Variational Data Assimilation of Satellite Remote Sensing Observations for Improving Methane Simulations in Chemical Transport Model

Atmospheric concentrations of methane (CH4), the second most important anthropogenic greenhouse gas, have been rapidly rising since 1850, however, the rate of increase has varied in recent decades. In order to attribute these trends, significant effort has been put into characterizing CH 4 surface emissions using inverse modelling ("top-down") approaches, which rely on the ability of chemical transport models (CTMs) to simulate atmospheric CH4 fields. However, systematic errors in models can significantly reduce the quality of the CH4 simulation and result in biased CH4 emission estimates. Until now, errors in models have been poorly characterized. The objective of this thesis was to characterize and investigate the origin of the CH4 model errors in the GEOS-Chem CTM and quantify their impact on inferred CH4 emission estimates. The weak constraint four-dimensional variational data assimilation scheme in GEOS-Chem, together with CH4 data from the Greenhouse gases Observing SATellite (GOSAT), were used to characterize model errors in GEOS-Chem at the horizontal resolutions of 4x5 and 2x2.5. Large biases in CH4 were found in the stratosphere and in vertical transport in the troposphere at mid-latitudes. The identified errors were significantly larger at 4x5 than at 2x2.5. It was determined that a major cause of the biases at 4x5 is excessive mixing due to increased numerical diffusion manifested in enhanced stratosphere-troposphere exchange, and stronger quasi-isentropic mixing through the edges of the "tropical pipe" and the polar vortex in the stratosphere. Coarsening of the model grid also weakened vertical transport in the troposphere due to the loss of advective air mass fluxes and sub-grid tracer eddy mass fluxes. A key outcome of this work is the recommendation that the 4x5 version of GEOS-Chem should not be used for inverse modeling of CH4 emissions. The thesis also investigated the sensitivity of North American CH4 emission estimates in the nested version of GEOS-Chem (at the 0.5x0.67) to biases in boundary conditions from the coarse global resolution model. It was shown that biases not fully mitigated in the global CH4 simulation could result in biases as large as 30-35% in monthly mean surface emission estimates on local to regional scales.