A Prototype Earth Observation-based Framework for Simulating Urban Intervention Effects on Land Surface Temperature

The surface urban heat island phenomenon poses critical challenges for climate-resilient urban planning, due to its well-known negative impact on thermal comfort and increased health risks during heatwave events. This work presents a prototype Earth Observation-based framework for the simulation of Land Surface Temperature (LST) to support scenario-driven assessment of urban transformations. The framework integrates Landsat 8/9 LST observations from 2015 to 2024 with Sentinel-2 reflectance data, used to derive surface material fractions through spectral unmixing, together with a set of Urban Canopy Parameters including sky view factor, building height, tree canopy height and soil imperviousness. These inputs form a comprehensive set of predictor layers that describe urban morphology and surface materials, and which can be modified in typical urban planning or retrofitting interventions. Random Forest regression models are trained to predict seasonal average LST from the predictor variables. The Metropolitan City of Milan is used as a case study. A Python module is designed to allow users to delineate intervention areas and adjust predictor layers to generate what-if scenarios that mimic real urban interventions. The summer model achieves the best performance, with a mean absolute error lower than 2 K, and the accuracy assessment shows good agreement with Landsat observations (R² > 0.74) and moderate agreement with ECOSTRESS data (R² > 0.63). An example application illustrates the potential of the framework for the analysis of urban heat mitigation strategies and its future integration into operational urban digital twin systems.