You are on page 1of 44

- 2.

6 –
Water Quality Modelling
(optional)
2-6 WATER QUALITY MODELLING
(optional)

Peter Kelderman
UNESCO-IHE Institute for Water Education

Online Module Water Quality Assessment


Overview

• Some Mathematics (background)


• Mass balances
• BOD-DO modelling in a river
• River Danube calamities model

3
Introduction: general
• What is a model?
– a simplified representation of a real system

• Answer “why?” questions:


– study cause-effect relationships
– find out lack of knowledge
– gain insight into the system

• Answer “what if?” questions:


– see effects of "measures"
– see effect of changes in conditions

– Also used as part of monitoring (complicated processes,


calamities,..)
4
Dimension of a model

• Looks at the length co-ordinates x, y z:


• If x, y, z are not included in the model à 0-D model
– Example: pollutant in a completely mixed lake, or:

• One length co-ordinate: 1-D model


– Example: BOD/DO model along river length (x)

• 2-D and 3-D models: two or three length co-ordinates


– E.g. “Delft 3-D model” for pollutant concentration in the North
Sea at different locations + water depths à very complicated !

5
1D model

6
3D model- salinity in an estuary

7
Basic terms

• Input data:
– geometry input data (“model”)
– boundary conditions: values of unknowns at the boundaries of the
spatial domain covered by the model
– initial conditions: values of the unknowns at the start of the
simulation (t=0)
– external variables or “forcing functions”: values of quantities
affecting the model results, which are not predicted by the model.
Their values are taken from outside (e.g. meteorology data)
– model parameters: acceleration of gravity, bottom friction, decay
rate, …

8
Some remarks
• A model can only be as good as the ideas/assumptions behind it:
“Garbage in à garbage out”
• All essential processes should be included; don’t be over-
impressed by many sophisticated models à often lot of “window-
dressing”. (Example: P models for shallow lakes that ignore P sediment-water
exchange à maybe completely wrong results!)

• But, keep the model as simple as possible à simple is beautiful

• In understanding/evaluating models in practice, look for the


theory behind the models;
• And look at the real water system behavior!

• The proof of a good model is the agreement between the model


and the real situation (calibration/validation).

9
Balances and differential equations

• In (Water quality) modelling, balances (mass, energy, momentum) are


commonly used:

• Mass balance:

dMass
“change/sec” = Massin - Massout ± processes
dt
E.g.:In a lake, if 10 kg/s of a pollutant come in, 5 kg/s go out and 3 kg/s go the bottom à change = +2 kg/s

• So modelling works with solving differential equations(exactly or as an


approximation (numerical solutions))!

10
Simple decay calcultation
In words: “decrease/sec is proportional to amount present”
(If radioactivity decreases with 20 g/s for 100 kg, then it will decrease with 10*20=200 g/s for 10*100 kg = 1000 kg)

k = decay rate (1/d)


dM
- = kM M = mass (g)
dt
t = time (d)
(Note that both sides are positive!)

• Used for many processes in water quality modelling, e.g.:


• Decay of BOD
• Radioactive decay
• Dieing of bacteria (e.g. E-coli)
• Mortality of algae
• etc.
11
Simple decay calculation (2)
dM
Decay formulation: - = kM
dt

Separating variables: 1
- dM = kdt
M
M t
1
Take the definite integral on both -ò dM = k ò dt
M0
M
sides: 0

æM ö æ M ö - kt
Will lead to: ln ç ÷ = - kt ® ç ÷=e
è M0 ø è M0 ø

Will finally lead to: M (t ) = M 0 e- kt


12
Decay Rate
1
M (t ) = M 0 e- kt
Slow rate (low k value)
0.8 High rate (high k value)

0.6

0.4

0.2

0
0

Exercise: Check that it wil take 6.9 days for the Mass to reach
50% of its original value M0 , for k = 0.1 day-1

13
Micropolllutants in Rivers
W = 500 kg/day

Q = 10 (m3/s)

Input chlorobenzene concentration = 580 μg/L (check yourself)

(rapid) Degradation by volatilization (loss to atmosphere) ; kv = 8.5 day-1


700

600
Similar to BOD degradation: concentration (ug/l)

500

400
Cpoll. = C0 exp (-kv t) 300

200
(t = travel time in river) 100

-
(13 hours to reduce to 1% of C0) 0 0.2 0.4 0.6 0.8 1
14
Tim e (days)
The essence of water quality modelling!
volume
V

discharge Qout
discharge Qin
concentration Cout
concentration Cin

reaction rate k

How are the concentrations changing, in space


and in time?
15
BOD – DO in a river

16
Oxygen Concentration pattern
BOD
load

DO
Diss.
oxygen

River axis
• Major WQ problem, already investigated in 19th century
• Major mathematical description: Streeter-Phelps model
(1925)

17
Major Processes
• Sources of DO:
• Re-aereation of oxygen from the atmosphere
• Oxygen production/consumption by plants, algae

• Sinks of DO:
• Oxidation of BOD
• Sediment oxygen demand
• Oxygen consumption by respiration of algae and plants

18
Oxidation of BOD (1)
• Assume CBOD and NBOD (NH4+ à NO3-) have same
degradation rate.
• According to first order reaction:

L = BOD (g/m3)
dL
= -k1L t = time (day)
dt k1= BOD degradation rate constant (day-1)
BOD

- k1t
Solution: L = L0 e
Time (days) 19
Oxidation of BOD (2)
• Degradation rate constant k1 for BOD :
• k1 generally between 0.1 – 0.4 (day-1)
• Dependent on e.g. type of organic material, river
characteristics, wastewater purification, time after discharge
• Increase of about 4% per oC

k = 0.2
k = 0.4
BOD

Tim e (days)

20
Re-aeration of Oxygen
• Driving force: Difference between O2 concentration and
saturation O2 concentration, in the water:
k2= re-aeration constant (day-1)
dc
= k2 ( csat - c ) c = Oxygen concentration(g/m3)
dt
csat= Saturation oxygen
concentration(g/m3)
air

water

21
Re-aeration of Oxygen (2)

• Solution: c = c – {(csat – c0 ) exp(-k2t)} (If you like: try yourself)


DO concentration

k = 0.6
k=1
Sat

tim e (days)

22
Re-aeration in rivers
• Value k2 dependent on:
• Velocity river : k2 proportional to about V0.5
• Water depth : k2 proportional to about 1/H1.5
• Temperature: about 2% increase per 0C

Empirical graph from different UK-


USA reseaches
Example:
• Velocity = 0.6 ft/sec = 0.18 m/sec
• Depth = 4.5 ft = 1.4 m

• àk2 = 1.0 day-1

(see Thomann and Müller (1987))


23
1 foot= 0.3 m

24
Sediment Oxygen demand
SOD (g/m2/day)
Bottom type and location
Range Average
Sphaerotilus (10 g dry wt/m2) 7

Municipal sewage sludge – outfall vicinity 2-10 4

Municipal sewage sludge –“aged”, downstream outfall 1-2 1.5

Estuarine mud 1-2 1.5

Sandy bottom 0.2-1.0 0.5

Mineral soils 0.05-0.1 0.07

(see Thomann and Müller (1987))

• Especially important for organic-rich sediments


• Increase about 7% per oC
• SOD can be neglected for deep rivers
25
Streeter-Phelps DO model for a river
BOD

U (m/s)
V
dx x

– Ignore SOD
– Ignore effect algae and water plants
– Ignore differences between CBOD and NBOD degradation

• Mass Balance for oxygen over segment dx:

dc
V = rear - oxidation + Inflow - Outflow
dt 26
Model equation
dc k1 x dc
= k 2 (c s - c) - k1 L0 exp(- ) -U
dt U dx
Re- BOD decay inflow-
aeration outflow
• (L0 is the BOD concentration in the river after mixing)

• BOD decay term expresses:


• Decrease in BOD that has already taken place between the
discharge point and the segment V (L = L0 exp(-k1 x/U)); and
the BOD decay in segment V itself: dL/dt = -k1 L

• Last term: “plug-flow” equation


27
Solution
• Assume steady state conditions: dc/dt =0 à solution:

k1 L0
c = cs - [exp(-k1t ) - exp(-k2t )] - (cs - c0 ) exp(-k2t )
k 2 - k1

t=travel time river,


equivalent with “distance”

Dc = “Critical” DO deficit”

(see Thomann and Müller (1987))


28
“Critical distance”
• “Critical distance" x*c or travel time t*c = x*c/U, i.e. the distance/travel
time where the minimum O2 concentration in the river is reached.
By differentiation; dc/dx = 0:

1 é k 2 æ (cs - c0 )(k 2 - k1 ) ö ù
tc=
*
ln ê çç1 - ÷÷ ú
k 2 - k1 ë k1 è k1 L0 øû
(k2 ¹ k1 ; for equal values assume a small difference, say 0.01, in the
calculations).

1 é k2 ù
For cs = co : t *
c = ln ê ú
k 2 - k1 ë k1 û

Example: for k1 = 0.3 day-1 and k2=0.8 day-1, tc* = 1/0.5 (ln 8/3) = 2.0 days

29
Minimal Oxygen concentration

k1
C s - C min . = L0 e - k1tc
k2

• à “oxygen deficit” roughly proportional to BOD loading L0 !


(see Thomann and Müller (1987)
30
Calculation Example
k1
Characteristics river/BOD load: C s - C min . = L0 e -k1tc
k2
• k1 = 0.3 day-1
• k2 = 0.9 day-1
• L0 = 30 mg/L (BOD conc. in river after mixing)
• cs = 8.0 mg/L (DO saturation concentration)
• tc = 2.5 days

Minimum DO will be: 3.3 mg O2/L

Till what value should L0 be reduced to reach DO standard = 5.0 mg/L?


(Answer: L0 < 19 mg/L)
31
BOD-DO models in practice

• Often much more complicated:


• A wide variety of point and non-point sources...
• Different tributaries and river branches, all with different k1
and k2 values ..
• Take into account SOD, effect algae, differences between
CBOD and NnBOD.......

• à BOD-DO models such as “SOBEK”, or “Mike11 Ecolab”

32
Numerical modelling
• Streeter-Phelps model could mathematically be solved exactly

• For more complicated equations, it will be impossible to find


exact solutions of the differential equations à “numerical
solutions”

• Approximation dC/dt à ΔC/Δt :”change of C during (small) time


step Δt’’

• With this kind of numerical modelling, the trend in C can then


be followed (time)step by step. See e.g. Euler method

33
Example: Black River USA
• Main river with two tributaries

• River system was divided into 59


segments, all with their specific values
for „input parameters" K-CBOD, K-
NBOD, K2, K-SOD

• Euler method was used to model DO


over the whole river

34
Calibration and Verification

– Calibration of a model: to fit the Calibration


model with the real situation/data,
by “trying out” (realistic !) values Not
of input parameters k1, k2, kSOD , ok
etc. Verification

Ok
– Verification/validation: to check
that the model then also works for Use ok
of model
another, independent data set

35
Calibration and Verification (2)

• Black River BOD-DO model:


• Calibration with the help of data set
August 14, 1973 (:“very good!”)

• Verification with data set November 1,


1973 (:“unusually good result!”)

• Finally the model can then be applied;


in this case for finding out effect of dam
construction

36
Modelling accidental spills

37
Accidental spills
• Short event of intensive pollution, due to an accident; examples:
– Sandoz, Rhine, 1986: some 20 tonnes of pesticides were
spilt into the river Rhine, due to extinguishing water for fire at
Sandoz factory à massive fish kills

– Baia Mare, Danube, 2000 (see later)

• accidental spills are important, because they attract a lot of


publicity and concern in the society
• often the starting point of action plans to improve pollution
situation

• Modelling: the peaks of discharge will gradually get wider (non-


ideal river plug-flow due to back-currents, shallow/deep, etc.)
38
Sandoz, measured concentration downstream

ref: Dietrich (2008), lecture notes Sandoz Accident, University of Konstanz, Germany
http://www.umwelttoxikologie.uni-konstanz.de/Lehre/Lecture__Environ__Tox__I_/02_29042008OkotoxIBS4Sem.pdf 39
Accidental spill model (basic version)

• 1-dimensional
• advection diffusion equation
¶c
transport ( g / s) = Qc - DA
¶x
• Analytical (exact) solution, for a mass M of pollutants released
at t=0 at x=0 (Taylor formula):

C (x,t) =
M æ -
e x p çç
( x - U t ) ö÷
2

A 4pD t 4Dt ÷
è ø

• Q, U = river discharge/flow; D = “mixing; diffusion” parameter


40
Case Study: Danube River
The Danube River is 2857 km long and the basin covers 817,000
square km in 18 countries in the heart of central Europe. The basin is
characterised by large socio-economical differences. It stretches out
from rich Western-European states to some relatively poor former
Soviet Union Republics.

The river has a number of very large tributaries. The Danube water is
used extensively by the 85 million inhabitants of the basin. The basin
includes many important natural areas, including the Danube delta - the
second largest wetland area in Europe.

41
Accidental Spill

42
Cyanide Spill in Danube river

• Source: mining waste in Rumania


• 100,000 m3 wastewater release due to breaking dike
in mine
• Much damage to Danubecosystem
• Danube Basin Alarm Model (DBAM) was used to
predict (maximum) levels of cyanide in the river,
travel time, restoration..
• No cyanide decay assumed over time scale used à
“worst case scenario” (is often done in modelling).

43
Results from “hindcasting” application : “quite ok, but
14
some extremes missed and time differences”

12
Balsa computed Balsa observed
Kiskore computed Kiskore observed
10
Tiszasziget computed Tiszasziget observed
Concentration (mg/l)

0
31-jan 2-feb 4-feb 6-feb 8-feb 10-feb 12-feb 14-feb
44

You might also like