An 8-year, high-resolution reanalysis of atmospheric carbon dioxide mixing ratios based on OCO-2 and GOSAT-ACOS retrievals

This talk presents an overview of results from the GEOS-Carb reanalysis of retrievals of average-column carbon dioxide (XCO2) from the Orbiting Carbon Observatory 2 (OCO-2) and Greenhouse Gases Observing Satellite (GOSAT) satellite missions. The reanalysis is a Level 3 (L3) product: a collection of 3D fields of carbon dioxide (CO2) mixing ratios every 6 hours beginning in April 2009 going until the present on a grid with a 0.5 degree horizontal resolution and 72 vertical levels from the surface to 0.01 hPa. Using an assimilation methodology based on the Goddard Earth Observing System (GEOS) atmospheric data assimilation system (ADAS), the L3 fields are weighted averages of the two satellite retrievals and predictions from the GEOS general circulation model driven by assimilated meteorology from the Modern-Era Retrospective analysis for Research and Applications, Version 2 (MERRA-2). In places and times where there are a dense number of soundings, the observations dominate the predicted mixing ratios, while the model is used to fill in locations with fewer soundings, e.g., high latitudes and the Amazon. Blending the satellite observations with model predictions has at least two notable benefits. First, it provides a bridge for evaluating the satellite retrievals and their uncertainties against a heterogeneous collection of observations including those from surface sites, towers, aircraft, and soundings from the Total Carbon Column Observing Network (TCCON). Extensive evaluations of the L3 reanalysis clearly demonstrate both the strength and the deficiency of the satellite retrievals. Second, it is possible to estimate variables from the reanalysis without introducing bias due to spatiotemporal variability in sounding coverage. For example, the assimilated product provides robust estimates of the monthly CO2 global growth rate. These monthly growth rate estimates show significant differences from estimates based on in situ observations, which have sparse coverage, and those based on model surface fluxes, which imperfectly represent key processes. This presentation discusses the implications of this finding as well as ongoing strategies to extract more information from the satellite retrievals in future L3 reanalyses.