Residual analysis of UTCI predictions on outdoor thermal sensation survey data

The Universal Thermal Climate Index UTCI assesses the interaction of ambient temperature, wind, humidity and radiant fluxes on human physiology in outdoor environments on an equivalent temperature scale. It is based on the UTCI-Fiala model of human thermoregulation and thus also allows for thermal comfort prediction. Comparing UTCI predictions to thermal sensation votes recorded on the 7-unit ASHRAE scale in outdoor comfort surveys with 1685 respondents in Curitiba and 567 in Glasgow, respectively, yielded negligible bias and less than one unit root-mean squared error (rmse) for Curitiba, but a noticeable underestimation of actual thermal sensation votes (bias=-0.73) with increased rmse=1.44 in Glasgow. Residual analyses revealed that the factors age, gender, body composition, site morphology (open space, street canyon), climatic state (comfort/discomfort) and clothing behaviour only explained a small portion of the error variance, which was dominated to over 95% by residual inter-individual variability. Adding historical weather information from the previous day to the Glasgow data gave superior information compared to longer time lags and helped to reduce the residual variance to 89%. Those numbers underpin current limitations in individual thermal comfort prediction, while UTCI performance appears reasonable at the population level.