On biases in atmospheric CO inversions assimilating MOPITT satellite retrievals

Abstract. CO inverse modelling studies have so far reported significant discrepancies between model concentrations optimised with the Measurement of Pollution in the Troposphere (MOPITT) satellite retrievals and surface in-situ measurements. Here, we assess how well a global CTM fits a large variety of independent CO observations (surface and aircraft air sample measurements and ground-based column retrievals) before and after assimilating MOPITTv6 total column (X_CO) retrievals to optimise CO sources/sinks. Consistent negative prior biases to all types of observations in all sensitivity tests suggest an underestimation of current surface emissions in the Northern hemisphere. In contrast, prior simulations fit the surface air sample observations well in the Southern hemisphere but underestimate CO in the free troposphere and on average in the column. Positive biases in MOPITT retrievals are identified in the Northern mid- and high latitudes, highlighting the importance of proper bias-correction of those satellite retrievals. Biases in representing vertical CO profiles are found over the ocean and most significantly in the Southern hemisphere, suggesting errors in the vertical distribution of CO chemical sources/sinks or in the vertical mixing to be improved in future modelling studies. Varying model-data differences are found in the vertical between CTM and MOPITT retrieved vertical profiles after having assimilated MOPITT X_CO; these bias structures indicate that the posterior model differences to in-situ observations would be even larger if the near surface retrievals were assimilated instead of X_CO. In addition, given the higher long-term stability of the MOPITT X_CO retrievals as opposed to divergent temporal bias drifts found in the vertical profile retrievals and a lower sensitivity to model errors in the total column quantity than at a certain altitude, we recommend assimilating the column rather than the profiles or the partial profiles at the current stage.