Spatial differentiation and driving mechanisms of Tibetan and Qiang traditional villages in northwestern Sichuan, China

Northwestern Sichuan Province in China, characterized by its unique geography and rich ethnic culture, hosts numerous Tibetan–Qiang Traditional Villages (TTVNS) of significant conservation value. These villages exhibit pronounced macro-level spatial inequalities, making the study of their spatial differentiation and the quantification of driving mechanisms essential for implementing targeted conservation strategies. This study examines 412 TTVNS using Kernel Density Estimation (KDE), accessibility-based clustering, Optimal Parameter Geographical Detector (OPGD), Multiscale Geographically Weighted Regression (MGWR), and Random Forest with SHAP (RF–SHAP) to identify spatial patterns and quantify the mechanisms influencing them. The results indicate that: (1) TTVNS display a “multi-core clustering with contiguous distribution” pattern, forming two northeast–southwest aggregation belts, with significant positive spatial autocorrelation (Moran’s I = 0.175); (2) in single-factor detection, Cultivated Land (X3, q = 0.3019), Population (X19, q = 0.2510), and Road Density (X13, q = 0.2303) are the strongest explanatory factors, while the strongest interaction occurs between Temperature (X9) and Population (X19, q = 0.406), demonstrating that spatial differentiation is driven by the coupling of environmental suitability and social support rather than any single factor; (3) model comparison shows that MGWR substantially outperforms OLS and GWR, with R² increasing from 0.3049 and 0.3246 to 0.6568, confirming pronounced spatial non-stationarity in the driving mechanisms of TTVNS; (4) different categories of drivers exhibit distinct pathways: nature and geography influence aggregation through cultivated land, river systems, and temperature; location and transportation shape a “node–corridor” settlement pattern along valley corridors; humanities and religion affect clustering through the coupling of religious nodes and local social structures; society and economy reshape contemporary spatial patterns primarily via tourism development and population dynamics. These findings demonstrate that the spatial differentiation of TTVNS emerges under long-term natural constraints and religious-cultural networks, and continues to be reshaped by modern transportation and tourism development. The study provides a quantitative basis for designing differentiated and context-specific conservation strategies.