Validation and Correction of PM_2.5 Mass Concentrations from Consumer-Grade Personal Exposure and Outdoor Fixed...

Consumer-grade (CG) air quality monitors with low-cost sensors that measure different pollutants and environmental factors are used to assess personal exposure and indoor and outdoor air quality in many communities. These devices empower citizens to actively participate in understanding their air quality and reducing their exposure to air pollution by accessing data in real time. As part of the Social Justice and Environmental Quality Denver project (SJEQ-D), we evaluated the performance of Atmotube Pro personal exposure monitors (ATMs) and stationary QuantAQ Modulair-PM outdoor monitors (QAQs) by colocating them with a GRIMM EDM180 (GRIMM) PM_2.5 reference instrument four times for multiple weeks before their use within communities. The significance of this evaluation is that it is one of the first to colocate a large number of CG air quality monitors (50 ATMs and 5 QAQs) over two years during different seasons in the Mountain West region of the U.S. This study is intended as a large-scale, field-based performance evaluation to establish fitness-for-purpose use in community exposure studies, rather than as a sensor development, mechanistic, or regulatory equivalence analysis. The mean correlation coefficient and root mean squared error (RMSE) between the 1-h averaged GRIMM and raw ATM data were ρ = 0.770 and RMSE = 5.12 μg m^-3; the corresponding values for the raw QAQ data were ρ = 0.720 and RMSE = 5.51 μg m^-3. These correlation coefficients indicate that both ATM and QAQ monitors capture a substantial portion of the temporal variability in PM_2.5 measured by the reference instrument. However, both monitors underestimated PM_2.5 concentrations, particularly above 30.0 μg m^-3. In addition, temporal drift was observed in the raw ATM and QAQ data sets, with individual monitors showing measurable changes in response across colocation time periods, indicating differences in long-term performance stability. The mean intramonitor correlation for the ATMs and QAQs was ρ = 0.980 and ρ = 0.780, respectively. When compared to each other, 5-11% of the ATMs were not working. A multilinear regression correction model incorporating relative humidity and dew point was applied to the ATM and QAQ colocation data. This correction increased the mean correlation coefficient by 6.5-8.3% and decreased RMSE by 42-50% relative to the reference instrument, while also reducing the temporal drift observed in the raw data. Overall, our colocation results show that ATMs and QAQs provide useful PM_2.5 measurements for personal exposure and outdoor air monitoring, and that applying appropriate corrections improves agreement with reference instruments. In situations where reference data are unavailable, colocating CG monitors with one another can provide valuable insight into device performance and data reliability.