Urban temperature increases present significant challenges, necessitating the identification of key influencing factors and developing high-resolution temperature maps for effective climate analysis. This study constructs a 100 m-resolution near-surface air temperature map for Seoul, South Korea, by integrating “a distance decay regression selection strategy (ADDRESS)” with “regression Kriging (RK)” to enhance spatial temperature modeling. The analysis incorporates land use and land cover, building morphology, and geographical and topographical factors to determine the spatial influence range of each variable. The study uses smart Seoul urban data sensor (S-DoT) network data to ensure dense, spatially continuous observations, improving model accuracy. Results indicate that temperature influences vary across spatial extents, with green spaces, water bodies, and agricultural land exerting strong cooling effects, while built-environment factors such as gross floor area and building coverage ratio contribute to urban heat. While RK effectively reduces spatial bias in residuals, it does not significantly enhance numerical accuracy compared to the regression model. The final model achieved an RMSE of 0.4704°C and an MAE of 0.3517°C based on 20-fold cross-validation. These findings underscore the importance of optimal variable selection and spatial influence assessment in urban temperature modeling. The study provides valuable insights for urban planners and policymakers in mitigating urban heat and identifying heat-vulnerable regions. However, limitations such as incomplete airflow representation and data gaps in suburban areas remain. Future research should incorporate additional environmental factors, such as shading effects and wind dynamics, and expand sensor coverage to improve predictive accuracy.
Lee et al. (Tue,) studied this question.