Estimation of Global Surface Carbon Fluxes Using Advanced Data Assimilation

Kang et al., (JGR, 2011, 2012) developed an advanced data assimilation methodology able to recover (in a simulation mode) the surface carbon fluxes from the assimilation of realistic atmospheric CO2 measurements. After a spin-up of 3 months, the carbon fluxes were estimated accurately at a grid-size resolution and captured quite well the natural annual cycle and the anthropogenic sources without using any prior information. In our current research, led together with co-authors, we are assimilating both simulated and real observations into the GEOS-CHEM global model coupled with the VEGAS vegetation model, using strongly coupled LETKF Data Assimilation, with the ultimate goal of estimating global real carbon fluxes. The coupled GEOS-CHEM/VEGAS model reproduces very realistically the observed seasonal cycle of the CO2 flask data for April 2012-July 2014 at many Global View stations, after a spin-up of about 6 months. We confirmed that the LETKF Data Assimilation could also be used to estimate unmeasured parameters in VEGAS model, as long as their time scales are not longer than a few years. We are performing OSSE and OSE data assimilation experiments with this system, assimilating a) Global View observations, b) OCO-2 observations, and c) combining both set of observations in order to determine optimal data assimilation parameters for the estimation of surface carbon fluxes, and will present an overview of our results.