Urban forest fragmentation can weaken ecological connectivity and reduce the spatial continuity of cooling benefits, highlighting the need for structure-function integrated planning. Using 2020 as the baseline year, this study combined the Markov Chain-Future Land Use Simulation model, Morphological Spatial Pattern Analysis, the Minimum Cumulative Resistance model, and Kernel Density Estimation to simulate the evolution of Hangzhou’s forest-dominated ecological network under Business-as-Usual, Economic Development, and Ecological Conservation scenarios for 2030. A Random Forest model with Shapley Additive Explanations was further used to quantify the nonlinear relationship between landscape spatial context and land surface temperature. The results show that the Economic Development and Business-as-Usual scenarios intensify forest fragmentation and weaken the structural integrity of potential ecological corridors, whereas the Ecological Conservation scenario improves network stability through patch amalgamation, core area expansion, and reduced breakpoint density. Land surface temperature was strongly associated with built-up land adjacency. Built-up areas near forest patches showed a maximum relative cooling benefit of 8.04 degrees Celsius, and corridor cooling effects were most pronounced within 300 m, remaining detectable up to about 900 m under the Ecological Conservation scenario. These findings support structure-oriented planning for urban forest conservation and heat mitigation.