graphic
View largeDownload slide

The sponge city is a new concept of stormwater management for ecological city construction, which aims to restore water-cycle processes and reduce runoff. Cities in coastal districts are suffering from serious instability due to high population density, urbanization, and land-use changes. However, previous research contains few evaluations of balancing urban ecological indicators of sponge city performance, including geographical, environmental, economic, and social factors, and their effect on resilience at a macro level to develop low-impact development schemes. In this study, we developed an integrated framework using factor analysis, geographical statistics, multi-objective analysis, and remote sensing methods to extract the factors influencing sponge city resilience and to establish spatial pattern schemes. The results indicated that the urbanization degree and plant adaptability had the greatest impact on sponge city performance, with weights of 45 and 27%, respectively. Sponge city spatial pattern schemes performed the best in the combination scenario of 14.8–46.8% green roofs (by area ratio) supported by grooves and rain barrels +10% herbaceous basins divided into units by ecological tree pools +10% permeable pavements and sidewalks. This scenario balanced facilities and cost to optimize the spatial pattern, which improved sponge city adaptability and urban ecological conditions.

  • An integrated evaluation system analyzing urban resilience was proposed.

  • We first established a link between sponge city and urban ecological resilience.

  • Factors of urbanization degree and plant adaptability possessed the most impact.

  • Green roofs were suitable for applying in high urbanized areas with dense buildings.

  • The framework could provide designers with valid experience in urban management.

coastal area, ecological indicators, factor analysis, geographical statistics, sponge city, urban resilience

Rapid urbanization has caused wetlands and water bodies to disappear, which has disrupted natural water-cycle processes, reduced stormwater storage capacity, decreased groundwater recharge, increased runoff peaks, and led to urban waterlogging (Bah et al. 2023). Traditional urban engineering infrastructures (e.g., drainage pipeline networks) cope mainly with rainwater with a short return period of between one and 10 years, which is not enough to alleviate waterlogging (Jia et al. 2017). As urban stormwater problems become increasingly severe, the use of ‘hard’ projects such as impervious drainage ditches and underground pipeline networks has been found to be insufficient (Putri et al. 2023). At present, stormwater management has become a worldwide focus, and countries have developed concepts and natural facilities (e.g., green roofs, permeable pavements, rain barrels, herbaceous basins, and rain gardens) based on their own situation (Koc et al. 2021). Examples include low-impact development (LID) in the US, sustainable urban drainage systems (SuDS) in the UK, and water-sensitive urban design (WSUD) in Australia (Ferrans et al. 2022). In the same vein, the first ‘sponge city’ initiative in China was proposed in 2011.

About 2.4 billion (∼40%) people in the world live within 100 km of a seacoast and suffer from serious water stress and instability due to high population density, urbanization, and land-use changes (Bredes et al. 2023). Coastal cities tend to suffer more severe damage when disasters occur, and therefore applying improved and optimized sponge city strategies in these locations is a significant step (Pereira et al. 2017). However, the effectiveness of a sponge city strategy can be affected by several factors, such as the factors of geographical environment, budget, soil surface, land use, and socioeconomic conditions (Yuan et al. 2022). For instance, drainage through an urban drainage pipeline will be hindered by high ocean tidal water level, which will decrease the city's overall ability to control stormwater (Li et al. 2018). Other researchers have indicated that annual mean precipitation can reach 900–2,500 mm because of the monsoon system, but that various combinations of sponge city measures appear to offer benefits in this context (e.g., decentralized controls could be a promising strategy) (Jackisch & Weiler 2017). The ability of sponge cities to control stormwater generally depends on infiltration from soil, including permeable pavements that transform impervious surfaces and infiltrate runoff to the soil (Eckart et al. 2017). Storage and infiltration capabilities to reduce runoff could be increased by soil amendments such as compost, lime, or organic materials, which could alter the physical, chemical, and biological characteristics of soils to improve plant growth (Chen et al. 2023). Moreover, sponge cities are always costly to build, and developed countries could become more reluctant to deal with waterlogging and stormwater problems due to worsening socioeconomic conditions such as lack of funding, low environmental awareness, and lack of public acceptance (Ishaq et al. 2023).

Principal component analysis (PCA) is widely used to analyze the influence of a series of factors on a process; factor analysis is an extension of this approach. Previous studies of sponge cities and factor analysis optimization have been mostly limited to providing schemes based on analyzing the impact of each factor and of geographical features at a small scale; few have calculated the comprehensive benefits after implementing an optimal plan. Impact factors could be extracted using factor analysis, after which the benefits of sponge city schemes in mitigating damage to resilience could be quantified using geographical statistics such as Moran's I and Getis-Ord Gi*.

The present study first integrated factor analysis, geographical statistics, multi-objective analysis, and remote sensing into a new framework, which included and balanced both geographical and socioeconomic factors at a macro level and then optimized sponge cities in detail with remote sensing in coastal areas in China, taking Tianjin as an example. In this research, the most significant factors affecting sponge city benefits and their effect weights were obtained by factor analysis. The urban resilience distribution in the study area was then obtained through a ‘resilience damage’ score based on ecological factors in geographical statistics. Finally, the spatial patterns of optimal sponge city schemes and their comprehensive benefits represented by resilience promotion were analyzed. The proposed evaluation system could help to integrate sponge city planning into urban planning systems and reduce the waste of natural and social assets. Moreover, our study can serve as a reference to designers around the world for sponge city optimization and waterlogging risk reduction.

Study area

The study area is located in the central district of Tianjin City, a coastal area in China covering 668.54 km2 (Figure 1). It has a shallow depth of groundwater and a high degree of salinity, making it prone to seawater intrusion and secondary soil salinization, with a mean annual temperature of 12–15 °C and a mean annual precipitation of 550–600 mm (with 75% of rainfall occurring between June and August). The central district is highly urbanized, consisting mainly of grey roofs and road surfaces with few greenbelt and water areas and low overall water permeability.
Figure 1
Location of the study area.
View largeDownload slide

Location of the study area.

Figure 1
Location of the study area.
View largeDownload slide

Location of the study area.

Close modal

Research framework and data

The framework of this research includes impact factor analysis, geographical statistical analysis, remote sensing analysis, tendency to waterlogging, methods in stormwater management, multi-objective optimization, and discussion (Figure 2(a)). First, the likely geographical environmental and socioeconomic factors were analyzed according to research results on the 30 pilot sponge cities in China obtained from the CNKI (China National Knowledge Infrastructure, https://www.cnki.net/), Web of Science (https://www.webofscience.com/wos/woscc/basic-search), and Google Scholar (https://scholar.google.com/). After the factor analysis calculations, factors with significant influence on sponge city performance and their effect weights were determined. Urban resilience in Tianjin was also calculated by further elaborating the indicators based on the factor analysis results using spatial autocorrelation analysis. Once the resilience distribution had been analyzed, the effects of various factors on stormwater control were determined, and several optimal plans were formulated according to each factor and detailed indicator. The pattern of LID facilities was then designed and combined with actual urban structures by remote sensing. By comparing the urban resilience distributions after implementing an optimization strategy, the resilience promotion benefits of each optimal plan were analyzed, and the optimal plan was determined through the multi-objective methods used to balance each factor. To supplement geographical feature analysis and remote sensing, the Landsat 8 OLI_TIRS satellite digital product (resolution ratio: 30 m) and the GDEMV3 digital elevation data (resolution ratio: 30 m) from the Geospatial Data Cloud (http://www.gscloud.cn/) were also used.
Figure 2
Research framework for the study: (a) framework for formulating optimal strategy in Tianjin and (b) framework for analyzing factors and their effect size.
View largeDownload slide

Research framework for the study: (a) framework for formulating optimal strategy in Tianjin and (b) framework for analyzing factors and their effect size.

Figure 2
Research framework for the study: (a) framework for formulating optimal strategy in Tianjin and (b) framework for analyzing factors and their effect size.
View largeDownload slide

Research framework for the study: (a) framework for formulating optimal strategy in Tianjin and (b) framework for analyzing factors and their effect size.

Close modal

Factor analysis

Generally, the operation of a sponge city infrastructure is affected by the geographical environment, the economy, and society. Geographical environment factors include climate, hydrology, topography, soil properties, air quality, and land-use types. Economic factors can be divided into residents’ income, consumption level, and real estate development. Additional social factors can be categorized as population distribution, urban traffic conditions, and the state of scientific research. Macro-indicators with strong correlation can be selected for factor analysis when analyzing impact factors on sponge cities at a national scale. Moreover, in this research, the word ‘indicator’ is used to represent an influence on sponge cities before factor analysis, and the phrase ‘impact factor’ is used to show the indicators in each category (i.e., each factor) after classification.

Figure 2(b) shows the factor analysis framework. Compared with PCA, factor rotation and overall score calculation are added, which makes it more reasonable to classify each indicator into the final factors and to calculate their effect sizes (i.e., their weight values).

With the overall score calculated, the weight of each factor was also determined. The factor results were then applied to the study area, and further detail indicators were selected by analyzing conditions in Tianjin.

Formulas and calculations

PCA and factor rotation

PCA is a multivariate statistical technique that is widely used to reduce high dataset dimensionality and improve interpretability while minimizing information loss. Relevant studies have shown that PCA can comprehensively evaluate the effects on sponge city performance of environmental, economic, and social concerns by analyzing the weight of every factor (Zhang et al. 2021). Nevertheless, the factors extracted by PCA differ little in the loading of each variable, making the results difficult to explain and define in a professional manner. Factor analysis has extended PCA by adding factor rotation to change the projection area of each common factor in the direction of the original variable, so that the factors can be explained and named appropriately (Johnson & Wichern 2014).

Indicator standardization

The dimensions and orders of magnitude associated with the variables are not uniform, and it is often necessary to conduct standardized processing for the collected data, also called normalization.

First, to unify the effect of different indicators on sponge city performance (promoting or hindering), evaluation indicators are often divided into positive, negative, and moderate indicators because different types of indicators require different standardization approaches:

  • (1)

    Positive indicators should not be processed.

  • (2)
    Negative indicators should be calculated as the reciprocal:
    formula
    (1)
    where is the original variable, is the variable after standardization, n is the number of indicators, and p is the number of variables in each indicator.
  • (3)
    Moderate indicators are processed by the following formula:
    formula
    (2)
    where is the moderate value of the variables in this indicator.

After initial processing, the variables in each indicator should be standardized again by z-score standardization to reduce the impact of dimensional differences between different variables:
formula
(3)
formula
(4)
formula
(5)
where is the variable after z-score standardization, is the mean of the variables in each indicator, is the standard deviation of the variables, and is the variance of the variables in each indicator.
Eigenvalue calculation and principal component extraction
formula
(6)
formula
(7)
formula
(8)
where R is the correlation coefficient matrix, E is the unit matrix, is the eigenvalue of R, is the variance contribution corresponding to each principal component, is the cumulative variance contribution of multiple principal components, and m is the number of principal components extracted.

Generally, the number of principal components can be determined as those with an eigenvalue >1 and a cumulative variance contribution >50%.

Component matrix and factor rotation
The component matrix can be calculated as:
formula
(9)
formula
(10)
formula
(11)
where is the principal component loading matrix and is the eigenvector of each principal component corresponding to the eigenvalue. The Kaiser-Varimax rotation measure was used to rotate the component loading matrix (Kaiser 1958):
formula
(12)
where is the loading of the variables on the principal component (factor) matrix after rotation.
Overall score calculation
formula
(13)
formula
(14)
formula
(15)
formula
(16)
where is the scoring coefficient of a factor on the entire rotated component matrix, represents the score of on ⁠, is the ratio of the eigenvalue corresponding to the factor to the sum of eigenvalues of the extracted factor, is the composite score of each indicator, is the combined score of each factor, and ⁠, ⁠, and represent the sum of the composite score of all indicators contained in a given factor.
should be normalized to eliminate negative values for more accurate comparison with other measures:
formula
(17)
formula
(18)
where is the combined score of each factor after being normalized, D is the standard deviation of all indicators, and is the weight of each factor.

Geographical statistics

Global Moran's I
Moran's I is widely used to test for the presence of spatial dependence in observations taken on a lattice and is defined as (Li et al. 2007):
formula
(19)
where is the observed value of the variable at the spatial position or spatial unit i, is the average value of the variable at all points, is the spatial weight matrix, and n is the number of data points. The range of Moran's I is [−1,1], where a value less than 0 means negative correlation, one greater than 0 means positive correlation, and one equal to 0 means that each spatial object unit in the research area is independent of all others. The closer I is to 1, the more significant the agglomeration effect of an attribute in the spatial distribution of the research object. The closer I is to −1, the more significant the divergence of an attribute of the research object.
Local Moran's I
In Cliff & Ord (1981), the global Moran's I is a global parameter for measuring spatial autocorrelation, whereas the local Moran's I examines individual locations and identifies hotspots based on comparison with neighboring samples. Local Moran's I at spatial position i is defined as:
formula
(20)
The range of local Moran's I is not limited to [−1,1]; the statistic for the local Moran's I test is:
formula
(21)
where is the mathematic expectation of ⁠. When ⁠, means a high degree of clustering, and means a low degree of clustering in the vicinity of high values. When ⁠, means a low degree of clustering, and means a high degree of clustering in the vicinity of low values. is the variance of ⁠.
Getis-Ord Gi*

Global spatial autocorrelations such as global Moran's I and the Getis-Ord general G (a method similar to Global Moran's I that reveals whether high or low clustering tendencies are more significant) measure overall clustering or dispersion. In contrast, Getis-Ord Gi* is a local spatial autocorrelation statistic that is widely used to investigate specific spatial distributions and local clusters and that is more intuitive than the local Moran's I (Ord & Getis 1995).

Getis-Ord Gi* is defined as:
formula
(22)
where S is the standard deviation of the variable at all data points.

Multi-objective optimization module

According to She et al. (2021), constraints for the optimization variables could be defined based on the sponge city construction, specifications, and other engineering experience:
formula
(23)
formula
(24)
formula
(25)
formula
(26)
formula
(27)
formula
(28)
where A is the area ratio (%) of hard surfaces, is the initial hard-surface ratio, is the decrease in this ratio by applying LID, B is the cost of the LID facility, is its unit price, is the area of LID implementation, C is the runoff coefficient in the study area, E is the area ratio of bare land, D is the green space ratio, is the initial green space area ratio, is the proportion added by implementing LID, and is the ratio of other LID facilities without green space, such as permeable pavements. is the lowest cost of implementing sponge cities, which is the same as the minimum cost Bmin.

Factor extraction and classification

According to relevant literature on sponge city performance (Peng 2017; Koc et al. 2021; Zhu et al. 2022), 12 indicators that could significantly impact sponge cities were selected (Figure 3). For these 12 indicators, ‘number of sponge city studies’ represents the research popularity and importance of sponge city concepts in a region, which could promote the development and benefit improvement of sponge cities (Kaykhosravi et al. 2018). Moreover, gross domestic product (GDP) and real estate price are positive for sponge cities in the aspect of economy because of the great expense in construction and the demand for better living conditions (Xu et al. 2019). ‘Annual average temperature’, ‘shallow soil organic carbon’, and ‘soil permeability coefficient’ could impact LID plant growth and rainwater infiltration, which could promote the function of sponge cities and are positive indicators for sponge city performance (Deeb et al. 2018). ‘Number of motor vehicles’ and ‘population density’ often reveal the degree of development in a district; a more developed city tends to be obviously more favorable to sponge city construction (Li et al. 2016). Apparently, the cost of incorporating sponge city concepts in construction and maintenance increases the financial burden and imposes disadvantages on sponge city building. ‘Annual mean precipitation’ and ‘impervious surface coverage’ determine the scale of regional runoff, which aggravates the stormwater problem. Bartlett's sphericity and Kaiser–Meyer–Olkin (KMO) tests were used to validate the suitability of the datasets for factor analysis. The KMO value was 0.542, and Bartlett's sphericity test value at the given significance of 0.05 (sig) was 0.043, which revealed that the datasets were appropriate for factor analysis (KMO value at least >0.5) and that there were significant relationships among the 12 indicators (sig at least <0.05; Mukherjee & Singh 2020). Indicators with similar loading values in one line of the factor matrix could be classified into one factor based on their practical definition (Johnson & Wichern 2014). Four factors were selected after PCA and factor rotation with the following factor loadings: A (Urbanization degree): 0.922, 0.705, 0.803, 0.644, 0.766; B (Financial burden): −0.118, −0.112; C (Rainwater characteristics): −0.086, −0.068; D (Plant adaptability): −0.043, −0.035, 0.705, which were defined and explained by how they affect the sponge city (Table 1). After factor extraction, the overall score of each indicator was first calculated. After this, the weight of each factor was obtained by adding up all the weights of their indicators (Table 2), which was then defined as the ‘weight’ that represented the impact of the factor on sponge cities. The results revealed that urbanization degree and plant adaptability had the greatest impact on sponge city performance, with weights A: 45% and D: 27% (Figure 3). Obviously, many urban roads and buildings are constructed during rapid urbanization, which not only change land use and increase impervious surfaces, but threaten the natural hydrological process and exert great pressure on the ecosystem. According to Strohbach et al. (2019), this was the main factor impacting the conditions of sponge city performance and construction. Green plants are the most significant component of a sponge city because they restore the natural water recycle and ecosystems by increasing rainwater seepage and recharging groundwater (Barbier et al. 2009), which impact the second Factor A, urbanization degree.
Table 1

Rotated factor matrix between 12 initial indicators and four extracted factors

 
 
View Large
Table 2

Overall score and weight of each indicator and factor

 
 
View Large
Figure 3
Impact factor classification and overall score weight calculation. Note: The hexagram shows the 12 indicators selected: (a) number of sponge city studies; (b) shallow soil organic carbon (%); (c) average price of nearby real estate (CNY); (d) soil permeability coefficient (m/d); (e) annual GDP per capita (10,000 CNY); (f) annual mean precipitation (mm); (g) annual average temperature (°C); (h) number of motor vehicles; (i) impervious surface coverage (%); (j) maintenance cost (CNY); (k) construction cost (CNY); and (l) population density (persons/km2). The right-hand side shows the four extracted factors: (A) urbanization degree; (B) financial burden; (C) rainwater characteristics; and (D) plant adaptability. The pie chart shows the overall score weight of each factor.
View largeDownload slide

Impact factor classification and overall score weight calculation. Note: The hexagram shows the 12 indicators selected: (a) number of sponge city studies; (b) shallow soil organic carbon (%); (c) average price of nearby real estate (CNY); (d) soil permeability coefficient (m/d); (e) annual GDP per capita (10,000 CNY); (f) annual mean precipitation (mm); (g) annual average temperature (°C); (h) number of motor vehicles; (i) impervious surface coverage (%); (j) maintenance cost (CNY); (k) construction cost (CNY); and (l) population density (persons/km2). The right-hand side shows the four extracted factors: (A) urbanization degree; (B) financial burden; (C) rainwater characteristics; and (D) plant adaptability. The pie chart shows the overall score weight of each factor.

Figure 3
Impact factor classification and overall score weight calculation. Note: The hexagram shows the 12 indicators selected: (a) number of sponge city studies; (b) shallow soil organic carbon (%); (c) average price of nearby real estate (CNY); (d) soil permeability coefficient (m/d); (e) annual GDP per capita (10,000 CNY); (f) annual mean precipitation (mm); (g) annual average temperature (°C); (h) number of motor vehicles; (i) impervious surface coverage (%); (j) maintenance cost (CNY); (k) construction cost (CNY); and (l) population density (persons/km2). The right-hand side shows the four extracted factors: (A) urbanization degree; (B) financial burden; (C) rainwater characteristics; and (D) plant adaptability. The pie chart shows the overall score weight of each factor.
View largeDownload slide

Impact factor classification and overall score weight calculation. Note: The hexagram shows the 12 indicators selected: (a) number of sponge city studies; (b) shallow soil organic carbon (%); (c) average price of nearby real estate (CNY); (d) soil permeability coefficient (m/d); (e) annual GDP per capita (10,000 CNY); (f) annual mean precipitation (mm); (g) annual average temperature (°C); (h) number of motor vehicles; (i) impervious surface coverage (%); (j) maintenance cost (CNY); (k) construction cost (CNY); and (l) population density (persons/km2). The right-hand side shows the four extracted factors: (A) urbanization degree; (B) financial burden; (C) rainwater characteristics; and (D) plant adaptability. The pie chart shows the overall score weight of each factor.

Close modal

Refined indicators and resilience damage scores

The initial indicators represent the macro-impact of sponge cities in the whole of China and the four factors used to categorize it. However, more detailed indicators should be refined from the four factors together with actual conditions in Tianjin, so that the effect on sponge city performance can be more clearly seen on the urban level. To this end, four indicators (one per factor) were refined to detail the four factors, and their influence on sponge city benefits in the study area was analyzed (Table 3). Hard surfaces represent the degree of urbanization in a district, which damages the natural environment and tends to have a negative effect on rainwater control. Cost has a negative effect on sponge cities from an economic standpoint because the great expense increases financial burden and causes disadvantages in actual planning and design work. Unlike the hard-surface coefficient, the runoff coefficient mainly depends on regional precipitation and land-use type, which reveals the urban ecological structure in another aspect and is closely related to its factors and to the initial indicators. According to the last factor, green space is the foundation of plant growth, and an insufficiently green condition usually represents a bad ecological environment. This provides detrimental conditions for all natural creatures, including plants and micro-ecosystems in the green infrastructure. Based on the theory of ‘simulating the natural conditions and maintaining or restoring ecological hydrological characteristics before the damage of urban development as far as possible’ of LID (USEPA 2000), a resilience damage score was used to represent the ecological resilience corresponding to the relation between each sponge city and its natural environment before development. The data on detailed indicators represent a degree of damage to resilience in each subdistrict divided by the study area (Table 3). This created a link between impact factors on the sponge city, urban ecological resilience, and stormwater management, where a higher score means severe damage and worsening rainwater problems. Thus, applying sponge city strategies would mitigate the disadvantages based on the urban ecological indicators, not only optimizing sponge city adaptation, but also promoting urban ecological resilience and the ability to cope with urban stormwater.

Table 3

Detailed indicators with their factors

Factor (Weight)NameRefined indicator
A (45%) Urbanization degree Hard-surface proportion (%) 
B (18%) Financial burden Total cost 
C (10%) Rainwater characteristic Runoff coefficient 
D (27%) Plant growth environment Inverse green space ratio (1/%) 
Factor (Weight)NameRefined indicator
A (45%) Urbanization degree Hard-surface proportion (%) 
B (18%) Financial burden Total cost 
C (10%) Rainwater characteristic Runoff coefficient 
D (27%) Plant growth environment Inverse green space ratio (1/%) 

Note: Indicators that could increase resilience damage are positive, whereas those that reduce damage are negative. The negative indicators were standardized by inversion so that the score variation trend could be consistent.

View Large

Spatial autocorrelation analysis in ecological resilience damage

The combined resilience damage scores in the 72 subdistricts were calculated from the data of the detailed indicators corresponding to their extracted factors and effect weights. Their spatial autocorrelation was then analyzed using global Moran's I and Getis-Ord general G. The results show that the global Moran's I was 0.37 and the z-score was 6.37, which indicated a positive correlation (I > 0) and a <1% likelihood that this clustered pattern could occur by random chance (z-score >2.58). This, in turn, revealed that the resilience damage distribution is strongly clustered in these subdistricts. Moreover, the Getis-Ord general G test indicated that resilience damage was concentrated in areas of high scores (i.e., existed significantly in the study area), with a G-value of 0.015 > 0 corresponding to a z-score of 5.36 > 2.58, which is consistent with the results of Moran's I.

In this research, the ‘high–high cluster, high–low outlier, not significant, low–high outlier, and low–low cluster’ classes in the results of local Moran's I were used to define the degrees of damage to resilience as ‘significant damage, damage, not significant, ecology, and high ecology’, whereas the Getis-Ord Gi* result classes of ‘hot spot – 99% confidence, hot spot – 95% confidence, hot spot – 90% confidence, not significant, cold spot – 90% confidence, cold spot – 95% confidence, and cold spot – 99% confidence’ were similarly represented by ‘high damage, medium damage, low damage, not significant, low ecology, medium ecology, and high ecology’. A high degree of resilience damage was identified in central districts such as s2–s10 (subdistricts 2–10) and s56, s59, and s62 (Figure 4(a)) using local Moran's I, which was consistent with the results (s2–s10 and s62) of further Getis-Ord analysis (Figure 4(b)). However, the distribution in s21–s30 and s15 indicated that the damage would be significantly low (high ecology) in suburban districts. It is not hard to understand that the central region in a city possesses a denser concentration of concrete buildings, hard-surface roads, and other development construction, with sparse ecological structures such as trees and grasslands. This contributed to the damage effect in the aspects of urbanization degree-hard surfaces and plant adaptability-green space ratio. As the distance from the center increased, it became apparent that existing buildings were fewer and that bare land awaiting construction accounted for more. Bare land could be defined with an ‘intermediate status’ between natural ecology and hard surfaces, which contributes less ecological damage than concrete surfaces, but more than natural spaces, to the factor-detailed indicator in the rainwater characteristics-runoff coefficient. This is the case because its natural soil structure has been damaged by urban construction, but it has not yet been fully changed into an impervious surface because it retains a certain amount of permeability (Li 2018). Therefore, the subdistricts between the city center and the suburbs showed moderate damage in s12, s34, s53, s57, s61, s63, and s65, low ecology in s41–43 with more green space surrounding, and insignificant damage in the others. In conclusion, to resolve the problems of damage to urban ecological resilience based on factor analysis and geographical statistics, the highly damaged districts as well as other districts surrounding them with high scores should all concentrate on implementing LID practices to improve land-use structure and to decrease the negative impact from the detailed indicators.
Figure 4
Urban ecological resilience damage distribution. The number in each subdistrict is its serial number, and a total of 72 subdistricts were classified: (a) distribution results for local Moran's I; (b) distribution results for Getis-Ord Gi*, with hot and cold spot analysis. The colored planes forming the background under the subdistrict boundaries are major buildings.
View largeDownload slide

Urban ecological resilience damage distribution. The number in each subdistrict is its serial number, and a total of 72 subdistricts were classified: (a) distribution results for local Moran's I; (b) distribution results for Getis-Ord Gi*, with hot and cold spot analysis. The colored planes forming the background under the subdistrict boundaries are major buildings.

Figure 4
Urban ecological resilience damage distribution. The number in each subdistrict is its serial number, and a total of 72 subdistricts were classified: (a) distribution results for local Moran's I; (b) distribution results for Getis-Ord Gi*, with hot and cold spot analysis. The colored planes forming the background under the subdistrict boundaries are major buildings.
View largeDownload slide

Urban ecological resilience damage distribution. The number in each subdistrict is its serial number, and a total of 72 subdistricts were classified: (a) distribution results for local Moran's I; (b) distribution results for Getis-Ord Gi*, with hot and cold spot analysis. The colored planes forming the background under the subdistrict boundaries are major buildings.

Close modal

Adaptability evaluation of LID facilities in the optimization scheme

The infiltration capacity of surfaces in Tianjin is only 0.03–0.05 m/d, and the water level and groundwater salinity are respectively 1–3 m and 1,000–3,000 mg/L (Wang et al. 2020), which are not appropriate for large-scale construction of infiltration facilities. However, it could also be inferred that the central regions of districts (Figure 4(b)) contain mainly existing construction, making ground-based retention and storage facilities like herbaceous basins difficult to implement widely. Therefore, the green roof supported by rain barrels (GR, Green roof-Rain barrel) could serve as the primary LID facility for recovering urban ecological structure by collecting stormwater instead of infiltrating most of it to groundwater. Permeable pavements and herbaceous basins could still be used in sponge cities as a supplementary measure to encourage rainwater absorption. Moreover, the area ratio of hard-surfaced roofs in central districts in the study area is commonly 75–85% as calculated by remote sensing analysis, which also indicates that GR is more suitable for central urban districts to restore natural structure and enhance resilience.

Feasibility of green infrastructure and spatial pattern design of sponge city

Based on the ‘Technical Guidelines for Sponge City Construction in Tianjin’ (Tianjin Urban & Rural Development Commission 2016), the runoff coefficient should be ≤0.45 in new and rebuilt urban infrastructure, which could be a minimum standard for recovering and promoting urban ecological resilience. However, sponge city facilities are expensive to build and maintain, which increases the financial burden of this aspect of Factor B. To address this, it should be noted that it is unnecessary to implement new LID green spaces in each subdistrict. Rather, they can be applied to significantly resilience-damaged districts and the others surrounding them, such as s2–s12, s16, s34, s53, s56, s59, s61–63, and s65. Hence, LID facilities with green space are preferred, not only to increase greenbelt ratios, but also to decrease hard-surface ratios. In plan 1, the GR was applied as the LID facility to transform hard surfaces to green spaces, especially in the aspect of roofs in high damage districts. An appropriate green space area ratio has been set in each of the significant districts to reduce its runoff coefficient to at least less than 0.45 but not excessively so, as calculated by the multi-objective formulas (23)–(28). The specific ratios for the addition of GR, ranging from 24.8 to 56.8% in each district, are presented in Figure 5(a). Figure 5(b) shows the ecological damage distribution after implementing plan 1. The hot spots in s2–s12 and other districts with resilience damage have all been transformed to cold spots and insignificant spots, but subdistricts in the northern area, such as s37, s38, and s48–s50, have appeared as new hot spots with high damage. The results indicated that the three mutually restrictive factors A, C, and D and their detailed indicators, ‘Hard-surface ratio’, ‘Runoff coefficient’, and ‘Inverse green space ratio’, have been reduced sufficiently, and the comprehensive damage scores calculated by the three detailed indicators has been decreased to significantly less than in the northern area. The new hot spots in this area appeared because it has a high green space ratio, but also contains a dense industrial zone. This is also a region undergoing urban development, which should concentrate on using more ecological infrastructure in future building construction to reduce resilience damage. Moreover, Factor B (financial burden) and its detailed indicator were not divided by the resilience damage score, but provided a contribution to damage with the account itself. The greatest benefit would be achieved by minimizing cost and finding a balance between cost and LID area ratio increase.
Figure 5
Spatial pattern and area ratio increase of LID and resilience promotion after implementation of sponge city schemes: (a) specific ratio increases and layout of GR in each subdistrict and (b) resilience damage distribution after GR implementation.
View largeDownload slide

Spatial pattern and area ratio increase of LID and resilience promotion after implementation of sponge city schemes: (a) specific ratio increases and layout of GR in each subdistrict and (b) resilience damage distribution after GR implementation.

Figure 5
Spatial pattern and area ratio increase of LID and resilience promotion after implementation of sponge city schemes: (a) specific ratio increases and layout of GR in each subdistrict and (b) resilience damage distribution after GR implementation.
View largeDownload slide

Spatial pattern and area ratio increase of LID and resilience promotion after implementation of sponge city schemes: (a) specific ratio increases and layout of GR in each subdistrict and (b) resilience damage distribution after GR implementation.

Close modal

Optimization in LID spatial pattern and detailed design in facilities

In Palla & Gnecco (2015), a combination of permeable pavements, herbaceous basins, and green roofs performed the best among combined LID facilities, but a reduction of at least 5% in hard surfaces through land-use transformation could contribute visible hydrologic benefits such as recovering urban ecological resilience. Moreover, permeable pavements and herbaceous basins were better added at a greater than 10% ratio in a reformed region with much existing construction and were preferred for application to non-automotive pathways and sidewalks with light traffic (Weiss et al. 2019). The unit prices of permeable pavement and herbaceous basins were 300 and 90 CNY/m2, respectively, which are significantly cheaper than the price of a green roof at 440 CNY/m2. Eventually, 10% of the area was intended for herbaceous basins, with analysis of remote images and current conditions of low infiltration and high water level and salinity in this coastal study area with high urbanization. However, the correlations in the results revealed that the balance between maximum resilience and minimum cost would be reached when the area ratio of permeable pavement was as low as possible and trending to zero. Hence, 10% permeable pavements were finally implemented in addition to the standard in Tianjin.

Figure 6 shows the spatial pattern optimization of LID facilities, with green roofs accounting for 14.8–46.8% of the subdistricts (24.8–56.8% of the total LID area ratio minus 10% of the herbaceous basins), which was calculated according to the multi-objective method. Moreover, herbaceous basins and permeable pavements accounting for 10% of the area were laid out on buildings in each residential area within the ‘rectangles’ formed by the road network (Tianjin Urban & Rural Development Commission 2016), and herbaceous basins were placed at the edges of the sidewalks using the unit of the ecological tree pool. Moreover, permeable bricks were used to replace the old sidewalks constructed with impervious concrete, which enabled the infiltration capacity to approach that of natural rainwater permeation processes. Nevertheless, it remains difficult to apply permeable materials to bicycle lanes because this involves a complex rebuild, including the transformation of surface, base, and sub-crust layers with perforated pipes set in at 2 m intervals (Liu & Lin 2014), requiring extensive work and causing conflicts with daily busy traffic. Therefore, methods of recovering ecological resilience on main roads in central regions with high traffic loads and difficulties in installing LID facilities should be further explored.
Figure 6
Optimal spatial pattern schemes of green infrastructure and their detailed design: (a) layout and structure of general green roofs; (b) schematic of triangular green roofs; and (c) schematic of a permeable sidewalk and a herbaceous basin formed by ecological tree pools.
View largeDownload slide

Optimal spatial pattern schemes of green infrastructure and their detailed design: (a) layout and structure of general green roofs; (b) schematic of triangular green roofs; and (c) schematic of a permeable sidewalk and a herbaceous basin formed by ecological tree pools.

Figure 6
Optimal spatial pattern schemes of green infrastructure and their detailed design: (a) layout and structure of general green roofs; (b) schematic of triangular green roofs; and (c) schematic of a permeable sidewalk and a herbaceous basin formed by ecological tree pools.
View largeDownload slide

Optimal spatial pattern schemes of green infrastructure and their detailed design: (a) layout and structure of general green roofs; (b) schematic of triangular green roofs; and (c) schematic of a permeable sidewalk and a herbaceous basin formed by ecological tree pools.

Close modal

Figure 6 shows the detailed structures of green roofs, herbaceous basins, and permeable pavements, all designed according to relevant research and engineering standards. Green roofs are still in an early stage in China, but have been widely used in Singapore, America, and various European countries (He 2020), which provides a representative experience for construction in highly urbanized cities. Layers of green roof can be divided into plants and base material, infiltration, drainage, protection, water resistance, drainpipes, and structural roofs (Figure 6(a)). In general, flat roofs can accept green roofs directly, but a triangular roof with two steep slopes needs a long groove at the bottom to collect rainwater and convey it to storage in the rain barrel (Figure 6(b)). Similarly, permeable pavement consists of permeable units, a bedding layer, a permeable base course, drainpipes, a permeable sub-base layer, and a soil subgrade. The pipes can help drain excess rainwater coming from the road surface and alleviate the stress of soil infiltration. Moreover, the permeable pavements were set at about 2 m wide to cover the sidewalk completely (Figure 6(c)). A typical herbaceous basin is composed of a storage layer (plant layer), an improved soil layer, the original soil layer, and an overflow well connected to a set of drainpipes. Ecological tree pools are used to break the long grassy expanse into several rectangular units, solving the problem that the herbaceous basin needs to occupy a long continuous stretch of land for storing, retaining, and infiltrating runoff and would be difficult to build in the central districts of cities. The ecological tree pools are designed in 1.2 × 1.2 m squares and are laid out over the permeable pavements (Tan & Zhu 2013). The pit surrounding each tree connects to a vertical rainwater collection pipe that is then linked to drainpipes in the permeable sidewalks. They can not only increase the green space ratio but also help permeable pavements to absorb stormwater. To further cope with the runoff generated by traffic roads, gutter inlets could be built on the side to connect roads and pools (Figure 6(c)), helping to collect road runoff and share the drainage pressure in old municipal gutters.

Limitations and future research directions

This paper has proposed a novel approach to evaluate the impact factors on sponge cities while also optimizing spatial patterns and dealing with damage to urban ecological resilience. An evaluation system for resilience restoration and sponge city benefits based on factor analysis, geographical statistics, remote sensing, and multi-objective optimization has been developed to evaluate the operating performance of various schemes. Moreover, contributions to urban resilience from the geographical environment as well as social and economic factors and the effect of sponge cities were all analyzed using the resilience damage score method, and detailed optimal sponge city plans were also designed and laid out. However, the practical performance of different optimal schemes for water infiltration, runoff control, and rainfall processing was still not addressed explicitly in this research. The benefits in practical cases should also be revealed to make the spatial pattern optimization and facility design more compelling.

In subsequent research, numerical simulation measures will be used to build a model that simulates actual operating conditions, the water retention process in actual storm events, and the layout of green roofs, herbaceous basins, permeable pavements, and underground pipes.

This study established a rapid and simplified framework for assessing the effects of sponge city strategies on urban ecological resilience recovery at a macro level. Various impact factors and their contributions to restoring resilience after damage were systematically integrated into the evaluation framework. Using factor analysis, geographical statistics, remote sensing, and multi-objective processing, maps were produced and used to visualize the benefits of sponge city strategies in the aspects of promoting urban resilience and decreasing stormwater trends, and specific LID facilities were designed for coastal areas. A number of conclusions were reached from this study using impact-benefit analysis:

  • (1)

    Factors A (urbanization degree) and D (plant adaptability) had the greatest impact on sponge city performance, with effect weights of 45 and 27%, respectively. Urban ecological resilience is positively related to distance from the city center because the damage scores of hard-surface ratio, runoff coefficient, and green space ratio, which are the detailed indicators corresponding to factors A, C (rainwater characteristics), and D are higher with more intense urban development.

  • (2)

    Results indicated that the best-performing sponge city spatial pattern scheme in addressing urban ecological resilience damage was the combination scenario of 14.8–46.8% green roofs (area ratio) +10% herbaceous basins +10% permeable pavements. This was the most appropriate application of LID facilities and achieved a balance between facility construction and cost.

  • (3)

    General LID practices that mainly depend on infiltration to underground or groundwater, or that need much land to operate, are not suitable for widespread use in prosperous coastal areas with large hard-surface roofs and high salinity and groundwater levels. Green roofs connected to a groove and supported by rain barrels could fit these conditions, including triangular roofs with a steep slope.

  • (4)

    Specific measures for urban ecological resilience recovery, including green roofs supported by grooves and rain barrels, permeable sidewalks, and herbaceous basins divided into units by ecological tree pools, were adopted in highly urbanized coastal areas. However, further research (such as numerical simulation) should be carried out to determine the operating effect of a sponge city under actual working conditions.

This study was supported by the National Natural Science Foundation of China (No. 41907149), the Tianjin Graduate Research Innovation Project (No. 2022SKY195) and the China Postdoctoral Science Foundation (No. 2018M631732). We thank International Science Editing (http://www.internationalscienceediting.com) for editing this manuscript.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.

Bah
A.
,
Hongbo
Z.
,
Bah
A.
,
Jufang
H.
 &
Zhumei
L.
2023
Study of the applicability of sponge city concepts for flood mitigation based on LID (low impact development) measures: A case study in Conakry City, Republic of Guinea
.
Water Science & Technology
88
(
4
),
901
–
921
.
https://doi.org/10.2166/wst.2023.251
.
Google Scholar
 
Barbier
S.
,
Balandier
P.
 &
Gosselin
F.
2009
Influence of several tree traits on rainfall partitioning in temperate and boreal forests: A review
.
Annals of Forest Science
66
(
6
),
1
–
11
.
https://doi.org/10.1051/forest/2009041
.
Google Scholar
 
Bredes
A.
,
Miller
J. K.
,
Kerr
L.
,
Gannon
K.
 &
Day
I.
2023
Developing guidance for the application of Natural and Nature Based Features (NNBF) on developed shores: A case study from New Jersey, USA
.
Regional Studies in Marine Science
62
,
102959
.
https://doi.org/10.1016/j.rsma.2023.102959
.
Google Scholar
 
Chen
F.
,
Zhang
Q.
,
Chen
S.
 &
Yuan
Y.
2023
Research on operation efficiency prediction and optimization of biological retention system based on GA-BP neural network model
.
Sustainability
15
(
14
),
10990
.
https://doi.org/10.3390/su151410990
.
Google Scholar
 
Cliff
A. D.
 &
Ord
J. K.
1981
Spatial Processes: Models & Applications
.
Taylor & Francis, London, UK
.
Google Scholar
 
Deeb
M.
,
Groffman
P. M.
,
Joyner
J. L.
,
Lozefski
G.
,
Paltseva
A.
,
Lin
B.
,
Mania
K.
,
Cao
D. L.
,
McLaughlin
J.
,
Muth
T.
,
Prithiviraj
B.
,
Kerwin
J.
 &
Chenga
Z.
2018
Soil and microbial properties of green infrastructure stormwater management systems
.
Ecological Engineering
125
,
68
–
75
.
https://doi.org/10.1016/j.ecoleng.2018.10.017
.
Google Scholar
 
Eckart
K.
,
McPhee
Z.
 &
Bolisetti
T.
2017
Performance and implementation of low impact development – A review
.
Science of the Total Environment
607
,
413
–
432
.
https://doi.org/10.1016/j.scitotenv.2017.06.254
.
Google Scholar
 
Ferrans
P.
,
Torres
M. N.
,
Temprano
J.
 &
Sánchez
J. P. R.
2022
Sustainable urban drainage system (SuDS) modeling supporting decision-making: A systematic quantitative review
.
Science of the Total Environment
806
,
150447
.
https://doi.org/10.1016/j.scitotenv.2021.150447
.
Google Scholar
 
He
J.
2020
Discussion on the development status and prospect of green roofs
.
Journal of Zhejiang Agricultural Sciences
61
(
09
),
1810
–
1815
.
https://doi.org/10.16178/j.issn.0528-9017.20200931
.
Google Scholar
 
Ishaq
S.
,
Rana
A.
,
Hewage
K.
,
Li
J.
 &
Sadiq
R.
2023
Environmental impacts and cost analysis of low impact developments in arid climate: A case study of Okanagan Valley (Canada)
.
Journal of Cleaner Production
419
,
138226
.
https://doi.org/10.1016/j.jclepro.2023.138226
.
Google Scholar
 
Jackisch
N.
 &
Weiler
M.
2017
The hydrologic outcome of a low-impact development (LID) site, including superposition with streamflow peaks
.
Urban Water Journal
14
(
2
),
143
–
159
.
https://doi.org/10.1080/1573062X.2015.1080735
.
Google Scholar
 
Jia
H.
,
Yu
S. L.
 &
Qin
H.
2017
Low-impact development and sponge city construction for urban stormwater management
.
Frontiers of Environmental Science & Engineering
11
(
4
),
1
–
2
.
https://doi.org/10.1007/s11783-017-0989-4
.
Google Scholar
 
Johnson
R. A.
 &
Wichern
D. W.
2014
Applied Multivariate Statistical Analysis
, Vol.
6
.
Pearson
,
London
,
UK
.
Google Scholar
 
Kaiser
H. F.
1958
The varimax criterion for analytic rotation in factor analysis
.
Psychometrika
23
(
3
),
187
–
200
.
Google Scholar
 
Kaykhosravi
S.
,
Khan
U. T.
 &
Jadidi
A.
2018
A comprehensive review of low-impact development models for research, conceptual, preliminary and detailed design applications
.
Water
10
(
11
),
1541
.
https://doi.org/10.3390/w10111541
.
Google Scholar
 
Koc
K.
,
Ekmekcioğlu
Ö.
 &
Özger
M.
2021
An integrated framework for the comprehensive evaluation of low-impact development strategies
.
Journal of Environmental Management
294
,
113023
.
https://doi.org/10.1016/j.jenvman.2021.113023
.
Google Scholar
 
Li
D.
2018
Seepage and Structural Evolution Characteristics of Undisturbed Soil Corroded by Landfill Leachate
.
Master's Thesis
,
Wuhan Polytechnic University
.
https://doi.org/10.27776/d.cnki.gwhgy.2018.000284
.
Google Scholar
 
Li
H.
,
Calder
C. A.
 &
Cressie
N.
2007
Beyond Moran's I: testing for spatial dependence based on the spatial autoregressive model
.
Geographical Analysis
39
(
4
),
357
–
375
.
Google Scholar
 
Li
X.
,
Li
J.
,
Fang
X.
,
Gong
Y.
 &
Wang
W.
2016
Case studies of the sponge city program in China
. In
Paper Presented at the World Environmental and Water Resources Congress
.
Google Scholar
 
Li
J.
,
Wu
S.
,
Zhao
X.
,
Xue
X.
 &
Li
M.
2018
Simulation of LID practices in coastal area based on PCSWMM
.
China Water & Wastewater
34
(
21
),
118
–
122
+ 126. https://doi.org/10.19853/j.zgjsps.1000-4602.2018.21.023
.
Google Scholar
 
Liu
Y.
 &
Lin
X.
2014
Comprehensive design application of permeable pavement in pedestrian space: Permeable pavement renewal demonstration of Lujiazui Ring Road as an example
.
Chinese Landscape Architecture
30
(
07
),
87
–
92
.
Google Scholar
 
Ord
J. K.
 &
Getis
A.
1995
Local spatial autocorrelation statistics: Distributional issues and an application
.
Geographical Analysis
27
(
4
),
286
–
306
.
https://doi.org/10.1111/j.1538-4632.1995.tb00912.x
.
Google Scholar
 
Palla
A.
 &
Gnecco
I.
2015
Hydrologic modeling of low-impact development systems at the urban catchment scale
.
Journal of Hydrology
528
,
361
–
368
.
https://doi.org/10.1016/j.jhydrol.2015.06.050
.
Google Scholar
 
Peng
C.
2017
Optimization of Stormwater Drainage System Planning and Design Based on Sponge City Theory
.
Doctor Thesis
,
Tianjin University (CNKI)
.
Google Scholar
 
Pereira
S.
,
Diakakis
M.
,
Deligiannakis
G.
 &
Zêzere
J.
2017
Comparing flood mortality in Portugal and Greece (western and eastern Mediterranean)
.
International Journal of Disaster Risk Reduction
22
,
147
–
157
.
https://doi.org/10.1016/j.ijdrr.2017.03.007
.
Google Scholar
 
Putri
F. K.
,
Hidayah
E.
 &
Ma'ruf
M. F.
2023
Enhancing stormwater management with low impact development (LID): A review of the rain barrel, bioretention, and permeable pavement applicability in Indonesia
.
Water Science and Technology
87 (9), 2345–2361. https://doi.org/10.2166/wst.2023.095
.
Google Scholar
 
She
L.
,
Wei
M.
 &
You
X. Y.
2021
Multi-objective layout optimization for sponge city by annealing algorithm and its environmental benefits analysis
.
Sustainable Cities and Society
66
,
102706
.
https://doi.org/10.1016/j.scs.2021.102706
.
Google Scholar
 
Strohbach
M. W.
,
Döring
A. O.
,
Möck
M.
,
Sedrez
M.
,
Mumm
O.
,
Schneider
A. K.
,
Weber
S.
 &
Schröder
B.
2019
The ‘hidden urbanization’: Trends of impervious surface in low-density housing developments and resulting impacts on the water balance
.
Frontiers in Environmental Science
7
,
29
.
https://doi.org/10.3389/fenvs.2019.00029
.
Google Scholar
 
Tan
H.
 &
Zhu
Q.
2013
The tree pool processing technology application analysis on the ecological landscape construction at the present stage
.
Guangdong Landscape Architecture
35
(
1
),
72
–
76
.
Google Scholar
 
Tianjin Urban and Rural Development Commission
2016
Technical Guidelines for Sponge City Construction in Tianjin [WWW Document]. https://zfcxjs.tj.gov.cn/ztzl_70/bzgf/xxbz/xxbzgf/202010/t20201029_4031124.html (accessed 22 February 2023)
.
USEPA
2000
Low-Impact Development (LID): A Literature Review
.
United States Environmental Protection Agency
,
Washington, DC
.
Wang
H.
,
Wen
W.
 &
Liu
X.
2020
Study on seepage storage capacity of vadose zone in construction of sponge city of Tianjin
.
Site Investigation Science and Technology
2020 (
5
),
6
–
12
.
Google Scholar
 
Weiss
P. T.
,
Kayhanian
M.
,
Gulliver
J. S.
 &
Khazanovich
L.
2019
Permeable pavement in northern North American urban areas: Research review and knowledge gaps
.
International Journal of Pavement Engineering
20
(
2
),
143
–
162
.
https://doi.org/10.1080/10298436.2017.1279482
.
Google Scholar
 
Xu
C.
,
Jia
M.
,
Xu
M.
,
Long
Y.
 &
Jia
H.
2019
Progress on environmental and economic evaluation of low-impact development type of best management practices through a life cycle perspective
.
Journal of Cleaner Production
213
,
1103
–
1114
.
https://doi.org/10.1016/j.jclepro.2018.12.272
.
Google Scholar
 
Yuan
Y.
,
Zhang
Q.
,
Chen
S.
 &
Li
Y.
2022
Evaluation of comprehensive benefits of sponge cities using meta-analysis in different geographical environments in China
.
Science of the Total Environment
836
,
155755
.
https://doi.org/10.1016/j.scitotenv.2022.155755
.
Google Scholar
 
Zhang
H.
,
Chang
Q.
 &
Li
C.
2021
Research on the evaluation of sponge city construction of Beijing from the perspective of the hydro-ecological environment, based on factor analysis and FAHP
.
Journal of Hebei University of Technology (Social Sciences Edition)
37
(
03
),
45
–
51
.
https://doi.org/10.16433/j.cnki.cn41-1379.2021.03.006
.
Google Scholar
 
Zhu
Y.
,
Xu
C.
,
Yin
D.
,
Xu
J.
,
Wu
Y.
 &
Jia
H.
2022
Environmental and economic cost-benefit comparison of sponge city construction in different urban functional regions
.
Journal of the Environmental Management
304
,
114230
.
https://doi.org/10.1016/j.jenvman.2021.114230
.
Google Scholar
 
© 2023 The Authors
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/).