Grey-green spatial morphology relates to summer daytime and nighttime land surface temperature across local climate zones in Wuhan.

Urban grey and green spaces jointly shape the urban thermal environment, yet their associations with land surface temperature (LST) are spatially heterogeneous and context-dependent. Existing studies have mainly focused on the city-wide scale or on individual spatial elements, with limited attention to the differentiated mechanisms through which grey-green space affects LST across Local Climate Zone (LCZ). This study developed a multidimensional grey-green indicator system integrating two-dimensional patterns, three-dimensional structures, and morphological spatial characteristics. Multiple machine learning models were then used to examine the marginal effects and nonlinear interactions of grey-green indicators on daytime LST, nighttime LST, and composite daytime-nighttime LST difference (LST_DN) across LCZ types. The optimal model explained more than 80% of LST variation. XGBoost-SHAP analysis revealed day-night transition in the contributions of grey-green indicators. Grey-space indicators contributed more strongly to model-predicted daytime LST and LST_DN, with B_PLAND, BD, and BH contributing 45.69% on average. By contrast, green-space indicators played a stronger role at night, G_PLAND contributed 30.3% and 28.1% in open low-rise and natural zones, respectively, and each 0.5 increase in G_PLAND reduced nighttime LST by about 0.6-1.2 °C. The combined effects of grey and green spaces differed significantly across LCZs. Compact built-up zones were mainly characterized by synergistic warming among grey-space indicators, whereas in open built-up and natural zones, increasing G_PLAND above approximately 0.6 reduced the positive SHAP contribution associated with high B_PLAND. These findings provide scientific support for LCZ-based thermal governance and refined grey-green space optimization in high-density cities.