Long-term evaluation of a low-cost air sensor network for monitoring indoor and outdoor air quality at the community scale

Abstract

Given the growing interest in community air quality monitoring using low-cost sensors, 30 PurpleAir II sensors (12 outdoor and 18 indoor) were deployed in partnership with community members living adjacent to a major interstate freeway from December 2017- June 2019. Established quality assurance/quality control techniques for data processing were used and sensor data quality was evaluated by calculating data completeness and summarizing PM2.5 measurements. To evaluate outdoor sensor performance, correlation coefficients (r) and coefficients of divergence (CoD) were used to assess temporal and spatial variability of PM2.5 between sensors. PM2.5 concentrations were also compared to traffic levels to assess the sensors’ ability to detect traffic pollution. To evaluate indoor sensors, indoor/outdoor (I/O) ratios during resident-reported activities were calculated and compared, and a linear mixed-effects regression model was developed to quantify the impacts of ambient air quality, microclimatic factors, and indoor human activities on indoor PM2.5. In general, indoor sensors performed more reliably than outdoor sensors (completeness: 73% versus 54%). All outdoor sensors were highly temporally correlated (r > 0.98) and spatially homogeneous (CoD<0.06). The observed I/O ratios were consistent with existing literature, and the mixed-effects model explains >85% of the variation in indoor PM2.5 levels, indicating that indoor sensors detected PM2.5 from various sources. Overall, this study finds that community-maintained sensors can effectively monitor PM2.5, with main data quality concerns resulting from outdoor sensor data incompleteness.