Using a random forest model to predict historical PM2.5 in Alaska

Speaker: Allison Baer | PhD Candidate
University of Maryland, Department of Geographical Sciences

The spatiotemporal coverage of regulatory-grade, ground-based air quality monitoring stations measuring PM2.5 concentrations is low across Alaska. Recently, there has been an increase in the number of low-cost air quality monitoring stations for PM2.5 that expand the spatiotemporal coverage of PM2.5 monitoring in Alaska and globally. This study uses a random forest model to predict PM2.5 concentrations from regulatory-grade data and corrected low-cost air quality monitoring data from the 2019 wildfire season (May through September) in Alaska. Results show that the model predicts a high amount of the variance at over 0.75. These results will inform mapping of PM2.5 continuous concentrations across Alaska.

ACCAP is partnering with NASA’s Arctic-Boreal Vulnerability Experiment (ABoVE) to highlight Alaska research results from this ongoing field campaign. ABoVE is a large-scale study of environmental change and its implications for social-ecological systems. ABoVE links field-based, process-level studies with geospatial data products derived from airborne and satellite sensors, providing a foundation for improving the analysis, and modeling capabilities needed to understand and predict ecosystem responses and societal implications. ABoVE also has occasional webinar series focused on research in Yukon Territory and Northwest Territories.