Ihe mosl potenl
processes using Ihese carcinogens
compounds
i -"_ ~.,_ "tf.- '0.
~ .--

700 PARTVII Toxics I.ECTURE40 Inlroduclion lo Toxic-Subslance Modeling 701

1.0 o Loadlng Outflow

I!
Fp 0.5 0.5 fd

./
o
. . 1.0
0.001 0.01 0.1 10 100 1000
Weak sorberl Slrong sorberl
lowsolids Kdm highsolids Settllng ~
t

FIGURE 40.1 FIGURE 40.2

Schematic 01 a solids budget lor a well-mixed lake wilh no
Relalionship 01 the Iraction 01 contaminant in particulate and
sedimenl leedback 01 solids.
dissolved lorm to the dimensionless number: K.m.

contamil1ant's partition coefficient and the lake's suspended solids concentration, In Ihis seclion we do Ihis in the absence of sedimenl-water interaclions. Then, in Ihe
as in following section, we add a sediment compartment and assess its effect.
1
Fd = --- (40.4)
1 + Kdm 40.3.1 Solids Budget
Kdm Recall Ihal for our conventional models, a waler budget was required to allow liS
and Fp = 1 + Kdm (40.5) to calculate the effect of waler transport on our pollulant budget. One illlplication of
solid-liquid partitioning is that a solids budget is now required. Just as Ihe waler-flow
where Kd = a partition coefficient (m3 g-l) and m = suspended solids concentration rate aIlowed us to determine how much poIlutant was carried along with Ihe water, a
(g m-3). Note, we will derive these formulas when we discuss sorption in the next solids budget allows us lo estimale how much conlaminant wilI be transported along
lecture. with the solids.
The partition coefficienl quantifies the tendency of the conlaminan! lo associate As in Fig. 40.2, solids are introduced lo a well-mixed lake via inflowillg tribu-
with solid matler,
taries and are purged from the system by settling and outflow. A solids balance can
Fd + Fp = 1 (40.6) be wriuen as
dm
In olher words the sum of the Iwo fractions is equallo l. V-
dt
= Qm'In - Q m - vs Am (40.7)
TJ1erelationship. of Eqs. 40.4 and 40.5 to suspended solids and the partition
úl('fficient is iIIustrated by Fig. 40.1. Suspended solids range from approximately whue min = inflowsuspendedsolids concentralion(g m-3) and Vs sctlling
I t m-3 for cJear lakes to 100 g m-3 for turbid rivers. Partition coefficients range velocily (m yr-I).
(mili about 10-4 to 10 m3 g-l. Therefore the dimensionless parameter group Kdm At sleady-slale, Eq. 40.7 can be solved for
would range from approximately 0.0001 lo 1000. As Kdm increases, an increas- i
ingly higher fraction of the toxicant will associate with solid matler. Thus for weakly m= Q/IIin
(40.0)
sorbing contaminanls (Iow Kd) in systems with low suspended solids (Iow m), Ihe Q + vsA
contaminant will be predominanlly in dissolved formoFor strong sorbers in turbid Altemalively the result can be expressed as the transfer coefficient
systems, the confaminant will be strongly associated with suspended solids.
Q
(40.9)
{3 = Q + vsA
4JU 'T9~tC§ rrlQPI:t- FQB p, C$TR
Because of the long time horizons exhibited by many toxic suhslunces, Ihe
?' ~W $M Wr t¡itve~way to repfe.~~ntsolid-liquid partitioning, we can use tflisknowl- steady-state assumption invoked in deriving Eq. 40.8 has conventionally been
~d~~fq{tevelop a simple mass ~¡dance of a toxicant for a well-mixed lake or eSTRo adopted. It should be noted, however, that this assumption may nol be accurale
702 PARTVII Toxics lECI1JRE40 Introductiol1 to Toxic-Substance Modeling 7M
....

for certain systems and substances. For example some lakes have highly variable ineludes a more mechanistic representation of settling and fue inc1usidn óf a
solid budgets. In addition highly reactive toxics would require that a subannual time volatilization loss. Together with Eqs. 40.4 and 40.5, the toxicant moclel iJrovideS
scale be adopted. For such scales a steady-state solids budget could represent an a way to determine how the mechanisms of flushing, sorption, volatilizatiort, and
unrealistic assumption. decay act to purge a toxicant from the well-mixed lake.

~ EXAMPLE 40.1. LAKE HURON PCD (WITHOUT SEDIMENTS). Mot/>ho-

40.3.2 Contamlnant Budget metric and hydrologic information for Lake Huron (Fig. 40.4) are summarized in Table
40.2. Other relevant data needed for the computation are
A mass balance for a toxic substance in a well-mixed lake is depicted in Fig. 40.3.
Notice that the toxic substance is partitioned into dissolved and particulate fractions
Parameter Symbol Value Unlts
and that volatilization and settling act selectively on these components. The mass
balance for Fig. 40.3 can be expressed mathematically as Suspended solids concentration m 0.531 g m-J
Suspended solids foad W" 6.15 X 10'2 g yr-I
dc 5.37 X 10'2
PCB load W, p.g yr-'
V dt = QCin - Qc - kVc - vvAFdC- v.AFpc (40.10)
where Vv = a volatilization mass-transfer coefficient (m yr-I). Notice how we have
If sediment feedback is neglected, determine the level of high-molecular-weight
utilized the parameters Fp and Fd to modify the mechanisms that act selectively on
PCBs in the lake. Such loxicants have the following characteristics: Kd = 0.0301 m3 g-I
the particulate and dissolved fractions, respectively.
At steady-state, Eq. 40. 10 can be solved for
Qc.
C=

~
In (40.11)
Q + kV + vvAFd + v.,AFp
The transfer coefficient can also be seen to be

(3 = Q (40.12)
Q + kV + vvAFd + v.AFp
These results are structurally identical to the solutions for conventional pollu-
tants. However, the toxicant model is certainly more mechanistic. For example it

Volatlllzatlon
Loadlng Outflow

FIGURE 40.4
lake Huron.

:II~

TABLE 40.2
Parameters for Lake Huron
-
Parameler Symbol Value Unlts

Volume V 3515 x lO" m'

Settling 11I. m2
Surface area A 59.570 x 10"
FIGURE 40.3 Dcpth H 59 m
Schematic 01 a toxic subslance budgel lor a well-mixed lake wilh no Outnow Q 161x lO" m' yr-I
sediment leedback 01 solids.
 ." .---

704 PARTVII Toxics LECfURE40 Inlroduclion lo Toxic-Subslance Modeling 705

which can be used 10 compule

and v. = 178.45 m yr-I. In panicu1ar delermine (o) Ihe senling velocily 01'solids,
(b) Ihe intlow concenlralion 01'conlaminanl, (c) Ihe transfer coefficienl and Ihe steady-
e = 0.149(33.35) = 0.497 ¡.tg In
stale concentration, and (ti) Ihe 95% response time.
Solution: (o) At sleady-state, Eq. 40.7 can be rearranged lo compute a senling velocity Thus, purging mechanisms allenuale Ihe inflow conú 1111 al ion by nearly two orders
for solids, 01'magnilude. The primary reason for Ihis is the high level 01'volalilizalion, as is evident
from inspection 01' Ihe Iransfer coefficienl. It is al so clearly shown in Fig. 40.5, where
= Qmi. - Qm = 6.15 x 1012 - [161 x 109(0.531)] = 19182 -1 /" vo1alilizalion accounls for approximalely 97% 01'Ihe PCB budgel.
V Am 59,570 x 106(0.531) . m yr (ti) The 95% response lime for Lake Huron can be eslimaled by
This value is equiva1enllo 0.526 m d-I, which is at Ihe 10wer end 01' the range 01' sel-
tiing ve10cilies reponed for fine-grained solids (Thomann and Mueller 1987, O'Connor
/95 = 3V = 3(3515 X 1()9) =fo9;y~
Q + kV + v,.FdA + v,F"A 10.8 x 1012 .;;
1988). One reason for Ihe low va1ue mighl be Ihat Ihis compulation does nol consider Ihe
fact Ihal resuspension might be occurring'concurrentiy wilh seuling. This mechanism is According to our mode11he lake would lake jusl under a year lo show 95% recovery
inc1uded when sediment-waler inleraclions are incorporaled inlo our framework 1aler in following a PCB load reduclion. As with Ihe steady-slale prcdiclion, Ihe magnilude 01'
this lecture. Ihe response lime is slrongly diclaled by volalilizalion.
(b) The values for Lake Huron PCB loading and tlow rale can be used lo calculale Ihe
inflow concentration,
W 5.37 X 1012
40.4 TOXICS MODEL FOR A CSTR WITH SEDIMENTS
ei. = Q = 161 X 109 = 33.35 ¡.tg m-3
Now that we have contrasted conventional pollutants and toxics for a simple eSTR,
(e) Equalions 40.4 and 40.5 can be used lo calculale
we incorporate another level 01'complexity into our mode!. This is done by simply
1
adding a sediment layer below the lake. As in Fig. 41).6,Ihe system is idealized as
Fd = 1 + 0.0301(0.531) = 0.9843 a well-mixed surface-water layer underlain by a well-mixed sediment layer. In Ihe
and Fp = 1- Fd = 1- 0.9843 = 0.0157 following discussion we designate these layers by the subscripls I and 2, respec-
tively.
Equalion 40.12 can Ihen be employed to eSlimale the Iransfer coefficient
Before developing our model, we should recall from Lec. 17thal special param-
161 x 109
eters are required to characlerize a sedimenllayer. These are Ihe porosilY,
(3 = 161 X 109 + [178.45(0.9843)59.570 x 106]+ [191.82(0.0157)59.570 x 106]
= 0.149(= 14.9%) Al cP = ~22 (40.13)

where Vd = volume 01' Ihe liquid parl 01'Ihe scdiment laycr (m,1) and V2 = lotal
5.20 X 1012 (96.8%) volume 01'Ihe sediment layer (m3), and the densilY,
5.37 x 1012 (100%) . 0.08 X 1012 (1.5%)
M2
.. .. p = V1'2 (40.14)

Ji where p = density (g m-3)

>:1 M2 = mass 01'Ihe solid phase in Ihe sedimenls (g)
V,,2 = volume 01'Ihe solid matter (m-3)

0.0 X 1012 (0%)

Water
(1)

0.09 X 1012 (1.7%) FIGURE 40.6

J:I A well-mixed lake underlain by a well-mixed
fiGURE 40.5
I Sed~~.I. I sediment layer.
Budget 01 PCB lor Lake Huron without sediment leedback.
I
706 PARTVII Toxics li!crtJRj¡
4b IntrodliéUoBIIf
1bxld¡¡ti~tllit!Modetlhi -1tff

As in Lec. t 7, these quantities can be used to express a "suspended solids" for which can be substittited ltito !he steady-stat~ vétSlblt or &tI .una dlrd iliete§dn
the sediments as solved for
m2 = (t - (/1}p (40.15) nll
J,¿ntln
m= (40.21)
Thus the sediment solids concentration is expressed in terms of parameters that are Q + v,A(I - Fr)
conventiollally used to measure porous media. We now use these parameters to de-
or expressing Eq. 40.21 as a transfer coefficient,
velop a solids budget for a sediment-water system.
Q
(40.12)
13 = Q + v,A(I _
40.4.1 Sollds Budget
where Fr = a resuspension factor that is defined as
Mass balances for solids in the water and the sediment layer (Fig. 40.7) can be wrlt- Vr 1
ten as Fr = - (40.23)
Vr+ Vb
dml
(40.16) Note that Eq. 40.21 is very similar to Eq. 40.8. The difference between the two
VI(j( = Qmin - Qml - v,Am¡ + vrAm2
formulations is isolated in the dimensionless parameter group Fr. This group rep-
dm2 resents the balance between the resuspension rate and the total tate at wh1ch llie
and (40.17)
V2(j( = v,Aml - vrAm2 - VbAm2 sediment purges itself of solids (that is, both burlal and resuspension). Thus it buriM
-
dominates resuspension (Vb » vr), Fr Oand Eq. 40.21reducesto Eq. 40.8. fa
where Vr = resuspension velocity (m yr-I) and Vb = burlaJ velocity (m yr-I).
Equation 40.15 can be used to express sediment suspended solids, m2, in terms
contrast if resuspension dominates burial (Vb « Vr), Fr -
t andEq.40.21 teduces
to m = minoIn other words when resuspension is relatively dominant, the water con-
of sediment porosity and density. In addition the sub~crl~t~f m I can be drop~d. The centration will approach the inftow concentration because everything that settles is
resulting solid balance equations are immediately resuspended.
dm The above solutions are in the simulation mode, where all the parameters are
(40.18)
VI dt = Qmin - Qm - v"Am + vrA(1 - rP)p known. Although the solids model can be used in this way, it is more conventionaJ
for the model to be employed to estimate some of the parameters.
and d(l - rP)
V2p J. = v"Am- vrA(I - rP)p- vbA(I - rP)p (40.19)
Parameter estimation. The parameters in the model are p, rP,m, min, Q, A,
v.. v" and Vb.For the steady-state case, Eqs. 40.18 and 40.19 represent a pair of
At steady-state, Eq. 40.19 can be sol ved for
simultaneous equations. Thus given seven of the parameters, they generally provide
~m us with a means for estimating the remaining two. Although an algorithm can be
(40.20)
(1 - rP)p = Vr + Vb developedforthisgeneralproblem,we takea differenttack-we trytoassesswhich
of theparametersare leastlikelyto be available.Then we showhowthe modelcan
be employed to estimate these two.
Loadlng Outflow Of the nine parameters, we assume that p and rPare known. Typical vaJues are
p = 2.6 X (06 g m-3 and rP = 0.75 to 0.95. We also assume that the ftow and area,
Q and A, are given.
/ Therefore we are left with five unknown parameters: m, min, v.. Vr, and Vb.
Amongthese,one is extremelydifficultto measure:Vr. This will be the focusof
our parameter estimation. There will be two situations that usuaJlyoccur.
Resuspenslon First, there is the case where m and mlnhave been measured. in addiHon the
settling velocity v, may have been measured directIy or estimated from literature
values. For example a value of 2.5 m d-I represents a typical value for organic
and cIay particles (O'Connor 1988). Ranges of settling velocities for organic and
FIGURE 40.7
inorganic particles are also listed in Table 17.3.At steady-state, Eqs. 40.18 and 40.19
Schematic 01 a solids budget
lor a well-mixed lake with can be added to give
sediment leedback.
O= Qm¡n - Qm - vbA(1 - rP)p (40.24)
...,~<I6..t.,.¿J-. ' -4¡-' ',_. \. - ..

LEClllRE40 Inlroduclion lo Toxic-Subslance Modcling 709

708 I'ARTVII Toxics

which can be use<! to estimate Vb. where the subscripts 1 and 2 indicate water and sediments, respectively. The fraction
Fd2 represents the ratio of the sediment pore-water concentration to the total concen-
Qmin-m
(40.25) tration of contaminant in the sediments. This quantity is related to more fundamental
Vb = A (1 - cf»p
sediment para!11eters by
Second, the burlal velocity is sometimes measured directly. This is best ac- 1
~owplis\1edIJsingsediment dating techniques. Once Vbhas been approximated. the Fd2 = cf> + Kd2(1 - cf»p (40.29)
resu~~psioq velocity can then be estimated by solving the steady-state version of
w6ere Kd2 = the contaminantpartitioncoefficientin the sediments(m3g-I). Wc
~. 4<}.J9fOf . willlkrive this formula when we discuss sorption in Lec. 41.

v, =
m
VS(1_ cf»p
_vb (40.26) The ditfusive mixing velocity Vdcan be estimated from the empirlcally derived
formula (Di Toro et al. 1981 as quoted in Thomann and Mueller 1987)
Vd = 69.35cf>M-2I3 (40.30) 4"1

1().i.~ fontam,nant Budget

where M = molecular weight of the compound and Vdhas units of m yr
~ mas~ \)alance for a toxic substance for the sediment-water system is depicted in
Steady-state solution. At steady-state (dcl/dt = dC2/dt = O),Eqs. 40.27 aud
ri~. 40.§. ~otice that the toxic substance is partitioned into dissolved and particulate
fraftípps in ~th the water and the sediments. Mechanisms such as burial and resus- 40.28 can be expressed as a system of two algebraic equations with two unknowns,
pension act .op,~oth components whereas diffusion acts selectively on the dissolved CI and C2:
fr!l~fip.n.The mass balance for Fig. 40.8 can be expressed mathematically as (Q + kl VI + vvAFdl + vsAFpl + vdAFdl)CI
dc¡ -(v,A+v,¡AFd2)C2 = QCin (40.31)
V¡ di = QCin - Qc¡ - k¡ V¡c¡ - vvAFdlCI - vsAFplCI and -(vsAF¡'1 + vdAFdl)CI + (k2V2 + v,A + VbA+ vdAFd2)C2= O (40.32)
" + V,AC2 + VdA(Fd2C2 - FdICI) (40.27)
Equation 40.32 can be solved for
dC2
a~d vsFI' I + VdFdl
(
V2d1 = -k2V2C2 + vsAFp¡cl - V,AC2 - VbAc2 C2 =
k2H2 + V, + Vb+ VdFd2
CI (40.33) .

+ vdA(FdIC¡ - Fd2CÚ (40.28) \

This result can be substituted into Eq. 40.31, which can be sol ved for

Volatlllzatlon QCin
CI = (40.34)
I! Loadlng
.
Outflow
. Q + k. V¡ + vvAFdl + (1 - F;)(vsFpl + vdF,II)A
where F; is the ratio of sediment feedback to total sediment purging,

F' = V, + VdFd2 (40.35)

, V, + Vb+ VdFd2+ k2H2
Thus the transfer coefficient for this case is

{3 = Q (40.36)
Q + kl VI + vvAFdl + (1 - F;)(vsF/11+ vdF,II)A
The effect of sediment-water interactions can be seen clearly by comparing Eq.
Dlffu,lol\ 40.34 with Eq. 40.11. Notice that the settling velocity has now been supplcmehted by I1
diffusion into the sediments. Also this total movement into the sediments i:-.modifico ¡
il :.! by the dimensionless number F;. If sediment feedback is negligible (that i". !'; - O), J
t'lu ¡I' ¡, .!. l'¡;- 1 Ihe!l Eq. 40.36 approaches pq. 40.12. If sediment feedback is very high (lhat is,
QjiT'lfi í .¡ ''; J
r;- I~
&111>,.&1
*
flGUBI;iCJ.!f . 40.36wouldapproach Q
f?chel]1ati<;01 a loxical)l budget for a well-mixed lake with sediment
feedback. I~ " I {3= Q + k IVI + VvAFdi (40.37)

.
 710 PARTVII Toxics LECJ1JRE40 Introduction to Toxic-Substance Modeling 711

In other words attenuation would be solely dictated by ftushing, reactions, and

\ A21CIO - (A¡ - AII)C20
(40.49)
volatilization because all settling losses would instantaneously be reintroduced into C2¡ = As - A¡

of"
the overlying water.
-(As - AII )C20 + A21clO
(40.50)
Time-variable solution. For the case where all loadings are terminated, the C2s = As - A¡
contaminant mass balances can be divided by volume and rearranged to give
It can be proven that for our model, A¡ will always be greater than As. Consequently
dCI A¡ and As are referred to as the "fast" and "slow" eigenvalues, respectively. This
dt = - AII CI + AI2C2 (40.38) refers to the rates at which they approach zero as time progresses.
dC2
dt = A21cl - A22C2 (40.39) * EXAMPLR14D.2) LAKE
level of hig~ular-weight
HURON PCO (WITH SEDlMENTS), Detennine the
PCBs in Lake Huron with sediment-water interactions.
where A11 = the parameter group that reftects how the water segment purges itself, In particular detennine (a) the burial and resuspension velocities of solids, (b) the
t,
transfer coefficient and the steady-state concentration, and (c) the temporal response
following the tennination of allloadings.
AII = ~ + kl + V.~;I + V~;I + V~dl (40.40) , 'iI Pertinent data for the system is corttained in Example 40.1. Assume that the settling
velocity of solids is 2.5 m d-I (912.5 m yr-I). In addition the sediment porosity, density,
where AI2 = the parameter group that reftects how the water segment is affected by and thickness are 0.9, 2.5 x 106g m-3, and 2 x 10-2 m, respectively. Recall from Exam-
the sediments, pIe 40.1 that the inftow concentration of PCB is 33.35 ¡.tg m-3. High-molecular-weight
Vr VdFd2 PCBs have the following characteristics: M = 305.6gmole,Kd = 0.0301g m-3, and
AI2 = -HI + -
HI
(40.41) Vv = 178.45 m yr-I .

Solutlon: (a) An inftow solids cOllcentration can be computed as

where A21 = the parameter group that reftects how the sediment segment is affected
by the water, 6.15 x 1012 _
min= 161xlO9 = 38.2gm3
A21 = vsFpl + VdFdl (40.42)
H2 H2 Equation 40.25 can then be employed to estimate
where A22 the parameter group that reftects how the sediment segment purges 161x 109 38.20- 0.53 _1
itself, Vb = 59.570 x 106(1 _0.9)2.5 x 106 = 0.000407 m yr
A22 = k2 + -Vr + -vb + -VdFd2 (40.43) The resuspension velocity can be determined by Eq. 40.26,
H2 H2 H2
If CI = CIO and C2 = C20,then the general solution can be developed with Laplace
v, = 912.5 (1 _ 0.9~;;'5 x 106 - 0.000407 =...QJ)!ll5.3m yr-I
lransforms as (recall Seco6.4)
(b) Equation 40.29 can be used to compute
CI = CI¡e-A¡1 + clse-A,., (40.44)
1
C2 = C2¡e-A¡1+ c2.,e-A,,1 (40.45) Fd2 = 0.9 + 0.0301(1- 0.9)2.5x 106 = 0.000133
where the A's are eigenvalues that are defined as Equation 40.30 can be employed lo calculale

A¡ -
A" _ (Al) + A22):!: )(AII + A22)2- 4(AIIA22 - A12A21)
2
(40.46)
Vd = 69.35(0.9)305.6-213
-
= 1.376 m yr-I

The values for Lake Huron can lhen be substituted jnlo Eq. 40.35 to compute
and thecoefficientsare )
F' 0.000407 O 192
(A¡ - A22)c1O - AI2C20 b = 0.00153 + 0.000407 + 1.37.6(0.000133) = .
CI¡ (40.47)
A¡ - A., which can be subsliluled into Eq. 40.36 lo calculale
AI2C20 - (As - A22)CIO f3 = 0.0149
CJ., = (40.48)
A¡ - As which in lum can be employed lo determine
; '1
1,.
... . '\ >;:¡,;~ ..,:';>"'!'i-':.~.

712 PARTVII Toxics LECfUKIi 40 Introduction to Toxic-Substance Mod<:ling 713

5.20 X 1012 CI = 0.0149(33.35) = 0.4971 JLg m-3

5.37 X 1012 0.08 x 1012
. . The sediment concentration can be determined with Eq. 40.33,

= 912.5(0.0157)+ 1.3757(0.9843) 04971 = 36873 -3

C2 0.000407+ 0.00153+ 1.3757(0.000133). . JLgm
The results of Ihe budget are depicted in Fig. 40.9.
Dlssolved I t!0rptlon / (c) The general solution is

CI = 0.4625e-3.296J + 0.0345e-O.0981r

0.336 X 1012 and C2 = -113.86e-3.296r + 3801.18e-O.9081r

These equations are used to produce Fig. 40.1Oa.Note Ihat the water experiences a rapid {
0.4255 X 1012 initial drop due primarily lo the high volatilizalion rateo Then the slower eigenva1ue, t---
which is primarily dictated by Ihe sediment response, dominates the latter part of the re-
Partlculate covery. Figure 40.1 Ob shows the recovery along with the recovery for the system trealed
as a single CSTR (Example 40.1).
0.0894 x 1012
FIGURE 40.9
Budget 91 PCB lor Lake Huron wilh sedimenl leedback.
40.5 SUMMARV
In lerms of analytieal solutions, we have now pushed our model framework lo its lim-
10000 its. As should be evident from Example 40.2, Ihe inc1usion of additional spatial seg-
0.5
ments (e.g., more segments) requires eomputers and numerieal solulion teehniques.
1000 However, we hope Ihat the simple models deseribed herein have provided you with
100
-1 0.4

0.3 1\ concentratlon
Water-column
some insight regarding toxie-substanee models. I urge you to implement this sim-
ple framework yourself (on a spreadsheet perhaps) so Ihat you ean experiment and
build greater intuition regarding loxie-substanee modeling. I believe that this is all
.,

01 01 important prerequisite for using more eomplieated eomputer paekages developed by

,g
o.:
-:1 0.2 T \ _WIIhoedlmentl others.
1°1
0.01
- ---- 0.11. \ / Whhoul oedlmenl8

PROBLEMS
0.001
40.1 Thomann and Di Toro (1983) presenled Ihe following dala related to the solids budgct
° 10 20 O 2 4 for the Western Basin of Lake Erie:

Volume = 23 x J09 m2 Area = 3030 x 106 m2

t (yr) t (yr)
Sol id loading = 11.4 x 1012 g yr-I Suspended solids = 20 mg L -1
(a) (b) Flow = 167 X J09 m3 yr-I

They assumed Ihat Ihe solids settle at a rale of 2.5 m/d (912.5 mlyr), and Ihal the sed.
FIGURE 40.10
iments have p = 2.4 g cm-3 and ti> = 0.9. Delermine the burial and resuspension ve-
Temporalresponse01Lake Huron sedimenl-walersyslem lollowing lerminalion 01 locities.
pca loading. (a) Semi-log plol 01Ihe recovery. (b) The response lor Ihe waler column
wil'" and wilnoul sedimel1l-waler interactions. 40.2 Suppose Ihat a toxic substance that is subject to volatilization (vv = 100 m yr-I) is
discharged to Lake Huron withan inflowconcentrationof 100 JLgL-l. I¿ the absence of -\
sediment feedback, determine the concentration for three cases: (a) weak-sorber (K,¡ =
0.002m3 g-I). (b) modl'fate sorber (Kd = 0.1),and (c) strongsorber(Kd = 2).Other
necessary informalion should be taken from Examples 40.1 and 40.2.
714 PART VII Toxics
~

40.3 A substance (Kd = 0.02 m3 g-l; M = 300) is discharged to a lake (ei.

having the following characteristics:
= 100ILgL-I) LECTURE 41 f
Volume = 1X 106 m3 Mean depth = 5 m
Residence time = I year Suspended solids = \O g m-3 Mass- Transfer Mechahisms:
Settling velocity = 50 m year-I Sediment deposition = 100g m-2 yr-I
Sedil1lent porosity = 0.85 Sediment density = 2.5 g cm-3 Sorption and Volatilization
(o) If sediment resuspension is negligible. compute the steady-slate concentration for
three levels of volatilization:
(i) highly soluble (11" = O)
(i) moderately soluble (11. = 10 m yr-I)
(i) insoluble (11. = 100 m yr-l)
(b) Repeat (o), but inelude the effect of sediment resuspension.

40.4. Repeat Example 40.1 for Lake Michigan. The parameters in Table P40.4 are needed
for the computalion.

TABLE P40.4
Parameters ror Lake Michigan

Parameter Symbol Value Unlts

Suspended solids concentration m 0.5 gm-J
Suspended solids load w., 4 x 1012 g yr I
PCB load W, 10X 1012 g yr-I
Volume V 4.616X 1012 mJ
Surface area A., 57.77X 109 m2
Depth H 82 m
Oulnow Q 50 x 109 mJ yr-I

40.5. Repeat Example 40.2 for Lake Michigan. See Table P40.4 for relevant parameters.

The next two lectures deal with the mechanisms that govem the sinks of toxic sub-
stances in natural waters. In the present lecture, we deal with transport-oriented
processes. First, I describe models of Ihe sorption of toxics on solid matter. This
mechanism is considered Iransport-related because solid-associated tox.ics are sub-
ject to settling losses. Second, I describe volatilization losses across the air-water
interface. Then. in the following lecture, we'lI tum to reaction losses that involve
transformations of the toxicant.
liiI

41.1 SORPTION
Sorption is a process whereby a dissolved substance is transferred lo al1d becomes
associated with solid material. It ineludes bolh the accumulation of dissolved sub-
stances on Ihe surface of solids (adsorption) and the inlerpenetration vr intermin-
gling of substances with solids (absorption). The substance that is sorbed is called

715
 ....

716 PARTVII Toxics LECTURE 41 Mass-Transfer Mechanisl11s: Sorption and Volatili:lation 717

the sorbate and the solid is called the sorbent. Desorption is the process whereby a
sorbed substance is released from a partiele.
For neutral organics, several mechanisms underlie the sorption process. These Cd
inelude (Schwarzenbach et al. 1993): (mg m-3) / Equlllbrlum cd

l. Hydrophobic effects cause the sorbate to associate with organic maUer in the par-
ticulate phase because of an unfavorable free-energy cost of staying in solution.
v "-
, (mgg-l) Equlllbrlum v
2. Weak surface interactions via van der Waals, dipole-dipole, induced dipole, and
other weak intermolecular torces.
3. Surface reactions where the sorbate actually bonds to the solid. < 1 hr Time
i;
.;1,'
For charged toxins, the additional mechanism of ion exchange can occur. It (e)
should also be mentioned that, aside from organically rich materials, neutral organic
FIGURE 41.1
chemicals are also sorbed by solids with liule or no organic content. For such cases,
Experiment to determine sorption. (a) Solids and contaminant are mixed
the sorbent consists of inorganic maUer such as elays. Such inorganic solid sorption in a vessel. A sample is taken and centriluged to separate the solids IrOlI1
is usually significant only when the organic carbon content of the solids is quite low the water. (b) The contaminant content 01 each phase is measured. The
(Schwarzenbachand Westall1981). . temporal progression 01 such an experiment is depicted in (e). An equilibriul11
In the following sections I first introduce some simple sorption models that are is typically reached rapidly.
used to describe the process in standard water-quality models. l then demonstrate
how these models can be employed to derive formulas to predict the fraction of toxi- solid concentration increases with dissolved concentration until the available siles
cant in dissolved and particulate formoFinally I review methods for estimating sorp- begin lo become saturated. At this point the curve levels off to a maximum value
tion parameters for particular toxicants. V¡n(mg m-3).
Several models have been proposed to malhemalically represenl Ihe isolherm.
Among the most popular are the
41.1.1 Isotherms '!l.
l. Langmuir isotherm
An experiment to measure sorption starts by mixing a mass of solids in a well-stirred vmbc"
vessel with a quantity of dissolved chemical (Fig. 4I.1a). At intervals, samples are (41.1)
taken from the vessel, centrifuged, and the amount of chemical on the solids and
v = I + bc"

in solution measured (Fig. 4I.1b). For most chemicals, an equilibrium is rapidly 2. Freundlich isotherm
reached in minutes to hours (Fig. 41.1c). The experiment can be repeated for differ- v = Kfc,~/1I (41.2)
ent contaminant levels and the results ploued. 3. BET isotherm
Figure 41.2 shows the type of plot that typically results from such an experi-
v= vmBc"
ment. The concentration of contaminant on the solids, v (mg g-I), is graphed versus (41.3)
the dissolved concentration, Cd(mg m-3). The plot, called an isotherm, indicates that (cs - cd)[1 + (B - 1)(Cd/c's)]
where b, B, cS, Kf' and n are coefficients used to calibrate the curves to Ihe data.
If'
Note that althoughthe shape depicted in Fig. 41.2 occurscommonly,olher pat-
Contamlnant soUds terns of sorption can occur for specific cases. For example Schwarzenl. l. el al.
(sorbate) (sorbant) Cd= mg m-J
(1993) describe situations where previous sorption modifies the surface and leads to
v ..
.8.. .. . ·. .
. .
.
v = mg g_1
(a) FIGURE 41.2
(b) Cd Sorption data and isotherm.
FIGURE 41.1
718 PARTVII Toxics I.ECTURE41 Mass-Transfer Mechanisms: Sorption and Volatilization 719

enhanced sorption. In such instances the isotherm could show an increasing slope v
with increasing dissolved concentration. The exponent in the Freundlich isotherm
allows increasing (exponent > 1),decreasing « 1), and constant slopes (= 1) to be
t v.
modeled. The BET isotherm, as well as other models, are capable of mathematically
capturing such effects (Weber and DiGiano 1996).
Because of its mathematical simplicity, the Langmuir isotherm has utility in
understanding the shape of the saturating isotherm in Fig. 41.2. It can be derived by
recognizing that the sorption process represents an equilibrium between the rate of
adsorption and desorption, kd.lko4 Cd
R"d = Rd. (41.4) (a)
C «kd.
wherethe massadsorbedper unittime,Rad(mgs-1), canbe representedas d ko4 Cd » kd.
ko4
Rad = kadM..Cd( Vm - v) (41.5) v~ ko4v.. Cd = KdCd 1/-+ v",
kd.
in whichk"d = mass-specific volumetric rate of adsorption [m3(mg s)-I ] and M.. =
mass of solids (g).
The mass desorbed per unit time, Rd. (mg s-I), can be represented as
Rd. = kd.M..v (41.6) v

whcre kd. = a first-order desorption rate (S-I).

Equations4 \.5 and 4\.6 can be substitutedintoEq. 4\.4 and the resultsolved
for ------------------------.
V
Kd= C;;
v = vmcd (41.7)
(kd.lkad) + cd o
FIGURE 41.3
o Cd
which is the general form ofthe Langmuir model (compare with Eq. 41.1). (a) A Langmuir isotherm indicatlng asymptotes.
(b) (b) Linearization 01 isotherm.
Equation 41.7 is plotted in Fig. 41.3a. Note that the curve manifests two asymp-
tatic regions. For relatively high levels of contaminant, the curve becomes zero-order
(Ir constant. For low levels of contaminant, the curve is first-order or linear. It is this
latter case that is employed in most toxic substance models because of the assump- 41.1.2 Fractlon Sorbed
tion that in most natural waters, toxicant concentrations are relatively low. Using the framework developed above, we can now derive expressions relating the
The linear region can be modeled by realizing that when v « vm, plenty of
sites are available and
contaminant fraction in dissolved (Fd) and particulate (Fp) forms to the partition
coefficient. To do this, individual mass balances can be written for each fraction in
Rad = k"dM..Cd(Vm- v) -+ k~dM..Cd (41.8) a eSTRo For example for the dissolved portion of the contaminant,
where k~d = kadVm[m3(g S)-I]. In other words the rate of adsorption becomes inde- d~ , (41.10)
pendent of solid concentration and is driven only by dissolved concentration. When Vd( = QCd.in - QCd + kd.Msv - kadMsCd
Eq. 41.8 is substituted along with Eq. 41.6 into Eq. 41.4, the result is
and for the particulate portion,
v = KdCd (41.9)
where Kd = a partition coefficient = k~,¡Ikd. (m3g-I). As in Fig. 41.3b, the partition (41.11)
vd:: = QCp.in- Qcp - v..Acp - kd.Msv + k~dMsCd
coefficient rcpresents the slope of the linear portion of the isotherm.
The linear isotherm implies that if the dissolved concentration is increased or where the subscripts d and p signify the dissolved and particulate contaminant, re-
decreased a certain proportion, the solid concentration wil\ manifest the same pro- spectively. To keep this derivation simple, we have excluded decay reactions from
portional increase or decrease. As discussed next, this means that the fraction of the these balanc'es. AIso, recognize that all time units are now expressed as years to be
contaminant in the dissolved and particulate form remains constant. compatible with the time scale of interest for systems such as lakes.
 720 PARTVII Toxies LECTURE41 Mass-Transfer Meehanisms: Sorption and Volatilization 721

Because sorption kinetics are so much faster than the other input-output terms for by using porosity in the derivation. As with Ihe water, Ihe 10lal eoneentration is
in the model, Eqs. 41.10 and 41.11 can be summed (recall our general discussion of divided into partieulate and dissolved eomponents (Eq. 41.13), where the dissolved
this "local equilibrium" assumption in Seco38.1) to yield eoneentration is now defined as

dc Cd = 4>Cdl' (41.21)
V dt = QCin - Qc - vsAcp (41.12) in which Cdp = eoneentration of the pore water (mg m-3). Similarly the particulate
where eoneentralion is defined as

C = cd + cp (41.13) Cp = (1 - 4»pv (41.22)

and cin = Cd.;n+ Cp,in (41.14) For the sediments, Eq. 41.9 is
Notice that the adsorption and desorption terms have been canceled by summing v = Kd2Cdl' (41.23)
the two equations. In addition, note that the right side of Eq. 41.12 is not exelusively
expressed in terms of the dependent variable c. That is, it also ineludes the variable
Substituting Eqs. 41.21 Ihrough 41.23 inlo 41.13 yields e'
= 4>Cdl'+ (1 - 4»PK./2C,/1' (41.24)
cp. Const:'I'I~ntlyit canoot be solved without additional infOlmation.To express the C

equation exelusively in terms of the unknown, recognize that particulate concen- whieh can be solved for
tration is related to suspended solids and mass-specific contaminant concentration
by
_ _C'/I' I
F'12- - C - 'I'+P,/2-
A: K (1 4» (41.25)
Cp = mil (41.15) Thus we have derived the fraetion Ihat was inlroduced previously as Eq. 40.29.
where m = suspendedsolidsconcentration(g m-3).
Equations41.9and41.15canbe combinedwith41.13toyield
C = Cd + mKdcd 41.1.3 Estimation 01 the Partition Coefficient
which can be sol ved for
A variety of investigators have attempted to relate partitioning to specific properties
Cd = FdC of the chemical. Karickhoff et al. (1979) have developed the following scheme for
where neutral (that is, noncharged) organic chemicals.
- \O" The partition coefficient for organic contaminants, Kd (m3 g-I), is assumed to
Cd 1 be a function of the organic-carbon conten! of the solids, as in
Fd = - = -
C 1 + Kdm
Kd = j;".K"c (41.26)
This result can in tum be substituted into Eq. 41.13 and the result sol ved for
where Koc = organic-carbon partition coeflicient «mg gC-I)(mg m-3)-'1 and
Cp = Fpc j;,c = weight fraction of the total carbon in the solid matter (gC g-I).
where The organic-earbon partition coefficient KOl"in turn, can be estimated in terms of
_ cp _ Kdm the contaminant's octanol-water partition coefficient Km"[(mg m;;-c~an()I)(mg m;;;~I<r)]
Fp - -C - 1 + Kdm (41.20) by (Karickhoff et al. 1979)
Koc = 6.17 X 10-7 Km" (41.27)
Thus we have derived the fractions that were introduced previously in Eqs. 40.4 and
40.5.
Equations 41.26 and 41.27 can then be combined to yield
Similar values can be derived for the sediments. As described in Box 41.1, these
relationships differ from Eqs. 41.18 and 41.20 because of the significant fraction of Kd = 6.17 X 10-7j;,cKow (41.28)
me sediment volume that is solid.
Thomann and Mueller (1987) suggest a range of 0.00 1 to 0.1 for fOl" How-
ever, it should be noted that the top end should be raised to account for autochtho-
nous partieles, such as living phytoplankton cells, that have an organie carbon of
ftPI41.1. Contamlnant Fractlon. for the Sedimenta about 0.4.
ib~ d~rivation of.~ofbed fraetio~s for the sedi~e?ts is ~omewhat e?mplie~ted beeause
The octanol-water partition coefficient can be obtained in a variety of ways.
~ ~!~mdieantpo!1lon of the sedlment volume IS IDsolld formo Thls faet ISaeeounted First, it can be measured directly. Second, it can be obtained from published tables.
 722 PART VII Toxics LEcruRE 41 Mass-Transfer Mechanisms: Sorption and Volatilization 723

PCB t-e=I:J---I Otherformulasare availábleforestimatingKdatirlkoc fromkbw andSwlSticlt

regression approaches improve as they are applied to smaller groupings básedbd
Phthalateesters f---I I H
similar chemical structure. Schwarzenbach et al. (1993) present an exceIleht arid
comprehensive review of the various approaches.
la endosuHan' ~ If ¡oe is assumed to be approximately 0.05, E<j.4 í .28 can be used to estimate
!f;:'~endosuHan~
'1r ~ Endosulfansullate
PAHs
Pesticides
the partilioncoefficient,as in
Kd = 3.085x 1O-8Kow (41.31)
x hexachlorobenzene
MAHs l-Q--1 This estimate along with suspended solids concentration and tabulated Kow's can be
x pentachlorophenol
substituted into Eqs. 41.18 and 41.20 to estimate the dissolved and particulate frac-
Phenols I~ tions. The result is displayed in Fig. 41.4. Note that we have also included box-and~
Hexachloroethane whisker plots for some major classes of toxicants (recall Table 40.1). As describe<!
Nilrosamines~
Halogenated~
aliphatics '1..U'
¡¡=
x xx
,~
~
Hexachlorobutadiene
Hexachlorocyclopentadiene
in Fig. 41.5, box-and-whisker plots (Tukey 1977, McGiII et al. 1978) utilize robust
statistics to summarize the pertinent features of a data sample. In particular they
provide a one-dimensional representation of a data distribution.
Ethers ~ I t---j Inspection of Fig. 41.4 yields some interesting conclusions. In particular, five of
the groups (ethers, halogenated aliphatics, nitrosamines, phenols, and monocyclic
1.00
-
aromatics) are overwhelmingly in dissolved form (that is, Fd 1).In faet if a value
of Kow < 104(which corresponds to Kd < 3.2 X 10-4 m g-t) is conservatively chó:.
~,
0.80 ,

sen as the criterion for assuming that the contaminant is predominantly dissolvecl,
Fp 0.60 about 65% of the 114priority toxicants (Table 40.1) would be classified as dissolved.
0.40
" Thlckness"of box Is
0.20 proportlonal to
1 ' 0.00
JNumber ofvalues Insample
-2 O 2 4 6 8 10
We~k sorber log (K"w) Strong sorber
FIGURE 41.4 _ X
Plot 01 Iraction dissolved, F., and Iraction particulate, F., versus partition
coefficienls (Chapra 1991). Each curve represents a particular suspended
solids concentration. Box-and-whisker diagrams (see Fig. 41.5) lor nine classes 01
organic contaminants are drawn above the plol.
1
Lower bound of
t
Median
f Outsider
Upper bound of
range of data range of data
(excludlng outslders) (excludlng outslders)
Third, it can be computedfromformulasbasedon morereadilyavailablecharac- 1
teristics of the contaminant. For example Chiou et al. (1977) offer the following Lower quartlle Upperquartlle
formula: bound bound

log Kow = 5.00 - 0.670 log S:v (41.29j FIGURE 41.5

Tukey's (1977) box-and-whisker plot uses robust slatistics such as tlle
whereS~ e: solubility in ¡LmoleL-1, which can be computed by median and Ihe inlerquartile range lo summarize Ihe dislributlon 01a
dala sel. In particular it provides a visual depiclion ollocallon, spread,
S' = Sw X 103 (41.30) skewness, taillenglh, and outlying dala poinls (Hoaglin el al. 1983). As
w M in Ihe above diagram, an "outsider" relers lo values that are lar outside
the normal range. Addilional inlormation on box-and-whisker plots al1d
where Sw = solubility (mg L-1) and M = molecular weight (g mole-I). Published other inleresling aspects 01 robust and exploralory data analysis cal1
tables (e.g., Lyman et al. 1982, Milis et al. 1982, Schnoor et al. 1987, etc.) can be be lound elsewhere (Tukey 1977, Hoaglin el al. 1983, Reckhow and
ltsed to oblain solubility. Chapra 1983, ete.).
 ____ : : -.a....

724 PARTVII Toxics LECfURE41 Mass-Transfer Mechanisms: Sorplion and Volalilizalion 725

For such pollutants the analysis becomes greatly simplified because sediment-water I "Suspended
interactions can be neglected and purging becomes solely a function of ftushing, F--- soUds
volatilization, and decomposition. For example Eq. 40.36 reduces to d - 1 +' Kd' m basla"

Q ,-L,
{3 = Q + k IV I + VvAF dI (41.32) K.cfocm
' '
Such simplifications will be explored in further detail at the end of this lecture. ,-L,
KocPOC

41.1.4 Sorptlon Revlsited (The "Third-Phase" Effect) X

IO-6K.O'POC
In Lec. 17,we noted that suspended solids can be categorized as allochthonous (orig-
inating in the drainage basin) 01'autochthonous (prou", ~din the water via photosyn-
thesis). To this point our sorption model would seem to be more compatible with
. 1 "Orgenlc
FIGURE 41.7
Depiction 01 how an organic-carbon basis has
supplanted the original 'suspended solids" basis
Fd carbon employed in early water-quality models 01 toxic
the former perspective. That is, we characterize solids as a loading and use partition 1 + 10-6 K.O' POC besls" organic sorption.
coefficients to attach some of the toxicant lo ¡he suspended partic1es.
At least for toxic organic compounds, this viewpoint must be broadened. In f--
the previous section, we have seen that the partition coefficient is a function of the
partic1e's organic carbon conten!. Therefore, rather than modeling suspended solids We have also seen Ihat the partition coefficient can be related to the octanol-waler
concentration, a more direct approach might be to model particulate organic carbon parlition coefficienl by
(POC). This would be particularly appropriate for systems with large quantities of Kd = 6.17 X 10-7 f"cKmv (41.34)
autochthonous solids. Further, an organic carbon perspective raises the issue that
productive systems also tend to have significant concentrations of dissolved (and Thomann and Mueller ( 1987) and olhers have noted thal this relationship can
colloidal) organic carbon (DOC). Because toxics would tend to associate with both be approximated by
dissolved and particulate forms, the DOC would represent a "third phase" in addition
to the particulate and dissolved phases (Hg. 41.6). Kd = (l x 1O-6K"w)f"c (41.35)
To quantify the effect of this third phase recall that the dissolved fraction is ~
modeled as (Eq. 41.18)
. ",,- Substituting this approximation into Eq. 41.33 yields
1;-'" 1 1
(41.33) Fd = (41.36)
Fd = I+Kdm l + (1 x 1O-6K"w)f"cm

01'recognizing that POC = f"cm,

l
(41.37)
e
Fd = 1 + (l x 1O-6K"w)POC
E
=' The process of converting from a suspended solid to an organic carbon basis is de-
"8 DOC DISSOIVed..... picted in Fig. 41.7.
phase

~
i Now if it is assumed that the toxicant associates similarly with DOC, the tbird
phase can be incorporated and three fractions derived,
!
Fdl = 1
1 + (1 x 10 6K"w)(POCI + DOC¡)
(41.38)

J!I
e
I DOC DISSOIVed..... Fpl = (1 X 1O-6K"w)POCI
1 + (l x 10 6Kow)(POC. + DOC¡)
(41.39)

ti
i phase
FIGURE 41.6
Depiction 01 how a DOC "third phase"
competes with POC 101'dissolved
1'1',,~ toxicar¡t in the (a) water and in the
Fol = (l X 1O-6Kow)DOCI
1 + (l x 10 6Kow)(POCI+ DOC¡)
(41.40)
(b) sediments.
 726 PARTVII Toxics
LECTURE
41 Mass-Transfer Mechanisms: Sorption and Volatilizatl~iI 721

where the subscript 1 denotes the watercolumn and Fd\, Fp(, and Fol = the fractions where Cop .. concentration of the toxic in the pore w~1etDdc(mft m~3,¡1iiA;PhrJ'
associated with dissolved, POC, and DOC, respectively. titioning between the dissolved and DOC forms in tbe pore water would be defined
A similar analysis can be used to deve\op fractions within the sediments (see as
Box 41.2),
1 Ko = coplDOCp (Jt.52)
Cdp
FdlJ2= .L . " " .,,_/ir; ,.)(POC2+ t/JDOCp2) (41.41) Combining Eqs. 41.51 through 41.52,

(1 x 1O-6Kow)pOC2 Co = </JKoCdpDOCp (41.53)

Fp2 = t/J + (1 x 1O-6Kow)(POC2
+ t/JDOCp2) (41.42) or substituting Eq. 41.35,
I
_ (1 X 10-6 Kow)DOCp2 Co = </J(Ix 1Q-6Kow)CdpDOCp (41.54)
I
F"p2 - t/J + (1 x 1O-6Kow)(POC2+ t/JDOCp2) (41.43) Finally Eqs. 41.46, 41.50, and 41.54 can be entered into Eq. 41.45, which can be
solved for
where the subscript 2 denotes the sediments; Fdp(, Fp2' and Fop2 = fraction dis- 1
">1 solved in the pore water, associated with POC, and dissolved in pore-water DOC, (41.55)
F'¡p2 = </J+ (1 x 1O-6Kow)(POC2 + </JDOCp2)
respectively; DOC2p = dissolved organic carbon concentration in the pore water
I (gC m-3); and POC2 = particulate organic carbon on a whole sediment volume ba-
sis (gC m-3). Note that the last quantity is often measured as a dry-weight fraction Further manipulation can be used to determine Eqs. 41.42 and 41.43.
of carbonoTherefore it is often convenient to make the following substitution in the '1
above equations:
What are the implications of the foregoing analysis for model calculations?
POC2 = foc,2(1 - t/J)p (41.44)
They mean that the presence of dissolved organic carbon draws toxicant away from I

where foc.2 = dry-weight fraction of organic carbon in the sediments (gC g-I). the particulate and dissolved pools. Thus mechanisms that influence these pools
would be diminished. For example in the water, higher DOC would detract from the I

volatilization and sedimentation mechanism because less toxicant would be associ-

ated with the dissolved phase and settling partic1es, respective\y, In the sediments
BOX 41.2. Derlvatlon o, Sedlment Contamlnant Fractlons I
higher DOC would enhance the amount of toxicant in the pore waters. Thus sediment I
feedback to the water column would be increased. I
As for the water column, the sediment contaminant can be partitioned into three phases,
In conc1usion, this has been a brief introduction to sorption. We have deliberately
I
C = c,¡ + CI. + Co (41.45)
limited the discussion to the leve\ of detail employed in commonly applied water-
The dissolved fraction can be related to the dissolved contaminant in the pore water by quality models. Further nuances and a more complete description can be found in
{lJ
c,¡ = </JCdp (41.46) references such as Schwarzenbach et al. (1993) and Lyman et al. (1982).

where Cdp = cOllcentration of the pore water (mg m-3). In a similar fashion to the
derivation of Eq. 41.25, particulate concentration is defined as ti.
41.2 VOLATILlZATION
Cp = (1 - </J)pKdCdp (41.47)
Many toxic substancesmove across a lake's surface by volatilization. From our dis-
Substitutillg Eq. 41.35 gives
cussion of oxygen gas transfer in Lec. 20, we are already familiar with how the
cp = (1 - </J)p(1x 1Q-6Kow)focCdp (41.48) process can be quantified. The flux of a gas across the air-water interface can be
modeled as
Now recognizing that

POC = (1 - </J)pfor (41.49) (41.56)

we rewrit{' Eq. 41.48 as
J = vv(~:- CI)

where J = mass flux (mole m-2 yr-I)

Cp = (1 X 1Q-6Kow)POCc,¡p (41.50)
Vv = net transfer velocity across the air-water interface (m yr-I)
Por the DOC,
Pg = partial pressure of gas in the air over the water (atm)
Co = tPcop (41.51) He = Henry's constant (atm m3 mole-I)
Cl = concentration of the gas in the water (mole m-3)
-- - ~~ ~. .....

tJ
728 PARTVII Toxics LEcrURE 41 Mass-Transfer Mechanisms: Sorption and Volalilizalion 729

The transfer velocity can be calculated based on the two-film theory as in the almosphere (g m-3) and ° = concenlration of oxygen in Ihe waler in mass
He units (g m-3). (
Toxics differ from oxygen in Iwo ways. First, we cannol generally conclude Ihal
Yv = K¡ He + RTa(K¡IKg) (41.57)
Ihl'Vare allliquid-film controlled. Second, we can assume Ihal, in many cases, Ihey
where K¡ and Kg = mass-transfer coefficients for the liquid and gaseous films an: not abundanl in Ihe atmosphere. The latter assumption means Ihat the gaseous
(m yr-I), respectively concentration in Eq. 41.56 can be assumed lo be zero. If Ihis is the case, Eq. 41.56
R = universal gas constant [8.206 x 10-5 alm m3 (K mole)-I]
~omes
Ta = absolule lemperalure (K) J = -VvCd (41.62)
Thus, as depicled in Fig. 41.8, Ihe transfer is dependent on a conlaminanl- If Ihis equation is multiplied by lake surface area and the loxic subslance's molecular
specific properly (Henry's conslant) and two environment-specific properlies (the weighl, it can be expressed in Ihe formal of our mass balance as
film-transfer coefficients).
The transfer coefficients can be computed based on more fundamental parame-
de
V dt = -VvAsCd (41.63)
.',
ters with Iwo standard approaches. For quiescent syslems such as lakes, a slagnanl
film approach is employed, The represenlation of Eq. 41.63 differs markedly from Ihal for dissolved oxy-
D¡ Dg
gen (Eq. 41.61). Ralher than approaching an equilibrium allhe saluralion point, ~-
K¡ =- Kg =- (41.58) Eq. 41.63 represents a one-way loss from Ihe syslem. In Ihis sense it is jusl like
Z¡ Zg
a settling mechanism. However, ralher Ihan representing it as a velocily transporl
where D¡ and Dg = diffusioncoefficientsfor the toxicantin the liquid and gas, through the bottom of the lake, we represenl volatilizalion as a velocity IranspOrloul
respectively, and Ihe z's = slagnant film thicknesses. the 10p.
For systems such as rivers, where the liquid flow is lurbulenl, Ihe surface re-
pewal model applies,
41.2.1 Parameter Estimation
K¡ = Jr¡D¡ Kg = JrgDg (41.59)
where the r's are renewal rales. At Ihis point Ihe key lo representing volalilizalion losses hinges on estimaling vv. As
The volatilizalion model can be integraled into a mass-balance by multiplying in Eq. 41.57, Ihis means Ihat we must delermine values for Henry's conslant and the
the flux (Eq. 41.56) by surface area lo give liquid and gas-film-Iransfer coefficienls. ,
I
dc pg Molecular weigbt and Henry's constant. Values of molecular weighl and
e'
V dt = YvAs He - cd ( ) (41.60) Henry'sconslanlhave been delermined for many loxic compounds. Compendiums
of this dala can be found in a variety of sources (Callahan et al. 1979, Mabey el al.
where V = volume of Ihe water being modeled (m3) and As = surface area. Note
that we have changed the subscript for c to denole "dissolved." 1979, Milis el al. 1985, Lyman et al. 1982, Schnoor el al. 1987, Schwarzenbach et
Recall that when Eq. 41.60 was written for dissolved oxygen (Eq. 20.32), the al. 1993, etc.).
result could be simplified because (a) oxygen is abundant in Ihe atmosphere and (b) Of the priority pollutants, molecular weights range from a low value of 50.6 for
oxygen is liquid-film controlled. The result is the halogenated aliphatic chloromelhane to a high of 430 for the pesticide toxaphene.
do The lighlest groupsof compounds« 200) are halogenaledaliphatics,phenols.
monocyclic aromalics, elhers, and nilrosamines. The heaviest are Ihe pesticides
V dt = K¡As(os - o) (41.61)
followed by Ihe phlhalale eslers and PCBs (> 200). The PAHs are inlermediale.
where Os = saturation concentration corresponding to the parlial pressure of oxygen Box-and-whisker diagrams for Henry's conslanl are depicled al Ihe 10p of
Fig. 41.9. )n general, three classes of compounds are highly insoluble: halogenated
aliphatics, monocyclic aromatics, and PCBs. The olher classes h¡¡vesimilar ranges
He of solubility, with some members being relatively soluble lo olhers being moderalely
Vv = K¡ insoluble.
He + RTa(K¡lKg)
FIGURE 41.8 Film-transfer coefficients. The transfer coefficients are convenlionally esli- .,
/
Contamlnant-
speclflc
\
Envlronment-
The volatilization mass-transler velocity is computed
as a lunction 01 contaminant-specilic (Henry's
constant) and environment-specilic (Iilm-Iransler
i"
mated by correlating them with commonly studied transfer processes. The liquid-
film coefficienlis typically correlaled with the dissolved-oxygen-Iransfer coefficienl.
speclflc coefficients) properties. The gas-film coefficienl is correlaled wilh Ihe evaporative Iransporl of water vapor.
730 PARTVII Toxics LECrURE41 Mass-Transfer Mechanisms: Sorption and Volatilizatloll 731
..
As previously noted in Lec. 20 (Eq. 20.58), the liquid-film mass-transfer coef- Analysls ot volatillzatlon. The impáct of vólatilizatiori éáfi be líS~ssed b~
ficient (m yr-I) can be related to the oxygen-transfer coefficient KL by (Milis et al. computing the net mass-transfer coefficientfora range of Henry's ,constánts,.As
1982) in Fig. 41.9, lie ranges in value from approximatety 10-8 to t atm m3lnole-I¡:f:he
results indicate that at high values of the lÍenry's constari\' the nel mass-&áitSter
32,°.25 coefficient approaches the liquid-film coefficient. Thus, as the terminology implies,
K¡ = KI.02 (M J (41.64) "insoluble" gases (high He) are purged rapidly from the water whereas "soluble"
where M = molecular weight of the compound and KI.02 = oxygen-transfer coeffi- gases (low He) stay in solution.
cientoThis transfer coefficient can be estimated using the reaeration formulas from
Lec. 20. EXAMPLE 41.1. KINETIC PARAMETER ESTIMATION FOR PCR. Estimate
For the gas-film rate, Milis et al. (1982) have suggested (a) the sorption coefficient and (b) the volatilization mass-transfer velocity for high-
molecular-weight PCBs (> 250 gmole). Note Ihat parameters for these compounds can
be summarized as
18,°25
KK = 168Uw M (J (41.65)
Toxlcant MW log (K.w) log (H,)
where Uw = windspeed(mps).Thisformulais basedon the gas-filmtransfercoef- - 3.48
Aroclor 1016 257.9 5.58
licient for water vapor. As expressed in Eq. 41.65 the units for Kg are m d-I. Aroclor 1242 266.5 5.29 -242
Aroclor 1248 299.5 5.97 -2.45
Aroclor 1254 328.4 6.14 -2.84 .,,,,\
-.... ...
Halogenated,~ Aroclor 1260 375.7 6.99 -0.60
aliphatics ~ I Mean 305.6 5.99 -2358
MAHsf---Q-i
PCBs ¡-{TI! , Assume that the wind speed is approximately 10 mph (= 4.47 m 8-1) and that the tem-
Aroclor 1260
perature is 10°C (= 283 K).
Ethers ~
Solution: (a) Substitution of the octanol-water partition coefficient into Eq. 41.31 yields
,
Pesticides ~ Toxaphene
Kd = 3.085 X 10-8(105.99) = 0.0301 g m-3
PAHs ~
Remember that Eq. 41.31 is predicated on the assumption that foe = 0.05.
2.4 Diniirophenol 1--[J-; Phenols (b) The liquid-film exchange coefficient is based on the oxygen-transfer coefficient. For
NitrosaminesI C:I=:::J- example, using the Banks-Herrera formula (Eq. 20.46),
Phthalate esters H::I:J--t
KI.02 = 0.728(4.47)050 - 0.317(4.47) + 0.0372(4.47)2 = 0.8655 m d-I
300.00
This value can be 8ubstituted into Eq. 41.64 (with appropriate conversions) to give
",,~~" "-"--"---
32 f'25
200.00 /' 110 MPH 1...___

FIGURE 41.9
K¡ (
= 365 305.6J 0.8655 = 179.7 m yr-t
"v //::-_00 00___.__-- 100
Plot 01 volatilization transler The gas-film exchange coefficient can be computed with Eq. 41.65,
200
400
coefficient Vv versus Henry's

1
18 \0.25
100.00 constant H. (Chapra 1991).
(
0.00
;/
#.~f',
1
t
MW
Each curve represents
a particular molecular
weight. Box-and-whisker
diagrams (see Fig. 41.5)
Kg = 61.320 305.6J 4.47 = 135,038m yr-I
These values along with the Henry's constant can be substituted into Eq. 41.57 to give

-10 -4 -2 O 2 10-2.358 -1
lor nine classes 01 organic Vv = 179.7.~ He. .~ AA'.. '~_"An~".~~ ~,,~~ n~n" = 178.45 m yr
Soluble Insoluble contaminants are drawn
10g(H,.) above the plot.
 ...

732 PARTVII Toxics LEl.'TURE

41 Mass-Transfer Mechanisms: Sorplion and Volalilization 733

41.3 TOXICANT-LOADING CONCEPT I 2. Water sorption is equal to sediment sorption (thut is. Kdl = Kd2). As shown by
Chapra (1991) this means that diffusive sediment-water transfer will either be
In Figs. 41.4 and 41'1 we have developed two r show how some classes of
negligible or result in a loss from the water to the sediments. Consequently the
toxic substances sorl, .,lIdvolatilize. Although th. ,IS provide t1.ht,they suffer exclusion 01'the diffusion mechanism also results in an upper bound prediction.
in that tloeyprovillc ¡IJI¡ncomplete picture of the substance's trunspon and fate in a
natural water. As in the elephant parable at the beginning of this book, the focus on Applying these simplifying assumptions reduces Eq. 41.66 to
the individual mechanisms does not iIIuminate their combined effect. QCin (41.72)
To assess the "big picture," we now integrate our knowledge regarding sorp- /'- CI Q + VI'A =
tion and volatilization with tllt' simple two-Iayer sediment-water model described in
Seco40.4. We call this a "tOxl.:ant-loadingmodel" because it is similar in spirit to where VI' = net toxicant loss rate (m yr-I), which representsall purging mechanisms
the "phosphorus-Ioading plots" developed by Vollenweider to model eutrophication aside from ftushing,
(recall Seco29.1). Details of the following analysis can be found in Chapra (1991). VI' = VvFdl + (1 - F,)vsF¡'1 (41.73)
Recall that our mass balance for the water and sediment layer of a well-mixed
Thomann and Mueller (1987) proposed this approach as a way to separate the
lake yielded the following model of the water column (Eq. 40.34):
lake-specific mechanism 01'flushing from the more toxicant-specific mechanisms. If
QCin desired, Eq. 41.72 can always be used to interpret how panicular values OfVI' would
CI = (41.66)
Q + kl VI + vvAFdl + (1 - F;)(v.Fpl + VdFd¡}A effect water concentration for lakes with different levels of flushing.
I
However, for our present purposes, we focus on VI' itself. As depicted in
whereF; = the ratioofrecyclesedimentpurgingto totalsedimentpurging, Fig. 41.10, the net loss rate provides a means to integrate the effects 01' sorption,
volatilization, and sediment feedback into a single number. If VI' is large, it suggesls
F' = v, + VdFd2 (41.67) that the toxic will be purged from the water at a rapid rateo If it is small, purging
, v, + Vb+ VdFd2+ k2H2
would be weak.
Now a problem with gaining insight into this model is that it depends on 14 rnde- The impact of sorption and volatilization on VI'can be assessed by mapping il
pendent parameters. Therefore we will perform some manipulations and make some on a logarithmic space defined by sorption (as represented by Kdm) and volatiliza-
simplifying assumptions to make the model more manageable. First, we reexpress tion (as represented by He). To do this we assume values of Tu = 283 K, M
the sediment feedback term F; in terms of the resuspension ratio F, (Eq. 40.23). To 200 g gmole-I, Uw = 2.235 m S-I (5 mph), and Vs = 91.25 m year-I. Based 011
do this, we define such an analysis, Cliapra (1991) suggested that the sorptionlvolatilization space can
v, F,Vsb = (41.68) be treated as three distinct regions (Fig. 41.11):
...~. - ;~d Vb = (1 - F,)Vsb (41.69)
. Air zone (insoluble weak sorbers). COlltaminants that are insoluble and sorb
where Vsb = a scaledsettlingvelocity, weaklyalwayshave high removalrates (that is, VI' > 10m year-I). This is be-
m cause they are almost completely in dissolved form where they are subject to
(41.70) strong volatilization.
Vsb = (1 -ifJ)p Vs . Water zone (soluble weak sorbers). At the other extreme, compounds that are solu-
By using these terms, Eq.41.67 becomes ble and sorb weakly have extremely low removal rates (that is, VI' < 10m year-I ).
This is because they are completely in dissolved form yet are not subject lo
F' = F,Vsb + VdFd2 (41.71) volatilization.
, V.b+ VdFd2+ k2H2 . Sediment zone (strong sorbers). These substances are independent 01'vulatiliza-
Thus the model is no longer expressed as a function 01'the interdependent parameters tion because they are not in dissolved formo Consequently their removal depends
v, and Vb,but rather is now formulated in terms 01'the single parameter F,. directly on sediment-water interactions. 11'resuspension is negligible, they exhibil
NexI we make two simplifying assumptions: high removal. If resuspension is high, removal is low.
1. Decayreactionsare negligible(that i.r. kl = k2 = O).Thus we ignoreremoval Sorptlon
/llechanisms such as photolysis, hydrolysis, and biodegradation (to be discussed
in the next lecture) that act to remove the contaminant from the system through
/ ~ FIGURE 41.10
¡
reactions. This is a conservative assumption in that any decay would tend to re- j
I
duce the amount of toxicant in the system. Consequently the exclusion 01'such
mechanisms means that our analysis will yield an upper bound on concentration.
/
vr =VvFdl +(I-F,>ZsFp1
Volatlllzetlon
Sedlment
Intaractlona
The net loss rate provides an integrated measure 01
Ihe eftects 01 sorplion, volalilizalion. and sedimanl
leedback on loxicanl purging.
1
 734 PARTVII Toxics LECTURE41 Mass-Transfer Mechanisms: Sorption and Volatilization 735

Strong
4 4
sorber SEDIMENTZONE
2
Asslmllatlon depends on sedlment recycle
-- o 2
E:
... -2
'-' TCDD
01 -4 WATER ZONE AIR ZONE
O .
.S!
-6
Poor asslmllatlon Strong asslmllatlon
DDD . DDT rI' DDE
-8
W.sk FIGURE 41.11 Endrin. Aldrin l. . Chlordane
sorber -10 Three regions on a -2 g-Hexachlor. B-Hexachlor H
e
-10 -6 -2 2 logarilhmic space delined
Dieldrin_
.~ a-Hexachlor. _ eptachlor.
-1 _ Lindane Toxaphene
by sorption (K.m) versus O>
Soluble
log (H.) Insoluble
volatilization (H.). .2-4 _ . Chlorpyrfiros -
Endrinaldehyde Methylparathionl - Heptachlor.epoxide
Isophoron:'
The implications of the foregoing for the organic priority pollutants (recall Ta-
-6-
ble 40.1) are explored in Fig. 41.12. Here, we locate each of the pollutants on the Acrolein
sorption/volatilization space. For this case we need to specify the suspended solids . r
concentration.Weuse a valueof5 g m- 3 as being representative of mildly eutrophic
lake conditions. The plot indicates that of the 114organic priority pollutants, a sig-
--8 . Endosulfan sulfate ."
A-B-Endosulfan
a-Endosu1Ian
nificant portion would be subject to rapid purging due to volatilization.
l' ve (Chapra 1991) taken the analysis further by showing how each of the or- -10
ganic priority pollutant categories are arranged on the space. For the analysis a value -10 -6 -2 2
of 5 g m-3 was again used for the suspended solids concentration. Although some log (H.)

(a)

. .
Strong
sorber 4 FIGURE 41.13
{,
I Mappings lor (a) pesticides, (b) PCBs, and (e) PAHs.
2
.¡i

O
. ...
. ...
... . .Sediments
-..
I
of the larger groups fell across several regions (pesticides, PCBs, and PAHs), oth-

.. . .
ers tended to cluster within a single region. For example the halogenated aliphatics
I '" and many of the monocyclic aromatics and ethers fell within the air zone and ex-
¡ -2 hibited strong volatilization losses. In contrast significant numbers of the phenols,
...,.. rI'..
..- -..... . . nitrosamines, and phthalate esters fell within the water zone where purging was
C>
.. . .
.. ...... ..
.2 weak.

-41 . .. -. : ........ Figure 41.13 provides some examples of how the plots could be employed to
-6
-. provide further insight into the behavior within the classes. Figure 41.13a shows

Weak
-8 . how the individual pesticides fall on the space. Such plots might prove useful in
screening applications such as determining which chemicals among a large class of
toxicants would have particular assimilation characteristics.
sorber
-10
I Water
, . , ,
Air Figure 41.13b and e takes the idea farther by attaching additional information
to the points. In Fig. 41.13b the numbers in parentheses indicate the percent chlo-
-10 -6 -2 2 rine for each PCB Aroclor. This plot suggests a positive correlation between sorp-
Soluble 10g(H,) Insoluble
tion, volatilization, and percent chlorine. Together the two effects are manifested as
FIGURE 41.12 a "northeast" movement as chlorine increases. Thus the benefits due to increased
Mapping ler organic priority pollutants on logarithmic space delined by sorption volatilization are somewhat negated (particularly for systems with large resuspen-
(K.m) versus volatilization (H.). A suspended solids concentralion 01 5 9 m-o was sion) by the fact that the higher chlorinated PCBs associate more strongly with par-
used to develop the plol. ticulate matter.
I
 ,

736 PARTVII Toxics

LECTURE41 Mass-Transfer Mechanisms: Sorption and VolatiliLation 737

4 for Kowand He. The lauer are a compilalion of estimales from a variely of sour<:es.
As such they c1earlyexhibit some uncertainty. At the least they could undoubtedly
2
be improved by additional measurements. However, because I have takcn a global
1254,(54%) 1260 (60%) approach, I believe that the uncertainties in the data will not obviate ¡he general
O 8
\. 1248 (48%) trends that have emerged from the analysis.
E -2 1016 (41%)81242 (42%) In conc1usionI have auempted to present a simple, visualIy oriented framework
1232 (31%)
-4 ,. 2Chloronaphthalene (22%) fuI' assessing the assimilation of toxicants for lakes. As was done by VolIenweidcJ'
1221 (21%) in his work on lake eutrophication, my iotent is to draw broad conc1usionsbased on
-6 a global perspective.

-6
PROBLEMS
-10
-10 -6 -2 2 41.1. A sorption experiment yields the following data:
log(H,)
~,.-

(b) ~. cd(¡.o.gL-') 5 10 15 20 25 30 35
4 v (rng g-I) 8 11.5 16 17 19 19 21

O
++ (a) Fit the data with the Langmuir
with the data.
isotherm. Display your resulting isotherm along
+ . (b) Use the results 01' (a) to determine Kd.

..
E .2 (e) Calculate the fractions associated with particulate and dissolved form in the walcr
? .4 I I
..
ee
based on the results from part (b) 1'01'a system with a suspended solids concenlratlOn
01' 10 g m-3. ,
-,:
.. ....-
-6 i · 2 rings
e 3 rings 41.2.
(d) Fit the data with a Freundlich isotherm. Display your resulting isotherm along with
the data.

A lake and its underlying sediments have the following characteristics: m = 5 mg L 1.

-8 . 4rings t/J = 0.8, p = 2.5 g cm-3. The water and sediment solids have fraction organiccarbons
+ 5 rings 01' ¡;/C = 0.2 amI 0.05, respectively. (a) Determine the fraction dissolved in the lake
-10
.10 .6 -2 2 water (Fdl) and the sediment pore waters (F,11'2)1'01'a toxic compound with a K"". =
1Q6. (b) If the total sediment contaminant concentration is 10 fJ.g L -1, compute (he
log(H,)
concentration in particulate form expressed in units 01' fJ.g g-I .
(e) 41.3. The particulate maller in a lake has the following characteristics: m = 5 mg L- I and
¡;/C = 0.4. For a toxicant with a K".. = 105.5: (a) determine the fraction in particulate
FIGURE41.13 (continued)
form; FI'I' (b) repeat part (u), bUI inelude the additional elTect 01'5 mg L -1 01' DOC.

Figure 41.13c c1assifiesthe PAHs according to their number of aromatic rings. 41.4. Determine the volatilization velocity 1'01'DDT (MW = 350, H, = 10 -4.4 atm m ]
For this case volatiliL..lliondecreases as sorption increases. Consequenlly Ihe Irend is mole-I) in a lake where U.. = 2 mps and T = 20°C.
to the "northwest" as Ihe number of rings incauses. For lakes wilh IiUle01'no resus- 41.5. A well-mixed lake has a suspended solidsconcentration 01'0.5 mg L -Ithat has acarhuII
pension, Ihis means Ihat strong assimilalion would slill occur but Ihal sedimentalion content 01' 5% and that sellles at arate 01' 0.2 m d-I. It also has a dissolved organic
would supplant volatilization as the primary removal mechanism. For lakes with carbon concentration 01' l mgC L -1 and its temperature is IOOC. ;
/1ighresuspension, assimilation would decrease as the number of rings increased. A contaminant in the lake has the following characteristics: H, = 10-5 atm m3 1
The foregoinl! ,lIwlysis is offered as an effort lo draw general conc1usions re- mole-I and Koc: = 1 m3 gC-I. Also the liquid and gas-film exchange coefficients are
garding the impl Ili".., of sorption and volatilization on lake toxicant models for 179.7 m yr-I and 135,038 m yr-I, respectively. 1
Determine whether sellling 01' volatilization will be the dominant purging mech-
the organic priurily puuutants. This analysis is based on many assumptions. It is
anism. Do this by de¡ermining the ratio 01'sellling to volatilization. Note: The fraction
strongly predicated on our choices for parameter values and in particular, on values
 '\

738 PARTVII Toxics

.!iI
I
11
of contaminant that is associated with the DOC neither volatilizes nor associates with LECTURE 42
I settling particles.

41.6. A well-mixed sediment layer has the following characteristics: DOCp = 10 mg L-1,
q, = 0.9, Vb = 0.1 mm yr-I, and p = 2.5 g cm-3. The solids in the sediments are 1% Reaction Mechanisms:
carholl. The lake has a DOe conccntration of I mg L -1 and a poe concentration of
0.5 mg L 1 that seltles at a rale of 0.1 m d -l. It al so has a contaminant concentration
of 1 mg m-J. If the log Km. for a contaminant is 5, determine the concentration of
Photolysis, Hydrolysis,
contaminant in the sediment pore water and on the sediment solids. Note that the con- and Biodegradation
taminant has a molecular weight of 300. Also there is no sediment resuspension, but
therc is diffusion between the sediment pore W31er and the overlying water.

'
,f
1
', 41.7. A lake has the following parameters: v" = 100 m yr-I, m = 5 mg L -1, loe = 0.1 and
T = 20oe, and V", = 2 m S-l. (a) Determine the apparent seltling velocity (Fpv,,) and
the apparent volatilization rates (FdV.) for the contaminants listed below:

log (Kv.) log(H.) Molecular weight

.~
TCOD 6.84 -2.68 322
,~¡; DOD 6.12 -7.66 320
,~
~; j
,~ Endrin aldehyde
Endosulfan sulfate
Toxaphene
3.15
_.1.30
3.30
-8.70
-4.59
-0.68
381
422.9
430
-4" ..

tL
~.
(h) In which zones would the contaminants fall on Fig. 41.11.

~
ili
~
41.8. Three new pesticides of equivalent toxicity are being evaluated for use in a lake wa-
tershed. The lake is shallow and has both high suspended solids (foe = 0.05; m = 5
" ~, mg L -1) and significant sediment resuspension. If your only concern is protecting the
R lake ecosystem, determine which toxicant you would employ. Base your analysis on
the toxicant loading concept.
A number of different processes act to transform contaminants in naturat waters.
Some decompose due to exposure to sunlight (photolysis), through chemical reac-
tions (e.g., hydrolysis), or by bacterial degradation. In the following lecture I provide
Pesticide log(Kv,,) log(H,) an overview of these three processes.
J Sans-a-Roach
Gnalmare
7
2
-4
-8
1, Bugs 'r Toasl I -1 42.1 PHOTOLYSIS

Photolysis refers to the breakdown of chemicals due to the radiant energy of lighl.
Light transforms toxicants by two general modes. The first, called direct photoly-
sis, occurs through absorption of light by the compound itself. The second, talled
indirect or sensitized photolysis, represents a group of processes tha! are iniWited
through light absorptionby intermediary compounds. Direct decompositiorlpredom-
inates in systems with little extraneous dissolved or particulate matter such as cIear
lakes. Indirect photolysis was discovered when researchers noticed that some com-
pounds degraded faster in natural water than in distilled water. Thus in more turbid
or highly colored systems, sensitized photolysis could be a very significant decom-
position mechanism for certain contaminants.

739

.. "... ,.-~ .
 L.::'-"'_.~ llrr......_ ... __o

740 PARTVII Toxics LECTUKE

42 Rt:aclion Mt:chanisms: PholOlysis, Hydrolysis, and Biodegradalion 741

e
FIGURE 42.2
Wavelength is delined as the
distance between peaks.

42.1.1 Light
To this point we have covered a number of cnvironmental processes involving solar
2. Llght attenuatlon Partlcle radiation. In Lec. 27 we described how light influences the die-off of bacteria. Then
in Lec. 30 we learned how solar radiation contributes lo lhe heat budgel of a water
body. Finally in Lec. 33 we modclcd the effect of light on planl growth. In each case
3. Absorptlon we treated lighl as a simple homogeneous entity quantified in energy UllilSsuch as
the langley (cal cm -2). Now we lake a c10ser look al solar radialion lo undersland
better how it affects toxics.
4. Quantum yleld
Light has both wave and particle properties. From the wave perspeclive il can
be conceptualized as oscillating electric and magnelic fields that are oriented per-
AGURE 42.1 pendicular to each other. One way lo characterize the light wave is via Ihe dislance
A simplilied schematic 01 the lour processes between maxima. This distance, called Ihe wavelength (Fig. 42.2), can be related lo
influencing direct photolysis in natural watars. olher fundamenlal P'''I,crties by
A= e (42.1)
v "
, . In this lecture I focus on direct photolysis because it is much better understood
where A = wavelength (nm)
than indirect photolysis. The direct photolysis rate depends on a number of factors
e = speed of light in a vacuum (3.0 X 108m S-I)
related to both the characteristics of the compound and the environment (Fig. 42.1):
JI = frequency,definedherc as the numberof complelecycles lhal pass a
l. Solar radiation. Depenc1i;,¡,.10the time ofyear, the weather, and the geographical point in a second (s-I)
position of the water bu.1)',dilTerent levels of incoming solar radialion will be
delivered to the surface 01'lhe water. ~ The particle characterizalÍon of light relales to the fact thal light energy consisls
of discrete quanlities called photons. The amount of energy 01'a photon is relaled lo
2. Light attenuation in the water. Suspended matter, color, and other factors influ-
ence the penetration and attenuation of the lighl. Photolysis in a quiescent, turbid
lake might be limited to a thin surface layer, whereas it could extend to great 500
depths in relatively clear water.
3. Absorption spectrufll he chemical. Due to its chemical structurc Ihe loxicant >o
1:1I......400
will absorb lighl energy lo different degrees from various wavelenglhs. ~L
eo
4. Quantum yield. This refers to the fraction of absorbed photons that result in a CIIE 300
che!Dicalreaction. -g~
0_
'n the fo11owingsections I will elaborate on each of these factors. Then I
ID 200 FtGURE 42.3
0-0 Relationship 01 bond energy to
discuss mathematical m(Jdeb Ihat predict their combined effect on compounds. wavelength with the corresponding
100
ijowever, before embarking on these subjects, 1'11first present some general infor- energy lor some single bonds
O 200 400 600 800 1000
!D~Qoqconceming lighl. shown at the right (data Iram
Wavelength (nm) Schwarzenbach et al. 1993).
742 "ART VII Toxics
LECI1JRE42 Reaction Mechanisms: Photolysis, Hydrolysis, and Biodegradation 743

inelude solar altitude, atmospheric scattering and absorption, reflection, and shading.
, Inlra-
ul!ra- .visible
viole! red In contrast to our other models involving light, photolysis requires that these factors
e be considered as a function of wavelength. For example the fact that the atmosphere
o
:;: selectively reduces various wavelengths must be considered (Fig. 42.4).
.! edge 01 outer atmosphere As was discussed in the lectures on temperature modeling, reflection is impor-
'C tant only when the solar altitude is low.Thus it is usualIy negligible « 10%). How-
ea
... ever, it could be significant when performing both seasonal (during certain times of
... the year at high latitude) and diurnal (early and late in the day) caIculations.
ea
'O
en
42.1.3 Llght Extlnctlon In Natural Waters
As depicted in Fig. 42.5, natural waters selectively attenuate different wavelengths
o 1000 2000 3000 of light. The attenuation of light in natural waters can be quantified by the Beer-
Lambert law,
Wavelength(nm) l(l.,A) = 1(0, A)e-ao(A)z (42.4)
FIGURE 42.4
,..
The eflecl 01 almospheric absorplion and scattering on delivery 01 solar where I(l., A) = light intensity (einslein cm-2 S-I) at depth l. (cm) and aD(A) =- .
radialion lo Ihe earth's surlace (redrawn Irom Hulchinson 1957. Milis el al.
1985). apparent or diffuse attenuation coefficient (cm-I). The attenuation coefficient can
be determined empiricalIy by measuring the light intensity at the surfaee and at a
wavelength, as in particular depth H and rearranging Eq. 42.4 to compute

e I /(0, A)
(42.5)
E = ¡'V = hA (42.2) aD(A) = ¡¡ In [ I(H, A)]
where E has units of kJ and h = Planck's constant (6.63 X 10-34 J s). Because
Before proceeding further we must introduce the notion that solar radiation in
we are dealing with chemical eompounds, light can be expressed on a molar basis.
water does not folIow a path that is strictly vertical. This is why the attenuation
l'hus, an eillsteill is defined as 6.023 x 1023photons. Consequently the energy can
he reexpressed as coefficient is calIed "apparent." In fact the solar irradiance in a water body actualIy
consists of both collimated (that is, traveling as a paralIeI beam) and diffuse light.
E = 6.023 X 1023h~ = 1.196 x 105 Perfectly diffuse light has a mean path length through water that is twice that of a
A (42.3)
beam of light. To quantify this effect the distribution function D has been developed
as
where E has units of kJ mole-l.
According lo both these formulas, shorter wavelengths have higher energy. This
J'('sultis depicted in Fig. 42.3. D(A) = t(A)
H
(42.6)
Finally we should mention the light source that we are dealing with in the envi-
ronment: the sun. As displayed in Fig. 42.4, the solar energy spectrum spans a wide where t(A) = mean path length of the light. For partiele-free water, D has a value
nmge of wavelengths. For photolysis we will be mostly concerned with the visible of approximately 1.2. Miller and Zepp (1979) have reported a mean value of t.6
and ultraviolcl portions of the speetrum (uv/vis). for partiele-Iaden water. The distribution function can be used to relate the apparent
coefficient to the attenuation coefficient per unit pathlength a, as in
42.1.2 Solar Radlation aD(A) = D(A)a(A) (42.7)
Now that we have some general background on light, we can begin to work our way where a(A) has units of cm-l.
Ihrough the scheme outlined in Fig. 42.1. As already discussed in Lec. 30, the amount
Now we can determine the total amount of light absorptionover a layer of thick-
(Ir solar radiation delivered to a natural water depends on a number of factors. These
ness H. This can be done by taking the difference of the incident light at the top of
 LÚ "..

744 PART
VII Toxics lECTURE 42 Reaction Mechanisms: Photolysis, Hydrolysis, and Biodegradation 745

100 104 h ,,- 1014106 "

_,j 1014
Sunlight I I
I I
spectrum , ,,/
1t I " I
I I
103 I 1013 105 I 1013
c: I Q)
o I o'
§:: 10 r:i.
=í
E
I
,,
I
,
I
.
'O
.S! É
'C '.
(1)0

'k102
I
,
1012 104 1012 .:
"'.
t:(I)
.,0 I .!!!o
~~ 1 oE I
I

=5 E¿
<J

A
I
I
I
-';8

i~
10 I 1011 103 1011
J

0.1
(/) 1. . . 1. .10." . . , l.. .. 102
200 250 300 400 500 200 250 300 400 500
0.01 Jli1r A, wavelength (nm) A. wavelength (nm)
(a) (b)

FIGURE 42.6
0.001 . Electronic absorption spectra 01 (a) benzene. a compound that does not directly photolyze,
and (b) naphthacene, a compound that does (110mMilis et al. 1985). The dashed line is a
sunlight spectrum taken Irom Burns et al. (1981).
350 450 550 650 75
I(f, A) = 1(0, A)e-[a(Á)+«Ák]f (42.9)
Wavelength (nm) ..
where I( f, A) = light emerging from the system (einstein cm-2 s-()
FIGURE 42.5
. ... 1(0, A) = incident light intensity (einstein cm-2 S-I)
<' The attenuation ollight in natural waters. Spectral irradiance at different depths
, e(A) = molar extinction coefficient of the compound (L mole-I cm-I)
in Lake San Vincente (Irom Tyler and Smith 1970).
e = concentration of compound (mole L-1)
e = pathlengthof light in the solution (cm)
the layer and the light at depth H. The result can be divided by H to yield the rate of An experimentcan be performedfor each individual compoundand an ab.
light absorbanceper unit volume, ¡ sorbancespectrumdeveloped.Knowingthe mediumattenuationcoefficientand the
pathlength,we can determinethe molar extinction spectrum.As in Fig.42.6, the
(42.8)
1 = I(~ A) (l _e-ao(Á)H) spectrum is usually expressed logarithmically because the attenuation coefficient
often varies over several orders of magnitude.
where 1 has units of einstein cm - 3 S-( . Whencompared with the incident light spectrum, the absorption spcclrum can
be used to crudely assess whether photolysis is possible. For example benzene would
not photolyze because the two spectra do not overlap (Fig. 42.6a). In conlrast, naph-
42.1.4Absorptlon thacenecouldphololyzebecauseit absorbsradiationat wavelenglhsIhalaresignif-
icant in natural waters (Fig. 42.6b).
Now that we know how mlldl light is absorbed in a layer, we describe how it affects Now that we know how much light is absorbed by the compound as a fllnctilHl
compounds. The first step in this process is to characterize how compounds absorb of wavelength, we must determine how much of the light in our layer of water is
light. To do this we employ an experiment that is used to quantify light absorption absorbed by the contaminant. A simple way to do this is based on recognizing that ,
by the compound itself. the totallight absorbance is a composite of the effects of the contaminant and other ~.
Suppose that we introduce a chemical into particJe-free water and place the re- factors such as water, color, and particJes. Further, in most cases, it is reasonable
sulting solution in a transparent container such as a quartz cuvette. The transmission to assumethat the absorbance due to the contaminant is dwarfed by the other factors.
of light through such a system can be quantified by the Beer-Lambert law,
 - _.__..-

746 PARTVII Toxics LECTURE42 Reactlon Mechanisms: Photolysls, Hydroiysls, and Blódegradátillft.'141

Thus we can define the fraction absorbed by the contaminant as to another moiecule aS in photosensitization. lb 411the~~ phehOnténA th~ o"flftilili1
\~ e(A)c e(A)
Fc(A) = a(A) + e(A)c :;; a(A)e
compound
is notchangedchemically.
11te extent of the chemical transformátlons
.
ls éxpressed
, .'
y~
bytmfúlHt1üdi
. .

1 r~ (42.10)
Thisisdefinedas '
where Fc(A) = fraction of absorbance due to the contaminant. This equation can
then be combined with Eq. 42.8 to determine the amount of light absorbed by the ... ' _ number of moles transformed
'V, (1\ ) - ( 42.13)
contaminant in the volume, total moles of photons of wavelength ~ absorbed
/a(A) = ka(A)c (42.11)
\Vhere Note that <I>,(A)is referred to as the reaction quantum yield (mole einstein-I) be-
cause it reflects the total effect of all the photolytic transformations acting on the
e(A)/(O,A)(I - e-ao(A)H) compound. At present it must be determined experimentally.
'íi ka(A) = Ha (A) (42.12) The final step in completing the process outlined in Fig. 42.I is to multiply
11:
in which ka(A) = specific rate of light absorbance (103 einstein mole-I S-I). Eq.42.12 by Eq.42.13 to give
e(A)/(O.A){1 - e-ao(A)H)
- ..- 42.1.5 Quantum Yleld kp(A) = <I>(A)ka(A) = <I>(A) Ha (A) (42.14)
.~
Just because a compound absorbs light does not mean that it will undergo signifi- where kp(A) = a first-orderrateconstant(s-I) forphotolysisas a functionof wave- ....
"IJ cant photolysis. In fact, as depicted in Fig. 42.7, a variety of physical or chemical length. Note that the light intensity / must be expressed as 10-3 einstein cm-2 s-~... .
.11
I t phenomena can occur. for the units to work out as s-l.
The chemical changes are those that lead to an actual modification in the com-
.1
I 'tI pound. Thus these processes are what we call photolysis. In contrast the physical
~ processes involve energy losses. For example the molecule could retum to its ground 42.1.6 Dlrect Photolysls Model
I
~ "
slate via a loss of heat. Further, the chemical can release the energy in the form
I " We can now build a direct photolysis model based on the information from the pre-
of light. This is referred to as luminescence. Finally the energy can be transferred
ceding sections. This is done by merely recognizing that we can simply integrate
Eq. 42.14 to "sum up" the total photolysis rate across the relevant wavelengths,
. I
,. c _
I

f
A, 8(A)/(O,A){I- e-ao(A)H)

,f;
:1
c.
¡- hv
kp - Aa<I>(A) Ha(A) dA

This equation can be simplified by substituting Eq. 42.7 and assuming that the dis-
tribution function and quantum yield are wavelength-independent,
(42.15)

_
kp - <l>DAaf A, 8(A)/(0. A){I - e-ao(A)H)
HaD(A) dA (42.16)
PHYSICAL PROCESSES CHEMICALPROCESSES or in discrete form,
.
Heallransfer ..
Fragmenlallon 700
.Lumlnescence .. Intramolecular
Isomerlzallon
rearrangemenl
kp = <l>D ¿ eA/o.AlIHaD- e-ao(A)Hj (42.17)

.Pholosensltlzatlon ..
Hydrogen alom exlracllon
Dlmerlzallon
Electron transfer
A=290

Now, although this certainly provides a framework for calculating a single, first-
order photolysis rate, it requires a great deal of information. In particular the light
! ~ extinction, yield, and solar radiation must all be supplied as a functionof wavelength.
c Product(s) Further the solar radiation changes as a function of time of year and day. Although
FIGURE 42.7
computer codes are available to facilitale the process (Zepp and Cline 1977, Zepp
1988), simplificalions are also available lo obtain order-of-magnitude estimates. The
Once a chemical compound is excited, a variely 01 physical and chem-
ical processes can occur (redrawn lrom Schwarzenbach et al. 1993).
next section describes the most commonly used approach.
 748 PARTVII Toxics LECJ'URE42 Reaclion Mechanisms: Photolysis, Hydrolysis, and Biodegradalion 749

42.1.7 Near-Surface Approach where kpO = direct near-surface photolysis rate (d-I)
I = total solar radiation at the top of the layer (ly di)
A problem with the framework developed hc previous section is that it requires
the acquisition and manipulation of large quantities of information. Although this
lo = totalsolarradiationat whichk/1IJwas meaSIlI"'!(ly d- 1)
0100 = ratio of the radiance distribution function to Ihe radiance dis-
information is sometimes available, more often experimental data are reported as
tribution function at the surface (= approximately 1.33)
near-surface rate constants. It would be useful to be able to extrapolate this data to r
uo(A.) = extinction coefficient (m-I) evaluated at the wavelength 01'
ottler contexts. To do this let us first look at what happens to Eq. 42.17 if uD(Á)H is
/ .' maximum light adsorption A*
very small, as would be the case near the surface of a natural water. For such cases,
Therefore, to use this approach, three parameters are required: k/¡(J,lo, ami
I - e-ao(A)H==uD(Á)H (42.18) uD(A.). Table 42.1 summarizes some values. In addition the light extinction can be
Therefore. Eq. 42.17 becomes ~
calculated by
700 ",1, uo(A) = O[ulV(A)+ u,,(Á)a + u,,(A)DOC + Us(Á)III] (42.22)
kpO = ct>00 ¿ 8A/0.A (42.19) where 0= 1.2 to 1.6
A=290
ulV(A) = attenuation coeflicient for water (m-I)
The ratio of Eq. 42.16 to 42.17 can be formed as u,,(A).u,.(A), and us(Á) = concentration-specilic attenuation coefficients
(L mg- 1m-1) for chlorophyll a, dissolved organic cal'''

~
ct>0 ~
A=290
8A/o.A(1 - eao(A)H)
HUD(Á)
bon, and inorganic suspended solids, respectively
(42.20) a, DOC, and m = concentrations (mg L-1) 01'chlorophyll a, dissolved oro
700 !,.
kl'o ganic carbon, and inorganic suspended solids
ct>00 ¿ 8A/0.A
A=290 Table 42.2 contains values 01'the u 's as a function 01'wavelength.
Figure 42.8 shows the effect of depth and water clarity on the atlenuation 01
Now to simplify further. we assume that most of the photolysis takcs place at a the surface rate. Other references (e.g., Lyman et al. 1982, Milis et al. 1985) con-
.-1
narrow wavelength region. If this is the case, Eq. 42.20 can be simplified to give tain parameter values for a number of constituents as well as descriptions of other
I O I - e-ao(A')1I techniques for estimating the photolysis rate.
kl' = k/¡/'¡-¡~n ,L. "*' (42.21)
"
. -~. TABLR42.2
TABLE 42.1 Attentuatlon coefficientsparameterizing the dependency of light
Near-suñace direct photolysis parameters for a select extinction on wavelength. Polynomial interpolation can be used to
group of toxic organics (extracted from Milis et al. determine Intermediate values. or a more complete version of the
table can be found in Miliset al. (1985)
1985) -----
k". l. A' Wavelength
Compound (d-I) (Iyd 1) (nm)
center a. a. a, a,
(nm) (m-') (Lmg-1m 1) (Lmg 'm') (Lmg1m ,)
Naphthalene 0.23 2100 310
300 0.141 69 6.25 0.35
I-Methylnaphlhalene 0.76 2100 312
320 0.0844 63 4.68 (1.35
2-Methylnaphlhalcne 0.31 2100 320
340 0.0561 511 3.50 0.35
Phenanthrcne 2 2100 323
Anthraccne 360 0.0379 55 2.62 0.35
22 2100 360
130 3110 0.0220 46 1.96 0.35
9-Melhylanlhracene 2100 380
400 0.0171 41 1.47 0.35
9,IO-Dimelhylanthracene 411 2100 400
24 440 0.0145 32 0.1121 0.35
Pyrene 2100 330
400 0.0257 20 0.344 0.3;
Chrysene 3.11 2100 320
490 2100 440 550 0.0638 10 0.167 0.35
Nnphthacene
31 2100 380 600 0.2440 6 0.0111 0.35
Benzo(a)pyrene
28 650 0.3490 8 - (1.35
Benzo(n)anthracene 2J()() 340
-
0.32 2100 700 0.6500 3 0.35
Carbary1 313
-
0.46 600 318 750 2.47()() 2 0.35
Pentachlorophenol (anion)
800 2.0700 O - 0.35
3.3' -dichlorobenzidine 670 2()()() 28(}...330
750 PARTVII Toxics . LECTURE
42 Reaction Mechanisms: Photolysis, Hydrolysis, and Biodegradation 751

Chis DOC .. . 500 1 - e-8.36(IO)

Watertype (mgL-')(mgL-') (mgL-')' 1 k,. = 0.2321001.33 H. = 0.OOO87d-1
A Pure water O O O which corresponds to a half-life of 0.693/0.00087 = 795 d.
B Lake Tahoe 0.001 0.1 0.5
C Eutrophlc 0.01 0.5 S
D Hypereutrophlc 0.1 2 20
As in the previous example Ihe produclivily of Ihe syslem lends lo diminish Ihe
effecliveness of phololysis as a removal mechanism. This is due lo Ihe increased lur-
bidily and color Ihal reduces lighl penelralion for such syslems. However, il should
:t be noled Ihal such syslems are more likely lo exhibil indirect or sensilized phololysis
1~ because of Ihe high concenlralions of organic matter in eUlrophic walers.
<; i 0.5
"'f..~ 42.2 SECOND-ORDER RELATIONSHIPS
...I."t
To Ihis poinl we have used firsl-order kinelics lo characlerize any reactions in our
models. Such reaclions are formulaled malhematically as
Reaclion = kV e (42.23)
5 10 This reaclion is a special case of Ihe more general formulalion ,...
.
Depth (m) Reaction = kn Ven (42.24)
FIGURE 42.8
where 11= orderof Ihereaclion.WehaveusedIhenomenclalureknto indicalethat
The effecl 01 deplh and waler clarity on Ihe attenualion 01 Ihe sur/ace
phololysis rale. This plol corresponds lo a wavelenglh 01 maximum
Ihe reaclion rale will have differenl unils depending on ils order. For example Ihe
light absorplion 01 340 nm.
second-order reaclion
Reaclion = k2 Ve2 (42.25)
EXAMPLE 42.1. NEAR.SURFACE DIRECT PHOTOLYSIS RATE. Determine has the rale conslanl k2. which has units (L3 M-' T-1). Thus Ihe 10lalexpression
the photolysis rate for naphthalene in the epilimnion (mean depth = 10m) of (a) an olig- has unils of (M T-1).
otrophic and (b) aeutrophic lake. The average light inlensity at the surface is 500 Iy d"'.
Although Eq. 42.25 represents one way lo specify a second-order reaction. it is
Solutioli: (a) From Table 42.1, Ihe parameters for naphthalene are k,.o = 0.23 d-I, nol Ihe Iypical formulalion used in loxic-substance modeling. Ralher such rales are
lo = 2100 Iy d-I , and ;r = 310 nm. The data from Table 42.2 can be evaluated with usually expressed as a funclion of Ihe producI of two concenlrations,
cubic interpolation to estimate the exlinction coefficients for A* = 310 nm, with the .. Reaclion = k2Vbe (42.26)
results being Uw = 0.108, Un = 65.9, a,. = 5.41, and as = 0.35. These rcsultscan
be used to determine Ihe total extinction coefficienl. Because the lake has a lower solids where b = a concenlralion (M L-3) upon which Ihe reaction depends in addition to e.
content, we will assume that D = 1.2. Applying Eq. 42.22 and using the values for Lake In Ihe nexl Iwo sections we discuss Iwo mechanisms-biotransformation and
Tahoe from Fig. 42.8 gives hydrolysis-Ihal are somelimes modeled with second-order reaclions.
an(31O) = 1.2[0.108 + 65.9(0.001) + 5.41(0.1)+ 0.35(0.5)J
= 0.13 + 0.079 + 0.65 + 0.21 = 1.068 m-I 42.3 BIOTRANSFORMATION
Thus, because of the low solids and DOC, the extinction is low. Now substituting this Biotransformation refers lo Ihe microbially medialed Iransformalion of organic
along with the other parameters into Eq. 42.21 yields
conlaminanls. Typical degrading microorganisms are helerolrophic bacteria, actino-
500 I - e-I.068(1n) myceles. aUlolrophicbacteria. fungi, and prolozoa.
kp = 0.232Iool.:n _._ = 0.00682d-1 The label"biolransformalion" encompasses a number of dislinctly differenl pro-
cesses, including (Alexander 1979):
This rate can be pUl into perspective by expressing it as a half,life of 0.693/0.00682 =
102d. l. Mineralizalioll. The conversion of an organic compound lo inorganic products.
(b) The calculation can be repeated for the eutrophic case from Fig. 42.8, 2. Deloxiealioll. The conversion of a loxicanl lo innocuous by-products.
3. COllletaholislII.The melabolism of a compound Ihal Ihe organisms cannol use
anCHO) = 1.6[0.108 + 65.9(0.01) + 5.41(0.5) + 0.35(5)]
as a nulrienl (which does nol result in mineralization, and organic metaboliles
= 0.173 + 1.05 + 4.33 + 2.80 = 8.36 m'l remain).
~~ ".

752 PARTVII Toxics LECTURE42 Reaclion Mechanisms: Phololysis, Hydrolysis, and Biodegradalion 753

TABLE 42.3 When (as is often the case) c « Ks, Eq. 42.9 simplifies to the linear form,
Second-order biodegrad rate
constants ror a select g/v.. , toxic k b -- /J-",."
- X -- k b2X (42.28)
organics (extracted rrom Milis et al. Yk,
1985) where kb2 = a second-order biotransformation coefficient with units of
k61 m3 (celIs yr)-I. For systems where the microbial population is relatively constant,
Compound (mL (''ell-I d-I) ~. 42.28 reduces to a first-order rateoTable 42.3 tists values of kb2 for a number
2,4-D BUloxyelhyl eSler 1.2 x \O-s of toxicants. Table 42.4 provides some estimates of bacterial concentration foe a
Malalhion l.l X \0-6 variety of aquatic environments.
Chloropropham 6.2 X \0-'0 Extrapolation of lab data to the environment can pose a significant problem in
Furadan 2.4 X \0-8 estimating biotransformation rates. Lyman et al. (1982), MilIs el al. (1982), and
Alrazine 2.4 X \0-8
2-Chlorololuene 6.5 X \0-8
Schwarzenbach et al. (1993) discuss this problem. They also contain summaries of
Dimelhyl 1.2 x \0-' parameter values for selected contaminants and additional information on biotrans-
Di-elhyl 7.7 X \0-8 formation.
Di-n-bulyl 7.0 X \0-7
Di-n-OClyl 7.4 X \0-9 EXAMPLE 42.2. BIODEGRADATION RATE. Delermine Ihe hiodegradation rak
Di-(2-elhylhexyl) 1.0 X \O-lO of di-n-butyl in (a) an oligolrophic and (b) a eutrophic lake. The second-order decay rate
Phenanlhrene 3.8 X \0-6 foe this peslicide is approximalely 7 x \0-7 mL cen-' d-I (Milis et al. 1985).
Solution: (a) Feom Table 42.3 we will assullle a bacterial concentralion of 100 cells
4. Activation. The conversion of a nontoxic substance to a lOxicone, or an increase mL -1 to be represenlalive of an oligolrophic lake. The firsl-order value can be deler-
lIlined as
in a substance's toxicity, by microbial action.
5. Defusing. The conversion of a potential toxicant into a harmless metabolite before kb = 7 X 10-7(100) = 0.(J()(J()7d-I
its potential is reatized.
which correspolIds to a half-life of 0.693/0.(J()(J()7 = 9902 d (27 yr).
If the microbiological community has adapted to the contaminant, a Michaetis- (b) From Table 42.3 we will assume a baclerial concenlralion of 5000 cells mL-1
Menten equation can be used to represent the rate of biotransformation as to be representalÍ"leof a eutrophic lake. The firsl-order value can be delerlllined as
f
. -~- kb = 7 x 10-7(5000)= 0.0035d-I
k - /J-maxX (42.27)
b - Y(ks + c) which corresponds lo a half-life ofO.693/0.oo35 = 198 d.

where /J-max= maximumgrowthrateof theculture(yr-I)

X = biomass concentration of the microorganisms (celIs ID- 3) As in the previous example, the water body's productivity tends to im:eease the
Y = yield coeffi¡'nl (celIs produced per mass toxicant removed) effectiveness of biodegradation as a removal mechanism. This is due to the IIIcreased
ks = half-saturalion constant (/J-gm-3) bacterial populations for such systems. This result is in contrast to the result for direct
photolysis in Example 42.1.
TABLE 42.4
Typical bacterial population densities in natural
waters (Milis et al. 1985) 42.4 HYDROLYSIS
Water-body type Bacterial numbers
Hydrolysis refers to reactions in which the bond of a molecule is cJeaved and
Surface walers 50-1 x 106 cells mL -, a new bond is formed with the hydrogen a/ld the hydroxyl componenls of a waler
Oligolrophic lake 50-300 eells mL -, molecule. Hydrolytic reactions are catalYLedby acids or bases and, to a more lim-
Mesolrophic lake 450-1400 cells mL-'
EUlrophic lake 2000-12,000 eells mL-'
ited extent, by water.t These catalytic effecls depend 0/1the type of reaction and
EUlrophic reservior 1000-58,000 eells mL -,
Dyslrophic laket 400-2300 eells mL -1
Lake surficial sedimenls 8 x \09-5 X \0'0 cells g-dry wl-I
Slream sediments 'Inorganic calalysls olher Ihan water (such as sulfale. nilrale. biearbonale, etc.) can ind"" .ransforma-
\07-IOS cells g-dry w¡-'
lions in a similar manner lo hydrolysis. Allhough Ihese are oflen dwarfed by hydrolysis. Ihey cal1 be
t Dystrophic lakes are hog-like with high humic acid eonlen!. importanl in some syslems (see Schwarzenbach el aL 1993 for delails).
754 PARTVII Toxics LECTURE
42 Reaction Mechanisms: Photolysis, Hydrolysis, ami Biodegradation 755

2
pH
pH
O 2 4 6 8 10 12 14
O
... 4 6
-lo!
el) -2 "'-1.
..5! :~ ' ~..................
-4
10-=.1:" .. ....... FIGURE 42.9 .... 0.1
Effecl 01pH on Ihe hydrolysis rale.
The case shown is lor k. = 1, k. =
-6 ... . .. 103, k. = 10', and K.. = 10-".
""
.:o:
0.01

the chemical structure of !he compound. The pH and temperature of the solution
have an influence on !he rate of reaction.
The rate uf hydrolysis can be expressed as
k¡, = kb[OW] + kn + ka[H+]
where kn, ka, and kb = coefficients parameterizing the neutral, acidic, and basic
dependencies of hydrolysis. The neutral rate kn has units of d-I , whereas ka and kb
have units of L mole-I d-I.
If the equilibrium relationship for the dissociation of water,
KIV= [H+I[OW]
is substituted intu Eq. 42.29, the rate can be expressed in terms of pH as
k¡, = k¡,I~;H + kn + kalO-pH
Thus, depending on the coefficient values and the pH, the rate can either be second-
or first-order (Fig. 42.9).
Hydrolysis rates range from 10-7 to 10-1 d-I. Lyman et al. (1982), Milis et al.
(1985), and Schnoor et al. (1987) can be consulted to obtain parameter values for
selected contaminants and additional information on hydrolysis.
EXAMPLE 42.3. HYDROLYSIS RATE. Determine the hydrolysis rate of pen-
tachlorophenol (k" = 5.8 x 10-.1d-I, k. = 1.1 x 1ü4 L mole-I d-I, and kb =
3.3 L mole-I dO') for the epilimnion (pH = 7.5 to 8.5) and hypolimnion(pH =
6.5 to 7.5) of a lake.
Solution: Using the mean pH for the epilimnion, we can estimate thc hydrolysis rate as
(Eq. 42.31)
10-14
kh = 3.3 10-8 + 5.8 x 10-.1+ (1.1 x 104)10-8 = 0.0059 d-I
The range can also be determined as 0.00615 to 0.00585 d-I, which corresponds to
half-lives ranging from 113 to 119 d. Thus it is fairly constant over the range of pHs
encountered in the epilimnion.
For the hypolimnion. the rale at the mea n pH of 7 is 0.0069 and has a varia-
tion bctween 0.00928 and 0.00615 di, which corresponds to half-lives ranging from 75
0.001 -L FIGURE E42.3
(42.29)
to 113. Conscquently it seems to vary much more in the bottom waters. A plot of the
decay rate versus pH provides an explanation of these results (Fig E42.3). ,.",.
As shown, the epilimnion pH variations accur on the nat, neutral par! of the functioQ.. .
whereas the hypolimnion varies across the range where the acidic constant is becoming
significan!. At lower pH's the effect of pH would become even more pronounced.
(42.30)
42.5 OTHER PROCESSES
(42.31)
In addition to the above processes, other mechanisms can effect Ihe transport and
fate of toxicants in aquatic environments:
Oxidatioll/reductioll. Chemical oxidation reactions occur in natural waters
whcna sufficientlevelof oxidantis present.CommonIypesarechlorineand
ozone. The general form of the chemical oxidation rate is
ko = k¡¡Ox (42.32)
where kij = a second-order rate constant and Ox = concentration of the oxi-
dant. Ifthe oxidant concentration is constant, Eq. 42.32 represents a first-order
rateo
Acid-base effects. The disposition of organic acids and bases (e.g., phenol, ben-
zidine, etc.) can be strongly effectcd by the hydrogen ion concentration of!he
water. Milis et al. (1985) can be consulted to obtain additional information on
this mechanism.
Biocollcelltratioll. The capacity of a chemical to be taken up by organisms is
termed bioconcentration. This effect is treated similar to the linear sorption
mcchanism. It is quantified by a bioconcentration factor that is analogous to
the partition coefficicnt (Lec. 45).
Additional information on these transformations can be found in references such
as Lyman et al. (1982) and Schwarzenbach et al. (1993).
756 PARTVII Toxics

PROBLEMS LECTURE 43
.-- - -- ----
42.1. You measure values of 500 Iy d-I and 100 Iy d-I at 1 m and 9 m below the water
surface of a lake. respectively. Determine the apparent attenuation coefficient.
Radionuclides and Metals
42.2. Repeat Example 42.1, but for the compound pyrene.

42.3. Repeat Example 42.2, but for the compound 2,4-0 butoxyethyl estero

42.4. Determine the combined half-life due to hydrolysis and biodegradation for 2,4-0 bu-
toxyethyl ester in a hypereutrophic lake (X = 50,000 cells mL -1) for two cases: p~ =
7 and pH = 7.5. Note that for 2,4-0 butoxyethyl ester, ka = 1.7 L mole-I d-I, k. =
O d-I, and kb = 2.6 x I(}ÓL mole-I d-I.

42.S. Repeat Example 42.3, but for the pesticide parathion (k. = 3.6 x 10-3 d-I, ka =
1.3 x 1()2 L mole-I d-I, and kb = 2.46 X 103 L mole-I d-I).

42.6. Determine the 95% response time for a contaminant in a lake having the following
characteristics: Q = 50 x 1(}Óm3 yr-I, V = 200 X 106 m3, H = 10 m. The lake has
an inflow solids concentration of 20 mg L - I and a suspended solids settling velocity
of 0.25 m d-I . The contaminant has the following characteristics: log Kaw = 6, ka =
L mole-I d-I, k. = 8.16 X 10-4 d-I, and kb = 9182.7 L mole-I d-I. Note that only
the dissolved contaminant is subject to hydrolysis. Volatilization, sediment feedback
and other decay mechanisms are negligible. The pH of the system is 7, the suspended
solids are 5% organic carbon, and the dissociation constant for water is approximately
10-14.
q preyiQj1s lectures it1.tJ1at it partiti()ns !he metal into particulate and disso!ved
T'
LI!CTURE OVl!RVlEW; 1describe!hetransportandfateof ínorganictoxicsin natu-
ral waters, After an overvíew of such substances, J develop and apply a model
framework for radíonuctides to slmulate !he levels of cesium- l 37 in Lake
trtic~¡g¡¡Q. J fol!oW "lis witll a de¡¡cription of two !pode!S pf !fIeta}s in natural
water~..The j1rst is patfel'lled a(ter the toxic org¡¡nic framewpr}¡;s ~escriJ>ed in
I
fOfl11~'The secon4 adds metal precipitation in !he sediment layer. Thís m()det
j
I
nl~~tra~~ ~ow cheD11cal equilibrium models can pe il1te$11Ite4 wi~ ma~~
balances in or4er tO more effectively model metals. r

To this point we have devcloped frameworks lo model toxic organics in nalural

waters. We'JI now cover how inorganic loxicanls, namely metals and radionuclides,
are simulated.

43.1 INORGANIC TOXICANTS

In Ihis leClureI describe models for Ihe two mosl common categories: hcavy metals
and radionuclides (Tablc 43.1). Heavy meta/s usually refcr to melals bctween alomic
numbers 21 and 84. A few nonmetals (arsenic and selcnillm) as well as the lighter
:2.-'
metal aluminum are also included in this category. Most of these occllr naluraJly.
{' However, they can bc greally enhanced by mining and by induslrial activities sm;h
as electroplating, smelting, and manufacturing.
Radiolluclides (or radíoactivc substances) result from nuclear energy genera-
tíon, nuclear weapons development, and some industrial applícations. In addition
certain radionuclides occur naturally (e.g., lead-21D)and are employed by environ-
r mental scicntists as tracers.

757
758 PARTVII Toxics LECTURE
43 Radionuclides and Metals 759

TABLE 43.1
Listing of some metals and radionuclldes of interest In water-quallty
modellng
Loadlng Outflow
.
Metals Radlonuclldes Water
Name Symhol Name Symbol Name Symbol (1) Decay
Aluminum Al Cesium Cs Cesium-137 mCs
Chromium Cr Barium Ba Plutonium-239. 240 2J9.240Pu
Nickel Ni Silver Ag Stronlium-90 90Sr
Copper Cu Mercury Hg Lead-210 210Pb
Zinc Zn Arsenic As
Cadmium Cd SeJeniul11 Se Dlffuslon
Lead Pb
Sedlments
(2)

43.2 RADlONUCLlDES FIGURE 43.1

Radionuclides are similar to organic pollutants in that they sorb to particulate matter. Schematic 01 a radionuclide budget for a well-mixed lake with sediment feedback.
,-.-
However, they differ in two fundamental ways: -
·. The radioactive substances of the type listed in Table 43.1 do not volatilize.
They decompose by simple first-order radioactive decay kinetics. They are not
A simple framework for a radionuclide in a well-mixed lake is depicted in"
Fig. 43.1. As with organic pollutants, a solids budget is required. After this is done,
mass balances for each of the layers can be written as
subject to the variety of decomposition mechanisms, such as photolysis, biodegra- dCI
dation, etc., that act to break down organic compounds. VId( = QC¡n - QCI - kVlcl - vsAFpJcJ + VrAC2 + vdA(Fd2c2 - FdICI)
(43.1)
In addition they are measured in radioactivity units (Ci or curie) rather than
dC2
in mass units. One curie corresponds to the disintegration of 3.7 X 1010atoms per and V2- -kV2C2 + vsAFplcJ - VrAC2- VbAc2+ VdA(FdICI- Fd2C2)
dt
second. This value was chosen because it is approximately the decay rate of 1 g of
ladium. A related unit is the becquerel (Bq), where 3.7 X 1010Bq are equivalent (43.2)
lo a CL Note that typical radioactivity magnitudes encountered in natural waters are where subscripts 1 and 2 indicate water and sediments, respectively, and
quite low when expressed in Ci units. Thus, beyond the common SI prefixes used
to this point (recall Table 1.1),additional prefixes corresponding to tiny magnitudes t = time (yr)
me necessalY (Table 43.2), C = concentration (Ci m-3)
V = volume (m3)
Cin = inflow concentration (Ci m-3)
TABLE 43.2 k = first-order decay coefficient
SI (lnternatlonal System A = sediment surface area (m2)
of Units) prefixes Vs = settling velocity of solids (m yr-I)
commonly used In Vd = sediment-water diffusion mass-transfer coefficient (m yr-I)
water-quallty modeling Vr = resuspensionvelocity(myr-I)
of radionuclldest Vb = burialvelocity(myr-I)
!:l"
Preflx Symbol Value The dissolved and particulate fractions can be calculated as previously derived in
pico- p 10-12 Lec. 41,
'ernto- , 1015 I
(43.3)
allo- a 10..1. Fdl =I- Fpl = I + Kdlm
t A complete list of prefixes ap- 1
and (43.4)
Fd2 = f/J + Kd2(1- f/J)p
pea'" in App. A.
 ~...~ Le.-",

760 PARTVII Toxics LECTUKE

43 Radiolluclides alld Metals 761

TABLE 43.3 111addition a solids budgel for Ihe lake can be developed. The resullillg paramelers are
Parameters Cor some radionuclides oC interest in water-quality modeling t/J = 0.9 p = 2.5 X 106 g m-3 In = 0.5 g m-3
Parameter VoUs 13..z.wPu lJ7Cs "Sr lIOPb W5 = 3.8 X (012 g yr-I v5 = 1.25 m d-I v, = 0.1565 mm yr-I
Half-life yr 4.5 x \O' 30 28.8 22.3 Vb = 0.756 mm yr-I
Diffusion coefficienl cm2 S-I 12.1 x \0-6 12.1 X \0-6 3.8 X 10-6 4.7 X \0-6
panilion coefficienl m) g-I 0.5 0.5 2 x \0-4 \O During Ihe lale 1950s and early 1960s, nuclear weapons lesling resulted in a fallout
(L kg-I) (5 x lOS) (5 x lOS) (2 x 102) (1 x (01) flux of a number of radionuclides to the surface of the earlh (recall Example 5.3). The
values for 239.240pUcan be tabulaled (Table 43.4). These values can be con verted lo fluxes
Of90Sr and I37Cs by multiplying by approximalely 57.9 and 60.9, respectively. Predicl
where Kd = a partitioncoefficient(m3g-I) the response of Lake Michigan lo Ihe cesium flux.
m = suspended solids concentration (g m-3) Solution: Equalions 43.1 alld 43.2 can be inlegraled and the resulls displayed iu
p = sediment density (g m-3) Fig. E43.1.
cf> = sediment porosity

The decay rate is related to the half-life by

T¡¡ble4}3 lists key parameters for some common radionuclides.

= 0.693
tso
(43.5)
TABLE 43.4
Fluxes OC2.W.240pU
Year Flux Year
-
in CCi m-2 yr-I to the Great Lakes
Flux Year Flux
36.0
Year Flux
11.5
1954 51.5 1960 37.0 1966 1972
1955 60.7 1961 53.0 1967 253.0 1973 4.6
EXAMPLE43.1. CESIUM-137IN LAKE MICHlGAN. LakeMichigan(excluding 1956 73.6 1962 166.0 1968 20.0 1974 4.6
Green Bay) has Ihe following physical characleristics: 1957 60.7 1963 345.0 1969 25.0 1975 4.6
4.6
Q = 40X 109m3yr-1 Al = 53.500X 106m2 VI = 4820X 109m3 1958 128.8 1964 255.0 1970 39.1 1976
1959 161.0 1965 96.6 1971 32.0
A2 = 20.000X 106m2 H2 = 2 cm

43.3 METALS
r (8) Inflow "
. .., 30,000
As with toxic organics and radionuclides, the fate of metals is inextricably tied 10-
20,000 gether with the fate of solid matter. Thus transport mechanisms such as sorption
O and settlinglresuspension significantly affect the fate for metals. Although there is
10,000 this similarity, metals ditfer from organic pollulants and rarlionuclides in several
oS
....
O
ways:
l. Naturallevels. Most heavy metals occur naturally. Thus we must often consider
400 T (b) Water background levels when assessing anthropogenic sources.
f" 2. Lack of decay mechan;sms. Most metals are conservative in that the tOlalamounI
E
O 200 of metal is not broken down by processes such as biodegradation, photolysis, or
....
radioactive decay. The absence of these mechanisms simplifies their modeling.
O In addition, although most do not have gaseous phases, some of the nonmelallic
inorganic toxicants (e.g., mercury) can be lost by volatilization.
3. lnorgan;c sorpt;oll. Although metals associate with particulate maUer,the nature
60 1 (e)Sedlments of their sorption ditfers from loxic organics. Recall that hydrophobic organics
1;; have an affinity for organic carbon particles. Therefore, as expressed by Schnoor
30
g et al. (1987), there is a "Iikes-dissolving-likes" quality to the mechanism. Metals
can be complexed by organic ligands and, thus, can be sorbed to organic solids in
?
O
a fashion similar to toxic organics. However, several other processes are signifi-
1950 1960 1970 1980 cant including (a) physical adsorption to solid surfaces, (b) chemical sorption or
Year FIGURE E43.1 binding by ligands. and (c) ion exchange.
 762 PARTVII Toxics
LECTURE
43 Radionuclides and Metals
. 763

4. Chemical speciation. Aside from sorption, metals can further speciate into dif- Solutlon: The fraclions can be calculaled as
ferent chemical forms. These species can be important in thal Ihey can exhibit 1
differing lransport and fate as well as differing toxicity. FJI =. . ~ ~.~ _, = 0.909 Fpl = I - 0.909 = 0.0909
Although some of Ihe above fealures mean thal melals models could be simpler 1 _
Ihan loxic organic models (e.g., lack of decay mechanisms), Ihe olher points mean FJ2 = 0.9 + [0.2(1 _ 0.9)2.5 x I06J = 2 x 10 5
Ihal Ihe melals frameworks could be more complex. In Ihe following sections I' 11first Then the recycle fraclion can be delerrnined as (Eq. 43.11)
show how a simple melals model can be pattemed after our toxic organic framework.
, 0.000157 + 1(2 x 10-5)
Then 1'11describe how equilibrium chemistry can be inlegrated wilh a mass balance
Fr = 0.000157 + 0.000756 + 1(2 x 1(}-5) = 0.1893
lo allow a more complex represenlalion lo be developed.
and Ihe nel loss rale as (Eq. 43.10)

Vr = (1 - 0.1893)[456.25(0.(J909) + 1(0.909)1 = 34.36 rn yr-I

43.3.1 Simple Partltlonlng Models
The concenlralions in Ihe waler and the sedilllenls can be deterrnined as
As was the case for radionuclides, a very simple model of melal dynamics in nalural
4 x I(}IO(I) _1
walers can be developed by assuming Ihat the metal partitions into dissolved and CI = ~.n. ._Oo' =0.055¡.tgL
particulate forms. In a fashion similar to Eqs. 43.1 and 43.2, mass balances can be
and 10-1¡.tg
developed for a well-mixed system,
(/cl
V'-dt = QCin - QCI - v,AF"ICI
. + VrAC2 + vdA(Fd2c2 - FdIC)) (43.6)
\' =
2 (1 - 0.9)2.5x 10610.0()OI
IO¡.tgg-1
456.25(0.0909)+
57 + 0.000756+ 1(2x 1(}-5)J.
1(0.9091) 0055
( mg ) , -1

dC2
and V2-;¡¡ = v.,AFI'ICI- VrAC2- V"AC2+ v"A(Fdlcl - Fd2C2) (43.7)
NOlicetha! these equations are idenlical to Ihe radionuclide model wilh Ihe exceplion 43.3.2 Chemleal Equlllbria and Mass-Balanee Approaehes
I

I Ihal decay is omitted. (Advaneed Topie)

./ Al sleady-slale, Ihese equalions can be solved for
r QCin
Although Ihe simple sorplion approaches described previously have some utilily for
CI = (43.8) firsl-order eslimales, modeling frameworks musl usually address key aspecls of mel-
Q + VrA
als chemistry lo oblain more refiued prediclions. One way lo do Ihis is lodireclly inle-
v,FI'I + V"Fdl grate large equilibrium codes inlo mass-balance frameworks. For example, Runkel et
'lIId 1'2 =' CI (43.9)
(1 - ¡JJ)p(vr+ v" + 1'"F,I2) al. (1996a, b) combined EPA's MINTEQ model wilh an adveclion-dispersion model
where I'r = net loss rale (m yr-I), lo sludy melals dynamics in a stremn. Because of compuling advances, such appli-
calions should become more commonplace in Ihe fulure.
Vr = (1 - F;)(v.,Fpl + V"Fdl) (43.10) An alternalive approach, which is closer in spirit lo olher models in Ihis lec-
'\lid F; is Ihe ralio of sedimenl feedback lo total sediment purging, lure, is to identify a few key equilibrium-chemislry mechanisms and inlegrale Ihem
v + Vd Fd2 inlo Ihe mass-balance approach. Recenlly Dilks el al. (1995) have deveIoped such a
F; = Vr +r VI>+ ~'dF,12
(43.11) model for waler-sedimenl systems. As was Ihe case for toxic organics, this model di-
vides Ihe melal inlo parliculale and dissolved forms (Fig. 43.2). However, in anoxic
Thomann (19H5)compiled waler-column partition coefficients for copper, zinc, sedimenls Ihe melals are subject lo precipitalion wilh hydrogen sulfide. This model
cadmium, chromium, lead, and nickel from 15 rivers. The values ranged from 1 X is useful in iIIuslrating howchemical equilibria mechanisms can be incorporaled into
10-4 lo 0.1 mJ g-I (102 lo HP L kg-I). On Ihe basis of this data, Thomann and a mass-balance approach lo more effeclively model melals.
Mueller(1987)concludedthatFtll ;;;; 0.8 :t 0.2 formetals.Theyalsonoledthatthe Mass-balance equalions can be wriUen for Ihe waler and the sedimenls in a
partilion coeffkient could be less in Ihe sediments. fashion idenlical lo Eqs. 43.6 and 43.7,
EXAMPLE 43.2. HEAVY METALS IN LAKE MICHIGAN. Using Ihe same pa- dCI
ntmeters as in Example 43.1. determine Ihe sleady-slale concentrations of zinc in Ihe wa- V'd/ = QCin- QCI -".,AF"lcl - F,flCI) (43.12)
+ I',.AC2+ 1,,,A(F,I2C2
ter and sedilllenls of Lake Michigan for a hypolhelical inl10w concenlntlion of I fJ.g L -l.
Assume (hal Ihe partition coel1icienl for zinc is 2 x lOS L kg 1 (0.2 m-' g-I) and Ihe and dC2
V2-(, 1 = ,..AF,}lcl - VrAC2- 1'"AC2+ v"A(f~/ICI- Fd2C2) (43.13)
scdilllcnl-waler diffusion lIIass-transfcr cocflicicnl is 1 11Iyr 1.
 ---~--- ~ " ~--- -~ ~~-~ _.~- -- ---

764 PARTVII Toxics LECTURE

43 Radionuclides and Me¡als 765

Loadlng OuUlow
. KI = [H+][S.!
[HS-l

[H2S]
1 (43.19)

Water K2 = [H+][HS-l (43.20)

(1) K = [cd/</J][HS-] (4321 )
s [H+] .
Twoolherequalionsare providedby massbalances.The firslis the sulfurbal-
ance,
AVS = [S2-j + [HS-j + [H2S]+ Cs (43.22)

Sedlments where AVS = acid-volatile sulfide. This is a measure of ¡he reactive sulfide in the
I I
system (AIIen et al. 1993).
(2)
The second is the metal balance expressed by Eq. 43.14. This balance can be
Prec¡pltate formulated in terms of Ihe model variables and parlilion coefficienls by recognizing
Ihal an equilibrium partition coefficient can be defined as (recall Eq. 41.9)
_ cl' </J
FIGURE 43.2 Kd - -cJ (1 - </J
)p (43.23)
Schemalic 01 a melal's budgel lor a well-mixed lake wilh sediment-water inleraclions. which can be rearranged to give
K,/(I - </J)p
However, now it should be nOled Ihat the total melals concentration in the sediment
cp = </J c" (43.24)
is written as
which can be substiluled inlo Eq. 43.14 to yield
C2 = Cd2 + Cp2 + cs2 (43.14)
(1 - </J)p
where Ihe Ihird form, Cs2,is metallhat has precipitated with hydrogen sulfide. In C2 = C" + </J K"c" + c., (43.25)
- . .ddition. as will shortly be made evident, the factor Fd2 is not a simple function
of sorption but will also depend on precipitation reactions. AII other parameters are Thus if pH, C2,and AVS are given. Eqs. 43.19 lo 43.22 and 43.25 represent
similar to those used in the conventional melals model. five equations Ihat can be solved for Ihe five unknowns. Although computer codes
Note that to simplify the folIowing development the sediment concentrations are available to do this (e.g., Dilks et al. 1995), Ihe presenl ( . ~imple enough

are aIl expressed on a molar basis (mole L-(). In addition we drop the subscripl 2 to be approached direclly. Firsl, as was done in the earlier Jc "qllilibrium
because the foIlowing derivation focuses solely on the sediments. chemistry (recalllhe derivalion of Eqs. 39.20 through 39.22), Eqs. 43. jl) alld 43.20
The chemical relationships goveming the sulfur compounds are can be substituted into Eq. 43.22 to solve for

H+ + S2- ~HS- sulfur ion reaction (43.15) AVS - Cs K¡[Wj

[HS-] = FHs (43.26)
H+ + HS- ~H2S hydrogen sulfide reaction (43.16) </J where FHS = 1 + K,[H+] + K1K2[H+j2 ~ I

AVS - Cs l
MS + H+ ~M2+ + HS- melal dissolutionlprecipilation reaclion (43.17) [S2-] - Fs (43.27)
- </J where Fs = 1 + K, [H+] + K, K2[H+ F
NOleIhat the dissolved metal ion. M2+. can be related to the terms in Eq. 43.14 by \1
AVS - Cs K¡K2[H+)2
[H2S] = FH2S cP = 1+ K¡[Ht] + K1K2IH"¡ f
(43.20)
[M2+ j =~ (43.18) whae FH2S

where division by porosity is nceded to convert the concentration lo a pore-water Equation 43.21 can be solved for
volume I>asis. Ks</J[W]
Now assuming thatwe know the pH, Eqs. 43.15 to 43.17 have five unknowns: (43.29)
C" = [HS-]
rS2-]. [HS-], [H2S].[MS], and [M2+]. Therefore five independenl equations are re-
quired. Three are supplied by the equilibrium re1ationshipsfor Eqs. 43.15 to 43.17, and the result, along with Eq. 43.26, can be substituted into Eq. 43.25 to give
 ~ .. ~-
g

766 PARTVII Toxics o" I.ECruRI!43 RadionuclláesándMebiTh 767

_ Kd(1 - ,p)p Ks[W],p2 = 2.14X \O-15(0.8)(lb-7) = 5. 13 X 1()-19

(43.30)
C2 - [ ,p + 1] FHS(AVS _ cs) + Cs Cd 3.34 X 10-4 '"
which is negligible.
which can be solved as a roots-of-equations problem for c". This value can then be
The calculation can then be repeated for the case where C2 = 0.002.tn thisInstancE!
used in conjunction with Eqs. 43.26 to 43.29 to compute the other four unknowns. c,. is fixed at the level of the AVS (= 0.001) and Eq. 43.25 can be solved ror
Now the foregoing solution holds up to the point where C2exceeds the A VS. At
this juncture the sulfur available for precipitation wil\ be exhausted and Cswil\ be- C2- c,. 0.002- 0.001 = 1.27x \0-7
Cd =
come fixed at A VS. Thereafter the five simultaneous equations reduce to four equa- (1- </J)pK¡ 1+ (1 - 0.8)25
__ X \060.0125
lions that can be solved for the remaining unknowns. I + ;¡;-- ,
The waler-quality implications of increasing the metals beyond the A VS are Thus. because the excess metal is not being precipitated, the pore-water levels increase
grea!. First, it means that the excess metal, C2- A VS, wil\ not be fixed in precip- over 11 orders of magnitude. Figure E43.2 shows how the porewater concentration
itates and, depending on the metal's sorption characleristics, a significant portion changes over a range of total metals concentration.
could build up in the pore waters. Beyond increasing the direct exposure to benthic
organisms, higher pore water levels also means increased mobility as the dissolved
metal becomes subject to diffusion. The foregoing model can be combined with a water balance to relate environ-
EXAMPLE 43.3. PRECIPITATION AND METAL LEVELS IN THE SEDI- mentallevels back to loadings. For example, Dilks et al. tl995) integrated it into
MENTS. A sediment has a pH of 7. a total lead concentration of 0.0005 mol L-1, a stream model (Box 43.1). It can also be applied to more than one metal. In such
.,""1
and an AVS of 0.001 mol L -l. Determine the dissolved lead concentration for this cases the metals precipitate in sequence depending on their solubility products-
case. In addition calculate the levels if the total metal concentration is increased to the ones with the lowest solubility precipitating firs!. As reported by Di Toro et ar .
0.002 mol L -l. Use the following parameters in your calculations: Ks = 2.14 X \0-15, (1992), the sequence is
Kd = 0.0125 m3 g-I, </J= 0.8, KI = 6.58 X 10'2, and K2 = 8.73 X 106.
-log(Ks) precipitates
SoIution: Parameter values can be substituted into Eq. 43.26 that can be solved for HgS 38.50 first
F HS = 0.5339. This value can be substituted into Eq. 43.30 along with other parameters CuS 22.19
togive PbS 14.67
2.14 X 1O-15(10-7)(0.W CdS 14.10
_ 0.0125(1- 0.8)2.5x 106 1
C2 - 0.8 + ] 0.5339(0.001 _ cs) + c,. ZnS 9.64 j
[ NiS 9.23 last
which can be solved numerically for c,. = 0.0005 mol L -l. Thus almost all the lead
precipitates with sulfur, This result can be substituted into Eq. 43.26 to calculate This sequence continues unlillhe A VS is exhausted. As described in Box 43.1, this
can lead to some interesting results.
[HS-] = 0.5339°.001 ~.~.0005 = 3.34 x 10-4
which can be substituted into Eq. 43.29 to give
BOX 43.1. Multiple Metals In Stream Sediments

Dave Dilks, Joseph Helfand, and Vic Bierman of Limno-Tech have applied the metals-
Total metal concentratlon, C2(mal L-') A VS model to streams. In addition they have included competition among five metais
0.0010 0.0100 (Ni, Zn, Cd, Pb, and Cu) for the available sulfide.
0.0001
O Figure B43.1 shows two simulations they made for a stream subject to á point
AVS di source of copper and cadmium. At river mile 1, a point source enters the streant. For
the first simulation the metal discharge is low and, hence, the standards for both metals
-5+ Cs are not violated. Because it precipitates first, the copper levels are lower.
... I In the second simulation the copper discharge is increased. As would be expected
~ -10 ./ cd the copper in the sediment increases; but it does not exceed the standard. In contrast
c2<AVS the cadmium increases to the point that a violation occurs. Why did cadmium increase
-15 when more copper was added? By increasing the copper the A VS was exceeded. Sul-
fides that formerly were avai1able to bind cadmium instead precipitate as copper sul-
-20 FIGURE E43.2 fideo Thus the liberated cadmium causes an increase in the pore water.
 ,.~
,.-,.'. "..~

768 PAKI'VII Toxics

LECTURE 44
J --"".- &-_-
Rlv_
(a) Cu standard
Cd standard
Toxicant Modeling in Flowing Waters
Cd
- 1x1o-e

!. 1x10-"
............. .............
Cu
GI
! (6)
.r.:r.:'.~.~.': .~.~ ... ...~.:'. :.~
.:
1x1o-e
............
.1
1x10-1Z I I I I I
O 1 2
Rlver mlle FIG. 643.1

PAOBLEMS
LECfU~ OVERVIEW: We cover some material related to the modeUng of Joxic~nts
43.1. Repeal Example 43.1, bul for 9OSr. Compule the inflow concentralion for 90Sr based on in Qowing waters. !11the first part we apply the lake p10deJs from the previQus
Table 43.4 and the factor (57.9) from Example 43.1.
lectures tOJivers and streams. We first describe stream solid's budgets and in-
43.2. Repeat Example 43.2 for copper. According lo Milis el al. (1985), copper has a smaller troduce tJte notion that sediment-water interactions are simpUfied because tbe
partilion coefficienl (KJ = 6 x 10" L kg-I) Ihan zinc. sedifJl~1}fco¡npat1mept is stationary. We then develop a contaminant balance
.' . "3.3. Delermine how much the zinc inflow concentralion lo Lake Michigan musl be reduced ailct ~c:~9rjbe some ~il1!ple sll';"I\'-~tate solutioQs for the mixillg zon!'. Next we
lo altain a sedimenl concenlralion of 5 J.Lgg-I. Whal is Ihe resuIting water concentra- P!!rfQtm ~ simHar~t'Hdysis H.. I!Hlries. We then develop a general cQntrol-
tior? volume approach that is applicable to both streams and estuaries. Along with
pme-variable solutions, we illustrate how the steady-state syslem respons!! ma-
43.4. Repeal Example 43.3, bul for cadmium (Iog K. = 14.1).
trix can be applied to toxics.
~3.S. Using the parameler values from Examples 43.1 and 43.2, (a) delermine Ihe inflow
concentralion of copper lo Lake Michigan lo altain a water concentralion of O.1 J.LgL -l.
(b) Whal is the resulting sedimenl concenlration in J.Lgg-I?
We now inlegrale horizonlal lransport inlo ollr loxics modeling schemc. Our t:IH-
phasis in this lecture will be on one-dimensional streams and eSluaries. Howevcr,
the control-volume approach described in a lalcr section is generally applicable 10
multidimensional water bodies.

44.1 ANALYTlCALSOLUTIONS
Before developing numerical approaches,l'lI tirsl describe some analylical schel1les.
As in other parts of this text, such solutions provide insight and can be used 101'quick
back-of-the-envelope estimation. 1'11firsl describe nondispersive, plug-flow syslems
Ihat are applicable to many streams. Then l' 11add dispersion to broaden Ihe potential
applications 10encompass one-dimensional estuaries and rivers, where dispersion is
significan!.

7(1)
 LEcruRE 44 Toxicant Modeling in Flowing Waters 771
770 !'¡\RTVII Toxics

mi
44.1.1 Plug-Flow Systems
As with the lake modelspresentedin Lec.40. we deve10pboth solidsandcontami-
nant balances for plug-f1owrivers and streams.

Solids budget. In a fashion similar to our lake model (recall Seco40.4.1), a v =0

m,(O) .. .. .. -. .. .. .. .. -. -- .. .. ______:_ -. _. __. FIGURE 44.1
steady-state solids budget can be written for a plug-f1owsystem with constant hy- Suspended solids
drogcol11ctriccharacteristics as versus dislance lor the
dml v, Vr v. < O case where sediment
0= -U- - ~ml + -m2 (44.1) solids concentration is
dx HI HI )( constan!.

and for the botlom sediments as

o = vsml- vrm2 - Vbm2
(44.2) · V" < O. In deeper streams,
accumulation of sediments.
sediments will be deposited and there will be a net
Thus the water solids decline as the sediment solids
build up.
where U = streamve10city(md-I) · V" > O. During certain periuds of time, scour will occur and there will be a net
11I1and /112 = suspended solids in the water (1) and sediment (2) layers (g m-3)
H = depth (m) loss of bottom sediments. Therefore suspended solids concentration may increase _,..1
in the water without external sources such as loadings, tributaries, or nonpoint", .
1'".Vr. and Vb = settling, resuspension, and burial ve10cities(m d-I) runoff of solids.
Now a key insight results from recognizing that because ~hesediments do not
Delos et al. (1984) and O'Connor (1985) have looked at the case where sediment
move horizontally (Eq. 44.2 has no advection term), Eq. 44.2 is identical to the sed-
solids concentration is a constant over the study stretch, m2 = (1- f/J)p. For this case
iment balance for a lake. Thus it can be solved for the solution for the water is
~ml (44.3) _-'Lx
= ml(O)e_-'L, ~'r(l - f/J)p
m2 = ~'r+ Vr

In other words the sediment concentration wiII always be a constant fraction of the
mi H,U +
(1- e HIU
) (44.7)

If the initial solids concentration m I(O)is small, then the downstream solids concen-
concentrationin the overlyingwater.As we will see shortly,this also applies to tration should approach a steady value of
steady-state contaminant budgets and has beneficial ramifications for time-variable
computations. ml(oo) = vr(l -
V.,
f/J)p (44.8)
Equation 44.3 can be substituted into Eq. 44. 1,
dml v, Vr Vs 8ased on this result, O'Connor (1985) suggested that solid profiles could be used to
0= -u- - ~/III + --mi (44.4)
dx HI HI Vr + Vb
estimate Vrby extrapolating downstream to a stable value of mi (00).We explore this
idea in Probo 44. I at the end of this lecture.
or by collecting terms,
Contaminant budget. Now we extend the analysis to toxics. To do this we

O = -U dx
dml _ HI
~ml (44.5)
/
assume that the suspended solids are constant across the stretch of interest. Ir this is
true a steady-state contaminant budget can be written for a plug-f1ow system with
constant hydrogeometric characteristics as
where VII= the net setlling velocity,
VII = vs(1 - Fr) (44.6) dc, Vv V, Vd Vr
O = -U dx - k1cI - H, F'¡ICI - HI FplCI + HI (Fd2C2 - FdICI) + HI C2
in which Fr = ratio of the resuspension velocity to the total purging velocity for the
sediment layer = Vr/(vr + Vb). (44.9)
Three general cases can occur (Fig. 44.1): and for the bottom sediments as
. VII = O. In shallow streams, there is often a negligible accumulation of sediments.
0= V..FpICI + V,¡(FdICI - F'¡2C2)- k2H2C2 - VrC2- VbC2 (44.10)
Therefore water solids remain constant as resuspension balances settling.
 772 PARTVII Toxics urWRE 44 Tuxicant Modeling in Fluwing Walcr~ 773

where k = a first-order decomposition rate (d-I) and the Fs are fraction of contam- Assuming instantaneous mixing at the injection poinl, we would compute Ihe
inant in (d)issolved and (p)articulate form in the two layers, initial concentralion with a simple mass balance as
1 Kdlm 1
CI(O) = QrClr
Qr
+ QwClw
+ Qw
(44.20)
Fd = 1 + Kdim Fpl = 1 + Kdlm Fd2 = cp+ Kd2(1 - cp)P (44.11)
Again, because the bed does not advect downstream, Eq. 44.10 establishes a where the subscripts w and r designate the waste outfaIl and the receiving river,
direct relationship between the sediment and the overlying water concentration, respectively.
Toxicant water-quality standards are often expressed in terms of a mass-specific
_ _ v.FI'1 + VdFdl sediment concentration. The initial value for the sediments can be determined as
C2 - R21CI - vd Fd2 + k 2H2 + Vr + Vbcl (44.12)
. R21 QrClr + QwClw
or in terms of a mass-specific sediment concentration as V2 (O) = (44.21)
(1 - cp)p Qr + Qw
R21
(44.13) Now these can be applied to determine the required loading to meellh.. '.Iandard.
V2 = (1 _ cp)pCI One way to do Ihis is to calculale the required waste concentration to attain a desired
Equation 44.12 can then be substituted into Eq. 44.9, water concentration CI(O),

dCI Vr Qr + Qw Qr
0= -u- - -CI (44.14) Clw = /> CI(O) - - ('I.r (44.22)
dx HI w Qw

where Vr = a totalloss term expressed as a settling velocity (m d-I), Alternatively we could calculale the required waste concenlration lo attain a desired
sediment concenlration V2(0),
Vr = klHI + VvFdl+ (v,Fpl + VdF,II)(1- F;) (44.15)
Clw -
_ Qr -
+ Qw (1 cp)pV2(O) _ -Qr elr
. (44.23)
I !

in whichF; = the ratio of sediment feedback to total sedimem purging, Qw R21 Qw

II
11
F' = Vr+ VdFd2 (44.16) EXAMPLE 44.1. POINT-SOURCE ANALYSIS (PLUG-FLOW). A loxic poinl
r Vr+ Vb+ VdFd2+ k2H2 source discharges lo a slream having Ihe following characlerislics:
11

L' Given a boundary condition of CI = CI(O),we can solve the water balance for Q, = 0.99 m3 S-I CI" = O mg m-3 Qw = 0.01 m3 S-I C!,w = I(~)O IlIg m
v, = 0.25md-' v, = 0.1md-I U = 0.1mps <p = O.l! !
t I

-~x I
CI = cl(O)e H,U (44.17) P = 2.5 x 1ü6g m-3 H. = 2 m
1I
In Ihe water, Ihe loxicanl is losl by phololysis al a rale of 0.1 d- '. AII olher los~es arc ¡
Then Eq. 44.12 can be used to compute the sediment concentration, zero and Ihe loxic associates slrongly wilh solid malter (all FJ's = O). Nole Ihallhe sus- I
I
- -!L.x ~nded solids in Ihe waler is a conslanl lO g m-3. (a) Delermine Ihe water and sedimenl
C2 = R2Icl(O)e H,U (44.18) concenlrations downSlream frum Ihe source. (b) Calculale Ihe needed inflow concentra-
lion so Ihat the maximum sediment concenlration is maintained at 250 ¡.¡.gg '.
or in terms of sediment solid concentration,
Solution: First, we have lo parameterize the solids budget. Because lb,' suspendcd
solids are conslant, we can assume that Ihe burial velocily is zero. Thcl ,,,'Ie seltling
V2 = R21 - Vr (44.19)
(1 _ cp)p cl(O)e H,Ux must balance resuspension,

Thus the water concentration foIlows a simple exponential decay. The sediment v,m, = v.1I - O)p
traces the identical shape, but is scaled as in Eq. 44.18 and 44.19. wluch can be ~ulved for

Critical concentration. Now an interesting caIculation involves computing the v, = v, (1 :~)p = 0.250= 0.l!)2.5x 1()6= 5 x 10-6m d
initial concentrations. They are particularly germane because the maximum value
The initial water concentration can be computed as
occurs at the outfall. Thus this concentration represents the critical value upon which
assimilative capacity calculations would be based (recall Lec 1, and particularly the . O - 0.99(0)+0.01(1000)- 10
discussion of Fig. 1.1). (I( ) - --0:99 + 0.01 -
774 PARTVII Toxics
LECTURE 44 Toxicant Modeling in Flowing Waters 775

The sediment-to-water concentration ratio can be computed as (note that because v,

Fd2 = O, k2 = O,and Vb = O, F; = 1) (44.26)
Fr = v, + vb
0.25
R21 = 5 X 10-6 = 50,000 For cases where sediment resuspension is negligible, the model simplifies further to
which can he used to determine
O= - lj dCI
-
dx
Fplv,
- --:"CI
HI
(44.27)
50,000 _J IOJ¡.tg
)
_1
~'2(0)= (1 _ 0.8)2.5 X 10610 mg m . ( = 1000 ¡.tg mg g
If the suspended solids are constant, the solution to this equation is merely
Afterdetermining that Vr = 0.1(2) = 0.2 m d-I , Ihe downslream profiles can then Eq. 44.17 with VT = Fpl v.. However, when solids are changing, the fraction par-
be compulcd with Eqs. 44.17 and 44.19, ticulate will also change with distance. For this case Milis et al. (1985) provide the
0.2 folIowing solution,
CI = 1Oe - 2(R64¡¡¡' V2 = IOOOe - 2(~~)'
and displayed in Fig. E44.1.
CI = clOe [ 11
'\ Kdmo+t/;:~ )-ln<Kdmo+ 1)- ;:~] (44.28)
I
10 1000
EXAMPLE 44.2. FLINT RIVER COPPER. Milis et al. (1985) present !he following

:s
-~ data for suspended solids and water copper concentrations in the Flint River, Michigan,
in August, 1981: ~ - ,~I
01
~ 5 ~
....
x (km) 1 2.5 7 12 21 30 43 61 63
.;' >
m(mgL-I) 11.75 10 10 8.5 6.75 5.5 11.5 13.5 11.75 i

(4-18) (5-15) (7-13.5) (6-11) (5.3-8) (5.3-5.7) (3-19.8)(10-16.5)

o O CI (lLg L -) 3 4.2 5.7 5.5 4.2 4.7 8 6 5.75
O 10 20 30 40 50 (2.8-3.4) (3.2-5.4) (4.5-7) (4.9-5.8) (2.4-4.8) (4-5.2) (4-10) (4.8-7.3) (4.5-7)
x (km)
FIGURE E44.1
where the numbers in parentheses represent observed ranges. The river has the following
boundary condilions and point sources:
(b) Equation 44.23 can be employed to compute

Suspended
= 0.99 + 0.01 (1 - 0.2)2.5 x 106025 _ 0.990= 250 L-1 FIow sollds
CI... 0.01 50,000 . 0.01 ¡.tg Copper
Kllometers (cros) (mg L -1) (¡.tg L -1)
Consequently the wasle concenlration must be reduced from 1000 to 250 ¡.tg L -1 lOmeel
Upslream boundary O 2.66 13.5 2.9
the standard.
Ainl WWTP 1.3 1.68 4.1 8.3
RagoneWWTP 30.9 0.69 58.7 28.5

Metals. For metals the model simplifies because of the omission of decay and
volatilization (with the exception of mercury) reactions. In addition, if it is assumed The depth and the velocity in the first stretch (km 1.3 to 30.9) are assumed to be 0.5
that sorption is the same in the water and sediments, sediment-water diffusion can m and 0.2 m S-l. (a) Use the suspended solids to determine the settling velocity. (b) 1f
be omitted and the steady-state model again can be represented by copper has a partition coefficient ofO.06 x 1Q6L kg-I, compute the copper contentrat¡on
in the stream.
dCI VT
0= -lj- - -CI (44.24) Solution: (a) First, inspection of the suspended solids data suggests !hat there is nel
dx HI
sedimentalion in the first stretch and zero in the second. The data for the first stretch
but with the total removal rate simplified to can be used to estimate the settling velocity by determining the slope of a semi-Iog plot
(44.25) of susPended solids versus distance. The resulting settling velocity is 0.213 ID d-I. This
VT = v.,F",(I - Fr)
value can be used lo simulate the suspended solids with the results shown in !he following
where p10t:
"""---~~...; c.. ~ "---'- ~ .~I- .3-'''1;''' ~.. _. . ~

U!CI"IIIW44 Toxicalll Modclillg in Flowing Walcrs 777

776 PARTVII Toxics

where the total removal velocity VT is defined as in Eq. 44.15, and the concenlra-
tion of the underlying sediments can he cümpuled as in Eqs. 44.12 and 44.13. The

. solulion can be computed as (reca)) Secolí.I.I)

CI_ = CI(O)eA-x x$O (44.30)

CI+ = cl(O)eA,X x>O (44.31 )

where
o Á- U
1 + 4VTE (44.32)
O 20 40 60 Á+ = 2E (I::!: HIU2 )
x (km) FIGURE E44.2-1
in which the subscripls "+" and "-" designale the system upslream and downstream
Note how the solids are reduced due to settling in the first stretch. A mas s balance is from Ihe source, respectively.
then used to determine the concentration jump at the confluence with the Ragone WWTP If it is assumed Ihal the waste ftow is much smaller than the estuary ftow,a mass
effluent. Thereafter the concentration is constant since net settiing is zero. balance at the outfall can be determined as
(b) For copper, Eq. 44.28 can be employed to simulate the water concentration for W
the first stretch. A mass balance is then used to determine the constan! concentration for CI(O) = - (44.33j
Q
the second stretch. The results along with the data are displayed below: 1 + 4vrE
HU2
and for the bottom sediments as
10
I
V2(0) = R21 W (44.34)
4vrE
1
1 + HU2
~ 5 The waste concentration required to meet a water standard can be computed in
-r
a similar fashion to Eq. 44.22. If it is assumed that the receiving water upstream of
the discharge has negligible toxic concentration, the result is (
. .-. O I

O 20 40 60 (44.35)
x (km) FIGURE E44.2-2 Clw = iw cl(O»)1 + 41J
I
where 1Jis the estuary number (recall Eq. 10.41), which in the present conlext would
In this section we have neglecterl rlistributed ftow increases along the stream. be defined as vrE/(H U2). Altematively we could ca1culate the required inftow con- i I
centration to attain a desired sediment concentration,
Although this can be a usefuI aSSUI", 11,many cases involve distributed ftow. ~

O' Connor ( 1988c) has addressed this issue for both organics and metals and has de-
veloped a number of use fui analytical solutions for some idealized cases. In the last Clw
-
-
JL
Qw
(1 - cP)pv2(0»)1
R21
+ 41J (44.36)

section of this lecture we present a numerical approach that allows such phenomena
to be incorporated into our framework. Thus we see thal the term )1 + 41Jreftects the effect of dispersion on the assimila-
tive capacity (compare Eq. 44.36 with 44.23).

44.1.~ Mixed.Flow Systems EXAMPLE 44.3. POINT-SOlJRCE ANALYSIS (MIXED-I<'l,OW). l{cpcal par1lv)

Now we extend the foregoing analysis to systems where dispersion cannot be ne- of Example 44.1, but assume that the system has an estuary number of l.

glected. These inelude estuaries and certain rivers. If suspended solids are constant, Solution: Substituling Ihe appropriate values inlo Eq. 44.36 gives
~ steady-state contaminant budgel can be written for a mixed-ftow syslem wilh con- ." I (1 - 0.8)2.5 X 106 ~ -1
st~nt ~ydrofeometric chlU"ucteristicsas Clw = 0.01 50.000 0.25 ,¡ l + 4(1) = 250(2.236)= 559¡LgL

. O = Ed2c1
dx2
_ UdC1
dx
_ Vr CI
H,
(44.29) Thus Ihe dispersive mixing means Ihal over Iwice as much loxicant can be discharged.
 778 PARTVII Toxics LEcrUIÚÍ ToJticant Modellng in t:lowltit \Vateif 3~

44.2 NUMERICAL SOLUTIONS Equation 44.41cári be wnttenfol' all thé water ~6lukes aftd ei~tes§ed ln ftti&1Jt
formatas(recatiEct. 11.26), u " ',~
Although the analytical solutions we developed are handy for obtaining back-of-
tAHc¡}= tW} t~U~)
the-envelope solutions, numerical approaches provide a more general approach. In
particular they provide a means to analyze multiple sources. and solved as
Mass balances can be developed for a toxicant in a control volume as (Pig. 44.2) {C.} = [Arl{W} (41.J3)
dCI'
Pinally it can be recognized that the sediment concentratlon can be detcrmirieJ as
VI,i d;' = W¡ + Q¡-I.;(ai-I,iCI,i-1 + 13¡-I,¡CI,i-¡)
1 · i' ·
- Q¡.i+1(ai.i+ICI.i + l3i,¡+1CI.i+1) + Et-l.i(CI,i-1 - Cl.i) {V2} =" h _ [R2JJ[A)-I{W} (44.44)
+ Et,i+1(c1,i+1 - CI,i) - kl,iVl,icl,i - vv,¡A..,¡Fdl.iCI.¡
or collecting terms,
- v..,¡A..,¡F"I.;cl.i
{V2} = [Srl{W} (44.45)
+ vr,¡A..,;c2,i+ Vd,iA...i(Fd2.;C2,i- Fdl,iCI,i) (44.37)
\. dC2 . (44.46)
where [Srl =" 1 h. [R2JJ[Arl
and V2,i dt" = -k2,iV2,ic2,i + V,.iA.,,¡Fpl,¡CI,i- vr,;A..,¡c2.¡
- ~'h,1 AC.t,' 2,1. + Vl,II A ,f,I. (Fdl ,1'C ItI. - FtI2,1'C 2 ,1.) (44.38) I
Thus, the matrix inverses [A]-I and [Sr I provide the vehic1es forsolving inverse as- I
These equations can be written for all the volumes in the system and then solved similativecapacity calculations for multidimensional systems with multiple sources. ...,
simultaneously to determine the distribution of contaminant lhroughout the system. In other words Eqs. 44.43 and 44.45 are the multidlmensional analogs of Eqs. 44.22- -
For lime-variable solutions the approaches oullined in Lees. 12 and 13 can be used. and44.23.
Por the steady-slate case, Eq. 44.38 can be solved for each volume in a similar
fashion to Eq. 44.12,
44.3 NONPOINT SOURCES
C2,i = R21.iCI,i (44.39)
or in matrix form as Although point sourcesare certainly important, many toxics enter streams and estli-
{c2} = IR21 J{c¡} (44.40)
aries in a nonpoint or diffuse fashion. Por example both urban and agricultural runoff
can carry toxics in significant concentrations.
where (R211 is a diagonal matrix. Equation 44.39 can be subsliluled inlo Eq. 44.37,
and the reslllt manipulated to yield
dcl.i 44.3.1 Low-Flow Nonpolnt Sources
VI ;--
. ti , = W; + Qi-I ."(
; a;_1 ;CI i-I + ai_1 iCI ;-1
1-'..
)
Pornonpointsourcesthatcontributenegligible flow,the watermassbalanceforthe
- Qi,i+1 (a;,¡+lcl.i + 13i.i+ICl.i+¡) + Et-l,i(cl,i-I - CI.i) plug-flow, conslant-parameter case becomes (compare with Eq. 44.14)
dCI Vr (44.47)
+ Et.;+I (cl.i+ I - CI.;) - Vr.iA.,,;CI,i (44.41) 0= -U---CI+Sd
dx HI
where I'r is calculated for each volume in a similar fashion to Eq. 44.15.
where the total removal velocity Vr is defined asinEq.44.15andSdis the distributed
loading terro (mg m-3 d-I). The closed-form solution is (recall Eq. 9.48)
-Jr...x
~
,t !,
~ CI = cl(O)e
-;,r,.,x
Hlu +
SdH.
~ (1- e NIU ) (44.48)
1,;-1 : 1,; : 1,;+1
The sediment concentrations follow directly from Eqs. 44.12 and 44.13.

:r-D
¡
,
,
~¡ 2,;
,
,
44.3.2 Flow-Contrlbutlng Nonpolnt Sources
: I " : FIGURE 44.2
Segmentation scheme lor control-volume
A1though nonpoint sourcescan contribute negligible flow, it is more likely that they
approach lor toxics. are accompanied by significant volumes of water. Por such casesan analysis similar
to that described previously in Seco22.3 can be developed.
 780 PARTVII Toxics LECTURE 44 Toxicanl Modeling in Flowing Waters 781

For example a ftow balance would be implemc¡;¡ , It ¡ rmine segment ftows 50

and geometries in a similar fashion to that employed in S"c J. Then a steady-state, I
(a)
control-volume approach can be applied 10the contaminant in each volume as
40
I
0= W¡ + Q¡-l.i(a¡-J,¡cJ,¡-J + #3¡-J,¡cJ,¡-.)- Q¡,i+1(a¡,i+ICl.i+ #3i.i+,CI,i+J)
+ EI-I,¡(Cl.i-I - cu) + EI,i+I (cl.i+I - CI,¡)- vT,¡As,¡cJ,¡+ QeCd.¡ (44.49)
where Qe = distributed inftow for segment i (m3 d-I) and Cd,¡= concentralionof ... 30
the contaminant in this inftow (mg m-3).
This equation can be written for the n elements of a reach. Together with ap-
20
propriate boundary conditions, it can be solved for the water concentralions in each
element. Then Eqs. 44.12 and 44.13 can be used to determine the sediment concen- l'
tration. The approach is ilIustrated in the foIlowing example.
10 11
EXAMPLE 44.4. FLOW-CONTRIBUTING DISTRIBUTED SOURCES OF I

TOXICS. A demonstration of the model can be developed for the case illustrated I
io Fig. 44.3. For Ibis simulaliuo a poiot source marks Ihe beginning of Ihe problem. For O
-10 O 10 20 30 40 50 ,1

3
I
(b)

/
Water
2 /lg L-1
c:
o

- ..' / 10 km
i
e
" r'Charmelslope fj
! Bottom wldth, m
0.002
25 I
. c:
o
U
IRoughness 0.035
!~onpolnt flow, m2 S-1 0.00025
L.~~__, ._____.j 1 1
Seds
""20 km
/lg g-1

0L-
-10 o 10 20 30 40 50
Downstream distance (m x 1000)

FIGURE 44.4
Simulation resulls. (a) Hydraulic variables; (b) concentrations.

Channel slope 0.0002 the next 10 km, Ihere is a constant nonpoint inUow of clean water. Then, inlhe stretcll
Bottom wldth, m 25 10 lo 30 km downstream from Ihe point source, runoff from a landlill adds additional
Roughness 0.035 toxicant in a diffuse manner. Finally, for the last 20 km, the slream again receives clean
Nonpolnt flow, m2 S-1 ~.0005 waler. Other model parameters are c/>= 0.9, p = 2.5 g cm-3, v, = 0.05 m d-I, Vr =
0.0006 m d-I, and Vb = 0.0002 m d-I.

FIGURE 44.3
Solutloo: The results of Ihe simulalion are shown in Fig. 44.4. Water and sediment con-
centrations are expressed in J.LgL -1 and J.Lgg-I, respectively. Two contaminant decay
Example problem.
 782 PART VII Toxics
LECfURE44 Toxicant Modeling in Flowing Waters 783

!f scenarios are ineluded. In the first the contaminants are assumed to be conservative (k = In a similar fashion in Example 44.2, simulate the distribution of zinc in the Flint Rivér.
~. O) in both the water and the sediments. In the second the contaminant is allowed to decay Use a sorption coefficient of 0.2 X 1ü6 L kg-I for your calculation.
in the water at arate of 2 d-I . Such might be the case if the substance were subject to
volatilization. 44.3. A point source of a toxic substance (Kow = 1(6) is discharged to an estuary having the
following characteristics:
As can be seen, the concentrations abruptly increase to high levels due to the point
source. Thereafter they decrease due to dilution and, in the case of the second scenario,
due to decay. The concentrations begin to rise as the water f10ws past lhe nonpoint souree. Value Vnits
Once past the landfill it again decreases in response to dilution and decay. As expected Dispersion coeflicient 100x 1()6 m! d-I
from Eq. 44.13, the sedimellts follow Ihe same pattern as the water. The addition of decay Aow 15x 10-' m) d-I
lo the water leads to depletion of both water and sediment concentrations. Width 75 m
Depth 10 m

The foregoing analysis provides a model that could be useful in assessing the
impact of flow-contributing diffuse sources on steady-state, stream water quality. It
should be particularly useful in simulating the impact of chronic nonpoint sources,
such as landfills and contaminaled groundwater, Onreceiving waters. It also could
have utility in evaluating the long-Ierm integraled effecl of urban and agricultural
nmoff.
In concIusion this leclure has been intended as an introduction lo Ihe modeling
of toxicant dynamics in flowing waters. The main point of the leclure is !hat for the
sleady-state case, the sediments map directly on the overlying water concentration.
This result is due to the fact that the sediments are stationary.
Clearly there are problems that require more sophisticated approaches than were
discussed here. In particular, time-variable cases (such as stormwater overflows)and
Ihe inelusion of toxicants in slream and esluary eutrophication frameworks represent
I1loreadvanced applications.
PROBLEMS
The estuary has a suspended solids concentration of 5 mg L -1 with an f oc = 5%.
These solids settle at arate of 0.25 m d-I. The bottom sediments have the following
characteristics: thickness = 10 cm, burlal velocity = 0.825 mm yr-I, resuspension
velocity = 1mmyr-I, porosity= 0.9,density = 2.5 x 1ü6g m-3. Thereare noother
loss mechanisms other than a volatilization rate of 0.2 m d-I .
What mass loading of a pollutant could be input to this system under steady-state
conditions if the maximum allowable sediment concentration at the outfall is I par! "
per million (that is, I p.g g-I)? Express your results in kg yr-I. Assume complete--
". and instantaneous lateral and vertical mixing at the outfall. Also assume that diffusive
sediment-water interactions are negligible.
44.4. A stream passes by a toxic landfill for an 8-km stretch. The landfillleaches a toxicant
at arate of 50 mg per meter of stream length per day. The toxicant has ah bCtilttol-
water partition coefficient of lOS, a molecular weight of 300, and is overwhelmingly
liquid-film controlled. The stream has about I mg L -1 of organic carbon !hat is carried
in suspension. It also has a velocity of 0.1 m S-I , a depth of 2 m, and a width of 30 m.
If sorption is instantaneous, calculate the concentration in the water for a distance of
16km.

44.1. A river downstream to the conf1uence with a tributary manifests the following sus-
pended solids concentralion during 10w-f1ow conditions. If il is assumed that the solids
settle al a rale of 0.25 m d-I, use this data to determine the resuspension velocity in
the stretch. Other relevant parameters are 4> = 0.8, p = 2.5 X 106 g m-3, HI = 1m,
and U = O.1 mps.

Distance (m) I O 20 40 60 80 100

Suspended solids (mg L -1) 10 6 5 3.5 2.5 2.5
44.2. In addition to copper, Milis et al. (1985) also present data on zinc for the Flint River.
The river has the following boundary condilions and point sources:

Zinc
Kilometers (lLg L -1)
Upstream boundary o 7.7
Flint WWTP 1.3 55
Ragone WWTP 30.9 84

".
 ~ -~.., ...

lECTURE45 ToxicantlFood-Chain Interaclions 785

LECTURE 45 food. Biomagni(il'l/Iion is the process wherein bioaccumulalion ¡ncreases along the
food chain. Thc wing sections review modeling frameworks that have been de-
veloped lo quantify Ihese processes.
ToxicantlFood-Chain
Interactions 45.1 DIRECT UPTAKE (BIOCONCENTRATION)
/Before writing equations we should first detennine an appropriate way to quantil'y
the toxicant within an organismo In prior models we have normalized mass either
to volume or to mass. The fonner was appropriate for water concentration, whereas
the latler was employed for the sedimenls. Clearly the mass-specific approach is
preferable for organisms.
However, now Ihe question arises: What mass do we normalize In? In early
toxicant-organism models, the mass of organisms was expressed on a wet-weight
basis. This is in contrast to particles thal, as we've seen in previous lectures, are ex-
pressed as dry weight. This pcrspective has been further refined for organic chem-
icals. For these toxicants it has been recognized that the association is governed by
the quantity of lipids in the organism (Mackay 1982, Connolly and Pedersen 1988,
and Thomann 1989). Thus, one way to quantify the concentration of a conlaminanl
in an organism is

IJ..~.oR E QVE~VlfW l J. describe !Ww ~oxicants interact wilh ofganisrnS' TJ1iS. ¡,,-
v= ?: (45.1)
.
Jrt1~U~p.'~k~s
. . .
. .
two prim¡iry fonns. Fiest, the organism cal1dtrecf}y
. . t*~
.
. . úp
. .. .
where v = organism toxicant concentration on a lipid basis (¡.tg-chemical kg-
lipid- J)
IV.ij~~lxe<t J9~ic~~tfrpm the w"ter. Second, the orga'11smcan CQI1SUme 9oQt~m-
'mat~af9Pd., also show how both types of inter¡1ction can be integrated jnto a VWI = organismtoxicanlconcenlrationon a wet-weightbasis(¡.tg-chemic¡.I
IlY'l!h mq~~.I>!!lal1ce framewQrk to relate food-chain Jevels b¡lck W loadings. kg-weCI)
-"'... ; !L = fraction lipid (kg-lipid kg-weCI)
..,
Bioconcentration is often measured by placing an organism into a vessel where
To this point our models provide a way to predict toxicant concentration in the wa- the concentration of a contaminant is kept at a constant level. The process can be
ter and the sediments. Such information is certainly valuable because water-quality modeled by writing a mass balance around an individual organismoAs in Fig. 45.1,
standards are often expressed in terms of concentralion. For example il makes sense the primary processes are (1) direct uplake by diffusive exchange across organism
to express drinking water standards as concentralion in the water. membranes and (2) losses,
However, toxics pose addilional threats that extend beyond human consumption e/VIII
of drinking water. Fírst, the loxic can reach harmful levels in organisms that serve di = kuWL(cd - cn) - Kv", (45.2)
as human food, such as fish and shellfish. Second, the toxic can grow lo levels that Accumulation Uptake Losses
inlerfere with the funclioning of Ihe ecosystem ¡Iself.Although both examples could
be related back to a water-quality standard for concentration, a more direct approach where VIII = loxicant whole-body burden (¡.tg-toxic per organism)
would be to assess the concentration ofthe toxicanl within the organisms Ihemselves. ku = chemical uptake rate (L kg-lipid- J d-I)
This lecture is devoled to inlroducing you lo how Ihis can be done. We first wL = organism lipid weight (kg-lipid)
examine how individual organisms lake up and purge contaminants direclly from Cd = "freely available" dissolved chemical in Ihe surface 01'porc walers (J.lI~
the water. Then we add food-chain transfers to show how organisms are impacted L-1)
by Ihe consumplion of conlaminaled food. Cn = "free" concentration of chemical in the organism's blood (lLg L 1)
Before proceeding we should introduce some nomenclature. Bioconcenlra- K = losses (that is, losses from skin surface, excretion, and chernical
lion refees to the direct uptake of a toxicant through the gills and epilhelial tissues metabolisrn) (d-I) ~
(skin) of an aquatic organismoBioaccumulalion is a more comprehensive term re- Next, if it is assumed that the partitioning between Ihe tissue concentrati()nand
ferring to direct bioconcenlration from the water as well as ingestion of contaminated the blood concentration is proportional to Ihe octanol-waler partition coefficient, we
can define
7114
...

" 786 PART VII Toxics

LECruRE45 ToxicantIFood-Chain Interactions 787

a "dilution" effect because the increase means that the mass of contaminant wi\l be
normalized to a large biomass (and lipid content).
Equation 45.8 can be simplified further as
dv , k (45.10)
dt + K v = uCd
FIGURE 45.1 where K' = the sum of the membrane losses, the other losses, and the growth dilution
A simplilied schemalic 01 an experimenl to delermine rates,
bioconcenlration. Inpuls and outpuls are shown lo indicate
k
Ihat the dissolved loxicant level in the eSTR is kept at a K' = - u +K+G (45.11)
constanllevel. Knw

v For zero K and G, this result suggests that the totallosses would be inversely pro-
- = Nw/ "" Knw (45.3) portional to the octanol-water partition coefficient. That is, higher Kowcompounds
Cs would be retained better than lower ones.
where Nw/ = a lipid concentration to blood concentration partition coefficient Assuming that the organism is uncontaminated at the beginning of the experi-
(1, kg-lipid-I). This resu1tcan be substiluted into Eq. 45.2 to give ment (v = O at t = O), we can solve Eq. 45.10 for
I
dVm V ku K',
dt = kuWL ( - )-
Cd Kow KVm
(45.4) V = K' CdO - e- )
Thus the mass-specific concentration starts at zero and asymptotically
(45.12)
approaches a
"." '1
The whole body burden is rclaled to the chemical concentration by steady-state concentration of
Vm = VWL (45.5) ku
(45.13)
which can be subslituted into Eq. 45.4 to give v = K,Cd
d(vwd V Note that this steady-state level can be expressed as
;¡¡- + kuWL ( - )-
Cd Kow KVm
(45.6)
Nw =~ = ku = ku (45.14)
Ir the organism's lipid mass is constant, we can take it outside of the differential.
Cd K' kuIKow+K+G
However, for organisms such as fish, this is not a good assumption. Therefore, we . ¡.tg-contaminant
kg-lipid-I = L-water
tnl!st expand the differential, umts: . ;-;-
[ ¡.tg-contammant
L-water-I kg-hpld]
v-dWL
dt
dv
+ WL- d =
t
k"WL
( Cd - _Kow)-
V
KVm (45.7)
where Nw = a bioconc~ntration factor (BCF). This representation is directly analo-
01'rearranging lerms, gous to a partition coefficient. However, rather than representing the assoeiation of
a toxicant with suspended solids, it parameterizes the association of a contaminant
dv V
with an organismo
WI. dI
Accumulation
= k"WL ( - )-
Cd
Uptake
Kow KWLV

Losses
- GWLV

Growth
(45.8) Note that an interesting case results when losses and growth are zero. For this
case Eq. 45.14 reduces to
dilut;o"
Nw = Kow (45.15)
where G = growth rate ofthe organism (d-I),
EXAMPLE 45.1. 8IOCONCENTRATION. A zooplankton is placed in an aquarium
G = dwtldt with a contaminant concentration of 1 [.LgL -l. Compute its steady-stale concentration
(45.9)
Wt given the following parameters: Knw = 105,ku = 31.000L kg-lipid-Id-I, G = 0.15
d-I, and K = 0.01 d-I.
The differenlial equation is now written to determine the temporal changes of
1/11'mass-specilic cOllcenlralion(ralher than the conlaminant mass as in Eq. 45.2). 80lution: The organism 's totalloss rate can be computed as (Eq. 45.11)
Consequenlly, along with uplake amI depuration, the change is also dictated by
(he rale al which Ihe organislll gains biomass. This latter erfect can be Ihought of as K' = 31,':> + 0.01 + 0.15 = 0.47 d-I
 ,, ,~

788 PARTVII Toxics LECfURE45 ToxicantlFood-Chain Inleracli"lIs 789

and Ihe BCF can be ca1culaled wilh Eq. 45.14, dV2 ,

dt = ku2Cd+ a21h2vI - K2v2 (45.17)
Nw = 3~:~ = 65.957L kg-lipid-I where a j,k = chemical assimilation efficiency (p.g-chemical absorbed per ¡.tg-
Finally Ihe organism's concenlralion can be delermined as
chemical ingesled) for jth predalor feeding on klh prey and h2 = Ihe IipiJ-specific
consumption rate (kg-lipid prey per kg-lipid predator per d).
Al sleady-slale, Eq. 45.16 can be solved for
v = NwCd = 65.957 L kg-lipid-I (IIL~) = 65.957 ILgkg-lipid-I
kU.1
VI = K;Cd (45.18)
or, as a waler bioconcenlralion faclor,
45.2 FOOD-CHAIN MODEL (BIOACCUMULATION)
k/ll
Nlw = VI (45.19)
Although bioconCenl¡. ,n faClorsare cerlainly useful, they ignore a second way in C,¡ K¡ {
which organisms can take up toxic subslances, namely by ingesling conlaminated
For Eq. 45. 17, Ihe sleady-slale solution is
prey.
A number of possible mathemalical formulalions characlerize inleractions V2 = N2wCd + g21VI (45.20)
among the elements of the food chain. As described in Lec. 35 fairly elaborale
where
nonlinear models are Iypically used lo simulate seasonal dynamics 01' biomass.
However, for long-Ierm compulalions, these seasonal models are coslly lo imple- N2 = V2 = ku2 (45.21)
meni and available field data for loxics rarely support Ihe use of sllch fine-scaled IV C,¡ K2
frameworks. One altemalive is lo inlegrale a seasonal model over the annual cycle to and g21 is a dimensionless food-chain multiplier,
eSlimate Iransfer rates for a long-Ierm food-chain model. Hydroscience (1973)'used
a21 h2
such an approach lo calibrale a linear, donor-dependent model for cadmium in wesl- (45.22)
g21=~ 2
em Lake Erie. A second allemalive is to direclly develop a long-Ierm model based
on conslant average annual rales. The following descriplion follows Ihis approach NOleIhal Eq. 45.18 can be substituled inlOEq. 45.20 so thal Ihe level in Ihe zoo-
and is derived primarily flOm Ihe research 01'Thomann and Connolly (Thomann planklon is directly related lo Ihe waler concentration, as in J
1981, 1989; Connolly 1991; i'l1omannel al. 1992; Thomann and Connolly 1992). (45.23)
- ",- As a first approximalion, Ihe pelagic food chain can be Llividedinlo four com- V2 = N2Cd
l."

partments (Fig. 45.2). The phyloplanklon are Irealed in a fashion similar lo other where N2 = overall bioaccumulation faClorfor Ihe zooplankton,
small particles. That is, membrane exchange is omitted and Ihe following balance (45.24)
reslllts: N2 = N2w+ g2lNIIV
dVI NOlice Ihal for phyloplanklon,
(45.16)
di = kulCd - Klvl NI = NIIV (45.25)
where Ihe subscript l denotes Ihe phytoplanklon. since Ihese organisms do nol receive contaminani by grazing, whereas IÚrLOoplank.
for the higher organisms, a balance accounting for membrane exchange as well Ion,
a~ grazing gains from Ihe phyloplankton can be written as
N2 'i' N2w (45.2(),

Similar equalions can be developed for Ihe olher compartmenls 01'the f(H,LI
chain.
For example Ihe small fish can be represented as
N3 = N3w + g32N21V + g32g21NI (45.211

and Ihe large fish as

N4 = N4w + g43N31V + g43g32N21V + g43g32g21NI (45.28) ~
FIGURE 45.2
A simplilied representation 01 a pelagic lood chain originally proposed by A generalized rclalionship can be expressed as
Thomann (1981). Note that Ihe lerm pelaglc relers lo Ihe open-water region
beyond Ihe inlluence 01the shore and bottom.
790 PART VII Toxics LECTURE4S ToxicantIFood-Chain Inleractions 791

N" = N"w + ~[(

)=1
,11
,=)+1
g¡.i-I
)
NiW
]
(45.29)

...
where N" is Ihe bioaccumulation factor (BAF) for the IIlh organismo 0.5
-.;...

45.3 PARAMETER ESTIMATION

O
In Ihe previous rramework there are chemical and organism parameters. We now 2 4 6 8
describe how each of these is eslimaled. We then apply Ihem lo compule the con- logK...
cenlration of a toxicanl in a natural water.
FIGURE 45.3
Ralio 01chemical lo oxygen transler efficiency (EcIE.) versus
45.3.1 Chemical Parameters oclanol-waler partition coefficient. The chemical transler
efficiency (a) is assumed lo follow Ihe same lunclion.
Uptake. The chemical uplake rate depends on the organism's respiralion rale
and the efliciency 01 transrer across the organism membrane. Connolly (1991) has As displayed in Fig. 45.3, the ef6ciency increases, leve1s off, and Ihen deereases
suggested with Kow.
.J
k" = II¡(' r",{/edP E, (45.30)
{/w,1JiP", E" Membrane and excretion losses. Using Eq. 45. 11, the membrane and exere-
tion losses are de6ned as
where r"e = oxygen-to-earbon ratio (= 2.67 kgO kgC-1)
(/,d = carbon-to-dry-weight ratio (= 0.4 kgC kg-dry weighC1) ku[Ec(Kow), p] + K (45.32)
Kow
P = organism oxygen respiration rate (d-I)
a "',1 = ralioof wet-to-dry-weight (kg-wet kg-dry-I) where Ihe nomenclalure ku[Ec(Kow), p] denotes Ihat the uplake is eomputed with
0", = oxygen eoneentration in the water (mg L -1) Eq. 45.30.
Ee = chemical transfer efficiency
E" = oxygen transfer efficiency for the organism Chemical assimilation efficiency. The assimilation of ehemical tbrough food
ingeslion should be similar lo Ihe faelors goveming gill membrane Iransport. Con-
Note thal Ihe 10(' is included so thal Ihe unils of k" work oul lo L kg-I. sequenlly Thomann and Connolly (1992) have assumed that a is a funclion of Kow.
It has been suggesled Ihal Ihe chemical Iransfer efficiency depends on a num- As a 6rsl approximation Ihey employed Ihe same funelion as in Fig. 45.3.
ber of factors. inclllding Ihe lipid parlition coefficient. moleclllar weight, ete. (see
McKim et al. 1985 for a review). Thomann (1989) has developed some piecewise
[unctions to compule the chemicallransfer efficiency basedon the octanol-water par- 45.3.2 Organism Parameters
t¡tion coefficient and organism size. Thomann and Connolly (1992) have developed
a simplilied vcrsion rÓr the ratio EJE" that is independent of organism size. It can
Growth and respiration. Bolh organism growlh and respiralion rales can be
he represented mathematically as deve10ped lo a 6rsl approximation by correlation wilh organism weight. Thomann
(1989) has proposed Ihe following relalionships for growlh,
E" G = 0.OO2512w-O.2 (45.33)
log - = -2.6 + 0.510gK",\, 2 S 10gK"", S 4
E"
E" and for respiration,
- = -3.339 + 0.8976IogK,,1I' 4 S logKow s 4.5 p = 0.OO9043w-O.2 (45.34)
E"
E. = 0.7
where w = organism weight (kg-wet weight).
4.5 s logK"". s 6.5 (45.31)
E"
Ee Consumption rateo Aecording to Connolly (1991), the organism's energy us-
E"
= 3.3 - O.4log K"... 6.5 S logK"w s 8 age rate can be computed, as in
Ee Pi = Á¡(p¡+ G¡) (45.35)
log-
T:"
= 7 - log K"", 8 S 10gK"". s 9 where Pi = energy usage rate of organism ¡(cal kg-wet-I d-I) and A¡ = calorie

.r ...
 ... ..~- ......

792 PARTVII Toxics LECTURE45 ToxicantIFood-Chain Interactions 793

11

density of organism ¡(cal kg-weC 1).The energy intake rate is this usage rate divided and the totalloss rate as
by the fraction of ingested energy that is assimilated. The food consumption rate is
then given by K2 = 31;~4 + 0.025 = 0.3433d-I

l¡ = ~ p¡ + G¡
Á¡_I~ 1,,-1
(45,36)
which can be used to compute the bioconcentration factor as

N2w = ~.13:~~ = 92,682L kg-lipid-I

where the subscript i - 1 designates the prey and a¡,¡_1= fraction of ingested prey
The food-chain multiplier can be calculated as
that is assimilated, If it is assumed that the caloric density of dry tissue is the ~ame
fol' the pl'edator and the prey, differences in caloric density should be related to the = 0.7(0.154) = 031418
821 0.3433 .
wet-weight-to-dry-weight ratio. Thus the lipid-specific consumption rate can be cal-
culated as This value, along with the BCF, can be used to determine the zooplankton concentra-
tion as
lu = awd,¡-I !L.¡-I p¡ + G¡ (45,37) V2 = 92,682(0.9804)+ 0.31418(98.039)= 90.865+ 30.802 = 121,667'Lgkg-lipid-I
awd,¡ fL.¡ a¡,¡_1
Results for the rest of the food chain can be computed in a similar fashion, and the results
where awd = wet-to-dry-weight l'atio (kg-wet kg-dry.I). are summarized in the following table and plot:

EXAMPLE 45.2. TOXICANTIFOOD CHAIN. Thomann et al. (1992) presented the

following data for the pelagic food chain in Lake Ontario: VI<
(components)

G N. vI< N
Weight A P h
v
Organism No. (kg-wet) (%lipid) a a.d (d-I) (d-I) (d-I) Organism (BCF) v. g Phyto Zoo Forage (total) (BU')

Phyto/detritus 1 - 1 - 10 Phyto/detritus 98,039 98,039 - - - - - 98,039 100,O()()

2 1 x lO-s 5 0.3 5 0.025 0.09 0.154 92,682 90,865 0.314 30,802 - - 30,802 121,667 124,1()()
Zooplankton Zooplankton
Forage fish 3 0.1 8 0.8 4 0'<J04 0.014 0.016 90,821 89,040 0.289 8890 35,117 44,006 133,047 135.708
Forage fish
Piscivorous fish 4 1 20 0.8 4 0.0025 0.009 0.0058 Piscivorous fish 79,830 78,265 0.325 2887 11,403 43,205 57,496 135,751 138,476

..- Note that the last three columns of the table have been calculated with Eqs. 45.33, 45.34,
IL' and 45.37. (a) Determine the levels of contaminant in the tour organisms for a toxicant
with a Kow = HP and a total water concentration of 1 J.tg L -1, Assume that the phyto-
plankton behave as simple partic1es, and

N1w "" Kow

The lake has a phytoplanktonldetritus concentration of about 1 mg L -1 with a carbon

content of 20%. Also assume that excretion losses, K, are negligible. (b) Plot the con-
centration in the four compartments versus log Kow from 2 to 9,

Solution: Using relationships for partitioning developed in Lec, 41 (e.g., Eq. 41.37),
we can first calculate the concentration in dissolved form as

Cd =
1 + \0-6/\05)0.2(1) 1 = 0.9804 J.tg L-I
Phyto
Recall that the \0-6 in the denominator is necessary because the suspended ~olids are
zoo
expressed as mg L -l. The solids concentration of the small organic particles (that is, Small
phytoplankton, detritus) can be computed as flsh
Larga
VI = NlwCd = 1<P(0.9804) = 98.039 J.tgkg-lipid-I Ilsh
Next we can determine the zooplankton concentration. First, using a value of 0.7
for EJEo (from Fig. 45.1), we can calculate the uptake rate as
4 6 8
k = \062.67(0.4)0.0907 = 31814Lk -1' 'd-I d-I log Kow FIGURE E45.2-2
u 5(0.05)8.5' , g Ip.
 794 PART VII Toxics LECTURE45 ToxicantIFood-Chain Interactions 795

(b) Results for a range of Kvw' s are displayed in Fig. E45.2.2. Notice how the peak Cdl = FdlCI (45.44)
occurs al a value of aboul Kvw = 6 to 7. This result is determined by Ihe relalionship of
and FplCI
bolh a and Ec to K"... (Fig. 45.3) along wilh the fact Ihal more loxicanl is associaled with VI (45.45)
Ihe food as K"... increases.
!oe.lm,

EXAMPLE 45.3. TOXICANTIFOOD.CHAIN MODEL AND MASS BALANCE.

It should be noted that it is believed that phytoplankton do not actually behave as
simple particles, and that NI\\' is not equal to Kmv for all organic chemicals. For exam-
pie Thomann (1989) suggests that the proportionality holds up to about log Ko\\' = 6.
Ahove that point the NI,,' appears to plateau at a level of aboullog NI\\' ==6.5. We
will explore the implications of this leveling off in one of the problems at the end of
Ihis lecture.
Lake Onlario has a 1101'1of Q = 212 X 109 m3 yr-I and a deposition zone of approx-
imalely 10.000 x 106 m2. Assume thal the suspended solids in the lake settle at arate
of 100 m yr-I. Also assume that Ihe cOl:taminant is subject to no losses other than set-
tling and nushing and Ihal resuspension and sedimenl-water diffusion are negligible. Al!
othcr paramelers and assumplions are Ihe same as Example 45.2. Ir the contaminanl has
a K"... of 10\ compule Ihe inllow concentration required so Ihat large fish concenlration
is maintained al a level of HP JLg kg-lipid-I.

Solution: For no resuspension, sediment-water diffusion, and decay reactions, Eq.

45.4 INTEGRATIONWITHMASS BALANCE 45.38 simplifies lo I

QCin I
In the previolls seclion we have iIIuslraled how food-chain-toxicant conccntrations
el =
Q + v..FpIA
c<Jnbe detenninedas a functionof a givenwaterconcenlration.(ntegrationof the I

I From Example 45.2 we have determined a BAF of 138,476 for the top predators. Thus .~
I f(,od-chain model into a mass-balance framework would have the added utility of
I
tying Ihe biotic concentralions back 10toxicant loadings.
this result can be multiplied by the dissolved concentration in the water to determine Ihe-
Ir
I If it can be assumed Ihat the toxicant mass relained in Ihe nonphytoplankton parts
large fish concenlralion,
Qe¡n
of the foodchain are negligible, we can merely apply the toxicant mass-balance mod- V4 = N4Fdl-
Q + VsF pl A
I I

I
els developed 10this point to calculale the toxicant concentrations in the dissolved which can be rearranged to determine the target inllow concentration
and particulate phases. If the laHeris assumed to represent food (that is, phytoplank-
ein -_ V4(Q + vsFplA) _
I

v.,(1 - Fdl)
II Ion and organic detritus), these levels can then serve as Ihe forcing functions for the
lood-chain model.
- -
N4Fdl Q
V4
N4Fdl [1
+
qs ]
I
For example recall that the steady-state solution for a well-mixed lake underlain where qs = Q/A, which for LakeOntariois 212 x 109/10,000x 106= 21.2 m ycl.
by a single sediment layer can be formulaled as (Eqs. 40.33 through 40.35) Substituting Ihis along wilh Ihe other parameter values yields
Io-~ 100(0.0196) _1
CI =
QCin
Q + kl VI + v,.AF"1 + (1 - F;)(v,.FfII+ v,¡F"I)A
(45.38 )
ein = 138,476(0.9804)I + [ 21.2 ] = 0.668JLgL

and e2 = V,.Ffll+ v"F"1

k'2 H2+ Vr + vI> + ""f "2
el - (40.39)
which represenls abollt a ~ reduclion.

where F; is lhe raljo of sedimenl feedback lOtOlalscdimenl purging,

45.5 SEDIMENTS AND FOOD WEBS (ADVANCED TOPIC)
F' = Vr + VdFd2 (45.40) The framework presented above can be extended to incorporate sediment-water in-
r Vr + VI> + I'dFd2 + k2H2
teractions and food webs. Figure 45.4 il\ustrates how both can be implemented by
aud the fractions are defined in terms of octanol-water partition coefficients as including an organism that feeds on bottom sediments.
Balances for the five food-chain compartments can be written as
Fdl (41.41)
1+(1 x 1O-6K""')!,,cllll O = kulCd - K;vl
(1 X 10-6 K"w)f,,,.1II1 (41.42)
F¡'I = O = ku2Cd+ U211L2VI- KZV2
1+ (1 x 1O-6Kow)f,,,1Il1
( O= ku3Cd + P32u32h..3V2+ P35u35h.jV5 - KJV3 (45.46)
(41.43)
Fdl'2 = ,1. + (1 x 1O-6Kol\.).f"c2(1-ljJ)p O = ku4Cd + u43h.Av3 - K~V4
Thus the Iwo quanlities needed to drive Ihe food-chain model can be computed as
O = [ku5(b5sc. + b5dCd)) + tp5.,u5slLocvs + P5tU5th5V,] - K$vs
 796 PARTVII Toxics lECTURE45 ToxicantIFood-Chain Interactions 797

as metals and radionuclides follow a similar approach. That is, direct uptake am!
predation provide the pathways by which chemicals move and concentrate in thc
biota.

PROBLEMS
dllluslon
45.1. Repeat Example 45.1, but compute the BCF and lipid-specific concentration for the
forage fish (see Example 45.2 for parameters).
WATER
--....-- 45.2. Repeat par! (a) of Example 45.2 for a contaminant with a log K"... of 6.
SEDIMENTS 45.3. Repeat par! (b) of Example 45.2. but use the following relationship for NI...:
~ NI... = K"... 2 s log K"w s 6.5

---¡bUria. N1w = 106.5 6.5 s log K"w s 9

FIGURE 45.4 45.4. Repcat Example 45.3, but for a compound with a log K"", = 6.
A five-compartment food-web model including sediment-water interactions (modified from Thomann
et al. 1992).
.1
where b5s and b5d = the fraction of the uptake from sediment and overlying water
(b5s+ b5d = 1); P32and PI, preferenee factors of forage tish for zooplankton and
benthic invertebrates, respectively (P32+ P35 = 1); P5s and P5J = preference factors
ofbenthic invertebrates for sediments and phytoplankton, respectively (P5' + P51 =
1); and hoc = rateofkilogramsof organiecarboningestedper kilogramlipidofthe
predator (kg-orgC kg-lipid-I d-I) that can be computed with (compare with Eq.
45.37)
'
/
IL0("-. - -foc.i Pi + G
,.' (45.47)
fu a;'¡-Iawd,i

We have highlighted the terms in Eq. 45.46 that make this model different from
the purely pelagic model dcpieted in Fig. 45.2. These differences are
. The use of prderence factors to split the grazing of both the benthic invertebrates
and the forage fish between different food sources. It is this splitting that makes
the system a food web rather than a food chain.
. The connection of the benthic macroinvertebrates with the sediment pore water
(uptake) and the sedimcllt solids (ingestion). In the former case the b terms are
used to divide the uptake between pore and overlying waters.
. The reformulation ofthe benthic organism consumption rate (Eq. 45.47) to reflect
the fact that the solid contaminant in the sediments is expressed on a carbon rather
than a lipid basis.
Thomann et al. (1992) applied this framework to Lake Ontario and illustrated how
the benthic pathway can contribute signiticantly to the contaminant levels in the
pelagic organisms.
In summary we have attempted to introduce you to how contaminantlfood-
chain interactions can be modeled and integrated with the mass-balance approach.
Although our focus has been on organic toxicants, models of other contaminants such
 AI'PENDIX
A Conversion Factors 799
"
APPENDIX A Veloclty Concentratlon
1 meterlseeond (m SO') = 3.2808 ft sol IgL-1 =Ippt
= 86.4 km d-I = I~
Conversion Factors 1 footlseeond (fI SO')
= 2.237 mph
= 0.6818mph
1mg L-1 = 8.34lb",(10"gal)-'
1gm-3
= 0.3048 m sol 1ppm
= 16.364 mi d-I IJLgL-I = Imgm-1
I mile per hour (mph) = 1.467 fI sol 1 ppb
= 0.4470 m sol 1 ng L-, = 1 JLgm-3
= 1.609 km hr-' I pptr

Energy
Dlsperslon/diffusion 1ealorie (cal) = 4.1868J
= 0.003968Btu
1 square eentimeterl = 3.336 x 10-6 mF d-I
1joule (1) = 1W-s
second (em2 SO,) = 93 ft2d-' = 1N-m
1 squaremile/day = 0.2998 x 10" em2 sol 1 kilowall-hour = 3600kJ
(mi2 d-I) = 860 keal
= 2.54cm :...¡
1 ineh(in) = 3412Btu
Metric and SI Preflxes = 1.609344km
1 mile(mi) I Britishtherrnalunit = 1.055056kJ
1 nautical mile = 1.852km
Factor Prefix Symbol (Blu) = 252cal
1 meter (m) = 3.2808ft Flow = 0.2930W-hr
lO" tera- T = 39.37in
= 778 fI-lb
lO' giga- G 1 angstrom (A) = 10-10 m 1 eubicmeterl = 35.3147efs
10" M = 0.1 nm 1erg = 10-10kJ
lO'
mega- seeond (mJ SO') = 2118.9 ftJ min-I (efm)
kilo- k = 22.8245 MGD
102 hecto- h = 70.07 ae-ft d-' Temperature
lO' deka- da Area 1 eubie feetl = 0.028316m's-'
10' deci- d seeond (efs) Degrees W(OF) - 32J
= 7.4806 gal mino'
10 2 eenti- e 1 square ineh (in2) = 6.452 em2 = 0.0646 MGD Celsius ("C)
10 ., milli- m 1 heetare (ha) = 10,000 m2 [T("F)+ 40) _ 40
10" micro-
1 million gallons/day = 0.0438 m' s-, 1.8
IJ. = 2.4710 acres (MGD) = 1.5472 cfs Degrees = ~ T(°C) + 32
iO · nano- n 1 square kilometer (km2) = 0.3861 mi2 1 gallonlminute = 6.30902x lO-s m' s-,
10" 1 acre (ae) = 43,560ft2 Fahrenheit ("F)= 1.8[T("C) + 40J - 40
pieo- p (gpm)
lO IS femto- f = 0.404685 ha Kelvins (K) = T(°C) + 273.15
10 IR allo- a Degrees ,,; T(°F) + 459.67
Rankine (OR)
Volume Mass Degrees Celsius DegreesFahrenheit
Time 1 eubie meter (m') = 35.3147 ft1 ("C) (OF)
1 kg = 2.2046 lb",
= 264.172 gal (U.S.) = 35.274 ounee O 32
= 1000 lilers (L) 5 41
I year (yr) = 31.536x 10"s 1 metric ton (Ionne) = 1000kg
1 eubic foot (fl1) = 7.480 gal 10 50
I average month (mo) = 30.42 d = 2204.6 lb",
1 aere-feel (ae-ft) = 43,560 ftJ 15 59
I day (d) = 1440 min 1 pound (lb) = 453.6 g
= 1,233.49 mJ 20 68
= 86,400 seeonds (s) = 325,851 gal (U.S.) 25 77
1 gallon (U.S.) = 3.785L 30 86
IMG = 10" gallons 35 95
Length = 0.15899 m' Density
1 barrel oil (bhl) 40 104
1 rnot (ft) = 0.3048 m = 42 gal (U.S.) 45 113
1 "uid ounee 1 g em-J = IO()()kg m-1
I yard (yd) = 0.9144 m = ot. gal (U.S.)
= 62.428 Ih", ft .1
50 122

1')R
 ... ----

800 APPENDlXA

Heatflux Vlscoslty (dynamlc)

I
I APPENDIX B

Ilangley/d (ly d-I) = 1cal cm-2 d-I IPa, 1kgm-IS-I I - - - --

1 BIUfl-2 hr-I = 3.154591 W m-I = 1000 centipose (cp)
= 0.672Ibmft-I S-I I
Oxygen Solubility
Pressure I
Heat capaclty
1 kPa = 0.0098692 almospbere Temperalure Chloride concenlration (g L 1)
= 0.296134 inHg 1 kJ (kg ° C)-I = 0.2388 cal (g oC)-1 (OC) O 5 10 15 20 25
= 7.5006151Iorr (mmHg) = 0.2388 Blu (lbm0F)-I
O 14.621 13.726 12.885 12.096 I 1.356 10.660
= 0.1450377 psi
= 0.01 bar 1 14.216 13.354 12.544 11.782 11.(168 10.396
2 13.830 12.999 12.217 11.482 10.792 10.143
= 0.0101972 kgr cm-2 Thermal conductivlty 3 13.461 12.659 11.904 11.195 IO.52!! 9.900
4 13.108 12.334 11.605 10.920 10.275 9.668
1 W (m K)-I = 0.5778Blu (hr flOFtl 5 12.771 12.023 11319 10.656 10.032 9.445
= 2.388X 10-) cal(scm°C)-I
Force 6 12.448 11.726 11.045 10.404 9.KOO 9.231
7 12.139 11.441 10.7!!2 10.161 9.577 9.025
1 newlon (N) = 1 kg m S-2 8 11.843 11.167 10.530 9.929 9.362 8.828
= 0.2248 Ibr Heat transfer coefflclent 9 11.560 10.906 10.288 9.706 9.157 8.639
= 7.2331b'D fl8-2 10 11.28K 10.654 10.056 9.492 K.959 K.456
= 1 x lOSdyne 1W (m2 K)-I = 0.17611 Blu (f12hr 0F)-I I1 11.027 10.413 9.834 9.2K6 K.769 8.281
1 dyne (dyn) = 1 g cm S-2 12 10.777 10.182 9.620 9.089 K.587 8.113
13 10.537 9.960 9.414 8.898 8.41 I 7.950
14 10.306 9.746 9.216 8.715 8.242 7.794
Gas law constant 15 10.084 9.540 9.026 8.539 8.079 7.643
Power 16 9.870 9.342 8.843 8.370 7.922 7.498
R = 8.206x lO-s m) alm (molK)-I 17 9.665 9.152 8.666 8.206 7.770 7.358
1 Wall (W) = 1joule s-a (J S-I) = 8.3143J(molK)-I \
18 9.467 8.968 8.496 8.048 7.624 7.223
= 1kg m2s-) = 1.9872cal (molK)-I ; 19 9.276 8.791 8.332 7.896 7.483 7.092
= 3.412 Blu hr-I = 10.73 ft) psi (lb...1 R) 20 9.092 8.621 8.173 7.749 7.347 6.966
= 1.34 x 10-3 hp = 0.7302 fl3 alm (lb""" R)-I 21 8.915 8.456 8.020 7.607 7.215 6.843
22 8.744 8.297 7.872 7.470 7.0K8 6.725
23 8.578 8.143 7.729 7.337 6.'J64 6.611
24 8.418 7.994 7.591 7.208 6.845 6.499
25 8.263 7.850 7.457 7.083 6.729 6.392
26 8.114 7.710 7.327 6.962 6.616 6.287
27 7.968 7.575 7.201 6.845 6.507 6.186
28 7.828 7.444 7.079 6.731 6.401 6.087
29 7.691 7.316 6.960 6.621 6.298 5.991
30 7.559 7.193 6.845 6.513 6.198 5.898
31 7.430 7.073 6.733 6.409 6.100 5.807
32 7.305 6.956 6.623 6.307 6.()()5 5.7111
33 7 1!!3 6Ji42 6.517 6.208 5.91 \ 5.6 \.'
34 7.11Ó5 6.731 6.414 6.111 5.812 5.5.IX
35 6.949 6.623 6.313 6.016 5.734 5.4(,5
36 6.837 6.518 6.214 5.924 5.648 5.3M5
37 6.727 6.415 6.118 5.834 5.564 5.J06 I
38 6.620 6.315 6.024 5.746 5.481 5.229 I
39 6.515 6.217 5.932 5.660 5.401 5.153 \ f
40 6.413 6.121 5.842 5.576 5.322 5.079

~'III
 APPENDIX D
APPENDIX e

Chemica1 E1ements
Water Properties

-'1
Atomlc Atomic Álomlc: Áriilc .-
Absolute Kinematic Vapor
pressure Symbol number welght' Symbol number weíght'
Temperature Density vlscoslty vlscosity
("C) (g cm-3) [10-2 gm (cm sr') 10-2cm2S-I (mmHg) Actinium Ac 89 [227) Europium Eu 63 151.96
4.58 Aluminum Al 13 26.9815 Fennium Fm 100 [257)
o 0.99987 1.7921 1.7923
5.29 Americium Am 95 [243J FIuorine F 9 18.9984
2 0.99997 1.6740 1.6741
6.10 Antimony Sb 51 121.75 Francium Fr 87 [223)
4 1.00000 1.5676 1.5676
7.01 Argon Ar 18 39.948 Gadolinium Gd 64 157.25
6 0.99997 1.4726 1.4726
8.04 Arsenic As 33 74.9216 Gallium Ga 31 69.72
8 0.99988 1.3872 1.3874 72.61
9.21 Astatine At 85 [210) Gennanium Ge 32
10 0.99973 1.3097 1.3101
10.52 Barium Ba 56 137.33 Go1d Au 79 196.9'57
12 0.99952 1.2390 1.2396
11.99 Berkelium Bk 97 [247) Hafnium Hf 72 178.49
14 0.99927 1.1748 1.1756
13.63 Beryllium Be 4 9.0122 Helium He 2 4.0026
16 0.99897 1.1156 1.1168
15.48 Bismuth Bi 83 208.980 Holmium Ho 67 164.930
18 0.99862 1.0603 1.0618
17.54 Boron B 5 10.811 Hydrogen H 1 1.0079
20 0.99823 1.0087 1.0105
19.83 Bromine Br 35 79.904 Indium In 49 114.82
22 0.99780 0.9608 0.9629
22.38 Cadmium Cd 48 112.41 lodine 1 53 126.9045
24 0.99733 0.9161 0.9186 192.22
25.21 Calcium Ca 20 40.078 lridium Ir 77
26 0.99681 0.8746 0.8774 26 55.847
0.8394 28.34 Califomium Cf 98 [251) lron Fe
28 0.99626 0.8363 C 6 12.011 Kr 36 83.80
0.8039 31.81 Carbon Krypton
30 0.99568 0.8004 58 140.12 Lanthanum La 57 138.91
Cerium Ce
Cesium Cs 55 132.905 Lawrencium Lr 103 [260)
.'/ I Chlorine
Chromium
CI
Cr
17
24
35.453
51.996
Lead
Lithium
Ph
Li
82
3
207.2
6.941
Cobalt Co 27 58.9332 Lutetium Lu 71 174.97
Copper Cu 29 63.546 Magnesium Mg 12 24.305
Curium Cm 96 [247) Manganese Mn 25 54.9380
Dysprosium Dy 66 162.50 Mendelevium Md 101 [258)
Einsteinium Es 99 [252) Mercury Hg 80 200.59
Erbium Er 68 167.26 Molybdenum Mo 42 95.94
, Valuesin bracketsdenotethe massnumberof Ihelongestlived or best-knownisolope.

803
R02
~
 "

804 APPENI>IX
() Chemical Elements

Atomic Atomic Atomic Atomic APPENDIX E

Symbol number weigbt' Symbol number weight'
Nd Silicon Si 14 28.086
- -
Neodymium 60 144.24
Neon Ne 10 20.180 Silver Ag 47 107.868
Sodium Na 11 22.9898
Neptunium
Nickel
Np
Ni
93
28
(237)
58.69 Strontium Sr 38 87.62 Numerical Methods Primer
Niobium Nb 41 92.906 Sulfur S 16 32.066
Nitrogen N 7 14.0067 Tantalum Ta 73 180.948
Nobelium No 102 (259) Technetium Te 43 (98)
Osmium Os 76 190.2 Tellurium Te 52 127.60
Oxygen O 8 15.9994 Terbium Tb 65 158.925
Palladium Pd 46 106.4 Thallium TI 81 204.38
Phosphorus P 15 30.9738 Thorium Th 90 232.038
Platinum Pt 78 195.08 Thulium Tm 69 168.934
Plutonium Pu 94 12441 Tin Sn 50 118.71
Polonium Po 84 [2091 Titanium Ti 22 47.88
39.098 W 74 183.85 r
Potassium K i9 Tungsten
Pmseodymium Pr 59 140.9077 Unnilhexium Unh 106 [263]
Promethium POI 61 (145) Unnilpentium Unp 105 [262]
Protadinium Pa 91 (231) Unnilquadium Unq 104 [261] I
Radium Ra 88 (226) Unnilseptium Uns 107 [2621
Radon Rn 86 (222) Uranium U 92 238.ü3
Rhenium Re 75 186.2 Vanadium V 23 50.94 Numerical methods involve solving mathematical problems with the computer. The
Rhodium Rh 45 102.905 Xenon Xe 54 131.29 following material summarizes some simple numerical methods that are use fuI in en-
Rubidium Rb 37 85.4678 Ytterbium Yb 70 173.04
Ruthenium Ru 44 101.07 Yttrium Y 39 88.9059 vi ron mental modeling. Additional details on the following methods, plus additional
Samarium SOl 62 150.4 Zinc Zn 30 65.39 techniques, can be found in Chapra and Canale (1988). The reader is advised 10con-
Scandium Se 21 44.956 Zireonium Zr 40 91.22 suli this and other references to understand nuances of the following techniques that
-Selenium Se 34 78.96 are beyond the scope 01'this primer.
!. Valuesin braekelsdenoteIhemassnumberor Ihelongesllivedor besl-knownisolOpe.

ROOTS OF EQUATIONS

The root of an algebraic or transcendental equation is the value 01'a variable that
makes the equation equal to zero. That is, it is the value of x that yields
f(x) = O (E.I)
Numerical methods for finding roots are based on making a guess (or guesses)
at the root and having the computer systematically refine the estimate. There are
two fundamental approaches for accomplishing this objective: bracketing and open
methods.

8racketing Method: Bisection

As the name implies, bracketing methods begin with two guesses that bracket the
root. That is, they define an interval within which there is a single root. This can
be done by determining whether a sign change occurs in the interval (a sign change ~
pleans that the function passes through zero within the interval),
f(x¡)f(xu) <O (E.2)

805
~-.___o
 .4-......................

806 APPENDIX E APPENDlX

I! Numerical Methods Primer 807

where XI and Xu = lower and upper guesses. If two such guesses can be developed, OpenMethods:Newton-AaphsoÍ'l
and Modlfll!dsecanl
a root estimate can be simply computed as occurring at the middle of the interval, Open methods do not require guesses that bracket the root. As a tónsequence Utey
XI + Xu often converge faster than bracketing methods. Howevet, they can also diverg~ from
(E.3)
xr= 2 the root. The speed and possibility of convergence usuaUy dependí; on fue quatlt}t bt
the initial guess. To a lesser extent it depends on the type of function being eváÍuated.
and Xr is the root estimate. The most commonly employed and (when it works) the most efficient open
Once the root estimate is obtained, a new bracket must be defined that encom- method is the Newton-Raphson approach.1t is based on using the function's deriva-
passes the rool. To do this, it is determined whether a sign change occurs between the ti ve to define a path to the next root estimate. This process can be expressed by the
lower guess and the root estimate. If this is true the new upper bound is defined as the following iterative formula,
rnot estimate. If it is not, the new lower bound is defined as the root estimate. Once ¡(Xi)
fhe bracket is redefined, a new midpoint can be computed and the process repeated X¡+I =X¡- f'(x¡) (E.S)
tu obtain finer estimates of the root. This formula is repeated until the percent relative error (computed with a formula
The process is iterated until an error estimate falls below some prescribed stop- like Eq. EA) falls below a specified leve\. Some pseudocode to implement this algo-
ping criterion. One such way to estimate the error is rithm is displayed in Fig. E.2.
old One problem with the Newton-Raphson approach is that it requires that the
x~w - Xr (100%) (E.4 ) derivative of the function be determined. An altemative that does not require a
Ea = I
x~ew I derivative evaluation is the modified secant method. This approach uses a finite dif- ., Wj
ference to estimate the derivative,
where Ea = approximate error and x~ew and ~Id present and previous root esti- =
mates. Some pseudocode to implement this algorithm is displayed in Fig. E.I. f'(x¡) E: ¡(Xi + EX¡)
EX¡
- ¡(Xi) (E.6)
II
where E = a small perturbation fraction (E: 0.01). Equation E.6 can be substituted
I SUB Bisect(xl, XU, xr, iter. ea)
into Eq. E.S to give the modified secant formula,
Ex;f(x¡)
Xi+I = Xi - (E.7)
~
This subroutine must have access to
¡(Xi + EX¡) - ¡(Xi)
I
I the {unction being evaluated, ((x) The Newton-Raphsonformula(Eq.E.S)can be replacedby Eq.E.7 to convertthe
algorithmin Fig.E.2tothe modifiedsecantmethod.
ASSIGN imax, es
iter = O
DO SUB Newton(xr, iter, ea)
xrolél xr =
xr = (xl+xu) /2 This subroutine must have access to the
iter = iter + l {unction being evaluated, {(x), and its
IF xr;tO THEN derivative, {'(x).
ea = abs((xr-xrold)/xr)*IOO
END IF ASSIGN imax, es
test = f(xl)*f(xr) iter = O
IF test = O THEN 00
ea = O xrold xr =
IF test<O THEN xr = xr - f(xr)/f' (xr)
xu = xr iter = iter+l
ELSE IF xr;tO THEN
xl = xr ea = abs((xr-xrold)/xr)*100
END IF END IF
LOOP UNTIL ea<es OR iter>imax LOOP UNTIL ea<es OR iter>imax

END Bisect END Newton

FIGURE E.1 FIGURE E.2

Bisection algorithm. Newton-Raphson method.

~~
 ~~~..,.; ~i_
.
;,
808 APPENDIXE
t APPENDIX
E Numerical Methods Primer 1;09

LINEARALGEBRAIC EQUATIONS often the case in water-quality modeling, \he Gauss-elimination algorithm can be al-
tematively expressed as what is called an LV decomposition fonnal. In this expres-
This cJass of problems involves solving n linear algebraic equations with n un- sion, the constant factors used to eliminate unknowns are first multiplied by the coef-
knowns. An example of such a system is ficients on the left-side of the equations. They are then saved so that they can be used
alJXt + al2X2 + al3x3 = bt later to multiply the individual right-hand-side constants. This can be done by a for-
a2tXt + a22x2 + a23x3 = b2 (E.8) " ward substitution. Thereafter a back substitution can be employed to obtain solutions.
/' An algorithm to implement an LU decomposition expression of Gauss elimina-
a31XI + a32x2 + a33x3 = b3 tion is listed in Fig. EA.
where the aij's are constant coefficients in which i denotes the row (or equation) '" Three features of this algorithm bear mention:
andj denotes the column (or unknown), the b's are constants, and the x's are the
un~nowns. There are two fundamental approaches for solving for the unknowns:
elimination (or decomposition) methods and iterative methods.
. The factors used for the forward pass are stored in the lower part of the matrix.
This can be done because these are converted to zeros anyway and are unnecessary

Ellmlnatlon Methods: Gauss Ellmlnatlon and LU Decomposltlon

Gauss elimination is based on multiplying the equations by constant factors and then
subtracting them. As the name implies, the object is to eliminate unknowns from the
equations. After eIimination, the last equation will have one unknown, the next-to-
the-Iast two unknowns, etc. Back-substitution can then be used to compute values
for the unknowns. The process is depicted in Fig. E.3. To avoid division by zero, a
partial pivoting strategy can be incJuded in the algorithm. This involves switching
equations to avoid division by zero.
Although Gauss elimination certainly represents a sound way to solve simulta-
neous equations, it becomes inefficient when solving equations with the same coeffi-
cients (the a's), but with different right-hand-side constants (the b's). Because this is
. for the final solution. This storage saves space.
This algorithm keeps track of pivoting by using an order vector o. Tltis greally
speeds up the algorithm because only the order vector (as opposed to the whole
row) is pivoted.
. The algorithm checks for singularity and near-singularity. If it passes back a value
of er = -1, a singular matrix has been detected and the computation shollld be
~ terminated.
One important application of the algorithm in Fig. EA is to caIculate the matrix
inverse. This can be done by attaining solutions with unit vectors as the right-hand-
side constants. For example if the right-hand-side constant has a I in the firstposition
and zeros eIsewhere,
1 t
(E'))

1:"' lb} ~ W
all~ +auXz +al3x3 =b.
a21x. +a11Xz +a13x3 =bz the resulting solution will be the first column of the matrix inverse. Similarly if a
unit vector with a I at the second row is used, {b}T = {O lO}, the result will be
a3.x. + a3ZxZ + a33x3 = b3
the second column of the matrix inverse. Pseudocode to generate the matrix inverse
in this way is shown in Fig. E.5.
The final elimination method involves the solution of tridiagonal systems,
all~ +auxz +al3x3 =b. flxl + g)X2 = rl
~
auxz +anx3 =bi
e2x) + hX2 + g2X3 = r2
a»x3=b; = r3 ,.
e3x2 + j3x3 + g3X4 ~
:>

b"
"3 =--1...
a» e,,-)X,,-2 + f,,-IX,,-) + g"-IX,, = rll-I
bi- an"3
Xz=---:-- e"x,,-) + f"x" = rll
i:
a11 (E,IO)
FIGURE E.3
". -auXz -al3"3
~= Schematic representation of the solution strategy underlying
all Gauss elimination. Such equations apply to several areas of water-quality modeling, notably the solution

----
 8\ O APPENDlX E ¡.JI APPENDIX
É Numerical Methods Priittet au

SUB Decomp (a(), o(), n, eps, er) Driver program to generate matrix inverse', ai ()
sets up order vector ENTER a (), n
DOFOR i = 1 TO n eps = .000001
o (i) = i CALL Ludcmp(a(), al), n, eps, er)
END DO 11' er = O THEN
pertorrnsdecomposition FOR i = 1 TO n
DOFOR i = 1 TO n -1 FOR j = 1 TO n
CALL Piuot(a(), o(), n, i)
DOFOR k = i + 1 TO n 11' i = j THEN
factor = a(o(k), i) / a(o(i), i) b (j) = 1
a(o(k), i) = factor ~
ELSE
DOFOR j = i + 1 TO n b(j) = O
a(o(k), j) = a(o(k), j) - factor' a(o(i), j) END 11'
END DO NEXT j
END 00 CALL Substitute(a(), a(), n, b(), xl), eps)
END DO FOR j = 1 TO n
checks tor singularity ai(j, i) = x(j)
DOFOR i = 1 TO n NEXT j
11'ABS(a(o(i), i» < eps THEN NEXT i I
er = -1
EX1T DOI'OR Output ai!), if desired I
END 11' ELSE
END DO PR1NT
END 11'
"ill-canditianed system" .J
END Decomp
END
SUB Pivat (a(), o(), n, i)
I
p = i FIGUREE.5
big = ABS(a(o(i), i» Mairixinverse. I

DOFOR
dummy
k = i + 1 TO n
= ABS(a(o(k), i»
I
TI'dummy > big THEN
big = dummy of partial differential equations. Because so many elements are zero, the algorithms
p = k
END U' for solving such systems are extremely efficient. In addition such systems often do
END DO not require pivoting. An algorithm for solving tridiagonal systems, which is in the
dummy = o(p) LU decomposition format, is shown in Fig. E.6.
o(p) = o(i)
o(i) = dummy
END Pivot

SUB Subst (a(), o(), n, b(), x(), eps) SUB Decamp (e ( ), f ( ), 9 (), n)
torward substitution DO FOR i = 2 TO n
DOFOR i = 1 TO n - 1 e(i) = e(i)/f(i)
DOFOR k = i + 1 TU n f(i) = f(i) - e(i)*g(i-1)
factor a(o(k), i) = ENDDO
b(o(k» = b(o(k» - factor' b(o(i» END Decamp
END DO
END DO
back substitution
SUB Subst(e(), f(), g(l, r(), xl), n)
DOFOR i = n TO 1 STEP -1
farward substitution
sum = O DO FOR i = 2 TO n
DOFOR j = i + 1 TO n r(i) = r(i) - e(i)*r(i-1)
sum = sum + a(o(i), j) . x(j) \1
ENDDO
END DO back substitution
xli) = (b(o(i» - sum) / a(o(i), i) x(n) = r(n)/f(n)
END DO DO FOR i = n-1 TO 1 STEP -1
END Subst
Xli) = (r(i) - g(i) *x(i+l) If(i)
ENDDO
FIGUREE.4 END Subst
Gausselimination/LUdecomposition.
FIGURE E.6
Tridiagonal solver.

~
 .~_-->..~ ~~'-- - . '-~ ..

812 APPENOIX E
AI'I'I'NI>lX
ti NUIl1t:ri~al M~lh,,\h !'lImer KIJ

Iteratlve Methoda: Gauss-Seldel Furlher, it is very well suited for what are calIed sparse syslems. These are
sets of simultaneous equations that have mostly zero elements. For such cases the
The Gauss-Seidel method provides an approximate approach to solve linear simulta- Gauss-Seidel approach requires only that the nonzero elements be storcu and maHip-
neous equations that is very well suited for many environmental modeling problems. ulaled. Again, becaus(,uny environmental modeling problems exhibit sparseness,
It can be iIIustrated for the three simultaneous equations from Eq. E.S. '1'0implement this makes the technil)ue usefu\.
the tec~nique. we merely solve the first equation for XI, the second for X2, and the
third for X3.
. CURVE FITTING
XI =
bl - al2x2 - al3X3
all Fitting a curve to data points is widely employed in environmental englOeering.Two
X2 = b2 - a21 XI - a23x3 (E.II)
different approaches are among the most commonly employed. For cases where the
a22 data are uncerlain (that is, they exhibit variability), regression is used lo generale
a "best fit" line through the points. When Ihe data are known precisely (e.g., as
X3 = b3 - a31xI
a33
- a32x2
in a table of physical properties of water), interpolation is used lo generate curves
connecting the points.ln this case the object is lo estímate values belween the poinls.
These equations can be solved iteratively until the estimates converge on the solu-
tion. Convergence can be tested by a formula like Eq. E.4. Acode for making these
computations is listed in Fig. E.7. Linear Regression
The Gauss-Seidel method wil\ be assured of converging only when the system JI
Given a set of points, (XI. YI). (X2, Y2) (x". y,,), estimales 01' Ihe slope and inter-
being solved is diagonally dominant. This means that the absolute value of the di- cept of a straighlline that minimizes Ihe sum 01'the squares of the residuals can be
agonal element in each equation must be greater than the summation of the absolute computed. The equation for the straightline is
values of the off-diagonal elements. Although this might appear to be asevere limita-
tion, itjust happens that many linear systems confronted in environmental modeling y=ao+a¡x (E.I.!)
I
(particularly those related to partial differential equations) have this attribute. Thus where the slope is computed by
the Gauss-Seidel approach has broad utility.
nLX¡Y¡ - ¿x;¿y; (
al = (E.13) 1,

,~
SUB Gseid(a(), b(), n, xi)~
n ¿ xJ - (L X¡)2 f
ASSIGN imax, es
whereal1the summationsare fromi = Itu 11.The interceptis calculatedwilh
iter =O ao = y - al X (E. 14)
DO
flag = O
where x and y = means of the x's and y's, respectively.
iter = iter + 1
DOFOR j 1 TO n Note that two statistics can be used lo assess the goodness of 111Tlae first is
old = xii) the coefficient of determination ,2, which represents the fraction of li" error Ih¡1Iis
sum = b(i) explained by Ihe regression (,2 = I for perfect fit and,2 = Ofor no tit),
DOFOR j = 1 TO n
IF i ~ j THEN sum = sum - a(i,j)*x(j) ,2 SI-S,.
(E.I ")
END DO SI
x(i) = sum/a(i,i)
IF flag = O AND x(i) ~ O THEN where SI = total sum of the squares, I
ea = ABS(x(i) - old)/x(i»*lOO " ¡
END IF
IF (ea> es) THEN flag = 1
SI = L(Y;
;=1
- y)2 (E.I\,)
I
END 00
LOOP UNTIL flag = O OR iter > imax
and Sr = residual sum of squares around the regression !ine,
END Gseid n
Sr = L(Y; - ao - alx;)2 (E.I?) ~
FIGURE E.7 ;=1
Gauss-Seidel algorithm.
 814 APPENDlX E
APPENDlX
E Numerical Methods Prilrtet fti5
SUB Regres(x(), y(), n, al, aO, syx, r2) FUNCTION Lagrng(x(), y(), n, x)

sumx = O: sumxy = O: st = O sum =O

sumy = O: sumx2 = O: sr = O DOFOR i = O TO n
DO FOR i = 1 TO n product = y(i)
sumx = sumx + xli) DOFOR j = O TO n
sumy = sumy + y(i) IF i "# j THEN
sumxy = sumy + x(i)*y(i)
sumx2 = sumy + x(i)*x(i)
product = product*(x - x(j»/(x(i)-x(j»
ENDIF
ENDDO ENDDO
xm = sumx/n
sum = sum + product
ym = sumy/n ENDDO
al = (n*sumxy*sumx*sumy)/(n*sumx2-sumx*sumx) Lagrng = sum
aO = ym - al*xm
DO FOR i = 1 TO n END Lagrng
st = st +(y(i)-ym)2
sr = sr + (y(i) - al*x(i) - aO)2 FIGUREE.9
END DO
sy x = ( sr / (n - 2) o .5
Lagrangeinlerpolalion.
r2 = (st - sr)/st

END Regres Integration -1

...

FIGURE E.S Integrationinvolvessolvingthegeneralproblem

Linear regression.

The second statistic is the standard error ofthe estímate Sylx,

rs;- (E.IB)
J = r f(x)dx
(E.21)

Sylx = V n=2 As depicted graphically in Fig. E.I Oa,the problem can be visualized as determining
Acode to implement linear regression is lisled in Fig. E.S. the area under the eurve detined by the function.
A simple numerieal approach eonsists of approximating the area as a series of
trapezoids (Fig. E.I Ob).Thus the sum of the trapezoids represents an estimate of the
Polynomiallnterpolation area,
An IIth-orderpolynomialcan be tit to a given set of 11+ I points: (xo,Yo), -
J = (XI - Xo) f(xo) + f(xl) + (X2 - XI )
f(xt) + f(X2)
(XI. YI), ..., (x". YII)' This polynomial can then be used to calculate intermedi- 2 2
ate values within the range. The Lagrange interpolating polynomial provides a
(
f(x,,-,) + f(x,,)
convenient computational formula for doing this,
" + . . . + x" - X,,_I) 2 (E.22)

f,,(x) = L
¡=o
L¡(x)f(x¡) (E.19) Acode formakingthisevaluationis listedin Fig.E.I l.
where
" X-
L¡(x) n
= j-~ X¡ -
j~,
Xj
Xj
(E.20)

in which n designates "the produet of." Acode to implement the algorithm and FIGURE E.10
evaluatethe funetionat a speciticvalueofXis displayedin Fig.E.9. (a) The inlegral of a lunctlon
can be represented as Ihe
area under Ihe funcllon.
I x
COMPUTER CALCULUS a b (b) The trapezoidal rule
Xo XI Xl ... XII.IX,. approximales the Inlegral
The simplest forms of eomputer ealculus involve simple integration and differenti- (a) (b)
by summing Irapezolds Ihal
ation of tabular dala. approximate the area.
._ _"... .oto...:.
-!J-" " "!" ....-..

816 APPI!NDIX B

FUNCTION Area (X(), y(), n)

APPENDIX F
FIGURE E.11 .- - -- --- - - - - - -
area =O . Pseudocode
DOFOR i = 1 TO n lar the
area
ENDDO
= area + (x(i)-x(i-l»*(y(i-l)+y(i)/2 multiple- Bessel Functions
segment t
trapezaidal .
END Area rule. .

Two other widely used for fonnulas (for equally spaced points) are S;mpson's
1/3 rule,
1:; (X2 - xo/(xo) + 4f(xl) (E.23) I
6 + f(X2) !
I
and S;mpson's 3/8 rule,
1:; (X3 _ xo/(xo) + 3f(xd + 3f(X2) + f(X3) (E.24) I

I
I
I
.

Dlfferentlatlon
The followingpseudocodes(abstractedfrom programsdevelopedby Press et al.
:I
I'
Numerical differentiation involves detennining a derivative (or slope) estimate from 1992) define program functions to evaluate the modified Bessel funclÍons with a
~
discrete data points. For example for equally spaced data, a simple approach is a computer programo In addition I have included tables of values that are useful for : I
centered difference, solving simple problems.

f'(XI) :; f(xÜ - f(xo) (E.25)

[ I

X2 - Xo
Although this approach has utility, it is limited because (a) it is limited to equally I

waced data and (b) it yields the derivative estimate only at a discrete point XI. A
more powerful method involves fitting an interpolating polynomial to the points and I

differentiating the resulting fonnula. The result for the case of a quadratic applied to )
three points is

f'(x):; f(xj-I) 2x - Xj - Xi+1 + f(xj) 2x - Xj-l - X¡+I

. (Xj_1 - Xj)(x¡_1 - X¡+I) (x¡ - Xj-l)(X¡ - xi+d
2x - X¡-l - X¡
+ f( Xi+1) (E.26)
(Xi+1- x¡-d(x¡+1 - Xi)
Note that in contrast to Eq. E.25, this fonnula can be used for unequally spaced
points and provides a derivative estimate anywhere within the interval containing
the points. However, it should be noted that it provides the best results near the mid-
dIe of the interval. A simple code to evaluate the formula is provided in Fig. E.12.

FUNCTION Deriv(xO, yO, xl, yl. x2, y2. x)

FIGURE E.12
tl = yO * (2*x-xl-x2)/(xO-xl)/(xO-x2) Pseudocode
t2 = ,1
yl * (2*x-xO-x2)/(xl-xO)/(xl-x2) lar numerically
t3 = y2
* (2*x-xO-xl)/(x2-xO)/(x2-xl) estimating the
Deriv = tl + t2 + t3 derivative
lar unequally
END Ded v
spaced data.
XI7
 00
-;¡ "TI
",- 00
Q
c:c: .. .. .. ...

"" . . . .
>
... ."
o.::u ... U'lW"""'¡UlW""
Rm li li i"""""""""""""''''''<=
"..lInIIMK"" .... . " " .
...

G;i""""""""""""'< ¡; " """ ".. >i

."
O""
0.. ... ti> .OQOowcw'
. .. .. &; &; &; &; ...
o "
CO....
(f) §n
>i
...
... ..
~¡Q"'.a.aQ.aQ.aw
o ~ w w ~ ~ ~ ~ = ~ =
" + + + + + . + + ~ ~
.
....10..,..,..,'010--
0","'''''''''U'l0'I0)4
" + + + + + + , "
- QWOO\DOG\O....
WO\\OWCD..UlCD
\Dw....W\DU'I\D\O'O
'<. CD.. .. z
i
;<
.."
O"
co .. '< '< '< '< '< '< '< '< , ~ ""'<'<'<'<'<'< "" ""
NU1O\U'I...CD-...I\QW
....U'lwww...w. ""
'" ..., ..., ..
.....
<
D>
E"
>c
-,o t:»
~
- . . '"...,...,
. WCD"'WWNNII
....at-&\DCD
... - - ... -. ..lO "" o
t:=
"
'"
c.o.o.g.Q.IO-UI .a .a .a .a "' U, ."" ."" -
...
D> ..., lO .. 2-
¡¡;
3
~""""""""""""""~
> :a
.., z..
CD 0'10... N
" . " '"'"..,
o"" . '"
.. .., . .
lO lO d'
:1
.oI .oI .oI...o
'" .., !l
(ir
a.
..... w o w
0\ o ... w
o o
..., '"
'" ...,
.'" .'" o'
:1
'"
... VI U'I el)
-..J ....a UI CD ..
CJ
co
(f)
en -..1 U'I ID
",",00\'"
"" '" '"
'"
.., .'" .'"
(f)
...,...,
2'
'='
!l

. . ... .. . . .I
o' ...., VI W ... ....aU'I w ....
'=' .. .. ...
...

. .. ..."
(f)
li tr""""""""""""'< &; tr""""'" """''''''< . ...
.. ... ... ...
.. 11II
.. .. ID
....
íi>
a. ~fDUIIH'.II'
... .. lA
o o . o
QONNOQWUI
o o .., '<
. ""
... .. .
D>
-o
¡¡)
~.a.a.a"'Q¡QN
o~W""""U'lO\
........ ,
O...ww....U'lO\
:O-

" + + + + + +
. .., o VI .... U'I o
UlCDCDWOcnO\....a
w....a\DWCW\DW
w O ....a
'" :i
....
a. NCDUI...-..JO\....J.... ..., ...,
..'<'<'<'<'<'< o:o-
a3 >c
...
~ -'<'<'<'<'<'<
I , =
>i
... ... ... .. .. ... .... =
O-..Ja......."'!JIU'I
- .. ..
CDWCD..... CD'" '" - - '" ."" .""
'7'c,Qlo.c.o..a -"""""'''''''' ..., '" .. "" '&'2.'0 E
"ti
<D
~ ..., ,.., . . ..,. . " ..,. ""
.lOlO
.
(f)
(f)

i..
I I
.I o O . ""
.'" .'"
o o
o.., o..o ""
.. lO
..,
..... '" o..'" o...,
CD
CD..,
o CD ...,
.... o ..'" ..,
lO lOlO
""
..,
""
..., '"
'" '" ..CD
.., .'" .'"
<D
<D - - '" '" CD
""
..., ...,
+
""

¡¡

TABLE F.l
Modified Bessel Cunctions oC the first kind, Io(x)

X O 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

O 1.00000o 1.002502 1.010025 1.022627 1.040402 1.063483 1.092045 1.126303 1.166515 \.212985
1 \.266066 1.326160 1.393726 1.469278 \.553395 1.646723 \.749981 1.863965 1.989560 2.127741
2 2.279586 2.446283 2.629143 2.829606 3.049256 3.289839 3.553268 3.841650 4.157296 4.502747
3 4.880790 5.294488 5.747203 6.242625 6.784807 7.378196 8.027676 8.738607 9.516876 10.36894-
4 11.30191 12.32355 13.44244 14.66795 16.01041 17.48114 19.09259 20.85842 22.79363 24.91473
5 27.23981 29.78879 32.58352 35.64802 39.00869 42.69453 46.73743 51.17239 56.03794 6\.37637"
6 67.23420 73.66256 80.71765 88.46126 96.96131 106.2925 116.5369 127.7849 140.1356 153.6984
7 168.5932 184.9522 202.9205 222.6579 244.3400 268.1601 294.3308 323.0860 354.6829 389.4044-
8 427.5620 469.4983 515.5903 - 566.2526 621.9417 683.1595 750.4587 824.4476 905.7950 995.2379
9 1093.586 120\.733 1320.659 145\.446 1595.284 1753.481 1927.479 2118.866 2329.388 2560.967

TABLE F.2
Modified Bessel Cunctions oC the second kind, Ko(x)

X O 0.1 0.2 0.3 0.4 OS 0.6 0.7 0.8 0.9

O '" 2.427069 - \.752704 1.372460 1.114529 0.924419 0.777522 0.660520 0.565347 0.48673Oc.
1 0.421024 0.365602 0.318508 0.278248 0.243655 0.213806 0.187955 0.165496 0.145931 0.128846>
2 0.113894 0.100784 0.089269 0.079140 0.070217 0.062348 0.055398 0.049255 0.043820 0.039006
3 0.034740 0.030955 0_027595 0.024611 0.021958 0.019599 0.017500 0.015631 0.013966 0.0t24
4 0.011160 0.009980' 0.008927 0.007988 0.007149 0.006400 0.005730 0.005132 0.004597 0.004l'19
5 0.003691 0.00330S- 0.002966 0.002659 0.002385 0.002139 0.001919 0.001721 0.001544 0.(0)386
6 0.001244 0.001117 0.001003 0.000900 0.000808 0.000726 0.000652 0.000586 0.000526 O.()O04:¡':>
7 0.000425 0.000382 0.000343 0.000308 0.000277 0.000249' 0.000224 0.00020t 0.00018l - 0.000J6a
00 8 0.000146 0.000132 0.000118 0.000107 O.()()()()960 0.00008&- 0.000078 0.000070 0.000063 .O.()()()M¡I'
o,Q
9 0.000051 O.{)(J()()46' 0.000041 0.000037 0.000033- 0.000030. 0.000027 0.000024- 0.000022:' O.OOOO2()
e

,
- _ _ __----!l- - --
 " -----

APPENDIX G <t
__._..__ - - _" ..ti....

PITOr fQnction and Complement References

Ahrens, C. D. 1988. Meleorology Today. Wesl Pub. Co., S\. Paul, MN.
Alberts, 1. J. and Wahlgren, M. A. 1981. "Concenlrations 01'239.240pU,I37Cs,and "'Sr in the Waler, "i'
the Laurenlian Oreat Lakes. Comparison 01'1973 and 1976 Values." Environ. Sci Tech. 15:94-')8.
Alexander, M. 1979. "Biodegradation 01'Toxic Chemicals in Water and Soil." In Dynamic~'. Expo.wre
and Hazard Assessmenl ofToxic Chemicals, R. Haque, ed., Ann Arbor Scienee, Ann Arbor, MI.
Allen, D.L. 1973.Wolves of Mi"'mg: Their Vi/al Role in a Wild Community. Houghton-Mifftin, Boslon
Allen,H. E., Fu, O.,and Deng, B. 1993. "Analysis 01'Acid Volatile Sulfide (AVS) and Simultaneously
Extracted Metals (SEN) for Ihe ESlimalion 01'Polenlial Toxicity in Aquatie Seclimtllts." Envi",,".
Toxicol. and Chem. 12(8):1441-1453.
American Herilage Diclionary. 1987. Houghlon-Mifftill, BOSlon.
APHA (American Public Health Association). 1992. Sllmdard Me/hodsfor /he Exalllllla/ioll I1/WtIl..,.
." b erf(b) erfc(b) b erf(b) erfc(b) and Waslewater. 18 ed., Washington, De.
0.00 0.00000o 1.00000o 1.00 0.842701 0.157299 Auer, M. T. 1979. "The DYllamics 01'Fixed I110rgallieNilrogen Nutrition in Two Speeies 01'Chlow-
0.05 0.056372 0.943628 1.05 0.862436 0.137564 phycean Algae." Ph.D. Dissertation, University 01'Michigan, Ann Arbor, MI.
0.10 0.112463 0.887537 1.10 0.880205 0.119795 Auer, M. T. 1982. Ecology 01'Filamentous Algae, special issue. J. Grea/ Lakes Res. 8( 1).
0.15 0.167996 0.832004 1.15 0.896124 0.103876 Baity, H. O. 1938. "Some Factors Affeeling the Aerobie Decomposition 01'Sewage Sludge Deposih"
0.20 0.222703 0.777297 1.20 0.910314 0.089686 Sewage Works J. 10(2):539-568.
0.25 0.276326 0.723674 1.25 0.922900 0.077100 Ballks, R. B. 1975. "Some Features 01'Wind Aetion on Shallow Lakes." J. Environ. fllgr. Div. ASCt
0.30 0.328627 0.671373 1.30 0.934008 0.065992 IOI(EE5): 813-827.
0.35 0.379382 0.620618 1.35 0.943762 0.056238 Banks, R. B. and Herrera, F. F. 1977. "Efteet 01' Wind and Rain on Surfae.: Reaeration."
0.40 0.428392 0.571608 1.40 0.952285 0.047715 J. Environ. Engr. Div. ASCE 103(EE3): 489-504.
0.45 0.475482 0.524518 1.45 0.959695 0.040305 Bannister, T. T. 1974. "Prediction Equalions in Terrns 01'Chlorophyll Concentration, Quantum Yid.!,
0.50 0.520500 0.479500 1.50 0.966105 0.033895 and UpperLimil10Produetion."Unmol. Oceanogr. 19(1):1-12.
0.55 0.563323 0.436677 1.55 0.971623 0.028377 Barnes. R. O. and Ooldberg, E. D. 1976. "Methane Production and Consumplioll in Anoxic Mari""
0.60 0.603856 0.396144 1.60 0.976348 0.023652 Sediments." Geology 4:297-300.
0.65 0.642029 0.357971 1.65 0.980376 0.019624 Bartsch, A. F. alld OakslaUer, 1.H. 1978. Management Decisions for Lake Syslelll', ,,,. a Survey ,.1'
0.70 0.677801 0.322199 1.70 0.983790 0.016210 Trophic Stalus, Limiling Nutrienls, and Nutrienl Loadings in American-Soviel Symposium on u,~
0.75 0.711155 0.288845 1.75 0.986672 0.013328 01'Malhematical Models 10Ol'timize Waler Quality Managemem. 1975, U.S.En'¡.ollmelllal Pm-
0.80 0.742101 0.257899 1.80 0.989091 0.010909 leelion Agency Office 01'Research and Development, Environmental Research l.aboralory, Gulf
0.85 0.770668 0.229332 1.85 0.991111 0.008889 Breeze, FL, pp. 372-394. EPA-60019-78-024. 1
0.90 0.796908 0.203092 1.90 0.992790 0.007210 Bartsch, A. F. and Ingr.un, W. F. 1967. Biology ofWater Pollulion, U.S. Depl. oflnterior, Water Pollution
0.95 0.820891 0.179109 1.95 0.994179 0.005821 Control Adminislralion.
Bales, S. S. 1976. "Effects oi' Lighl and Ammonium on Nilrate Uptake by Two Species 01'Esluarine
PhylOplankton." Limnol. OceOllOg.21:212-218.

821 ¡
820
 ':'11

.t
822 References .~
J "
Refereri~ Ai:J

Reeton. A. M. 1958. "Relationship Between Secehi Disk Readings and Light Penetration in Lake Cerco, C. E and Cole, T. 1993. ''111ree-Dimensional Eutrophiealion Mudel of Chesapeake Bay," J,
Buron:' Amer;can F;sher;e.vSociely Trans. 87:73-79. Env;ron. Engr. ASCE 119(6): 1006-1025.
Bemer, R. A. 1980. Ear/y D;ag.."e.v;.v.Prineeton University Press, Princeton. NJ. Cerco. C. F. and Cole, T. 1994. Three-Dimensional EUlrophieation Mudel of Chesapeake Bay. Vol. 1:
Rierm~n. V. J.. Jr. 197/\. "Mathematieal Model of the Seleetive Enriehment nf Blue-Green Algae by ~
Nutrient Enriehment. "In Modelil/g '!{B;ochem;ca/l'mce.v.ve.v;n Aqua/;c Ecosyslel/LV,R. P.Canale, Main Report. U.S. Anny Corps of Engineers, Waterways Experiment Station. Tech. Report
ed. Ann Amor Sdence. Ann Arbor. MI. p. 1. EL-94-4. I
Bowie. G. L.. Milis. W. B., Porcella. D. B., Camphell. C. L.. Pagenkopf. 1. R., Rupp. Chapra. S. C. 1975. "Comment on 'An Empirical Methud ofEstimaling the Retention ofPhosphorus in (
G. L.. Johnson. K. M., Chan. P. W. B., Gherini. S. A. and Chamberlin. e. E. 1985. Rates. Lakes,' by W. B. Kirchner and P. J. Dillon." Warer Resour. Res. 11: 1033-1034.
<.onstants, and Kinetic Formulations in Surfaee Water Quality Modeling. U.S. Environmental Chapra. S. C. 1977. ''Total Phosphorus Mudel for the Great Lakes." J. Env;ron. Engr. D;v. ASCE
I
Pro'eetion Ageney. ORD. Alhens. GA. ERL. EPN600/3-85/040. 103(EE2): 147-161.
Boyee. F. M. 1974. "Some AspeCISof Great Lakes Physies of Importanee to Biologieal and Chemieal Chapra, S. C. 1979. "Applying Phosphorus Loading Mudels to Embayments." Limnol. Oceanogr.
PllIeesses." J. F;.vh.Re.v.Bd. Cal/. 31:689-730. 24(1): 163-168.
Boyee. F M. IInd Hamhlin. P. E 1975. "A Simple Diffusion Mudel of the Mean Field Dislribution o} Chapra. S. C. 1980. "Appliealion of Ihe Phosphorus Loading Coneept lO the Great Lakes."ln Phosphorus
SI.luhleMllterialsin IheGreat Lakes."Lil/mo/.OceOlwgr.20(4):511-517. \ Managelllenl Slraleg;e.v for Lakes, R. C. Loehr et al., eds., Ann Arbor Science, Ann Arbor. MI.
Bmdhury. J. and Waddinglon. 1. C. B. 1973. "Stratigraphie Record ofPollulion in Shagawa Lake. North- Chapra. S. C. 1982. "Long-Tenn Models of Interaetions Between Solids and Conlaminants in Lakes."
eas{em Minnesot¡,:' In Syml'0.v;l/m 01/Q/Il/rl.."ary 1'1t1l11
Eco/agy, H. J. B. Birks and R. G. West, Ph.D. Dissertation. The University of Michigan. Ann Arbor, MI.
elis.. Blackwell. London. pp. 289-307. Chapra. S. C. 1991. "A Toxieanl Loading Concept for Organic Conlaminanls in Lakes."
Brmly. n. K.. Graves. W. L.. and Oeyer. 1. C. 1969. Surfaee Heat Exehange al Puwer Plllnt Cooling J. Env;ron. Engr. 117(5):656-677.
j
,
L¡lles. Cooling Water Discharge Projeet Report. No. 5. Edison Eleetric (nst. Publication No. 69- Chapra. S. C. and Canale, R. P. 1988. Numer;ca/ Melhods for Engitleers. 2d ed.. MeGraw-HiII. New
'JOI. N..w York. -').) York.
Bras, R. L. 1990. Hydm/ogy. Addison-Wesley. Reading, MA. Chapra. S. C. and Canale. R. P. 1991. "Long-Tenn Phenomenological Model ofPhosphorus and Oxygen .w

Broe\'k"I. H. c.. Pelermann. J.. and Siems, W. 1978. The In"uenee of Wind on C02 Exchange in a in Stmtified Lakes." Waler Re.vearch 25(6):707-715.
Wind-Wave Tunnel. J. Mar;I/e. Re.v.36(4):595-610. Chapra. S. C. and Di Toro. D. M. 1991. ''111e Della Methud for Estimating Comri1Unity Production.
Brown. L. C. and Barnwell. T. O.. Jr. 1987. The Enhaneed Stream Water Qualily Models QUAL2E and Respiration and Reaeration in Streams." J. Env;ron. Engr. 117(5):640-655.
QUAL2E-UNCAS: Documentation and Users Manual, U.S. Environmental Proteetion Ageney, Chapra. S. C. and Reekhow, K. H. 1983. Eng;neer;ng Approaclresfor Lake Managemenl, Vol. 2: Mech-
Athens. GA. Report EPi\l600/3-87/007. al/;SI;C Modelil/g. Butterworth. Wobum. MA.
Bun".1. A.. Cline. D. M.. and Lassiter. R. R. 1981. Exposure Analysis Modeling System (EXAMS): Chapra. S. C. and Sonzogni, W. C. 1979. "Greal Lakes Total Phosphorus Budget forthe Mid 1970s." J.
User Manual and System Doeumentation. Environmental Researeh Laboratory, Environmental Waler 1'01/. Conlro/ Fed. 51:2524-2533.
P'olection Agency. Athens. GA. Chapra. S. C. and Tarapchak. S. J.. 1976. "A Chlorophyll a Mudel and lis Relationship lo Phosphorus
Burns. N. M. and Rosa. E 1980. "In Silu Measurement of the Settling Velocity of Organie Carbon Loading Plots for Lakes." Waler Resour. Res. 12: 12W--1264.
!'anicles and 10 Species of Phytoplankton." Lim/lo/. OceOlwgr. 25:855-864. .. ~1 Chaudry. M. H. 1993. Open-Channe/ F/ow. Prentice-Hall. Englewood Cliffs. NJ.
Butb. T. A. and Ev¡ms, R. L. 1983. Effeets of Channel Dllms on Dissolved Oxygen Concentrations Chen, C. W. 1970. "Coneepts and IItilities of Ecologieal Models." J. San, Etlgr. D;v. ASCE 96(SA5):
in Nnrtheastern lIIinois Streams, Circular 132. State of lIIinois, Dept. of Reg. and Educ., lIIinois 1085-1086.
11
Waler Survey. Urhana. IL. Chen. C. W. and Orlob, G. T. 1975. "Ecological Simulalion for Aquatie Environments." In Syslems
C;,llahan. M. A., el al. 1979. Waler-Re/aled Fale of 129I'r;or;/)' Pol/I/lal/l.v, U.S. Environmenlal Pro- Alla/y,v;.v al/d S;I/III/O/;on ;n Ec%gy, Vol. /l/, B. C. Pallon. ed., Academie. New York, p. 475.
lec'ion Agene)'. Washinglon. DC. Reporl EPN440/4-79-029b. 2 vols. 11
Chiou, C. T., et al. 1977. "Partition Coefficient and Bioaeeumulation of Selected Organic Chemicals."
Can~lc. R. P.and Auer. M. T. 1982. "Ecological Studies and Malhematical Modeling of C/adophora in Env. Sci. Tech. 11:47~"8.
L¡.ke HUfOn:7. Mode! Verilication and System Response." J. Greal Lake.vRes. 8( 1):134-143. Chow. V. T. 1959. Open-Channe/ Hydraulic.v. McGraw-HiII. New York.
Carwle. R. P. and Emer. S. w. 1996. A Model for the Impael of Zcbru Mussels on River Water Qualily Chow. V. T.. Maidmenl. D. R.. and Mays, L. W. 1988. Applied H)'dr%gy. McGraw-HiII. New York.
Id...n manuseripl). Churchill, M. A.. Elmore, H. L., and Buckingham. R. A. 1962. "Prediclion ofStream Reaeration Rates."
Canale. R. P. and Vogel. A. H. 1974. "Effeets of Temperature on Phytoplankton Growlh." J. San. Engr. D;v. ASCE SA4:I. Proc. Paper 3199.
.1.EI/v;rO/I.Engr. /Jiv. ASCE IOO(EEI):231-241. Q Cole, T. M. and Buehak, E. M. 1995. CE-QUAL-W2: A Two-Dimensional. Laterally Averaged. Hydro-
Cm,.,le, R. P., DePalma, L. M , and Vogel. A. H. 1976. "A Plankton-Based Foud Web Model for Lake dynamic and Waler Qualily Mudel, Version 2. U.S. Anny Corps of Engineers. Waterways Exper-
Miehigan." In M(}{¡,,[;ngB;ochem;ca/ Proce.v.ve.v;1/Aql/al;c Ecosy.vlem.v,R. P. Canale. ed., Ann iment Stalion, Tech. Report EL-95-x.
AII>orSeienee. Ann Arbor. MI. p. 33. Connolly. J. P. 1991. "Application of a Food Chain Mudel to Polychlorinated Biphenyl Contamiilation
C;...oolc.R. P., Binemann. D. E, and Nachippan, S. 1974. "A Biologieal Produetion Model for Grand of the Lobster and Winler Flounder Food Chains in New Bedford Harbor." Env;ron. Sci. Techno/.
1 raverse Bay." Universily of Michigan Sea Granl Programo Teehnical Report No. 37 (Ann Arbor, 25(4):7W--770.
1\11:Universily of Michigun Sea Grant Program). Connolly. l. P. and Pedersen, C. J. 1988. "A Thennudynamic Based Evaluation ofOrganic Chemical
Canak. R. P., Owens. E. M.. Auer. M. T., and Efner. S. W. 1995. "The Validation of a Water Quality Aecumulation in Aqualie Organisms." Env;rrlll. Sci. Technol. 22(1):99-103.
Model for the Seneea River. New York." J. Water Re.wl/r. P/atl. MmragellleIll121(3):241-250. Covar, A. P. 1976. "Selecling Ihe Proper Reaeration Coefficient for Use in Water Quality Mudels."
Can"lhan. A.. Luther. 11.A.. and Wilkes. 1. O. 1969. Applied NUlller;ca/ Mel/lOd.v.Wiley. New York. Presenled at the U.S. EPA Conference on Environmental Simulation and Mudeling, Apri119-22,
('alsla"'. 11.S. and lue!'er. J. C. 1959. Crmdtlcl;ml of I/eal ;/1Solid.,. Oxford Universily Press, London. Cincinnali, OH.
CEI) f1)7K.1"/II';nmmC'llla/QI/(/liIy, 7'/1t'Nilllh AI/II//a/ R"/lOrlol/lre COl/IlCiI01/E/II';ml/l/ll'llla/ Qlll/lily.
, ' S. (jovcmmcnt Prinling Oflicc. Washinglon. De.
Dailey, J. E. and Harleman, D. R. F. 1972. Numerieal Model for the Predietion
in Estuary Networks. R.M. Parsons Laboratory. Tech. Report No. 158. MIT. Cambridge,
ofTransient Water Quality
MA.
DIIUy. J. W. und Harlernun. D. R. E 1966. F/I/;d Dytlalllic.v. Addison-Wesley, Reading. MA.
 : ..."' - - ~.1__ MiI8# Idii

824 References I{eferences ~25

Dalton, 1. 1802. "Experimental Essays on the Constitution of Mixed Gases; on the Force of Steam or
Effier, S. W. and Siegfried, C. 1994. "Zebra Mussel Populations in the Seneca River: l!TIpacton Oxygen
Vapor from Water and O!her Liquids, Both in a Torricellium Vacuum and in Air; on EV¡lporation; Resources." EI/viron. Sci. Teclmol. 28(12):2216-2221.
and on the Expansion of Gases by Heal." Manches/er Li/erary and Philosophical Socit'/Y I'".aed- Emerson, K., et al. 1975. "Aqueous J\mmonia Equilibrium Calculations: Effect of pH and TemperalUrc."
ing.f 5:536-602. J. Fish. Re.'. Bd. Cl/n. 32:2379.
Danc k werts, P. V. 1951. "Significance of Liquid-Film Coefficients in Gas Absorption." II/d. EI/g. Chem. Emerson, S. 1975. "Gas Exchange in Small Canadian Shield Lakes." l.iml/ol. Oceo//ogr. 20:754.
43(6):1460-1467.
Engelhardt, W. V. 1977. The Origil/ 01Sedimeflls l"III Sedimefllury Rock.I'.Halstead, New York.
l'
'-l Danckwens, P. V. 1953. "Conlinuous
Sci.2(1):I-13.
Flow Systems. Distribution of Residence Times." Chem. Engr.
EPA. 1989. Nalional Primary Drinking Water Regulations; Filtration and Disinfection; Turhldity;
Giardialllmbliu. Viruses, Legionella, and Heterotrophic Bacteria. Fed. Regis/. 54:27486-27541.
Deininger, R. A. 1965. "Water Quality Management- The Planning of Economically Optimal Control Eppley, R. W. 1972. "Temperature and Phytoplanklon Growth in the Sea." Fishery 811I1e/iI/70(4):1063-
Syslems." Proc. of the First Annual Meeting of Ihe American Water Resources Assoc. 1085.
Delos, C. G., et al. 1984. Technical Guidance Manual for Perforrning Wasleload Allocations, Book 11. Erdmann. J. B. 1979u. "Systematic Diurnal Curve Analysis." J. Water Pollu/. COI//rol Fed. 15(1),
Streams and Rivers, chapter 3, Toxic Substances. USEPA, Washlllgton, O.l, EPA-440/4-84-022. 78-86.
deRegnier, D. P., el. al. 1989. "Viability of Giardia Cysts in Lake, River, and Tap Water." Appl. Environ. :l Erdmann,l B. 1979b. "Simplilied Diurnal Curve Analysis." J. Ellviron. EI/gr. Div. ASCE 105(6). 1063-
Microbio/. 55(5): 1223-1229. 1074.
I
Dickinson, W. T. 1967. Accuracy of Discharge Determinati0!1s. Hydrology Paper 20, Colora- Fair, G. M., Moore, E. w., and Thomas, H. A. 1941. "The Natural Purification (¡f River Muds ulld
do State University, Fon Collins, CO. Pollulional Sediments." SelVuge "'"ksJ. 13(2):270--307; 13(4):756-779; 13«('¡:1209-122H.
Dilks, D. W., Bierman, V. J., Jr., and Helfand, 1. S. 1995. "Development and Application of Models to Fillos, J. and Molof, A. H. 1972. "Effect of Benthal Deposits on Oxygen and Nutrient Econnmy nI'
Determine Sediment Qualily Crileria-Driven Permil Levels for Metals." In Toxic Subs/al/ces in Flowing Wmers." J. Water Pollo Control Fed. 44(4):644-662. ,I
"~/er Environmen/s: Assessmell/ al/d COI//rol. WEF Special1y COI/! Series Proceedil/g~'. Cincin- 0(" ..') Fischer, H. B. 1968. "Dispersion Predictions in Natural Streams." J. Sal/. EI/gr. Di,'. ASCE 94(SA5): ,1
nati, OH, Water Environ. Fed., Alexandria, VA. 927-944.
DilIon, P. J. and Rigler, F. H. 1974. "The Phosphorus-Chlorophyll Relationship for Lakes." Liml/ol. Fischer, H. B., List, E. 1., Koh, R. C. Y, Imberger, J., and Brooks. N. H. 1979. Mi.rillg il/ 1l//{/IIt/l/lld ,1
Oceanogr. 19:767-773. Coastal Wa/en. Academic, New York.
DilIon, P. 1. and Rigler, F. H. 1975. "A Simple Method for Predicting the Capacity of a Lake for Devel- Fogler, H. S. 1986. E/emell/s olChemical Reactiol/ EI/gineering. Prentice-Hall, Englewood Cliffs, NJ.
opment Based on Lake Trophic Status." J. Fish. Res. Bd. Can. 31(9):1519-1531. Ford, D. E. and Johnson, L. S. 1986. "An Assessment of Reservoir Mixing Processes." U.S. Army c.Jrps
Di Toro, D. M. 1972a. "Recurrence Relations for First-Order Sequential Reactions in Natural Waters." of Engineers, Waterways Experimelll Station, Tech. Report. EL-86-7.
Water Resour. Res. 8(1):50--57.
FWPCA. 1966. Delaware ESlUary Comprehensive Study, Preliminury Repon ami Findings. Federill
Di Toro, D. M. I972b. "Line Source Distribution in 1\vo Dimensions: Applications 10 Water Quality." Water Pollution COlllrol Federation, Dep.. of the Inlerior, Philiadelphia, PA.
Water Resour. Res. 8(6):1541-1546.
J
Gameson, A. L. H. and Gould, D. J. 1974. Effects of Solar Radiarion on the Monality of Some Terrcs-
Di Toro, D. M. 1976. "Combining Chemical Equilibrilllll and Phytoplankton Models." In Modeling tial Bacteria in Sea Water. Proc. Inl. Symp. on Disclwrge 01SelVugeImm Seu Vut/all." Lol/dol/,
Biochemica/ Processes in Aqua/ic Ecosystems, R. P. Canale, ed., Ann Arbor Science, Ann Arbor, !(
Pergamon, Great Britain, Paper No. 22.

.
,.
MI, pp. 233-255.
o¡ Toro, D. M. 1978. "Optics
Res. 12:1059-1068.
of Turbid EstUarine Walers: Approximalions and Applications." Water
!I
Gardiner, R. D., Auer, M. T., and Canale, R. P. 1984. "Sediment Oxygen Demand in Green Bay (Lakc
Michigan)." Proc. 1984 Specialty Conf., ASCE, 5]4-519.
Gassmann,L. and Schwanzbrod,1. 1991."Waslewaterand GiardiuCysts." Wu/.:;ei. Tech. 24(2):
Di Toro, D. M. and Fitzpatrick, 1. J. 1993. Chesapeake Bay Sediment Flux Model. U .S. Arrny Corps of 183-186.
Engineers, Waterways Experiment Station, Tech. Repon EL-93-2. 't: Gaudy, A. F. and Gaudy, E. T. 1980. Micmbi%gy lor EI/virol/mel/lul Sciell/i'w' al/d El/gi/ll'ef.I'.
Di Toro, D. M., Mahony, J. D., Hansen, D. 1., Scon, K. 1., Carlson, A. R., and Ankley, McGraw-HiII, New York.
G. T. 1992. "Acid Volatile Sulfide Predicts the Acute Toxicity of Cadmium and Nickel in Sedi- ,~ Gelda, R. K., Auer, M. T., Effler, S. W., Chapra, S. C., and Storey, M. L. 1996. "D,'tcrminatilln lIf
ments." Environ. Sci. Technol. 26(1):96-101.
Reaeration Coefficienls: A Whole Lake Approach." J. Envimn. El/gr. (in press)
Di Toro, D. M., O'Connor, D. J., Thomann, R. V., and SI. John, J. P. ]98]. Analysis ofFate ofChemicals Goldman, 1. c., Oswald, W. 1., and Jenkins, D. 1974. "The Kinetics of Inorganic Carbon Lllnited Algal
in Receiving Waters. Phase 1. Chemical Manufacl. Assoc., Washington, Growth." J. Water PolloCOII/rolFed. 46:554-573.
D.C. Prepared by HydroQuallnc., Mahwah, N1. Goldman, J. C., Tenore, D. R., and Stanley. H. 1. 1973. "Inorganic Nilrogen Removal from Wastewater:
Di Toro, D. M., Paquin, P. R., Subburamu, K., and Gruber, D. A. 1990. "Sediment Oxygen Oemand Effecl on I'hytoplankton Growth in Coastal Marine Waters." Science 180:955-956.
Model: Methane and Ammonia Oxidation." J. Environ. Engr. Div. ASCE 116(5):945-986. Gutham, 1. 1. and G. Y Rhee. 1981. "Comparative Kinelic Studies of Phosphate I itllikd Growth ;;n,1
Di Toro, D. M., Thomann, R. V.,and O'Connor, D. J. 1911. "A Dynamic Model ofPhytoplankton Popula- Phosphate Uptake in Phytoplankton in Continuous Culture." J. Phycol. 17:2'\7 _:',).
tion in the Sacr.lmento-San Joaquin Delta." In Advances in Chemis/ry Series 106: Nonequilibrium Grady, C. P. L., Jr. and Lim, H. C. 19110.Biologil'llIIIJtII/ a/er Treu/ment. Marce! t k,.I.,'I, Nn\ ).. ' .
Systems in Natura/ Water Chemis/ry, R. F. Gould, ed., American Chemical Society, Washington, Graham,1. M., Auer, M T., Canalc, K. 1>.,and Hnl'ftllanll. J. P. 1982. "Ecological SI.I¡., ,,,"1' ,,,1'...
De, p.l3I. ti' matical Modeling of C/ado[l/wrlI in Lake Huron: 4. Photosynlhesis and Respiration as FUIlCI"""
Droop, M. R. 1974. "The Nutrient Status of Algal Cells in Continuous Culture." J. Ml/r. Bio/. Assoc. of Light and Temperature." J. Great Lakes Res. 8( 1):100--111.
(U.(C.) 54:825-855. Grant, R. S. and Skavroneck, S. 1980. "Comparison 01'Tmcer Methods and Predictive F.quations fur
~gleson, P.S.'1970.DynamicHydr%gy. McGraw-HilI,NewYork. . Delerrnination of Stream Reaeration Coefficients on Three Small Streams in Wisconsin." USGS.
~nger,~. a., ~rady, D. K., and Geyer, 1. C. 1974. Heat Exchange and Transpon in the Environmenl. Water Resources Investigatíon 80-19. Madison, WI.
Report No. 14, EPRI Pub. No. EA-74-049-OO-3, Electric Power Research Institute, Palo Alto, CA. Grenney, W. J., Bella, D. A., and Curl, H. C., Jr. 1973. "A Mathematical Model oflhe Nutrienl Dynamics
Edmond,~. M. and Gieskes, J. A. T. M. 1970. "On !he Calculation of !he Degree of Saluration of Sea of Phytoplankton in a Nitrate-Limited Environmenl." Biotech. BioEngin. 15:331-358,
Wl}~r With ResJ>I=Ctto Calcium CarJ¡onate Under In Silu Conditions." Geochim. Cosmochim. Actl/ Gundelach,1. M. and Castillo, J. E. 1970. "Natural Stream Purification Under Anaerobic Conditions."
H:P61+1291. J. Water PolloContr. Fed. 48(7):1753-1758.
!!26 References References . Ai1

Gupta. R. S. 1989. Hydmlogy m,d Hydralllic SY.fIEm.~.Prentice-Hall, Englewood ClifTs, NJ. Pholosynlhesis and Respiralion in an Undislurbed River Communily." Umnol. Oceanog. 19(2),
lIall. C. A. and Porshing, T. A. 1990. Nllmerical Allaly.~Üof Parlial Differelllial Eqllolioll.V.Prenlice- 305-312.
Hall, Englewood ClifTs, NJ. Kenner, B. A. 1978. "Fecal Streplococcallndicalors." In Indicators of Virnses In Waler and Food, O,
lIansen, J. S. and Ongerth, 1 E. 1991. "Effects of Time and Watershed Characterislics on Ihe Con- Berg,ed., Ann ArborScience.AnnArbor,MI. .
cenlration of Cryplo.~poridillm Oocysls in River Water:' Appl. E/II'iIYJII.Microbiol. 57(10): Kielland, J. 1937. "Individual AClivily Coefficienls of lons in Aqueous Solulions," J. Am. Chem. Soc.
2790-27lJ5. 59:1675-1678.
lIarned, 11.S. ami Davis. R.. .Ir. 1943. "The lonizalion Constanl of Carbonic Acid in Water and Ihe King, D. L. 1972. "Carbon Limilation in Sewage Lagoons." In Nulriellts and EUlrophication Speeial
Soluhility ofCarhon Dinxide in Waler ami Aqueous Sal! Solutions from O to 50°C." J. Am. Clrem. Sympo.,ia, Vol./(Am. Soco Limnol. and Oceanog.). pp. 98-112.
Soc.65:2030-2037. King, D. L. and J. T. Novak. 1974. 'The Kinelics oflnorganic Carbon-Limiled AIgal Growlh," J. Waler
lIal'l1ed.lI. S. amI lIamer. W. J. 19.n. "The lonization Constant ofWater:' J. Am. CI,em. SOl'.51:2194. Poli. Colltrol Fed. 46:1812-1816.
1I"l'I1ed,lI. S. and Seholes. S. R. 1941. "The lonization Const:1II1of IICO.,." J. Am. Cll<'m.SOl'.63: 17116- Kreider, 1 F. 1982. TIre Solar Healing Design Process. McGraw-HiII, New York.
171J9. Kreith, F. and Bohn, M. S. 1986. Prillciple.~ of Heal Transfer. 4th ed. Harper & Row. New York.
Ilarvey. U. S. 1955. 711<'CI/{'miSII:"lIIrelFal;/ily of S"illI'illt'r.Cambridge University Press, Cambridge, Lahlou, M., Choudhury, S., Wu, Yin, and Baldwin, K. 1995. QUAL2E Windows Inlerface Users Manual.
Englmul. EPA-823-B-95-003. U.S. Environmenlal Proteclion Agency, Washinglon, De.
lIenderson-Sellers, B. 1984. ElIgillet'l'illg LimllolollY. Pitman. Boston. Lam, D. e. L.. Schertzer, W. M., and Fraser, A. S. 1984. Modeling Ihe EfTects of Sedimenl Oxygen De-
llighie. R. 1935."The Rateof Adsorplionof a Pure Gas Intoa Still LiquidDuringShort Periudsof mand in Lake Erie Waler Qualily Condilions Under Ihe Innuence of Pollulion Conlrol and Wealher
Exposllre." 7'rlllU.Ame,: 11I.f1. CI,em.ElIgill.3I:365-389. Variations. As quoled in Bowie el al. (1985).
lIoffm:III, 1 D. 19'.12.Nllmerical M,'IIIOI!.,.rorEII¡¡;lIeerHllldS";"III;.fI.f.McGraw-lIill. New York. Langbien, W. B. and Durum. W. H. 1967. The Aeralion Capacily of Streams. USGS, Washington, De,
Ilornberger. G. M. atUlKdly. M. (1. 1972. "111eDetermination of Primary Pruduetion in a Slream Using Circo542.
:111Exael Soll1lion to Ihe Oxygen Balanee Equation." Wil/er Re."'IIr. 81111.8(4).795-801. Langelier, W. F. 1936. "The Analylical Conlrul of Anti-Corrosion Water Treatment." J. Am. Water Works
lIornberger. G. M. and Kelly. M. G. 1975. "Atmospheric Reaeration in a River Using Productivily Anoc.28:1500.
Analysis." J. E/II';IYIII.ElIgr. f)¡'c ASCE 101(5),729-739. Larsen, D. P. and Malueg. K. W. 1976. "Limnology of Shagawa Lake, Minnesota, Prior to Reduclion of
I1l1tehinson. G. E. 1')57. A 7'rt'llli.«'011Limllology. \f,I. l. Geoll/'llplr.l~PIrY.fic.~m,d Clremi.~lry.Wiley. Phosphorus Loading." Hydrobiol. 50:177-189. "":1
New York. Larsen, D. P. and Malueg. K. W. 1981. "Whalever Became ofShagawa LakeT' In Restoratlon ofLalces
lIydroscience. Ine. 1971. Simplilied Mathematical Modeling of Water Quality. prepared for Ihe Mitre and Inland Waler.~. IlIternalional Symposium olllnland Waters and Lalce Restorarion. U,S. Envl-
Corporation and the USEPA. Water Programs. Washington, D.e., Mar. 1971. ronmenlal Proteclion Ageney, EPA-440/4-8I-oIO, pp. 67-72.
lIydroscienee.lne. 1972. Addendllm lo Simplified Mathematieal Modeling of Water Quality, USEPA. Larsen, D. P. and Mercier. K. W. 1976. "Limnology of Shagawa Lake. Minnesola. Prior to Reduction af
Washington, D.C. Phosphorus Loading," Hydrob;ologia 50(2): 177-189.
lIydroscienee. Ine. 1973. Lil11nologiealSystel11sAnalysis of the Great Lakes: Phase I-Preliminary Larsen, D. P., Malueg, K. W., Schults, D. W.. and Brice, R. M. 1975. "Response of Eutrophic Shá-
Model Design. Hydroseienee. Ine., Westwood, NJ (now HydroQual, Ine., Mahwah. NJ). gawa Lake, Minnesota, U.S.A., lo Poinl Source Phosphorus Reduclion," Verlr. 11It. Verein. Umnol.
Ijima. T. and Tang. F. L. W. 1966. "Numerieal Calclllations of Wind Waves in Shallow Waler." In Proc. 19:884-892.
IOIIr Co",,'ol ElIgt: Cm,J., Tokyo, pp. 3!!-45. Larsen, D. P.. Mercier, H. T., and Malueg. K. W. 1973. "Modeling Algal Growlh Dynamics in Shágawa
hnboden. D. M. 1974. "Phosphorus Models of Lake ElItrophicalion." Limllol. OCt'lIIllIgr.19:297-304. Lake. Minnesota, wilh Commenls Conceming Projecled Resloralion of the Lake,"ln Modeling the
Imhoden, D. M. and Gaehter. R. 1978. "A Dynal11icModel for Trophie Slate Predietion." Ecol. Model. Elllroplricalioll Proces.f, E. J. Middlebrooks, D. H. Falkenborg. and T. E. Maloney, eds., Ann Arbor
4: 77-9R. Science. Ann Arbor, MI, p. 15.
Interna!ional Joint C'oml11ission.1979. "Oreat I.akes W:lter Quality 1978, Appendix D, Radioactivily Larsen, D. P., SchullS, D. w., and Malueg, K. W. 1981. "Summer Intemal PhosphorusSupplies in Sha-
SII!lcommittee Report." Intemational Joint Coml11ission.Windsor. Ontario. gawa Lake. Minnesola," Umnol. Oceanogr. 26:740-753.
lamil, A. 1971. "Raw Sewage Charaeteristies in Greater Beirut." M.S. Thesis. American Universily of Larsen. D. p" Van Siekle, l, Malueg. K. W., and Smith, P. D. 1979. 'The Effecl of Waslewaler Phospho-
Beirut. Lehanon. rus Removal on Shagawa Lake, Minnesola: Phosphorus Supplies, Lake Phosphonls, and Chloro-
Kang, S. W.. Sheng. y. P., and Lick, W. 19!!2."Wave Aetion and Bottom Shear Slresses in Lake Erie." phyll a." Waler Res. 13:1259-1272.
J. Grrilll""k,'.' Re.~.8(3 ):4X2-494. Laws. E. A. and Challlp. M. S. 1990. "A Microalgal Growlh Model," Umnol. Oceanogr. 35(3):597-608,
Kariekhoff. S. w., Brown, D. S., and Scott. T. A. 1979. "Sorption of Hydrophobic Pollulants on Natural LeChevallier. M. W. and Norton, W. D. 1995. "Giardia and Cryplo.fporidium in Raw and Finished
Sedil11cnls:' \VillerRe.,emy'lr 13:24 I-248. Waler," J. Am. Waler Work.f Assoc. 87(9):54--{i8.
Kavanallgh, M. C. and Trussell, R. R. 1980. "Design of Aeralion Towers to Remove Volatile Conlami- LeChevallier, M. W., el al. 1991. "Occurrence of Giardia and Cryplo.fporidillm spp, in Surface Water
nants from Drinking Water." J. Am. ~"er Work.<A.f.mc 72( I2):684--{i92. Supplies," Appl. Environ. Microbiol. 57(9):2610-2616.
KecJing. e. D.. et ¡¡l. 1982. As presenled in M. e. M¡¡eCracken and H. Moses, The Firsl Deleclion Lehman, T. D., BOlkin, D. B., and Likens, G. E. 1975. 'The Assumptions and RationaIes
of Carhon Dioxide Effects: Workshop SlImnmry R-IO June 1981, Harpers Ferry, WV. 81111.Am. of a Compuler Mudel of Phyloplanklon Populalion Dynamics," Umnol. Oceanogr. 20:
MelCtlml. So('. 63: 1165. 343-364.
Kceling. e. D.. et al. 1989. "A Three Dimensional Mode! of Atmospheric CO2 Transporl Based on Leopold,L. B. and Maddock,T. 1953.TIreHydraulic Geomelry Channels and Some Plrysiograplric
Ohserved Winds: ObservationaJ Dat¡¡ and Preliminary Analysis." Aspecl.~ of Clima/e Var;abilily Implicalioll.f. Geological Survey Professional Paper 252, Wa.~hinglon,D.e.
illll,e Pacifi<'""d 11,1'~"",'m Amer;('(f,\'.Oeophysic¡¡1 Monograph. American Geophysical Union, Lerman. A. 1972. "Slronlium 90 in the Great Lakes: Concenlralion-Time Mudel." J_ Geophys. Res.
Vol. 55. 77:3256--3264.
Keilh, L. ~I.and Telli¡¡rd. W. A. 1979. "Priority Pollutants. l. A Perspective View." Env;r"". Sci. Teclrnol. Lewis. G. N. and Randall, M. 1921. 'The Aclivily Coefficienl of Slrong Eleclrolyles." J. Am. Clrem.
I.l(4):416--42.l Soc.43:1112-1154.
Kelly. M. (i.. lIurnberger. G. M.. and Cushy. B. J. 1974. "Cuntinllous Automated Measllrement of

~~ ": ......
828 References
Rcfcrences ¡>2')

Lewis, W. K. and Whitman, W. G. 1924. Principies 01'Gas Absorption.ITld. ETlg. Chem. 16(12):1215-
1220. Mortimer, C. H. 1975b. "Modeling 01'Lakes as Physico-Biochemical Systems-Present Limitation, un..!
Needs." In Modeling of Marine Systems. J. C. 1. Nihoul. ed., Elsevier, New York, p. 217.
Liss, P. S. 1975. "Chemistry 01' the Sea Surface Microlayer." In Chem. OceaTlogr.. J. P. Riley and G.
Mueller, 1. A. 1976. Accuracy 01'Steady-Slate Finite Diffcrence Solutions. Technical Mcmol'anduII!.
Skirrow, eds., Academic, London.
Hydroscience, Inc. (now HydroQual, Inc., Mahwah, NJ).
Loucks, D. P., Revelle, C. S., and Lynn W R. 1967. "Linear Programming Models for Water Pollution
Munk, W. H. and Anderson, E. R. 1948. "Notes on a Theory 01' Ihe Thermocline." J. Mar;ne Re,.
Control." ManagemeTll Science 1.1(, I B 166-B 181. 7:276-295.
Ludyanskiy, M. L., McDonald, D., and MacNeil, D. 1993. "Impact 01' the Zebra Mussel, A Bivalve
Invader." Bioscience 43:533-544. Murthy, C. R. 1976. "Horizomal Difl'usion Characteristics in Lake Omarío." J. Physic. Oceallo¡;r.
6:76-84.
Lung. W. S. 1994. Water Quality Modeling, Vol. 3 Applica/ion to Es/uaries. CRC, Boca Ralon, FL.
Nakamura, Y. and Stefan, H. G. 1994. "Effect 01'Flow Vclocity on Sedimenl Oxygen Demand: Theory,"
J.,yman, W. J., Reehl, W. F., and Rosenblatt, D. H. 1982. Handbook of Chemical Properry Estima/ion J. Environ. Engr. 120(5):996-1016.
Methods. Environmental Behavior of Organic Compounds. McGraw-HiII, New York.
Neftel, A., Moor, E., Oeschger, H., and Stauffer, B. 1985. "Evidence from Polar Ice
Mabey, W. R., et al. 1979. Aquatic Fate Process Da/a for Organic Priority Pol/u/ants, U.S. Environ-
Cores I'or Ihe Increase in Almospheric C~ in the Past Two Centuries." Na/ure. J 15:
mental Protection Agency, Washington, DC, Report EPA/440/4-81-014. May2.
MacCormack, R. W. 1969. 'The Effecl 01'Viscosity in Hypervelocity Impact Cnllering." Am. Inst. Aero- O'Connor, D. J. and Dobbins, W. E. "Mechanism 01'Reaeration in Natural Streams." ASCE TI'IIII\'.
naut. AstronaU/. Paper 69-354. 86(SA3):35-55.
Mackay, D. 1977. "Volatilization 01'Pollutants from Water." In Aqua/ic Pol/utaTlls: Transformations and
O'Connor, D. J. 1960. "Oxygen Balance of an Estuary." J. San. Eng/: Div. ASCE 86(51\3):
Biological Effects, O. Hutzinger el al., eds., Pergamon, Amsterdam, p. 175. 35-55.
Mackay, D. 1982. "Fugacily Revisited." Environ. Sci. Teclmol. 16(12): 654A-660A.
O'Connor, D. J. 1962. "The Baclerial Dislribution in a Lake in the Vicinily 01'a Sewage Discharge." Iu
Mackay, D. and Yeun, A. T. K. 1983. "Mass Transfer Coefficients Correlations for Volatili¿¡¡tion 01'
ProceediTlgsofthe 2nd Purdue /ndustrialllilS/e Conference. WeSILafayeue.IN.
Organic Solures from Water." Environ. Sci. Technol. 17:211-233.
O'Connor, D. 1. 1967. 'The Temporal and Spalial Dislribulion 01'Dissolved Oxygen in Streams." lliller
Mackie, G. L. 1991. "Biology 01'!he Exotic Zebra Mussel, Dreissena polymorpha, in Relation to Native Resour. Re,'. 3(1):65-79.
Bivalves and lis Potential Impact on Lake SI. Clair." Hydrobiologia 219: 251-268. O'Connor, D. J. 1976. "The Concenlration 01' Dissolved Solids and River Flow." IVlI/e,.
Madore, lyf. S., el al. 1987. "Occurrence 01' Cryptosporidium Oocysts in Sewage Efftuenls and Select Resour. Res. 12(2):279-294.
Surface Waters." J. Parasitol. 73:702.
O'Connor, D. 1. 1985. Modeling Frameworks, Toxic Substances NOles, Manhauan College SummCl'
Malueg, K. W., J.,arsen, D. P., Schults, D. W., and Mercier, H. T. 1975. "A Six Year Water, Phosphorus
Inslitule in Waler Pollution Conlrol, Manhauan College, Bronx, NY.
and Nitrogen Budget 01' Shagawa Lake." J. Envirrm. Qual. 4:236-242. I
O'Connor, D. 1. 1988a. "Models of Sorplive Toxic Substances in Freshwaler SystenlS.
Mancini, J. L. 1978. Numerical Estimales 01' Coliform Mortalily Rates Under Various Condilions. J.
1: Basic Equations." J. Envir. Engr. 114(3):507-532.
Water Pol/. Control Fed. 50(11). ¡I
O'Connor, D. 1. I988b. "Models 01' Sorptive Toxic Substances in Freshwaler Systems.
McGiII, R., Tukey, 1. W., and Larsen, W. A. 1978. "Variations 01' Box Plots." Am. Stat. 32: 1
11:Lakes and Reservoirs." J. Ellvir. Engr. 1/4(3):533-551.
12-/6.
O'Connor, D. J. I988c. "Models of Sorptive Toxic Subslances in Freshwater Systcms {
McKim,1. P., Schmeider. P., and Veith, G. 1985. "Absorplion Dynamics 01'Organic Chemical Transport III: Streams and Rivers." J. Envir. Engr. 114(3):552-574.
Across Troul GiIIs as Relaled to Octanol-Water Partition Coefficients." Toxicol. Appl. Pharacol. O'Connor, D. 1. 1989. "Seasonal and Long-Term Varialions 01'Dissolved Solids in Lakes and Re,er-
~. 77:1-10.
voirs." J. Ellviron. Engr. Div. ASCE 115(6): 1213-1234.
McQuivey, R. S. and T. N. Keel'er, 1974. Simple Me!hod for Predicling Dispersion in Streams, J. Envi-
O'Connor, D. 1. and Di Toro. D. M. 1970. "Pholosynthesis and Oxygen Balance in Streams." J. Sil'/.
ron. Engr. Div. ASCE IOO(EE4):997-101l.
Engr. Div. ASCE 96(2),547-571.
M, 'alf and Eddy,lnc. 1979. Was/ell'ater Engineering. McGraw-Hill, New York.
O'Connor, D. 1. and Dobbins, 1958. "Mechanism 01'Reaeration in NatUral Streams." Tram. Am. S""
I
Mll~alf and Eddy, Inc. 1991. Was/ewater Engineering. McGraw-HiII, New York.
!
Civil Engin. 123:641-666. I
Miller, G. C. and Zepp, R. G. 1979. "Effects 01'Suspended Sediments on Photolysis Rales 01'Dissolved
O'Connor, D. J. and Mueller, 1. A. 1970. "A Water Qualily Model 01'Chloride in Ihe Great Lakes." 1.
pollutants." Water Res. 13:453--459. I
San. Engr. Div. ASCE96(SA4):955-975.
Millero, F. J. and Poisson, A. 1981. "Intemational One-Atmosphere Equalion 01' State for Sea Water."
O'Loughlin, E. M. and K. H. Bowmer, 1975. "Dilurion and Decay 01'Aqualic Herbicides in Flowll1g
Deep-Sea Res. (Part A). 28(6A):625-629. Channels." J. Hydml. 26:217-235
Milis, W. B., et al. 1985. Water Quality Assessment: A ScreeniTlg Procedur" for Toxic and Conventional
O'Melia, C. R. 1972. "An Approach to rhe Modeling 01'Lakes." Schweiz. Z Hydml. 3~: 1-34. ¡t
Pol/utanu. Par/l. Tetra Tech, Inc., Env. Res. Lab., Office 01'Research and Devel., USEPA, Alhens,
Odum, H. T. 1956. "Primary Produclion in Flowing Waters." Limnol. Oceanog. 1(2). IO~. 117.
GA, EPA-600/6-82-004a.
Officer, C. B., 1976. Pllysical Oceanography of E.l"luarie..(and Associll/ed Coas/al Wa/ers). Wiley, N<w
Moran, J. M., Morgan, M. D., and Wiersma, 1. H. 1986. ITlIroduction /0 EnvironmeTllal York.
Science. 2d oo. Freeman, New York. 1:
Officer, C. B. 1983. "Physics of Estuarine Circularion." In Ecosy,,/ems ()f tlle "".,Id. E.l"luarie"",uJ
Mortimer, C. H. 1941. ""be Exchange 01' Dissolved Subslances Between Mud and Waler. I and 11." J.
Ecol.29:280-329. ft. Enclosed Seas. B. H. Ketchum, ed., Elsevier, Amslerdam, pp. 15--41.
Okubo, A. 1971. "Oceanic Dil'fusion Diagrams." Deep-Sea Res. 18:789-802. I
t-for¡imer,!::. H. 1942. 'The Exchange 01' Dissolved Substances Between Mud and Water. lit" J. Ecol.
30:147-201.
~ Omemik, J. M., 1977. Non-Point Source-Stream Nutrient Level Relationships: A Nationwide Study.
t-fortill1er, C. ti. 1974. J.,ake Hydrodynamics.
Mortimer, C. H. 1975a. "Environmental
Mitt. Internat. Verein. Limnol. 20: 124-197.
Status of Lake Michigan Region, Vol. 2. Physical Limnology
f Corvallis ERL. ORD. USEPA, Corvallis, OR. EPA-600J3-77-105.
Oreskes, N., Shrader-Frechette, K., and Belitz, K. 1994. "Verification, Validalion and Confirrnation 01'
Numerical Models in !he Earth Sciences." Science 263(5147):641-646.
of ~e Michigan, Par! l. 'Physical Characterislb 01' Lake Michigan and lis Response lo AppliOO
fQrces." Argonne National Laboratory, Argonne, lL ANUEs-40, vol. 2.
.
l
Orlob, G. T. 1977. "Mathematical Modeling of Surface Water Impoundments, Vols. I and 11.'"NTIS,
PB-293-204.
Oswald, T. and Roth, C. 1988. "Make Us a River." Rod and Reel May/June:28.
830 References Reference~ A31'

Owens, M., Edwards, R., and Gibbs, 1. 1964. Some Reaeration Sludies in Streams. Inl. J. A;r Waler Small Streams: 1. Development of an Equilibrium-Based Simulation Modelo" Waler Resour. Rd,
Poli. 8:469-486. (in press).
Pete/son, R. O.. Page, R. E., and Dodge, K. M. 1984. "Wolves, Moose, and the Allometry of Population Runkel, R. L., McKnight, D. M., Bencala. K. E.. and Chapra, S. C. 1996b. "Reactive Soluté Trahspon
C'ydcs." Sciel/ee 224: 1350-1352. in Small Streams: 11. Simulation of a pH-Modification Experiment." Waler Resour. Res. (in llréss).
Pielou E. C. 1969. AI/ Inlrr"ll/el;OIllo MlIlhrmal;cal F.roIOllY.Wiley-Inlerscience, New York. Russell, L. L. 1976. "Chemical Aspects of Groundwater Recharge with Wastewalers." Ph.D. Disserta-
Poncc. V. M. 1989. EI/llit",eril/g Hydmlogy: Pril/ciplr.f a",II'mel;ce.f. Prenlice-Hall, Englewood Cliffs, tion, Univ. Calif., Berkeley, CA. 11
NJ. Salvato, J. A. 1982. Ellv;rrJllmental ElIg;neer;lIg and Sall;lal;oll. Wiley. New York.
Press W. H.. Teukolsky, S. A., Veuerling, W. T., and Flannery, B. P. 1992. Nl/meriCllIRecipe,f;n FOR- Sawyer, C. N.. McCarty, P. L., and Parkin, G. F. 1994. ChemÜlry for Env;rollmenta/ Eng;neering.
7RAN, 2d ed. Cambridge University Press, New York.
Rasl, W. and Lee, G. F. 1978. Summary Analysis of the North American Project (US portion) OECD
:t
f
McGraw-Hill,
Scavia,
New York.
D. 1979. ''The Use of Ecological Models of Lakes in Synlhesizing Available Infonnation and
1'lIlrophicalion Pmject: NUlrient Loading-Lake Response Relalionships and Trophic State Indices, Identifying Research Needs." In Perspecl;ves on wke Eco.fyslenr Modeling,
USEPA Corvallis Environmental Research L.1bomlory,Corvallis, OR, EPA-600/3-78-008. D. Scavia and A. Robertson. eds., Ann Arbor Science, Ann Arbor, MI. p. 109.
Rallol",", R. E., Slephens, D. W., Shuhz, D. J., and Tai. D. Y. 1978. "Lahoratory Sludies ofGas Tracers Scavia. D. 1980. "An Ecological Model for Lake Ontario." Eco/. Mude/. 8:49-78.
fm Reaeration." J. Envirrm. EI/gr. D;v. ASCE III4(EEI):2 15-229. Scavia, D. and Bennell, J. R. 1980. "Spring Transition Period in Lake Ontario-A Numerical Study
Raudl:ivi. A. J. 1979. Hydrology. Pergamon, Oxford, England. of the Causes of Ihe Large Biological and Chemical Gradients." Can. J. F;sh. Aqual. Sci. 37(5):-
Rawllc, C.. Loucks, D. P., and Lynn. W. R. (1967) "A Management Model for Water Qualily Conlrol." 823-833.
./. lVala Poli. Contml Fed. 39(7):1164-1183. Schmidt, G. D. and Roberts, L. S. 1977. FOl/lldal;on.f of Para.fil%gy. C. V. Mosby, SI. Louis, MO.
RawsolI. D. S. 1955. "Morphometry as a Dominant Faclor in Ihe Produclivity of Large Lakes." Verh. Schnoor. J. L., Sato, c.. McKechnie, D.. and Sahoo, D. 1987. Proce.fses. Coefficients. and Mude/sfor
11/1.Ver.Umno/. 12:164-175. Sinru/alillg Tox;c Orgall;c.r and Heavy Mela/s ;11Suiface Waler.f, U.S. Environmental Proteclion
Reckloow. K. H. 1977. Phosphorus Models for Lake Managemenl. Ph.D. Dissertation, Harvard Univer- Agency, Athens, GA, Report EPN600/3-87/015. 1987.
S;ly, Camhridge. MA. Schurr,1. M. and Ruchli. J. 1975. "Kinetics of02 Exchange, Photosynlhesis. and Respiration in Rivers
..~
Redho\V, K. H. 1979. "Empirical Lake Models for Phosphorus Development: Applicalions, Limitations, Deterrnined from TIme-Delayed Correlalions Between Sunlight and Dissolved Oxygen." Schwe;z. W

ami Uncertainty."ln Per.vpeclive.f01/wke ECO.fy.flemMode/itlg. D. Scavia and A. Robertson, eds.. Z Hydrol. 37(1), 144-174. I
Ann Ar!>orScience, Ann Arbor, MI. pp. 183-222. Schurr. J. M. and Ruchli. J. 1977. "Dynamics of02 and CO2 Exchange, Photosynlhesis, and Respira!ion
Redhnw. K. H. and Ch¡lpra, S. C. 1979. "Em>r Analysis for a Phosphorus Relention Model." Waler in Rivers From Time-Delayed Correlations wilh Ideal Sunlight." Unrllo/. Oceallog. 22(2): 208-
{<e.">IIr.Re.f. 15:1643-1646. 225.
Red'!¡ow. K. H. and Chapra, S. C. 1983. Engilleeril/lI ApI,roaehe.ffor ú,ke MlIl/agenrel/l. \f,lo 1: DlIIlI Schuster. R. J. 1987. Colorado River Simulation System Documentation: System Overview. U.S. Bureau I
I
A""ly.f;.f (11I(1
Emp;riCll/ Modelillll. Buuerworth, Woburn, MA. of Reclamation. Denver, CO.
Retlticld. A. c., Ketchum. B. 11.,and Richards, F. A. 1963. 'The Innuence of Organisms on the Com- Schwarzenbach, R. P. and Westall, 1. 1981. ''Transport of Nonpolar Organic Compounds from Suñace I
p"silion of Seawatcr." In The Sea, M. N. Hill, ed.. Wiley-Interscience. New York, pp. 26-77. Water to Groundwater: Laboratory Sorption Studies." Ellv;roll. Sci. Techno/. 15: 1360-1367. I
Re,,:!i, s., el al. 1991. "Modeling Ihe Risk from Gial'dia and Vinlses in Drinking Waler."J. Am. Schwarzenbach, R. P., Gschwend, P. M.. and Imboden, D. M. 1993. Ellvironmenlal Organ;c Chem;slry. I
Waler "hrh As.fOc.80:74-77. Wiley-Interscience, New York.
ReYIIC'lds,T. D. 1')82. Unil OpemliOlI.f alltl Pmeeue.f ;n Ellv;rrmnrental ElIgil/eeril/g, Brooksl Shah, 1.,1970. Ta/es oflhe Derv;.fhe.f. Dullon. New York.
Cole, Monlerey, CA. Shapiro,1. 1973. "Blue-Gre'?n Algae: Why They Became Dominanl." Science 179:382-384. I
Rkhardson, L. F. 1926. "Atmospheric Diffusion Shown on a Distance-Neighhorhood Graph." Pmc. Shugart, H. H., Goldstein. R. A., O'Neill, R. V.. and Mankin. J. B. 1974. ''TEEM: A Terrestrial Ecosys-
T/oya/ SOl'.'(A) 110:709-727. lem Energy Model for Forests." Oeco/. 1'/0111.9(3):231-264.
I
R;¡cy. G. A. 1946. "Faclors Controlling Phytoplankton Population on Georges Bank." J. Mar. Re.f. 6:- Simons, T. J. and Lam, D. C. L. 1980. "Some Limitations ofWater Quality Models for urge Lakes: A
104-113. Case Sludy of Lake Ontario." Waler Resour. Res. 16: 105-116.
Rilcy. G. A., 1956. Oceanography of Long Island Sound 1952-1954. 11.Physical Oceanography, 81/11. Smilh. J. W. and Wolfe. M. S. 1980. "Giardiasis." Amer. Rev. Med. 31 :373-383.
/li1/1I'mnr.Ocelllwg. CoUecl;oll 15, pp. 15-] 6. Smith, V. H. and Shapiro, J. 1981. A Retrospective Look at the Effecls of Phosphorus Removal in Lakes,
Rilcy, G. A. 1963. "Theory of Food-Chain Relalions in Ihe Ocean."ln The Sell, M. N. HiII, ed., Wiley- in Restoration of Lakes and Inland Waters. USEPA, Office of Waler Regulations and Standards,
Inlerscience, New York, pp. 438-463. Washington, DC. EPA-440/5-81-01O.
R..hHns, J. A. and Edginglon, D. N. 1975. "Determination of Recent Sedimentation Rates in Lake Snodgrass. W. J. 1974. A Predictive Phosphorus Model for Lakes: Developmenl and Testing. Ph.D.
MichiganUsingPb-21Oand Cs-137." Geoc!lim. Co.fllwcl,;m. Aela 39:285-301. Dissertation, Universily of North Carolina. Chape] HiII. NC.
R","ner, L. A.. Giguerre, P. R., and Evenson, D. E. 1981{l. Cnmputer Program Documentation for Snodgrass, W. J. and O'Melia. C. R. 1975. Predictive Model for Phosphorus in Lakes. Env;ron. Sci:
Stream Qualily Modeling (QUAL-II), U.S. Environmental Prolection Agency, Athens, GA, Techno/. 9:937-944.
Report EPA/600/9-81-014. Steele, J. H. 1962. "Environmental Control of Photosynlhesis in the Sea." Umno/. Oceanogr. 7: 137-150.
ROI'slIer,L. A., Giguerre. P. R., and Evenson, D. E. 1981h. Users Manual for Slream Quality Modeling Steele, J. H. 1965. "Notes on Some Theoretical Problems in Production Ecology."ln Primary Prodacl;on
(QlJAL-II), U.S. Environmenlal Protection Agency, Athens. GA, Report EPA/600/9-81-015. ;n Aqualk Ellvirollnrellls. C. R. Goldman, ed., University of California Press. Berkeley. CA.
R",c. J. B. 1988. "Occurrence and Significance of Cl'yplo.fl'oridil/m in Water." J. Am. Waler Wtlrh /' Stefan, H. G.. Cardnni, 1. J.. Schiebe, F. R.. and Cooper, C. M. 1983. "Model of Light Penelration in a
;I"""c. 80:53-58. Turbid Lake." Waler Re.four. Re.f. 19( 1): 109-120.
U"'e.1. 11.,el al. 1991. "Risk Assessmcnt al1(llhc Contrul of Waterhorne Palhogens." AIII.J. P"hlie Streeter, H. W. and Phelps, E. B. 1925. A Study of Ihe Pollulion and Natural Purification of th~ Ohio
fll'<llll,.81:709. River, 11I. Faetors Concerning the Phenomena of Oxidation and Reaeration. U.s. Public Heallh
Rl1l1kcl.R. L. 1996. Personal communicalinn. Service, Pub. Heallh Bulletin No. 146, February, 1925. Reprinted by U.S., DHEW, PHA, 1958,
Ul1l1kel.R. L.. lIencala, K. E.. Rroshears. R. E.. and Chapra, S. C. 1996(/."Reactive Snlute Transport in '\, Slomm. W. and Morgan, 1. J. 1981. Aql/mk Chemi.flry. Wiley.lnterscience, New York.
<,.,.,
, , .1
832 References
SvenJrup, H. l/., Johnson, M. W., and Fleming, R. H. 1942. The Ocealls. Prenlice-Hall,
Englewood Cliffs, NJ.
Sysk.ora, J. L., et al. 1991. "Distribulion of Giordio Cysls in Waslewater." WOlerRes. 24(2): 187-192.
Taylor, G. 1. 1953. "Dispersion 01"Soluble Maller in Solvent Flowing Slowly Through a Tube." Proc.
Royal Soco Lolldoll Ser. A 219: 186-203.
Texas Waler Development Board 1970. Simulation of Water Qualily in Slreams and Cauals, Program
Documentalion and l/ser's Manual, Auslin, TX.
Thackston, E. L. and Krenk.el P. A. 1966. "Reaeralion Predictions in Nalural Slreams." J. Sall. ElIgr.
Div. ASCE 95(SA 1):65-94.
Thomann, R. V. 1963. "Malhemalical ModelforDissolvedOxygen."J. Sall. ElIgr. Div. ASCE 89(SA5):
1-30.
Thomann, R. V. 1972. Syslems Allalysis olld Waler Quolily Ml/IlIIgemenl. McGraw-Hill, New York.
Thomann, R. V. 1981. "Equilibrium Model of Fate of Microcontaminanls in Diverse Aqualic Food
Chains." Call. J. Fish. Aqul/I. Sd. 38:280-296.
Thomann, R. V. 191!5."A Simplilied Heavy Metals Model for Slreams." (in preparalion). Manhattan
College, Bronx, NY.
Thomann, R. V. 1989. "Bioaccumulation Model of Organic Chemical Dislribulion in Aqualic Food
Chains." EIII'iroll.Sci. Techllol. 23:699-707.
Thomann, R. V. and Connolly, J. P. 1992. "Modeling Accumulation of Organic Chemicals in Aquatic
Food Webs." In Chemical DYllamics ill Freshwaler Ecosyslem~'. F. A. P. C. Gobas and J. A. Mc-
Corquodale, eds., Lewis, Boca Raton, FL, pp. 153-186.
Thomann, R. V. and Di Toro, D. M. 1983. "Physico Chemical Model of ')()xic Subslances in Ihe Great
Lakes." J. GrecllLokes Re.I'.9(4),474-496.
Thomann, R. V. and Filzpalrick, J. F. 1982. Calibration and Verificalion of a Mathematical Model oflhe
EUlrophication of Ihe Potomac Estuary; report by Hydroqual, Inc., Mahwah, NJ, lODES, Disl. Col.
Thomann, R. V. and Mueller, J. A. 1987. Prillciples of Surj;,ce lliller Qualily Modelillg olld COll/rol.
Harper & Row, New York.
Thomann, R. V. and Sobel, M. J. 1964. "Esluarine Waler Qualily Management and Forecasling." J. SOIl.
ElIgr. Div. ASCE 90(SA5):9-36.
Thomann, R. V., (onnolly, J. P., and Parkerton, T. F. 1992. "An Equilibrium Model ofOrganic Chem-
ical Accumr,ialion in Aqualic Food Webs \Vilh Sedimenl Interaclions." Ellvircm. Toxicol. Chem.
11:615-629.
Thornton, K. W. and Lessem, A. S. 1978. "A Temperature Algorilhm for Modifying Rales." Tralls. Am.
Fish. Soco 107(2):284-287.
Tsivoglou, E. C. and Neal, L. A. 1976. "'{'racer Measurements of Reaeralion: 111.Predicling Ihe Reaera-
lion Capacily of Inland Slreams." J. Wtller Pol/. COlllrol Fed. 48(12):
2669-2689.
Tsivoglou, E. C. and Wallace, S. R. 1972. Characlerizalion of Stream Reaeralion Capacily, USEPA,
Report No. EPA-R3-72-012.
Tukey, J. W. 1977 Explof{llory Dala Allalysis. Addison-Wesley, Reading, MA.
Tyler, J. E. and Smilh, R. C. 1970. Measllremelll of S/1eclrallrradiallce UlUler lliuer. Gordon and
Breach, New York.
Velz, C. J., 1931!. "Deoxygenalion and Reoxygenalion." Proc. Am. Soco Civ. ElIg/: 65(4):
677-6!10.
Velz, C. J., 1947. "Factors Inlluencing Self-Puriticalion and Their Relation lOPollution Abatemenl."
Sewage Works J. 19(4):629-644.
Vollenweider, R. A. I961!."The Scientitic Basis of Lake and Stream EUlrophication wilh Parlicular Ref-
erence 10Phosphorus and Nitrogen as EUlrophication Faclors." Technical Report DASIDSI/61!.27,
Organizalion for Economic Cooperalion and Development, Paris.
Vollenweider, R. A. 1975. "lnpul-Outpul Models wilh Special Reference 10Ihe Phosphorus Loading
Concepl in Limnology." Schweiz. Z. Hydrol. 37:53-84.
Vollepweider, R. A. 1976. "Advances in Defining Critical Loading Levels for Phosphorus in Lake Eu-
!\'Ophicalion." Mem. Isl. I/l/I. Idrobiol. 33:53-83.
Walk~r, W. W., Jr. 1977. "Some Analylical Melhods Applied 10Lake Waler Qualily Problems." Ph.D.
pissertation, Harvard UniversilY, Cambridge, MA.
References (,.13
Walker, W. W., Jr. 191!0."Variability ofTrophic Slate Indicalors in Reservoirs." In Re.,loralioll (1 r..ke.\'
alld IlIlalld llillers. l/.S. Environmental Proleclion Agency, EPA-440/5-1!1-010, pp. 344-3411.
Walker, W. W., Jr. 191!3.Personal communication <¡uoledin Brown and Barnwell (191!7).
Walsh, J. J. and Dugdale, R. C. 1972. "Nulrienl Submodels and Simulation Models 01'Phytoplanklon
Produclion in Ihe Sea." In NlIIrielllS ill Naluralllillers, H. E. Allen and J. R. Kramer, eds., Wilcy.
New York.,p. 171.
Wannink.hof, R., Ledwell, J. R., and Crusius, J. 1991. "Gas Transfer Velocilies on Lak.es Measurccl
wilh Sulfur Hexafluoride." In Symposium !'hltlme of Ihe Se('olld IlIIemalÜml/1 Crmferellce 011Gos
Trallsfer al Waler Surfaces. S. C. Wilhelms and J. S. Gulliver, eds., Minneapolis, MN.
Weber, W. J., Jr. 1972. Physicochemical Processes for Waler Qualily COlllrol. Wiley, New York..
Weslrich, J. T. and Berner, R. A. 1984. "The Role 01'Sedimentary Organic Maner in Baclerial S"ltalc
Reduclion: The G Mode! Tesled." Limllol. Oceollogr. 29(2):236-249.
Wetzel, R. G. 1975. Limllology. Saunders. Philadelphia.
Welzel, R. G. 1983. Limllology. 2d ed. Saunders, Philadelphia.
Whitman, W. G. 1923. "The Two-Film Theory of Gas Absorplion." Clcem. Melal/tlrg. ElIg. 29(4): 146-
148.
Wick.ramanayak.e,G. B., Rubin, A. 1, and Sproul, O. J. 1985. "Effecl ofOzone and Sturage TemperaHnc
on Giardia Cysls." J. Am. WOlerWorks Anoe 77:74-77.
Wilcox, R. J. 1984a. "Melhyl Chloride as a Gas.Tracer for Measuring Slream Reaeration Coefficiellls
-1. Laooratory Sludies." Waler Res. I I!(1):47-52.
Wilcox, R. J. 1984b. "Melhyl Chloride as a Gas-Tracer for Measuring Stream Reaeration CoeffiuclIls
-11. Strcam Sludies." Waler Re~'. 11!(1):53-57.
Wrighl, R. M. and McDonnell, A. J. 1979. "111-slreamDeoxygenaliol\ Rale Predicliul1."ASCE J. /:'''''.
ElIg. Div. 105:323-335.
YOlsukura, N. 1968. As referenced in preliminary repnrt Techlliqtle~'of lliller Rc'sotlrcc'sIlIve~"igl"it"lI
of Ihe U.S. Geological Survey, Measurement 01'Time of Travel and Dispersiol1 by Dye TraclI.;!.
Book 3, Chapler A9, by F. A. Kilpalrick., L. A. Martens, and J. F. Wilson, 1970.
Zepp, R. G., "Environmental pholoprocesses involving nalural organic maner." In Humic SlIbs/(/"cn
l/lid Tlceir Role ill Ihe Ellvirollmell/, F. H. Frimmel and R. F. Chrislman, eds., Wiley, New y", ~.,
1988 pp. 193-214.
Zepp, R. G. and D. M. CHne, "Rates of direcl phololysis in aquatic environmenl." Ellvirrm. Sd. Ted/lwl.
1I:359-366 (1977).
Zison. S. w., Milis, W. B., Diemer, D., and Chen, C. W. 1978. Rates, Constants, and Kinetic FomnulaliulI:.
in Surface Waler Quality Modeling. U.S. Environmental Proleclion Agency, ORD, Alhens, GA,
ERL, EPN600/3-78-105.
~
 f~,f;

ltidex
; I( :r: "¡

Acknowledgments Absorption, 7 I 5 BaselIow, 236

Accumulation, 48 Batch reactor, 24, 596-597 I!
Accuracy, 2 I2 reactions with feedback in, 114-116
Acid-base effects, 755 Bathymetry, 278
Acid-volalile sulfide, 765-767 Becquerel (Bq), 758
Activation, 752 Bed effects, and BOD removal, 359-360 .,1
Activity, 670-672 Beer-Lambert law, 507, 611, 661,744
Adsorplion, 715 Bemou\li's distribution, 179 I
Advection, 138, 180 Bessel functions, 288-292, 817-8 19
I
impulse spills, 180 Binomial distribution, 179 !
QUAL2E, 488--489 Bioaccumulation, 784, 788-790
For complete citations, see References, pages 821-834.
Advection number, 216 Biochemical oxygen demand (,<ee BOD)
AHOD. 452, 543-544 Bioconcentration, 755, 784, 785-788
ri¡¡.2.6 Reprinted from Chapra and Canale (1988; Pig. 17.7, p. 534) by permission of McGraw-
Algae, 635-636 Biomagnification, 785
Hill, New York, NY,
growth of, on a limiting nutrient, 600-601 Biotransformation, 751-753
Fig 5.7 From Fogler (1986; Fig. 9-5, p. 464); Prentice-Hall, Englewood Cliffs, NJ. in QUAL2E, 653-658 "Blue babies," 430
Fór. B 18.1 From Canale and Auer (1982; Figs, 3 and 8, p. 137); reprinted with permission of Algebraic method, 57 BOD, 353-355
JOllrll,¡1 o!Grea/ Lake.< Re.<earch; International Association for Great Lakes Research, Alkalinity, 684 carbonaceous, 355
Toronto, Ontario, Canada. Allochthonous, 298 5-day,357
I irs. 24.3-24.7 From Chapraand Di Toro(1991;Figs. J, 2. 4-6, p. 642, 644, 647, 649);Jouma/ Ammonia toxicity, 420--421, 430--431, loadings, concentrations, and rates,
680-682 357-360
o! EI/I'irollmellla/ ElIgineerÍl¡g, American Society of Civil Engineers, New York,
NY. Anaerobic, 399--40 I model for a stream, 355-357
Analytical solutions: nitrogenous, 355
I"¡¡¡s. 25.4, 25.5, 25.9 From Di Toro et al. (1990; Figs. 2-4, p. 950, 951, 960). Jauma/ o! Envi- for a diffuse source Ihat contributes both no-lIow sources, 407--409
ronmen/a/ Engineering, American Society of Civil Engineers, New York, lIow and mass, 410--415 BOD5,357
NY,
and organic decomposition, 428--429 Bode diagrams, 76
/'ir" 29.10-29,13 From Chapra and Canale (1991; Pigs. 4-7, p. 712, 713;); "Long-Term Phe- and QUAL2E, 649-653 Bottom scour, 313
nornenologicalModelof Phosphorus andOxygenin StratifiedLakcs." Water and total SOD, 468--470 Bottom sediments, 302-304
Research,25 (6): 707-715, wilh kind perrnissionfrom EIsevier ScienceLId" and water quality, 419--421 as a distributed syslem, 307-311
The Boulevard,Langford Lane, Kidlington OX5 1BG, lJK. Annual hydrograph, 236 Boulder Creek, CO, 23, 246, 257-259, 449
I ig. 33.3 From Canale and Vogel (1974; Fig, 1, p. 239); JOllrnal o! Environmelllal Engineering, Areal hypolimnetic oxygen demand, 452, Boundary conditions, 194-195
Americall Society of Civil Engineers,New York, NY. 543-544 Dirichlet, 194
Arrhenius equation, 40 Neumann, 194
I irs. 41.4, 41.9, 41.11-41.13 FrornChapra(1991: Figs. 2, 3, 5, 6. 9, p. 664,666,668,668,669); Assimilation factor, 12 Box-and-whisker plot, 723
JOllrna/ o! Enl'ironmen/a/ Engineering, American Society of Civil
Atmospheric moislure, 566-568 Brackish water, 260
Engineers, New York. NY.
AUached plants (.<eeFixed plants) Budget:
Fig 42.7 From Schwarzenbachet al. (1993; Fig. 13.6,p. 447); Wiley-Interscience,NewYork,NY, Auer, Marty, 326 contaminant, 702-705, 708-713, 771-772
Fir 45.4 FromThomannetal. (1992; Fig. 1, p. (16); "An Equilibrium Model ofOrganicChemical Autochthonous, 298 heat, 560-576
Accumulation in Aquatic Food Webs with Sediment Interaclions." Environmelllal Tox- Autotrophic, 347 solids, 701-702, 706-707, 77()-771
icology and Chemi,rtry, II :615-629, from Pergamon Press, Elsevier Science LId., The Avogadro's number, 668 Budget models, 536-538
Boulevard, Langford Lane, Kidlington OX5 1GB, UK. AVS (.<eeAcid-volatile sulfide)
Calcite precipitation, 298
Fi,: F 1 From Press et al. (1992; p. 230, 231); Cambridge University Press, New York, NY.
"Back-of-the-envelope" calculations, 79 Calibration, and Streeter-Phelps model,
Backward difference, 193 398-399
Bacteria, 289, 504-5 I1 (See a/so Model calibration)
Bacterial growth, 590--591 Canale, Ray, 326, 453
Bacterialloss rate, 506-510 Carbono 526-527
light, 507-508 nonliving organic, 636
natural mortality and salinily. 506 Carbonaceous SOD, 466-468
seuling, 508-509 Carbondioxide, 686-689
lotal loss rate, 509-510 "Cartoon" modeling, 54
834
835
 ;..:..: ,. - --

836 Index Index ~17

Cascade, 8'1-91 Cramer's mle. 102 lateral, 246-247 Estllary number, 1H9-190. J09
Cascad" 1> 'rvoir, ID, 293 Cl"ank-Nicolson melhod, 227 longiludinal, 245-246 Euler's melhod, 121-123
CBOD. 1'ryplosporidiosis, 512 numerical,201-203 EUlrophication, 420, 519. 663
CelerilY, Cryplmporidium. 512-516 QUAL2E, 489 cultural, 521
Cenlered-dllkrellcc .approach, 198-201 parvllll/.512 slream. 245-247, 489 in lIowing waters, 644 663
CE-QUAL-W2.321 CSTR, 47. 598-600 Dispersion coefficienl. 263-270 modeling with QUAL2E, 649-6SH
Channelization. etTecl of, on hydrographs. comparison wilh plug-l1ow reactor, Dissolved oxygen, 345-500 nUlrients.521-533
238 159-160 experimenl, 351-353 lemperalure. 560-5S9
Characlerislic. 414 reaclions wilh feedback in, 116 nilrogen and, 421-423 Evaporalion, 567. 571
Characlerislic curve, 175 loxics model for, 700-713 no-l1ow sources, 409 in lakes, 2S3-2S5
Chemistry.665-691 Curie (Ci). 758 saturaljon, 361-364 Explicit melhod, 212-214
equilibrium. 25, 667-682 Cysts.512 Dissolved oxygen deficil, 390 slabilily, 214-215
inorganic carbon, 683-686 balance at Ihe discharge poinl. 391-392 Exponenlialloading.71-73 ~
and mass balance, 677-682 Dalton's law of evaporation, 284, 567, 571 Dissolved oxygen sag, 348-351 Exlended Debye-HUckel
unils and conversions. 667-669 Dam, effecl of, on stream reaeralion, 382 Dislribuled-parameler syslems: approximalinn. 670
Chlorophyll. 540-543 Damkohler number. 163 time-variable solutions, 173-190 EXlensive property. 6, 561
Secchi-disk deplh correlalions. 541-543 Danckwerts, 162 sIeady-slale solmions. 156
variable models. 615-621 Danckwerts boundary condilions, 162 Distributed sOllrce Fall overturn. 579
Churchill reaeralion formula. 380 Debye-HUckellimiling law. 670 in estuaries, 170-171 Fecal coliform (FC), 50i 50S
Civil engineers. 4 Decay. of impulse spills, 1113 paramelerizalion of. 405-407 Fecal slreptococci (FS), 505
C/adophora, in Greal Lakes. 325-327 Decomposilion, 347 in slreams, 166-167 Feedback. 113-116
Closed syslems. reaclions wilh feedback in. nitrilicalion and, 428-429 Slreeler-Phelps model, 405-417 (Se" a/so Reversible reactions)
114-116 Defusing. 752 Diurnal oxygen swings, 387 Feedback syslems of reactors, 101-116
Closed-form solulions. 83 Delaware eSluary, 110 Oreissena pO/YII/OIpha.453 Feedforward reaclions in a single reaCh>l.
Colifonns, 504-511 Delta method. 442-448 "Drunkard's" walk. 177-180 95-99
Colorado River. 118-119 Denilrilicalion. 420. 478 Feedforward syslems of reactors, Sf>-9')
Column. 104 Density Eigenvalue, 58, 111, 710 mass balance aOlIsleady-slale. 8f>-91
Column veclors. 104 in boUom sediments, 303-304 Einslein (mole of pholons). 742 lime vari¡¡ble, 91-95
Comelabolism. 751 particulale maUer, 299 Electroneulrality condilion. 675 Fick, Adolf, 141
Complelely mixed syslems. 1-133 water, 271, 299 Elemenl malrix. 104 Fick's lirsllaw. 141-143
Compuler melhods, 482-500 Dependent variable, 51 Embaymenl model. 143-149 Fick's second law, 181
I
", .' QUAL2E model. 486-487 Deplh, 110w.and velocity, 247-250 eSlimalion of diffusion, 143-145 Finile-difference analysis, estuary
I sleady-slale system response malrix. Design mode. 11 steady-slale solulion. 145-146 dispersion. 264-26S
482-486 Desorplion. 716 lime-variable solulion. 146--149 Finile-dilTerence equalion
well-mixed reaclors, 120-131 Detoxification. 751 Engineers, and water qualily, 4-6 Crank-Nicolson, 227
1 "ucenlnuion. 6-8 Dew-point temperature. 567 Entrainment rate, 313-315 forward-timclcenlered-space (FTCS J. 214
Concenlration condilion. 675 Diagenesis, 457-459 Environments, 233-344 Firsl-order error analysis. 327-332
Condensalion. 571 DitTerential melhod, J 1 Ephemeralllow regime, 236 Firsl-order reaclions, 27
Conduclion. 151. 570 Diffuse sources: Equilibrium: Fixed plan!s. 326--32S. 434, 65S-662
Conduclivily. 670 with l1ow.410-417 calculations. 673-675 Flinl River, 775-776
Conservalion of mass. 13 with no l1ow.407-410 chemislry, 25, 667-682 Flow, depth and vclocilY, 247-250
Conslanl-parameler analysis. estuary Dift'usion, 137-152, 149-150. 177, 180 local, 677-682 Flow regime:
dispersion. 263-264 defined, 138. 139 Equilibrium approach, 115 ephemeral, 236
Conlinuily equation. 9 experimenl foro 1'18-141 Error funclion. 185, 820 perennial, 236
Conlinuous spills. 183-186 impulse spills, 1110 ESluaries, 260-272 Flux. 8-9
Conlinuously stirred lank reaclor (See CSTR) molecular, 138 applicalion of Ihe MFR modello. 168-171 Focusing, 297
Conlrol-volume approach: lurbulent, 138. 149-150 comparison wilh Slreams, lakes. and Ihe Food chain. 635-636
exlension lo Iwo- and Ihree-dimensional vertical. 580-585 ucean, 268-270 Food webs. 795-797
systems, 208-209 Diffusion number. 214 dispersion coefficienl. 263-270 Food-chainltoxicanl inlcrallions (,~~e
sleady-state Solulions. 192-209 Dimension. 104 nel 110w,262 Toxicanllfoud-chain inleraclions)
lime-variable solulions. 21~216 Dirdc della funclion, 66 reaeralion. 383-385 Forcing funclion, 51
Convection.570-571 Dirichlel boundary condilions, 194 slack-lide sampling, 268 Forward-lime/centered-space (FTCS)
forcel\.152 Discharge coefficienls. 248 .í\1} in. 479 difference equalion, 214
free. 152 Dispersion. 150-151. 181 ;,Ireeler-Phelps model, 40 1-403 Fourier series. 80-83
Courant condilion. 216 esluary. 263-267. 269 Iransport in, 260-262 Fundamental frequency, 80
Co~rant number. 216 impulse spills, 181 vertical slralificalion. 270-272 Fundamental qualilies, 6--10
 ...

t "
111

R3R Index t tndex 839

Gas lransfer. 369-388 Inslanlaneous or "impulse" spills. 18\1-183 Linear loading, 70-71 Methane bubble formation, 474-475
experiment, 389-391 Integralmelhod. 30 Linear prograrnming, 110 Methemoglobinemia, 430
al1<lplants. 686-{¡89 Inlensive property. 6. 561 Lineweaver-Burk plot, 593 Method of excess, 35
surf"ce renewal model. 375-378 In verse. 106 I Loading,8, 13,49 Melhod 01' half-lives, 34
Whitman's two-lilm theory. 371-.175 lonie slrenglh. 670 exponenlial. 71-73 Melhod 01' initial rates, 34
Gauss 'Iuadrature. 24() loni7.ation produet 01'water. 672 impulse, 66-68 MFR (see Mixed ftow reaclor)
Gencral solution. to mass balance e<Juation. Irreversible reaelion. 25 linear ("ramp"), 70-71 Michaelis-Menlen model, 592, 607-610,
58 (~I Isle Royale. 631-632 sinusoidal, 73-76 616,627,752
G¡tI";;,,. 512-516 Isolherms. 716-719 slep, 68-69 (See also Hall' saluralion constanls)
Gimdias¡". 512 Local equilibrium, 116.677-682 Mierobelsubstrale modeling, 590-602
Gradi"I]I. 14() Kinemalie wavc. 251-257 and ehemical reaclions. 680-682 Mieronutrienls, 522
Gral1<IRiver. MI. 447 Kinelie approach. 115 Longiludinal mixing, 245 Mineralizalion, 751
Gleal I.akes. 10, 93-95. 99-1()(). 117.325 Kinelic segmentalion. 557-559 Lolka- Vollerra equalions, 622-626 MINTEQ, 763
Gre"11l\'Iy. 154 Kinctics, 26. 669 Low-ftow analysis, 243-244 Mixed ftow reaclor, 160-164
QUAL2E.489-491 Luminescenee, 746 appliealion of, lo esluaries, 168-171
Ualf-lilc. 34. 61
IIalt .¡"turation consl:mls. 592 Lakc. 276-293
I Lumped-parameler syslems, 156 Mixed-ftow syslems, 776-717
Mixed-Iayer approach, for vertical
al!!"e. 1>27 evaporation. 283-285 MacCormack melhod, 229-230 lemperalure dislribulions, 586
liJ!ht. (, 10. 61(, morphomelry. 278-282 Maeronulrienls, 522 Mixing, 245-247
mIl' ients. ('()7-6()9 near-shore modcls. 287-293 earbon, 526-527 laleral, 246-247
to"ie:..752 reaemtion. 383-385 nilrogen, 524-526, 530-533 Mixing length, 147
Ha"..s "<Jlmlion.593 W¡lterbalance. 282-287 phosphorus, 522-524, 530-533 Model: ./
Harn'"nics. 80 Lake Erie. 292. 589. 713 silicon, 527 assessing performance, 335-338
IIeal '1,"1temperalure. 561-562 Lake Uuron. 144-147.703-705.711-713 Maerophyles (see Fixed plants) defined, 10
IIeal balance, 563-565 Lake Michigan. 2(). 78. 154.317.714. Manning equalion, 248-250, 252, 413, empirical, 12
7611-761. 762-763 488-489

,
mllltilayer.585-588 implemenlalions, 11
1;'I!l' variable. 57.'-575 Lake Onlario. 12. 63. 278. 308. 555-556. Marks, Dave, 11O mechanislic, 12
lIeill '\IIdget. 560-576 580.582-583.637-640 Mass,6-8 Model calibralion, 336-338, 398-399
He:,1c"hange. 565-571 Lake Sammamish. WA. 559 Mass balance, 47-51 Model resolulion, 339-341
IIcm)", law. 360-361 Lake SI. Clair. HXI for a diffuse souree Ihal eonlribules Model seleclion, 318-322
Ilclc1lll'eneous reaelions. 25 Lake Washington. 132 ~ both ftow and mass, 410-415 numerieal specificalion and validation,
He!e'''trnl'hic. 347 Lake Youngs. WA. 293 I for a well-mixed lake, 47 321-322
I,arge syslems 01'reaetors. 103-107
lIoblel' c<Jualion.593
U(Huog(Oneous rCilclinn.25
110""" method. 124-126
IIum.'11water Usc.el"ICct01".on hydrugraphs. 238
Laleml mixing. 245
I.aw 01'mass action. 26. 669-670
Lead-210. scdiment daling. 312
;
.
Mass balance equation, 52
feedback syslems 01' reaelors,
particular solutions, 65-83
solutions for feedforward syslems
10 1-116

of
Iheorelieal development,
Model sensilivily, 327-335
Monle Carlo analysis,
"Modeling" environmenl,
319-321

332-335
318-342
Ilydr,,!,cnmetry. 238 Leontirf. W. w.. 110 reaelors, 86-99 Molar solulion, 668
IIvdrnlysis.753-755 Liebig's law 01'Ihe minimum. 6118 sleady-stale, 86-91, 101-103 Monod model, 592
Light. 435-437. 741-742 sleady-stale solulions, 52-57 Monle Carlo analysis, 333-336
absorption.744-746 10lal solulion, 76-79 Morphomelry, lake, 278-282
Idcal g"'s law. 36(1-361
effeet 01'.on bacteria. 507-508
¡
Ideal "',,elors. 156-164 Mass ftux rale, 9 Mulliple point sources, Streeler-Phelps
l"IIg-l1ow reaclor. 157-159 extinelion in mllural waters. 441. 507. Mass loading, 8 model, 393-396
JclcII'.ilymalrix. 104 586.611,743-744 rale, 8
Iml'lic;1 approaehes. 223-228 and phyloplanklon growth, 439-442, volumetric ftow, 9 Nalural mortality, 506
Iml'oonding. 238 609-612. 615-621 Mass-lransfer mechanisms, 715-738 NBOD (see Nitrogenous BOD)
1"'l'll1mdments. 276-293 Light-bottle/dark-bollle leehnique, Malhemalieal models, 10-14 Near-shore models, 287-293
''''Plllse function. 66 438-439 Malrix, 104 steady-slale case in a bounded ftuid,
1""lIlillg.66-68 Lil1lits. 6113-605 algebra, 104-107 291-293
'p,lIs. 18(1-183 and nonpredatory losses, 613-615 inversion, 106 sleady-slale case in an infinile ftuid:
III,"''''plelcly mixl'd systel1ls, 135-294 and nutrienls. (~J7-609 mulliplieation, 105 adveclion along Ihe shoreline, 290
IIInelllcnt function. 126 lo phytoplanklon growth. 603-621 Measuremenl melhods, pholosynlhesisl no adveelion, 288-290
""I<'pendent variables. 51 mte l1Iodel,612-613 t respiration, 437-448 Nel esluarine ftow, 262
IIIeI;e;¡l"rorgani'l1Is. 504-5()6 and tcmperature, 605-607 Melalimnion, 579 Neumann boundary eondilions. 194
(1"" cnlrations. 5055()6 Limnologist. 276 , Metals, 757, 761-768, 774-776 Nilrale. 430, 477-478
1"1'"1oulput modcling. 110 Linear algebraic c<Juations. 103 Melhane balance, 4611-464 Nilrale pollution, 420, 430

I
 ,
I

1
~

840 Index Indn 841

Nitrification, 39, 355,420,421-423 Paniculate organic maller (POM), 457 Plug-lIow reaclor (PFR), 157-160 Rate, 11-10
inhibition, 427-428 Parlilion coefficient, 508, 697, 700, 721-724 comparison with CSTR, 159-160 mass l1u~,9
modeling, 426-428 Pathogens, 503-516 model, 164-167 Rate data, analysis of, 29-38
sedimen\S, 458-459, 468-474, 477 bacterial loss rale, 506-510 time-variable solulion, 173-177 differential method, 31
Nitrobacler, 422 indicator organisms, 504-506 Plug-l1ow systems, 770--776 inlegral melhod, 30
Nitrogen, 419-432, 458-459, 524-526, protozoans, 512-516 Point estimate, 238 melhod of excess, 35
530--533 sedimenl-waler interactions, 510--511 Point source, 8, 207-208 method of half-lives, 34
ammonium, 637 Peclel number, 163 in eSluaries, 168-170 melhod of inilial rotes, 34
ni trate, 637 Penelration Iheory, derivalion, 376-377 segmentalion around, 207-208 Rate law, 26
Nitrogen:phosphorus ratio, 530--533 Perennial lIow regime, 236 in streams, 164-166 Reach eSlimates, 242-243
Nitrogenous BOD, 355, 424-425 Periphylon, 658 Slreeter-Phelps model, 389-403 Reaclion, 13, 50
Nitrosonwnas, 421 (See (l/so Fi~ed plants) Poinl-slope method, 122 with feedhack, 113-116
No-l1ow sources: Penllrbation, 327 Pollulanl, muting, 256-257 heterogeneous, 25
BOD, 407-409 PFR (see PllIg-lIow reactor) Pollutant reduclion homogeneous, 25
dissolved o~ygen, 409 pH: lemporal aspects, 57-{j2 irreversible, 25
Normal distribution, 333 and iooizatioo of water, 672-{j73 Porosily: reversible, 25
Normality, 668 modeling, 683-{j91 bouom sediments, 302-303 Reaclion kinetics, 24-42
Null zone, 297 Phosphate, 477, 478-479 defined, 303 defined, 26
l Numerical differentiation, 31 Phosphorus, 522-524, 530--533 Porous media, 303 Reaction mechanisms, 734-756
Numerical dispersion, 201-203, 231 budget models, 53ó--U8 as sources of BOD, 407 Re¡IClionorder, 26
in fTCS, 216-221 loading concept, 534-559 PosilivilY, 203-205 Reaclion lypeS, 25-26
Numerical methods, 120 sedimenls, 478-479 Post-audit, 324-327 Reaeralion, 349, 376-386
for diffuse sources with 110w, 415-417 soluble reactive, 637 Predalor-prey inleraclions, 622-{j32 e~lrapolating other gases, 3115-3K6
Nutrientlfood-chain interactions, 62~32 lrophic-state correlalions, 539-544 Predictor-correclor melhod, 125 measurement of, wilh Iracers, 386- .\xK
modeling, 633--M3 Phololysis, 739-751 Pressure, elfect of, on oxygen saturalion, 0~ygen,369-388
Nutrients, 522-527, 607-609, 637 direct model, 747 363-364 photosynthesis/respiralion, 445
in QUAL2E, 653-{j58 near-surface approach, 748-751 Priority pollulants, 698-{j99 QUAL2E,491
Photons, 741 Problem specificalion, 318 slanding walers and estuaries,
Ockham, 322 Photoperiod, 436 Product of two malrices, 105 383-385
O'Connor, Doll8ld, 83 Photosynlhesis, 347,433-448 Prolozoans, 512-516 Recovery from pollulanl, 57 -62
O'Connor-Dobbins formula, 379-380 deha melhod, 442-448 concentrations in nalural waters, 513 Residence time, 55
Okubo ditlusion diagram, 150,269,343 fundamenlals, 433-434 drinking-water Irealment and acceplable Respiration, 347, 433-448
- Onondaga Lake, 22, 44 lighl, 435-437 risk levels, 514 deha melhod, 442-448
\ Oocysts, 512 measuremenl methods, 437-448 Gi(lrdialCryplosporidium model, 514-516 fundamental s, 433-434
Open systems, reactions with feedback in, 116 Photosynthesis/respiralion, 442-448 loadings, 512-513 measuremenl methods, 437-4411
Orbital velocity, 313 measuremenl melhods, 437-448 units and paramelerizalion, 434-43\
Origin, of lake, defined, 277 units and parameterization, 434-435 QUAL2E, 486-487, 649-{j58 Response time, 60--62
Outl1ow, 49 Phyloplanklon, 603 application of, 493-500 Resuspension, 312-315
Owens-Gibbs formula, 380 stream and esluary models, 647--M9 kinelics, 489-491 bouom scour, 313
Oxidationlreduction, 755 Phytoplankton growlh, 603-{j21 numerical algorithm, 491-493 orbilal velocilY, 313
O~ygen and lighl, 609-{)12 reaeralion, 491 shear Slress, 313
(see a/so Sediment oxygen demand; BOD) limits lO,603-{j05 spatial discrelization and model overview, suspended solids concclllralion ami
areal hypolimnetic demand, 452, 543-544 and nonpredatory losses, 613-{j 15 487-488 entrainment rale. .\ 11 .\15
Oxygen demand: and nUlrients, 607-609 Qnantum yield. 746-748 wave height, period, and knglh.
areal hypolimnelic, 452 and temperature, 605-{jQ7 Q,o,41 312-.\13
biochemical, 353-360, 407-409 Phytop)¡mktonlnulrient interactions, stream, Reversible reaclion. 25
sediment, 450-479 644--M9 Radialion, 565-566 in balch reaclion (dosed system),
Oxygen equivalents, in SOD process, 458 Planimeter, 278 almospheric longwave, 570 114-116
Oxygen reaeration, 369-388 Plant: net absorbed, 568-570 in CSTR, 116
formulas, 379-386 fixed, in streams, 658-662 solar shonwave, 568-569 Richardson number, 578, 5811
and gas Iransfer, 686-{)89 water longwave, 570 Rivers, 235-257
Parameter, 51 Planl activity, 434 Radionuclides, 93-95, 312, 757-761 reaeration, 379-382
penurbation, 328 Plant growth, 603-{j21 "Ramp" loading, 70--71 Iypes, 235-238
Panicular solutions for mass balance and nonpredatory losses, 613-{j 15 Random walk, 177-180 Robust, 324
equation, 65-83 Plant stoichiometry, 527-530 mathematics, 179-180 Roughness coefficients, 249
 -- - -- -

\142 Index hufel~:t

Rouling: nilrogen and, 468-470 Slreeler-Phelps equation: Tolal balance, 50 )"'! '".1),
pollulanls, 256-257 numerical model, 470-474 poinl-source, 391 Total colifonn (fe), 504 1:- '
waler, 251-256 oxidalion in Ihe aerobic zone, 464-466 sleady-stale syslem response matrix, Total Kjeldahl nitrogen, 4:14. ,1 ,~ 'y
and waler qualily, 250-257 phosphale, 477, 478-479 482-486 Total solution, to mass bálaiíce. equá~¡illf .
Rule 01'72, 62 lime-variable calculalions, 475-476 Slreeler-Phelps model, 389-403 76-79 ! -, ,"
RUlIge 1<ulla melhods, 126-128 water boundary layer, 476-477 anaerobic condilion, 399-401 Toxicantlfood-chain Interactions, 784--797 '!
Sodium sulfile, 387 analysis 01'. 396-397 chemical parameters, 790-791
Saginaw Bay, 144-147 Solar radiation, 568-570, 742-743 dislribuled sources, 405-417 integralion with mass balance, 794-795
Salinily' Solid-liquid partiliuning, 697-700 esluary, 401-403 Toxics, 693-797 ..
effel"1"r. on bacleria. 506 Solids budgel: poi nI sources, 389-403 classification01',697..{j99 "
effetloll oxygen saluralion, 363 simple. 304-307 SOO, 455-457 contrasled wilh convenlional "
Salumliol) cOllcenlralion 01'oxygen. 361 parameler eSlirn<llion,305-307 10Ial.409-410 pollution, 695..{j97
Sea~onal cyele, simulalion. fi37--{í41 Solids. suspended, 297-302 ,
StreploctJcCU.f: inorganic, 757 !
Secolld-ordcr reaclions. 29 (Su a/.ro Suspended solids) bovis, 505 ~.. .
loading concepl, 732-737'
Second.orrlrr. 751 Sorbale, 716 equinu.f, 505 model for a eSTR, 700-713 ('
(See a/.", Reversible reacliuns) Sorbenl, 716 faecalis, 505 modeling, 695-714
Sedimelll. 296-316. 795-797 Sorplinn. 697. 700, 715-727 Slronlium 90, 93-95 modeling in flowing walers. 769-783 .
Sedimelll daling. 311 "third-phase" effect. 724-727 Surface heal exchange, 565-571 nonpoint sources, 779-782
Sedimelll diagenesis, 457-459 Splraerolilus, 452, 453 Surface renewal mo<lel,375-378 Tracer studies, 186-189,384-386
Sedi",",,¡ oxygcn demalld (SOO). Spill models, 180-186 Surface renewal theory, 377 Transfer funclion, 55
450-479 inslantaneous, 180-183 Suspended solids, 297-302, 313-315 Transport, 13 .-
ob,ervali..ns, 451-455 impulse spills, 180-183 in bollom sedimenl~, 303-304 in eSluaries, 260-262
(See (//\0 SOO 1I10del) Spills.66-68, 180-183 properties, 297-300 Triangular dislribution, 333 '

Sediment solids coneenlralilln, in bollom conlinuous. 183-186 sellling and Stokes' law, 300-302 Trophic-stale correlations, 539~544
,1 sedimcnls, 303-304
Sedimenl-walcr inleraclions, 545-550
Square malrices. 104 Syslem response matrix, 107-110, 197-198, Turbulenl diffusion, 149-150 .
SI. Venant equalions. 251 482-486 approach to tnodeling tem~ratui't " ~.'. \
bacteria. 510-511 Slabilily, 212, 214-215 Syslems 01'equalions, numerical melhods, dislribulion, 586 .;,; . .;I~j'"
hypolil1l11e'icoxygen model. 546 explicit melhod, 214-215 128-130
1I 1\vo-film Iheory, 371-375 I ,\~""
lolal phosphnrus 1I10del.545 Slanding walers, reaeration, 383-385
.,cgmelllalion, 14, 340-342 Slalc-space represcnl<llion, 625 Temperate lakes. 577-580 Uniform distribution, 333
kinetic. 1\34-637 Steady-slule, 14 Temperalure: Upslream difference, 193
sratia!. 633-634 bollom sedimenls. 309-310 effecl 01',on oxygen saturalion, 362 Uptake, 609 . ' ;1.. '
SClIsiti\'ily amtlysis. 322. 327 cunlrol-volul11e<lppro<lch,192-209 modeling, 571-575 Urbanizalion, effeet 01', on hydrographs;238'
lirsl-ordcr.127-332 dislribuled-parameler syslems. 156-171 TemperalUre effecls on reaction kinelics,
;elllil1g. 50. 300 embuyment model. 145-146 40-42 Vapor pressure, 284, 567
111'bacleria. ~O8-509 near-shore models. 288-290 Temporal aspecls 01'pollulanl reduclion, 57- Velocity, flow, deplh, and, 247-250
01 parliel 301 plug-now reaclor (PFR), 157-159 62 Vertical transport, 580-585
rlTecl' 011BOJ) removal. 359 solulions. 52. 145-146 Thermal slralification. 577-589 Volatilizalion, 369-376, 727-731
7QIO.243 syslel11response matrix, 107-110, 482-486 . in temperale lakes, 577-580 Vollenweiderloadingplols,534-536 I \
"Se"'age IlIlIgUS."453 for 111.'0
eSTRs wilh feedbuck, 101-103 Thermocline, 579 Volumelric flow rate, 9 . ",:,)
:;hagav:a I..~ke.43, 293. 546-551 Stefan-Boltzmann law, 566 Thomann, Bob, 110
Shape. nI' I<lkc.dcfined. 277 SICp function, 68 Time-domain represen!ation, 624 Washout, 598
Shcar slrl"ss. 113 Slep inpul. 68 Time sbiflS, 76-79 Waler balance, steady-slate solution, fo¡'.lake.
Silicon. 52'7 Slep loading, 68-69 I
, Time-variable calculations, 475-476 282-283 .. ';
. Simulalion mode. 11 Sloichiomelry, 38-40 Time-variable solulions. 146-149 Waler qualily: .1',
Sillusoidallo'lding, 73-76 plunt, 527-530 advanced, 223-232 engineers and, 4-6
Size. or lakc defined. 277 variahle, 609 bollom sedimenls, 309-311 inlroduclion, 3-20
S"hal, Mall. 110 Slokes' law, 301. 316, 515 control-volume approach, 215-216 nitrogen and, 419-421
SOO model: "Slre,.m 01'consciousness" modeling, 54 dislributed-parameter syslems, 173-190 routing ando 250-257
al1:tlylical modeling, 45'J-470 Slreal11.235-257 embaymenl model. 146-149 Waler-qualily models, historieal
carbnll:tccoI1S,466-46R applicalion or Ihe plug-nuw reaclor model plug-now reaclor. 173-177 developmenl 01', 14-1 g .
melhal1c o<ll<luce,460.-464 lo. 164-167 simple, 212-221 Waler rouling, 251-256
mclhal1e hubblc fonllalion, 474-475 BOD model foro 355-357 111.'0
reactors, 111-113 density, 271, 579
"n'l;\''''' Slrceler-Phelps. 455-457 hydrngeomelry. 238-243 water bal<lnce for lakes. 285-287 Walerfalls, effect 01', on stream reaeration,
n;lrale. ~17 -478 rcacralion, 379 -382 1'011Reservoir. WA, 280-281 382
--~ :,;.':;;
......
- '" '- ", _-. ~_ w_ ' _.,_ _ ' , , ..:.....

844 Index

Water-quality-mndcling process, 317-327 "Whitings," 300

calibralion, 322-324 Whitman's two-lilm thcnry,
confirmalion and robustness. 324 371-375
managemcnl applicalions, 324
model selection, 318-322 Zebra mussel, 453 \.
post-audit, 324-326 Zebra mussel oxygcn demand (ZOD),
preliminary application, 322 453-454
problem specificatíon, 318 Zero-order reaclions, 27
Wave, height, period, and length, 312-313 Zooplankton:
Wavelenglh, 741 . carni voruus, 636
Well-mixed reaclors, compuler mcthnd" hcrbi vorous, 636
120-131 intcractions with phytnplankton, 626-629

,-