Identifying Model and Observational Biases in Inverse Modelling of CH4 Observations
Monteil, Guillaume1; Houweling, Sander2; Aben, Ilse3; Hasekamp, Otto3; Röckmann, Thomas4
1IMAU/SRON, NETHERLANDS; 2SRON (Netherland Institute for Space Research) / IMAU, NETHERLANDS; 3SRON (Netherland Institute for Space Research), NETHERLANDS; 4IMAU (Universiteit Utrecht), NETHERLANDS

In the past 10 years, the growing availability of satellite observations of methane (in particular from SCIAMACHY, on-board ENVISAT, and more recently from GOSAT), has led to the development of inverse-modelling techniques for constraining the global methane emissions using these observations. The presence of systematic errors in the observations, and in the global transport model that is used to simulate the measurements, is estimated to contribute significantly to the error in the optimized emissions. It is therefore crucial to identify, and if possible correct, these systematic errors. While it is relatively easy to verify the accuracy of conventional in situ measurement techniques, satellite retrievals are affected by the local atmospheric conditions, and therefore the quality of the retrieval dataset varies in space and time. Comparisons with modelled CH4 concentrations can be used to identify and correct biases, in an iterative process where both the emission estimates and the observations are improved. In such a process, transport model errors could lead to an incorrect estimation of the observational biases. So far, however, they have been neglected in applications involving satellite data since they were assumed to be small in comparisons with errors in the observations. This approach may be valid for inversions using SCIAMACHY retrievals, with an estimated retrieval uncertainy of 2% (Frankenberg et al., 2008). However, retrievals from the more recent GOSAT instrument are of much better quality, such that model errors are becoming an important term in the error budget and should therefore be taken into account. We performed a series of inverse modelling simulations using TM5-4DVAR (Merink et al., 2008; Bergamaschi et al., 2009) GOSAT CH4 retrievals (Schepers et al., 2011) and surface observations from the NOAA/CMDL network (http://www.esrl.noaa.gov/gmd/dv/iadv/). We find that it is difficult for the model to correctly reproduce both surface and satellite observations, in particular in the tropics, and that fitting them in a combined inversion leads to inconsistencies in the optimized emissions. We have investigated the causes of these inconsistencies using various independent observation datasets, and by performing sensitivity tests targeting uncertainties in various model parameters (such as the parametrisation of CH4 chemistry, and the speed of inter-hemispheric transport). We show that, with the increased accuracy of GOSAT retrievals, and the expected accuracy of future instruments, it should be possible, to use the observations to improve the Chemistry-Transport Models.