CO2 Flux Inverse Modeling with OCO-2 Retrievals, Ground-Based Observations and High-Resolution Tracer Transport

We developed a CO₂ flux inversion system that is based on the high-resolution Lagrangian-Eulerian coupled tracer transport model and is designed to estimate surface fluxes from atmospheric CO₂ data observed at surface stations and by NASA's OCO-2 satellite. To increase the transport models spatial resolution for global simulation, we use the Lagrangian particle dispersion model (LPDM) FLEXPART to estimate surface flux footprints at 0.1-degree spatial resolution for the OCO-2 observations aggregated to 1 second averages. The LPDM is then coupled to a global atmospheric tracer transport model (NIES-TM) driven by JRA-55 reanalysis. The adjoint of the coupled transport model is used in an iterative optimization procedure to derive surface flux corrections for terrestrial biosphere and ocean. High-resolution prior fluxes were prepared for anthropogenic emissions (ODIAC), biomass burning (GFAS), and the terrestrial biosphere (VISIT model). The terrestrial biosphere flux was constructed using a vegetation mosaic map and simulations of CO₂ fluxes for each vegetation type present in a grid. The prior flux uncertainty is scaled proportionally to monthly mean GPP by the MODIS product for land, and OTTM model for ocean. Use of the high-resolution transport leads to improved representation of the anthropogenic plumes, often observed at continental continuous observation sites. Combining surface (Obspack) and satellite data for use in inversion requires correcting for biases present in satellite observation data. Before including satellite observations in the inversion, the monthly varying latitude-dependent bias is estimated by comparing satellite observations with column abundance simulated with surface fluxes optimized by surface inversion for the period of 2014-2016. Monthly mean model to OCO-2 difference by 10 degree latitudinal band is below 1 ppm between 30° South and 50° North. Inverse modeling reduces mean mismatch between OCO-2 and model to below 0.1 ppm in the latitude bands from 20° to 40° (both North and South) for every month. Bias correction to satellite data improves consistency between the flux estimates based on ground-based and satellite observations.