Assimilation of AIRS CO2 observations with an EnKF in a Carbon-Climate model

Accurate monitoring of carbon cycle, especially estimation of surface carbon flux, is an urgent problem for both of scientific understanding and policy-making purposes. Lack of CO2 observations is a major obstacle to reach these goals, since there are only about 100 surface CO2 stations worldwide. The recent effort of retrieving CO2 observations from infrared satellite sensors has greatly improved the CO2 observation coverage. Retrieving CO2 from AIRS is one of them (Chahine et al., 2005; 2008). AIRS CO2 observations have the advantage of dense coverage (about 1000 CO2 observations within 6 hours), but its weighting functions peaks around middle troposphere, which implies that the signal of surface carbon flux is weaker in AIRS CO2 observations than in surface CO2 observations.

In this study, we have assimilated the AIRS column weighted CO2 retrievals along with all the meteorological observations assimilated in NCAR/DOE Reanalysis 2 in the NCAR carbon-climate model. We discussed a proper way to obtain CO2 vertical profiles based on column-integrated data. An initial test has been carried out to estimate surface carbon flux based on both AIRS CO2 observations and conventional surface CO2 observations within the LETKF data assimilation framework. Surface carbon flux is estimated by appending to the CO2 dynamical vector during data assimilation. It is updated by using the covariance between CO2 concentration and surface carbon flux.

The results show that the accuracy of CO2 vertical profiles have been improved by assimilation of the column weighted AIRS CO2 observations, when compared to independent aircraft CO2 observations. The preliminary surface carbon flux estimation shows that AIRS CO2 observations have more impact on surface carbon flux over tropical region where AIRS CO2 observations is more related to surface carbon flux due to strong vertical mixing.