PNNL

Abstract

Annual burned areas in the United States have increased twofold during the past decades. With more large fires resulting in more emissions of fine particulate matters, an accurate prediction of fire emissions is critical for quantifying the impacts of fires on air quality, human health, and climate. In this study, we present a machine learning (ML) model developed to predict monthly fire emissions over the contiguous US (CONUS) at 0.25-degree resolution with game-theory interpretation. We then compare the predicted PM2.5 fire emissions from the interpretable ML model with the GFED observations and those from process-based models in the Fire Modeling Intercomparison Project (FireMIP). Results show promising performance for the ML model, Community Land Model (CLM), and JULES-INFERNO in reproducing the spatial distributions, seasonality, and interannual variability of fire emissions over CONUS. Regional analysis shows that only the ML model and CLM simulate the realistic interannual variability of fire emissions for most of the subregions (r>0.95 for ML and r=0.14~0.70 for CLM), except for Mediterranean California, where all the models perform poorly (r=0.74 for ML and r