Tight physics and chemistry coupling improves air quality forecasts: a case study of the 2020 North American wildfires with the Rapid Refresh model coupled to Chemistry (RAP-Chem v1.0)

Abstract. The Rapid Refresh model coupled to Chemistry (RAP-Chem) is an experimental forecast model built and operated by the NOAA Global Systems Laboratory (GSL). The RAP-Chem is a full chemistry extension to the operational smoke forecasting model, RAP-Smoke. RAP-Chem’s chemistry includes a reduced complexity, carbon-bond, gas-phase mechanism and an aerosol scheme that are coupled together to produce secondary aerosols and perform heterogeneous chemistry. The model includes chemical lateral boundary conditions, hourly anthropogenic emissions and online emissions from biomass burning, dust, sea salt, and biogenic sources. RAP-Chem also includes the direct aerosol effect on radiation and photolysis and a simplified indirect aerosol effect that couples prognostic aerosols to representative aerosols in the aerosol-aware microphysics scheme. The default ‘offline’ vertical mixing implemented in WRF-Chem is updated to instead mix chemical species alongside other scalars in the model’s planetary boundary layer (PBL) scheme, which tends to improve predictions across species – increasing surface ozone (O₃) and decreasing less reactive tracers (e.g., CO). Sub-grid boundary layer clouds in the PBL scheme are also coupled to a WRF-Chem photolysis scheme for the first time, overall reducing O₃ biases in the eastern US. Here we present the ability of the RAP-Chem model to simulate the extreme North American wildfires in 2020. We demonstrate that the relatively simple gas-phase chemical mechanism alongside our model physics developments and coupling with chemistry can predict maximum daily 8-h average O₃ with exceptional skill (Normalized Mean Bias (NMB) = 6.6 %) averaged over CONUS, improving on simulations that neglect the direct aerosol effect (NMB = 9.5 %). Ozone biases are large in regions with extreme aerosol loading, but the NMB improves more significantly. The model produces the observed extreme aerosol loading with less skill; tending to underpredict extreme wildfire induced concentrations while overpredicting urban particle pollution. As with O₃, daily average PM2.5 is predicted with better skill when the direct aerosol feedback is included (-15.8 % vs. -16.8 %). Overall we find that improving the coupling between chemistry and physics processes in the air quality model improves its predictions of both weather and atmospheric composition.