Performance Comparison of Three Low-cost Particulate Matter Sensors in an Ambient Environment

Field Sampling, Measurement & Sensor Technology
Oral Presentation

Prepared by

Contact Information:; 919-541-1800


Traditionally, air quality monitoring has been restricted to organizations operating expensive and complex Federal Reference Method (FRM) or Federal Equivalent Method (FEM) equipment and other research-grade instrumentation. The emergence of low-cost (<$2,500) particulate matter (PM) sensors has created new opportunities for professional researchers, community groups, and citizen scientists to engage in community level monitoring. The low-cost, compact size, and portability enables the deployment of multiple sensors to assess temporal and spatial trends in micro-environments at high-time resolution. Despite these advantages, known issues occur when deployed across a range of ambient conditions resulting in the need for detailed evaluations to better understand the performance of low-cost PM sensors. The U.S. Environmental Protection Agency (EPA) has focused on the discovery, evaluation, and application of low-cost PM sensors and has published these results on the Air Sensor Toolbox (
Here, we present the performance of three PM sensors, TES 5322 Air Quality Monitor, Plantower PMS 7003, and Aeroqual Portable Particulate Monitor, deployed in triplicate for at least 30 days between October 2017 and July 2018 at the Ambient Air Innovation Research Site (AIRS) on EPA’s campus in Research Triangle Park, NC. Results identified variability between replicate PM sensors indicating the need for individual sensor calibrations. When compared to collocated references analyzers (GRIMM 180 and Teledyne T640) the lowest 1-minute average root mean square error (RMSE) was observed by the Aeroqual sensors (RMSE=2.6-9.8 µg/m3) with similar error observed by the Plantower sensors (RMSE=4.4-6.4 µg/m3) and much greater error by the TES sensors (RMSE=27.8-51.3 µg/m3). Errors were explored as a function of environmental parameters (temperature and relative humidity), averaging times, and reference instrument. Techniques useful in the exclusion of raw sensor data due to the unique impact of environmental conditions on each PM sensor model were examined to develop best practices for future studies.