An Ensemble Nonlinear Least Squares Four-Dimensional Variational Data Assimilation System to Estimate surface CO2 Flux from space-borne CO2 observations

Abstract: In this paper, a satellite carbon data assimilation system (Tan-Tracker) based on nonlinear least squares four-dimensional variational data assimilation algorithm (NLS-4DVar) and atmospheric chemical transport model (GEOS-Chem) is constructed, which adopts an efficient localization scheme and a dual-pass optimization strategy. The atmospheric CO₂ column concentrations with high spatial and temporal resolution, observed by the CO₂ observing satellite (OCO-2), are used to accurately obtain surface carbon fluxes (CFs). A group of observational system simulation experiments (OSSEs) are designed and performed to verify the robustness of the system. The following conclusion can be drawn: The optimized CO₂ concentrations (directly forced by optimized fluxes) are closer to the observations meanwhile the result of the optimized fluxes have closer total amount, spatial and temporal distribution to the real fluxes, meaning that Assimilating satellite carbon data can be used effectively to optimize fluxes. Subsequently, a real data assimilation experiment on the CO₂ column concentration of OCO-2 observations shows that: the CO₂ concentration forced by the optimized fluxes are closer to the independent satellite observations and surface observations (TCCON), indicating that the optimized fluxes are more reasonable. The experimental results show that Tan-Tracker can provide a more accurate surface flux estimation, which provides a better tool for the study of carbon cycle and provides a useful strategy for top-down methods.

Key words: NLS-4DVar; surface carbon flux; atmospheric CO₂ concentration; OCO-2; GEOS-Chem;