This study examines changes in summer thermal comfort across 12 European cities in 2000, 2010, and 2020 by integrating multi-source geospatial and demographic data. We introduced the urban eco-climatic livability index (UECLSI), which combines 19 indicators related to climate, urban form, vegetation, and population. Collinearity was tested using the variance inflation factor and tolerance, and principal component analysis was applied for dimensionality reduction. To model spatial patterns, we compared three approaches: (i) geographical Gaussian process regression (GGPR), (ii) multiscale geographically weighted regression (MGWR), and (iii) Extreme Gradient Boosting (XGBoost). Model interpretation was supported by SHapley Additive exPlanations (SHAP/GeoShapley), while spatial associations were analyzed with GeoDetector, bivariate classification, and spatial standard deviation ellipses. Results reveal a general decline in surface urban heat island (SUHI) intensity over two decades in most cities (e.g., Budapest, Vienna), though some cities (Bratislava, Ljubljana) showed fluctuations. UECLSI was highest in 2000 for Sofia, Bucharest, and Prague, decreased in 2010, and improved again by 2020 in cities such as Bucharest and Belgrade, while Ljubljana recorded the lowest values in 2020. Model evaluation showed that GGPR provided the highest accuracy (R2 ≈ 0.99), followed by MGWR, while XGBoost performed the worst. SHAP analyses highlighted a growing role of climatic variables and a declining importance of vegetation indices by 2020. GeoDetector confirmed significant interactive effects between UECLSI and SUHI. Overall, the findings emphasize the importance of city-specific strategies focused on climate adaptation, vegetation planning, and urban form design to mitigate SUHI and improve urban livability.
The alternative text for this image may have been generated using AI.