Green Space Optimization Strategy to Prevent Urban Flood Risk in the City Centre of Wuhan

: Changing the water permeability ratio of urban underlying surface helps alleviate urban ﬂood. This paper designs the swale identiﬁcation experiment to modify the ﬂood-submerging simulation experiment based on the SCS-CN model and proves that the results generated by the modiﬁed experiment better reﬂect the realities. The modiﬁed ﬂood-submerging simulation experiment is then applied to downtown Wuhan to obtain the quantitative data. The data are used to quantify the catchment capacities of the lots. Based on the rainfall collection capacities, the maximum surface rainfall runoff volume that would not cause ﬂood is arrived at using the rainfall runoff formula. The maximum runoff volume represents the rainwater storage capacities of the lot based on the proportion of the green space that is identiﬁed within the study area. The results suggest that this rainwater storage capacity evaluation model works efﬁciently to identify the urban areas with ﬂood risks and provides the rainwater runoff thresholds for different areas. Adjustments in the spatial patterns and proportions of the green space help ensure that the rainwater runoff volume is below the thresholds, thus contributing to the prevention and control of the urban ﬂood risks.


Introduction
Urban flood disasters resulting from heavy rainfall are on the increase over the past few years, causing enormous losses to most of China's cities. One of the root causes of this problem is the rapid and undifferentiated expansion of the impermeable urban surfaces due to urbanization. The process significantly changes the catchment hydrology, increasing runoff rates and volumes on one hand and undermining infiltration and baseflow (provided there is no additional source of baseflow) on the other hand. The original hydrologic and ecological environment in urban areas is damaged as well [1][2][3][4][5].
Fundamentally, the essence of non-engineering measures to alleviate flood in urban areas aim twofold: (1) to restore the pre-development hydrologic patterns of a site by regulating the volumes and rates of the urban hydrologic processes and (2) to ensure the independent absorption of the rainfall within a basin so as to avoid flow concentration among basins [6][7][8]. The urban flood can be effectively avoided by regulating the hydrological process in an ecological way [9,10]. Therefore, the accurate simulation or prediction of storm runoff is one of the most important bases of water resource management [11].
Hydrologic models, as the basic tools to estimate the peak volumes and flood peaks [12], have been widely studied by scholars in the field. There are two types of hydrologic models today. One is the performance evaluation models and the stormwater-management models, procedure is as follows; the detailed operation procedure is in the Supplementary Material Tables S1 and S2.

Catchment-Capacity Simulation Experiment
The catchment capacity refers to the water volume that can be accommodated by a catchment after the depression hydrological process without infiltration. This indicator was used to reflect the accumulated precipitation and the capacity to collect the rainwater from the surrounding area of the catchment. It can be expressed by the accumulated water volume per unit of projected area when a catchment is filled up with the rainwater. The formula is shown as below: where β represented the catchment capacity (m 3 / m 2 ), V the maximum rainwater volume (m 3 ) a catchment can collect, and S the projected area (m 2 ) of the catchment on the horizontal plane.
To calculate the catchment capacity, the target return period should be identified first. This paper calculated the catchment capacity during the 100-year return period (decomposed into 6 grades: namely 1-year, 5-year, 10-year, 20-year, 50-year, and 100-year return periods). The values of β for different return periods were drawn based on the swale identification experiment and the flood-submerging simulation experiment.

Swale Identification Experiment
In this experiment, all swales were identified based on the topographic features. The difference between the overflow point (the highest point) and the bottom point (the lowest point) of each swale was arrived at to reflect the depth of the swale. The catchment capacity was calculated based on the swale depth.
This experiment was conducted in two steps.
Step 1: Identified the swales. Processed the elevation data in Wuhan with the flow-direction tools to obtain the flow-direction data; used the hydrological confluence tools to identify all swales in Wuhan; and divided these swales into watersheds with the watershed tools.
Step 2: Calculated the swale depth. Employed the regional analysis-zonal statistics tools to get the minimal elevation of each swale; used the regional analysis-region filling tools to identify the overflow point of each swale; and adopted the raster calculator to obtain the difference between the overflow point and the bottom point of each swale, i.e., the depth of the swale.
For a city, detailed threshold division in the swale-identification experiment would better reflect the catchment capacity of all lots within the study area. This experiment took the depression process into consideration only. The results reflected the relationship between the terrains and the hydrologic patterns and did not represent the actual rainwater storage capacity. The impact of elevation on the catchment capacity within small watersheds was not taken into account either.

SCS-CN-Based Flood-Submerging Simulation Experiment
The SCS-CN model was adopted to simulate the precipitation process and the floodsubmerging area during various return periods, thus identifying the area with the largest catchment capacity.
This experiment was composed of three steps.
Step 1: Divided the watersheds. Obtained the flow data in downtown Wuhan with the Flow Direction tools and the Flow tools. By setting the Flow Threshold for each watershed to be 1,500,000, the downtown area was divided into 24 watersheds. Step 2: Calculated the daily submerging volume during extreme conditions. Employed the SCS-CN model to calculate the submerging volume of each threshold during various return periods.
Step 3: Calculated the floodsubmerging elevation. Used the surface volume tools and the dichotomy (0.001) to estimate the flood-submerging elevation of all watersheds during various return periods. A detailed explanation of the dichotomy: The submergence elevation of each watershed in different recurrence periods is estimated by using dichotomy method (accuracy 0.001). First, we can estimate a range of submergence height (a,c), then calculate the submergence volume of mid-point c, and then compare the result with that of the submergence volume of each basin; if it is small, the range of values of the elevation becomes (c,b). Repeat the above steps until the calculated elevation is infinitely close to the submerged volume of each watershed.
This experiment took precipitation and infiltration processes into consideration, examining the flood-submerged conditions under different rainfall. However, the catchment capacity of the regions other the flood-submerged areas was not simulated. After getting the rainfall, the rainwater was filled into the thresholds from the lowest point until they were submerged completely, which deviated from the actual precipitation process.

Catchment-Capacity Calculation
To sum up, the SCS-CN-based flood-submerging simulation experiment took more factors into consideration and better reflected the actual hydrologic process. The experiment principle is as shown in Figure 1. However, this experiment cannot reveal the catchment capacity without the aid of the swale identification experiment ( Figure 2). Data from the swale identification experiment was further processed because it did not take the precipitation amount into consideration. For regions where the precipitation amount was larger than the swale volume, the catchment capacity was expressed as the ratio of the swale volume to the horizontal plane projection of the swale; for regions where the precipitation amount was smaller than the swale volume, the simulation experiment for the catchment was repeated before the catchment capacity was calculated. The fusion principle of two experiments was as shown in Figure 3. Based on the interpretation of the flooded area from the Wuhan satellite image data and the monitoring data of the relevant technical departments in Wuhan in recent years, the actual flooded area of Wuhan was obtained ( Figure 4). Comparison between the experiment results and the actual flood-submerged conditions suggested that the fused results were more accurate.
the flood-submerging elevation of all watersheds during various return periods. A de tailed explanation of the dichotomy: The submergence elevation of each watershed in dif ferent recurrence periods is estimated by using dichotomy method (accuracy 0.001). Firs we can estimate a range of submergence height (a,c), then calculate the submergence vo ume of mid-point c, and then compare the result with that of the submergence volume o each basin; if it is small, the range of values of the elevation becomes (c,b). Repeat th above steps until the calculated elevation is infinitely close to the submerged volume o each watershed.
This experiment took precipitation and infiltration processes into consideration, ex amining the flood-submerged conditions under different rainfall. However, the catch ment capacity of the regions other the flood-submerged areas was not simulated. Afte getting the rainfall, the rainwater was filled into the thresholds from the lowest point unt they were submerged completely, which deviated from the actual precipitation process.

Catchment-Capacity Calculation
To sum up, the SCS-CN-based flood-submerging simulation experiment took mor factors into consideration and better reflected the actual hydrologic process. The exper ment principle is as shown in Figure 1. However, this experiment cannot reveal the catch ment capacity without the aid of the swale identification experiment ( Figure 2). Data from the swale identification experiment was further processed because it did not take the pre cipitation amount into consideration. For regions where the precipitation amount wa larger than the swale volume, the catchment capacity was expressed as the ratio of th swale volume to the horizontal plane projection of the swale; for regions where the pre cipitation amount was smaller than the swale volume, the simulation experiment for th catchment was repeated before the catchment capacity was calculated. The fusion princ ple of two experiments was as shown in Figure 3. Based on the interpretation of th flooded area from the Wuhan satellite image data and the monitoring data of the relevan technical departments in Wuhan in recent years, the actual flooded area of Wuhan wa obtained ( Figure 4). Comparison between the experiment results and the actual flood submerged conditions suggested that the fused results were more accurate.   the flood-submerging elevation of all watersheds during various return periods. A detailed explanation of the dichotomy: The submergence elevation of each watershed in different recurrence periods is estimated by using dichotomy method (accuracy 0.001). First, we can estimate a range of submergence height (a,c), then calculate the submergence volume of mid-point c, and then compare the result with that of the submergence volume of each basin; if it is small, the range of values of the elevation becomes (c,b). Repeat the above steps until the calculated elevation is infinitely close to the submerged volume of each watershed.
This experiment took precipitation and infiltration processes into consideration, examining the flood-submerged conditions under different rainfall. However, the catchment capacity of the regions other the flood-submerged areas was not simulated. After getting the rainfall, the rainwater was filled into the thresholds from the lowest point until they were submerged completely, which deviated from the actual precipitation process.

Catchment-Capacity Calculation
To sum up, the SCS-CN-based flood-submerging simulation experiment took more factors into consideration and better reflected the actual hydrologic process. The experiment principle is as shown in Figure 1. However, this experiment cannot reveal the catchment capacity without the aid of the swale identification experiment ( Figure 2). Data from the swale identification experiment was further processed because it did not take the precipitation amount into consideration. For regions where the precipitation amount was larger than the swale volume, the catchment capacity was expressed as the ratio of the swale volume to the horizontal plane projection of the swale; for regions where the precipitation amount was smaller than the swale volume, the simulation experiment for the catchment was repeated before the catchment capacity was calculated. The fusion principle of two experiments was as shown in Figure 3. Based on the interpretation of the flooded area from the Wuhan satellite image data and the monitoring data of the relevant technical departments in Wuhan in recent years, the actual flooded area of Wuhan was obtained ( Figure 4). Comparison between the experiment results and the actual floodsubmerged conditions suggested that the fused results were more accurate.     Therefore, the rainfall-collection capacity within the 100-year return period was calculated differently depending on the actual conditions ( Figure 5). (i) When the catchment was within the simulated flood-submerged area, the catchment capacities during different return periods were calculated as below: where βa represented the catchment capacity (m 3 / m 2 ) of the simulated a-year flood-submerged area, Sa the horizontal plane projection area (m 2 ) of the simulated a-year floodsubmerged area, and Qa the simulated a-year rainfall (mm). The simulated flood inundation area is different in different years ( Figure 6). For example, calculation of catchment capacity for simulated 1-year flood-submerged area (S1):   Therefore, the rainfall-collection capacity within the 100-year return period was calculated differently depending on the actual conditions ( Figure 5). (i) When the catchment was within the simulated flood-submerged area, the catchment capacities during different return periods were calculated as below: where βa represented the catchment capacity (m 3 / m 2 ) of the simulated a-year flood-submerged area, Sa the horizontal plane projection area (m 2 ) of the simulated a-year floodsubmerged area, and Qa the simulated a-year rainfall (mm). The simulated flood inundation area is different in different years ( Figure 6). For example, calculation of catchment capacity for simulated 1-year flood-submerged area (S1): Therefore, the rainfall-collection capacity within the 100-year return period was calculated differently depending on the actual conditions ( Figure 5).
Water 2021, 13, x FOR PEER REVIEW 5  Therefore, the rainfall-collection capacity within the 100-year return period was culated differently depending on the actual conditions ( Figure 5). (i) When the catchment was within the simulated flood-submerged area, the ca ment capacities during different return periods were calculated as below: where βa represented the catchment capacity (m 3 / m 2 ) of the simulated a-year floodmerged area, Sa the horizontal plane projection area (m 2 ) of the simulated a-year flo submerged area, and Qa the simulated a-year rainfall (mm). The simulated flood inundation area is different in different years ( Figure 6). For example, calculation of catchment capacity for simulated 1-year floodmerged area (S1): (i) When the catchment was within the simulated flood-submerged area, the catchment capacities during different return periods were calculated as below: where β a represented the catchment capacity (m 3 /m 2 ) of the simulated a-year floodsubmerged area, S a the horizontal plane projection area (m 2 ) of the simulated a-year flood-submerged area, and Q a the simulated a-year rainfall (mm). The simulated flood inundation area is different in different years ( Figure 6).  Therefore, the rainfall-collection capacity within the 100-year return period was calculated differently depending on the actual conditions ( Figure 5). (i) When the catchment was within the simulated flood-submerged area, the catchment capacities during different return periods were calculated as below: where βa represented the catchment capacity (m 3 / m 2 ) of the simulated a-year flood-submerged area, Sa the horizontal plane projection area (m 2 ) of the simulated a-year floodsubmerged area, and Qa the simulated a-year rainfall (mm). The simulated flood inundation area is different in different years ( Figure 6). For example, calculation of catchment capacity for simulated 1-year flood-submerged area (S1): For example, calculation of catchment capacity for simulated 1-year flood-submerged area (S 1 ): Calculation of catchment capacity for simulated 5-year flood-submerged area (S 5 ): Calculation of catchment capacity for simulated 10-year flood-submerged area (S 10 ): Calculation of catchment capacity for simulated 20-year flood-submerged area (S 20 ): Calculation of catchment capacity for simulated 50-year flood-submerged area (S 50 ): Calculation of catchment capacity for simulated 100-year flood-submerged area (S 100 ): (ii) When the catchment is beyond the simulated 100-year flood-submerged area, then compare the volume of the catchment with the rainfall amount during the 100-year return period. If the former is smaller than the latter, then the catchment capacity of each catchment is calculated as below.
where β represents the catchment capacity (m 3 / m 2 ), V the maximum rainfall amount (m 3 ) that can be accommodated by a small watershed, S the horizontal plane projection (m 2 ) of a small watershed, r the radius (m) of the underside of a small watershed approximate to a cone, h the height or depth (m) of a small watershed approximate to a cone, and π the constant. When the volume of the catchment was greater than the rainfall amount during the 100-year return period, then the simulation experiment would be repeated. The formula for the simulated flood-submerged area was adopted to calculate the catchment capacity.

Rainwater Storage Capacity Assessment
The rainwater storage capacity refers to the maximum water volume that can be accommodated by a catchment after the depression and infiltration processes during the return periods without causing flood. In other words, the rainwater storage capacity assessment is made based on the assumption that all the excessive surface rainfall runoff can be infiltrated underground. In this paper, the assessment results of the rainwater storage capacity were expressed as the maximum surface runoff coefficient. The maximum surface runoff coefficient of a catchment with no flood was formulated as below.

Ratio of Green Space Calculation
Different urban lots vary in the underlying surfaces, which have different surface runoff coefficients. These coefficients depend on the green spaces (i.e., the gently-spanning green spaces and the sunken green spaces that retain water) as well as hard roofs and roads. These underlying surfaces totaled over 90% in all lots. Moreover, the surface runoff coefficient of green spaces is about 0.2 while that of the hard roofs and roads is about 0.95. The green space ratio is the ratio between green space and hard ground. The runoff coefficient corresponding to different green space ratio is calculated in Table 1. The minimum green coverage when flood does not take place (i.e., the minimum green coverage at which a lot can give the fullest play of its rainwater storage capacity) can be obtained based on the comprehensive surface runoff coefficient, i.e., the assessment results of rainwater storage capacity, of each lot in Wuhan.

Experiment and Assessment Results
A total of 25,894 swales were identified in downtown Wuhan (Figure 7). The depth of each swale equaled the difference between the overflow point and the bottom point (Due to the poor resolution of the urban-space data and the urban-property data, the accuracy of the swale depth was 1 m. The error in the results was significant.) ( Figure 8). The simulated results of the 1-year, 5-year, 10-year, 20-year, 50-year, and 100-year flood-submerged area were obtained, with the flood-submerged elevation of each watershed in downtown Wuhan listed in Appendix A Table A3.    According to the simulation results, the Nanhu residential area According to the simulation results, the Nanhu residential area, the Qingshan industrial area, the southeast Donghu high-tech development zone, the Sixin district, and the northwest Hankou suffered severe submerging disasters in all return periods. There were differences between the simulation results and the actual rainfall amount. The simulation experiment did not take the confluence among watersheds into consideration. Therefore, the simulated results for areas with low elevation were smaller than the actual amount while those for areas with high elevation were larger than the actual amount. Based on the actual situation of city, the spatial data were modified in Arcgis to make up for the poor resolution of the urban space data and the urban property data or the shortcomings with data. Nonetheless, errors remained during the simulation process, making the results less accurate and reasonable. Figure 9 reveals the simulated catchment capacity of the flood-submerged areas during various return periods, Figure 10 reveals the catchment capacity of areas where the catchment capacity of the swales is smaller than the rainfall volume of 100-year return period, and Figure 11 reveals the catchment capacity of areas where the catchment capacity of the swales is greater than the rainfall volume of 100-year return period. The catchment capacity of downtown Wuhan, as shown in Figure 12, is the combination of results of Figures 9 and 10. There were errors when calculating the catchment capacity of swales, which were approximated to the cones. The actual hydrologic processes were not taken into consideration when the simulated catchment capacity of the flood-submerged areas was calculated. The difference in the infiltration speed of various areas would inevitably result in rainfall concentration. For instance, if there is rainfall flowing from S 5 to S 1 , then the actual value of β 5 is smaller than the calculation value, while the actual value of β 1 is greater than the calculation value. But these slight errors were neglected in the calculation process, i.e., the infiltration speed of all areas were taken as the same. The infiltration process was taken into consideration in the simulated flood-submerging experiment. The simulated catchment capacity was therefore smaller than the actual value.
, 13, x FOR PEER REVIEW period, and Figure 11 reveals the catchment capacity of a ity of the swales is greater than the rainfall volume of 10 ment capacity of downtown Wuhan, as shown in Figure  of Figures 9 and 10. There were errors when calculating t which were approximated to the cones. The actual hydr into consideration when the simulated catchment capac was calculated. The difference in the infiltration speed o result in rainfall concentration. For instance, if there is ra the actual value of β_5 is smaller than the calculation va is greater than the calculation value. But these slight erro tion process, i.e., the infiltration speed of all areas were ta process was taken into consideration in the simulated flo simulated catchment capacity was therefore smaller than        The maximum comprehensive surface runo period standard of the lots in downtown Wuhan capacity simulation experiment, as shown in Figu final evaluation results of the rainwater storage surface runoff coefficient is, the larger that rain cussed before, the comprehensive surface runoff c the green coverage. The minimum green coverage minimum ratio between the green area and the to age capacity gives the fullest play, would be obtai factors, were employed to define the green cover prehensive Plan 2006-2020, as shown in Figure 1 the results can be used to analyze the green-space land-use types and to adjust the use types accord The maximum comprehensive surface runoff coefficient under the 100-year return period standard of the lots in downtown Wuhan was obtained based on the catchmentcapacity simulation experiment, as shown in Figure 13. The coefficient was viewed as the final evaluation results of the rainwater storage capacity, i.e., the smaller the maximum surface runoff coefficient is, the larger that rainwater storage capacity will be. As discussed before, the comprehensive surface runoff coefficient is functionally correlated with the green coverage. The minimum green coverage under this evaluation standard, i.e., the minimum ratio between the green area and the total area of a lot when the rainwater storage capacity gives the fullest play, would be obtained. The results, together with the other factors, were employed to define the green coverage of the 1530 lots in the Wuhan Comprehensive Plan 2006-2020, as shown in Figure 14. When compiling future urban plans, the results can be used to analyze the green-space distribution characteristics in different land-use types and to adjust the use types accordingly.       Figure 15 is the result of combining the actual flood-submerged areas and the water storage capacity of south downtown Wuhan (On 9 July 2016, Wuhan was hit by the most severe flood over the past three years. The waterlogged areas on this day were obtained by recognizing the water bodies on the satellite image via GIS. Limited by data access, this study recognized and analyzed the water bodies in the south downtown Wuhan). The comparison between the results, satellite images, and field observation revealed three main cases: 1 When the rainwater storage capacity was the same, most areas within the floodsubmerged areas often had lower green coverage than those outside the floodsubmerged areas. This shows that increasing the proportion of green space can effectively reduce urban flood. 2 A few areas with greater rainwater storage capacity and green coverage were within the flood-submerged areas because the rainwater collected exceeded the infiltration speed of the green space. 3 There were also areas with greater rainwater storage capacity and smaller green coverage not within the submerged areas or areas with smaller rainwater storage capacity and greater green coverage within the submerged areas. This was caused by experiment errors.

021, 13, x FOR PEER REVIEW
As a result of the rain flood storage evaluation model analysis type (the smaller the comprehensive runoff coefficient value in t the stormwater storage potential value is, the larger the proport between it and the submerged area. In other words, areas with gr capacity (smaller comprehensive runoff coefficient values in the fi able to be flooded. The flood-storage evaluation results proved to Therefore, compared with SCS-CN submersion simulation re lation results from this model have a higher coincidence rate of sub ing that results have higher accuracy and smaller errors.  Therefore, compared with SCS-CN submersion simulation results (Figure 16), simulation results from this model have a higher coincidence rate of submerged areas, indicating that results have higher accuracy and smaller errors.

Conclusions
This study designs the rainwater storage capacity evaluation model. Firstly, the SCS-CN model based on hydrologic flood-submerging simulation experiment is improved by developing the swale recognition experiment. The improved results reflect the actual floodsubmerged conditions better than the SCS-CN model. Secondly, based on the hydrologic process principles, the model translated the quantitative spatial data obtained from the flood-submerging simulation experiment into the comprehensive surface runoff coefficient and evaluated rainwater storage capacity quantitatively, finally proposing rainwater storage capacity to indicate the responsiveness of the urban flood catchments.
In downtown Wuhan, the rainwater storage capacity evaluation model has great evaluation results. Additionally, the evaluation results are translated into the green coverage, which is applied to determine the land-use types in the Wuhan Comprehensive Plan. Also, in hydrology, the evaluation results are quantitative references for the plan compilation at the current stage.
This study compares the identified urban flood areas in Wuhan on 9 July 2016 with the rainwater storage capacity evaluation results and combines a spatial characteristics analysis of the flood-submerged areas based on the day's satellite images and survey. It was found that the model, compared with the SCS-CN model, had a higher submerged coincidence rate of simulated and actual submerged area, which proves the effectiveness of this model.

Conflicts of Interest:
The authors declare no conflict of interest.         Table A4. A list of simulated submergence elevation of SCS-CN.

Appendix A
No.