Impacts of Urbanization on the Spatio-Temporal Patterns of Trade-Offs and Synergies Among Climate-Related Ecosystem Services

Abstract

Under the context of rapid urbanization and climate change, urban ecosystem services (ES) have undergone dramatic transformations. Elucidating the trade-off and synergy relationships among ES and quantifying how urbanization mediates these relationships are critical to achieving urban sustainability. Focusing on Shanghai during 2000–2020, we quantified three climate-related ES—water yield (WY), urban cooling (Heat Mitigation Index, HMI) and carbon storage (CS)—with the InVEST model. We then examined the spatio-temporal evolution of these services, analyzed their trade-offs and synergies, and examined the underlying urbanization drivers. Results show that total WY increased by 76%, with peak volumes concentrated in the central districts; HMI declined, with low-value zones spreading inward; CS rose and became spatially more homogeneous. WY–HMI trade-offs intensified, whereas CS–HMI were synergistic (r = 0.33–0.61) except in core districts where built-up expansion created trade-offs. CS–WY trade-offs weakened, becoming synergistic in most districts by 2020. HMI loss was driven by GDP and industrial output (p < 0.05). Per-capita green-space area was positively correlated with HMI but exerted no significant influence on CS or WY, highlighting the limitations of ecological interventions focused on single ES enhancement.

1. Introduction

Since the Industrial Revolution, cities have emerged as hotspots for human economic activities and population growth. Among the global population of 8 billion, over 55% reside in urban areas [1,2]. Anthropogenic alterations to the land surface driven by human activities within cities have shaped unique urban climates and resource demands, modifying hydrological and material cycles in urban regions [3]. Besides carbon dioxide emissions from urban activities, land-use changes associated with urbanization also promote the release of greenhouse gases such as methane (CH₄) and nitrous oxide (N₂O) from soils, further exacerbating global warming [4]. Climate change, induced by rising global average temperatures, has garnered worldwide attention and represents a common challenge confronting the entire world [5]. Against this backdrop of climate change, the increasing frequency of extreme climatic events (e.g., heatwaves, intense precipitation) elevates risks faced by cities [6,7]. The continued advancement of urbanization leads to greater populations being exposed to such extreme climates, further intensifying climate risks [8,9,10].

Ecosystem services (ES) refer to all benefits humans derive from ecosystems, encompassing provisioning services (e.g., water, food), regulating services (e.g., flood and disease control), cultural services (e.g., recreation, cultural values), and supporting services (e.g., maintenance of life-supporting environments) [11]. Urban climate-related ecosystem services play a crucial role in helping cities cope with the risks and challenges posed by climate change and urbanization.

Existing research demonstrates that rapid urban development exerts profound impacts on global ecosystem services, with land-use change—primarily driven by urbanization—significantly influencing these services [12,13]. The impact of land-use change on ES is diverse and heterogeneous; it may contribute to overall ES increases in some contexts [14]. However, further studies indicate that urbanization more commonly leads to regional ecological degradation and a reduction in ecosystem service provision on a global scale [15,16]. Different ES and the trade-offs and synergies among them respond dissimilarly to land-use change, and spatio-temporal variations in land use significantly affect future ES provision [17]. Urban spatial heterogeneity and rapid development induce notable shifts in ES trade-offs and synergies. Concurrently, interventions like urban renewal and ecological restoration projects can enhance urban ES; research shows that ES trade-offs and synergies are influenced by land-use management policies [18,19]. Therefore, investigating urban ES trade-offs and synergies and their relationship with urbanization processes is key to formulating effective land-use policies and achieving sustainable urban development.

Current research on urban ES trade-offs and synergies primarily focuses on static relationships among ES stocks [20,21,22], often overlooking the relationships between changes during dynamic development processes. For instance, ecosystem services exhibiting an overall trade-off relationship might display synergistic changes within specific spatial and temporal scales—a dynamic relationship that has not been adequately explored in existing studies. Some research indicates that shifts in land use and policy can trigger transformations in ES trade-offs and synergies [23], highlighting the significant influence of changes in driving ES dynamics. This research selects three key climate-related ES—water yield (WY), urban cooling Effect (represented by the Heat Mitigation Index, HMI), and carbon storage (CS)—under the dual pressures of climate change and urbanization. It constructs indicators to analyze the impact of urbanization on ES changes and their dynamic trade-offs and synergies from 2000 to 2020. Furthermore, Principal Component Regression (PCR) is employed to dissect the driving effects of urbanization on these ES changes.

2. Materials and Methods

2.1. Research Area

Shanghai is situated in East China on the west coast of the Pacific Ocean, along the eastern edge of the Eurasian continent. It lies at the midpoint of China’s north-south coastline, at the confluence where the Yangtze River and Qiantang River meet the sea. Shanghai constitutes part of the Yangtze River Delta alluvial plain. Its geographical coordinates span 120°52′ to 122°12′ east longitude and 30°40′ to 31°53′ north latitude. The city is bounded by the Yangtze River to the north, the East China Sea to the east, and Hangzhou Bay to the south and borders Jiangsu and Zhejiang provinces to the west. Shanghai is one of the most populous and economically dynamic metropolises in China. Between 2000 and 2020, Shanghai’s resident population grew from 16.4 million to 24.9 million, and its gross regional product rose from 0.48 to 3.87 trillion CNY. The sixteen districts can be broadly grouped into three functional areas: the metropolitan core (Huangpu, Jing’an, Xuhui, Changning, Hongkou, Putuo, Yangpu), characterized by the densest urban fabric and commercial functions; the emerging development areas (Pudong, Minhang, Baoshan, Jiading), which have experienced the most rapid urban expansion since 2000; and the peri-urban zone (Songjiang, Qingpu, Fengxian, Jinshan and the island district of Chongming), where ecological land and agriculture still predominate.

The average elevation of Shanghai is 2 m. Dajinshan Island represents the highest point within Shanghai, with an elevation of 103.7 m. The city features a humid subtropical monsoon climate. Its river network is primarily formed by the Huangpu River, which flows through the urban center, and its tributaries: Suzhou River (Wusong River), Chuanyang River, and Dianpu River, among others. The study area is shown in Figure 1.

2.2. Data Sources

The primary datasets employed in this study include land-use data, elevation data, soil data, meteorological data, and vegetation data, among others. Parameters required for the calculation of ecosystem services were determined in accordance with the InVEST User’s Guide (https://storage.googleapis.com/releases.naturalcapitalproject.org/invest-userguide/latest/en/index.html, accessed on 10 March 2025). Detailed information regarding the specific data and their sources is provided in

Table 1. Data and sources used in the study.

Data.	Resolution	Time	Data Sources
Land use and land cover	30 m	2000, 2010, 2020	GlobeLand30 (https://www.globallandcover.com/, accessed on 8 March 2025)
DEM	12.5 m	\	EARTHDATA (https://search.asf.alaska.edu/, accessed on 8 March 2025)
Soil	1 Km	\	Harmonized World Soil Database (HWSD) (https://www.fao.org/soils-portal/soil-survey/soil-maps-and-databases/harmonized-world-soil-database-v12/en/, accessed on 8 March 2025)
Meteorological data (temperature, humidity, sunlight, precipitation)	\	2000, 2010, 2020	Resource and Environment Science and Data Center (RESDC) (https://www.resdc.cn/, accessed on 8 March 2025)
Economic and social statistics data	\	2000, 2010, 2020	Shanghai Statistical Yearbook 2000, 2010, and 2020 (https://tjj.sh.gov.cn/, accessed on 8 March 2025)
Carbon density		2010 s	Public Dataset [24]

.

Land-use patterns in Shanghai in 2000, 2010, and 2020 are illustrated in Figure 2.

2.3. Ecosystem Services (ES) Assessment

Under climate change, the capacity of ecosystems to regulate water resources and mitigate the urban heat island effect is particularly crucial for Shanghai. Carbon storage also stands as one of the ecosystem services of greatest concern to policymakers. This study employs the InVEST model to quantify these three ecosystem services.

2.3.1. Water Yield

Water yield (WY) serves as a critical indicator of an ecosystem’s response to water resources, reflecting its capacity to regulate and sustain hydrological functions. This metric quantifies the ecosystem’s carrying capacity for water by measuring the net volume of water generated within a given spatial unit over an annual cycle.

WY is calculated using the Annual Water Yield module within the InVEST model.

$$Y_j=\left(1-{\displaystyle \frac{{AET}_j}{P_j}}\right)\times P_j$$

(1)

where $Y_j$ is the water yield (mm) of pixel $j$, ${AET}_j$ is the actual evapotranspiration (mm) of pixel $j$, and $P_j$ is the precipitation (mm) of pixel $j$.

2.3.2. Urban Cooling

The urban cooling potential is quantified through an integrated assessment of three biophysical parameters, (1) Canopy Shading Index, derived from the proportional coverage of arboreal vegetation (canopy height > 2 m) within each land use category; (2) Evapotranspiration Index (ETI), representing a normalized measure of latent heat flux relative to potential evapotranspiration; and (3) Surface Albedo, a dimensionless coefficient [0, 1] characterizing solar reflectance efficiency of specific land use/land cover (LULC) types.

$$ETI={\displaystyle \frac{K_c\times {ET}_0}{{ET}_{max}}}$$

(2)

where $ETI$ is evapotranspiration index, $K_c$ is crop coefficient, ${ET}_0$ is potential evapotranspiration, and ${ET}_{max}$ is the maximum value of potential evapotranspiration.

In the InVest model, three factors are used to calculate the cooling capacity (CC):

$${CC}_i=0.6\times Shade+0.2\times albedo+0.2\times ETI$$

(3)

where ${CC}_i$ is the cooling capacity of pixel $i$, Shade is shading index (0~1), and albedo is between 0 to 1.

The InVEST model quantifies the radiative influence of green patches exceeding 2 hectares (large green spaces) on adjacent areas. For pixels beyond the influence radius of any large green space, the Heat Mitigation Index (HMI) is equated to the CC of the local pixel. Conversely, pixels within the influence zone are assigned a distance-weighted composite index integrating the CC values of both proximal large green spaces and the target pixel.

$${GA}_i={cell}_{area}\times \sum _{j\in dradiusfromi}{g}_i$$

(4)

$${CC}_{{greenspace}_i}=\sum _{j\in dradiusfromi}{g}_i\times {CC}_j\times e^{\left({\displaystyle \frac{-d\left(i,j\right)}{d_{cool}}}\right)}$$

(5)

where ${GA}_i$ is the area of green spaces within a search distance $d_{cool}$ around each pixel, $d_{cool}$ is the distance over a green area that has a cooling effect (450 m in this study, which is set according to the INVEST User Guide), $d\left(i,j\right)$ is the distance from pixels $i$ to $j$, ${cell}_{area}$ is the area of a cell in ha, $g_i$ equals 1 when the pixel $j$ is green area or 0 if it is not, and ${CC}_{{greenspace}_i}$ is the distance weighted average of the CC values attributable to green spaces.

The HMI of pixel $i$ is calculated as follows:

$${HMI}_i=\left\{\begin{array}{c}{CC}_{i}if{CC}_i\ge {CC}_{{park}_i}or{GA}_i<2ha\\ {CC}_{{greenspace}_i}otherwise\end{array}\right\}$$

(6)

2.3.3. Carbon Storage

Given the practical constraints in acquiring carbon density data for dead plant matter—compounded by the routine clearance of senesced vegetation in urban green spaces—this study quantifies urban carbon storage through three core reservoirs: aboveground biomass carbon, belowground biomass carbon, and soil organic carbon.

$$C_{total}=C_{above}+C_{below}+C_{soil}$$

(7)

where $C_{total}$ is the total carbon storage, $C_{above}$ is aboveground biomass carbon, $C_{below}$ is belowground biomass carbon, and $C_{soil}$ is soil organic carbon.

In this study, carbon-density values for each land-use type were assigned according to publicly available datasets [24], and the specific values are provided in

Table 2. Carbon density of different land use types (t/ha).

Land Use Type	${\mathit{C}}_{\mathit{a}\mathit{b}\mathit{o}\mathit{v}\mathit{e}}$	${\mathit{C}}_{\mathit{b}\mathit{e}\mathit{l}\mathit{o}\mathit{w}}$	${\mathit{C}}_{\mathit{s}\mathit{o}\mathit{i}\mathit{l}}$
Cultivated Land	5.00	1.00	25.60
Forest	47.80	9.94	120.80
Grassland	0.25	1.11	18.20
Shrubland	9.30	2.00	25.60
Wetland	1.00	0.00	33.00
Water Bodies	0.00	0.00	0.00
Artificial Surfaces	0.00	0.00	25.30
Ocean	0.00	0.00	0.00

.

This study did not include the carbon storage of oceans and water bodies because carbon in water bodies and oceans is primarily stored as dissolved organic carbon and sediment organic carbon. These carbon pools are highly dynamic and largely controlled by hydrodynamic processes rather than municipal land-use decisions. The objective of this study is to inform urban land-use management; we restricted carbon storage calculations to terrestrial and wetland ecosystems whose extent and management fall within the direct purview of city planners. This exclusion may lead to a slight underestimation of total urban carbon stocks

2.4. Tradeoff and Synergy Analysis

The interrelationships between ecosystem services (ES) were determined quantitatively using the following function. Tradeoffs represent negative correlations, indicating competing resource demands, and synergies states represent positive correlations, reflecting mutual enhancement.

$$f\left(X\right)=\left\{\begin{array}{c}Tradeoff,ifX\le 0\\ Synergy,ifX>0\end{array}\right.$$

(8)

$$X={∆ES}_i\times {∆ES}_j$$

(9)

where $f\left(X\right)$ is the tradeoff-synergy discriminant function, ${∆ES}_i$ represents the variation magnitude of the $i-th$ ecosystem service, and ${∆ES}_j$ denotes the variation magnitude of the $j-th$ ecosystem service.

Trade-offs are recorded when one service increases while the other decreases; conversely, when both services decline simultaneously, the product ${∆ES}_i\times {∆ES}_j$ remains positive, and the pair is classified as a negative synergy rather than a trade-off.

2.5. Principal Component Regression (PCR)

This study utilized Principal Component Regression (PCR) to investigate the driving effects of urbanization on these ES changes. The influence of urbanization on changes in climate-related ecosystem services was analyzed by quantifying urban development through six indicators reflecting urban construction, economic activity, and population metrics. The six indicators are per capita public expenditure (X₁), per capita green space area (X₂), per capita GDP (X₃), per capita industrial output (X₄), per capita completed housing area (X₅), and population density (X₆).

Per-capita public expenditure and per-capita completed housing area quantify urban construction intensity; per-capita GDP and per-capita industrial output capture economic vitality; per-capita green-space area reflects municipal ecological investment; and population density measures demographic pressure [25,26].

To eliminate scale effects, all variables are standardized to achieve a zero mean and unit standard deviation.

$$X_{scaled}={\displaystyle \frac{X-\mu }{\sigma }}$$

(10)

Here, $X$ denotes the original variable, $X_{scaled}$ represents the standardized variable, $\mu$ is the feature mean, and $\sigma$ is the standard deviation.

The covariance matrix of the standardized data is then computed as follows:

$$\Sigma ={\displaystyle \frac{1}{n-1}}{X}_{scaled}^T{X}_{scaled}$$

(11)

where $\mathsf{\Sigma }$ denote the correlation matrix of variables.

The correlation matrix $\mathsf{\Sigma }$ is decomposed as follows:

$$\Sigma =V\wedge V^T$$

(12)

where $\wedge$ is the diagonal matrix of eigenvalues, and $V$ is the orthogonal eigenvector matrix.

Select the top k eigenvectors corresponding to the largest eigenvalues to form the projection matrix $P_k$.

$$P_k=\left[V_1,V_2···,V_k\right]$$

(13)

The dimensionally reduced data $Z$ is then

$$Z=X_{scaled}$$

(14)

where the columns of $Z$ are mutually orthogonal.

The PCR model regresses the standardized response variable on the principal components:

$$Y_{scaled}=Z\beta +ϵ$$

(15)

where $Y_{scaled}$ is the standardized dependent variable, $\beta$ denotes the regression coefficient vector of principal components, and $ϵ$ is the intercept term.

The spatial mapping in this study is conducted on the ArcGIS 10.8 platform with kernel density estimation and spatial interpolation, and the statistical analysis is performed with R 4.3.1.

3. Results

3.1. Changes in Ecosystem Services

3.1.1. Water Yield (WY)

In 2000, water yield (WY) rose along a clear southwest–northeast transect, peaking in the Huangpu-Jing’an commercial core and the adjacent Huangpu River corridor; the lowest values were recorded on the northern Baoshan and southern Fengxian peri-urban fringe. By 2010, high-yield zones remained anchored in these central districts, whereas declines appeared in the newly industrializing western belt of Jiading and the exurban polders of Chongming Island. Between 2010 and 2020, WY contracted again in the intensively redeveloped core of Huangpu and Xuhui, yet continued to climb in the rapidly urbanizing Pudong New Area (Figure 3).

Across the two decades, cultivated land in Qingpu and Songjiang supplied most of the initial WY, but artificial surfaces in Pudong and Minhang became the dominant source after 2010, adding 1171.5 Mt, while cropland contributions fell by 193.6 Mt (

Table 3. Ecosystem services across different land use types.

LULC	2000			2010			2020
	Carbon Storage (Mt)	HMI	Water Yield (Mt)	Carbon Storage (Mt)	HMI	Water Yield (Mt)	Carbon Storage (Mt)	HMI	Water Yield (Mt)
Cultivated Land	12.23	0.81	2430.89	10.05	0.76	1980.15	8.52	0.76	2818.09
Artificial Surfaces	4.08	0.41	1646.12	6.14	0.33	2459.83	7.03	0.33	3989.08
Water Bodies	0.00	0.35	144.20	0.00	0.36	57.22	0.00	0.36	359.15
Ocean	0.00	0.27	66.83	0.00	0.26	58.71	0.00	0.26	189.98
Wetland	0.39	0.27	51.28	0.63	0.26	73.66	0.63	0.26	141.66
Grassland	0.04	0.65	10.32	0.05	0.66	11.82	0.25	0.66	102.23
Forest	0.04	0.59	0.66	0.21	0.85	2.81	1.98	0.85	62.88
Shrubland	\	\	\	0.00	0.68	0.00	0.00	0.68	0.77
Total	16.78	\	4350.30	17.07	\	4644.21	18.40	\	7663.84

).

3.1.2. Heat Mitigation Index (HMI)

In 2000, the strongest urban-cooling capacity (HMI) was found in the leafy western districts of Songjiang and Qingpu, the northern peri-urban forests of Baoshan, and Chongming Island’s extensive wetlands, whereas the densely built Huangpu–Jing’an core already registered the city’s lowest HMI values.

By 2010, the low-cooling zone had pushed south-westwards into Minhang and north-eastwards along the Yangtze River corridor in Pudong, yet small pockets of high HMI persisted on Chongming’s eastern shoreline. During 2010–2020, the sharpest losses occurred in the rapidly redeveloped waterfront of Yangpu and the new industrial parks in eastern Pudong; conversely, isolated gains appeared around Dianshan Lake in western Qingpu.

Across the two decades, cultivated land on the city fringe and artificial surfaces in the metro core experienced the largest cooling declines, while Chongming’s wetlands and the Huangpu River itself were the only land-cover types to register modest HMI improvements (Figure 4, Table 3).

3.1.3. Carbon Storage (CS)

Between 2000 and 2010, city-wide carbon storage (CS) climbed from 16.78 Mt to 17.07 Mt, reaching 18.40 Mt by 2020 (Figure 5). The gains were geographically concentrated: the largest increments were recorded in the newly afforested hills of western Songjiang, the restored wetlands along Chongming’s eastern shoreline, and the peri-urban green belts of Qingpu and Jiading. Conversely, traditional polders in northern Baoshan and southern Fengxian lost 0.25 Mt as cropland was converted to industrial parks. Although artificial surfaces in central Huangpu and expanding Pudong still supplied the bulk of CS in 2020, their share dropped from 97% (2000) to 84% (2020), while the proportion held by ecological land—forest, wetland, and grassland—rose sharply, reflecting Shanghai’s large-scale afforestation programs on its urban fringe (Table 3).

Notably, the disparity between the high-carbon peri-urban belt and the low-carbon metropolitan core narrowed markedly between 2000 and 2020. Limited further urban expansion in the core allowed small-scale green interventions—street trees, pocket parks, and rooftop vegetation—to raise local stocks, while planned new towns in the outer districts incorporated extensive greenways and suburban forests, maintaining high carbon densities despite land-use conversion.

3.2. Spatio-Temporal Characteristics of Ecosystem Services

From 2000 to 2020, significant changes occurred in the spatio-temporal patterns of climate-related ecosystem services in Shanghai. Kernel density analysis revealed that the distribution of CS shifted from a bimodal to a trimodal pattern (Figure 6). Since the peak near zero corresponds to areas excluded from CS calculation, this study focuses on the peaks within the intermediate-value range. During 2000–2020, the frequency of high CS values decreased, while the frequency of relatively low values also declined. From 2010 to 2020, the inter-peak distance in the intermediate-value range narrowed, indicating reduced spatial heterogeneity in CS distribution across Shanghai.

For the HMI, kernel density distributions in low-value areas remained similar, but the frequency of high HMI values decreased significantly and exhibited a persistent downward trend over time. WY displayed a multimodal distribution, with both the magnitude and frequency of high WY values increasing from 2000 to 2020.

All climate-related ecosystem services in Shanghai showed significant correlations, with dynamics evolving between 2000 and 2020 (as shown in Figure 6):

2000: All three services were positively correlated. The strongest correlation was between WY and CS (r = 0.71), while the weakest was between WY and HMI (r = 0.09).

2010: CS maintained positive correlations with HMI (r = 0.53) and WY (r = 0.61). However, the relationship between WY and HMI shifted from positive to negative (r = −0.11).

2020: CS remained positively correlated with HMI (r = 0.33) and WY (r = 0.28), while WY and HMI continued a negative trend (r = −0.19)

3.3. Tradeoffs and Synergies of Ecosystem Services

3.3.1. At the 1 Km Grid Scale

This study quantified tradeoff-synergy relationships among three ecosystem services at both the 1 km grid scale and the urban district scale. At the 1 km grid scale, WY and HMI primarily exhibited tradeoffs. In 2000, synergistic areas were mainly located in Chongming, Baoshan, Jing’an, Yangpu, and Putuo districts, with minor portions in Pudong, Jinshan, Fengxian, and Qingpu. By 2010, synergistic areas decreased significantly, particularly in Chongming and central urban districts (though other central districts resembled 2000 patterns). In 2020, synergistic areas expanded markedly compared to 2010, especially in central urban zones, exceeding both 2000 and 2010 levels (Figure 7).

CS and HMI primarily showed synergies, with tradeoffs concentrated in central urban areas. From 2000 to 2010, synergistic areas increased substantially in peripheral districts but minimally in central districts. Between 2010 and 2020, tradeoff areas expanded in peripheral districts, forming a ring around the central urban zone (Figure 8).

CS and WY mainly displayed tradeoffs. From 2000 to 2020, spatial patterns of tradeoffs and synergies remained largely similar. Synergistic areas outside Chongming District were concentrated in Qingpu, with minor clusters along the coasts of Pudong, Jinshan, and Fengxian. Chongming shifted from predominantly synergistic (2000) to significantly increased tradeoffs (2010–2020) (Figure 9).

3.3.2. At the Urban District Scale

At the urban district scale, WY and HMI in Shanghai’s major districts primarily exhibited tradeoffs. In 2000, only Baoshan District showed synergies. By 2010, synergies were confined to Chongming District. In 2020, synergistic districts increased to six: Chongming, Baoshan, and central urban districts including Jing’an, Yangpu, Hongkou, and Xuhui (Figure 10).

For CS and HMI, tradeoffs and synergies displayed distinct spatial patterns. In 2000, an east-west divergence emerged: tradeoffs concentrated in eastern districts (Huangpu, Hongkou, and Pudong). By 2010–2020, a north-south contrast prevailed. Synergies in 2010 covered Chongming, Jiading, Jinshan, and Fengxian; by 2020, synergies expanded to include Baoshan, Jing’an, Yangpu, Hongkou, and Xuhui, while Jiading shifted to tradeoffs (Figure 11).

Tradeoff-synergy dynamics between CS and WY shifted significantly from 2000 to 2020. In 2000, tradeoffs dominated, with synergies only in Baoshan, Jing’an, Yangpu, Pudong, and Fengxian. By 2010–2020, synergies prevailed across most districts. In 2010, tradeoffs persisted only in Jiading, Jinshan, and Fengxian; by 2020, tradeoffs were limited to Huangpu, Jinshan, and Fengxian, with all other districts synergistic (Figure 12).

3.4. Principal Component Regression (PCR) Results

The influence of urbanization on changes in climate-related ecosystem services was further analyzed by quantifying urban development through six indicators reflecting urban construction, economic activity, and population metrics. The six indicators are per capita public expenditure, per capita green space area, per capita GDP, per capita industrial output, per capita completed housing area, and population density, denoted as X₁–X₆. Due to administrative boundary changes and data availability constraints, the period 2010–2020 was selected to analyze urbanization’s impact on climate-related ecosystem service dynamics.

To mitigate multicollinearity effects among indicators, Principal Component Analysis (PCA) was applied to X₁–X₆, with results presented in

Table 4. PCA results.

	PC1	PC2	PC3	PC4	PC5	PC6
X1	0.53	0.03	−0.11	0.15	−0.75	0.32
X2	0.39	−0.59	−0.12	0.12	0.003	−0.68
X3	0.31	0.54	0.42	−0.45	−0.099	−0.45
X4	0.38	−0.44	0.46	−0.35	0.30	0.46
X5	−0.42	−0.34	−0.13	−0.68	−0.46	−0.05
X6	0.35	0.18	−0.74	−0.40	0.32	0.11
Standard deviation	1.58	1.19	0.93	0.76	0.65	0.43
Proportion of Variance	0.41	0.23	0.14	0.097	0.071	0.03
Cumulative Proportion	0.41	0.65	0.80	0.90	0.97	1.00

. PCA revealed that PC₁ to PC₄ collectively explained 90% of the total variance (cumulative variance = 0.90). Consequently, these four components were used in subsequent regression analyses.

PCR results are shown in

Table 5. PCR results.

CS2020_2010
R2 = 0.214
	Estimate	Std. Error	t_value	Pr(>|t|)
(Intercept)	0.0000	0.2709	0.0000	1.0000
PC1	−0.2219	0.1773	−1.2518	0.2391
PC2	0.2205	0.2354	0.9368	0.3709
PC3	−0.1550	0.2987	−0.5188	0.6152
PC4	−0.0328	0.3665	−0.0894	0.9305
WY_2020_2010
R2 = 0.342
	Estimate	Std. Error	t_value	Pr(>|t|)
(Intercept)	0.0000	0.2478	0.0000	1.0000
PC1	0.1673	0.1622	1.0317	0.3265
PC2	−0.0371	0.2154	−0.1721	0.8668
PC3	0.4133	0.2733	1.5122	0.1614
PC4	0.4520	0.3354	1.3477	0.2075
HM_2020_2010
R2 = 0.652
	Estimate	Std. Error	t_value	Pr(>|t|)
(Intercept)	0.0000	0.1801	0.0000	1.0000
PC1	−0.3463	0.1179	−2.9382	0.0148
PC2	0.4465	0.1565	2.8525	0.0172
PC3	0.2757	0.1986	1.3881	0.1952
PC4	−0.0666	0.2437	−0.2733	0.7902

. For CS, PC₁, PC₃, and PC₄ exhibited negative trends, while PC₂ showed a positive trend. None were statistically significant. For WY, PC₂ demonstrated a negative trend, with other components showing positive trends. All results were non-significant. For HMI, PC₁ had a significantly positive effect, PC₂ a significantly negative effect, while PC₃ and PC₄ were non-significant.

4. Discussion

4.1. Uncertainty Analysis in Ecosystem Service Quantification

The InVEST model has been widely applied in quantifying ecosystem services. Based on land use data, climatic data, and vegetation data, its results hold significant value for regional and urban governance [27,28,29]. However, the InVEST model quantifies ecosystem services using static values rather than dynamic processes, which confers advantages in parameter accessibility and computational simplicity but also introduces uncertainties. For example, carbon density parameters based on land use cannot reflect intra-land-use heterogeneity or carbon density variations within identical land use types. In quantifying WY, the use of annual averages fails to capture impacts of extreme weather events induced by climate change. For HMI quantification, influences of vegetation morphology and species are not considered. These factors may contribute to model uncertainties. Additionally, temporally, all three ecosystem services are analyzed at annual resolution, introducing temporal-scale uncertainties.

This study selects Shanghai as the research area. Located on the alluvial plain of the Yangtze River Delta with flat topography, differences within identical land use types are relatively small. Model parameters from existing literature in similar geographical contexts [30] partially mitigate uncertainties. Furthermore, analyses at the 1 km grid and urban district scales focus on informing land use management policies for communities and urban areas. These scale selections reduce sensitivity to spatiotemporal uncertainties in the model.

4.2. Trade-Offs and Synergies in Urban Climate-Related Ecosystem Services

Trade-offs and synergies are central to ecosystem service research [31]. Results indicate that trade-offs and synergies are influenced by spatial scale, with relationships potentially shifting significantly across scales—a finding consistent with existing studies [19,32,33]. At finer scales, local parameter variability exerts greater influence, while broader scales reflect regional averages. This study also reveals temporal shifts in trade-off/synergy relationships, aligning with prior research, though directionality varies [23]. At community scales, urban renewal and green space renovations increase local parameter variability and ecosystem service fluctuations. From 2000–2020, correlations between CS and HMI/WY weakened, likely due to industrial-to-service sector transitions increasing green coverage per land unit while urban expansion reduced ecological lands (e.g., forests). HMI and WY shifted from positive to negative correlation, intensifying over time, possibly because urban expansion increased impervious surfaces (raising WY) while reducing evapotranspiration (lowering HMI). Such local-scale dynamics are more detectable at finer resolutions.

At broader urban district scales, trade-offs and synergies exhibit greater stability, with 2000–2020 changes substantially smaller than those at the 1 km scale. Large-scale relationships show reduced sensitivity to local variations but stronger responses to policies like ecological restoration projects and urban planning [18]. Compared to 2000–2010, post-2010 changes in trade-offs/synergies diminished, suggesting stabilization at urban district scales as urbanization slowed. Urban planning and land management policies should prioritize these relationships to achieve integrated optimization of ecosystem services.

4.3. Impact of Urbanization on Climate-Related Ecosystem Services

This study employs PCA to derive four principal components (PC1–PC4).

PC1 functions as a “comprehensive development index” reflecting socioeconomic intensity. Positive loadings (X1, X2, X3, X4, X6) indicate high scores corresponding to high public expenditure, green space coverage, GDP, industrial output, and population density—characteristics of urban cores or mature zones. The negative loading of X5 (−0.42) suggests low per capita housing completion in high-PC1 areas, aligning with urban lifecycle theory where central districts face construction saturation [34].

PC2 represents an “economy-environment trade-off.” High scores link to high GDP (X3: +0.54) but low green space (X2: −0.47) and industrial output (X4: −0.44), indicating service-led growth in commercial hubs at the expense of environmental resources and traditional industry.

PC3 captures “industrial productivity,” with high scores correlating to high GDP (X3) and industrial output (X4) but low population density (X6: −0.74), typifying industrial zones (e.g., Baoshan) where scale economies reduce labor dependence [35].

PC4 reflects an “underdevelopment index,” with high scores indicating low housing completion (X5: −0.68), GDP (X3: −0.45), population density (X6: −0.40), and industrial output (X4: −0.35), marking peripheral or emerging transitional zones.

PCR results reveal distinct urbanization impacts.

For WY, R² = 0.342 (34.2% explained), with no significant effects of all four principal components. This implies that climatic or hydrological variables (e.g., rainfall, evapotranspiration) likely dominate water yield variations beyond socioeconomic factors [36,37]. For CS, R² = 0.214 (21.4% explained), and principal components had no significant effects; this is likely attributable to carbon storage changes being primarily influenced by land-use transitions and forest management policies [38]. For HMI, R² = 0.652 (PC1–PC4 explain 65.2% variance). The negative effect of PC₁ on heat mitigation (−0.346) indicates significantly reduced heat mitigation capacity in high-PC₁ zones (urbanized mature areas). This may result from intensified heat island effects due to high socioeconomic activity (PC₁ loadings: high population density, industrial output). The negative loading of the completed housing area (X₅ = −0.42) suggests that urban expansion reduces permeable surfaces and enhances thermal radiation, aligning with existing studies [39,40]. The positive effect of PC₂ on heat mitigation (0.447) indicates improved heat mitigation capacity in high-PC₂ zones (service-dominated, low green space coverage). This may stem from reduced anthropogenic heat emissions during economic transitions (industry-to-services), coupled with policy-driven green renovations (e.g., rooftop greening) partially offsetting baseline green space deficits. Existing research confirms that urban renewal in central and older districts enhances heat mitigation capacity [41,42].

5. Conclusions

This study quantified the spatiotemporal evolution characteristics and trade-off/synergy relationships of WY, HMI, and CS in Shanghai from 2000 to 2020 and deciphered key urbanization-driven mechanisms. The key conclusions are as follows:

(1) Quantitative evidence of urban restructuring. Between 2000 and 2020, WY rose by 76.1% and became concentrated in the metropolitan core (Huangpu-Jing’an-Yangpu), transforming these districts into dominant stormwater sources. Simultaneously, CS in the historically low-carbon core increased by 9% via incremental greening, whereas planned new towns in Songjiang, Qingpu, and Pudong maintained or elevated CS despite rapid land conversion. This demonstrates that mature urban cores can regain carbon stocks after saturation and that strategic green-space quotas can reconcile urban expansion with sustained carbon sequestration. While artificial surfaces do not deliver on-site climate regulation, their water yield instead signals the volume of stormwater that must be managed. As the climate becomes wetter and more extreme, the high water yield from artificial surfaces increasingly signals flood risk rather than climate regulation. Expanding green infrastructure such as bioswales, permeable pavements, and rooftop vegetation can help to slow and absorb the runoff, easing the pressure on urban drainage, and can also increase the carbon storage.

(2) Dynamic trade-off trajectories. The antagonism between WY and HMI intensified with impervious-surface expansion, but localized synergies emerged where roadside and rooftop vegetation were introduced. CS–HMI and CS–WY relationships shifted from widespread conflict (2000) to broad synergy (2020) once peri-urban afforestation exceeded cropland losses, indicating that land-use zoning, rather than economic intensity per se, governs multi-service outcomes.

(3) Socio-economic drivers versus land-use conversion. Per-capita GDP and industrial output were significant negative predictors of HMI (p < 0.05), corroborating the adverse effect of urban expansion on heat mitigation. Conversely, per-capita green-space area was positively associated with HMI improvement yet exerted no significant influence on CS or WY, underscoring the limited efficacy of single-service ecological interventions. Urbanization indicators explained only a modest fraction of the variance in CS and WY, indicating that alterations in land-cover type, rather than aggregate socio-economic metrics, remain the primary determinant of carbon storage and water-yield dynamics.

This study provides spatial guidance for Shanghai’s land-use managers under compound climate hazards. The observed convergence of carbon storage between the historically low-carbon core and the high-carbon periphery shows that even mature districts can increase carbon storage through incremental greening, while planned new towns can maintain high stocks despite rapid development. Overlaying these patterns with flood-risk zones generated by high WY on artificial surfaces pinpoints where green infrastructure—bioswales, permeable pavements, and rooftop vegetation—will yield the greatest co-benefits for heat mitigation and storm-water retention. Thus, the integrated maps and district-specific trade-off metrics translate global ecosystem-service frameworks into locally actionable priorities.