Multimodal Urban Heat Island Mitigation via Knowledge Reasoning and Geospatial Optimization

Urban Heat Island (UHI) mitigation demands curated strategies that are both ecologically effective and spatially feasible. This paper presents a dual-mode framework to support data-informed urban cooling interventions, combining passive knowledge-driven recommendations with active geospatial Land Use/Land Cover (LULC) optimisation. The passive approach leverages a Hybrid Retrieval-Augmented Generation (RAG) pipeline that fuses dense and sparse retrieval over a curated corpus of peer-reviewed research and policy literature. This enables the generation of evidence-based, transparent, and domain-validated textual recommendations for generalised UHI mitigation strategies. Complementing this, the active mitigation pathway integrates computer vision, graph-based learning, and evolutionary optimisation to modify LULC configurations of the region of interest for targeted cooling of upto 3°C. The workflow begins with a cooling efficiency model that quantifies the additional green cover for the user-defined temperature reduction. Subsequently, morphological image processing generates a diverse population of feasible LULC variants by simulating realistic vegetation growth of the estimated amount within feasible urban areas. Finally, a Genetic Algorithm (GA), guided by a Graph Neural Network (GNN) trained to predict UHI intensity from spatial graphs, evolves this population to find optimal layouts that maximise cooling, preserve spatial coherence, and minimise infrastructure disruption. The entire framework is deployed as a web-based geospatial intelligence platform, featuring an interactive 3D visualisation engine for rendering GA-optimised landscapes, quantitative dashboards for assessing the mitigation effectiveness, and an integrated suite of insight tools offering city-to-city comparative analysis, country-wise hotspot identification, and automated report generation. Overall, this work presents an end-to-end, scalable solution that empowers urban planners, policymakers, and researchers with actionable intelligence to build more climate-resilient cities.