To test this hypothesis, we selected eight dust storm events over the western U.S. For each case we ran two seven-day WRF-Chem simulations: in one simulation we assimilated SM content retrievals from the Soil Moisture Active Passive (SMAP) satellite into WRF-Chem via direct insertion (“Insert SMAP”), while the other simulation had no SM assimilation (“No SMAP”). The SM representation in WRF-Chem was generally improved in many locations in the western U.S. by assimilating SMAP retrievals. Additionally, the Insert SMAP simulations consistently had higher values of aerosol optical depth (AOD) due to increased atmospheric dust loading; in cases where WRF-Chem under-predicted dust loading, the AOD simulations were improved in these regions, but in the cases where WRF-Chem already over-predicted dust loading, the additional dust in Insert SMAP produced slightly worse simulations of AOD. Overall, adjusting the soil moisture representation in WRF-Chem using SMAP data had a relatively small impact on AOD during these dust storms, which points to other larger sources of model error, such as the erodibility input dataset or dust emission parameterization in WRF-Chem, that improved SM alone cannot resolve.
To explore whether assimilating SMAP SM retrievals adds more value if dust emission errors are reduced, we also tested scaling down the dust emissions across the entire domain by various factors in a series of one-year WRF-Chem simulations over the conterminous U.S. We found that while the magnitude of the impacts of assimilating SMAP data were reduced with reduced emissions (as expected), skill scores showed more consistent improvements to WRF-Chem AOD simulations when assimilating SMAP data with more accurate dust emissions.

