Rethinking mixed land use measurement and its driving mechanisms: Beyond traditional frameworks and linear assumptions

Spatial fragmentation and resource misallocation stemming from functional segregation have emerged as critical challenges to sustainable urban development. As a strategy that combines efficiency with sustainability, mixed land use (MLU) has garnered significant attention. However, existing MLU quantitative models are mainly based on traditional entropy calculations and its result interpretations also follow linear attribution paradigms. Herein, we introduced a measurement framework that combined gravity model and entropy index to quantify MLU among different land use types, considering the diversity of both spatial structure and interaction intensity. Then, we applied the XGBoost model with SHAP value to decouple the contributions of multiple driving factors, and used restricted cubic spline (RCS) to unveil their nonlinear effects. Results revealed a prominent east–west gradient in the degree of MLU across Chinese urban built-up areas. Notably, cities such as Beijing and Chongqing exhibited disruption effects driven by policy and topography factors, respectively, highlighting significant local heterogeneity. Urban shape index and annual housing prices were identified as core drivers among the nine selected factors, signifying that compact spatial forms and moderate economic density contributed most to MLU. Significant Shapley Interaction observed among the factors influencing MLU, indicates that their effects are interdependent rather than isolated. Moreover, inverted U-shaped curves by RCS method showed a nonlinear response between most factors and MLU, demonstrating that their optimization effects on MLU occurred within a specific range, while excessive intensification might suppress the diversity. The proposed framework could offer quantitative insights for understanding MLU, thereby providing actionable “decision coordinates” for resolving land use conflicts and promoting urban sustainability.