Moulins on the southwestern Greenland Ice Sheet (GrIS) can drain large volumes of surface meltwater to the base of the ice sheet in summer, affecting basal water pressures and ice motion. However, the spatiotemporal moulin distributions (including numbers, densities and locations) and the drainage-versus-storage competition between moulins and terminal lakes (two outlet types of internally drained catchments) remain unclear. We manually map moulins and terminal lakes on the southwestern GrIS and analyze their interannual distributions during late summer 2016-2021 using Sentinel-2 images and ArcticDEM. Our findings include: (1) moulin numbers vary significantly with an annual mean of 346, ranging from 264 to 418 (densities of 0.02–0.03 km -2 ), and stronger meltwater runoff induces more moulins; (2) moulins drain 91.7% of surface meltwater runoff on average and up to 97.8% in the warm 2016; (3) stable moulins (recur ≥4 in 6 years) account for ~60% of moulins on average, and drain ~70% of surface meltwater runoff; (4) terminal lakes increase in number as runoff decreases, storing 8.3% of surface meltwater runoff on average and 21.0% in the cool 2017; (5) the majority of moulins (~66%) and terminal lakes (~99%) can be captured by surface topographic depressions modeled by ArcticDEM, whereas >70% of topographic depressions contain neither of these features. In summary, this study produces multi-year moulin maps on the southwestern GrIS, and reveals that moulins are denser in stronger-melting years with higher percentages of meltwater drainage into the ice sheet, thereby providing new constraints for the GrIS hydrological models.

Identifying the spatiotemporal patterns and future trends of carbon budgets in Chinese provinces is crucial for achieving the dual-carbon targets, yet comprehensive analyses incorporating provincial carbon sink projections remain limited. This study used panel data from 30 Chinese provinces (2000-2019) to characterize carbon budget patterns and developed province-specific system dynamics models integrating the energy, economy, carbon emissions, and land-use subsystems. Three policy-informed scenarios were designed to simulate provincial carbon-budget trajectories and progress toward the dual-carbon targets. The results showed that 2011 marked a shift in carbon emissions, with panel regression analysis showing that industrial structure, energy structure, and economic development promoted emissions, while urbanization, population density, and temperature negatively affected carbon sinks. At the provincial level, under the Baseline Scenario (BS), nearly half of the provinces could peak by 2030 and achieve carbon neutrality by 2060. In the High Development Scenario (HDS), most provinces would fail to achieve dual-carbon targets on schedule, while the High Improvement Scenario (HIS) showed the opposite trend. At the national level, the provincial simulation results were summed to obtain total and net carbon emissions for China. Under the HIS, BS, and HDS, national total carbon emissions were projected to peak in 2026, 2029, and 2032, with corresponding peak values of 11,480.54, 13,302.55, and 14,565.69 Mt CO2, respectively. National net carbon emissions were projected to decline to zero in 2058, 2060, and after 2060, respectively. This study provides data reinforcement and differentiated emission-reduction strategies for low-carbon development in Chinese provinces.

Ecotourism suitability mapping is essential for balancing conservation and development in protected areas; however, existing methods often overlook geospatial heterogeneity, limiting their effectiveness for ecological indicator assessment and spatial planning. This study aims to map ecotourism suitability in Shennongjia National Park, China, using a geospatial heterogeneity-constrained diversity regularization ensemble learning (GSH-DREL) framework that integrates global and local spatial information. We constructed a comprehensive evaluation system incorporating regional culture, endangered species, and local folklore. Training samples were derived from GPS coordinates, Amap points of interest, and geotagged photographs. The GSH-DREL framework combined geographically weighted logistic regression, a spatially weighted ensemble neural network, and diversity regularization via mutual information-based hierarchical clustering. The suitability map classified the study area into four categories: Highly suitable (1158.38 km2, 35.83%), moderately suitable (790.41 km2, 24.45%), marginally suitable (520.81 km2, 16.11%), and unsuitable (763.17 km2, 23.61%). Suitable areas were linearly distributed along transportation corridors and river valleys, shaped by resource uniqueness and topographic constraints. Feature importance and SHAP analyses identified NDVI, soil conservation, land cover, and precipitation as dominant drivers. Thus, the GSH-DREL framework provides a robust and interpretable approach for ecotourism suitability mapping, offering practical insights for sustainable development, environmental protection, and land-use management in ecologically sensitive regions. These findings directly support ecological indicator-based planning and conservation decision-making in protected areas

Continuous land cover changes in remote sensing observations can provide crucial data support for evaluating urban spatial planning and sustainable development, but their acquisition remains challenging because it must overcome missing observations caused by clouds and cloud shadows while accurately characterizing the spatiotemporal dynamics of land cover semantics. Therefore, we proposed a continuous change detection method fusing transformer and bi-directional long short-term memory (LSTM) via cross-attention (CCD-TBLCA) in land cover using Sentinel-2 imagery. Specifically, it employed a dual-branch encoding design for spectral time-series data, enabling deep information fusion of long-term dependencies and local dynamics. First, we used the optical temporal information from raw irregular time series, achieving monthly mapping from spectral anomalies to land cover change events based on CCD-TBLCA. Subsequently, we proposed a breakpoint detection criterion that enforced semantic consistency before and after changes, thereby forming a more comprehensive framework for evaluating breakpoint detection alongside spatial existence and temporal accuracy. Across seven urban regions, CCD-TBLCA outperformed existing methods in both temporal classification and breakpoint detection, with semantic-consistent breakpoint detection accuracy generally improving by over 15%. In spatial mapping, CCD-TBLCA achieved F1 accuracy above 0.9 for most land cover types, surpassing existing products. CCD-TBLCA's robustness to sequence updates and irregular time series had also been validated. The results show that CCD-TBLCA provided in this study can accurately describe continuous spatiotemporal land cover changes and help to understand how human activities reshape land cover patterns.

Urban vitality (UV) is a concept reflecting development of cities, and rich connotations make its assessment require the consideration of multiple dimensions. Existing studies on UV assessment mostly focused on development of the region itself, and seldom deemed spatial interaction within different regions as an independent evaluation dimension, which referred to the comprehensive representation of mobilities, exchanges, and perceptions of geography elements. Meanwhile, non-linear relations between UV and proposed indexes were not given sufficient attentions. This study combined eight types of earth observation data and eight types of geographic big data to construct a UV assessment index system, which was divided into density, diversity, livability, accessibility, and interaction dimensions. The gravity model was mainly used to quantify the interaction among regions and location-based big data was to quantify between people and region. Then XGBoost regressor was applied to construct UV assessment model based on these five dimensions, where the reference samples were established using expert scoring method. Finally, UV in four typical Chinese cities were calculated at a spatial resolution of 1 km, including Nanjing, Beijing, Wuhan, and Shenzhen. Results revealed that the UV assessment results were consistent with overall urban development patterns, and PCCs with validation samples were all beyond 0.7 in four cities, higher than linear methods such as entropy weighting method. Based on SHAP values, the interaction dimension was proved to have the highest average index importance of 5.07% among 24 indexes, and the removal of indexes from interaction dimension would cause obvious accuracy reduction of UV assessment. The model also had low level of the mean absolute percentage error, <18%, and slight overestimation on average, where the over/underestimations were mainly in regions like urban villages and new towns. We believe the proposed UV assessment model can help to improve the understanding of regional development imbalance and guide scientific urban planning.

Understanding the evolution of land systems is a prerequisite for optimizing territorial spatial patterns and promoting sustainable development. Guided by the Element–Structure–Function theory, this study classified land systems in the Yangtze River Delta (YRD) using hierarchical clustering and random forest algorithms. A LightGBM model was then applied to analyze constituent elements and element structures, while the land function pattern was systematically evaluated by constructing a land function evaluation index system. Additionally, the geo-informatic Tupu method was used to reveal the spatio-temporal characteristics of these land systems. The results indicated profound spatial differentiation and significant transformations in the YRD from 2000 to 2020, prominently featuring the expansion of settlement system (LS1), the structural optimization of cropland system (LS2), and the ecological recovery of forest system (LS3). During this period, human activities exerted a strong multi-dimensional influence on land system transitions. Crucially, the constituent elements of these systems have shifted from landscape-dominated elements to the joint influence of landscape and human activities, with socio-economic elements increasingly becoming key determinants of micro-level differences of the land systems. Functionally, the agricultural production function concentrated in the core areas of the Huang-Huai and Yangtze-Huai Plains, while the urban living function evolved into a polycentric network centered on Shanghai. The regional ecological function remained generally stable despite facing localized pressures. These findings can provide a scientific basis for territorial spatial planning and support refined territorial space governance.

Point-of-interest (POI) data are widely used as urban big data to sense the functional organization of urban land and associated socioeconomic activities. Yet, POI-based studies often treat these records as temporally stable, even though different POI categories vary in turnover and responses to external shocks, affecting the comparability of urban sensing over time. Using POI datasets for Changzhou, China, from 2016, 2019, 2022, and 2025, this study examines category-specific changes before, during, and after the COVID-19 disruption. The results identify three patterns: government-driven POIs (government and residential) show high stability and retention; government–market hybrid POIs (financial and educational) exhibit moderate stability; and market-driven POIs (catering and shopping) have low stability. Across periods, change intensity is inversely related to stability. During the pandemic, turnover increased across all categories, retention declined markedly, and most replacements occurred within the same category, indicating functional continuity despite entity-level change. Although the overall spatial framework of urban functions remained broadly stable, new hotspots emerged during the pandemic and some persisted into the recovery period. These findings demonstrate that POI-based urban sensing is temporally contingent and that category-specific dynamics should be considered when interpreting urban land-use functions and supporting land-use planning under disruptive events.

As the world’s population expands and urbanization accelerates, the interconnectedness of water, energy, and food (WEF) resources becomes increasingly critical. In the Yangtze River Delta (YRD) of China, rapid urban expansion has heightened WEF demand and environmental pressures, posing challenges to WEF system resilience (WEFSR). To clarify how new-type urbanization (NU) affects WEFSR across dimensions, spatial effects, and periods, this study constructed a multidimensional index system for NU (covering demographic, economic, social, spatial, ecological, and urban-rural integration dimensions) together with a WEFSR index capturing resistance, recoverability, and adaptability, based on panel data from 41 cities in the YRD from 2010 to 2022. We measured city-level WEFSR and NU, analyzed their spatial correlations, and applied a spatial Durbin model to quantify the direct and spillover effects of NU on WEFSR. Results showed that while NU and WEFSR increased, a negative spatial correlation between them highlighted an imbalance in the distribution of NU and WEF resources. NU exerted a negative local effect (−0.057) but a positive spillover effect (0.308) on WEFSR. Population, economic, and social urbanization played a decisive role in the impact of NU on WEFSR. Spatial urbanization negatively impacted the WEFSR of both local and neighboring areas. In contrast, eco-environmental urbanization and urban-rural integration enhanced WEFSR locally and in neighboring areas, with their positive effects being most pronounced in Zhejiang and Anhui provinces. NU reduced local resilience in the WEF subsystems but benefited adjacent regions. These findings provide valuable insights for refining NU strategies and WEF resource management policies in highly urbanized areas.

While urban polycentric development reshapes regional spatial structure and significantly impacts cultivated land allocation and fragmentation, the underlying mechanisms are still poorly understood. Moreover, existing studies on the drivers of cultivated land fragmentation often overlook spatial effects, failing to accurately capture the spatial heterogeneity of predictive factors. To address these gaps, this study employed panel data from 217 Chinese cities (2002-2022) to quantify urban polycentricity and the cultivated land fragmentation index (CLFI). By comparing multiple predictive models and integrating SHAP and GeoShapley explainable AI (XAI) techniques, we analyzed the impact of polycentric development on CLFI. The results indicated that CLFI showed a fluctuating rise with significant spatial divergence. Polycentric development was the second-largest driver, where Population concentration positively correlated with CLFI, while the Population of the largest center and Aggregated central area exhibited nonlinear negative correlations. Significant interaction effects were observed among these three variables, with Population concentration and the other two jointly suppressing CLFI. Furthermore, the effect of polycentric variables on CLFI displayed notable spatial heterogeneity. The driving effect of population concentration was stronger in the southeast and weaker in the northwest, whereas the suppression from the other two was concentrated in the North China and Northeast Plains. This study reveals the complex mechanisms through which polycentric structures affect cultivated land fragmentation from a macro-spatial strategic perspective, offering critical insights for balancing urban growth and cultivated land protection, as well as policy insights for formulating differentiated polycentric development strategies for cities of different population sizes.

With the rapid development of urbanization, urban spatial structures have become increasingly complex, making effective urban spatial analysis increasingly important. However, existing methods primarily adopt single-task modeling strategies, which not only ignore the intrinsic correlations among urban spatial analysis tasks but also face challenges such as high computational resource consumption and imbalanced sample labels. Therefore, this study proposes a multitask learning method and applies it to the analysis of land use (LU) classification and urban villages (UVs) identification, taking the central urban area of Changzhou as a case study. First, the shared embedding layer bridges the semantic gap between remote sensing imagery (RSI) and street view imagery (SVI), enabling effective cross-modal alignment and multitask shared feature extraction. Second, the task-specific layers capture the distinct characteristics of LU and UV tasks, enhancing task-specific feature representation. Third, the task interaction layer facilitates cross-scale spatial knowledge sharing and deep feature interaction between the two tasks. Fourth, the study trains and outputs classification results for both tasks through the output layer. From the experimental results, the LU classification achieved an overall accuracy (OA) of 0.872, a Kappa coefficient of 0.843, and an F1 score of 0.868, while the UV classification attained an OA of 0.906, a Kappa of 0.857, and an F1 score of 0.916. Compared to single-task learning models that train separate networks for each task independently, our multitask learning framework achieved OA improvements of 6.7% and 5.0% for LU classification and UV identification, respectively. Our proposed shared embedding layer, task-specific layer, and task interaction layer outperformed existing methods (IPCAM, Task-Adapter, and MAInt), demonstrating superior performance in multimodal fusion, task-specific representation, and multitask synergistic complementarity for urban spatial analysis. The framework can provide new insights for urban spatial modeling and offer reliable technical support for urban planning. Data are available at https://doi.org/10.6084/m9.figshare.31054927

Understanding how soil moisture drought transitions into hydrological drought is essential for effective monitoring of water resources and assessing drought risk. Traditional methods for constructing drought indices often overlook regional heterogeneity and fail to adequately capture the dynamic dependencies among drought variables. Thus, this study proposes an integrated framework to characterize the propagation characteristics from soil moisture drought to hydrological drought. The k-means algorithm was first used to conduct the rough spatial zoning considering precipitation and evapotranspiration, and the standardized soil moisture index (SSMI) and standardized runoff index (SRI) were constructed. The sliding window-based Spearman's correlation analysis on these two drought indices was subsequently applied to estimate the drought propagation time. Finally, Dynamic Bayesian Network (DBN) and Copula joint distribution functions were combined to model their dynamic relationships and infer the drought propagation probability. Results indicate that, during 1982-2023, drought propagation from soil moisture to hydrological droughts in the Yangtze River Basin is highly non-linear and demonstrates substantial regional heterogeneity. The average drought propagation time is 7.39 months, with pronounced lag effects evident at higher elevations. Drought propagation probability also revealed spatial distribution differences and distinct hierarchical response patterns, with averaging 19.13% in the high-elevation upper reaches compared to 10.95% in the middle-lower reaches, and 16.52% under extreme soil moisture droughts compared to 12.33% under light droughts. Comparative evaluations demonstrated that spatial zoning based SSMI and SRI can more accurately reflect regional drought dynamics, and DBN can also enhance the sensitivity to the spatial-temporal changes of drought variables compared to conventional BN. The proposed probabilistic reasoning framework can help to strengthen the foundation for decision-making in agricultural production and ecological conservation.

The monitoring method from a two-dimensional (2D) perspective achieves limited precision in accurately reflecting spatial allocation. This study proposes a three-dimensional (3D) urban building space utilization efficiency (3D-UBSE) analytical framework, using multi-source geospatial big data to measure the balance of urban 3D functional space allocation. First, the framework proposes a data fusion-based feature extraction method that identifies urban spatial functions at a fine-grained level. Second, it calculates 3D spatial user density at the scale of individual buildings. Third, it analyzes the degree of balance in 3D functional space allocation at both local and global scales. Results from Nanjing show that the urban functional zone identification method achieves an overall accuracy of 0.921 at a 54 m resolution and reaches 0.926 when transferred to Wuhan. Under normal weekday conditions, the 3D-UBSE exhibits a production–living daily migration zonal structure, which shifts to a predominantly production-oriented structure at the peak of infection, dominating 96.724% of the total building volume. Furthermore, the 3D analysis reveals a 34.499% higher degree of imbalance in functional space allocation, whereas the 2D method fails to capture local imbalances. The findings of this study highlight the greater imbalance in 3D spatial functional allocation from both global and local perspectives during public health emergencies, with important implications for urban planning and emergency response.

Understanding the evolution of land systems is essential for supporting sustainable territorial spatial governance. Guided by the Element–Structure–Function theory, this study established an integrated analytical framework to investigate land system evolution in the Yangtze River Delta (YRD). Land systems were classified using Divisive Analysis (DIANA) and random forest (RF) algorithms. A LightGBM model was employed to quantify constituent elements and element structures, land functions were evaluated using a Production–Living–Ecological function evaluation system, and the geo-informatic Tupu method was used to reveal spatiotemporal transition patterns. The results showed that the land systems of the YRD exhibited pronounced spatial heterogeneity and hierarchical characteristics. Between 2000 and 2020, the settlement system (LS1) expanded from 6024 km2 to 25,988 km2, whereas the cropland system (LS2) maintained the largest spatial extent despite substantial internal restructuring. Human activities became increasingly important in shaping land systems, with cropping intensity (CI) becoming the most influential constituent element in 2020 (20.80%), while the feature importance of impervious surface coverage (ISC) increased from 6.75% to 9.03%. During the study period, LS1, the forest system (LS3), and the grassland system (LS4) showed stable element structures dominated by single landscape element, with feature importance exceeding 70%. In contrast, LS2 was dominated by cropland coverage (CLC), which accounted for more than 50% of total feature importance. Land system transitions were dominated by the conversion from LS2 to LS1 (15,383 km2), while frequent bidirectional transitions within LS2 and between LS3 and the cropland & forestland mosaic system (LS7). Functionally, agricultural production became increasingly concentrated in the Huang–Huai and Yangtze–Huai Plain, the urban living function evolved into a multicentric network centered on Shanghai, and the regional ecological function remained generally stable. The findings comprehensively reveal the integrated characteristics of land systems and provide valuable support for territorial spatial planning and territorial spatial governance.

Anthropogenic pressures from irrigation and human usage (i.e. domestic and industrial use) pose growing challenges to river sustainability under future global change. However, how these pressures evolve across river networks at fine spatial scales remains poorly understood. Here we project the anthropogenic pressures on 129 429 river reaches in China from 2020 to 2050 under the four CMIP6 scenarios (SSP1-2.6, SSP2-4.5, SSP3-7.0, and SSP5-8.5), using a proximity-based index that integrates built-up and irrigated areas. The results show that the river reaches in the Hai and Huai River basins (RBs) are projected to face high and intensifying anthropogenic pressures, which are further exacerbated by meteorological drought threats under the SSP1-2.6 and SSP2-4.5 scenarios. In contrast, the Southwest and Inland RBs exhibit approximately 5% of the anthropogenic pressures observed in the Huai RB, but meteorological drought remains an important stressor threatening the potential future river sustainability. Furthermore, anthropogenic pressures also vary by river size. Large rivers (stream orders >6) are projected to experience the highest anthropogenic pressures, followed by medium rivers (stream orders 4–5), whereas small rivers (stream orders 1–3) remain comparatively less impacted across all scenarios. Our reach-scale analysis provides a refined understanding of human-river interactions and offers critical insights for targeted water management under climate change.

Understanding the drivers of carbon emissions from the water-food-energy (WFE) system during urbanization is essential for achieving China's carbon peak and neutrality goals; yet, systematic quantification in this domain remains limited. This study develops a consumption-based accounting framework to estimate WFE carbon emissions for 41 cities in China's Yangtze River Delta (YRD) from 2000 to 2023. By integrating restricted cubic splines, boosted regression trees, and piecewise structural equation modeling, we systematically uncover the nonlinear impacts and multi-path transmission mechanisms of comprehensive urbanization on WFE emissions. The key findings are as follows: (1) WFE carbon emissions (WFEC) in the YRD exhibit a fluctuating pattern with an overall upward tendency since 2000, mainly driven by industrial and residential water use. (2) A significant nonlinear relationship exists between the composite urbanization index (UI) and both WFEC and WFE carbon intensity (WFECI), with a turning point at approximately UI = 0.2. (3) The marginal effects of population, spatial, and social urbanization on WFEC intensify over time, while that of ecological urbanization weakens; effects on WFECI are predominantly negative. (4) Population, economic, and spatial urbanization exert significant direct effects on emissions, alongside indirect effects mediated by resource use intensity and behavioral consumption structure, marking them as the most pivotal and complex dimensions. The present study provides new evidence on consumption-driven WFE emissions and offers theoretical and policy insights for low-carbon transitions in resource-intensive regions.

Ecological compensation (EC) is crucial for balancing ecological protection and economic growth, particularly in river basins. However, existing studies often overlook regional disparities and ecosystem service (ES) supply-demand dynamics. This study integrated remote sensing and geospatial data into a 'source-flow-sink' framework, simulating ES flows using the D8 algorithm for water yield and wind-flow models for carbon sequestration. An ecological-economic synergistic compensation model (EESCM) assessed EC ratios in the Yangtze River Delta (YRD) from 2000-2020. Results showed carbon sequestration supply declined by 1.83%, while demand surged 191.95%, causing mismatches in 85% of high-demand areas. Water-yield gaps peaked, forming dual-core networks around Yangtze River-Taihu Lake (57%) and Chaohu Lake-Xin'an River (29%). Spatial polarization was evident: the Huanghuai Plain, bearing only 23.15% of ecological compensation intensity, received 74.75% of funds; the Dabie Mountains, generating 89.67% of carbon sinks, received just 18%, revealing a vast gap. Atmospheric transport redirected 60% of carbon sinks downstream, intensifying spillover-compensation conflicts. The proposed EC framework can offer scientific guidance for sustainable regional development.

Current understanding of low-carbon food distribution is limited by the lack of observed grain flow data and forward-looking scenario analysis. In this study, we construct China’s interprovincial rice distribution network for 2020 using 30,524 observed trade records. We then project provincial rice supply and demand dynamics for 2030 under four Shared Socioeconomic Pathway (SSP) scenarios, employing the Global Agro-Ecological Zones (GAEZ) and Future Land Use Simulation (FLUS) models. Subsequently, a scenario-based linear programming framework is applied to explore optimized low-carbon trade configurations. Results indicate that, compared to simulations focused solely on minimizing transport costs, observed trade flows capture substantial long-distance interprovincial exchanges that cost-based models tend to overlook. Moreover, optimized flows could reduce total carbon emissions from rice distribution by 16.1%-20.5% in 2030. These findings offer robust empirical and modeling evidence to inform the reconfiguration of grain distribution systems and support region-specific adaptation strategies under future climate change.

Rapid urbanization in China has led to spatial antagonism between urban development and farmland protection and ecological security maintenance. Multi-objective spatial collaborative optimization is a powerful method for achieving sustainable regional development. Previous studies on multi-objective spatial optimization do not involve spatial corrections to simulation results based on the natural endowment of space resources. This study proposes an Ecological Security–Food Security–Urban Sustainable Development (ES–FS–USD) spatial optimization framework. This framework combines the non-dominated sorting genetic algorithm II (NSGA-II) and patch-generating land use simulation (PLUS) model with an ecological protection importance evaluation, comprehensive agricultural productivity evaluation, and urban sustainable development potential assessment and optimizes the territorial space in the Yangtze River Delta (YRD) region in 2035. The proposed sustainable development (SD) scenario can effectively reduce the destruction of landscape patterns of various land-use types while considering both ecological and economic benefits. The simulation results were further revised by evaluating the land-use suitability of the YRD region. According to the revised spatial pattern for the YRD in 2035, the farmland area accounts for 43.59

Ecological restoration is a complex process characterized by dynamic, bidirectional feedbacks within social-ecological systems. Public attitudes, as a critical nexus between natural and social subsystems, have rarely been quantitatively examined in terms of their impact on restoration performance. Taking the Yellow River Basin as a case study, this study employed the Chinese-RoBERTa-wwm-ext model to analyze social media data and characterize public attitudes toward ecological conservation. Furthermore, a Latent Dirichlet Allocation model based on Term Frequency–Inverse Document Frequency was used to perform a fine-grained analysis of sentiments in public posts. A coupling coordination model was utilized to explore the relationship between ecological restoration effectiveness (ERE) and public enthusiasm for ecological conservation (PEC). Additionally, the XGBoost-SHAP model and SOM-K-means algorithm were used to develop differentiated governance strategies. The results showed that ERE and PEC exhibited a fluctuating upward trend, with significant spatial gradients. The middle reaches consistently exhibited the highest coordination levels. Overall, 94.59% of cities displayed a benign synergistic relationship between ERE and PEC. However, ecological governance in fragile upstream areas still faces substantial challenges. Meteorological disasters associated with low vegetation cover and arid climates have elicited strong public dissatisfaction. The study revealed that population density and educational attainment are the primary drivers shaping PEC. A targeted governance framework based on spatial heterogeneity and threshold effects is therefore proposed to mobilize endogenous social drivers of ecological restoration. By integrating social sensing methods to support environmental recovery, this study provides a novel perspective for advancing regional sustainability and the modernization of global ecological governance.