Challenges in urban stormwater management in Chinese cities: A hydrologic perspective

Abstract For managing the worsening urban water disasters in China, the Government of China proposed the concept of “Sponge City” in 2013 and initiated the strategy in 30 pilot cities from 2015. Despite the promise of the concept, there have been many challenges in implementing the “Sponge City” program (SCP). In this manuscript, we discuss the hydrology-related challenges in implementing the SCP. In particular, we consider two key challenges: (1) Determination of the “Volume Capture Ratio of Annual Rainfall” (VCRAR), as controlling urban stormwater runoff is one of the core targets of the SCP; and (2) Estimation of a proper rainfall threshold, which influences the layout of green-infrastructures in the SCP to achieve the core VCRAR target. To discuss these challenges, we consider the city of Beijing, the capital of China, as a case study. Our analysis shows that the trade-offs between the investment for the SCP and its potential economic benefits should be considered by undertaking a proper determination of VCRAR. The VCRAR estimated for Beijing from the present analysis is 0.73. This value is more reasonable than the empirical value of 0.80 that is presently used, as it can guarantee the positive rate of return on the investment. We also find that the nonstationary characteristics of rainfall data and their spatiotemporal differences are important for the estimation of the rainfall threshold in SCP. For instance, even using the daily rainfall data over a period of 30 years (1983–2012) in Beijing, as required by the National Assessment Standard, the estimated rainfall threshold of 27.3 mm underestimates the reasonable rainfall threshold that should at least be larger than 30.0 mm. Thus, the former cannot ensure the VCRAR target of 0.80. Based on these results, we offer proper approaches and key suggestions towards useful guidelines for delivering better SCP in the Chinese cities.


Abstract:
For managing the worsening urban water disasters in China, the 23 Government of China proposed the concept of "Sponge City" in 2013 and initiated the 24 strategy in 30 pilot cities from 2015. Despite the promise of the concept, there have 25 been many challenges in implementing the "Sponge City" program (SCP). In this 26 manuscript, we discuss the hydrology-related challenges in implementing the SCP. In 27 particular, we consider two key challenges: (1) Determination of the "Volume Capture 28 Ratio of Annual Rainfall" (VCRAR), as controlling urban stormwater runoff is one of 29 the core targets of the SCP; and (2) Estimation of a proper rainfall threshold, which 30 influences the layout of green-infrastructures in the SCP to achieve the core VCRAR 31 target. To discuss these challenges, we consider the city of Beijing, the capital of 32 China, as a case study. Our analysis shows that the trade-offs between the investment 33 for the SCP and its potential economic benefits should be considered by undertaking a 34 proper determination of VCRAR. The VCRAR estimated for Beijing from the present 35 analysis is 0.73. This value is more reasonable than the empirical value of 0.80 that is 36 presently used, as it can guarantee the positive rate of return on the investment. We 37 also find that the nonstationary characteristics of rainfall data and their spatiotemporal 38 differences are important for the estimation of the rainfall threshold in SCP. For 39 instance, even using the daily rainfall data over a period of 30 years  in 40 Beijing, as required by the National Assessment Standard, the estimated rainfall 41 threshold of 27.3 mm underestimates the reasonable rainfall threshold that should at 42 least be larger than 30.0 mm. Thus, the former cannot ensure the VCRAR target of 43 pavements, grass swales, etc.) to control stormwater pollution, reduce runoff volume, 66 and relieve combined sewer overflow (Li et al., 2016). However, the stormwater 67 problems encountered in many mega cities around the world are too complex to be 68 solved only depending on those source control facilities. In 2000, the GIs were 69 proposed as a strategic framework for environmental, social, and economic 70 sustainability of cities (Benedict, 2000). Presently, the WSUD is gaining popularity. It 71 is an integrated concept for urban planning and designs based on water environments, 72 for balancing different land use types and for protecting the water cycle, so that the 73 city will be sustainable and ecologically friendly (Coombes et al., 2000). Apart from 74 these concepts, many other new city concepts and ideas have also been proposed, 75 including "sustainable cities", "livable cities", "intelligent cities", "eco cities", "low 76 In this study, we address two key hydrology-related challenges, which play vital 106 roles and, hence, are top priorities in designing the SCP. The first is concerned with 107 the determination of the "Volume Capture Ratio of Annual Rainfall" (VCRAR), as the 108 core target of the SCP; and the second is the estimation of a proper rainfall threshold, 109 which influences the layout of Green-infrastructures (GIs) to achieve the above core 110 VCRAR target (Randall, 2019). As shown in Figure 1, we denote the total rainwater 111 of a region as R0 (sum of all blue (or red) rainfall intensities in Figure 1) and denote 112 the rainwater magnitude that can be controlled (through infiltration, storage, and 113 evaporation) by the SCP as R1 (the part below the rainfall threshold, T * (blue) (or 114 T * (red)), in Figure 1); a VCRAR target requires that the ratio between R1 and R0 115 should be no smaller than VCRAR, and thus it is the core target of the SCP. To 116 achieve this, a proper rainfall threshold T * should be estimated. To be specific, the 117 suitable layout of GIs (green roofs, bio-retention cells, permeable pavements, etc.) 118 should be designed and implemented, based on which all the rainfall intensities below 119 T * should be controlled, and their sum should be no smaller than R1, to achieve the 120 above core VCRAR target. However, in the current implementation of the SCP in 121 Chinese cities the VCRAR is empirically determined, which lacks reliable scientific 122 basis. Furthermore, the nonstationary variability of rainfall is not given adequate 123 consideration in the estimation of proper rainfall threshold. For example, the 124 estimated rainfall threshold to ensure the same VCRAR target is different when using 125 the two rainfall samples (blue and red) in Figure 1. 126

<Figure 1> 127
To address the above two issues here, we consider the city of Beijing, the capital of 128   Table 1. 175 Therefore, the choice of the data period (or record length) would inevitably influence 176 the estimation of a proper, or at least a reasonable, rainfall threshold, leading to a 177 biased rainfall threshold that cannot ensure the VCRAR target. More details on this 178 will be discussed in Section 4. 179 <Table 1> 180

Volume Capture Ratio of Annual Rainfall (VCRAR) 181
The VCRAR is the core target in the design of the SCP, and a key index to 182 quantify its hydrological effects (from rainfall). Therefore, determination of a proper 183 VCRAR is a vital issue in the implementation of the SCP. Presently, the VCRAR is In this study, we propose the following approach for the determination of the 197 proper VCRAR by considering these trade-off issues: 198 (1) estimate the total investment of the SCP under different VCRAR targets; 199 (2) use historical hydrological and natural disasters data, and establish the 200 economic loss curve of urban stormwater disasters in the concerned study area; 201 (3) use the above curve to estimate the annual economic losses without the SCP; 202 (4) use the same curve to estimate the annual economic losses under different 203 VCRAR targets and compare their differences with and without the SCP, aimed at 204 identifying the annual economic benefits from the SCP; 205 (5) normalize the total investment for the SCP and its annual economic benefits and 206 calculate the change rates for analyzing the trade-offs (The normalization is needed, 207 since the investment and benefits cannot be directly compared); and 208 (6) identify the VCRAR value below which the increased rates of annual economic 209 benefits stay higher than the total investment. This is regarded as the proper VCRAR 210 value (This is reasonable, since a higher benefit with lower investment is always 211 desirable and the expectation). 212 For the city of Beijing, considered here as a case study, the economic loss curve of 213 urban stormwater disasters is obtained from Yang et al. (2016). Based on this, the 214 trade-offs between the investment for the SCP and its potential benefits in Beijing are 215 analyzed using the above approach. Figure 3(a) indicates that with an increase in the 216 VCRAR target, the total financial investment continues to increase and the annual 217 economic loss continues to decrease in this urban area, as expected, corresponding to 218 the increase in its annual economic benefits. It also shows that the change rate of the 219 total investment exponentially increases with an increase in the VCRAR target. This 220 is different from the change rates of its annual economic benefits. 221

Rainfall threshold 240
For achieving the VCRAR target determined from the approach proposed above, 241 estimation of a proper rainfall threshold is important. That is, through a rational layout 242 of GIs (green roofs, bio-retention cells, permeable pavements, etc.) and their 243 constructions, rainfall intensities below a certain rainfall threshold are required to 244 intercept and control, based on which the VCRAR target is expected to achieve (as 245 shown in Figure 1). to be estimated using the following approach: 251 (1) select the daily rainfall data samples (denoted as X0) with more than 30 years 252 and remove those data samples with the intensities smaller than 2 mm/day. Then, take 253 the residual as the effective rainfall data samples (denoted as X and denoted its total 254 amount as R); 255 (2) set a small rainfall threshold Tj and divide the effective rainfall data samples X 256 into two parts (above and below Tj), with the amount Rj a and Rj b (i.e., R = Rj a + Rj b ), 257 respectively, in order to evaluate the ratio Sj = Rj b /R; 258 (3) increase the value of Tj and repeat the above steps to obtain the time series Sj; 259 and 260 (4) the T * when its Sj is equal to the determined VCRAR is the estimated rainfall 261

threshold. 262
There are two major drawbacks in the above approach. First, it does not consider 263 the influence of data period (and also record length) of the rainfall data samples used. 264 Note that the record length always has some influence, even if it is longer than 30 265 years; in general, the longer the data, the better and more reliable are the outcomes. These problems can be explained with the case study of the city of Beijing. Figure  280 4 shows the results for two different scenarios at the Guanxiangtai meteorological 281 station: one with rainfall data over different time periods (Figure 4(a)) and the other 282 with rainfall data at different temporal resolutions (Figure 4(b)). More specifically, 283 five different time periods (1951-1965, 1966-1996, 1997-2016, 1966-2016, and 284 1951-2016) and five different temporal resolutions (1-hr, 2-hr, 6-hr, 12-hr, and 24-hr) 285 are considered. As seen from Figure 4(a), the estimated rainfall thresholds for a given 286 VCRAR are different for the five periods, due to the nonstationary characteristics of 287 rainfall variability (as shown in Table 1). Furthermore, the rainfall thresholds 288 corresponding to a given VCRAR also differ for different temporal resolutions of 289 rainfall data samples (see Figure 4(b)). These different rainfall thresholds require 290 different layouts of GIs and their distinct construction standards, for guaranteeing the 291 VCRAR targeted. 292 The rainfall threshold used in the present implementation of the SCP in Beijing was 293 estimated as 27.3 mm (as explained in Section 2), to ensure the VCRAR target of 0.80 294 that is presently used. However, a reasonable rainfall threshold should at least be 295 larger than 30.0 mm, no matter considering longer data periods (see Figure 4 This implies that more efforts are needed to improve the SCP in Beijing and to 300 improve the urban water management in the region. 301

<Figure 4> 302
By further considering the spatial heterogeneity of statistical characteristics of 303 extreme rainfall events, rainfall thresholds should also have spatial changes (Sang and 304 Yang, 2016), which cannot easily be estimated by the uniformed approach from the 305 National Assessment or Guidance standard of the SCP. Indeed, the rainfall threshold 306 estimated from the uniformed approach would have large numerical bias or errors, 307 influencing the layout of the GIs, and causing inaccurate investment budget and low 308 effects for the SCP. 309 For obtaining more explicit and precise estimation of the rainfall threshold, it is 310 important to analyze the nonstationary characteristics of rainfall by considering the 311 influences of the above two factors, even though rainfall data over long periods and at 312 high temporal resolutions are not easily available in China, and globally more broadly. 313 It is suggested, therefore, that a set of rainfall thresholds be estimated to reflect 314 different rainfall situations, with at least the maximum, average, and minimum rainfall 315 thresholds estimated, for supporting the rational layout of green-infrastructures and 316 the estimation of investment budget for the SCP. 317 318

Closing Remarks 319
The Sponge City program (SCP), which is based on an accurate understanding of 320 hydrological characteristics, is an important direction of development in urban 321 stormwater management in China. However, there are some key issues in the present 322 implementation of the SCP in different Chinese cities. This study discussed two key 323 hydrology-related issues, which are also common problems in the current SCP 324 implementation in Chinese cities, that need to be considered prior to the design and 325 implementation of the SCP: (1) determination of Volume Capture Ratio of Annual 326 Rainfall VCRAR); and (2) estimation of a proper rainfall threshold. With a case study 327 of the city of Beijing, the present study proposed new approaches to address the above 328 two issues. The results from the proposed approaches are certainly encouraging. 329 Application of the approaches to other Chinese cities (as shown in Figure 2), where 330 similar situations and many other complex problems exist, would help realize their 331 suitability for the SCP, and urban stormwater management, more broadly. 332 It is important to note, at this point, that approaches for estimation of proper 333 VCRAR and rainfall threshold are still in their early stages. There is still a long way 334 to go in our efforts to mitigate the problems associated with urban stormwater 335 management (e.g. waterlogging) in Chinese cities, which continues to be a very 336 complex and challenging issue. With the anticipated impacts of climate change, the 337 increasing trend (i.e. more frequent and greater magnitude) of regional hydroclimatic 338 extremes (especially floods) is very likely to continue, and even accelerate in the 339 future. Therefore, their potential risks and influences should be further evaluated in annual rainfall (VCRAR) and the rainfall threshold (T * ) to ensure the VCRAR target. 474 Here, the blue and red time series represent two rainfall samples. Using the two 475 rainfall thresholds based on these samples to ensure the same VCRAR are different. (VCRAR) and the total investment, and the annual economic losses and benefits by 485 considering the effects of the SCP in Beijing. (b) Relationship between the VCRAR 486 and the change rates of the normalized investment and its annual economic benefits. 487 Here, the daily rainfall data measured at the representative Guanxiangtai station in 488 Beijing, with the measured period from 1951 to 2016, are used for the calculation. 489 ensure the VCRAR target of 0.80 that is presently used in Beijing.

498
Tables in the manuscript 499 500