A new Fused L3 OCO-2/GOSAT product performs equally well in Data Assimilation as the pooled L2 products

The spatial and temporal variations in the concentrations of atmospheric trace gases like carbon dioxide (CO₂) are currently measured from aircraft, in-situ (surface) and satellite-based sensors. These measurements and retrievals are generally combined with atmospheric transport models to estimate CO₂ fluxes within Data Assimilation (DA) systems. Satellite retrievals of the atmospheric CO₂ concentrations help in reducing: (1) the sparsity of measurements, and (2) the uncertainty in the CO₂ fluxes estimated from DA. Due to their widespread spatial coverage, they provide new quantitative insights into the global carbon cycle. However, records of satellite retrievals of CO₂ only cover the previous 13 years, and any given CO₂ satellite at present has operational products for either the full column average CO₂ (e.g. XCO₂) or the free tropospheric (FT) CO₂ but not both. Thus, combining data from satellites covering overlapping periods is a natural step that addresses the first problem (the sparsity of measurements); further, combining data from different sensors could potentially solve the second. While creating such a fused satellite record of CO₂ has benefits, it is challenging due to different sensor geometries, revisit times and measurement accuracies of instruments. In light of these limitations, we have developed gap-filled Earth System Data Records (ESDR) of CO₂ concentrations by integrating observations from the Atmospheric CO₂ Observations from Space (ACOS) retrievals from the Greenhouse Gases Observing Satellite (GOSAT) and Orbiting Carbon Observatory-2 (OCO-2). As part of this product, we provide a compressed form of the full error covariance (not just standard errors) that can be used to inflate uncertainties in DA systems. Rigorous comparative evaluation of these products with Level-2 observations within DA systems show that in they provide similar estimates of fluxes as those provided by Level-2 products. Discrepancies between them, if any, can be harmonized by adjusting for information content in Level-2 and fused product.