Transport of Mineral Dust Into the Arctic in Two Reanalysis Datasets of Atmospheric Composition

Two three-dimensional reanalysis datasets of atmospheric composition, the Copernicus Atmosphere Monitoring Service reanalysis (CAMSRA) and the Modern-Era Retrospective Analysis for Research and Applications, version 2 (MERRA-2), are analyzed for the years 2003-2018 with respect to dust transport into the Arctic. The reanalyses agree on that the largest mass transport of dust into the Arctic occurs across western Russia during spring and early summer, but substantial transport events occasionally also occur across other geographical areas during all seasons. In many aspects, however, the reanalyses show considerable differences: the mass transport in MERRA-2 is substantially larger, more spread out, and occurs at higher altitudes than in CAMSRA, while the transport in CAMSRA is to a higher degree focused to well-defined events in space and time; the integrated mass transport of the 10 most intense 36-hour dust events in CAMSRA constitutes 6 % of the total integrated dust transport 2003-2018, whereas the corresponding value for MERRA-2 is only 1 %. Furthermore, we compare the reanalyses with surface measurements of dust in the Arctic and dust extinction retrievals from the Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) satellite data. This comparison indicates that CAMSRA underestimates the dust transport into the Arctic and that MERRA-2 likely overestimates it. The discrepancy between CAMSRA and MERRA-2 can partially be explained by the assimilation process where too little dust is assimilated in CAMSRA while in MERRA-2, the assimilation process increases the dust concentration in remote areas. Despite the profound differences between the reanalyses regarding dust transport into the Arctic, this study still brings new insights into the spatio-temporal distribution of the transport. We estimate the annual dust transport into the Arctic to be within the range 1.5-31 Tg, where the comparison with observations indicates that the lower end of the interval is less likely.