Global dust simulations in the multiscale modeling framework

Abstract

[1] This study investigates the role of subgrid vertical transport in global simulations of soil-dust aerosols. The evolution and long-range transport of aerosols are strongly affected by vertical transport. In conventional global models, convective and turbulent transport is highly parameterized. This study applies the superparameterization (SP) framework in which a cloud-resolving model (CRM) is embedded in each grid cell of a global model to replace these parametric treatments with explicit simulation of subgrid processes at the cloud-system scale. We apply the implementation of the SP framework in the National Center for Atmospheric Research community atmospheric model (CAM) denoted by SPCAM for dust simulations. We focus on the effects of subgrid transport on dust simulations; thus, the sources and sinks of dust are calculated in the large-scale CAM grids, and the vertical transport of dust is computed in the CRM. We simulate present-day distributions of soil-dust aerosols using CAM and SPCAM operated in chemical transport mode with large-scale meteorological fields prescribed using the same meteorological reanalysis. Therefore, the differences of dust fields between two models caused by explicit versus parameterized treatments of convective transport are examined. Comparison of dust profiles shows that SPCAM predicts less dust in the low to mid troposphere but relatively higher concentration in the upper troposphere. The larger dust mass in upper troposphere in SPCAM may be related to the dust implementation approach in this study, in which the larger resolved updrafts in CRM for deep convection transport more dust aloft but are not accounted by the removal processes in the CRM grid scale. A slightly higher mobilization flux of less than 5% on an average is shown in SPCAM when compared with CAM. Similar patterns of elevated dry deposition are also produced with increases larger than 100% in some areas. For wet deposition, on average CAM is ∼31% higher than SPCAM. The average burden of dust in the simulated year for SPCAM and CAM is 14.8 and 19.7 Tg, respectively. The time-scale analysis shows the predicted dust lifetimes in SPCAM are shorter than CAM by approximately 1 day. The differences between CAM and SPCAM demonstrate that process-oriented treatments of convection can significantly affect the distributions, sources, and sinks of global soil-dust simulations.