Regional CO2 Inversion Study for Europe and West Siberia

Recently, the number of stations to continuously measure atmospheric CO2 concentration increased. These stations bring us important information about regional and short-term variations of CO2 fluxes through atmospheric transport. To take advantage of this new information, we developed a Bayesian synthesis inverse scheme to infer daily CO2 flux at a grid cell resolution over a pre-defined region from measured daily mean CO2 concentration. This inverse method was applied to AEROCARB European network measurements for the year 2001. To calculate atmospheric transport, we used the global circulation model LMDZ with variable grid size. We used enhanced spatial grid resolution over Europe up to 50x50 km2, therefore improving atmospheric transport modeling over this region. For each grid cell the flux contribution to simulated CO2 concentration at observation sites was calculated by LMDZ running in adjoint mode. Whereas still under discussion, we believe the flux errors at these small spatial and temporal scales don't vary independently from one another. Then we developed and tested three different spatial correlation schemes on flux errors in the inversion. In a first attempt, we used synthetic data to analyze the potential of the European network. These data were simulated by forward atmospheric transport of daily CO2 fluxes from the biogeochemical model ORCHIDEE for the year 2001. This allowed us to determine over which area the fluxes are retrieved most accurately in case of a perfect atmospheric transport. We also estimated at which minimal scales, in space and in time, fluxes should be aggregated to compare favorably with true fluxes from ORCHIDEE model. With actual observations, flux estimates appear less realistic. Temporal and spatial aggregations, as well as proper data selection, are shown to improve the fluxes considerably. Moreover, by 2005, NIES has equipped five towers in West Siberia to continuously measure atmospheric CO2 concentrations, in addition to previously established airborne observations. These towers are located relatively close from one another, covering 786000 km2, and allow us to study CO2 fluxes at regional scale by inverse modeling.. We developed regional inverse model setup with LMDZ zoomed over West Siberia for 2005, and tested the forward simulations of CO2 transport with ORCHIDEE model fluxes against observations.