In order to understand the carbon cycle and to project the impact of atmospheric CO2 concentrations on future climate, it is essential to quantify the temporal and spatial distribution of CO2 sources and sinks at the Earth’s surface. Due to the lack of direct flux measurements and of accurate models of the carbon cycle, however, many studies have estimated surface CO2 fluxes using the observations of atmospheric CO2 concentration, known as a “top-down” approach. Most of the top-down approaches require a-priori information of surface CO2 fluxes that have to be pre-calculated based on independent observations or by numerical modeling. Moreover, these methods do not explicitly account for transport errors in the atmospheric CO2 forecasts and the flux estimation when they use transport models in order to link the surface carbon fluxes with the atmospheric CO2 concentrations.

We developed for the first time a simulation Ensemble Kalman Filter system to analyze simultaneously meteorological and CO2 variables, so that the system estimates the background error covariance between the atmospheric CO2 and wind fields. After implementing several new methodologies into the Local Ensemble Transform Kalman Filter (e.g., localization of variables, adaptive inflation, vertical localization of simulated CO2 column OCO-2/GOSAT retrievals, and others) we succeeded in estimating rather accurately the evolving surface carbon fluxes, without any a-priori information. The results from simulation experiments show the potential for estimation of evolving surface fluxes without direct observations or a priori information. We have also successfully tested this methodology for estimating surface fluxes of heat and moisture using simulated AIRS retrievals.