This may be the author’s version of a work that was submitted/accepted for publication in the following source:
Liu, An, Guan, Yuntao,Egodawatta, Prasanna, &Goonetilleke, Ashantha (2016)
Selecting rainfall events for effective Water Sensitive Urban Design: A case study in Gold Coast City, Australia.
Ecological Engineering, 92, pp. 67-72.
This file was downloaded from: https://eprints.qut.edu.au/95139/
c
Consult author(s) regarding copyright matters
This work is covered by copyright. Unless the document is being made available under a Creative Commons Licence, you must assume that re-use is limited to personal use and that permission from the copyright owner must be obtained for all other uses. If the docu-ment is available under a Creative Commons License (or other specified license) then refer to the Licence for details of permitted re-use. It is a condition of access that users recog-nise and abide by the legal requirements associated with these rights. If you believe that this work infringes copyright please provide details by email to [email protected]
Notice: Please note that this document may not be the Version of Record (i.e. published version) of the work. Author manuscript versions (as Sub-mitted for peer review or as Accepted for publication after peer review) can be identified by an absence of publisher branding and/or typeset appear-ance. If there is any doubt, please refer to the published source.
Selecting rainfall events for effective Water Sensitive Urban Design:
A case study in Gold Coast City, Australia
An Liu1,2, Yuntao Guan1,3*, Prasanna Egodawatta4, Ashantha Goonetilleke4
1
Institute of Environment, Graduate School at Shenzhen, Tsinghua University, 518055 Shenzhen, People’s Republic of China;
2
College of Chemistry and Environmental Engineering, Shenzhen University, Shenzhen 518060, People’s Republic of China
3
School of Environment, Tsinghua University, Beijing100084, People’s Republic of China
4
Science and Engineering Faculty, Queensland University of Technology (QUT), P.O. Box 2434, Brisbane, QLD 4001, Australia
*Corresponding author:
E-mail: [email protected]; Tel: 86-755-26036702; Fax: 86-755-26036702
Highlights
• Presents approach for selecting rainfall events for stormwater treatment design • Selecting high intensity-short duration events is the most effective for design • Stormwater treatment design based on more frequent events is the most feasible
Abstract: The current approach for protecting the receiving water environment from
urban stormwater pollution is the adoption of structural measures commonly referred to as Water Sensitive Urban Design (WSUD). The treatment efficiency of WSUD measures closely depends on the design of the specific treatment units. As stormwater quality is influenced by rainfall characteristics, the selection of appropriate rainfall events for treatment design is essential to ensure the effectiveness of WSUD systems. Based on extensive field investigations in four urban residential catchments based at Gold Coast, Australia, and computer modelling, this paper details a technically robust approach for the selection of rainfall events for stormwater treatment design using a three-component model. The modelling results confirmed that high intensity-short duration events produce 58.0% of TS load while they only generated 29.1% of total runoff volume. Additionally, rainfall events smaller than 6-month average recurrence interval (ARI) generates a greater cumulative runoff volume (68.4% of the total annual runoff volume) and TS load (68.6% of the TS load exported) than the rainfall events larger than 6-month ARI. The results suggest that for the study catchments, stormwater treatment design could be based on the rainfall which had a mean value of 31 mm/h average intensity and 0.4 h duration. These outcomes also confirmed that selecting smaller ARI rainfall events with high intensity-short duration as the
threshold for treatment system design is the most feasible approach since these events cumulatively generate a major portion of the annual pollutant load compared to the other types of events, despite producing a relatively smaller runoff volume. This implies that designs based on small and more frequent rainfall events rather than larger rainfall events would be appropriate in the context of efficiency in treatment performance, cost-effectiveness and possible savings in land area needed.
Keywords: Multivariate analysis; Stormwater pollutant processes; Stormwater
quality; Stormwater treatment; Water Sensitive Urban Design (WSUD)
1. Introduction
Treatment efficiency of Water Sensitive Urban Design (WSUD) structural elements are closely dependent on the design specifications (Hatt et al., 2007). Additionally, effective treatment design not only entails efficient treatment performance, but is also related to cost-effectiveness such as device size and land area required (Hatt et al., 2006). Treatment system design should take into account the influential factors in relation to stormwater quality, with rainfall characteristics playing a significant role (Dechesne et al., 2004; Kleinman et al., 2006; Smullen et al., 1999; Liu et al., 2012; Brodie and Rosewell, 2007; Alias et al., 2014). For example, Brodie and Rosewell (2007) noted the influence of rainfall intensity on stormwater quality and used the square of the rainfall intensity to measure the kinetic energy available in rainfall for the wash-off process. Alias et al. (2014) reported that rainfall depth and rainfall intensity are two key rainfall characteristics which influence the wash-off process compared to the antecedent dry period. Therefore, the selection of appropriate rainfall events for design is critical to ensure the effectiveness of the treatment system.
In treatment design, the conventional approach is to consider stormwater quality as a stochastic variable, irrespective of the nature of the rainfall event (Wong and Somes, 1995; Wong et al., 2002). Such an approach can diminish the overall performance of the treatment system since stormwater quality can be influenced by a range of rainfall
types such as high intensity-short duration and low intensity-long duration events (Liu et al., 2012). Other than stormwater quality, quantity characteristics need to be
considered in treatment design. The design volume being too small can lead to a large number of rainfall events exceeding the capacity of the treatment device. Alternately, there will be increased cost for limited gain in efficiency if the design volume is too large (Guo and Urbonas, 1996). These facts highlight the need for the prudent selection of design rainfall events for stormwater treatment design based on stormwater quality and the appropriate runoff volume.
The current research study was underpinned by the knowledge created in previous research undertaken by Liu et al. (2012) in order to provide a technically robust approach for the selection of rainfall events for stormwater treatment design. In the research conducted by Liu et al. (2012), three rainfall types were defined based on monitored rainfall events and the resulting stormwater runoff quality as given by pollutant event mean concentration (EMC). The three rainfall types were, high intensity-short duration (Type 1), high intensity-long duration (Type 2) and low intensity-long duration (Type 3). It was found that Type 1 and Type 2 tend to generate relatively higher EMC values in stormwater runoff while the Type 3 tends to produce relatively lower EMCs. Additionally, Liu et al. (2012) have pointed out the
correlationship between stormwater quality and rainfall characteristics such as
average rainfall intensity, rainfall duration and antecedent dry days. Average intensity plays a more important role in relation to stormwater quality compared to antecedent dry days, while long rainfall durations have a dilution effect and can lead to relatively low EMCs.
The knowledge created by Liu et al. (2012) can provide guidance in relation to stormwater treatment design in the context of water quality. However, the impact of runoff volume generated by rainfall events was not investigated in their research study. As noted above, design of a stormwater treatment system should also take into
account the runoff volume captured. In this context, modelling is an appropriate approach since it can generate the required stormwater quality and quantity data. Accordingly, the research study was based on extensive hydrologic and stormwater quality modelling and underpinned by field data collected from four urban residential catchments. The research outcomes provide guidance for efficient and cost-effective stormwater treatment design.
2. Materials and methods
2.1 Study areasFour urban residential catchments were selected for the modelling study. They were, Alextown, Gumbeel, Birdlife Park and Highland Park, which are located at Gold Coast, Australia. Gold Coast is situated to the south of Brisbane, the state capital and is the sixth largest city in Australia. Gold Coast has a subtropical climate. The
average summer temperature ranges from 19oC to 29oC and winter from 16oC to 21oC. The mean of annual rainfall depth is around 1300 mm. Alextown, Gumbeel and
Birdlife Park are in effect subcatchments of the larger Highland Park catchment. The study catchments are characterised by the similar geology based on the Neranleigh-Fernvale metasediments and have similar predominant soil type (mainly Kurosols) and topography. This ensured that these factors would not differently influence the stormwater runoff quality characteristics. A small tributary of the Nerang River,
Bunyip Brook is the study catchment’s primary stormwater drainage, which starts from the westward hilly region and flows towards the Nerang River. The integrated pipe and channel network connecting various parts of the catchment to the tributary further facilitates stormwater drainage. The four study catchments have been continuously monitored for water quantity, quality and rainfall using automatic monitoring stations established at catchment outlets for runoff flow measurement and collection of samples for laboratory analyses for water quality parameters (see Fig. S1 and S2 in the Supplementary Information (Goonetilleke et al., 2005). The four
catchments are shown in Figure 1 while the data monitoring program is provided in the Supplementary Information.
Figure 1 Study catchments
2.2 Model description
Commercially available stormwater quality models are typically not adequately sensitive to effectively simulate all of the important rainfall characteristics such as, antecedent dry days and rainfall intensity. This leads to difficulties in investigating relationships between different rainfall characteristics and stormwater quality. In order to overcome this constraint, the three-component model illustrated in Figure 2, which incorporates the relevant hydrologic processes (quantity) and pollutant
processes (quality) in relation to build-up and wash-off was adopted for this research study.
Highland Park: 105.1 ha extent; 40% impervious fraction; mixed land use with significant residential fraction; located close to busy highway; large extent of through roads
Birdlife Park Alextown Gumbeel
Gumbeel Alextown Birdlife Park Birdlife Park: 7.5 ha extent;
45.8% impervious fraction; residential - detached housing; Low population density
Alextown: 1.7 ha extent; 70% impervious fraction; residential -townhouses ; High population density Gumbeel: 1.2 ha extent; 70% impervious fraction; residential - duplex housing developed around cul-de-sac; High population density
Catchment boundary Catchment outlet Rain gauge
Quantity simulations Build-up estimations Wash-off estimations
Total runoff volume Total pollutant load
EMC=
Total pollutant load Total runoff volume
Component 1 Component 2 Component 3
Computer model Equations
Pollutant EMC
Figure 2 Three-component model simulation approach
The simulation of stormwater quantity was undertaken using MIKE URBAN (2008). This model was selected after a rigorous review of commonly used models in
Australia. Detailed information on the stormwater quantity estimation can be found in Liu (2011), while information regarding model setup and the calibration procedure adopted (see Fig. S3-S5 and Table S3-S5) is provided in the Supplementary
Information.
Pollutant build-up and wash-off analysis were undertaken using the mathematical equations developed by Egodawatta and Goonetilleke (2006) and Egodawatta et al. (2007), respectively, as shown in Equation (1) and (2) below. Detailed information on the derivation of coefficients can be found in Table S1 and S2 in the Supplementary Information.
B= aDb (1)
Where
B -Build-up load (g/m2) D -Antecedent dry days (days)
a -Multiplication build-up coefficient (dimensionless) b -Power build-up coefficient (dimensionless)
𝐹𝐹𝐹𝐹 = 𝑊𝑊/𝑊𝑊0 = 𝐶𝐶𝐹𝐹(1 − 𝑒𝑒−𝑘𝑘𝑘𝑘𝑘𝑘) (2)
Where
Fw -Wash-off fraction (%)
W - Weight of the material washed-off after time t (g/m2) W0 - Initial weight of the material on the catchment surface (g/m2)
CF -Capacity factor (dimensionless)
I - Rainfall intensity (mm/h)
k -Wash-off coefficient (dimensionless)
The simulated pollutant loads, runoff volumes and pollutant EMCs for each rainfall event were used in the subsequent analysis. Total solids (TS) was adopted as the indicator pollutant. This is due to the importance of solids which act as a mobile substrate in the transport of other pollutants (Nelson and Booth, 2002; Vaze and Chiew, 2002; Rossi et al., 2005).
2.3 Data analysis
Data analysis was undertaken as follows:
1) Runoff volume, pollutant build-up and wash-off were simulated using the three-component model discussed above for the same 41 rainfall events investigated by Liu et al. (2012) to derive pollutant EMC values.
2) The three-component model was validated for its sensitivity to important rainfall characteristics such as antecedent dry days, average rainfall intensity and rainfall duration to assess its ability and accuracy for developing the approach for rainfall event selection. This was undertaken by comparing the relationship between simulated pollutant EMC (obtained from Step 1) and rainfall characteristics to the relationship between monitored pollutant EMC (Liu et al., 2012) and rainfall characteristics.
3) Rainfall events from a selected representative year for the study region were simulated using the validated three-component model for pollutant loads and runoff volumes and the results were analysed to develop the approach for rainfall event selection for stormwater treatment design.
3. Results and discussions
3.1 Validation of the three-component model
The 41 rainfall events were simulated using the three-component model (Figure 2) for each catchment. Accordingly, runoff volume and TS EMCs were estimated. In order to identify the relationship between simulated TS EMCs and rainfall characteristics, Principal Component Analysis (PCA) was employed. The use of PCA was due to its versatility for investigating the possible correlations between variables and objects (Kokot et al., 1998). StatistiXL software (StatistiXL, 2007) was used for the PCA.
The TS EMC values estimated for each of the 41 rainfall event at each catchment were considered as objects whilst average rainfall intensity (AgI), rainfall duration (RD), antecedent dry days (ADD) and simulated TS EMCs were considered as
variables for the PCA. Consequently, a matrix (164×4) was generated. Figure 3 shows the resulting PCA biplot.
Figure 3 PCA biplot for simulated data
(TS=Total solids; AgI=Average rainfall intensity; ADD=Antecedent dry days; RD=Rainfall duration)
As shown in Figure 3, most of the Type 3 rainfall events (low intensity-long duration) are positioned on the negative PC1 axis and clustered together whilst Type 1 (high intensity-short duration) and Type 2 (high intensity-long duration) rainfall events are positioned on the positive PC1 axis and are relatively scattered. Additionally, the average rainfall intensity (AgI) vector forms an acute angle with the TS EMC vector, followed by antecedent dry days (ADD) vector whilst the rainfall duration (RD)
vector is almost opposite to the TS EMC vector forming an obtuse angle. Furthermore, both AgI and ADD as well as TS EMC vectors are projected on the positive PC1 axis whilst RD is projected on the negative PC1 axis.
These observations mean that the TS EMC is strongly correlated with the average rainfall intensity, followed by antecedent dry days whilst rainfall duration is
negatively correlated with TS EMC. Additionally, the PCA outcomes imply that Type 1 and Type 2 events tend to produce relatively higher pollutant EMC than Type 3. The results are in agreement with the conclusions derived by Liu et al. (2012) as discussed above. This means that the three-component model is sensitive to the important rainfall characteristics, including average rainfall intensity, antecedent dry days and rainfall duration and is able to adequately represent the relationship between rainfall characteristics and stormwater quality. Therefore, the model was suitable for the envisaged analysis.
3.2 Selecting rainfall events in a representative year
Due to the fact that the annual pollutant load is one of the important parameters employed to define pollutant export from a given catchment (Skinner et al., 2009), for further analysis, it was important to select a representative year where the rainfall
0 RD ADD AgI TS EMC -2 -1 0 1 2 3 4 5 -4 -2 0 2 4 6 8 PC A 2 ( 25. 6% ) PCA 1 (41.9%)
Type 1 and Type 2 Type 3
characteristics are typical for the study region. The rainfall events for the selected representative year needed to have moderate characteristics rather than extreme characteristics such as extremely high rainfall amounts or rainfall frequency. Additionally, as rainfall events can be classified into three types in the context of stormwater quality (Liu et al., 2012), the representative year needed to include all three types. Evaluation of the selected representative year is discussed below, whilst detailed information including the reasons for the representative year selection and rainfall data for the selected representative year is given in the Supplementary Information.
After careful investigation of historical records from 1992-2010, the rainfall for 2003 was tentatively deemed suitable as the events for that year was the closest to the average annual rainfall depth and the average number of rainy days for the study catchments (see rainfall data for 2003 in Figure S6 of Supplementary Information). It was found that there were 65 rainfall events where the rainfall intensity was less than 5 mm/h. Such low intensity does not generate appreciable pollutant loads due to the low kinetic energy and resulting inability to detach pollutants from catchment surfaces (Egodawatta et al., 2007). Therefore, these 65 rainfall events were not considered in the analysis undertaken, whilst the remaining 66 rainfall events were included.
Furthermore, these 66 events were also investigated to ensure the presence of the three rainfall types. This was undertaken by plotting these events in an Intensity- Frequency-Duration (IFD) plot as shown in Figure 4. It can be observed that 15 events (shown as circular data points) have a relatively high average intensity (> 20 mm/h; 31 mm/h on average) but short duration (<1.5 h; 0.4h on average), whilst 50 events (shown as square data points) have a lower average intensity (<20 mm/h; 10.2 mm/h on average) and most of them have relatively long duration. One event (shown as a triangular data point) has an average intensity of 27.3 mm/h and 7.4 h duration. Therefore, it was concluded that the rainfall events in 2003 encompassed the three rainfall types under consideration. Accordingly, 2003 was considered appropriate as the representative year.
Figure 4 Rainfall events of 2003 in IFD plot
3.3 Analysis of modelling results
The identified 66 rainfall were simulated using the validated three-component model to generate runoff volume and TS loads for the four study catchments. The cumulative simulated runoff volumes and TS loads for the different rainfall types were
determined by the summation of runoff volumes and loads derived for the four catchments as shown in Figure 5.
Figure 5 Comparison of three types of rainfall events
5 10 15 20 25 30 35 40 45 50 55 0 1 2 3 4 5 6 7 8 Av era ge ra in fa ll i nt en sit y ( mm/ h) Rrainfall duration (h) 0 10 20 30 40 50 60 70 80 90 100
No. of rainfall events TS load Runoff volume
Per cen ta ge % Type 3 Type 2 Type 1 Type 2 Type 3 Type 1 6-month ARI Type 2 Type 3 Type 1 6-month ARI 22.7% 75.8% 37.0% 58.0% 51.8% 29.1% 1.5% 5.0% 19.1% 9
It was found that although the number of Type 1 events account for only 22.7% of all rainfall events, they generated 58.0% of the TS load. The total runoff volume from Type 1 events only accounts for 29.1%. This means that compared with the other two types, Type 1 rainfall events are responsible for most of the pollutants export from the catchments, although they occur relatively less frequently. Type 2 events occur even more rarely and only generated 5% of TS load but generated 19.1% of the runoff volume. Type 3 events occur more frequently, but do not generate high pollutant loads. Based on these observations, it can be concluded that focusing on Type 1 events which are high intensity-short duration rainfall events can lead to greater efficiency in stormwater treatment system performance due to their greater pollutant generation ability. This can be further supported by the relatively smaller runoff volume generated by Type 1 events. This is due to the fact that a smaller runoff volume can be effectively captured by the treatment system, whilst a larger runoff volume may significantly by-pass the treatment system without achieving the desired removal of pollutants.
As evident in Figure 4, most of the rainfall events are less than 6-month ARI. This means that an ARI less than 6-month would be appropriate in terms of stormwater quality treatment design. Figure 6 below provides a comparison of the total runoff volume and TS loads for rainfall events larger than (4 events) and smaller than 6-month ARI (62 events). It was found that smaller events generate a greater cumulative runoff volume (68.4% of the total annual runoff volume) and TS load (68.6% of the TS load exported) than the larger events in all the four catchments. This means that the rainfall events less than 6-month ARI is responsible for a major fraction of the runoff volume together with most of the pollutants load. The outcomes confirm that stormwater treatment design based on large rainfall events may not provide a satisfactory cost-benefit outcome. Treating small and more frequent rainfall events should be the preferred option since its benefits include efficiency in treatment performance, cost-effectiveness and possible savings in land area needed.
Figure 6 Comparison based on 6-month ARI for all four catchments combined
0 10 20 30 40 50 60 70 80 90 100
Runoff volume TS load
Per
cen
ta
ge %
Larger than 6-month ARI Smaller than 6-month ARI
31.6% 31.4%
68.4%% 68.6%
Accordingly, for the four investigated catchments, stormwater treatment design could be based on Type 1 events for 2003 which had a mean value of 31 mm/h average intensity and 0.4 h duration, as these events generated a greater pollutants load than the other types of rainfall events. Furthermore, treatment design based on these rainfall parameters can be expected to capture a major portion of the annual runoff volume as other smaller events (including Type 3) can also be captured. Although this set of design rainfall parameters were based on the four catchments, this approach provides a technically robust methodology for the selection of a threshold rainfall event for stormwater treatment design.
4. Conclusions
The study outcomes provide an innovative approach for the selection of rainfall parameters for designing stormwater treatment systems taking into consideration both, stormwater quality and quantity. Selecting smaller ARI events with high intensity-short duration as the threshold for the design of treatment systems is the feasible approach since these events cumulatively produce a major portion of the annual pollutant load compared to the other types of rainfall events, despite producing a relatively smaller runoff volume. This confirms that designs based on small and more frequent rainfall events rather than larger rainfall events would be appropriate in the context of efficiency in treatment performance, cost-effectiveness and possible savings in land area needed.
Supplementary Information
Supplementary information provides further details about the three-component model, the derivation of the coefficients for pollutant build-up and wash-off, model
calibration and validation for stormwater quantity and quality and the procedure adopted for the selection of the representative rainfall year.
References
Alias N., Liu A., Egodawatta P. and Goonetilleke A. 2014 Sectional analysis of the pollutant wash-off process based on runoff hydrograph. Journal of Environmental Management, 134, 63-69.
Brodie I. and Rosewell C. 2007 Theoretical relationships between rainfall intensity and kinetic energy variants associated with stormwater particle wash-off. Journal of Hydrology, 340, 40-47.
Dechesne M., Barraud S. and Bardin J.P. 2004 Indicators for hydraulic and pollution retention assessment of stormwater infiltration basins. Journal of Environmental Management, 71(4), 371-380.
Egodawatta P. and Goonetilleke A. 2006 Characteristics of pollutants build-up on residential road surfaces. The 7th International Conference on Hydroscience and Engineering (ICHE), Philadelphia USA. Available for downloading at
http://eprints.qut.edu.au/12660/
Egodawatta P., Thomas E. and Goonetilleke A. 2007 Mathematical interpretation of pollutant wash-off from urban road surfaces using simulated rainfall. Water Research, 41(13), 3025-3031.
Goonetilleke A., Thomas E., Ginn S. and Gilbert D. 2005 Understanding the role of land use in urban stormwater quality management. Journal of Environmental
Management, 74(1), 31-42.
Guo J. C. Y. and Urbonas B. 1996 Maximized detention volume determined by runoff capture ratio. Journal of Water Resources Planning and Management, 122(1), 33-39.
Hatt B. E., Deletic A. and Fletcher T. D. 2006 Integrated treatment and recycling of stormwater: A review of Australian practice. Journal of Environmental Management, 79(1), 102-113.
Hatt B. E., Deletic A. and Fletcher T. D. 2007 Stormwater reuse: designing biofiltration systems for reliable treatment. Water Science and Technology, 55(4), 201-209.
Kleinman P.J.A., Srinivasan M.S., Dell C.J., Schmidt J.P., Sharpley A.N. and Bryant R.B. 2006 Role of rainfall intensity and hydrology in nutrient transport via surface runoff. Journal of Environmental Quality, 35(4), 1248-1259.
Kokot S., Grigg M., Panayioyou H. and Phuong T.D. 1998 Data interpretation by some common chemometric methods. Electroanal, 10(16),1081-1088.
Liu A. 2011 Influence of rainfall and catchment characteristics on urban stormwater quality. PhD thesis, Queensland University of Technology, Brisbane, Australia.
Liu A., Goonetilleke A. and Egodawatta P. 2012 Taxonomy for rainfall events based on pollutant wash-off potential in urban areas. Ecological Engineering, 47, 110-114.
MIKE URBAN 2008 MIKE URBAN Model User Manual, Danish Hydraulic Institute.
Nelson E. J. and Booth D. B. 2002 Sediment sources in an urbanizing, mixed land-use watershed. Journal of Hydrology, 264(1-4), 51-68.
Rossi L., Krejci V., Rauch W., Kreikenbaum S., Fankhauser R. and Gujer W. 2005 Stochastic modeling of total suspended solids (TSS) in urban areas during rain events. Water Research, 39(17), 4188-4196.
Skinner L., Townsend S. and Fortune J. 2009 The impact of urban land-use on total pollutant loads entering Darwin Harbour, Report 06/2008D, Department of Natural Resources, Environment, the Arts and Sport, Darwin, Australia.
Smullen J.T., Shallcross A.L. and Cave K.A. 1999 Updating the U.S. Nationwide urban runoff quality data base. Water Science and Technology, 39(12), 9-16.
StatistiXL 2007 StatistiXL manual, Version 1.10, Western Australia
Vaze J. and Chiew F. H. S. 2002 Experimental study of pollutant accumulation on an urban road surface. Urban Water, 4(4), 379-389.
Wong T. H. F., Fletcher T. D., Duncan H. P., Coleman J. R. and Jenkins G. A. 2002 A model for urban stormwater improvement conceptualisation. The 9th International Conference on Urban Drainage. Portland, Oregon.
Wong T.H.F. and Somes N.L.G. 1995 A stochastic approach to designing wetlands for stormwater pollution control. Water Science and Technology, 32(1), 145-151.
SUPPLEMENTARY INFORMATION
Selecting rainfall events for effective Water Sensitive Urban Design:
A case study in Gold Coast City, Australia
An Liu1,2, Yuntao Guan1,3*, Prasanna Egodawatta4, Ashantha Goonetilleke4 1
Institute of Environment, Graduate School at Shenzhen, Tsinghua University, 518055 Shenzhen, People’s Republic of China;
2
College of Chemistry and Environmental Engineering, Shenzhen University, Shenzhen 518060, People’s Republic of China
3
School of Environment, Tsinghua University, Beijing100084, People’s Republic of China
4
Science and Engineering Faculty, Queensland University of Technology (QUT), P.O. Box 2434, Brisbane, QLD 4001, Australia
*Corresponding author:
E-mail: [email protected]; Tel: 86-755-26036702; Fax: 86-755-26036702 14
Data monitoring and collection
The catchment data relevant to hydrologic model setup were obtained from Gold Coast City Council (GCCC) as MapInfo GIS software (MapInfo, 2006) files. The data files included:
• Digital Elevation Model (DEM) • Land use
• Detailed drainage networks including waterways, pipes, channels and locations of gully pits (manholes)
• Aerial photographs
Automatic stream flow and stormwater quality monitoring stations were established at the outlet of each of the study catchments to record stream flow and for stormwater sample collection for subsequent laboratory analysis. The monitoring program started in 2001 and was managed by Queensland University of Technology. These stations are equipped with:
• Depth gauges fixed to a V-notch weir to record water depth at 15 min intervals and thereby determine stream flow using calibrated rating curves (Figure S1); • Water quality probes to automatically measure selected quality parameters
such as pH and turbidity;
• Automatic water samplers for sample collection for laboratory testing for additional pollutant parameters (see Figure S2).
The sample collection equipment was set to trigger when the flow reached a pre-determined depth. This depth was dependent on the downstream structure at the collecting point. Hence, it varied from sampling station to sampling station. Samples could be collected in 15-minute intervals once the sampler was triggered. Each sampler had the capacity to collect 24 samples. The sampler can create a log of the sampling times.
Figure S1 Depth gauge and V-notch
V-notch
Depth gauge
Figure S2 Automatic stormwater sample collection equipment
The “three-component” model-Component 1, Quantity simulation
MIKE URBAN (2008) was selected to undertake the quantity simulation. MIKE URBAN was selected for this study after a rigorous review of available models. The hydrologic module in MIKE URBAN is based on the time-area method, which is the same procedure employed in developing the wash-off equation (Component 3) used in the analysis (Egodawatta et al. 2007). The time-area method needs time ofconcentration, reduction factor, initial loss and time-area curves for calibration. Reduction factor and initial loss parameters influence the runoff volume whilst time of concentration and time-area curves influence the hydrograph shape. An initial
value was decided for each calibration parameter and then adjusted by “trial and error” until an acceptable agreement was obtained between simulated and observed results. MIKE URBAN was developed by the Danish Hydraulic Institute (DHI). The
modelling software is flexible and capable of modelling wastewater and stormwater systems.
In order to simulate the runoff volume for the four catchments, the MIKE URBAN model was setup accordingly for the four catchments. Detailed maps of the drainage network including sizes of gully pits and pipe diameters were supplied by the Gold Coast City Council (GCCC).
Alextown catchment model
Alextown consists of an efficient drainage system with rectangular gully pits with steel mesh lids placed in the middle of the road to collect road runoff. For modelling, the catchment was divided into 28 subcatchments so that its distributed nature can be adequately represented. The Alextown catchment model consisted of 28 nodes, 1 outlet and 28 pipes.
Other than the input data obtained directly from GCCC datasets, a number of
assumptions were adopted during model setup. Firstly, a constant value was used for impervious surfaces percentage based on the assumption that impervious surfaces are distributed equally over a catchment surface. Secondly, MIKE URBAN considers the
Data logger
Water samples
gully pits as cylindrical and a diameter is assigned as the primary dimensional parameter. However, all the catchment gully pits were of rectangular shape. An equivalent diameter which was considered as the diagonal length of the rectangular manhole was adopted for modelling. According to the MIKE URBAN manual (MIKEURBAN 2008), the assignment of gully pits of different shapes does not introduce error to the modelling outcomes.
Gumbeel catchment model
Gumbeel is located along a ridge. Therefore, the runoff is primarily contributed by roads, which means that the percentage of connected impervious surfaces is relatively low. The drainage system in Gumbeel catchment is relatively simple and short, compared with the other study catchments. For modelling, the catchment was divided into two subcatchments. The Gumbeel model consisted of 3 nodes, 1 outlet and 3 pipes.
Birdlife Park catchment model
Birdlife Park is in a valley with a relatively greater slope than other study catchments. The runoff is collected by side manholes. Furthermore, the runoff contributed from pervious surfaces is also combined with the road runoff as most of the pervious surfaces are located in the front of houses and are relatively extended. For modelling, Birdlife Park was divided into 77 subcatchments, which consisted of 78 nodes, 1 outlet and 79 pipes.
Highland Park catchment model
The main drainage line in Highland Park is a small tributary, the Bunyip Brook. The integrated pipe and channel network connecting various parts of the catchment area to the tributary, further facilitates stormwater drainage. For modelling, Highland Park catchment was divided into 625 subcatchments. The Highland Park model consisted of 633 nodes, 1 outlet, 39 channels and 596 pipes. In the case of the drainage channels, the cross sections were obtained from the Digital Elevation Model (DEM) supplied by GCCC.
The “three-component” model -Component 2, Build-up estimation
The build-up estimation was undertaken using the equations developed byEgodawatta and Goonetilleke (2006) based on the same catchments as the study sites used in this research study. The two sets of empirical coefficients, a and b, were obtained for two types of urban form; high population density (such as townhouses and duplex housing) and low population density (such as detached housing). The two sets of coefficients are given in Table S1.
Table S1 Build-up coefficients (Egodawatta and Goonetilleke, 2006)
Urban form a b
High population density 2.90 0.16
Low population density 1.65 0.16
Alextown and Gumbeel catchments are townhouse and duplex housing developments, respectively, whilst Birdlife Park catchment is a detached housing development. Therefore, coefficients applicable for high population density were used for build-up estimations for Alextown and Gumbeel catchments, whilst coefficients applicable for low population density were applied for build-up estimation for Birdlife Park
catchment. Highland Park is a mixed land use catchment. Therefore, the Highland Park catchment was divided into a number of subcatchments based on high and low population densities. The two sets of coefficients were then used for the different subcatchments and the total build-up loads were the summation of the different subcatchment build-up loads.
The antecedent dry days was used in Equation (1) to calculate the build-up load per unit area. The total build-up load for each catchment was obtained by the build-up load per unit area multiplied by the total catchment area and the percentage of impervious surfaces. The outcomes were further used for estimating pollutant wash-off loads.
The “three-component” model –Component 3, Wash-off estimation
Pollutant wash-off estimations were undertaken using the wash-off equationdeveloped by Egodawatta et al. (2007). The capacity factor CF indicates the capacity
of a rainfall event to mobilise pollutants and has a value ranging from 0 to 1,
primarily depending on the rainfall intensity. According to Egodawatta et al. (2007),
CF varies based on three rainfall intensity classes. For the 5 to 40 mm/h intensity
range, CF increases linearly; for the 40 to 90 mm/h intensity range, CF is a constant
and for the above 90 mm/h intensity, it varies linearly to a maximum value (Table S2). This approach was used for determining the coefficients for Equation (2).
Table S2 Wash-off parameters (Egodawatta et al., 2007)
Parameters Intensity range (mm/h) Equation/Value
Capacity factor (CF)
5-40 (0.01×I) +0.1
40-90 0.5
>90 (0.0098 × I) -0.38 Wash-off coefficient (k) All intensities 8×10-4 I-rainfall intensity
Quantity model (Component 1) calibration and validation
Based on trial-and-error, the calibrated parameters obtained are given in Table S3. The Root Mean Square Error (RMSE) and Coefficient of Determination (CD) values for all discharge calibrations and validations are given in Table S4. Figure S3 shows the comparison between simulated and observed peak discharge values for the calibration and validation study, whilst Figure S4 shows the comparison between simulated runoff volumes and observed runoff volumes from the calibration and validation study.As shown in Table S4, the RMSE values range from 0.003 to 0.228 for all of the four catchments. These relatively small RMSE values indicate that the simulated results are close to the observed results. In addition, most of the CD values are close to 1. This further confirms that the calibration and validation results are reasonable. According to Figure S3 and Figure S4, it can be seen that both, peak discharge and total runoff volume show good agreement between observed and simulated results. It can be concluded that the runoff volumes were simulated accurately by the calibrated models.
Table S3 Calibration parameters
Catchments
Time of concentration Tc
(min)
Reduction factor Initial loss (mm) Time-area curve
Alextown 22 0.7 1.0 TA Curve 1
Gumbeel 30 0.1 1.0 TA Curve 1
Birdlife Park 20 0.9 0.1 TA Curve 1
Highland Park 40 0.75 1.0 TA Curve 1
Table S4 RMSE and CD values for calibrations and validations
Catchment Task Rainfall event RMSE CD
Gumbeel Calibration 2002-04-28 0.003 0.64 2001-12-29 0.007 0.63 Validation 2002-05-03 0.003 0.49 Alextown Calibration 2003-12-06 0.003 0.74 2003-11-24 0.006 0.66 2002-04-12 0.010 0.48 2002-10-27 0.006 0.69 Validation 2002-11-13 0.010 0.82 2002-08-21 0.005 0.72 2002-08-25 0.002 0.83 2002-11-15 0.007 0.82 Birdlife Park Calibration 2002-02-01 0.009 0.64 2002-10-27 0.013 0.78 2002-06-02 0.011 0.65 Validation 2001-12-31 0.006 0.54 2002-02-02 0.006 0.87 Highland Park Calibration 2002-04-12 0.079 0.83 2002-08-21 0.122 0.81 2002-04-28 0.102 0.76 2002-05-03 0.137 0.67 Validation 2002-06-04 0.057 0.77 2002-09-20 0.127 0.53 2001-03-21 0.228 0.61 2002-11-15 0.219 0.63 19
a. The Gumbeel, Alextown and Birdlife Park
b. Highland Park
Figure S3 The comparison of observed and simulated peak discharges
0 0.05 0.1 0.15 0.2 0.25 0 0.05 0.1 0.15 0.2 0.25 Si mu la te d p ea k d isc ha rg e ( m 3/s )
Observed peak discharge (m3/s)
1:1
Gumbeel Alextown Birdlife Park 0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5 Si mu la te d p ea k d isc ha rg e ( m 3/s )Observed peak discharge (m3/s)
1:1
Highland Park
a. Gumbeel, Alextown and Birdlife Park
b. Highland Park
Figure S4 The comparison of observed and simulated total runoff volumes
Quality model (Component 2 and 3) calibration and validation
Since total solids (TS) was adopted as the indicator pollutant for stormwater quality estimation in the research study, TS event mean concentration (EMC) estimation procedure using the build-up and wash-off equations are given below:Estimation approach
In order to estimate TS EMC, the TS off load was estimated initially. TS wash-off load from the catchment surface was obtained by multiplying the TS build-up load (Equation 1) by the wash-off fraction (Equation 2) for the rainfall event. For different rainfall intensities, the capacity factor CF for the wash-off fraction is different and can be calculated by the equations given in Table S2. TS EMC (mg/L) for each rainfall
0 100 200 300 400 500 600 700 800 900 1000 0 200 400 600 800 1000 Si mu la te d ru no ff v ol ume (m 3)
Observed runoff volume (m3)
1:1
Gumbeel Alextown Birdlife Park 0 5000 10000 15000 20000 25000 30000 35000 40000 0 5000 10000 15000 20000 25000 30000 35000 40000 Si mu la te d ru no ff v ol ume (m 3)Observed runoff volume (m3)
1:1
Highland Park
event was obtained by the TS wash-off loads divided by the total runoff volumes simulated by the stormwater quantity model.
Simulation results
Figure S5 shows the comparison of estimated and measured TS EMC individually for the four catchments. It can be noted that Alextown and Gumbeel displays a relatively good agreement between estimated and measured EMC values whilst for Birdlife Park and Highland Park, the EMC values are under estimated.
0 20 40 60 80 100 120 0 20 40 60 80 100 120 Si mu la te d E M C (mg /L ) Measured EMC (mg/L)
1:1
0 50 100 150 200 250 300 350 0 50 100 150 200 250 300 Si mu la te d E M C (mg /L ) Measured EMC (mg/L)1:1
a. Alextown b. Gumbeel c. Birdlife Park 22Figure S5 The comparison of estimated and measured TS EMC for the four catchments
Additionally, it can be observed that the under estimation of TS EMC by the model for Highland Park catchment is relatively more significant than for Birdlife Park. This can be attributed to the following reasons:
1. Only the impervious surfaces were considered in both, the runoff volume simulation and pollutant wash-off loads estimation. However, in a catchment, both impervious and pervious surfaces may contribute to the pollutant load.
Alextown and Gumbeel have the highest percentage of impervious surfaces (70%) among the four catchments, followed by Birdlife Park at 45.8% and Highland Park catchment with 40%. Therefore, the pollutant EMC estimations performed relatively better for Alextown and Gumbeel due to the high percentage of impervious area, but poorly for Birdlife Park and Highland Park catchments. 2. Sampling from storm events for subsequent laboratory analysis for pollutant
concentrations was done by pumping from a constant height above the drainage
0 10 20 30 40 50 60 0 20 40 60 80 100 120 Si mu la te d E M C (mg /L ) Measured EMC (mg/L)
1:1
0 10 20 30 40 50 60 70 80 90 100 0 100 200 300 400 500 Si mu la te d E M C (mg /L ) Measure EMC (mg/L) d. Highland Park 1:1 23channel bed, irrespective of the flow depth. This could lead to the collection of non-representative samples with high TS loads.
3. EMC samples were prepared from the runoff samples typically collected 15 min apart. Due to the relatively smaller size of the catchments and the resulting “flashy” nature of stormwater runoff, water quality can vary significantly during this time interval. This can lead to non-representative EMC results.
4. The computer model and build-up and wash-off equations used for estimating TS EMC are based on a number of assumptions and conceptual descriptions of pollutant processes. This may introduce errors to the estimation results.
Although the estimation of TS EMC were not satisfactory for all of the four
catchments, these results were accepted since the primary function of the study was to validate the three-component catchment model (hydrologic, build-up and wash-off) in terms of its sensitivity to important rainfall parameters rather than to predict the TS EMC. Therefore, the estimated TS EMC values for the four catchments were further analysed on the sensitivity of the three-component catchment model to important rainfall characteristics as discussed in Section 3.1.
Selection of the representative rainfall year
The selection of the representative rainfall year was to ensure that rainfall events in that year had moderate characteristics rather than extreme characteristics such as extremely high rainfall amounts or rainfall frequency. Accordingly, historical rainfall records of the study catchments were analysed. For this purpose, annual rainfall depth and the number of rainfall events which occurred each year were extracted from the rainfall records for the period 1992-2010. Figure S6 shows the annual rainfall depth and the number of rainfall events from 1992 to 2010. It was found that the average annual rainfall depth for that period was 1352.7 mm, whilst the average number of rainfall events occurring in one year was 132. The characteristics of rainfall events in 2003 are the closest with an average annual rainfall depth of 1363.7 mm and the number of rainfall events being 131events. Therefore, the rainfall events in 2003 were initially selected and further analysed in order to ensure that the three rainfall types were present within the data set as discussed in Section 3.2.
a. Annual rainfall depth
600 800 1000 1200 1400 1600 1800 2000 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 An nu al ra in fa ll de pt h ( mm) 2003 24
b. No. of rainfall events
Figure S6 Rainfall events in 1992-2010
References
Egodawatta P. and Goonetilleke A. 2006 Characteristics of pollutants build-up on residential road surfaces. The 7th International Conference on Hydroscience and Engineering (ICHE), Philadelphia USA.
Egodawatta P., Thomas E. and Goonetilleke A. 2007 Mathematical interpretation of pollutant wash-off from urban road surfaces using simulated rainfall. Water Research, 41(13), 3025-3031.
MapInfo. 2006 MapInfo professional, Version 8.5 - User Guide.
MIKEURBAN. 2008 MIKE URBAN Model User Manual, Danish Hydraulic Institute.
80 100 120 140 160 180 200 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 N o. o f r ain fa lll ev en ts 2003 25
