Predicting the impact of the urban texture on the Urban Heat Island intensity using Machine Learning: The case for the Iberian Peninsula

Urbanization increases impervious surfaces that store heat, concentrates residual heat sources, and alters atmospheric ventilation. This makes cities warmer than surrounding rural areas, especially at night, a phenomenon known as Urban Heat Island (UHI). UHI intensity depends on meteorology and urban morphology, including building height, population density, green cover, and built-up area. However, estimating each factor’s contribution remains challenging due to variability. Furthermore, defining the intensity of the UHI as the temperature difference between urban and rural areas introduces arbitrariness, since factors such as altitude influence the results. To address these issues, we propose a systematic reference temperature definition and a machine learning model to predict daily maximum UHI intensity (UHII). Trained on UHII data from multiple rural–urban pairs, the model uses 100×100 meter hourly raster data sets of meteorological and urban morphology features from 11 Iberian Peninsula cities in 2017. Assuming weak temporal and spatial correlations, it operates independently of past time steps and neighboring cells, ensuring a fully local approach. The performance of the test across cities and a hybrid synthetic region exceeds R 2 = 0 . 8 . The main findings indicate that the average rate of change on the daily maximum UHII is 0.09 °C per meter of building height, 0.34 °C per 0.1 point increase in build-up fraction, 0.08 °C per 1000 people/m2 and-0.11 °C per 0.1 point increase in vegetation fraction. Focused on the Iberian Peninsula, the model accounts for climatic vulnerabilities, helping urban planners in designing strategies to protect public health and improve resilience. • Systematic reference temperature defined to improve UHI intensity estimation. • Machine learning model accurately predicts daily maximum UHII across cities. • Findings support urban planning strategies for heat mitigation and resilience. • Role of building height, greenness and built-up fraction on UHII are estimated. • PCA identifies grouped features based on their role on increased temperatures.