You are here:
Development and Performance Validation of U.S. Wide Correction Equation for PurpleAir Sensor Data
Citation:
Johnson, K., A. Holder, AND A. Clements. Development and Performance Validation of U.S. Wide Correction Equation for PurpleAir Sensor Data. AirNow Sensor Data Pilot Webinar, Research Triangle Park, NC, September 16, 2020.
Impact/Purpose:
Interest and concern about air quality has grown in recent years. Simultaneously, growth in the popularity and use of air sensors across the US has also occurred especially during wildfires. This work evaluates and develops correction methods for a popular low-cost PM2.5 sensor (PurpleAir) across the U.S.. It seeks to inform users of the AIRNow smoke map how the PurpleAir correction was developed and some of the limitations on the map. This presentation will be made to an audience of state, local, tribal, regional, and public audience members on a webinar on Sept 16.
Description:
PurpleAir sensors are widely used by individuals, community groups, and other organizations including state and local air monitoring agencies. The massive, global network of more than 6,000 sensors provide a rich source of crowdsourced PM2.5 data. EPA Office of Research and Development has worked to develop a U.S.-wide correction equation for PurpleAir PM2.5 sensor data based on 50 collocations conducted by air quality monitoring agencies at 39 unique sites in 16 states. This presentation will review the process by which the U.S.-wide correction equation was developed and will present several metrics describing its performance. Further, the presentation will discuss additional work to validate the performance of the correction equation for 1-hr data collected at both ambient and smoke impacted sites and will demonstrate how the correction can be used to generate the NowCast AQI to assist State/Local/Tribal air quality agencies in communicating health impacts to communities. Lastly, we will discuss the operational performance of the correction equation and known/suspected issues and will address some of the commonly expressed QA/QC concerns of this crowdsourced data.