Using GEOS-5 Atmospheric Transport Simulations to Test the Consistency of Land- and Ocean- Carbon Fluxes with CO2 Observations

Many components of the carbon cycle are constrained by a variety of remote sensing measurements. Observations of land surface parameters constrain estimates of carbon flux from terrestrial biosphere models while estimates of oceanic carbon fluxes are informed by satellite observations of ocean color and ocean properties. Atmospheric CO2 concentrations, which are governed by the balance of terrestrial, oceanic, and anthropogenic fluxes, are observed from space by an expanding suite of instruments (AIRS, TES, and GOSAT) in addition to being monitored by an extensive global network of surface stations. Additionally, atmospheric transport patterns simulated by NASA's GEOS-5 data analysis system are strongly influenced by observations of atmospheric state variables. NASA's Carbon Monitoring System Flux Pilot Project was created to quantify the constraints placed on carbon flux estimates by the current observing system and to assess what additional observational needs are required for future monitoring and attribution efforts. To this end, we have conducted an ensemble of GEOS-5 modeling studies using different combinations of two sets of land (NASA-CASA, CASA-GFED) and two sets of ocean (NOBM, ECCO2/Darwin) fluxes. Results from this ensemble of simulations are sampled at locations consistent with NOAA GMD and TCCON surface networks as well as locations of AIRS, TES, and GOSAT overpasses to quantify how surface flux uncertainty may be observed by different observing systems. Additionally, an ensemble of GEOS-5 simulations with alterations to subgrid-scale transport parameterizations is analyzed to compare model transport uncertainty with flux uncertainty. Our results indicate that uncertainty in both land and ocean flux estimates can introduce a large degree of variability into atmospheric CO2 distributions and that the magnitude of these differences is observable by existing satellite and in situ platforms. In contrast, transport uncertainty introduced by subgrid-scale parameterizations has a much smaller impact on CO2 mixing ratios because the assimilated observations of state variables place a strong constraint on atmospheric transport patterns.