Extended Data Fig. 1: Schematic diagram of the TROPOMI XCO inversion procedure.
(a) Ensembles of inversions are performed based on three different flux inventories. (b) To quantify the sensitivity to systematic error sources, four inversions are performed that differ in observational error constraints, prior error constraints, and temporal optimization frequency. (c) Bayesian posterior error estimates are estimate for 2023 by following the Monte Carlo approach for 4D-Var of Chevalier et al.71. (d) The posterior best estimates are taken as the average maximum a posteriori estimate across inversion configurations while the uncertainty is taken to be the sum-of-squares of the error components estimated in (b) and (c).
