[A-CG36] Satellite Earth Environment Observation

[ACG36-05] Global high-resolution inverse model for application to space-based CO₂ observations

*Shamil Maksyutov¹, Tomohiro Oda², Jiye Zeng¹, Rajesh Janardanan¹, Yukio Yoshida¹, Johannes W Kaiser³, David Crisp⁴, Tsuneo Matsunaga¹ (1.National Institute for Environmental Studies, 2.Universities Space Research Association, 3.Deutscher Wetterdienst, 4.Jet Propulsion Laboratory/California Institute of Technology)

Keywords:carbon dioxide, inverse modeling, satellite

We report revision of the global inverse model NTFVAR (NIES-TM-FLEXPART-variational) towards its application to estimating the regional carbon dioxide fluxes using data by GOSAT and OCO-2 satellites. To improve CO₂ simulation, the interhemispheric and vertical mixing rates in the transport model NIES-TM uses ERA-5 reanalysis winds, interpolated to 42 hybrid-pressure levels with a horizontal resolution of 2.5 and 3.75 degrees, and revised 3rd order upwind scheme for advection in the transport model and its adjoint. The validation of the transport model shows an improved match with the observed interhemispheric gradient of SF₆ and vertical gradient of radon. In CO₂ simulations, land biosphere fluxes are provided by the global upscaling product based on tower flux data at the spatial resolution of 0.1 degrees and temporal resolution of 10 days. Fossil emissions are provided by ODIAC, fire emissions by GFAS, and oceanic fluxes by the upscaling product based on surface ocean pCO₂ observations (Landschützer et al. 2015). The respiration and oceanic fluxes are scaled to match the global CO₂ growth rate. The corrections to the prior fluxes are estimated on a bi-weekly time step. With this set of prior fluxes, the model is close to reproducing the observed seasonal cycle at most global monitoring sites even before the inversion. Another problem to address is the data size. Recent space-based observations of the greenhouse gases produce a large number of data leading to a need to input into an inverse model an even larger volume of surface flux sensitivity matrixes prepared by the high-resolution Lagrangian transport model. In the revised model, storing the matrixes out of core leads to a significant reduction of memory demand, at the expense of a need to reread the data from disk on each iteration, but allowing the use of general-purpose computer systems for satellite data inversion. To test the system, inversions of 1.3 sec averaged OCO-2 (version b10) data were performed for a period of 2015-2019. The CO₂ retrieval data by GOSAT (v.02.95) and OCO-2 were compared with forward simulation driven by fluxes optimized with surface data inversion and showed a good match with model data, typically within 0.5 ppm for monthly average data over 5-degree latitudinal bands.

Landschützer, P., et al. The reinvigoration of the Southern Ocean carbon sink, Science, 349, 1221-1224, 2015.