Evaluation methods for low-cost particulate matter sensors

Abstract

Understanding and improving the quality of data generated from low-cost sensors represent a crucial step in using these sensors to fill gaps in air quality measurement and understanding. This paper shows results from a 10-month-long campaign that included side-by-side measurements and comparison between reference instruments approved by the United States Environmental Protection Agency (EPA) and low-cost particulate matter sensors in Bartlesville, Oklahoma. At this rural site in the Midwestern United States the instruments typically encountered only low (under 20 µg m−3) concentrations of particulate matter; however, higher concentrations (50–400 µg m−3) were observed on 3 different days during what were likely agricultural burning events. This study focused on methods for understanding and improving data quality for low-cost particulate matter sensors. The data offered insights on how averaging time, choice of reference instrument, and the observation of higher pollutant concentrations can all impact performance indicators (R2 and root mean square error) for an evaluation. The influence of these factors should be considered when comparing one sensor to another or when determining whether a sensor can produce data that fit a specific need. Though R2 and root mean square error remain the dominant metrics in sensor evaluations, an alternative approach using a prediction interval may offer more consistency between evaluations and a more direct interpretation of sensor data following an evaluation. Ongoing quality assurance for sensor data is needed to ensure that data continue to meet expectations. Observations of trends in linear regression parameters and sensor bias were used to analyze calibration and other quality assurance techniques.