Improving U.S. fine particulate matter air quality forecasts during wildfires based on chemical data assimilation

Abstract:

Fine particulate matter (PM_2.5) continues to be a major air quality problem in the U.S. especially during wildfires. To improve PM_2.5 forecasts during fire seasons, this study develops a chemical data assimilation system by coupling the Weather Research and Forecasting (WRF) model, the Community Multiscale Air Quality (CMAQ) model and the community Gridpoint Statistical Interpolation (GSI) system to assimilate the Moderate Resolution Imaging Spectroradiometer (MODIS) and Geostationary Operational Environmental Satellite (GOES) aerosol optical depth (AOD) retrievals. The WRF-CMAQ modeling system follows the EPA model configuration and uses the National Emissions Inventory (NEI), with a horizontal grid spacing of 12 x 12 km². The background covariance matrix used in the assimilation is generated by contrasting two simulations using different WRF physics configurations and different anthropogenic and biomass burning emissions. We select the 2018 summer fire season as a case study considering data availability and quality of GOES AOD retrievals. We find that the WRF simulation generally captures the meteorological fields. Before assimilation, the WRF-CMAQ first- and second-day forecasts significantly underestimates AOD (mean bias of -0.1 and correlation of < 0.2) compared with MODIS and GOES retrievals, and surface PM_2.5 concentrations (mean bias of -1.3 and correlation of 0.1) compared with the EPA AirNow in-situ network measurements. With assimilation of MODIS AOD at 15 Z, 18 Z, and 21 Z (UTC) every day, the model forecasts substantially improve AOD (mean bias of -0.05 and correlation of > 0.7) and surface PM_2.5 (mean bias of 0.2 and correlation of 0.2), with similar improvements for the first- and second-day forecasts. The diurnal cycles of surface PM_2.5 forecasts are largely improved particularly during afternoon and night. The assimilation of GOES AOD every 3 hours on each day shows similar but slight smaller improvements in AOD and PM_2.5 forecasts, likely due to the relatively less accurate AOD retrievals of GOES than MODIS. We are also evaluating model forecasts against fire-related field campaign (e.g., WE-CAN) measurements, and quantifying the effect of assimilating MODIS and GOES AOD together to investigate the benefit gained from using the high temporal resolution of geostationary satellite data.