Spatial patterns and natural–social associations of ethnic minority toponyms in Xizang, China

Ethnic Minority Toponymic Cultural Heritage (EM-TCH) preserves local knowledge, cultural memory, and human–environment relationships, yet its semantic structure and spatial organization remain insufficiently understood. Based on 714 township-level place names in Xizang, this study integrates morphemic decomposition, semantic trace extraction, spatial statistics, and explainable machine learning to examine spatial patterns and natural–social associations. Thirteen semantic traces are grouped into four toponymic types: Nature Observation, Economic Activity, Social Organization, and Faith. Kernel Density Estimation, Standard Deviational Ellipse, and K-means clustering reveal differentiated patterns: Nature Observation toponyms are widely dispersed, Economic Activity toponyms form river-valley clusters, and Social Organization and Faith toponyms concentrate near administrative, transport, and religious centers. SHAP-interpreted XGBoost models show that vegetation, elevation, river proximity, grazing intensity, nighttime light intensity, and religious-site density are closely associated with these patterns. Township-level toponyms in Xizang act as socio-ecological archives, informing ethnic minority toponymic heritage protection.