Causal and threshold-dependent interactions of blue-green patterns in urban thermal regulation

Inconsistent findings regarding the impact of urban blue-green space patterns on regional thermal environments hinder the practical application of research in landscape ecology. This gap largely stems from a predominant focus on the global effects of landscape metrics, while their nonlinear inflection points, pairwise interactions, and underlying causal pathways have been insufficiently explored. This study aimed to (1) quantify the nonlinear and threshold effects of blue-green pattern metrics on land surface temperature (LST), (2) identify critical interactions between these pattern metrics, and (3) assess the direct and indirect pathways through which key drivers influence LST. The study was conducted in Suzhou, a humid subtropical canal city in China. We employed XGBoost-SHAP to decipher nonlinear effects and interactions, and further constructed a structural causal model (SCM) to quantify the direct and indirect pathways influencing LST. Blue-green pattern metrics effectively predicted seasonal land surface temperature (R 2 = 0.667–0.830). Water coverage (PLAND W ) was the dominant cooling factor across all seasons, followed by grassland coverage (PLAND G ) except in winter. Both water and forest coverage exhibited activation thresholds (2% for water; 5.32%–8.14% for forest) beyond which their cooling effects increased linearly. Crucially, PLAND W and PLAND G acted as key moderating factors. Exceeding 10–12% PLAND W shifted the effect of forest metrics from negative to positive (indicating changes in the sign of the marginal effects), while reversing the effect of patch densities from positive to negative. Similarly, exceeding 26–29% PLAND G inverted the effect of water metrics from negative to positive. SCM results further confirmed that PLAND W exerted the strongest direct cooling effects on LST, consistent with the SHAP-based findings. The identified thresholds and causal pathways provide quantitative references for landscape planning and support more precise urban thermal management to mitigate heat.