Promoting or inhibiting? Unraveling the nonlinear and spatially heterogeneous effects of land use efficiency on ecological resilience: Evidence from the Taihang Mountains Region, China

Abstract

Ecological resilience is fundamental to sustainable urban development, yet how land use efficiency affects ecological resilience remains insufficiently understood. Prior research has rarely accounted for the nonlinear dynamics and spatial heterogeneity embedded in ecological resilience response mechanisms. To address these gaps, this study develops a geospatial explainable artificial intelligence framework that integrates CatBoost, SHAP, and GeoShapley, applying it to 22 cities in the Taihang Mountains Region. Ecological resilience is assessed through the “Scale–Density–Morphology” dimensions. Findings reveal a spatial gradient of “high in the central–north and low in the southeast,” with an overall decline from 2000 to 2020, most pronounced in rapidly urbanizing cities. The CatBoost model achieved strong predictive accuracy (R² = 0.953). COHESION and population efficiency were identified as dominant associated factors, while introducing spatial location enhanced the explanatory power of land use efficiency variables. Nonlinear and spatially heterogeneous responses were evident: GDP efficiency was positively associated with ecological resilience within an moderate range (approximately 0.2–0.9), while exhibiting a U-shaped tendency at higher values. K-means clustering distinguished three mechanism patterns, providing the basis for differentiated policy pathways. These insights underscore that land use efficiency and landscape configuration are closely associated with ecological resilience, and demonstrate the potential of the geospatial explainable artificial intelligence framework to support spatial planning, land governance, and targeted regional development strategies toward sustainable urban futures.