The GEOS-CHEM 4D-Var data assimilation system provides a powerful framework to estimate the global state of several chemical species as well as their source strengths as long as transport errors don't introduce systematic errors in the prediction of the species under investigation. Modelling and observational studies have shown that the depiction of atmospheric carbon monoxide is negatively impacted by errors introduced by the convective parameterization as well as from large-scale convective mass fluxes. The estimation of CO sources using the GEOS-CHEM data assimilation system thus suffers from the problem outlined above.
In this talk, I will present an extended version of the 4D-Var formalism which allows for the joint estimation of sources as well as model error terms. I will further discuss preliminary results of an application of this formalism to the problem of estimating CO source strengths using GEOS-CHEM.