HR: 11:35h
AN: A51H-06    [PDF]
TI: A 4-D Variational Data Assimilation Approach for Estimating Time-Varying Sources and Sinks of CO$_{2}$
AU: * Baker, D F
EM: dfb@ucar.edu
AF: National Center for Atmospheric Research, 5800 Table Mesa Dr., Boulder, CO 80305
AU: Stephens, B
EM: stephens@ucar.edu
AF: National Center for Atmospheric Research, 5800 Table Mesa Dr., Boulder, CO 80305
AU: Doney, S
EM: sdoney@whoi.edu
AF: Woods Hole Oceanographic Institute, 360 Woods Hole Road, Woods Hole, MA 02543
AU: Schimel, D
EM: schimel@ucar.edu
AF: National Center for Atmospheric Research, 5800 Table Mesa Dr., Boulder, CO 80305
AB: Inversions of the historical atmospheric CO2 measurement record provide robust estimates of CO2 sources and sinks at generally coarse scales: at continental and ocean basin scales in space, and at seasonal to monthly time scales. To help choose between the different possible physical mechanisms driving the sources and sinks, finer-scale results are desired ? e.g., daily-average fluxes at the resolution of 100s of km. Global inversions at these scales are currently computationally feasible only using adjoint methods. Using the adjoint of a global 3-D atmospheric transport model (with 2x2.5 degree resolution), we present daily CO2 flux estimates at the model resolution using a variational data assimilation method (4-D Var). With two different sets of modeled surface CO2 fluxes (LPJ land/NCAR ocean vs. CASA land/Takahashi ocean), we perform observing system simulation experiments (OSSEs), both in a perfect model context and using two different transport models, to assess the total errors expected at these fine scales when using CO2 observations from a greatly expanded version of our current network (no satellite-based measurements, but with hourly-resolved measurements from hundreds of in situ analyzers). In the perfect model case, the OSSE total errors show corrections at coarser scales, but persistent errors at finer spatial scales. Apparently, the inversion can improve the fluxes only at scales dictated by the density of the observations, rather than the underlying resolution of the solution method. [This result implies, for example, that a network equivalent to 100 fixed sites across the continental U.S. would be limited to a resolution of about 400 km, at best.] The OSSE errors for the inversion using different transport models suggests that even this resolution limit is quite optimistic. A dense observing network, coupled to a transport model that accurately represents synoptic-scale (perhaps even meso-scale) mixing, is required; the correlations assumed in the inverse procedure must be tuned well, also. The adjoint transport model is used, independent of the inverse procedure, to show that the surface fluxes for the mid- to high-latitudes are constrained much better than the tropical fluxes by the assumed network. This suggests that we will have to be satisfied with much coarser resolution in the tropics, unless a highly dense observing network is established there.
DE: 0312 Air/sea constituent fluxes (3339, 4504)
DE: 0315 Biosphere/atmosphere interactions
DE: 0322 Constituent sources and sinks
SC: Atmospheric Sciences [A]
MN: 2003 Fall Meeting