Integrating the PLUS-InVEST Model to Project Water Conservation Dynamics and Decipher Climatic Drivers in the Chengdu–Chongqing Economic Zone Under Multiple Future Scenarios

Abstract

Identifying the evolutionary trends of water conservation functions and their climatic impacts under future scenarios is crucial for enhancing regional ecological security. This study integrates the PLUS and InVEST models with projected land use and meteorological data to analyze water conservation patterns in the Chengdu–Chongqing Economic Zone during 2030–2050 under natural development (ND) and ecological protection (EP) scenarios. Key findings include the following: (1) during 2000–2020, low-value areas decreased from 60% to 40%, while high-value zones expanded from 27.32% to 40.35%; (2) both the ND and EP scenarios project lower water conservation volumes compared to 2020 levels; (3) under the ND scenario, the combined proportion of high and extreme importance zones fluctuates at 0.51% (2030), 0.11% (2040), and 3.97% (2050); (4) spatial heterogeneity shows high-value clusters concentrated in Chengdu’s urban core and northeastern regions, contrasting with midland low-value areas; (5) the SSP1-1.9 climate scenario yields higher water conservation capacity with stronger spatial aggregation compared to SSP2-4.5. This integrated modeling of PLUS and InVEST provides scientific support for regional ecological security and sustainable development strategies.

1. Introduction

As the core of the upper Yangtze River Basin, the Chengdu–Chongqing Economic Zone plays a pivotal role in maintaining ecological security for the middle–lower Yangtze regions. Water conservation functions, serving as critical components of ecosystem services, regulate hydrological cycles through vegetation interception and soil water retention, thereby ensuring regional water resource equilibrium, flood mitigation, and ecological safety. Against the background of global climate change and accelerated regional urbanization, the spatiotemporal dynamics of water yield not only directly affect ecosystem stability but also play a critical role in policymaking for inter-basin water resource allocation and high-quality development, highlighting its significant scientific research value and practical governance implications. Current research focuses primarily on functional assessment, spatiotemporal patterns, and driving mechanisms. Quantitative evaluations employ water balance models (e.g., the Soil and Water Assessment Tool, SWAT) and ecosystem service models (e.g., Integrated Valuation of Ecosystem Services and Tradeoffs, InVEST). Notable studies by Li [1,2] (2021, 2024), Qing [3], Gao [4], and Lei [5] have identified spatiotemporal evolution patterns of water conservation using these models across diverse regions including the Qinling-Danjiang River Basin, Three-River Headwaters Region of the Tibetan Plateau, Yunnan Province, and Hainan Island. These investigations, combined with geographical detector analysis, have successfully elucidated key driving factors, establishing a robust scientific foundation for related research.

Recent studies have begun exploring water conservation functions under future development scenarios. For instance, Zhou [6] conducted scenario-based analyses of water conservation patterns, but these efforts remain limited to a single timeframe (2030) and scenario (natural development), crucially omitting climate change projections, and did not account for the impacts of future climate change. Consequently, these analyses exhibited limitations in both dynamic simulation accuracy and policy support capacity. In recent years, some scholars have investigated regional water conservation functions under different scenarios. For instance, Wang [7], using the FLUS-InVEST model, found that, under a natural development scenario, the water conservation capacity of the Fujian Delta urban agglomeration in 2030 would decrease by 1.5% compared to 2015. Ke [8] applied the LANDSCAPE and InVEST models to predict and assess water conservation in Wuhan under a farmland protection scenario for 2030, suggesting that the implementation of strict farmland protection policies in rapidly urbanizing areas could reduce water conservation capacity. However, systematic studies focusing on the “mountain–hill–plain” transitional zones in complex geomorphic regions of Southwest China—such as the Chengdu–Chongqing Economic Zone—are still lacking, and current research is insufficient to meet the needs of refined ecological management and differentiated regional governance. Consequently, two critical research gaps persist: (1) the need for multi-scenario comparisons across extended timelines (2030–2050), and (2) the imperative to integrate climate change impacts into scenario modeling. Addressing these limitations through systematic long-term scenario simulations will provide essential scientific support for regional ecological conservation and sustainable development strategies.

In land use simulation, conventional models like the Conversion of Land Use and its Effects at Small regional extent (CLUE-S) [9] and cellular automata–Markov (CA-Markov) [10] have demonstrated widespread application but face inherent constraints in simulation accuracy and spatial adaptability. The Patch-generating Land Use Simulation (PLUS) model has emerged as a superior alternative by integrating multi-objective optimization with cellular automata theory. This framework employs random forest algorithms and patch-level simulation strategies to effectively resolve spatial non-linearity and heterogeneity in land use dynamics [11], thereby achieving enhanced predictive reliability for future scenario analyses [12].

For water conservation assessment, the InVEST model offers distinct operational advantages through its multi-functional modularity, high-resolution outputs, and computational efficiency [13]. The synergistic coupling of PLUS and InVEST models [14,15] enables comprehensive scenario-based evaluation of water conservation functions, capitalizing on PLUS’s spatial simulation fidelity and InVEST’s ecosystem service quantification capabilities. This integration facilitates bidirectional simulation from land use change processes to the dynamic responses of ecosystem services, effectively addressing the limitations of the traditional “static land use–static service” modeling paradigm. It provides a new modeling framework for understanding land–ecosystem feedback mechanisms and their response pathways under future scenarios.

This study employs the PLUS and InVEST models to simulate land use patterns under multiple development scenarios in the Chengdu–Chongqing Economic Zone [16]. Leveraging simulation outputs and projected climate data, we quantify water conservation functions across the 2030–2050 timeframe. Spatial–temporal dynamics and clustering patterns are systematically analyzed using transition matrices and hot spot analysis techniques, with further evaluation of climate scenario impacts. The findings establish an empirical foundation for advancing ecological conservation and sustainable water resource management in this strategic economic region and promote the translational planning and empirical application of ecosystem service models in regional ecological governance.

2. Materials and Methods

2.1. Study Area

The Chengdu–Chongqing Economic Zone (27°39′ N–33°03′ N, 101°56′ E–110°11′ E), a strategic hub in southwestern China, constitutes a pivotal component of the Yangtze River Economic Belt and the Belt and Road Initiative. Encompassing 27 districts/counties in Chongqing Municipality and 15 prefecture-level cities in Sichuan Province, the zone spans 185,000 km². Its topography features mountainous hills in the west and basin-hill complexes in the east, traversed by major rivers including the Yangtze, Jialing, and Minjiang. The region receives an average annual precipitation of approximately 1000–1350 mm, with marked spatiotemporal variability. Land cover is dominated by forest, cropland, and built-up areas (Figure 1), with forests mainly concentrated in the western mountainous areas, playing a critical role in water conservation. As the core economic driver of western China, the region recorded a GDP exceeding 8 trillion CNY (2024) while serving as an essential ecological barrier and water conservation base for the upper Yangtze Basin. However, the region faces pronounced hydrological vulnerability. The steep terrain of the western mountainous and hilly areas, coupled with concentrated and torrential rainfall, frequently triggers soil erosion, undermining water conservation capacity. In the eastern basin–hill transition zone, population and industrial density are high, and agricultural non-point source pollution, along with industrial wastewater discharge, pose significant threats to river water quality. Additionally, some hydraulic infrastructure is outdated, leading to insufficient water resource allocation and limited capacity for flood and drought management. Recent collaborative ecological initiatives between Sichuan and Chongqing have maintained superior water quality, with the Yangtze mainstream in this section consistently achieving Class II status (2023) [17]. This unique dual role as an economic powerhouse and ecological safeguard necessitates scenario-based analysis of water conservation functions to optimize spatial governance under the “Three Zones and Three Lines” framework. Systematic quantification of mountain–watershed regulatory capacities will inform riparian restoration strategies to enhance runoff regulation [18].

2.2. Methodology

Drawing upon previous research [19] and preliminary findings [20], this study selected DEM, slope, NDVI, soil type, distance to water bodies, and distance to roads as driving factors for land use change, with water bodies designated as the restriction factor. Utilizing land use data from 2010 and 2020, this study first validated the feasibility and accuracy of PLUS model simulations at 30 m resolution within the Chengdu–Chongqing Economic Zone. Five-fold cross-validation was used to optimize the ntree and mtry parameters of the random forest algorithm, and sensitivity analysis of patch expansion threshold and minimum patch size was conducted in the CA module. The Kappa coefficient was employed to evaluate simulation accuracy. Based on the calibrated model, land use simulations for the Chengdu–Chongqing Economic Zone in 2030, 2040, and 2050 were carried out under two scenarios: ND and EP. The LULC raster outputs from the PLUS model were used both as biophysical input parameters for the InVEST water conservation module and as the basis for aggregating and analyzing water yield results at the watershed and 50 km² grid scales using watershed boundaries derived from DEM. The simulated land use data for 2030–2050 were then integrated with precipitation and evapotranspiration projections from CMIP6 under SSP1-1.9 and SSP2-4.5 scenarios to calculate water conservation capacity. Finally, the Moran’s I index was applied to quantify the degree of global spatial autocorrelation in water conservation under ND and EP scenarios within the Chengdu–Chongqing Economic Zone. Based on the county-level scale, the Getis-Ord Gi* method in ArcGIS Pro was used to identify high-value hot spots and low-value cold spots, enabling analysis of the spatiotemporal evolution and spatial clustering characteristics. The overall technical roadmap is shown in Figure 2.

2.2.1. PLUS Model

The PLUS (Patch-generation Land Use Simulation) model, developed by the HPSCIL@CUG laboratory research team, represents an advanced land use change simulation framework with two distinctive advantages over conventional models such as CLUE-S [19] and CA-Markov [21]:

Land Expansion Analysis Strategy: The model employs a sophisticated approach to identify driving factors for land use changes. Specifically, it utilizes the random forest algorithm to systematically analyze the causal relationships between land use expansion types and their driving forces. This process generates development probability surfaces for each land use category while quantifying the contribution weights of individual drivers [22,23,24]. By integrating the strengths of existing transformation analysis and spatial pattern analysis methodologies, this strategy enhances the model’s capacity to interpret land use change mechanisms within defined temporal intervals.

Multi-class Seed Growth Mechanism: The model incorporates a novel patch-level simulation mechanism that combines stochastic seed generation with a threshold-degradation algorithm. This dual approach enables dynamic simulation of patch evolution under the constraints of development probability surfaces, effectively capturing the spatial heterogeneity of land use transitions [22,23,24].

Building upon prior research [24], our simulation for the Chengdu–Chongqing Economic Zone followed a two-phase protocol:

• Phase 1: Calibration and Validation

Using land use data from 2000 and 2010 as baselines, we selected water bodies as exclusion zones and identified nine driving factors: distance to transportation networks (including highways, railways, and primary/secondary roads), proximity to water bodies, soil type, DEM, slope, and NDVI. The model simulated 2010 and 2020 land use patterns, which were compared against observed data through Kappa coefficient validation. Following established criteria [25,26], Kappa coefficients exceeding 0.75 indicate excellent agreement between simulated and actual land use distributions.

• Phase 2: Predictive Simulation

Upon achieving satisfactory validation accuracy, we employed the integrated Markov Chain module within the PLUS framework to project future land use demand. Based on 2020 baseline data, this module generated predictive land use maps for 2030, 2040, and 2050, ensuring temporal continuity in scenario analysis.

2.2.2. InVEST Model

The water conservation assessment based on the InVEST model primarily involves two functional modules: water yield estimation and water retention quantification.

(1)

Water Yield Calculation Module

Water yield refers to the total precipitation-derived water resources available for utilization within a defined geographical unit (e.g., watershed or catchment area) over a specific period. In this study, it specifically represents the water yield at the individual grid scale. The calculation was conducted through the water yield module of the InVEST model [27,28], which integrates land use patterns, soil characteristics, climatic conditions, and other factors to accurately quantify the spatial distribution of water yield. The model framework is established on the Budyko theory [29] and water balance principles, with its computational formula expressed as follows:

$$Y_{\mathrm{x}}=\left(1-{\displaystyle \frac{AE{T}_x}{P_x}}\right)\times P_x$$

(1)

$${\displaystyle \frac{AE{T}_{\mathrm{x}}}{P_x}}=1+{\displaystyle \frac{ET{O}_x}{P_x}}-{\left[1+{\left({\displaystyle \frac{ET{O}_x}{P_x}}\right)}^w\right]}^{{\displaystyle \frac{1}{w}}}$$

(2)

$$ETO=\mathrm{x}{K}_c\left(l_x\right)\cdot E{T}_{0x}$$

(3)

$$w=Z\times {\displaystyle \frac{AW{C}_x}{P_x}}+1.25$$

(4)

$$AW{C}_x=Min\left(Max.Soil.depth,Root.depth\right)\times PAWC$$

(5)

$$\begin{array}{c}PAWC\left(\%\right)=54.509-0.132SAN\%-0.003{\left(SAN\%\right)}^2-0.055SIL\%-0.006{\left(SIL\%\right)}^2\\ -0.738CLA\%+0.007{\left(CLA\%\right)}^2-2.688OM\%+0.501{\left(OM\%\right)}^2\end{array}$$

(6)

In the hydrological model framework, key variables are defined as follows: Y_x denotes the annual water yield (mm) at grid cell x; AET_x represents the actual annual evapotranspiration (mm); P_x indicates the annual precipitation (mm); ETO_x refers to the potential annual evapotranspiration (mm) at grid cell x; w signifies the ratio of vegetation water demand to precipitation; ET_0x describes the reference evapotranspiration (mm) governed exclusively by local climatic conditions; Kc(l_x) denotes the vegetation evapotranspiration coefficient (decimal value 0–1.5) for land type l_x at grid x, determined by vegetation coverage density; Z corresponds to the dimensionless seasonal constant (Zhang coefficient), with a value range of 1 to 30; AWC_x quantifies the plant-available water content (mm) derived from the minimum value between Max.Soil.depth (maximum soil depth, mm) and Root.depth (vegetation root depth, mm), scaled by PAWC% (plant-available water capacity percentage), with a value range of 0 to 200 mm. Soil composition parameters include SAN% (sand content), SIL% (silt content), CLA% (clay content), and OM% (organic matter content), expressed as mass percentages.

(2)

Water Conservation Capacity Calculation Module

Building upon water yield outputs, the water conservation capacity is derived through systematic corrections employing three key hydrological parameters: the topographic index (TI), soil saturated hydraulic conductivity (Ks, mm/d), and flow velocity coefficient (V). The computational framework integrates these parameters as follows:

$$WR=\min \left(1,{\displaystyle \frac{249}{V}}\right)\times \min \left(1,{\displaystyle \frac{0.9\times TI}{3}}\right)\times \min \left(1,{\displaystyle \frac{K_s}{300}}\right)\times Y_x$$

(7)

$$TI=\lg {\displaystyle \frac{D_{area}}{soi{l}_{dep}\times P_{slope}}}$$

(8)

In the above formula, WR represents water conservation capacity per unit area (mm); V denotes the flow velocity coefficient; TI signifies the topographic index; Ks indicates soil saturated hydraulic conductivity (mm/d), calculated using the Soil Water Characteristics module within the SPAW (Soil Plant Atmosphere Water) software v6.02; Darea corresponds to the watershed grid count (dimensionless); soil_dep refers to soil layer depth (mm); P_slope represents percentage slope (%).

To identify spatial distribution patterns of critical water conservation areas, this study implements the ecosystem service importance classification criteria specified in the Technical Guidelines for Ecological Protection Redline Delineation, employing ArcGIS’s Natural Breaks (Jenks) method to categorize water conservation service functionality into five hierarchical tiers. This statistically driven classification optimizes inter-group variance while preserving spatial continuity of hydrological service capacities, thereby objectively demarcating priority conservation zones through data-driven threshold determination.

In addition, this study adopts two scenarios—SSP1-1.9 (low emissions) and SSP2-4.5 (moderate emissions)—which together encompass the potential extremes of emission reduction and land use policies. These scenarios reflect differences in precipitation, evapotranspiration, and land use assumptions under varying mitigation pathways, thereby facilitating a comparative assessment of the spatiotemporal responses of water conservation functions across different trajectories.

2.2.3. Spatial Autocorrelation

Global spatial autocorrelation analysis (Moran’s I) aims to quantify whether spatial data exhibit a clustered or dispersed pattern and further reveal the strength and significance of such a pattern. The value of Moran’s I ranges from [−1, 1]. A value greater than 0 indicates positive spatial autocorrelation, and the higher the index, the stronger the spatial correlation and the more evident the spatial clustering. A value less than 0 indicates negative spatial autocorrelation, where a lower index suggests greater spatial heterogeneity. A value of 0 implies a random spatial distribution with no apparent correlation.

At the county level, the spatial autocorrelation (Moran’s I) tool in the ArcGIS platform was used to calculate the Z-score and p-value for each element. These are key indicators for determining statistical significance. Specifically, |Z| > 1.96 and p < 0.05 indicate statistically significant spatial autocorrelation; otherwise, the result is considered not significant. The calculation formula is as follows:

$$M\mathrm{oran}’sI={\displaystyle \frac{{\displaystyle \sum _{i=1}^n{\displaystyle \sum _{j=1}^n{w}_{ij}(x_i-\overline{x})}}(x_j-\overline{x})}{({\displaystyle \sum _{i=1}^n{\displaystyle \sum _{j=1}^n{w}_{ij}}}){\displaystyle \sum _{i=1}^n(x_i-\overline{x})}}}$$

(9)

$$Z={\displaystyle \frac{I-E[I]}{\sqrt{Var(I)}}}$$

(10)

In the formula, x_i and x_j represent the attribute values of grid cells i and j, respectively; $\overline{x}$ is the mean of the attribute values; w_ij is the spatial weight matrix; and n is the number of grid cells. I is the actual computed statistic (e.g., the Moran’s I value); E\[I] is the expected value under the null hypothesis of no spatial autocorrelation (with the expected value of Moran’s I being −1/(n − 1)); and Var(I) denotes the variance of the statistic.

2.2.4. Hot Spot Analysis

Cold and hot spot analysis is one of the methods for testing global clustering. It calculates the local spatial autocorrelation index (G) for elements within a dataset. In this study, the Hot spot Analysis tool based on the Getis–Ord Gi* statistic in the ArcGIS platform was used to explore spatial autocorrelation through local indicators. The calculation formula is as follows:

$$G_{\mathrm{i}}^{*}(d)={\displaystyle \frac{{\displaystyle \sum _j^n{W}_{ij}(d)X_j}}{{\displaystyle \sum _j^n{X}_j}}}$$

(11)

In the above formula, E(Gi*) and Var(Gi*) represent the mathematical expectation and variance of Gi*, respectively; W_ij(d) denotes the spatial weight. The greater the absolute value of Gi*, the less likely the result is due to random chance, indicating stronger statistical significance. A Gi* value greater than 0 indicates that the area is a clustered hot spot; a value less than 0 indicates a negatively clustered cold spot; and a value of 0 suggests the result is randomly generated and not statistically significant. By conducting a significance test on the obtained Gi* values, cold and hot spot areas within specific confidence intervals can be identified.

2.3. Data Sources

The data used in this study include basic geographic data, land use data, environmental-driven data, climate data, scenario-driven data, and abrupt event data. Detailed descriptions and sources of these data are provided in

Table 1. Summary of Data Information.

Dataset Type	Name	Source	Resolution (m)	Model Role	Preprocessing Steps
Basic Geographic Data	DEM	Geospatial Data Cloud	30	Slope extraction; one of the driving factors for PLUS model	Extracted study area using clipping function in ArcGIS
	Study Area Boundary Data	National Geographic Information Public Service Platform	/	Define boundary scope	Selected and exported study area boundary using ArcGIS
Land Use	LULC Imagery (2000, 2010, 2020)	Resource and Environmental Science Data Center, Chinese Academy of Sciences	30	Training and validation of PLUS model; raster input for InVEST water model	Reclassified into five categories (cropland, forest, grassland, water body, built-up land, unused land) and clipped to study area using ArcGIS
Environmental-driven Data	Road Data	OpenStreetMap	/	Random forest driving factor for PLUS model	Imported and clipped in ArcGIS
	NDVI (Normalized Difference Vegetation Index)	Geospatial Data Cloud	500	Driving factor for PLUS model	Atmospheric correction, projection, and other processing using ENVI
Climate Data	Annual Precipitation (2000, 2010, 2020)	National Qinghai-Tibet Plateau Scientific Data Center	1000	Input for InVEST model to calculate water yield	Raster projection and raster calculation using ArcGIS
	Potential Evapotranspiration (2000, 2010, 2020)
Scenario-driven Data	Annual Precipitation (SSP1-1.9, SSP2-4.5)	National Qinghai-Tibet Plateau Scientific Data Center	1000	Input for InVEST model to calculate water yield	Raster projection and raster calculation using ArcGIS
	Potential Evapotranspiration (SSP1-1.9)
Soil Data	Soil Type	World Soil Database	1000	Input for InVEST model and water conservation calculation	Raster projection and raster calculation using ArcGIS
	Root Restriction Layer Depth	Literature-shared data [30]	1000	Input for InVEST model	Extracted by mask and raster calculation in ArcGIS

. All datasets underwent spatial harmonization through 30 m resampling using bilinear interpolation and coordinate system unification to CGCS2000 in ArcGIS Pro 3.0.1.

3. Results and Analysis

3.1. Analysis of Land Use Evolution Trends in the Chengdu–Chongqing Economic Zone

3.1.1. Evolution of Land Use Status

As shown in the land use change map (Figure 3) and transfer matrix from 2000 to 2020, the land use pattern in the Chengdu–Chongqing Economic Zone exhibits significant spatiotemporal heterogeneity. In 2020, forest cover accounted for the largest proportion (approximately 29.81%), concentrated in ecological barrier zones such as the Qinba Mountains and Wuling Mountain Range. Construction land coverage increased 2.4-fold from 1.48% in 2000 to 3.58% in 2020, expanding linearly along the Chengdu–Chongqing development axis (Chengdu–Chongqing–Yongchuan–Jiangjin) with an annual growth rate of 0.11%, primarily encroaching on cultivated land (annual decrease of 9.69%) and grassland (annual decrease of 4.05%). Grassland area declined from 18,216.11 km² in 2000 to 16,523.33 km² in 2020, with degraded areas partially converted to forest through ecological restoration. Water bodies increased by 0.12% after 2010 but experienced localized shrinkage due to urbanization. Unused land slightly expanded from 211.65 km² in 2000 to 307.03 km² in 2020.

The PLUS model demonstrated high reliability in simulating future land use for the Chengdu–Chongqing Economic Zone. Validation against actual data from 2010 and 2020 yielded Kappa coefficients of 0.81 and 0.82, respectively, exceeding the 0.75 threshold for high consistency. This confirms the model’s applicability for simulating 30 m resolution land use data at large regional scales.

3.1.2. Land Use Evolution Trends Under Future Scenarios

Simulation results from the PLUS model (Figure 4) reveal significant spatiotemporal differentiation in land use patterns under the ND scenario in the Chengdu–Chongqing Economic Zone from 2020 to 2050. Cultivated land area continued to decline, with an average annual decrease of 0.11% between 2030 and 2050, primarily converted to forest land (51.01% of total transfers) and construction land (24.31%). Forest land area showed sustained growth, increasing by 0.02% annually during the same period. Construction land expanded rapidly, reaching 6.73% by 2050—1.55 times higher than in 2030—with expansion hot spots concentrated in the core areas of the Chengdu–Chongqing dual-city Economic Circle and along major transportation corridors. Grassland and water areas exhibited slight increases, mainly derived from conversions of cultivated and forest lands.

Under the EP scenario (Figure 5), forest land area in 2050 increased by 0.49% compared to the ND scenario, while water area decreased by 0.03%. The expansion rate of construction land slowed by 0.04%, exhibiting a “two-points-one-axis” intensive spatial pattern with strict limitations on development intensity in ecologically fragile areas. The loss rate of cultivated land decreased to an annual average of 0.07%, influenced by the cultivated land requisition–compensation balance policy, with lost areas shifting to hilly and gentle slope zones. Forest land expansion surpassed ND scenario levels, while construction land growth diminished due to ED constraints. Unused land showed no significant proportional changes. Land use transition trends revealed a 1.86% increase in bidirectional conversion between forest and cultivated lands compared to the ND scenario, demonstrating that policy interventions under the EP scenario significantly enhanced land system resilience.

3.2. Analysis of Water Conservation Function Evolution Trends

3.2.1. Evolution of Water Conservation Function

Using 2020 as the baseline year for natural break classification, due to its highest water conservation values and availability of precise data, water conservation capacity was classified into five importance levels: non-significant (0–94.040 mm), moderately significant (94.040–182.548 mm), moderately high (182.548–304.247 mm), highly significant (304.247–464.668 mm), and extremely significant (>464.668 mm). From 2000 to 2020, significant spatial–temporal changes in water conservation grades occurred (Figure 5), characterized by a gradient differentiation pattern of “low-value areas contracting and high-value areas expanding.” The mean water conservation capacity showed continuous growth, with low-grade zones shifting from near 60% coverage in 2000 to approximately 40% concentrated in the northwestern arid belt. Medium–high- and high-grade zones formed stable cores in the southeastern monsoon humid region, increasing their combined area proportion from 2.50% to 3.40%. Medium–low-grade zones expanded from 27.32% to 40.35% of the total area through ecological restoration projects, notably in farmland-to-forest conversion areas and riparian wetland restoration zones where grade improvements were most pronounced.

3.2.2. Water Conservation Function Levels Evolution Trends Under Future Scenarios

Under both ND and EP scenarios, water conservation function levels from 2030 to 2050 remained lower than those in 2020. In the ND scenario, the proportions of highly significant and extremely significant grades in 2020, 2030, 2040, and 2050 were 5.02%, 0.51%, 0.11%, and 3.97%, respectively. Significant divergences in water conservation grade evolution trends were observed between ND and EP scenarios (Figure 6), exhibiting a spatiotemporal differentiation pattern characterized by “high-value grades concentrated in Chengdu’s main urban areas and northeastern regions, while low-value grades diffused across peripheral mountainous zones and central areas.”

The grade evolution trends under the EP scenario largely mirrored those of the ND scenario (Figure 7). From 2030 to 2050, the mean water conservation capacity per unit area increased by 35.55% and 36.12% under the ND and EP scenarios, respectively. High-grade zones clustered in the Three Gorges Reservoir urban agglomeration of northeastern Chongqing, the coordinated development zone of eastern Sichuan, and Chengdu’s periphery, showing significant spatial coupling with construction land expansion. During 2030–2050, the combined proportions of highly significant and extremely significant grades remained notably lower than in 2020, exhibiting a fluctuating downward trend. Under the ND scenario, the proportions of highly significant and extremely significant grades declined from 3.40% and 1.62% in 2020 to 0.48% and 0.03% in 2030, respectively, before recovering to 3.40% and 0.57% by 2050. Conversely, the proportions of medium–low significance grades (moderately significant, moderately high, and non-significant) surpassed 2020 levels, displaying a fluctuating upward trend. Under the ND scenario, the proportions of moderately significant, moderately high, and non-significant grades shifted from 10.85%, 40.35%, and 43.79% in 2020 to 2.93%, 23.29%, and 73.28% in 2030, respectively, and further adjusted to 8.78%, 32.49%, and 54.77% by 2050.

3.3. Analysis of Spatial Aggregation Characteristics of Water Conservation Function

3.3.1. Spatial Autocorrelation Analysis of Water Conservation Function in the Chengdu–Chongqing Economic Zone

The results of spatial autocorrelation analysis (

Table 2. Calculation results of the global Moran’s I index for water conservation function in the Chengdu–Chongqing Economic Zone.

Scenario Models	Year	Moran’s I	Z-Score	p Value	Variance
Current Status	2000	0.424	7.543	0	0.004
	2010	0.463	8.263	0	0.003
	2020	0.282	5.192	0	0.003
ND Scenario under SSP1-1.9	2030	0.448	8.061	0	0.003
	2040	0.509	9.127	0	0.003
	2050	0.441	7.856	0	0.003
EP Scenario under SSP1-1.9	2030	0.446	8.027	0	0.003
	2040	0.508	9.114	0	0.003
	2050	0.436	7.780	0	0.003

) indicate that the global Moran’s I index for water conservation functions in the Chengdu–Chongqing Economic Zone during 2000, 2010, 2020, and 2030–2050 under the SSP1-1.9 scenario—both ND and EP—remains around 5%, showing significant spatial correlation. All corresponding Z-scores exceed 1.96, and p-values are below 0.05, suggesting a strong spatial dependency in water conservation functions across the region.

The trend of the Moran’s I index under both ND and EP scenarios from 2000 to 2050 is generally consistent, with values first decreasing and then increasing. This indicates that the overall spatial clustering degree of water conservation functions in the Chengdu–Chongqing Economic Zone initially weakens and then strengthens over time.

To further reveal local spatial heterogeneity and accurately identify core areas of clustering, this study employed the Getis–Ord Gi* hot spot analysis at the county level as the basic spatial unit to generate hot spot maps, thereby identifying cold and hot spot regions of water conservation.

3.3.2. Current Spatial Aggregation Characteristics of Water Conservation Function in the Chengdu–Chongqing Economic Zone

Aggregation analysis results (Figure 8) indicate that, from 2000 to 2020, the spatial aggregation characteristics of water conservation function in the study area exhibited a differentiated pattern of “east–west polarization and gradient transition.” Hot spot areas remained persistently clustered, primarily concentrated in northeastern Chongqing. Cold spot areas, initially sporadically distributed in the central-left region in 2000, gradually migrated toward the northwestern arid belt. By 2020, cold spot areas contracted compared to 2000, mainly aggregating in Mianyang City, Deyang City, Suining City, Tongnan District, Dazu District, Tongliang District, Hechuan District, and Guang’an City. The shifting distribution of cold and hot zones reflects a transformation in the spatial heterogeneity of water conservation function from “simple high–low clustering” to “complex diversification.”

3.3.3. Spatial Aggregation Characteristics of Water Conservation Function Under Future Scenarios in the Chengdu–Chongqing Economic Zone

Spatial clustering analysis of water conservation function under the ND and EP scenarios was conducted using SSP1-1.9 (Figure 9,

Table 3. Summary of the area proportions of cold and hot spot regions for water conservation in the Chengdu–Chongqing Economic Zone under ND and EP scenarios from 2030 to 2050.

Year	Region	Natural Development (ND) Scenario	Ecological Protection (EP) Scenario
2030	Hot spot	28.32%	28.32%
	Cold Spot	12.83%	12.83%
2040	Hot spot	22.57%	22.57%
	Cold Spot	12.82%	12.82%
2050	Hot spot	18.79%	18.79%
	Cold Spot	20.72%	19.29%

). Under the ND scenario, the cold and hot spot pattern exhibited “low-value diffusion and high-value fragmentation.” From 2030 to 2050, hot spot areas decreased in size, concentrating in northeastern and southwestern Chongqing, including Meishan City, Leshan City, Liangping District, Wanzhou District, and Kaizhou District. Cold spot areas formed a continuous belt along the Mianyang–Suining–Yongchuan–Jiangjin corridor, expanding in area by 2050 compared to 2030. These shifts reflect an exacerbated spatial imbalance between ecosystem service supply and urbanization demands.

Under the EP scenario (Figure 9), the cold and hot spot pattern demonstrated “low-value integration and high-value synergy.” While hot spot area changes aligned with the ND scenario, cold spot areas in 2050 excluded Beibei District, Shapingba District, Yuzhong District, and Nan’an District compared to the ND scenario. These transformations underscore the supporting role of ecological priority policies in fostering sustainable development for Chengdu–Chongqing’s “dual-core” urban system, providing a regional exemplar for the “collaborative conservation” strategy in the Yangtze River Economic Belt.

4. Discussion

Based on the dynamic land use simulation results from 2000 to 2050, the PLUS model revealed a continuous conversion of cropland to forest and built-up land in the Chengdu–Chongqing Economic Zone and demonstrated greater land system resilience under the ecological protection scenario, consistent with the findings of Wang [31], Zhi [32], and others. This finding not only confirms the profound influence of policy regulation and spatial competition mechanisms on the evolution of land use patterns but also highlights the advantages of coupled simulation in research related to ecological barrier construction.

Building on this, the temporal analysis of water conservation functions under the SSP1-1.9 scenario shows an overall downward trend in water conservation capacity between 2030 and 2050, accompanied by increasing fluctuations. Meanwhile, the proportion of regions classified as having a highly important level of water conservation experienced a decline, followed by a rebound, reflecting a sensitive response to the balance between water recharge and evapotranspiration under climate change. This result underscores that evaluating water conservation under a low-emission pathway alone is insufficient to meet the region’s comprehensive water security demands amid economic growth.

Therefore, introducing the moderate-emission SSP2-4.5 scenario, which better aligns with current policy trajectories, holds significant importance. Under the SSP2-4.5 scenario (Figure 10), the spatial clustering of water conservation functions exhibits an imbalanced pattern characterized by fragmentation of high-value hot spots and belt-like expansion of low-value cold spots. Although hot spots remain concentrated in the core western catchment areas, their boundaries contract significantly and do not form large-scale contiguous zones. In contrast, cold spots spread in a linear pattern along river plains and hilly zones, showing relatively high spatial stability, particularly in the central part of the region, where they persist over time. Compared to the stable configuration under SSP1-1.9, the SSP2-4.5 scenario shows more pronounced hot spot contraction and cold spot expansion, indicating a decline in water conservation capacity under the moderate-emission pathway. This aligns with the findings of Ma [33], reflecting the continued ecological pressure caused by the coupled effects of emissions and land use under the SSP2-4.5 pathway. These results provide precise spatial references for dynamic optimization of ecological redlines, zonal management and control, and targeted water replenishment, thereby directly supporting regional planning policies and water resource governance practices.

Although previous studies—such as those by Luo [34], Wen [35] and Zhang [36]—have applied the coupled PLUS-InVEST model to evaluate land use dynamics and ecosystem services (e.g., carbon storage and habitat quality) in regions such as Xi’an City, the Pinglu Canal Economic Belt, and the Yellow River water conservation area, targeted research on the feedback mechanisms between watershed water conservation functions and land use in complex geomorphic regions remains limited. Given the Chengdu–Chongqing Economic Zone’s distinctive “mountain–hill–plain” transitional landform and its critical role as an ecological barrier in the upper Yangtze River, this study’s coupled PLUS-InVEST modeling approach provides bidirectional feedback between land use dynamics and water conservation services in the region, thereby enhancing insight into the evolution of ecological barriers under various scenarios.

However, this coupled modeling framework still bears uncertainties, such as mismatches between future climate projections and land use assumptions, insufficient responsiveness of static vegetation parameters to extreme events, and simplified parameter transmission processes. Future research may improve simulation accuracy and enhance practical decision support for multi-scale ecological spatial planning and resource management by integrating multi-model ensembles, quantifying uncertainties, and exploring the incorporation of models like Land-N2N and dynamic vegetation modules.

5. Conclusions

This study systematically revealed the spatiotemporal evolution patterns of future water conservation function in the Chengdu–Chongqing Economic Zone by integrating the PLUS and InVEST models to simulate land use data for 2030–2050 and incorporating future climate data. The study overcomes the limitations of traditional single-factor simulations and addresses the gap in systematic research on the complex “mountain–hill–plain” geomorphology. It achieves dynamic simulation of water conservation under the interaction of climate and land use, offering new perspectives for enhancing model accuracy and flexibility. Key findings include the following:

(1)

The PLUS model demonstrated high accuracy (Kappa coefficient: 0.81–0.82) in simulating 30 m resolution land use data for the Chengdu–Chongqing Economic Zone, providing reliable support for large-scale water conservation function predictions.

(2)

From 2000 to 2020, the water conservation function exhibited a gradient differentiation characterized by “contraction of low-value zones and expansion of high-value zones.” Under both ND and EP scenarios, water conservation function during 2030–2050 remained lower than 2020 levels.

(3)

Under the ND scenario, the proportions of highly important and extremely important grades in 2020, 2030, 2040, and 2050 were 5.02%, 0.51%, 0.11%, and 3.97%, respectively. Spatiotemporal differentiation manifested as “high-value clustering in Chengdu’s urban core and northeastern regions, with low-value diffusion in peripheral mountainous and central areas.”

(4)

The ND scenario showed a cold and hot spot pattern of “low-value diffusion and high-value fragmentation,” while the EP scenario exhibited “low-value integration and high-value synergy.”

(5)

Divergences in water conservation function between SSP1-1.9 and SSP2-4.5 scenarios underscored the interactive effects of climate policy intensity and human activities. These findings reveal the evolutionary characteristics of water conservation function under varying development pressures, offering critical insights for regional ecological security and sustainable development.

Based on the above findings, this study not only provides a scalable coupled approach for modeling water conservation under future scenarios but also offers data support and decision-making references for regional spatial planning, such as ecological compensation zoning, dynamic optimization of ecological redlines, and water resource allocation strategies. However, certain limitations remain. This study does not fully incorporate ground-based measurements of runoff and evapotranspiration, and the simulation of responses to extreme climate events involves uncertainties. Future research could enhance model accuracy and practical value through multi-model integration, fusion of observational data, and the development of scenarios with higher spatiotemporal resolution.