You are on page 1of 27

ECOLOGICAL

ENGINEERING
ELSEVIER Ecological Engineering 4 (1995) 249-275

An approach toward rational design of constructed


wetlands for wastewater treatment
Steven G. Buchberger *, George B. Shaw
Department of Civil and Environmental Engineering, Universityof Cincinnati, Cincinnati,
OH 45221-0071, USA
Received 25 June 1993; revision received 19 May 1994; accepted 12 October 1994

Abstract

A simulation-based design approach is used to find the optimal size a wetland con-
structed for wastewater treatment. The simulation scheme synthesizes submodels describing
variable wastewater loadings, atmospheric moisture fluxes, contaminant fate and transport,
and effluent release and recycle rules. The wetland is modeled as a lined densely vegetated
prismatic open channel that behaves as a non-ideal plug flow reactor (PFR) with first-order
temperature-dependent kinetics. The hydrodynamic equation for gradually varied unsteady
free surface flow is linked with the one-dimensional advection diffusion decay equation for
a nonconservative nonsorbing substance. The coupled equations are solved using an implicit
finite difference method. The flexibility of the simulation approach is illustrated with four
hypothetical examples which compare treatment performances among alternate wetland
configurations and operating strategies. Results show that ambient temperature and the
effluent release and recycle rule are much more important than precipitation and evapora-
tion in determining the size of the wetland required to meet permit limits.

Keywords: Constructed wetlands; Wetland hydrodynamics; Wastewater treatment; Optimal


design; Monte Carlo computer simulation

I. Introduction

Wetlands represent the transition zone between terrestrial and aquatic environ-
ments. M a n y studies have shown t h a t w e t l a n d s p r o v i d e effective n u t r i e n t sinks a n d

* Corresponding author.

0925-8574/95/$09.50 © 1995 Elsevier Science B.V. All rights reserved


SSDI 0925-8574(94)00053-0
250 S.G. Buchberger, G.B. Shaw/Ecological Engineering 4 (1995) 249-275

buffering sites for organic and inorganic pollutants (Hammer and Bastian, 1989).
This capability is the catalyst behind "constructed wetlands" or man-made marshes
designed to simulate a natural wetlands environment for the singular purpose of
treating wastewater (USEPA, ]988).
Similar to steel and concrete treatment systems, constructed wetlands achieve
wastewater treatment through various physical, chemical and biological processes.
Conventional facilities rely on compact energy-intensive operations with short
residence times. In contrast, wetlands are large passive systems with long detention
times. Owing to extensive surface areas, wetland performance is affected by
rainfall, evapotranspiration and temperature. Because these factors change over
time, wetlands systems do not operate as a steady-state process (Kadlec, 1989).
About 200 wetland treatment systems have been built in North America to
renovate effluent from municipal, industrial and stormwater sources (Knight et al.,
1993). Due to favorable economics, use of constructed wetlands for wastewater
treatment will likely increase. Optimal design and operation of constructed wet-
lands are hindered, however, by a poor representation of biogeochemical processes
and hydrologic mechanisms that affect wetland performance. Empirical design and
operating criteria for constructed wetlands have evolved to bridge these gaps
(WPCF, 1990). These criteria offer some guidance for preliminary design of simple
systems. However, they provide little insight into the expected performance of
complex wetland configurations subject to variable wastewater loads and intricate
release and recycle schemes.
In a broad sense, two critical processes control the treatment performance of
constructed wetlands: microbial dynamics and hydrodynamics. While microbial
processes are crucial in renovating wastewater at constructed wetlands, we do not
attempt here to model biogeochemical cycling. Instead, for illustrative purposes,
microbial and other natural processes are lumped into a black box where transfor-
mations follow simple first-order temperature dependent kinetics. We point out,
however, how more sophisticated biogeochemical models can be incorporated into
the proposed design method.
Our objective is to demonstrate a conceptualdesignprocedure for optimizing the
size of a wetland system constructed for wastewater treatment. Here optimal
design can be viewed as the least cost wetland system which produces effluent
concentrations that meet NPDES regulatory standards. For example, the proposed
procedure could help to identify the smallest wetland area required to satisfy
permitted releases of a particular contaminant to receiving surface waters. The
design strategy has three steps:

(1) identify the neighborhood of near optimal wetland candidates,


(2) simulate operation and compare performance among wetland candidates, and
(3) select a preferred choice from the competing alternatives.

The simulation step involves repeated solution of the one-dimensional advection-


diffusion decay equation describing the transport of an arbitrary nonsorbing
S.G. Buchberger, G.B. Shaw/Ecological Engineering 4 (1995) 249-275 251

nonconservative substance through the wetland. The model can accommodate


seasonal temperature effects, atmospheric moisture fluxes, variable wastewater
loadings, and site specific release and recyle schemes. We illustrate application of
the proposed method at a hypothetical location and show how it can be used to
complement current wetland design approaches.

2. Wetland design approaches

2.1. Conventional practice

Constructed wetlands are classified according to whether the water level is


above or below the substrate surface (USEPA, 1988). Free water surface (FWS)
systems transport wastewater at low velocities and shallow depths through pris-
matic open channels. Subsurface flow systems (SF) convey wastewater through
gravel filled trenches. Both systems contain dense stands of emergent aquatic
vegetation and are usually lined to prevent excessive seepage. Long narrow
channels have been advocated in the belief that slender wetland configurations
encourage ideal plug flow behavior (USEPA, 1988).
FWS and SF wetland systems are used for removal of BOD, TSS, fecal coliform
and more recently ammonia nitrogen (Conley et al., 1991). From a design perspec-
tive the problem of constructed wetlands is to determine the surface area of the
treatment cell needed to achieve a specified removal efficiency. The design
procedure is identical for both types of wetland systems and is based on two major
premises. First, constructed wetlands are assumed to be attached growth biological
plug flow reactors operating with first-order kinetics (WPCF, 1990). The governing
equation for removal of BOD is

C2
- - = ~b exp( - k r t ) (1)
C1

where C~ and C 2 a r e the influent and effluent concentrations, th is the percent of


influent mass that does not settle in the inlet area, k r is a temperature-dependent
first-order reaction rate and t is the wetland detention time. With FWS wetlands,
prevailing practice has tb ~--0.50 and kr=20 -~ 0.50 whereas with SF wetlands ~b
= 1.00 but kr=20 --- 1.10 (Reed et al., 1988).
The second major premise involves wetland hydrodynamics: flow at a depth h
and a rate Q is assumed to move through the constructed wetland under steady
uniform conditions with no short circuiting so that the hydraulic detention time is
given by

~Thlw
t= (2)
Q
252 S.G. Buchberger, G.B. Shaw/Ecological Engineering 4 (1995) 249-275

where rt is the porosity of the substrate material, l is the length and w is the width
of the treatment cell. Combining Eqs. (1) with (2) gives the required surface area
of the wetland cell,

A=lw = Q-~Thrk ln[~bc~ ] (3)

Final dimensions for the length and width of the treatment cell are chosen to
satisfy Eq. (3) subject to a specified aspect ratio, A = l:w. A n inventory of 215
treatment cells shows that the aspect ratio ranges from 0.2 to 83 at existing FWS
wetlands (Knight et al., 1993). It is now believed, that small aspect ratios are
preferable because they offer reduced construction costs and improved hydraulic
control (USEPA, 1993).
Current wetland design based on Eq. (3) is expedient. The most important input
parameter is the reaction rate constant k T. Since reaction rates decrease as
temperatures drop, winter conditions produce the largest required area. For the
same removal efficiency, treatment areas may be 300% greater for winter opera-
tion than for summer operation. Prevailing practice takes winter conditions as the
control and so significant portions of the resulting wetland system may be under-
utilized much of the year. The excess size provides a factor of safety and
operational flexibility. If the wetland is partitioned into parallel cells, then routine
maintenance on dormant cells can be performed during times when the treatment
system is not fully utilized.
The design given by Eq. (3) has many serious drawbacks though. The premises
regarding BOD removal and hydraulic detention time are an overly simplistic
description of actual conditions and are not likely to exist in practice. The
assumption of first-order BOD reaction kinetics is an empirical expedient which
has not been universally validated. Wetland systems, in fact, generate BOD by
internal decomposition processes and therefore, a production term should be
included in the governing equation. In addition, tracer studies have shown that
constructed wetlands do not exhibit ideal plug flow behavior. Pulse inputs often
lead to broad asymmetric responses (Kadlec et al., 1993) which may indicate short
circuiting (Choate et al., 1989; Fisher, 1990) or significant mixing in the longitudi-
nal a n d / o r transverse directions. In view of the observed divergence between
assumed and actual behavior, design of constructed wetlands based on the ideal-
ized conditions embodied in Eq. (3) could lead to unrealistic performance expecta-
tions.
There is yet another significant shortcoming with the current design approach:
it provides only a static solution to a dynamic problem. In reality constructed
wetlands receive variable wastewater loadings and random inputs of atmospheric
moisture and energy. Daily influent concentrations can vary by an order of
magnitude or more (Cooper and Hobson, 1989). Communities with aging wastewa-
ter collection systems frequently experience infiltration and interception problems
which flush enormous effluent loads from the collection system into the treatment
facility during rainfall events. In addition, the variability of these inputs is driven
by natural processes which operate over a wide range of time scales. For instance,
S, G. Buchberger, G.B. Shaw / Ecological Engineering 4 (1995) 249-275 253

moisture and energy fluxes to the wetland change on an hourly basis due to diurnal
processes, on a daily basis due to the passage of storm fronts and on a monthly
basis due to the annual march of the seasons.
Even the most simple constructed wetland represents a complex system of
interconnected physical, chemical, biological and hydrodynamic processes which
vary over time and across space. It seems unlikely, if not impossible, that these
simultaneous dynamic processes can be adequately represented by a single explicit
wetland design equation. These attributes, however, make the problem of wetland
design a potential candidate for solution with computer simulation using classic
Monte Carlo techniques. Simulation methods have a rich history in hydrologic
analysis where they have been used with success in the design and operation of
complex water resources systems subject to variable supplies and demands (Loucks
et al., 1981). Variations of this approach have been applied to natural wetlands
subject to wastewater inputs (Parker, 1974; Hammer, 1984). Evidently, however,
Monte Carlo simulation has seen little if any application in design and operation of
constructed wetlands. In part, this may reflect the limited understanding of
contaminant removal mechanisms and the difficulty in specifying reaction rates in
natural systems. Provided that reasonable biogeochemical models of wetland
systems can be formulated and verified, then computer simulation holds promise
as a tool for investigating the performance of wetlands constructed for wastewater
treatment.
With an eye on the continuing evolution of improved biogeochemical models,
our intent here is to demonstrate how Monte Carlo simulation can be exploited to
help optimize wetland design and operation. As our point of departure, we
consider a simple case dealing with a nonsorbing nonconservative contaminant
subject to first-order temperature-dependent reaction kinetics. More elaborate
kinetic models could be used, but this complication is neither necessary nor
justified for purposes of illustrating the conceptual approach.

2.2. Simulation-based design

Current practice yields the required size of the wetland treatment cell as output
from the design process. In contrast, simulation-based methods take wetland size
as an input parameter and then mimic operation and monitor performance. The
simulation is carried out for a sufficient period of time to estimate the long-term
treatment efficiency expected from the wetland. This entire exercise, then, is
repeated for a suite of wetland configurations in order to define the trade-off
between cell size and expected performance. From those wetland configurations
which meet permit limits, least cost alternative(s) can be selected.
Simulation-based design has shortcomings, too. The procedure imposes a signif-
icant computational burden, though increases in the speed of desk-top worksta-
tions have largely defused this issue. Simulation generates large amounts of output.
It is, therefore, essential to implement a thoughtful experimental design which
targets appropriate output variables. In the case of constructed wetlands, impor-
254 S.G. Buchberger, G.B. Shaw/Ecological Engineering 4 (1995) 249-275

p~ ~e p~ ~e Q:2

Q0 Q2 QP
m
Co

QR, C2
c*
RECYCLE
RIVER
Fig.1.Conceptualreactormodelofconstructedwetlandsforwastewatertreatment,
tant variables may include effluent concentrations and flowrates, water depths in
the pretreatment lagoon and recycle rates.
When applied properly, Monte Carlo simulation can produce a wealth of
information about the range, sensitivity, variability and reliability of treatment
efficiencies for a proposed wetland system. Simulation schemes are process synthe-
sizers. Owing to their modular structure, simulation methods are quite flexible and
easily updated. Hence, future progress on fundamental mechanisms governing the
transformation and removal of nutrients and other substances in aquatic environ-
ments can be readily incorporated into the simulation approach.

3. Problem formulation

3.1. Conceptual wetland reactor model

Many different wetland configurations for wastewater treatment have been


proposed (Steiner and Freeman, 1989). For purposes of illustration, we adopt the
simple arrangement shown in Fig. 1. The wetland system is represented as two
reactors in series, a stabilization lagoon followed by a constructed free water
surface wetland. The stabilization lagoon provides primary treatment and serves as
an equalization basin by moderating incoming municipal wastewater flows (Di
Toro, 1975) and collecting recycled wetland effluent whenever the wetland is
unable to satisfy permit limits. The chief function of the FWS wetland is to meet
permit requirements. For modeling purposes the lagoon is treated as a lined
continuously stirred tank reactor (CSTR) (Ferrara and Harleman, 1981) while the
FWS wetland is modeled as a lined one-dimensional non-ideal plug flow reactor
(PFR) (Watson et al., 1989). Other configurations could be considered, including
non-ideal reactors partitioned into active and dead storage zones. However, such
embellishments are not essential to demonstrate the proposed simulation ap-
proach.
S.G. Buchberger, G.B. Shaw/Ecological Engineering 4 (1995) 249-275 255

Depending on local climate, waste strength, reaction kinetics, and other factors,
wetland treatment efficiency may change during the course of a year. If P F R
effluent concentrations exceed permit limits, then (as shown in Fig. 1) some or all
of the effluent is recycled back to the CSTR. The recycle feature anticipates the
seasonal performance of a constructed wetland, adds operational flexibility, and
injects an element of realism (Wieder et al., 1989). With two reactors in series the
problem of wetland design involves finding the minimum combined surface area of
the CSTR and the PFR to assure that effluent concentrations comply with permit
limits, subject to reasonable restrictions on the depth of wastewater in the CSTR.

3.2. Optimal combination of CSTR and PFR

When the PFR is preceded by a CSTR which receives wastewater at a concen-


tration Co, the steady-state concentration of effluent from the combined system is
C2 Q ( O2A2 )
C0 Q + OIA1 exp Q (4)

where O i = rlikih i represents an overall net loss rate [L/T] for reactor i and all
other terms have been previously defined. Minimizing the surface area required by
the dual reactor configuration is a variation of a well-known exercise in chemical
reaction engineering where the objective is to minimize the volume of the com-
bined reactors (Fogler, 1992). Rearranging Eq. (4) gives the required lagoon area
in terms of the wetland area

al = ~1 [ oxo( Q ] (5)

This expression is valid provided

A2 < O2 In ~22 (6)

If the constraint in Eq. (6) is violated, the lagoon area A 1 becomes negative which
implies the wetland alone is sufficient for treatment purposes. The total treatment
area of the wetland system A T is the sum of the CSTR and PFR surface areas
A T ---A 1 + A 2 (7)
Eqs. 5 and 7 are plotted in Fig. 2 versus wetland area A 2 for O~ = 0.30 m / d ,
02 = 0.10 m / d , Q --- 500 m 3 / d and a ratio of influent to effluent concentrations of
C o / C 2 = 10. The total treatment area of the wetland system is minimized near
A 2 = 6000 m 2. The P F R area corresponding to the minimum is found in the usual
way by setting the derivative of Eq. (7) to zero and solving for A 2. Three distinct
cases arise and each is summarized in Table 1.
The family of seven curves in Fig. 3 illustrates the sensitivity of total treatment
area to the wetland net loss rate O 2 and provides some insight to the three
outcomes presented in Table 1. Aside from changes in the parameter O2, Fig. 3 is
256 S.G. Buchberger, G.B. Shaw /Ecological Engineering 4 (1995) 249-275

I.B

1.6~-

1.4

1.2
lal

I--
1

I,-
C
tad
0.8
e~
I.- FI1--I.~OON

O.B .....
FIT- -T{ITFL
0.4 -

0.2

o 2~
__1.
o 0.2 0.4 0.6 0,8 1 1.2 1.4 1.6 1.0 2

WEILFINtl ttIRER (ha)


Fig. 2. Minimizing the total surface area of combined lagoon-wetland treatment system.

based on the same conditions as those used to construct Fig. 2. The upper region
of Fig. 3 (above curve 2) identifies case 1 where the wetland is inefficient relative
to the lagoon (curves 1, 2 with (92 = 0.00 and 0.03 m / d ) . In this region, the

Table 1
Reactor combination that minimize the total treatment area
Case Condition Minimum area, A 1 Minimum area, A 2
02 C2 Q 1 C0 ]
(1) Lagoon only ®~ < C--o A1 = Oil [ ~ 2 -- 11 A2= 0

< < ] .,, Q r< ol


(2) Lagoon and Wetland C--o-< 0--7_<1 A, = O1 L o2 - 1 2= ~ , n [ ~ J

(3) Wetland only


o2
--®1 > 1 A 1= 0 A2 = O2
o ln[Co]
L~22]
S.G. Buchberger, G.B. Shaw/Ecological Engineering 4 (1995) 249-275 257

2.T

2.4

•~ 2.1
r-

e
,,, 1.8
c
F.-
Z
ta 1.5
E
I--
I,tl
CC
~- 1.2
-I
C
I--
C)
i- 0.9

B.8

0.3

B 8.2 0.4 O.B 0.8 1 1.2 1.4 1.6 1.8 2

Fig. 3. Effect of wetland net loss rates on total treatment area. (1) 0 2 = 0.00 m / d , (2) 02 = 0.03 m / d ,
(3) 0 2 = 0.05 m / d , (4) ®2 = 0.10 m / d , (5) 0 2 = 0.20 m / d , (6) 0 2 = 0.30 m / d , and (7) O 2 = 0.50 m / d .

minimum total treatment area occurs at the left edge of the plot where the wetland
area is zero. Hence, the smallest treatment system is a lagoon only.
Conversely, the lower region (below curve 6) corresponds to case 3 where the
lagoon is inefficient relative to the wetland (curves 6, 7 with O2 = 0.30 and 0.50
m/d). In this lower region, the total treatment area is minimized by eliminating
the lagoon and using only the wetland. The shaded region in the middle of Fig. 3
corresponds to the case where both the lagoon and the wetland effectively
compete to treat effluent (curves 3, 4 and 5 with 0 2 = 0.05, 0.10 and 0.20 m / d ,
respectively). Here the minimum total area consists of a lagoon and a wetland
whose relative sizes depend on the overall system efficiency and the individual net
loss rates. In practice, the lagoon/wetland combinations in the shaded region of
Fig. 3 are the most useful since natural treatment operations are usually designed
to include both units.
Based on the expressions in Table 1, Table 2 summarizes the combinations of
lagoon and wetland areas that minimize total treatment areas for the seven cases
258 S.G. Buchberger, G.B. Shaw/Ecological Engineering 4 (1995) 249-275

Table 2
Condition for minimum size of total treatment areas in Fig. 3
Number (92 / (91 A1 A2 Ar Case
1 0.000 15 000 0 15000 (1) Lagoon only
2 0.100 15000 0 15000 (1) Lagoon only
3 0.167 8330 5110 13 440 (2) Lagoon and wetland
4 0.33 3330 6020 9350 (2) Lagoon and wetland
5 0.667 830 4740 5570 (2) Lagoon and wetland
6 1.000 0 3840 3840 (3) Wetland only
7 1.667 0 2810 2810 (3) Wetland only

given in Fig. 3. These are not final values but rather starting points for a more
detailed analysis of required wetland size.

3.3. Release and recycle rule

T h e wetland release and recycle ( R / R ) rule is intimately c o n n e c t e d to the


operating policy of the t r e a t m e n t system and has a p r o f o u n d impact on the design.
The R / R rule specifies the rate at which wastewater is passed from one reactor to
a n o t h e r including off-site releases m a d e during an "excursion" or instance where
the final effluent c o n c e n t r a t i o n exceeds the permitted limit. It is not our intention
to identify an optimal R / R rule from m a n y possible options. Instead, we c o m p a r e
three simple rules to give an idea of the different approaches which can be
considered:

R / R Rule No 1: Constant release/full recycle - Flow from C S T R to P F R is


constant; if an excursion occurs, all effluent is recycled back to the C S T R .
R / R Rule No 2: Constant release/equivalent permitted load on recycle - Flow
from C S T R to P F R is constant; if an excursion occurs, effluent is released to
surface waters at a rate to match the downstream concentration that would
result if the wetland were discharging at the maximum permitted concentration.
T h e balance is recycled back to the CSTR.
R / R Rule No 3: Seasonal release/equivalent permitted load on Recycle - Flow
from C S T R to P F R is seasonal; excursions handled according to rule 2.

T h e first rule is easy to visualize. It is, however, quite rigid and tends to g e n e r a t e
very high volumes of recycled water. Consider, for example, two scenarios that may
occur at a wetland operating u n d e r release rule n u m b e r one with a permitted
c o n c e n t r a t i o n of say C e = 20 m g / l :

Scenario No 1 : 0 2 = 500 m 3 / d a y and C 2 = 19.99 m g / l < C,, (no excursion) and


Scenario No 2." Q2 = 500 m 3 / d a y and C 2 = 20.01 m g / 1 > C e (excursion).
S.G. Buchberger, G.B. Shaw/EcologicalEngineering4 (1995)249-275 259

Under the first scenario, all effluent is released to surface waters while in the
second scenario all effluent is recycled back to the CSTR. These are dramatically
different responses and yet the total mass released to surface waters differs by only
5 mg per day (0.1%).
The second R / R rule attempts to balance this discrepancy. Suppose that the
wetland is discharging Q2 = 500 m 3 / d a y at C 2 = Ce -- 20.00 m g / l into a river. This
scenario corresponds to the maximum permitted discharge under rule one. Assum-
ing complete and instantaneous mixing, the wetland effluent will produce a
maximum concentration C * in the river given by
QsCs + QzCp
c* = (8)
Qs + Qz
where Qs and Cs are the flow and concentration in the river upstream from the
effluent discharge point. The second R / R rule finds the maximum rate of
discharge Qp that is permissible during an excursion at concentration C 2 > Cp so
that the resulting well-mixed concentration in the river will not exceed C * given in
Eq. (8). Under this condition the fraction of effluent flow that can be released
from the treatment system to the river is given by

QP PIP3( 1 -P2)
= (9)
Q2 (1 + P l ) --(I+plPz)P3
where Pl is the ratio of streamflow to total discharge from the wetland (Qs/Q2),
p2 is the ratio of stream concentration to the permitted concentration (C s /Cp),
and P3 is the ratio of the permitted concentration to the effluent concentration
(Cp/C2).
Eq. (9) is plotted in Fig. 4a,b for P2 = 0.00 and P2 = 0.50 for six values of /91.
Permissible wetland discharges during an excursion drop as the flow in the river
decreases or as the ambient concentration in the river increases. The most
favorable conditions for R / R rule number 2 correspond to high river flows
(Pl ~ 10) with low ambient concentrations in the river (P2 ~'~0). Under these ideal
conditions, the permitted discharge to the river is roughly Qp = P3Q2. Following
the example above (Q2 = 500 m 3 / d a y and Cp = 20 m g / l ) suppose C 2 = 25 rag/l,
Qs = 5000 m 3 / d a y and Cs = 10 rag/1. With Px = 10, P2 = 0.50 and P3 = 0.80, Eq.
(9) gives the permissible discharge to the river as Qe = 323 m 3 / d a y . The balance
of the wetland discharge (177 m 3 / d a y ) is recycled back to the lagoon. R / R rule 2
is quite forgiving and leads to downstream loadings similar to those expected from
rule 1 at fully permitted levels.
While rule 2 is an improvement over rule 1, permitted effluent concentrations
may be subject to seasonal constraints. When cold weather extremes dictate the
wetland design, seasonal effects can be moderated with a seasonal release rule.
The idea is to make above average releases from the CSTR during warm weather
in order to deplete storage and make room for effluent that accumulates from
below average releases during cool weather. R / R rule 3 offers this refinement in
conjunction with the equivalent permitted load from rule 2.
260 S.G. Buchberger, G.B. Shaw / Ecological Engineering 4 (1995) 249-2 75

1
0.9
0.8
0.7

v
0.8
0,5

(21
0.4
0.3
0.2
0.1
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0,7 0.8 0.9 1
C(P)/C(2)

0.9

O
\
0.8
0.7
o.s
0.5
0.4
0.3
0.2
!ii!iii i!ii!!i!ir i
0.1

0 i . . . . i . . . . t . . . . J . . . . i . . . . i . . . . i . . . . i . . . . i . . . . i . . . . t

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1


C(P)/C(2)
Fig. 4. M a x i m u m e f f l u e n t d i s c h a r g e from c o n s t r u c t e d w e t l a n d to river b a s e d on e q u i v a l e n t p e r m i t t e d
load for (1) Px = 10.0, (2) P1 = 1.00, (3) Pl = 0.50, (4) Pl = 0.25, (5) Pa = 0.10, a n d (6) Pl = 0.02.

The release rules considered here illustrate just three of many possible operat-
ing policies that may improve wetland operation. Implicit in these rules is the need
to monitor the receiving water body and the wetland effluent on a continuous
basis. Real time control of constructed wetlands is a significant departure from
prevailing practice. However, the necessary technology is available to provide
continuous monitoring and to adjust off-site releases accordingly. More work is
needed to optimize release and recycle rules. The potential for improving treat-
ment performance of constructed wetland systems is promising.
We have described a conceptual dual reactor wetland model, demonstrated a
method to minimize total treatment area and compared three R / R rule options.
In the next sections, we present the governing transport equations for the wetland
system and then show how simulation links these concepts and aids in the design
of constructed wetlands for wastewater treatment.
S.G. Buchberger,G.B. Shaw/EcologicalEngineering4 (1995)249-275 261

3.4. Conservation equations for pretreatment lagoon

Simulation of wetland performance requires simultaneous solution of the conti-


nuity and momentum equations for water flow and the constituent mass balance
equations for the CSTR and the PFR. In what follows, the subscript 0 denotes the
wastewater source, while 1 and 2 denote the first reactor (CSTR) and second
reactor (PFR) in series. Referring to Fig. 1, the water balance for the lined CSTR
subject to precipitation p and evaporation e can be written
dhl Q0 - Q1 + Qn
= +p -e (10)
dt A(hl)
where A(h 1) is the CSTR surface area at wastewater depth h~, Q0 is the
wastewater supply to the CSTR, 01 is the release from the CSTR to the PFR, and
QR is the recycle rate from the P F R back to the CSTR. If the final effluent
concentration from the wetland meets water quality standards, then QR = 0;
otherwise 0 < QR < Q2 depending on the adopted release rule.
The constituent mass balance for a nonconservative substance subject to first-
order reaction kinetics in the CSTR is
dC 1 1
- -dt = V( hl----
~ [Q0(C0 - C1) + QR( CR -- C1) ] - [~bI + k~ ]C 1 (11)

where
A(hl)
qJl V(hl-----~(P- e ) (12)

and V(h~) is the wastewater volume in the CSTR at depth hi, C O is the concentra-
tion of the influent to the CSTR and C R = C 2 is the recycle concentration. The
temperature-dependent first-order decay rate, k l, reflects losses due to settling,
decay and other processes in the CSTR.
The term qs1 defined in Eq. (12) represents a climate-dependent dilution rate
and can be positive or negative depending on the net flux of atmospheric moisture
at the CSTR. Whenever precipitation exceeds evaporation, 01 is positive and
hence concentrations in the CSTR are reduced by the dilution effect of rainfall
added to the water column. Conversely, when evaporation exceeds precipitation,
I]/1 is negative and concentrations are increased by the net loss of water from
storage in the lagoon. Hence; Eq. (11) accounts for mixing of influent wastewater
with CSTR storage and P F R recycle and also reflects changes in CSTR wastewater
concentrations due to first-order reaction kinetics and net evaporation.

3.5.Conservation equations for FWS constructed wetland

The continuity equation for gradually varied unsteady flow through a wide lined
prismatic FWS wetland subject to precipitation and evaporation is
i}h2 i)q2
r/ at + ax = p - e (13)
262 S.G. Buchberger, G.B. Shaw/Ecological Engineering 4 (1995) 249-275

in which h e is the depth of flow and q2 is the discharge per unit width at any
cross-section along the PFR. The porosity term r/ accounts for the reduction in
cross-sectional area due to dense stands of vegetation in the FWS wetland.
Estimates of porosity values in existing FWS wetlands range from 86 to 98%
(Herskowitz, 1986; Watson and Hobson, 1989). A simplified form of the momen-
tum equation for gradually varied unsteady flow without lateral inflow is
0h 2
= S o - ST (14)
0x
where S o is the channel slope and S f is the friction slope. Combining Eqs. (13) and
(14), leads to the diffusion wave approximation for gradually varied unsteady open
channel flow (Stephenson and Meadows, 1986; Ponce, 1989)
~q i)2q ~q
-- =D -c k +ck(p-e ) (15)
Ot Ox 2 -~X
where D is the hydraulic diffusivity and c k = a/3h t~-I is the wave celerity given as
the slope of the stage-discharge rating curve. The parameters a and /3, respec-
tively, represent the flow resistance and the flow regime (i.e. laminar, transition or
turbulent). Rough estimates of a and /3 values for wetland environments are
available in the literature (Turner et al., 1978; Shih and Rahi, 1982; Kadlec, 1990;
Maheshwari and McMahon, 1992). Although the expressions given here apply to a
FWS wetland system, a similar strategy is readily formulated for a SF wetland
system where Eq. (15) would be replaced with the Boussinesq equation (Strack,
1989) describing transient shallow unconfined groundwater flow.
The one dimensional transport of a nonsorbing nonconservative solute through
the wetland is given by the advection-diffusion equation including first-order
effects for decay and dilution
aC 2 02C2 aC 2
Ot = E 2 ~x 2 u2 0-'x-- [0z+k2]C2 (16)

Here E 2 is the longitudinal dispersion coefficient, u 2 is the mean flow velocity in


the x direction, k 2 is a temperature-dependent first-order reaction rate and 02 is
a climate-dependent first-order dilution rate to account for the effects of net
evaporation at the wetland,
p--e
02 (17)
h2(x)
Again, it should be emphasized that in Eqs. (11) and (16), we have adopted a
first-order decay rate f o r illustrative purposes only. Many other kinetic models are
available and could be used to describe solute decay (or growth) over time. The
important point here is that the formulation has the flexibility to account for these
mechanisms and other natural processes, provided the appropriate rate constants
can be identified.
As presented here, the governing equation includes current design practice as a
special case. If longitudinal dispersion, precipitation and evaporation are neglected
S.G. Buchberger, G.B. Shaw/Ecological Engineering 4 (1995) 249-275 263

(E 2 = 0; ~/2 = 0), then Eq. (16) reduces to the plug flow reactor with first-order
kinetics. The steady state solution for this simple PFR model appears in Eq. (1).
The wetland treatment system has been visualized as a series of two reactors
that responds to continuous external inputs, including variable municipal wastewa-
ter loadings and random fluxes of atmospheric moisture and energy. The following
section describes a numerical simulation approach for solving Eqs. (10) through
(16) in a way to help identify optimal configurations and operating strategies for
cost effective natural treatment systems.

3.6. Monte Carlo simulation

The governing equations are solved with a program called HydroSim which uses
Monte Carlo simulation imbedded in an implicit numerical scheme. Some key
features of the simulation program are mentioned here. Depths and concentra-
tions in the CSTR (Eqs. 10 and 11) are estimated with a modified Euler method.
Flows and concentrations along the wetland (Eqs. 15 and 16) are found using a
Crank-Nicolson implicit finite difference method (Chapra and Canale, 1988) linked
to a root finding subroutine for non-linear equations. Upstream boundary condi-
tions at the wetland are given by releases from the pretreatment lagoon. Down-
stream boundary conditions use a zero gradient for concentrations and a flow
rating curve for discharges at the wetland outlet. Municipal effluent loadings to the
wetland system are generated using an autoregressive time series model (Box and
Jenkins, 1976). Site specific daily air and water temperatures are obtained using
methods of Richardson (1981) and Edinger et al. (1968). Daily evapotranspiration
is simulated with a Priestley-Taylor type equation (Jensen et al., 1990) and then
distributed over daylight hours with a truncated sine function. Night-time evapo-
transpiration is assumed to be negligible (Malek, 1992). Daily precipitation is
modeled as a Markov chain mixed exponential process (Hanson et al., 1992) and
then disaggregated to hourly amounts (Hershenhorn and Woolhiser, 1987). To
preserve negative cross-correlation between evaporation and precipitation, both
are taken to be mutually exclusive on an hourly basis. Further details on simulation
of site-specific climatic fluxes at the constructed wetland are given by Shaw (1993).
The general procedure followed by HydroSim to simulate a FWS wetland system is
summarized in 10 steps:

(1) Specify input parameters, initial conditions and wetland release and recycle
rule.
(2) Simulate air and water temperatures, evapotranspiration and precipitation
rates.
(3) Generate wastewater Q0 and contaminant C O loadings from the city to the
lagoon.
(4) Solve Eq. (10) to estimate the new depth of wastewater h~ in the lagoon.
(5) Solve Eq. (11) to estimate the new concentrations C 1 in the lagoon.
(6) Release wastewater at rate Q1 and concentration C~ from lagoon to the FWS
wetland.
264 S. G. Buchberger, G.B. Shaw/Ecological Engineering 4 (1995) 249-275

(7) Solve Eq. (15) to find depth h 2 and flow Q2 along the FWS wetland.
(8) Solve Eq. (16) to find effluent concentrations C2(x) along the FWS wetland.
(9) If an excursion occurs, determine amount of effluent to recycle back to the
lagoon.
(10) R e p e a t steps (2) to (9) to simulate and monitor several years of wetland
operation.

4. Simulation: examples and discussion

Four examples follow to demonstrate the simulation approach for wetland


design. Each assumes two lined reactors in series, a lagoon (CSTR) followed by a
free water surface wetland (PFR) which discharges to a river. Wastewater from the
town is generated at a rate of 500 m 3 / d a y . Influent concentration are 200 m g / I
for a nonconservative nonsorbing substance subject to temperature-dependent
first-order decay. The permit limit is 20 m g / l .
The climate is warm with an average annual water temperature of 20°C and a
minimum winter temperature that averages 10°C. At 20°C, the net loss rates are
O~ = 0.20 m / d a y in the lagoon and ~}2 = 0.10 m / d a y in the wetland. Using a
t e m p e r a t u r e correction factor 1.05 for both reactors, the net loss rates adjusted for
winter conditions are:

®I(T = 10°C) = 0.20(1.05) 1°-2° = 0.1228 m / d a y .


O2(T = 10°C) = 0.10(1.05) 1°-2° = 0.0614 m / d a y .

4.1. Case no. 1: Conventional design - no recycle

Assuming that the CSTR achieves a minimum removal efficiency of f l = 75%


(Metcalf and Eddy, 1991), the maximum effluent concentration from the CSTR is
about C 1 = 50 mg/1. This implies that the wetland must achieve a minimum
removal efficiency of f2 = 60% to meet the permit limit of C 2 = 20 mg/1. The
areas required for the lagoon and the wetland are

AI = -, t . , i 0.1228 1-0.75 1 = 12215 m 2

A 2 = O 2 2 In - 0.0614 In 1 - 0 . 6 0 =7462m 2

The total required treatment area is A T = 19677 m 2.

4.2. Case no. 2: Conventional design - m i n i m u m total area, no recycle

From the given information, 0 2 / 0 1 = 0.50 and C 2 / C o = 0.10. This corresponds


to case (2) in Table 1 where the minimum total treatment area includes a lagoon
S.G. Buchberger, G.B. Shaw/Ecological Engineering 4 (1995) 249-275 265

and a wetland with individual areas given by

hi= 1 = 0.~812- 1]=4072m 2

e [ 0 2 C o] 500 l n [ ( O . 5 ) ( l O ) ] = 1 3 1 0 6 m 2
A2 = ~22 In[ O---~] 0.0614

The minimum total treatment area is A T = 17 178 m e. This estimate is 13% less
than the previous approach. Depending on the nature of the contaminant, it may
be necessary to limit the areal loading to the lagoon. If, for instance, the maximum
allowable loading is 125 kg ha-1 d a y - 1 , then the lagoon area must be increased to
at least A 1 = 8000 m 2. The corresponding wetland area drops to A 2 = 9900 m:
giving a total treatment area of A T = 17900 m 2, still 9% less than case no. 1.

4.3. Case no. 3: Simulation-based design - R / R rule no. 2

In contrast to the first two cases, here the size of the wetland system must be
specified as input to the simulation program. In addition, more details are needed
about the local climate, wastewater characteristics and the release rule. Average
precipitation at the site is taken as 1400 m m / y r and average evapotranspiration
(ET) is 1250 m m / y r . Simulated sequences, plotted in Fig. 5, show that daily
precipitation is relatively uniform throughout the year while daily E T exhibits a
distinct seasonal pattern. Their joint distribution preserves negative cross-correla-
tion since daily E T tends to fall below the seasonal average on days with rainfall.
Wastewater characteristics are summarized in Table 3. Variability has been
added to both the inflows and the concentrations while autocorrelation has been
added to the inflows only. The effects are clear in Fig. 6 which shows simulated
sequences of normally distributed daily flow rates and daily concentrations enter-
ing the lagoon for a 1-year period. Both sequences fluctuate about their respective
means with identical coefficients of variation at 20%. The influent concentrations
with a lag-1 autocorrelation coefficient of zero are purely random over time. By
comparison, the daily inflows with a lag-1 autocorrelation coefficient of 0.70 show
a tendency to cluster. That is, daily inflows that are above-average (e.g. days 110 to
135) or below-average (e.g. days 300 to 315) tend to occur in groups. These
properties of the influent series are included here to further illustrate the flexibil-
ity of the simulation approach.
Based on hypothetical loading limits identified in case 2, we keep the lagoon
area at A 1 = 8000 m 2 and try a wetland size of A2 = 9 0 0 0 m 2. This combination
gives a total area of 17000 m 2, close to the minimum treatment area from the
previous example. Since cell dimensions must be specified for simulation, we
assume a square lagoon (89.5 m on a side) with a nominal depth of 4 m and a
wetland 225 m long by 40 m wide with a nominal operating depth of 0.20 m.
Porosities in the lagoon and in the wetland are taken as unity. These specifications
imply a first-order decay rate of k I = 0.05 day-1 in the lagoon and k 2 = 0.50 day-1
in the wetland, both at 20°C. The first R / R rule is not considered here since
266 S.G. Buchberger, G.B. Shaw/Ecological Engineering 4 (1995) 249-275
100 "

90 ]

00

,0 , i ''r ' --
80 :-

0 , • . . . . • , , • , . , , , .

0 30 80 90 120 150 100 210 240 270 300 330 360


DRY

10 _.' ?..i ! ~ : ! ' ' ' i i ? i i.! : 7 ! ~ ? .I i . ~ ~. i ~ ! ' T I.i ' ! _


b
9 . . . . . . . . . . . . . . . . . . . . . . . . . . : .......... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

! .

iz °4
5

I.-
0
3
0 .

i . . . . .~ , , . i . . .I . i . , i . . . . . . . "; :i i

0 30 80 90 120 150 100 210 240 270 300 330 380


DAY
Fig. 5. Simulated daily atmospheric moisture fluxes for a 1-year period. (a) Precipitation and (b)
evapotranspiration.

previous work revealed that it generates enormous recycle volumes (Shaw, 1993).
To implement R / R rule 2, we take Pl = 10.0, P2 = 0.50 (see Eq. 9) and fix the
lagoon release at 510 m 3 / d a y , slightly higher than the mean inflow to account for
effluent recycle and net input from precipitation minus ET. The simulations start
in the spring season and are performed over a 3-year horizon.

Table 3
Wastewater characteristics used in HydroSim examples
Characteristic Mean Standard Lag-1 A C F
deviation
Inflow rate 500 m 3 / d a y 100 m 3 / d a y 0.70
Influent concentration 200 m g / l 40 m g / l 0.00
S.G. Buchberger, G.B. Shaw/Ecological Engineering 4 (1995) 249-275 267

88e

700

<
E
v
500
..d
kl.

3Be
p..
--t 200
IM
I---
O~
.............. .............. :.......... i
100 : . . . . . . . . . . . . . i . . . . . . .

71 i
0 . . l , , i , , i , , l . , i . . i . , i , , i , . i . , i . , t . , i

0 30 60 90 120 150 180 210 ;)40 270 300 330 360


DRY

400
~b
350

7"

c~ 250

i....................................i..............
H

200
z
tu
(J
150

100

50
.... [ ....................... : ..............................................................................................................
i • . i , . , . , t . , i , • L , • i , . , . . , , . , . . t • . t . , ,

0 30 60 90 120 150 180 210 240 270 300 330 380


DRY
Fig. 6. S i m u l a t e d a v e r a g e d a i l y i n f l u e n t f o r a 1 - y e a r p e r i o d . ( a ) F l o w r a t e a n d ( b ) c o n c e n t r a t i o n s ,

Simulated daily average wastewater releases from the wetland and daily recycle
rates from the wetland back to the lagoon are shown in Fig. 7a. Outflows range
from a seasonal base of about 400 m3/day to spikes reaching 1400 m3/day during
rain events. In many instances, the peak outflow more than doubles the dry
weather flow and demonstrates that rainfall input to the wetland often exceeds the
inflow from the lagoon. These outflow spikes represent only direct precipitation
into the wetland. Runoff from surrounding areas and storm surges from infiltration
and interception are not included in this example.
While the effects of daily ET are less pronounced than rainfall, ET nonetheless
does introduce a noticeable seasonal cycle to the wetland release. As expected, the
period of minimum outflow from the wetland coincides with the season of
maximum ET. Recycle episodes occur each winter and last about 45 days with
average rates of 80 m3/day. The difference between the wetland outflow and the
268 S.G. Buchberger, G.B. Shaw/Ecological Engineering 4 (1995) 249-275

1600

1400
x

(
E 1200
v

lOOO
000

600

400

w
200
0 i .... i .... i .... J .... i .... i .... i .... i .... ~ .... t .... i .... J

0 100 200 300 400 500 800 700 000 900 1000 1100

DRY

i ,¸.,, , .... i .... i .... , .... i .... ~ .... i .... i .... , .... i .... i
25 • ....... .... ....... ........ : ..... ..... :

20

0 100 200 300 400 500 600 700 800 900 1000 1100

DRY

Fig. 7. Simulated average daily effluent for case 3 for a 3-year period. (a) Wetland releases and recycle
and (b) concentrations.

lagoon recycle is the rate at which effluent is released from the treatment system
to the river.
Fig. 7b shows simulated effluent concentrations leaving the wetland. The
random fluctuations of the influent (Fig. 6b) are effectively damped out during
residence in the C S T R and PFR. The primary source of variability in the effluent
is seasonal with frequent but temporary dilution from rainfall. As expected from
the problem set-up, seasonal t e m p e r a t u r e effects dominate the treatment perfor-
mance of this wetland system. Most of the year, the wetland complies with the 20
mg/1 limit. During warm months, effluent concentrations drop below 2 mg/1.
Depending on the type of contaminant, this value may be unrealistically low since
natural processes in the wetland may actually generate residual levels of the
contaminant. The wetland experiences an extended excursion each winter which
coincides with the recycle episodes identified in Fig, 7a.
S. G. Buchberger, G.B. Shaw / Ecological Engineering 4 (1995) 249-275 269

4.4. Case no. 4: Simulation-based design - R / R rule no. 3

This case differs from the previous case in two key respects. First, a seasonal
release rule is used to determine the flow of wastewater from the lagoon to the
wetland. Essentially the seasonal rule links release with temperature - warm
months have above average outflows while cool months have below average
outflows. Second, the trigger on the recycle rule is reduced from 20 to 16 mg/l. In
this way, the treatment system initiates recycling before the wetland experiences an
excursion. The net effect of these two operational changes is to dampen the
seasonality in the effluent concentrations noted in Fig. 7b. All other aspects of this
example, including system configuration, simulated loadings, rainfall and ET are
identical to case no. 3. As before, the simulation starts in spring and extends for 3
years. Daily average values for simulated effluent releases and concentrations are
in Fig. 8.
As expected, simulated daily effluent releases in Fig. 8a follow a seasonal
pattern with most flows ranging from about 300 to 1000 m3/day. Superimposed on
the seasonal release pattern are deviations at both the high and low ends. At the
high end, rainfall events produce temporary spikes approaching 1600 m3//day. At
the low end, a contingency plan in the seasonal release rule reduces minimum
outflows to about 200 m3/day. The contingency plan curtails releases to the
wetland whenever effluent levels in the lagoon drop below a specified threshold.
This precaution is necessary to prevent the lagoon from draining completely.
Seasonal effects due to daily ET are present but not easy to discern since they
are masked by the annual cycle of daily effluent releases. Recycle episodes occur
each winter and last about 120 days at an average rate of 65 m3//day. Compared to
case 3, this case has a slightly lower recycle rate but a much longer recycle time
and, hence, its total recycle volume is greater.
In Fig. 8b, simulated effluent concentrations display a moderate seasonal
pattern. The range of 2 to 24 mg/1 noted in case 3 has been compressed mainly to
the interval 10 to 20 mg/l. At the upper end, simulated effluent levels comply with
the permit limit. At the lower level some effluent concentrations drop below 10
m g / l when the seasonal release rule is superseded by the contingency release plan.
The resulting low outflows from the lagoon lead to long wetland detention times
which reduce concentrations to low levels seen in the plots. As before, influent
variability has been smoothed. Precipitation events generate brief dilution blips on
the seasonal pattern of effluent concentrations. The magnitude of the rain-induced
dilution can be significant, but it is short-lived. The overall performance of this
wetland system is controlled by the interaction of seasonal temperature effects and
the seasonal release and recycle rule.

4.5. Further comparisons

Two additional 3-year runs with HydroSim were made to investigate the
performance of the wetland configurations obtained from cases 1 and 2. In both
instances, the lagoon release was held at 500 m3/day, the mean effluent inflow.
270 S.G. Buchberger, G.B. Shaw/Ecological Engineering 4 (1995) 249-275
1600 _' .~iii'''i~'!i'' ''~ i! . . . . ''i!7 ' .... ' .... Ii'' _

1400

~ 1200

1000

800

600
400
200
-#

0 i . . . . i . . . . i J . . . . J , . , , , , , , , , , , i , . . , i . . . . i , , , ,

0 100 200 300 400 500 600 700 000 900 1000 1100

DRY

r . . . . i . . . . , . . . . , . . . . , . . . . r . . . . f . . . . i . . . . i . . . . ~ . . . . f . . . . f
25 . . . . . . . . . . . . ". . . . . . .
b •
_3
20

o
P-I 15

I--
Z
ul 10
(J
Z
o

. . . . , . . . . , . . . . i . . . . , . . . . J . . . . J . . . . , . . . . , . . . . , . . . . , . . . . ,

0 100 200 300 400 500 600 700 800 900 1000 1100
DAY
F i g . 8. S i m u l a t e d a v e r a g e d a i l y e f f l u e n t f o r c a s e 4 f o r a 3 - y e a r p e r i o d . ( a ) W e t l a n d r e l e a s e s a n d r e c y c l e
and (b) c o n c e n t r a t i o n s .

Recycle was not permitted. Simulated daily average effluent concentrations, plot-
ted in Fig. 9a,b, show episodes of rainfall dilution superimposed on a seasonal
cycle. Both wetland systems comply well with the permitted effluent limit, though
case 2 has more excursions than case 1.
Recycle versus non-recycle treatment options are compared in Table 4. The top
four entries in this Table represent input parameters to HydroSim; the remaining
entries are simulation results. Case 1 with the largest treatment area and longest
detention time performs exceptionally well having only five simulated excursions in
3 years. Case 2 may not be feasible due to the disproportionately small pretreat-
ment lagoon. Case 3 may suffer from too many excursions, though none are too
severe and the release rule is controlled to mitigate the downstream impact. Case 4
with the shortest detention time and only four effluent excursions offers a
competitive alternative. Although recycle is required, the total treatment area for
S.G. Buchberger, G.B. Shaw/Ecological Engineering 4 (1995) 249-275 271

25

z is

~ 10

0
0 100 200 300 400 500 800 700 800 900 1 0 0 0 1100
DAY

2i i .... i
20
v

z is

lO

0
0 100 200 300 400 500 800 700 800 900 1000 1100
DAY
Fig. 9. Simulated average daily effluent concentration for cases 1 and 2 for a 3-year period.

case 4 is 13% less than the treatment area for case 1. Continued refinements in the
seasonal release and recycle rule are likely to improve treatment performance and
may lead to further reductions in total area. Ultimately, economic factors or other
project constraints will determine which of the competing wetland candidates is
the preferred treatment option at a particular site.
Some general observations about FWS wetlands can be gleaned from this
exercise. From a hydrologic perspective, rainfall events cause temporary increases
in flow and dilutions in concentrations relative to the prevailing seasonal outflow
and effluent patterns. Evapotranspiration, while not pronounced on a daily basis,
can measurably reduce effluent outflows and increase effluent concentrations over
the course of a season. However, the overall impact of these hydrologic factors on
effluent concentrations is minor compared with seasonal temperature effects.
From an operating perspective, wetlands that receive nearly constant flows and
272 S.G. Buchberger, G.B. Shaw/Ecological Engineering 4 (1995) 249-275

Table 4
Simulation results from common 3-year runs with HydroSim
Feature Units Case 1 Case 2 Case 3 Case 4
Lagoon surface area m2 l 2100 4096 8000 8000
Wetland surface area m2 7480 13 120 9000 9000
Total treatment area m2 19580 17216 17000 17000
Maximum lagoon depth m 4.62 5.07 4.59 4.81
Average lagoon depth m 4.20 4.08 3.96 2.21
Minimum lagoon depth m 3.81 3.26 3.36 0.25
Average system storage m3 52 300 19 300 33 500 19 500
Average through flow m3/day 500.4 501.2 511.7 520.6
Average detention time day 104.5 38.5 65.5 37.5
Number of excursions day 5 27 133 4
Minimum effluent output mg/l 1.35 0.62 1.39 2.11
Average effluent output mg/l 8.68 7.62 9.90 13.62
Median effluent output mg/l 7.01 5.36 7.81 13.37
Maximum effluent output mg/l 20.68 21.95 24.09 2 l. 10
Average annual recycle volume m3/yr 0 0 3500 8000
Average annual recycle time day/yr 0 0 44 122
Average recycle rate m3/day 0 0 79.5 65.6

maintain steady depths appear susceptible to large swings in removal efficiency


dictated by seasonal changes in water temperature. If sized to meet critical winter
conditions, these wetland systems may be inefficient for most of the remaining
year. In such cases, it appears possible to improve wetland efficiency or to reduce
cell size by adopting seasonal release and recycle rules that moderate effluent
concentrations as demonstrated in case 4. Ultimately, optimal design and opera-
tion of constructed wetland treatment systems may require strict on-site control of
effluent loadings, real time monitoring of receiving water quality and accurate
near-term forecasts of precipitation and temperatures.

5. Conclusions

Simulation-based procedures have been used with success to gain insight to the
behavior of complex water resources systems. Such techniques may also offer
potential advantages in design and operation of wetlands constructed for wastewa-
ter treatment. The formulation presented here is based on fundamental conserva-
tion principles. Individual submodels describing municipal wastewater loadings,
hydrologic moisture and energy fluxes, contaminant fate and transport, and efflu-
ent release and recycle rules have been synthesized in a conceptual model of
wetland behavior.
Some elements of the simulation model are reasonably well represented and
calibrated. For example, progress in hydrology over the past two decades has
S.G. Buchberger, G.B. Shaw/Ecological Engineering 4 (1995) 249-275 273

produced tractable and reliable schemes for generating precipitation, evapotran-


spiration, and solar radiation over short and long time horizons in diverse climates.
These advances provide a very good picture of site specific atmospheric moisture
and energy fluxes. However, other critical components of the simulation model are
still very primitive. There is a great need to improve our understanding of
fundamental hydro-biogeochemical processes which mediate the transformation
and transport of substances placed into wetlands. Owing to the modular construc-
tion of simulation-based procedures, it is relatively easy to update the model by
incorporating relevant advances in biogeochemical dynamics.
The simulation approach demonstrated here has drawbacks too. It carries a
high computational burden and generates lots of output. While meaningful output
providing insight to system performance can be generated with thoughtful experi-
mental design, simulation results may lead to a false sense of legitimacy. On the
question of validity, it is very important to recognize that successful execution of
the model does not imply that the simulation is correct or that it even properly
accounts for the key physical processes. At best, simulation-based analyses provide
only a rough approximation of the real system behavior. Like any other tool, the
burden rests with the user to recognize the inherent strengths and limitations of
simulation-based design procedures. Considering, the great uncertainty that still
surrounds wetland treatment mechanisms, it is probably most prudent at this point
to interpret simulation output in a relative way and use the information as a basis
to compare competing alternatives.
Simulation-based design offers tremendous flexibility to the user. Without
having to build a treatment system, it is possible to compare the effectiveness of
different release and recycle rules, to investigate the efficiencies of various reactor
configurations, to examine the loading impact of a steady versus a growing town
population, to evaluate the feasibility of reliability-based effluent standards, and to
carry out sensitivity analyses on how treatment performance is affected by assumed
treatment mechanisms, wetland operating depths or cell aspect ratios. The list of
potential applications is large. In the right circumstances, simulation approaches
can contribute significantly to improved decision making in the design and opera-
tion of constructed wetlands for wastewater treatment.

Acknowledgements

This work was supported in part by grants from the US Department of Energy
E R / W M Young Investigator Program and from the State of Ohio Water Re-
sources Center. Insightful comments from several reviewers improved our initial
draft of this paper. We gratefully acknowledge the assistance of Clayton Hanson
and David Woolhiser, both of the USDA Agricultural Research Service, for
providing advance copies of ARS software to generate rainfall, evaporation and
temperatures at sites across the US. Ranjit Jadhav and Srikant Hemmady assisted
us in developing the HydroSim source code.
274 S.G. Buchberger, G.B. Shaw / Ecological Engineering 4 (1995) 249-275

References

Box, G.E.P. and G.M. Jenkins, 1976. Time Series Analysis. Holden-Day, Oakland, CA.
Chapra, S.C. and R.P. Canale, 1988. Numerical Methods for Engineers. McGraw-Hill Book Company,
New York, 811 pp.
Choate, K.D., G.R. Steiner and J.T. Watson, 1989. Demonstration of constructed wetlands for
municipal wastewater, March to December 1988. First Semiannual Monitoring Report,
T V A / W R / W Q - 8 9 / 5 , Chattanooga, TN.
Conley, L.M., R.I. Dick and L.W. Lion, 1991. Assessment of the root zone method of wastewater
treatment. Res. J. WPCF, 63: 239-247.
Cooper, P.F. and J.A. Hobson, 1989. Sewage treatment by reed bed systems: the present situation in
the United Kingdom. In: D.A. Hammer (Ed.), Constructed Wetlands for Wastewater Treatment.
Lewis Publishers, Chelsea, MI, pp. 153-171.
Di Toro, D.M., 1975. Statistical design of equalization basins. ASCE J. Environ. Eng., 101: 917-933.
Edinger, J.E., D.W. Duttweiler and J.C. Geyer, 1968. The response of water temperatures to
meteorological conditions. Water Resour. Res., 4: 1137-1143.
Ferrara, R.A. and D.F. Harleman, 1981. Hydraulic modeling for waste stabilization ponds. ASCE J.
Environ. Eng., 107: 817-830.
Fisher, P.J., 1990. Hydraulic characteristics of constructed wetlands at Richmond, NSW, Australia. In:
P.F. Cooper and B.C. Findlater (Eds.), Constructed Wetlands for Water Pollution Control, Perga-
mon Press, Oxford, pp. 21-31.
Fogler, H.S., 1992. Elements of Chemical Reaction Engineering, Prentice Hall, Englewood Cliffs, New
Jersey, 838 pp.
Hammer D.A. and R.K. Bastian, 1989. Wetlands ecosystems: natural water purifiers?. In: D.A.
Hammer (Ed.), Constructed Wetlands for Wastewater Treatment. Lewis Publishers, Chelsea, MI.
Hammer, D.E., 1984. An Engineering Model of Wetland/Wastewater Interactions. Ph.D. Dissertation,
University of Michigan, MI, 139 pp.
Hanson, C.L., K.A. Cumming, D.A. Woolhiser and C.W. Richardson, 1992. Microcomputer Program
For Daily Weather Simulation in the Contiguous United States. USDA-ARS Technical Report,
Submitted to U.S. Department of Agriculture, Springfield, VA.
Hershenhorn J. and D.A. Woolhiser, 1987. Disaggregation of daily rainfall. J. Hydrol., 95: 299-322.
Herskowitz, J., 1986. Listowel Artificial Marsh Project Report. Ministry of the Environment, Ontario,
Canada, 253 pp.
Jensen, M.E., R.D. Burman and R.G. Allen, 1990. Evapotranspiration and Irrigation Water Require-
ments. ASCE Manuals & Reports Engineering Practice No.70.
Kadlec, R.H., 1989. Hydrologic factors in wetland water treatment. In: D.A. Hammer (Ed.), Con-
structed Wetlands for Wastewater Treatment. Lewis Publishers, Chelsea, MI.
Kadlec, R.H., 1990. Overland flow in wetlands: vegetation resistance. ASCE J. Hydraul. Eng., 116:
691-706.
Knight, R.L., R.W. Ruble, R.H. Kadlec and S.C. Reed, 1993. Wetlands for wastewater treatment:
performance database. In: G.A. Moshiri (Ed.), Constructed Wetlands for Water Quality Improve-
ment. Lewis Publishers, Boca Raton, FL.
Loucks, D.P., J.R. Stedinger and D.A. Haith, 1981. Water Resource Systems Planning and Analysis.
Prentice Hall, Englewood Cliffs, N J, 559 pp.
Maheshwari, B.L. and T.A. McMahon, 1992. Modeling shallow overland flow in surface irrigation.
ASCE J. Irrig. Drain., 118: 201-217.
Malek, E., 1992., Night-time evapotranspiration vs. daytime and 24 hour evapotranspiration. J. Hydrol.,
138: 119-129.
Metcalf & Eddy, Inc., 1991. Wastewater Engineering: Treatment Disposal and Reuse. McGraw-Hill,
Inc., New York, 1334 pp.
Parker, P.E., 1974. A Dynamic Ecosystem Simulator. Ph.D. Dissertation, University of Michigan, MI,
193 pp.
Ponce, V.M., 1989. Engineering Hydrology. Prentice Hall, Englewood Cliffs, NJ, 640 pp.
S.G. Buchberger, G.B. Shaw / Ecological Engineering 4 (1995) 249-275 275

Reed, S.C., E.J. Middlebrooks and R.W. Crites, 1988. Natural Systems for Waste Management and
Treatment. McGraw-Hill Book Company, New York.
Richardson, C.W., 1981., Stochastic simulation of daily precipitation, temperature and solar radiation.
Water Resour. Res., 17: 182-190.
Shaw, G.B., 1993. Effect of Climate on Performance of Constructed Wetlands for Wastewater
Treatment. MS thesis, University of Cincinnati, 100 pp.
Shih, S.F. and G.S. Rahi, 1982. Seasonal variations of Manning's roughness coefficient in a subtropical
marsh. Trans. Am. Soc. Agric. Eng., 25: 116-120.
Steiner G.R. and R.J. Freeman, 1989. Configuration and substrate design considerations for con-
structed wetlands wastewater treatment. In: D.A. Hammer (Ed.), Constructed Wetlands for
Wastewater Treatment. Lewis Publishers, Chelsea, Michigan, pp. 363-377.
Stephenson D. and M.E. Meadows, 1986. Kinematic Hydrology and Modelling. Elsevier, Amsterdam,
247 pp.
Strack, O.L., 1989. Groundwater Mechanics. Prentice Hall, Englewood Cliffs, NJ.
Tchobanoglous G., 1993. Constructed wetlands and aquatic plant systems: research, design, operational
and monitoring issues. In: G.A. Moshiri (Ed.), Constructed Wetlands for Water Quality Improve-
ment. Lewis Publishers, Boca Raton, FL.
Turner, A.K., K.J. Langford, M. Win and T.R. Clift, 1978. Discharge-depth equation for shallow flow.
ASCE J. Irrig. Drain., 104: 95-110.
U.S. EPA, 1988. Design Manual: Constructed Wetlands and Aquatic Plant Systems for Municipal
Wastewater Treatment, EPA/625/1-88/022, Cincinnati, Ohio.
U.S. EPA, 1993. Subsurface Flow Constructed Wetlands for Wastewater Treatment. A Technology
Assessment, EPA/832-R-93-001, Washington, D.C.
Water Pollution Control Federation, 1990. Natural Systems for Wastewater Treatment. Manual of
Practice FD-16, Alexandria, Virginia, 270 pp.
Watson J.T. and J.A. Hobson, 1989. Hydraulic design considerations and control structures for
constructed wetlands for wastewater treatment. In: D.A. Hammer (Ed.), Constructed Wetlands for
Wastewater Treatment. Lewis Publishers, Chelsea, MI, pp. 379-391.
Watson, J.T., S.C. Reed, R.H. Kadlec, R.L. Knight and A.E. Whitehouse, 1989. Performance expecta-
tions and loading rates for constructed wetlands. In: D.A. Hammer (Ed.), Constructed Wetlands for
Wastewater Treatment. Lewis Publishers, Chelsea, MI, pp. 319-351,.
Wieder, R.K., G. Tchobanoglous and R.W. Tuttle, 1989. Preliminary considerations regarding con-
structed wetlands for wastewater treatment. In: D.A. Hammer (Ed.), Constructed Wetlands for
Wastewater Treatment. Lewis Publishers, Chelsea, MI, pp. 297-305.

You might also like