Comparison of CMIP6 model performance in estimating human thermal load in Europe in the winter season

In this work, historical simulations of CMIP6 GCMs are evaluated with respect to the ERA5 reanalysis dataset, in order to examine their ability to assess human thermal load in Europe in the winter season. The period of 1981–2010 is chosen for the analysis, and thermal load is expressed via the clothing resistance index (rcl index; expressed in clo). It is found that the GCMs are able to reproduce the areal differences of thermal load satisfactorily, the spatial correlation with the reanalysis is greater than 0.95 in all cases. The effects of the main geographical constraints (latitude, continentality and elevation) are shown by all GCM simulations, as rcl index values are greater at higher latitudes, away from the ocean and in mountainous areas, although GCMs only capture major mountains (the Caucasus, the Armenian Highlands, the Scandinavian Mountains, the Alps). The root‐mean‐square error (RMSE) is around 0.2 clo in all cases, GCMs generally perform better in homogenous lowland areas, while results are less accurate in highlands and mountains owing to the coarse horizontal resolution of GCMs (~1°). The smallest errors occur over central and western Europe and the Mediterranean region, while results tend to be less accurate over the northeastern part of Europe. Biases in the estimation of heat deficit can mainly be attributed to biases in temperature, but biases in wind speed and atmospheric downward radiation seem to be important factors as well. In this study the suitability of CMIP6 GCMs is examined for estimating human thermal load in Europe in the winter season via the clothing resistance index. It is found that the selected GCMs are able to reproduce the areal differences of thermal load satisfactorily, and they also capture the magnitude of it quite well. Simulations are most accurate and reliable over central and western Europe and the Mediterranean region, especially in lowland regions, while they appear to be more ambiguous in mountainous areas with complex topography.