Evaluating high-resolution mean radiant temperature within an urban street canopy: Resolving spatiotemporal variations with LiDAR/thermal infrared scanning and data-driven simulation

As extreme weather intensifies, the increasing frequency of summer heatwaves make the study of human thermal exposure essential. Mean radiant temperature (T_mrt) is crucial for assessing human thermal exposure, because it quantifies the largest sources of spatial variability in pedestrian-perceived thermal stress and comfort in complicated urban environments. Despite the availability of existing methods for evaluating T_mrt through measurements and numerical simulations, the lack of detailed urban three-dimensional (3D) models and spatially and temporally resolved pedestrian-level irradiance from urban surfaces poses a significant challenge in obtaining high-resolution T_mrt data. This paper introduces a methodology that combines LiDAR and thermal infrared scanning with data-driven simulations. The approach was applied to a street canyon segment in Salt Lake City during the summer, to enable assessment of high spatial resolution (0.3 m²) shortwave and longwave radiant fluxes of urban surfaces at different periods. Based on the refined 3D radiation field, different irradiance sources received by the human body at different locations was sampled and then a high-spatial-resolution field (0.5 m²) of pedestrian-level sampled T_mrt (T_{mrt_sampled}) was generated. Such a method for calculating T_{mrt_sampled} is efficient, requiring only 30 seconds of computational time for each simulated instant. Results indicated significant variations of  across heterogeneous urban spaces, with the largest difference exceeding 35 ℃. Spatiotemporal variations in longwave irradiance from urban surfaces significantly influenced T_{mrt_sampled}. Exposure of ground and wall materials to direct sunlight, coupled with their substantial thermal  inertia, drove peak human thermal stress by 17:00. Furthermore, T_{mrt_sampled} was compared with SOLWEIG-simulated  (T_{mrt_simulated} ) for the same meteorological conditions. Due to differences in mesh and mechanism for quantifying T_mrt,  T_{mrt_sampled} values typically were 4 ~ 6  ℃  higher than T_{mrt_simulated} over sunlit surfaces, and their root mean square error reached 4.71  ℃ when the solar elevation was high and ground shadows were minimal.