%0 journal article
%@ 0148-0227
%A Ott, L.,Pawson, S.,Collatz, J.,gregg, W.,Menemenlis, D.,Brix, H.,Rousseaux, C.,Bowman, K.,liu, J.,Eldering, A.,Gunson, M.,Kawa, S.R.

%D 2015
%J Journal of Geophysical Research : Atmospheres
%N 2
%P 734-765 
%R doi:10.1002/2014JD022411
%T Assessing the magnitude of CO2 flux uncertainty in atmospheric CO2 records using products from NASA's Carbon Monitoring Flux Pilot Project
%U https://doi.org/10.1002/2014JD022411
2 
%X NASA's Carbon Monitoring System (CMS) Flux Pilot Project (FPP) was designed to better understand contemporary carbon fluxes by bringing together state-of-the art models with remote sensing datasets. Here we report on simulations using NASA's Goddard Earth Observing System Model, version 5 (GEOS-5) which was used to evaluate the consistency of two different sets of observationally informed land and ocean fluxes with atmospheric CO2 records. Despite the observation inputs, the average difference in annual terrestrial biosphere flux between the two land (NASA Ames CASA and CASA-GFED) models is 1.7 Pg C for 2009-2010. Ocean models (NOBM and ECCO2-Darwin) differ by 35% in their global estimates of carbon flux with particularly strong disagreement in high latitudes. Based upon combinations of terrestrial and ocean fluxes, GEOS-5 reasonably simulated the seasonal cycle observed at northern hemisphere surface sites and by the Greenhouse gases Observing SATellite (GOSAT) while the model struggled to simulate the seasonal cycle at southern hemisphere surface locations. Though GEOS-5 was able to reasonably reproduce the patterns of XCO2 observed by GOSAT, it struggled to reproduce these aspects of AIRS observations. Despite large differences between land and ocean flux estimates, resulting differences in atmospheric mixing ratio were small, typically less than 5 ppm at the surface and 3 ppm in the XCO2 column. A statistical analysis based on the variability of observations shows that flux differences of these magnitudes are difficult to distinguish from inherent measurement variability, regardless of the measurement platform.