Multiscale nonlinear patterns of traditional villages across the Yangtze River Economic Belt

Abstract The Yangtze River Economic Belt covers approximately 2.05 million km², or 21.4% of China’s land area, yet contains approximately 58.1% of the country’s nationally recognized traditional villages. Their conservation is situated within contrasting environmental gradients, development contexts, and heritage–tourism resource configurations. Single-scale and linear frameworks are insufficient for characterizing nonlinear associations and conditional heterogeneity. Using 4,741 nationally recognized traditional villages and 83,584 5-km grids, we combined spatial statistics, adaptive geographically weighted regression and interpretable machine learning to characterize spatial patterns, contribution structures and response forms. Village catalog density was significantly clustered, with major high-value areas in interprovincial mountain areas upstream and in hilly cultural landscapes downstream. Long-term mean annual precipitation, gross domestic product (GDP) per capita, elevation and terrain relief retained relatively high contributions in temporal and scale comparisons, although their rankings and responses varied across river reaches and density strata. Key variables showed nonlinear responses. Compared with weighted scenic accessibility, the colocation of cultural relic protection-unit density with village catalog density was stronger, while intangible cultural heritage richness provided complementary information on the living-heritage context. These findings support an interpretation of the Yangtze River Economic Belt as a segmented cultural-landscape network and provide evidence for monitoring, value review and differentiated planning and heritage responses for traditional villages.