Apparent temperature is the preferred measure of hotness or coldness expressed to depict the human sense. Spatially explicit measurement of the hourly apparent temperature is essential for capturing the threats to bioclimatic comfort and preventing potential mortality/morbidity risk from heat or cold. However, existing apparent temperature products only provide daily observations at the spatial resolution of several dozen kilometers, resulting in some substantial underestimations for some life-threatening thermal stresses highly localized in space and time. Furthermore, some data-driven models lack mechanical constraints on the turbulent exchange between the surface and the atmosphere, making some unsatisfactory accuracy. Here, we propose Humidex reconstruction model incorporating atmospheric dynamics theory and aerodynamic parameters (i.e., heat and momentum roughness lengths for natural surfaces and three urban canopy geometry parameters for artificial surfaces), capable of developing an hourly dataset at fine-grained spatial resolution (0.01° × 0.01°). In this study, a total of 2952 h in four seasons were selected to test the seasonal performance of this model, taking the Yangtze River Delta as an example. The results show that the Humidex products from this model generally outperform the existing comparable products, with the hourly population root mean square error (RMSE) ranging from 1 to 2 °C in winter and autumn and 2–3 °C in spring and summer. Moreover, the constraint of aerodynamic parameters can reduce RMSE with a significant margin for each season, up to 2 °C, especially in areas with dense woodlands or buildings. In addition, the results demonstrate the excellent performance of this model in capturing short-lived thermal health threats, which are easily overlooked when observed data only provides a daily variation. This indicates that the model can allow researchers and practitioners investigate the fine-grained spatial and temporal evolution of thermal stress and its impact on public health, tourism, learning, and work performance.
Read full abstract