Predicting thermal satisfaction as a function of indoor CO[formula omitted] levels: Bayesian modelling of new field data

Abstract

In the pursuit of mitigating the performance gap between model-predicted thermal comfort and measurements of perceived thermal comfort, recent studies have investigated the multi-domain nature of thermal comfort and IEQ. This paper presents an update to a recent work that applied Bayesian logistic regression to examine possible independent correlations between perceived thermal comfort of building occupants and metrics of indoor environmental quality (IEQ) such as CO2 concentrations. The prior work made use of the COPE dataset, a 20-year-old field database of objective and subjective IEQ measurements collected from approximately 800 occupants of open-plan offices in large Canadian and US cities. This study updates that work with a new IEQ field study of 141 workstations carried out at the University of British Columbia across 2019 and 2020. In addition, a new, hierarchical (multi-level) Bayesian logistic regression model is formulated and applied. The new results not only increase the credibility of observing measurements of CO2 concentrations to improve predictions of thermal satisfaction in the original COPE dataset, but that the credibility of this evidence is strengthened on the addition of the new data. At indoor temperatures of 23.5 °C, the probability of an occupant feeling thermally satisfied at measured CO2 levels of 550 ppm was 0.62 [0.54 - 0.69, 95% CrI]. This decreased to 0.28 [0.17-0.42, 95% CrI] at 750 ppm.This study does not suggest these observations are generalizable outside the COPE and UBC datasets, nor does it suggest that CO2 must be affecting occupants physiologically or psychologically to cause these observations. However, it is suggested that CO2 levels may remain overlooked as a contextual, or latent variable for the real-time prediction of perceived thermal comfort in buildings. For instance, prior research has observed a relationship between indoor CO2 levels and perceived air quality as well as a relationship between multiple perceptual factors of IEQ, such as air quality and thermal comfort. If the findings of this paper continue to be found in future studies, measurements of CO2 concentrations may be used to improve the accuracy of thermal comfort prediction models. The core recommendation for future work is therefore to expand measurements in future thermal comfort field studies to include, at least, measurements of indoor CO2 levels. A greater dataset is needed to determine whether the findings of this study are possibly universally applicable, and/or whether knowledge of indoor CO2 concentrations may improve personalized models of thermal comfort that are building- and context-specific.