Signature Redacted

Water Quality and Hydrological Assessment and Modeling of
Bioretention Basins in Tropical Cities

by
OF TECHNOLOGY
Jia Wang
MAR 3 02017
B.S. Biomedical Engineering, second major in Economics
Johns Hopkins University (2009) LIBRARIES
S.M. Civil and Environmental Engineering
Massachusetts Institute of Technology (2011) ARCHIVES
Submitted to the Department of Civil and Environmental Engineering in
partial fulfillment of the requirements for the degree of
Doctor of Philosophy
in Civil and Environmental Engineering
at the
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
February 2017
2017 Massachusetts Institute of Technology. All rights reserved.
Signature redacted
Au th or ................................................................................................................................................
Department of Civil and Environmental Engineering
January 2017
Certified by ....................................
Signature redacted
Peter Shanahan
Senior Lecturer of Civil and Environmental Engineering, MIT
Thesis Co-supervisor
Signature redacted
C ertifi ed b y ........................................................................................................................................
Lloyd H.C Chua
Assistant Professor of Civil and Environmental Engineering, Nanyang Technological University
Associate Professor of Civil and Environmental Engineering, Deakin University (current)
Thesis Co-supervisor
Accepted by ............................................ ............. Signature redacted

f Jesse Kroll
Professor of Civil and Environmental Engineering, MIT
Chair, Graduate Program Committee
2
Water Quality and Hydrological Assessment and Modeling of
Bioretention Basins in Tropical Cities
by
Jia Wang
Submitted to the Department of Civil and Environmental Engineering

on January 20, 2017, in partial fulfillment of the
requirements for the degree of
Doctor of Philosophy in Civil and Environmental Engineering
ABSTRACT
The bioretention basin is a form of stormwater best management practice that originated in
temperate regions and is becoming increasingly popular in tropical regions. Nevertheless, it is a
poorly evaluated system in the tropics due to the limited availability of long-term, high-
resolution hydrological and multi-component water quality data on completed basins. The
resulting gaps in data and knowledge limit the ability of environmental managers to control non-
point source pollution by stormwater. In this thesis, we provide a survey of basin performance in
treating 15 water quality parameters in the Balam Estate Rain Garden in Singapore. We then use
a combination of field observations and model simulations to recommend practical engineering
strategies to design and manage bioretention basins in the tropics.
First, field observations on water quality performance in six sampled events show that
removal targets for total nitrogen, total phosphorus, and total suspended solids are generally met
only for low-depth rainfall events and not high-depth events. We attribute this to the low influent
event mean concentration and weak first flush resulting from the frequent and intense rainfall of
the tropical climate.
Second, field observations on hydrological performance in 80 events show that a lack of
storage capacity and resulting high culvert overflow is the main driver in reducing removal
efficiency for large, but still common, storm events. We recommend that design guidelines in the
tropics be specified in terms of the more definitive quantity, water quality depth (WQD), instead
of average return interval (ARI).
Third, we investigate design configurations for improved hydrological performance
efficiency by first demonstrating the applicability of the RECHARGE numerical hydrological
model for continuous simulation over a half-year period at the field scale. Sensitivity analysis
using the model suggests design configurations that improve basin hydrological performance.
Lastly, we develop a process-based numerical water quality model using the
mathematical formulations from CW2D to describe the bacterial dynamics in contributing to
nitrogen removal in the subsurface soil media. Sensitivity analysis using the model produces
bioretention basin design curves that are suitable for the tropics.
3
Thesis Co-supervisors:
Dr. Peter Shanahan

Title: Senior Lecturer of Civil and Environmental Engineering, MIT
Prof. Lloyd Chua

Title: Assistant Professor of Civil and Environmental Engineering, Nanyang Technological
University of Singapore; Associate Professor of Civil and Environmental Engineering, Deakin
University, Australia (current)
4
Acknowledgements
This thesis would not have been possible without the guidance and help of many individuals who
in one way or another contributed and extended their valuable assistance in the preparation and
completion of this study.
First and foremost, I would like to offer my sincerest gratitude to my advisors, Dr. Peter
Shanahan and Prof Lloyd Chua, who have supported me throughout the thesis with their patience
and knowledge. I appreciate all their contributions of time, ideas, and funding to make my PhD
experience productive and stimulating. Their enthusiasm and meticulousness for details are
contagious and motivational for me. I thank them for the countless hours they have put into
discussions, manuscript editing, and even personal consultation. They are always behind me to
give me support whenever I run into problems. They are my pillar of strength throughout this
journey. I am forever grateful for their guidance.
Next, I would like to thank my thesis committee: Dr. Eric Adams, Prof. Janelle Thompson, and
Prof. Elfatih Eltahir for their willingness to serve in my committee, for their guidance and useful
feedback and ideas, and for their availabilities in last-minute meetings and many cross-three-
time-zones committee meetings. I am extremely grateful for their kind service. I thank them for
giving me opportunities and capacities to carry out independent work and to grow.
I am also thankful for the assistance rendered by numerous colleagues from MIT (Parsons Lab
and Pierce Lab), SMART-CENSAM (Singapore MIT Alliance for Research and Technology
-
Center for Environmental Sensing and Modeling), NTU (Nanyang Technological University of
Singapore), and NUS (National University of Singapore).
I would like to especially thank Alwi Alkaff and Maria Stefanie Suryadharma for their assistance
in field work; MIT Master of Engineering students Halle Ritter and Riana Kernan for sharing
their field results at the same rain garden; Veronique Blanc for her assistance in budgeting and
making funds available for research; Ong Geok Suat and Yau Wing Ken from PUB (Public
Utilities Board of Singapore) as well as Veera Sekaran of Greenology Pte Ltd for their insightful
5
discussions on the design and construction of the Balam Estate Rain Garden; and Adeline Wong
from Marine Parade Town Council for her support of the field monitoring program. Special
thanks also extend to Alejandro Dussaillant-Jones for sharing the RECHARGE model with me
for it offers a solid foundation for me to build upon. Special mention goes to PUB for their
generous support in undergraduate study (PUB Overseas Undergraduate Scholarship) as well as
for their utmost kindness in allowing me to concentrate on my pursuit of graduate degrees.
I am thankful of the following colleagues whom I'm very fortunate to meet: Prof Janelle
Thompson's group (Kyle, Adam, Samodha, Jia Yi, Ju Hyoung, Jean-Pierre) for the enjoyable lab
time during my Masters years. Dr. Eric Adam's group (Cindy, Adrian, Zhao Bing, Ruo-qian,
Wang Yuan, Jenn Wei, Mariana) for always making me feel at home. Prof. Elfatih Eltahir's
group (Noriko, Mariam, Mekki, Eun-Soon, Suchul) and Prof Karina Gin's group (Jingjie, Lay
Leng) for their insightful discussions. Last but not least, Prof. Lloyd Chua's group (Alwi, Song
Ha, Eveline) for their utmost support in field work and all aspects of related research.
Special thanks also go to the Juanes Group for the memorable year spent together. I would like to
thank Prof. Ruben Juanes for giving me the chance to be immersed in the welcoming and
stimulating research environment of his group (Zhibing, Ruby, Mathias, Jane, Amir, Robin, Ben,
Peter, Marie-Julie, David, Luis, and Na-Hyun).
My stay at MIT and Singapore would also not be as enjoyable without the care and concern from
relatives, friends and colleagues: Aunt Bihua and family, Serene, Li Na, Xiaoxuan, Laura,
Momoe, Lanbing, Anni, Hsiang-He, Cheng Dan, Benjamin, Kuanbo, Liu Xuan, Mengli, Gui Lin,
Yuepeng, Meghali, Shakil, Kian Wee, Chen Zhi, Grace, Jane, Joo Siong, Xianxiang, Bing, Kang
Hao, Tan Rui, Wendy, Yoong Keok, Xiaosai, Hongye, Ninghan, Shengyong, Chyan Ying, Teck
Chuan, Mingjuan, Yinjin, and Richie. Thank you friends for being with me through the many
ups and downs.
Last but not least, I owe my deepest gratitude to my parents, Zhaokang Wang and Yuzhen Chen,
my brother Xingsheng Wang, and my grandparents for their love, encouragement and
unconditional support in all my pursuits. Without you, I would never have gone this far.
6
Contents
Chapter 1 Introduction............................................................................................................. 15
1.1 Research m otivations ......................................................................................... 15
1.2 Background ............................................................................................................ 16
1.2.1 Best Managem ent Practice (BM P) ....................................................................... 16
1.2.2 Our study site - Balam Estate Rain Garden in Singapore .................. 17
1.2.3 D esign guidelines of bioretention basins .............................................................. 18
1.2.4 Hydrological and water quality models: the need for a process-based model........ 19
1.3 Research objectives ............................................................................................ 20
1.4 Thesis overview ...................................................................................................... 21
Chapter 2 Field evaluation of w ater quality perform ance ................................................... 23
2.1 Abstract .................................................................................................................. 23
2.2 Introduction ............................................................................................................ 24
2.3 M ethodology .......................................................................................................... 28
2.3.1 D ata collection ........................................................................................................ 28
2.3.2 D ata analysis ........................................................................................................... 31
2.4 Results .................................................................................................................... 33
2.4.1 Hydrologic perform ance ..................................................................................... 33
2.4.2 Pollutant rem oval efficiency ............................................................................... 34
2.4.3 W ater quality of influent..................................................................................... 37
2.4.4 Rainfall characteristics....................................................................................... 39
2.5 D iscussion .............................................................................................................. 40
2.5.1 Frequent heavy flushing in tropical catchment appears to lead to low pollutant
concentration .......................................................................................................... 40
2.5.2 Lack of storage capacity affects rem oval efficiency........................................... 41
2.5.3 Design guidelines for sizing bioretention basins in the tropics need to be revised 44
2.6 Conclusions ............................................................................................................ 47
2.6.1 Study impact and future work............................................................................. 49
2.7 A cknow ledgem ents ............................................................................................. 49
2.8 Supplem entary materials ..................................................................................... 62
Chapter 3 Field evaluation of hydrological perform ance..................................................... 81
3.1 A bstract .................................................................................................................. 81
3.2 Introduction ............................................................................................................ 82
7
3.3 M ethodology ....................................................................................................... 83
3.3.1 Site description...................................................................................................... 83
3.3.2 Data collection ..................................................................................................... 84
3.3.3 Data analysis ....................................................................................................... 85
3.4 Results and discussion......................................................................................... 86
3.4.1 W ater balance and w ater quality volum e (W QV)............................................... 86
3.4.2 Surface and subsurface storage........................................................................... 89
3.5 Conclusions ......................................................................................................... 93
3.6 Acknow ledgem ents ............................................................................................. 94
Chapter 4 Hydrological m odel ............................................................................................. 99
4.1 Abstract .................................................................................................................. 99
4.2 Introduction .......................................................................................................... 100
4.3 M ethodology ........................................................................................................ 103
4.3.1 Flow and w ater level m easurem ents ..................................................................... 103
4.3.2 Bioretention geom etry .......................................................................................... 103
4.3.3 Hydrological model - RECHARGE ..................................................................... 104
4.3.4 Data analysis ......................................................................................................... 108
4.4 Results and discussion.......................................................................................... 109
4.4.1 Model calibration and validation .......................................................................... 109
4.4.2 Sensitivity to m odel param eters............................................................................ 112
4.4.3 Simulation results in response to calibrated param eter set ................................... 116
4.4.4 Model application: infiltration percentage as a proxy for sufficiency of basin
storage .................................................................................................................. 119
4.5 Conclusion............................................................................................................ 122
4.6 A cknow ledgem ents .............................................................................................. 123
Chapter 5 W ater quality model.............................................................................................. 138
5.1 Abstract ................................................................................................................ 138
5.2 Introduction .......................................................................................................... 139
5.3 M ethodology ........................................................................................................ 143
5.3.1 Field data for calibration and validation............................................................... 143
5.3.2 CW 2D ................................................................................................................... 143
5.3.3 Coupling of RECHARGE and CW 2D .................................................................. 148
5.4 Results and discussion.......................................................................................... 149
5.4.1 Model calibration and validation .......................................................................... 149
5.4.2 Simulation results.................................................................................................. 154
8
5.4.3 Model response to param eter choice - a sensitivity analysis ............................... 159
5.4.4 Model application: predicting TN removal at various basin configurations as part
of design curve developm ent................................................................................ 163
5.5 Conclusion............................................................................................................ 168
5.6 Acknow ledgem ents .............................................................................................. 169
Chapter 6 Im plications for future bioretention basin design ................................................. 215
6.1 Current design guidelines..................................................................................... 215
6.1.1 ARI.............................................................................................................. ......... 2 1 5
6.1.2 W Q V / WQ D ........................................................................................................ 216
6.2 Perform ance under current Singapore design guidelines ..................................... 217
6.2.1 ARI........................................................................................................................ 217
6.2.2 W QV / W QD ........................................................................................................ 218
6.3 Recom m ended design guidelines ......................................................................... 218
6.4 Conclusion............................................................................................................ 221
Chapter 7 Conclusions and future work ................................................................................ 224
7.1 Conclusions .......................................................................................................... 224
7.2 Future w ork .......................................................................................................... 230
References ............................................................................................................................... 23 2
9
List of Figures
Figure 2-1. Hyetographs and hydrographs for the inlet, outlet, and culvert of six sampled
storm events at per-minute intervals. Time 0 indicates the start of rainfall.
Sampling times for water quality analysis are also shown on the hydrographs.
Sampling schedule and locations are shown in Table 2-S-1 and Figure 2-S-1
in the supplementary materials respectively. Events are arranged in order of
ascending rainfall depth.......................................................................................... 54
Figure 2-2. Mass load (kg) at the inlet, culvert, and outlet; removal rate (%) with
uncertainty calculated using Eq. (S-2); culvert flow volume as percentage of
inlet flow volume; and EMC of each of the six sampled events for TN, TP,
an d TS S ...................................................................................................................... 55
Figure 2-3. Culvert discharge volume as percentage of total inflow volume, highlighting
the water quality performance of the six sampled events (top) and
categorization of events in the monitoring period into < 3-month and < 1-year
ARI events (bottom) (adapted from Wang et al.(2016)). ...................................... 56
Figure 3-1. a) Histogram showing water balance for 80 events. b) Plots of water balance
against rainfall depth (mm) and ADP (days) for 80 events.................................... 95
Figure 3-2. a) Culvert discharge volume and b) outlet discharge volume as percentage of
total inflow volume, highlighting the water quality performance of the
sampled events; c) categorization of events in the monitoring period into >
and < 3-m onth A RI categories.............................................................................. 96
Figure 3-3. Time series of cumulative storage, surface storage, and subsurface storage for
a) a small event (Event #63, 2.6 mm), b) a medium event (Event #61, 8.2
mm), and c) a large event (Event #47, 29.4 mm). Time series are separated
into three stages (A , B, and C).............................................................................. 97
Figure 3-4. Histograms of ponding duration (of 80 events), Stage B duration (of 59
events), increase in subsurface storage in Stage B (of 59 events), and
drainage rate in Stage C (of 59 events)................................................................. 98
Figure 4-1. Cross section of the Balam Estate bioretention basin. ............................................. 126
Figure 4-2. Relationships between pressure head (h) and relative hydraulic conductivity
(Kr), pressure head (h) and soil moisture (0), and pressure head (h) and soil
moisture capacity (96/ ah) as described in van Genuchten-Mualem equations
(Equation 4-2 to Equation 4-4), using parameters in Table 4-2. ............................. 127
Figure 4-3. Normalized root-mean-square-error (nRMSE), coefficient of determination
(R2), and Nash-Sutcliffe Efficiency (NSE) values (pond water level, WLp
(cm); outlet discharge rate, qo (cm/hr); and culvert discharge rate, qc (cm/hr))
for model calibration of weir coefficient (C,) and saturated hydraulic
conductivity (Ks) using data from August, September, and November. ................. 128
10
Figure 4-4. Modeled profiles for pond water level, WLp (cm); outlet discharge, qo (cm/hr);
and culvert discharge, qc (cm/hr), at various weir coefficients, Cw = 0.5, Cw
0.2, and Cw = 0.05 compared to observed values. Time t = 0 hr is the
beginning of the field monitoring period with discharge records in June 2013...... 129
Figure 4-5. Modeled profiles for pond water level, WLp (cm); outlet discharge, qo (cm/hr);
and culvert discharge, qc (cm/hr), at various saturated hydraulic conductivity
of soil filter media layer, Ks = 0.5 cm/hr, Ks =1 cm/hr, and Ks =1.3 cm/hr
com pared to observed values................................................................................... 130
Figure 4-6. Modeled profiles for pond water level, W<Lp (cm); outlet discharge, qo (cm/hr);
and culvert discharge, qc (cm/hr), at Ks = 0.5 cm/hr and various a values, a =
0.005, a = 0.075, and a = 3 compared to observed values. ..................................... 131
Figure 4-7. Ponding water level, WLp (cm), for the six months................................................. 132
Figure 4-8. Culvert discharge, qc (cm/hr), hydrograph for the six months................................. 133
Figure 4-9. Outlet discharge, qo (cm/hr), hydrograph for the six months................................... 134
3 3
Figure 4-10. Spatial-temporal profile of soil moisture, 0 (cm /cm ) for the six months. ........... 135
Figure 4-11. Percentage of infiltration, I (%) and precipitation, P (mm) of 80 modeled
e v en ts ...................................................................................................................... 136
Figure 4-12. Predicted basin performance at various detention depth, hd (cm); ratio of
drainage area to bioretention area, R; and saturated hydraulic conductivity,
Ks (cm/hr), as assessed using ID (infiltration index).............................................. 137
Figure 5-1 A conceptual model for nitrogen removal: summarizing the bacteria species
and chemicals involved in the nine reaction processes (identified as RC 1
through RC 9) that govern nitrogen species transformation from organic
nitrogen to ammonia, to nitrite, to nitrate, and to molecular nitrogen...................... 170
Figure 5-2 a) General scheme and b) numerical implementation for the coupling of
RECHARGE hydrological model and CW2D reactive biokinetic model................ 171
Figure 5-3. Results from model calibration and validation, showing observed against
modeled values of event mean outlet concentration and removal rate.................... 181
Figure 5-4. Time-depth profiles of various water quality parameters: 02, NH3-N, N02-N,
N03-N, N2, IP, CR, CS, CI (mg/L), bacteria, XH, XANs, XANb (mg/kg), soil
moisture, 0 (cm 3/cm 3), and flow rate, q (cm/hr) for Event #1. ................................ 184
11
Figure 5-10 . Ponding water level, WLp (cm), for Events #1 - #6............................................... 195
Figure 5-11 . Culvert discharge, qc (cm/hr), hydrograph for Events #1 - #6............................... 196
Figure 5-12 . Outlet discharge, qo (cm/hr), hydrograph for Events #1 - #6. (The observed
peaks in outlet discharge due to leakage in the outlet box are not modeled.)....... 197
Figure 5-13 . Mean outlet concentration and removal rate of various pollutants under
cases of high, medium, and low bacterial growth and decay rates in Event
# 1 ........................................................................................................................... 203
Figure 5-14 . Time-series and depth profiles of various pollutants when the model is
subjected to high, medium, and low bacterial growth and decay rates................. 206
Figure 5-15 Mean effluent concentration and removal rate of various pollutants under
cases of high, medium, and low bacterial count in Event #1. ............................... 210
Figure 5-16 Time-series and depth profiles of various pollutants when the model is
subjected to high, medium, and low bacterial count............................................. 213
Figure 5-17 Predicted TN (total nitrogen) removal and basin performance as assessed
using PEI (performance efficiency index) at various detention depth, hd
(cm); ratio of drainage area to bioretention area, R; and saturated hydraulic
conductivity, Ks (cm /hr)........................................................................................ 2 14
Figure 2-S- 1. Plan view of Balam Estate Rain Garden with on-site instrumentation. ........... 64
Figure 2-S- 2. Pollutographs of TN, TP, and TSS for each of the six sampled events............. 72
Figure 2-S- 3. Plots of cumulative mass against cumulative runoff volume indicating the
magnitude of first flush for TSS, TN, and TP................................................... 76
Figure 2-S- 4. Histograms and cumulative probability curves of event-based rainfall
amount (mm), event duration (hr), and ADP (days) for the 96 events
observed during Apr-N ov 2013. ...................................................................... 77
Figure 2-S- 5. Rainfall intensity of events of various average return intervals (ARI) in
Singapore, Melbourne, and selected cities in the U......................................... 80
12
List of Tables
Table 2-1. Event statistics and hydrologic performance metrics of six sampled events ....... 57
Table 2-2. Concentration (mg/L) of all samples collected during all the six events at the
inlet, culvert, and outlet; event-specific removal rates (%); and influent EMCs
(mg/L) for each of 15 water quality parameters .................................................... 58
Table 2-3. Presence of first flush (FF) in the six events based on three criteria of different
level of stringency .................................................. . . . . .. . .. .. . . . . . . . . . . . . . . .. . . . . . . . . . .. . . . . . . . . . . 60
Table 2-4. Comparison of pollutant removal rates in this study to literature values for
b ioretention b asins ....................................................................................................... 61
Table 4-1. Mass balance of 80 events in the six months. .......................................................... 124
Table 4-2. Hydraulic parameters of bioretention basin soil media............................................. 126
Table 5-1. Summary of different bacteria species in process-based model................................ 170
T able 5-2. M odel param eters ...................................................................................................... 172
Table 5-3. The overall reaction rate term, dCi/dt [mgi/(Lxh)], following the Gujer matrix
structure for each of the 12 components i (02, CR, CS, CI, XH, XANs, XANb,
N H3, N 0 2, N 0 3, N 2, and IP)........................................................................................ 174
Table 5-4. Initial concentrations within the soil m edia............................................................... 178
Table 5-5. Influent concentrations .............................................................................................. 180
Table 5-6. Model evaluation statistics for the results from calibration and validation:
coefficient of determination (R2), Nash-Sutcliffe efficiency (NSE), mean
absolute error (MAE, mg/L), relative root-mean-square-error (RRMSE),
RMSE-observations standard deviation ratio (RSR), and percent bias (PBIAS,
%) for the average outlet concentration.................................................................... 182
Table 5-7. Model results showing the modeled mass (g) at culvert, and outlet and the
resultant removal rate, R% compared to the observed values. ................................. 198
Table 5-8. Literature values for pXH, gXANs, [IXANb (p is the maximum growth rate)................. 199
Table 5-9. List of parameter values set for the different levels of bacterial growth rate and
decay rate (day-')....................................................................................................... 199
Table 5-10. Effects of increasing or decreasing growth rate (IH, IDN, [tAns, PANb) and cell
lysis rate (bH, bAns, bANb) on various pollutants in the model as compared to
the case of m edium rates (Table 5-9). ..................................................................... 200
Table 5-11. Statistical evaluation of effects of bacterial dynamics on mean outlet
concentration: coefficient of determination (R2 ), Nash-Sutcliffe efficiency
(NSE), mean absolute error (MAE, mg/L), relative root mean square error
(RRMSE), RMSE-observations standard deviation ratio (RSR), and percent
bias (PBIAS, %) for the event mean outlet concentration....................................... 204
13
Table 5-12. Literature values of bacterial count for denitrifiers and nitrifiers ........................... 207
Table 5-13. List of parameter values set for the different levels of initial bacterial count in
the different soil layers ............................................................................................ 207
Table 5-14. Statistical evaluation of effects of bacterial count on mean outlet
concentration: coefficient of determination (R2), Nash-Sutcliffe efficiency
(NSE), mean absolute error (MAE, mg/L), relative root mean square error
(RRMSE), RMSE-observations standard deviation ratio (RSR), and percent
bias (PBIAS, %) for the event mean outlet concentration....................................... 211
3
Table 6-1. Predicted infiltration percentage I (%), water quality volume, WQV (m ), and
water quality depth, WQD (mm) for various basin configurations (varying hd
and R) that similarly constitute to approximately TN removal rate (TNR) Of
4 5 % ........................................................................................................................... 223
Table 6-2. Comparison of TN removal rate (TNR %) between predictions made by six-
month water quality simulation in this study and by MUSIC curves (Public
Utilities B oard of Singapore 2014)........................................................................... 223
14
Chapter 1
Introduction
1.1 Research motivations
Due to large-scale urbanization and development, much of the natural landscape has been altered.
One prominent modification is the construction of buildings and roads which increases the
proportion of surface imperviousness in an urban catchment and hence results in hydrologic
modification. For instance, as more precipitation fails to infiltrate locally and recharge
groundwater in a pre-development manner, more surface overflow or stormwater runoff is
created. In order to prevent flooding, traditional stormwater controls used in urban areas are
designed to collect, convey, and discharge water quickly and efficiently. Unfortunately, such
concrete-based drainage systems further worsen the problem of hydrologic modification since
peak discharge is raised significantly and time to peak discharge is shortened considerably while
local groundwater recharge is also impaired.
Not only does this pose serious hydrologic problems, urban overland runoff is a major cause of
non-point source pollution as stormwater washes off car parks, roads, and lawns. Water quality
of receiving water bodies is greatly compromised since large amounts of runoff and fast
discharge carries with it high loads of natural and manmade pollutants like sediments, nutrients,
heavy metals, fertilizers, pesticides, toxic chemicals, and pathogens. There is an urgent need to
15
find ways to manage urban runoff effectively with respect to both quantity and quality as these
pollutants impose harmful effects on receiving water bodies such as drinking water reservoirs.
1.2 Background
1.2.1 Best Management Practice (BMP)
In order to mitigate the problems associated with urban stormwater runoff, the concept of best
management practice (BMP), also known as low impact development (LID), water-sensitive
urban design (WSUD) or sustainable urban drainage system (SUDS) has recently been
introduced (Prince George's County 2002). BMP is a term used to describe structured and
engineered control devices and systems, usually environmentally friendly green features that
detain and treat polluted stormwater at its source.
Various new hydrologic practices which embody the BMP concept have been developed. They
fall broadly into two groups. The first group of BMPs aims at reducing runoff by promoting
infiltration into the ground through the use of infiltration ponds and bioswales. The second group
of BMPs aims at removing particulate matter and bound contaminants and nutrients through
sedimentation or biological processes. These include BMPs such as retention and detention
ponds, bioretention basins, sand filters, and constructed wetlands.
As the concept of the BMP became widely accepted, its adoption under the notion of BMP-
improved watersheds appeared in many places. Studies that compare watersheds with and
without BMPs further promote the practice of BMP (Bedan and Clausen 2009).
16
1.2.2 Our study site - Balam Estate Rain Garden in Singapore
Singapore as an island state has long been facing the problem of water scarcity. In an effort
towards building a sustainable nation, two-thirds of Singapore functions as local catchment areas
where rainfall in these areas is collected and channeled through drainage systems and eventually
retained in service reservoirs for water purification and subsequent use as drinking water. Thus,
urban runoff for Singapore serves as an extremely important source of drinking water.
Under the Active, Beautiful and Clean Waters (ABC Waters) Program, the Singapore
government, Public Utilities Board (PUB), plans to transform canals and drains into clean
flowing rivers and streams that could be open for public recreational activities (PUB 2014).
Therefore, out of concern over public health and cleaner drinking water sources, pilot projects
have been underway as BMPs of various types have been installed all over the country. Our
study site is one such installation.
One type of BMP, the bioretention basin, also commonly referred to as a biofiltration basin or
rain garden, is an excavated depression planted to look like a garden. They are sized to detain
and infiltrate stormwater runoff that would otherwise flow into the stormwater drainage system
and receiving reservoirs directly. Bioretention basins are also appealing because they beautify
landscape, encourage wildlife and biodiversity, reduce flow rate and are inexpensive, easy to
install and require minimal maintenance.
The first bioretention system in Singapore and also the study site for this research, Balam Estate
Rain Garden (240 m2 ), was built in 2008 and is located in the central-east region of Singapore
17
(Wang et al. 2009). Sized at 1.4% of the drainage area, it captures and treats stormwater runoff
from its neighboring high-rise residential area of about 16,800 in2 . Treated stormwater is
channeled into the Pelton Canal and eventually into the Marina Reservoir. The bioretention basin
not only acts as a sustainable treatment system for urban runoff using landscaping and plants, but
also adds to the greenery and improves the quality of life in the Balam residential estate (Figure
2-S-1, Chapter 2).
1.2.3 Design guidelines of bioretention basins
A well-designed bioretention basin ensures sufficient capture and treatment of pollutants to
achieve targeted pollutant removal rates. Design guidelines have largely originated from studies
carried out in countries that have predominantly temperate climates, including the United States
and Australia. For instance, in the U.S., bioretention basins are generally designed to capture and
treat a runoff volume that contains the first flush or the first portion of stormwater runoff from a
rainfall event that supposedly contains the highest pollutant load. The required amount of runoff
depth or "water quality depth (WQD)" to be retained by the bioretention basin varies from place
to place but is often in the range of 12.7 - 25.4 mm (Massachusetts Department of
Environmental Protection 2008; Minnesota Pollution Control Agency 2008; New Hampshire
Department of Environmental Services 2008). In other places, design guidelines are written in
terms of average recurrence interval (ARI). For instance, in some parts of Australia and
Singapore, bioretention basins are regulated to have a critical design flow rate of a 3-month ARI
storm (Australian Capital Territory Planning & Land Authority 2008; Public Utilities Board of
Singapore 2014).
18
Currently, due to the lack of field-scale studies carried out in tropical cities, design guidelines in
Singapore are adopted after those of temperate places, primarily Australia. However, due to the
distinctly different rainfall-runoff characteristics, guidelines originated from temperate places
may not be entirely applicable in the tropical context. Optimizing design of bioretention basins
for the urban tropical cities becomes a challenging problem that this study attempts to tackle.
Chapters 2 and 3 aim to evaluate performance of the Balam Estate Rain Garden in terms of
pollutant removal efficiency and hydrological capture efficiency respectively.
1.2.4 Hydrological and water quality models: the need for a process-based model
In order to optimize basin design for the tropics, it is important to have well-calibrated and
validated hydrological and water quality models that accurately predict pollutant effluent
concentrations in the bioretention basin outflow under different basin design configurations in
the tropical climate.
Out of all the currently available models, RECARGA and RECHARGE (the more complex
version of RECARGA) are the only hydrological models designed especially for bioretention
basins (Dussaillant et al. 2004). As of now, there is no water quality model that is designed for
bioretention basins. Current water quality modeling practice often treats the whole bioretention
basin as a black box and lumps all the biochemical processes under a single first-order decay
constant. This approach is applicable for macroscopic watershed-level modeling but is too low in
resolution as a method for improving basin designs and performance with respect to individual
pollutant and biochemical process. For instance, in considering nitrogen removal, the bacterial
kinetics of a whole chain of biochemical reactions consisting of ammonification, nitrification,
19
and denitrification should be considered in order to test for the sensitivity of each process to
different basin designs. The need for a process-based mechanistic model is what this study
attempts to fulfill. Chapters 4 and 5 aim at modeling the hydrological and water quality aspects
of the study site respectively.
1.3 Research objectives
This study aims to fulfill the research objectives below:
1. Assess whether a bioretention basin built according to temperate guidelines (i.e. Balam
Estate Rain Garden) is able to remove pollutants at targeted rates in a tropical setting;
2. Analyze the factors behind observed removal efficiency in terms of rainfall
characteristics such as rainfall depth, rainfall intensity, and antecedent dry period (ADP)
as well as runoff characteristics such as event mean concentration (EMC) and first flush;
and,
3. Investigate the suitability of current design guidelines for the tropics and suggest sizing
guidelines for basins in the tropics in terms of WQV or WQD.
4. Explore the applicability of the RECHARGE hydrological model for continuous
simulation at the field scale.
5. Predict bioretention basin hydrological performance in terms of infiltration percentage as
a function of engineering design parameters such as detention depth, ratio of drainage
area to basin area, and saturated hydraulic conductivity of the soil media.
6. Recommend basin configurations for improvement of hydrological performance
efficiency.
20
7. Develop a process-based model to describe the bacterial dynamics in contributing to
water quality improvement in the bioretention basin subsurface soil media.
8. Predict bioretention basin water quality performance in terms of pollutant removal
percentage and effluent concentration at various levels of bacteria growth kinetics.
1.4 Thesis overview
In accordance with the above research motivation and objectives, the thesis is divided into four
major chapters, each corresponding to a paper that has been published or is in preparation.
Chapter 2 of this thesis presents an evaluation of pollutant removal efficiency of the bioretention
basin under study based on a field monitoring program. Implications for stormwater management
in tropical cities are also discussed. This chapter has been published in Environmental Science:
Water Research & Technology, Vol 3, Issue 1, 78-91 (DOI: 10.1039/C6EW00285D).
Chapter 3 of this thesis, published as "Field evaluation of hydrological performance of a
bioretention basin in an urban tropical catchment" in WIT Transactions on the Built Environment,
Vol 165, 177-188 (DOI: 10.2495/UW160161), characterizes the hydrology of the study site
based on the field results from 80 events.
Chapter 4 of this thesis presents continuous hydrological modeling of 80 events in June

-
November 2013 using RECHARGE, an existing hydrological model developed specifically for
bioretention basins. Model calibration and validation are made based on comparisons to
observed field data that are presented in Chapter 3.
21
Chapter 5 of this thesis couples the outputs from Chapter 4 with a model developed based on the
mathematical formulations in CW2D (Constructed Wetland 2D model) to carry out event-wise
water quality modeling. Model calibration and validation are made based on comparisons to
observed field data that is presented in Chapter 2.
In Chapter 6, we draw design-specific findings from the previous chapters to make
recommendations for bioretention basin design in the tropics.
In Chapter 7, we summarize the key findings and contributions of this thesis. Some future
research directions are also suggested.
22
Chapter 2
Field evaluation of water quality performance
2.1 Abstract
Non-point source pollution is a prevalent problem throughout the world. Bioretention basins
have been deployed worldwide to treat stormwater runoff and alleviate eutrophication in
downstream water resources. However, basin performance in the tropics is poorly understood.
Given the distinctly different rainfall-runoff characteristics of tropical climates, whether basins
that are built according to temperate design guidelines are effective is questionable. There have
been no field studies based on continuous, high-resolution, long-term monitoring in the tropics.
In this study, 96 storms were monitored in the first bioretention basin in Singapore. Of these,
flow measurements were made during 80 events and samples were collected and analyzed for 15
water quality parameters (including nitrogen and phosphorus species, total suspended solids, and
chemical oxygen demand) during six events. The mean removal rates were 25%, 46%, and 53%
for total nitrogen, total phosphorus, and total suspended solids respectively. Results show that a
lack of storage capacity and resulting high overflow reduce pollutant removal efficiency for
high-rainfall-depth events. During nearly half of the events, treatment would not meet removal
targets for bioretention basin performance in treating surface runoff in Singapore. The transition
from efficient to non-efficient removal occurs at a rainfall depth between 10 and 30 mm. Low
EMC (event mean concentration) and weak first flush as a result of frequent and intense rainfall
in the tropics also contribute to low removal rate. The results suggest a need to revise
23
bioretention basin design guidelines for the tropics to be based on WQV or WQD (water quality
volume or depth) instead of ARI (average recurrence interval). A larger basin volume (WQD
between 10 to 30 mm) is recommended.
2.2 Introduction
Hydrologic modification brought about by traditional concrete-lined drainage systems due to
urbanization have unwanted consequences such as raised peak discharge, reduced groundwater
recharge, and non-point source pollution (Barron et al. 2013; He et al. 2009). BMPs (best
management practices) or LID (low impact development) entail decentralized strategies that aim
to reinstate pre-development hydrological features through detention and treatment of polluted
runoff at its source (Bedan and Clausen 2009; EPA 2000; Prince George's County 2002; Xuan et
al. 2013).
BMPs such as bioretention basins (also known as rain gardens) have been shown to be effective
in reducing runoff volume and removing pollutants (Davis et al. 2009; DeBusk and Wynn 2011;
Hunt et al. 2008; Roy-Poirier et al. 2010; Tang et al. 2016; Zhang and Guo 2014). Tang et al.
(2016) showed that even if bioretention basins occupy a small fraction of the city area, they can
substantially reduce the adverse hydrological effect brought about by imperviousness. Zhang
and Guo (2014) derived an analytic probabilistic expression to approximate the efficiency of
bioretention basins in capturing stormwater over the long term. Hunt et al. (2008) found such
basins achieved significant reduction of most pollutant concentrations including total nitrogen
(TN), total Kjeldahl nitrogen, ammonia, biochemical oxygen demand, total suspended solids
(TSS), copper, zinc, lead, fecal coliform, and Escherichiacoli. DeBusk and Wynn (2011) found
24
over 99% cumulative mass reductions for sediment, TN, and total phosphorus (TP), 97% flow
volume reductions, and 99% peak flow rate reductions. Reviews by Davis et al. (2009) and Roy-
Poirier (2010) both recognized the need to conduct further research to answer persisting design
questions such as basin sizing and soil media composition as adoption of bioretention basins
grows worldwide. Roy-Poirier (2010) also urged the development of bioretention models that
include both hydrologic and water quality processes as a tool to compare and verify the
suitability of currently inconsistent design guidelines.
Field studies have been carried out in various climatic conditions around the world to evaluate
the effectiveness of bioretention basins in removing pollutants. Early studies have primarily been
conducted in temperate regions such as North Carolina (Brown and Hunt 2011; Brown and Hunt
2011; Passeport et al. 2009), Maryland (Li and Davis 2009), Connecticut (Dietz and Clausen
2005; Dietz and Clausen 2006), and Virginia (DeBusk and Wynn 2011) in the USA; Victoria
(Hatt et al. 2009) and Queensland (Mangangka et al. 2015) in Australia; and Auckland in New
Zealand (Trowsdale and Simcock 2011). More recent studies include semi-arid climate regions
like Utah (Houdeshel et al. 2015) in the USA and cold climate regions like Quebec (Geheniau et
al. 2015) and Calgary (Khan et al. 2012) in Canada. For the tropics, there has been a lack of field
studies although there are several lab and mesocosm studies that investigated the impact of
media composition on nutrient removal (Goh et al. 2015; Lee et al. 2013; Ong et al. 2012). Our
field study attempts to bridge the gap in understanding basin removal efficiency in the tropics
and aims to suggest effective management guidelines.
25
Bioretention basins are generally required to remove 80-92% of TSS, 30-65% of TN, and 30-90%
of TP in various parts of the world (Brisbane City Council 2006; Department of Irrigation and
Drainage 2011; Dierkes et al. 2015; Massachusetts Department of Environmental Protection
2008; Minnesota Pollution Control Agency 2008; New Hampshire Department of Environmental
Services 2008; Public Utilities Board of Singapore 2014). In order to achieve these pollutant
removal targets, stormwater regulations usually require that basins be designed to capture a
certain volume of runoff so as to ensure most of the annual pollutant load is treated (WEF and
ASCE 2012). The term "water quality volume" or WQV refers to the runoff volume specified to
be captured and treated. Water quality depth (WQD) is the unit WQV per catchment area. These
recommendations have largely originated from countries that have predominantly temperate
climates, including the United States and Australia.
In most of the U.S., basins are set to capture a WQD that corresponds to the first half to one inch
(12.7 - 25.4 mm) of runoff depending on the water quality of the discharge and the requirements
of the receiving water body (Massachusetts Department of Environmental Protection 2008;
Minnesota Pollution Control Agency 2008; New Hampshire Department of Environmental
Services 2008). These requirements are based on results from Novotny's (1995) initial studies in
Florida that the first flush contains 90% of total pollutant load in an event. New Hampshire
standardized one inch of runoff for all discharges on the basis of a "90 percent rule" that
supposes that in the northeastern U.S., the cumulative volume from the first inch of runoff from
each event will amount to 90% of the total runoff in a year (New Hampshire Department of
Environmental Services 2008; Roesner et al. 1991). In Australia, stormwater guidelines are
written in terms of both WQD and ARI (average recurrence interval). In western Sydney, Upper
26
Parramatta River Catchment Trust (UPRCT) (2004) recommends retaining a WQD of 15 mm for
a 24-hour period and 20 mm for a 48-hour period to capture 60% of the average yearly rainfall.
The Australian Capital Territory requires that models such as MUSIC (Model for Urban
Stormwater Improvement Conceptualization) (Cooperative Research Centre for Catchment
Hydrology 2003), which generate WQVs, be used to analyze large catchments. For smaller
catchments, runoff resulting from a 3-month ARI event must be retained and discharged over
one to three days (Australian Capital Territory Planning & Land Authority 2008). In Europe,
stormwater management authorities generally do not have a regulatory procedure for approving
green sustainable urban drainage system (SUDS) infrastructure such as bioretention basins
(Dierkes et al. 2015; Eriksson et al. 2007).
Currently, design guidelines in Singapore are modeled after those of temperate areas, primarily
Australia (Facility for Advancing Water Biofiltration 2009). For instance, basins in Singapore
must be designed to accommodate a critical design flow rate for a 3-month ARI event and WQV
generated with the MUSIC model (Public Utilities Board of Singapore 2014). The targeted
removal rates are 80% of TSS, 45% of TN, and 45% of TP. However, sustainable practices from
temperate regions may not apply to tropical climates such as Singapore with its distinctly
different rainfall-runoff characteristics. Nevertheless, the null hypothesis for this study is that the
implicit assumption by Singapore: that bioretention basins built according to existing guidelines
from temperate regions are able to meet the intended removal targets in tropical climates.
Recent field studies have mostly focused on basin performance. Little attention has been paid to
the relationship between observed removal rates and hydrology in the context of rainfall and
27
runoff characteristics. There has also been a lack of field studies in the tropics. However, field
studies based on continuous, high-resolution, long-term rainfall and actual flow records (i.e. non-
synthetic runoff) as well as complete pollutographs (i.e. non-composite sampling) are important
for developing appropriate management and design guidelines, evaluating their suitability in the
tropics, and validating hydrological and water quality models. There is also an urgent need to
develop management measures to meet removal targets in the tropics.
In this paper, we present research on bioretention basin performance in terms of removal
efficiency of various pollutants in six storms of varying rainfall characteristics and hydrology in
a basin in Singapore. The objectives of this study were to:
1) assess whether a bioretention basin built according to temperate guidelines is able to
remove pollutants at targeted rates in a tropical setting;
2) analyze the factors behind observed removal efficiency in terms of rainfall characteristics
such as rainfall depth, rainfall intensity, and antecedent dry period (ADP) as well as
runoff characteristics such as event mean concentration (EMC) and first flush; and,
3) investigate the suitability of current design guidelines for the tropics and suggest sizing
guidelines for basins in the tropics in terms of WQV or WQD.
2.3 Methodology
2.3.1 Data collection
2.3.].] Flow and level measurement
A real-time monitoring system that consisted of flow and water-level measurement equipment
and automatic water quality samplers was installed and operated from April to November 2013
28
at Balam Estate Rain Garden in Singapore. A full description of the study site can be found in
the supplementary material. Four Doppler ultrasonic area-velocity level sensors (ISCO-2150
Area Velocity Module with 2105 Interface, Teledyne Isco, Lincoln, Nebraska, USA) were used
to continuously monitor flow rates at the inlet, outlet, and overflow culvert as well as ponding
water level at one-minute intervals. Figure 2-S-1 shows the placement of these four sensors.
Weekly basin maintenance was carried out to prevent sensor blockage by fallen leaves and trash.
2.3.1.2 Event statistics
Rainfall intensity (1-min interval) was measured with a tipping bucket rain gauge (Model TB4,
Hydrological Services PTY Ltd., Liverpool, NSW, Australia) at 3 m above ground. Per-minute
precipitation records from the rain gauge were used to calculate the total rainfall, event duration,
and ADP for each event. Two neighboring events are defined to be separated by an inter-event
time of at least six hours without rainfall (Driscoll et al. 1989). Intermittent light drizzles with a
total rainfall amount of less than 2.54 mm were excluded from the event count consistent with
Driscoll et al. (1989).
2.3.1.3 Sampling strategy and water quality
Water quality was monitored by collecting water samples throughout the storm duration and
subsequent ponding and drainage of the basin using four autosamplers (two ISCO-Avalanche
refrigerated portable samplers each equipped with fourteen 950-ml polypropylene bottles and
two ISCO-3700 portable full-size samplers equipped with 24 one-liter bottles). Samples were
conveyed to the autosamplers via 1.0-cm-diameter Teflon-lined suction tubing connected to
sample-collection strainers. The autosamplers automatically evacuated the tubes once prior to
29
collection of the first sample of each storm. The four sampling points were at the inlet, outlet,
outflow culvert, and on the basin floor (Figure 2-S-1).
Table 2-S-1 shows the time-based discrete sampling scheme followed. Surpassing a pre-defined
threshold level triggered the inlet sampler which in turn triggered the other three samplers with a
time delay of 0, 5, and 10 min for the outlet, pond, and culvert samplers respectively. Table 2-S-
2 shows a comparison with sampling strategies used in other studies reported in the literature.
Discrete sampling, instead of commonly practiced composite sampling, was followed in this
study. Discrete sampling has a clear advantage over composite sampling in terms of resolving
data variability over time by delineating an entire pollutograph. Although there are studies that
have incorporated discrete sampling, most have been done in the laboratory and few have been
done in the field. Analysis of discrete samples comes at considerable expense, but capturing full
pollutographs is important to ensure that the data collected are suitable for future research
involving calibration and validation of hydrological and water quality models. Preliminary
sampling trials were conducted during events in April 2013 to ensure that the total sampling time
imposed by the time-based sampling scheme in Table 2-S-1 was suitable for capturing full
pollutographs during both small and large events.
During six events of varying characteristics (Table 2-1), discrete water samples were collected
and then tested for concentrations of 12 parameters including various species of nitrogen and
phosphorus (Table 2-S-3). Three additional parameters (organic nitrogen, ON; total particulate
inorganic phosphorus, TPIP; and total particulate organic phosphorus, TPOP) were derived (Eq.
S-1). Heavy metals are not measured as the main motivation for this study was eutrophication of
30
the downstream reservoir that receives the effluent from this bioretention basin. The total number
of samples analyzed from the six events for the different parameters range over 102-121, 122-
144, and 48-60 for the inlet, outlet, and culvert respectively (Table 2-S-4). Table 2-S-4 also lists
the number of samples analyzed for each parameter in each event. Samples were transported
from the site to the laboratory within 30 hours after the start of the event and were preserved and
processed according to Standard Methods (APHA et al. 2009). For every 20 samples, a duplicate
and a standard curve check were made for all parameters and their recovery rates were within
acceptable limits (89-110% for sample duplicates and 83-117% for standards). Blanks (deionized
water) for all the parameters in all the events were below method detection limits.
2.3.2 Data analysis
2.3.2.1 Pollutantloading
Total cumulative pollutant loads at the inlet, outlet, and culvert were estimated for each sampled
event by aggregating flow measurements over the time scale of samples according to Equation
2-1:
-1 M/2
V= Qi At + QCAt + YQ At
i=-N/2 S i=1 Equation 2-1
Load = >jVjcj
j=1
where Qc (L/s) is the flow rate at the current sampling time point; Qj (L/s) is the flow rate at the
current sampling time point plus or minus iAt; At is the time interval between two readings of
flow rate which in this study is one minute; N and M are the number of readings of flow rate
between the current and previous sampling time point and between the current and next sampling
31
time point respectively; C; (mg/L) is the concentration measured at sampling time point j; V (L)
is the total flow volume assigned to thej' sample; and, S is the total number of samples.
2.3.2.2 Removal rate as a measure ofperformance efficiency
The mass removed is taken as the mass that is either retained in the basin or eliminated by
biochemical reactions in the basin. The removal rate was calculated by comparing the difference
in mass between the inflow and outflow (via both the culvert and the outlet) of the basin, as
shown in Equation 2-2:
Inlet mass - (Culvert mass + Outletmass) X100%

Removal rate (%) = Equation 2-2
Inlet mass
2.3.2.3 Event mean concentration (EMC)
The EMC of the influent measures the flow-averaged concentration of runoff reaching the basin
and can be taken as a single index that represents the water quality of runoff during a particular
event. It is defined according to Equation 2-3:
Cumulative inf luent mass EL1E

EMC = jjEquation 2-3
Cumulative influent volume Z'= 1 Qi
where Qi (L/s) is the flow rate measured at one-minute intervals over the entire event and n is the
total number of flow-rate readings.
32
2.4 Results
2.4.1 Hydrologic performance
Figure 2-1 shows the hyetographs and hydrographs for the inlet, outlet, and culvert discharges of
the six sampled events, arranged in order of increasing rainfall depth. The hydrologic
effectiveness of the bioretention basin in managing stormwater runoff often lies in its ability to
reduce peak flow rate, delay the time of peak outflow, and delay hydrologic volume via runoff
storage and holdup. These three aspects are quantified using three metrics defined by Davis
(2008): Rpeak (peak flow reduction ratio, Rpeak = qpeak-out / qpeak-in); Rdelay (peak delay ratio Rdelay
tq-peak-out / tq-peak-in); fV24 (fraction of input water leaving the bioretention basin within 24 h, fV24
Vout-24 / Vin) where qpeak-out, qpeak-in, tq-peak-out, tq-peak-in, Vin and Vout-24 are peak outflow (L/s), peak
3
inflow (L/s), time to peak inflow (hr), time to peak outflow (hr), total inflow volume (m ), and
total outflow volume (in 3 ) after 24 hours respectively. The target values recommended by Davis
are Rpeak < 0.33, Rdelay > 6, and fv24 < 0.33 respectively (Davis 2008). In this study, for the six
sampled events, mean Rpeak (0.06) meets the target while mean Rdelay (3.3) and meanfv24 (1.08)
fail to meet targets (Table 2-1). This shows that the basin is effective in reducing peak flow rate
(Rpeak) across events of varying rainfall depths (Table 2-1). The ability to delay peak timing
(Rdelay) is better for small events (Event #3, #2) than large events (Event #6, #4). The ability to
store and hold up runoff (fV24) is poor in general.
In terms of hydraulic retention time, for the 96 storm events that occurred during the monitoring
period, statistics based on 80 events with usable flow data indicate that within the bioretention
basin, the ponding duration is 12.5 5.5 hr (Wang et al. 2016). For the 59 events with sufficient
influent to cause a culvert overflow, overflow stops at a mean pond water level of 13.9 cm or
33
equivalently a surface storage of 40 m3 during the receding phase. The mean ponding duration
between zero culvert overflow and zero pond water level during this phase is 10.1 hr. Drainage
of surface ponding of 40 m 3 over a basin area of 240 m 2 in 10.1 hr gives an estimated infiltration
rate of 1.7 cm/hr (Wang et al. 2016). This phase of infiltrating ponded water results in an
increase in subsurface storage of 23.7 in 3 , corresponding to about 100% of the total available
porous space in the unsaturated zone (estimated to be about 23.3 M3 ) (Wang et al. 2016). This
implies that the unsaturated zone is likely to stay saturated under a ponding condition. The
infiltration rate of 1.7 cm/hr is within the same order of magnitude as field measurement of 4.4
cm/hr using a double-ring infiltrometer (Wang et al. 2016). As water continues to drain from the
basin, the mean gravity drainage rate is 0.3 cm/hr under an unsaturated condition (Wang et al.
2016).
2.4.2 Pollutant removal efficiency
2.4.2.1 Parameterremoval rates
Treatment of runoff is achieved by various physical and biochemical processes such as
sedimentation (Roy-Poirier et al. 2010), filtration (Liu and Davis 2014), adsorption (Lucas and
Greenway 2011), biodegradation (Chen et al. 2013), and plant uptake (Henderson et al. 2007)
within the basin. Table 2-2 summarizes the mean and one standard deviation of the inlet, outlet,
and culvert concentrations of the 15 parameters in all samples across the six sampled events.
Pollutant concentration decreases successively from the inlet to culvert to outlet which reflects
the improvement in water quality as runoff is routed through the basin. Pollutographs of TN, TP,
and TSS for each of the six sampled events are shown in Figure 2-S-2. Greater improvement in
water quality is seen if water is filtered through the soil media (outlet concentrations) than if it
34
passes through the surface basin only (culvert concentrations). Reduced concentrations of TSS,
TPIP, and TPOP at the culvert reflect sedimentation of particulates in the surface basin (Li and
Davis 2009; Roy-Poirier et al. 2010). Reduced concentration at the outlet reflects additional
subsurface treatment such as filtration (Liu and Davis 2014), adsorption (Henderson et al. 2007;
Lucas and Greenway 2011), and plant uptake (Henderson et al. 2007) for phosphorus species
(Sample et al. 2012) as well as aerobic and anaerobic biochemical transformations by soil
bacteria for nitrogen species (Chen et al. 2013). Reduced COD concentration reflects possible
removal mechanisms including biodegradation, sorption to bioretention media, and plant uptake
of organic compounds (likely sources being natural organic materials as well as petroleum
hydrocarbons from carparks and the main road servicing the residential estate) (LeFevre et al.
2015).
The total mass of each of the 15 water quality parameters passing the inlet, culvert, and outlet
during the six sampled events is shown in Table 2-S-5. The resultant net removal rate for each
parameter is shown in Table 2-2. TSS has the highest mean removal rate of 52.8%, followed by
ON (52.7%), TP (46%), and P04-P (45%). Although there is a net removal of TN of 25%, there
is net export instead of removal of N02-N (-56%) which suggests incomplete denitrification.
Although the presence of a saturated anoxic zone should improve nitrogen removal (Zinger et al.
2013), denitrification is found to occur mainly during dry periods (Subramaniam et al. 2016).
The export of NOx-N was also observed in other field studies (Hatt et al. 2009; Line and Hunt
2009). Among the 15 parameters, three species (ON, TDP, and P04-P) show positive removal
rates (6-80%, 4-66%, and 4-75% respectively) in all six events. COD removal rate (15%) is low
although bioretention basins have been reported to be 85% or more effective in removing
35
hydrocarbons (e.g. polycyclic aromatic hydrocarbons including naphthalene) (Diblasi et al. 2009;
LeFevre et al. 2015).
2.4.2.2 Event removal rates
Basin performance is often assessed in terms of the removal rate of three water quality
parameters: TSS, TN, and TP. Bioretention basins are typically designed with removal targets of
80-92% for TSS, 30-65% for TN, and 30-90% for TP (Brisbane City Council 2006; Department
of Irrigation and Drainage 2011; Dierkes et al. 2015; Massachusetts Department of
Environmental Protection 2008; Minnesota Pollution Control Agency 2008; New Hampshire
Department of Environmental Services 2008; Public Utilities Board of Singapore 2014). Figure
2-2 singles out removal rates of TN, TP, and TSS during individual events for closer evaluation.
The removal rate and its associated uncertainty are computed according to Eqs. (2) and (S-2),
respectively.
The removal targets for TN, TP, and TSS are generally met only for small events (Events #3, #2,
and #1) but not for large events (Events #5, #6, and #4) (Figure 2-2). The mean removal rates for
TN, TP, and TSS are 54%, 72%, and 90% respectively for small events, compared to -3%, 19%,
and 16% for large events. For all three pollutants, as rainfall depth increases, removal rate
generally decreases. This is mirrored by the upward trend seen in the fraction of culvert-overflow
as event size increases. The high culvert-overflow fraction during large events reduces the
removal rate since the overflowed portion of runoff failed to infiltrate and receive soil treatment.
A noticeable drop in removal rate is seen in the transition from Event #1 (3-month ARI) to Event
#5 (< 1-year ARI) for all three pollutants. This drop is especially evident in phosphorus species
36
and TSS, which are predominantly in particulate form. These results point to an undersized basin
with insufficient treatment efficiency.
EMC might play a role in affecting removal rate. Higher removal rates for TN and TP are
observed when the influent concentration is greater (McNett et al. 2011). However, the extent of
the effect of EMC could not be distinguished in this study (Figure 2-2). Although all the large
events with low EMC (Events #4, #5, and #6) show low removal rates, a small event with a low
EMC (Event #2) still achieved a high removal rate since the majority of inflow was contained
within the basin and received full treatment. Based on these results, a high culvert-overflow
fraction appears to be the driver in reducing removal rate.
2.4.3 Water quality of influent
2.4.3.1 EMC of influent
The influent EMCs of 15 parameters for the six sampled events are shown in Table 2-2. TN, the
most comprehensive measure of the nitrogen pollutant load, has a mean EMC of 2.20 mg/L. The
majority of nitrogen comes into the basin in the form of ON (1.03 mg/L) which is the least
available form and is most likely in particulate form. Some of the nitrogen load comes in as the
dissolved and reduced form NH3-N (0.42 mg/L) while the rest is in the oxidized dissolved forms
of N03-N (0.52 mg/L) and N02-N (0.09 mg/L).
The EMCs of phosphorus species are in general lower than those of nitrogen species in the
runoff. Most of the phosphorus exists in particulate form as the mean EMC of TP is 0.19 mg/L
compared to 0.07 mg/L of TDP. Among the four phosphorus species analysed (TPIP, TPOP,
37
P04-P, and TDOP), the inorganic phosphorus species exhibit higher EMCs than the organic
species. Specifically, the dissolved (P04-P) and particulate (TPIP) inorganic phosphorus species
have mean EMCs of 0.05 mg/L and 0.09 mg/L respectively. The dissolved (TDOP) and
particulate (TPOP) organic phosphorus species have mean EMCs of 0.01 mg/L and 0.03 mg/L
respectively.
A comparison between EMCs in this study and published data (Table 2-S-6) shows that EMCs
are generally lower in the tropics than in temperate climates. The EMC of TSS (40 mg/L) found
in this study is comparable to those in the literature for tropical studies but is 61 - 79% lower
than those found in temperate studies. Similarly, the EMC of COD (54 mg/L) is lower by 18 - 46%
(Table 2-S-6).
2.4.3.2 Firstflush of influent
Bioretention basins are typically designed in temperate countries with the expectation of
capturing the first flush of runoff. However, given that rainfall in tropical climates tends to be
frequent and intense, a strong first flush might not exist. Thus, the presence or absence of a
strong first flush in an urban catchment in the tropics may be important to assessing whether a
bioretention basin will perform as intended.
Figure 2-S-3 shows the event-wise cumulative-mass-cumulative-runoff curves of TSS, TN, and
TP. Among the three parameters, TSS shows the most prominent first-flush phenomenon,
followed by TP and TN. Table 2-3 uses established literature criteria to evaluate the presence of
first flush for all the water quality parameters (Bertrand-Krajewski et al. 1998; Deletic 1998;
38
Geiger 1987). The parameters that most frequently exhibit first flush (in at least five out of six
events) according to Geiger's (1987) criterion are particulate forms (TN, TKN, ON, TP, TOP,
TPIP, TPOP, TSS, and COD). Le and Chua (2014) similarly see a stronger first flush for
particulates than dissolved substances. TSS exhibits first flush in all events while TN, TKN, ON,
TP, and TPIP exhibit first flush in all but the smallest event (#3). The dissolved species like
NH3-N, N03-N, TDP, P0 4-P, and TDOP exhibit first flush less frequently, in at most three out of
six events. When more stringent criteria are used to assess the presence of first-flush behavior,
fewer water quality parameters exhibit first flush during fewer events. Only ON exhibited first
flush in as many as four out of six events under Deletic's (1998) criterion of 40% mass in 20%
volume while the other parameters exhibit first flush in at most three out of six events. NH3-N is
the only parameter that exhibited first flush, and then in only one event, based on Bertrand-
Krajewski et al.'s (1998) criterion of 80% mass in 30% volume. Therefore, the first-flush
phenomenon is not found to be prominent in this study if a stringent criterion is imposed.
2.4.4 Rainfall characteristics
The low pollutant removal rates seen in this study can be attributed to the distinct rainfall
patterns of the tropics. Histograms and cumulative probability curves of rainfall characteristics
observed during this study are shown in Figure 2-S-4. A total of 96 storm events occurred during
the eight-month monitoring period of April-November 2013. The recorded rainfall tends to be
heavy with 46% of the storms having a total rainfall > 10 mm; 19% storms > 30 mm; and 3%
storms > 70 mm. The storm events also tend to have short durations. Of the 96 events, the
maximum event duration is 11 hours with 10% of the events < 0.25 hr; 50% < 1.5 hr; and 90% <
6 hr duration. Storms were frequent during the monitoring period and 42% of storms have an
39
antecedent dry period (ADP) < 1 day; 71% < 2 days; and 99% < 9 days. The mean rainfall
intensity, event duration, and ADP are 7.49 mni/hr, 2.8 hr, and 2 days respectively (Table 2-S-7).
These characteristics are all typical of a tropical climate, in which frequent, short, intense
thunderstorms are the norm.
2.5 Discussion
2.5.1 Frequent heavy flushing in tropical catchment appears to lead to low pollutant
concentration
High-intensity rainfall events are less common in temperate climates than in the tropics. For
instance, the 10-min peak rainfall intensities of the six sampled events in this study are below the
1-year ARI for Singapore (Table 2-S-8). However, if these tropical storms were to occur in
temperate regions, they would be assigned higher ARIs (Table 2-S-8). For example, if Events #4
and #5 were to occur in Seattle, they would be considered 100-year events.
The regular flushing of the Balam Estate catchment by frequent and intense storms appears to
have led to reduced pollutant accumulation. Table 2-2 shows that the events (Event #2, #6, #5,
and #4) with either short ADP (i.e. recent flushing) or high rainfall depth tend to have lower
EMCs compared to events (Event #1 and #3) with long ADP (i.e. sufficient time for pollutant
accumulation) and low rainfall depth (Table 2-1).
The high EMCs of the two long-ADP events (Events #3 and #1) increase the overall mean EMC
for each of the 15 water quality parameters (Table 2-2). Event #1 (25 June 2013) is unique
because it was the first storm after a historically severe trans-boundary air pollution episode in
Singapore. The air quality Pollutant Standards Index on 21 June 2013 surpassed the previous
40
record in 1997. Coupled with a long ADP of 8.25 days, Event #1 has the highest EMC for almost
all of the parameters including TN, TKN, ON, NH3-N, TP, TDP, P04-P, and TPIP. Event #3 is
also unusual because it is a very small event (3.2 mm total rainfall) preceded by a very long ADP
of 9.08 days (Table 2-1). Event #3 has the highest EMC of N02-N, TOP, TDOP, TPOP, TSS,
and COD. Even with the two long-ADP high-EMC events, the overall mean EMCs in this study
are low compared to temperate studies (Table 2-S-6).
2.5.2 Lack of storage capacity affects removal efficiency
Removal efficiency is an important assessment criterion for bioretention basins as it reflects the
basin's performance in improving water quality. Results from this study demonstrate that
performance of the bioretention basin in Singapore is sub-optimal for large storm events. The
main reason is insufficient storage capacity which causes most runoff to bypass the basin with
minimal treatment during large, but still common, storm events.
Comparison of three events (#2, #1, and #5) illustrates the need for greater storage capacity in
tropical basins (Figure 2-2). In Event #2 (a small event of 8.0 mm), the storage capacity is
sufficient since only 28% of the inflow exits the basin by culvert overflow (Table 2-1). As most
of the inflow is soil-filtered and fully treated, a satisfactory removal rate is achieved (91%, 58%,
and 75% for TSS, TN, and TP respectively, in Figure 2-2 and Table 2-2). The removal targets of
80%, 45%, and 45% are also fully met. The rainfall characteristics of Event #2 (8.0 mm of total
rainfall, 32 min of event duration, and 2.06 days of ADP, in Table 2-1) lie close to the median
storm characteristics observed in this study (8.2 mm of total rainfall, 1.5 hours of event duration,
41
and 1 day of ADP, in Figure 2-S-4) and the basin is able to treat these smaller storms adequately.
This shows that a tropical basin can perform acceptably when adequately sized.
Event #1 is similar in rainfall depth (10.2 mm) to Event #2 but has a significantly higher EMC
(Table 2-2) and load (Figure 2-2, Table 2-S-5) due to the severe haze during the ADP of 8.25
days. The basin storage capacity is sufficient to retain 63% of the inflow (Figure 2-2). Despite
the higher pollutant load, the removal targets are still generally met and within expectations
(86%, 35%, and 69% for TSS, TN and TP respectively, in Figure 2-2, Table 2-2). Therefore,
removal targets are still achievable for a heavily-polluted storm if sufficient storage capacity is
provided. Hence, the inherent pollutant removal mechanism in a sufficiently-sized basin is not
the limiting factor in affecting removal rates.
In contrast to Events #2 and #1, Event #5 is a large event (29.4 mm of rainfall). Storage capacity
is insufficient for Event #5 since almost all of the inflow (95%) exits the basin by culvert
overflow (Figure 2-2). Hence, the removal rates for Event #5 are below targets (31%, 10%, and
33% for TSS, TN, and TP respectively, in Figure 2-2 and Table 2-2). However, events of such
rainfall depth are not uncommon in Singapore as 19% of the events in this study have > 30 mm
of rainfall (Figure 2-S-4).
Extrapolating from the hydrological performance during events during the monitoring period, at
the current storage capacity, the Balam Estate Rain Garden would not have been able to meet
pollutant removal targets during many events (Figure 2-3). The transition between the sampled
events that meet removal targets and those that do not occurs at a rainfall depth between 10 mm
42
and 30 mm. Based on the cumulative probability curve of rainfall amount in Figure 2-S-4, this
range of rainfall depth translates to 19% (> 30 mm events) to 46% (> 10 mm events) of events.
Therefore, at the current storage capacity, as many as 46% of the events during this monitoring
period would not have met pollutant removal targets.
A comparison between mean removal rates in this study and published data for basins under
various climate conditions and configurations (Table 2-4) shows that removal of TSS in this
basin is lower while TN and TP are within observed ranges. The rates for TN and TP in this
study are close to those found by Brown and Hunt (2011) for undersized basins, which highlights
the adverse effect of overflow on removal rates. Li and Davis (2009) similarly noted the close
relationship between hydrologic and water quality performance. Ong et al.'s (2012) study at the
same site yields higher removal rates than this study, however, they used a different definition of
removal rate. In particular, they did not deduct culvert mass load from inlet mass load in Eq. 2,
thus treating culvert mass bypass as removal too. Further studies need to be conducted to
determine the removal efficiencies of basins in other tropical regions. In Europe, the
bioretention basin is considered a green sustainable urban drainage system (SUDS) and Dierke et
al. (2015) reported that SUDS as a group was able to achieve removal of 50-90% of TSS, 30-60%
of TP, and 10-40% of TN.
The presence of a saturated internal water storage zone (IWS) seems to enhance denitrification
and TN removal. An IWS not only provides additional storage volume and increases hydraulic
retention time, it also provides an anoxic environment which promotes denitrification. In all
basins with IWS cited in Table 2-4, but not in this study, the general target of > 30% removal of
43
TN is achieved. In contrast, only 55% of the bioretention basins without IWS are able to meet
the target.
The removal rate of TP shows more variability than TN and TSS. A larger deviation in removal
rates is seen across various studies (Table 2-4). In several studies, export of TP is observed and
this could be attributed to leaching of P from the bioretention media itself which often consists of
mulch and other organic matter. For effective P removal, soil media with high cation exchange
capacity and low P-index has been recommended (Hunt et al. 2006).
Although TSS removal in this study (53%) is the lowest among all the studies in Table 2-4, it has
the highest mean removal rate among the water quality parameters assessed in this study. Li and
Davis (2008a,b) note that as soil media capture TSS, finer particles become trapped in the upper
layer leading to media stratification and reduced hydraulic conductivity. This could partially
explain the reduction in TSS removal in this study (five years after construction) compared to
Ong's (2012) (one year).
2.5.3 Design guidelines for sizing bioretention basins in the tropics need to be revised
Bioretention basins were originally designed based on studies done and experience gathered in
temperate places. Current design guidelines in the tropics are adopted from those developed for
temperate climates. As discussed above, in Singapore the basin size or WQV is based on water
quality performance predicted by design curves generated using MUSIC, a model developed
based on studies done in temperate Australia (Cooperative Research Centre for Catchment
Hydrology 2003; Public Utilities Board of Singapore 2014). Following after Australia, basins in
44
Singapore are also designed to treat a critical design flow rate equivalent to the 3-month ARI
storm (Public Utilities Board of Singapore 2014). The performance of Balam Estate Rain Garden,
as shown in this study, does not meet the pollutant removal rates forecasted for large but still
common storm events (Figure 2-3). These results suggest that design guidelines and sizing
curves developed for temperate settings cannot simply be transferred to tropical regions. Rather,
there is a need to develop design guidelines specifically for tropical applications.
2.5.3.1 WQV or WQD is a betterparameterfor sizing basins than ARI in the tropics
This study recommends that design guidelines for the tropics be given in terms of the more
definitive quantity, WQV or WQD, instead of AR.
ARI is an inferior design parameter because it covers a wider spectrum of events with different
combinations of intensity and duration and thus includes distinctly different rainfall depths.
Singapore's IDF curves show that the 3-month ARI corresponds to an approximate range of 5 to
35 mm of rain (Public Utilities Board of Singapore 2011). A 3-month ARI event could be intense
and short (e.g. Event #1) or mild and long (e.g. Event #6). The term, ARI, fails to distinguish
events such as these even though they generate vastly different runoff amounts. For instance,
events with rainfall depth > 12 mm could be either < 3-month ARI or < 1-year ARI and they lead
to enormously different culvert overflow rates (Figure 2-3). Thus, events that are mild in rainfall
intensity but high in rainfall depth (e.g. Event #6) could still exceed the basin capacity and fail to
meet removal targets.
45
ARI is ambiguous in describing events in the tropics because a modest increase in ARI can lead
to a more significant increase in rainfall intensity in the tropics than in temperate regions. Figure
2-S-5 compares the 10-min rainfall intensity of events of various ARIs in Singapore, Melbourne,
and selected cities in the U.S. The 10-min rainfall intensity in Singapore is higher than that of
temperate regions in the U.S. and is similar to that of the sub-tropical city of Miami. In this study,
storms that are only slightly higher in ARI terms (Events #4 and #5) are in fact much more
intense and generate much higher runoff volumes (Table 2-1).
Since ARI is not fully definitive, it fails to pinpoint a definite runoff volume (WQV) that needs
to be retained in order to achieve the desired removal rate. However, this is the volume that
ultimately determines basin performance. As demonstrated in this study, in tropical events of
slightly higher ARI, the higher runoff volume results in a sharp drop in removal rates (Figure 2-2,
Table 2-2). Therefore, WQV is a better parameter for sizing retention basins than ARI in the
tropics.
2.5.3.2 Recommended value of WQ V or WQD for the tropics
Based on water quality considerations, this study recommends a WQD range of 10-30 mm. This
is the same as that recommended by Wang et al. (2016) based on a hydrological analysis of 80
events (Figure 2-3). The exact amount of runoff (WQV or WQD) that needs to be retained for a
tropical catchment in order to meet removal targets requires further analysis based on hydrologic
and water quality modeling. In the U.S., Roesner et al. (1991) found that capturing a WQD of
one inch (25.4 mm) of runoff is sufficient to capture 90% of the annual runoff in a broad cross
sample of U.S. cities. However, from results of this study, if Singapore were also to capture 90%
46
of annual runoff, a WQD of about 32 mm would be required. This equates to a WQV of 540 m 3
for the Balam Estate Rain Garden, which is about 16 times the volume of the current basin (33
m3 ). Basins of this size would be difficult to site in land-limited urban catchments. Therefore,
other strategies to increase storage volume would need to be explored or pollutant removal
targets would need to be re-evaluated. Alternatively, as Strecker et al. (2001) point out, a
maximum effluent concentration instead of removal targets could be imposed for basin
discharges. In contrast to concentration targets, removal percentages can underestimate basin
effectiveness in the case of cleaner influent.
2.6 Conclusions
Performance of the first bioretention basin in Singapore was assessed based on 96 storms, of
which 80 storms yielded useful recorded flow data and six storms were sampled for 15 water
quality parameters using a discrete time-based sampling strategy. A combination of hydraulic
data and rainfall-runoff characteristics was used to assess system performance in a tropical
setting and recommend practical engineering strategies to design and manage bioretention basins
in the tropics. The following may be concluded from this study:
1. The Balam Estate Rain Garden is very effective in reducing peak flow rate (mean Rpeak --
0.06) but is not effective in delaying peak timing (mean Rdelay = 3.3) and reducing hydrologic
volume via runoff storage and holdup (meanfV24 = 1.08). The ponding duration is 12.5 5.5
hr and the estimated infiltration rate based on statistics of 59 overflow events is 1.7 cm/hr.
This implies that the unsaturated zone is likely to stay saturated under a ponding condition.
Under an unsaturated condition, the mean gravity drainage rate is 0.3 cm/hr.
47
2. The Balam Estate Rain Garden is able to improve water quality. Higher removal is seen at
the outlet when runoff is allowed to filter through the soil media than if it overflows via the
culvert. The highest mean removal rate of 53% was seen in TSS. The dissolved specie N02-
N most often showed net export instead of removal (mean removal of -56%). This might
suggest extensive nitrification but incomplete denitrification within the soil media. ON, TDP,
and P04-P show positive removal rates in all six events.
3. In terms of event-specific removal rates, removal targets for TN, TP, and TSS are generally
met only for events with small rainfall depths (Events #3, #2, and #1) and not for those with
greater rainfall depths (Events #5, #6, and #4). The mean removal rates for TN, TP, and TSS
are 54%, 72%, and 90% respectively for small events, compared to -3%, 19%, and 16% for
large events. A lack of storage capacity and resulting high culvert overflow is the main driver
in reducing removal efficiency for large, but still common, storm events. The transition from
efficient to non-efficient treatment occurs at a rainfall depth between 10 and 30 mm. At the
current storage capacity, the basin performance during as many as 46% of the events in this
monitoring period would not meet the pollutant removal targets.
4. The range of influent EMCs observed in this study are comparable to those in tropical
catchments presented in the literature but are lower than temperate counterparts. The frequent
and intense rainfall of the tropical climate likely leads to lower EMCs than in temperate
climates. The first-flush phenomenon of the influent is not found to be prominent if a
stringent criterion is imposed. Low influent EMC and weak first flush appear to contribute to
low pollutant removal rate. Further research is required to confirm this.
5. There is a need to revise design guidelines for sizing bioretention basins in tropical
catchments. This study recommends design guidelines in the tropics be specified in terms of
48
the more definitive quantity WQV or WQD instead of ARI. A larger basin storage volume
(WQD in the range of 10 to 30 mm) is needed in the tropics in order to treat the intense but
common events in the tropics.
2.6.1 Study impact andfuture work
Results from this study could be of use to water resource managers who are utilizing bioretention
basins and other BMPs to improve the water quality of stormwater runoff in a similar tropical
climate setting. Basin designers should especially take high-rainfall-depth events into
consideration when sizing for tropical bioretention basins since they tend to produce basin
bypass which leads to poor pollutant removal efficiency. The exact amount of runoff (WQV or
WQD) that needs to be retained in order to meet removal targets requires additional research
employing long-term water quality and hydrological models that are specific to bioretention
basins. The long-term, continuous, high-resolution, and complete hydrographs and pollutographs
available from this study can serve as sources of observed measurements for calibration and
validation of such models.
2.7 Acknowledgements
This research was funded by the National Research Foundation (NRF), Prime Minister's Office,
Singapore under its Campus for Research Excellence and Technological Enterprise (CREATE),
through the Singapore MIT Alliance for Research and Technology's Center for Environmental
Sensing and Modeling (CENSAM) research program. The authors would also like to thank Alwi
Alkaff of Nanyang Technological University of Singapore and Maria Stefanie Suryadharma of
National University of Singapore for their assistance in field work; Ong Geok Suat and Yau
49
Wing Ken from Public Utilities Board of Singapore as well as Veera Sekaran of Greenology Pte
Ltd for their insightful discussions on the design and construction of the Balam Estate Rain
Garden; and Adeline Wong from Marine Parade Town Council for her support of the field
monitoring program.
50
0.0 0.5 1.0 1.5 2.0 2.5 3.0 0 2 4 6 8 10 12 14 16 18 20 22 24
fi 01 0.0. . I . 1. I . I . I .
I . I . I
I . I . I
I 0
I. Event #3
0.5- Event #3
0- Rainfall -300
Rainfall K;
- Inlet discharge - Outlet discharge
1.0 - Inlet sample 1.0- Outlet sample
- - Culvert discharge [250
2) -20 -J
E 1.5 E 1.5-
E 200
_) E
-
-
C) Co
(
2.0 2.0 -15 -=
C
a) 0
- 150 C2-
2.5 .)
S2.5
-10
100
3.0 3.0-
-5
3.5 50 3.5-
4.0 n 4. - I .~. - -- -;- - - - -I -0

.0 0.5 1.0 1.5 2.0 2.5 3.0 0 2 4 6 8 10 12 14 16 18 20 22 24
.
Time (hr) Time (hr)

0 .0 0.5 1.0 1.5 2.0 2.5 3.0 0 2 4 6 8 10 12 14 16 18 20 22 24
0.0
i i i i i I 350 0. 0. I I I I . I . I .
i 30
.
0.5 I I Event #2
0.5
Event #2
Rainfall - 300
. Rainfall -25
1.0 * Inlet sample 1.0
- 250 Outlet sample
- Culvert discharge
* Culvert sample -20
E 1.5 E 1.5 (D
E - 200 E 2)
(U
2.0 * Co = 2.0 -15 -C
-150 .C CU
C
2.5 m 2.5 0
.-10 .5
- 100
3.0 3.0 I I
-50
-5
3.5 3.5
4.0 -0 4.0 -0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 0 2 4 6 8 10 12 14 16 18 20 22 24
Time (hr) Time (hr)
51
0.0 0.5 1.0 1.5 2.0 2.5 3.0 0 2 4 6 8 10 12 14 16 18 20 22 24
0.0 3 50 0.0, 30
Event #1 Event #1
0.5 0.5-
Rainfall -3 00 Rainfall
-25
1.0 * Inlet sample Outlet sample
- Culvert discharge 2 50
- Culvert sample -20
E 1.5 E1.5
Cc)
-
E -2 00 E
2)0= 2E
2.0 2 2.0.- -15
50-
Co . i
2.5 . K 2.5- 0
-10 -89
00
3.0 3.0
-
-5
50 3.5
3.5
- -r - - -0
4.0 -0 4.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 0 2 4 6 8 10 12 14 16 18 20 22 24
Time (hr) Time (hr)
0.0 0.5 1.0 1.5 2.0 2.5 3. 0 2 4 6 8 10 12 14 16 18 20 22 24
0.0 4= 3 50 0.0. 30
10111 1 I Event #5
0.5 Rainfall - 300 0.5 Event #5
-
- Inlet discharge - 25
* Inlet sample Rainfall
1.0 1.0-
I - Culvert discharge Outlet discharge
-
-250 -
o Culvert sample Outlet sample -20
E 1.5 E 1.5- a>
-
E -200 a)E 2)
2.0 0) 2.0- -15 -C

-
-C
-150
CU 2.5- 0 2.5-
10
- 100 *0
3.0- 3.0-
-5
- 50
3.5- 3.5-
40
- - *0 - 0
-
4.0-
.
0.0 0.5 1.0 1.5 2.0 2.5 3.0 0 2 4 6 8 10 12 14 16 18 20 22 24

Time (hr) Time (hr)
52
0.0 0.5 1.0 1.5 2.0 2.5 3.0 0 2
I 4 6 8 10 12 14 16 18 20 22 24
0.0- I . I
350 0.0- ~~.I ~'I
1~~~~
30
I i
0.5 0.5- Event #6
- -300
Event #6 Rainfall - 25
I I - Outlet discharge
1.0- Rainfall 1.0-
- 250 Outlet sample
- Inlet discharge
* Inlet sample 20 C,)
1.5- -j 1.5
E - Culvert discharge E
-
* Culvert sample 200 CU
2.0- 0)
I.- 2.0 -15
-
4-- (U U.)
C 150 -C C:
0 -o
2.5- 2.5-
- 10 -'
-100 10
3.0- 3.0-
-5
-50
3.5- 3.5
-
tA .A-- - -
-0 -0
-
4.0- An
,I ,I
.
0.0 0.5 1.0 1.5 2.0 2.5 3.0 0 2 4 6 8 10 12 14 16 18 20 22 24
Time (hr) Time (hr)
53
0.0 0.5 1.0 1.5 2.0 2.5 3.0 0 2 4 6 8 10 12 14 16 18 20 22 24
0.0 0.0 .
4 350 30
.
Event #4
0.5-11 Rainfall
-300 0.5
- Outlet discharge 25
1.0- MillI Event #4 1.0 Outlet sample
- 250
Rainfall 20 -J
1.5- - Inlet discharge '"E1.5
E Inlet sample -200'-1 E
)
- Culvert discharge CU
2.0- * Culvert sample D) 2.0 15
-
C C,,
4-
150 -C
.
2.5- L 2.5
10*
-100 - I., a --..
3.0- 3.0 0
5
3.5 -50 3.5
-
4.0- 0 4.0 0
0.0 0.5 1.0 1.5 2.0 2.5 3 .0 D 2 4 6 8 10 12 14 16 18 20 22 24
Time (hr) Time (hr)
Figure 2-1. Hyetographs and hydrographs for the inlet, outlet, and culvert of six sampled storm events at per-minute intervals. Time 0
indicates the start of rainfall. Sampling times for water quality analysis are also shown on the hydrographs. Sampling schedule and
locations are shown in Table 2-S-1 and Figure 2-S-1 in the supplementary materials respectively. Events are arranged in order of
ascending rainfall depth.
54
1.0-
[
Cnulvert
-
0.8 Outlet
-
0) - Target removal % (PUB 2014)
0.6- + Culvert / Inflow (VN)
%
E Removal
%
I-]
0.4- A EMC
0.2-
E.
K
U (D 1 0- - 10
30- T -j8
30- 6
4 E
0- AA
2
?0- to - -- t - 0
A
-
-2 -2
40 - #3 #2 #1 #5 #6 #4 -4
-
0.1 0 __
TP
0.0 8.
0.0 6.
0.04 L
0)
0.02
100- 1.0
8 0.8
6*0-- - - 0- - - - - - - - - --0.6
-
4) 4 0- A - -I [ -- 0.4E
2 0- A 0.2
10
0.0
-2
0 #3 #2 #1 #5 #6 #4
20- _
0)
6-
1
2-
U)
(U 8-
4-
00 100
80- - - - - - - - -- - - - - - - - - - - 80
0)
0)
(U
60 -6 0 Y
AA
4 0)
0 20 -2 0 2
a-
0 0
-2 20
#3 #2 #1 #5 #6 #4
Figure 2-2. Mass load (kg) at the inlet, culvert, and outlet; removal rate (%) with uncertainty
calculated using Eq. (S-2); culvert flow volume as percentage of inlet flow volume; and EMC of
each of the six sampled events for TN, TP, and TSS.
55
100- S- - Asymptotic fit
0 * Event meeting removal targets
&
80- * Event not meeting targets
< 3-month ARI events
60- S< 1-year ARI events
~40-
7 20-
0 0
5 10 15 20 25 30 35 40 45 50 55
100-
80- a 0 0
40-
20- le
,
0-
I 5 1 1i ' 2 '0 ' ' 4 I 55
0 5 10 15 20 25 30 35 40 45 50
Total rainfall (mm)
Figure 2-3. Culvert discharge volume as percentage of total inflow volume, highlighting the water quality performance of the six
sampled events (top) and categorization of events in the monitoring period into < 3-month and < 1-year ARI events (bottom) (adapted
from Wang et al.(2016)).
56
Table 2-1. Event statistics and hydrologic performance metrics of six sampled events
26/08/2013 04/08/2013 25/06/2013 09/10/2013 27/10/2013 29/09/2013
Event date and number #3) (#2) (#1) (#5) (#6) (#4)
Rainfall (mm) 3.2 8.0 10.2 29.4 33.2 40.2
Storm duration 10 min 32 min 26 min 1.7 hrs 2 hrs 5.1 hrs
10-min peak intensity (mm/hr) 16.8 22.8 38.4 86.4 44.4 69.6
ADP (days) 9.08 2.06 8.25 2.88 5.42 2.75
ARI + 3-month 3-month 3-month < 1 yr 3-month < 1 yr
No. of inlet samples 8 24 24 20 21 24
No. of outlet samples 24 24 24 24 24 24
No. of culvert samples 0 12 7 14 14 13
No. of basin samples 1 14 10 14 14 14
Inflow volume (m 3 ) 15 73 94 424 527 667
Outlet volume (M 3) 24 34 60 55 60 76
Culvert volume (m 3 ) 1 20 34 404 502 543
Water balance (%) t 63 -26 1 8 7 -7
Culvert/inflow (%) 5 28 37 95 95 81
Outlet/inflow,fv24 (m 3 / M 3 ) X 1.63 0.74 1.01 1.08 1.07 0.93
Rpeak(Ls- 1/Ls-) - 0.10 0.04 0.05 0.03 0.04 0.11
Rdelay(hrs/hrs) ' 5.8 5.6 2.2 4.2 0.9 1.0
Ponding duration (hrs) 0.8 10.0 5.8 9.9 10.4 13.4
TBased on Intensity-Duration-Frequency (IDF) curve from PUB (2011); tWater balance (%)= (Total outflow - Total inflow)/Total inflow; 'Hydrologic
performance metrics:fy,24 Rpeak, Rdejay (Davis 2008).
57
Table 2-2. Concentration (mg/L) of all samples collected during all the six events at the inlet, culvert, and outlet; event-specific
removal rates (%); and influent EMCs (mg/L) for each of 15 water quality parameters
Concentration (mg/L)
Removal rate (%)
Inlet Culvert Outlet
Mean SD Mean SD Mean SD #3 #2 #1 #5 #6 #4 Mean SD
TN 2.45 1.93 1.76 1.42 0.61 0.48 68 58 35 10 -24 4 25 35
TKN 1.25 1.42 0.60 0.76 0.36 0.37 72 63 61 58 -46 12 37 46
ON 0.97 1.03 0.36 0.37 0.19 0.18 75 80 74 69 6 12 53 34
NH 3-N 0.48 0.62 0.24 0.43 0.17 0.28 61 -23 36 96 -120 12 10 76
N0 3-N 0.81 0.58 0.78 0.53 0.05 0.16 96 82 -55 -14 -31 -22 9 63
NO2 -N 0.08 0.12 0.04 0.08 0.05 0.11 98 69 -118 -234 -129 -25 -56 128
TP 0.20 0.17 0.11 0.04 0.04 0.01 73 75 69 33 -7 32 46 32
TDP 0.08 0.04 0.05 0.02 0.02 0.01 55 58 59 4 4 66 41 29
P0 4-P 0.06 0.04 0.04 0.02 0.01 0.01 66 54 60 10 4 75 45 30
TOP 0.04 0.05 0.02 0.01 0.01 0.01 74 74 64 -13 -15 2 31 44
TDOP 0.01 0.01 0.01 0.00 0.00 0.00 35 73 53 -65 -22 56 22 54
TPIP 0.10 0.13 0.05 0.02 0.01 0.01 73 83 77 51 -25 -7 42 47
TPOP 0.03 0.04 0.02 0.01 0.01 0.00 82 74 67 5 2 -70 27 59
TSS 45 64 23 12 2 2 92 91 86 31 -12 29 53 43
COD 57 74 32 20 24 15 55 60 56 22 -94 -8 15 60
t Total number of samples used is indicated in Table 2-S-4
58
Table 2-2. (continued) Concentration (mg/L) of all samples collected during all the six events at the inlet, culvert, and outlet; event-
specific removal rates (%); and influent EMCs (mg/L) for each of 15 water quality parameters
Influent EMC (mg/L)
#3 #2 #1 #5 #6 #4 Mean SD
TN 4.02 0.58 5.09 1.24 1.43 0.81 2.20 1.88
TKN 3.07 0.39 4.13 0.36 0.18 0.30 1.41 1.74
ON 2.41 0.14 2.74 0.46 0.15 0.26 1.03 1.21
NH 3-N 0.66 0.26 1.39 0.10 0.05 0.04 0.42 0.53
N0 3-N 0.46 0.00 0.54 0.77 1.01 0.31 0.52 0.35
N0 2-N 0.36 0.00 0.14 0.00 0.01 0.00 0.09 0.14
TP 0.31 0.04 0.40 0.19 0.11 0.11 0.19 0.14
TDP 0.08 0.02 0.16 0.05 0.06 0.07 0.07 0.05
P0 4-P 0.04 0.02 0.12 0.05 0.05 0.05 0.05 0.03
TOP 0.13 0.01 0.08 0.02 0.01 0.02 0.05 0.05
TDOP 0.02 0.00 0.02 0.00 0.00 0.01 0.01 0.01
TPIP 0.14 0.01 0.20 0.11 0.04 0.04 0.09 0.07
TPOP 0.11 0.00 0.06 0.02 0.01 0.01 0.03 0.04
TSS 65 3 64 51 26 30 40 24
COD 104 20 80 59 22 38 54 34
59
Table 2-3. Presence of first flush (FF) in the six events based on three criteria of different level of stringency'
Geiger (1987) Deletic (1998) Bertrand-Krajewski (1998)
Criterion: above bisector line Criterion: 40% mass in 20% volume Criterion: 80% mass in 30% volume
#3 #2 #1 #5 #6 #4 #3 #2 #1 #5 #6 #4 #3 #2 #1 #5 #6 #4
TN
TKN
NH3-N
N03-N
N02-N
ON
TP
TDP
P0 4 -P
TOP
TDOP
TPIP
TPOP
TSS
COD
Shaded cells indicate the presence of FF and unshaded ones indicate the absence
60
Table 2-4. Comparison of pollutant removal rates in this study to literature values for bioretention basins
Saturated
Koppen climate classi fication Location Media TN TP TSS
Izone present?
This study Tropical Af Singapore Sandy loam, fine sand, wood Yes 25% 46% 53%
chip and hard rocks.
Ong et al. (2012) Tropical Af Singapore Sandy loam, fine sand, wood Yes 64% 51% 73%
chip and hard rocks.
Brown and Hunt (2011) Temperate Cfa USA, North Carolina Sandy clay loam Yes 88% 85% 95%
Brown and Hunt (2011) Temperate Cfa USA, North Carolina Sand Yes 99% 99% 99%
Brown and Hunt (2011) Temperate Cfa USA, North Carolina Loamy sand (0.6 in) No 12% 5.3% 71%
Brown and Hunt (2011) Temperate Cfa USA, North Carolina Loamy sand (0.9 m) No 13% 44% 84%
Passeport et al. (2009) Temperate Cfa USA, North Carolina Expanded slate fines, sand, Yes 56% 58%
organic matter (South cell)
Passeport et al. (2009) Temperate Cfa USA, North Carolina Expanded slate fines, sand, Yes 47% 63%
organic matter (North cell)
Li and Davis (2009) Temperate Cfa USA, Maryland Sandy loam (College Park cell) No -3% -36% 96%
Li and Davis (2009) Temperate Cfa USA, Maryland Sandy clay loam (Silver Spring No 97% 100% 99%
cell)
DeBusk and Wynn (2011) Temperate Cfa USA, Virginia Medium sand, clay, silt fines, Yes 99% 99% 99%
leaf compost
Dietz and Clausen (2005) Temperate Cfa USA, Connecticut Loamy sand, shredded No 32% -111%
hardwood bark mulch.
Dietz and Clausen (2006) Temperate Cfa USA, Connecticut Loamy sand, shredded Yes 69% -98%
hardwood bark mulch.
Trowsdale and Simcock Temperate Cfb New Zealand Topsoil (pumice sand, No 90%
(2011) horticultural soils), subsoil (clay
and weathered limestone)
Hatt et al. (2009) Temperate Ctb Australia, Victoria Cell 1: sandy loam, Cell 2: No -7 72% -398 + 559% 76 25%
sandy loam, vermiculite, perlite,
Cell 3: sandy loam, compost,
hardwood mulch.
Hatt et al. (2009) Temperate Cfa Australia, Queensland Sandy loam No 37 21% 86 3% 93 4%
Mangangka et al. (2015) Temperate Cfa Australia, South East No 48 31%a 75 14% a 81 a 8% a
Queensland 39 3 3%b 36 37%h 62 25%'
Houdeshel et al. (2015) Semi-arid BSk/Dfa USA, Utah Top soil (sand, silt, clay) No 59% 55%
/
Continental (Wetland cell)

Houdeshel et al. (2015) Semi-arid BSk/Dfa USA, Utah Top soil (sand, silt, clay) No 22% 60%
/
Continental (Upland cell)

Geheniau et al. (2015) Cold (continental) Dfb Canada, Quebec Coarse sand (sand, silt, clay) Yes -65.3%c 74.5%'
Khan et al. (2012) Cold (continental) Dwb Canada, Alberta Sandy loam No 92% 95% 99%
a Long dry period. bShort dry period. C Only changes in concentration are reported
61
2.8 Supplementary materials
S-1. Site description
The field monitoring program was carried out at the Balam Estate Rain Garden (Figure 2-S-1) in
2
the tropical setting of Singapore. With an effective treatment area of 240 m , this basin receives
runoff from a 16,800 m2 residential catchment (about 88% surface imperviousness) with a time
of concentration of about 10 min (Wang et al. 2009). The basin was designed to capture 3-month
ARI events following Australian guidelines (Facility for Advancing Water Biofiltration 2009).
The current surface storage volume (33 m3 ) corresponds to a WQD of about 2 mm. More than 14
native tropical plants such as Umbrella sedge, Alligator flag, and Cattail have been planted with
a planting density of more than 8 plants per square meter (Wang et al. 2009). Plants were
selected based on their extensive root systems and ability to withstand intermittent periods of
flooding. Field measurement using a double-ring infiltrometer indicated an infiltration rate of 4.4
cm/hr (Wang et al. 2016).
Catchment runoff converges at the inlet and is then distributed across two basin cells which are
hydraulically interconnected by ten pipes such that the whole basin operates as a single unit
(Figure 2-S-1). Runoff spread on the basin floor infiltrates through four soil layers consisting of
(from surface to bottom): a 40-cm sandy-loam filter layer; a 10-cm fine-sand transition layer; a
30-cm saturated anoxic zone of hard rocks (average size of 50 mm and 65% by volume) and
wood chips (average size of 5 mm and 35% by volume); and a 15-cm drainage layer of fine
gravel (average size of 2-5 mm) (Ong et al. 2012; Wang et al. 2009). Treated runoff is collected
by underdrains that drain to an outlet pipe that discharges to an adjacent drainageway (Pelton
Canal). As the underdrains are at a lower elevation than the outlet, a 35-cm-deep saturated
62
anoxic zone favoring denitrification is created. The basin also has a 1.1-m-high-by-1.2-m-wide
outflow culvert with a 13.7-cm high overflow weir that limits the amount of surface ponding in
the basin. Surface ponding decreases over time due to infiltration but some ponding persists for
12.5 : 5.5 hours after a storm event. The entire soil filter media is underlain by a high-density
polyethylene liner to prevent exfiltration into the surrounding soil.
63
POND OUTLET PIPE RAIN GAUGE
(Depth measurement) (Subsurface treated water)
A Velocity and level sensor

} U Sampling location
M Surface flow direction
Connecting pipes
INLET (09 m x 0.6 mcilvert)
Figure 2-S- 1. Plan view of Balam Estate Rain Garden with on-site instrumentation.
64
S-2. Sampling strategy
Table 2-S-1 lists the sampling scheme adopted in this study. Table 2-S-2 shows a comparison of different sampling strategies in the
literature. Table 2-S-3 lists the test method and detection limit of the 12 water quality parameters that were chemically analyzed. Table
2-S-4 shows the number of water samples that were tested for the concentration of each of the 15 water quality parameters in each
event at four locations (inlet, outlet, culvert, and basin floor).
Table 2-S- 1. Sampling scheme
Inlet Outlet Culvert Basin

No. of Interval Duration No. of Interval Duration No. of Interval Duration No. of Interval Duration
bottles bottles bottles bottles
6 2 min 12 min 8 30 min 4 hrs 4 5 min 20 min 3 5 min 15 min
12 4 min 48 min 12 60 min 12 hrs 6 10 min 60 min 3 10 min 30 min
6 10 min 60 min 4 120 min 8 hrs 4 30 min 2 hrs 8 60 min 8 hrs
24 Total 2 hrs 24 Total 24 hrs 14 Total 3h20m 14 Total 8h45m
65
Table 2-S- 2. Comparison of sampling strategies from literature
Event-wiseSytei
n monitoring? Composite sample? Synthetic Note
mmrmrunoff? Nt
Brown & Hunt (2011) 18 Yes yes, flow-
proportional
Passeport et al. (2009) 18 Yes yes, flow-weighted
Passeport et al. (2009) 16 Yes yes, flow-weighted
Li and Davis (2009) 12 Yes yes, flow-weighted
Li and Davis (2009) 9 Yes yes, flow-weighted
DeBusk and Wynn single influent and effluent

(2011) 28 Yes yes, flow-weighted sample for each event. 8
months monitoring.
Trowsdale and Simcock culvert is not sampled and
(2011) 12 Yes no, discrete samples assumed to have the same
water quality as inlet
complete pollutograph is
available for 3 events (each 7-
Hatt et al. (2009) 7 Yes yes, composite for 16 discrete samples for
most events effluent). 6 months
monitoring.
Hatt et al. (2009) 4 yes yes, flow-weighted yes
number of samples per event
Mangangka et al. 12 yes no, discrete samples and the type of sampling
(2015) strategy (flow-weighted or
time-based) is not reported.
Houdeshel et al. (2015) 49 yes no, discrete samples yes 1-year monitoring
Khan et al. (2012) 24 yes yes, flow-weighted yes 3 flow-weighted composite
samples per event
Dietz and Clausen 47 no yes, weekly
(2005) composite
Dietz and Clausen 72 no yes, monthly
(2006) composite
Geheniau et al. (2015) 17 no grab samples
66
Table 2-S- 3. Water quality parameters (APHA et al.(2009))
Standard Test Detection Limit

Test Parameter
Method (mg/L)
Total Nitrogen, TN 4500-N (C) 0.01
4
Total Kjeldahl Nitrogen, TKN 50 0 -Norg 0.23
Ammonia, NH 3-N 4500- NH 3 (H) 0.003
Nitrate, N0 3-N 4500- NO 3 (I) 0.003
Nitrite, N0 2-N 4500- NO 2 (I) 0.003
Total Phosphorus, TP 4500- P (H) 0.003
Total Dissolved Phosphorus, TDP 4500- P (H) 0.003
Inorganic Phosphorus, P0 4-P 4500- P (G) 0.01
Total Organic Phosphorus, TOP 4500- P (H) 0.003
Total Dissolved Organic Phosphorus, TDOP 4500- P (H) 0.003
Total Suspended Solids, TSS 2540D 2
Chemical Oxygen Demand, COD 5220B 5
67
Table 2-S- 4. Number of water samples analyzed for each parameter in each event
Event TN TKN NH
-N 3 NO 3 No TP TDP P0 4 -P TOP TDOP TSS COD
-N
Inlet 8 8 8 8 8 8 8
Outlet 24 24 24 24 24 24 24
#3
Culvert 0 0 0 0 0 0 0
Basin 1 1 1 1 1 1 1
Inlet 24 24 24 24 24 24 24
Outlet 24 24 24 24 24 24 24
#2
Culvert 12 12 12 12 12 12 12
Basin 14 14 14 14 14 14 14
Inlet 24 24 24 24 24 24 24
Outlet 24 24 24 24 24 24 24
#1
Culvert 7 7 7 7 7 7 7
Basin 10 10 10 10 10 10 10
Inlet 20 11 20 11 11 20 20
Outlet 24 24 13 13 13 13 13
#5
Culvert 8 8 8 8 14 14 8
Basin 8 8 8 8 14 8 8
Inlet 21 11 21 11 11 21 21
Outlet 24 24 13 13 13 13 13
#6
Culvert 8 8 8 8 14 14 8
Basin 8 8 8 8 14 8 8
Inlet 24 24 24 24 24 24 24
Outlet 24 24 24 24 24 24 24
#4
Culvert 13 13 13 13 13 13 13
Basin 14 14 14 14 14 14 14
Inlet 121 121 102 121 102 10

2
121 121 121 121 121 121
All Outlet All 144 144 144 122 122 12 122 122 122 122 122 133
2
Culvert 48 48 48 48 48 60 60 60 60 48 60 60
Basin 55 67 55 55 55 67 55 55 55 55 67 67
68
S-3. Calculation of ON, TPIP, TPOP
Three additional parameters (organic nitrogen, ON; total particulate inorganic phosphorus, TPIP;
and total particulate organic phosphorus, TPOP) were derived from their relationships with the
measured parameters according to
CON - CTKN - CNH,-N
CTPIP = CTP - CTOP - PO-PS-)
CTPOP = CTOP - CTDOP
where CON, CTKN, CNH3.- N, CTPIP, CTP, CTOP, CPO4 -P, CTPOP, and CTDOP are the concentrations of
organic nitrogen, total Kjeldahl nitrogen, ammonia, total particulate inorganic phosphorus, total
phosphorus, total organic phosphorus, inorganic phosphate, total particulate organic phosphorus,
and total dissolved organic phosphorus respectively.
S-4. Uncertainty in the removal rate calculation
Uncertainty in the removal rate determination is calculated based on propagation of the errors in
quantifying loading and mass. The overall uncertainty in the removal rate arises from two
sources: 1) that associated with concentration measurement using chemical analyses and 2) that
associated with flow rate measurement using area-velocity sensors. Uncertainty in chemical
analyses encompasses gross error and instrumental error which are reflected in the standard
deviations of the duplicate recovery rate (uD, range 0.2 - 6.4%) and standard curve recovery rate
( as , range 0 - 8.5%) respectively across six sampled events. Uncertainty in flow rate
measurement, c-F, is taken as the accuracy of the instrument (ISCO-2150 Area Velocity Module),
which is 2% of readings from 1.5 m/s to 6.1 m/s as provided by the manufacturer. The
propagation of removal rate error for each water quality parameter is calculated using Eq. (S-2):
69
2
SRemoval =
Qculvert
inletQiniet)
2 2
C r
((inlet
Qoutlet
inlet
2 2
C
+(CculvertQcuivert + CoutietQoutlet 2 .2
\Cinlet 2Qinlet Cne
('CinletQinlet)
Cculvert -2 2
Qcuivert
(CculvertQculvert
2
+ CoutletQoutlet
Qinlet Cinlet
) 2
2
Qinet
+( Coutlet) 2 2
\CinletQinet) outet
(S-2)
0 0
7removal rate error = 2Removal
where o2 Removal is the variance of removal rate; Qculvert, Qoutlet, Qinlet (L/s) are the mean flow
rates, Ccuivert, Coutlet, Cinlet (mg/L) are the mean parameter concentrations for each event;
0Cculert 02 Coutet , 2 Cinet (mg/L)2 are the variances of concentration measurements; and
a Qcuivert, a Qoutlet a 2 Qiniet (L/s)2 are the variances of flow measurements for each event.
70
S-5. Pollutographs of TSS, TN, and TP for the six sampled events
350- 8 350 --0 450 350 *0 -1.6 -450

Event #3 Inlet cscharg (Di) Event # 2 - Ie dacharge (LI) Event #1 -Inlet idiarge (/s)
300- TN (m -7 .7 400 300 - TN (ngfL) -7 1.4 -400
-U-TP (mgL) 300- a- TP (rngfL)
-- TP (mg/L)
25-0- TSS (M9&L) a- TSS (mAg/L) 350 * * -U- TSS (mg/I) -1, 350
250 . 6 6 12
.
250- 250 <
300 U Se .300
a * ** Em"* .5 1,0
200 - o 200 250 200 250
.4 U 4 0.8
150- 200
150 150 200
3
8% .3 150 -3 -0.6
100- -150
100 100
52 .2 100 ,2 -0.4
100
50- 0 aaa.... U a a 0 aa
J'i!m au A.Ra
a so 11MUSSE'R a a a a a a 1 -0.2 -50
I I
0, 0 0
0- .. +o 0 0 -00 -0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0
Time (hr) Time (hr) Time (hr)
350 350 a -450 350 1.6 -450
.
Event #3 -Culert decherg (Ls) Event #2 -Culert d arge (1.1s) Event #1 -cuivertdchlarg (Lis)
-
TN(motL
l7 300- T 300- -- TN (m"Ig) .7 -400 TN (mgIL) 1.4 -400
---- TP (11V -eTP (MOIL) -a- TP (mg/IL)
H. I
)
TW(eg) TSS (mgL) -350 -a- TSS (mgrI)

250 -. S TW 12 - 350
250- 86
.
250
.5 -300
a a -1.0
300
200- 200-
-250 200 - -250
4 4
150- 150- 150
-200 200
-
3 3 -0.6
100- -150 150
100- 10
2 - 0.4
2
.
-100 -100
501
50
so U aUU a a 50 aa a -0.2 0
0 0 0 0 0 .0 -0.0 0
.
0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0
dF Ir (LIS)
-450
Event #3 -Oult Event #2 discharge (Li, .6 450
+TN (ffg) -7
-Ouet
TN (mg/L) Event #1 -- Outet dghar (L)
I
7 -400
-uTP (MgOIL) .1.4 400
-- TSS (mg&L) -350 --a-TO (mA/) ,S7m1L
)6 -350
-1.2
-300
5 5 -300
64 -1.0
250
4 4 250
200
200
3 3
150 Em -34 -0.6
150
2 2
-100 -0.4
m a6 ma. a 2 100
-50 -02
50
Waft I 1
a aa~ag a us .. 0 -0,0
0 2 4 8 8101214 6 8 2 2240 0 2 4 6 8 10 12 14 16 18 20 22 24 0 2 4 6 8 10 12 14 16 18 20 22 24
Time (Nr) Time (hr) Time (hr)
71
350 - a -1.6 -45 -450 350 8 -1.6 -450
1 0 Event #5 -InletdlechIrg (L/) Event #6 Inletdo () Event #4 -Ilet dsdhmw ().
0
-
TN (rnWL) . 7 -1.4 -40 -- TN (nm-) 1.4 400 300 ) .1.4 -400
--
300 30 --- N(m
,
300-
.
-a- TP (mg.) -u-TP (mg.)) -- TP (mL)
----TSS (Mig&.)
-1.2
-35 0 -- Tes (not-) -350 - MIS (ntgo 6 -1.2
-350
-6 -1.2
250. 25D . -- 250-
30 -300 -300
.5 1.0 0 -5 -1.0 .5 1.0
*
200 20 250 200- 250
.
4 -0.8 ,4 -0.4 4 0.8

150- 2C 0 150 200 150- me -200
-3 0.6 S3 %
3 0.6
11 0 U 150 150
100 100 100-
-
U -2 0.4 -2 0.4
10 0 * .2
100 -100
0.2 50 mm a -0.2 50- so . 1 0.2

Sc0 I * mUEEU EDE 1 50 -50
so-
50 , o 0 U'UaI . I 1i*
0.0 -0.0 0 0.0 40

0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0
Time (hr) ime (ha) Time (hr)
3 I -450 -450 35 8
I-Ie Event 05 CU vrt decharg (US)
30 Is1 1.6 -450
Event 08 d-hrg (L) E ven t # 4 - C ~ -ta r
I-TN(n 1 .7 -400 -C8
TL
ftLJ -1.4
"
300. T(mwl.)T 450 -- TN (nvgf .7 1.4 -40D
300- --- 7 -1.4 300-
-TP(MO6 +TP (mgL)
-350 Tag~~
(UI -350 -- TSS (nvIL -350
250- -1.2 -6 -1.2
-300 2 -300 - 300
-5 -1.0 .5 1.0 .5 -1.0
.4 0.8
250 2 -- 250 200- 250
-4 -0.8 4 -0.8
150. 200 ISO. 200 150. - 200
.3 0.6 -3 -0.6 3 -0.6
150 s 150 -150
10
.2 0.4 -2 -0.4 * -2 -0.4
-100 1500 1so- -100
50
-1 0.2 10 -0 050.- 1 -0.2 50 -0.2
* * * *-* 1 .50
0 e - ... I. . ...
-0
0 0.0 0. -0.0
.
.0
"
-0.0
-
0. 0.5 1.0 1.5 2.0 2.5 3.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0
Tkneso Time (hr) Time (hr)
Event #5 -F thg() I 1.6 *450

. Event 0 - OIt dShr(
S1.6 -450
Event 04 --- OU h
450
(
-U-T (mWL) 400 --- TN (mlOL) 7 400
1.4 -- TN (MgFL) .7y -1.4 -400 -1.4
I-u--P(mR1 TP (nOL)
)
+ TP (mg L
L-- TS (mg STSS (M9&)
J 1.2
-350 ---- TSS (gl
-1.2
350 6 -1.2
-350
20 .300 2C - 300 '0 300

-
1.0 -5 -1.0 5
250 250 250
0.8 0.8 4
200 200 200
0.6 -3 0.6 3 -0.6
150
150 150
* t
1
.2 0.4
0.2
0.0
.100
-0
2
0
0.4
0.2
0.0
100
.50
.0
e -
2 -0.4
-0.2
-0.0
50
0
0 2 4 6 8 10 12 14 16 18 20 22 24 0 2 4 6 8 10 12 14 16 18 20 22 24 0 2 4 i 8 10 12 14 16 18 20 22 24
Tane (hr) ime (hr) Time (hr)
Figure 2-S- 2. Pollutographs of TN, TP, and TSS for each of the six sampled events.
72
S-6. Mass load of 15 water quality parameters
Table 2-S-5 shows the mass load of 15 water quality parameters at inlet, culvert, and outlet.
Table 2-S- 5. Mass load (g) of 15 water quality parameters at inlet, culvert, and outlet.
Inlet
#3 #2 #1 #5 #6 #4 Mean SD
TN 43.23 101.57 464.04 459.11 667.14 461.35 366.07 241.73
TKN 32.97 58.77 376.87 132.16 81.83 172.21 142.47 125.38
ON 25.89 49.17 249.94 171.94 69.63 150.61 119.53 85.94
NH 3-N 7.08 9.60 126.92 38.19 24.81 21.60 38.03 44.97
N0 3-N 4.94 29.30 49.61 286.18 470.57 177.19 169.63 181.89
N0 2-N 3.87 2.82 12.82 0.82 5.71 1.78 4.64 4.35
TP 3.35 12.64 36.20 68.93 49.31 63.43 38.98 26.73
TDP 0.86 3.62 15.03 18.03 27.95 37.82 17.22 14.12
PO 4 -P 0.44 2.53 10.53 16.82 23.68 29.40 13.90 11.54
TOP 1.39 2.64 7.31 9.17 5.71 10.80 6.17 3.67
TDOP 0.25 0.79 1.72 1.67 1.79 6.15 2.06 2.09
TPIP 1.52 7.47 18.35 42.28 18.64 23.24 18.58 14.13
TPOP 1.14 1.85 5.59 7.50 3.91 4.66 4.11 2.36
TSS 701.82 3760.15 5798.28 18848.62 12242.24 17237.34 9764.74 7462.16
COD 1113.79 2721.95 7329.68 21959.25 10030.91 21393.40 10758.16 9038.26
73
Table 2-S-5 Continued.
Culvert
#3 #2 #1 #5 #6 #4 Mean SD
TN 0.00 22.01 166.52 390.91 806.86 419.17 300.91 304.78
TKN 0.00 7.68 76.21 48.23 112.20 133.40 62.95 54.40
ON 0.00 4.84 34.07 47.62 58.92 115.09 43.43 42.09
NH 3-N 0.00 2.83 42.13 0.61 53.27 18.30 19.52 23.11
N0 3-N 0.00 5.21 46.53 322.86 615.01 216.92 201.09 240.46
NO 2-N 0.00 0.81 6.29 0.61 9.46 2.07 3.21 3.81
TP 0.00 1.83 6.28 44.15 50.52 40.62 23.90 23.52
TDP 0.00 0.79 3.18 16.84 25.29 11.67 9.63 10.14
PO 4-P 0.00 0.59 2.27 14.68 21.55 6.92 7.67 8.72
TOP 0.00 0.44 1.50 9.93 6.15 9.70 4.62 4.58
TDOP 0.00 0.09 0.28 2.62 2.01 2.12 1.19 1.19
TPIP 0.00 0.81 2.50 19.54 22.81 24.00 11.61 11.63
TPOP 0.00 0.35 1.22 6.80 3.63 7.58 3.26 3.30
TSS 0.00 245.72 605.03 12879.79 13626.92 11989.16 6557.77 6895.25
COD 0.00 378.52 1015.23 15639.88 18338.23 20276.02 9274.65 9768.04
Outlet
#3 #2 #1 #5 #6 #4 Mean SD
TN 13.80 20.26 133.39 21.99 22.58 24.09 39.35 46.21
TKN 9.17 13.85 70.99 7.05 7.65 18.02 21.12 24.78
ON 6.45 4.87 31.89 6.17 6.37 17.34 12.18 10.69
NH 3-N 2.73 8.99 39.10 0.89 1.28 0.68 8.94 15.10
N0 3-N 0.22 0.06 30.25 3.74 0.33 0.13 5.79 12.07
N0 2-N 0.08 0.05 21.61 2.12 3.62 0.15 4.60 8.45
TP 0.92 1.30 4.79 1.89 2.14 2.23 2.21 1.36
TDP 0.38 0.75 2.92 0.39 1.43 1.06 1.16 0.95
P0 4-P 0.15 0.58 1.94 0.40 1.19 0.43 0.78 0.67
TOP 0.36 0.26 1.17 0.45 0.40 0.89 0.59 0.36
TDOP 0.16 0.12 0.52 0.12 0.18 0.55 0.28 0.20
TPIP 0.41 0.46 1.68 1.04 0.55 0.92 0.84 0.48
TPOP 0.20 0.13 0.65 0.33 0.22 0.33 0.31 0.18
TSS 55.86 103.81 227.91 150.32 60.56 172.01 128.41 67.46
COD 501.32 699.54 2211.16 1553.36 1136.09 2807.69 1484.86 893.06
74
S-7. Comparison of EMC of this study to literature values
Table 2-S-6 shows that EMC values from this study are comparable to those from another tropical study (Chua et al. 2009). Compared
to world-wide data, EMCs are generally lower in tropical Singapore catchments than in temperate catchments (Brezonik and
Stadelmann 2002; Duncan 1999; EPA 1983; Francey et al. 2010; McLeod et al. 2006; Smullen et al. 1999; Taylor et al. 2005).
Table 2-S- 6. Comparison of EMCs in this study to literature values
Location Catchment % Imp TN TKN TP TDP TSS COD

Tropical climates
This study Singapore High density residential 88 2.20 1.41 0.19 0.07 40 54
Chua et al. (2009) Singapore Residential 68 1.41 0.08 0.03 66
Temperate climates
Duncan (1999) Worldwide Urban 2.60 0.40 155 78
Taylor et al. (2005) Australia Urban 2.13 1.39
Francey et al. (2010) Australia High density residential 74 0.35 102
McLeod et al. (2006) Canada Residential 0.53 190 100
Smullen et al. (1999) USA Urban 2.51 1.67 0.34 0.10 174 66
Brezonik and Stadelmann (2002) Minnesota, USA Urban 3.08 0.58 0.20 184
75
S-8. First flush M(V) curves for TSS, TN, and TP
1.0-
0.8-
E 0.6-
= 0.4- -- Bisector
E ---
-
Event #1
-d- Event #2
-U- Event #3
0.2- ' -Event #4
-
--- Event #5
Event #6
0.0-
0.0 0.2 0.4 0.6 0.8 1.0
Cumulative runoff volume
1.0- 1.0-
TN
0.8- 0.8-
E 0.6- E 0.6-
U)
CU
7 0.4- - - - Bisector 0.4- - - - Bisector
E -e- Event #1 E -4- Event #1
AEvent #2 -- Event #2
-- w- Event #3 1--w- Event #31
0.2 -6-Event #4 0.2- -- Event #4
-
-
-
-U- Event #5 -c- Event #5

-*- Event #6 --X- Event #6
0.0- U.U P..

0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
Cumulative runoff volume Cumulative runoff volume
Figure 2-S- 3. Plots of cumulative mass against cumulative runoff volume indicating the
magnitude of first flush for TSS, TN, and TP.
76
S-9. Rainfall characterisitcs in this study
-100
-
-90
50
-80
-
40 -70
-
-
-60
30 -50 rL
-
-40
20
- -30 3E
-20
-10
1 0
0 10 20 30 40 50 60 70 80 90 100
Event based rainfall amount (mm)
100
35 .90
-
30- -80
-70
25-
60
20 50
-
15- 40
-30
10-
20
5 10
-
9 10 11 12 13
Event duration (hrs)
1AR 1i 00
.9
40- 0
.8 0
35.
-7 0
30.
60
25-
50
20-
40
15- 30
10- 20
5. 1i 0
0-
) 1 2 3 4 5 6 7 8 9 10 11
ADP (days)
Figure 2-S- 4. Histograms and cumulative probability curves of event-based rainfall amount
(mm), event duration (hr), and ADP (days) for the 96 events observed during Apr-Nov 2013.
77
S-10. Frequent heavy flushing of tropical catchment
Compared to temperate climates such as the United States, the tropical climate in Singapore
produces storms that are more frequent and have higher intensity and shorter duration. Table 2-
S-7 compares the mean statistics of the 96 events observed during this study to similar statistics
by Driscoll et al. (1989) for a selection from fifteen regions in the U.S. Table 2-S-7 shows a wide
range of rainfall patterns in the U.S., from the dry West Inland region to the wet East Gulf region.
The equivalent mean rainfall intensity observed in Singapore during this study is 66% higher
than that of East Gulf, the region in the U.S. identified by Driscoll et al. (1989) to have the
highest rainfall intensity. The mean event duration in Singapore is 56% shorter than that of East
Gulf, the region in the U.S. with the shortest mean event duration. The mean time interval
between event midpoints (delta) in Singapore is 61% shorter than in the Pacific Northwest, the
region in the U.S. with the shortest mean delta.
78
Table 2-S- 7. Storm event statistics in Singapore (this study) compared to various
regions in the U.S. (Driscoll et al. (1989))

Mean Event Rainfall Mean Event Mean Delta*
Intensity (mm/hr) Duration (hr) (day)
Singapore (this study) 7.49 2.8 2.00
Mid Atlantic (U.S.) 2.34 10.1 5.96
East Gulf (U.S.) 4.52 6.4 5.42
North Central (U.S.) 2.21 9.5 6.96
West Inland (U.S.) 1.40 9.4 32.8
Pacific Northwest (U.S.) 0.89 15.9 5.13
Pacific Southwest (U.S.) 1.37 11.6 19.8
*Delta is defined as the time interval between event midpoints by Driscoll et al.(1989)
Table 2-S- 8. ARI (years) if the six sampled events were to occur in Melbourne, Australia or
various cities in the U.S
08/26/13 08/04/13 06/25/13 10/09/13 10/27/13 09/29/13

(#3) (#2) (#1) (#5) (#6) (#4)
Singapore (this study) 3-month 3-month 3-month < 1 3-month <1
Melbourne <1 < 1 1-2 10-20 1-2 5-10
Baltimore, MD (Mid Atlantic) <1 <1 < 1 1-2 <1 <1
Miami, FL (East Gulf) <1 <1 <1 < 1 < 1 <1
Detroit, MI (North Central) <1 <1 < 1 2-5 <1 < 1
Reno, NV (West Inland) <2 1-2 5-10 > 100 10-25 50-100
Seattle, WA (Pacific Northwest) < 0.5 0.5-2 5-10 > 100 10-20 > 100
Los Angeles, CA (Pacific Southwest) <1 < 1 1-2 25-50 2-5 10-25
Source of Intensity-Duration-Frequency (IDF) curves: (Public Utilities Board of Singapore 2011); (NOAA 2015;
Pilgrim 1987); (City of Seattle 2014)
79
S-11. Rainfall intensity of events of various ARIs across the world
300 -__ Singapore

-El- Melbourne
-0- Baltimore, MD
'~' 250 ---- Miami, FL
.c --- Detroit, MI
E , ---X--- Reno, NV
E -- y-- Los Angeles, CA
200 - -0- Seattle, WA
Z%.
*150
-
*ia 10
A- El
100-
0-... ...... I
0.2 0.5 1 2 5 10 20 50 100

ARI (years)
Figure 2-S- 5. Rainfall intensity of events of various average return intervals (ARI) in Singapore,
Melbourne, and selected cities in the U.S. (Sources of Intensity-Duration-Frequency (IDF)
curves: (Public Utilities Board of Singapore 2011); (NOAA 2015; Pilgrim 1987); (City of Seattle
2014).
80
Chapter 3
Field evaluation of hydrological performance
3.1 Abstract
Bioretention basins or rain gardens are constructed green infrastructure in developed areas that
retain and treat stormwater runoff. This study evaluates whether a tropical basin in Singapore
that is built according to design guidelines from temperate countries has sufficient storage and
infiltration capacity to handle tropical storms. Flow measurements from 80 storm events which
varied in rainfall depth (2.6 - 71.8 mm), event duration (10 min - 2.8 hr), and antecedent dry
period (ADP) (0.25 - 9.08 days) were recorded for half a year. Hydrological performance in
terms of surface and subsurface storage for each storm event has been analyzed. Results show
that the basin storage capacity is inadequate since in about 50% of the events, more than 50% of
the runoff overflows the basin without soil treatment. In a storm event, the subsurface storage
space remains fully saturated throughout the ponding period and the infiltration rate estimated
based on the aggregate statistics of 59 events with extended surface infiltration is 1.7 cm/hr. This
is within the same order of magnitude as that found by direct field measurement using a double-
ring infiltrometer (4.4 cm/hr) but is lower than a recommended rate of >10 cm/hr. The
information drawn from this study could be of use to stormwater managers for sustainable
planning and development. For instance, this study suggests that in order to meet pollutant
removal rate targets, a tropical basin should be sized to retain a critical water quality volume
(WQV) in the range of 10 to 30 mm.
81
3.2 Introduction
BMPs (best management practices) are sustainable strategies aimed at reinstating pre-urban
development hydrological features including storage, infiltration, runoff, and groundwater
recharge through detention and treatment of polluted runoff at its source. BMPs such as
bioretention basins and permeable pavement have been shown to be effective in reducing runoff
volume (DeBusk and Wynn 2011) and removing pollutants (Glass and Bissouma 2005;
Marchioni and Becciu 2015) in temperate regions. Studies have been carried out for investigating
important field parameters such as saturated hydraulic conductivity (Ksat) for basins in tropical
and cold climates (Facility for Advancing Water Biofiltration 2009; Paus et al. 2016). Both
studies recommended a Ksat value of more than 10 cm/hr for cells so as to maintain satisfactory
performance. Another important factor that determines performance is the amount of spilled or
overflow volume. Becciu and Raimondi (2015) have used a probabilistic model to evaluate the
ability of detention basins to manage overflows.
However, there are limited hydrological studies based on continuous, high resolution, long-term
rainfall and flow records for basins in a tropical urban setting, especially for bioretention basins.
The rainfall-runoff characteristics in tropical regions such as Singapore are very different from
those in temperate places. For instance, rainfall in the tropics is often of higher intensity, having
shorter event duration and antecedent dry period. These climatic differences also suggest that
tropical basins may not have sufficient storage and infiltration capacity to handle tropical storms
if they have been designed according to temperate design guidelines. In this context, the
objective of this study is to assess the hydrological performance of a bioretention basin in the
setting of a highly urbanized tropical catchment and to attempt to recommend appropriate sizing
82
guidelines. To investigate the storage capacity, an event-based water balance is computed to
compare the fraction of runoff that is able to infiltrate and hence gets treated to the fraction of
runoff that overflows the basin without treatment. Additionally, hydrological performance in
terms of time evolution of cumulative storage, surface storage, and subsurface storage for each
storm event is analyzed statistically. Infiltration rate is also estimated from aggregate statistics as
well as from direct field measurement. The field results from this study could be used for
validating models such as RECARGA (Dussaillant et al. 2003) for bioretention basins in tropical
applications.
3.3 Methodology
3.3.1 Site description
The study site, Balam Estate Rain Garden (Figure 2-S-1), which includes 240 m 2 of effective
treatment area and 760 m 2 of landscaped area (Wang et al. 2009), is designed to receive
stormwater runoff from a highly impervious (about 88%) residential catchment (16,800 m 2 ), with
a time of concentration of about 10 min. The treatment area is partitioned into two above-ground
cells that are hydraulically interconnected by ten pipes.
The basin has one entrance (a 0.9-m-high by 0.6-m-wide inlet) and two exits (a culvert for
surface overflow and an outlet for subsurface drainage) (Figure 2-S-1). The culvert (1.1-m-high-
by-1.2-m-wide) has an overflow weir (about 13.7 cm above the basin bottom) that limits the
amount of surface ponding in the basin by discharging overflow into the neighboring drainage
way (Pelton Canal). Two subsurface outlet pits collect runoff treated by soil filtration from
underdrains (10-cm-diameter high-density polyethylene (HDPE) perforated pipes). The outlet
83
pits then discharge the filtrate into Pelton Canal through a 30-cm diameter polyvinyl chloride
outlet pipe. The soil filter media is underlain by a HDPE liner to prevent exfiltration into the
surrounding soil and maintain a 35-cm thick saturated anoxic zone that is formed by raising the
outlet to a 35-cm-higher elevation relative to the underdrains. Non-perforated goose-neck
discharge pipes (10-cm-diameter) convey the filtrate from the underdrains to the outlet pipe. The
subsurface basin treatment area consists of four-layered soil filter media (from surface to bottom):
a 40-cm sandy-loam filter layer; a 10-cm fine-sand transition layer; a 30-cm saturated zone of
hard rocks (average size of 50 mm and 65% by volume) and wood chips (average size of 5 mm
and 35% by volume); and a 15-cm drainage layer of fine gravel (average size of 2-5 mm) (Ong et
al. 2012; Wang et al. 2009).
3.3.2 Data collection
A real-time monitoring system that consisted of flow and water-level measurement equipment
was installed in the Balam Estate Rain Garden from June 2013 to November 2013. Four Doppler
ultrasonic area-velocity level sensors (ISCO-2150 Area Velocity Module with 2105 Interface,
Teledyne Isco, Lincoln, Nebraska, USA) were used to continuously monitor flow rates at the
inlet, outlet, and culvert as well as ponding water level at one-minute intervals. Figure 2-S-1
shows the placement of these four sensors at the inlet; at an arm's length into the outlet pipe;
downstream of a 10-cm grated trash trap at the culvert; and on the basin floor. Rainfall intensity
(1-min interval) was measured with a tipping bucket rain gauge (Model TB4, Hydrological
Services PTY Ltd., Liverpool, NSW, Australia) which was installed on the basin edge at 3 m
above ground. In-situ infiltration rate of the soil filter media was measured with a double-ring
84
infiltrometer (IN 10-W: 12 & 24-inches-diameter and 20-inches-height, Turf-Tec International,
Tallahassee, Florida, USA) using the constant head (10 cm) testing method.
3.3.3 Data analysis
Per-minute precipitation records from the rain gauge installed at the field site were used to
calculate the total rainfall and antecedent dry period (ADP) for each storm event. The rainfall in
a particular minute was assigned to an event in progress if it was within six hours from the
previous recorded incidence of rainfall. Intermittent light drizzles with a total rainfall amount of
less than 2.54 mm were excluded from the event count (Driscoll et al. 1989).
Equation 3-1 was used to compute the difference between the total inflow volume and the total
outflow volume, which we define as the water balance, WB. When the total inflow exactly
equals the total outflow, WB = 0%.
WB = (Z V 0 - Z Vin)/ Vi X100% Equation 3-1
where: Vout = Voutlet + Vcuivert and Vin = Vrain + Viniet , Vin and Vout are the total discharge
(m 3 ) into and out of the basin respectively, Vrain is the direct rainfall onto the basin (M3), Vinlet
(m 3 ) is the catchment runoff entering the basin, Voutiet (M 3) is the subsurface water discharge, and
Vcuvert (m 3 ) is the discharge through the overflow culvert that bypasses the basin untreated.
To compute cumulative storage, a mass balance of water into and out of the basin requires the
storage rate to be the difference between the inflow and outflow rates in Equation 3-2.
(Qrain + Qiniet) - (Qcuvert + Qoutlet) = Qponding+ Qsubsurface Equation 3-2
85
where Qponding and Qsubsurface are the surface storage rate (L/s) and subsurface storage rate (L/s)
respectively, Qrain is the direct rainfall onto the basin (L/s), Qiniet (L/s) is the catchment runoff
entering the basin, Qoutiet (L/s) is the subsurface water discharge rate, and Qcuivert (L/s) is the
discharge rate through the overflow culvert that bypasses the basin untreated. Cumulative storage
is the storage rate accumulated over time. To compute surface ponding, the basin horizontal
cross-sectional area at different elevations was computed from as-built drawings for the basin.
From these, volume-area-stage curves were computed.
3.4 Results and discussion
3.4.1 Water balance and water quality volume (WQV)
An event-based water balance was computed according to Equation 3-1. Figure 3-la is a
histogram showing the range of water balance over 80 events (-75% to 118% with a root mean
square error (RMSE) of 34%). Events with a water balance of 29% to -39% are within one
standard deviation of the mean.
Figure 3-lb shows plots of water balance as a function of rainfall depth and ADP. There is no
obvious trend in how either factor separately affects the water balance because the water balance
is affected by both rainfall depth and ADP combined. However, some explanations for the poor
water balance seen in large events can be offered. In large events, direct runoff from adjacent
landscaped area is not accounted in the measured inflow. There is also on occasion backflow of
water from the bordering Pelton Canal into the outlet and culvert, hence affecting the water
balance. This phenomenon is observed in 11 events (Figure 3-ib). Errors in field measurements
due to blockage of sensors by floating objects like fallen leaves and trash especially at the culvert
86
trash grate that is located upstream of the culvert sensor could be another source of error.
However, weekly cleaning of the basin was carried out to minimize such instances. In addition,
the length of ADP affects the water balance because it determines the degree of initial soil
moisture in the unsaturated filter media layer at the start of an event. Hence, the available
subsurface storage capacity for retention of infiltrated water varies from event to event. For
events with very short ADP, there is also a possibility that the next event arrives before the
ponded water from the prior event is able to infiltrate completely. This phenomenon is observed
in three events with ADPs of 0.42, 0.79, and 1.25 days.
Figure 3-2 plots the percentage of total inflow volume that exits the basin as surface overflow
(via the culvert) as compared to subsurface flow (via the outlet). Due to the complications in the
water balance computation, six events with poor water balance (water balance ranging from 29%
to 57%) are excluded from the computation of the culvert fraction and two events (water balance
of 93% and 118%) are excluded from the outlet fraction computation. As the outlet and culvert
are the only two exits from the basin, the two figures would add up to 100% if there was a
perfect water balance for each event.
Figure 3-2 shows that as rainfall depth increases, there is an exponential increase in the
percentage of runoff exiting through the culvert without receiving soil treatment (Figure 3-2a)
and a corresponding exponential decrease in the percentage of runoff receiving soil treatment by
exiting through the outlet (Figure 3-2b). These asymptotic trends reflect the tendency of
additional inflow to route through the culvert rather than the subsurface soil filter and outlet. This
is expected because in a large event, as the available subsurface space is filled up and the soil
87
filter media becomes saturated, a larger proportion of any additional inflow would have to exit
through the culvert instead of the outlet. The limit in the subsurface flow is also reflected in the
asymptotes seen in Figure 3-2b and Figure 3-2c in which the outlet fraction continues to
diminish but at a slower pace when rainfall depth increases beyond the 30-mm mark.
Results from Wang et al. (2017) for six events during which water quality was analyzed show
that events with a higher outlet fraction are able to meet the pollutant removal rate targets (80%
removal in total suspended solids, 45% in total nitrogen, and 45% in total phosphorus, (Public
Utilities Board of Singapore 2014)) while events with a lower outlet fraction fail to meet the
removal rate targets (Figure 3-2a-b). This is expected since additional removal processes like soil
adsorption and plant uptake occur in the subsurface for phosphorus species while biochemical
transformations like ammonification, nitrification, and denitrification are carried out by bacteria
residing within the soil filter media for nitrogen species. Hence, basin treatment is more efficient
when a higher percentage of runoff is retained, infiltrated, and exits through the outlet.
The term "water quality volume" or WQV refers to the runoff volume specified to be captured
and treated for adequate basin performance in removing pollutants. Figure 3-2 shows that in
about 50% of the events, more than 50% of the runoff overflows the basin via the culvert without
soil treatment. For events with > 20 mm rainfall depth, less than 30% of the runoff gets
infiltrated, soil-treated, and discharged via the outlet. This reflects the inadequacy of basin
storage capacity which compromises the efficiency of basin.
88
Since out of the six sampled events in the study by Wang et al. (2017), the largest event that still
exhibits satisfactory pollutant removal rate has a rainfall depth of about 10 mm while the
smallest event not meeting the pollutant removal standards has a rainfall depth of about 30 mm, a
suitable WQV will be in the range of 10 to 30 mm. The same conclusion can be drawn from
Figure 3-2c which illustrates that the range of 10 to 30 mm includes both events of higher
rainfall intensity (i.e. > 3-month annual recurrence interval (ARI) events) and events of lower
rainfall intensity (< 3-month ARI events). The range of 10-30 mm includes both mild, long-
duration events of high total rainfall depth but low rainfall intensity (orange dots in Figure 3-2c)
and intense, short-duration events of low rainfall depth but high rainfall intensity (green dots in
Figure 3-2c). Therefore, this range span points to where the critical WQV lies but exactly where
it is within this range depends on the characteristics of the rainfall event. Hence, a basin designer
could be flexible in deciding the appropriate WQV on a case-by-case basis, with 30 mm being
the more conservative estimate and 10 mm being the more stringent estimate. Moreover, Wang
et al. (2017) estimated that in order to capture 90% of the total runoff in the six months of the
field monitoring period, a WQV of 32 mm would be required. This amount corresponds to the
conservative estimate of 30 mm in Figure 3-2 which marks the beginning of the asymptotes.
3.4.2 Surface and subsurface storage
Equation 3-2 is used to compute the storage rate which is the difference between the inflow and
outflow rates. To partition the amount of water stored between the surface (ponded water) and
subsurface (within soil media) compartments over time, the storage rate is cumulated over time
to give rise to the cumulative storage term. The surface storage term is found from the pond
storage-stage relationship. The subsurface storage is found by taking the difference between the
89
cumulative storage and surface storage. Figure 3-3 illustrates the time evolution of the
cumulative storage, surface storage and subsurface storage terms during three events with
varying degrees of rainfall depth: a small (Event #63, 2.6 mm), a medium (Event #61, 8.2 mm),
and a large (Event #47, 29.4 mm) event.
Figure 3-3 is analyzed more closely by segregating the time series into three defined stages, A, B
and C, which are the three phases in basin drainage. In Stage A, both cumulative storage and
surface storage increase sharply due to the high inflow of catchment runoff (inlet) at the start of
the event. Once the pond water level exceeds the extended detention depth of 13.7 cm and
surcharges the basin, water overflows and exits the basin through the culvert. High culvert
overflow leads to quick recession in cumulative storage and surface storage. Thus, this sharp
initial drop reflects the fast drainage from the basin as the culvert is actively channeling away
surcharged water. This phenomenon is only seen in events that are sufficiently large to cause a
culvert overflow (Figure 3-3b and c). In small events where there is no culvert overflow, inflow
is retained entirely within the basin and eventually infiltrated (Figure 3-3a). Thus, there is no
significant change in cumulative storage and surface storage. The subsurface storage declines
sharply in medium and large events as water is discharged through the outlet. Since the total
available pore space in the unsaturated zone is estimated to be about 23.3 in 3 , the number of
times that the unsaturated zone is flushed out can be estimated from the reduction in subsurface
storage. For instance, the unsaturated zone is flushed out once during the selected medium event
and about four times during the large event in Figure 3-3b-c.
90
The start of Stage B corresponds to the time point when pond water level falls below the
extended surface detention depth of 13.7 cm (i.e. culvert overflow just ceased). The end of Stage
B corresponds to the termination of ponding in the basin (Figure 3-3). Therefore, throughout this
stage, water infiltrates into the soil under a ponding head. By this stage, inflow has also ceased.
For events with multiple rainfall peaks, which might result in multiple instances of pond water
level rising above and then falling below the detention depth of 13.7 cm, Stage B is defined as
the last instance. This definition facilitates statistical analysis in the next section. However, for a
small event, since there is no culvert overflow, there is no distinction between Stage B and Stage
A. There are 21 such small events out of the entire 80 events.
Since there is zero culvert overflow in Stage B, Stage B represents a duration when there is
slower drainage of ponded water as water infiltrates into the soil media under a ponding head as
compared to the faster channeling off via culvert overflow in Stage A. This phenomenon is
reflected in the gentler slope of decline of surface storage curve in Stage B compared to Stage A.
The gentler decline of surface storage curve reflects the amount of ponded water that is entering
the unsaturated soil zone. Moreover, since there is no culvert discharge or inflow during Stage B,
the influence of culvert discharge and inflow discharge on the storage term is removed from
Equation 3-2. Hence, the decline in the cumulative storage curve in this stage entirely represents
the amount of water leaving the basin through the outlet. Therefore, a decreasing cumulative
storage curve decreases subsurface storage since water is flowing out of the basin while a
decreasing surface storage curve increases subsurface storage since water is transferred from one
storage compartment (the pond) to another compartment (the unsaturated soil zone). The relative
slope of cumulative storage compared to that of surface storage determines the net change in the
91
subsurface storage in Stage B. Since the slope of the surface storage curve is steeper than that of
the cumulative storage curve, this results in increasing subsurface storage during Stage B (Figure
3-3). In this stage, more water infiltrates into the soil under a ponding head than exits the basin
from the outlet.
The start of Stage C is marked by zero surface-ponding head until the end of drainage or the start
of the next event, whichever comes earlier. Thus, Stage C represents subsurface drainage by
gravity from the saturated soil media to an unsaturated state. In this stage, since there is no
surface storage, the drainage of water from the basin as represented by the subsurface curve
coincides with the cumulative storage curve.
Figure 3-4 presents statistics of some observed features. For the 80 events, the average ponding
duration (Stage A and B) is 12.5 hours with a standard deviation of 5.5 hours. Out of the 80
events, only 59 events have sufficient rainfall depth to result in culvert overflow and hence a full
Stage B, similar to the plots in Figure 3-3b and c. For these 59 events, the start of Stage B that
corresponds to zero culvert discharge occurs when the average pond water level is 13.9 cm. This
average of 13.9 cm is not exactly equal to the extended detention depth of 13.7 cm because
accumulating trash at the culvert causes the zero-discharge elevation to vary from event to event.
A level of 13.9 cm corresponds to a surface storage of 40 in 3 . The average ponding duration in
Stage B is 10.1 hours during which about 40 m3 stored in the surface basin needs to infiltrate into
the soil. This gives an estimated infiltration rate of 1.7 cm/hr. This infiltration rate is within the
same order of magnitude as the infiltration rate measured in the field using a double-ring
infiltrometer (4.4 cm/hr). However, these fall below the hydraulic conductivity (>10 cm/hr)
92
recommended for tropical regions by Facility for Advancing Water Biofiltration (2009). In this
case, an infiltration rate of 1.7 cm/hr under a saturated condition could indicate some degree of
surface clogging and soil compaction that could have occurred since the basin was constructed.
The average amount of increase in subsurface storage during Stage B is 23.7 m3 which
corresponds to about 100% of the total available porous space (estimated to be about 23.3 m 3 ) in
the unsaturated zone (Figure 3-4). This means that the unsaturated zone is likely to stay saturated
during ponding in Stage B as the porous volume is fully utilized to store water. However, under
the unsaturated condition in Stage C, the average gravity drainage rate is 0.3 cm/hr (Figure 3-4).
3.5 Conclusions
In this study, a six-month field monitoring program has been conducted at a bioretention basin in
an urban tropical catchment. Whether this basin has sufficient storage and infiltration capacity to
handle tropical storms has been investigated over 80 storm events. The main findings are:
1. The basin storage capacity is inadequate since in about 50% of the events, more than 50% of
the runoff overflows the basin without soil treatment. During events with >20 mm of total
rainfall depth, less than 30% of the runoff gets infiltrated and treated in the soil column.
2. The information and results drawn from this study could be of use to stormwater managers
for sustainable planning and development. For instance, water quality volume (WQV) is an
important concept and tool for sizing bioretention basins to ensure that pollutant removal
targets are met by retaining a certain amount of runoff. This study suggests that an
appropriate WQV for a tropical basin falls in the range of 10 to 30 mm.
3. Based on the aggregate statistics of the 59 events with extended surface infiltration, the
subsurface soil filter media is found to remain fully saturated under a ponding condition. The
93
infiltration rate is estimated to be 1.7 cm/hr which is within the same order of magnitude as
that found by direct field measurement using a double-ring infiltrometer (4.4 cm/hr). The
gravity drainage rate under unsaturated conditions is 0.3 cm/hr.
4. The observed infiltration rate is lower than that intended (>10 cm/hr) which indicates some
degree of surface clogging and soil compaction after five years of operation. A regular basin
maintenance routine could be set up to ensure that the basin maintains an intended infiltration
rate.
5. Field data, particularly for large storms, showed inconsistencies that made it difficult to close
the water balance on all storms. This problem could be at least partially addressed in future
monitoring programs by installing more instrumentation and collecting redundant flow
measurements.
This research was funded by the National Research Foundation (NRF), Prime Minister's Office,
Singapore under its Campus for Research Excellence and Technological Enterprise (CREATE),
through the Singapore MIT Alliance for Research and Technology's Center for Environmental
Sensing and Modeling (CENSAM) research program. The authors would also like to thank the
following for their insightful discussions and field support: Alwi Alkaff, Maria Stefanie
Suryadharma, Ong Geok Suat, Yau Wing Ken, Veera Sekaran, and Adeline Wong.
94
a)
22-~
20-
18-
16
Co 14.
12.
0 10-
6 8.
z 6-
4--
2-
0- i i i i i i
-80 -60 -40 -20 0 20 40 60 80 100 120 140

Water balance (%)
b) 150 * Backflow events

-
0
100- 0
50 0 0 0
0 0 0
-
0
a) -50 - P10
0
* 0
*
CCO
-100 I 1I 2 I
0 10 20 30 40 50 60 70 80
150 Rainfall (mm)
0
*- 100 0
cc
50 JI~m~
0 0
3: 0
0
0 0
0
C 0
-50 XnDc 00
0
-100
0 4567 8 9 10
ADP (days)
Figure 3-1. a) Histogram showing water balance for 80 events. b) Plots of water balance against
rainfall depth (mm) and ADP (days) for 80 events.
95
- Asymptotic fit
a) 120 * Event meeting removal targets
Event not meeting targets
-0 SCa * < 3-month ARI events
80 * > 3-month ARI events
C
40
74 events
0
b)
10 20 30 40 50 60 70 8
)
120
80
b) s 78 events
40 2
0 0
10 20 30 40 50 60 70 8
120
80
78 events
40
0
0 10 20 30 40 50 60 70 80
Total rainfall (mm)
Figure 3-2. a) Culvert discharge volume and b) outlet discharge volume as percentage of total
inflow volume, highlighting the water quality performance of the sampled events; c)
categorization of events in the monitoring period into > and < 3-month ARI categories.
96
-_:
--- - - - I_ q
a) 100
-
A+ part ofB C
80- cumulative storage
60 - surface storage
0) subsurface storage
E 40
20-
0 2 4 6 8 10 12 14 16 18 20 22 24
b) 100-
1 A C
CO>
60 _ _ _ _ _ _ __ _ _ _ _ _
40-
E
> 0
-20
0 2 4 6 8 10 12 14 16 18 20
c)
80-
0_C
40-
a)
E
-5
-80
0 1 2 3 4 5 6 7 8 9 10 11 12
Time (hrs)
Figure 3-3. Time series of cumulative storage, surface storage, and subsurface storage for a) a
small event (Event #63, 2.6 mm), b) a medium event (Event #61, 8.2 mm), and c) a large event
(Event #47, 29.4 mm). Time series are separated into three stages (A, B, and C).
97
I
'U
18
-
U)
14- p = 12.5 hrs Cn 16
12- = 5.5 hrs 14
C aO
a) 12
00 10-
8- 10
C: C
8
6- J
4- 6
4-
0 0 4
z 2- z 2
0
0 2 4 6 8 10 12 141618 20 22 2426 2 4 6 8 10 12 14 16 18 20
Ponding duration (Stages A+ B) (hrs) Stage B duration (hrs)
55-
3 18
50- p = 23.7 m U, p = 0.3 cm/hr
45-
C 16-
a = 6.4 m 3 a) a = 0.5 cm/hr
40- a) 14
a')
35- 12
30- 10
0 25- 8
6.. 20- U)
U, U) 6
15- 0
10 4
5- 2
0. 0
-10 0 10 20 30 40 50 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Increase in subsurface storage in Stage B (M 3 Drainage rate in Stage C (cm/hr)
)
Figure 3-4. Histograms of ponding duration (of 80 events), Stage B duration (of 59 events),
increase in subsurface storage in Stage B (of 59 events), and drainage rate in Stage C (of 59
events).
98
Chapter 4
Hydrological model
4.1 Abstract
Flash floods, non-point source pollution, and eutrophication are perpetual urban stormwater
management problems that cities across the world face. To mitigate such problems, an emerging
green infrastructure, the bioretention basin, has been deployed world-wide to reduce peak flows,
encourage infiltration, and treat pollutants. However, insufficient storage capacity in a basin
impairs its treatment potential. This need prompts the development and validation of a suitable
hydrological model for bioretention basins. In this study, an existing numerical model,
RECHARGE, has been adopted to simulate hydrological performance of a basin in a tropical
climate over a half year that included 80 storm events. Results show that RECHARGE could
successfully simulate the basin hydrology of 80 events of varying rainfall characteristics with an
event-wise mass balance error (MBE) of 5.1 7.5% and an overall half-year-MBE of 0.3%. The
verified model is then used to predict the infiltration percentage and basin performance
efficiency index over the half year as a function of detention depth, hd; ratio of drainage to
bioretention area, R; and saturated hydraulic conductivity of the basin soil media, Ks. Increasing
any of these basin characteristics improves basin performance, but the rate of improvement drops
significantly with increases beyond hd = 40 cm, R = 23.3, and Ks = 10 cm/hr.
99
4.2 Introduction
Traditional stormwater management strategies based on concrete-lined channels, together with
increasing urbanization and surface imperviousness, induce hydrologic modifications that result
in rising instances of flash floods in urban areas. Flash floods can be mitigated by encouraging
local infiltration through Best Management Practices (BMPs) such as bioretention basins
(Nehrke and Roesner 2004). More importantly, such green infrastructure also reduces non-point
source pollution and eutrophication by providing a soil treatment medium to remove pollutants
such as nitrogen, phosphorus, and suspended solids (Payne et al. 2014). However, insufficient
storage capacity in the basin impairs its treatment potential. An undersized basin will experience
a high rate of overflow and allow stormwater to bypass storage and treatment, leading to
pollutant removal rate below targets (Wang et al. 2017). How to size basins effectively to meet
pollutant removal targets therefore is an important issue to address.
With the accumulation of experience in working with BMPs across the world over the past 20
years, BMPs have become part of a regulatory and planning requirement for land developers.
Planners are required to demonstrate that with the installation of BMPs, the post-construction
runoff will be able to replicate the pre-construction hydrology as closely as possible. Therefore,
it becomes necessary to quantify accurately the effects of BMPs in encouraging infiltration and
reducing runoff generation. However, current quantification of these effects, and hence design
guidelines that arise from such quantification, have mostly been based on "rules of thumb" rather
than specific design curves. The need to predict basin performance at projected dimensions
prompts the development and validation of a suitable hydrological model for bioretention basins.
100
However, there have been few hydrological models specifically for bioretention basins. Heasom
et al. (2006) used a simplified way to model bioretention basin infiltration and culvert bypass
that involves a few lumped constant rates. Infiltration is represented as a two-stage process with
two rates: a rapid infiltration stage at the start of an event in the absence of surface ponding (as
represented by RPILoss); a second steady infiltration stage when infiltration occurs under a surface
ponding condition (as represented by a rate C). RPLOSS is taken as a diversion element in the
HEC-HMS model to represent the capacity of the bioretention basin to rapidly infiltrate a portion
of each incremental inflow. C is used in the post-processing of the outflow generated by storage
indicator routing from the HEC-HMS model. As noted by the authors themselves, their method
of modeling could be improved by relating infiltration to physical characteristics such as soil
properties or by treating saturated/unsaturated flows like a ground water model.
The HYDRUS-1D and HYDRUS (2D/3D) software packages are finite element variably-
saturated models that numerically solve for the movement of water using the Richards equation
for saturated-unsaturated water flow (Simiinek and Van Genuchten 2006). Although this code
models infiltration with a more complicated and higher resolution approach than using lumped
constants as in Heasom et al. (2006), it does not have a surface water balance to account for
ponded infiltration and neither it is specifically built for bioretention basins.
Another complex model, RECHARGE, could be considered the state-of-the-art hydrological
model for bioretention basins, due to its high resolution modeling of unsaturated flow by the
Richards' equation as well as coupling that with a surface water balance (Dussaillant et al. 2004).
Since its development, RECHARGE has been verified against single events in a controlled
101
experimental setting, but not at field-scale or over a continuous time period that includes
multiple storms and inter-storm periods.
A finite element model was developed by He and Davis (2011) specifically for bioretention
basins but this model was only verified with local rainfall data or literature values and not
validated using field measurements.
The current literature lacks a field-verified, bioretention-basin-specific, and high-resolution
model to simulate ponded infiltration. In this paper, we aim to bridge that gap by applying the
RECHARGE model in a continuous fashion for half a year, in which the model is validated
against 80 events of wide variety of rainfall depths and intensities in the tropics. We present
research on high-resolution single-event and continuous (half-year) modeling of hydrological
behavior including time series of discharges at the subsurface outlet and surface overflow culvert
as well as surface ponding depth and soil moisture of the filter media. We then apply an
infiltration index (ID) to describe predicted model performance for a tested range of engineering
parameters.
The objectives of this study were to:
1. Validate the applicability of the RECHARGE hydrological model for continuous
simulation at field scale.
2. Attempt to predict bioretention basin performance in terms of infiltration capability as a
function of engineering design parameters such as detention depth, ratio of drainage area to
basin area, and saturated hydraulic conductivity of the soil media.
102
3. Recommend basin configurations that improve performance efficiency.
4.3 Methodology
4.3.1 Flow and water level measurements
In April - November 2013, a real-time monitoring system that consisted of flow and water-level
measurement equipment was installed at Balam Estate Rain Garden in Singapore (Wang et al.,
2016). Four Doppler ultrasonic area-velocity level sensors (ISCO-2150 Area Velocity Module
with 2105 Interface, Teledyne Isco, Lincoln, Nebraska, USA) were used to continuously monitor
flow rates at the inlet, outlet, and overflow culvert as well as ponding water level at one-minute
intervals. Weekly basin maintenance was carried out in September - November 2013 to prevent
sensor blockage by fallen leaves and trash.
4.3.2 Bioretention geometry
Figure 4-1 shows a cross section of the bioretention basin. With an effective treatment area of
240 M2 , the basin receives runoff from a 16,800 m 2 residential catchment (about 88% surface
imperviousness) with a time of concentration of about 10 min (Wang et al. 2017; Wang et al.
2009). Basin inflow infiltrates through four soil layers consisting of (from surface to bottom): a
40-cm sandy-loam filter layer; a 10-cm fine-sand transition layer; a 30-cm saturated anoxic zone
of hard rocks (average size of 50 mm and 65% by volume) and wood chips (average size of 5
mm and 35% by volume); and a 5-cm drainage layer of fine gravel (average size of 2-5
mm)(Ong et al. 2012; Wang et al. 2009). Treated runoff is collected by underdrains that drain to
an outlet pipe that discharges to an adjacent drainageway. As the underdrains are at a lower
elevation than the outlet, a 35-cm-deep saturated anoxic zone favoring denitrification is created.
103
The basin also has a 1.1-m-high-by-1.2-m-wide culvert with a 13.7-cm-high overflow weir that
restricts the amount of surface ponding in the basin. The entire soil filter media is underlain by a
high-density polyethylene (HDPE) liner to prevent exfiltration into the surrounding soil.
4.3.3 Hydrological model - RECHARGE
4.3.3.1 Governing equations
In the RECHARGE model, the variably-saturated subsurface flow is simulated with a one-
dimensional mixed form of the Richards' equation which models both saturated and unsaturated
conditions in the basin during wet and dry periods (Dussaillant et al. 2004). Under the "mixed"
formulation of the Richards' equation, 0is a function of both h and z. The governing equation is
ao(h,z)K(h, z) + 1 - S(h, z) Equation 4-1

at ah at -z [K (az
where 0 = soil moisture content (cm3 /cm3 ); t = time (hr); z = vertical depth (cm); h = pressure
head (cm); K = unsaturated hydraulic conductivity (cm/hr); and S = depth-averaged
evapotranspiration rate (hr ').
Solving the Richards' equation (Equation 4-1) requires relationships among h, 0, K, and -. The
aft
Van Genuchten-Mualem (Equation 4-2 - Equation 4-4 ) are used in RECHARGE to estimate soil
properties, assuming that there is no hysteresis (van Genuchten 1980). Figure 4-2 shows the
relationships illustrated in these equations with the parameters for the four soil layers listed in
Table 4-2.
0(h) Os -r _+0 Equation 4-2

[1 + (a |h|)n]m r
104
K(h) = Ks (a IhI)n- 1 [1 + (a IhI)'] 7 '} 2 Equation 4-3
[1 + (a lhI)n]m/ 2
Me am(O, - 0r) {1 - [1 + (a Ih|)E]-l}m

M(h) = -hBh= 1I-M[+(alln, Equation 4-4
-rn [1 +(a |h)]"1
where Os = saturated soil water content (cm 3/cm 3); Or = residual soil water content (cm 3/cm 3); jhl
= absolute value of the pressure head (cm); Ks = saturated hydraulic conductivity (cm/hr); a is a
van Genuchten parameter (cm-'), with a = hb-1, where hb is the air-entry pressure; and n is a
measure of the pore-size distribution (with n >1 and m = 1- 1/n).
One prominent feature of the RECHARGE model is that it considers the characteristic wetting
and drying conditions in a bioretention basin system by including a surface water balance. The
time-dependent depth of water ponded in the surface basin, hs (cm), is found from the water
balance Equation 4-5.
A = Qrain + Qin low - Qin-~ Qcuivert Equation 4-5

dt
where A = bioretention basin area (M 2 ); Qrain (m 3/hr) = direct rain onto the basin; Qinfow (m 3/hr)
catchment runoff entering the basin; Qif (m3 /hr) = water infiltrated into the soil media; Qculvert
(m 3/hr) = basin surface overflow. To compute the unit infiltration rate, qitf (cm/hr), the Darcy-
Buckingham law for unsaturated conditions is used (Equation 4-6).
qinf = K(hz) -+1 Equation 4-6
where K(h,z) is a representative spatial average.
105
4.3.3.2 Numerical scheme
In the RECHARGE model, Equation 4-1 is discretized using a fully implicit (weight, w =1)
Crank-Nicholson finite difference numerical scheme. The discretized form of Equation 4-1 is
shown in Equation 4-7 (Dussaillant et al. 2004).
o/
-
At
A
i = _-
AZ
1
' (h'
Ah
wK
-1/2
(+1AZ
-
++1K
+ 1 +(1-w)K
1-1/2
'1
AZ
+ 1
- [
wK
1+1/2
AhV1Ah
AZ
+ 1 + (1 - w)K! 1
1+1/2
(
AZ
+ 1
Equation 4-7
- S|
where i= spatial step of 1 cm; j= time step.
The system of equations is solved using the Thomas algorithm, with a modified Picard iteration
for mass balance (Dussaillant et al. 2004) . To improve the computation speed and convergence
of the model by minimizing finite-difference discretization truncation error, a variable time-
stepping scheme (Dussaillant et al. 2004) is adopted, with an initial At of 30 sec. The minimum
and maximum At are 10-6 hrs and I min respectively.
4.3.3.3 Modification of RECHARGEfor current study
Due to site specific conditions, modifications have been made to the inflow, culvert surface
overflow, boundary conditions, and initial conditions of the original RECHARGE formulations.
Instead of computing basin inflow as done in the original RECHARGE model, we used per-
minute measured inlet flow rates recorded during the monitoring period. The sum of this inlet
106
discharge and the direct rain that fell onto the basin area was taken to be the total inflow into the
basin. The total inflow was then assumed to be spread uniformly across the basin.
For the basin surface overflow, once pressure head of the surface node, hs (cm), exceeds a culvert
weir height of hd (13.7 cm) as shown in Figure 4-1, ponded water in the basin overflows through
the surface overflow culvert. A rectangular weir equation (Equation 4-8) was used to estimate
the amount of culvert discharge Qcuzer (m 3/s) (International Organization of Standards 1980;
Kindsvater and Carter 1959). In Equation 4-8, Kb = -0.0009 m and Kh= 0.0001 m account for
effects of viscosity and surface tension. The weir coefficient (Cw) was set by calibration and
validation, each against three months of data.
Qcuivert = 2s-2) (0.075) +C Equation 4-8
(0.01hs - 0.01hd + Kh) 3 / (b + Kb)
The original RECHARGE model contains three soil layers: a root zone, a storage zone and an
underlying layer of urban subsoil. We added one additional soil layer to the model to
accommodate an additional transition layer. We also converted the original layer of urban subsoil
to a drainage layer. The soil properties of the four layers are set according to Table 4-2. The
RECHARGE model assumes that water from the bioretention cell exfiltrates into the underlying
natural soil. However, the Balam Rain Garden is enclosed in a HDPE liner and water exits via
drains rather than exfiltration (Figure 4-1).
Due to the presence of a permanently saturated anoxic zone, a pressure head of 35 cm (as shown
in Figure 4-1) is imposed as the bottom boundary condition. This pressure head is equivalent to
107
the outlet pipe elevation. For the top boundary condition, the same decision tree in the original
RECHARGE model that incorporates a surface water balance is used to select among four cases
of flux and head boundary conditions (Dussaillant et al. 2004). At the soil surface, a specified-
flux boundary condition is taken if there are rainfall or evapotranspiration fluxes during the non-
ponding periods at the beginning and end of an event. A specified-head boundary condition is
taken if the basin surface is ponded with water during precipitation and after it has ceased.
For the initial conditions, a uniform increase in pressure head from -45 cm to 0 cm through the
soil thickness of 50 cm is assumed for the unsaturated sandy loam soil media and transition
zones. Another uniform increase in pressure head from 0 cm to 35 cm through the soil thickness
of 35 cm is assumed for the saturated anoxic and drainage zones.
4.3.4 Data analysis
The absolute mass balance error (MBE) was computed according to Equation 4-9 for individual
storm events and a half-year continuous simulation to find the difference between total inflow
and total outflow.
C100% Equation 4-9 MBE= 1 - x

Vrain + Vin
where Vo is the modeled outlet discharge volume; Vc is the modeled culvert discharge volume;
Vrain is the measured direct rain on the basin; and V is the measured inlet discharge volume. The
MBE calculation was also checked by cumulatively summing up fluxes over all time steps.
A sensitivity analysis was carried out to estimate the percentage of inflow that infiltrates the
basin, I (%), according to Equation 4-10.
108
I( = V x100% Equation 4-10
Vrain + Vin
ID (infiltration index) is an improvised measure that indicates the gain in I (%) per unit increase
in an engineering parameter: detention depth, hd (cm); ratio of drainage area to bioretention area,
R; and saturated hydraulic conductivity, Ks (cm/hr). ID is calculated according to Equation 4-11.
dl dI dl
ID = or - or Equation 4-11
dhd dR d Ks
4.4.1 Model calibration and validation
The field monitoring program of six months (June - November 2013) covers the Southwest
Monsoon (June - September) and the inter-monsoon period (October - November). To separate
half of the field data for model calibration and the other half for model validation, two months in
the Southwest Monsoon period and one month in the inter-monsoon period are selected for each
of calibration and validation. Field data from August, September, and November are selected for
calibration and those from June, July, and October are used for model validation.
To evaluate model prediction capability, three indices are used in model calibration and
validation: Nash-Sutcliffe Efficiency (NSE) (Nash and Sutcliffe 1970), normalized root-mean-
square-error (nRMSE), and coefficient of determination (R 2). These indices are used to compute
the difference between the observed (Oi) and modeled (M) results during the model parameter
calibration process, according to Equation 4-12 to Equation 4-14.
NSE = 1 - 1= 1 (o - Mi) 2 Equation 4-12

1=N(Q - 0)2
109
( MEquation 4-13
nRMSE = N
0
2I: 1(0, - &)(M, - )R)]2

R = 1 - Equation 4-14
where 0 and M are the observed mean and modeled mean value respectively and N is the
number of comparisons made between observed and modeled values.
The nRMSE, R 2 , and NSE values for pond water level, WLp (cm); outlet discharge rate, qo (cm/hr);
and culvert discharge rate, qc (cm/hr) at various values of Cw and Ks are shown in Figure 4-3 for
the calibration period. The test range for Cw and Ks is 0.05 - 0.5 and 0.5 - 1.4 cm/hr respectively.
As shown in Figure 4-3, as Cw decreases, the model prediction more closely matches the
observed value of qc, as shown by the monotonically decreasing nRMSE and increasing R 2 and
NSE. For qo, however, as Cw decreases, an increasing R2 value indicates an improvement in the
match between modeled and observed but not the decreasing NSE and increasing nRMSE. In
terms of WLp, at C. = 0.2, the nRMSE and NSE values of WLp are at their lowest, representing
the best fit between observed and modeled WLp for the months of August, September, and
November. Since the ponding level, WLp, is governed by both the surface overflow through the
culvert (qc) and the subsurface infiltration and drainage (qo), the decision on the overall best fit
for C. (Cw= 0.2) is made primarily based on WLp.
For the range of Ks tested, the lowest nRMSE and highest NSE for qo are seen when Ks = 0.5
cm/hr but the maximum R2 occurs when Ks = 1.1 cm/hr. This apparent disagreement can be
explained by the small values of qo throughout the simulation period. During the dry inter-event
110
periods, there is often zero outlet discharge (qo = 0 cm/hr) and these zero values introduce bias
when they are averaged with the positive outlet discharges during the events. Therefore, R2 is a
better indicator of goodness of fit for qo. Unlike NSE, the index R 2 relies on the point-to-point
comparisons between observed and modeled values. The maximum R 2 value for qo occurs at Ks =
1.1 cm/hr, indicating that the correlation between observed and modeled values of outlet
discharge is the highest at Ks = 1.1 cm/hr. For the effect of Ks on qc, the modeled values are the
closest to observed values when Ks = 1.2 cm/hr. For the effect of Ks on WLp, a maximum for R2
and NSE and a minimum for nRMSE occurs when Ks = 0.9 cm/hr. Although no single value of Ks
is seen to give the best match between observed and modeled values, Ks = 1.0 cm/hr appears to
produce the best overall results and is selected to be the final calibrated parameter value. At Ks=
1.0 cm/hr, the overall values of the three indices (nRMSE, R 2, and NSE) indicate a good balance
among the affected factors (WLp, qo, and qc) under consideration.
The calibrated parameters (C. = 0.2 and Ks = 1.0 cm/hr) were validated against an independent
set of field data from the months of June, July, and October. The prediction capability of the
model was evaluated with the same three indices (nRMSE, R 2 , and NSE) for WLp (cm), qo
(cm/hr), and q, (cm/hr). The statistical values show that the model is able to predict well: the
nRMSE values for the validation dataset for WLp, qo, and q, are 1.11, 1.79, 7.06 respectively; the
R 2 values are 0.86, 0.72, 0.77 respectively; and the NSE values are 0.86, -0.51, and 0.77
respectively.
111
4.4.2 Sensitivity to model parameters
Six hydraulic parameters were determined by calibration of water level in the bioretention basin
(WLp) and outlet and culvert discharge rates (qo and qc) against observed values. Ks, CW', Os, Or, a,
and n. Aside from Cw, which mainly affects the culvert discharge rate, the other five model
parameters define the shapes of the soil water retention curves (Equation 4-2 to Equation 4-4)
which describe the relationships among the unsaturated hydraulic conductivity, K(h); saturation,
0(h); and pressure head, h.
4.4.2.1 Effects of weir coefficient, C
The culvert weir is considered as a fully suppressed rectangular thin-plate with a width of 1.2 m.
The effect of changing the weir coefficient is shown in Figure 4-4. The weir coefficient has a
pronounced effect on pond water level, WLp, and culvert discharge rate, qc (cm/hr). As Cw
decreases from 0.5 to 0.05, WLp increases. The amount of increase is proportional to the
observed peak, with a larger increase seen in large events with higher observed peaks. This
differing rise brought about by C. makes Cw = 0.05 most suitable for fitting large events with
higher peaks and least suitable for fitting those with lower peaks. The converse is true for C =
0.5. A higher Cw value increases culvert discharge by allowing more surface water to overflow
the basin. Therefore, the reduction in surface overflow is too prominent in large events to create
a good fit for small events with a high weir coefficient of Cw = 0.5. In terms of the ponding
duration, increasing C reduces ponding duration due to higher surface overflow and shorter
period of residence within the basin. However, the average reduction in ponding duration of
about 1.5 hours is not too pronounced to warrant a reduction of C from 0.5 to 0.05. Cw= 0.2 on
112
the other hand, gives an overall best fit for both high- and low-peak events, as well as giving a
realistic ponding duration.
In terms of the culvert discharge, q, (cm/hr), C has a significant influence on the peak of qc but
less on the duration of time over which there is flow in the culvert. A higher C, of 0.5 results in
higher peak values of qc than observed values but lower values in non-peak regions (Figure 4-4).
A medium C. of 0.2 is thus more appropriate for calibrating culvert discharge, qc.
Effect of C, on outlet discharge, qo (cm/hr), is negligible (Figure 4-4). (As explained in Section
4.4.3.3, the spikes in flow seen in Figure 4-4 are not modeled and thus not considered in the
calibration.) Even though a higher ponding height induced by a higher C, might result in a
higher driving head from standing ponded water for outlet discharge, the modeled values of qo
remain the same for C.= 0.5 and 0.2 as well as a negligible deviation from that of C,= 0.05.
4.4.2.2 Effects of saturatedhydraulic conductivity, Ks
Effects of saturated hydraulic conductivity on WLp (cm), qo (cm/hr), and qc (cm/hr) are shown in
Figure 4-5. As Ks decreases from 1.3 to 0.5 cm/hr, the peak value of WLp remains unaffected.
This is expected since the peak period most likely corresponds to the period when there is high
culvert overflow (high qc). The culvert is thus the dominating controller of WLp. As culvert
discharge dissipates and surface water recedes in the basin, Ks, which controls the infiltration
capacity of the soil media, will take over as the dominating factor. Therefore, an extended period
of ponding is seen with the lowest tested value of Ks of 0.5 cm/hr. As shown in Figure 4-5, once
113
several rainfall events occur close to one another, the basin is predicted to be perpetually ponded
if this low infiltration rate is specified.
Since Ks affects the infiltration capacity of soil media, it has its most pronounced effect on the
outlet discharge, qo (cm/hr). As Ks is reduced from 1.3 to 0.5 cm/hr, the peak values of qo
decrease and approach the observed values. However, the extended duration seen in WLp also
becomes apparent for qo at Ks = 0.5 cm/hr. There is a need to consider simultaneously fitting
peak values and durations of qo and WLp because favoring one factor would require a
compromise of the other factor. In this case, a Ks value of 0.5 cm/hr gives the best fit for peak
values of qo but also gives unwanted extended duration in both WLp and qo (Figure 4-5). As part
of the second phase of model calibration (described below), adjustments were made to the other
hydraulic parameters: Os, Or, a, and n to explore the possibility of resolving the extended duration
seen in modeled profiles for Ks = 0.5 cm/hr. The effect of Ks on culvert discharge, qc (cm/hr), is
negligible (Figure 4-5).
4.4.2.3 Effects of model parameters:Os, Or, a, n
Effects of a on WLp (cm), qo (cm/hr), and qc (cm/hr) when Ks = 0.5 cm/hr are shown in Figure
4-6. As a decreases from 3 to 0.005 cm-1 , WLp for large events is unaffected but for small events
is over-estimated. The ponding duration is slightly reduced for large events but a value of a as
high as a = 3 is insufficient to eliminate the extended duration due to the low Ks of 0.5 cm/hr.
The ponding duration for small events is greater with a = 0.005 compared to higher values. At a
lower a, the air-entry suction is higher. Therefore, a larger air pressure is needed to overcome the
surface tension which makes air penetration more difficult and prolongs ponding duration.
114
As a decreases from 3 to 0.005, the rise and fall in qo becomes sharper and the profile of qo vs.
time more angular. The extended duration of outlet flow due to a low Ks of 0.5 cm/hr is slightly
reduced but not eliminated even at a low value of a = 0.005. The peak value of qo is slightly
increased but is within acceptable range. However, numerical instability in peak values of qo
occurs with a = 3. The angular profile reflects the ease of release of soil moisture during the
drainage phase. A sharper drop in qo during the inter-event drainage period for a = 0.005
compared to a = 3 is because soil moisture is held more tightly to the soil particles and requires a
higher air-entry suction or bubbling pressure to break through the continuous moisture films
between soil particles when a = 0.005. Therefore, a pressure head h that enables air entry and
leads to a positive qo for a = 3 is insufficient when a = 0.005. The soil-moisture-content profile
also reflects prolonged saturated soil moisture conditions in modeled profiles for a = 0.005 (not
shown in figures). Since perpetually high soil moisture content during inter-event periods is
impossible, a higher value of a (e.g. a = 0.075) is more appropriate for the model. The effect of a
on culvert discharge, q, (cm/hr), is negligible (Figure 4-6) which is expected since q, is most
affected by Cw.
The effect of increasing n, a measure of the pore-size distribution, is similar to that of decreasing
a. For instance, increasing n leads to be a more angular profile with higher rise and fall in qo. The
extended duration of ponding is not eliminated and a prolonged soil moisture content results.
This is expected since the air-entry suction, a, is dependent on the uniformity of soil particle, a
metric that n measures. The effect of changing 0r is negligible (not shown in figures). Large
adjustments in the value of 6s lead to poor calibration of WLp and qo and such adjustments should
115
be avoided since they would essentially change the type of soil assumed in the model from that
used in the field.
Based on the analysis above, the model parameters 0s, 0, a, and n do not seem to have a
significant effect on resolving the extended duration of ponding predicted by the model.
Therefore, model calibration depends more on the two more controlling parameters Ks and Cw.
4.4.3 Simulation results in response to calibrated parameter set
The calibrated and validated parameter set (C, = 0.2 and Ks = 1.0 cm/hr) is used in modeling six
months (June - November 2013) of bioretention basin hydrology with RECHARGE. Time series
of monthly simulation results against observed values for the ponding water level, WLp (cm), the
culvert discharge, qc (cm/hr), and the outlet discharge, qo (cm/hr) are shown in Figures 4-7, 4-8,
and 4-9 respectively.
4.4.3.1 Ponded water level, WLp
Figure 4-7 shows that the simulated WLp agrees well with measured values. The simulated
maximum ponding height (WLpax= 25.3 11.8 cm) for the 80 events is comparable to that of
measured values (WLpmax = 26.1 16.4 cm), with a 3% under-estimation for the mean. The
simulated ponding duration (WLpd = 14.3 i 4.2 hrs) for the 80 events is slightly prolonged by 14%
compared to the measured values (WLpd = 12.5 5.5 hrs). Multiple ponding peaks were
simulated and well represent intermittent rain events. More frequent instances of over-estimation
for WLp are observed in the inter-monsoon period (October - November), especially in
November. One possible reason is that the basin weeding and pruning that took place during
116
major basin maintenance in November removed the top layer of vegetation, loosened up the
surface soil layer, and allowed for a higher infiltration rate.
4.4.3.2 Culvert discharge, qc
Figure 4-8 shows that the simulated qc hydrograph agrees well with the observed hydrograph.
The simulated peak discharge, qcmax = 119 155 cm/hr, for the 80 events is comparable to that
of measured values (qcax = I 177 cm/hr), with a 7% over-estimation for the mean. The
simulated culvert discharge duration (qcd = 2.4 i 2.1 hrs) for the 80 events is about 17% shorter
than measured values (qcd= 2.9 2.5 hrs). qc is largely over-estimated in the months of June
-
August, which could be attributable to the lack of basin maintenance and litter removal in the
first half of monitoring period. Basin maintenance in September - November reduced instances
in which the culvert discharge was slowed by trash accumulated on the trash rack situated
upstream of culvert discharge point. As a result, predicted values of qc are closer to observed
values during the second half of the monitoring period in Figure 4-8. Two peaks in the culvert
discharge, qc, occurred during the Is', 2 "d, and 4' largest events (on 2 0 th October, 2 5th September,
and 1 5 th October respectively). The second peak is not captured by the model and thus the model
underestimates the culvert discharge in these events. One possible reason is that during such
large events there are significant backflows from the neighboring Pelton Canal such that water is
held back from one-time discharge. After the canal water has receded, the detained water in the
basin is then allowed to be released freely. Since this phenomenon is not modeled, the second
peak for qc is absent from the simulation which results in an apparent under-estimation.
117
4.4.3.3 Outlet discharge, qo
Figure 4-9 shows that the time series of simulated qo agrees well with modeled values. However,
spikes in qo at the start of each large event are not modeled. While the 1 00-mm opening in the
outlet box (Figure 4-1) was intended to be sealed, there was in fact leakage which created the
observed spikes in qo. The leaks were not explicitly modeled. Therefore, during large events
where the ponded water level was greater than the elevation of these openings, WLp > 20 cm,
spikes in qo were seen although they do not represent true subsurface flow. These spikes were
excluded from our statistical comparisons by setting a cut-off outlet discharge value of 0.9 cm/hr.
With the spikes exlcuded, the mean modeled value, pqo = 0.350 cm/hr, is found to be a 28%
over-estimation of the mean observed value, Uqo =0.274 cm/hr. Dussaillant et al. (2004) found a
similar 30% over-estimation in simulated values in an experimental rain garden lysimeter.
Dussaillant et al. (2004) questioned if a long-term simulation would be able to reduce the over-
prediction in qo since that would allow for better characterization of the initial conditions such as
initial soil moisture and storage which tend to govern short-term simulation results. However,
results from this study show that the RECHARGE model might have other limitations that
restrict its performance.
4.4.3.4 Soil moisture, 0
Figure 4-10 shows that the basin becomes saturated soon after the start of an event, remains
saturated during the event, and returns to unsaturated state shortly after the event. This pattern
also agrees with the finding that the soil filter media remains saturated during events based on an
evaluation of hydrological performance of 59 field-observed events (Wang et al. 2016).
118
4.4.3.5 Mass balance, MBE
Measurements from 80 storm events which varied in total rainfall depth (P = 3 - 55 mm), 10-min
rainfall intensity (Iho = 4.8 - 122.4 mm/hr), total rain duration (ED = 10 - 120 min), and
antecedent dry period (ADP 2.25 - 9 days) are shown in Table 4-1. The modeling discrepancy,
as measured by the mass balance error (MBE) between the measured total inflow (Vin + Vrain) and
simulated total outflow (V, + Vc) for each event, is also shown in Table 4-1. There is a good
agreement in mass balance, with a mean of MBE = 5.1% and a standard deviation of 7.5% for the
80 events. The RMSE of MBE is 9.0%. In terms of the hydrology for the entire period of six
months, the total inflow volume, (Vin + Vrain)six-mth, and total outflow volume (V + Vc)six-mth, are
17,068 m 3 and 17,126 m 3 respectively, with an overall MBEsix-mth being 0.3% or 58m3 in absolute
amount.
4.4.4 Model application: infiltration percentage as a proxy for sufficiency of basin storage
The computed percentage of infiltration, I (50 34%), for each of the 80 events (4.2 < I < 100%)
using the calibrated parameter set is shown in Figure 4-11. I decreases exponentially with rainfall
depth, P (mm), with the largest drop in I(100% to 20 - 30 %) seen for small events (P < 10 mm).
This indicates that the current basin storage is insufficient for larger events (P > 10 mm). Figure
4-11 also shows that the histogram of I follows a bimodal distribution, showing that although
there are as many as 16 events with very high infiltration rate (I> 90%), there are also 23 events
with very low infiltration rate (I < 20%). This distribution implies that a disproportionately large
increase in basin storage might be required in order to maximize infiltration of inflow from the
large number of large events that have minimal I. With the current basin configuration, only
about 20% of the total inflow is able to infiltrate. How to effectively optimize basin
119
configuration in order to maximize basin storage in a land-scarce urban environment is therefore
a question that the validated model could be applied to answer.
Figure 4-12a-b evaluates the impact on basin capture and performance as engineering parameters
such as detention depth, hd (cm); ratio of drainage area to bioretention area, R; and saturated
hydraulic conductivity, Ks (cm/hr) vary. I is used as a proxy for the sufficiency of basin storage
capacity while ID (infiltration index as defined by Equation 4-11) (Figure 4-12c-d) is used as a
measure to illustrate the marginal gain in I given a unit increase in either hd, R, or Ks.
4.4.4.1 Detention depth, hd
Figure 4-12a shows that as hd increases from 10 to 80 cm, I increases from 18% to 55% when R
= 70 (current basin configuration). This increase is non-linear with a higher initial increase in I at
lower hd and gradually slower rate of increase at higher hd values. The non-linear slope is
captured in the ID plot (Figure 4-12c) in which a higher ID = 0.96 is achieved at hd =10 cm than
at hd= 75 cm (ID = 0.35). This trend indicates diminishing improvement in I per unit increase in
hd. This implies that once the detention depth, hd, increases beyond about 40 cm, the benefit in
terms of increasing the percentage of infiltration I by increasing hd would be less.
Similar trends are observed at other R values (R = 35, 23.3, 17.5). Comparing the ID plot (Figure
4-12c) at the four values of R, a lower R value, e.g. R = 35 (i.e. a larger basin area, A) starts off
with a higher ID than that of a higher R = 70 (i.e. smaller A). This implies that for a bigger basin,
the unit gain in I by increasing detention depth hd will be higher. However, as A increases further
from R = 70 to R = 35, 23.3, and 17.5, the marginal gain in I reduces, as indicated by a higher
120
jump in ID from 0.96 to 1.31 for R = 70 to R = 35 than the slight increase in ID from 1.53 to 1.58
for R = 23.3 to R = 17.5. This implies that at the four values of R tested, decreasing R beyond R =
23.3 will produce a low marginal gain in basin infiltration performance (ID). This trend is also
evident from the decreasing upward shift in the I curves as R decreases in Figure 4-12a.
Greater infiltration (higher value of 1) could be achieved with an increase in either detention
depth, hd, or basin area, A. Figure 4-12a shows that for the same surface storage volume (hd x A),
an increase in A results in higher I than would an increase in hd. For instance, with the same
surface storage of 144 m3 , a basin with hd = 20 cm and A = 720 m 2 (or R = 23.3) results in I=
57.6% which is higher than that of a basin (I= 47.9%) with hd = 60 cm and A = 240 m 2 (or R =
70). This is expected since with the increase in A, there will be a proportional increase in
subsurface storage besides the evident increase in surface storage.
4.4.4.2 Saturatedhydraulic conductivity, Ks
Figure 4-12b reflects the same trend of diminishing marginal gain in I as Ks increases from 0.1 to
60 cm/hr. The highest gain in performance, ID, is observed for the increase from Ks = 0.1 to Ks =
1 cm/hr (ID = 34) when R = 17.5. Beyond Ks = 10 cm/hr, there is less gain in performance per
unit rise in Ks as indicated by a steep drop in ID below Ks = 10 cm/hr (Figure 4-12d) and a rapid
tendency to plateau after Ks = 10 cm/hr (Figure 4-12b) when R = 17.5. This shows that basin
performance is most affected at low values of K (e.g. Ks < 10 cm/hr). Comparing the marginal
gain in I across the four values of R, the larger the basin area, the faster the benefit of
incremental benefit of I diminishes as the curve plateaus off. (Figure 4-12b). This is expected
121
since with a larger basin area, the amount infiltrated, I, is close to reaching its maximum, placing
a limitation to how much further I can be increased.
Combining the results from the performance evaluation above, I suffers from diminishing
marginal gain for hd, R, and Ks, but is most sensitive in the low range of Ks < 10 cm/hr. This
study suggests that to be cost effective in increasing I and hence hydrological performance,
basins need to be configured to have a detention depth hd of about 40 cm, a ratio of drainage area
to bioretention area R of about 23.3, and a saturated hydraulic conductivity Ks of about 10 cm/hr.
Further increasing Ks or hd or decreasing R would result in a further increase in I but the marginal
rate of return would be less. Before reaching these values, there is a high incentive to increase
parameter values up to the those suggested if space, design rules, and other practical concerns
allow.
4.5 Conclusion
The following may be concluded from this study:
1. The ability of the existing numerical model, RECHARGE, to successfully simulate the
hydrology of a bioretention basin in a tropical climate has been demonstrated with
continuous field measurements from 80 events over a half-year period.
2. The culvert weir coefficient, C.. and saturated hydraulic conductivity, Ks, are the main
parameters affecting discharge from the basin surface and subsurface respectively. The
effects of other model parameters which describe soil retention capability (0s, Or, a, n) are
less. Based on three indicators (nRMSE, NSE, and R2 ) for model calibration and validation,
the calibrated values for C,, and Ks are 0.2 and 1.0 cm/hr respectively.
122
3. The verified model was used to simulate time series of ponded water level WLp(cm), culvert
discharge qc (cm/hr), outlet discharge qo (cm/hr), and soil moisture 0 over a half-year period.
The simulated peak values and durations of these variables are close to the measured values.
There is also good agreement between total inflow and outflow for event-wise (MBE = 5.1
7.5%) and long-term mass balance (MBE = 0.3%).
4. There is a need to strike a balance between building a basin that has a sufficient size for
treating most storm events but which minimizes land usage. The verified model was used to
predict the performance of basins to infiltrate incoming stormwater as a function of detention
depth hd, ratio of drainage to bioretention area R, and Ks. The marginal improvement in basin
performance drops significantly with increases in basin dimensions beyond hd = 40 cm, R =
23.3, and Ks = 10 cm/hr. As a reference, by increasing current basin configurations (hd = 13.7
cm, R = 70) to a larger surface storage volume (hd = 40 cm, R = 23.3), the infiltration
percentage is estimated to increase from I= 21.6% to 75%.
This research was funded by the National Research Foundation, Prime Minister's Office,
Singapore through the Singapore MIT Alliance for Research and Technology's Center for
Environmental Sensing and Modeling research program. The authors also thank Alejandro R.
Dussaillant for providing the original RECHARGE codes.
123
Table 4-1. Mass balance of 80 events in the six months.
Event Start End Vin + Vrain V. vc
P (mm) (t ED (hrs) (D MBE (%)
Wm)
6/2/2013 4:40:00 PM 6/3/2013 5:22:00 PM 3 8.4 0.79 1.92 18.0 16.5 0.0 8.5
6/5/2013 2:43:00 PM 6/6/2013 5:21:00 PM 13.2 22.8 5.67 1.83 182 54.3 125 1.6
6/6/2013 5:22:00 PM 6/7/2013 10:08:00 AM 32 81.6 2.23 0.875 346 44.2 296 1.7
6/25/2013 3:35:00 PM 6/26/2013 12:17:00 PM 10.2 38.4 0.43 8.25 93.6 38.4 50.8 4.7
6/26/2013 12:18:00 PM 6/28/2013 11:10:00 AM 18.8 72 1 0.83 148 44.8 105 1.5
6/28/2013 11:11:00 AM 7/1/2013 2:45:00 PM 8 34.8 0.43 2 55.2 41.5 14.8 1.9
7/2/2013 10:37:00 AM 7/4/2013 11:15:00 AM 3.2 8.4 1.22 4 23.1 23.3 0.0 0.8
7/4/2013 11:16:00 AM 7/7/2013 6:55:00 AM 3.8 9.6 11 2 30.5 31.0 0.0 1.7
7/9/2013 4:48:00 AM 7/11/2013 11:54:00 AM 49.4 66 3 4.25 462 70.7 393 0.4
7/11/2013 11:55:00 AM 7/14/2013 4:23:00 AM 9 31.2 0.42 1.88 107 42.4 64.9 0.1
7/14/2013 10:27:00 PM 7/17/2013 11:29:00 AM 4.2 10.8 0.87 0.42 26.4 27.9 0.0 5.7
7/17/2013 11:30:00 AM 7/18/2013 6:18:00 AM 11 45.6 0.5 2.5 95.7 38.3 52.9 4.6
7/18/2013 6:19:00 AM 7/23/2013 1:30:00 PM 6 12 6 0.75 44.0 43.6 4.9 10.1
7/26/2013 3:49:00 PM 7/27/2013 9:21:00 AM 4.4 20.4 0.73 8.13 42.4 34.7 4.3 7.9
7/27/2013 9:22:00 AM 7/29/2013 10:47:00 AM 23.8 78 7 0.71 253 52.9 211 4.2
7/31/2013 8:18:00 AM 8/1/2013 3:18:00 AM 5.8 32.4 1.5 1.92 69.9 37.7 28.5 5.3
8/1/2013 3:19:00 AM 8/2/2013 12:49:00 AM 19 49.2 7 0.75 282 55.3 226 0.1
8/2/2013 12:50:00 AM 8/4/2013 9:30:00 AM 16 58.8 2 0.6 229 81.1 154 2.6
8/4/2013 9:31:00 AM 8/5/2013 8:38:00 AM 8 22.8 1.5 2.06 80.0 43.6 34.0 3.0
8/5/2013 8:39:00 AM 8/7/2013 9:31:00 AM 35.4 33.6 3 0.9 575 52.9 527 0.7
8/7/2013 9:32:00 AM 8/8/2013 5:51:00 AM 12.8 45.6 0.5 1.92 184 40.5 142 0.7
8/8/2013 5:52:00 AM 8/10/2013 4:01:00 AM 3.2 8.4 1 0.79 28.8 34.2 0.0 18.8
8/10/2013 4:02:00 AM 8/11/2013 5:40:00 AM 4.2 10.8 1 1.29 42.3 37.8 2.5 4.8
8/11/2013 5:41:00 AM 8/12/2013 2:05:00 AM 39.2 55.2 3 1 560 47.0 513 0.0
8/12/2013 2:06:00 AM 8/13/2013 7:12:00 AM 6.4 27.6 2 0.71 40.9 41.3 2.2 6.4
8/13/2013 7:13:00 AM 8/14/2013 10:58:00 AM 11.6 32.4 4 1.13 130 46.0 84.4 0.1
8/14/2013 10:59:00 AM 8/17/2013 2:54:00 PM 3.4 20.4 0.13 1 57.0 62.2 6.6 20.9
8/17/2013 2:55:00 PM 8/24/2013 3:37:00 PM 10.8 38.4 0.75 3.17 127 43.1 85.5 1.3
8/26/2013 5:11:00 PM 8/28/2013 1:44:00 PM 3.2 16.8 0.17 9.08 13.9 16.3 0.0 17.3
8/28/2013 1:45:00 PM 8/29/2013 1:22:00 PM 9.8 46.8 0.3 1.21 114 39.8 73.5 0.3
8/29/2013 1:23:00 PM 9/1/2013 10:06:00 AM 4.8 27.6 0.07 1 23.0 23.5 0.0 2.2
9/4/2013 5:12:00 AM 9/5/2013 8:16:00 AM 15.2 27.6 7 0.5 202 61.5 143 1.1
9/5/2013 8:17:00 AM 9/6/2013 10:06:00 AM 27.6 36 10 0.83 338 76.8 250 3.1
9/6/2013 10:07:00 AM 9/6/2013 11:25:00 PM 2.8 4.8 3 0.65 35.3 38.9 0.2 10.7
9/6/2013 11:26:00 PM 9/8/2013 2:13:00 AM 43 42 5 0.42 771 62.8 719 1.4
9/8/2013 2:14:00 AM 9/9/2013 12:42:00 PM 27.6 39.6 7.5 0.92 555 66.7 490 0.2
9/10/2013 6:37:00 AM 9/12/2013 2:49:00 AM 27.4 46.8 6 0.75 540 64.1 478 0.4
9/12/2013 2:50:00 AM 9/15/2013 7:54:00 AM 3 9.6 0.38 0.58 18.8 19.1 0.0 1.7
9/15/2013 7:55:00 AM 9/18/2013 8:24:00 AM 18.2 50.4 6 0.29 159 45.1 116 1.2
9/18/2013 8:25:00 AM 9/22/2013 11:34:00 AM 3.2 18 0.08 2 21.7 20.9 0.0 4.0
9/25/2013 9:56:00 AM 9/26/2013 10:29:00 AM 55 76.8 2 7.06 771 47.1 718 0.6
9/26/2013 10:30:00 AM 9/29/2013 5:25:00 AM 5.8 21.6 1.5 0.92 80.8 46.8 36.1 2.6
9/29/2013 5:26:00 AM 9/30/2013 1:20:00 AM 40.2 69.6 5 2.75 603 52.0 544 1.1
9/30/2013 1:21:00 AM 10/2/2013 6:26:00 AM 5.8 20.4 5 0.63 80.3 62.1 25.3 8.8
10/2/2013 6:27:00 AM 10/5/2013 6:00:00 AM 9.4 21.6 5 1.1 78.6 51.6 31.1 5.3
10/5/2013 6:01:00 AM 10/6/2013 4:05:00 AM 6 8.4 3 2.79 71.7 44.4 22.7 6.5
124
Event ADP Vi + Vrain vc
Start End P (mm) ED (hrs) MBE (%)
No (mm/hr) (days)
47 10/9/2013 1:44:00 AM 10/10/2013 6:26:00 PM 29.4 86.4 1.72 2.88 377 44.5 331 0.4
48 10/10/2013 6:27:00 PM 10/11/2013 10:51:00 AM 29.2 70.8 0.67 1.63 453 39.9 409 0.8
49 10/11/2013 10:52:00 AM 10/13/2013 3:05:00 PM 3.4 9.6 1.62 0.67 64.7 48.5 21.1 7.6
50 10/14/2013 6:48:00 AM 10/15/2013 4:50:00 AM 17.6 61.2 0.6 0.67 245 42.3 201 0.6
51 10/15/2013 4:5 1:00 AM 10/18/2013 10:24:00 PM 43.2 64.8 6 0.875 671 65.3 609 0.4
52 10/18/2013 10:25:00 PM 10/19/2013 10:02:00 AM 5.6 13.2 0.85 3.46 62.7 28.1 20.4 22.7
53 10/19/2013 10:03:00 AM 10/20/2013 2:32:00 PM 15.2 51.6 3 0.46 183 52.8 141 6.1
54 10/20/2013 2:33:00 PM 10/22/2013 12:50:00 AM 71.8 122 5 1.04 1260 54.3 1200 0.1
55 10/22/2013 12:51:00 AM 10/27/2013 12:38:00 PM 7.8 14.4 1.5 1.21 81.5 44.3 38.4 1.4
56 10/27/2013 12:39:00 PM 10/28/2013 3:03:00 PM 33.2 44.4 4 5.42 474 43.9 428 0.5
57 10/28/2013 3:04:00 PM 10/29/2013 6:12:00 PM 34.2 58.8 2.5 0.92 637 48.7 587 0.2
58 10/29/2013 6:13:00 PM 10/30/2013 12:44:00 PM 5.2 30 0.22 1 69.7 39.4 29.3 1.5
59 10/30/2013 12:45:00 PM 11/1/2013 3:39:00 PM 4.4 21.6 0.5 0.75 53.2 43.7 18.8 17.4
60 11/1/2013 3:40:00 PM 11/2/2013 2:55:00 PM 8 27.6 0.68 2.08 103 40.1 60.4 2.2
61 11/2/2013 2:56:00 PM 11/3/2013 12:07:00 PM 8.2 20.4 0.65 0.938 138 42.2 106 7.1
62 11/3/2013 12:08:00 PM 11/4/2013 4:14:00 PM 6 27.6 0.37 0.875 48.4 42.0 9.9 7.2
63 11/4/2013 4:15:00 PM 11/5/2013 5:51:00 PM 2.6 8.4 0.92 1.17 20.7 20.4 0.0 1.7
64 11/5/2013 5:52:00 PM 11/9/2013 2:07:00 PM 29.2 50.4 3 1.21 590 77.4 512 0.0
65 11/9/2013 2:08:00 PM 11/10/2013 8:48:00 AM 12.6 54 1.5 3.42 234 41.5 191 0.4
66 11/12/2013 6:39:00 AM 11/13/2013 6:15:00PM 16.6 30 5.5 0.27 331 57.0 276 0.4
67 1/13/2013 6:16:00 PM 11/14/2013 4:37:00 AM 3.4 18 0.75 1.25 52.1 26.3 14.8 21.1
68 11/14/2013 4:38:00 AM 11/15/2013 10:50:00 AM 30.4 38.4 4 0.4 622 58.2 577 2.2
69 11/15/2013 10:51:00 AM 11/16/2013 3:08:00 PM 2.6 4.8 3.5 1.13 20.7 20.5 0.0 0.8
70 11/16/2013 3:09:00 PM 11/18/2013 4:14:00 PM 4.8 25.2 0.2 1 32.9 35.3 0.0 7.2
71 11/20/2013 4:45:00 AM 11/21/2013 2:58:00 AM 3.2 14.4 0.5 1.5 21.1 22.2 0.0 5.2
72 11/21/2013 2:59:00 AM 11/22/2013 9:47:00 PM 6.2 28.8 0.25 0.92 45.3 40.8 5.5 2.1
73 11/22/2013 9:48:00 PM 11/23/2013 4:27:00 AM 5.2 31.2 0.15 0.79 48.7 15.5 9.9 47.8
74 11/23/2013 4:28:00 AM 11/24/2013 1:01:00 PM 31.2 80.4 4 0.25 528 59.4 493 4.6
75 11/24/2013 1:02:00 PM 11/25/2013 2:09:00 PM 4.2 9.6 0.92 1.19 58.2 40.7 17.2 0.5
76 11/25/2013 2:10:00 PM 11/26/2013 4:01:00 PM 14.6 45.6 5 1 164 50.3 115 0.8
77 11/26/2013 4:02:00 PM 11/27/2013 6:16:00 PM 4.2 16.8 0.38 0.875 72.9 42.2 35.6 6.7
78 11/28/2013 4:22:00 AM 11/28/2013 3:09:00 PM 6 16.8 3 0.42 69.7 29.9 26.2 19.5
79 11/28/2013 3:10:00 PM 11/29/2013 5:51:00 PM 6 12 0.33 0.3 91.5 49.6 59.3 19.0
80 11/29/2013 5:52:00 PM 11/30/2013 1:47:00 PM 5.4 20.4 2.5 1 72.4 40.2 34.2 2.7
P (mm): rainfall depth; I/o (mm/hr): highest 1 0-min rainfall intensity; ED (hrs): event duration; ADP (days): antecedent dry period; Vi,
(m3): measured inlet volume; Vrain 3

): measured direct rain on the basin; V0 (m3 ): modeled outlet volume; Vc (M3): modeled
culvert volume; MBE (%): mass balance error.
125
Table 4-2. Hydraulic parameters of bioretention basin soil media.
Unsaturated Transition Saturated Drainage

layer layer layer layer
Soil Sandy loam Fine sand Hard rock & Fine
wood chips gravel
Thickness [cm] 40 10 30 5
Os [cm3 /cm3 ] 0.3 0.35 0.4 0.4
Or [cm 3/cm 3] 0.12 0.1 0.4 0.4
Ks [cm/hr] 1 2 100 200
a 0.075 0.075 0.075 0.075
n 1.89 1.89 1.89 1.89
m 0.47 0.47 0.47 0.47
Culvert overflow (surface exit)

300m
(
/
Outlet box 863 mm
f 100 mm
00 mm opening .A#,o
4
137 mm culvert weir
600 mmM
180 mm 77q
60 M300
M 220 mm
t163 mu,
At
W.-
Figure 4-1. Cross section of the Balam Estate bioretention basin.
126
1 1...
0.8-
0.6-
0.4-
3 2
0.2-
01
-10 5 -10 4 -103 -102 -101 -100
Pressure head, h (cm)
. I
0.25
E
' 0.2
E
0.15
U. 1
-10 5 -104 -103 -10 2 -101 -100
0-3
5 x
4-
3-
02-
1-
U--
-10 5 -104 -103 -102 -101 -10 0

Pressure head, h (cm)
Figure 4-2. Relationships between pressure head (h) and relative hydraulic conductivity (Kr),
pressure head (h) and soil moisture (0), and pressure head (h) and soil moisture capacity ( ) as
described in van Genuchten-Mualem equations (Equation 4-2 to Equation 4-4), using parameters
in Table 4-2.
127
1,10- 5 8
-- WLP (nRMSE) 1,10-
WL, (nRMSE) 1.8
1.05- -- q. (nRMSE) -1.8
1.05- -- q (nRMSE) .57
-- q. (nRMSE)
1.00- - 1.8 -5.6 1.00-
q. (nRMSE)
U U
0.95- -1.4 0.95.
0.90- -1.2 -5.4 0.90- 1.2 -5
a a
0.85- -1.0 0.85- -U- 1.0
0.80- -0.8 -5.2 0.80- 0.8
0.75- -0.0 0.75- 0.8
54
0.70. -0.4 -5.0 ne i I
.
0.0 0.1 0.2 0.3 0.4 0.5 0.4 0.6 0.8 1.0 12 1.4
K (cm/hr)
0.86- -0.68 -0.80 0.86. - 0 80

-0.88
-0.66 -0.78
0.78 0.84-
0.84-
- 0.64 -0.64
-0.76 -0.76
-0.62
0.82- -0.62 0.82-
0.74 -0.60 -0.74

-0.80
0.80- 0.80-
-0.58 -0.58 -0.72
0.72
2
2 -0.56 0.78- ~U- WL, (R -0.56
0.78- WL-(R----
)
-0.70
-0.70
)
0.54
- (R 2 -0.54
-U- q (R2 -
)
0.76- q4 (R 2 -0.8
)
0.76- 2 0.8
-- q (R -0.52
)
- 0.52
)
0.74' ? -0.88 0.74- -.D--A

-
0.0 0.1 0.2 0.3 0.4 0.5 0.4 0.6 0.8 1.0 1.2 1.4
C., K., (cm/hr)
-1 -0.78 0.86. -1 -0.78

0.86.
0.84- -0.77 0.84- -0.77

-0 -0
0.82- -0.78 0.82- U -0.76
-
--1
0.80- 0.80-
0.75 -0.75
0.78- 0.78-
- -2 -0.74 - -2 -0.74
0.76- 0.76-
-0.73 -0.73
- -3 - -3
0.74- 0.74-
-- WL, (NSE) -0.72 - WL, (NSE) -0.72
0.72- 0.72-
n (NSE) --4 q., (NSE) --4
0.71 0.70. 0.71
0.70- --- q (NSE) q, (NSE)
L0.70
n An 1 -5 -0.70 U3 -
.
0.0 0.1 0.2 0.3 0.4 0.5 0.4 0.6 0.8 1.0 1.2 1.4
K, (cm/hr)
Figure 4-3. Normalized root-mean-square-error (nRMSE), coefficient of determination (R2), and

Nash-Sutcliffe Efficiency (NSE) values (pond water level, WLp (cm); outlet discharge rate, qo
(cm/hr); and culvert discharge rate, qc (cm/br)) for model calibration of weir coefficient (Cw) and
saturated hydraulic conductivity (Ks) using data from August, September, and November.
128
I Irve
50 - -observed
-
CW =0.5
40 -C =0.2
-
.C=0.05
I~E. 30
-.-
20
101
0
920 925 930 935 940
t (hrs)
-observed
Cw = 0.5
1.5 -C = 0.2
W
.C w
= 0.05
E 1
0
0r
I- 0.5 I I
0
1540 1560 1580 1600 1620 1640 1660 1680 1700
t (hrs)
-observed
300 Cw = 0.5
-C = 0.2
250
..... C = 0.05
-w
.C 200
E
$ 150
100
50
0
923.5 924 924.5 925 925.5 926 926.5 927 927.5
t (hrs)
Figure 4-4. Modeled profiles for pond water level, WLp (cm); outlet discharge, qo (cm/hr); and
culvert discharge, qc (cm/hr), at various weir coefficients, Cw = 0.5, Cw = 0.2, and Cw = 0.05
compared to observed values. Time t = 0 hr is the beginning of the field monitoring period from
June - November 2013.
129
50
-observed
_K =0.5
S
40 _K =1
S
K S =1.3
E 30
C.
. 0
20
10
2300 2320 2340 2360 2380 2400

t (hrs)
-observed
K =0.5
2
.K =1
- K =1.3
1.5
E
0
-.... -..............-.................
0.5
0
2300 2320 2340 2360 2380 2400
t (hrs)
350
-observed
300 - K = 0.5
S
-K = 1
250
K =1.3
-
-
200
E
0' 150
100
50
0
923.5 924 924.5 925 925.5 926 926.5 927 927.5
t (hrs)
culvert discharge, qc (cm/hr), at various saturated hydraulic conductivity of soil filter media layer,
Ks = 0.5 cm/hr, Ks =1 cm/hr, and Ks =1.3 cm/hr compared to observed values.
130
I
-observed
40 a = 0.005
-
-a = 0.075
..-=3
30-
E.
C
20
10-
0
550 600 650 700 750 800 850
t (hrs)
I I
1.2 -observed
-
a 0.005
1 = 0.075
3
0.8-
E 0.6
-
00.4-
a'4
0.2
0
L I I I I I I I J
3100 3150 3200 3250 3300 3350 3400 3450 3500

t (hrs)
-observed
300 a=0.005
= 0.075
250 .........................
a=3
1~
200
EC.,
150
()
----- I=
100
50
0
918 920 922 924 926
t (hrs)
culvert discharge, qc (cm/hr), at Ks = 0.5 cm/hr and various a values, a = 0.005, a = 0.075, and a
= 3 compared to observed values.
131
70 70
m Modeled - Modeled Modeled
Obserd so -Observed l o -- Observed s
60 ~50 6 60
~40, 030
0
150 5 ~10 50 (20

- 20
E 0 40
-V
0 a)4 0 5.0 8.2 8.4 8.6 8. - 10 11 12 13 14 15
9.0 9.2 9.4
Time(fth
4.5 5.0 5.5 6.0 6.5 7.0 7.0 6.0
Tkm. (day)
o 30 03 0
-
z CD 0
20 20
0 < 20
-
10 10 10
0 0 0
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30
Time (day) Time (day) Time (day)
70
Modeled ~s 100
Modeled
50
Modeled
60 0 Observed - observedi
130
Observs i 4m 80 40
- 0 -
-3 4 5 E
I
E.
CM 40 60 - 0o0 y) 30
S
00 00
CL
o 30
40 |
-~.( 20
0 0
C 20
z
20 10
10
0 0 0
I
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30
Figure 4-7. Ponding water level, WLp (cm), for the six months.
132
- ____-
300 Modeled a 350
- Modeled Modeled
Observed - Obsered -
-
Observed
250
.200
250 300 35 300 0
100 '3W0
2W0 100
80
- o 250 250 -0
200
E ~40
z20 2001 200 4.5s .0 5. 6.0 6.5 7'.0
150 01 5D 5 8.2 8.4 8.6 8.8 8.8 9.2
0
24.5 25.0 25.5 26.0 00
.1 50 Time (da) 150
-
Tm,. (day)
0 (D
b
z 100 -j
00 100
50
0
-A 50 50-
0T o. (day)
Modeled] 0 . . . . . . . . ..
.
o 2 4 6 8 10 12 14 16 18 20 22 24 26 28 3 0 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30
700 1000 600
i *00 - Modeled 250 - Modeled
Observedl 350 - Observed E200 I--Observed
600 500
250 800
000 500 w C
C, 400
-10
i600 28.s 27.0 27.5 288 0 2 4 10. 13. 132 13 13.8 13.2
400
t: 0 300
-5 8.0 U. 7.0 7.5
300 T80 (day) 400
0 > 200
(L F-
w~ 200 0 z
o 200
100
100
0 -1L.. 0 i i (Ii 0
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30
Figure 4-8. Culvert discharge, qc (cm/hr), hydrograph for the six months.
133
10 10
- Modeled - Modeled 2.0
12
Modeled 1 2.01 II I H I
- Observed - Observed - Observed 1.5
10 1.5
8 8
10.0 10
0.5 iu 005
0 0 11
E6 E6 L ~ ~ 12
(:
13 14 15 16 171
-b
. 13 14 15 16 17 18 19
01 ..
0 4.0 45 50 55 60 65 7.0 0
-3 Thre (day)
0 4 0
8
-J
2 2
2
0 0 0
0 2-0 4 6 8 10 12 14 16 18 20 22 24 26 28 30 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30
Ii I IL IK
aIkl I I 16
Modeled 14 Modeled 23
14 --- Observed F= Observed E-2.0
14
12 01.0 I > I
12 Outlet cover leakage
12 0.0:K (not modeled) 0 JIidiL..
10 S10
E
2 a 10
E10 EU
Una. (day) 0 1 2 s
467
8 Tho (day)
06 0 6
0
0
I:4 04 Z4
2 2 2
0 0 0
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 0 5 10 15 20 25 30 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30
Figure 4-9. Outlet discharge, qo (cm/hr), hydrograph for the six months.
134
Soil moisture content (0) for JUL Soil moisture content (0) for AUG
0.4 0 0.4 0 0.4
10 10 10
0.35 0.35 0.35
20 20 20
30 0.3 30 0.3 30, 0.3

E
40 40 40
0.25 EL 0,25 CL 0.25

o50
C 0
o50 50
60 0.2 60 02 60 0.2
70 70 70
0.15 0.15
80 80 80
-
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30
Day Day Day
Soil moisture content moisture content (0) for OCT R-1" "'"A"' e f^,*"
n Mn
Hi
III I
0 0.4
0 0.4 0 0.4
10 10 10
0.35 0.35 0.35
20 20 20
30 0.3 30 0.3 30 0.3
40 40 40
0.25 K. 0.25 0. 0.25
50 o50 50
60 0.2 60 0.2 60 0.2
70 70 70
0.15 0.15 0.15
80 80 80
. . . . . . . . . .
.
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30
Day Day Day
Figure 4-10. Spatial-temporal profile of soil moisture, 0 (cm 3/cm 3) for the six months.
135
120 1=1=1* 1=1* I.
14-
100
12-
80 c 10-
A
60 8-
-7
40
20
W
6-
4.
2-
I
0 : I
0 10 20 30 40 50 60 70 80 0- I
0 20 40 60 80 100
P (mm) I (%)
Figure 4-11. Percentage of infiltration, I (%) and precipitation, P (mm) of 80 modeled events.
136
a) 100 b) 100
90 . . 90
.
AA
80 VA
*I0 0
80 /**
70 VI/ Aw 70 -0
-
60 XA 60 -U
-
50 0/ .'- 50
40 .n- K, = 1 cm/hr 40 h = 13.7 cm
-
30 S -U- R = 70.0 30 -- R = 70.0.
-
U-.- --- R = 35.0 -- R = 35.0
20 20
-
-- R = 23.3 - R = 23.3
10 -v- R = 17.51- 10 S-v-R=17.5
0 0
C 10 20 30 40 50 60 70 80 9C 0 10 20 30 40 50 60
hd (cm) K. (cm/hr)
c) 29n d) 40
R= 70.0. 5
1.8 35 . * R= 70.0
-
35.0
.
R=
1.6 4--e R= 35.0
R= 23.3 -
-
30 a-R = 23.3
1.4 -v-R= 17.5-
-
3- R= 17.5
-
1.2 K= 1 cm/hr 25 -
-
h = 13.7 cm.
-
-
2-.
1.0 S20 0~ 10 203 705
-
0.8- --
A A 15
0.6-
0.4- 10 -0 10 20 30 40 50
-
0.2 5 - K, (cm/hr)
0.0 0
0 10 20 30 40 50 60 70 80
0 0 20 30 40 50
hd (cm)
K. (cm/hr)
Figure 4-12. Predicted basin performance at various detention depth, hd (cm); ratio of drainage area to bioretention area, R; and
saturated hydraulic conductivity, Ks (cm/hr), as assessed using ID (infiltration index).
137
Chapter 5
Water quality model
5.1 Abstract
Rapid urbanization and increasing urban overland runoff exacerbate non-point source pollution
which impairs freshwater bodies and challenges stormwater management. The bioretention basin
is a type of green infrastructure that ameliorates downstream eutrophication through treatment of
runoff at its source. However, there is a limited understanding of its treatment potential with
different design configurations. Current black-box modeling practice of such systems further
limits the representation of bacteria-mediated biochemical reactions that are the main
contributors of nitrogen removal. This study provides a more comprehensive reaction-based
model of water quality treatment processes using multiple bacterial Monod-kinetics under a
variably saturated flow pattern for predicting the dynamics of nitrogen and organics in a
bioretention system. The model was calibrated and validated based on water samples collected in
a field monitoring program. The mean effluent concentration of ammonia, nitrate, nitrite, and
inorganic phosphorus are calibrated with coefficient of determination (R 2 ) of >0.90, Nash-
Sutcliffe efficiency (NSE) of >0.81, root-mean-square-error (RMSE) of <0.54, absolute percent
of bias (PBIAS) of <34%, while that of validation are R2 > 0.93, NSE >0.01, RMSE <1.17, and
PBIAS <76% respectively. Final simulations with sensitivity analysis on the effects of bacterial
kinetics and bacterial count were also conducted.
138
5.2 Introduction
Current design guidelines for bioretention basins lack consistency across geographical location,
partly due to variations in climatic conditions. The ability to optimize characteristics of a
bioretention basin to meet specific removal targets relies on the predictive power of a
comprehensive water quality model. Such a model could be used to verify the appropriateness of
design guidelines by evaluating the response of the system to different design parameters.
Current bioretention basin modelling efforts (both hydrological and water quality models) have
been limited. In terms of hydrological models, RECARGA (Atchison and Severson 2004;
Dussaillant et al. 2003) and RECHARGE (Dussaillant et al. 2004) (the more complex version of
RECARGA),are the only existing bioretention basin-focused models. In terms of water quality
models, there has been no process-based, system-scale mechanistic model catering to
bioretention basins. Some of the widely-used models are SWMM (Storm Water Management
Model) (Huber et al. 1988; Rossman 2004), SUSTAIN (System for Urban Stormwater Treatment
and Analysis Integration Model) (Lai et al. 2009; Lai et al. 2006; Shoemaker et al. 2009),
WinSLAMM (Source Loading and Management Model for Windows) (Myllyoja et al. 2001; Pitt
and Voorhees 1995), P-8 (Program for Predicting Polluting Particle Passage through Pits,
Puddles, & Ponds) (Walker 2007), and MUSIC (Model for Urban Stormwater Improvement
Conceptualization) (Wong et al. 2002; Wong et al. 2006). These water quality models are
applicable in general to BMPs such as bioretention basins, detention ponds, bioswales, wetlands,
and porous pavements. The emphasis of these models is often on the optimal number and
placement of the different types of BMPs across the watershed in order to achieve a targeted
level of water quality improvement as well as on the life cycle cost and economic benefits.
139
Not only do these models lack a specific module for bioretention basins, they do not model the
removal processes occurring in the bioretention basins soil filter media and thus cannot predict
the effluent quality. Instead, much of the complexity of these models is involved in making
influent quality evaluations based on physical input factors like hydrologic characteristics
(hyetographs, hydrographs, evaporation, and temperature); system hydraulics (conduit, manhole,
pump, weir, and orifice geometry); catchment characteristics (percent pervious or impervious,
depression storage, overland flow length, and initial abstraction rate); and soil properties
(hydrologic type, infiltration rate, moisture content, and antecedent moisture condition). In
contrast, pollutant removal is modeled in terms of an average removal efficiency for a specific
pollutant. Therefore, each BMP in these models is considered as one "black box".
There is a growing body of literature attempting to understand the main processes leading to
contaminant removal in bioretention basins. These studies seek to find a process-based
alternative to treating the whole bioretention basin as a black box and lumping all reaction
processes under a single first-order decay constant, k. Besides the physical processes such as
sedimentation, filtration, and adsorption that are commonly modeled in current models, chemical
and biological mechanisms are simultaneously active and mutually influence each other. In fact,
they play a key role in nutrient and pathogen removal. For instance, organic nitrogen removal is
bacteria-mediated through biochemical reactions such as ammonification that converts organic
nitrogen to ammonia, nitrification that converts ammonia to nitrate, and denitrification that
converts nitrate and nitrite to nitrogen. Such a detailed mechanistic breakdown is necessary to
track elemental transformation and transport within each zone of the soil filter media. Models
140
that take all these biochemical processes into account will give a more complete picture of the
complexity occurring in bioretention basins. The relative importance of the various processes
and design elements in contributing to overall pollutant removal could also be quantified through
sensitivity analysis. The model thus developed will help to clarify design optimization problems
such as the optimal ponding depth and soil hydraulic conductivity for pollutant load reduction.
Due to the lack of mechanistic process-based models for describing biochemical transformations
among nitrogen species within the soil filter media, current models focus more on pollutants that
rely on physical processes for removal. For instance, Li and Davis (2008c) developed a one-
dimensional model that represents filtration and capture of total suspended solids. Other
analogous models look at filtration and adsorption of particulate and dissolved heavy metals as
well as the fate of polycyclic aromatic hydrocarbons such as naphthalene and pyrene in
bioretention basins (He and Davis 2009; Li and Davis 2008a). Recently, there is a deeper
understanding of nitrogen composition and fate in bioretention basins (Li and Davis 2014) and
quantification of the nitrifying and denitrifying bacteria that are responsible for nitrogen removal
(Chen et al. 2013). Such studies prompt the development of nitrogen process-based mechanistic
water quality models as the next step in bioretention basin research.
One good starting point for bioretention basin modelling is to look at systems such as
constructed wetlands (CWs) in which similar biochemical transformations take place. The
microbial community including nitrifiers and denitrifiers that are responsible for nitrogen
removal in CWs play an equally key role in bioretention basins. Moreover, the hydrological
characteristics of bioretention basin are analogous to those of a surface flow (SF) CW in terms of
141
the presence of a ponded surface; to those of a horizontal subsurface flow (SSF) CW in terms of
the presence of internal water storage (IWS) and variable saturation; and to those of a vertical
SSF-CW in terms of the flow direction and variable saturation.
Extensive research on process-based numerical models of varying complexity has been
conducted for CWs. One group of such models incorporate flow models that use the Richards'
equation for variably-saturated flow (Langergraber and Simunek 2005; Ojeda et al. 2008; Wanko
et al. 2006). CW2D (Constructed Wetlands 2D) is one such model that couples the flow model
(HYDRUS) with multiple Monod-type kinetic equations to create a multi-component reactive
transport model (Langergraber and Simunek 2005; Langergraber and Simninek 2006; Simi'nek et
al. 1998; Simnunek and Van Genuchten 2006). Saeed and Sun (2011) recommend that models for
wetland systems be developed based on multiple Monod kinetics instead of first-order kinetics or
single Monod to predict the removal of nitrogen and organics.
In order to better evaluate how bioretention basin design affects treatment performance and
hence explore management options for solving pollution problems, it is imperative to develop a
process-based model which can be used to make reliable predictions and assist in formulating
appropriate controls. Therefore, this study aims to fill in the knowledge gap in water quality
modelling for bioretention basins by coupling the existing flow model for bioretention basin,
RECHARGE, with mathematical formulations adopted from the CW2D model. The resulting
combined model is calibrated and validated using field data from six tropical storm events.
142
5.3 Methodology
5.3.1 Field data for calibration and validation
Hydrologic measurements and water quality samples used in this study were obtained from a
real-time monitoring system that consisted of flow and water-level measurement equipment and
automatic water quality samplers installed from April to November 2013 at Balam Estate Rain
Garden in Singapore. A full description of the study site, sample collection scheme, data
collection and analysis can be found in (Wang et al. 2016; Wang et al. 2017). The model was
calibrated using three events (Events #4, #1, #6) during the sampling campaign and validated
using another three events (Events #2, #5, #3) from the same sampling campaign.
5.3.2 CW2D
Constructed Wetlands 2D (CW2D) was developed as an extension of the HYDRUS software
package for variably-saturated water flow and multi-component reactive solute transport
(Langergraber and imnnek 2006). CW2D models the bacteria-mediated biochemical
transformation and degradation processes for organic matter, nitrogen, and phosphorus in SSF
CWs. Its mathematical formulation has its origin in the Activated Sludge Models (ASMs) in
which biochemical process rates are expressed in Monod-type equations (Henze et al. 2000).
CW2D has been adopted to model water quality in bioretention basins in this study.
5.3.2.1 Conceptual model
Nitrogen removal via the conversion of organic nitrogen (ON) to nitrogen gas (N2) in
bioretention basins is the end-result of a chain of reactions which is mediated by multiple
143
bacteria species as shown in Figure 5-1. The bacteria species utilize different sets of electron
donors and electron acceptors and generate different final products as summarized in Table 5-1.
Organic nitrogen, the starting material that enters the bioretention basin, is modeled as part of
COD (chemical oxygen demand) (which in turn consists of readily biodegradable COD (CR),
slowly biodegradable COD (CS), and inert COD (CI)). Either through the hydrolysis of COD or
cell lysis of heterotrophs (XH) and autotrophs (XANs and XANb), ON is mineralized and
converted to NH4' during ammonification. NH4' could be incorporated back into bacteria cells as
part of aerobic growth for heterotrophs together with inorganic phosphorus (IP or P04 3-) in the
presence of 02. Alternatively, NH4' undergoes nitrification which is modeled as a two-step
aerobic process, from NH4' to NO2- (as mediated by autotrophic bacteria Nitrosomonas (XANs))
and then from N02- to NO3- (as mediated by autotrophic bacteria Nitrobacter (XANb)).
Nitrification occurs in the presence of 02 which is the preferred electron donor. In the absence of
free 02 but the presence of bound oxygen (N02- and N03~) (anoxic condition as defined by
WeiBbach et al.(2017)), N02 and NO3- replace 02 as the next thermodynamically preferred
electron acceptors that can be utilized by heterotrophic denitrifiers (such as Paracoccus
denitrificans and various pseudomonads). Organic matter (COD) in this case acts as the electron
donor and the denitrification product is molecular N2 which is lost from the bioretention system
and enters into the atmosphere. Figure 5-1 summarizes the nine biochemical reactions that are
responsible for nitrogen removal in CW2D: hydrolysis of COD (RC 1), aerobic growth of
heterotrophs (RC2), growth of heterotrophic denitrifiers on NO3- (RC3), growth of heterotrophic
denitrifiers on N02- (RC4), lysis of heterotrophs (RC5), aerobic growth of Nitrosomonas on
144
NH4' (RC6), lysis of Nitrosomonas (RC7), aerobic growth of Nitrobacteron NO2- (RC8), as well
as lysis of Nitrobacter(RC9).
Since the presence of 02 determines the type of growth that bacteria undergo, it directly affects
the abundance of bacteria species in the different zones of soil filter media in bioretention basins.
Table 5-1 shows the zone of primary residence for the different bacteria species. Autotrophic
bacteria, Nitrosomonas and Nitrobacter, would mainly reside in the upper unsaturated zone of
sandy loam. Deeper into the soil media and approaching the saturated anoxic zone of IWS, the
number of autotrophs reduces. Heterotrophs, in contrast, reside throughout the entire soil media
as the denitrifiers are facultative anaerobes.
5.3.2.2 Mathematicalformulationof reaction rates
The CW2D model considers the interactions among twelve components (02, CR, CS, CI, XH,
XANS, XANb, NH4', NO2-, N03-, N2, IP) in nine processes (RC1 - RC9). Its mathematical
structure is based on that of ASMs where these nine bacteria-facilitated transformation processes
are modeled using Monod-type bacterial growth and decay equations (Henze et al. 2000). The
overall reaction rate term, [mgi/(Lxh)], for component i follows the Gujer matrix structure as
shown in Equation 5-1).
R
dCj (0Y
)(RC) Equation 5-1
j=1
where Ci is the aqueous concentration; 0 [n 3 W/m 3 s] is the volumetric water content; V,i [mgi/mgj]
is the stoichiometric factor for component i = 1., N (N is the number of components) and
process j = 1, ... , R (R is the number of processes); and RC; [mgj /(Lxh)] is the Monod-type
145
kinetic reaction rate for process j in the aqueous phase. The reaction rates, -i, are the resultant
equations from element-by-element multiplications between V,j and RC for each of the twelve
components modeled listed in Table 5-3.
The stoichiometric matrix, Vii, consists of composition parameters (which describe the nitrogen
and phosphorus content in COD and bacteria biomass) and stoichiometric parameters (which
describe the generation of COD from hydrolysis and bacteria lysis, as well as the bacteria
biomass yield coefficients from the consumption of COD). Therefore, V,i represents the extent to
which a given microbial-mediated process, RC, affects the concentration of component i, Ci.
The Monod-type kinetic reaction rate, RC;, for modeling bacterial growth and decay consists of
the following general structure Equation 5-2):
RC(growth) = Yx,max (K Cs
1 Kis
K) + s1 Kinh,sz + Csz)
CNH4 CIP _
kKx,NH4 + CNH4 Kx,1 + Cp

p ( SY) Equation 5-2
RC(decay) = byCy
where x = XH, XDN, XANs, XANb represent heterotrophic bacteria in general, heterotrophic
denitrifiers, autotrophic Nitrosomonas, and autotrophic Nitrobacter respectively; Ux,nax is the
maximum specific growth rate of bacteria; Csi [mg/L] is the concentration of the limiting
substrate S 1 for growth; Ksi [mg/L] is the half-saturation coefficient for substrate S 1; Cs2 [mg/L]
is the concentration of the limiting substrate S2 that inhibits growth; Kinh,s2 [mg/L] is the half-
inhibition coefficient for substrate S2; CNH4 [mg/L] is the concentration of the limiting substrate
ammonium; Cip [mg/L] is the concentration of the limiting substrate inorganic phosphate; Kx,NH4
146
[mg/L] is the half-saturation coefficient for ammonium; Kxip [mg/L] is the half-saturation
coefficient for inorganic phosphate; p is the soil bulk density [kgs /dms3 ]; Y= XH, XANs, XANb
represent heterotrophic bacteria (including heterotrophic denitrifiers XDN), autotrophic
Nitrosomonas, and autotrophic Nitrobacterrespectively; Sy is the concentration of bacteria in the
solid phase [mg/kgs]; Cy [mg/L] is the concentration of bacteria in the aqueous phase; and by [1/h]
is the bacteria decay rate.
The term x Cj'

Ks 1 + Cs 1
) represents the specific growth rate of bacteria by utilizing substrate SI
KinhS2
for cellular respiration. The term represents the inhibitory effect on the growth of
inh,s2+
-%S2
bacteria caused by substrate S2. The term CNH4 CjP represents the potential for an
Kx,NH4 + CNH4 Kx,jp+ Crp
insufficiency of either of the limiting nutrients NH4+ and IP to reduce the rate of growth. These
terms are multiplied together as more than one nutrient or growth factor has the potential to limit
the growth rate of bacteria.
The CW2D model relies on a number of assumptions. All reactions are assumed to occur in the
aqueous phase and all components except bacteria are soluble and mobile. Bacteria associate
exclusively with the solid phase and their lysis (i.e, decay rate) is set as a constant and does not
depend on environmental conditions. Moreover, following the ASM models, it is assumed that
pH, rate equation coefficients, and stoichiometric factors are constant (Table 5-2). CW2D
defines kinetic parameters in RC (px,max, Ks, Kinh,s, and by) using literature values taken from
studies of wastewater treatment. However, since this study is on treatment of stormwater, the
kinetic parameters are set by calibration instead.
147
5.3.3 Coupling of RECHARGE and CW2D
Since the twelve components in CW2D are interdependent in a complex manner (e.g. the product
of one reaction becomes the reactant of another reaction), their concentrations must be evaluated
simultaneously at each spatial and temporal step. This can be done by coupling the
hydrodynamic model in RECHARGE with the Monod-type biological degradation kinetics in
CW2D.
The general coupling scheme between RECHARGE and CW2D to model 1D reactive transport
of the twelve components is shown in Figure 5-2. The advection of influent through the ID
column of soil filter media is carried out with a spatial step of 1 cm in RECHARGE. Two
important outputs from RECHARGE which become the inputs to the CW2D model are the time
evolution of flow rate, q(z) [cm/hr], between two neighboring spatial cells and the soil moisture,
O(z) [m 3 w/m 3s], within each spatial cell. Reaction within each spatial cell is assumed to be
uniform (that is, each cell is modeled as a continuously stirred tank reactor (CSTR)). Thus, the
ID column of soil media can be considered as a vertical series of CSTRs at any time step. The
mass balance for each component i in a single-cell CSTR reactor can be expressed as:
d Ci(z) 1 1
dt + - Ci,pre(z) = - Ci,post(z) Equation 5-3
dt T T
where T is the hydraulic retention time within the single-cell CSTR reactor z in this study; Cipre(z)
is the concentration of component i before undergoing simultaneous biokinetic reactions and is
calculated according to Equation 5-4; dCi(z) is the rate of change of the concentration of
dt
component i in the current spatial cell as governed by the set of Monod-type kinetics equations in
Table 5-3; and Cipost is the concentration of component i of the single-cell CSTR reactor z after
undergoing the simultaneous biokinetic reactions and will reinitialize to be the new influent
148
concentration for the next spatial cell z+1 within the same time step. The overall concentration of
component i before undergoing biokinetic reactions, Cipre, is calculated by summing up the
influent mass (C, 7 (z)q(z)dt) and initial mass (Ci,,njt(z)O(z)dz) within the single-cell CSTR
reactor z as shown in Equation 5-4. Cipre is a virtual concentration used for computational
convenience. In the actual system, mass inflow and reactions occur simultaneously.
Cipre (z) = C (z)0(z)dz + C 1, ,f1 (z)q(z)dt Equation 5-4

6(z)dz+ q(z)dt
At the end of the spatial stepping from the first spatial cell at the surface to the last cell at the
bottom of the bioretention basin, C1,post(z) is reinitialized to be the initial concentration at the
next time step as shown in Figure 5-2. The process is iterated until the end of the simulation time.
The time series for effluent concentration can be taken as that of the bottom CSTR.
5.4.1 Model calibration and validation
The six storm events sampled during the monitoring program as described in Chapter 2 were
used for model calibration and validation. These six events consist of two high EMC events
(Events #1 and #3) and four lower EMC events (Events #6, #5, #4, #2). Influent EMC is given
more emphasis than rainfall depth in this model verification since EMC has a direct effect on
biokinetic reaction rates. To separate half of the events for model calibration and the other half
for model validation, the two high EMC events are placed in the two different categories.
Consequently, the model was calibrated using three events (Events #4, #1, #6) and validated
using another three events (Events #2, #5, #3).
149
5.4.].] Model evaluation indices
To evaluate model's predictive capability, six indices are used: coefficient of determination (R2 ),
Nash-Sutcliffe efficiency (NSE) (Nash and Sutcliffe 1970), mean absolute error (MAE), relative
root-mean-square-error (RRMSE), RMSE-observations standard deviation ratio (RSR), and
percent bias (PBIAS, %) for the mean outlet concentration, according to Equation 5-5 to
Equation 5-10. (Khan et al. 2013; Moriasi et al. 2007; Woznicki and Nejadhashemi 2012).
2 ZE=1( - 0)M q-)]2

R= i2-1=_ )(Mi - ) Equation 5-5
_ - 0 1 (M - I)
NSE = 1 - 1 (O - - Equation 5-6

E1 _- 0)2
MAE = Equation 5-7

N
RRMSE - RMSE N Equation 5-8

PM, M
RMSE J - M)2
RSR= Equation 5-9
0 O- 6)2
_E=
1 (O - Mi)
PBIAS = N 100% Equation 5-10
where Oi, Mi, 0, and M are the observed, modeled, mean of observed, and mean of modeled
values respectively; N is the number of comparisons made between observed and modeled values.
These indices are chosen for model evaluation as they represent three major categories of model
evaluation techniques: standard regression, dimensionless, and error index (Moriasi et al. 2007;
Moriasi et al. 2015). Standard regression such as R2 determines the strength of the linear
150
relationship between modeled and measured data. R2 describes the extent to which the model
explains the variance of the observed data. R2 ranges from 0 to 1, the higher the value, the
smaller the error variance and the stronger is the collinearity between modelled and observed
values. A value of R2 >0.50 is considered acceptable (Santhi et al. 2001). The dimensionless
Nash-Sutcliffe efficiency, NSE, shows how well a plot of observed against modeled values fits
the 1:1 line by comparing the relative magnitude of the residual variance ("noise") to the
measured data variance ("information") (Nash and Sutcliffe 1970). NSE ranges from -oo to 1,
with 1 being the optimal value, and values <0.0 indicating that the observed mean value is a
better predictor than the modeled value. Values above 0.0 are generally taken as an acceptable
level of performance while Moriasi et al. (2015)recommend for watershed-scale models to
predict nitrogen load the following criteria: NSE >0.35, >0.5, and >0.65 for satisfactory, good,
and very good performance respectively. The criteria developed by Moriasi et al. is are an
imperfect fit with the model evaluated here since they are directed to a somewhat different type
of model but nonetheless they provide a useful touchstone. Error indices such as MAE, RRMSE,
RSR, and PBIAS quantify the error in the units of the observed values. RRMSE and RSR modify
RMSE based on a scaling factor which is the modelled mean and the observed standard deviation
respectively. RRMSE and RSR range from 0 to oo with the optimal value being zero. Moriasi et al.
(2007) recommend RSR <0.70, <0.60, and 50.50 for satisfactory, good, and very good watershed
model performance respectively. PBIAS shows the tendency of the modeled data to be higher or
lower than the observed values. PBIAS has an optimal value of 0% with positive values
indicating that the model tends to underestimate and negative values indicating that the model
tends to overestimate (Green and Stephenson 1986; Gupta et al. 1999). Moriasi et al. (2007)
151
recommended PBIAS < 30%, < 20% and < 15% for satisfactory, good, and very good model
performance respectively for prediction of nitrogen and phosphorus watershed load.
5.4.1.2 Model performancefor the calibrationdataset
Figure 5-3 shows the model performance using a calibrated parameter set (Table 5-2, Table 5-4,
and Table 5-5). The observed values are shown against the modeled values on a 1:1 line for the
event mean outlet concentration of six water quality parameters: NH3-N, N03-N, N02-N, ON,
TN, and IP. The model evaluation statistics (R 2, NSE, MAE, RRMSE, RSR, PBIAS) for each of
these water quality parameters is shown in Table 5-6.
Figure 5-3 shows there is generally good agreement between the observed and modeled values in
the calibration simulations for all the water quality parameters except ON. ON aside, the model
evaluation statistics for the rest of the water quality parameters achieved R 2 >0.90, NSE >0.81,
MAE <0.12 mg/L, RRMSE <0.54, RSR <0.44, and IPBIASI <34% (Table 5-6). The values of NSE
and RSR especially have passed the threshold for very good model evaluation performance
according to Moriasi et al. (2007) of NSE >0.65 and RSR <0.50 for these water quality
parameters. The correlation found in this calibration is satisfactory compared to those found by
Saeed and Sun (2011) for a multiple-Monod CSTR model for a vertical flow subsurface wetland
(R2 > 0.7, RRMSE < 1.0). Visually, Figure 5-3 shows that the three points representing the three
events used for calibration for each water quality parameter lie near the bisector line. Negative
values of PBIAS in Table 5-6 indicate that the modeled values are overestimating the observed
values. This is also evident in Figure 5-3 as most data points lie above the bisector line.
152
Figure 5-3 also shows the removal rate of the six water quality parameters: NH3-N, N03-N, N02-
N, ON, TN, and IP as an additional indicator of model predictive performance. A strong
correlation is seen between the modeled and observed removal rates as the data points lie closely
to the bisector line.
Poorer correspondence is seen in the calibration for ON. For ON, although an R 2 value of 0.80
still indicates an acceptable agreement between modeled and observed values, the other indices
show a poorer correspondence (NSE = -4.92, RSR = 2.30, PBJAS = -76%). The reason for the
less-than-satisfactory performance for ON could be partly due to its method of estimation in
CW2D. Unlike the other water quality parameters like NH3-N, NO3-N, NO2-N, and IP which are
directly computed in their reaction kinetic equations (Table 5-3), ON is estimated as a rough
fraction of CODs in the CW2D model. ON consists of 0.03 mgN/mgCOD-CR (iNCR)' 0.04
mgN/mgCOD-CS (iNCS), and 0.01 mgN/mgCOD-CI (iNCy) as represented by the composition
parameters in Table 5-2. Therefore, ON is indirectly inferred from the concentrations of CR, CS,
and CI.
5.4.1.3 Model performanceforthe validation dataset
The model is validated using the calibration parameter set (Table 5-2, Table 5-4, and Table 5-5)
on the set of three validation events (Events #2, #5, #3) independent of the calibration events.
Figure 5-3 shows that model is able to predict the mean outlet concentration and removal rates of
NH3-N, N02-N, and IP well. The correlation between modeled and observed values is strong
with a R 2 >0.93 and the error deviation is small with NSE >0.76, MAE < 0.03 mg/L, RRMSE <
0.38, RSR <0.49, and PBIAS <18%) (Table 5-6). However, model performance is slightly weaker
153
for N03-N and more so for ON and TN. The poor performance in TN could be linked to the
rough estimation method used for computation of ON as ON constitutes a large fraction of TN.
5.4.2 Simulation results
Based on the calibrated and validated model parameters which consist of kinetic parameters,
composition parameters, stoichiometric parameters, and others (Table 5-2), the simulation results
for each of the six events are shown in Figure 5-4 to Figure 5-9. The initial conditions within the
soil media and the influent concentrations which are the inputs to the CW2D water quality model
are shown in Table 5-4 and Table 5-5 respectively. Pollutographs from Figure 2-1 were taken to
be the time-dependent inputs as influent concentration for this model (Wang et al. 2017). The
time-depth profiles in Figure 5-4 to Figure 5-9 reflect the variations in concentration of the
twelve components in the CW2D model: oxygen, nitrogen species (NH3, N02, N03, N2),
phosphorus specie (IP/PO 4), CODs (CR, CS, CI), heterotrophic and autotrophic bacteria species
(XH, XANS, XANb), as the storm event progresses following the start of rainfall (t = 0 hour).
Table 5-7 shows the modeled outlet and culvert mass that contribute to the resultant removal
rates of six components (TN, ON, NH3, NO3, NO2, IP) compared to their observed values in the
six sampled events.
5.4.2.1 Water quality at the start qf an event (initial condition)
5.4.2.1.1 Oxygen
Before a storm event, the aqueous phase of the upper 40-cm unsaturated zone is assumed to be at
equilibrium with the atmosphere and is saturated with oxygen at 29'C (02= 7.67 mg/L, oxygen
profile in Figure 5-4 to Figure 5-9). In the bottom 35-cm saturated IWS (internal water storage)
154
zone, free oxygen is almost unavailable while bound oxygen in the form of N03 and N02 are
present. According to Ritter's (2013) dry-day measurement at the same study site, the
concentration of dissolved oxygen was found to be low: 1.75, 1.72, and 1.55 mg/L at 65cm,
75cm, and 85cm depths respectively. These measurements were imposed as the initial condition
for dissolved oxygen in the anoxic IWS zone in the model (Table 5-4). In the transition zone (40-
50cm depth), dissolved oxygen in the medium range of 4 mg/L was assumed.
5.4.2.1.2 Bacteria
Since this water quality model focuses on modelling the growth-decay dynamics of bacteria that
help to remove nutrients from stormwater runoff as part of biomass incorporation for energy and
carbon needs, the availability of electron acceptors in the form of 02, N02, or N03 for cellular
respiration becomes the determining factor for the distribution of bacterial species within the
different zones of the soil media. Nitrifying bacteria (XANs and XANb) are obligate
chemolithoautotrophs as well as obligate aerobes, which means that these nitrifiers fix C02 and
are not able to oxidize NH3 or N02 for energy needs in the absence of oxygen. As nitrification
ceases to occur at low oxygen levels (< 2 mg/L), the bottom 35-cm saturated IWS zone becomes
uninhabitable for nitrifiers. Since in this model, bacteria are assumed to be immobile and do not
advect with the infiltrated flow, the abundance of XANs and XANb is limited to the unsaturated
aerobic zone throughout the duration of storm events (XANs and XANb profiles in Figure 5-4 to
Figure 5-9). On the other hand, denitrifiers, a special group of heterotrophs (XH), are facultative
aerobes (autotrophic denitrifiers such as Thiobacillus denitrificans are not included in this
model). This means that denitrifiers are able to respire N02 and N03 whenever 02 becomes
unavailable. As denitrifiers are included as part of the general population of heterotrophs which
155
are found throughout the soil media at all depths, the initial heterotroph concentration, SxH-init
(mg/kg), is assumed to be higher in the anoxic IWS zone to account for denitrifiers' competitive
advantage in utilization of N03 and N02 as electron acceptors for cellular respiration. The
bacteria concentrations, SXH-init, SXANs-init, and SXANb-init (mg/kg), are computed based on literature
reports of nitrifying and denitrifying gene copies per gram of soil (Chen et al. 2013), (Table
5-12) together with gene copy numbers per genome of bacteria species (Kandeler et al. 2006).
5.4.2.1.3 Nitrogen, phosphorus, COD
The initial concentration distribution of NH3, N02, N03, N2, IP, CR, CS, and CI varies from
event to event because it is dependent on the extent of biochemical reactions that occurred in the
residual water during the dry period since the previous event. The streamlines in Figure 5-4 to
Figure 5-9 show that pollutant mass from the incoming stormflow reaches a depth of about 50-60
cm by the end of the simulation period of about 24 hours. The pollutants would then undergo
biochemical transformations throughout the intervening dry period and stay within the system
before it is flushed out by infiltrated water during the subsequent event. Since we are simulating
single events only, the initial condition could not be assumed to be a long-term average for all.
Thus, the initial condition for each event is different and needs to be individually calibrated.
However, field measurements by Ritter (2013) were used as a guideline to approximate the
initial conditions. We also noted that the anoxic IWS zone should be enriched with organics and
carbon leached from woodchips within the layer. Degradation and hydrolysis of woodchips
would supply the zone with higher concentrations of organics in the form of COD (CR, CS, CI).
Thus, a higher concentration of COD in the anoxic IWS zone was imposed in the model. To
improve the specification of initial conditions in future studies, there is a need to carry out a
156
program of continuous monitoring of the water quality that stretches over consecutive events and
the intervening dry periods.
5.4.2.2 Water quality during an event
During an event, influent water infiltrates into the unsaturated zone with a distinct wetting front
as indicated in the time-depth profiles for soil moisture (0, cm 3/cm 3 ) and flow rate (q, cm/hr) as
well as shown by the black arrows indicating linear velocity and red streamlines indicating the
flow path of water particles in Figure 5-4 to Figure 5-9. The flow rate is the fastest in the first
hour after the start of a storm event as shown by the more downward arrows. During this initial
phase of infiltration, the wetting front reaches a depth of about 10-20 cm. This downward
movement of the penetrating water front mobilizes the pollutants along its flow path and forms a
"plume" of pollutants in z-t space. Pollutant plumes bring with them fresh supplies of oxygen
and nutrients (organic and inorganic nitrogen and phosphorus) and change the microenvironment
in which bacteria live and act as an event progresses.
5.4.2.2.1 Nitrification
An occurrence of the nitrification process is evident in Event #2 (Figure 5-5). During Event #2,
from t = 6.7 hrs and z = 1.0 cm to t = 20.7 hrs and z = 37.4 cm (i.e. following the marked
streamline), the oxygen concentration drops by 23% from 7.45 mg/L to 5.72 mg/L. There are
two factors forcing the oxygen level. The first factor is an increase in the oxygen level due to re-
aeration from the atmospheric oxygen. After surface ponding ends at t = 20.7 hrs (Figure 5-10),
the soil moisture saturation decreases from full saturation (0 = 0.3 in the first soil layer, Figure
5-5) and allows for re-aeration to occur deeper into the soil column and not just at the surface.
157
The second factor is a decrease in the oxygen level due to bacterial utilization in cellular
respiration. During the time between t = 6.7 and 20.7 hours during Event #2, the rate of oxygen
consumption by bacteria for nitrification is higher than the rate of re-aeration and the oxygen
concentration decreases.
As part of the two-step nitrification process, NH3 is converted to N02 by Nitrosomonas (XANs)
which is then converted to N03 by Nitrobacter (XAN) (Figure 5-1). Following the pathways
marked in Figure 5-5, NH3 is reduced by 93% from 0.18 mg-N/L to 0.012 mg-N/L which
equates to a NH3 reduction rate of 0.012 mg-N/L/hr. N02 increased by 62%, from 0.058 mg-N/L
to 0.094 mg-N/L which results in a N02 net production rate of 0.0026 mg-N/L/hr. N03 increased
less, by 12%, from 0.38 mg-N/L to 0.42 mg-N/L which results in a N03 production rate of
0.0031 mg-N/L/hr. On average, for all six sampled events, the average reductions in NH3 and 02
are 33% and 22% respectively and the average increases in N02 and N03 are 280% and 5%. The
large build-up in N02 and the slight increase in N03 indicate that the first step of the nitrification
process, which is the conversion of NH3 to N02 as mediated by Nitrosomonas (X4Ns), is
proceeding at a faster rate than the second step which is the conversion of N02 to N03 as
mediated by Nitrobacter(XANb).
5.4.2.2.2 Denitrification
In the anoxic IWS zone, instead of oxygen, N02 and N03 are consumed as the final electron
acceptor during the anaerobic cell growth of denitrifiers. Nitrogen is produced as the end
product. Two instances of denitrification occurring in the anoxic IWS zone are evident in Event
#1 (Figure 5-4). In its nitrogen profile, two regions show extensive nitrogen generation mediated
158
by denitrifying bacteria at very low oxygen levels: between t = 0 and t = 12 hr at a depth of about
50 - 85 cm, as well as between t = 14 to t = 20 hr at a depth of 50 - 60 cm. In the first region (i.e.
following the streamline from t = 0.4 hr and z = 60 cm to t = 9 hr and z = 83 cm as marked by the
pathway in Figure 5-4), N02 decreased by 87% from 0.59 to 0.078 mg/L, N03 decreased by 55%
from 0.40 to 0.18 mg/L, N2 increased by 64% from 1.0 to 1.7 mg/L, and 02 remains low at <
1.48 mg/L. In the second region (i.e. following the streamline from t = 13.8 hr and z = 50 cm to t
= 20 hr and z = 55 cm as marked by the pathway in Figure 5-4), N02 decreased by 98% from
0.15 to 0.0024 mg/L, N03 decreased by 64% from 1.3 to 0.46 mg/L, N2 increased by 88% from
0.98 to 1.8 mg/L, and 02 remains low at < 0.49 mg/L.
In the zone following the streamline marked by the pathway in Figure 5-4 from t = 2 hr and z
30 cm to t = 12 hr and z = 65 cm, oxygen originally from the unsaturated zone is advected
downwards by the infiltrated inflow and reaches the IWS zone. Along this oxygenated pathway,
denitrification is prevented. This implies that even within the anoxic IWS zone where
denitrification is expected to occur consistently, there is a need to have enough basin storage
volume and thus adequate residence time for oxygen depletion first before denitrification can
proceed.
5.4.3 Model response to parameterchoice - a sensitivity analysis
5.4.3.1 Effects of bacterialkineticparameters:p, b
As CW2D is a model developed using parameters adopted from the Activated Sludge Model that
originates from the field of wastewater treatment, some parameters need to be adjusted for
stormwater applications in this study. Composition and stoichiometric parameters in Table 5-2
159
are left unaltered from the original ASM values as they describe universal fractions of nitrogen
and phosphorus in biomass, CR, CS, and CI, which should be applicable to both wastewater and
stormwater treatment. However, effects of some sensitive parameters such as those that describe
bacterial growth (y) and decay (b) dynamics are considered in this sensitivity analysis.
Table 5-8 lists literature values for growth rates of various bacteria species including nitrifiers
and denitrifiers. They serve as a guide for setting high, medium, and low levels of growth rate
(pJH, jUDN, fAns, PANb) and decay rate (bH, bAns, bANb) used in this sensitivity analysis, as shown in
Table 5-9. The literature values for growth rates are often based on controlled lab studies of a
specific bacteria strain growing on specific culture media with abundant nutrients for growth.
However, in stormwater treatment, lower pollutant concentrations undergo treatment. Therefore,
the literature values in Table 5-8 are used as an upper bound for the bacterial growth rate (pUH=
19.2 day-, pDN= 3.84 day-', UAns = PANb= 3.24 day-1) and cell lysis rate (bH= bAns = bANb = 1
day-'). The medium rates are the calibrated and validated values in this study (Table 5-2) and are
used for comparison purposes while the low rates are set to an order of magnitude lower.
The effects of increasing and decreasing bacteria growth and decay rates on various pollutants as
compared to the case of medium rates are conceptually shown in Table 5-10. The effects on
mean outlet concentration and pollutant removal rate for a demonstrative event (Event #1) as
compared to observed values are conceptually shown in Figure 5-13. The time-depth profiles for
modeled concentrations of 02, NH3, N03, NO2, N2 are shown in Figure 5-14. The comparison
between modeled and observed mean outlet concentration is quantified statistically in Table 5-11.
160
Comparing the left-hand column (high rates) with the middle column (medium rates) in Figure
5-14 shows that the onset of denitrification is earlier when biokinetic rates (pH, PDN, PAns, PANb, bH,
bAns, bANb) are high. With both high and medium rates, denitrification occurs from t = 4 hr at z =
10 cm to t = 20 hr at z = 50 cm. During this time, reduction in N03 and N02 as well as
concurrent generation of N2 is seen. Spatially, the upper unsaturated zone is usually where
oxygen-essential processes such as nitrification occur. However, with high kinetic rates, this
zone becomes so deficient in oxygen that denitrifiers (facultative anaerobes) have to utilize N03
and N02 for cellular respiration instead. This occurs because the dissolved oxygen in the
infiltrated water is rapidly used up by intensified biological activities such as the growth of
nitrifiers and other heterotrophs. In these activities (reaction numbers RC 2, 6, and 8), 02, IP, and
NH3 are consumed, as are N03 and N02 by denitrification. IP and NH3 are shown at later hours
(t = 14 to 20 hr at z = 50 cm) to be increasing in concentration due to another high-rate biological
activity: cell lysis (bH, bAns, bANb). Combining the effects brought about by the different processes,
as shown in Table 5-10, the overall effects are an increase in NH3 and IP and a decrease in N03
and N02. These overall effects are also reflected in Figure 5-13 for mean effluent concentration
and conversely for removal rate. Statistically, values of PBIAS are becoming more negative for
NH3 and IP (- 190% and -670%) indicating an over-prediction in the high rate case as compared
to medium rate case scenario. For N03 and N02, values of PBIAS are turning positive (65% and
77%) indicating a reduction in concentration compared to the medium rate case scenario. Figure
5-14 shows that the converse is true for the low biokinetic rates (PH, PDN, PAns, /ANb, bH, bAns, bANb)
case scenario. Lower biological activities have led to insignificant instances of nitrification and
denitrification as well as little changes to the pollutant concentrations and removal rates.
161
5.4.3.2 Effects of initialbacteriaconcentration, Sx
Initial bacteria concentrations (SXH, SXANs, SXANb (mg bacteria/kg soil)) are quantified from copy
numbers of nitrifying genes (ammonia monooxygenase, amoA) and denitrifying genes (nitrite
reductase cytochrome cdl, nirS; nitrite reductase copper-containing enzyme, nirK; nitric oxide
reductase, norB; nitrous oxide reductase, nosZ) from the literature (Table 5-12) and an assumed
gene copy number per genome of one for denitrifiers (Kandeler et al. 2006). Nitrifiers (SXANs,
SxANb) are also assumed to be lower in abundance than heterotrophs (SxH) due to their longer
generation times and lower energy yield from their oxidation reactions.
The effects of different levels of bacteria concentrations (SXH, SXANs, SXANb in Table 5-13) on
mean outlet concentration and pollutant removal rate for a demonstrative event (Event #1) are
compared to observed values in Figure 5-15. The time-depth profiles for modeled concentrations
of 02, NH3, N03, N02, and N2 are shown in Figure 5-16. The comparison between modeled and
observed mean outlet concentration is quantified statistically in Table 5-14. The effects of
different levels of bacteria concentrations are similar to those caused by increasing bacterial
kinetic parameter values as described in the previous section. In particular, earlier onset of
denitrification occurs with high initial bacterial count due to increased N03 and N02 removal.
Low initial bacterial counts imply little biological activity and thus insignificant changes in
pollutant concentrations.
162
5.4.4 Model application: predicting TN removal at various basin configurations as part of
design curve development
Performance of a bioretention basin is often assessed in terms of its ability to remove pollutants
such as total nitrogen (TN). Bioretention basins are typically designed with removal targets of
30-65% for TN (Brisbane City Council 2006; Department of Irrigation and Drainage 2011;
Dierkes et al. 2015; Massachusetts Department of Environmental Protection 2008; Minnesota
Pollution Control Agency 2008; New Hampshire Department of Environmental Services 2008;
Public Utilities Board of Singapore 2014). Currently, design guidelines in Singapore are modeled
after those of temperate areas, primarily Australia (Facility for Advancing Water Biofiltration
2009). Under the regulations of Singapore's Public Utilities Board (2004), basins in Singapore
must be designed to accommodate a critical design flow rate for a 3-month ARI event and water
quality volumes (WQVs) generated from the MUSIC model (Model for Urban Stormwater
Improvement Conceptualization) (Cooperative Research Centre for Catchment Hydrology 2003).
The removal rates promised in the MUSIC-generated design curves are yet to be field-validated
in the tropical context due to a lack of field-scale studies. Instead, the removal rates are assumed
to be achievable given the basin configurations (and WQVs) determined from the MUSIC-
generated design curves.
This study aims to resolve this uncertainty in design curves for the tropics by using the water
quality model developed here to make predictions of pollutant (TN) removal potential with
various basin configurations (and WQVs). However, the model developed in this study has only
been validated against individual events, but single storms do not necessarily show how well a
basin performs in terms of pollutant removal over longer time periods. Therefore, to predict the
163
pollutant removal potential across events of varying sizes and intensities over a half-year period,
an extension from single event simulation to long-term simulations is made.
To evaluate performance over multiple events, a hydrological numerical simulation of 80 storms
captured during a six-month monitoring program by Wang et al. (2016) is coupled with a
complete pollutograph of a selected event from Wang et al. (2017) in a half-year water quality
simulation. Due to the limited availability of observed water quality information for all the
events that occurred during the monitoring period, the same pollutograph of Event #1 in the
study by Wang et al. (2017) is assumed to have occurred during all the events. Event #1 is
selected because its runoff contains the highest EMC (event mean concentration) among the six
sampled events in the study. Design curves generated under this assumption are presumed to be
conservative because if a basin designed according to these curves is able to treat the most
polluted event, it can be assumed to perform equally well in less polluted events.
The design curves resulting from a half-year water quality simulation are shown in Figure 5-17a-
b. The bioretention basin performance in terms of half-year TN removal (TNR, %) is evaluated in
terms of engineering parameters such as detention depth, hd (cm); ratio of drainage area to
bioretention area, R; and saturated hydraulic conductivity, Ks (cm/hr). PEI (performance
efficiency index) (Figure 5-17c-d) is an improvised measure to illustrate the incremental gain in
TNR given a unit increase in either hd, R, or Ks:
dTNR dTNR orEqain51

dTNR Equation 5-11
PEI = or
Ad dR d Ks
With the current basin configuration, the estimated TNR is 28% which falls short of the target rate
of 45%. Comparing the estimated long-term TNR to that of the six individual events as shown in
164
Table 5-7, TNR = 28% falls within the range of 5 - 84% seen in the individual events, although
28% is closer to the TNR values of the large events (5 - 29%) than of the small events (46 - 84%).
5.4.4.1 Detention depth, hd
Figure 5-17a shows that at the current basin area (R = 70), TNR increases from 26 % to 36% as hd
increases from 10 to 80 cm. This increase in TNR is non-linear with a higher initial increase seen
at lower hd and gradually slower rate of increase at higher hd values. The non-linear slope is
captured in the PEI plot (Figure 5-17c) in which a higher PEI = 0.39 is achieved at hd = 10 Cm
than at hd = 75 cm (PEI= 0.04). This trend indicates that the rate of improvement in TNR drops
significantly with increases in hd beyond a certain point. At R = 70, this point occurs at about hd
= 30 cm, beyond which the benefit in improving TNR by increasing hd becomes minimal. This
tendency is consistently observed at other basin area ratios (R = 35, 23.3, 17.5) as well.
As relative basin area increases from R = 70 to R = 35, 23.3, and 17.5, the corresponding gain in
TNR at any particular hd reduces, as indicated by a higher jump in PEI from 0.39 to 0.66 for R =
70 to R = 35 than the slight increase in PEI from 0.80 to 0.91 for R = 23.3 to R = 17.5 (Figure
5-17c). This shows that at the four values of R tested, increasing basin area beyond R = 23.3
would result in only small incremental gain in basin performance (PEI). Such trend is also
evident from the decreasing upward shift in the TNR curves as R decreases sequentially in Figure
5-17a.
The incremental gain in TNR resulting from an increase in detention depth hd is greater for larger
basin area (A). For instance, a similar increase in hd from 10 to 80 cm results in an increase of
165
10%, 16%, 19%, and 19% in TNR for basin area ratios of R = 70, 35, 23.3, 17.5 respectively
(Figure 5-17a). Therefore, a smaller basin area (A) needs disproportionately higher detention
depth hd to achieve the same current TN removal target of 45% in Singapore (Public Utilities
Board of Singapore 2014). Design curves in Figure 5-17a show that for decreasing basin area
ratios (R = 17.5, 23.3, and 35.0), the detention depth required to achieve TNR = 45% are hd = 12,
18, and 45 cm respectively. The corresponding values of WQV range from 240 to 360 m 3 and
those of WQD (water quality depth) from14 to 22 mm. At the current basin area (R = 70), the
basin is projected to be unable to meet the TN removal target of 45% with hd < 80 cm (Figure
5-17a). Although different basin surface storage configurations (hd x A) could equally meet a TN
removal target rate of 45%, extra subsurface volume created along with a larger basin surface
area contributes more towards TN removal than the extra surface volume created by a greater
detention depth. This is expected since a larger basin area inherently involves a larger subsurface
volume and treatment capacity. Since the soil filter media is the main driver behind nitrogen
removal through denitrification over the inter-event period, the larger subsurface storage volume
in larger basin area induces a non-linear increase in TN removal compared to a smaller basin
area.
5.4.4.2 Saturatedhydraulic conductivity, Ks
Figure 5-17b shows the effect of varying saturated hydraulic conductivity, Ks, on TN removal,
TNR. As Ks increases from 0.1 to 5 cm/hr, TNR increases sharply, indicating a gain in TN removal.
However, TNR decreases at first rapidly from Ks = 5 to 10 cm/hr, then gradually from Ks = 10 to
60 cm/hr. Therefore, the best basin performance in terms of TN removal is seen when Ks = 5
cm/hr. Poorer performance at low Ks could be explained by the reduced volume of infiltrated and
166
hence soil-treated stormwater. Poor removal at high Ks could be explained by the rapid drainage
of infiltrated water and hence lower hydraulic residence time within the system. Infiltrated water
would be washed out from the system without receiving complete treatment by nitrification and
denitrification. The superior performance at Ks = 5 cm/hr is also reflected in the PEIplot (Figure
5-17d). PEI is most negative at Ks = 5 cm/hr, indicating a reduction in basin performance if
otherwise. Therefore, it is not recommended from the water quality point of view to increase Ks
beyond 5 cm/hr, even if that allows more water to be infiltrated.
5.4.4.3 Recommended basin design values
Combining the results from the above analysis of design curves which evaluate bioretention
basin TN removal (TNR) potential at various basin configurations (detention depth hd, catchment
area to basin area ratio R, and saturated hydraulic conductivity Ks), some design
recommendations can be made. The incremental benefit of TNR drops as hd increases beyond hd
= 30 cm and R decreases beyond R= 23.3 such that further changes in these two engineering
parameters provide little improvement. Before reaching these values, there is a high incentive to
increase them if space, design rules, and other practical concerns allow. For Ks, a maximum
value of TNR is seen for Ks = 5 cm/hr. This value reflects the best balance between maximizing
the volume of stormwater infiltrated and ensuring adequate treatment. It is important to allow
sufficient hydraulic residence time for the infiltrated amount to receive biochemical treatment for
nitrogen removal before it is flushed out of the system. The design curves presented in Figure
5-17a-b are appropriate for use by basin designers. For instance, to achieve TNR = 45% at the
current soil hydraulic conductivity of Ks = 1 cm/hr, the WQV required to be retained ranges from
167
240 - 360 m 3 with equivalent WQD (water quality depth) of 14 - 22 mm depending on basin area
A and detention depth hd.
5.5 Conclusion
A numerical water quality model has been built based on the CW2D model for wetlands to
model the subsurface bacteria-mediated biochemical processes occurring in bioretention basins.
1. This study has demonstrated that the mathematical formulations originating from the
activated sludge model (ASM) used in wastewater treatment are also applicable for modeling
the role that bacterial growth and decay play in stormwater treatment in bioretention basins.
However, some biokinetic parameters need to be calibrated for usage in bioretention basins.
The model was successfully calibrated and independently validated against field data
collected by (Wang et al. 2017).
2. Simulation results based on calibrated and validated parameter values show that the high
build-up of N02 and slight increase in N03 during an event indicate that the first step of the
nitrification process (the conversion of NH3 to N02 mediated by Nitrosomonas (XANs)) is
proceeding at a faster rate than the second step (the conversion of N02 to N03 mediated by
Nitrobacter(XANb)).
3. A sensitivity analysis of bacterial kinetic parameters (PH, PDN, UAns, /JANb, bH, bAns, bANb) and
initial bacteria concentrations (SxH, SXANs, SXANb) indicates that the model is sensitive to both.
For instance, the onset of denitrification will be earlier when bacterial kinetic rates or initial
bacteria concentrations are higher. These parameter settings need to be better defined to
ensure accurate simulations of real scenarios. To do that, future studies need to focus on
168
quantifying the depth-dependent microbiological activities occurring over time in the
subsurface of bioretention basins.
4. Well-defined initial conditions are important for an accurate short-term event-based
simulation. To improve the calibration of initial pollutant concentrations across different soil
depths, there is a need to carry out continuous monitoring of the water quality that stretches
over consecutive events. A more intensive pre- and post-event sampling program over a
longer time duration would give a better representation of the initial conditions at the start of
each event.
5. The developed water quality model was used to evaluate total nitrogen removal (TNR) with
various bioretention basin configurations (detention depth hd, catchment area to basin area
ratio R, and saturated hydraulic conductivity Ks) over a six-month simulation period. The
simulation results show that the rate of improvement of TNR decreases as hd and R increase
beyond hd = 30 cm and R= 23.3 such that further increase in these two engineering
parameters become less cost-effective. For Ks, a maximum value of TNR is found for Ks = 5
cm/hr. With the current basin configuration, the estimated TNR is 228% which falls short of a
target rate of 45%. The design curves developed using the model are useful as a reference for
future bioretention basin designs in the tropics.
This research was funded by the National Research Foundation, Prime Minister's Office,
Singapore through the Singapore MIT Alliance for Research and Technology's Center for
Environmental Sensing and Modeling research program.
169
CS, CI, CR:
Hydrolysis (RC 1) XANs: Aerobic cell XANb: Aerobic cell XH: Anaerobic cell
Organic trogen XH, XANs, XANb Ammonia growth Nitrite growth Nitrate growth: Denitrification
(oN Cell lysis (RC 5,7,9) Nitrification (RC 6) nitrogen Nitrification (RC 8) nitrogen (RC 3)
CS, CI, CR (NH+ - N) (N02 - N) ' (N03--N) _ N2
+
+ P04 - + P04
(COD) + P04 -
-
+02 +02
+ P04 - + PO43-
+ 02
XH: Aerobic cell XH: Anaerobic cell growth:

growth (RC 2) Denitrification (RC 4)
Figure 5-1 A conceptual model for nitrogen removal: summarizing the bacteria species and chemicals involved in the nine reaction
processes (identified as RC 1 through RC 9) that govern nitrogen species transformation from organic nitrogen to ammonia, to nitrite,
to nitrate, and to molecular nitrogen.
Table 5-1. Summary of different bacteria species in process-based model
Electron Reaction
Electron donor acceptor Product number in Zone of primary residence
the model
Heterotrophs (in general), XH Organic matter (COD) 02 Biomass RC2 Both unsaturated and saturated zones
Autotrophic Nitrosomonas,XAN NH4 + 02 NO2- RC6 Unsaturated zone
Autotrophic Nitrobacter,XAN NO 2 - 02 NO 3- RC8 Unsaturated zone
Heterotrophs (denitrifiers growing on NO 3-), XH Organic matter (COD) NO 3- N2 RC3 Saturated zone
Heterotrophs (denitrifiers growing on NO 2 ), XH Organic matter (COD) NO 2- N2 RC4 Saturated zone
170
a)
RECHARGE
hydrological model
Measured influent Initial

Outputs:
concentration: concentration:
O(z, t), q(z, t)
Cij (z, t) C,it('it(z, t)
CW2D Monod-type biokinetic model for water quality
Modeled effluent
concentration:
c- posetz, 0
b)
Compute virtual concentration before undergoing biokinetic reactions:

b.
C-(zt-1)0(zt-1)dz + Cj(z-,t-1)q(z-1,t-1)dt
Cire(Z, 0) = ! O(z,t)dz+ q(z,t)dt ; i = 1,.., 12
Solve the set of zero-order Monod-type biokinetic equations:

(z, t) = (Zt) (Vj)( R Cj; i = 1,..,12; j = 1..,9 R
k dC =
dt I
R j= 4
z = z +1
I Compute the mass balance within a single-cell CSTR:

Ci,post(z, t) = Ci,pre(Z, t) +d Ci(z, t)1R ; i = ],..,12
Z = Zboilom
T
Re-initialize initial concentration of the next time step:
Ci(z, t + 1) = Cipost(z, t); i = 1,..,12
t =t+1
Figure 5-2 a) General scheme and b) numerical implementation for the coupling of RECHARGE
hydrological model and CW2D reactive biokinetic model.
171
Table 5-2. Model parameters
Symbols Parameter Unit Value Reference

Kinetic parameters (at 29'C)
Hydrolysis
Kh Hydrolysis rate constant 1/day 4.22 Henze et al. 1995
Kx Saturation/inhibition coefficient for hydrolysis mgCOD-CS / mgCOD-BM 0.05 Langergraber 2005
Heterotrophic bacteria (aerobic growth)
PH Maximum aerobic growth rate on CR 1/day 1.5 Henze et al. 1995, calibrated
bH Rate constant for lysis 1/day 0.1 Henze et al. 1995, calibrated
KHet 0 2 Saturation/inhibition coefficient for 02 mgo2/L 0.2 Henze et al. 1995
KHetCR Saturation/inhibition coefficient for substrate mgCOD-CR/L 2 Henze et al. 1995
KHetNH4 Saturation/inhibition coefficient for NH 4 (nutrient) mgNH4-N/L 0.05 Henze et al. 1995
KHetp Saturation/inhibition coefficient for P mgip/L 0.01 Henze et al. 1995
Heterotrophic bacteria (denitrification)
PDN Maximum denitrification rate 1/day 1 Henze et al. 1995, calibrated
KDN02 Saturation/inhibition coefficient for 02 mgo 2 /L 0.2 Henze et al. 1995
KDNNO3 Saturation/inhibition coefficient for NO 3 mgNo3-N/L 0.5 Henze et al. 1995
KDNNO2 Saturation/inhibition coefficient for NO 2 mgNO2-N/L 0.5 Henze et al. 1995
KDNCR Saturation/inhibition coefficient for substrate mgCOD-CR/L 4 Henze et al. 1995
KDNNH4 Saturation/inhibition coefficient for NH4 mgNH4-N/L 0.05 Henze et al. 1995
KDNip Saturation/inhibition coefficient for P mgip/L 0.01 Henze et al. 1995
Ammonia oxidizing bacteria (Nitrosomonas)
PANs Maximum aerobic growth rate on NH 4-N 1/day 0.75 Langergraber 2006, calibrated
bANs Rate constant for lysis 1/day 0.1 Langergraber 2006, calibrated
KANsO2 Saturation/inhibition coefficient for 02 mgo 2 /L 1 Langergraber 2006
KANSNH4 Saturation/inhibition coefficient for NH 4 mgNH4-N/L 0.07 Langergraber 2005
KANSip Saturation/inhibition coefficient for P mgip/L 0.01 Henze et al. 1995
Nitrite oxidizing bacteria (Nitrobacter)
PANb Maximum aerobic growth rate on N0 2-N 1/day 0.75 Langergraber 2006, calibrated
bANb Rate constant for lysis 1/day 0.1 Langergraber 2006, calibrated
KANb 02 Saturation/inhibition coefficient for 02 mgo2/L 0.1 Langergraber 2006
KANbNO2 Saturation/inhibition coefficient for NO 2 mgNO2-N/L 0.1 Langergraber 2006
KANbNH4 Saturation/inhibition coefficient for NH4 mgNH4-N/L 0.05 Henze et al. 1995
KANbip Saturation/inhibition coefficient for P mgip/L 0.01 Henze et al. 1995
172
Composition parameters
iNCR Nitrogen content of CR mgN / mgCOD-CR 0.03 Henze et al. 1995
iNCS Nitrogen content of CS mgN / mgCOD-CS 0.04 Henze et al. 1995
'NCI Nitrogen content of CI mgN / mgCOD-Cl 0.01 Henze et al. 1995
iNBM Nitrogen content of biomass mgN / mgCOD-BM 0.07 Henze et al. 1995
iPCR Phosphorus content of CR mgP / mgCOD-CR 0.01 Henze et al. 1995
ipcs Phosphorus content of CS mgp / mgCOD-CS 0.01 Henze et al. 1995
iPC, Phosphorus content of CI mgp / mgCOD-CI 0.01 Henze et al. 1995
t
qm Phosphorus content of biomass mgp / mgCOD-BM 0.02 Henze et al. 1995
Stoichiometric parameters
fHydci Fraction of CI generated in hydrolysis mgCOD-CI / mgCOD-CS 0 Henze et al. 1995
fBMCR Fraction of CR generated in biomass lysis mgCOD-CR / mgCOD-BM 0.05 Henze et al. 1995
fBMCI Fraction of CI generated in biomass lysis mgCoD-CI mgCOD-BM 0.1 Henze et al. 1995
YN Yield coefficient for heterotrophs mgCOD-BM mgCOD-CR 0.63 Henze et al. 1995
ANS Yield coefficient for Nitrosomonas mgCOD-BM / mgNH4-N 0.24 Langergraber 2006
YANb Yield coefficient for Nitrobacter mgCOD-BM / mgNO2-N 0.24 Langergraber 2006
Other parameters
0
2-sat Saturation concentration of oxygen at 29'C mgo2/L 7.67 Luckner and Schestakow 1991
Kaer-O2 Oxygen reaeration rate 1/hr 25 McBride and Tanner 2000
EaKh Activation energy for K in the Arrhenius equation J/mol 27,980 Henze et al. 1995
Ea x Activation energy for K, in the Arrhenius equation J/mol -53,000 Langergraber 2007
EaKANs-NH4 Activation energy for KANs-NH4 in the Arrhenius equation J/mol -160,000 Langergraber 2007
173
Table 5-3. The overall reaction rate term, dc-
' [mgi/(Lxh)], following the Gujer matrix structure for each of the 12 components i (02,
CR, CS, CI, XH, XANs, XANb,NH3, N02, N03,N2, and IP).
dC0 :! 1 )(PC) C )2 CCR -) ( CNH 4 ()( IP ( S
)
dt YH) Ket 0 2 + C02 KHetCR + CCR KHetNH4 CNH4 KHet~p IP
ANS ANs KANsO + KANSNH + CNH4 (KANS IP)

)
02 XANS
+(A -Nb (KAANbNbN CN0 NH4 KANbC0 (X ANb) + ( 6 sat - O)KaeO2 (Co 2 sat
N2) (K NH 4 +CH)(K~j I
C02) YAb KA~b0 + C02KANbN02 +
- C02
)
Equation 5-12
CCs
dC = (- H Kb) CXH C0 2 CCR CNH
4 ip XH
dt Kx+ YH )CS (KHeto 2 + C02 KHetCR + KHetNH4
CCR+ CNH4 KHe + CIP
CX H
YH
(IDN) KDN 0 2
KDN 0 2 + C0 2
CNO 3 KDNNO2
KDNNO3 + CNo ) \KDNNO2
CCR
CNo 2) \KDNCR +
( CCRI
CNH4 ip
lKDNNH4 + CNH 4 ) \ KDNip + Eip)
SXH
(
+ PD)KDN0
DN DC
KDN
CN0
CDNN
(C N4Ci S+fIMR
NO 2 N
(KzKDNCRC
'CXH
CCR /KDNNH NH4 KDNp CIP sXH BMCR)HCXH)
Y D02+ C02 NN CN 4
!
+ (fBMCR)(bANSCXANs) (JBMCR)(bANbCXANb)
Equation 5-13
d C C/ CCCs
XH
= E(-Kh)K C CXH +(1-BMCR - fBMCI (bHCXH) + ( ~BMCR fBMCI )(bANsCXANs) + (1 ~ BMCR ~BMCI (bANbCXANb)
CX H
Equation 5-14
dC,=d(fHydC
CS ) (Kh) (
1 CCS CXH + (fBMc, ) (bHCXH) + (fBMC, )(bANsCXANS) + (fBMCI )(bANbCXANb)
+ CX-
Equation 5-15
EqaKi
174
dCXH (H C0 2 ( CR eNH 4 C ) K etjXHp
dt (KHet2 + CO,) \KHetCR + cCR) (KHetNH4 + INH4 KHetp (
)
+ (PDN) KDN 02 CNO, KDNNO2 CCR CNH
4 ip XH
\KDN 0 2 + C 02 KDNO 3 + CNO 3)KDNNO2 + CNO 2 I KDNCR + CCR \KDNNH 4 + CNH4 KDNIp C CIP (
)
+ (MDN) KDN 0 2 CN0 2 CCR CNH4 i ( SX) bC
DDN + NO 2 KDNN CNO2) KDNCR CCR /KDNNH4 CNH 4 KDNIp +IP X
02
Equation 5-16
dCXANs =ANs A 02 KAsCNH 4 Ci p SXANS - (bANS CXANs)
dt \KANSQ, + C0 2 ] KANSNH 4 + CNH 4 ) KANSIp + CjpJ 05As
Equation 5-17
dCXANb
d =(Mb)KANb 0
CANb
C
K 02 C2
KANbN02
NO 2
CNO?
)
KANbNH+
(NH4 CN
Kb XANb ANbCXANb)
b)bl + (bAPbXA19
Equation 5-18
dCNH 4 [iNCS fHydCDINCR (fHydC 1 )iNCI](Kh) CCS CXH
d Kx + CX H
+ (+NCR - NBM) (PH)

\C KHet 0
C02
+i02
CCR
HetCR
(
C R \KHetNH
CNH4
NH 4
'\(
K ) (_______
(XH)
+ INCR
iNBM (PDN) 02
(KDN 0 D2 +C2 KDNNo3NO+ CNo3
3 KDNNO
(KDNN0 + CN02
2 + CCR)KDN4+CH CNH4
)KDNCR CCR Cip
DNp+IP9 (LXH)
NCR ~ iNBM) (PDN)

(HKD0 DDNN0
+ C0 2
2 KDN( CN O )(
KDNNo2 C2KD
CC R
CCR) (KDNH
N CNH H4 CN ip p SXH)
)(
+ [iNBM - (1 fBMCR - fBMCI) NCS - (fBMCR)iNCR - (fBMCI)iNCI (bH CXH)
+ ( Y - iNBM) (lPANs) (
2ANs ANs2
CN~0 02
+C02
CNH 4
KANSNH + CNH 4
Cjp
KANS+
(P SAN
CIP 0XANs
1
" [iNBM - (1 - fBMCR - fBMCI)'NCS (BMCR) NCR - (fBMCI)NCI](bANsCXANs)
" (-iNBM)GlANb) C02

A 0 2 +C0
(CN
KAb2
02 NH4 CN 4 (KANb2 ) (I SXANb)
[iNBM - (1 - fBMCR - fBMCI) iNCS - (fBMCR)LNCR - (fBMCI)iNCI ](bANbCXANb)
Equation 5-19
175
dCNO 2 ~ ( YH KDNO ( CNo 2 CCR CNH4 ( ip
- (PDN) ( SXH)
dt 1.71YH KDN 0 + C0 2 ) \KDNNo2 + CNOZJ KDNCR CCR KDNNH4 CNH 4 ! \ KDN 1 p CIP
(As
ANs
S0 z +02
) CN
\KANSNH4 +4 NH
C
KANSIp CIP
) XANs) + (fBMCR) (bH CXH)
502 C -NHO (N4 I
+ (
YANb
)IANb)
ANb0
(K
(KANbo,
C 02
2 + CO,
) CN 2
KANbN02+
C_ 2
07
(CNH
KANbNH4 + CNH4)
4)(Kbp+
KANbip + CIP
C)(pSXANb)
1 A~
Equation 5-20
dC
dt
~H
2.86YH (DN) KDN 0 2

KDNY_
0 (02 CNo3
KDNNO N0 3 KDNN 2
KDNNO2
+ NO 2 KDNCR + CCR
CCR CNH 4
KDNNH4 + NH4
_ ip_____
KDNIp + IP
YANb
K( 1
b) KANb
+ C2
02 + 02 )
CN2
KANbNOZ+ CNOz) (KANbNH4
NH4
+ NH 4 ) \KANbp
ip 'K
)IP CN+
sxANb)
Equation 5-21
dC
dt 2.H DN
K +0 2 )( CNO 3
C
KDNN02 CNH4 C
KDNCR + CCR
K H4
KDNNH 4 + CNH4
)p(
0 2 02)\ KDNN 3 DNNO2 CN2 KDNIP + XH
1-YHKDNO, CN02 CCR CNH4 cip
1.71YH
y) (PDN)
KDNO
CNH KDNIp +jp SXH)
2 + C02 KDNNO2 + CN0z KDNCR +CR KDNNH4 NH4 IP
Equation 5-22
176
CCs
dCp
(1 -
~ H dC 1
JHydLCI) P C
PCR -x
(~ H y dCji PCI EK h) C
+fy~ ~s C XN c H)
dt PC ~ 1
K+CXH
+ iPCR - iPBM) (
YH
KH)(K C
K~et 0 2 + 02 C02 ]
K~ ( CCR K~e(NCNH4
KHetCR + CCR! \\KHetNH 4 + CNH4 I
(7- CI
KHetlp + CIP0
(XH
+ iPCR
KDNNO CCR - NH4 - IP

CNO 3 - LSH
)
PBM) (DN) KDN02
D KDN 0 2 + C 02 KDNNO3 + CNO 3 ) KDNNO2 + CNo2) \KDNCR + CCR) (KDNNH4 + CNH 4 ) ( KDNIp + Cip. X
+ 1PCR - 1PBM)QJDN) KDNO

CK +2 c02 J\KDNNO CNO 2
+ CNO2 KDNCR
CCR
CCR
CNH
\KDNNH 4 +4 CNH 4 I \ KD
ip ( C
+ (1 (fBMCR) PCR (bH NXH+
(
[iPBM - - fBMCR - fBMCI)iPCS - - (fBMCI)iPCI)
"_____M)______C2 CNH4 _CIP )( As
BM)ANS KANso 4 0 2) KANSNH 4 CNH 4 )( KANSI IP sXANs
+ IPBM (1
)(.SXANb)
- - JBMCR - fBMCI PCS (fBMCR)IPCR - (JBMCI)iPCI) I(ANsCXANS)
+ (-ipBM) (PAN) (C
BKANbO+
0 2 "-
CO 02
( KANbNO
N2
+ CNO 2 KANbNH
CNH, _
CNH 4I
(
( KANbj
i
+ CIP
(p
X
+ I[PBM - (1 - fBMCR - fBMCi )PCS - (fBMCR) PCR - (fBMCI)iPCI bANbCXANb)

Equation 5-23
177
Table 5-4. Initial concentrations within the soil media
Symbol Parameter Value Source
Unsaturated zone Transition zone Anoxic zone
#1 #2 #3 #1 #2 #3 #1 #2 #3
CO2-init Initial 02 conc (mg/L) 7.67 7.67 7.67 4 4 4 1.55-1.75 1.55-1.75 1.55-1.75 Ritter 2013
CCR-init Initial CR conc(mg/L) 10 10 10 10 10 10 15 15 15
Ccs-init Initial CS conc(mg/L) 1 1 1 1 1 1 1 1 1
Cc-init Initial CI conc(mg/L) 1 1 1 1 1 1 1 1 1
SXH-init Initial heterotroph conc 10 10 10 10 10 10 20 20 20 Chen et al. 2013,
(mg/kg) calibrated
SXANs-init
SX~-nt
Inta irSMnSChen
Initial nitrosomonas 0.1 0.1 0.1 0.01 0.01 0.01 0 0 0 Ce etalrated
al. 2013,
conc (mg/kg) calibrated
SXANb-init Initial nitrobacterconc 0.1 0.1 0.1 0.01 0.01 0.01 0 0 0 Chen et al. 2013,
(mg/kg) calibrated
CNH4-init Initial NH 4 conc 0.12 0.2 0.1 0.12 0.2 0.1 1 0.2 0.1 Ritter 2013
(mg/L)
CNO3-init Initial NO 3 conc 0.1 0.01 0.01 0.1 0.01 0.01 0.4 0.01 0.01 Ritter 2013
(mg/L)
CN02-init Initial NO 2 conc 0.04 0.01 0.005 0.04 0.01 0.005 0.6 0.01 0.005 Ritter 2013
(mg/L)
CN2-init Initial N 2 conc (mg/L) 0 0 0 0 0 0 1 0.6 0.6
CI>-init Initial IP conc (mg/L) 0.03 0.01 0.001 0.03 0.01 0.001 0.03 0.02 0.001
178
Table 5-4. (continued) Initial concentrations within the soil media
Symbol Parameter Value Source
Unsaturated zone Transition zone Anoxic zone
#4 #5 #6 #4 #5 #6 #4 #5 #6
Co2-init Initial 02 conc (mg/L) 7.67 7.67 7.67 4 4 4 1.55-1.75 1.55-1.75 1.55-1.75 Ritter 2013
CCR-init Initial CR conc(mg/L) 10 10 10 10 10 10 15 15 15
Ccs-init Initial CS conc(mg/L) I 1 1 1 1 1 1 1 1
Cciiit Initial CI conc(mg/L) I 1 1 1 1 1 1 1 1
Initial heterotroph conc 10 10 10 10 10 10 20 20 20 Chen et al.
SxH-init
(mg/kg) 2013, calibrated
Initial nitrosomonas Chen et al.
SXANs-init 0.1 0.1 0.1 0.01 0.01 0.01
conc (mg/kg) 2013, calibrated
Initial nitrobacterconc Chen et al.
SXANb-init 0.1 0.1 0.1 0.01 0.01 0.01
(mg/kg) 2013, calibrated
CNH4-init Initial NH 4 conc (mg/L) 0.0015 0.0015 0.0015 0.0015 0.0015 0.0015 0. 0015 0.0015 0.0015 Ritter 2013
CNO3-init Initial NO 3 conc (mg/L) 0.0015 0.05 0.01 0.0015 0.01 0.005 0. 0015 0.15 0.01 Ritter 2013
CNo2-init Initial NO 2 conc (mg/L) 0.0015 0.25 0.3 0.0015 0.01 0.05 0. 0015 0.04 0.05 Ritter 2013
CN2-init Initial N 2 conc (mg/L) 0 0 0 0 0 0 0.6 0.6 0.6
CP-init Initial IP conc (mg/L) 0.001 0.001 0.01 0.001 0.001 0.01 0.001 0.001 0.01
179
Table 5-5. Influent concentrations
Symbol Parameter Unit Value Source

Co2-inf Influent 02 concentration mg/L 7.67 for influent and ponded water
CCR-inf Influent CR concentration mg/L 80% of influent COD concentration measured
Ccs-inf Influent CS concentration mg/L 10% of influent COD concentration measured
CCL-nf Influent CI concentration mg/L 10% of influent COD concentration measured
SXH-inf Influent heterotrophic bacteria concentration (including denitrifiers) mg/kg 0
SXANs-inf Influent autotrophic bacteria concentration (nitrosomonas) mg/kg 0
SXANb-inf Influent autotrophic bacteria concentration (nitrobacter) mg/kg 0
CNH4-inf Influent NH 4 concentration mg/L Values from the pollutograph measured
CNO3-inf Influent NO 3 concentration mg/L Values from the pollutograph measured
CNO2-inf Influent NO 2 concentration mg/L Values from the pollutograph measured
CN2-inf Influent N 2 concentration mg/L 1 for influent and ponded water
Cw-inf Influent IP concentration mg/L Values from the pollutograph measured
180
Mean effluent concentration for calibration (mg/L) Mean effluent concentration for validation (mg/L)
2 2-
1.8 1.8-
'
1.6 0 1.6
-
-
-
1.4 1.4-
-
1.2 (D 1.2-
(D --- bisector --- bisector
-
'a 1 * NH
1 * NH
-
3 0 3
0
F 0.8 N * NO E 0.8 * NO
3
-
3
-
* NO 2 0.6 - * NO 2
0.6
'
* IP * IP
0.4 0.4
- -"
* TN - * TN
-
0.2 * ON 0.2 *1' * ON
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
observed observed
Removal rate for calibration (%) Removal rate for validation (%)
200- 200-
150- 150-
100- 100-
or
19
o 50 50-
-
i- 'YE U-'
7> 0 0-
-
'V - -- bisector -0 ,U-

E-50-
U ,I - bisector
- -
,' * NH -50- * NH3
U 3 0
-100- * NO
.
U,-' E -100-
3 * NO
3
-150 * NO * NO
2 -150-
.
2
-200 *IP * IP
-200-
* TN U,-' * TN
-250- * ON -250- * ON
-300' ' '2 ' 1 '5 -300
-300 -250 -200 -150 -100 -50 0 50 100 150 200 -300-250-200-150-100 -50 0 50 100 150 200
observed observed
Figure 5-3. Results from model calibration and validation, showing observed against modeled values of event mean outlet
concentration and removal rate.
181
Table 5-6. Model evaluation statistics for the results from calibration and validation: coefficient of determination (R2 ), Nash-Sutcliffe
efficiency (NSE), mean absolute error (MAE, mg/L), relative root-mean-square-error (RRMSE), RMSE-observations standard
deviation ratio (RSR), and percent bias (PBIAS, %) for the average outlet concentration
Calibration (Events #4, #1, #6) Validation (Events #2, #5, #3)
TN ON NH3 N03 N02 IP TN ON NH3 N03 N02 IP
R2 1.00 0.80 1.00 1.00 1.00 0.90 0.37 0.37 0.94 0.99 0.99 0.93
NSE 0.94 -4.29 0.84 1.00 0.81 0.84 -0.89 -7.69 0.76 0.01 0.97 0.89
MAE 0.12 0.21 0.07 0.01 0.02 0.00 0.10 0.21 0.03 0.02 0.00 0.00
RRMSE 0.20 1.11 0.54 0.06 0.39 0.18 0.26 1.20 0.38 1.17 0.23 0.15
RSR 0.25 2.30 0.40 0.05 0.44 0.40 1.37 2.95 0.49 1.00 0.17 0.33
PBIAS -16 -76 -34 -1 -26 -8 -8 -115 18 -76 0 6
182
Depth (cm) Depth (cm)
co -4 N x co -4 cn 01 AC wA N) -
0 0 0 0 ox
x
CO
3
IN)
C)
-4c J )c
-I,
Depth (cm) Depth (cm) Depth (cm)

So o o o o o o 0 0 0 C> 0 C5o 0 0 8r
3.
- CD
00 3
- )
0 0
0 0) 0 0
rl z
Deptm (cm) Depth (cm) Depth (cm)

co -4 a) L1 A w N) - co -4 0) (iA CA w ) -. co -Q mD (CP A. CAw )
o o 0o
11
o( 0e z
0, ri0
z
-4
Af,
CD
o 0 P 0 0 o 0 0
C:?
C)
0 0 0 0 0
0 0 03 C> CD )
CI (mg/L), Inert COD, Event #1
14
10
100 12
20 20
30 80 -30 10
40 8
60 40
50 50 6
40
60 4
20 70
2
80
0
0 2 4 1U 1z 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Time (hr) Time (hr)
4lture, 0, Event #1
0.4
4.5
10
4
20 0.35
3.5
-30
3
9 40 0.3
2.5
CL I
050 2
60 0.25 1.5
70 1
-
80 0.2 0.5
Li 0
0 2 4 6 8 10 12 14 16 18 20
Time (hr) Time (hr)
Figure 5-4. Time-depth profiles of various water quality parameters: 02, NH3-N, N02-N, N03-N, N2, IP, CR, CS, CI (mg/L), bacteria,
XH, XANs, XANb (mg/kg), soil moisture, 0 (cm 3/cm 3), and flow rate, q (cm/hr) for Event #1.
184
co __ 0) U) -t W. IN) - )o -4 CD 4 w N
x 0
0 CD 0C: 0 0 0
1zo
N)
co i, 4
=~! el.
2
44
co
y rl 'r
w L 41
III,t/111
co Y,,
C) 0
-4-
Cm m
Depth (cm) 2o Depth (cm)

c -14 C TI U) 1. wON)
00 04 0m e.i 00C 0 0 0 010
0
Pi
z 0
Wa
z
101 00
z
-4-
(J)
~1~/ ,0 -. 0
0 0 0 0 0 0 N)
0D '0c 0 - NJ a)N ) 4, (n

co -4 CD U1 0. CW N) 9R 'A 0) C Et 0. C. N - co a) CP Z w 3
0
0
z
0
z
0
-4
2
G (D
00Y~ ni
0
o ~ i / L-,~
0 0C>
w 46 0co
0 0 0 0 0 9
CR (mg/L) Readily degradable COD, Event
#
45 5.5
I
10 40 5
-
20 35 -4 4.5
-30 30 ju 30 4
25 40 3 j40 3.5
40,
C.
3
50 20 50 2 0 50
2.5
60 15 60 60
2
10 70 1 70
70
15
5
8() 80 0
80
0 2 4 6 8 10 12 14 16 18 20 22 0 2 4 6 8 10 12 14 16 18 20 22
U 2 14 16 18 20 22
Sol I mnlaam. I vmnt 12
-0.4
10 5
20 0.35 2.)
4
30 -30
0
40
0.3 E40 3
c50- 050
2
60
0.25
70- I
0.2
80
0
0 2 4 6 8 10 12 14 16 18 20 22 0 2 4 6 8 10 12 14 16 18 20 22
Time (hr) Time (hr)
3 3
XH, XANs, XANb (mg/kg), soil moisture, 0 (cm /cm ), and flow rate, q (cm/hr) for Event #2.
186
02 (mg/L), Event #3 0
I0.2
0.6 -0.4
10 -7 10 0.35
6 0.5
20~ 20
-0.3
30 5 -30 0.4
0.25
40 4
0 03 0.2
CL
)5 -5~- 50
3
0.15
60
0.1
70 70 01
0.05
80 80
0
0 2 4 6 8 10 12 14 16 18 20 22 0 2 4 6 8 10 12 14 16 18 20 22
IP (mg/L), Event #3
0
-1 0.045
10 3 10
*0.9 -04
20 -~ 0 98 20 -25 20
0.035
30 07 30 2-S30 0.03
40 0.6 40 40 0.025
0.5
50 50 1.5 0 0.02
0.4
60 1 0.015
03
0.2 0.01
70 0,5
80
0.1 0.005
I I 0
0 2 4 6 8 10 12 14 16 18 20 22 0 2 4 6 8 10 12 14 16 18 20 22 0 2 4 b 3 10 12 14 lb 16 20 22
XANb (mg/kg), Nitrobacter, Event #3 XANs (mg/kg), Nitrosomonas, Event #3
22 0 ________________ 0
- - - - - 0.12
10 10--- 0.12- - -
-
20 -_
20 20 0.1 20 --- 0.1
-
30 18 30 0.08 30 0.08
-
r-40
16
-40
0.06 0 0.06
------- (0
50 0)5
14 0.04
60. >-Z7~-'~--- 60 60 0.04
70--- 12 0.02 70 0.02
80- - 80
-
10 0 0
0 2 4 6 8 10 12 14 16 18 20 22 0 2 4 6 8 10 12 14 16 18 20 22 0 2 4 6 8 10 12 14 16 18 20 22
187
0 -11
10 101 10
20 -9
-8 2
8
-30 30
7
=40 6 40
6
(50 050 5
60 S60 4
0
70 2 70 3
2
80 80
0
0 2 4 6 8 10 12 14 16 18 20 22 0 2 4 6 8 10 12 14 16 18 20 22
Time (hr) Time (hr)
Event #3 Flow
1 -4
20 3.5
I 3
I
0
-00 2.5
0 2
60 0 25 20 1.5
70- 70
0.2 80
LO 0.5
80
0
0 2 4 6 8 10 12 14 16 18 20 22 24 0 2 4 6 8 10 12 14 16 18 20 22 24
Time (hr) Time (hr)
XH, XANs, XANb (mg/kg), soil moisture, 0 (cm 3/cm 3), and flow rate, q (cm/hr) for Event #3.
188
I . .
0? OD
co (D
000 w
* ,,ArN'f ~ ')1~
.0
LU
A -
-i
r o
-w)qdC 0 0 ; 0 ) 0 0 0 c
0
E
N (D z
(N M- 17L Cc)
(wo) tidaeO (wo) 41daC]
I CO 0
co
(D
(P
0 0
CN
0 o Q 0 0 0
OD
/ ff
04) 00
CN W)
SE~/ fy 4
co
0
CN '
Cc
I
C 0 CD 0 0 0
r- CID
o
f-
o00
0D M -I C) to
(wo) qida@i (wo) i~ldag (wo) q1da40
0 Go to C- (N c
N
CD
(D
GCD
AV//
0N 04
- w 1 "
Cc E4 0 , 0 C E
cn/IT I D-0
(w / t ~/ eA
to CD
V,
07
0D
0 00 00 0 0, 0 0
-c in 0~ r l C - WD
(wo) t~jdG0 (wo) q~deoi
_CR (mg/L), Readily degradable COD. Event #4
0 1
-
-16
116
20 120 10 10
14
20 20 114
20
100 12
-30 12
30
10
40 10
40
8
50 8
6
608 60 60 6
4
70 70 70 4
2
802 80 80 2
S 2 4 6 8 10 1 14 16 18 0 2 4 6 8 10 12 14 16 18 0 2 4 6 8 10 12 14 16 18
Soil molstui Event #4
0 - M 0.4
4 l
100
20 0.35 2
30
40 0.3
0) 50- |I
0
60 0.25 0
2
70-
80 0.2
0
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Time (hr) Time (hr)
XH, XANs, XANb (mg/kg), soil moisture, 0 (cm 3/cm 3 ), and flow rate, q (cm/hr) for Event #4.
190
0 -4 0 . D. wA NJ O0 -4 0 0 ( NJ -
00 0 0 0 0) 0 OX
0
oo
=1
3
CD
0 el /0 3
I r r0
!li;.i/;i It il I IiI
NI~
0 P0 04 0 0 00

C(P 4- (I
00 - A, w " N g -
C)0 0 0 0 0C 0 0
Ti
0)
Al 0
0-
IC 0
0
z -4
CD 3
0 (D
U
0
3
0
8U
U
m 0
NJ
0
NJ
NJ
<0
P9-0 0 0 0 0 0 0 0 P 0 0 0 0
0 02a C Ili - 0 D
-p ( (ma)
N) a

o0 0 0 0 0
z
AP-
4- 0
C- 3
115,
cD-
=r
0~
1
0 16 0o
AI
0 0 P
0
P 0 0 0 0 0) 0 9 P P
*o P 0 1%j
CwJ 6 CA07 a
IF ; - -
200 25 25
10
180
160
20 20 20
20
140 30 -30
120 40 15 E 0 15
100
050
80 10
10
60
60
40 70 5 70 5
20 80 80
U z 4 6 8 10 12 14 16 18 20 22 0 2 4 6 8 10 12 14 16 18 20 22 0 2 4 6 8 10 12 14 16 18 20 22
Flo* q (cmlhr), Event #5
i0.4 0 8
101 10 -7
7
201 0.35 20
6
-30 -30
5
401 0.3 -40
4
3
60 0.25 60
2
70 70
80 0.2 80
0
0 2 4 6 8 10 12 14 16 18 20 22 24 0 2 4 6 6 10 12 14 15 18 20 22 24
Time (hr) Time (hr)
XH, XANS, XANb (mg/kg), soil moisture, 0 (cm 3/cm 3 ), and flow rate, q (crn/hr) for Event #5.
192
01 -4 0) Cln -9 (a P) - 01 (D) C" A. N) -.
8$o --j0 8~ ~ ~ _ n - U W "
o -
000 0 0 0 0 0 C) O 0 0
01
co01 r " / ' lgr
Kr
0 0 , '
3a
CD-
Z
re A m
~
N j/
~~/j//f'/
-. I 0
N
N
0 -. N)NN
01 *01 Q

01 D 01 01 C) c' NI 0 - 0 P P
01~~~~ =r0 3 (
~ ) - 010
010
~0 3
1
43
01
I "I
0 0 03 0 . 0 M OD 0 .0 0 0 0 00
*0 * 0 0 N) ~ ~ ~ ~ Z Cn M10 -4~ 0000

01 -4 0) b (a N) 01 -4 0 01P (aw -o 01-14
0 0131w
0 0 0> C3 0D 0 0) 0 0 0 0 0 0D 0 0
0) 0 0 0 a 0 0 0
PooI 0
jz 0
3
co-
V11
iM
I It/
I
Ka
CD
o0 0 0 o
0 0 0 0 0 0 00c 0 -. - N) 7'
C) ca C~ a' 0) C)o
17 0 7
55 10
50 6
20 20
45 5
40 30 30 5
E
35 -40 4
30
50
25
3
60
20
15 70 2
10 80 00
2 4 6 8 10 12 14 16 18 20 22
0 2 4 6 8 10 12 14 15 18 21 22 S0
Time (hr)
Time (hr)
Event #6 0.4
0 m-
10 10 7
201 0.35 20 6
30E -30 5
401 0.3 E40 4
d) 50 3
60
0.25 -60
2
70 70
-
0.2
80
0
0 2 4 6 8 10 12 14 16 18 20 22 24 0 2 4 6 8 10 12 14 16 18 20 22 24
Time (hr) Time (hr)
3 3
XH, XANs, XANb (mg/kg), soil moisture, 0 (cm /cM ), and flow rate, q (cm/hr) for Event #6.
194
70- 70- 70-
Event #3 - Modeled Event #2 - Modeled Event #1 - Modeled
60- - Observed -- Observed - Observed
60- 60-
E E
50- Io 50- 50-
C)
0)
:E5c 40- c
40- 40-
0 0 0
a-
C'Y) 30- 04 30- IC
30-
4)C: a,
W
20- 20- w 20-
10- 10- 10-
U 1 a c . . . . . . . . n- 0
.
.
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
70- Time (hr) 70- Time (hr) 70- Time (hr)
60- Event #5 - Modeled 60- Event #6 - Modeled 60. Event #4 - Modeled

-- Observed - Observed - Observed
E E
50- 50- 0 50.
C)
0 C:
c
I) 40- C: 40- 40-
LO 0 0
f5
W 30- (- 30-1 30-
Wr
a-
20- 20- 20-
10- 10- 10-
0 0- -S . . m . . . . I 0 I. . . .
-
.
0 2 4 6 8 10 12 14'16'18 20 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Figure 5-10. Ponding water level, WLp (cm), for Events #1 - #6.
195
400- 400- 400-

350- Observed 350- - Observed 350- - Observed
300- 300- 300-
E E E
250- 250- 250-
200- C, 200- 200-

0
150- 150- 150-
w 100 w 100- 100-
50- 50- 50-
0- 0
-
-
1 2 3 4 0 1- , , 0 1 2 3 4
350- 350- Modeled 350- Event #4- Modeled

Event #5- Modeled Event #6 -
-- Observed -- Observed
- Observed -C -c
300- 300- 300-
E E E
250- 250- 250-
U
200 200 200-
0
-
C.
LO) CD
150 C: 150- 150-
-
a)
w 100- Lu 100- 100-
50- 50- 50-
U 0 -! 0- I
0 1 2 3 4 0 1 2 3 4 0 1 2 3 4
Figure 5-11. Culvert discharge, qc (cm/hr), hydrograph for Events #1 - #6.
196
12- 12- 12-
- Observed 10- - Observed - Observed
10- 10.
E 8-
0 0 6- 0
C:
4- a) 4- U)
4-
w
2- 2-
n A . . . 0-
-
.
o 5 10 15 20 0 5 10 15 20 0 5 10 15 20
10- Event #5 - Modeled 10- Event #6 - Modeled 10- Event #4 - Modeled

- Observed - Observed - Observed
E 8- E 8- E 8-
a)
0 6- 4-
0 6- 6-
LO
C C: -"lo
a)
4- a) 4- a)
4-
LU
2- 2- 2-
n I
0- 0 I
6
.
0 5 10 15 20 10 15 20 0 5 10 15 20
Figure 5-12. Outlet discharge, qo (cm/hr), hydrograph for Events #1 - #6. (The observed peaks in outlet discharge due to leakage in the
outlet box are not modeled.)
197
Table 5-7. Model results showing the modeled mass (g) at culvert, and outlet and the resultant
removal rate, R% compared to the observed values.
Observed Mass (g) and Removal (%) Modeled Mass (g) and Removal (%)
Event Inlet Culvert Outlet R% Culvert Outlet R%
#3 TN 43.23 0.00 13.80 68 0.00 7.13 84
#2 TN 101.57 22.01 20.26 58 20.18 32.96 48
#1 TN 464.04 166.52 133.39 35 198.62 49.76 46
#5 TN 459.11 390.91 21.99 10 301.92 25.41 29
#6 TN 667.14 806.86 22.58 -24 607.35 25.15 5
#4 TN 461.35 419.17 24.09 4 373.02 26.98 13
#3 ON 25.89 0.00 6.45 75 0.00 4.77 82
#2 ON 49.17 4.84 4.87 80 6.91 18.22 49
#1 ON 249.94 34.07 31.89 74 53.36 11.98 74
#5 ON 171.94 47.62 6.17 69 37.86 17.38 68
#6 ON 69.63 58.92 6.37 6 52.24 18.05 -1
#4 ON 150.61 115.09 17.34 12 121.18 19.44 7
#3 NH3 7.08 0.00 2.73 61 0.00 2.34 67
#2 NH3 9.60 2.83 8.99 -23 4.34 14.72 -99
#1 NH3 126.92 42.13 39.10 36 64.62 36.90 20
#5 NH3 38.19 0.61 0.89 96 0.50 6.27 82
#6 NH3 24.81 53.27 1.28 -120 49.63 6.74 -127
#4 N13 21.60 18.30 0.68 12 20.93 7.48 -32
#3 N03 4.94 0.00 0.22 96 0.00 0.02 100
#2 N03 29.30 5.21 0.06 82 7.74 0.01 74
#1 N03 49.61 46.53 30.25 -55 70.44 0.67 -43
#5 N03 286.18 322.86 3.74 -14 263.07 1.47 8
#6 N03 470.57 615.01 0.33 -31 497.91 0.08 -6
#4 N03 177.19 216.92 0.13 -22 228.51 0.04 -29
#3 N02 3.87 0.00 0.08 98 0.00 0.00 100
#2 N02 2.82 0.81 0.05 69 1.18 0.00 58
#1 N02 12.82 6.29 21.61 -118 10.20 0.21 19
#5 N02 0.82 0.61 2.12 -234 0.50 0.29 4
#6 N02 5.71 9.46 3.62 -129 7.56 0.28 -37
#4 N02 1.78 2.07 0.15 -25 2.40 0.02 -36
#3 IP 0.44 0.00 0.15 66 0.00 0.36 19
#2 IP 2.53 0.59 0.58 54 0.91 2.68 -42
#1 IP 10.53 2.27 1.94 60 3.45 2.38 45
#5 IP 16.82 14.68 0.40 10 12.24 1.80 17
#6 IP 23.68 21.55 1.19 4 18.88 2.32 10
#4 IP 29.40 6.92 0.43 75 7.26 2.33 67
198
Table 5-8. Literature values for pXH, UXANs, pXANb (u is the maximum growth rate)
Bacteria species p, growth rate (day')

Koike & Hattori (1975) Pseudomonas denitrificans 3.84 (anaerobic), 19.2 (aerobic)
Garcia-Plazaola et al. (1993) Rhizobium meliloti (13 strains of denitrifiers) 0.25
Hommes et al. (2003) Nitrosomonaseuropaea 1 to 3.24 (depending on culture substrate)
Table 5-9. List of parameter values set for the different levels of bacterial growth rate and decay rate (day-)
Symbol Description High Medium Low

PH Maximum aerobic growth rate on CR 19.2 1.5 0.1
pDN Maximum denitrification rate by denitrifiers 3.84 1 0.1
pANs Maximum aerobic growth rate on NH4-N by Nitrosomonas 3.24 0.75 0.01
pUANb Maximum aerobic growth rate on N0 2-N by Nitrobacter 3.24 0.75 0.01
bH Rate constant for cell lysis of heterotrophs 1 0.1 0.01
bANs Rate constant for cell lysis of Nitrosomonas 1 0.1 0.01
bANb Rate constant for cell lysis of Nitrobacter 1 0.1 0.01
199
Table 5-10. Effects of increasing or decreasing growth rate (pH, PDN, ,UAns, 1 ANb) and cell lysis rate (bH, bAns, bANb) on various pollutants
in the model as compared to the case of medium rates (Table 5-9).
Reaction number in the model NH 3 NO 3 NO 2 N2 IP 02

Higher bH, bA, bANb RC5, 7, 9 +
+
Higher pH RC 2
-
Higher pAns RC 6 + -
-
Higher pJANb RC 8 + - -
-
HigherpDN RC 3,4 - - +
-
Reaction number in the model NH 3 NO 3 NO 2 N2 IP 02
Lower bH, bAns, bAn RC5, 7, 9
Lower pH RC 2 + +
+
Lower pAns RC 6 + +
+
Lower pAvb RC 8 - + +
+
Lower pDN RC 3,4 + + -
200
Event mean outlet concentration, TN (mg/L) Removal rate, TN (%)
2.5 150 r
- - - bisector - - - bisector 10
U highp,b U highp,b
medp,b medy,b
,
2 lowu,b
/
lowu,b
100 F
.
1.5
* -4 24.-'
4.
(D
0 50 F
0
E E
I
C?
0 -0
O. F ''
-50 -- I I --- I I
0 0.5 1 1.5 2 2.5 -5 0 0 50 100 150

observed observed
Event mean outlet concentration, ON (mgIL) Removal rate, ON (%)

1 150 r
- - - bisector - - - bisector
0.9 * highp,b high/A,b
-
medp,b 4 of
medyA,b
0.8 Iowp,b lowp,b
-
100 F
0.7
-
0
0.6
-
~0 'a
0.5 4. 50 F 4.4.
-
0 0
E E
0.4
It
0.3 0 ,4
0
.
0.2
-
0.1
0 -50 '-
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 -50 0 50 100 150
observed observed
201
Event mean outlet concentration, NH3 (mg/L) Removal rate, NH 3 (%)
250
1.5
-
- - - bisector
- - - bisector 200 * highp,b
-
* highp,b
medy,b
medy,b 11111 150
1 Iowp,b
-
Iowp,b
100
-
50
0.5
-
D
-
II)
A a) 0
-
0
0
0 10
E E
0 -50
4*
-100
4*
4.
-0.5 I- -150 - U
-200
, , , , , 250~
-
150 200 250
-
-1 -0.5 0 0.5 1.5 -250 -200 -150 -100 -50 0 50 100

observed observed
Removal rate, NO 3(%)
Event mean outlet concentration, NO 3 (mg/L)
250
-
0.5 r
- - - bisector 4.
- - - bisector 200
* highp,b
-
E highp,b -4
0.4 medyI,b
medyb 150 lowp,b
-
owp,b 4* .0
0.3 F- 100
-
4*
"0
50
-
D 0.2
-
0
0
0
E 0.1 F=
-50
-
M
-
0
-100
0
-
-
-150
-
-0.1 F
-200
-0.2 ' ' ' ' ' ' -250 6

'
-0. 2 -0.1 0 0.1 0.2 0.3 0.4 0.5 -250 -200 -150 -100 -50 0 50 100 150 200 250
observed observed
202
Event mean outlet concentration, NO 2 (mg/L) Removal rate, NO 2 N
1 250 r
U highp,b s 200
- highp,b
medp,b medy&,b
lowp,b 150
lowo,b
100
0.5 I-
-
50
'a
0 1-
U
#-
0
0 0
E E 1-0
-50
0
-100
-150
-200
-
-0.5 -250 -0
'
-0.5 0 0.5 1 -250 -200 -150 -100 -50 0 50 100 150 200 250
observed observed
Removal rate, IP (%)

Event mean outlet concentration, IP (mg/L) 250
0.15
- - - bisector
200
highpb
U highp,b I medy4,b
med,bI 150
- -b
low/A,b
*
'I
0.1 100
-
50
'a
(D
-) (D
"a 0
0.05 0
0
E E
-50 U
-
0 [ -100
-150
-
-
U
-200 -
-0.05 -250
-0.05 0 0.05 0.1 0.15 -250 -200 -150 -100 -50 0 50 100 150 200 250
observed observed
Figure 5-13. Mean outlet concentration and removal rate of various pollutants under cases of high, medium, and low bacterial growth
and decay rates in Event #1.
203
Table 5-11. Statistical evaluation of effects of bacterial dynamics on mean outlet concentration: coefficient of determination (R2),
Nash-Sutcliffe efficiency (NSE), mean absolute error (MAE, mg/L), relative root mean square error (RRMSE), RMSE-observations
standard deviation ratio (RSR), and percent bias (PBIAS, %) for the event mean outlet concentration
High bacterial growth and decay rate

TN ON NH3 N03 N02 IP
R2 0.97 0.03 0.96 0.85 0.95 0.51
NSE -0.06 -13.37 -1.62 0.29 0.07 -197.73
MAE 0.41 0.31 0.32 0.04 0.04 0.08
RRMSE 0.70 1.73 2.00 1.38 1.17 6.74
RSR 1.03 3.79 1.62 0.84 0.97 14.10
PBIAS -66.7 -163.8 -192.4 64.7 76.9 -670.7
Medium bacterial growth and decay rate
R2 0.93 0.12 0.95 0.96 0.98 0.89
NSE 0.89 -5.41 0.83 0.95 0.87 0.87
MAE 0.11 0.21 0.05 0.01 0.01 0.00
RRMSE 0.22 1.15 0.51 0.37 0.44 0.17
RSR 0.33 2.53 0.41 0.23 0.36 0.36
PBIAS -12.9 -94.7 -13.2 -17.8 -20.6 -2.0
Low bacterial growth and decay rate
R2 0.95 0.06 0.96 0.99 0.92 0.82
NSE 0.55 -8.87 0.76 0.78 -2.02 -0.07
MAE 0.22 0.26 0.05 0.03 0.05 0.00
RRMSE 0.46 1.43 0.61 0.77 2.10 0.49
RSR 0.67 3.14 0.49 0.47 1.74 1.03
PBIAS -36.5 -135.9 -18.3 -49.0 -91.7 7.7
204
0D
0 CD 0 (co -4 0m u1 D. w Ili
00 01 0 N o 0 0 0 0 0 0 0 0 0
0000 n
r
3 0) D/I0 3 XC
C7/ ~CD0
CD0
m
=rn m
2S
a,
co
0) c " c -j
0 -~ N, WC r. (A -4
'CO NC 4. a, co NC

co - I 0) n A. ) w , di o -- 4 a) C Un .1 w K .)
itI I, j~
IM 1)
(JI
di III.A/ ~It Is
0 C0) 0 0 - - ~ 0 Pi w -1 0 0) -4 co
N) 4 0) cc r) Z. mD C N) C wD N ~
Di epth (cm) Depth (cm) Depth (cm)

co -4 a) Pl C - I> . ) w , di 'J -4 C C (n P .w NJ,0 -4 D ) Ln di - w- 1 1 -
S 0 0 0 D 0 0 0 0 0 C :) C D C C C) 0 C 0
0 0 0
r CD d N
9g
CD0
0 ~ Ii Y/1 N,f 0
0,
e oo
N,1
a 0 W 4- M CD c
;o * c *m
IT1

A g. w So 0 0 0 0 0 0
C N N
.0
'.
-. 3
=r
0
CD mr
aD
00
P PP o 0 0 0o ____
*CJ *0
V3 i: 0l

0o -4 0D (IC -9 wJ N
0
z
-t
0
4 8 N
C 3 3 CD
k~ 0-
0- m
a
a
00 PP 0 0 0 0
ir
*. N
P
C 2

CD~ 0 0) 0 8
C iV~L D
[~i
41
CD 0 If
3
3) 0
m
a
3
aD
-j
0 0--N)N)W 0 * P P
00 00 o o oq CJC (IC (IC
N)
Table 5-12. Literature values of bacterial count for denitrifiers and nitrifiers
Reference Target genes Gene copies/ g soil Bacterial count, Sx (mg/kg)

Chen et al. (2013) Denitrifying: nirS, nirK, norB, nosZ 105 - 108 0.1 - 100
Chen et al. (2013) Nitrifying: amoA 10 4 - 106 0.01 - 1
Henry et al. (2006) nosZ 10 5 -10 7 0.1-10
Henry et al. (2004) nirK 9.7x10 4 - 3.9x10 6 0.1 -4
Table 5-13. List of parameter values set for the different levels of initial bacterial count in the different soil layers
High bacterial count Medium bacterial count Low bacterial count

Unsaturated Transition Anoxic Unsaturated Transition Anoxic Unsaturated Transition Anoxic
zone zone zone zone zone zone zone zone zone
SXH 100 100 100 10 10 20 0.1 0.1 0.1
SXAN 1 0 0.1 0.01 0 0.01 0.01 0
SXANb 1 1 0 0.1 0.01 0 0.01 0.01 0
207
Event mean outlet concentration, TN (mg/L) Removal rate, TN (%)
2.5
150
- - - bisector
- - - bisector
* high count
U high count
med count
med count
2 low count
( low count 100 F
-
0
1.5 F p0
50 |-
0 0
E E
1
0
0
07-I
0
0.5 F
0 0. -50
'
'
0 0.5 1 1.5 2 2.5 -5 0 0 50 100 150
observed observed
Removal rate, ON (%)

Event mean outlet concentration, ON (mg/L) 150 r
I
- - - bisector
- - - bisector U high count
0.9 * high count
-
med count
med count low count
0.8 low count 100 F
-
0.7
-
0.6
-
-
50 F
0.5 0
-
0
E E
0.4
0.3 0
0
0.2
0.1
0 ' ' ' ' ' ' -50

'
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 -50 0 50 100 150
observed observed
208
Event mean outlet concentration, NH 3 (mg/L) Removal rate, NH 3 (%)
1.5 r 250
-
200
* high count U high count
-
med count med count
1 low count 150
low count
-
0 100
-
0.5 F 50 F
4)
0 0
0 10.
E E
0 -50 F
-100 F a
,~
-0.5 F -150
-
-200
L'
-
-1 ' ' ' '
-
-250 -'
'
-1 -0.5 0 0.5 1 1.5 -250 -200 150 -100 -50 0 50 100 150 200 250
observed observed
Removal rate, NO
Event mean outlet concentration, NO3 (mg/L) 250 3
-
0.5
-
- - - bisector
- - - bisector 200
U high count
-
* high count
0.4 med count
med count 150
-
low count
-
low count-
10
0.3 100
-
.
.1 50
-
0.2 40
-
0
0
0
0, E
-
E 0.1F -50
-
0
0 [ Ad
-100
-150
F
I-
-0.1
-200
-0.2 -250
-
-0.2 -0.1 0.1 0.2 0.3 0.4 0.5 -250 -200 -150 -100 -50 0 50 100 150 200 250
observed observed
209
Removal rate, NO (%)
Event mean outlet concentration, NO 2 (mgIL) 250
2
-
1
- - - bisector
- - - bisector 200
U
-
high count
* high count
med count
med count 150
-
low count
low count
100
10
-
0.5 1
50
(, 0
0 -U
S
'o
0 0
E E
-50
0 -100 O.
-150
-200
-
-0.5 -250
-0.5 0 0.5 1 -250 -200 -150 -100 -50 0 50 100 150 200 250
observed observed
Event mean outlet concentration, IP (mgIL) Removal rate, IP (%)

0.15 250
U 200
high count U high count
med countl med count
low count 150 low count
0.1
100
50
0.05 .1 0 0
0
E I E
-50 0
-100
0
-150
-20-200
-
-0.05 -250
-
-0.05 0 0.05 0.1 0.15 -250 -200 -150 -100 -50 0 50 100 150 200 250
observed observed
Figure 5-15. Mean effluent concentration and removal rate of various pollutants under cases of high, medium, and low bacterial count
in Event #1.
210
Table 5-14. Statistical evaluation of effects of bacterial count on mean outlet concentration: coefficient of determination (R 2), Nash-
Sutcliffe efficiency (NSE), mean absolute error (MAE, mg/L), relative root mean square error (RRMSE), RMSE-observations standard
deviation ratio (RSR), and percent bias (PBIAS, %) for the event mean outlet concentration.
High bacterial count, Sx

TN ON NH3 NO3 N02 IP
R2 0.97 0.00 0.95 0.77 0.96 0.65
NSE 0.74 -6.60 0.28 0.32 0.16 -51.53
MAE 0.21 0.22 0.16 0.03 0.04 0.04
RRMSE 0.35 1.25 1.05 1.35 1.11 3.47
RSR 0.51 2.76 0.85 0.82 0.92 7.25
PBIAS -24.2 -111.4 -93.1 60.0 72.9 -344.2
Medium bacterial count, Sx
R2 0.93 0.12 0.95 0.96 0.98 0.89
NSE 0.89 -5.41 0.83 0.95 0.87 0.87
MAE 0.11 0.21 0.05 0.01 0.01 0.00
RRMSE 0.22 1.15 0.51 0.37 0.44 0.17
RSR 0.33 2.53 0.41 0.23 0.36 0.36
PBIAS -12.9 -94.7 -13.2 -17.8 -20.6 -2.0
Low bacterial count, Sx
TN ON NH3 NO3 N02 IP
R2 0.96 0.41 0.96 0.99 0.92 0.81
NSE 0.50 -9.66 0.76 0.77 -2.18 -0.41
MAE 0.23 0.27 0.06 0.03 0.05 0.01
RRMSE 0.48 1.49 0.61 0.78 2.16 0.57
RSR 0.71 3.26 0.49 0.48 1.78 1.19
PBIAS -37.8 -142.9 -14.0 -49.7 -94.4 9.9
211
00 -4 0) 0 1 w Ili N CID -- 4 0m Ln At w. N -
CD CD m 0)
C)
" 0
00
CD0
'-N) co
A
CA
Ar ?c
f~iCD 0~ M
PQ!'_- __ __ _ __
co
0 - fN) w. A' -.4 co Cii M

o l J, o n0o 0-K- 0 0 0O
Depth (cm) Depth (cmn)

-0 0 0 C> C) )
C"C
CDQ 0C
m A (A
-CD
K . ) to3 *1 0 01 -4
K0 0 b0 0 0 ... K3

0o -- 4 0) (1A 4 C.) N) CID0 -4 0) 01 At c.) NJ - co -4 0) C11 A' 8. ", -8
C) C) C CD C) C) C) C D C C) c) C) 0) CD C3 C)0
i/,/9 Au'/ z
EOE
CD ~iIIIII
~I IIiz
CI IlII ii' ~I)IIIL0I 0 w 0 0 0
Z AfJf14 / 4. 'rT~h
.~ "1 x
CD
*I/
OD '
01
PQ IV/
CD
o 0 0 0 7- . -
N)~ D
~~~~ N) ~ 0 CDC ka 0. m o b C"
0
C -t

-) n P) N PQ a J 0
c o -4 0) Cr J, w 0 0 0 0000 8 o
C)
0
C-
CD
3
Es
m4
r4j m
6)D 1'1/21
CIOt
00 0 * 9 71 N !) w * 0 0

Ta) 1! T k R :E, S C) C C,) C) 8 c
F/k z
OD1
CD
?c
CL ~fr
(A
?A
00 m m m
0 0 0 0
0 0 0 9 0 0
In Im

00 a D Ln St w P e C3
I, aCD
C
z
0
3e
10 IF c
-4
0
1,'CD
0/(A
C,,
HI 'Cm
0 0
* *l.
0
LO
o po
a) 70 I I I I I I I I b) 45
60 40 -
-
A-A
-
V -AA A-A
35
-
050
A- - ~ N 30
VA.V'AA-- A- ._-----
C40 > 25
0 0
E 30 E 20
2C K, - 1 cm/hr U) R= 70
-20 . --- R = 70.0 I Z 15 h =10cm
-
-- R = 35.0 10 -- = 30 cm
.
10 .- -A-R= 23.3 -A- hd = 50 cm
-v-R = 17.5 5 --- v- hd= 80 cm
0 0
0 10 20 30 40 50 60 70 80 90 o 10 20 30 40 50 60
hd (cm) K, (cm/hr)
c) 1 d) 20
.
.A ----R = 70.0 R - 70 3h= 10 CM

2
--- R = 35.0 +h,= 30 cm
0.8 R = 23.3 15 hd= 50 cm
-
-
A
1
- = 17 .hI
R -V'- h, = 80 CM
i02
0.6 K= 1 cm /hr 10- 0
wU izi
-
-1
0.4 5-
A
0 10 2 mih io sb
L-
0.2 0-
0.0 -5-
0 10 20 30 40 50 60 70 80 0 10 20 30 40 50
hd (cm) K, (cm/hr)
Figure 5-17. Predicted TN (total nitrogen) removal and basin performance as assessed using PEI (performance efficiency index) at
various detention depth, hd (cm); ratio of drainage area to bioretention area, R; and saturated hydraulic conductivity, Ks (cm/hr).
214
Chapter 6
Implications for future bioretention basin design
6.1 Current design guidelines
Current design guidelines for bioretention basin design by various regulatory authorities lack
consensus in terms of the type of metric used for quantification and the amount of runoff that
needs to be retained by the bioretention basin (Chapter 2, Section 2.2). Design guidelines broadly
fall under two groups: WQV or WQD (water quality volume or depth) and ARI (average
recurrence interval). WQV or WQD specifies the runoff amount to be retained by the basin while
ARI describes the rainfall intensity and duration of an event that determines the potential runoff
volume.
6.1.1 ARI
As described in Chapter 2 (Section 2.5.3.1), ARI as a term is too ambiguous to specify the
rainfall depth and hence the runoff volume in the tropics. A single ARI value can encompass a
broad range of rainfall depths with a varying mix of rainfall intensity and duration. The rainfall
depths within this broad range result in different runoff volume. While specification of ARI may
be adequate in temperate climates, it is not in tropical climates because a modest increase in ARI
can lead to a more significant increase in rainfall intensity in the tropics than in temperate
regions (Figure 2-S-5). Therefore, due to the the differing climate conditions and rainfall
215
characteristics in different parts of the world, design guidelines do not work uniformly well at all
places and particularly need to be adjusted for the tropics.
6.1.2 WQV / WQD
Unlike ARI, which encompasses a broad range of rainfall depths, WQV or WQD is more
specific as it refers to a particular runoff volume. By retaining a specified amount of WQD, it is
assumed that a corresponding level of pollutant removal can be achieved. Therefore, as the water
quality requirement of the receiving water body varies, the amount of WQD required to be
retained also varies.
Guidelines in terms of WQD often are expressed in two ways. The first way is more of a rule-of-
thumb guideline that simply specifies the upper bound, or the most conservative WQD, and thus
guarantees a high pollutant removal rate. An example of such a guideline is that in some parts of
the U.S. under which basins are set to capture a WQD that corresponds to the first one-half to
one inch (12.7 - 25.4 mm) of runoff (Massachusetts Department of Environmental Protection
2008; Minnesota Pollution Control Agency 2008; New Hampshire Department of Environmental
Services 2008).
The second way to express guidelines in terms of WQD is to provide a set of design curves that
promise pollutant removal levels at a given basin dimension (often in terms of the surface
detention depth, hd, and the ratio of catchment area to basin area, R) and soil filter media
property (often in terms of the saturated hydraulic conductivity, Ks). For example, the Australian
Capital Territory requires basins to be designed with models such as MUSIC (Model for Urban
216
Stormwater Improvement Conceptualization) (Cooperative Research Centre for Catchment
Hydrology 2003) to generate the design curves for analyzing large catchments.
6.2 Performance under current Singapore design guidelines
The current design guidelines in Singapore are modelled after the Australian guidelines and
include both the ARI and WQV metrics. This study made use of both field measurements and
models to look at how these design guidelines work in the tropics as exemplified by the Balam
Estate Rain Garden in Singapore. Design implications for the next generation of bioretention
basins in the tropics are then drawn.
6.2.1 ARI
As the Balam Estate Rain Garden was the first bioretention basin in Singapore, there were no
formal Singapore design guidelines to follow at its time of construction in 2008. However, at
present, the Australian guideline (3-month ARI) is used in Singapore (Public Utilities Board of
Singapore 2014). Singapore's IDF curves show that the 3-month-ARI event corresponds to
approximately 5-35 mm of rain. In Chapter 2, Figure 2-3 explores the ability of the Balam Estate
Rain Garden to retain 3-month ARI events during the monitoring period of this study. Runoff
retention at the lower end of the 3-month ARI (e.g. 5 mm rain) is satisfactory but that at the
higher end of the 3-month ARI (e.g. 33.2 mm for Event #6) is poor. Hence, this study has
demonstrated that the use of ARI in guiding basin sizing is not definitive in a tropical context.
217
6.2.2 WQV/ WQD
In terms of WQV, Singapore's design guidelines assume that basins with dimensions (and hence
WQVs) delineated by the MUSIC model-derived design curves are expected to deliver the
removal rates defined by the curves. However, the current curves from the Australia-originated
MUSIC model are not field-validated in the tropics. In Chapter 4, a six-month hydrological
simulation estimated that the infiltration percentage, I, at the current basin configuration (hd =
13.7 cm, R = 70, Ks 1 cm/hr), is only 22%. In Chapter 5, the hydrological model is coupled
with a water quality model to simulate the effluent quality during the six-month period. The
model shows that the total nitrogen removal, TNR, by the basin at its current configuration is 28%,
which falls short of the 45% target rate. The corresponding prediction of TNR by the MUSIC
design curves for the current basin configuration (Ks = 1 cm/hr) is not able to be read off as the
curves are generated with Ks = 18 and 36 cm/hr. However, it is evident that the bioretention
basin in its current configuration would not have delivered the expected performance in TN
removal during the monitoring period. In the absence of field-validated design curves, this study
used the water quality model developed in Chapter 5 to generate design curves that are more
applicable to the tropics.
6.3 Recommended design guidelines
In this study, we gathered information from field observations to first suggest a range of WQD
that is suitable for the tropics. We then coupled a hydrological model and a water quality model
to narrow that range.
218
First in Chapter 2 and 3, we showed that a lack of storage capacity and resulting high overflow in
Balam Estate Rain Garden reduce its pollutant removal efficiency for high-rainfall-depth events.
We further observed from field measurements that a transition from efficient to non-efficient
pollutant removal occurs at a rainfall depth between 10 and 30 mm. In terms of ARI, this range is
also the transition from < 3-month ARI to < 1-year ARI events. In light of these results, we
recommend that a larger basin storage volume (WQD in the range of 10 - 30 mm) is needed in
the tropics in order to treat the intense but common rainfall events in the tropics.
With respect to the quantity of water, Roesner et al. (1991) found that capturing a WQD of one
inch (25.4 mm) of runoff is sufficient to capture 90% of the annual runoff in a broad cross
sample of U.S. cities. However, from results of this study, a WQD of about 32 mm would be
required to capture 90% of the annual runoff in Singapore. It would require too much space to
construct bioretention basins large enough to capture 90% of the runoff from the frequent and
heavy rain storms in the tropics., Therefore, this study investigated if it is possible to achieve
removal rate targets (e.g. 45% for TN) while retaining less than 90% of runoff. To answer this
question, we developed a water quality model with the aim of narrowing down the WQD range
further from that suggested in Chapter 2 and 3 (10 - 30 mm).
Therefore, in Chapter 4, we employed a hydrological model, RECHARGE, to simulate the
hydrology of the rain garden during a half-year period. After verifying the model with
continuous field measurements from 80 events, we used the model to predict the ability of the
basin to infiltrate runoff as a function of the basin configuration in terms of hd, R, and Ks. We
found that increasing basin dimensions beyond hd= 40 cm, R = 23.3, and Ks = 10 cm/hr provides
219
little improvement in the fraction of incoming water infiltrated (1). Therefore, in order to improve
infiltration, we recommend future basins be built no larger than hd= 40 cm, R = 23.3, and Ks
10 cm/hr.
To recommend design guidelines with respect to the quality of water, in Chapter 5 we adapted
the CW2D model to simulate the water quality of the subsurface soil media over a six-month
period. The water quality model showed that increasing basin dimensions beyond hd= 30 cm and
R = 23.3 provides little improvement in the total nitrogen removal (TNR). Moreover, the
maximum value of TNR occurs at Ks = 5 cm/hr. We recommend that future basins to be built
using the design curves constructed using the water quality model developed in this study
(Figure 5-17).
To investigate if it is possible to achieve the removal rate target of TNR= 45% while retaining
less than 90% of runoff as in the case of the U.S. guideline by Roesner et al. (1991), Table 6-1
shows that at the current saturated conductivity of Ks 1 cm/hr, a WQV of 239 - 362 m 3 or
equivalently a WQD of 14 - 22 mm is required. This corresponds to an infiltration percentage of
I= 51 - 67% which is less than the 90% capture in the U.S. This shows that it is possible to
achieve removal rate targets (e.g. 45% for TN) without retaining 90% of runoff.
The design curves developed in this study have several advantages. First, as the design curves
are plotted using R, the ratio of catchment area to basin area, the curves are applicable to
catchments of varying area. WQD acts as a scaling factor between catchment area and WQV and
can be taken as the unit storage volume. Second, the process-focused water quality model takes
220
into account any changes in the inherent removal rate as the basin configuration changes in both
surface and subsurface storage capacity. Third, the design curves eliminate the ambiguity of
using ARI to define a runoff capture volume.
Table 6-2 compares the predictions of nitrogen removal rate TNR made by the design curves
developed in this study to those made by the MUSIC model (Public Utilities Board of Singapore
2014). There are two prominent differences. First, the MUSIC curves predict a higher TNR than
this study. Second, this study found that the maximum TNR occurs at Ks = 5 cm/hr while MUSIC
curves predict that TNR monotonically increases as Ks increases from 18 to 36 cm/hr. Reduced
TNR at high hydraulic conductivity could be due to insufficient hydraulic residence time for
complete biochemical reactions for pollutant removal.
6.4 Conclusion
The findings in this study imply the following for future bioretention basin designs in the tropics:
1. In Chapter 2 and 3, we recommended that a larger basin storage volume (WQD in the range
of 10 - 30 mm) is needed in the tropics in order to treat the intense but common events in the
tropics.
2. In Chapter 4, based on the hydrological simulation over half a year, we found that the
percentage of the inflowing water that is infiltrated by the basin (1%) increases only
marginally if the basin dimensions are increased beyond hd= 40 cm, R = 23.3, and Ks = 10
cm/hr,
221
3. In Chapter 5, based on the water quality simulation over half a year, we found that increasing
basin dimensions beyond hd = 30 cm and R= 23.3 provides little improvement in the total
nitrogen removal (TNR). Moreover, the maximum value of TNR occurs at Ks = 5 cm/hr.
4. It is possible to achieve a targeted total nitrogen removal rate of 45% while retaining less
than 90% of runoff. At the current saturated conductivity of Ks = 1 cm/hr, a WQV of 239 to
362 m 3 or equivalently a WQD of 14 to 22 mm is required to achieve TNR = 45%. This
corresponds to an infiltration percentage of I= 51 - 67% of the influent.
222
Table 6-1. Predicted infiltration percentage I (%), water quality volume, WQV (m3), and water
quality depth, WQD (mm), for various basin configurations (varying hd and R) that achieve TN
removal rate (TNR) of 45%
Basin hd R Ks I WQV WQD

Configuration (cm) (cm/hr) (%) (m3) (mm)
#1 10 17.5 1 53 249 15
#2 15 17.5 1 61 302 18
#3 15 23.3 1 51 239 14
#4 20 23.3 1 57 280 17
#5 45 35.0 1 64 335 20
#6 50 35.0 1 67 362 22
Table 6-2. Comparison of TN removal rate (TNR) between predictions made by six-month water
quality simulation in this study and by MUSIC curves (Public Utilities Board of Singapore 2014)
Basin hd R Ks TNR (%) TNR (%)

Configuration (cm) (cm/hr) This study muSIC
#1 10 70 18 26 27
#2 30 70 18 28 32
#3 10 70 36 24 28
#4 30 70 36 26 34
223
Chapter 7
Conclusions and future work
7.1 Conclusions
The bioretention basin is a form of stormwater best management practice that originated in
temperate regions and is becoming increasingly popular in tropical regions. Nevertheless, it is a
poorly evaluated system in the tropics due to the limited availability of long-term, high-
resolution hydrological and multi-component water quality data on completed basins. The
resulting gaps in data and knowledge limit the ability of environmental managers to control
pollution by stormwater.
In this thesis, a real-time monitoring system that consists of flow and water-level measurement
equipment and automatic water quality samplers was installed and operated from April to
November 2013 at Balam Estate Rain Garden in the tropical setting of Singapore. With an
effective treatment area of 240 M 2 , this basin receives runoff from a 16,800 m 2 residential
catchment (about 88% surface imperviousness) with a time of concentration of about 10 min
(Wang et al. 2009). Balam Estate Rain Garden was the first bioretention basin in Singapore,
originally constructed as a demonstrative project under the ABC (Active, Beautiful, Clean)
scheme by the Public Utilities Board of Singapore.
224
In this thesis, we provide a complete survey of basin performance in treating 15 water quality
parameters (nitrogen and phosphorus species, total suspended solids, and chemical oxygen
demand). We then combine the findings on water quality with those from hydrological
assessment of the system performance in 80 storm events to recommend practical engineering
strategies to design and manage bioretention basins in the tropics. Next, we calibrate a
hydrological model, RECHARGE; as well as formulate a water quality model (an adaption of
CW2D, a wetland model) to simulate the hydrological and water quality performance of the
bioretention basin. These models have also been validated against our field observations.
Following is a summary of the research highlights:
1. We assess whether a bioretention basin built according to temperate guidelines (i.e. the study
site) is able to improve water quality of stormwater runoff. We found that the Balam Estate
Rain Garden is able to improve water quality. Higher removal is seen at the outlet when
runoff is allowed to filter through the soil media than if it overflows via the culvert. The
highest mean removal rate was seen in TSS while N02-N most often showed net export
instead of removal which might suggest extensive nitrification but incomplete denitrification
within the soil media (Chapter 2).
2. We evaluate if the bioretention basin was able to remove pollutants at targeted rates in a
tropical setting. We found that removal targets for TN, TP, and TSS are generally met only
for events with small rainfall depths (Events #3, #2, and #1) and not for those with greater
rainfall depths (Events #5, #6, and #4). The mean removal rates for TN, TP, and TSS are
54%, 72%, and 90% respectively for small events, compared to -3%, 19%, and 16% for large
events (Chapter 2).
225
3. We analyze the factors behind observed removal efficiency in terms of rainfall characteristics
such as rainfall depth, rainfall intensity, and antecedent dry period (ADP) as well as runoff
characteristics such as event mean concentration (EMC) and first flush. We found that the
range of influent EMCs observed in this study are comparable to those in tropical catchments
presented in the literature but are lower than temperate counterparts. The frequent and intense
rainfall of the tropical climate likely leads to lower EMCs than in temperate climates. The
first-flush phenomenon of the influent is not found to be prominent if a stringent criterion is
imposed. Low influent EMC and weak first flush appear to contribute to low pollutant
removal rate (Chapter 2).
4. We investigate the suitability of current design guidelines for the tropics. We found that a
lack of storage capacity and resulting high culvert overflow is the main driver in reducing
removal efficiency for large, but still common, storm events. The transition from efficient to
non-efficient treatment occurs at a rainfall depth between 10 and 30 mm. At the current
storage capacity, the basin performance during as many as 46% of the events in this
monitoring period would not meet the pollutant removal targets (Chapter 2, 3).
5. We propose that there is a need to revise design guidelines for sizing bioretention basins in
tropical catchments. We recommended design guidelines in the tropics be specified in terms
of the more definitive quantity WQV or WQD instead of ARI. A larger basin storage volume
(WQD in the range of 10 to 30 mm) is needed in the tropics in order to treat the intense but
common events in the tropics (Chapter 3).
6. We explore the applicability of the RECHARGE numerical hydrological model for
continuous simulation at the field scale. We found that RECHARGE is able to successfully
simulate the hydrology of a bioretention basin in a tropical climate. We validated the model
226
with continuous field measurements from 80 events over a half-year period. There is a good
agreement between observed and simulated time series of ponded water level, WLp(cm);
culvert discharge, qc (cm/hr), outlet discharge, qo (cm/hr), and soil moisture, 0 over the entire
half-year period of simulation (Chapter 4).
7. We explore bioretention basin configurations for improved hydrological performance
efficiency using the validated model. We predicted bioretention basin hydrological
performance in terms of infiltration percentage as a function of engineering design
parameters such as detention depth (hd), ratio of drainage area to basin area (R), and saturated
hydraulic conductivity (Ks) of the soil media. We found that increasing basin dimensions
beyond hd = 40 cm, R = 23.3, and Ks = 10 cm/hr, the marginal improvement in basin
performance drops significantly (Chapter 4).
8. We adopt a process-based model, CW2D, to describe the bacterial dynamics that contribute
to water quality improvement in the bioretention basin subsurface soil media. We
demonstrated that mathematical formulations originating from activated sludge model (ASM)
in wastewater treatments are also applicable for modeling the role that bacterial growth and
decay play in stormwater treatment in bioretention basins. However, some biokinetic
parameters need to be calibrated for usage in bioretention basins (Chapter 5).
9. We predict bioretention basin water quality performance in terms of pollutant removal
percentage and effluent concentration at various high, medium, and low levels of bacterial
growth and decay kinetics (,UH, YDN, UAns, UANb, bH, bAns, bANb) and initial bacteria
concentrations (SxH, SXANs, SXANb). We found that the onset of denitrification is earlier for
cases of high bacterial kinetic rates or high initial bacteria concentration. Simulation results
based on calibrated and validated parameters also show that the high built-up of N02 and the
227
slight increase in N03 during an event indicate that the first step of the nitrification process,
which is the conversion of NH3 to N02 and mediated by Nitrosomonas (XANs), is proceeding
faster than the second step, which is the conversion of N02 to N03 and mediated by
Nitrobacter(XAN) (Chapter 5).
10. We explore bioretention basin configurations for improved water quality removal efficiency
using the validated water quality model. We predict bioretention basin performance in terms
of TN (total nitrogen) removal percentage (TNR) over a half-year period as a function of
engineering design parameters such as detention depth (hd), ratio of drainage area to basin
area (R), and saturated hydraulic conductivity (Ks) of the soil media. We find that increasing
basin dimensions beyond hd = 30 cm, R = 23.3, and Ks = 5 cm/hr improves basin
performance only slightly. We also develop a set of design curves for future bioretention
basin designs in the tropics. For example, to achieve TNR = 45% at the current soil hydraulic
conductivity of Ks 1 cm/hr, the required WQV ranges from 240 - 360 m 3 and WQD (water
quality depth) from 14 - 22 mm, depending on the basin area A and detention depth hd
(Chapter 5).
The findings in this study raise the question of whether it is practical to build bioretention basins
in the tropics. To achieve a TN removal target of 45% at the current saturated hydraulic
conductivity of 1 cm/hr, we would need to capture a WQD of 14 - 22 mm which translates to a
basin of 45 - 50 cm in detention depth and twice the current basin area (2.8% of catchment area).
This requirement is still practical if basins of this size are planned as a part of initial land
development process. The basin need not occupy a continuous piece of land. Several smaller
basins that run in parallel and are placed strategically around the catchment could achieve the
228
same treatment efficiency. Therefore, as long as urban developers factor the basin size
requirement into their plans, the practicality of bioretention basins should not be limited by a
space requirement. A detention depth of 45 - 50 cm may prolong ponding duration and raise
ponding depth to the extent that it may raise public health and safety concerns. Further research
needs to be done to investigate if a deep unfenced pond would pose public risks and if extended
standing water would result in mosquito breeding.
The hydrological and water quality models developed in this study provide tools for future use.
The coupled model could be used as a tool to generate design curves for any set of design
parameters desired by regulatory agencies. If necessary, the model should be calibrated to local
soil conditions and inflow characteristics. The design curves would then be available as simple
"off-the-shelf" curves for basin designers and landscape architects to read off from.
229
7.2 Future work
The following research tasks are suggested to further extend the research introduced in this thesis:
1. In Chapter 2, we recommended a larger basin storage volume (WQD in the range of 10 - 30
mm) is needed in the tropics in order to treat the intense but common events in the tropics.
Although we have narrowed this range to 14 - 22 mm with the water quality model built in
Chapter 5, several improvements could be made in the terms of design guideline
development. First, there is a need to improve the calibration of initial pollutant
concentrations across different soil depths. A more intensive pre- and post-event sampling
program over a longer time duration could give a better representation of the initial
conditions. Second, a more comprehensive study of microbial activity is also needed to
estimate growth and decay rates as well as bacteria count in the subsurface soil media. To do
that, future studies need to focus on quantifying the depth-dependent microbiological activity
occurring in the subsurface of bioretention basins. Third, to obtain the inputs to the water
quality model to represent the influent water quality from a catchment, a catchment runoff
model (e.g. SWMM) could be utilized. This would be an improvement over the simplified
uniform pollutograph used for all the events during the model application described in
Chapter 5. Fourth, a model needs to be built to simulate water quality treatment of water
ponded in the surface basin by physical processes such as sedimentation.
2. In Chapter 5, the water quality model is adopted from an existing model. Additional
biochemical processes involved in the nitrogen cycle such as DNRA (dissimilatory nitrate
reduction to ammonium) and Anammox (anaerobic ammonium oxidation) could be added to
the model as could other bacteria species like autotrophic denitrifiers and aerobic denitrifiers.
230
3. Another potential research area is to investigate how a bioretention basin perform compared
to other types of BMPs. A bioretention basin is more costly to construct and maintain than
other BMPs but has the best nitrogen removal capability. The IWS (internal water storage)
compartment plays a major role in denitrification during inter-storm periods. Therefore, if
nitrogen removal is high on the priority list of pollutants to be removed for a particular
catchment, bioretention basin is still the best BMP option.
231
References
APHA, AWWA, and WPCF (2009). Standard methods for the examination of water and
wastewater, American Public Health Association, Washington, D.C., USA.
Atchison, D., and Severson, L. (2004). RECARGA User's Manual, version 2.3, University of
Wisconsin - Madison, Civil & Environmental Engineering Department, Water Resources
Group.
Australian Capital Territory Planning & Land Authority (2008). "Waterways water sensitive
urban design general code." Australian Capital Territory Planning & Land Authority,
Canberra, Australia
Barron, 0. V., Donn, M. J., and Barr, A. D. (2013). "Urbanisation and Shallow Groundwater:
Predicting Changes in Catchment Hydrological Responses." Water Resour Manag, 27(1),
95-115.
Becciu, G., and Raimondi, A. (2015). "Probabilistic Modeling of the Efficiency of A Stormwater
Detention Facility." Int JSus Dev Plan, 10(6), 795-805.
Bedan, E. S., and Clausen, J. C. (2009). "Stormwater Runoff Quality and Quantity From
Traditional and Low Impact Development Watersheds." J Am Water Resour As, 998-
1008.
Bertrand-Krajewski, J.-L., Chebbo, G., and Saget, A. (1998). "Distribution of pollutant mass vs
volume in stormwater discharges and the first flush phenomenon." Water Res, 32(8),
2341-2356.
Brezonik, P. L., and Stadelmann, T. H. (2002). "Analysis and predictive models of stormwater
runoff volumes, loads, and pollutant concentrations from watersheds in the Twin Cities
metropolitan area, Minnesota, USA." Water Res, 36(7), 1743-1757.
Brisbane City Council (2006). "Water sensitive urban design technical design guidelines for
South East Queensland. Version 1.", Brisbane City Council and the Moreton Bay
Waterways and Catchments Partnership, Brisbane, Australia.
Brown, R. A., and Hunt, W. F. (2011). "Evaluating Media Depth, Surface Storage Volume, and
Presence of an Internal Water Storage Zone on Four Sets of Bioretention Cells in North
Carolina." World Environmentaland Water Resources Congress 2011, 405-414.
Brown, R. A., and Hunt, W. F. (2011). "Impacts of Media Depth on Effluent Water Quality and
Hydrologic Performance of Undersized Bioretention Cells." J Irrig Drain Eng, 137(3),
132-143.
232
Chen, X. L., Peltier, E., Sturm, B. S. M., and Young, C. B. (2013). "Nitrogen removal and
nitrifying and denitrifying bacteria quantification in a stormwater bioretention system."
Water Res, 47(4), 1691-1700.
Chua, L. H. C., Lo, E. Y. M., Shuy, E. B., and Tan, S. B. K. (2009). "Nutrients and suspended
solids in dry weather and storm flows from a tropical catchment with various proportions
of rural and urban land use." JEnviron Manage, 90(11), 3635-3642.
City of Seattle (2014). "City of Seattle stormwater manual April 2014 review draft." Department
of Planning and Development, Seattle, Washington, USA.
Cooperative Research Centre for Catchment Hydrology (2003). Model for urban stormwater
improvement conceptualisation (MUSIC) user guide, Cooperative Research Centre for
Catchment Hydrology.
Davis, A. P. (2008). "Field performance of bioretention: Hydrology impacts." J Hydrol Eng,

13(2), 90-95.
Davis, A. P., Hunt, W. F., Traver, R. G., and Clar, M. (2009). "Bioretention Technology:
Overview of Current Practice and Future Needs." JEnviron Eng, 135(3), 109-117.
DeBusk, K. M., and Wynn, T. M. (2011). "Storm-Water Bioretention for Runoff Quality and
Quantity Mitigation." JEnviron Eng, 137(9), 800-808.
Deletic, A. (1998). "The first flush load of urban surface runoff." Water Res, 32(8), 2462-2470.
Department of Irrigation and Drainage (2011). Urban stormwater management manual for
Malaysia, Department of Irrigation and Drainage Malaysia, Kuala Lumpur, Malaysia.
Diblasi, C. J., Li, H., Davis, A. P., and Ghosh, U. (2009). "Removal and Fate of Polycyclic
Aromatic Hydrocarbon Pollutants in an Urban Stormwater Bioretention Facility."
Environ Sci Technol, 43(2), 494-502.
Dierkes, C., Lucke, T., and Helmreich, B. (2015). "General Technical Approvals for
Decentralised Sustainable Urban Drainage Systems (SUDS)-The Current Situation in
Germany." Sustainability, 7(3), 3031.
Dietz, M. E., and Clausen, J. C. (2005). "A field evaluation of rain garden flow and pollutant
treatment." Water Air Soil Poll, 167(1-4), 123-138.
Dietz, M. E., and Clausen, J. C. (2006). "Saturation to Improve Pollutant Retention in a Rain
Garden." Environ Sci Technol, 40(4), 1335-1340.
Driscoll, E., Palhegyi, G., Strecker, E., and Shelley, P. (1989). "Analysis of storm event
characteristics for selected rainfall gages throughout the United States." U.S.
Environmental Protection Agency, Washington, D.C., USA.
233
Duncan, H. P. (1999). "Urban stormwater quality: A statistical overview." Cooperative Research
Centre for Catchment Hydrology, Melbourne, Australia.
Dussaillant, A. R., Cozzetto, K., Brander, K., and Potter, K. W. (2003). "Green-Ampt model of a
rain garden and comparison to Richard's equation model." Sustainable planning and
development, the sustainableworld, WIT, Southampton, U.K., 891-900.
Dussaillant, A. R., Wu, C. H., and Potter, K. W. (2004). "Infiltration of Stormwater in a Rain
Garden: Richards Equation Numerical Model and Field Experiment." International
Environmental Modelling & Software. Complexity and Integrated Resources
Management.Osnabruck, Germany.
Dussaillant, A. R., Wu, C. H., and Potter, K. W. (2004). "Richards equation model of a rain
garden." J Hydrol Eng, 9(3), 219-225.
Eriksson, E., Baun, A., Scholes, L., Ledin, A., Ahlman, S., Revitt, M., Noutsopoulos, C., and
Mikkelsen, P. S. (2007). "Selected stormwater priority pollutants - a European
perspective." Sci Tot Env, 383(1-3), 41-5 1.
Facility for Advancing Water Biofiltration (2009). Adoption guidelines for stormwater
biofiltration systems, Facility for Advancing Water Biofiltration, Monash University,
Melbourne, Australia
Francey, M., Fletcher, T. D., Deletic, A., and Duncan, H. (2010). "New Insights into the Quality
of Urban Storm Water in South Eastern Australia." JEnviron Eng, 136(4), 381-390.
Geheniau, N., Fuamba, M., Mahaut, V., Gendron, M. R., and Dugue, M. (2015). "Monitoring of
a Rain Garden in Cold Climate: Case Study of a Parking Lot near Montreal." J Irrig
DrainEng, 141(6), 04014073.
Geiger, W. F. (1987). "Flushing effects in combined sewer systems. Topics in Urban Storm
Water Quality, Planning and Management." Proceedings of the IV International
Conference on Urban Storm Drainage, XXII Congress InternationalAssociation for
Hydraulic Research, International Association for Hydraulic Research, Lausanne,
Switzerland.
Glass, C., and Bissouma, S. (2005). "Evaluation of a parking lot bioretention cell for removal of
storm water pollutants." Ecosystems and sustainabledevelopment V, WIT, Southampton,
U.K., 699-708.
Goh, H. W., Zakaria, N. A., Lau, T. L., Foo, K. Y., Chang, C. K., and Leow, C. S. (2015).
"Mesocosm study of enhanced bioretention media in treating nutrient rich stormwater for
mixed development area." Urban Water Journal, 1-9.
Green, I. R. A., and Stephenson, D. (1986). "Criteria for Comparison of Single Event Models."
Hydrolog Sci J, 31(3), 395-411.
234
Gupta, H. V., Sorooshian, S., and Yapo, P. 0. (1999). "Status of Automatic Calibration for
Hydrologic Models: Comparison with Multilevel Expert Calibration." J Hydrol Eng, 4(2),
135-143.
Hatt, B. E., Fletcher, T. D., and Deletic, A. (2009). "Hydrologic and pollutant removal
performance of stormwater biofiltration systems at the field scale." J Hydrol, 365(3-4),
310-321.
He, B., Wang, Y., Takase, K., Mouri, G., and Razafindrabe, B. H. N. (2009). "Estimating Land
Use Impacts on Regional Scale Urban Water Balance and Groundwater Recharge."
Water Resour Manag, 23(9), 1863-1873.
He, Z., and Davis, A. P. (2009). "Unit Process Modeling of Stormwater Flow and Pollutant
Sorption in a Bioretention Cell." World Environmental and Water Resources Congress
2009, 1-9.
He, Z. X., and Davis, A. P. (2011). "Process Modeling of Storm-Water Flow in a Bioretention
Cell." JIrrigDrainEng, 137(3), 121-131.
Heasom, W., Traver, R. G., and Welker, A. (2006). "Hydrologic modeling of a bioinfiltration
best management practice." JAm Water Resour As, 42(5), 1329-1347.
Henderson, C., Greenway, M., and Phillips, I. (2007). "Removal of dissolved nitrogen,
phosphorus and carbon from stormwater by biofiltration mesocosms." Water Sci Technol,
55(4), 183-191.
Henry, S., Baudoin, E., Lopez-Gutierrez, J. C., Martin-Laurent, F., Brauman, A., and Philippot,
L. (2004). "Quantification of denitrifying bacteria in soils by nirK gene targeted real-time
PCR." JMicrobiolMeth, 59(3), 327-335.
Henry, S., Bru, D., Stres, B., Hallet, S., and Philippot, L. (2006). "Quantitative detection of the
nosZ gene, encoding nitrous oxide reductase, and comparison of the abundances of 16S
rRNA, narG, nirK, and nosZ genes in soils." Appl Environ Microb, 72(8), 5181-5189.
Henze, M., Gujer, W., Mino, T., and van Loosdrecht, M. C. M. (2000). Activated Sludge Models
ASM], ASM2, ASM2D and ASM3, IWA Scientific and Technical Rep.9. Int. Water Assn.,
London.
Houdeshel, C. D., Hultine, K. R., Johnson, N. C., and Pomeroy, C. A. (2015). "Evaluation of
three vegetation treatments in bioretention gardens in a semi-arid climate." Landscape
Urban Plan, 135, 62-72.
Huber, W. C., Dickinson, R. E., Barnwell Jr, T. 0., and Branch, A. (1988). Storm Water
Management Model, version 4, U.S. Environmental Protection Agency, Environmental
Research Laboratory.
235
Hunt, W. F., Jarrett, A. R., Smith, J. T., and Sharkey, L. J. (2006). "Evaluating bioretention
hydrology and nutrient removal at three field sites in North Carolina." J Irrig Drain Eng,
132(6), 600-608.
Hunt, W. F., Smith, J. T., Jadlocki, S. J., Hathaway, J. M., and Eubanks, P. R. (2008). "Pollutant
removal and peak flow mitigation by a bioretention cell in urban Charlotte, NC." J
Environ Eng, 134(5), 403-408.
International Organization of Standards (1980). "Water flow measurement in open channels

using weirs and venturi flumes - Part 1: Thin plate weirs. ISO 1438/1-1980(E)."
Kandeler, E., DeigImayr, K., Tscherko, D., Bru, D., and Philippot, L. (2006). "Abundance of
narG, nirS, nirK, and nosZ genes of denitrifying bacteria during primary successions of a
glacier foreland." Appl Environ Microb, 72(9), 5957-5962.
Khan, U. T., Valeo, C., Chu, A., and He, J. X. (2013). "A Data Driven Approach to Bioretention
Cell Performance: Prediction and Design." Water-Sui, 5(1), 13-28.
Khan, U. T., Valeo, C., Chu, A., and van Duin, B. (2012). "Bioretention cell efficacy in cold
climates: Part 2-water quality performance." Can J Civil Eng, 39(11), 1222-1233.
Kindsvater, C. E., and Carter, R. W. (1959). "Discharge characteristics of rectangular thin-plate

weirs." Trans, Amer Soc Civil Eng, 24, Paper No. 3001.
Lai, F.-H., Zhen, J., Riverson, J., Alvi, K., and Shoemaker, L. (2009). "Multiple watershed scales
approach for placement of best management practices in SUSTAIN." Proc., ASCE
Environment and Water Resources Congress.
Lai, F.-H., Zhen, J., Riverson, J., and Shoemaker, L. (2006). "SUSTAIN. An Evaluation and
Cost-optimization Tool for Placement of BMPs." Proc., ASCE EWRI World Water and
Environmental Congress, 21-25.
Langergraber, G., and Simunek, J. (2005). "Modeling variably saturated water flow and
multicomponent reactive transport in constructed wetlands." Vadose Zone J, 4(4), 924-
938.
Langergraber, G., and Simnek, J. (2006). The multi-component reactive transport module
CW2D for constructedwelands for the HYDRUS software package. Manual - version 1.0,
Institute for Sanitary Engineering and Water Pollution Control, BOKU - University of
Natural Resources and Applied Life Sciences, Vienna; Dept. of Environ. Sciences,
University of California, Riverside.
Le, S. H., and Chua, L. H. C. (2014). "Comparison of first flush quantification methods in
tropical catchments." 13th International Conference on Urban Drainage.Sarawak,
Malaysia.
Lee, L. Y., Tan, L., Wu, W., Yeo, S. K. Q., and Ong, S. L. (2013). "Nitrogen removal in
saturated zone with vermicompost as organic carbon source." Sus Env Res, 23(2), 85-92.
236
LeFevre, G. H., Paus, K. H., Natarajan, P., Gulliver, J. S., Novak, P. J., and Hozalski, R. M.
(2015). "Review of Dissolved Pollutants in Urban Storm Water and Their Removal and
Fate in Bioretention Cells." J Environ Eng, 141(1).
Li, H., and Davis, A. P. (2008a). "Heavy metal capture and accumulation in bioretention media."
Environ Sci Technol, 42(14), 5247-5253.
Li, H., and Davis, A. P. (2008b). "Urban particle capture in bioretention media. I: Laboratory and
field studies." JEnvironEng, 134(6), 409-418.
Li, H., and Davis, A. P. (2008c). "Urban particle capture in bioretention media. II: Theory and
model development." JEnviron Eng, 134(6), 419-432.
Li, H., and Davis, A. P. (2009). "Water Quality Improvement through Reductions of Pollutant
Loads Using Bioretention." JEnviron Eng, 135(8), 567-576.
Li, L. Q.,and Davis, A. P. (2014). "Urban Stormwater Runoff Nitrogen Composition and Fate in
Bioretention Systems." Environ Sci Technol, 48(6), 3403-3410.
Line, D. E., and Hunt, W. F. (2009). "Performance of a Bioretention Area and a Level Spreader-
Grass Filter Strip at Two Highway Sites in North Carolina." J Irrig Drain Eng, 135(2),
217-224.
Liu, J. Y., and Davis, A. P. (2014). "Phosphorus Speciation and Treatment Using Enhanced
Phosphorus Removal Bioretention." Environ Sci Technol, 48(1), 607-614.
Lucas, W. C., and Greenway, M. (2011). "Phosphorus Retention by Bioretention Mesocosms

Using Media Formulated for Phosphorus Sorption: Response to Accelerated Loads." J
IrrigDrain Eng, 137(3), 144-153.
Mangangka, I. R., Liu, A., Egodawatta, P., and Goonetilleke, A. (2015). "Performance
characterisation of a stormwater treatment bioretention basin." J Environ Manage, 150,
173-178.
Marchioni, M., and Becciu, G. (2015). "Experimental Results On Permeable Pavements in Urban
Areas: A Synthetic Review." Int JSus Dev Plan, 10(6), 806-817.
Massachusetts Department of Environmental Protection (2008). Stormwater Management

Standards and Structural BMP specifications, Massachusetts stormwater handbook,
Massachusetts Department of Environmental Protection, Boston, Massachusetts, USA.
McLeod, S. M., Kells, J. A., and Putz, G. J. (2006). "Urban runoff quality characterization and
load estimation in Saskatoon, Canada." JEnviron Eng, 132(11), 1470-1481.
McNett, J. K., Hunt, W. F., and Davis, A. P. (2011). "Influent Pollutant Concentrations as
Predictors of Effluent Pollutant Concentrations for Mid-Atlantic Bioretention." J Environ
Eng, 137(9), 790-799.
237
Minnesota Pollution Control Agency (2008). Minnesota stormwater manual, version 2,
Minnesota Stormwater Steering Committee, Minnesota Pollution Control Agency, St.
Paul, Minnesota, USA.
Moriasi, D. N., Arnold, J. G., Van Liew, M. W., Bingner, R. L., Harmel, R. D., and Veith, T. L.
(2007). "Model evaluation guidelines for systematic quantification of accuracy in
watershed simulations." Transactions of the American Society of Agricultural and
BiologicalEngineers, 50(3), 885-900.
Moriasi, D. N., Gitau, M. W., Pai, N., and Daggupati, P. (2015). "Hydrologic and Water Quality
Models: Performance Measures and Evaluation Criteria." TAsabe, 58(6), 1763-1785.
Myllyoja, R., Baroudi, H., Pitt, R., and Paluzzi, J. (2001). "Use of SLAMM in evaluating best
management practices." Models and Applications to Urban Water Systems, Vol. 9, W.
James, ed., Computational Hydraulics International, Guelph, Ontario, 131-141.
Nash, J. E., and Sutcliffe, J. V. (1970). "River Flow Forecasting Through Conceptual Models
Part I - A Discussion of Principles." JHydrol, 10(3), 282-290.
Nehrke, S. M., and Roesner, L. A. (2004). "Effects of design practice for flood control and best
management practices on the flow-frequency curve." J Water Resour Plan Manage
130(2), 131-139.
New Hampshire Department of Environmental Services (2008). New Hampshire stormwater

manual, Vol 2, New Hampshire Department of Environmental Services, Watershed
Assistance section, Concord, New Hampshire, USA.
NOAA, U. S. (2015). "Precipitation frequency data server."

<http://hdsc.nws.noaa.gov/hdsc/pfds>. (Sept 2015).
Novotny, V. (1995). Nonpoint pollution and urban stormwater management, Technomic

Publishing Company, Lancaster, Pennsylvania, USA.
Ojeda, E., Caldentey, J., Saaltink, M. W., and Garcia, J. (2008). "Evaluation of relative
importance of different microbial reactions on organic matter removal in horizontal
subsurface-flow constructed wetlands using a 2D simulation model." Ecol Eng, 34(1),
65-75.
Ong, G. S., Kalyanaraman, G., Wong, K. L., and Wong, T. (2012). "Monitoring Singapore's first
bioretention system: rain garden at Balam Estate." 7th InternationalConference on Water
Sensitive Urban DesignMelbourne Cricket Ground, Melbourne, Australia.
Passeport, E., Hunt, W. F., Line, D. E., Smith, R. A., and Brown, R. A. (2009). "Field Study of
the Ability of Two Grassed Bioretention Cells to Reduce Storm-Water Runoff Pollution."
JIrrigDrainEng, 135(4), 505-510.
238
Paus, K. H., Muthanna, T. M., and Braskerud, B. C. (2016). "The hydrological performance of
bioretention cells in regions with cold climates: seasonal variation and implications for
design." Hydrol Res, 47(2), 291-304.
Payne, E. G. I., Fletcher, T. D., Cook, P. L. M., Deletic, A., and Hatt, B. E. (2014). "Processes
and Drivers of Nitrogen Removal in Stormwater Biofiltration." Crit Rev Env Sci Tec,
44(7), 796-846.
Pilgrim, D. H. (1987). Australian rainfall and runoff - A guide to flood estimation, revised
edition, Institution of Engineers, Barton, ACT, Australia
Pitt, R., and Voorhees, J. (1995). "Source Loading and Management Model (SLAMM)."
National Conference on Urban Runoff Management: Enhancing Urban Watershed
Management at the Local, County, and State Levels, Center for Environmental Research
Information, U.S. Environmental Protection Agency. EPA/625/R-95/003, Cincinnati,
Ohio., 225-243.
Prince George's County (2002). "Bioretention Manual." Prince George's County (MD)
Government, Department of Environmental Protection, Watershed Protection Branch,
Landover, MD.
PUB (2014). "Active Beautiful Clean Waters Design Guidelines 3rd Edition." Public Utilities
Board of Singapore.
Public Utilities Board of Singapore (2011). "Code of practice on surface water drainage. 6th ed.
with amendments under addendum no.1 ", Public Utilities Board of Singapore.
Public Utilities Board of Singapore (2014). "Chapter 7, Bioretention basins, Engineering

procedures for ABC waters design features." Public Utilities Board of Singapore.
Ritter, H. (2013). "Nitrogen Chemistry in an Urban Bioretention System in Singapore." Master

of Engineering, MIT.
Roesner, L., Burgess, E., and Aldrich, J. (1991). "The hydrology of urban runoff quality
management." Water Resources Planning and Management and Urban Water Resources
18th Annual Conference and Symposium, American Society of Civil Engineers, New
Orleans, Louisiana, US, 764-770.
Rossman, L. A. (2004). Storm Water Management Model User's Manual Version 5.0, U.S.
Environmental Protection Agency, Washington, DC, EPA/600/R-05/040.
Roy-Poirier, A., Champagne, P., and Filion, Y. (2010). "Bioretention processes for phosphorus
pollution control." Environ Rev, 18, 159-173.
Roy-Poirier, A., Champagne, P., and Filion, Y. (2010). "Review of Bioretention System
Research and Design: Past, Present, and Future." JEnviron Eng, 136(9), 878-889.
239
Saeed, T., and Sun, G. Z. (2011). "Kinetic modelling of nitrogen and organics removal in vertical
and horizontal flow wetlands." Water Res, 45(10), 3137-3152.
Sample, D. J., Grizzard, T. J., Sansalone, J., Davis, A. P., Roseen, R. M., and Walker, J. (2012).
"Assessing performance of manufactured treatment devices for the removal of
phosphorus from urban stormwater." J Environ Manage, 113, 279-291.
Santhi, C., Arnold, J. G., Williams, J. R., Dugas, W. A., Srinivasan, R., and Hauck, L. M. (2001).
"Validation of the swat model on a large river basin with point and nonpoint sources." J
Am Water Resour As, 37(5), 1169-1188.
Shoemaker, L., Riverson, J., Alvi, K., Zhen, J. X., Paul, S., and Rafi, T. (2009). SUSTAIN- A
Framework for Placement of Best Management Practices in Urban Watersheds to
Protect Water Quality, U.S. Environmental Protection Agency, EPA/600/R-09/095.
Sim'nek, J., Huang, K., and Van Genuchten, M. T. (1998). The HYDRUS code for simulating
the one-dimensional movement of water, heat, and multiple solutes in variably-saturated
media. Version 6.0, Research Report No. 144, California: U.S. Salinity Laboratory,
USDA, ARS, Riverside.
Sim'nek, J., and Van Genuchten, M. T. (2006). The HYDRUS software packagefor simulating
the two- and three-dimensional movement of water, heat, and multiple solutes in
variably-saturatedmedia: Technical manual. Version 1.0, PC-Progress, Prague, Czech
Republic.
Smullen, J., Shallcross, A., and Cave, K. (1999). "Updating the U.S. nationwide urban runoff
quality data base." Water Sci Technol, 39(12), 9-16.
Strecker, E. W., Quigley, M. M., Urbonas, B. R., Jones, J. E., and Clary, J. K. (2001).
"Determining urban storm water BMP effectiveness." J Water Resour Plan Manage,
127(3), 144-149.
Subramaniam, D., Mather, P., Russell, S., and Rajapakse, J. (2016). "Dynamics of Nitrate-
Nitrogen Removal in Experimental Stormwater Biofilters under Intermittent Wetting and
Drying." JEnviron Eng, 142(3).
Tang, S., Luo, W., Jia, Z., Liu, W., Li, S., and Wu, Y. (2016). "Evaluating Retention Capacity of
Infiltration Rain Gardens and Their Potential Effect on Urban Stormwater Management
in the Sub-Humid Loess Region of China." Water Resour Manag, 30(3), 983-1000.
Taylor, G. D., Fletcher, T. D., Wong, T. H. F., Breen, P. F., and Duncan, H. P. (2005). "Nitrogen
composition in urban runoff - implications for stormwater management." Water Res,
39(10), 1982-1989.
Trowsdale, S. A., and Simcock, R. (2011). "Urban stormwater treatment using bioretention." J
Hydrol, 397(3-4), 167-174.
240
Upper Parramatta River Catchment Trust (2004). Water sensitive urban design technical
guidelines for Western Sydney, Upper Parramatta River Catchment Trust, Parramatta,
NSW, Australia.
U. S. EPA (1983). "Results of the Nationwide Urban Runoff Program Volume I- Final Report."
U.S. Environmental Protection Agency, Washington, D.C., USA.
U. S. EPA (2000). "Low impact development (LID): A literature review." U.S. Environmental
Protection Agency, Washington, D.C., USA.van Genuchten, M. T. (1980). " A closed-
form equation for predicting the hydraulic conductivity of unsaturated soil." Soil Sci. Soc.
Am. J., 44, 892-898.
Walker, W. W. (2007). P8 Urban Catchment Model Version 3.4, U.S. Environmental Protection
Agency, Minnesota Pollution Control Agency, Wisconsin Dept of Natural Resources.
Wang, J., Chua, L. H. C., and Shanahan, P. (2016). "Continuous Numerical Hydrological
Modelling of a Bioretention Basin " In preparation.
Wang, J., Chua, L. H. C., and Shanahan, P. (2016). "Field evaluation of hydrological
performance of a bioretention basin in an urban tropical catchment." 3rd International
Conference on Design, Construction, Maintenance, Monitoring and Control of Urban
Water Systems, Urban Water 2016, WIT Press, San Servolo, Venice, Italy, 177-188.
Wang, J., Chua, L. H. C., and Shanahan, P. (2017). "Evaluation of Pollutant Removal Efficiency
of a Bioretention Basin and Implications for Stormwater Management in Tropical Cities."
Env Sci Water Res & Tech, 3(1), 78 - 91.
Wang, J., Wong, T. H. F., and Ong, G. S. (2009). "Design, construction and monitoring
Singapore's first bioretention systems." 6th InternationalWater Sensitive Urban Design
Conference, Perth, Australia.
Wanko, A., Mose, R., Carrayrou, J., and Sadowski, A. G. (2006). "Simulation of biodegradation
in infiltration seepage - Model development and hydrodynamic calibration." Water Air
Soil Poll, 177(1-4), 19-43.
WEF, and ASCE (2012). "Design of urban stormwater controls. Water Environment Federation
Manual of Practice No. 23, American Society of Civil Engineers Manual and Report on
Engineering Practice No. 87."Water Environment Federation, Alexandria, Virginia,
U.S.A.
WeiBbach, M., Criddle, C. S., Drewes, J. E., and Koch, K. (2017). "A proposed nomenclature for
biological processes that remove nitrogen." Env Sci: Water Res & Tech, 3(1), 10-17.
Wong, T. H. F., Fletcher, T. D., Duncan, H., Coleman, J. R., and Jenkins, G. A. (2002). "A
Model for Urban Stormwater Improvement Conceptualisation." Global Solutions for
Urban Drainage, 8-13.
241
Wong, T. H. F., Fletcher, T. D., Duncan, H. P., and Jenkins, G. A. (2006). "Modelling urban
stormwater treatment - A unified approach." Ecol Eng, 27(1), 58-70.
Woznicki, S. A., and Nejadhashemi, A. P. (2012). "Sensitivity Analysis of Best Management

Practices Under Climate Change Scenarios." JAm Water Resour As, 48(1), 90-112.
Xuan, Z. M., Chang, N. B., Wanielista, M. P., and Williams, E. S. (2013). "System Dynamics
Modeling of Nitrogen Removal in a Stormwater Infiltration Basin with Biosorption-
Activated Media." JEnviron Qual, 42(4), 1086-1099.
Zhang, S. H., and Guo, Y. P. (2014). "Stormwater Capture Efficiency of Bioretention Systems."
Water Resour Manag, 28(1), 149-168.
Zinger, Y., Blecken, G. T., Fletcher, T. D., Viklander, M., and Deletic, A. (2013). "Optimising
nitrogen removal in existing stormwater biofilters: Benefits and tradeoffs of a retrofitted
saturated zone." Ecol Eng, 51, 75-82.
242

Signature Redacted

Uploaded by

Copyright:

Available Formats

Signature Redacted

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Signature Redacted

Uploaded by

Copyright:

Available Formats

Water Quality and Hydrological Assessment and Modeling of

Bioretention Basins in Tropical Cities

2017 Massachusetts Institute of Technology. All rights reserved.

Accepted by ............................................ ............. Signature redacted

Submitted to the Department of Civil and Environmental Engineering

Dr. Peter Shanahan

Prof. Lloyd Chua

1.1 Research motivations

proportion of surface imperviousness in an urban catchment and hence results in hydrologic

groundwater in a pre-development manner, more surface overflow or stormwater runoff is

local groundwater recharge is also impaired.

1.2.1 Best Management Practice (BMP)

detain and treat polluted stormwater at its source.

ponds, bioretention basins, sand filters, and constructed wetlands.

study site is one such installation.

install and require minimal maintenance.

2-S-1, Chapter 2).

1.2.3 Design guidelines of bioretention basins

A well-designed bioretention basin ensures sufficient capture and treatment of pollutants to

to place but is often in the range of 12.7 - 25.4 mm (Massachusetts Department of

distinctly different rainfall-runoff characteristics, guidelines originated from temperate places

pollutant removal efficiency and hydrological capture efficiency respectively.

the tropical climate.

kinetics of a whole chain of biochemical reactions consisting of ammonification, nitrification,

of the study site respectively.

1.3 Research objectives

This study aims to fulfill the research objectives below:

2. Analyze the factors behind observed removal efficiency in terms of rainfall

guidelines for basins in the tropics in terms of WQV or WQD.

4. Explore the applicability of the RECHARGE hydrological model for continuous

simulation at the field scale.

5. Predict bioretention basin hydrological performance in terms of infiltration percentage as

a function of engineering design parameters such as detention depth, ratio of drainage

6. Recommend basin configurations for improvement of hydrological performance

water quality improvement in the bioretention basin subsurface soil media.

8. Predict bioretention basin water quality performance in terms of pollutant removal

percentage and effluent concentration at various levels of bacteria growth kinetics.

1.4 Thesis overview

Water Research & Technology, Vol 3, Issue 1, 78-91 (DOI: 10.1039/C6EW00285D).

Chapter 3 of this thesis, published as "Field evaluation of hydrological performance of a

based on the field results from 80 events.

Chapter 4 of this thesis presents continuous hydrological modeling of 80 events in June

observed field data that are presented in Chapter 3.

mathematical formulations in CW2D (Constructed Wetland 2D model) to carry out event-wise

observed field data that is presented in Chapter 2.

In Chapter 6, we draw design-specific findings from the previous chapters to make

recommendations for bioretention basin design in the tropics.

research directions are also suggested.

Field evaluation of water quality performance

between 10 to 30 mm) is recommended.

Hydrologic modification brought about by traditional concrete-lined drainage systems due to

to reinstate pre-development hydrological features through detention and treatment of polluted

suitability of currently inconsistent design guidelines.

and aims to suggest effective management guidelines.

Drainage 2011; Dierkes et al. 2015; Massachusetts Department of Environmental Protection

climates, including the United States and Australia.

of the receiving water body (Massachusetts Department of Environmental Protection 2008;

Minnesota Pollution Control Agency 2008; New Hampshire Department of Environmental

Stormwater Improvement Conceptualization) (Cooperative Research Centre for Catchment