Abstract

The storm water management models were established at three spatial scales (large, medium, and small) based on a sponge city pilot area in China to explore the hydrological and environmental effects of rainfall conditions and development modes. Results showed that: (1) total runoff reduction rates increased from 26.7% to 53.9% for the rainfall event of a 2-year recurrence period as the scale increased. For 5-year and above recurrence periods, total runoff reduction rates were 19.5–49.4%. These rates increased from the small to medium scale and slightly decreased from the medium to large scale. (2) The runoff coefficients were 0.87–0.29, which decreased from the small to medium scale and were basically constant from the medium to large scale. (3) The peak flow reduction rates decreased with increased recurrence periods. The rates increased initially and then decreased at the small scale, whereas the opposite trend occurred at the medium scale. (4) The reduction rates of pollutants were negatively correlated with recurrence periods under the three spatial scales. The pollution load reduction rates were 19.5–54.7%, which increased from the small to medium scale and were basically constant from the medium to large scale.

INTRODUCTION

In recent years, global climate change and urbanization have led to major changes in the hydrological environment. The challenges which are faced include drainage, non-point source pollution (NSP) control, and comprehensive utilization of water resources. The amount, intensity, and duration of rainfall have different impacts on urban water logging and NSP (Li et al. 2017). Changes in the global climate are increasing the frequency and intensity of storms, particularly during El Niño Southern Oscillation and Arctic Oscillation events (Jiang et al. 2014; Kuo et al. 2014). Rapid urbanization is changing the land use/cover in urban areas, which increases surface runoff, shortens the duration of runoff, and reduces the time to reach the peak runoff, which can cause urban water logging and increase non-point source pollutant exports (Owrangi et al. 2014). In addition, the heat and rain island effects will also exacerbate such urban problems (Dou et al. 2014).

In the face of the impacts of urbanization, a number of approaches have been internationally used to mitigate the adverse impacts. The widely used approaches include low-impact development (LID) in the United States, water-sensitive urban design (WSUD) in the Australian, sustainable urban drainage system (SUDS) in the UK and the Green Rainwater Infrastructure. An LID plays an important role in relieving the frequent water logging and the NSP caused by urban runoff (Davis et al. 2012). Based on the LID approach, China has proposed the concept of sponge city. This concept of urban development includes a range of measures such as source reduction, mid-level transfer, and end-of-pipe storage to maintain the pre-development hydrological cycle (Li et al. 2018). Thus, infiltration, storage, treatment, utilization, and managed discharges can all be used to improve, maintain, or restore urban ‘sponge’ functions.

The rainwater system can be divided into several scales according to planning design, scientific research, control targets, and so on. In recent years, many developed countries have based the scale of the rainwater system on the landscape scales. Among these countries, the United States divides its rainwater system into three scales, namely source (small), community (medium), and watershed (large) (Chris & Gordon 2010). In China, the rainwater system scale is mainly based on urban hydrology, which is managed separately by water conservancy and urban construction departments according to the large and small hydrological system scales (Zhao 2012). Along with the development of the sponge city in China, the scale division of the Chinese urban rainwater system has been gradually perfected on the basis of combining LID concept and national conditions. Much of the spatial scale research was focused on the ecological risk of landscape pattern and soil moisture change (Wang et al. 2012; Zhang et al. 2012). The impacts of rainwater systems and LID measures can vary as a function of scale. But a few studies were performed on hydrological and environmental effects of sponge city on different spatial scales. In this study, the storm water management model (SWMM) was selected to simulate the effects of three spatial scales (large, medium, and small). This study aimed to (i) simulate and analyze the effects of urban rainwater runoff and NSP under three spatial scales and (ii) quantify and analyze the influences of scale change on urban hydrological environmental effects. The results provided a reference and basis for simulation and evaluation of the effectiveness of the sponge city approach at the three spatial scales.

MATERIALS AND METHODS

Scale definition and study frame

An SWMM, namely Environmental Protection Agency (EPA)-SWMM, was used to simulate the urban stormwater quantity and quality. The SWMM is a widely used package which was developed by the EPA of the United States of America (Rossman 2012). It is a comprehensive computer model for predicting and assessing the impact of LID on storm water quantity and quality in urban areas.

Fengxi New City located in the Xixian New Area was selected as the study area. It is one of the first pilot sponge cities in China. It has a total area of 2,265 ha, of which six drainage systems span a total area of 2,067.5 ha, while other areas cover a total of 197.5 ha. The six drainage systems were selected for a large-scale regional model. The areas of the six drainage systems from 1# to 6# were 530.8, 618.5, 298.9, 295.9, 66.4, and 257.1 ha, respectively. The 4# drainage system with a high construction level was selected for the medium-scale regional model. It is located in the Weihe drainage basin. As the basic unit of human activities, the community occupies the majority of the urban area. The Xibu Yungu community was chosen as the small-scale study area with an area of approximately 6.8 ha. This research was carried out based on the research on drainage system planning. It was combined with a variety of rainwater system partitioning methods. We adopted the six drainage systems, 4# drainage system, and the Xibu Yungu community as large, medium, and small scales, respectively. The three study scale regions are shown in Figure 1. The Keifer and Chu rainfall pattern is an uneven design based on the intensity–duration–frequency relationship. The rainfall process is divided into two parts: pre-peak and post-peak. It expresses most types of rainfall and reflects the average characteristics of rainfall process. It was used to analyze the hydrological, hydraulic, and environmental effects with the changes of spatial scales and rainfall recurrence periods (2, 5, 10, and 20 years). The rainfall duration of the four recurrence periods was 2 h. Rainfall intensities were 13.67, 18.36, 21.90, and 27.33 mm/h, respectively. Figure 2 illustrates the structure of the study.

Figure 1

Generalization results of three scale regions.

Figure 1

Generalization results of three scale regions.

Figure 2

Study frame.

Figure 2

Study frame.

Establishment of the SWMM model

Regional generalization

This study followed the principle of regional generalization and was mainly based on the 3-year implementation plan for the pilot area. The research area was generalized according to the data on the drainage network and the underlying surface on which the traditional development (TD) mode of the research area was constructed (Li et al. 2014). Based on a map of regional land use, the planned commercial districts, residential areas, roads, public facilities, and green areas were divided into roof, road, and green space surface types. Therefore, the space of commercial districts, residential areas, and public facilities was classified as roof type. The proportion of roof, road, and green space in each sub-watershed could be obtained by using the GIS software to make statistics on the area of each land type. During the construction of the LID mode of the research area, the LID facility types, area ratios, and structural parameters were determined according to the regional LID special programming. Rain gardens, infiltration ditches, green roofs, permeable pavements, rain tanks, and sunken green space were selected as the six types of LID facilities. Each subcatchment area contained five types of LID facilities, including rain garden, sunken green space, infiltration canal, infiltration pavement, and rain bucket. There were 181 subcatchment areas with green roofs. For each subcatchment area, the five LID facilities were set according to the area ratio. According to the proportion of total green roof area, 181 subsets of water area were calculated. The layout percentages and parameters are shown in Table 1. Firstly, the total area of 181 subcatchment areas with a green roof was calculated. According to the proportion of green roof layout in the research area, the layout area was calculated. The proportion of green roofs on the 181 subcatchment areas could be obtained.

Table 1

LID setting situation of the study region

 LID1 LID2 LID3 LID4 LID5 LID6 Total 
Area of LID (ha) 65.84 65.47 64.99 97.11 66.06 28.85 388.33 
Area ratio (%) 3.18 3.17 3.14 4.70 3.20 1.40 18.78 
Structure parameter 
 Surface layer (mm) 400 200 200 100 – 100 – 
 Soil layer (mm) 700 300 – – – 150 – 
 Permeable layer (mm) – – – 500 – – – 
 Packing layer (mm) 500 – 400 500 700 – – 
 Plant density 0.6 0.6 0.5 – – 0.8 – 
 LID1 LID2 LID3 LID4 LID5 LID6 Total 
Area of LID (ha) 65.84 65.47 64.99 97.11 66.06 28.85 388.33 
Area ratio (%) 3.18 3.17 3.14 4.70 3.20 1.40 18.78 
Structure parameter 
 Surface layer (mm) 400 200 200 100 – 100 – 
 Soil layer (mm) 700 300 – – – 150 – 
 Permeable layer (mm) – – – 500 – – – 
 Packing layer (mm) 500 – 400 500 700 – – 
 Plant density 0.6 0.6 0.5 – – 0.8 – 

Note: LID1, rain garden; LID2, sunken green space; LID3, infiltration ditch; LID4, permeable pavement; LID5, rain tank; LID6, green roof.

The large-scale study area was divided into six drainage systems, a total of 190 subcatchment areas, 134 nodes, 14 outlets, and 134 pipes. The medium-scale study area was divided into 43 subcatchment areas based on the urban planning and drainage network for the study area. The total area of each subcatchment ranged between around 0.35 and 20.7 ha. The drainage system was divided into 46 sections, 46 nodes, and one outlet. The small-scale study area was divided into 136 subcatchment areas based on the urban planning and drainage network maps. The total area for each subcatchment area ranged between 0.009 and 0.262 ha. The drainage system was split into 84 sections (with pipe diameters that ranged between 300 and 500 mm), 85 nodes, two outlets, and six types of control measures. The underlying surface types were divided into roofs, road, green, and so on. Figure 1 presents the generalization results.

Parameter sensitivity analysis

The SWMM has many parameters, and parameter sensitivity analysis is mainly used for qualitative and quantitative evaluation of the influence of the model input error on the model simulation results. Through parameter sensitivity analysis, the important parameters of the output model can be identified and the efficiency of parameter calibration and model validation can be improved. It can guarantee the reliability and accuracy of the model (Crosetto & Tarantola 2001). In recent years, the study of the SWMM was mainly aimed at the sensitivity analysis of hydrological, hydraulic, and water quality parameters. The research methods include local and global sensitivity analyses. To facilitate parameter adjustment, the Morris screening method was used to analyze local sensitivity of the parameters that require calibration. Other model parameters were set according to the research data and experiences.

The Morris screening method is one of the common methods to analyze model parameter sensitivity. Firstly, variable parameters should be selected, and the conditions of other parameters should be kept constant. The selected parameters were changed regularly in the range of values, and different results were obtained by simulation and calculation. Thus, the influence values of the parameter change on the operation result of the model were identified (Lenhart et al. 2002). Independent variables were used to change the fixed step length according to the modified Morris screening rule. The average value of the Morris coefficient was obtained by multiple disturbances. It was adopted as the discriminant factor of parameter sensitivity. Generally, parameter sensitivity could be divided into four categories according to sensitivity values (Goldstein et al. 2010). When the sensitivity value was greater than 1, it was a high sensitivity parameter. Sensitive parameters were between 0.2 and 1, and medium sensitive parameters were between 0.05 and 0.2. When it was greater than 0 and less than 0.05, it was an insensitive parameter.

In this study, the modified Morris screening method was used to evaluate the sensitivity of the SWMM parameters. The hydrological, hydraulic, and water quality parameters were disturbed at a fixed step length of 5% with values of −20%, −15%, −10%, −5%, 5%, 10%, 15%, and 20%. The other parameters remained constant when the value of a certain weighting was changed. Under such a condition, the sensitivity of the total runoff correlation parameters, such as peak runoff and pollutant load, was analyzed for two rainfall conditions (3-h short-duration rainfall under the 2-year recurrence period and 12-h long-duration rainfall under the 20-year recurrence period). The parameter sensitivities of total runoff, peak flow, and pollution load (taking the main control target suspended solid (SS) as an example) are shown in Figure 3. The parameter sensitivities were sorted by combining graphics. The sensitivity values in the figures were taken in absolute terms.

Figure 3

Results of parameter sensitivity analysis. Note: Hydrologic and hydraulic parameters: 1. Width; 2. N-perv is the Manning coefficient of pervious area; 3. N-imperv is the Manning coefficient of impervious area; 4. D-perv is the depression storage of pervious area; 5. D-imperv is the depression storage of impervious area; 6. pipe roughness; 7. Max. infil. rate is the maximum infiltration rate; 8. Min. infil. rate is the minimum infiltration rate; 9. Infiltration attenuation coefficient; 10. Grade. Water quality parameters: 1. Maximum road surface accumulation; 2. Road surface half full and constant; 3. Pavement Scour coefficient; 4. Pavement scour Index; 5. Maximum roof cumulative amount; 6. Roof half-saturation constant; 7. Roof scour coefficient; 8. Roof scour index; 9. Maximum accumulated amount of green space; 10. Green half-saturation constant; 11. Green space scour coefficient; 12. Green space scour index. (a) Sensitivity of total runoff under a 2-year recurrence period. (b) Sensitivity of total runoff under a 20-year recurrence period. (c) Sensitivity of runoff peak under a 2-year recurrence period. (d) Sensitivity of runoff peak under a 20-year recurrence period. (e) Sensitivity of pollution load under a 2-year recurrence period. (f) Sensitivity of pollution load under a 20-year recurrence period.

Figure 3

Results of parameter sensitivity analysis. Note: Hydrologic and hydraulic parameters: 1. Width; 2. N-perv is the Manning coefficient of pervious area; 3. N-imperv is the Manning coefficient of impervious area; 4. D-perv is the depression storage of pervious area; 5. D-imperv is the depression storage of impervious area; 6. pipe roughness; 7. Max. infil. rate is the maximum infiltration rate; 8. Min. infil. rate is the minimum infiltration rate; 9. Infiltration attenuation coefficient; 10. Grade. Water quality parameters: 1. Maximum road surface accumulation; 2. Road surface half full and constant; 3. Pavement Scour coefficient; 4. Pavement scour Index; 5. Maximum roof cumulative amount; 6. Roof half-saturation constant; 7. Roof scour coefficient; 8. Roof scour index; 9. Maximum accumulated amount of green space; 10. Green half-saturation constant; 11. Green space scour coefficient; 12. Green space scour index. (a) Sensitivity of total runoff under a 2-year recurrence period. (b) Sensitivity of total runoff under a 20-year recurrence period. (c) Sensitivity of runoff peak under a 2-year recurrence period. (d) Sensitivity of runoff peak under a 20-year recurrence period. (e) Sensitivity of pollution load under a 2-year recurrence period. (f) Sensitivity of pollution load under a 20-year recurrence period.

Figure 3 shows the sensitivity of runoff amount, runoff peak, and pollution load parameters under 2a and 20a, respectively. The value in the figure represented sensitivity. The weaker the sensitivity was near the center and the higher the sensitivity was near the edge. In Figure 3(a) and 3(b), the parameter closest to the edge was width (1), followed by the minimum infiltration rate (8). The results of model parameter sensitivity analysis revealed that the parameters that had the greatest impact on the total runoff volume in order from large to small were characteristic width, minimum infiltration rate, and Manning roughness for pervious areas, and depression storage in pervious areas. The parameters that had the greatest impact on the peak runoff in order from large to small were pipe Manning roughness, width, Manning roughness for impervious areas, and grade. The parameters that had the greatest impact on the pollutant loads were the relevant accumulation and scouring parameters for roads and roofs. The relevant accumulation and scouring parameters for green space were secondary. The results were basically consistent with those of Li et al.’s (2016) simulations in the northern plain of China.

Parameter calibration and model validation

The research area is currently in the planning and construction stages, and the drainage network is only partially built. Therefore, rainfall and runoff data are not available to calibrate parameter values and validate models. Consequently, parameter values calibrated using the method that was put forward by Liu (2009) for the selection of urban rainfall runoff model parameter values in the absence of calibration data. The SWMM5.1 user manual and relevant research in this field were used to guide the initial selection of parameter values (Rossman 2015). The Horton model was selected for the infiltration process of the surface runoff subsystem, and the nonlinear reservoir surface overland model was selected for the confluence process. The kinematic-wave and complete mixing first-order attenuation equations were selected for the simulation of the flow and transport through the drainage system. The kinematic-wave equations are defined as the following (Wang et al. 2018). 
formula
(1)
 
formula
(2)
where A is the flowing water cross-sectional area; Q is the cross-sectional water quantity; S0 is the longitudinal slope of the river; and Sf is the friction slope.

A target runoff coefficient value of 0.45–0.60 was adopted based on the density of the study area. Two-field design rainfalls (0.5 and 2 years) were selected to determine the parameter values. The parameter calibration process is shown in Table 2.

Table 2

The process of parameter calibration

Parameter Initial value Parameter adjustment
 
First Second Third Fourth Fifth 
Manning roughness for pervious area 0.17 0.1615 0.153 0.1445 0.14 0.13 
Manning roughness for impervious area 0.016 0.0152 0.0144 0.0136 0.013 0.012 
Depression storage for pervious area 4.73 4.494 4.257 4.021 3.85 3.80 
Depression storage for impervious area 2.3 2.3 2.06 2.06 2.06 2.05 
Max. infiltration rate 25.6 25.6 25.4 25.4 25.4 25.4 
Min. infiltration rate 3.7 3.65 3.6 3.56 3.56 3.56 
Pipe roughness 0.016 0.0152 0.0144 0.0136 0.013 0.013 
Simulation value of runoff coefficient 
 0.5a 0.414 0.427 0.446 0.471 0.518 0.523 
 2a 0.452 0.466 0.484 0.502 0.521 0.531 
Parameter Initial value Parameter adjustment
 
First Second Third Fourth Fifth 
Manning roughness for pervious area 0.17 0.1615 0.153 0.1445 0.14 0.13 
Manning roughness for impervious area 0.016 0.0152 0.0144 0.0136 0.013 0.012 
Depression storage for pervious area 4.73 4.494 4.257 4.021 3.85 3.80 
Depression storage for impervious area 2.3 2.3 2.06 2.06 2.06 2.05 
Max. infiltration rate 25.6 25.6 25.4 25.4 25.4 25.4 
Min. infiltration rate 3.7 3.65 3.6 3.56 3.56 3.56 
Pipe roughness 0.016 0.0152 0.0144 0.0136 0.013 0.013 
Simulation value of runoff coefficient 
 0.5a 0.414 0.427 0.446 0.471 0.518 0.523 
 2a 0.452 0.466 0.484 0.502 0.521 0.531 

The result of the fifth parameter adjustment satisfied the target. The calibrated value was used as the simulation parameter. According to the large-scale parameter sensitivity analysis and the calibration method, the values of middle- and small-scale parameters were obtained. Table 3 summarizes the parameter values at the three scales.

Table 3

Calibrated values of model parameters

Scales Parameters Value Parameter Value 
Large N-perv 0.13 Min. infiltration rate 3.56 
N-imperv 0.012 Pipe roughness 0.013 
D-perv 3.80 Simulated runoff coefficient (0.5 year) 0.523 
D-imperv 2.05 Simulated runoff coefficient (2 years) 0.531 
Max. infiltration rate 25.4 Comprehensive runoff coefficient 0.45–0.60 
Medium Area (ha) 0.35–20.7 D-imperv (mm) 2.02 
Width (m) 59.79–306.4 D-perv (mm) 3.80 
Average grade (%) 0.2 I-imperv (%) 25 
Impermeability (%) 20–100 Max. infiltration rate (mm/h) 25.4 
R-imperv 0.013 Min. infiltration rate (mm/h) 3.56 
R-perv 0.14 Infiltration attenuation coefficient 
Small Area (ha) 0.009–0.262 D-imperv (mm) 2.03 
Width (m) 13.8–53.3 D-perv (mm) 3.81 
Average grade (%) 0.2 I-imperv (%) 25 
Impermeability (%) 20–100 Max. infiltration rate (mm/h) 25.4 
R-imperv 0.013 Min. infiltration rate (mm/h) 3.56 
R-perv 0.15 Infiltration attenuation coefficient 
Scales Parameters Value Parameter Value 
Large N-perv 0.13 Min. infiltration rate 3.56 
N-imperv 0.012 Pipe roughness 0.013 
D-perv 3.80 Simulated runoff coefficient (0.5 year) 0.523 
D-imperv 2.05 Simulated runoff coefficient (2 years) 0.531 
Max. infiltration rate 25.4 Comprehensive runoff coefficient 0.45–0.60 
Medium Area (ha) 0.35–20.7 D-imperv (mm) 2.02 
Width (m) 59.79–306.4 D-perv (mm) 3.80 
Average grade (%) 0.2 I-imperv (%) 25 
Impermeability (%) 20–100 Max. infiltration rate (mm/h) 25.4 
R-imperv 0.013 Min. infiltration rate (mm/h) 3.56 
R-perv 0.14 Infiltration attenuation coefficient 
Small Area (ha) 0.009–0.262 D-imperv (mm) 2.03 
Width (m) 13.8–53.3 D-perv (mm) 3.81 
Average grade (%) 0.2 I-imperv (%) 25 
Impermeability (%) 20–100 Max. infiltration rate (mm/h) 25.4 
R-imperv 0.013 Min. infiltration rate (mm/h) 3.56 
R-perv 0.15 Infiltration attenuation coefficient 

Note: N-perv, Manning roughness for pervious area; N-imperv, Manning roughness for impervious area; D-perv, depression storage for pervious area; D-imperv, depression storage for impervious area; R-perv, permeability roughness coefficient; R-imperv, impervious roughness coefficient; I-imperv, the area ratio of impervious area without depression storage.

RESULTS AND DISCUSSION

Hydrological effect analysis

The total runoff and runoff coefficients were used to analyze the hydrological impact of the spatial scales in the study area. As the main control target for a sponge city, the runoff depth was an important reference value for urban rain-flood regulation and the evaluation of the effectiveness of LID measures. Table 4 provides the results of the regional hydrological and hydraulic simulations at the three spatial scales.

Table 4

Hydrological and hydraulic simulation results

 TD
 
LID
 
Δ
 
Recurrence period (years) Total runoff (mm)
 
20.93 18.09 17.86 15.34 8.37 8.24 26.7% 53.8% 53.9% 
30.03 26.50 26.42 23.11 13.42 13.52 23.0% 49.4% 48.8% 
10 36.98 33.00 33.06 29.10 17.50 17.76 21.3% 47.0% 46.3% 
20 47.71 39.57 39.80 38.39 21.71 22.16 19.5% 45.1% 44.3% 
 Runoff coefficient
 
 
0.77 0.65 0.64 0.56 0.30 0.29 27.3% 53.9% 54.697% 
0.82 0.70 0.70 0.63 0.36 0.36 23.2% 48.6% 48.6% 
10 0.84 0.74 0.74 0.66 0.39 0.40 21.4% 47.3% 46.0% 
20 0.87 0.76 0.76 0.70 0.42 0.43 19.5% 44.7% 43.4% 
 Peak flow (m3/s)
 
 
0.43 9.86 – 0.36 5.99 – 16.3% 39.3% – 
0.64 12.59 0.52 8.95 18.8% 28.9% 
10 0.79 14.03 0.72 9.64 8.9% 31.3% 
20 1.30 14.74 1.26 10.10 3.1% 31.5% 
 TD
 
LID
 
Δ
 
Recurrence period (years) Total runoff (mm)
 
20.93 18.09 17.86 15.34 8.37 8.24 26.7% 53.8% 53.9% 
30.03 26.50 26.42 23.11 13.42 13.52 23.0% 49.4% 48.8% 
10 36.98 33.00 33.06 29.10 17.50 17.76 21.3% 47.0% 46.3% 
20 47.71 39.57 39.80 38.39 21.71 22.16 19.5% 45.1% 44.3% 
 Runoff coefficient
 
 
0.77 0.65 0.64 0.56 0.30 0.29 27.3% 53.9% 54.697% 
0.82 0.70 0.70 0.63 0.36 0.36 23.2% 48.6% 48.6% 
10 0.84 0.74 0.74 0.66 0.39 0.40 21.4% 47.3% 46.0% 
20 0.87 0.76 0.76 0.70 0.42 0.43 19.5% 44.7% 43.4% 
 Peak flow (m3/s)
 
 
0.43 9.86 – 0.36 5.99 – 16.3% 39.3% – 
0.64 12.59 0.52 8.95 18.8% 28.9% 
10 0.79 14.03 0.72 9.64 8.9% 31.3% 
20 1.30 14.74 1.26 10.10 3.1% 31.5% 

Note: Δ = (TD–LID)/TD.

The results showed that the total runoff reduction rates decreased under the three spatial scales with increased recurrence periods. The total runoff reduction rates increased under the 2-year recurrence period with increased spatial scales. Under 5-year and above recurrence periods, the total runoff reduction rates increased from the small to medium scale. However, a slight downward trend occurred in the large scale. This was related to the complexity of the underlying surface and arrangement of LID facilities. The small and medium scales that underlie the surfaces were independent drainage systems. The large scale was composed of six drainage partitions with a different functional location for each drainage partition. Thus, the layouts of the underlying surface and the LID were different. Figure 4 depicts the analysis of the regional total runoff under the three spatial scales.

Figure 4

Analysis of total runoff under the three spatial scales.

Figure 4

Analysis of total runoff under the three spatial scales.

The runoff coefficient was mainly affected by the topography, slope, and vegetation of the underlying surface. A high infiltration rate would cause more rainwater loss in the confluence process and less rainwater converted into runoff. The slope increases confluence velocity and reduces confluence time. The chance of runoff seeping into the ground would be correspondingly reduced, and more rainfall would be converted into runoff. Therefore, the steeper slope would increase runoff coefficient. In the relationship between rainfall and runoff, the runoff and confluence process could be affected by plant interception, thus affecting the runoff coefficient. The larger the runoff coefficient, the greater the flow in the drainage ditch. The runoff coefficient could be used as a design parameter or an evaluation index for the management of urban storm. It could be divided into rainfall runoff coefficient and flow runoff coefficient (Tang et al. 2009). The rainfall runoff coefficient was used to calculate runoff thickness according to the rainfall, which could be used to analyze the total amount of rainfall runoff. The flow runoff coefficient was calculated according to the rainfall. The maximum flow rate could be used to determine the size of the designed drain. This study used the event rainfall runoff coefficient, that was, the ratio of the total runoff to rainfall produced in one rainfall event. Table 4 provides the regional runoff coefficients under the three spatial scales.

With the expansion of the scale, the underlying surface, vegetation cover, and slope conditions would become complex. As the confluence area increased, the confluence time increases. The runoff coefficient decreased with the increase of scale. Under the TD mode, the runoff coefficient decreased as the spatial scale increased within the 2-year recurrence period. However, due to the large rainfall and fast flow in the recurrence period of 5 years and above, the scale change had few effects on the runoff. But, with the scale of increase, the rate of reduction was still decreasing, from the medium to large scale remained basically constant. Under the LID mode, the variation in the runoff coefficient with the spatial scale was similar to that in the TD mode. But the runoff coefficient had a slight upward trend during the large recurrence periods from the medium to large scale. The variation amplitude of the runoff coefficient increased with an increased spatial scale under the small recurrence periods (2 and 5 years). Meanwhile, this variation displayed a downward trend under the large recurrence periods (10 and 20 years) from medium to large scales, as shown in Figure 5.

Figure 5

Analysis of runoff coefficient under the three spatial scales.

Figure 5

Analysis of runoff coefficient under the three spatial scales.

Hydraulic efficiency analysis

The large scale of the research area (sponge pilot) consists of six drainage systems with a total of 14 outlets. Therefore, analyzing its peak flow was extremely difficult. Only the peak flow reduction rates in the small and medium scales were analyzed, as shown in Table 4.

The peak flow reduction rates decreased with increased recurrence periods. The reduction rates in the small scale increased initially and then decreased with the increased recurrence periods, whereas the opposite trend occurred at the medium scale. These results related with the regulatory capacities and the layout of LID facilities.

Environmental effect analysis

To evaluate the reduction of pollutant exports in the three spatial scales, four pollutants, namely SS, chemical oxygen demand (COD), TN, and TP, were selected, that is shown in Table 5. The reduction rates of pollutants were negatively correlated with recurrence periods under the three spatial scales. In the small recurrence periods (2 and 5 years), the reduction rates of the four types of pollutant load increased from small to large scales, then decreased slightly. In the large recurrence periods (10 and 20 years), the load reduction rates of SS, TP, and TN showed a trend of significantly increasing, then decreasing. The reduction rates of the COD load had the same regularity as those in the small recurrence periods. The trend of pollution load reduction rates displayed a considerable relationship with the trend of the total runoff reduction rate, which was influenced by LID facility regulation. Figure 6 illustrates the analysis of pollutant reduction in the three spatial scales for the four types of pollutants. The figure displayed the trend of the reduction rate.

Table 5

Reduction rate of pollutants under the three spatial scales

Pollutant Recurrence period (years) Pollution load under TD (kg)
 
Pollution load under LID (kg)
 
Δ
 
SS 74 3,249 23,876 54 1,548 11,937 26.8% 52.4% 50.0% 
106 3,783 28,079 82 2,304 17,133 23.0% 39.1% 39.0% 
10 131 4,096 30,641 103 2,687 19,944 21.3% 34.4% 34.9% 
20 168 4,363 32,857 136 2,999 22,259 19.5% 31.3% 32.3% 
COD 15 2,042 15,372 11 924 7,348 27.0% 54.7% 52.2% 
21 2,245 17,187 16 1,289 9,980 23.1% 42.6% 41.9% 
10 26 2,344 18,027 21 1,437 11,147 21.4% 38.7% 38.2% 
20 34 2,414 18,592 27 1,547 11,957 19.6% 35.9% 35.7% 
TN 10 535 3,918 251 1,946 26.8% 53.1% 50.3% 
14 615 4,592 11 371 2,783 23.0% 39.6% 39.4% 
10a 17 659 4,971 14 428 3,211 21.3% 35.1% 35.4% 
20 22 696 5,277 18 473 3,547 19.5% 32.1% 32.8% 
TP 51 368 25 187 26.8% 52.2% 49.1% 
61 448 38 279 23.0% 37.9% 37.8% 
10 67 496 45 329 21.3% 33.1% 33.8% 
20 72 538 50 370 19.6% 30.0% 31.2% 
Pollutant Recurrence period (years) Pollution load under TD (kg)
 
Pollution load under LID (kg)
 
Δ
 
SS 74 3,249 23,876 54 1,548 11,937 26.8% 52.4% 50.0% 
106 3,783 28,079 82 2,304 17,133 23.0% 39.1% 39.0% 
10 131 4,096 30,641 103 2,687 19,944 21.3% 34.4% 34.9% 
20 168 4,363 32,857 136 2,999 22,259 19.5% 31.3% 32.3% 
COD 15 2,042 15,372 11 924 7,348 27.0% 54.7% 52.2% 
21 2,245 17,187 16 1,289 9,980 23.1% 42.6% 41.9% 
10 26 2,344 18,027 21 1,437 11,147 21.4% 38.7% 38.2% 
20 34 2,414 18,592 27 1,547 11,957 19.6% 35.9% 35.7% 
TN 10 535 3,918 251 1,946 26.8% 53.1% 50.3% 
14 615 4,592 11 371 2,783 23.0% 39.6% 39.4% 
10a 17 659 4,971 14 428 3,211 21.3% 35.1% 35.4% 
20 22 696 5,277 18 473 3,547 19.5% 32.1% 32.8% 
TP 51 368 25 187 26.8% 52.2% 49.1% 
61 448 38 279 23.0% 37.9% 37.8% 
10 67 496 45 329 21.3% 33.1% 33.8% 
20 72 538 50 370 19.6% 30.0% 31.2% 
Figure 6

Analysis of pollutant reduction under the three spatial scales. Note: In this figure, the medium-scale data were reduced by a factor of 10, and the large-scale data were reduced by a factor of 100.

Figure 6

Analysis of pollutant reduction under the three spatial scales. Note: In this figure, the medium-scale data were reduced by a factor of 10, and the large-scale data were reduced by a factor of 100.

CONCLUSIONS

The scale of the study area was divided according to the definition method of the rainwater system scale. The hydrological, hydraulic, and environmental responses of different rainfall conditions and development modes were studied by establishing SWMMs in the three spatial scales. The rainfall conditions were 2-, 5-, 10-, and 20-year recurrence periods, which were used in the traditional and LID modes. The following conclusions were obtained.

  1. With the increased recurrence periods, the total runoff reduction rates decreased under the three spatial scales. The total runoff reduction rates increased under the 2-year recurrence period with increased spatial scales. For the 5-year and above recurrence periods, the total runoff reduction rates increased from the small to medium scale. A slight downward trend occurred in the large scale. However, the overall trend still increased by 23.0–48.8%. This result was related to the complexity of the underlying urban surface and the arrangement of LID facilities.

  2. Under the TD mode, the runoff coefficients in the 2-year recurrence period were 0.77, 0.65, and 0.64 for the small, medium, and large spatial scales, respectively. During the 5-year and above recurrence periods, the runoff coefficient decreased from the small to medium scale and remained basically flat from the medium to large scale. Under the LID mode, the variation in the runoff coefficient with the spatial scale was similar to that in the TD mode. However, the runoff coefficient displayed a slight upward trend during the large recurrence periods (10 and 20 years) from the middle to large scale. The variation amplitude of the runoff coefficient increased with an increased spatial scale in the small recurrence periods (2 and 5 years). In the large recurrence periods, a downward trend occurred from the middle to large scale.

  3. With increased recurrence periods, the peak flow reduction rates in the small scale showed a trend of firstly increasing, then decreasing. Under the middle scale, it showed a trend of initial decrease and then increase. This change agreed with the regulatory capacities and the layout of LID facilities.

  4. With increased recurrence periods, the reduction rates of the four types of pollutant load in the three spatial scales showed a downward trend. For the small recurrence periods (2 and 5 years), the reduction of the pollutant loads firstly increased, then slightly decreased from the small to large scale. For the large recurrence periods (10 and 20 years), the reduction rates of SS, TP, and TN loads gradually increased from the small to large scale, but the increase amplitude decreased. The reduction rates of COD load had the same regularity with those in the small recurrence period. The change in the trend of the pollution load reduction rates had a substantial relationship with the change in the trend of the reduction in total runoff, which was influenced by LID facility regulation.

ACKNOWLEDGEMENT

This research was financially supported by the National Natural Science Foundation of China (grant no. 51879215).

REFERENCES

REFERENCES
Chris
P.
&
Gordon
A.
2010
SUDS for Roads: Final Version
.
Glasgow
,
UK
.
Crosetto
M.
&
Tarantola
S.
2001
Uncertainty and sensitivity analysis: tools for GIS-based model implementation
.
International Journal of Geographical Information Systems
15
(
5
),
415
437
.
Davis
A. P.
,
Traver
R. G.
,
Iii
W. F. H.
&
Olszewski
J. M.
2012
Hydrologic performance of bioretention storm-water control measures
.
Journal of Hydrologic Engineering
17
(
5
),
604
614
.
Dou
J. J.
,
Wang
Y. C.
,
Bornstein
R.
&
Miao
S. G.
2014
Observed spatial characteristics of Beijing urban- climate impacts on summer thunderstorms
.
Journal of Applied Meteorology and Climatology
54
(
1
),
94
105
.
Goldstein
A.
,
Giovanni
K. D.
&
Montalto
F.
2010
Resolution and sensitivity analysis of a block-scale urban drainage model
. In:
World Environmental and Water Resources Congress: Challenges of Change. ASCE
2010
, pp.
4270
4279
.
Jiang
R.
,
Gan
T. Y.
,
Xie
J.
&
Wang
N.
2014
Spatiotemporal variability of Alberta's seasonal precipitation, their teleconnection with large-scale climate anomalies and sea surface temperature
.
International Journal of Climatology
34
(
9
),
2899
2917
.
Kuo
C. C.
,
Gan
T. Y.
&
Hanrahan
J. L.
2014
Precipitation frequency analysis based on regional climate simulations in Central Alberta
.
Journal of Hydrology
510
(
3
),
436
446
.
Lenhart
L.
,
Eckhardt
K.
&
Fohrer
N.
2002
Comparison of two different approaches of sensitivity analysis
.
Physics and Chemistry of the Earth
27
(
2
),
645
654
.
Li
C. L.
,
Hu
Y. M.
,
Liu
M.
&
Chen
T.
2014
Local sensitivity analysis of parameters in storm water management model
.
Chinese Journal of Ecology
33
(
4
),
1076
1081
(in Chinese)
.
Li
J. K.
,
Li
Y.
&
Li
Y. J.
2016
SWMM-based evaluation of the effect of rain gardens on urbanized areas
.
Environmental Earth Sciences
75
(
1
),
17
30
.
Li
J. K.
,
Deng
C. N.
,
Li
Y.
,
Li
Y. J.
&
Song
J. X.
2017
Comprehensive benefit evaluation system for low-impact development of urban stormwater management measures
.
Water Resources Management
31
,
4745
4758
.
Li
J. K.
,
Zhang
B.
,
Mu
C.
&
Li
C.
2018
Simulation of the hydrological and environmental effects of a sponge city based on MIKE FLOOD
.
Environmental Earth Sciences
77
(
2
),
32
.
Liu
X. P.
2009
Parameter calibration method for urban rainfall-runoff model based on runoff coefficient
.
Water and Wastewater Engineering
35
(
11
),
213
217
(in Chinese)
.
Owrangi
A. M.
,
Lannigan
R.
&
Simonovic
S. P.
2014
Interaction between land-use change, flooding and human health in Metro Vancouver, Canada
.
Natural Hazards
72
(
2
),
1219
1230
.
Rossman
L. A.
2012
Storm Water Management Model User's Manual Version 5.0
.
U.S. Environmental Protection Agency
,
Washington, D.C., USA
.
Rossman
L. A.
2015
Storm Water Management Model User's Manual Version 5.1
.
National Risk Management Research Laboratory, U.S. Environmental Protection Agency
,
Cincinnati, OH
,
USA
.
Tang
N. Y.
,
Che
W.
&
Pan
G. Q.
2009
Runoff coefficient analysis for urban storm water and flood control and utilization
.
China Water & Wastewater
25
(
22
),
4
8
(in Chinese)
.
Wang
X. Z.
,
Jiao
F.
,
Liu
Y. X.
,
Zhu
L. T.
&
Lin
K.
2012
Relationships between soil moisture and environmental factors at different spatial scales
.
Chinese Journal of Ecology
31
(
2
),
319
323
(in Chinese)
.
Wang
T.
,
Ding
X.
,
Cai
T.
,
Liu
W. R.
,
Tu
J.
,
Wu
F. F.
,
Wen
G. X.
,
Shi
C. J.
,
Wu
T.
,
Li
C. Q.
,
Sun
Z.
&
Guo
X. P.
2018
Influence of routing methods on outflow of SWMM model
.
China Water & Wastewater
15
,
133
138
(in Chinese)
.
Zhang
Y.
,
Lei
G. P.
,
Lin
J.
&
Zhang
H.
2012
Spatiotemporal change and its ecological risk of landscape pattern in different spatial scales in Zhalong Nature Reserve
.
Chinese Journal of Ecology
31
(
5
),
1250
1256
.
Zhao
J.
2012
Local Flood Characteristics and Stormwater Regulation: Storage Potential of Landscape in Beijing Urban Area
.
Peking University
,
Beijing
(in Chinese)
.