Spatial Diagnosis of Climatic and Landscape Controls on Forest Leaf Area Index Across China Using Interpretable Machine Learning

Forest cover condition is a key determinant of ecosystem functioning and ecological resilience, yet its spatial variability across large and environmentally heterogeneous regions remains insufficiently understood. Leaf area index (LAI) provides a continuous and physically meaningful indicator of forest canopy condition, reflecting variations in canopy density associated with climate and landscape structure. Here, we develop a spatially explicit and interpretable analytical framework to diagnose the dominant climatic and landscape controls on forest cover condition across mainland China during 2000–2020. By integrating machine-learning modelling with SHapley Additive exPlanations, GeoDetector interaction analysis, and nonlinear dependence diagnostics, we quantify the relative contributions and interactions of precipitation, temperature, topography, and forest landscape structure to spatial patterns in forest LAI. The results reveal pronounced spatial heterogeneity in forest cover control regimes. Precipitation dominates forest cover condition in humid regions but exhibits nonlinear saturation, whereas forest fragmentation strongly constrains canopy development and moderates climate-LAI relationships in arid and semi-arid forested landscapes. In high-elevation regions, topographic and thermal factors exert primary control. Overall, the findings demonstrate that forest cover condition reflects climate-conditioned and landscape-dependent control regimes, providing a transparent basis for large-scale forest cover assessment and ecological monitoring.