EVALUATION OF FINE PARTICULATE MATTER AIR SENSOR DATA AND INTERPOLATION MODELS IN PHOENIX, ARIZONA

Abstract

Chronic and acute exposure to fine particulate matter in ambient air has increasingly been linked through epidemiological studies to negative health outcomes. The rise in popularity of low-cost air sensors presents an opportunity to potentially improve interpolation predictions for fine particulate matter by combining air quality monitor data with air sensor data. However, the use of air sensor data presents data quality concerns when compared to air monitor data which must conform to federal methods. Recently, U.S. EPA researchers have developed a correction equation to reduce bias in PurpleAir sensor measurements. This study aims to evaluate the potential for using public PurpleAir sensor data and interpolation models to generate fine particulate matter concentration prediction surfaces for Phoenix, Arizona. This study uses geostatistical analyses to evaluate interpolation model performance using three datasets: 1. air monitors, 2. air monitors and uncorrected air sensors, and 3. air monitors and corrected air sensors. This study also evaluates interpolation performance by ranking cross-validation statistic scores for model bias, prediction accuracy, precision, worst-case error, and standard error accuracy for each data scenario. Models for data processing, cross-validation, and interpolation automation will be made available for the purpose of reproducing these analyses in other locations. Based on fine particulate matter concentration data collected from the study area between December 17 - December 31, 2022, the ordinary kriging optimized interpolation technique achieved the highest average rank with the “air monitors” and “air monitors and corrected air sensors” data scenarios achieving an equal weighted average rank.