Identifying spatiotemporal patterns and associated determinants of stable natural vegetation resilience across China

Vegetation resilience (VR) is a critical factor in ecosystem stability. Previous studies in China have mainly focused on vegetation restoration monitoring rather than resilience dynamic identification and their heterogeneous drivers, which may be insufficient for ensuring ecosystem restoration efficacy as well as afforestation management and optimization. Therefore, the present work quantified VR as the negative first-order autoregressive coefficient (−AR(1)) of detrended and deseasonalized vegetation anomalies obtained from MODIS kNDVI (kernel normalized difference vegetation index) from 2000 to 2025. We combined 60-month sliding windows with interpretable machine learning, ecosystem-specific models, and geographically weighted regression to examine spatial patterns, temporal trajectories, and environmental determinants of VR. VR was generally higher in southern and eastern China, with relatively low values in parts of the Inner Mongolian and Loess Plateau. Grassland resilience declined substantially with latitude, whereas the corresponding relationship for forests was weak. Although most of the vegetated areas lacked monotonic trends, nonlinear trajectories were widespread. The dominant increase-then-decrease mode of VR change accounted for approximately one-third of the classified area, indicating that earlier improvements of VR were often unsustainable. Climatic variables accounted for 61.8% of the total model importance for VR, with mean soil moisture contributing to approximately one-third. Soil moisture, vapor pressure deficit, and topography showed threshold-dependent or non-monotonic associations with VR. Grasslands were more sensitive to vapor pressure deficit variability, soil moisture, and soil organic carbon than forests, and the local regression coefficient revealed strong spatial nonstationarity. These findings reveal the complex climatic drivers of VR and provide a scientific basis upon which ecosystem restoration efforts can be optimized across China and beyond.