Improving annual fine mineral dust representation from the surface to the column in GEOS-Chem 14.4.1

Accurate representation of mineral dust remains a challenge for global air quality or climate models due to inadequate parametrization of the emission scheme, removal mechanisms, and size distribution. While various studies have constrained aspects of dust emission fluxes and/or dust optical depth, annual mean surface dust concentrations still vary by factors of 5-10 among models. In this study, we focus on improving the annual simulation of fine dust in the GEOS-Chem chemical transport model, leveraging recent mechanistic understanding of dust source and removal, and reconciling the size differences between models and ground-based measurements. Specifically, we conduct sensitivity simulations using GEOS-Chem in its high performance configuration (GCHP) version 14.4.1 to investigate the effects of mechanism or parameter updates on annual mean concentrations. The results are evaluated by comparisons versus Deep Blue satellite-based aerosol optical depth (AOD) and AErosol RObotic NETwork (AERONET) ground-based AOD for total column abundance, and versus the Surface Particulate Matter Network (SPARTAN) for novel measurements of surface PM2.5 dust concentrations. Reconciling modelled geometric diameter versus measured aerodynamic diameter is important for consistent comparison. The two-fold overestimation of surface fine dust in the standard model is alleviated by 39 % without degradation of total column abundance by implementing a new physics-based dust emission scheme with better spatial distribution. Further reduction by 20 % of the overestimation of surface PM2.5 dust is achieved through reducing the mass fraction of emitted fine dust based on the brittle fragmentation theory, and explicit tracking of three additional fine mineral dust size bins with updated parametrization for below-cloud scavenging. Overall, these developments reduce the normalized mean difference against surface fine dust measurements from SPARTAN from 94 % to 35 %, while retaining comparable skill of total column abundance against satellite and ground-based AOD.