Urban Heat Island Mitigation through AI-Optimized Green Infrastructure Planning

摘要

Urban heat islands (UHIs) elevate city-center temperatures by 2-8°C relative to surrounding rural areas, increasing cooling demand, mortality, and energy consumption. We develop an AI-driven green infrastructure optimization framework that integrates high-resolution land surface temperature (LST) satellite data, urban morphology parameters, and a graph neural network (GNN) surrogate model trained on 15,000 CFD thermal simulations. Applied to three megacities — Jakarta, Madrid, and Houston — the framework identifies optimal spatial configurations of urban trees, green roofs, and cool pavements that maximize UHI reduction per unit investment. The optimized plans achieve predicted daytime LST reductions of 1.8-3.2°C in hotspot districts, corresponding to 12-22% reductions in peak cooling electricity demand, while maintaining equitable green space distribution across low- and high-income neighborhoods.