Bioretention Model for Urban Runoff Treatment in a Tropical Climate: A Case Study at the Universiti Sains Malaysia


Abstract
Model for Urban Storm Water Improvement Conceptualism (MUSIC) software has been widely used to predict the treatment and performance of stormwater Best Management Practices (BMPs) such as bioretention for decision-making purposes in stormwater management. However, the calibration of bioretention models based on pollutant runoff characteristics in the tropics is rarely studied. This paper presents the calibration of bioretention model parameters using MUSIC software to treat polluted runoff in a tropical climate. The bioretention model was simulated based on a pilot study at the Universiti Sains Malaysia (USM) engineering campus to evaluate the flow rate and pollutant’s reduction performance. Two stages of calibration were conducted, with the first stage to calibrate the inflow and pollutant concentrations, and the second stage to further calibrate the k-C* model to fit the experimental results. The validation of the model was done using the percentage bias between modeled and experimental data to evaluate the accuracy of bioretention modeling using MUSIC software. Overall, the accuracy of this model increased after calibration and can be accepted, as the performance of bioretention models for total suspended solids, total nitrogen, and total phosphorus removal percentage are good or very good (-13%, -4%, and -39% respectively), whereas the flow rate reduction is satisfactory (17%).
1 Introduction
Urbanisation can be defined as the development of residential, commercial, and industrial areas, as there is a projected increase in human population of 60% by 2050 (United Nations 2015). Unplanned urbanisation results in large, vegetated areas converting to impervious areas, such as roads, car parks, roofs, and other paved surfaces (Mangangka et al. 2013). Thus, it can cause an increase in the amount of surface runoff and brings negative effects to the balance of the water cycle (Gatwaza et al. 2016). The climate change issue also contributes to the problems such as rising sea levels increasing the amount of runoff and urban flooding (Miller and Russell 1992). In addition, the increase of pollutants in runoff due to human activities such as farming, and the leaching of agriculture will contaminate the receiving water bodies. According to Osman et al. (2019), water pollution is a major challenge that will deteriorate our environment. Stormwater Best Management Practices (BMPs) were introduced in the 1970s to transform conventional stormwater management devices from an engineering-based approach to a nature-based approach in water quality and quantity control (Zakaria et al. 2003). Hence, stormwater BMPs can be considered as sustainable innovations to mitigate urban flooding, reduce runoff pollution and protect the environment (Fletcher et al. 2015).
Bioretention (also referred to as rain gardens, stormwater biofilters, or biofiltration systems), as one of stormwater’s BMPs, is the landscaped depression which can cater to runoff from the catchment, especially from an impervious surface (Shafique 2016). In previous studies, bioretention has been proven to help to reduce the peak flow and treat the pollutants effectively (Hunt et al. 2008; Jia et al. 2015; Wang et al. 2017). For stormwater quantity control mechanisms, both the exfiltration and evapotranspiration (ET) processes play a crucial role in performance of bioretention hydrology (Meng et al. 2014). However, the pollutant removal of bioretention can be affected by several mechanisms, such as filtration, adsorption, denitrification, and plant uptake (Skorobogatov et al. 2020). Thus, the process or mechanism may alter under different climates and level of development. Generally, tropical climate regions have warm temperatures, with intense and frequent rainfall throughout the whole year. Bioretention under tropical climates is expected to have better treatment performance as higher temperatures promote higher microbial activity (Trang et al. 2010). Apart from that, ET which is linked closely to wetting and drying periods, impacts nutrient capture by the plants in bioretention (Skorobogatov et al. 2020). However, it is difficult to study a complete bioretention system due to the restriction of land use, different laboratory test conditions, and unexpected results from experimental studies (Spraakman et al. 2020; Meng et al. 2014). Hence, modeling software or simulation can be used to solve those limitations from the field-based study, as the computational models simplify the complex process of bioretention by using mathematical equations and predict the performance of bioretention (Liu et al. 2014). As long-term performance of the bioretention system can be predicted via modeling, policymakers can adjust design standards that can fit local climate, vegetation, pollutant characteristics, and other local characteristics (Lisenbee et al. 2021).
Model for Urban Storm Water Improvement Conceptualism (MUSIC) software developed by Wong et al. (2002) acts as a conceptual and decision-making tool, which aids to conceptualize possible stormwater BMPs with suitable sizing to meet specified water quality goals. As the most-used numerical stormwater model software in Australia, MUSIC provides a visualisation interface and allows both single event-based and continuous simulation (Li et al. 2021). Natural or synthetic events can be created as rainfall inputs for event-based simulation, and it is very useful to study how bioretention will respond under certain storm events (Lisenbee et al. 2021).
In MUSIC, water quality simulation is performed based on rainfall data, catchment data, and soil parameters, whereas the water quality is simulated using event mean concentration (EMC) based on the study by Duncan (1999). MUSIC was used by Brisbane City Council to assess the performance of stormwater management facilities based on the reduction of pollutants (Wong et al. 2002). A study on the accuracy of bioretention modeling using MUSIC based on field studies located at Brisbane and Melbourne, Australia revealed that the prediction of bioretention performance had good accuracy for outflow rate, but was varying for the removal efficiency of TSS, TP, and TN (Imteaz et al. 2013). Dotto et al. (2011) found that effective impervious fraction (EIF) and Muskingum Cunge translation factor are sensitive to the rainfall runoff model in MUSIC. Based on a simulation done by Gagrani et al. (2014), 70% of runoff from existing stormwater control measures in Beaverdam Creek, USA would be required to divert into bioretention basins to meet the targeted removal percentage for TSS and TP. Moreover, sensitivity analysis was performed, and it was found that the bioretention model was sensitive to first-order decay rates of TSS, TN, and TP.
The simulation of pollutant load reduction in MUSIC involves both hydrologic routing algorithms and first-order kinetic models to simulate the pollutant generation and removal process (Darwin 2009). USTM (Universal Stormwater Treatment Model) is commonly used in water quality treatment modeling because it can represent the removal process that acts to reduce water quality constituents as they pass through the BMPs (Shojaeizadeh et al. 2021). The first order kinetics (or k-C*) model is expressed in Equation 1:
![]() |
(1) |
where:
Cin | = | input concentration (mg/L), |
Cout | = | output concentration (mg/L), |
C* | = | background concentration (mg/L), |
k | = | rate constant (m/year), and |
q | = | hydraulic loading (m/year). |
The k-C* model is used to describe the behaviour of pollutants, by which the pollutants concentration tends to move towards an equilibrium value or background concentration (C*) by an exponential decay process (Wong et al. 2002). Based on the principle of the k-C* model, the k values could reflect the settling velocities of targeted sediment size, which are predominantly influenced by physical removal processes and may be associated with biological and chemical processes in stormwater BMPs (eWater 2013). While for C*, this parameter reflects the minimum concentration, which is not removed by the treatment system. Higher k values / lower C* applied means the better performance of BMPs to treat pollutants. Thus, the default values of k and C* in MUSIC may not reflect the performance of the bioretention system for urban runoff treatment in the tropics.
Across Australia, MUSIC has been utilized by local government and the urban planning industry to enhance stormwater management since 2001. MUSIC has also been adopted by other countries such as the United States and Singapore, for preliminary design of stormwater control measures such as swales, bioretention, and wetlands (Gagrani et al. 2013; PUB 2018). However, according to Water Sensitive Urban Design (WSUD) Stormwater Quality Modeling Guide, the calibration should be based on local conditions with elements of climate data, pollutant parameters, and stormwater treatment parameters (Darwin 2009).
Pollutant sources and the transport mechanism in tropical regions are different from those in temperate regions due to the difference in rainfall regimes and management approaches (Chow et al. 2012). Moreover, the difference in the level of development affects the runoff characteristics. As reported by Chow and Yusop (2014), the average concentration of pollutants in Malaysia such as TSS, TN, and TP are 204 ± 182 mg/L, 3.0 ± 1.2 mg/L and 0.9 ± 0.2 mg/L respectively, which are higher than developed countries. According to Lucke (2018), the concentration of TSS (36.5–54.4 mg/L), TN (1.36–1.57 mg/L), and TP (0.21– 0.34 mg/L) in Australia is lower. Hence, for application in tropical countries, calibration is needed based on local conditions since the urban runoff characteristics are different due to the rainfall patterns, seasons, land use, and the level of development (Goh et al. 2019). Moreover, the error prediction of the size of bioretention, and other parameters for the design of bioretention, bring negative effects to the actual performance on site. The k-C* model adopted, which represents the treatment behaviour of bioretention, may not represent how the bioretention behaves in real life, especially under different climate conditions. For the bioretention module, the recommended k value for TSS is 4000-15000 m/yr, whereas for C* the value is 10-30 mg/L, which is in a wide range of values (eWater 2013). The calibration of k and C* values which are better suited to the bioretention under a tropical climate, is still lacking. Therefore, the accuracy of modeling needs to be evaluated, especially in the preliminary design stage via the calibration based on the local condition.
This study aims to (1) calibrate the bioretention model based on the local urban pollutant runoff characteristics and k-C* model, and (2) evaluate the accuracy of MUSIC models for bioretention systems in the tropics. The study, using MUSIC, based on local conditions, can help it to fit with the design requirements or guidelines in Malaysia such as the Urban Stormwater Management Manual (MSMA). The MUSIC software was used to gain a better understanding of the calibration based on local conditions. The literature gap can be answered via the research on the modeling based on local conditions such as climate, pollutant runoff characteristics, and treatment and performance of bioretention systems in the tropics.
2 Methodology
2.1 Site Description
A bioretention pilot site constructed at Universiti Sains Malaysia (USM) Engineering Campus, Nibong Tebal, which is located at the mainland of Penang state, was modeled in this study (Figure 1). As reported by Tan et al. (2022), the average annual rainfall of Penang is 2200 mm, and the temperature is around 23.2°C – 32.8°C, which is characterised by intense annual precipitation and hot temperature. For the pilot study, the natural runoff was collected from the main drain at Parit Buntar and was released to the bioretention site weekly. The collected runoff for dosing consists of a mixture of pollutant sources from runoff from the street, municipal grey water, and urban farming, all contributing to the high value of TSS, TN, and TP. The bioretention pilot study was designed to simulate the actual conditions to treat the runoff from the catchment of Parit Buntar. The volume of weekly dosing at the inlet of the bioretention system was 6 m3, which is equivalent to a threshold of 25 mm of runoff depth for the first flush captured by the system (Sage et al. 2015). The dimensions of the filter area are 3 m width and 4 m length. The design properties of the bioretention system located at the pilot site were used as the input data for modeling.
Figure 1 Geographical location and actual view of bioretention site
2.2 Pilot site setup
The bioretention site setup was constructed to be freely exposed to natural sunlight, rainfall, and weather conditions which are hot and humid. A layer of 150 mm of gravel as a drainage layer, followed by 150 mm of sand as a transition layer, was placed at the base of the pilot site, where the platform level is higher than the surrounding layers to prevent the ingress of surface runoff, such as pavement, roads, and parking lots. A half-inch diameter of perforated pipe was installed at the bottom of the site to collect the outflow. Then, a piece of geotextile was placed on the top of the perforated pipe to prevent the filter media from entering the pipe. The mixture of filter media with 600 mm depth was filled, and the dimension was based on the criterion suggested in MSMA (between 450 mm to 1000 mm). The layer of filter media consisted of a mixture of 20% topsoil, 60% fine river sand, and 20% compost, as suggested in MSMA (DID 2012). Tropical plants (Red Hot Hibiscus (Hibiscus rosa-sinensis), Amaryllis (Hippeastrum), Singapore Daisy (Sphagneticola trilobata), Lobster claw (Heliconia rostrata), and Alternanthera (Alternanthera cultivar)) were used as vegetation in this bioretention system. The selected plants satisfied the criteria suggested in MSMA, which can adapt to the local climate and soil, and rapid growth, and can be tolerant to the pollutants in the stormwater runoff. The selected plants were set out at the pilot site for three months prior to the experiment to ensure the plants grew well and adapted to the environmental conditions. The presence of these plants was expected to help with the removal of nutrients such as TN and TP. To allow the water to enter the system, a 50 mm PVC pipe was connected from an inlet tank to the inlet of the bioretention site. Collected runoff was released from the inlet tank with the control of the valve.
2.3 Data collection
The meteorological data collection, including rainfall and evapotranspiration data, was conducted before proceeding to the creation of a meteorological template for the rainfall-runoff simulation. The raw rainfall data at Jalan Matang Buloh, Bagan Serai (Station ID: 5005003) was acquired from the Department of Irrigation and Drainage (DID). Meanwhile, the monthly evaporation data of Bayan Lepas was acquired from Department of Meteorology, Malaysia. Apart from this, the secondary data, including the bioretention design parameters and the experimental data from the USM pilot study, were required. Six sets of water samplings were taken at the inlet and outlet of the bioretention system weekly. The concentration of pollutants from laboratory tests are important as data for the simulation and calibration of the bioretention modeling. The water samples at the outlet were collected at the half hour, second hour, fourth hour, and eighth hour in empty bottles and sent to the laboratory for testing. Three water quality parameters were monitored throughout the pilot study, which include total suspended solids (TSS), total nitrogen (TN), and total phosphorus (TP), using the methods APHA 2540 D, APHA 4500–N B, and APHA 4500 P C, respectively.
2.4 Development of the bioretention model
The model was simulated according to the meteorological data saved in the mlb file, the input information of the catchment (source node), and the bioretention system (treatment node). The setting of source node data includes the input of catchment area, rainfall-runoff parameters (impervious area properties, pervious area properties, and groundwater properties) and pollutant generation parameters. The bioretention system was simulated using MUSIC, based on the input properties in the treatment node, which include inlet properties, storage properties, filter media properties, infiltration, lining properties, vegetation properties, and outlet properties. The modeling of the bioretention system was based on the design parameters of the pilot site study. Those properties were identified for hydrologic routing of stormwater runoff and the prediction of treatment effectiveness.
2.5 Calibration
Calibration was done by adjusting the model parameters to ensure the modeling is performed as close as possible to the real conditions of the bioretention system located at the pilot site. Calibrations of this model can be divided into two stages, the calibration of the inflow, and the calibration of rate constant, k, and background concentration, C* for each pollutant. To model the flow rate generation at the source node, which is close to the pilot study, the catchment characteristics were calibrated by adjusting the imperviousness percentage. For the pollutant generation, the calibration of the inflow can be done by changing the parameters of log-normal distribution of each pollutant type. The mean and the standard deviation of TSS, TN, and TP were adjusted based on the sampling data from the inlet at the bioretention pilot site. Table 1 below shows the mean and standard deviation values that were used to calibrate the inflow concentrations for each pollutant at the source node. To ensure the simulated inflow concentrations are constant for each time of simulation, the mean estimation method was applied instead of using the default estimation method (stochastic generation method).
Table 1 Mean and Standard Deviation Value used for inflow calibration (sampling between 10/3/2021-24/3/2021).
Pollutant | Mean Concentration (mg/L) | Log10 (Mean) | SD | Log10 (SD) |
TSS | 193 | 2.29 | 25.01 | 1.398 |
TN | 18.7 | 1.27 | 0.68 | -0.167 |
TP | 4.49 | 0.65 | 0.79 | -0.102 |
For the calibration of k and C* values, the values used were based on the range used for the bioretention system as shown in Table 2. Based on eWater (2013), the suggested values have been derived from the swales’ experiment. However, the bioretention system of this study was designed like a basin. Hence, the recommended range of k and C* for the infiltration basin can be taken as a reference for the calibration (Table 2). As mentioned by Imteaz et al. (2013), most of the default parameter values are based on field experiments conducted in Brisbane and Melbourne. Thus, it is important to do the calibration to ensure the k-C* model suits local conditions, because both k and C* values are affected by pollutant characteristics, particle size distributions, and settling velocity distribution (Wong et al. 2002).
Table 2 Recommended range of k and C* for bioretention and infiltration basin (eWater 2013).
Type of Pollutants | TSS | TN | TP | |
Bioretention | k (m/yr) | 4000-15000 | 250-1000 | 3000-12000 |
C*(mg/L) | 10-30 | 1.1-1.7 | 0.08-0.18 | |
Infiltration Basin | k (m/yr) | 200-1000 | 30-50 | 150-500 |
C*(mg/L) | 12-15 | 0.7-1.3 | 0.05-0.13 |
First order kinetics graphs were plotted for each pollutant to obtain the suitable k values based on Equation 2 below. To convert the k' (hour-1) obtained from ln(Cout – C*) versus time graph into k (m/yr) values used for calibration, convert factor (depth of bioretention pond x 365 x 24) is needed to multiply.
![]() |
(2) |
where:
Cin | = | inflow concentration (mg/L), |
Cout | = | outflow concentration (mg/L), |
C* | = | background concentration (mg/L), |
k' | = | rate constant (hour-1), and |
t | = | time (hour). |
The laboratory data for outflow concentration from the bioretention site as shown in Table 3 (under sampling date of 10/3/2021,17/3/2021, and 24/3/2021) were used to plot the first order graphs. By adjusting the C* values based on the recommended range, the different combinations of k and C* values were obtained for each pollutant.
Table 3 Outflow concentration for TSS, TN, and TP (sampling between 10/3/2021-24/3/2021).
Sampling Name | Outflow Concentration of Pollutants (mg/L) | |||||
TSS | TN | TP | ||||
Mean | SD | Mean | SD | Mean | SD | |
BRS S0.5 | 82.0 | 31.1 | 5.10 | 0.22 | 2.41 | 0.58 |
BRS S2 | 78.7 | 23.9 | 4.67 | 0.77 | 1.97 | 0.80 |
BRS S4 | 66.7 | 22.3 | 4.60 | 0.50 | 1.54 | 0.49 |
BRS S8 | 46.7 | 17.9 | 4.03 | 0.25 | 1.32 | 0.35 |
2.6 Validation
The model was validated after calibration with different combinations of k and C* values. The best-fit combination of k and C* to the experimental results were determined. Validation of the model was done based on the flow rate reduction and pollutants concentration removal percentage for TSS, TN, and TP. The percentage bias (PBIAS) of simulation and experimental data was used as the method to validate the model. PBIAS can be calculated by using Equation 3 below:
![]() |
(3) |
where:
Qexp | = | experimental data, and |
Qmod | = | modeled data. |
The performance rating of the model was evaluated based on Table 4, using PBIAS statistical methods, in which the flowrate parameter, TSS parameter, TN, and TP parameters were evaluated under streamflow, sediment, and NP respectively.
Table 4 Performance rating for models (Moriasi et al. 2007).
Performance Rating | PBIAS (%) | ||
Streamflow | Sediment | N, P | |
Very Good | PBIAS<±10 | PBIAS<±15 | PBIAS<±25 |
Good | ±10≤PBIAS<±15 | ±15≤PBIAS<±30 | ±25≤PBIAS<±40 |
Satisfactory | ±15≤PBIAS<±125 | ±30≤PBIAS<±55 | ±40≤PBIAS<±70 |
Unsatisfactory | PBIAS≥±25 | PBIAS≥±55 | PBIAS≥±70 |
3 Results and discussion
3.1 Modeling results without calibration
Table 5 shows the comparison between laboratory results and modeling results without any calibration. Overall, the estimation of flow rate and pollutant concentrations were much lower than those that were measured from the pilot site. The low estimated inflow flow rate from the model explained that the water quality simulation was not only governed by rainfall data, but the catchment data and soil parameters also affected the model. However, the effect of soil parameters in this model can be neglected as the urban catchment (imperviousness greater than 30%) were not sensitive to soil parameters such as soil capacity and field capacity (Dotto et al 2011). The comparison of inflow pollutant concentrations revealed the simulation of water quality performance using default parameters of log-normal distribution of each pollutant in MUSIC was not suitable for this study.
Table 5 Experimental data from study site and modeled data without calibration.
Flow (x10-6m3/s) | TSS (mg/L) | TN (mg/L) | TP (mg/L) | |||||
Exp. Data | Modeled Data | Exp. Data | Modeled Data | Exp. Data | Modeled Data | Exp. Data | Modeled Data | |
Average Inflow | 100 | 51.8 | 193 | 45.7 | 18.7 | 2.21 | 4.49 | 0.20 |
Average Outflow | 35.0 | 11.5 | 46.7 | 8.29 | 4.03 | 0.31 | 1.32 | 0.03 |
Percentage Reduction (%) | 65 | 78 | 76 | 82 | 78 | 86 | 71 | 85 |
3.2 Modeling results in the first stage of calibration
Table 6 shows the modeling results for flow rate under different impervious conditions. Different imperviousness percentages (50%, 60%, 70%, 80%, 90%, and 100%) were used in the model to obtain the closest inflow flow rate. The results showed the increasing of imperviousness conditions caused a higher inflow flow rate. Also, the higher the imperviousness, the lower the percentage reduction of flow rate. Hence, the change of imperviousness affected the impact of the performance of the bioretention system. In this case, 60% imperviousness with 0.21 ha of catchment area was adopted, as it showed the closest inflow flow rate from the pilot study.
Table 6 Comparison of pilot site flow rate and modeling flow rate results with different imperviousness.
Pilot Site | Average inflow flowrate (m3/s) | Average outflow flowrate (m3/s) | Flowrate percentage reduction of bioretention (%) |
0.000100 | 0.000035 | 65 | |
Modeling with imperviousness (%) | Inflow flowrate (m3/s) | Outflow flowrate (m3/s) | Flow rate percentage reduction of bioretention model (%) |
100 | 0.000118 | 0.0000715 | 39 |
90 | 0.000112 | 0.0000626 | 44 |
80 | 0.000107 | 0.0000561 | 48 |
70 | 0.000105 | 0.0000472 | 53 |
60 | 0.0000961 | 0.0000385 | 60 |
50 | 0.0000907 | 0.0000302 | 68 |
Table 7 shows the comparison of data with the first stage of calibration, and it reflects the importance of understanding the rainfall-runoff process and pollutant generation process in modeling based on local conditions. Obviously, the simulation of pollutant generation was close to the experimental results after changing the parameters of log-normal distribution for each pollutant including TSS, TN, and TP. The pollutant removal efficiency also increased when compared with the model that simulated higher pollutant concentration. This resulted in the same conclusion obtained from the study by McNett et al. (2011), which concluded that the removal efficiency of the bioretention system increases with the increase of pollutant concentration in the runoff. Thus, the bioretention system has better performance for the treatment of more polluted runoff. Moreover, further calibration based on k and C* values is necessary, as the modeled outflow concentration as shown in Table 7 for TSS, TN, and TP cannot fit the experiment data.
Table 7 Comparison of data from first stage calibration.
Parameters | Experimental Data | Modeled Data1 | Modeled Data 1st Calibration | |
Flow (m3/s) | Average Inflow (10-6) | 100 | 51.8 | 96.1 |
Average Outflow (10-6) | 35 | 11.5 | 38.5 | |
Reduction Percentage (%) | 65 | 78 | 60 | |
TSS (mg/L) | Average Inflow | 193 | 45.7 | 195 |
Average Outflow | 46.7 | 8.29 | 16.4 | |
Reduction Percentage (%) | 76 | 82 | 92 | |
TN (mg/L) | Average Inflow | 18.7 | 2.21 | 18.6 |
Average Outflow | 4.03 | 0.31 | 3.08 | |
Reduction Percentage (%) | 78 | 86 | 83 | |
TP (mg/L) | Average Inflow | 4.49 | 0.20 | 4.47 |
Average Outflow | 1.32 | 0.03 | 0.39 | |
Reduction Percentage (%) | 71 | 85 | 91 |
3.3 Modeling results from the second stage of calibration
For this calibration, the combination of k and C* values used for simulation were identified. Since the bioretention site showed less effective treatment performance in removing pollutants as compared to the modeled bioretention system, the rule of thumb is choosing lower k or higher C* values. Lower k values represent a slower approach to equilibrium and hence a lower treatment efficiency. Table 8 summarizes the sensitivity of the model based on the resulting concentration towards the changes of k and C* values. The sensitivity of the model was observed when changing either one of parameters only. The results showed the sensitivity of the model towards the changes of k and C* is different based on different pollutants.
Table 8 Sensitivity of outflow concentration toward k and C* values.
Pollutants | Changes of Outflow concentration (mg/L) | |
Changes of k by 100m/yr | Changes of C* by 1 mg/L | |
TSS | 0.400 | 0.033 |
TN | 0.060 | 0.020 |
TP | 0.010 | 0.025 |
After several trials using different combinations of k and C* values, the combination of k and C* values which fit the experimental results and were suited to the bioretention pilot conditions were chosen and summarized in Table 9. The k value (850 m/yr) for TSS reflects the removal of TSS and was modeled as an infiltration basin (200-1000 m/yr), which means that the runoff was assumed to be subjected to pre-treatment to remove coarse sediment (eWater 2013). This can be further explained by Barrett et al. (2013), which shows that the role of vegetation does not significantly affect the removal of TSS in bioretention systems. In general, the TSS can be removed via filtration and sedimentation with the higher levels being related to physical process, and thus the TSS removal by bioretention with plants and without plants is similar. The adopted k value in this model was 350 m/yr, and it was within the range of suggested k for bioretention (250-1000 m/yr). The k value for TN reflects the role of plants in the bioretention site, and thus the model can be simulated as the behaviour of bioretention to remove N. Meanwhile, the adopted k value for TP in this model was 1200 m/yr and this value was out of the recommended range for bioretention (3000-12000 m/yr). The result reflected that the role of plants did not fully perform as bioretention treatment at the site. As mentioned by Vijayaraghavan et al. (2021), TN and TP removal can be up to 80-90% with the presence of plants. However, the reduction percentage of TN and TP from experimental results were only 78% and 71%. A logical inference can be made from the modeling results, which is that the plant density at the bioretention site, especially plants that are capable to remove P, was insufficient. Based on the literature review, study of the k and C* was rare in the tropics, especially for bioretention. Trang et al. (2010) reported that k values of TSS, TN, and TP for wetland treatment were 31-115 m/yr, 12-24 m/yr, and 40-66 m/yr. However, these values were not applicable to bioretention systems as the physical, chemical, and biological process for both systems are different.
Table 9 Summary of k and C* values used.
Pollutants | k (m/yr) | C* (mg/L) |
TSS | 850 | 12 |
TN | 350 | 1.2 |
TP | 1200 | 0.13 |
The modeling results were validated by comparing the reduction in performance from the bioretention site study. The modeling results after the calibration of k and C* values based on experimental data were validated. The performance rating of the modeling, based on the percentage bias between the experimental and modeling results, is shown in Table 10. Without calibration of k and C* values, the results revealed that the bioretention performance was overestimated for TSS, TN, and TP under the default setting. Hoban and Gambirazio (2021) also reported that MUSIC modeling overestimated the concentration in TN and TP reduction compared to experimental results without any calibration. The authors also concluded the performance parameters for bioretention in MUSIC modeling are required to re-calibrate. After calibration, the accuracy of the model increased and the performance rating for TSS and TN were very good (-13% and -4%, respectively) whereas the accuracy of the TP parameters was good (-39%). For the estimation of flow performance, the performance rating of the MUSIC modeling was satisfactory. The flow reduction from the pilot study (72%) was better than the estimated flow reduction from the MUSIC modeling (60%). This is because the simulation of evapotranspiration (ET) in the MUSIC modeling does not consider the transpiration of vegetation (Lisenbee et al. 2021). The ET losses contributed by the vegetation can be up to 52%, depending on the vegetation types (Skorobogatov et al. 2020). Hence, the modeling results were acceptable.
Table 10 Performance rating of modeling.
Parameters | Percentage Bias Between Experimental and Modeling without k-C* model calibration (%) | Percentage Bias Between Experimental and Modeling after calibration (%) | Performance Rating of Modeling after calibration |
Flow Rate | 17 | 17 | Satisfactory |
TSS | -24 | -13 | Very Good |
TN | -4 | -4 | Very Good |
TP | -49 | -39 | Good |
4 Conclusions
From the findings, the rainfall-runoff process in this modeling was not only contributed to by local meteorological data, catchment characteristics such as effective impervious area (EIA) based on local conditions also affected the flow entering the bioretention model. Calibration based on the pollutant runoff is essential, instead of using the default pollutant generation parameters due to the runoff characteristics varying with the climate condition, level of development, and type of land use. Moreover, the stormwater quality treatment performance for bioretention modeling was overestimated. Hence, the modeling results showed that the treatment behaviour of bioretention modeling highly relied on k and C* values. The prediction of stormwater quality treatments using MUSIC are accurate based on the performance rating of the model, which is good for TP (-39%), very good for TSS and TN (-13% and -4% respectively) and satisfactory for flow rate (17%). To obtain significant experiment data for bioretention performance, the bioretention site setup and data collection process for a long-term performance study is suggested with appropriate plant density. Future research can focus on how the long-term performance of bioretention in the tropics is vital and hence, the model can be re-calibrated in future to reflect the real capacity of bioretention to treat stormwater runoff. The quantity of data collected is vital for continuous stormwater modeling. Moreover, future research can also focus on other pollutants such as Biochemical Oxygen Demand (BOD), Chemical Oxygen Demand (COD), and Ammoniacal Nitrogen (AN), because they also contribute to water pollution. In short, MUSIC software can be used to estimate the performance of bioretention systems in the tropics.
Acknowledgments
Thanks to all the technical staff from River Engineering and the Urban Drainage Research Centre for their support and help throughout the study. This study was funded and supported by the Ministry of Higher Education Malaysia under the Fundamental Research Grant Scheme entitled "Phytoremediation performance of tropical plants in bioretention system for urban runoff treatment." with grant number FRGS/1/2019/TK10/USM/02/5.
References
- Barrett, M.E., M. Limouzin, and D.F. Lawler. 2013. “Effects of Media and Plant Selection on Biofiltration Performance.” Journal of Environmental Engineering 139 (4): 462-470.
- Chow, M.F., Z. Yusop, and M.E. Toriman. 2012. “Modelling runoff quantity and quality in tropical urban catchments using Storm Water Management Model.” International Journal of Environmental Science and Technology 9 (4): 737–748. https://doi.org/10.1007/s13762-012-0092-0
- Chow M., and Z. Yusop. 2014. “Characterization and source identification of stormwater runoff in tropical urban catchments.” Water Science Technology 69: 244–252.
- Darwin Harbour WSUD. 2009. Water Sensitive Urban Design Stormwater Quality Modeling Guide. Department of the Environment, Water, Heritage, and the Arts. Australian Government.
- DID (Department of Drainage and Irrigation). 2012. Urban stormwater management for Malaysia, 2nd ed. Malaysia Department of Drainage and Irrigation (DID), Malaysia, Kuala Lumpur.
- Dotto, C.B.S., A. Deletic, D.T. McCarthy, and T.D. Fletcher. 2011. “Calibration and Sensitivity Analysis of Urban Drainage Models: MUSIC Rainfall/Runoff Module and a Simple Stormwater Quality Model.” Australian Journal of Water Resources 15 (1): 85-94.
- Duncan, H.P.1999. ”Urban Stormwater Quality: A Statistical Overview.” Report 99/3, Cooperative Research Centre for Catchment Hydrology, February 1999.
- eWater. 2013. MUSIC v6 Documentation and Help Home. Available at: https://wiki.ewater.org.au/display/MD6 Accessed 27.03.21.
- Fletcher, T.D., W. Shuster, W.F. Hunt, R. Ashley, D. Butler, S. Arthur, S. Trowsdale, et al. 2015. “SUDS, LID, BMPs, WSUD and more – The evolution and application of terminology surrounding urban drainage.“ Urban Water Journal 12 (7): 525–542. https://doi.org/10.1080/1573062X.2014.916314
- Gagrani, V., J. Diemer, J. Karl, and C. Allan. 2013. “Assessing the Hydrologic and Water Quality Benefits of a Network of Stormwater Control Measures in a SE U.S. Piedmont Watershed.” JAWRA Journal of the American Water Resources Association 50 (1): 128-142.
- Gatwaza, O., X. Cao, and M. Beckline. 2016. “Impact of Urbanization on the Hydrological Cycle of Migina Catchment, Rwanda.” Open Access Library Journal 3, 1-12.
- Goh, H.W., K.S. Lem, N.A. Azizan, C.K. Chang, A. Talei, C.S. Leow, and N.A. Zakaria. 2019. “A Review of Bioretention Components and Nutrient Removal under Different Climates—Future Directions for Tropics.” Environmental Science and Pollution Research 26 (15): 14904–14919.
- Hoban A., and C. Gambirazio. 2021. “Bioretention performance: a review of field studies.” Australasian Journal of Water Resources, 25 (2): 202-210. https://doi.org/10.1080/13241583.2021.1984190
- Hunt, W.F., J.T. Smith, S.J. Jadlocki, J.M. Hathaway, and P.R. Eubanks. 2008. “Pollutant Removal and Peak Flow Mitigation by a Bioretention Cell in Urban Charlotte, N.C.” Journal of Environmental Engineering 134 (2): 403-408. https://doi.org/10.1061/(asce)0733-9372(2008)134:5(403)
- Imteaz, M.A., A. Ahsan, A. Rahman, and F. Mekanik. 2013. “Modeling stormwater treatment systems using MUSIC: Accuracy.” Resources, Conservation & Recycling 71: 15–21.
- Jia, H., X. Wang, C. Ti, Y. Zhai, R. Field, A.N. Tafuri, H. Cai, and S.L. Yu. 2015. “Field monitoring of a LID-BMP treatment train system in China.” Environmental Monitoring and Assessment 187, 373. https://doi.org/10.1007/s10661-015-4595-2
- Li, G., J. Xiong, J. Zhu, Y. Liu, and M. Dzakpasu. 2021. “Design influence and evaluation model of bioretention in rainwater treatment: A review.” Science of the Total Environment 787 (13): 147592. https://doi.org/10.1016/j.scitotenv.2021.147592
- Lisenbee, W.A., J.M. Hathaway, M.J. Burns, and T.D. Fletcher. 2021. “Modeling Bioretention Stormwater Systems: Current Models and Future Research Needs.” Environmental Modeling & Software 144, 105146.
- Liu, J., D.J. Sample, C. Bell, and Y. Guan. 2014. “Review and Research Needs of Bioretention Used for the Treatment of Urban Stormwater.” Water (Switzerland) 6 (4): 1069-1099.
- Lucke T., D. Drapper, and A. Hornbuckle. 2018. “Urban stormwater characterisation and nitrogen composition from lot-scale catchments—new management implications.” Science of the Total Environment 619: 65–71.
- Mangangka, I.R. 2013. “Role of Hydraulic Factors in Constructed Wetland and Bioretention Basin Treatment Performance.” Ph.D. thesis, Queensland University of Technology.
- McNett, J.K., W.F. Hunt, and A.P. Davis.2011. “Influent Pollutant Concentrations as Predictors of Effluent Pollutant Concentrations for Mid-Atlantic Bioretention.” Journal of Environmental Engineering 137 (9): 790–799.
- Meng, Y., H. Wang, J. Chen, and S. Zhang. 2014. “Modeling Hydrology of a Single Bioretention System with HYDRUS-1D.” Scientific World Journal, 2014. https://doi.org/10.1155/2014/521047
- Miller, J.R., and G.L. Russell. 1992. “The Impact of Global Warming on River Runoff.” Journal of Geophysical Research 97, D3: 2757-2764.
- Moriasi, D.N., J.G. Arnold, M.W. Liew, R.L. Van Bingner, R.D. Harmel, and T.L. Veith. 2007. “Model Evaluation Guidelines for Systematic Quantification of Accuracy in Watershed Simulations.” Transactions of the ASABE 50 (3): 885–900.
- Osman, M., K.W. Yusof, H. Takaijudin, H.W. Goh, M.A. Malek, N.A. Azizan, A.A. Ghani, and A.S. id. Abdurrasheed. 2019. “A review of nitrogen removal for urban stormwater runoff in bioretention system.” Sustainability (Switzerland) 11 (19). https://doi.org/10.3390/su11195415
- PUB (Public Utilities Board of Singapore). 2018. Bioretention Basin, Condensed Booklet on Engineering Procedures for ABC Water Design Features, Chapter 7, Public Utilities Board of Singapore.
- Sage, J., E. Berthier, and M. Gromaire. 2015. “Stormwater Management Criteria for On-Site Pollution Control: A Comparative Assessment of International Practices.” Environmental Management 56. https://doi.org/10.1007/s00267-015-0485-1
- Shafique, M. 2016. “A review of the bioretention system for sustainable storm water management in urban areas.” Materials and Geoenvironment 63 (4): 227–236. https://doi.org/10.1515/rmzmag-2016-0020
- Shojaeizadeh, A., M. Geza, C. Bell, J. McCray, and T. Hogue. 2021. “A Site-Scale Tool for Performance Based Design of Stormwater Best Management Practices.” Water 2021 (13): 844.
- Skorobogatov, A., J. He, A. Chu, C. Valeo, and B. van Duin. 2020. “The Impact of Media, Plants and Their Interactions on Bioretention Performance: A Review.” Science of the Total Environment 715, 136918. https://doi.org/10.1016/j.scitotenv.2020.136918
- Spraakman, S., T.F.M. Rodgers, H. Monri-Fung, A. Nowicki, M.L. Diamond, E. Passeport, M. Thuna, and J. Drake. 2020. “A Need for Standardized Reporting: A Scoping Review of Bioretention Research 2000–2019.” Water (Switzerland) 12 (11): 1–35.
- Tan, M.L., F. Zhang, C.J.C. Derek, K.H. Yu, S.M. Shaharudin, N.W. Chan, and A.R. Asyirah. 2022. “Spatio-temporal analysis of precipitation, temperature and drought from 1985 to 2020 in Penang, Malaysia.” Water Supply 22 (5): 4757–4768. https://doi.org/10.2166/ws.2022.140
- Trang, N.T., D. Konnerup, H.H. Schierup, N.H. Chiếm, L.A, Tuan, and H. Brix. 2010. “Kinetics of pollutant removal from domestic wastewater in a tropical horizontal subsurface flow constructed wetland system: Effects of hydraulic loading rate.”Ecological Engineering 36: 527-535.
- United Nations. 2015. “World Urbanization Prospects: The 2014 Revision” (ST/ESA/SER.A/366), Department of Economic and Social Affairs, Population Division, New York.
- Vijayaraghavan, K., B.K. Biswal, M.G. Adam, S.H. Soh, D.L. Tsen-Tieng, A.P. Davis, S.H. Chew, P.Y. Tan, V. Babovic, and R. Balasubramanian. 2021. “Bioretention Systems for Stormwater Management: Recent Advances and Future Prospects. Journal of Environmental Management, 292 (May), 112766.
- Wang, J., L.H.C. Chua, and P. Shanahan. 2017. “Evaluation of Pollutant Removal Efficiency of a Bioretention Basin and Implications for Stormwater Management in Tropical Cities.” Environmental Science: Water Research and Technology 3 (1): 78–91.
- Wong, T H.F., T.D. Fletcher, H.P. Duncan, J.R. Coleman, and G.A. Jenkins. 2002. “A Model for Urban Stormwater Improvement Conceptualization.” A. Rizzoli, and A. Jakeman, editors. Integrated assessment and decision support. Proceedings of the 1st biennial meeting of the International Environmental Modeling and Software Society, Vol. 1, p. 48–53.
- Zakaria, N.A., A. Ab Ghani, R. Abdullah, L. Mohd. Sidek, and A. Ainan. 2003. “Bio‐ecological Drainage System (BIOECODS) for Water Quantity and Quality Control.” International Journal of River Basin Management 1 (3): 237-251.