The Pioneer Valley Healthy Air Network: Strategies for Establishing a Community-driven Low-cost Sensor Network

DONG GAO, Jiarong Qi, Anna Woodroof, Mahea Heimuli, David Bloniarz, Alexander Sherman, Kayla Fennell, Samantha Hamilton, Yoni Glogower, Sarita Hudson, Krystal Godri Pollitt, Yale University

     Abstract Number: 308
     Working Group: Aerosol Exposure

Abstract
Air pollution continues to be a global public health threat. With the development of sensor and internet technologies, low-cost air sensors have emerged as an effective tool for hyperlocal air quality monitoring and can help answer community-driven and locally motivated questions. The Pioneer Valley Healthy Air Network was recently created in Western Massachusetts through a partnership between local community groups, city officials, non-for-profit organizations, health care facilities, academic researchers, and residents in the cities of Springfield, Chicopee and Holyoke. The network currently includes 60 Purple Air PA-II-SD and Tetrad AirU sensors that measure real-time PM2.5 mass and ozone concentrations. The goal of the network is to conduct high-quality air quality measurements in environmental justice communities known to have the highest rates of asthma in the country and to educate and inform residence of local air quality issues and actions to mitigate exposure. We developed a formal approach for optimally locating a dense network of low-cost sensors and derived an area-specific calibration model. Combined bottom-up and top-down approaches were used to determine locations for sensor deployment. Residents identified sites of concern/interest and this information was paired with a weighted site selection analysis. This site selection analysis was performed by assigning varying importance levels to different factors such as traffic density, proximity to emission hotspots, and distribution of at-risk vulnerable populations. To ensure high-quality data is shared with the community, pre- and post-deployment calibrations are conducted. For pre-deployment calibration, all sensors were collocated at a regulatory station for a minimum of two weeks. Sensor data was compared to reference methods, and calibration factors were derived for each sensor. Inter-comparison of sensor data showed good correlation between sensors (R2=0.98~0.99). An interactive dashboard was created using Power BI to monitor the sensor health and produce real-time charts, average levels, and alerts. Successes and challenges with establishing this air monitoring network will be shared.