Ecological sensitivity measures an ecosystem’s reaction and restoration difficulty to exogenous disturbances. Regional ecological and environmental challenges can be identified using ecological sensitivity evaluation. The regional characteristics and environmental challenges of plateau lakes were quantified to create four assessment indexes: landform, natural conditions, human activities, and ecological security. Eleven ecologically sensitive characteristics were chosen. We evaluated the landscape ecological sensitivity of the Erhai Lake Basin, China, using ArcGIS and the analytical hierarchy process (AHP). The results showed that the Erhai Lake Basin was mostly forest from 1990 to 2020 and that the dynamic land-use attitude increased and then reduced. The cultivated land became mostly woods and grassland. The largest building area was tied to reverting farms to woods and urban construction. The highest weight was from single-factor ecologically sensitive vegetation covering, followed by river systems and roadways, and lowest was from landslides and collapses. The ecologically sensitive areas with more vegetation and a greater distance from roads and woodland are harder to recover from harm. According to the thorough sensitivity analysis, the study region’s high sensitivity area is 1102.36 km² (26.16%) and the higher sensitivity area is 1177.10 km² (27.93%). The ecological sensitivity increases from neighborhoods to nature reserves. High-sensitivity areas were in hilly woods and grasslands with few people. The low-sensitivity area was around water and homes. The dynamic stability of the area ecological environment influenced ecological sensitivity. This study aims to provide sustainable land use solutions for the Erhai Lake Basin and a scientific basis for managing and protecting ecologically vulnerable areas.

Continuous urban expansion and the resulting land use and land cover (LULC) changes significantly exacerbate the urban heat island effect and intensify heatwaves. While the cooling effects of blue–green spaces are widely documented, most studies focus on single landscape types or specific time frames. Few investigations systematically explore the comprehensive thermal regulation mechanisms of gray–green spaces, or their nonlinear driving factors and interactive effects across coupled seasonal and diurnal scales. To address these gaps, this study focuses on Chengdu, a typical expanding city in China, to establish a comprehensive indicator system for urban gray–green spaces. This system encompasses four key dimensions: coverage, fragmentation, aggregation, and morphological spatial pattern. After evaluating 12 machine learning models, the optimal model was selected for further analysis using SHapley Additive exPlanations (SHAP) and Partial Dependence Plots (PDP). This research investigates the nonlinear thresholds and interactive effects of composite gray–green space indicators on land surface temperature (LST) across varying seasonal and diurnal cycles. The results indicate that: (1) The impact of gray–green spaces on LST varies significantly across seasonal and diurnal contexts. Green spaces primarily exert a cooling effect during spring, summer, and autumn, whereas gray spaces dominate heat retention during winter and across all nocturnal periods. (2) The morphological spatial pattern of green spaces (GMSPA) outperforms traditional coverage indicators (G1) in providing cooling benefits across multiple scenarios. (3) The cooling efficiency of GMSPA peaks between −0.8 and −0.5, reaching saturation at 0.53. Conversely, LST exhibits a sharp, step-like increase when gray space aggregation (B3) exceeds −0.58. (4) Optimizing areas with high GMSPA can significantly mitigate heat exposure risks in expanding cities. These findings offer robust theoretical insights and actionable guidelines for spatial planning aimed at thermal resilience, urban thermal environment management, and building energy conservation in rapidly growing urban areas.