Modelling of detention-sedimentation basins for stormwater treatment using SWMM software

gayoldMécanique

21 févr. 2014 (il y a 3 années et 8 mois)

121 vue(s)

11
th
International Conference on Urban Drainage, Edinbu rgh, Scotland, UK, 2008

Zawilski and Sakson
1


Modelling of detention-sedimentation basins for stormwater
treatment using SWMM software

M.Zawilski
1*
, G.Sakson
1

1
Department of Environmental Engineering, Technical University of Lodz
Al.Politechniki 6, 90-924 Lodz, Poland

*Corresponding author, e-mail: mrkzaw@p.lodz.pl



ABSTRACT
In the paper, application of the model for total su spended solids removal with the use of
SWMM software is analysed. Two types of kinetics we re checked: first order and second
order. In the SWMM model two types of detention-sed imentation tank were input: a single
completely stirred reactor and a reactor of the same mixing conditions but divided into four
subsequent parts. The model was calibrated according to field investigations of a real tank.
The modelling results show a good accordance of the modelled and measured parameters,
especially of the outflow TSS load. The best result s were obtained for the single CSTR and
the second order kinetics. The proposed model is se nsitive to changes of decay rate constant.
Removal of TSS depends also on the tank area and ou tflow intensity. However, these changes
are essential for intense storm events. Despite some simplifications, the SWMM model can be
used for practical calculations.


KEYWORDS
Detention tanks; total suspended solids; sedimentat ion; urban drainage; SWMM software.


INTRODUCTION
Detention-sedimentation basins are frequently used for stormwater treatment prior to its
draining into natural waters. The random character of rainfall events and unsteady flow
conditions make it difficult to evaluate the technological effect of such facilities. It is well-
known that elimination of suspended solids is essential for stormwater treatment and allows
for elimination of other pollutants (like heavy met als and organic contaminants) associated
with suspended solids.

In detention-sedimentation basins, various processes occur: detention of water as a
consequence of throttling of the outflow, flocculat ion and sedimentation of (mainly mineral)
particles, resuspension of sediments, flotation of oils, light particles and so on. In practical
engineering analyses, there is an urgent need to ev aluate the efficiency of such tanks
depending on their size, hydraulic loading, pollutant level of influent stormwater and other
factors. Long-term computer simulation is one of techniques used for this purpose. A
detention tank usually is an element of sewerage ne twork and often is situated at the outlet of
a main stormwater sewer. The tank receives all the catchment runoff and depending of the
tank size (area, active depth) a removal of certain portion of incoming load of suspended
solids is possible. So far, several original simple and sophisticated models for such tanks have
been elaborated. However, the popular SWMM software can be used for this purpose but the
Modelling of detention-sedimentation basins 

2

user should introduce a treatment function. In the paper, a proposal for using such SWMM
option is presented. The proposed algorithm resulte d from own investigation of a real tank.

METHODS
The principle of detention-sedimentation tanks is c apturing storm runoff together with
transported pollutant load. Usually, this can be ac hieved by installation of a throttling device
or a regulator at the tank bottom end. Capturing of inflow or pollutant load from a certain
period (for instance the whole year) may be a basic design criterion (Vaes and Berlamont,
2004; Birch et al., 2006; Calabro and Viviani, 2006). Another option is limiting the annual
volume or load input from a catchment into receivin g water body. Computer modelling is an
effective method for assessing the technological ef ficiency of stormwater detention-
sedimentation tanks situated at outlets from urban catchments.

Models of detention-sedimentation tanks
The functioning of the detention-sedimentation tank can be described either as a hydraulic or
a technological efficiency (i.e. reducing the flow, volume of stormwater or the pollutant load).
Long-term simulations can be used for this purpose. Having a well-calibrated model for
calculation of runoff, stormwater flow and load fro m a catchment, it is necessary to introduce
certain treatment efficiency formula for a possible tank.

Sedimentation of suspended solid particles was reco gnised as the main technological process
for traditional and also modern sediment trap devic es (Phipps et al., 2005). Removing of
turbidity of stormwater is especially difficult (Ha n and Mun, 2007). However, the
sedimentation process occur in non-standard conditi ons, for instance in comparison with
typical primary settling tanks used in wastewater t reatment plants. The main difference is
unsteady inflow, outflow and active volume which ma ke impossible a direct application of
known simple sedimentation models for typical clari fiers. Additionally, inflow and (in
consequence) outflow pollutant concentration are hi ghly variable in time as well as a flow
turbulence undergo rapid changes, especially in rea l tanks of non-ideal shapes (Vaes, 1998).
Models for describing objects are formulated with s ome simplifications. For instance,
Takamatsu et al. (2006) assumed the plug-flow inside the tank which made possible
adaptation of the known Camp theory of particle mov ement in the tank active volume. If a
particle settling velocity distribution is known, t his theory makes possible calculating
technological effect as a portion of removed load o f solid particles in relation to the inflow
load. Other assumptions take non-ideal flow conditi ons, i.e. longitudinal mixing into
consideration (Jensen, 2005). This effect, as known, can be modelled as a flow through a
completely stirred tank reactors (CSTRs) in series and usually few (4-10) such reactors can be
assumed. This model, however, does not allow for de tailed calculating of particle trajectories
during settling process and a kinetic model for pol lutant removal has to be applied instead.
Nonetheless, such models may give satisfactory resu lts (Zawilski, 1997; Person et al.,
1999;Wong et al., 2002).

EPA SWMM model for treatment facilities
EPA SWMM (Stormwater Management Model) is the widel y known and free software used
for modelling of sewerage systems. Among others, it contains a procedure for modelling the
technological effect taking place at any node of a sewer network. The node can be declared as
a storage unit with an assumed treatment function w hich the option can serve for modelling
detention-sedimentation facilities. The user, howe ver, should input his own formula (or a set
of formulas) calculating the removal efficiency usi ng some standard technological parameters.
Some general hints for the algorithm are known only. Basically, a treatment function for the
11
th
International Conference on Urban Drainage, Edinbu rgh, Scotland, UK, 2008

Zawilski and Sakson
3

pollutant concentration (C) in a stream entering th e tank can be used. The following process
variables may be used in the formula:


￿ FLOW for flow rate into node (in user-defined flow units )
￿ DEPTH for water depth above node invert (ft or m)
￿ AREA for node surface area (ft
2
or m
2
)
￿ DT for routing time step (sec)
￿ HRT for hydraulic residence time (hours)

Also, a fractional removal (R) of some other tracer pollutants (usually associated with TSS)
can be formulated. In this case a simple proportional formula type R= a×R(TSS) is proposed.
It is logical that the used treatment formula should be verified according to investigations of a
pilot or real object.

Proposed treatment formula
First of all, it was checked whether a reasonable simple formula for a single CSTR can be
applied. Formally, the process of static sedimentation can be modeled using first order kinetic.
However, as it was stated earlier for the investigated real tank, also other phenomena affect
the sedimentation efficiency:

· characteristics of inflowing TSS strongly depends on rainfall intensity; there is a strong
correlation between TSS concentration and settling velocity distribution, i.e. smaller TSS
concentration always is correlated with smaller settling velocities and vice-versa
(Zawilski, 1996, 1998). However, only a crude relationship can be derived from real data
because of random variation of TSS composition transported from the catchment after
each rainfall event,
· As it was proved earlier, a certain portion of TSS of low settling velocity cannot be
removed during dynamic conditions with variable water level inside the tank (Takamatsu
et al., 2006),
· Turbulence of flow may have greater influence on removal of particles of low settling
velocity.

Therefore the influence of TSS concentration cannot be ignored as it dominates the kinetics
especially for low concentration inflows. Similar relationship was found for settling tanks at
wastewater treatment plants (Lessard and Beck, 1988).
Taking this into consideration, two types of kinetic reactions for inflow TSS concentration
were checked:

· First order kinetic
in
kc
dt
dc
= (1)
· Second order kinetic
2
in
kc
dt
dc
= (2)

These kinetic equations result in following formulas for pollutant removal:

· First order kinetic
HRT)kEXP(1R
×


=
(3)
Modelling of detention-sedimentation basins 

4


· Second order kinetic
HRTkc1
HRTkc
R
in
in
×+
×
= (4)
where hydraulic residence time HRT is assumed as a driving parameter.

A dependence of settling velocity on TSS concentration was stated earlier by the author and
considered for the first-order kinetics (Zawilski 1996). A correction in the formula for TSS
removal efficiency was proposed. In this paper, a similar attitude for formula (3) is applied,
resulting in the following expression:














×=
n
1000
TSS
HRTkEXP1R (5)
In other words, in this corrected formula low TSS concentration causes the same effect as
shortening of HRT. The above formula (5) can be called the improved first order kinetic. In
the SWMM algorithm, HRT is represented by the instantaneous volume divided by outflow
intensity. The outflow intensity can be obtained from the water balance (as it will be
described later).

Additionally, the model for 4 CSTR in series was checked, too. This option was supposed to
give better fitting to measured data as it represents longitudinal mixing of the tank content
better. For this purpose, a series of four sub-tanks were input. The sub-tanks, each of 25%
area of the total area were connected with short virtual channels of big dimensions in order to
avoid artificial throttling effects. Outflow was modeled with the input of outlet characteristics
from the last tank only. Inflow was input to the first tank in series. For each of the sub-tanks
the same kinetic equation was assumed but the decay rate k could be different in comparison
with a single CSTR and should be adjusted separately.

Description of the research object
Lodz city (central Poland) was equipped with a separate sewerage system in the city districts
outside the centre where still the combined system is operated. The stormwater separate
outlets direct runoff into local urban rivers. Therefore, some of the outlets were equipped with
detention-sedimentation tanks in order to remove most of stormwater pollutants prior to their
diverting into receiving waters. One of the biggest tanks was built in 1970s at the outlet of a
sewer serving for a 300 ha urban catchment. The geometrical parameters of the tank are given
in Table 1.

Table 1. Geometrical parameters of the investigated tank
Category Unit
Value
Bottom area m
2
4233
Minimum water depth m 0.16
Water level area increase m
2
m

depth 477
Crest height of emergency overflow
above bottom level
m 1.52
Inflow channel
(long.slope 7%)
m
2.60 1.60
Outflow pipes (29) mm 100
Box filter length m 2G20
Box filter width m 0.5
11
th
International Conference on Urban Drainage, Edinburgh, Scotland, UK, 2008

Zawilski and Sakson
5


The open triangular tank is situated at a local urban river. During first years of functioning,
outflow from the tank was taking place through 29 pipes of 100 mm diameter. Afterwards, a
special filter as an open-work box made from bricks filled with wood-wool was constructed
few meters in front of the end wall (Fig.1). The role of the filter is retaining oils. The filter,
however, functions as a ponding device because of its significant flow resistance. In the
middle of the box filter a concrete wall with small openings at the tank bottom was
constructed (Fig.1). The inlet to the tank was constructed as a short open channel with the
slope of 7%. This allows avoiding of backwater in the inlet rectangular sewer. The tank
collects sediments which are removed from the bottom once a year, usually at the end of the
year (November-December) when no frequent storms over the catchment are observed.




Figure.1. Scheme of the investigated tank

Experimental data
The investigations were performed on the basis of older but still valuable measurements made
for the five years research campaign in 1987-1991 (Table 2). The general results were
already presented (Zawilski 1996).

Inflow measurements were carried out with the use of a limnigraph situated in the side well at
the last segment of the inflow sewer. Simultaneously, water depth in the tank was measured
with the time interval depending on the water level variations. During the research period,
Modelling of detention-sedimentation basins 

6

activating of the emergency overflow was observed only once but not for an investigated run.
Also, in most cases the receiving water level was well below the tank bottom level.
The TSS concentration was determined in grab samples taken: - at the inlet at the lowest point
of the inflow channel (where strong turbulence of flowing stormwater is observed), -at the
outlet at the middle of the filter wall for samples taken from below 20 cm depth; if the active
depth was less than about 40 cm, samples were taken from the outflow opening in the wall.

Table 2. Characteristics of investigated runoff events
Inflow Rainfall

Run
Q
max
(m
3
/s)
TSS
max
(g/m
3
)
C
w
#
(g/m
3
)
V
(m
3
)
L
TSS
(kg)
h
(mm)
t
(min)
i
mean
(mm/h)
1
5.11 1900 1044 10061 10507

7.98 22 21.76
2
3.40 1326 294 5902 1736

7.39 26 17.05
3
0.69 1230 699 2571 1796

10.26 430 1.44
4
3.64 5875 933 5528 5158

6.56 135 2.91
5
1.58 3105 2486 2972 7388

3.78 124 1.83
6
0.13 640 195 5904 1149

- - 1.1*
7
0.23 164 83 9216 761

13.94 1493 0.58
8
0.59 916 639 1856 1186

3.49 62 3.38
9
1.91 2257 549 3488 1916

5.53 40 8.31
10
0.29 296 89 3325 296

7.64 306 1.51
11
0.19 503 183 1041 191

2.31 323 0.43

# TSS concentration weighted by flow for the whole event
* snowmelt event; intensity of equivalent rainfall caused the same average outflow

The effect of the wood-wool filter was not investigated although apart from oils it is likely
able to capture some portion of fine suspended solids.

For the SWMM implementation, the obtained flow, depth and TSS measurements were input
into text files using the necessary format.


RESULTS
Calibration of SWMM model
For all measured runs, separate calculations were carried out in order to adjust model
parameters. At first, for each run outflow characteristics were found. The outflow from the
tank as the flow through the wood-wool filter was not measurable in details. However, its
value can be determined from the water balance:
dt
dV
QQ
inout
= (6)

This equation was solved twice: - with the use of the SWMM procedure as fitting the outflow
parameters (see below) so that the measured and calculated water level in the tank were in
close agreement, and for control  with the use of own calculation procedure numerically
determining values of dV/dt for any time point.
The SWMM formula for outflow intensity is:
B
out
hAQ ×= (7)

11
th
International Conference on Urban Drainage, Edinburgh, Scotland, UK, 2008

Zawilski and Sakson
7

This formula was checked for its applicability and satisfactory results were obtained (Figure
2). However, the outflow parameter A had to be adjusted for each run independently and
turned out to be much greater after the periodical exchange of wood-wool in comparison with
the periods of partly clogged material. During few months of filter functioning, the parameter
A was gradually decreasing from 3.0 to 0.3. The value of parameter B was around 2.8.

For the calibration of TSS removal model, several computation runs were carried out for
different values of kinetic coefficients. The main optimisation fitting criterion was the
correlation between the measured and modelled outflow TSS load. This relationship was
assumed as linear:
meas
out
mod
out
LaL ×= (8)
with ideal fitting for a=1 and R
2
=1.

It was stated that the pure first order kinetics (Equation 3) gives wrong model results.

All calculations were made for the time step of 30 seconds. This parameter proved to be
essential for model results to some extent, decreasing the outflow TSS load with a decreasing
of time step. However, this effect was not studied in detail since input of very short time steps
results in a substantial increase of calculation time. This would be very unfavorable in the
case of simultaneous modeling the runoff from a catchment, flow in sewer networks and
treatment in the tank.

One run had to be rejected from the whole experimental set. For this run, a considerable
scouring of TSS from the tank bottom was detected. Unfortunately, this effect cannot be
directly modeled as the SWMM removal formula allows for positive reduction of pollutant
only. During the rejected run, the maximum inflow was equal to 5 m
3
/s and was reached soon
after a very intense rainfall and with a corresponding rapid increase of the inflow.
The results of calibration are presented in Table 3.

Table 3. Results of calibration of the SWMM model
Type of tank Improved first order kinetics
(formula (5))
Second order kinetics
(formula (4))
Single CSTR k = 0.42; n =2
a = 1.007
R
2
= 0.91
k = 0.0011
a = 0.978
R
2
= 0.97
Four CSTRs k = 0.98; n = 2
a = 1.015
R
2
= 0.92
k = 0.00085
a = 1.002
R
2
= 0.92

In Figure 2 some modelling results are presented. Surprisingly, the best results were obtained
for a single CSTR and assumption of four CSTRs in series did not improve the general fitting.
This result can be explained as following: the SWMM model is able to predict the removal of
the total TSS load from a single event and a series of events. However, simplified calculation
formula for pollutant removal cannot describe outflow pollutographs ideally. Especially, the
TSS concentration is underestimated in its late part, i.e. the concentration tail is not
represented and the TSS level decreases to zero. Most of the TSS load is being removed
during first phases of the tank filling when it works in a manner similar to a single CSTR. The
later rest of TSS load is not decisive for the result.

Modelling of detention-sedimentation basins 

8

Modelling of COD concentration and load
The COD parameter is one of most important ones for evaluating of the influence of
stormwater on receiving waters. This parameter can be modelled in similar way like TSS.
However, it would be necessary to gather a COD database similar to that of TSS and to
calibrate the model. Other simplified method is using the relationship between the two
parameters. The relationship can serve for recalculating of COD from TSS at the outlet of the
tank. For the investigated catchment and the tank, the following relationship was found:

70TSS2.08COD
0.66
+×= (9)

in which the value 70 represents the average soluble fraction of COD (non-degradable in the
tank). The formula is valid both for the inflow, as and the outflow COD, therefore can be
applied in the SWMM removal model. The relationship may be input as the treatment formula
for any node downstream of the tank.



Figure 2. Examples of modelling results for run no.4. Water depth and outflow TSS with
dots as measurement data (above), inflow - outflow discharge and inflow - outflow TSS
(below).








11
th
International Conference on Urban Drainage, Edinburgh, Scotland, UK, 2008

Zawilski and Sakson
9

0
10
20
30
40
50
60
70
80
90
100
1 2 3 4 5 6 7
Multiplication factor for tank bottom area
Percentage of outflow TSS load
Run 1
Run 7
Sensitivity analysis of the model
For the purpose of evaluation of the model, a sensitivity analysis was made. The analysis was
carried out for the single tank, second-order model and the best fitted decay rate k=0.0011
was taken as a base value.
















Figure 3. Sensitivity analysis of the model for the decay rate


Figure 4. Effect of hydraulic loading and outflow discharge onto outflow TSS load

From Figure 3 it can be concluded that the model is very sensitive to changes of decay rate
constant. Therefore, the decay coefficient should be determined in field tests, possibly with
the use of a real or pilot installation.

Removal of TSS depends also on the tank area and outflow intensity (Figure 4). However, the
appropriate changes of TSS load are essential for intense storm events. For weak rainfalls, the
efficiency of the tank is usually poor; therefore the effect of the tank area and outflow
discharge is relatively less.




0.4
0.6
0.8
1
1.2
1.4
1.6
0 0.0005 0.001 0.0015 0.002 0.0025
Decay rate constant
Fitting coefficient
70
80
90
100
110
120
130
0 0.5 1 1.5 2 2.5
Outflow coefficient
Percentage of outflow TSS load
Run 1
Run 7
Modelling of detention-sedimentation basins 

10

CONCLUSIONS
The SWMM model proved to be able to predict TSS removal in detention-sedimentation
tanks properly. However, it is necessary to use a proper formula for TSS removal and to
determine kinetic coefficients. For the investigated real tank, the second order kinetics and
single CSTR reactor turned out to be the best model option. Alternatively, the first order
kinetics with a correction for TSS concentration can be used. In both cases, best results were
obtained for modelling the TSS outflow load, the TSS outflow pollutographs are modelled
with a less precision. For further practical use of the removal model a credible and not
complicated test methodology for determining the decay rate of TSS removal in dynamic flow
conditions should be elaborated.


REFERENCES
Birch G., Matthai C. and Fazeli M. (2006). Efficiency of a retention/detention basin to remove contaminants
from urban stormwater. Urban Water Journal, 3(2), 69-77.
Calabro P.S. and Viviani G. (2006). Simulation of the operation of detention tanks. Water Research, 40, 83-90.
Han M.Y. and Mun J.S. (2007). Particle behaviour consideration to maximize the settling capacity of rainwater
storage tanks. Wat.Sci.Tech., 56(11), 73-79.
Jensen M. (2005). Sediment retention in rectangular storage tanks: Computations with a mass transport model
and comparison with laboratory experiments. Proc. of the 10
th
International Conference on Urban
Drainage (10ICUD), Copenhagen., a CD-ROM collection of papers.
Lessard P. and Beck M. B. (1988). Dynamic Modeling of Primary Sedimentation. JEED, ASCE, Vol. 114. no.4,
pp. 753-769.
Person J., Somes N.L.G. and Wong T.H.F. (1999). Hydraulics efficiency of constructed wetlands and ponds.,
Wat.Sci.Tech., 40(3), 291-300.
Phipps D.A. and Alkhaddar R.M., Faram M.G. (2005). Pollutants retention in stormwater treatment chambers.
Proc. of the 10
th
International Conference on Urban Drainage (10ICUD), Copenhagen, a CD-ROM
collection of papers.
Takamatsu M., Barrett M. and Charbeneau R.J. (2006). Modeling of Sedimentation in Stormwater Detention
Basins. WEFTEC06 (Water Environmental Foundation Conference), 5559-5577.
Vaes G. (1998). Design of storage sedimentation basins. Proc. of the NOVATECH98 conference Innovati ve
technologies in urban drainage, Lyon, a CD-ROM col lection of papers.
Vaes G. and Berlamont J.(2004). Source control design using contiuous long term modeling. 6
th
Int. Conf. on
Urban Drainage Modelling, Dresden, 599-608.
Wong T. H. F., Fletcher T. D., Duncan H.P., Coleman J. R. and Jenkins G. A (2002). A Model for Urban
Stormwater Improvement Conceptualization. 9
th
International Conference on Urban Drainage (9ICUD),
Portland., 598, and a CD-ROM collection of papers.
Zawilski M. (1996). Technological modelling of detention-sedimentation tanks for stormwater treatment. Proc.
7th.Int.Conf. on Urban Storm Drainage. Hannover, t.III, pp.1275-1280
Zawilski M. (1997). Prognosing of amount and pollutant loads of stormwater from urbanised areas. Zeszyty
Naukowe Politechniki ódzkiej, no 792 (in Polish).
Zawilski M. (1998). Distribution of settling velocity - a tool of predicting the stormwater solid contamination
characteristics. Proc. of the NOVATECH98 conference Innovative technologies in urban drainage,
Lyon, a CD-ROM collection of papers, vol.2, 231-238.