A41G-3138:
Development of the inverse model for estimation of the surface CO2 fluxes at grid scale and high resolution with GOSAT data
Thursday, 18 December 2014
Shamil S Maksyutov1, Akihiko Ito1, Tomohiro Oda2,3, Johannes W Kaiser4, Dmitry A Belikov1,5, Rajesh Janardanan Achari1, Alexey Yaremchuk6, Ruslan Zhuravlev7,8, Alexander Ganshin7,8 and Vinu Valsala9, (1)National Institute for Environmental Studies, Tsukuba, Japan, (2)Colorado State University, Boulder, CO, United States, (3)Earth System Research Laboratory, Boulder, CO, United States, (4)Max Planck Institute for Chemistry, Mainz, Germany, (5)NIPR National Institute of Polar Research, Tokyo, Japan, (6)N. N. Andreyev Acoustics Institute, Moscow, Russia, (7)Central Aerological Observatory, Moscow, Russia, (8)Tomsk State University, Tomsk, Russia, (9)IITM, Pune, India
Abstract:
We develop an iterative inversion method to estimate surface CO2 fluxes at resolutions up to 0.1 degree using atmospheric CO2 data collected by the global in-situ network and GOSAT. The atmospheric transport model and its adjoint are made by coupling the Eulerian grid model (NIES-TM) to Lagrangian particle dispersion model FLEXPART. The inverse model calculates corrections to the prior fluxes at a weekly time step and spatial resolution of the FLEXPART model (1 or 0.1 degrees). The terrestrial biosphere fluxes are simulated with VISIT model at hourly time step using CFSR reanalysis. Ocean fluxes are calculated using a 4D-Var assimilation system of the surface pCO2 observations. In the high resolution mode, prior fluxes of fossil emissions (ODIAC) and biomass burning (GFASv1.1) are given at a model resolution, while ocean and terrestrial ecosystem fluxes are interpolated from a coarser resolution. The surface flux footprints for in-situ and GOSAT observations are simulated with Flexpart. Precalculated flux response functions are then used in forward and adjoint runs of the coupled transport model. We apply the truncated singular value decomposition (SVD) of the scaled tracer transport operator A=R-1/2HB1/2, where H - tracer transport operator, R and B - uncertainty matrices for observations and fluxes, respectively. The square root of covariance matrix B is constructed by directional splitting in latitude, longitude and time, with exponential decay scales of 500 km on land, 1000 km over oceans and 2 weeks in time. Once right and left singular vectors of ATA are obtained, the prior and posterior flux uncertainties are evaluated. Numerical experiments of inverting the surface CO2 fluxes showed that the high resolution (Lagrangian) part of the flux responses dominates the solution so that patterns from the coarser resolution NIES TM (10x10 degree) are not visible in flux singular vectors and the optimized flux. The reconstruction of the fluxes at highest resolution of 0.1 degree has been tested to complete successfully for shorter time window of 7 months. We demonstrated that application of a coupled tracer transport model in adjoint-based assimilation provides an efficient way to increase resolution of the inverted flux, which can be extended to match current resolution of the global operational forecast models.
