Assessing seasonal building thermal adaptation through the acclimatization distance: a GIS-based machine learning framework

Introduction Enhancing the capacity of residential buildings for maintaining thermal comfort under changing climate conditions is crucial to reducing energy demand and increasing urban resilience. While extensive research has been conducted on energy efficiency at the building scale, less attention has been paid to how the building’s characteristics and the urban morphological context interact to modulate adaptive comfort in different seasons. This study aims to develop and evaluate a spatial machine learning framework to quantify seasonal building thermal adaptation. Methods To this end, a novel indicator termed acclimatization distance ( d acc ) is introduced to measure the thermal gradient between outdoor and adaptive comfort temperatures that incorporates correction factors for internal loads and thermal inertia. Using a dataset of 4,954 residential buildings in Zaragoza (Spain), intrinsic indicators and extrinsic urban morphological variables were processed using Geographic Information Systems (GIS) and integrated within a Random Forest Regression (RFR) to evaluate seasonal drivers of thermal adaptation. Results The RFR models achieved uneven predictive performance ( R 2 = 0.86–0.87 for summer, R 2 = 0.25–0.29 for winter), revealing a clear seasonal asymmetry: morphological factors dominated thermal adaptation during cooling periods, whereas intrinsic building attributes became more influential in heating conditions. Discussion The proposed framework bridges adaptive comfort, urban morphology and building performance, offering a transferable approach to assess thermal adaptation capacity at the city scale. These findings highlight the need to integrate morphological design strategies with building retrofits to strengthen urban resilience and mitigate cooling demand under future climatic stress.