Decoding the Nonlinear and Interactive Associations of Urban Renewal: Evidence from Explainable Machine Learning in the Yangtze River Economic Belt, China

Urban renewal has become increasingly important as Chinese cities shift from rapid expansion toward the improvement of existing urban areas. However, existing studies have paid less attention to how urban renewal evolves within interconnected regional urban systems and how its associated factors vary across spatial contexts and development conditions. Using panel data for 108 cities in the Yangtze River Economic Belt from 2012 to 2023, this study combines spatial analysis with explainable machine learning to jointly examine regional disparities, spatially conditioned transitions, and nonlinear factor contributions that are difficult to capture within a single analytical framework. The results show that the average urban renewal level increased by 13.85%, while overall disparities remained relatively stable. However, the underlying spatial structure changed substantially, shifting from a pronounced upper–middle–lower regional gradient toward greater within-region heterogeneity and cross-regional overlap. Urban renewal also exhibits strong state dependence, with transition probabilities varying across different neighborhood conditions. Financial development and urbanization show the largest model-based contributions and display nonlinear turning patterns, while technological innovation shows diminishing marginal contributions and industrial co-agglomeration follows a J-shaped pattern. These findings extend the understanding of urban renewal beyond static and spatially isolated assessments by revealing its path dependence, spatial conditionality, and nonlinear contribution patterns. The results highlight the need to align renewal priorities and resource allocation with cities’ existing renewal conditions, development stages, and surrounding urban contexts rather than relying on uniform renewal strategies.