Revealing urban–rural sustainability dynamics based on multi-source data and machine learning: spatio-temporal hotspots, driving mechanisms, and non-linear thresholds in Suzhou, China

Rapid urban expansion has transformed rural land use, generating multiple sustainability challenges, including environmental pollution, habitat degradation, and crises of community identity. However, existing studies lack a quantitative framework that captures the multidimensional coupling among economic vitality, ecological integrity, and social wellbeing during urban–rural transitions, limiting guidance for integrating urban and rural landscapes. To respond to this issue, this study introduces an integrated Urban-Rural Sustainability Index (URSI) that encompasses economic, ecological, and social dimensions. Additionally, it develops a methodological framework that combines a Random Effects Model (REM), an Extreme Gradient Boosting (XGBoost) model, and Shapley Additive Explanations (SHAP) to identify both linear and non-linear relationships, interaction effects, and threshold conditions, while ensuring predictive accuracy and interpretability. Using geospatial data from Suzhou, China, this study systematically examines the spatial-temporal dynamics of urban-rural integration and associated sustainable outcomes between 2012 and 2024. Three key findings were revealed: (1) Urban-Rural Sustainability Index (URSI)-based conflict exhibits significant spatial clustering, evolving from a monocentric hotspot into a polycentric, multi-nodal pattern as urban expansion proceeds; (2) REM results reveal heterogeneous effects across land use types: built-up area is positively associated with conflict, whereas farmland, water, and forest are negatively associated, indicating a buffering role of ecological and agricultural stocks; (3) XGBoost–SHAP analysis enhances these findings by identifying segmented effects and indicative operating ranges, including an inflection around a built-up area about 60%, diminishing marginal mitigation from water coverage around 15%, and conflict alleviation when contiguous farmland is maintained near 20%. Building upon these empirical insights, this study proposes targeted strategies to enhance urban-rural sustainability, including spatially differentiated land-use management, prioritising blue-green infrastructure in urban planning and balancing sustainability pressures within urban areas. The proposed framework and analytical methodology provide a transferable foundation for future research and practical urban-rural sustainability planning across diverse geographical contexts.

1 Introduction

Urbanisation is one of the primary drivers of land-use transformation in contemporary society, reshaping both urban and rural landscapes throughout the 20th century (Nuissl and Siedentop, 2021). Although urbanisation drives economic growth and infrastructure advancement, it simultaneously leads to extensive transitions from rural to urban land uses, inevitably posing profound challenges to the sustainable development of land-use systems, particularly in peri-urban areas (Fu et al., 2021). These transitional zones, situated between urban and rural systems, integrate land, ecological resources, and human activities related to production and living. They are critical components of regional development (Zhang et al., 2020) and have significant impacts on human health and wellbeing (Barton, 2009). In the Chinese context, these processes unfold against a long-standing institutional separation between urban and rural areas, shaped by the household registration (hukou) system, differentiated land-use practices, economic structures and socio-cultural characteristics (Li, 2012). Rapid urbanisation has accelerated land-use transformations in peri-urban areas, shifting landscapes from primarily agricultural functions toward residential, industrial, and commercial uses. Development models focused on economic growth, to some extent, overlook the land-use conflicts resulting from such expansion and exacerbate their negative impacts (He et al., 2020).

Existing studies on urban–rural land-use transitions can be broadly categorised into two perspectives. The first perspective examines how land-use change in peri-urban areas reshapes economic, ecological and social outcomes. Empirical work has assessed the efficiency and effectiveness of land-use transformation (Fu et al., 2021), the ecological consequences of land-use changes and land-use intensity (Wu et al., 2024; Xu et al., 2020), and broader sustainability implications, including environmental degradation, habitat fragmentation and the loss of cultural landscape identity (Shao et al., 2020; Xu et al., 2020). International research further highlights that unsustainable land-use transformations in rural areas during urbanisation can lead to displacement (Salem et al., 2020), infrastructure deficits, and widening socioeconomic disparities (Dadashpoor and Ahani, 2019). Taken together, this perspective demonstrates that peri-urban land-use transitions are crucial to the sustainability of urban-rural regions and have a direct impact on human wellbeing. However, most studies address economic, ecological and social dimensions separately or treat them as loosely connected outcome indicators, paying limited attention to how these dimensions interact and co-evolve over time. As a result, the coupled dynamics of “coordination-conflict” between land-use transitions and multi-dimensional sustainability outcomes remain insufficiently understood, particularly in rapidly urbanising contexts such as China.

A second perspective focuses on governance arrangements, planning instruments and assessment frameworks for managing land-use change in urban-rural interfaces. Scholars have highlighted that current land-use transformation processes are often governed by fragmented and overlapping institutional regimes, leading to overlapping jurisdictions, inefficient decision-making and weak coordination across administrative boundaries (Perrin et al., 2018). Additionally, Follmann et al. (2021) and Ahani and Dadashpoor (2021) note that existing peri-urban land-use assessment and management frameworks fail to address the dynamic realities of urbanisation, creating loopholes for informal land speculation, unauthorised urban expansion, and environmental degradation. Although this line of research has significantly advanced the understanding of governance deficits and institutional complexity, it tends to emphasise qualitative institutional analysis or single-dimensional indicators. However, the foundation of sustainable development lies in the coordination between economic, social, and environmental factors, with the core being the balance between ecological considerations and other dimensions (Gan et al., 2020; Li et al., 2022). There is a lack of attention to operational, quantitative frameworks that can systematically link land-use structures to economic vitality, ecological quality, and social wellbeing, and that can identify when and how urban-rural systems shift from coordinated development to conflictual outcomes (Buzási and Csete, 2017; Keith et al., 2023). Consequently, despite a substantial body of work on land-use transformation, ecological impacts, planning and control mechanisms, a comprehensive evaluation framework that integrates these dimensions to support sustainable development remains lacking (Salem et al., 2025).

In addition to these theoretical gaps, there are notable limitations in the methodological approaches employed in the existing research. Currently, three categories of quantitative approaches are commonly used to evaluate urban–rural sustainability and land-use dynamics. The first category comprises statistical and spatial econometric models, including fixed-effects and random-effects models, as well as spatial panel specifications. The Spatial Durbin Model (SDM) has been widely used to identify spatial spillover effects and the impacts of neighbouring areas, serving as a key tool for evaluating the efficiency and driving factors of land use structure (Luge et al., 2025). These models offer clear parameter interpretations and a solid theoretical foundation for analysing directional relationships and spillover effects. However, their reliance on linearity and global stationarity limits their capacity to capture non-linear interactions, high-order effects and complex spatial heterogeneity, and model stability can be compromised when the underlying processes are strongly non-stationary.

The second category centres on geographically weighted regression (GWR) and its extensions, such as multi-scale regression methods (MGWR), which are used to address spatial non-stationarity by allowing coefficient estimates to vary across space (Wu et al., 2025). These approaches are effective for uncovering localised spatial mechanisms and scale-dependent relationships and have been used to explore the spatial heterogeneity of land-use drivers in urban-rural regions. Yet, they are less suited to handling high-dimensional variables and intricate interaction structures, and their performance often depends on subjective choices of bandwidths and kernel functions. As noted by Lessani and Li (2024), many extensions of GWR still have these limitations and thus only partially resolve the challenges of modelling complex, multi-dimensional land systems.

The third category encompasses predictive models based on remote sensing and machine learning, including random forests, support vector machines, XGBoost, and deep learning. These models have demonstrated strong capabilities in extracting complex spatiotemporal patterns from large-scale remote sensing and multi-source big data, and can achieve high predictive accuracy in land-use classification, change detection and risk assessment. However, they often suffer from limited interpretability and weak connections to established theoretical frameworks in land-use and sustainability research (Argota Sánchez-Vaquerizo and Helbing, 2025). Recent advances in explainable artificial intelligence (XAI), including SHAP and partial dependence plots (PDPs), have enhanced the interpretability of such models by quantifying global and local feature contributions and interactions, leading to composite frameworks such as XGBoost-SHAP. These approaches have helped to bridge, to some extent, the gap between “spatial processes” and “black-box predictions” (Alvarez-Garcia et al., 2025; Cui et al., 2025). Nonetheless, their application in urban-rural sustainability assessment remains fragmented, and XAI tools have rarely been embedded within systematic evaluation frameworks that integrate economic, ecological and social dimensions. Overall, existing approaches either prioritise statistical inference under restrictive assumptions or emphasise predictive performance at the expense of interpretability, with few attempts to integrate their respective strengths into a coherent analytical workflow.

Building on the above review, it is clear that existing research has made significant contributions to understanding the spatial patterns, structures, and efficiency of urban-rural land use, as well as the governance challenges of peri-urban transformation. However, several key gaps remain. First, at the conceptual and analytical levels, there is a lack of systematic exploration of the coupling mechanisms between land-use transitions and their impacts on economic vitality, ecological quality, and social wellbeing. This results in an absence of a quantitative basis for evaluating the dynamic transformation processes of “coordination-conflict” outcomes in urban-rural regions, and for identifying when development pathways remain sustainable or become problematic. Second, the non-linear threshold relationships and interaction effects among urban-rural land-use factors have not been clearly identified, making it difficult to determine how different land use structures and combinations influence sustainability outcomes, and under what critical conditions marginal changes in land-use configurations can trigger regime shifts. Third, at the methodological level, the integration of multi-source spatial data and machine learning models into urban-rural sustainability research frameworks remains underdeveloped. While single methods have been fruitfully applied, there is a lack of integrated workflows that can simultaneously capture multi-dimensional coupling, spatial heterogeneity and non-linear threshold effects, and translate these insights into policy-relevant indicators for sustainable land-use governance.

In response to these gaps and the associated practical needs, this study addresses the following core research question: How do land-use transitions in evolving urban-rural systems interact and co-evolve across the economic, ecological and social dimensions, and under what critical conditions do these processes shift from coordination to conflict? To answer this question, three interrelated research objectives are proposed: (1) To develop an Urban–Rural Sustainability Index (URSI) that integrates economic, ecological, and social dimensions, thereby quantifying the multi-dimensional coupling characteristics of urban-rural sustainability; (2) To employ the CRITIC method and the Coupling Coordination Model (CCM) to characterise the spatial coordination mechanisms between human activity intensity and ecological sensitivity, and to identify the linear effects of different land use types and structures on urban–rural sustainability; (3) To develop a quantitative analytical framework for urban–rural sustainability centred on “multi-dimensional coupling-nonlinear drivers-threshold identification”. Building on the interpretability of traditional panel models, the framework incorporates a Random Effects Model (REM) together with an XGBoost-SHAP approach to form a composite analytical pathway that combines statistical explanation, non-linear pattern recognition and segmented effect profiling.

Correspondingly, the study makes three main contributions. First, at the theoretical level, it presents an analytical framework that reveals the mechanisms of “multi-dimensional coupling and non-linear segmentation” within urban-rural systems. This framework extends existing discussions beyond single-dimensional linear effects to include cross-dimensional coupling and critical threshold mechanisms. Second, at the methodological level, it establishes a quantitative workflow that links statistical models (panel and random effects models, CCM) with machine learning models (XGBoost and SHAP), offering an operational technical route for identifying nonlinear relationships and threshold effects in complex land systems. Third, at the practical level, using Suzhou as an empirical case, the study identifies policy-relevant land use thresholds and key driving mechanisms. It provides measurable and testable evidence that supports spatial optimisation and sustainable governance in rapidly urbanising regions, with potential applications to other urban-rural contexts.

2 Materials and methods

2.1 Evaluation indicator system for urban-rural sustainable integration

Urban–rural land-use transformation is both a consequence of urbanisation and a critical driver of urban and rural development (Somanje et al., 2020). In this process, a dynamic system emerges as people, capital, resources, and land shift between urban and rural areas, with these factors continuously interacting over space and time (Zhao and Wan, 2021; Yan et al., 2018; He et al., 2017). Historically, this transformation, driven by rural labour migration, profit-seeking capital, and technology diffusion, has often increased urban-rural disparities by skewing the allocation of resources toward cities (Yang et al., 2021). Baffoe et al. (2021) argued that sustainable urban–rural land use should instead optimise the spatial distribution of capital, labour, technology, and industry, fostering balanced development on both sides of the urban–rural spectrum. Achieving this balance requires maintaining a dynamic equilibrium across economic, ecological, and social dimensions, thereby narrowing long-standing gaps between urban and rural areas.

To monitor progress toward this goal, it is necessary to develop a comprehensive framework for assessing the sustainability of land use in integrated urban and rural contexts. In this regard, the Urban-Rural Sustainability Index (URSI) serves as a crucial tool for assessing the sustainability of regional land-use patterns. The construction of the URSI in this study is based on the human–land system coupling and coordination theory (Long, 2022) and the three-dimensional framework of sustainable development (Un, 2015). This theory posits that sustainable development in urban and rural systems is the result of a dynamic balance between economic vitality, ecological carrying capacity, and social wellbeing, and should be understood as a continuously adjusting and evolving process (Yang et al., 2021). These three dimensions interact and constrain each other, collectively determining the level of coordination and sustainability in regional development. Economic vitality drives growth, ecological capacity ensures sustainable resource use, and social wellbeing ensures quality of life and social equity. Only through the coordination of these three components can sustainable urban-rural integration be achieved (Liu et al., 2020).

Building upon this theoretical foundation, this study defines the logical path for the URSI evaluation as follows: from the Sustainable Development Goals to System Dimensions and Indicator Factors. The sustainable development goal in this context is to achieve efficient resource allocation, ecological security, and improvements in living standards within the framework of urban-rural integration. This goal focuses not only on economic development but also on the balanced advancement of ecological protection and social equity. Therefore, based on the urban-rural sustainable development goal, the index is divided into three dimensions: economic, ecological, and social. The economic dimension reflects the development dynamics and competitiveness of urban and rural areas, the ecological dimension focuses on environmental quality, resource utilisation efficiency, and environmental resilience, while the social dimension measures social welfare, the accessibility of public services, and social equity (Liu, 2018; Long and Tu, 2018).

Therefore, from a measurable perspective, this study selects 17 multi-source data indicators that reflect the characteristics of the three dimensions to assess the URSI. The selection of indicators is guided by three principles. First, the theoretical relevance indicates that each indicator corresponds to a specific dimension and goal within the sustainable development framework. Second, regarding data availability, the indicators are based on publicly accessible or remotely sensed multi-source data, such as Points of Interest (POI), nightlight imagery, and Normalised Difference Vegetation Index (NDVI). This approach ensures repeatability and spatiotemporal comparability. Third, to address non-redundancy, the study employs the CRITIC method to evaluate the variability and correlation of the indicators. This ensures that each indicator in the system is independent and contributes valuable information. As shown in Table 1, each indicator is linked to economic, ecological or social dimensions, and to the overarching goal of urban-rural sustainable integration.

Table 1

Table 1. Evaluation indicator system of URSI.

The economic dimension of the URSI assesses the capacity for regional economic development and its sustainability, focusing on the transitions in land use between urban and rural areas. It is operationalised through four target layers, including economic vitality, industrial agglomeration, innovation and production, business mobility and the nighttime economy, a set of seven indicators (A1-A7). Economic vitality is measured by the presence of retail facilities, financial institutions, and food and beverage outlets. Existing studies highlight that retail facilities such as shopping malls, supermarkets and convenience stores reflect the vibrancy of urban and rural consumer markets and their responsiveness to demand for goods and services (Zhou and Li, 2020). Similarly, restaurants, cafés and related food and beverage outlets are indicative of the consumption-driven economy and its potential for growth (Zikirya et al., 2021). Financial institutions, such as banks, securities firms, insurance companies, and ATMs, represent the economic services infrastructure and financial accessibility, which are crucial for economic integration between urban and rural areas (Ma et al., 2020). Industrial agglomeration, innovation and production are captured through the distribution of industrial parks and office premises, encompassing high-tech industrial zones, development zones, corporate headquarters, and office buildings. These indicators jointly reflect the degree of industrial concentration, innovation capacity and production sustainability, as well as the extent of economic diversification and the ability of a region to adapt to evolving industrial structures (Hu and Han, 2019). Furthermore, accommodation facilities (e.g., hotels, guesthouses) are included as proxies for business mobility and the region’s capacity to support economic activities such as tourism and business conferences (Hidalgo, 2024). Finally, the nighttime economy is represented by nighttime facilities, such as night markets, bars, and 24-h convenience stores, complemented by night-time light data, which provide an additional layer of evidence on economic vibrancy after dark (Shi et al., 2020).

The ecological dimension of URSI assesses the sustainability of land use in terms of ecological infrastructure, biodiversity, and environmental governance capacity, using four indicators (B1-B4). Ecological infrastructure is represented by parks, green spaces, and bodies of water. These include urban parks, street greenery, wetland parks, lakes and rivers, which are key indicators for assessing the extent of urban greening and its role in improving air quality, biodiversity, and overall environmental health (Pan et al., 2023; Pan et al., 2024). Protected natural areas, including scenic and forest parks, are vital for maintaining ecological functions and providing ecosystem services like carbon sequestration and water filtration, reflecting the region’s biodiversity (Alvarez and Larkin, 2010). Additionally, environmental protection facilities, such as monitoring stations, waste sorting systems, and wastewater treatment plants, reflect the region’s ecological governance capacity and its efforts in pollution control and resource recycling. These facilities are integral to maintaining environmental quality in the face of urbanisation and industrialisation (Fu et al., 2021).

The social dimension of URSI encompasses three interrelated aspects of social sustainability: residents’ quality of life, public governance and social welfare, and the diversity of cultural and community life, represented by six indicators (C1-C6). Quality of life is primarily reflected in the accessibility of basic public service facilities, especially education and healthcare services, which directly affect residents’ everyday wellbeing and health security (Chen et al., 2022; Zhang, 2025). Social welfare facilities (e.g., community service centres, nursing homes and disability service stations) and public service facilities (e.g., government offices, police stations, and post offices) reflect the region’s social governance capacity and the depth of welfare provision (Yang et al., 2021). These facilities are core elements of social equity in the process of urban-rural integration (Saraswat et al., 2025). Furthermore, religious and cultural venues (e.g., churches, temples, and mosques) and cultural and recreational venues (e.g., libraries and cultural centres) are treated as key indicators of the diversity of cultural and community life (Ives and Kidwell, 2019; Yang et al., 2021). Instead of directly measuring religiosity, the presence and density of these venues are used as indicators of community organisations and shared spaces for interaction. These places serve as crucial venues for forming and maintaining social connections, as well as for expressing cultural and religious practices. Literature on social sustainability and social capital consistently highlights that these venues contribute to social cohesion, foster a sense of belonging, and promote inclusive community development (Sumi et al., 2025; Maselko et al., 2011). Therefore, they play a vital role in supporting the social foundations necessary for sustainable integration between urban and rural areas.

Unlike other studies on the sustainability of urban-rural land use, the URSI proposed in this research encompasses 17 evaluation indicators across the economic, ecological, and social dimensions, which can dynamically reflect the core mechanisms of urban-rural integration to some extent. Furthermore, this new assessment framework can serve as a systematic diagnostic tool to identify the strengths and weaknesses in the sustainability of urban-rural land use. The results derived from this assessment framework systematically reflect the overall elements and structural functions of urban-rural areas, providing policymakers with valuable decision-making support to promote sustainable development.

2.2 Research methods and procedure

Building on the foregoing review of literature on urban-rural land-use sustainability, this study develops a comprehensive framework that combines economic, ecological and social dimensions to evaluate the Urban-Rural Sustainability Index (URSI). Figure 1 illustrates the three sequential stages of the framework. First, recognising that urban-rural sustainability is multilayered, the 17 selected indicators are categorised into two composite components: (i) Human Activity, represented by economic facility (EMF) density, economic activity (EA) intensity, and social facility (SF) density, and (ii) Ecological Sensitivity, represented by ecological facility (EGF) density and ecological quality (EQ) index. To capture both cross-sectional and temporal dynamics, longitudinal data for Suzhou were compiled for five benchmark years, including 2012, 2015, 2018, 2021, and 2024. Second, sustainable-index assessment proceeds in two steps. The Criteria Importance Through Inter-criteria Correlation (CRITIC) method is first employed to assign time-specific weights to all indicators. The Coupling Coordination Model (CCM) is then applied to derive the coupling index and identify URSI-based conflicts. Subsequently, the Random Effects Model (REM) and an Extreme Gradient Boosting (XGBoost) model supplemented by SHapley Additive exPlanations (SHAP) are employed to rank variable importance, reveal variable interactions, and identify non-linear thresholds that shape URSI-based conflicts. This two-stage REM-XGBoost approach combines the interpretability of linear models with the non-linear discovery capability of machine learning, while SHAP visualisations quantify threshold values for individual variables, clarifying priority levers for sustainable urban-rural development under varying conditions. The final stage translates the integrated analytical results into targeted urban-rural sustainability strategies. These measures include spatially differentiated land-use controls, prioritisation of blue-green infrastructure, and actions to balance intra-urban development pressures.

Figure 1

Figure 1. Research framework.

2.3 Study area

This study selected Suzhou City in Jiangsu Province, China, as the research case. As of 2023, Suzhou administers six districts (Gusu, Wuzhong, Xiangcheng, Wujiang, High-tech, and Suzhou Industrial Park) and four county-level cities (Changshu, Zhangjiagang, Kunshan, and Taicang), with a permanent population exceeding 10.68 million and an urbanisation rate of 82.48% (Figure 2). Suzhou is a central city in the Yangtze River Delta and one of the earliest regions in China to undergo rapid urbanisation. It has long been characterised by high-intensity factor concentration and spatial expansion (Ma et al., 2007). Since the reform and opening-up, Suzhou has expanded through the integration of towns and the establishment of township development zones, leading to the reconstruction of its urban functions and spatial structure. Agricultural land has been continuously converted into residential, industrial, and service land, resulting in significant changes in vegetation cover and landscape patterns (Zhang et al., 2003). As a key node in China’s national urban agglomeration and metropolitan circle system, Suzhou has long endured the dual pressures of economic development and population aggregation, making it a representative case of the land use transformation and land use competition typically observed in urban-rural areas undergoing rapid urbanisation. Additionally, Suzhou encompasses a full spectrum of urban areas, ranging from historical urban centres and mature districts to industrial new towns and rapidly developing townships, as well as high-value ecological agricultural spaces. This comprehensive urban-suburban-rural gradient, combined with significant internal heterogeneity, presents an ideal case for studying the differentiated trajectories of land use and functional space across various stages of development. Moreover, with the recent emphasis on high-quality urban development in China, Suzhou is facing dual pressures from rigid ecological protection and farmland preservation policies, which are commonly seen in many rapidly urbanising regions of China. The simultaneous pursuit of economic development goals alongside environmental and resource conservation “red lines” is a policy combination increasingly prevalent in fast-growing urban regions. Therefore, the findings derived from Suzhou as a case study have broader applicability and can offer valuable insights for other cities facing similar challenges.

Figure 2

Figure 2. Boundaries and administrative divisions of the study area, Suzhou, China.

From the perspectives of urbanisation intensity, the complete urban-suburban-rural gradient, and multi-objective policy constraints, Suzhou represents a typical case of the urban-rural sustainable development issues emerging in China’s current rapid urbanisation process, thus demonstrating strong representativeness. Research on Suzhou helps identify the shortcomings in urban-rural land use transformation models and provides targeted recommendations to advance sustainable urban development goals, offering valuable insights for the development of other cities in China and beyond.

2.4 Methodologies

2.4.1 The criteria importance through intercriteria correlation (CRITIC) weighting method

The CRITIC weighting method is an objective weighting approach proposed by Diakoulaki et al. (1995). It quantifies the relative importance of indicators by considering their variability (contrast intensity) and interdependence (conflict). This method is particularly suited for Multi-Criteria Decision Analysis (MCDA), offering a systematic approach to objectively determine indicator weights through the following steps:

Step 1: Data Standardisation

The original data matrix $X={\left(x_{ij}\right)}_{m\times n}$ (with m samples and n indicators) is standardised using min-max normalisation, defined as follows:

For positive indicators (where higher values indicate better performance):

$r_{ij}=\frac{x_{ij}-\mathit{min}\left(x_j\right)}{\mathit{max}\left(x_j\right)-\mathit{min}\left(x_j\right)}$

For negative indicators (where lower values indicate better performance):

$r_{ij}=\frac{\mathit{max}\left(x_j\right)-x_{ij}}{\mathit{max}\left(x_j\right)-\mathit{min}\left(x_j\right)}$

Step 2: Calculation of Contrast Intensity (Standard Deviation)

The contrast intensity reflects the variability of each indicator and is computed as follows:

${\sigma }_j=\sqrt{\frac{1}{m}{\displaystyle \sum _{i=1}^m}{\left(r_{ij}-{\overline{r}}_j\right)}^2}$

Where ${\overline{r}}_j$ is the average value of indicator $j$.

Step 3: Determination of Conflict (Intercriteria Correlation)

Construct a correlation coefficient matrix $R={\left(r_{jk}\right)}_{n\times n}$, where $r_{jk}$ is the Pearson correlation coefficient between indicators $j$ and $k$. Higher conflict values imply lower redundancy among indicators, calculated as:

$f_j={\displaystyle \sum _{k=1}^n}\left(\left.\left.1-\right|r_{jk}\right|\right)$

Step 4: Computation of the Information Content ($C_j$)

The information content integrates both the variability and independence of indicators, defined as:

$C_j={\sigma }_j{\times f}_j$

Larger values of $C_j$ indicate a greater amount of unique information provided by the corresponding indicator.

Step 5: Determination of Weights ($w_j$)

Finally, indicator weights are computed by normalising the information content:

$w_j=\frac{C_j}{{\displaystyle {\sum }_{k=1}^n}{C}_k}$

2.4.2 The coupling coordination model (CCM)

Building upon the URSI assessment framework, two remote sensing data indicators, radiance-calibrated nightlight and Normalised Difference Vegetation Index (NDVI), were incorporated as supplementary measures in the construction of the Human Activity (HA) and Ecology Sensitivity (ES) indices, respectively. The Coupling Coordination Model (CCM) is adopted to measure the interactive relationship between HA and ES. Table 2 shows that the index evaluates HA using economic facilities, economic activities, and social facilities, while ecological facilities and quality are used for assessing ES.

Table 2

Table 2. Indicators composition of CCM.

$HA=w_1\times EMF+w_2\times EA+w_3\times SF$

$ES=w_1\times EGF+w_2\times EQ$

The coupling degree $C$ between HA and ES is calculated based on physics-derived coupling concepts:

$C=\sqrt[2]{\frac{HA\times ES}{{\left(HA+ES\right)}^2}}$

The coupling degree $C$ ranges between [0,1]. A value approaching 0 indicates imbalance and increasing deviation, whereas a value approaching 1 signifies stronger and improving coordination between the two systems. However, the coupling degree alone cannot effectively distinguish between scenarios of low-level yet highly coupled systems and those of high-level coordination. Therefore, this study employs the CCM indicator, adapted from relevant existing literature (Shen et al., 2023), which is calculated as follows:

$CCM=\sqrt{C\times T}$

$T=HA+ES$

2.4.3 Random effects model (REM)

To link the Urban–Rural Sustainability Index (URSI) and its coupling coordination measure (CCM) with land-use characteristics in a statistically robust manner, this study constructs a spatial panel dataset at the level of regular grid units. Specifically, the entire study area of Suzhou is divided into 1 km × 1 km grid cells, and for each grid cell, we calculate the URSI/CCM and associated land-use, ecological and socio-economic indicators for each year from 2012 to 2024. Using a grid-based design offers several advantages over traditional administrative units. First, it minimises the modifiable areal unit problem, which can skew data analysis. Second, it enables a more detailed identification of variations within districts and highlights localised conflict hotspots. Lastly, a grid-based approach remains consistent over time, making it essential for effectively analysing dynamic land-use changes.

In panel data analysis, to simultaneously account for the dual dimensions of time-series and cross-sectional individual characteristics, this study employs the Random Effects Model (REM) to examine the relationship between the Urban-Rural Sustainability Index and land use characteristics. Unlike the Fixed Effects Model, which assumes that individual effects are fixed constants, the REM posits that individual effects are not fixed but are random variables drawn from a larger population. This assumption enables the model to account for both temporal variation and random individual differences, making it particularly suitable for research settings with large sample sizes or when the goal is to capture variations within and between regions (Wooldridge, 2010).

The general form of the Random Effects Model used in this study can be expressed as:

$y_{it}=\alpha +\beta X_{it}+u_i+{\epsilon }_{it}$

where, $y_{it}$ represents the dependent variable (the CCM value derived from the URSI framework) for grid unit $i$ in year $t$; $x_{it}$ is the vector of explanatory variables for grid $i$ in year $t$, including land-use structure indicators (e.g., proportions of built-up land, farmland and water bodies), ecological, and socio-economic indicators; $\alpha$ is the constant term, $\beta$ is the vector of regression coefficients, reflecting the marginal impact of each explanatory variable on the CCM; $u_i$ represents the grid-specific random effect, assumed to follow an independent normal distribution $u_i\sim N\left(0,{\sigma }_u^2\right)$, capturing time-invariant unobserved heterogeneity across grid units; and ${\epsilon }_{it}$ is the random error term, assumed to follow ${\epsilon }_{it}\sim N\left(0,{\sigma }_{\epsilon }^2\right)$, representing time-varying shocks specific to unit $i$ and year $t$.

The core assumption of the REM is:

$\mathit{Cov}\left(X_{it},u_i\right)=0$

This implies that the unobserved individual-specific effects are uncorrelated with the explanatory variables. Under this assumption, the REM yields consistent and more efficient estimates than the fixed effects model, and allows the estimation of both within-unit (over time) and between-unit (across grid cells) variation, thereby supporting broader generalisation to similar urban–rural contexts. In practice, the model parameters are estimated using the Generalised Least Squares (GLS), which provides efficient estimates in the presence of heteroskedasticity and serial correlation in panel data. Furthermore, to test the appropriateness of the REM, a Hausman test is conducted to compare the results from the Fixed Effects and Random Effects Models, ensuring that the REM is suitable for the data at hand.

2.4.4 XGBoost-SHAP modelling

To systematically quantify the complex, non-linear, and interactive influences of land-use composition on the Urban–Rural Sustainability Index (URSI)-based conflict, this study employed the Extreme Gradient Boosting (XGBoost) modelling in conjunction with SHapley Additive exPlanations (SHAP). Compared to conventional linear models, XGBoost has demonstrated superior predictive performance by effectively capturing higher-order interactions and non-linear threshold effects between variables (Yang et al., 2025; Li and Miao, 2025). This methodological strength makes it well-suited for exploring the complex and multidimensional relationships associated with the research question in this study.

XGBoost iteratively constructs an ensemble of decision trees, optimized via a regularised objective function defined as:

$\mathcal{L}\left(\varphi \right)={\displaystyle \sum _{i=1}^n}l\left(y_i,{\widehat{y}}_i^{\left(t\right)}\right)+{\displaystyle \sum _{k=1}^t}\mathrm{\Omega }\left(f_k\right)$

with the regularisation term defined as:

$\mathrm{\Omega }\left(f_k\right)=\gamma T+\frac{1}{2}\lambda {‖\omega ‖}^2$

Here, $l\left(.\right)$ denotes the prediction loss function, $y_i$ and ${\widehat{y}}_i^{\left(t\right)}$ represent the actual and predicted values, respectively, for the $i-th$ observation at the $t-th$ iteration, $T$ is the number of leaves, and $\omega$ indicates leaf weights. Regularisation parameters $\left(\gamma ,\lambda \right)$ control model complexity, thereby enhancing generalization performance. By iteratively fitting residuals from previous stages, XGBoost incrementally refines its predictive accuracy.

In this study, the input variables consisted of areas from five distinct land-use categories: farmland, forest, water, built-up, and other land. The dependent variable was the CCM results evaluated based URSI framework. Model predictive performance was assessed by the coefficient of determination (R2) and root mean squared error (RMSE).

To address the interpretability challenge associated with complex machine learning models, SHAP was utilised. This method applies principles from cooperative game theory to provide a transparent decomposition of each individual prediction into feature-level contributions (Xi et al., 2024). Specifically, the SHAP explanation for the $i-th$ prediction can be represented as:

${\widehat{y}}_i={\varphi }_0+{\displaystyle \sum _{j=1}^M}{\varphi }_j^{\left(i\right)}$

Where ${\varphi }_0$ represents the baseline model prediction, and ${\varphi }_j^{\left(i\right)}$ is the marginal contribution (SHAP value) of feature $x_j$ to the prediction for observation $i$. By analysing the mean absolute SHAP values, the key global drivers of land use were identified. Additionally, SHAP dependence and interaction plots offered valuable insights into the non-linear marginal effects and interactions between different land-use variables.

2.5 Data sources

This study utilises multiple datasets for assessing the sustainability of urban-rural land use, including remote sensing data, publicly accessible datasets, the Gaode Map platform, and supplementary auxiliary data (Table 3). Data were collected at 3-year intervals, covering the years 2012, 2015, 2018, 2021 and 2024. Indicators for the economic domain were obtained from two sources. First, the Gaode (Amap) web-mapping API offered geocoded details on the distribution and density of commercial locations, financial institutions, transport hubs, and other aspects of the industrial structure. Second, radiometrically corrected Night-time Light Data (NLD) were obtained from the Harvard Dataverse, provided by Chen et al. (2020) and continuously updated through 2024. Night-time light intensity is widely recognised as a reliable proxy for urban economic vibrancy and built-up intensity (Lin et al., 2022). For indicators within the ecological dimension, land-cover characteristics were quantified using the 30 m annual China Land Cover Dataset (CLCD) developed by Huang and Yang (2025), which includes data from 1985 to 2024. This dataset classifies land into six primary categories: cropland, forest, grassland, water bodies, residential land, and unused land. Vegetation condition was measured using the 250-m resolution Normalised Difference Vegetation Index (NDVI), downloaded from the Resource and Environmental Science Data Platform and published by Xu (2024). For indicators within the social dimension, point-of-interest (POI) records retrieved via the Amap API (https://lbs.amap.com/) were used to assess public services. This database provides a consistent and high-density list of essential services, including medical, educational, cultural, entertainment, and other infrastructure facilities, enabling a systematic assessment of the spatial distribution of social infrastructure across urban-rural continua.

Table 3

Table 3. Data type and sources.

3 Results

3.1 Indicator analysis of the coupling coordination model: spatio-temporal characteristics of human activity and ecological sensitivity

In this study, two composite indices, Human Activity (HA) and Ecological Sensitivity (ES), were constructed to objectively reflect urban-rural land-use dynamics in Suzhou City during the period of urban expansion from 2012 to 2024. The HA index comprised three indicators: economic facilities density (EMF), economic activity intensity (EA), and social facilities density (SF), while the ES index included ecological facilities density (EGF) and ecological quality (EQ). The spatio-temporal distribution characteristics of these indicators during the study period are illustrated in Figure 3.

Figure 3

Figure 3. Spatio-temporal distribution of Human Activity and Ecological Sensitivity indicators in Suzhou (2012–2024).

Within the HA indicator system, the EMF showed a clear spatial clustering in 2012, primarily concentrated in the core urban area of Gusu District. This centralisation gradually expanded outward in subsequent years. From 2015 to 2018, a notable diffusion trend was observed, with the density of EMF in peripheral regions, such as county-level cities Kunshan and Changshu, experiencing significant increases. However, in 2021, despite the overall spatial configuration of the city remaining largely unchanged, EMF density experienced a marked decline. This reduction is likely attributable to the impact of the COVID-19 pandemic, aligning with findings from previous studies on Chinese cities during the pandemic (Han et al., 2024). By 2024, EMFs were more evenly distributed across Suzhou, forming a multi-centre structure. This trend reflects the city’s shift from a centralised economy development model to a more dispersed model, with multiple focal points across the region. Similarly, EA also exhibited a spatial expansion pattern. In 2012, the initial EA hotspots were primarily concentrated along the east-west axis, centred on Gusu District, Suzhou Industrial Park (SIP), and Kunshan County. From 2015 to 2021, these hotspots gradually extended towards the north-south axis, with further expansion to counties by 2024. This trend demonstrates a clear “core intensification-county expansion” characteristic. Moreover, the area of the EA hotspots significantly increased and began to merge into contiguous clusters, signalling a shift from a single-centre polarisation to a multi-centre structure, with spatial diffusion reaching the suburban areas. This finding supports previous findings from the Suzhou Industrial Park (Su et al., 2024) and further extends the understanding of economic development across the broader city region, reflecting the spatial extension of industrial diversification. In terms of SF density, in 2012, SF was initially concentrated within mature residential areas in central urban districts. Between 2015 and 2021, the number, density, and spatial coverage of SF steadily increased. By 2024, these facilities were spread across key urban areas, with notable improvements in SF density in the peripheral counties, indicating a trend toward greater spatial equity in the distribution of public services in Suzhou.

Regarding the ES system, the temporal and spatial evolution of EGF revealed a trend of increasing density from the periphery to the city centre. In 2012, the EGF were primarily concentrated in peripheral ecological protection zones, including areas surrounding Tai Lake, Kuncheng Lake in Changshu County, Lou River in Kunshan County, and the Grand Canal within Wujiang District. With the advancement of urban ecological infrastructure construction, the density of EGF within the city, especially in Gusu District, SIP, and Wujiang, increased significantly from 2015 to 2021. By 2024, a more cohesive ecological infrastructure network had emerged, strengthening the connectivity between urban ecological spaces and peripheral ecological areas, thus improving overall ecological resilience. In contrast, the EQ exhibited fluctuations between 2012 and 2021, primarily due to urban development and construction activities. Between 2015 and 2018, encroachment of agricultural land on water-based ecological areas, urban development on agricultural land, and fragmentation of ecological spaces due to land conversion led to a decline in the EQ within the central urban districts. However, since 2018, continuous urban ecological restoration efforts have led to significant improvements in the EQ of urban areas. Despite this, when comparing 2021 with 2015, the overall EQ showed a slight decline, supporting the conclusions of previous studies (Ben and Yu, 2022). By 2024, there was a significant improvement in EQ, especially in the central urban districts (Gusu, Xiangcheng, Wuzhong, Wujiang, SIP, and High-tech Zone), and a narrowing of the ecological quality gap between the central and peripheral regions (Kunshan, Changshu, Zhangjiagang, and Taicang counties).

Taken together, these indicators collectively reflect the multidimensional developments and transitions across economic, social, and ecological domains in Suzhou, laying the groundwork for establishing an Urban-Rural Sustainable Index (URSI). To objectively determine the influence weights of different indicators, the CRITIC method was utilised, effectively avoiding subjective bias by scientifically identifying key indicators. The calculated weights for each indicator fluctuated across the study periods, as summarised in Table 4.

Table 4

Table 4. Indicators’ weight allocation results are based on CRITIC.

Within the HA domain, EA contributes more than 4.5% to the total HA weighting, while EQ accounts for over 9.5% within the ES weighting system, marking them as the most influential indicators within their respective dimensions. Empirically, the HA index demonstrated a volatile yet upward trajectory, whereas the ecological sensitivity index remained relatively stable from 2012 to 2024, suggesting that Suzhou’s ecosystems retained a degree of stability and resilience amid urban–rural land transformations. These two-dimensional indicators were then introduced into the Coupling Coordination Model (CCM) to derive the URSI, which is subsequently used to examine the evolving coordination-conflict dynamics throughout the land transformation process.

3.2 Spatio-temporal dynamics of the urban-rural sustainable index (URSI) conflict based on the conflict and coordination model

This study employed the Coupling Coordination Model (CCM) to quantitatively evaluate the URSI for Suzhou from 2012 to 2024, revealing the spatial distribution and temporal evolution of urban-rural land-use conflict in the context of rapid urban expansion (Figure 4). Utilising Jenks’ natural breaks classification, URSI-based conflicts were categorised into five intensity levels: Low conflict (0.00<CCM<0.19), Relatively low conflict (0.20<CCM<0.28), Medium conflict (0.29<CCM<0.39), Relatively high conflict (0.40<CCM<0.53), and High conflict (0.54<CCM<1.19).

Figure 4

Figure 4. Spatial distribution and temporal evolution of URSI-based conflict in Suzhou.

In 2012, high-conflict regions were primarily concentrated within the urban core, particularly around the Gusu District, indicating a clear monocentric spatial pattern. By 2015, conflict hotspots had gradually diffused outward along major transportation corridors toward suburban areas, and new hotspots had emerged within secondary urban centres and newly developed regions, such as Kunshan, Changshu, and Zhangjiagang counties. This spatial diffusion aligns with the concept of urban-rural transition zones (Yang et al., 2021; Zhang, 2025), illustrating a shifting land-use conflict pattern driven by accelerated urban expansion (Zheng et al., 2022). By 2018, high-conflict areas (CCM >0.54) had notably increased in central urban areas, including Gusu, the High-tech zone, and SIP districts, as well as Kunshan County, with an expansion of 8.4% compared to 2015.

By 2021, medium-conflict areas (0.29<CCM<0.39) had become significantly more concentrated, exhibiting a 6.82% increase compared to 2018. In contrast, high-conflict areas (CCM>0.54) in Suzhou experienced a slight decrease in conflict intensity, with a reduction of 13.18%. In 2024, a multi-centric conflict pattern emerged, characterised by persistent hotspots within central urban districts and an intensified expansion of conflict hotspots in peripheral nodes and newly developed county areas. The emergence of multiple high-intensity conflict nodes underscores the spatial spillover effects of continuous urban development, driven by industrial relocation, population migration, and infrastructural expansion (Xu et al., 2020; Zhu and He, 2025; Pan et al., 2023). Overall, the evolution from a monocentric to polycentric conflict pattern in Suzhou illustrates the cumulative and diffusive nature of URSI-based conflicts, providing critical insights into the dynamic interplay between urban economic growth, spatial expansion and sustainable development. This highlights the necessity of adopting spatially differentiated, multi-nodal urban planning and ecological governance strategies.

3.3 Drivers of URSI-based conflict identified by ordinary least squares regression analysis

The Hausman test was conducted to compare the Fixed-effects (FE) estimator with the Random-effects (RE) estimator in order to select the appropriate model. As reported in Table 5, the test yields a statistic of −21,956.937 with 5 degrees of freedom and a p-value of 1.000 (p ≥ 0.05). The large p-value indicates that there is no statistically detectable difference between the FE and RE estimates. Hence, we do not reject the orthogonality assumption, implying that RE is consistent for our data. Conditional on this result, the RE model was selected for the main analysis due to its ability to achieve greater efficiency than the FE model, thereby improving the precision of the estimated effects.

Table 5

Table 5. Hausman.

Table 6 presents RE estimates for two specifications: URSI-based conflict level (Model 1) and its temporal change (Model 2). In the baseline model (Model 1), the Built-up area shows the largest positive coefficient among land-use categories (0.615***), indicating that more urbanised regions are associated with increased urban–rural conflict. This finding is consistent with agglomeration-driven land competition and pressure on shared resources (Zheng et al., 2022). By contrast, Farmland (−0.149***) and Water (−0.051***) present a significant negative sign, indicating that agricultural and ecological stocks are associated with lower levels of conflict, theoretically through buffering, livelihood support and ecological services (Yang et al., 2018). Forest also shows a small but significant negative association (−0.008***), suggesting a modest conflict-mitigating role at the urban–rural interface. Other land is weakly positive (0.004*). The overall fit is relatively high (R² = 0.510, N = 46,210), indicating that land-use composition explains a substantial portion of the variation in conflict intensity both cross-sectionally and temporally. Turning to Model 2, effect patterns are more complex. Built-up area change remains strongly positive (0.637***), indicating that urban expansion is a primary driver of worsening conflict over time. Water change is also positive and significant (0.242***), suggesting that short-term reconfiguration of blue-green systems, through construction or connectivity adjustment, can coincide with transitional URSI-based conflict intensification, even if steady-state stocks are conflict-reducing in their levels. In contrast, Farmland change is negative (−0.132***), implying that its expansion is associated with conflict alleviation, whereas the loss of farmland tends to exacerbate conflict. Forest change (0.052***) and Other land change (0.055***) are positive but small, consistent with modest increases in conflict when these categories expand. The lower fit (R² = 0.284, N = 10,318) reflects the expected results for the change model and highlights the impact of non-linear and temporal disturbances, which motivates the threshold analysis discussed later.

Table 6

Table 6. REM results.

This research re-estimates the models after excluding the top and bottom 5% of samples in the URSI-based conflict index (see Supplementary Table S1). The signs and significance remain consistent, confirming the robustness of the findings. Specifically, in the baseline model, Built-up area remains strongly positive (0.401***), Water and Forest remain negative (−0.037***, −0.009***). Farmland yields a weakly positive result (0.018*), but the magnitude is small and does not undermine the main argument that farmland is conflict-mitigating at both levels and in its changes. In the change model (Model 2), Built-up area change (0.249***) and Water change (0.160***) continue to be the principal positive drivers of conflict dynamics. The overall model fit (R² = 0.525 and 0.242) is similar to that of the main models, further verifying the stability of the results.

To further explore the impact of spatial heterogeneity, REM tests were conducted in nine districts and counties under the jurisdiction of Suzhou City, and the results are presented in Supplementary Tables S2, S3. Overall, district and county-specific random-effects estimates show consistent directions but heterogeneous magnitudes. In Model 1, Built-up area is the dominant positive correlate in most areas (e.g., Gusu = 1.300***, Wuzhong = 0.654***), reflecting stronger URSI-based conflict where urban concentration and development pressure are highest. Farmland is significantly negative in many areas (e.g., Huqiu = −0.370 ***, Changshu = −0.191 ***), indicating that agricultural land helps buffer urban-rural tensions. Forest and Water coefficients are predominantly negative, highlighting the conflict-reducing role of ecological stocks. In Model 2, positive effects of Built-up area change (e.g., Gusu = 1.396***) and Water change (e.g., Wuzhong = 0.433***) are pervasive, while the negative effect of Farmland change is particularly pronounced in Xiangcheng (−0.248***) and Taicang (−0.261***). The R² range of the model is between 0.13 and 0.71, reflecting the meaningful but place-specific heterogeneity in the urban-rural sustainability process.

These findings confirm that increasing urban development intensity remains the primary driver of escalating URSI-based conflict, echoing and expanding upon prior studies by Zheng et al. (2022) and Shen et al. (2023). Importantly, this research highlights that different types of land use, especially ecological and agricultural areas, play a significant role in reducing conflict levels. These land-use categories demonstrate a buffering effect and emphasise their value in balancing both conservation efforts and production needs. Moreover, in the short run, adjustments to water bodies may coincide with heightened conflict, which highlights the central role of blue-green infrastructure in urban-rural sustainability strategies and aligns with recent theoretical arguments that nature-based solutions is significant for mitigating urban ecological pressures (He et al., 2024). Collectively, these results reveal a dual driver of the URSI-based conflict, structural composition and temporal transformation, and provide an empirical basis for the subsequent non-linear interaction and segmented effect analysis.

3.4 Drivers of URSI-based conflict identified by XGBoost prediction

To further validate and enhance the results derived from REM, the XGBoost modelling and Shapley Additive exPlanations (SHAP) analysis were employed to explore non-linear thresholds and pairwise interaction effects among different land-use variables. Figure 5 illustrates the importance of variables and their contributions based on two model specifications: the cross-sectional model (Figure 5a) and the change model (Figure 5b). Specifically, the bar plots on the left display the ranked SHAP values, arranged from highest to lowest. These plots highlight the average contribution of each land-use variable to the model predictions. In contrast, the SHAP beeswarm plots on the right illustrate the distribution and direction of local impacts for individual observations.

Figure 5

Figure 5. Importance of the impact of each indicator on URSI-based conflict, (a) cross-sectional model, (b) change model.

In the cross-sectional model (Figure 5a), which demonstrates strong model performance (R² = 0.698, RMSE = 0.548), built-up area emerged as the dominant variable, with an average SHAP value of +0.61, more than three times greater than that of any other land-use category. Water bodies ranked second with a mean SHAP value of +0.18, followed by Farm (+0.15), Forest (+0.02), and Other land (+0.01), which exhibited minimal influence. This hierarchy of importance aligns closely with findings obtained from the REM, confirming the robustness of earlier REM results (i.e., β_built-up > β_water > β_farmland > β_forest). In the change model (Figure 5b), the explanatory power declined (R² = 0.311, RMSE = 0.819), indicating that recent variations in URSI-based conflicts were driven by a broader set of land-use changes rather than dominated by built-up land alone. Farmland change remained influential (mean SHAP = +0.21) but with reduced relative importance compared to the cross-sectional model, while built-up land change (mean SHAP = +0.19) gained prominence, reflecting its growing role in recent urban–rural land-use dynamics. Beeswarm plots further demonstrated that high built-up land values consistently increased conflict intensity (positive SHAP), whereas high farmland values consistently mitigated conflicts (negative SHAP). Water and forest maintained relatively modest but stable negative impacts on conflict, suggesting their limited yet consistent mitigating roles.

Figure 6 further clarifies the non-linear relationships and segmentation effects between land-use variables and URSI-based conflicts. Specifically, SHAP dependence plots (Figure 6a for the static model; Figure 6b for the change model) were employed to visualise how the marginal contribution of one variable, conditioned on the levels of another, impacts the Urban–Rural Sustainability Index (URSI)-based land-use conflict. In each SHAP dependence plot (Figure 6), the horizontal axis represents the actual values of the target variable, while the vertical axis indicates the corresponding SHAP values, thereby quantifying each variable’s positive or negative contribution to predicted land-use conflicts. Additionally, the colour gradient denotes the magnitude of another land-use variable, providing a direct visualisation of interactive effects.

Figure 6

Figure 6. Interaction effects of indicators on the URSI-based conflict, (a) results of the cross-sectional model; (b) results of the change model.

The cross-sectional model (Figure 6a) revealed a strong association between the URSI-based conflict and the proportion of built-up area, as well as significant segmentation effects between built-up land and other land-use categories. The built-up area shows a monotonic increase in SHAP values up to approximately 0.60, after which the curve flattens, indicating diminishing marginal contributions beyond this level. The steepest rise occurs in the 0.40–0.60 interval. Taken together, these patterns suggest an empirical operating threshold near 0.60 for planning attention, with particularly pronounced sensitivity as the built-up area moves through the 0.40–0.60 range. Interaction-coloured plots further reveal a turning point for the farmland area around a Z-score of approximately 0.40 (i.e., 0.40 standard deviations above its mean). Below this level, SHAP values rise sharply, indicating that low farmland availability fails to buffer urban pressure and amplifies conflict as the built-up area increases. Above a Z-score of approximately 0.40, the slope attenuates, implying that maintaining at least a moderate, contiguous farmland share mitigates conflict, an effect that is more pronounced in highly urbanised areas. By contrast, the water share curve is smooth and generally conflict-reducing, with diminishing returns as area increases. This profile is consistent with a stable but moderate buffering role: expanding water areas contribute to conflict alleviation, but the marginal benefit tapers as coverage grows.

The change model (Figure 6b) shows that year-to-year variation in the conflict index arises from the joint influence of land-use category, the rate of land-use change, and baseline urbanisation. Built-up expansion remains the dominant driver. The SHAP dependence curves display a V-shaped response around zero growth, with contributions being lowest near zero and rising as the growth rate departs from zero in either direction. The increase is steeper for expansion than for contraction. Within the intervals −0.25 ≤ built-up in z-standardised units ≤0, and 0 <built-up in z-standardised units≤ 0.50, SHAP values increase approximately linearly with the absolute growth rate, indicating that moderate acceleration in built-up change already elevates conflict risk. Changes in water area are asymmetric. Small losses have a limited effect, but larger losses exacerbate conflict. Moderate gains provide the most effective mitigation, particularly when paired with sufficient farmland, which enhances the buffering capabilities of the blue-green system. In contrast, significant increases in farmland (greater than 0.25 in z-standardised units) are linked to higher levels of conflict. This suggests that there are transitional frictions, such as reallocation, access issues, and compensation challenges, when agricultural land is rapidly reconfigured. Forest cover change shows relatively minor effects throughout its range, and significant negative impacts primarily arise during periods of rapid forest gains, aligning with challenges in transition management. Therefore, these findings highlight the need for context-sensitive land-use management, where sequencing is required, rather than relying on a single ecological mitigation measure.

These findings quantify model-implied marginal contributions, including the 60% limit for built-up areas, 15% coverage for water, and 20% preservation levels for farmland serve as referable targets for sustainable development and the integration of urban and rural areas. However, it is essential to note that the results derived from the integration of XGBoost and SHAP represent the segmental effects of features on model predictions, revealing nonlinear relationships and correlation thresholds. These results should not be interpreted as a causal relationship. Furthermore, the identified segmental effects may be influenced by geographical, social, and institutional contexts and should be validated against local planning practices.

4 Discussion

4.1 Policy implications based on the URSI assessment

In the 21st century, addressing the uneven and fragmented urban–rural development has become one of the most pressing challenges for China’s social and policy challenges (Yang et al., 2021). In this context, the effective integration of economic, ecological, and social resources between urban and rural areas is therefore a critical strategy for achieving sustainable regional development. The Urban-Rural Sustainability Index (URSI) developed in the study not only provides a comprehensive quantitative tool for identifying different patterns of coordination and conflict, but also offers a technical basis for formulating differentiated, place-based land-use management strategies.

Using Suzhou as a case study, the URSI-based conflict mapping reveals a clear evolution between 2012 and 2024: the spatial pattern of conflicts has shifted from a monocentric conflict hotspot to a polycentric, multi-nodal configuration. This spatial transformation reveals that as urban spillover persists and multi-core development advances, urban-rural conflicts are no longer confined to the traditional urban core. Instead, these conflicts are now appearing as overlapping and multi-scalar patterns across major urban districts, county-level sub-centres, and the surrounding rural areas. Based on the intensity of conflicts, the regions in Suzhou can be classified into three categories. First, high-conflict core urban areas, represented by Gusu, Wuzhong and High-Tech Zone. Second, county-level urban centres such as Kunshan, Changshu, and Zhangjiagang, which form medium-conflict peripheral sub-centres. Third, low-intensity rural areas and ecological core zones, represented by Wujiang District. Under such conditions, traditional control measures based solely on administrative boundaries or single indicators are insufficient to respond to heterogeneous regional challenges. By contrast, the URSI evaluation framework enables the construction of a results-oriented, tiered governance system, supporting a governance pathway that links problem identification, zoning, regulation, and monitoring.

Specifically, the proposed implementation pathway comprises four steps. First, a grid or cell-based diagnosis is conducted using the URSI-based conflict index to identify high-conflict, medium-conflict and coordinated areas, and to delineate the three types of regions in spatial terms. Second, for each type of region, differentiated land-use configuration intervals (e.g., upper limits for built-up land, target ranges for the proportion of farmland and water bodies) and flexible optimisation targets are defined based on the assessment results. Third, these configuration indicators and optimisation targets are embedded into control systems such as territorial spatial planning and land-use approval procedures, thereby forming an operational management process. Finally, by building on the existing URSI evaluation framework, key indicators can be updated regularly in light of development dynamics, enabling the establishment of a dynamic monitoring and early warning mechanism. When the conflict level in a given area intensifies beyond a predefined threshold, stricter land-use controls and ecological restoration measures can be triggered in a timely and targeted manner.

4.2 Implementation pathways for urban-rural sustainable development

Drawing upon the proposed URSI evaluation framework and empirical findings, targeted land-use management and planning measures can be designed along three main aspects to alleviate URSI-based conflicts.

4.2.1 Spatially differentiated land-use management strategies

Land-use control strategies can be differentiated according to the three spatial categories identified by the URSI assessment. First, in Category 1 high-conflict core urban areas, urbanisation is already mature, socio-economic structures are relatively stable, and the share of built-up land exceeds the critical threshold of around 60% revealed by the URSI analysis. Land use in these areas is characterised by high intensity and density, but also by pronounced spatial pressures and sustainability conflicts. In such contexts, further outward expansion of construction land is likely to exacerbate conflicts. Therefore, these areas should prioritise improving the quality and efficiency of the existing built environment as the main measure of conflict mitigation. In operational terms, additional development demand should, in principle, be met through urban renewal, redevelopment of underutilised land, and vertical intensification, while strictly controlling further outward expansion of urban construction land, an approach supported by the findings of Xu et al. (2020). Moreover, URSI impact assessments can be integrated into the planning and development approval processes. Major projects must demonstrate that the URSI-based conflict for the relevant unit does not increase after implementation. If there are negative impacts, these should be mitigated through measures such as demolishing low-efficiency buildings or increasing green and blue spaces. Differentiated land prices, taxation, and floor-area ratio policies can help discourage land-intensive, low-value activities while promoting the clustering of knowledge-intensive and innovation-driven industries. When these measures are paired with digital transformation initiatives, such as the digital upgrading of traditional manufacturing and data-driven production organisation, they can enhance economic output and social benefits for each unit of construction land (Kim and Feng, 2024).

Second, Category 2 county-level peripheral areas are characterised by rapid economic growth, strong pressure for urban expansion and the continued presence of agricultural and ecological spaces. For these areas, the primary task is to control the scale of construction land growth. The expansion of construction land exhibits diminishing marginal returns: once the scale of built-up land exceeds certain thresholds, further expansion may not generate additional economic benefits and can even lead to a decline in economic performance (Xu et al., 2020). Accordingly, these regions should focus on developing ecologically compatible industries and infrastructure, aligning industrial development models with local ecological conditions. The empirical analysis indicate that maintaining approximately 20% of the area as farmland and increasing the proportion of water bodies by 0%–15% can significantly mitigate URSI-based conflicts. In implementation terms, this implies that explicit targets for the preservation of farmland and water bodies should be incorporated as semi-binding requirements in spatial plans. At the same time, when new construction land is designated, land consolidation, land-use swaps and quota trading can be employed within or across jurisdictions to ensure that overall land-use composition remains within the “safe ranges” identified by the URSI. In addition, planning interventions in these areas should address the balanced and equitable spatial distribution of social infrastructure, such as education, healthcare and recreational facilities, in peri-urban neighbourhoods. Improving access to such services is an indispensable component of enhancing the URSI (Baffoe et al., 2021).

Third, Category 3 rural and ecological core areas (e.g., Wujiang District) are characterised by low land-use intensity and high ecological sensitivity, and they play a key role in safeguarding regional ecological security and ecosystem service provision. In these areas, new construction should be strictly controlled. Rural housing improvements and infrastructure provision should be achieved as far as possible through village consolidation, in-situ upgrading and the reuse of idle homesteads, rather than through the conversion of forests, water bodies or high-quality farmland. Remote sensing and long-term monitoring data can be used to establish ecological monitoring indicators, such as forest cover, farmland, water body areas, and composite ecological quality indices. These indicators can then be integrated with the URSI assessment to define early-warning thresholds. When the URSI or its ecological sub-component shows a sustained decline, ecological restoration programs and stricter development control measures should be implemented.

4.2.2 Prioritising blue-green infrastructure-oriented development pathways

The empirical analysis shows that increases in water bodies, forests and agricultural land have a significant positive effect in alleviating URSI-based conflicts. Prioritising blue-green infrastructure in urban-rural planning and development is therefore a key strategy for mitigating conflicts and enhancing resilience. However, the focus of interventions should differ across the three types of regions. In Category 1 core urban areas, blue-green infrastructure should be advanced primarily through nature-based solutions embedded in stock redevelopment. URSI-based conflict hotspot maps can effectively pinpoint high-conflict streets and blocks for prioritised blue-green retrofitting. In these units, interventions such as the creation of pocket parks, the installation of permeable pavements, the introduction of rainwater harvesting systems and the construction of green roofs should be implemented first, which is consistent with the findings of Chen et al. (2024). At the same time, blue-green infrastructure indicators should be incorporated into the planning conditions and approval criteria for urban renewal and regeneration projects, with explicit requirements that each project achieve a net increase in blue and green space within its footprint.

In Category 2 peripheral urban areas, the emphasis should be on constructing ecological networks. Operationally, rivers, canals, lakes and agricultural water systems can serve as the backbone for delineating riparian buffer zones and blue-green corridors in territorial spatial plans. Within these corridors, land use should be restricted to low-impact activities such as ecological conservation, recreation and eco-agriculture, thereby reducing the negative ecological impacts of outward urban expansion. In addition, URSI-based conflict results can be used to prioritise segments where conflicts are most intense and ecological fragmentation is most severe. These segments should be prioritised for restoration and connectivity projects, as modest investments can lead to significant reductions in land-use conflicts. In Category 3 rural and ecological core areas, blue-green infrastructure development should focus on ecological restoration and the conservation of traditional farmland and water systems. Restoring fragmented farmland, wetlands, and riparian zones should be a top priority. It's essential to re-establish traditional composite patterns of farmland, water systems, and woodlands to support agro-ecosystems that offer significant ecological benefits. Blue-green infrastructure projects in these areas should be closely integrated with rural revitalisation initiatives. Through eco-tourism, landscape agriculture, and similar activities, ecological restoration can also provide a new source of income for farmers and local governments.

4.2.3 Differentiated development pathways for factor flows and regional coordination

Urban-rural sustainable development relies not only on spatial land use structure but also on the organized movement of essential factors like land, capital, technology, and industry across various spatial units. Currently, factor flows between urban and rural areas in China are still dominated by the movement of population and land, whereas the flows of technology and capital remain relatively limited (Yang et al., 2021). To achieve sustainable urban-rural development, it is therefore necessary to coordinate factor allocation across areas with different land-use structures, ecological conditions and development stages, and to improve the rationality and complementarity of factor flows among them. In Category 1 core urban regions, pressures arising from intensive development should be alleviated through industrial restructuring and upgrading. Governments can use industrial guidance catalogues and differentiated land-leasing and pricing policies to promote the orderly relocation of landand energy-intensive industries to suitable Category 2 areas, while concentrating knowledge-intensive, high value-added industries and producer services in Category 1. At the same time, the URSI framework can be applied to assess redevelopment projects on existing industrial land, giving priority to transforming high-conflict industrial parcels into innovation districts, mixed-use communities or public service facilities, thereby improving both land-use efficiency and environmental quality. In Category 2 regions, development capacity envelopes should be clearly defined based on URSI assessment results and key land-use thresholds. Within these envelopes, peripheral centres can appropriately receive transferred capital, technology and industries from the core, under the condition that overall land-use configurations remain within sustainable ranges. Policy-guided coordination of factor flows between core and peripheral areas, and strengthened industrial linkages and collaboration along value chains, are important means of promoting sustainable development at the scale of the entire urban region (Su et al., 2024). Given that Category 3 areas function as providers of ecological assets and ecosystem services for the broader region, an ecological compensation mechanism should be established to ensure that a share of land value increments and fiscal revenues generated in Category 1 and 2 areas is reinvested in Category 3. These funds can be used to support ecological restoration, infrastructure improvement and the provision of basic public services. At the same time, policies should facilitate the movement of surplus labour from core urban areas to Category 3 regions where appropriate, for example, through the development of eco-agriculture, eco-tourism or decentralised service activities. This can help to prevent excessive depopulation and hollowing-out in Category 3 areas, which would otherwise undermine ecological management and the operation of local infrastructure.

Overall, the URSI evaluation framework developed in this study establishes a systematic governance pathway that spans conflict identification, spatial zoning, control measures and monitoring. It thus provides a replicable technical route and policy reference for optimising urban-rural spatial structures and advancing sustainable governance in rapidly urbanising regions.

4.3 Limitations and future directions

Although this study has made significant progress in both theoretical contributions and practical applications, several limitations remain that warrant further exploration and improvement in future research. Firstly, this study uses Suzhou as a single case study. While Suzhou is one of the key cities in the Yangtze River Delta region and features a comprehensive urban-rural structure, ranging from high-density urban cores to districts and county-level areas, the generalisability of the conclusions may be limited by spatial scale and regional context. Future research could apply this framework to different urban agglomerations with varying development backgrounds and spatial scales, such as the Beijing-Tianjin-Hebei regions and the Guangdong-Hong Kong-Macao Greater Bay Area, to test its applicability and generalisability under varying regional policy contexts. Particularly, given the differences in economic development stages, policy support, and regional resource endowments, urban-rural sustainability models may vary significantly across regions. Thus, conducting cross-regional comparative studies will help enrich the theoretical framework and broaden the practical application scenarios of this research.

Secondly, this study employs a static approach to set indicator weights using the CRITIC method. This approach somewhat overlooks the dynamic importance of each indicator during the urban-rural integration process. As urbanisation progresses, the sustainability goals and key factors at different stages of urban-rural development may shift, leading to changes in the influence of specific indicators on sustainability assessments. Consequently, this study did not fully capture this dynamic variation. Future research could consider the adoption of dynamic entropy-weight methods or Bayesian dynamic weighting techniques to more accurately reflect the temporal evolution of indicator weights (Almulhim, 2024), thereby improving the flexibility and precision of the model. Finally, while this study focuses on the impact of different land-use types on urban-rural sustainability conflicts, it remains an open question how to leverage these findings to offer differentiated governance solutions for various regions. Future research could incorporate additional policy variables into the analytical framework to explore how policy interventions affect urban-rural sustainability pathways. This would further enhance the specificity and operationality of policy recommendations, making them more applicable to diverse contexts and facilitating more effective governance for sustainable urban-rural integration.

5 Conclusion

This study develops an integrated analytical framework that combines economic, ecological, and social dimensions to assess land-use conflict and coordination within urban-rural sustainable development. In contrast to earlier work that examines sustainability from a single perspective, the framework proposed in this study delivers a more comprehensive and systematic understanding of how land systems shape sustainability outcomes and tensions. Methodologically, this study combines Random-Effects regression with Extreme Gradient Boosting (XGBoost) and Shapley Additive exPlanations (SHAP), to unite prediction and interpretation. This creatively hybrid design identifies the complex mechanisms through which land-use categories influence the URSI-based conflict index, including variable importance, non-linear responses, interaction effects, and segmented effects. The approach enhances both explanatory clarity and empirical traction relative to conventional linear models alone.

Applying the framework to Suzhou over 2012–2024 demonstrates its validity and identifies three key findings: (1) URSI-based conflicts diffuse and accumulates spatially as urban expansion proceeds, shifting from a monocentric hotspot to a polycentric pattern; (2) land-use categories exhibit differentiated and non-linear effects: expansion of built-up land is the principal driver of heightened conflict, whereas farmland, water, and forest stocks play conflict-mitigating roles; (3) SHAP reveals indicative operating ranges, including an inflection around built-up area approximately 60%, a diminishing-return mitigation effect for water coverage around 15%, and conflict alleviation when contiguous farmland is maintained near 20%. These thresholds should be treated as model-based heuristics that require local validation. Together, the findings provide new empirical support for the interaction between polycentric urban structure, land-use transition, and urban-rural sustainability. Targeted strategies have been proposed, including differentiated land-use management across the urban-rural gradient, prioritisation of blue-green infrastructure, and protection of cohesive farmland to buffer conflict while accommodating growth. The proposed multidimensional framework and analytical toolkit are transferable and offer a measurable, evidence-based foundation for future comparative research and for practice in sustainable urban-rural planning across diverse contexts.

Data availability statement

The original contributions presented in the study are included in the article, further inquiries can be directed to the corresponding authors.

Author contributions

WG: Conceptualization, Data curation, Formal Analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing – original draft, Writing – review and editing. HW: Data curation, Formal Analysis, Investigation, Methodology, Resources, Software, Validation, Visualization, Writing – original draft, Writing – review and editing. PL: Data curation, Formal Analysis, Investigation, Methodology, Resources, Software, Validation, Visualization, Writing – original draft, Writing – review and editing. JL: Conceptualization, Funding acquisition, Methodology, Project administration, Resources, Supervision, Validation, Writing – original draft, Writing – review and editing. JC: Conceptualization, Methodology, Project administration, Writing – original draft, Writing – review and editing.

Funding

The author(s) declared that financial support was received for this work and/or its publication. This work is supported by the Science Foundation of Zhejiang Sci-Tech University (ZSTU) under Grant No. 25082393-Y and the Education of Zhejiang Province General Research Project ‘Research on the Optimisation Path of Urban Spatial Quality Driven by Human Centred Perception and Multi-Source Data’ (Grant No. 25086176-F).

Conflict of interest

The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Generative AI statement

The author(s) declared that generative AI was used in the creation of this manuscript. An AI tool was used solely as an editing assistant to improve sentence flow and enhance grammatical accuracy.

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.2025.1712246/full#supplementary-material

Abbreviations

URSI, Urban-rural sustainability index; CRITIC, The criteria importance through intercriteria correlation; CCM, The coupling coordination model; REM, Random Effects Model; HA, Human activity; ES, Ecology sensitivity; EMF, Economic facility; EA, Economic activity; SF, Social facility; EGF, Ecological facility; EQ, Ecological quality; XGBoost, Extreme Gradient Boosting Model; SHAP, SHapley Additive exPlanations.

References

Ahani, S., and Dadashpoor, H. (2021). A review of domains, approaches, methods and indicators in peri-urbanization literature. Habitat Int. 114, 102387. doi:10.1016/j.habitatint.2021.102387

CrossRef Full Text | Google Scholar

Almulhim, T. (2024). Assessment of green development level performance in G20 countries: a dynamic evaluation framework. Heliyon 10, e37622. doi:10.1016/j.heliyon.2024.e37622

PubMed Abstract | CrossRef Full Text | Google Scholar

Alvarez, S., and Larkin, S. L. (2010). Valuing ecological restoration and recreational benefits in a Mountain protected area: the case of los nevados national park, Colombia. J. Sustain. Dev. 3, 3. doi:10.5539/jsd.v3n4p3

CrossRef Full Text | Google Scholar

Alvarez-Garcia, M., Ibar-Alonso, R., and Arenas-Parra, M. (2025). Mapping urban change: an explainable machine learning approach to census unit classification and evolution. Cities 167, 106295. doi:10.1016/j.cities.2025.106295

CrossRef Full Text | Google Scholar

Argota Sánchez-Vaquerizo, J., and Helbing, D. (2025). Spatial econometrics to estimate traffic reduction by transforming office space into housing and other land uses: the case for Barcelona. Cities 162, 105901. doi:10.1016/j.cities.2025.105901

CrossRef Full Text | Google Scholar

Baffoe, G., Zhou, X., Moinuddin, M., Somanje, A. N., Kuriyama, A., Mohan, G., et al. (2021). Urban–rural linkages: effective solutions for achieving sustainable development in Ghana from an SDG interlinkage perspective. Springer.

Google Scholar

Barton, H. (2009). Land use planning and health and well-being. Land Use Policy 26, S115–S123. doi:10.1016/j.landusepol.2009.09.008

CrossRef Full Text | Google Scholar

Ben, X., and Yu, C. (2022). Ecological effects of land use transformation and its driving force based on production-living-ecological space: a case study of suzhou city. J. Environ. Eng. Technol. 12, 1428–1436. doi:10.12153/j.issn.1674-991X.20210359

CrossRef Full Text | Google Scholar

Buzási, A., and Csete, M. S. (2017). Ex-ante assessment of urban development projects. Eur. J. Sustain. Dev. 6, 267. doi:10.14207/ejsd.2017.v6n4p267

CrossRef Full Text | Google Scholar

Chen, Z., Yu, B., Yang, C., Zhou, Y., Yao, S., Qian, X., et al. (2020). The global NPP-VIIRS-like nighttime light data (version 2) for 1992-2024. Harv. Dataverse. doi:10.7910/DVN/YGIVCD

CrossRef Full Text | Google Scholar

Chen, J., Pellegrini, P., Xu, Y., Ma, G., Wang, H., An, Y., et al. (2022). Evaluating residents’ satisfaction before and after regeneration. The case of a high-density resettlement neighbourhood in suzhou, China. Cogent Soc. Sci. 8, 2144137. doi:10.1080/23311886.2022.2144137

CrossRef Full Text | Google Scholar

Chen, J., Gan, W., Liu, N., Li, P., Wang, H., Zhao, X., et al. (2024). Community quality evaluation for socially sustainable regeneration: a study using multi-sourced geospatial data and AI-Based image semantic segmentation. ISPRS Int. J. Geo-Information 13, 167. doi:10.3390/ijgi13050167

CrossRef Full Text | Google Scholar

Cui, P., Abdel-Aty, M., Han, L., and Yang, X. (2025). Multiscale geographical random forest: a novel spatial ML approach for traffic safety modeling integrating street-view semantic visual features. Transp. Res. Part C Emerg. Technol. 179, 105299. doi:10.1016/j.trc.2025.105299

CrossRef Full Text | Google Scholar

Dadashpoor, H., and Ahani, S. (2019). A conceptual typology of the spatial territories of the peripheral areas of metropolises. Habitat Int. 90, 102015. doi:10.1016/j.habitatint.2019.102015

CrossRef Full Text | Google Scholar

Diakoulaki, D., Mavrotas, G., and Papayannakis, L. (1995). Determining objective weights in multiple criteria problems: the critic method. Comput. Oper. Res. 22, 763–770. doi:10.1016/0305-0548(94)00059-H

CrossRef Full Text | Google Scholar

Follmann, A., Willkomm, M., and Dannenberg, P. (2021). As the city grows, what do farmers do? A systematic review of urban and peri-urban agriculture under rapid urban growth across the global south. Landsc. Urban Plan. 215, 104186. doi:10.1016/j.landurbplan.2021.104186

CrossRef Full Text | Google Scholar

Fu, Y., Zhou, T., Yao, Y., Qiu, A., Wei, F., Liu, J., et al. (2021). Evaluating efficiency and order of urban land use structure: an empirical study of cities in Jiangsu, China. J. Clean. Prod. 283, 124638. doi:10.1016/j.jclepro.2020.124638

CrossRef Full Text | Google Scholar

Gan, L., Shi, H., Hu, Y., Lev, B., and Lan, H. (2020). Coupling coordination degree for urbanization city-industry integration level: sichuan case. Sustain. Cities Soc. 58, 102136. doi:10.1016/j.scs.2020.102136

CrossRef Full Text | Google Scholar

Han, H., Bai, X., Costanza, R., and Dong, L. (2024). The effects of COVID-19 on the resilience of urban life in China. Npj Urban Sustain. 4, 32. doi:10.1038/s42949-024-00167-8

CrossRef Full Text | Google Scholar

He, Y., Zhou, G., Tang, C., Tan, X., Chen, Z., and Fan, S. (2017). Theory on the spatial organization of urban-rural integration in urban agglomeration areas. Geogr. Res. 36, 241–252. doi:10.11821/dlyj201702004

CrossRef Full Text | Google Scholar

He, S., Yu, S., Li, G., and Zhang, J. (2020). Exploring the influence of urban form on land-use efficiency from a spatiotemporal heterogeneity perspective: evidence from 336 Chinese cities. Land Use Policy 95, 104576. doi:10.1016/j.landusepol.2020.104576

CrossRef Full Text | Google Scholar

He, Y., Wen, C., Fang, X., and Sun, X. (2024). Impacts of urban-rural integration on landscape patterns and their implications for landscape sustainability: the case of changsha, China. Landsc. Ecol. 39, 129. doi:10.1007/s10980-024-01926-9

CrossRef Full Text | Google Scholar

Hidalgo, A. (2024). Your room is ready: tourism and urban revival. Regional Sci. Urban Econ. 109, 104059. doi:10.1016/j.regsciurbeco.2024.104059

CrossRef Full Text | Google Scholar

Hu, Y., and Han, Y. (2019). Identification of urban functional areas based on POI data: a case study of the guangzhou economic and technological development zone. Sustainability 11, 1385. doi:10.3390/su11051385

CrossRef Full Text | Google Scholar

Huang, X., and Yang, J. (2025). The 30 m annual land cover datasets and its dynamics in China from 1985 to 2024. Earth Syst. Sci. Data 13, 3907–3925. doi:10.5281/zenodo.15853565

CrossRef Full Text | Google Scholar

Ives, C. D., and Kidwell, J. (2019). Religion and social values for sustainability. Sustain. Sci. 14, 1355–1362. doi:10.1007/s11625-019-00657-0

CrossRef Full Text | Google Scholar

Keith, M., Birch, E., Buchoud, N. J., Cardama, M., Cobbett, W., Cohen, M., et al. (2023). A new urban narrative for sustainable development. Nat. Sustain. 6, 115–117. doi:10.1038/s41893-022-00979-5

CrossRef Full Text | Google Scholar

Kim, J. S., and Feng, Y. (2024). Understanding complex viewpoints in smart sustainable cities: the experience of suzhou, China. Cities 147, 104832. doi:10.1016/j.cities.2024.104832

CrossRef Full Text | Google Scholar

Lessani, M. N., and Li, Z. (2024). SGWR: similarity and geographically weighted regression. Int. J. Geogr. Inf. Sci. 38, 1232–1255. doi:10.1080/13658816.2024.2342319

CrossRef Full Text | Google Scholar

Li, Y. (2012). Urban–rural interaction patterns and dynamic land use: implications for urban–rural integration in China. Reg. Environ. Change 12, 803–812. doi:10.1007/s10113-012-0295-4

CrossRef Full Text | Google Scholar

Li, H., and Miao, L. (2025). A study of the non-linear relationship between urban morphology and vitality in heritage areas based on multi-source data and machine learning: a case study of Dalian. ISPRS Int. J. Geo-Information 14, 177. doi:10.3390/ijgi14040177

CrossRef Full Text | Google Scholar

Li, J., Yuan, W., Qin, X., Qi, X., and Meng, L. (2022). Coupling coordination degree for urban green growth between public demand and government supply in urban agglomeration: a case study from China. J. Environ. Manag. 304, 114209. doi:10.1016/j.jenvman.2021.114209

PubMed Abstract | CrossRef Full Text | Google Scholar

Lin, V. S., Qin, Y., Ying, T., Shen, S., and Lyu, G. (2022). Night-time economy vitality index: framework and evidence. Tour. Econ. 28, 665–691. doi:10.1177/13548166211042970

CrossRef Full Text | Google Scholar

Liu, Y. (2018). Research on the urban-rural integration and rural revitalization in the new era in China. Acta Geogr. Sin. 73, 637–650. doi:10.11821/dlxb201804004

CrossRef Full Text | Google Scholar

Liu, J., Jin, X., Xu, W., Gu, Z., Yang, X., Ren, J., et al. (2020). A new framework of land use efficiency for the coordination among food, economy and ecology in regional development. Sci. Total Environ. 710, 135670. doi:10.1016/j.scitotenv.2019.135670

PubMed Abstract | CrossRef Full Text | Google Scholar

Long, H. (2022). Theorizing land use transitions: a human geography perspective. Habitat Int. 128, 102669. doi:10.1016/j.habitatint.2022.102669

CrossRef Full Text | Google Scholar

Long, H., and Tu, S. (2018). Theoretical thinking of rural restructuring. Prog. Geogr. 37, 581–590. doi:10.18306/dlkxjz.2018.05.002

CrossRef Full Text | Google Scholar

Luge, W., Tianjiao, Z., and Tiyan, S. (2025). Spatial-temporal evolution and influencing factors of urban land use structure efficiency: evidence from 282 cities in China. J. Clean. Prod. 500, 145275. doi:10.1016/j.jclepro.2025.145275

CrossRef Full Text | Google Scholar

Ma, X., Ma, R., and Pu, Y. (2007). The spatial pattern and evolvement of suzhou urbanization. Urban Probl., 20–24.

Google Scholar

Ma, L., Liu, S., Fang, F., Che, X., and Chen, M. (2020). Evaluation of urban-rural difference and integration based on quality of life. Sustain. Cities Soc. 54, 101877. doi:10.1016/j.scs.2019.101877

CrossRef Full Text | Google Scholar

Maselko, J., Hughes, C., and Cheney, R. (2011). Religious social capital: its measurement and utility in the study of the social determinants of health. Soc. Sci. and Med. 73, 759–767. doi:10.1016/j.socscimed.2011.06.019

PubMed Abstract | CrossRef Full Text | Google Scholar

Moldan, B., Janoušková, S., and Hák, T. (2012). How to understand and measure environmental sustainability: indicators and targets. Ecol. Indic. 17, 4–13. doi:10.1016/j.ecolind.2011.04.033

CrossRef Full Text | Google Scholar

Nuissl, H., and Siedentop, S. (2021). “Urbanisation and land use change,” in Sustainable land management in a European context: a co-design approach, 75–99.

Google Scholar

Pan, L., Gan, W., Chen, J., and Ren, K. (2023). An integrated model for constructing urban ecological networks and identifying the ecological protection priority: a case study of wujiang district, suzhou. Sustainability 15, 4487. doi:10.3390/su15054487

CrossRef Full Text | Google Scholar

Pan, L., Gan, W., Chen, J., and Xian, X. (2024). Ecological planning for mitigating urban heat based on thermal network simulation:a case study of guangzhou-foshan metropolitan area. J. Chin. Urban For. 22, 10–19. doi:10.12169/zgcsly.2024.06.26.000

CrossRef Full Text | Google Scholar

Perrin, C., Nougarèdes, B., Sini, L., Branduini, P., and Salvati, L. (2018). Governance changes in peri-urban farmland protection following decentralisation: a comparison between montpellier (france) and rome (italy). Land Use Policy 70, 535–546. doi:10.1016/j.landusepol.2017.09.027

CrossRef Full Text | Google Scholar

Salem, M., Tsurusaki, N., and Divigalpitiya, P. (2020). Remote sensing-based detection of agricultural land losses around greater Cairo since the Egyptian revolution of 2011. Land Use Policy 97, 104744. doi:10.1016/j.landusepol.2020.104744

CrossRef Full Text | Google Scholar

Salem, M., Ravetz, J., Sareen, S., Dong, T., Haque, M., Bayoumi, W., et al. (2025). Managing the urban-rural transition: a review of approaches and policies for peri-urban land use. J. Urban Manag. 14, 1115–1129. doi:10.1016/j.jum.2025.04.006

CrossRef Full Text | Google Scholar

Saraswat, A., Pipralia, S., and Kumar, A. (2025). Advances in urban sustainability assessment: a systematic literature review of methods, frameworks, and future directions. Front. Urban Rural Plan. 3, 15. doi:10.1007/s44243-025-00063-4

CrossRef Full Text | Google Scholar

Shao, Y., Lange, E., Thwaites, K., Xue, Z., and Xu, X. (2020). Understanding landscape identity in the context of rapid urban change in China. Land 9, 298. doi:10.3390/land9090298

CrossRef Full Text | Google Scholar

Shen, C., Shi, L., Wu, X., Ding, J., and Wen, Q. (2023). Exploring the coupling coordination and key factors between urban–rural integrated development and land-use efficiency in the yellow River Basin. Land 12, 1583. doi:10.3390/land12081583

CrossRef Full Text | Google Scholar

Shi, K., Chang, Z., Chen, Z., Wu, J., and Yu, B. (2020). Identifying and evaluating poverty using multisource remote sensing and point of interest (POI) data: a case study of chongqing, China. J. Clean. Prod. 255, 120245. doi:10.1016/j.jclepro.2020.120245

CrossRef Full Text | Google Scholar

Somanje, A. N., Mohan, G., Lopes, J., Mensah, A., Gordon, C., Zhou, X., et al. (2020). Challenges and potential solutions for sustainable urban-rural linkages in a Ghanaian context. Sustainability 12, 507. doi:10.3390/su12020507

CrossRef Full Text | Google Scholar

Su, B., Shen, X., Wang, Q., Zhang, Q., Niu, J., Yin, Q., et al. (2024). The evolution and performance response of industrial land use development in China's development zone: the case of suzhou industrial park. Land, 13, 2182. doi:10.3390/land13122182

CrossRef Full Text | Google Scholar

Sumi, A. A., Awang, J. B., Haque, R., Qazi, S., Rahman, N. F. A., and Senathirajah, A. R. B. S. (2025). Investigating the function of religion and social capital in shaping sustainable social development. Discov. Sustain. 6, 748. doi:10.1007/s43621-025-01622-x

CrossRef Full Text | Google Scholar

UN (2015). Transforming our world: the 2030 agenda for sustainable development, 1. New York: United Nations, Department of Economic and Social Affairs, 41.

Google Scholar

Wang, M., Zhao, J., Zhang, D., Xiong, Z., Sun, C., Zhang, M., et al. (2025). Assessing urban vitality in high-density cities: a spatial accessibility approach using POI reviews and residential data. Humanit. Soc. Sci. Commun. 12, 1119. doi:10.1057/s41599-025-05459-7

CrossRef Full Text | Google Scholar

Wooldridge, J. M. (2010). Econometric analysis of cross section and panel data. MIT press.

Google Scholar

Wu, M., and Chen, Y. (2022). Gourmet and vitality: analysing the restaurant and urban vitality in the old city of beijing based on multi-source big data.

Google Scholar

Wu, H., Wu, F., Cai, Y., and Li, Z. (2024). Assessing the spatiotemporal impacts of land use change on ecological environmental quality using a regionalized territorial impact assessment framework. Sustain. Cities Soc. 112, 105623. doi:10.1016/j.scs.2024.105623

CrossRef Full Text | Google Scholar

Wu, Y., Zhang, Y., and Zhang, B. (2025). The effect of urban agglomeration land use zoning on sustainable development goals. Sci. Rep. 15, 33427. doi:10.1038/s41598-025-18892-0

PubMed Abstract | CrossRef Full Text | Google Scholar

Xi, Y., Hou, Q., Duan, Y., Lei, K., Wu, Y., and Cheng, Q. (2024). Exploring the spatiotemporal effects of the built environment on the nonlinear impacts of metro ridership: evidence from xi’an, China. ISPRS Int. J. Geo-Information 13, 105. doi:10.3390/ijgi13030105

CrossRef Full Text | Google Scholar

Xu, X. (2024). China annual NDVI and EVI 250m datasets. Department of Resource and Environmental Science Data Registration and Publication. doi:10.12078/2024072901

CrossRef Full Text | Google Scholar

Xu, F., Wang, Z., Chi, G., and Zhang, Z. (2020). The impacts of population and agglomeration development on land use intensity: new evidence behind urbanization in China. Land Use Policy 95, 104639. doi:10.1016/j.landusepol.2020.104639

CrossRef Full Text | Google Scholar

Yan, J., Chen, H., and Xia, F. (2018). Toward improved land elements for urban–rural integration: a cell concept of an urban–rural mixed community. Habitat Int. 77, 110–120. doi:10.1016/j.habitatint.2018.01.007

CrossRef Full Text | Google Scholar

Yang, S., Zhao, W., Liu, Y., Wang, S., Wang, J., and Zhai, R. (2018). Influence of land use change on the ecosystem service trade-offs in the ecological restoration area: dynamics and scenarios in the yanhe watershed, China. Sci. Total Environment 644, 556–566. doi:10.1016/j.scitotenv.2018.06.348

PubMed Abstract | CrossRef Full Text | Google Scholar

Yang, Y., Bao, W., Wang, Y., and Liu, Y. (2021). Measurement of urban-rural integration level and its spatial differentiation in China in the new century. Habitat Int. 117, 102420. doi:10.1016/j.habitatint.2021.102420

CrossRef Full Text | Google Scholar

Yang, B., Yang, J., Tan, L., and Xiao, J. (2023). Understanding the change of land space utilization efficiency with different functions and its coupling coordination: a case study of urban agglomeration in the middle reaches of the yangtze river, China. J. Geogr. Sci. 33, 289–310. doi:10.1007/s11442-023-2083-0

CrossRef Full Text | Google Scholar

Yang, D., Lin, Q., Li, H., Chen, J., Ni, H., Li, P., et al. (2025). Unraveling spatial nonstationary and nonlinear dynamics in life satisfaction: integrating geospatial analysis of community built environment and resident perception via MGWR, GBDT, and XGBoost. ISPRS Int. J. Geo-Information 14, 131. doi:10.3390/ijgi14030131

CrossRef Full Text | Google Scholar

Zhang, A. (2025). Analysis of the sustainable development pathway of urban–rural integration from the perspective of spatial planning: a case study of the urban–rural fringe of beijing. Sustainability 17, 1857. doi:10.3390/su17051857

CrossRef Full Text | Google Scholar

Zhang, L., Wu, C., and Yao, S. (2003). The characters of urban land expansion in southern Jiangsu Province and its problems. Prog. Geogr. 22, 639–645. doi:10.11820/dlkxjz.2003.06.012

CrossRef Full Text | Google Scholar

Zhang, L., Zhang, L., Xu, Y., Zhou, P., and Yeh, C.-H. (2020). Evaluating urban land use efficiency with interacting criteria: an empirical study of cities in Jiangsu China. Land Use Policy 90, 104292. doi:10.1016/j.landusepol.2019.104292

CrossRef Full Text | Google Scholar

Zhao, P., and Wan, J. (2021). Land use and travel burden of residents in urban fringe and rural areas: an evaluation of urban-rural integration initiatives in beijing. Land Use Policy 103, 105309. doi:10.1016/j.landusepol.2021.105309

CrossRef Full Text | Google Scholar

Zheng, Y., Cheng, L., and Wang, Y. (2022). Measuring the spatial conflict of resource-based cities and its coupling coordination relationship with land use. Land 11, 1460. doi:10.3390/land11091460

CrossRef Full Text | Google Scholar

Zhou, L., and Li, L. (2020). Spatial distribution and format difference of large-scale retail business facilities: a case study of guangzhou based on POI data. Trop. Geogr. 40, 88–100. doi:10.13284/j.cnki.rddl.003164

CrossRef Full Text | Google Scholar

Zhu, L., and He, Y. (2025). Mitigating spatial land use conflicts through an ecological lens: insights from the chengdu-chongqing urban agglomeration. Environ. Dev. Sustain., 1–24. doi:10.1007/s10668-025-06230-3

CrossRef Full Text | Google Scholar

Zikirya, B., He, X., Li, M., and Zhou, C. (2021). Urban food takeaway vitality: a new technique to assess urban vitality. Int. J. Environ. Res. Public Health 18, 3578. doi:10.3390/ijerph18073578

PubMed Abstract | CrossRef Full Text | Google Scholar