A simple Earth system model is developed by coupling a box model of the glo
bal carbon cycle to an energy-balance approximation of global temperature.
The model includes a range of feedback mechanisms between atmospheric CO2,
surface temperature and land and ocean carbon cycling. It is used to assess
their effect on the global change being driven by anthropogenic CO2 emissi
ons from fossil fuel burning and land-use change. When tuned to reach the 1
990 level of atmospheric CO2, the model CO2 predictions for 1832-1990 are r
easonably close to ice-core and instrumental records, observed global warmi
ng of similar to0.6 K from 1860-1990 is accurately predicted and the land a
nd ocean carbon sinks for the 1980s are close to IPCC central estimates. Th
e ocean sink is reduced by similar to0.3 GtC yr(-1) when the ocean surface
is assumed to warm at the same rate as global surface temperature. Land and
oceanic carbon sinks are predicted to be growing at present and hence buff
ering the rate of rise of atmospheric CO2. In the basic model, the current
land carbon sink is assumed to be due to CO2 fertilisation of photosynthesi
s. The slight warming that has occurred enhances soil respiration (carbon l
oss) and net primary productivity (carbon uptake) by similar amounts. When
the model is forced with a "business as usual" (IS92a) emissions scenario f
or 1990-2100 followed by a linear decline in emissions to zero at 2200, CO2
reaches a peak of 985 ppmv in 2170 and temperature peaks at +5.5 K in 2180
. Peak CO2 is similar to 135 ppmv higher than suggested by IPCC for the sam
e forcing, principally because global warming first suppresses the land car
bon sink then generates a land carbon source. When warming exceeds similar
to4.5 K, soil respiration "overtakes" the CO2 fertilisation of NPP, trigger
ing a release of similar to 70 GtC from terrestrial ecosystems over similar
to 100 years. When the effects of temperature on photosynthesis, respirati
on and soil respiration are removed, peak levels of CO2 are reduced by simi
lar to 100 ppmv and peak temperature by similar to0.5 K. Distinguishing sep
arate soil carbon pools with different residence times does not significant
ly alter the timing of the switch to a land carbon source or its effect on
peak CO2, but it causes the source to persist for longer. If forest re-grow
th or nitrogen deposition are assumed to contribute to the current land car
bon sink, this implies a weaker CO2 fertilisation effect on photosynthesis
and generates a larger future carbon source. Peak CO2 levels are also sensi
tive by about +/- 80 ppmv to upper and lower limits on the temperature resp
onses of photosynthesis, plant respiration and soil respiration. By forcing
the model with a range of future emission scenarios it is found that the c
reation of a significant land carbon source requires rapid warming, exceedi
ng similar to4.5 K, and its magnitude increases with the rate of forcing. T
he carbon source is greatest for the most rapid burning of the largest rese
rve of fossil fuel. It is concluded that carbon loss from terrestrial ecosy
stems may significantly(similar to 10%) amplify global warming under "busin
ess as usual" or more extreme scenarios.