A two-dimensional global atmospheric transport model is used to relate estimated air to surface exchanges of CO2 to spatial and temporal variations of atmospheric concentration of CO2. This serves to illustrate the gross features of the carbon cycle and the measurement precision required of the burgeoning global observational network if the latter is to contribute useful information to the quantitative description of the cycle. Calculations are based on the atmospheric model coupled to a fixed, or variable depth, oceanic mixed layer. Zonally averaged air-sea fluxes of CO2 are estimated from a relationship between flux, partial pressure, temperature, and wind stress. The result is a time-dependent (pre-industrial to 1980) estimate of the meridional rates of net exchange of CO2 with the oceans. For example, it is estimated that, at present, the equatorial oceans release a net total of 1.3 Gt (Gt=1012 kg) of carbon into the atmosphere annually, while high latitude oceans take up a net total of 4.4 Gt. Associated with the increasing release of fossil-fuel derived CO2 into the northern hemisphere atmosphere, the model suggests that the interhemispheric difference in concentration of CO2 (high latitude north-high latitude south) has changed from ~-1 ppmv preindustrially to ~+1 ppmv in 1960 and to 4--5 ppmv at present. The 1960 distribution agrees with the limited observational data available for that time. Present day observations, although provisional, are in good agreement with model estimates of the annual mean global distribution of CO2. The influence of this changing interhemispheric distribution on the concept of the airborne fraction of CO2 is discussed. The sensitivity of the mean global distribution of CO2 to additional surface exchanges is demonstrated by modeling the hypothetical cases of a 1, 2, or 5 Gt yr-1 equatorial source (such a tropical deforestation) with a similar uptake by the temperate northern hemisphere forests. Such changes bring about a 20%, 40%, or 100% reduction, respectively, in the simulated interhemispheric concentration difference. Alternatively, the 1, 2, or 5 Gt yr-1 equatorial release is taken up by the global oceans. In this case, the south pole to equatorial concentration difference increases from 3 ppmv on average to 3.5, 4.0, and 5.5 ppmv, respectively. Taken in conjunction with presently available observations, which must be regarded as provisional, these results place constraints on the magnitude of any actual equatorial deforestation source. It is unlikely that such a source, combined with the sink, could amount to more than about 2 Gt yr-1. The difficulty of establishing more precisely the magnitude of such regional exchanges is discussed in light of the model limitations and required precision of the observational network.