Improving a process-based biogeochemistry model using an atmospheric transport chemistry model and in-situ and remotely-sensed terrestrial and atmospheric data.

摘要

To improve the quantification of methane emissions from natural wetlands, an integrated framework was developed. It includes an improved process-based model and a 4-D VAR inversion algorithm implemented with the adjoint GEOS-Chem model.;The new process-based model was tested comprehensively at two temperate peatlands in Michigan, USA, with historical measurement data for a 3-year period. The results showed that (1) the ebullition can be modeled as a mechanical process that is controlled by the atmospheric pressure, water table level and total pressure of the various dissolved gases; (2) the four-substance structure is the appropriate choice to correctly model the transport pathways to release wetland methane into the atmosphere; and (3) the estimated methane emission is very sensitive to the hydrological forcing used to drive the model and can be seriously biased due to an improper modeling of water tables. The wetland emission from wetland in 45°N north was thence quantified with the one-substance model, and was fed into the GEOS-Chem model for an assessment of forward transport and flux inversion for the year 2004.;The forward simulations were evaluated using observations including low frequency flask measurement, high frequency in-situ time series, atmospheric methane profiles from aircraft campaigns and retrievals from space instruments SCIMACHY and AIRS. It was found that simulations with the anthropogenic emission datasets compiled based on EDGAR4 dataset resulted in better agreement with the high precision surface measurements. The use of GFED2 and GFED3 datasets for methane fluxes from biomass burning lead to small differences in the forward simulations. Overall, the seasonal error characteristics of the model-AIRS misfit, and model-SCIAMACHY misfit, are around 1%-2% in most of the grid cells at the grid resolution for year 2004.;4D-Var inversions were conducted using the adjoint GEOS-Chem to obtain an improved estimation of the surface fluxes. The inversion experiments for year 2004 lead to posterior estimates of net surface methane emissions ranging from 496 Tg CH4 yr-1 to 585 Tg CH4 yr -1. The posterior estimates for the methane emission from vegetation, including wetland, rice paddy and wet tundra, range from 277 Tg CH4 yr-1 to 305 Tg CH4 yr-1.;This study was not possible to obtain a credible estimation for methane emissions from the natural wetland due to the significant uncertainties in the process-based modeling and the flux partition in the atmospheric inverse modeling. Further studies in both of these two regards are needed to improve our understanding of the global wetland methane dynamics.