Unveiling the factors influencing Sakya monastery distribution via interpretable machine learning

Tibetan Buddhism has deeply influenced the political, cultural, and spatial fabric of the Tibetan Plateau. Among its various traditions, the Sakya school is notable for its early theocratic regime. However, the temporal and spatial evolution of its monasteries—as key vehicles for the school’s spread—and the quantitative mechanisms behind their distribution remain unclear. This study combines GIS-based spatial analysis with interpretable machine learning (LightGBM-SHAP) to identify the key natural and human factors shaping Sakya monastery distribution. Emphasis is placed on capturing nonlinear interactions among variables to uncover deeper spatial patterns. Results reveal that human factors—particularly historical population, settlement density, and the presence of other sects—exert a stronger influence than natural conditions. The findings underscore the complex socio-political dynamics embedded in monastic diffusion and highlight the potential of explainable machine learning for advancing research in historical GIS (HGIS) and religious geography.