Long-term field evaluation of the Plantower PMS low-cost particulate matter sensors

Abstract

The low-cost and compact size of light-scattering-based particulate matter (PM) sensors provide an opportunity for improved spatiotemporally resolved PM measurements. However, these inexpensive sensors have limitations and need to be characterized under realistic conditions. This study evaluated two Plantower PMS (particulate matter sensor) 1003s and two PMS 5003s outdoors in Salt Lake City, Utah over 320 days (1/2016–2/2016 and 12/2016–10/2017) through multiple seasons and a variety of elevated PM2.5 events including wintertime cold-air pools (CAPs), fireworks, and wildfires. The PMS 1003/5003 sensors generally tracked PM2.5 concentrations compared to co-located reference air monitors (one tapered element oscillating microbalance, TEOM, and one gravimetric federal reference method, FRM). The different PMS sensor models and sets of the same sensor model exhibited some intra-sensor variability. During winter 2017, the two PMS 1003s consistently overestimated PM2.5 by a factor of 1.89 (TEOM PM2.5<40 μg/m3). However, compared to the TEOM, one PMS 5003 overestimated PM2.5 concentrations by a factor of 1.47 while the other roughly agreed with the TEOM. The PMS sensor response also differed by season. In two consecutive winters, the PMS PM2.5 measurements correlated with the hourly TEOM measurements (R2 > 0.87) and 24-h FRM measurements (R2 > 0.88) while in spring (March–June) and wildfire season (June–October) 2017, the correlations were poorer (R2 of 0.18–0.32 and 0.48–0.72, respectively). The PMS 1003s maintained high intra-sensor agreement after one year of deployment during the winter seasons, however, one PMS 1003 sensor exhibited a significant drift beginning in March 2017 and continued to deteriorate through the end of the study. Overall, this study demonstrated good correlations between the PMS sensors and reference monitors in the winter season, seasonal differences in sensor performance, some intra-sensor variability, and drift in one sensor. These types of factors should be considered when using measurements from a network of low-cost PM sensors.