Data-driven optimization of facade material combinations for enhancing urban microclimates and pedestrian thermal comfort

This study proposes a data-driven approach for optimizing building façade material combinations to enhance urban microclimates and pedestrian thermal comfort. In contrast to conventional analyses that focus on single materials, this study integrates ENVI-met simulations with a Random Forest Regressor machine learning model to quantitatively evaluate thermal conditions in dense urban settings. A case study was conducted in a high-density commercial corridor between Sinnonhyeon and Gangnam Stations in Seoul, South Korea. Results showed that brick façades increased pedestrian-level temperature owing to high heat storage, and aluminum façades reduced thermal comfort due to high solar reflectance. In contrast, green wall façades demonstrated significant cooling effects, by approximately 1.29 °C and 0.73 °C, respectively. Accordingly, a machine learning model was developed to predict the physiological equivalent temperature (PET), mean radiant temperature (T MRT ), and surface air temperature using meteorological conditions (air temperature, humidity, wind speed, and solar radiation) and façade material properties (albedo, thermal conductivity, and emissivity). The model achieved high predictive accuracy, with R 2 values exceeding 0.98 for PET and T MRT . The optimal façade material configuration for minimizing PET was determined as albedo of 0.27, thermal conductivity of 0.00296 W/m·K, and emissivity of 0.99. This study offers practical guidance for selecting optimal façade material that improves urban pedestrian thermal environments and demonstrates the utility of combining ENVI-met simulations with machine learning for urban microclimate control. The proposed approach can guide future material policies and guidelines, particularly in high-density pedestrian zones, and expanded through empirical validation in diverse urban contexts. • Coupled ENVI-met with Random Forest to optimize façade material impacts. • Brick increased pedestrian heat via storage; aluminum raised PET by reflection. • Green walls reduced wall temperature by 11.7 °C and PET by 1.29 °C. • Model predicted PET and T MRT accurately (R 2 ≈ 0.98 and 0.97); air temperature dominated. • Design target: higher albedo and emissivity, lower thermal conductivity.