Characterizing infiltration and indoor contribution of PM2.5 based on volunteer-generated monitoring data at large spatial and temporal scales

Abstract:

Personal PM_2.5 exposure may deviate from ambient PM_2.5 levels due to differential infiltration of PM particles and contribution of indoor PM_2.5 sources. It is important to quantify infiltration factors (F_inf) and the contribution of indoor sources. Research in this area has been historically limited in space and time due to the high labor and capital costs of deploying and maintaining collocated indoor/outdoor monitors. Recently, the growth of volunteer-generated PM_2.5 data provides an unprecedented opportunity to characterize F_inf and indoor contribution at large spatial and temporal scales. In this study, 93 volunteer-maintained PurpleAir indoor/outdoor PM_2.5 monitor pairs with ~1.1 million hourly measurements were identified within 500 m from each other during a 20-month period (November 2018 to June 2020) in California. A data-driven method was developed based on local polynomial regression to estimate site-specific F_inf with an assumption that indoor sources are negligible when outdoor levels are sufficiently high (only using the highest 20 outdoor levels for each site). The estimated F_inf had a mean of 0.25 (25^th, 75^th percentiles = [0.14, 0.33]) with a mean standard error of 0.04, the scale of which generally agreed with those reported in previous studies. A seasonal analysis (dry and wet seasons in California) showed a physically meaningful difference in F_inf across seasons where the dry season had a higher F_inf than the wet season by ~0.01 (~4%). The estimated F_inf was significantly higher for commercial buildings than residential houses by ~0.05 (~23%). By applying the F_inf estimates, we found the indoor-generated PM_2.5 contributed an average of 41% (25^th, 75^th percentiles = [26%, 66%]) of the total indoor PM_2.5 levels. For each site, we further identified an outdoor level threshold above which the indoor sources were not substantial for personal exposure assessment (with a mean of ~50 μg/m³). Due to the use of volunteer-generated data, this is the first time the indoor contribution for multiple sites monitored over thousands of hours were characterized at a large spatial scale.