Laboratory evaluation of low-cost PurpleAir PM monitors and in-field correction using co-located portable filter samplers

Abstract Low-cost aerosol monitors can provide more spatially- and temporally-resolved data on ambient fine particulate matter (PM2.5) concentrations than are available from regulatory monitoring networks; however, concentrations reported by low-cost monitors are sometimes inaccurate. We investigated laboratory- and field-based approaches for calibrating low-cost PurpleAir monitors. First, we investigated the linearity of the PurpleAir response to NISTUrban PM and derived a laboratory-based gravimetric correction factor. Then, we co-locatedPurpleAirs with portable filter samplers at 15 outdoor sites spanning 3 × 3-km in Fort Collins, CO, USA. We evaluated whether PM2.5 correction factors calculated using ambient relative humidity data improved the accuracy of PurpleAir monitors (relative to reference filter samplers operated at 16.7 L min−1). We also (1) evaluated gravimetric correction factors derived from periodic co-locationswith portable filter samplers and (2) compared PM2.5 concentrations measured using portable and reference filter samplers. Both before and after field deployment, a linear model relating NISTUrban PM concentrations reported by a tapered element oscillating microbalanceand PurpleAir monitors (“PM2.5 ATM”) had R2 = 99%; however, an F-test identified a significant lack of fit between the model and the data. The laboratory-based correction did not translate to the field. Over a 35-day period, time-averaged ambient PM2.5 concentrations and RHs measured during 72- and 48-h filter samples ranged from 1.5 to 8.3 μg m−3 and 47%–77%, respectively. Corrections calculated using ambient RH data increased the fraction of time-averaged PurpleAir PM2.5 concentrations that were within 20% of the reference concentration from 24% (for uncorrected measurements) to 66%. Corrections derived from monthly, weekly, and concurrent in-field co-locationswith portable filter samplers increased the fraction of time-averaged PurpleAir PM2.5 concentrations that were within 20% of the reference to 48%, 54%, and 72%. PM2.5 concentrations measured using portable filter samplers were within 20% of the reference for 69% of samples.