Desk‐Level CO ₂ Sensing for Monitoring Occupancy and Indoor Environmental Quality

Occupants of buildings with poor indoor environmental quality (IEQ) suffer from a lack of comfort, poor work performance, health effects and effects on wellbeing. IEQ can vary within a space, and building occupants both influence and are influenced by this, by generating carbon dioxide (CO ₂ ), heat and moisture. Monitoring both IEQ and occupancy enhances understanding of the indoor conditions and provides information on space utilisation. In this study, a novel method of utilising low‐cost, distributed CO ₂ sensing for monitoring both occupancy and IEQ was designed and validated. Spatiotemporal variation in IEQ was captured via the distributed system, and ground truth occupancy data were obtained via a thermal imaging camera and computer vision–based person detection model. Linking person detection and IEQ data, statistical analysis of the influence of occupant presence on IEQ was conducted and found that while office‐average CO ₂ , temperature and relative humidity were within proposed comfort criteria, desk‐level data demonstrated strong spatial heterogeneity, suggesting the presence of microenvironments. CO ₂ and temperature differed significantly between desks, and occupancy had a significant effect on office CO ₂ and temperature, with relative humidity comparatively uniform and uninfluenced. Rate of change and roughness metrics, derived from the CO ₂ data and defined as the mean absolute first difference and the standard deviation of the first difference of the CO ₂ time signal sampled at 1‐min intervals, respectively, demonstrated moderate correlations with occupancy ( r = 0.38–0.40). Threshold‐based classification at frequently occupied desks achieved good discriminative performance for both dynamic features, with the area under the curve (AUC) = 0.85–0.86, compared with moderate performance across all desks (AUC = 0.75). These results demonstrate that desk‐level CO ₂ sensing represents a viable low‐cost alternative to dedicated occupancy sensors at regularly used workstations, with simple feature‐based classifiers achieving good discriminative performance.