Spatial Resolution and Sensor Accuracy in Networks for Routine IAQ Monitoring: Are More Sensors Better?

Considerable focus on intelligent building control has expanded interest in incorporating more sensors into buildings, including ones for measuring indoor air quality. In commercial buildings with automation systems, adding more indoor air sensing in more locations could enable more effective ventilation control through dynamic and spatial fine-tuning, helping to save energy, improve air quality, or both. But how necessary and how useful is adding more than one air quality sensor per mechanical zone in typical commercial buildings with air-mixing mechanical systems, given the accuracy of available sensors? We assessed two approaches for locating from one up to 10 sensors for measuring carbon dioxide (CO₂) and the sum of volatile organic compounds (ΣVOC) in typical offices. One placement approach was intended to determine the best possible performance, using optimal clustering of locations by their concentration profiles; the second simply sited each additional sensor in the mechanical system’s return duct and averaged readings. Each approach was assessed at three levels of sensor accuracy in four different office cases that combined high or low air recirculation rates and open or private layouts. The analysis was conducted within a Monte Carlo framework in which dozens of other parameters—including spatially resolved emission rates, air mixing parameters, schedule profiles, and envelope and wind characteristics—were varied for each case. The results indicated that, in an office environment with at least some mechanical air recirculation, concentrations are reasonably homogeneous within a mechanical zone. The spatial coefficient of variation (the spatial standard deviation normalized by the mean concentration) was almost always less than 20%, even when spatial emissions variability was highly elevated. The spatial variation that did exist was generally overwhelmed by sensor error when typical building grade sensors were modeled. The implications are that: 1) for typical office spaces served by a common air-based system with at least a small amount of mechanical air recirculation, a well-mixed model is a reasonable representation of the bulk air concentration (though not necessarily personal exposure); 2) attempting to capture spatial distribution is less important than improving sensor accuracy, and without better accuracy can actually increase network inaccuracy; 3) if using and maintaining better sensors is not possible, placing more typical building grade sensors in the return duct to provide redundancy can help reduce overall error.