Land use change, driving factors, and multi-scenario projections of ecosystem service supply in the Yangtze River delta urban agglomeration

1 Introduction

Global urbanization is accelerating rapidly, with over half the global population residing in urban areas and this proportion projected to reach 68% by 2050 according to the United Nations World Urbanization Prospects. This trend is particularly pronounced in China, where major city clusters including the Yangtze River Delta have been positioned as core national development engines under strategic spatial planning frameworks (Jia et al., 2024; Zhou et al., 2025). While these policies boost economic agglomeration and infrastructure development effectively (Meury et al., 2025), rapid land use conversion exerts intense ecological pressure. Sustained built-up area expansion encroaches on croplands, forests and wetlands, fundamentally reshaping regional landscape patterns (Verma et al., 2024; Herdt and Wälty, 2026). This transformation fragments natural habitats, degrades ecosystem functions, and reduces the supply of key services such as water conservation, carbon sequestration and soil retention (Jessica et al., 2022; Bagstad et al., 2019; Sharma et al., 2024; Yi et al., 2022). Rapid urban expansion further intensifies pressure on ecosystem service supply, threatening sustainable development in densely urbanized regions (Wang et al., 2021). Therefore, clarifying the interactions between urban expansion, land use change and ecosystem service dynamics is essential to establish effective spatial governance and coordinate socioeconomic development with ecological security.

However, retrospective analyses of historical ES dynamics, though effective in detecting ecosystem service supply pressure (Yu et al., 2025), have limited ability to guide future sustainable development pathways (Sullivan et al., 2024). Urban systems are complex and nonlinear; future changes in ES supply will be dominated by dynamic adjustments in policies, land use strategies and climatic conditions. Thus, simulating the evolution of ecosystem service supply under diverse plausible scenarios is critical for proactive ecological governance (Ai et al., 2024). An integrated framework combining land use simulation models with ES assessment under multiple scenarios supports forward-looking analysis of these dynamics (Li et al., 2023). This multi-scenario approach helps clarify the complex mechanisms by which land use decisions affect ES cascades, and provides reliable scientific support to steer regions toward a more sustainable and ecologically secure development path.

Therefore, to investigate the ecosystem service provision capacity and ecological condition and sustainable development prospects of key regions such as urban agglomerations, this study selects the YRDUA as the research area. Applying ecosystem service theory and employing the InVEST and PLUS model, supported by multi-source data including meteorological observations, remote sensing imagery, and economic statistics, the following research was conducted: (1) assessment of land use changes and ecosystem service capacities in the region; (2) establishment of a driving factors framework for land use changes to identify key drivers of various land types; and (3) development of three scenarios—baseline, economic priority, and ecological priority—to conduct multi-scenario simulations. These simulations project and analyze the spatial patterns of land use changes and ecosystem service prospects in the YRDUA for 2030 and 2060, providing a reference for scientifically mitigating ecological risks in the region.

2 Materials and methods

2.1 Study area and data sources

2.1.1 Study area

The YRDUA (Figure 1), located in eastern coastal China, is a key engine of national economic growth and one of the world’s six major world-class urban agglomerations. Stretching from 116°E to 123°E longitude and 27°N to 33°N latitude, it has a subtropical monsoon climate with clear seasons and coordinated rainfall and temperature. Covering 2.12 × 10⁵ km², centered on Shanghai, it includes 26 prefecture-level cities across Jiangsu, Zhejiang and Anhui provinces. Dense water networks, including the Yangtze River, Taihu Lake and Qiantang River, sustain water supply and regional ecological stability. The region had a permanent population of about 225 million and a GDP of 31 trillion yuan in 2025 (Tang et al., 2025), making it China’s most economically developed and densely urbanized area. It should be noted that the historical land-use and ecosystem service analysis in this study is based on data from 2000–2020, and future projections are for 2030 and 2060.

2.1.2 Data

This study employed multi-source datasets to examine land use dynamics and ecosystem service changes in the YRDUA. Data included multi-temporal land cover from remote sensing, meteorological data, WorldPop population grids, and GDP and administrative boundaries from the Resource and Environment Science and Data Center of the Chinese Academy of Sciences. Socioeconomic statistics were collected from yearbooks and water bulletins. Transportation and river vectors, 30 m ASTER GDEM, and soil parameters from the Harmonized World Soil Database were also used. The data and sources used are as shown in Table 1.

TABLE 1. Description of data and sources.

It should be noted that the historical land-use analysis in this study covers 2000–2020, while ecosystem service assessment covers 2010–2020 (three time points: 2010, 2015, and 2020) due to data availability constraints for pixel-level socio-economic parameters before 2010. Future projections are for 2030 and 2060.

2.2 Land use transfer matrix

To accurately characterize the dynamic evolution of geospatial units during the study period, a land use transfer matrix needs to be constructed based on the changes in land use types of the same units between the start and end of each observation period.

This matrix visually reflects the transformation pathways and intensity of various land cover types within the study area. Assuming there are n distinct land use types, the land use transfer matrix can be represented as an n × n matrix. Each element in the matrix denotes the transition proportion or probability from the ith land use type to the jth land use type. This can be used to analyze and predict the changing trends and relationships between different land use types. The land use transfer matrix serves as a critical analytical tool in fields such as land use planning, resource management, and environmental protection, enabling decision-makers to better understand the patterns of land use change and formulate rational land use policies and spatial planning strategies.

2.3 Ecosystem services assessment

The YRDUA faces rapid population growth and urban expansion, which greatly suppresses local ecosystem service supply. As a core region for carbon control under China’s ‘Dual Carbon’ target, the region depends heavily on ecosystem carbon storage. This study selected four key ecosystem services: water conservation, carbon storage, soil conservation, and crop production, to assess regional ecosystem service provision capacity and ecological condition. Following China’s Technical Specification for Ecosystem Service Assessment (HJ 1173–2021) and previous studies, the InVEST model was applied to evaluate ecosystem service status.

Table 2 lists detailed assessment methods for each service. Specifically, the water yield module is based on the Budyko curve framework. Carbon storage is evaluated using the four-carbon-pool framework (aboveground biomass, belowground biomass, soil organic matter, and litter)(Xu et al., 2019). Soil conservation is calculated using the Revised Universal Soil Loss Equation (RUSLE)(Wischmeier, 1959; Zhang et al., 2022). Crop production was assessed using Normalized Difference Vegetation Index (NDVI) data-based spatial allocation method: total grain production for each city was obtained from statistical yearbooks, annual maximum NDVI were used as a proxy for vegetation productivity, and statistical production was allocated to pixels following Y i j = Y j × N D V I i j / ∑ N D V I i j. Crop type differentiation was not separately modeled; comprehensive NDVI was used for total grain production allocation. The ecosystem service assessment covers 2010–2020 (three time points: 2010, 2015, 2020) due to data availability constraints for pixel-level socio-economic parameters before 2010, while land-use change analysis covers 2000–2020. To ensure accuracy, simulation results were compared with official data from water bulletins and statistical yearbooks. Water yield was validated against observed runoff data, and crop production against statistical yearbook data, with errors of all indicators within 15%, consistent with the commonly accepted range (10%–20%) for ecosystem service models. The Zhang coefficient was calibrated to 27.2, meeting research accuracy requirements. For future crop production projections (2030 and 2060), the 2020 NDVI-based productivity conditions were held constant, and crop production was re-allocated spatially based solely on the projected land-use changes under each scenario. This assumption isolates the impact of land-use pattern change on crop production, without considering future improvements or declines in agricultural technology and climate-driven yield changes.

TABLE 2. Evaluation methods for ecosystem services.

2.4 Analysis of driving factors of land use change

Regional land use change is driven by a combination of socioeconomic demands, natural conditions, and infrastructure development. Infrastructure, such as roads and railways, enhances the efficiency of material and information flow, thereby influencing the spatial concentration or dispersion of resources. The driving factor analysis was conducted using the LEAS (Land Expansion Analysis Strategy) module of the PLUS model, which employs a Random Forest algorithm to extract patch-level land-use expansion data and calculate the contribution (feature importance) of each driving factor for each land-use type. To analyze these effects, this study establishes a driving factors system from three dimensions: infrastructure, socioeconomic conditions, and the natural environment. Specific indicators include proximity to major roads and railways; regional GDP and population; annual rainfall, temperature, and evapotranspiration; distance to major rivers; slope; and DEM. This framework is designed to systematically examine the influence of these factors on recent land use change in the YRDUA. Contribution is expressed as Random Forest feature importance, where higher values indicate higher Random Forest feature importance score for that land-use type’s expansion. The composition of the driving factors system, along with detailed data processing procedures, is provided in Table 3.

TABLE 3. Driving factors of land use type change.

2.5 Land use change prediction based on PLUS model

The PLUS model is a patch-level land use change simulation model driven by historical raster data. It contains two core modules: the Land Expansion Analysis Strategy (LEAS) and Cellular Automata based on Random patch Seeding (CARS). Built on the Cellular Automata (CA) framework, this model adopts the Random Forest (RF) algorithm to extract patch expansion data. It ranks driving factors sequentially and identifies regional land use trends, thus finishing patch-scale simulation. The PLUS model outperforms traditional models with high integration and superior prediction accuracy.

Figure 2 illustrates the model simulation workflow in this study. The target land-use demand for each category in 2030 and 2060 was determined using a Markov chain prediction based on historical land-use transition probabilities from 2000 to 2020 (Li et al., 2024; Yang et al., 2022). Three contrasting scenarios were designed: (a) Baseline scenario: land-use evolves according to historical transition probabilities from 2000–2020 without additional constraints; (b) Economic priority scenario: construction land demand is increased, natural lands are permitted to convert to cropland and construction land unidirectionally, and ecological land expansion is constrained; (c) Ecological priority scenario: construction land expansion is constrained, forest and grassland are strictly protected, and ecological restoration is encouraged. Each land use type was parameterized separately per scenario logic. The model achieved a Kappa coefficient of 0.928, overall accuracy of 0.951, and FoM (Figure of Merit) of 0.144. FoM specifically assesses the model’s ability to simulate the location of land-use changes; a value of 0.144 is within the acceptable range (0.1–0.3) commonly reported in similar studies. The confusion matrix was computed excluding permanent water bodies and marine areas (approximately 93,600 km²), resulting in approximately 116,984 valid land pixels at 1 km resolution. Detailed parameter settings and model accuracy verification are provided in Supplementary Appendix A1.

3 Results

3.1 The current situation of land use change and transfer

This study utilized land cover data from the YRDUA for the period 2000–2020. The analysis was structured around five-year intervals, specifically 2000–2005, 2005–2010, 2010–2015, and 2015–2020. For each interval, the land use transitions were calculated, the transfer matrices in 2000–2005 and 2015–2020 are presented in Tables 4, 5 respectively. The remaining results can be seen in Supplementary Appendix A2.

TABLE 4. Land type transfer matrix of YRDUA from 2000 to 2005 (unit: km²).

TABLE 5. Land type transfer matrix of YRDUA from 2015 to 2020 (unit: km2).

Between 2000 and 2005, the total cropland area in the YRDUA decreased from 110,517.60 km² to 107,185.69 km², representing a net loss of 3,331.91 km². During this period, 3,163.90 km² of cropland were converted to construction land, 655.38 km² to water, and 223.34 km² to forest land, with minor conversions to grassland and unutilized land. As the second largest land cover type, forest land experienced a total transition of 491.91 km², primarily to cropland (222.08 km²) and construction land (166.98 km²), while the remainder was converted to grassland, water, and unutilized land.

During the period from 2015 to 2020, land use types in the YRDUA continued to undergo significant changes. Cropland area in the region decreased consistently, from 99,435.39 km² in 2015 to 97,692.92 km² in 2020. In contrast, construction land area showed an opposite trend, increasing by 1,756.08 km² due to urban expansion. Meanwhile, with the growing emphasis on ecological conservation and the implementation of relevant policies, forest land and grassland areas increased by 62.26 km² and 219.47 km², respectively. The expansion of these land types helps enhance the provision of fundamental ecosystem services in the region, thereby contributing to alleviating ecosystem service supply pressure.

Figure 3 illustrates the conversion trends among the six major land types in the YRDUA from 2000 to 2020. The results indicate that the conversion of cropland to construction land was the most prominent change across all five-year intervals, with large-scale encroachment on cropland, particularly during 2005–2010, posing a serious threat to the stability of regional grain production services. Furthermore, frequent mutual conversions were observed between cropland and water. Although conversions of forest land and grassland to construction land also occurred, their scale was relatively limited. Overall, sustained economic activity, population growth, and urbanization are associated with the continuous expansion of construction land at the expense of cropland, forest land, and water. This process not only disrupts regional ecological balance but also directly undermines the capacity of natural ecosystems such as forests and grasslands to provide critical services. As these fundamental ecological land covers are progressively eroded, the ecological condition of the region faces significant risks, and the supply level of its ecosystem services is projected to decline markedly, thereby further intensifying ecosystem service supply pressure in the region.

3.2 Analysis of key driving factors of land type changes

To examine the differential impacts of various driving factors on distinct land types, a driving factor analysis was separately conducted for each of the six land types using Random Forest feature importance analysis from the PLUS-LEAS module. As shown in Figure 4, the influence of drivers varies significantly across different land categories: Changes in cropland are jointly associated with slope, GDP, and population. Topographical constraints act as the primary limiting factor, while economic and demographic factors indirectly shape its spatial pattern through grain demand and circulation efficiency, the latter being influenced by proximity to major roads and railways. Changes in forest land are primarily associated with natural factors such as annual rainfall and DEM, consistent with its concentration in the mountainous areas of Zhejiang and Anhui Province, though human activities also exert a notable influence. Changes in grassland are most strongly linked to GDP, indicating that the dynamics of the relatively scarce grassland resources in this region area are primarily associated with human activities like economic development and urban greening. Changes in water areas are mainly driven by natural processes like rainfall and evapotranspiration, with anthropogenic factors playing a comparatively weaker role. The expansion of construction land shows the strongest association with the population factor, directly reflecting how urban population agglomeration may contribute to land urbanization. The distribution of unutilized land is highly correlated with annual potential evapotranspiration, with other natural and proximity factors also contributing, highlighting the inherent environmental characteristics of this vegetation-scarce land type.

3.3 Multi-scenario simulation of changes in ecosystem services

3.3.1 Current situation of ecosystem services

This study assessed the water conservation, carbon storage, soil conservation, and crop production in the YRDUA. The spatial distribution, at a 1 km pixel resolution, is shown in Figure 5, and the ecosystem service supply levels for the 26 cities are summarized in Table 6.

TABLE 6. Ecosystem service status of 26 cities in the YRDUA (2010–2020).

In terms of water conservation, the maximum pixel-based water yield was 2345.1 mm in 2010 and decreased to 1995.3 mm in 2020. Spatially, pixels with high water yield became significantly more concentrated around Taihu Lake in 2020, primarily due to the expansion of construction land areas in cities like Shanghai, Suzhou, and Wuxi to meet urban development and population needs. Additionally, pixels with high water yield capacity were mainly located in southern Anhui Province and western Zhejiang Province, where forest ecosystems play a role in storing and conserving natural rainfall and surface water resources. In terms of physical volume, Hangzhou had the highest total water conservation volume, 135.48 × 108 m³ in 2010, followed by Jinhua and Taizhou. Zhoushan had the lowest, 1.95 × 10⁸ m³, due to its limited land area. Cities with high total water conservation volume were primarily in Anhui and Zhejiang provinces, where larger administrative areas and higher forest coverage contribute to greater aggregate water yield, while Jiangsu and Shanghai, with less forest and grassland coverage, lagged significantly. By 2020, Hangzhou remained the highest, 112.54 × 10⁸ m³, followed by Xuancheng and Anqing. Overall, total water conservation volume in the YRDUA decreased compared to 2010, largely due to urban expansion encroaching on vegetated areas.

For carbon storage, the spatial concentration of carbon storage pixels showed little change across the three time points. High-value pixels were consistently concentrated in the forested mountainous areas of southern Anhui and western Zhejiang, as forests and grasslands are primary natural carbon sinks. The maximum pixel value remained 1192.4 t/pixel for all 3 years. Pixels with significant carbon storage changes were mainly found around Shanghai, Suzhou, Hangzhou Bay, and Nanjing, resulting from urbanization replacing natural ecosystems. In terms of physical supply, Hangzhou and Xuancheng had the highest total carbon storage volumes among the 26 cities in both 2010 and 2020, largely due to their larger land areas and extensive forest coverage. A comparative analysis from 2010 to 2020 revealed a continuous decline in carbon storage in cities such as Yangzhou, Wuxi, Zhenjiang, Shanghai, Hangzhou, Jinhua, and Shaoxing. Shanghai experienced the largest decrease, 7.67 × 10⁵ t, reflecting the impact of economic and urban development on ecosystem services.

Regarding soil conservation, the soil conservation service calculated by the InVEST SDR module, higher soil conservation values indicate that the ecosystem retains more soil, i.e., avoids more soil loss, reflecting stronger service provision capacity. The maximum pixel value increased from 15,162.4 t/pixel in 2010 to 232,544.4 t/pixel in 2020, reflecting both enhanced soil retention capacity and increased potential soil erosion pressure due to changes in rainfall and land-use conditions. Pixels with high soil conservation capacity were concentrated in the mountainous areas of southern Anhui and western Zhejiang, where forest vegetation plays a crucial role in preventing soil erosion. In 2010, Hangzhou had the highest total soil conservation volume, 53.79 × 10⁶ t, followed by Anqing and Jinhua. By 2020, the spatial pattern remained similar, but the physical supply of soil conservation services increased across most cities due to rising potential soil erosion.

For crop production, the maximum potential yield per pixel remained relatively stable during the study period, peaking at 51.8 t/pixel in 2020. Pixels with high production potential were concentrated in the central part of the urban agglomeration, along the Yangtze River and Taihu Lake water networks, with particularly dense distribution in the Subei Plain. In contrast, pixels in Zhejiang Province were more scattered due to limited arable land. Overall crop production in 2015 was lower than in 2010 and 2020. In 2020, Yancheng had the highest total crop production, 39.39 × 10⁵ t, reflecting its extensive arable land in the Subei Plain, followed by Chuzhou, 30.72 × 10⁵ t. Zhoushan had the lowest, 1.15 × 10⁵ t, while Tongling and Wuxi also had relatively low supplies.

3.3.2 Multi-scenario prediction results of land use change

Based on 2020 land cover data from the YRDUA, the accuracy of the PLUS model for multi-scenario forecasting was validated (Supplementary Appendix A1). Building on the analysis of driving factors, land use changes under three scenarios—baseline (current situation), economic development priority, and ecological protection priority—were simulated for the years 2030 and 2060. The spatial distribution of the simulation results is shown in Figure 6, and a comparative area statistic by land use type is provided in Table 7. The previously established model for quantifying the supply of ecosystem services was applied to predict changes under these multiple scenarios.

TABLE 7. Projected area of land-use types under different scenarios in the YRDUA, unit: km².

Comparison with 2020 land use data shows certain deviations in the simulation results: cropland area is overestimated by 9.19 km², forest land is underestimated by 12 km², construction land area is overestimated by 5.57 km², and unutilized land is significantly reduced by 90.51 km². The primary prediction errors lie in the estimation of water bodies and grassland, with grassland underestimated by 406.91 km² and water bodies underestimated by 368.42 km² compared to the actual data.

Under the baseline scenario, urbanization and expansion within the YRDUA are the main drivers of reductions in other land types, with cropland experiencing the most significant loss. By 2030, construction land area is projected to increase by 3,039.37 km² compared to 2020, while cropland decreases by 3,034.43 km², forest land by 271.83 km², and water bodies by 32.79 km². Unutilized land remains relatively stable, increasing slightly by 2.28 km². The degradation of some forest areas leads to their conversion into grassland, which has a lower ecosystem service capacity, resulting in an increase of 326.34 km² in grassland area. By 2060, the expansion of construction land areas intensifies further, increasing by 9,153.89 km² compared to 2020, while cropland, forest land, and grassland decrease by 8,026.64 km², 655.50 km², and 340.49 km², respectively.

Under the economic development priority scenario, the rate of urbanization and expansion exceeds that of the natural development scenario. By 2030, construction land area increases by 3,354.47 km² compared to 2020, while cropland, forest land, and grassland decrease by 3,214.43 km², 49.62 km², and 194.42 km², respectively. By 2060, urban expansion further encroaches on other land types, resulting in a total increase of 9,378.89 km² in construction land area and decreases of 9,775.43 km² in cropland and 153.92 km² in grassland compared to 2020.

Under the ecological protection priority scenario, the implementation of protection policies for natural land types such as forests and grasslands effectively restricts urban encroachment and development, yielding significant ecological benefits. By 2030, the expansion of construction land area is considerably slower than in other scenarios, increasing by only 2,859.47 km² compared to 2020. Forest and grassland areas increase by 227.67 km² and 53.71 km², respectively, while cropland decreases by 3,079.43 km². By 2060, the total construction land area increases by 8,748.89 km², cropland decreases by 9,595.43 km², and forest and grassland increase by 306.96 km² and 95.11 km², respectively.

3.3.3 Prediction results of ecosystem services under different scenarios

Based on the simulated land use changes under various scenarios in the YRDUA, the physical supply of four ecosystem services was calculated using the previously established quantification method. This allowed for further analysis of the changing characteristics of ecosystem service provision in the region.

The results of the physical water conservation volume by prefectural city are shown in Table 8. The data indicate that in 2030, the total water conservation service under the natural development scenario is 53.5176 billion m³, compared to 51.3119 billion m³ under the economic development priority scenario and 57.6333 billion m³ under the ecological protection priority scenario. Compared to the historical baseline year of 2020, the region’s water conservation capacity shows varying degrees of decline under all three 2030 scenarios. This reduction is attributed to the expansion of urban built-up areas and the encroachment on forest land, which severely impairs the water conservation capacity of the original landscapes. For the 2030 and 2060 projections, climate variables (precipitation, temperature, evapotranspiration) are held constant at 2020 levels, and only land-use patterns change based on PLUS scenario projections. Thus, the projections isolate the effect of land-use pattern changes on ecosystem services under current climate conditions. However, by 2060, simulation results suggest an improvement in ecological condition and that water yield may surpass 2020 levels. It should be noted that an increase in water yield (annual water production) does not necessarily indicate improved water conservation service; it may reflect increased surface runoff due to expanded impervious surfaces, rather than enhanced ecological water storage capacity. Cities with significant forest coverage, such as Xuancheng, Hangzhou, and Anqing, exhibit notable variations in water conservation volumes across scenarios, with values under the ecological protection priority scenario being significantly higher than those under the economic development priority scenario. The multi-scenario simulations suggest that ecological protection policies and curbing excessive construction and environmental degradation in the YRDUA may contribute to ecological restoration and enhance the region’s ecosystem service capacity.

TABLE 8. Prediction results of water yield in the YRDUA unit: 10⁸ m³.

Table 9 presents the simulated physical supply of carbon storage services for cities in the YRDUA under three development scenarios. Similar to the water conservation results, the region’s carbon storage capacity in 2030 declines compared to the 2020 level across all scenarios. This reduction is attributed to the significant loss of vegetation, particularly forests and grasslands, key land cover types for carbon storage, due to continued expansion of built-up areas, which impairs the ecosystem’s ability to absorb and store carbon dioxide. In the economic development priority scenario, total carbon storage decreases by 60.78 × 10⁶ t; under the baseline scenario, it falls by 40.60 ×10⁶ t; while the smallest reduction of 24.64 × 10⁶ t occurs under the ecological protection scenario.

TABLE 9. Prediction results of carbon storage in the YRDUA unit: 10⁶ t.

Unlike the trend observed for water conservation services, the region’s carbon storage capacity does not recover over time. Damage to forest vegetation may be difficult to reverse, leading to a further weakening of overall carbon storage by 2060 under all three scenarios compared to both 2020 and 2030 levels. However, the trends in carbon storage capacity among cities vary significantly across scenarios. Under the ecological protection scenario, cities with stronger ecological foundations, such as Xuancheng, Anqing, Chizhou, and Hangzhou, are able to restore their carbon storage capacity to the level projected under the 2030 natural development scenario, or even exceed it in terms of physical supply.

Ecosystem soil conservation service is also directly linked to the type and density of surface vegetation cover. The simulation results for the YRDUA under the three development scenarios are shown in Table 10. Similar to the carbon storage simulations, the soil conservation capacity of the ecosystem exhibits a declining trend across all cities in the region due to the destruction of forest vegetation, with Jiangsu Province and Shanghai, areas with relatively poor forest resources, experiencing the most pronounced declines. For cities in Anhui and Zhejiang provinces with a higher proportion of mountainous forests, the decrease in the physical supply of soil conservation service under the economic development priority scenario is significantly greater than under the other two scenarios. By 2060, the decline in soil conservation becomes even more severe across all regions. However, under the ecological protection priority scenario, the decrease in soil retention capacity can be mitigated, and the supply provided by the ecosystem is significantly higher than under the economic development priority scenario.

TABLE 10. Prediction results of soil conservation in the YRDUA unit: 10⁶ t.

In the multi-scenario simulations of land use change, cropland patches show significant reduction across all three scenarios compared to 2020. Besides forests and grasslands, urban expansion primarily encroaches on cropland, leading to a decline in the region’s ecosystem service of crop production. The simulation results confirm this trend. As shown in Table 11, the physical crop production in the YRDUA decreases markedly under all three scenarios. Compared to 2020 levels, the total crop production under the natural development scenario decreases by 46.21 × 10⁵ t in 2030, while the economic development priority scenario results in a reduction of 66.07 × 10⁵ t, and the ecological protection priority scenario shows a smaller decrease of 32.16 × 10⁵ t. By 2060, the ecosystem’s crop production capacity further weakens. Under the baseline scenario, the region’s crop production decreases by 77.28 × 10⁵ t compared to 2020. Without ecological protection policies, accelerated urban expansion and further loss of cropland would lead to a reduction of 99.89 × 10⁵ t by 2060. However, by restricting the expansion of built-up areas and protecting natural land covers such as forests and grasslands, this declining trend can be mitigated. Under the ecological priority scenario, the total crop production in 2060 decreases by only 70.17 × 10⁵ t compared to 2020, performing better than the baseline scenario.

TABLE 11. Prediction results of crop production in the YRDUA unit: 10⁵ t.

4 Discussion

4.1 Urban expansion and regional ecological imbalance risks

From 2000 to 2020, the YRDUA witnessed a nonlinear, tightly coupled relationship between urban expansion and land use change. Continuous construction land expansion is associated with sharp cropland loss. The pattern of ‘urban encroachment on cropland’ was primarily associated with regional land use change throughout the study period. Later ecological policies restored small parts of forest and grassland to build ecological security barriers, but the scale and speed failed to offset cropland loss and habitat degradation caused by urbanization inertia. This unbalanced land competition reflects path dependence of the traditional growth model relying on land finance and spatial expansion.

From a landscape ecology perspective, this process reshaped regional landscape patterns and degraded ecological functions. It aggravated natural landscape fragmentation, reduced landscape connectivity, and strengthened construction land dominance, forming a sprawling urban spatial structure. Such changes broke ecological process integrity, blocked species migration, and disrupted water and nutrient cycles, posing long-term risks to regional ecological condition. This shift is associated with socioeconomic needs, infrastructure layout and natural conditions, reflecting trade-offs between economic growth, food security and ecological protection. Urban-driven land change is associated with multi-dimensional ecosystem service declines. Core cities saw declining water conservation and carbon storage; ecologically fragile areas faced higher soil erosion risks. Service supply concentrated in peripheral ecological zones, while core metropolitan areas experienced the most pronounced service declines, creating spatial pressure on ecosystem service supply. It should be noted that inter-city comparisons in this study are based on total ecosystem service volumes within each administrative boundary, reflecting each city’s aggregate contribution to regional supply an approach relevant to cross-jurisdictional ecological compensation and regional ecological security planning. Total-volume comparisons do not directly indicate ecological quality or service provision intensity per unit area. Coordinated multi-objective management of ecosystem services has become critical for regional sustainable development under the Dual Carbon goals.

4.2 Influencing factors and mechanisms of land use change

Land use transitions in the YRDUA reflect complex human-land interactions under rapid urbanization. Driving forces differ distinctly across land types. Cropland change is limited by slope but mainly associated with GDP growth and population agglomeration. Forest distribution is strongly associated with annual rainfall and topography, yet pressured by infrastructure expansion. Grassland dynamics correlate strongly with GDP, while construction land expansion links closely to population. This significant divergence in driving forces suggests that natural geographical elements form a “rigid” constraint framework for regional development, while human activities exert “steering” and “intensifying” effects upon this natural baseline.

This significant divergence in driving forces demonstrates that natural geographical elements form a rigid constraint framework for regional development, while human activities exert “steering” and “intensifying” effects upon this natural baseline. Through nonlinear interactions, these elements collectively shape the regional landscape pattern. Natural factors establish the possibility boundaries for land use conversion, while socioeconomic factors determine the direction and intensity of conversion by altering resource demands and circulation efficiency. This complex driving mechanism indicates that the land use system in urban agglomerations represents a typical multiple-stress system, whose evolutionary trajectory is co-determined by the continuous feedback between natural baseline conditions and socioeconomic processes.

4.3 Urban agglomerations ecological risk prevention based on multi-scenario prediction

Multi-scenario simulation of land use change and ecosystem services provides critical scientific support for ecological governance in the YRDUA. This study applied the PLUS model to project land use patterns and corresponding ecological dynamics for 2030 and 2060 under three targeted scenarios: natural development, economic priority, and ecological priority. Simulation results reveal sharp divergence in ecological outcomes. By 2060, the natural development and economic priority scenarios are projected to drive construction land expansion to approximately 9,150 km², triggering a loss of roughly 9,000 km² of cropland and widespread degradation of key ecosystem services, especially substantial declines in carbon storage. In contrast, the ecological priority scenario restrains construction land to 8,750 km², effectively protects and slightly expands forest and grassland areas, and significantly mitigates ecosystem service degradation. These findings show that while ecological policies cannot entirely prevent degradation during urbanization, they may contribute to mitigating its impacts, especially on carbon sequestration and soil retention. The study further reveals that ecosystem services exhibit varying sensitivities to land use changes, with carbon storage showing limited recovery potential once disturbed. Model underestimation of water bodies and grasslands highlights the need for improved parameterization of hydrological and vegetation processes. Future research should prioritize next-generation models that integrate multi-agent systems and deep learning to better simulate human-environment interactions. Developing quantitative frameworks for assessing cross-scale ecosystem service trade-offs, coupled with predictive early-warning systems, can enable adaptive spatial governance through dynamic ecological protection lines—supporting sustainable urban development.

4.4 Limitations and prospects

This study has several limitations. First, the 1 km resolution remote sensing data restricts land use simulation accuracy, especially for grassland and water bodies; interpolation uncertainties of meteorological data also affect assessment precision. Second, inter-city comparisons are based on total ecosystem service volumes within administrative boundaries, which are inherently influenced by city area size; higher total values should not be interpreted as stronger per-unit-area service provision capacity. Others, the FoM value of 0.144 suggests that the precise spatial location of future land-use changes carries some uncertainty, and the 2030 and 2060 projections should be interpreted as scenario-dependent trend projections rather than precise spatial predictions. Future projections do not consider climate change impacts, reflecting only the isolated effect of land-use pattern changes; actual future ecosystem services will also be influenced by climate change, CO₂ fertilization, and other factors. The crop production assessment did not differentiate among crop types (e.g., rice, wheat, maize) and relied on comprehensive NDVI for total grain production spatial allocation, which may introduce uncertainty in areas with diverse cropping systems.

Future research will address these gaps: expand to other urban agglomerations for cross-regional comparison; integrate longer-term data to deepen trend analysis; develop hybrid models coupling process-based methods and machine learning; optimize ecological project evaluation and spatial governance strategies, to support high-quality urban agglomeration development.

5 Conclusion

Based on a systematic analysis of land use dynamics (2000–2020) and ecosystem service assessment (2010–2020) in the YRDUA, core findings can be summarized as follows: (1) Land use change exhibited a clear pattern of cropland conversion to construction land. During 2000–2005, cropland experienced a net loss of 3,331.91 km², with 3,163.90 km² converted to construction land, marking the dominant conversion path. This trend persisted through 2020, while construction land expanded by 1,756.08 km² in 2015–2020 alone. (2) The drivers of land use change varied significantly across land types. Cropland changes were jointly influenced by slope and socioeconomic factors, while construction land expansion correlated most strongly with population. Forest distribution is primarily associated with natural factors including annual rainfall and DEM. (3) Multi-scenario projections revealed substantial ecological risks and mitigation potential. Continuing current trends would lead to construction land expansion exceeding 9,000 km² by 2060, associated with substantial carbon storage decline, e.g., 60.78 × 10⁶ t decrease by 2030 under economic priority scenario. However, compared with the economic priority scenario (60.78 × 10⁶ t decrease by 2030), the ecological priority scenario reduces carbon storage loss to 24.64 × 10⁶ t, representing a reduction of approximately 60%, while maintaining higher water supply capacity. These findings underscore the necessity of a diagnosis and simulation framework that can inform spatial governance and ecological regulation decisions, enable dynamic optimization of ecological security patterns while balance development needs in urban agglomerations.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fenvs.2026.1908887/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Merja H Tölle, University of Kassel, Germany

Reviewed by: Quanzhi Li, China University of Mining and Technology, Beijing, China

Jianyu Liao, Hunan University of Technology, China