Extended Abstract (124K)15A.3
CO2 Data Assimilation with the Coupled Atmosphere-Vegetation Model Using Local Ensemble Transform Kalman Filter
Ji-Sun Kang, University of Maryland, College Park, MD; and E. Kalnay, N. Zeng, J. Liu, and I. Fung
To improve the estimation of atmospheric CO2 concentration it has been proposed (Inez Fung/Eugenia Kalnay) to carry out a coupled atmospheric/CO2 data assimilation cycle with an advanced model (NCAR's CAM3.5 finite volume GCM) including CO2 transports and surface fluxes of CO2 from the land and ocean. For the data assimilation, the proposal included the use of the LETKF (Local Ensemble Transform Kalman Filter; Hunt et al., 2007), and the forthcoming OCO observations of total column CO2, and unlike other carbon data assimilations, the standard atmospheric observations and the CO2 observations are assimilated simultaneously.
Since the LETKF/CAM3.5 system is very complex and expensive, we plan to do Observing System Simulation Experiments (OSSEs) with the SPEEDY model (Simplified Parametrizations, privitivE-Equation DYnamics; Molteni, 2003), to which we added a prognostic CO2 variable with just advection and diffusion and surface fluxes. In addition, to create a “nature” run for the simulations, we have coupled the SPEEDY model with a model for land and vegetation known as VEGAS (VEgetation Global Atmosphere and Soil; Zeng et al., 2005). We also included fossil fuel sources of CO2 over land and monthly prescribed ocean fluxes (Takahashi et al., 2002). We plan to use the LETKF to assimilate “observations” with errors obtained from the SPEEDY-VEGAS nature run into the SPEEDY model with just simple fluxes and CO2 transport in order to test the feasibility of the approach, as well as the ability of the LETKF to estimate parameters and surface fluxes. Although the models used for this research are not state-of-art, they are realistic enough to provide guidance for the much larger experiment with real observations.
The results of the coupled atmospheric/CO2 data assimilation with the SPEEDY model, and its ability to reproduce the nature run obtained with the SPEEDY-VEGAS model will be reported at the conference.
Supplementary URL: http://ams2008/jskang
Session 15A, Data Impact Tests: Real and Simulated Observations-I
Thursday, 24 January 2008, 1:30 PM-3:00 PM, 204
Previous paper Next paper
Recorded presentation