Flood mitigation by permeable pavements in Chinese sponge city construction

It is important to evaluate the effectiveness of permeable pavements on flood mitigation at different spatial scales for their effective application, for example, sponge city construction in China. This study evaluated the effectiveness of three types of permeable pavements (i.e., permeable asphalts (PA), permeable concretes (PC), and permeable interlocking concrete pavers (PICP)) on flood mitigation at a community scale in China using a hydrological model. In addition, the effects of clogging and initial water content in permeable pavements on flood mitigation performance were assessed. The results indicated that in 12 scenarios, permeable pavements reduced total surface runoff by 1–40% and peak flow by 7–43%, respectively. The hydrological performance of permeable pavements was limited by clogging and initial water content. Clogging resulted in the effectiveness on total surface runoff reduction and peak flow reduction being decreased by 62–92% and 37–65%, respectively. By increasing initial water content at the beginning of the simulation, the effectiveness of total runoff reduction and peak flow reduction decreased by 57–85% and 37–67%, respectively. Overall, among the three types of permeable pavements, PC without clogging had the best performance in terms of flood mitigation, and PICP was the least prone to being clogged. Our findings demonstrate that both the type and the maintenance of permeable pavements have significant effects on their performance in the flood mitigation.


Introduction
Urban flooding/waterlogging has been a serious problem in many cities around the world.For example, urban flood hazards caused huge damage in the last decades in China [1][2][3].Due to climate change and rapid urbanization [4][5][6], the risks of urban waterlogging are very likely to increase in the near future.The traditional urban rainwater management approaches and existing flood disaster defense structures have been proven to be ineffective in extreme weather events [7].Development of new approaches to supplement existing urban water management is urgently needed in many cities in the world.In 2014, the Chinese government proposed a plan of sponge city construction at the The research area is in a residential block with an area of about 0.58 km 2 in Nanjing, China (Figure 1).It is adjacent to the Qinhuai River, a tributary of the Yangtze River.Because of its elevation, being lower than the water level of the Qinhuai River in the rainy season, it is one of the most seriously waterlogged areas in Nanjing City.The impervious underlying surfaces of the study area account for 73.8% of the total area.Among them, there are 0.15 km 2 of rooftops, 0.046 km 2 of parking lots and squares, 0.085 km 2 of non-busy roads and 0.143 km 2 of busy roads (Table 1).The non-busy roads are defined as roads less than 10 m wide with low traffic.A typical 20-year recurrence storm event which occurred in Nanjing from 19:00 June 12 to 7:00 June 13 in 1991 was selected for this study.The hourly rainfall data at Nanjing Meteorological Station was shown in Figure 2. The 12-h precipitation was 113.8 mm and 72% of this precipitation amount fell within two hours.

Flow Calculation
A hydrological model which includes impervious and pervious modules was used to simulate the hydrological processes.The impervious module is based on the soil conservation service (SCS) curve number method [37] and the pervious module is based on Horton's infiltration model [38].The model has the advantages of less requirements of parameters and observation data (only precipitation).

Impervious Module
The SCS curve number method has been widely used in permeable pavement studies with acceptable model performance [18,39].The key parameter of the SCS method is the curve number (CN) value, which is determined by a series of factors including land use, soil type and antecedent moisture condition.According to a previous study [18] and the reference CN values given in the manual of the model [37], a CN value of 94 was set for the impervious surfaces in this study.The terminal equation for impervious surface runoff computation is: where S is the potential maximum soil moisture retention, equal to − 254; is the initial abstraction, equal to 0.2 ; P is the precipitation (mm) and RF is the surface runoff volume (mm).
When P is less than 3.24 mm, there is no surface runoff.

Pervious Module
Horton's infiltration model [38] was applied to determine the infiltration rate: where f (t) is the infiltration rate (mm/h) at time t; k is the decay constant for the Horton curve; f0 is the maximum infiltration rate (mm/h) and fc is the minimum infiltration (mm/h).Soil moisture was from the infiltration by the force of capillary action [40].Then, the capillary action and gravity infiltration rates (f1 and f2) can be obtained using the following equations:

Flow Calculation
A hydrological model which includes impervious and pervious modules was used to simulate the hydrological processes.The impervious module is based on the soil conservation service (SCS) curve number method [37] and the pervious module is based on Horton's infiltration model [38].The model has the advantages of less requirements of parameters and observation data (only precipitation).

Impervious Module
The SCS curve number method has been widely used in permeable pavement studies with acceptable model performance [18,39].The key parameter of the SCS method is the curve number (CN) value, which is determined by a series of factors including land use, soil type and antecedent moisture condition.According to a previous study [18] and the reference CN values given in the manual of the model [37], a CN value of 94 was set for the impervious surfaces in this study.The terminal equation for impervious surface runoff computation is: where S is the potential maximum soil moisture retention, equal to 25400 CN − 254; I a is the initial abstraction, equal to 0.2 S; P is the precipitation (mm) and RF is the surface runoff volume (mm).When P is less than 3.24 mm, there is no surface runoff.

Pervious Module
Horton's infiltration model [38] was applied to determine the infiltration rate: where f (t) is the infiltration rate (mm/h) at time t; k is the decay constant for the Horton curve; f 0 is the maximum infiltration rate (mm/h) and f c is the minimum infiltration (mm/h).Soil moisture was from Water 2018, 10, 172 5 of 12 the infiltration by the force of capillary action [40].Then, the capillary action and gravity infiltration rates (f 1 and f 2 ) can be obtained using the following equations: where W (t) is the soil moisture (mm) at time t; WM is the storage capacity (mm).Then, surface runoff intensity (R s ) and subsurface flow intensity (R g ) have the following relations with precipitation intensity (I): If soil moisture (W) is known at the beginning of a precipitation event, surface runoff yield (h s ), subsurface runoff yield (h g ) and increased soil water (F) at any duration (x) of the event can be calculated using the following equations: In the pervious module, the mandatory parameters are f 0 , f c , k and W at the beginning of a precipitation event.

Permeable Pavement Scenarios and Model Setting
Three kinds of permeable pavements (permeable asphalts (PA), permeable concretes (PC) and permeable interlocking concrete pavers (PICP)) were installed into non-busy roads, parking lots and squares in the study area.Nanjing is one of the cities suffering air pollution, which resulted in a high concentration of PM2.5, thus the performance of permeable pavements is markedly affected by clogging over their service life.Considering the impact of clogging, two conditions (good and poor) of permeable pavements in terms of infiltration and void rate differences were evaluated.The good permeable pavements have high infiltration and void rates without the impact of fine particles.The poor condition of permeable pavements is due to the effect of the deposition of fine particles.The good and poor infiltration rates and void rates of all the three permeable pavements were set according to the recommendations by MHURD [12] and the previous studies [29,30,[33][34][35].The values are given in Table 2. Two settings of the initial water in permeable pavements are used (no water and 50% of maximum water storage capacity).All the simulated scenarios are listed in Table 3. Permeable pavements are implemented on all non-busy roads, parking lots and squares.The total implementation area is 0.13 km 2 .Permeable pavements are assumed to be continuous pavement systems and no infiltration water goes into soils under the layer of permeable pavements.In this study, the mandatory parameters for original condition simulation were obtained from Gao [40] who studied urban flooding with the same storm event in a watershed in which our study area is located.The values of f 0 , fc and k are 199.8mm/h, 12 mm/h and 1.98, respectively.The initial soil moisture at the beginning of the storm event was set to be half of the maximum water storage capacity.The parameters for permeable pavement scenarios were calculated based on the attribute values listed in Table 2.The value of f 0 was equal to permeability and k was calculated using Equations ( 4) and (12).All the mandatory parameters were listed in Table 4.
where VR is the void ratio of permeable pavements and TK is thickness (mm).

Effectiveness of Permeable Pavements on Flood Mitigation
Table 5 indicates the results of the total 12-h surface runoff under original conditions without permeable pavement implementation and 12 scenarios with different permeable pavements.Rainfall falling on permeable pavements and the flow from the impervious surface to permeable pavements Water 2018, 10, 172 7 of 12 could be stored in the pavements.The total runoff was reduced by 1-40% (Figure 3).This is consistent with other studies.For example, Luan et al. [41] reported that permeable pavements can reduce 2-12% of runoff for different rainfall intensities in a watershed in Beijing, China.Ahiablame and Shakya [42] found that permeable pavements reduced 15-41% of flood events under different implementation areas in a watershed in Illinois, USA.Also, it was found that peak flow decreased by 7-43% with the implementation of permeable pavements (Figure 4).Nine of the 12 scenarios had a lag time on peak flow, with the exception of Scenarios 8, 10 and 12 (Figure 5).Similar findings were also found by other studies [21,23].For example, Huang et al. [21] reported an approximately 29% decrease in peak flow with the implementation of permeable pavements in the new campus of Tianjin University, China.with other studies.For example, Luan et al. [41] reported that permeable pavements can reduce 2-12% of runoff for different rainfall intensities in a watershed in Beijing, China.Ahiablame and Shakya [42] found that permeable pavements reduced 15-41% of flood events under different implementation areas in a watershed in Illinois, USA.Also, it was found that peak flow decreased by 7-43% with the implementation of permeable pavements (Figure 4).Nine of the 12 scenarios had a lag time on peak flow, with the exception of Scenarios 8, 10 and 12 (Figure 5).Similar findings were also found by other studies [21,23].For example, Huang et al. [21] reported an approximately 29% decrease in peak flow with the implementation of permeable pavements in the new campus of Tianjin University, China.with other studies.For example, Luan et al. [41] reported that permeable pavements can reduce 2-12% of runoff for different rainfall intensities in a watershed in Beijing, China.Ahiablame and Shakya [42] found that permeable pavements reduced 15-41% of flood events under different implementation areas in a watershed in Illinois, USA.Also, it was found that peak flow decreased by 7-43% with the implementation of permeable pavements (Figure 4).Nine of the 12 scenarios had a lag time on peak flow, with the exception of Scenarios 8, 10 and 12 (Figure 5).Similar findings were also found by other studies [21,23].For example, Huang et al. [21] reported an approximately 29% decrease in peak flow with the implementation of permeable pavements in the new campus of Tianjin University, China.The flood mitigation performance varied with the types of permeable pavement.PA reduced 1-37% of total surface runoff and 7-36% of peak flow; PC reduced 3-40% of total surface runoff and 10-42% of peak flow; and PICP reduced 4-36% of total surface runoff and 12-32% of peak flow.Overall, PC showed the best performance (maximum decrease of 40% of total runoff and 42% of peak flow), followed by PA (maximum decrease of 37% of total runoff and 36% of peak flow) and PICP (maximum decrease of 36% of total runoff and 32% of peak flow) under good conditions.Similarly, Collins et al. [28] found different effects of four kinds of permeable pavements on runoff generation in Carolina, USA.

Impact of Clogging on the Performance of Permeable Pavements
The clogging of permeable pavement has a negative impact on flood mitigation.Permeable pavements under good conditions reduced total runoff by 15-40% (Scenarios 1-6), whereas under poor conditions runoff was only reduced by 1-14% (Scenarios 7-12).The effectiveness of runoff reduction decreased by 62-92%.Also, the peak flow was reduced by 37-65% under poor conditions.This is because that the fine particles in permeable pavements have reduced total and blocked the pores, with the result of reducing infiltration rates.Some studies have reported a significant reduction in infiltration rates of permeable pavements within several years due to The flood mitigation performance varied with the types of permeable pavement.PA reduced 1-37% of total surface runoff and 7-36% of peak flow; PC reduced 3-40% of total surface runoff and 10-42% of peak flow; and PICP reduced 4-36% of total surface runoff and 12-32% of peak flow.Overall, PC showed the best performance (maximum decrease of 40% of total runoff and 42% of peak flow), followed by PA (maximum decrease of 37% of total runoff and 36% of peak flow) and PICP (maximum decrease of 36% of total runoff and 32% of peak flow) under good conditions.Similarly, Collins et al. [28] found different effects of four kinds of permeable pavements on runoff generation in Carolina, USA.

Impact of Clogging on the Performance of Permeable Pavements
The clogging of permeable pavement has a negative impact on flood mitigation.Permeable pavements under good conditions reduced total runoff by 15-40% (Scenarios 1-6), whereas under poor conditions runoff was only reduced by 1-14% (Scenarios 7-12).The effectiveness of runoff reduction decreased by 62-92%.Also, the peak flow was reduced by 37-65% under poor conditions.This is because that the fine particles in permeable pavements have reduced the total void and blocked the pores, with the result of reducing infiltration rates.Some studies have reported a significant reduction Water 2018, 10, 172 9 of 12 in infiltration rates of permeable pavements within several years due to clogging [17,43,44].This could be reduced to certain levels, such as 1% of runoff reduction in our study within several years without proper maintenance.In addition, PICP was the least prone to clogging among the three types of permeable pavements.PICP showed the best performance under poor conditions (decreased by 4-14% of total runoff and 12-20% of peak flow), followed by PC (maximum decrease by 3-12% of total runoff and 10-22% of peak flow) and PA (maximum decrease by 1-9% of total runoff and 7-21% of peak flow).Similarly, Bean et al. [30] found that PICP was less prone to being clogged when compared with other permeable pavements in a survey of 40 permeable pavement sites in the USA.

Impact of Initial Water Content on the Performance of Permeable Pavements
The effectiveness of permeable pavements in terms of flood mitigation is affected by initial water content at the beginning of a storm event.The reduction in total runoff and peak flow decreased by 57-85% and 37-67% due to the increase of initial water content at the beginning of the simulation.For example, PA under good conditions with maximum void at the beginning (Scenario 1) reduced 37% of the surface runoff, which was more than twice the value (16%) in the case of 50% initial water content (Scenario 2).
In addition, the effect was intensified by the interaction of initial content and clogging.For example, the performance of PA on peak flow reduction decreased by 44% between Scenarios 1 and 2. This value increased to 67% between Scenarios 7 with 8. Similarly, there was a 42% decrease in PA effectiveness of peak flow reduction between Scenarios 1 and 7 and this value rose to 65% between Scenarios 2 and 8. Also, the changes in lag time of the peak flow were not evident under the integrated impacts of clogging and initial water content (Scenarios 8, 10 and 12, see Figure 5).

Implications and Limitations
As a case study in Nanjing, China, this research demonstrated the effectiveness of permeable pavements on flood mitigation and indicated the impact of type, clogging and initial water content on the hydrological performance of permeable pavements.It provides valuable information for decision makers regarding the effectiveness of permeable pavements for sponge city construction in Nanjing and other cities in China.With increasing challenges regarding water resources, the Chinese government has invested more money in environmental protection in recent years [45,46].However, to develop an achievable plan for permeable pavement implementation, economic considerations should be taken into account.
In line with numerous other studies, there are several limitations to the current research.This study lacks data for model calibration and validation, which would lead to biases in the absolute values of runoff reductions.However, model parameter values were obtained from the published literature and the main conclusions of the impact of types, clogging and initial water content were conducted from scenario comparisons.The impact of model uncertainties is not significant and will not change the conclusions of this study.Consideration of these factors in future studies will increase our understanding of urban floods in China and other countries.

Conclusions
This study evaluated the role of permeable pavements in mitigating urban floods in the context of sponge city construction in China.It was found that the hydrological effects of permeable pavements on flood mitigation were highly dependent on the types of permeable pavements, the clogging situation of permeable pavements, and the initial water content at the beginning of the storm event.The major findings include: (1) Permeable pavements reduced surface runoff by 1-40% and peak flow by 7-43% in a 12-h storm event with rainfall of 113.8 mm.
(2) The hydrological performance of permeable pavements varied widely depending upon whether their function was fully applied.Without the impact of clogging, PC showed the best performance, followed by PA and PICP.(3) Clogging reduced the hydrological performance of permeable pavements.PICP was the least prone to being clogged, followed by PC and PA.PA, PC and PICP reduced the total surface runoff by 77-92%, 71-82% and 62-73%, respectively; and reduced peak flow by 41-65%, 48-50% and 37-40%, respectively.(4) The hydrological performance of permeable pavements decreased with increasing initial water content.Compared to permeable pavements without initial water content, the effectiveness of the cases with 50% initial water content decreased the total runoff by 57-85% and peak flow by 37-67%, respectively.

Figure 1 .
Figure 1.Study area of the residential block in Nanjing, China.

Figure 1 .
Figure 1.Study area of the residential block in Nanjing, China.

Figure 2 .
Figure 2. Hourly precipitation at Nanjing Meteorological Station from June 12 to June 13 in 1991.

Figure 3 .
Figure 3. Percent of total surface runoff reduction by implementation of different permeable pavements.

Figure 4 .
Figure 4. Percent of peak flow reduction by implementation of different permeable pavements.

Figure 3 .
Figure 3. of total surface runoff reduction by implementation of different permeable pavements.

Figure 3 .
Figure 3. Percent of total surface runoff reduction by implementation of different permeable pavements.

Figure 4 .
Figure 4. Percent of peak flow reduction by implementation of different permeable pavements.

Figure 4 . 12 Figure 5 .
Figure 4. Percent of peak flow reduction by implementation of different permeable pavements.

Table 1 .
Hourly precipitation at Nanjing Meteorological Station from June 12 to June 13 in 1991.Underlying surface types and proportions in the study area.

Table 1 .
Underlying surface types and proportions in the study area.

Table 2 .
Parameters of three permeable pavements.
Note: Void ratios are the average values of all layers.

Table 3 .
Simulation scenarios of permeable pavements.

Table 4 .
Mandatory parameters for all scenarios in model simulation.

Table 5 .
Total 12-h surface runoff under original condition and all scenarios.

Table 5 .
Total 12-h surface runoff under original condition and all scenarios.

Table 5 .
Total 12-h surface runoff under original condition and all scenarios.