Assessment of Mean Radiant Temperature Using Reanalysis Data

Mean Radiant Temperature (MRT) is a key parameter for assessing thermal comfort and heat stress in urban environments, yet its large-scale estimation remains challenging. This study compares MRT estimates derived from ERA5-Land reanalysis data, processed on the Google Earth Engine (GEE) platform, with 1,786 in-situ reference measurements collected in urban canyons in Curitiba, Brazil. The estimation methodology employs a Stefan-Boltzmann-based model, using ERA5-Land radiative fluxes and simplified biometeorological parameters. Linear regression analysis shows a strong temporal correlation between the datasets (r = 0.87; R² = 0.758), indicating that the reanalysis approach successfully captures trends in MRT variation. However, systematic underestimation and considerable absolute error are observed (RMSE = 9.65°C), with an attenuated model response (slope = 0.58). The main source of discrepancy is the scale mismatch between mesoscale reanalysis data, which smooths thermal extremes, and the heterogeneous urban microclimate. Despite local-scale inaccuracies, the study validates the methodology as a strategic, accessible, and low-computational-cost tool for monitoring heat-stress trends at regional and global scales, with significant applications for climate planning and public health policies, particularly in data-scarce regions.