Colorado State University General Circulation Model

Latitudinal Gradient of Atmospheric CO2 Due to Seasonal Exchange with Land Biota.

A. Scott Denning*, Inez Y. Fung**, and David A. Randall*

NATURE, 376, 240-243.

**National Aeronautics and Space Administration, Goddard Space Flight Center, Institute for Space Studies, 2880 Broadway, New York, New York 10025, USA, and School of Earth and Ocean Sciences, University of Victoria, Victoria BC V8W2Y2, Canada

The mixing ratio of atmospheric carbon dioxide (CO2) is increasing due to fossil fuel combustion, but the rate of increase (about 3e12 kg/yr , GtC/yr) is only about half of the rate of emissions (6 GtC/yr). A complete accounting of the remaining 3 GtC/yr is not possible at this time. It is clear that some of the global sink is in the oceans and some is in the terrestrial biosphere, but the relative importance of each, as well as the geographic distribution of the sink and the mechanisms involved are still a matter of debate. The observed record of spatial and temporal variability of CO2 concentration at remote marine sites around the world contains information that has been used in conjunction with numerical models of atmospheric transport to deduce the location and nature of the carbon sinks. One of the most important constraints on such estimates is the observed gradient in CO2 concentration between the high latitudes of the Northern and Southern Hemispheres. Simulation models of tracer transport suggest that a significant north-south gradient is due to fossil fuel emissions, with small contributions arising from natural carbon exchanges. Using a full atmospheric general circulation model with a special representation of turbulent mixing near the ground, we investigated the transport of CO2, and found that the meridional gradient imposed by the seasonal terrestrial biota is nearly half as strong as that imposed by fossil fuel emissions. Such a strong gradient implies that the sinks of atmospheric CO2 in the Northern Hemisphere must be stronger than previously suggested.

Annual mean mixing ratio of CO2 (ppmv) in the PBL as simulated by the CSU GCM using surface fluxes representing the purely seasonal exchange with the terrestrial biosphere as calculated in ref. 15 and used in ref. 4. The value at the South Pole has been subtracted at every grid point. The concentration was initialized to be globally uniform, and the model was run on a coarse 7.2 x 9 grid with 9 levels for 10 years. The end of this *spin up: run was then used as the initial condition for a further 4 year integration on a 4 x 5 grid with 17 levels. All results presented here are for the final three years of the 4 x 5 run.

Annual mean mixing ratio (ppmv) of simulated CO2 (a) using the seasonal biotic fluxes of ref. 4, and (b) using the fossil fuel emission estimates of ref. 4. Values are annual means in the PBL at the grid cells containing the 21 flask sampling stations used in ref. 4. Where the grid cell was defined as land, we chose the adjacent cell off-shore because flask sampling protocol selects marine air only. Filled squares indicate values simulated by the CSU GCM, and open circles indicate values simulated by the GISS tracer model. The curves are cubic polynomials fit by least-squares to the station values (the solid curves were fit to the GISS results and the dashed curves were fit to the CSU GCM results). The value at the South Pole has been subtracted.

Seasonal variations of (a) prescribed surface flux of CO2 from the terrestrial biosphere to the atmosphere, (b) simulated mass flux due to cumulus convection at the top of the atmospheric PBL, and (c) simulated depth of the atmospheric PBL. Values are area-weighted means for land points north of 28 N, averaged for each calendar month.

Correlation coefficients of the relationship between prescribed monthly mean terrestrial surface CO2 flux and (a) simulated cumulus mass flux at the top of the PBL, and (b) simulated depth of the PBL. The correlation was calculated separately at each land grid cell in the model, and the resulting values were contoured to produce the maps in the figure. Note that the color scheme has been reversed (reds indicate negative values and blues indicate positive values) for easier comparison to Fig. 1.