Heat Stress Prediction in Glasgow: Integration of Historical Data With Machine Learning Models

ABSTRACT Urban heat stress is intensifying because of climate change, particularly in densely constructed areas where urban morphology significantly affects thermal conditions. This research examines how vegetation and built form influence heat stress in Glasgow by integrating historical geospatial data, Artificial Neural Network (ANN)‐based temperature prediction, and ENVI‐met microclimate simulations across multiple greening scenarios applied to vacant and derelict land. ANN notably outperformed Generalized Linear Regression in predicting Land Surface Temperature (LST) and Mean Radiant Temperature (MRT), reflecting the greater sensitivity of MRT to fine‐scale radiative and geometric processes. Essential predictors included Normalized Difference Vegetation and Built‐up Index, Sky View Factor, and Building Surface Fraction. Scenario‐based greening simulations showed that 100% vegetation on vacant lands significantly reduces local thermal conditions, with LST decreasing by 8.7°C and the Universal Thermal Climate Index (UTCI) by 3.3 K. In contrast, full conversion to built‐up surfaces increases LST by 1.2°C and raises UTCI by up to 6 K. Despite these pronounced local effects, the net cooling impact at the citywide scale remains limited. Results highlight the necessity of site‐specific greening strategies and high‐resolution environmental data to enhance predictive precision and support the development of resilient, thermally adaptive urban environments.