Quantitative Analysis of Sponge City Construction and Function in the Main Urban Area of Chengdu

Abstract

This study utilizes ArcGIS, the InVEST model, and the SCS model to analyze remote sensing data from the central urban areas of Chengdu. The analysis simulates water yield and runoff within the study area while calculating the water conservation capacity for each land use type using the water balance method. This study aims to address the challenges faced by Chengdu in implementing its sponge city initiatives. The results reveal that the spatial distribution of direct runoff generally follows a pattern of “low in the periphery and high in the center”. Transportation, commercial, industrial, and residential land types account for 74.7% of the total surface runoff within the study area, emphasizing their importance in urban rainwater management and sponge city development. Water yield varies across different land use types, with water bodies exhibiting the lowest capacity and artificial land exhibiting the highest capacity. This pattern initially exhibited a downward trend before increasing, with land use type, climatic factors, and vegetation coverage identified as the primary drivers of water yield. The water conservation capacity of the study area gradually decreased, with higher values observed in the east and south and lower values in the north and west. These trends and spatial differences can be attributed to urban expansion and alterations in land cover. Based on these findings, this study assessed the risk of urban waterlogging and provided recommendations for optimizing low-impact development (LID) strategies. This study provides a scientific foundation for the development of sponge city initiatives, urban waterlogging mitigation, and rainwater management strategies in Chengdu.

1. Introduction

With the acceleration of urbanization, urban scales and construction areas have progressively expanded [1,2]. Frequent waterlogging disasters in Chinese cities, alongside the prevalent “sea-watching” phenomenon, stem from inadequate water resource utilization and a deteriorating water ecological environment, severely impacting urban social and economic development and ecological construction [3]. To address the pressing issues of urban water ecology, the “sponge city” construction concept has been introduced and implemented [4]. Nevertheless, urban waterlogging has persisted since the construction of sponge cities. Strengthening control over urban surface runoff and enhancing water retention to alleviate postconstruction waterlogging has emerged as a critical research topic [5].

In the context of sponge city construction, numerous cities have made notable advancements. Among these, Changsha has applied a combination of gray and green infrastructure to regulate stormwater within its basin, aiming to mitigate urban stormwater-related disasters [6]. Shanghai has implemented cutting-edge green infrastructure and optimized its drainage system to efficiently manage stormwater. Wuhan has emphasized ecological restoration and adopted integrated water management strategies to bolster its flood resilience [7]. In contrast, Chengdu, a major city in southwestern China, faces significant urban waterlogging during the summer months due to heavy rainfall, dense urban development, and fragmented green spaces.

The construction of sponge cities emphasizes the analysis of urban water conservation functions, which encompass the regulation of runoff, purification of water quality, and storage of natural water resources [8,9]. Scholars worldwide have conducted extensive research on this topic. International research has primarily concentrated on models for calculating water yield, including the SWAT model [10] and InVEST model [11]. In China, the emphasis is on water conservation within simulated research areas, utilizing methods like the water balance approach [12]. Existing studies indicate that the water balance method effectively calculates regional water conservation, which is defined by the differential between precipitation, water yield, runoff, and other water consumption. The water yield module of the InVEST model calculates the difference between precipitation and evaporation, thereby determining the water yield. To calculate urban surface runoff, numerous hydrological simulation models have been developed globally for analyzing surface runoff, including the SUSTAIN model [13] and the SCS-CN model [14]. Notably, the SCS-CN model, which is part of the runoff curve models developed by the United States Department of Agriculture’s Soil Conservation Service in 1945, stands out as one of the most widely adopted models for runoff generation [15]. This model utilizes climatic rainfall, terrain, and surface data to calculate and simulate urban surface runoff during storms and examines the causes of urban waterlogging based on these data. The model is applicable to urban water resource management, surface runoff analysis, and various hydrological research endeavors [16,17].

In the context of sponge city construction, the selection of models is crucial for accurately assessing water conservation and runoff characteristics. The InVEST model is particularly well-suited for evaluating ecosystem services, such as water yield and runoff regulation, because it integrates multiple environmental factors and provides spatially explicit results. Furthermore, the SCS-CN model is particularly well-suited for urban environments due to its simplicity and effectiveness in simulating runoff based on land use and soil properties. The combination of these models facilitates a comprehensive analysis of urban water conservation and runoff dynamics, thus offering a robust framework for sponge city planning and management.

Despite extensive research on sponge city construction in natural ecological areas and cities in Southeast China, a knowledge gap persists regarding the analysis of urban water conservation and optimization of sponge construction across different land types in Chengdu. Current studies primarily focus on simulating urban rainstorms and optimizing urban drainage network structures. However, few studies have investigated the application of multiple models to analyze urban water conservation and optimize sponge construction in the unique urban environment of Chengdu. This study aims to analyze the rainwater storage capacity and runoff characteristics of various land types using the InVEST and SCS-CN models. It seeks to simulate the spatial distribution of urban runoff and water accumulation caused by summer rainfall in the study area over the past three years, assess the risk of urban waterlogging, optimize the functional effectiveness of sponge city initiatives, and provide a scientific foundation for sponge city planning and urban waterlogging control in Chengdu.

2. Study Area

Chengdu, the capital of Sichuan Province, is located at geographic coordinates ranging from 102°54′ E to104°53′ E and 30°05′ N to 31°26′ N. It is located in the southwestern part of Sichuan Province and is an important central city in China. The terrain is inclined from northwest to southeast, and the altitude is mostly between 1000 and 3000 m. Chengdu belongs to the subtropical monsoon climate zone, with abundant rainfall; the four seasons are distinct, the annual average temperature is 15.7~18.0 °C, the total annual precipitation is 734.8~1142.3 mm, and the rainfall is large, mainly concentrated from June to September, often causing urban waterlogging and water accumulation problems. This study focuses on the sponge city construction area in the main urban area of Chengdu, including Qingyang District, Jinjiang District, Chenghua District, Wuhou District and Jinniu District, with a total area of 463.3 km², as shown in Figure 1. In the main urban area of Chengdu, the rainfall concentration is high, building land patches easily form surface runoff, the vegetation structure area is relatively simple, and the water source storage is average. Therefore, it is of great significance to analyze the construction effect of sponge cities in the main urban area of Chengdu to optimize the urban water ecological environment.

3. Materials and Methods

3.1. Data Source

The basic data of this study includes the geospatial data of the main urban area of Chengdu in the three phases of 2021, 2022, and 2023. The original data are derived from remote sensing images of Landsat series data from China Geospatial Data (http://www.gscloud.cn/, accessed on 26 June 2024) with a spatial resolution of 30 m. The meteorological data were obtained from the National Earth System Science Data Center, and the rainfall data is based on the monthly rainfall data of the main urban area of Chengdu in 2021, 2022, and 2023 from the National Environmental Information Center (NCEI) under the National Oceanic and Atmospheric Administration (NOAA). This study is mainly based on land use type data, which comes from the 2020 10 m resolution land cover data of Professor Zhong Yanfei’s team at Wuhan University and the “Mapping of China’s Basic Urban Land Use Type” published by researchers from Tsinghua University in the journal Science Bulletin in December 2019. The 2018 Preliminary Results paper published a dataset of nationwide land-scale urban land use maps for 2018 [18].

In the simulation process of the SCS-CN model, the required data on urban spatial administrative divisions were obtained from the official website of Tiandetu, which is affiliated with the National Basic Geographic Information Center. Soil texture data were obtained from the Chinese soil dataset in the World Soil Database (Harmonized World Soil Database version 1.2; HWSD, Hamburg, Germany). Month-by-month soil moisture data were obtained from the National Tibetan Plateau Data Center.

3.2. Research Methods

3.2.1. Water Conservation Simulations

In this study, water conservation in the study area was simulated and calculated on the basis of the regional water balance equation. The water balance method conceptualizes the study area as a “black box”, considering atmospheric precipitation as the input and evaporation and runoff as the water outputs [19]. Water conservation in the study area was derived from data on precipitation, evaporation, and surface runoff.

$$W=P-ET-Ra$$

(1)

formula, W is the water conservation quantity (mm); P is the precipitation (mm); Ra is the surface runoff (mm); ET is the evapotranspiration (mm).

Subsequently, the InVEST model was used to calculate the average water yield (P-ET) within the study area, and the SCS-CN model was used to compute the surface runoff (Ra) for the study area. Furthermore, the raster calculator in ArcGIS 10.8 was used to superimpose the water conservation quantities across the study area.

3.2.2. InVEST Model

Model Principle

The water yield module within the InVEST model employs a water balance-based estimation approach, utilizing the difference between precipitation and actual evaporation to determine water yield [20], encompassing parameters such as surface runoff, soil water content and root depth. The primary algorithm underlying the model is as follows:

$$Yx=Px-AETx$$

(2)

$${\displaystyle \frac{AETx}{Px}}={\displaystyle \frac{1+\omega xRx}{1+\omega xRx+\frac{1}{Rx}}}$$

(3)

where, Y_x is the annual water yield of grid cell x (mm); AET_x represents the actual evapotranspiration of grid cell x; P_x is the annual average precipitation of grid cell x (mm); Rx is the dry index; ω_x is the ratio of the corrected vegetation annual available water to precipitation.

Data Sources and Processing

The necessary layers for the operation of the water yield module comprise precipitation (P/mm), potential evapotranspiration (E/mm), root restriction layer data (RRLD), plant available water content (PAWC/%), study area boundary, land cover/land use map, and sub-basin boundary; and the table parameters include the Z coefficient and biophysical parameters (

Table 1. Main data required for water conservation in the main urban area of Chengdu.

Date	Source and Treatment
Precipitation	National Earth System Science Data Center (http://www.geodata.cn, accessed on 18 July 2024)
Potential evaporation	National Qinghai-Tibet Plateau Science Data Center (https://data.tpdc.ac.cn/, accessed on 6 October 2024)
Land Cover Classification Data	2020 10 m resolution land cover data from Professor Zhong Yanfei’s team at Wuhan University (https://data.tpdc.ac.cn/, accessed on 22 August 2024)
Root Restriction Layer Data	1 km Chinese Soil Depth Map (https://www.nature.com/articles/s41597-019-0345-6, accessed on 22 August 2024)
Plant Available Water Content	ISRIC Global Dataset
Sub-watershed Data	Based on data from the Resource and Environment Science Data Platform (https://www.resdc.cn/, accessed on 8 October 2024) and generated sub-watersheds through GIS hydrological analysis tools
Biophysical Parameters Table	The coefficients of land use types in the biophysical table are obtained from literature [21] and parameters recommended by the InVEST model

).

3.2.3. Principles of the SCS-CN Model

The SCS model, established by American scientists, is based on the statistical analysis of precipitation-runoff data across various basins and is primarily utilized for calculating runoff in small watersheds. It holistically accounts for the interrelationships among soil type, regional rainfall, land use practices, antecedent soil moisture conditions, and surface runoff [22].

Based on the water balance equation:

$$P=Ia+F+Q$$

(4)

The rainfall-runoff relationship is satisfied as follows:

$${\displaystyle \frac{Q}{P-I_a}}={\displaystyle \frac{F}{S}}$$

(5)

where: P represents total rainfall (mm); Ia represents the initial loss value of rainfall (mm); F represents the cumulative infiltration during the runoff, i.e., the actual infiltration (mm); Q represents surface runoff (mm); S represents the potential water storage capacity (mm).

Empirical evidence and experimental data indicate that the initial loss value constitutes a portion of the soil’s potential maximum retention, and it is widely accepted that a linear relationship exists between Ia and S. The corresponding formula is as follows:

$$Ia=\lambda S$$

(6)

where λ represents the initial loss coefficient, a dimensionless parameter that encapsulates the collective conditions of regional hydrology and soil. This correlates with the geographical and climatic factors of the study area [23], and the value range for λ is 0.1 ≤ λ ≤ 0.3. In the existing SCS-CN model, for computational simplicity, λ is commonly assumed as 0.2. Consequently, the runoff formula for the SCS model can be derived as follows [24]:

$$\left\{\begin{array}{c}Q={\displaystyle \frac{{\left(P-Ia\right)}^2}{P-Ia+S}}, \left(P\ge 0.2S\right)\\ Q=0, \left(P<0.2S\right)\end{array}\right.$$

(7)

Since the amplitude of S is large, a dimensionless parameter—the runoff curve coefficient (CN value, denoted as CN in the formula), is introduced. In the SCS-CN model, it is a comprehensive parameter that reflects the characteristics of the basin’s underlying surface and rainfall, and the empirical conversion expression between related parameters S and CN value is as follows:

$$S=254\left({\displaystyle \frac{100}{CN}}-1\right)$$

(8)

where S denotes the maximum retention capacity or potential water storage capacity of the soil, and CN is a dimensionless parameter that encapsulates a comprehensive index of land use, soil type, and antecedent soil moisture conditions [25]. Theoretically, the value range is from 0 to 100 [26], and the specific value is optimized and adjusted based on the particular conditions of the study area. Given that the focus of this study is the main urban area of Chengdu, which features diverse land use types, the CN value is accordingly refined based on land use types and the presence of parks and green spaces within the study area. The SCS-CN model categorizes soil into types A, B, C, D, etc., based on variations in soil permeability and water conductivity, as delineated in

Table 2. Classification criteria for hydrological soil groups in the SCS model.

Soil Type	Soil Properties	Minimum Infiltration Rate/(mm·h−1)
A	Sandy soil, loamy sand, sandy loam	>7.26
B	Silt loam, loam	3.81~7.26
C	Sandy clay loam, silty clay loam, shallow sandy loam	1.27~3.81
D	Sandy clay, clay	0~1.27

3.2.4. Runoff Calculation Based on SCS-CN Model

To calculate runoff in the study area, the initial steps involve classifying land use types and adjusting the CN coefficient. The classification and nomenclature of urban land use types in the United States differ from those of Chinese standards, and similar discrepancies exist in the categorization and naming of the CN values. Considering the planned distribution of land use types in Chengdu’s main urban area and the local climatic conditions, this study aligned the land use nomenclature from the “Soil and Water Conservation Service Curve Number (SCS-CN) Method” with the specific urban land use types in the study area and adjusted the CN values accordingly [27]. The preprecipitation index API, as proposed by the Soil and Water Conservation Service of the United States Department of Agriculture, was utilized to delineate the antecedent soil moisture conditions [26]. This index relies primarily on the cumulative rainfall recorded over the 5 days preceding rainfall events. Based on the average rainfall within the study area, soil moisture was categorized into three levels: drought (AMCⅠ I), normal (AMC II), and humid (AMC III) [28]. By utilizing the normal soil wetness condition from the antecedent stage (AMC II) as a baseline, the CN value was determined in conjunction with the type of urban land use. The specific values are listed in

Table 3. CN value reference table for the main urban area of Chengdu.

Land Use Type	Soil Hydrology Group
	A	B	C	D
Industrial Land	81	88	91	93
Residential Land	77	85	90	92
Commercial Land	89	92	94	95
Public Utility Management Land	83	88	92	94
Water Area	98	98	98	98
Transportation Land	98	98	98	98
Park Green Land	39	61	74	80
Forest Land	36	60	73	79
Grass Land	52	70	80	84
Cultivated Land	66	75	82	85

Note: A, B, C, and D refer to soil types.

.

As documented in the Annals of Chengdu, the city’s soil types are diverse, predominantly consisting of sandy loam, loam, and yellow-brown soil, all of which are categorized under Class C; hence, the CN value for Group C is selected. By utilizing the SCS model’s calculation formula, the CN value is integrated into the model to determine the potential water storage capacity S and the surface runoff quantity Q. Following the computation of each parameter’s numerical values, GIS technology was employed for land vectorization processing, aligning with the spatial distribution data of land use types within the study area. This process yields a classified map of vectorized land planning, which aids in the GIS-assisted simulation of spatial runoff. Statistical analysis revealed 2956 distinct catchment areas across 10 land types within the study area, as depicted in Figure 2.

As summer and autumn are the primary seasons for rainfall in Chengdu, the average monthly rainfall from June to September for the years 2021, 2022, and 2023 are 198.0 mm, 137.9 mm, and 173.5 mm, respectively, representing the total monthly average rainfall for each respective year. The urban surface runoff in the main urban area of Chengdu post-rainfall was both calculated and simulated, and the results are presented in

Table 4. Calculation parameters of each parameter for different land use types under the SCS hydrological model.

Land Use Type	CN	S/mm	$\mathit{I}\mathit{a}$/mm	Q2021/mm	Q2022/mm	Q2023/mm
Industrial Land	91	25.1	5.0	170.8	111.8	146.6
Residential land	90	28.2	5.6	167.8	109.0	143.8
Commercial Land	94	16.2	3.2	179.8	120.2	155.5
Public Facilities Management Land	92	22.1	4.4	173.8	114.5	149.6
Water Area	98	5.2	1.0	191.9	131.9	167.4
Transportation Land	98	5.2	1.0	191.9	131.9	167.4
Park Green Land	74	89.2	17.8	120.5	68.9	99.0
Forest Land	73	93.9	18.8	117.6	66.6	96.3
Grass Land	80	63.5	12.7	138.0	83.1	115.3
Cultivated Land	82	55.8	11.2	143.8	88.0	120.8

.

The 10 land use categories in the CN value calculation table were cross-referenced with those in the land use planning map, and the area and number of plots for each land use category were meticulously tallied. The SCS model was subsequently applied to analyze direct runoff across different land use types. Utilizing the spatial analysis capabilities of ArcGIS and employing the natural breakpoint method, the data from Table 4 were imported, and GIS was leveraged to simulate the regional spatial distribution of direct runoff for different land use types over the past three years. The outcomes are depicted in Figure 3.

3.3. Statistical Analysis

In this investigation, statistical analyses were conducted using SPSS software (version 27), and a one-way analysis of variance (ANOVA) was employed to assess significant differences in the water conservation function and mean values across various land use types for 2021, 2022, and 2023. Prior to examining significant differences, Levene’s test was used to evaluate the homogeneity of variance, followed by the least significant difference (LSD) test for multiple comparisons, with significance set at p < 0.05. Lastly, data visualization and mapping were performed using Origin2021 software.

4. Results and Discussion

4.1. Surface Runoff Analysis of Different Land Use Types

The observations in Table 4 and Figure 3 reveal that surface runoff in the study area has exhibited a progressively increasing trend over the past three years. Considering the regional soil and rainfall conditions in Chengdu’s main urban area, compared with the SCS-CN model, the direct runoff ratios between different land use types in the main urban area of Chengdu over the past three years can be more intuitively determined. These ratios are primarily influenced by rainfall patterns and land use differences. To assess the spatial distribution of total surface runoff across various land types, the regional analysis capabilities of ArcGIS 10.8 software were employed for comprehensive statistical analysis and area calculations of patches for each land type. Surface runoff data from the past three years were used to determine the direct runoff volumes in the study area, with the total runoff for each land type calculated in accordance with the SCS-CN model. The data are presented in

Table 5. Water conservation of different land types in the main urban area of Chengdu over the past three years.

Land Use Type	Average Water Conservation Depth/mm			Standard Deviation	Proportion of Water Conservation/%
	2021	2022	2023		2021	2022	2023
Industrial Land	10.48	10.03	9.73	1.83	4.57	4.81	4.44
Residential Land	12.06	10.77	10.23	3.05	5.26	5.17	4.67
Commercial Land	1.72	1.38	1.16	2.67	0.75	0.66	0.53
Public Facilities Management Land	7.57	6.05	6.93	2.52	3.30	2.91	3.16
Water Area	0.54	0.27	0.49	9.60	0.24	0.13	0.22
Transportation Land	0.46	0.61	0.73	8.52	0.20	0.29	0.33
Park Green Land	52.50	46.79	49.26	9.37	22.90	22.45	22.48
Forest Land	61.69	53.27	57.32	4.32	26.91	25.55	26.15
Grass Land	45.04	42.07	44.47	6.58	19.65	20.18	20.29
Cultivated Land	37.19	37.22	38.86	6.84	16.22	17.85	17.73
Statistics	229.25	208.46	219.18	5.53	100.0	100.0	100.0

.

As indicated in

Table A1. Statistical table of land use types and runoff in the main urban area of Chengdu.

Land Use Type	Number of Plaques	Area/km2	Q/mm	Standard Deviation	Total Runoff/m3	Runoff Percent Ofevery Land Use Type/%
Industrial Land	156	48.8	143.1	24.26	7 × 106	10.73
Residential land	1086	193.3	140.2	24.14	2.7 × 107	41.65
Commercial Land	291	42.0	151.8	24.47	6.4 × 106	9.81
Public Facilities Management Land	302	36.5	146.0	24.34	5.3 × 106	8.19
Water Area	348	12.3	163.7	24.63	2 × 106	3.09
Transportation Land	32	49.9	163.7	24.63	8.2 × 106	12.55
Park Green Land	106	5.3	96.1	21.16	5.1 × 105	0.78
Forest Land	580	8.4	93.5	20.91	7.8 × 105	1.21
Grass Land	22	9.0	112.1	22.52	1 × 106	1.55
Cultivated Land	33	57.8	117.5	22.90	6.8 × 106	10.44
total	2956	463.3	1327.7	233.96	6.5 × 107	100.00

, residential land contributes the most total runoff within the study area, constituting 41.65% of the overall runoff, followed by traffic land, which accounts for 12.55% of the total runoff, with industrial and cultivated lands contributing 10.73% and 10.44%, respectively. GIS technology was employed to import the data from Table 5 to simulate the spatial distribution of direct runoff, with the outcomes depicted in Figure A1.

Integrating the direct runoff statistics from Figure A1 and Table A1, it is evident that traffic land, water areas, and commercial land in Chengdu’s main urban area are particularly susceptible to rainwater accumulation. The overall spatial distribution of the study area exhibited a pattern of “low around and high in the middle”. According to the surface runoff calculations in Table A1, traffic land in Chengdu’s main urban area has the highest direct runoff, with 163.7 mm of surface runoff and a total runoff volume of 8.2 × 10⁶ m³. This indicates that the land cover in this area has inadequate water absorption and drainage capabilities, primarily due to its impervious composition, suggesting the need to enhance the water pipe network infrastructure.

The analysis of the total runoff data in Table A1 reveals that residential land generates the highest surface runoff within the study area, attributed to its having the highest number of patches and the largest total area, specifically 1086 patches and 193.3 km². Residential land contributes a total runoff volume of 2.7 × 10⁷ m³. This indicates that the impervious surfaces of residential areas possess a low infiltration capacity for rainfall; therefore, it is necessary to augment the drainage pipe network in residential areas and bolster the sponge function of these areas through the incorporation of ecological rainwater infrastructure. Additionally, industrial land accounts for a total runoff of 7 × 10⁶ m³, while commercial land contributes 6.4 × 10⁶ m³, suggesting that the impervious pavements in industrial and commercial areas within the main urban area exhibit low permeability to rainwater. Surface runoff from traffic, commercial, industrial, and residential land types constitute 74.74% of the total runoff within the study area, highlighting its significance in urban rainwater management and sponge city construction.

4.2. Sponge Function Analysis

Sponge function analysis refers to the evaluation of the hydrological responses of urban and natural ecosystems to rainfall events using simulation and computational methods. This analysis primarily focuses on the ability of urban areas to infiltrate, store, and reuse rainwater. In this study, we assessed the water conservation functions of various land types within urban areas by analyzing the water yield and runoff data. Based on these findings, we propose measures for sponge city construction aimed at mitigating urban runoff and enhancing the infiltration and storage of rainwater.

4.2.1. Water Yield and Direct Surface Runoff

The model calculations showed that the average runoff in the study area over the past three years was 163.01, 103.52, and 139.44 mm, with the highest values reaching 191.9, 131.9, and 167.4 mm, and the lowest values being 117.6, 66.6, and 96.3 mm, respectively. Surface runoff in the study area is primarily influenced by precipitation and land use type. A comparison with three-year data reveals that surface runoff in the main urban area of Chengdu initially decreased, subsequently increased, and ultimately decreased again. The analysis indicates that changes in rainfall patterns over the past three years have significantly impacted surface runoff, with frequent and intense rainfall events exacerbating the runoff. Secondly, under similar precipitation conditions, surface runoff is predominantly affected by land cover type. The continuous expansion of urban areas has led to an increase in impervious surfaces (e.g., roads, roofs, and sidewalks). These surfaces reduce water infiltration into the soil, and the transition from natural land cover (e.g., grasslands and forests) to urban built-up areas diminishes the land’s water conservation capacity. Spatially (Figure 3), surface runoff increases from the periphery to the center, with the highest runoff volumes observed in the central and southern parts of the study area. To analyze the reasons for the spatial differences in runoff, the natural slope of urban topography influences runoff accumulation and concentration. Compared with sandy soil, loam soil in Chengdu’s urban areas exhibits lower permeability. Areas with dense vegetation cover exhibit higher permeability, thereby resulting in lower runoff. Changes in land use, including urban planning and human activities, affect soil permeability and its capacity to absorb and retain water, thereby promoting the formation of surface runoff. These findings are consistent with the analyses of runoff influencing factors by other scholars [29,30]. Under the same high-intensity rainfall conditions, there are significant differences in runoff generation and surface runoff among different land use types. In contrast, under low precipitation conditions, urban surface runoff is significantly reduced. These results indicate that the three-year runoff variation in the study area is influenced by land use type. Compared to the low density of green space patches, the large area and high density of built-up land patches tend to lead to greater runoff, a pattern that aligns with the results of previous studies [31,32].

Based on the statistical analysis from the third phase, the water yields for Chengdu’s main urban area in 2021, 2022, and 2023 are calculated to be approximately 1.14 × 10⁶ mm, 7.63 × 10⁵ mm, and 9.77 × 10⁵ mm, respectively. The corresponding average water yields for the study area for these years are 181.22 mm, 121.35 mm, and 155.37 mm, respectively. From 2021 to 2023, the total water yield initially decreased, then increased, and ultimately exhibited a decreasing trend, with a total reduction of 1.63 × 10⁵ mm. Analysis of annual precipitation data indicates that water yield is influenced by precipitation intensity to a significant degree.

Spatially, the water supply in the study area exhibited a distribution pattern characterized by higher values in the central regions and lower values at the periphery, as illustrated in Figure 4. Temporally, the three-year water yield distribution was analyzed using the natural breakpoint method. The overall water yield exhibited an initial increase, followed by a decrease. Specifically, the water yield from residential, commercial, and other built-up lands gradually increased, whereas the water yield from land types with high vegetation cover, such as woodlands, grasslands, park green spaces, and cultivated lands, remained relatively stable or exhibited a decreasing trend. The spatial distribution characteristics and trends are closely related to climatic factors, such as precipitation and evapotranspiration. High precipitation intensity years tend to result in higher water yields, while low precipitation leads to lower yields. However, high evapotranspiration reduces water yield, and the evapotranspiration rate is generally higher in densely vegetated areas, contributing to spatial variation. The three-year variation in water yield within the study area is predominantly influenced by rainfall and evapotranspiration, with land use types having a comparatively minor effect, a finding that aligns with the conclusions drawn from other scholars’ analyses and research on water yield disparities [33,34]. Huang Xin et al.’s research [35] revealed that climate factors account for 76.9% of the water yield in Yunnan Province, whereas land use types account for only 22.8%. The water balance method highlights that precipitation and actual evapotranspiration are pivotal in determining water yield [36,37]. High-value water yield areas are concentrated in built-up lands and grasslands, likely influenced by urban planning, increased impervious surfaces, and spatial topography. Most built-up land consists of impervious surfaces, leading to poor precipitation infiltration and the formation of runoff that is channeled into the drainage system. Additionally, plants in grasslands have shallow root systems, resulting in low water availability, reduced surface evaporation and increased runoff. Low-value water yield areas are concentrated in woodlands, cultivated lands, parks, and green spaces. Vegetation in these areas mainly consists of shrubs and trees with deep root systems, which enhance water availability and surface evapotranspiration. Consequently, runoff is reduced, resulting in a lower water yield. The stability of the water yield in these areas suggests that vegetation cover and soil compaction significantly influence water retention.

Through an in-depth analysis of the characteristics and interactions of land use and rainfall changes, this study not only reveals the challenges faced by the main urban area of Chengdu in the construction of sponge cities but also provides a scientific basis for future urban planning and water resource management. Firstly, although changes in land use types are minimal, significant differences in runoff generation and soil and water conservation capacity are observed among different land use types. For instance, a higher proportion of impervious surfaces in built-up areas (such as transportation, commercial, and residential land) results in greater runoff generation, thereby increasing the risk of urban flooding. Additionally, significant changes in vegetation coverage have been observed in the study area, with stronger soil and water conservation capacities in parks, green spaces, forests and grasslands. This finding indicates that even with relatively stable land use types, optimizing land use layout, enhancing vegetation coverage, and increasing green infrastructure can significantly improve the sponge function of cities. Secondly, although the rainfall data were primarily characterized by temporal differences, changes in rainfall intensity and frequency significantly affected soil and water conservation capabilities. During the study period, interannual fluctuations in rainfall intensity led to significant changes in runoff generation, particularly in areas with concentrated construction land. This result highlights the significant regulatory role of rainfall characteristics in urban hydrological processes and suggests that urban planners should fully consider the uncertainty of rainfall changes during sponge city construction. The study found that land use changes have a significant threshold effect on the regulation of ecosystem services. When vegetation coverage reaches a certain proportion, its effects on enhancing soil permeability and regulating runoff are the most significant. This finding provides important theoretical support for sponge city construction and offers specific guidance for urban planners regarding land use planning and green infrastructure layout. Future research should further expand the temporal and spatial scales to more comprehensively understand the mechanisms of ecosystem service changes under urbanization, thereby providing stronger support for sustainable urban development.

However, further analysis is needed to elucidate the underlying causes of these trends and spatial differences. The spatial patterns of water yield and runoff are influenced by factors such as land use, soil characteristics and topography. Meanwhile, differences in conservation potential across regions can be attributed to variations in vegetation cover, soil permeability, and human activities. Understanding these factors will enhance our comprehension of the mechanisms driving the observed spatial patterns and inform the development of more effective strategies for water resource management and conservation. However, this study only analyzed the water yield between different land use types in the past 3 years, and land use change was only manifested as a small increase in the impervious area. Therefore, the impact of land use change on water yield over time remains a key area for future research.

4.2.2. Analysis of Water Conservation

Based on the differential analysis of water conservation in Chengdu’s main urban area, the average water conservation depths for the study area in 2021, 2022, and 2023 are 21.63 mm, 18.76 mm, and 19.20 mm, respectively. The total water conservation measures 2.02 × 10⁷ mm. The water conservation for different land types is calculated based on water production and runoff data, and the results are presented in Table 5.

To validate the mean water conservation capacity over the past three years, a significance analysis was performed on the mean conservation capacity across different land use types for each year, as depicted in Figure A2. The land use types were categorized and numbered from 1 to 10, and the results revealed significant differences in the mean water conservation capacity among different land use types within the same year (p < 0.05). Over the past three years, a significant difference was observed between high-function woodlands and other land types (p < 0.05), whereas no significant difference was found between traffic land and water areas, which had the lowest functions.

Owing to the significant variations in vegetation distribution and land cover types across the study area, forestlands presented the greatest water conservation depths, measuring 61.69 mm, 53.27 mm, and 57.32 mm, accounting for 26.91%, 25.55%, and 26.15% of the total water conservation, respectively. Variations were observed in the spatial distribution of water conservation depths across different periods, with a general trend of higher values in the east and lower values in the west, as illustrated in Figure 5. The highest water conservation values were predominantly found in woodlands, parks, green spaces, and grasslands. Lower values were primarily concentrated in transportation areas, water bodies and commercial lands.

Synthesizing the data presented in Figure 5 and Table 5, the water conservation function over the past three years exhibited a trend of initial decline followed by an increase. The analysis suggests that land use planning initially led to a decline in water conservation capacity, while subsequent improvements in urban green infrastructure and land management practices positively impacted water conservation. Notably, water conservation for industrial, residential, and commercial land types progressively diminished, whereas water conservation in public facilities, transportation, park green spaces, forests, and grasslands exhibited a biphasic trend of an initial decrease followed by an increase. Similarly, water conservation in water bodies and cultivated land demonstrated a steady, upward trend. These variations are ascribed to the diversity in vegetation coverage and land use types and are also influenced by climatic factors [38], which is consistent with the findings of this study [39].

Considering the spatial distribution within the study area, the water conservation capacity is higher in the eastern region than in the western region because of the higher vegetation coverage in the northeast and southeast regions, which are predominantly forests and green lands. These areas feature dense vegetation and diverse plant species that enhance water conservation capacity. The deep root systems of plants in these regions increase soil permeability and water retention, particularly during heavy rainfall events. In contrast, the western and central regions are characterized by built-up land and water bodies with substantial water yields. Land use changes in these regions have led to increased impervious surfaces and reduced water conservation capacity. In terms of the overall water conservation ratio, the water conservation capacity of each land type is generally stable. However, the water conservation capacities of cultivated and forestlands significantly varies.

By examining the water conservation depths, it is evident that the ranking of different land types in the study area, in terms of water conservation capacity, is as follows: forestland > park green space > grassland > cultivated land > residential land > industrial land > public facility land > commercial land >, and water bodies > transportation land. This indicates that forestland, grassland, and park green spaces possess the strongest water conservation capacities, whereas commercial land exhibits the weakest water conservation capacity. Analyzing the reasons:

Forestland and park green spaces are predominantly covered by trees and shrubs, creating multilayered community structures. The canopy and ground litter of these plant communities strongly influence precipitation, thereby reducing surface runoff. However, grasslands have shallow root systems and cultivated crops have uniform spacing, resulting in a lower community level and coverage area than those of other vegetation types. This makes them more prone to generating runoff and susceptible to water resource loss, thus rendering their water conservation capacity relatively weak and their overall conservation effect suboptimal [40].

The water conservation capacity of built-up land is relatively low. Notably, commercial, transportation, and public facility lands exhibit a high degree of imperviousness, poor water permeability, substantial runoff, and limited rainfall infiltration. Additionally, these land types have low vegetation coverage, reduced plant evapotranspiration, and high water yields, resulting in a diminished water conservation capacity. In contrast, residential and industrial lands have more comprehensive underground pipe networks and greater vegetation coverage, leading to comparatively better water conservation capacity. For water bodies, high runoff and strong evapotranspiration capacity result in a low water yield and a relatively weak water conservation capacity.

The research findings hold significant importance for the theory and practice of sponge city construction. They reveal significant differences in water production and runoff among various land use types, providing scientific evidence for urban planners, particularly in optimizing land use layouts and increasing green infrastructure. The multi-tiered vegetation structures of woodlands and parks exert a significant interception effect on precipitation, thereby reducing the surface runoff. However, the water retention capacity of commercial land is compromised by its substantial impervious surface area, which supports the findings of this study. The increase in impervious surfaces during urbanization significantly reduces water retention capacity [41,42]. However, the study’s temporal scope, limited to the months of June to September over the past three years, does not fully capture the annual water conservation dynamics in the study area, thereby introducing seasonal biases into the analysis of the sponge city function. Future research should expand the temporal and spatial scales to better understand the change mechanism of ecosystem services in the context of urbanization, so as to establish a more scientific and reasonable basis for the construction of sponge cities [29].

4.3. Urban Waterlogging Risk and Suggestions for Improvement of LID Measures

Rapid urbanization in Chengdu has led to significant population growth and urban expansion, thereby exacerbating the flooding problem. The increase in impervious surfaces during urbanization reduces rainwater infiltration and enhances the surface runoff. In the face of heavy rainfall, the existing drainage system has insufficient carrying capacity. Utilizing the current planning data of the study area, GIS technology was employed to assess urban waterlogging risk by integrating elevation, rainfall, urban road pipe networks, river networks, and additional relevant data, as depicted in Figure 6. Considering urban land types and urban green space planning, the appropriate LID (low-impact development) technology should be selected to alleviate the problems of urban water accumulation, purification, and utilization [43,44]. Figure 6 illustrates that the spatial distribution of urban waterlogging risk gradients decreases from north to south, influenced predominantly by elevation and rainfall within the study area. The analysis of land use types indicates that the risk of waterlogging is higher in built-up land compared to natural green spaces like forests. The risk of urban waterlogging is primarily influenced by elevation, precipitation, and vegetation cover, which is consistent with the findings of previous studies [45,46]. Nevertheless, further analysis of the fundamental factors influencing urban waterlogging risk and its spatial distribution is essential. This analysis will facilitate a deeper exploration of the sponge city construction effect in the study area and provide a rational basis for optimizing LID facilities. Within the study area, impervious paving is predominantly utilized for built-up land, while permeable paving is employed for forests, grass, and green spaces. Nevertheless, the waterlogging issue cannot be effectively resolved solely through natural regulation and storage mechanisms. Consequently, LID facilities, including green roofs, permeable pavements, bioretention systems, and rain gardens, can be strategically implemented to enhance sponge city construction in the study area. Additionally, their combined use can be tailored based on specific land use contexts [47].

Residential land is characteristic of high density urban areas. For this type of land, common LID facilities primarily consist of green roofs and permeable pavements [48]. The residential land area in Chengdu’s main urban area is extensive and spatially concentrated, resulting in high surface runoff, low water conservation capacity, and a propensity for waterlogging. LID facility designs for residential areas can incorporate green roofs to collect and manage rooftop rainwater [49]. Bioretention facilities and vegetated swales are designed to enhance rainwater infiltration and recycling. Road runoff can be managed using permeable pavements. Green roofs have a high capacity for rainwater control, making them suitable for areas with high rainfall concentrations and intensities. When combined with permeable pavement, green roofs can maximize the control and optimization of summer rainstorms in Chengdu. Additionally, vegetation can be planted to increase surface roughness, deepen soil layers, enhance the water storage and purification capabilities of LID facilities, and reduce the burden on underground drainage networks. This approach helps to mitigate waterlogging in urban residential areas [50].

Transportation, commercial, industrial, and other land types are predominantly hard-paved, resulting in high surface runoff. During low rainfall events, water infiltrates the soil or drains into the pipeline; however, as the rainfall intensity increases, runoff increases significantly despite minimal improvement in permeability. Consequently, LID facilities, including permeable pavements and sunken green spaces, can effectively mitigate peak runoff, and infiltration wells within green spaces can increase the capture and utilization of rainwater [51]. Overall, various LID facility combinations can significantly reduce runoff from built-up land during summer rainstorms and augment rainwater harvesting, thereby improving water resource utilization.

For land types characterized by soft pavements, including woodlands, grasslands, and park green spaces, the LID facility system is centered on green spaces and can incorporate various LID facility combinations, such as rain gardens, porous pavements, and shallow vegetation ditches [52]. Grass ditches are situated primarily along the main garden roads within park green spaces, allowing rainwater to naturally infiltrate or be directed into pipelines [53]. Forests and grasslands possess strong water conservation abilities, increasing soil layer thickness and vegetation structure, bolstering plant interception and rainwater infiltration capabilities, and sustaining and enhancing the overall water conservation effect. The prevalence of construction land in Chengdu’s central urban area can be optimized by enhancing the fragmentation of construction land parcels and strategically expanding green spaces to strengthen urban water conservation capabilities. Park green spaces can be designed with concave features to create larger green expanses, thereby reducing flat green areas and enhancing rainwater capture efficacy [54]. Furthermore, the selection of plant configurations within sponge facilities is pivotal, necessitating the choice of plants that exhibit robust environmental adaptability and water resilience [55]. Employing diverse plant combinations can enrich plant structure, thereby reducing runoff and purifying water quality, ultimately enhancing the urban water ecological environment. Water bodies are optimized primarily to address water quality issues, with water purification achieved through ecological bank protection and the establishment of water circulation systems, thereby forming a comprehensive water ecosystem [56]. Grass and stone revetments are predominantly utilized along the coastlines of water bodies to increase their self-purification capacity and landscape aesthetics.

5. Conclusions

Water conservation within the study area was determined using the water balance method, further employing GIS for spatial simulation and zoning statistics to directly assess the water conservation capacity across various land uses in the main urban area. The conclusions drawn from this analysis are as follows:

It is noteworthy that significant differences in water yield and runoff were observed among different land use types in the main urban areas of Chengdu. In terms of water yield, the land use types ranged from weakest to strongest as follows: water bodies, forest land, park green spaces, arable land, grassland, and construction land. From 2021 to 2023, regional runoff first decreased and then increased, ultimately showing a downward trend. When ranking regional runoff from high to low, the regional distribution was characterized by water bodies, construction land, arable land, grassland, park green spaces, and forest land. These findings indicate that both climatic factors and land use types significantly influence water yield and the runoff.

Based on the water yield and runoff calculations, the study area’s water conservation capacity from 2021 to 2023 displays a spatial distribution with higher values in the eastern region and lower values in the western region. Ranked by water conservation capacity in descending order, the land use types are as follows: forestland, park green space, grassland, cultivated land, residential land, industrial land, public facility land, commercial land, traffic land, and water area. Over the three-year period, the overall water conservation function demonstrated a biphasic trend, characterized by an initial decrease, followed by an increase. Specifically, this pattern was observed for public facility land, transportation land, park green space, grassland, and forestland. Conversely, the water conservation capacity of cultivated land and water areas exhibited a steady increase, in contrast to residential land, industrial land, and commercial land, which experienced a consistent decline. Comparisons with findings from other scholars indicate significant spatial variations in water conservation capacities among different land use types.

To enhance the functional effectiveness of sponge city initiatives across different land types, the risk of urban waterlogging in the study area was analyzed. The results indicated that the spatial distribution of waterlogging risk gradually decreased from north to south. Natural land types exhibit superior water conservation capacity, thereby reducing the risk of urban waterlogging. However, building sites characterized by impervious surfaces and low vegetation cover exhibit a higher risk of waterlogging than other sites. Therefore, it is proposed to construct sponge cities primarily using a suitable combination of low-impact development (LID) facilities. For building land and other impervious surfaces, the main LID facilities should include permeable pavements, grass swales, sunken green spaces, and bioretention facilities. For natural land such as forests and grasslands, rain gardens and permeable pavements should be primarily utilized.