From Data to Decision: Evaluating Air Sensors with Regulatory Monitors in the Southeastern U.S.

Air sensors are increasingly deployed to supplement data from regulatory monitoring networks, yet concerns about their performance persist. Most particulate matter (PM) sensors rely on the light scattering principle, which can struggle with accuracy. To improve the reliability of sensor data, collocating these sensors with regulatory-grade monitors for performance evaluation and validation is a critical step. In collaboration with the Florida Department of Environmental Protection (FDEP), NCASI is conducting a collocation study to evaluate several commercially available PM2.5 sensors. At a site in north-central Florida, five sensors (AirGradient, Airlink, AirVisual, BlueSky, and PurpleAir), each with two duplicates, were collocated with a Federal Equivalent Method (FEM) monitor (Teledyne T640). Sensor performance varied across multiple metrics. In terms of Mean Absolute Error (MAE) for hourly averaged values, AirGradient sensors had the highest MAE at 3.58 µg/m³, while AirVisual showed the lowest MAE at 2.10 µg/m³. The PurpleAir sensors demonstrated a significant improvement after applying the EPA's correction approach, with an MAE of 1.64 µg/m³, reduced from 3.51 µg/m³. However, sensor performance was influenced by environmental conditions. For all sensors, the bias (defined as the difference between the average sensor and FEM PM2.5 concentrations) increased significantly, with large positive bias values observed when relative humidity (RH) exceeded 75%. This trend was consistent across all sensors, indicating that elevated RH caused an overestimation of PM2.5 concentrations. This observation is particularly relevant for regions with higher humidity, which contrasts with other collocation studies conducted in drier climates. Therefore, users should carefully interpret data from low-cost sensors to account for potential inaccuracies under varying environmental conditions.