Landscape-socioeconomic feedback mechanisms in Sichuan's river basins

Landscape fragmentation and socio-economic systems interact in complex, bidirectional ways. This study examines three Sichuan river basins with similar natural conditions but varying socio-economic gradients using land use and socio-economic data (2000-2020). Analyses combining PCA, rolling-window Granger causality, and Random Forest-SHAP models revealed distinct basin-specific patterns: Jialingjiang exhibited the highest fragmentation, Minjiang the optimal landscape integrity, and Tuojiang showed transitional traits. Causality tests identified land-use-dependent interactions, with cropland acting as a "stable buffer," forest as "context-sensitive," and grassland as "dynamically vulnerable." SHAP analysis highlighted population density as the main socio-economic driver of fragmentation, while edge density most affected socio-economic outcomes, with policy-driven population agglomeration altering traditional correlations. The findings provide a scientific basis for targeted sustainable management in river basins.