Nonlinear relationships between land surface temperature and influencing factors along the urban–rural gradient from a seasonal–diurnal perspective: A SHAP-based framework in the Gui'an region

Understanding the spatiotemporal differentiation of land surface temperature (LST) across urban-rural gradients and its complex driving mechanisms is critical for advancing urban thermal dynamics research and informing climate mitigation strategies. However, existing studies focus on linear analyses or singular temporal scales, neglecting the seasonal-diurnal evolution of LST and its nonlinear interactions within diverse topographical conditions. This study investigates the Guiyang-Anshun region, a representative karst terrain area, and develops an analytical framework to explore the driving mechanisms of LST through a “seasonal-diurnal” spatiotemporal perspective. Urban-rural gradient zones are delineated using nighttime light data (NTL) quantile regression, combined with Bayesian-optimized XGBoost models and SHAP to systematically uncover nonlinear driving patterns and spatial interactions. Key findings include: (1) The Gui'an region exhibits obvious urban-rural LST gradients, with urban cores exhibiting significant diurnal temperature differences and intense summer fluctuations, while suburban and rural areas experience smaller fluctuations and greater thermal stability; (2) LST influencing mechanisms display significant temporal variability across seasonal-diurnal scales; (3) The relative importance of factors shifts across the urban-rural gradient zones, with natural factors gaining prominence as urbanization decreases, reflecting a transition from “anthropogenic-driven” to “terrain-ecological control” mechanisms; (4) The influence of factors exhibit notable nonlinear and threshold effects, demonstrating varying effectiveness and response intensities across regions; (5) Interaction analyses reveal that LST variation arise from comprehensive multi-factor coupling interactions, rather than linear superposition of individual variables. This study provides a novel framework for understanding the intricate dynamics of LST and offers practical guidance for sustainable urban planning and climate adaptation.