Interpreting complex relationships between urban and meteorological factors and street-level urban heat islands: Application of random forest and SHAP method

Abstract

Urban and meteorological factors strongly affect street-level urban heat islands (UHIs), but few studies have considered their interactions under varying weather conditions. This study investigated the relationship between urban and meteorological factors and street-level UHI intensity in Seoul during daytime and nighttime in summer. UHI intensity was calculated from urban air temperatures measured by 568 street-level sensors. Random forest regression models and Shapley Additive exPlanations (SHAP) method were used to quantitatively analyze nonlinear relationships and interaction effects of the predictors. The results indicated that meteorological variables, particularly regional air temperature, significantly influenced UHI intensity during both daytime and nighttime. Furthermore, urban factors such as building coverage ratio and pervious ratio became more important during nighttime. Both meteorological and urban variables indicated nonlinear relationships with UHI intensity, with some showing threshold effects. Compared to total effects, main effects were significantly smaller in magnitude and range due to high parameter interactions among variables. For most variables, the sum of interaction effects outweighed main effects. In particular, notable interaction effects were observed within each meteorological and urban category. These results pinpoint that effects of urban variables are important individually and in combination with other variables. The findings highlight the importance of designing effective mitigation strategies that account for both nonlinear relationships of individual factors influencing UHI and interactive influences of multiple factors.