This study investigates a typical valley city, Lanzhou, China, to reveal the nonlinear relationships and interaction mechanisms between gray–green space morphology and seasonal diurnal land surface temperature (LST) using multi-source remote sensing and land use data. A comprehensive morphological indicator system encompassing scale, complexity, connectivity, and structural integrity was constructed through landscape metric screening and the CRITIC objective weighting method, combined with the XGBoost-SHAP explainable machine learning framework. The findings highlight that: (1) Gray–green space impacts on LST exhibit significant seasonal and diurnal variations—daytime LST is predominantly governed by gray space morphology (e.g., fragmentation degree), while nighttime LST is driven by green space morphology (e.g., coverage intensity). (2) Key indicators demonstrate pronounced nonlinear and threshold characteristics: the cooling effect of green space coverage intensity (GCI) saturates beyond 0.25; gray space morphological structure factor (GRMSF) demonstrates cooling potential when exceeding 0.25, mitigating its warming effect. (3) Significant synergistic interaction effects exist between gray and green spaces. Interaction analysis reveals that “high green coverage with low structural connectivity of gray space” produces optimal synergistic cooling effects, representing the most effective spatial configuration for nighttime LST mitigation. This study deepens theoretical and methodological understanding of the complex relationships between spatial morphology and thermal environments, providing quantified, temporally differentiated spatial optimization guidance for climate-adaptive planning in valley cities.
Li et al. (2026) studied this question.