Enhancing thermal resilience of heat–cold island networks in lake-ring urban agglomerations using machine learning and multi-scenario simulations

Under rapid urbanization and climate warming, urban heat island (UHI) effects in lake-ring urban agglomerations have intensified. Prior studies often emphasize single patches or local scales, obscuring regional heat–cold connectivity and limiting network-based resilience planning under future scenarios. This study integrates morphological spatial pattern analysis (MSPA) with circuit theory to construct a heat–cold island (HCI) composite network comprising sources, corridors, and nodes. In this framework, machine learning characterizes nonlinear thermal-connectivity resistance, while the patch-generating land-use simulation (PLUS) model simulates HCI network reorganization under multiple 2030 scenarios. A land surface temperature (LST)-derived surface thermal structural resilience framework was then developed to link historical and future network changes. Evidence from the Taihu Lake urban agglomeration indicates that from 2000 to 2020, heat island (HI) sources expanded rapidly across the eastern lakeside built-up area, with marked increases in corridors and nodes. In contrast, cold island (CI) networks became increasingly fragmented. HCI composite nodes shifted from the Taihu Lake shoreline toward the western ecological transition zone. By 2030, SSP585 formed a strong heat-connectivity network, and cooling connectivity pathways were nearly lost. Under SSP119, barriers to cold-source connectivity were reduced, and CI corridors increased. The surface thermal structural resilience improvement index (TR) showed strong structural dependence, remaining negative historically and declining further under SSP245 and SSP585. By contrast, SSP119 increased TR to −0.134 through cooling-corridor recovery and key-node reorganization, approaching a more balanced network structure. This study reframes LST-based surface thermal structural resilience from a whole-network HCI perspective and provides guidance for climate-adaptive spatial planning.