Carbon cycle models have indicated the potential for climate change to influence the rate of CO2 uptake by both land (Section 3.7.1) and oceans (Section 3.7.2) and thereby influence the time course of atmospheric CO2 concentration for any given emissions scenario. Coupled models are required to quantify these effects.
Two general circulation model simulations have included interactive land and ocean carbon cycle components (Cox et al., 2000; Friedlingstein et al., 2001). The Cox et al. (2000) model was driven by CO2 emissions from the IS92a scenario (Legget et al., 1992) and the Friedlingstein et al. (2001) model was driven by CO2 emissions from the SRES A2 scenario (IPCC, 2000b). Both simulations indicate a positive feedback, i.e., both CO2 concentrations and climate change at the end of the 21st century are increased due to the coupling. The simulated magnitudes of the effect differ (+70 ppm, Friedlingstein et al., 2001; +270 ppm, Cox et al., 2000). In the Cox et al. (2000) simulation, which included a DGVM, the increased atmospheric CO2 is caused mainly by loss of soil carbon and in part by tropical forest die back. The magnitude of the climate-carbon cycle feedback still has large uncertainties associated with the response of the terrestrial biosphere to climate change, especially the response of heterotrophic respiration and tropical forest NPP to temperature (Cox et al., 2000; see Sections 188.8.131.52 and 3.7.1). In the following section, simplified models are used to assess these uncertainties.
This section follows the approach of previous IPCC reports in using simplified, fast models (sometimes known as reduced-form models) to assess the relationship between CO2 emissions and concentrations, under various assumptions about their future time course. Results are shown from two models, whose salient features are summarised in Box 3.7. The models lend themselves to somewhat different approaches to estimating uncertainties. In the ISAM model, "high-CO2" and "low-CO2" alternatives are calculated for every emissions scenario, based on tuning the model to match the range of responses included in the model intercomparisons shown in Figure 3.10. Uncertainties cited from the ISAM model can be regarded as providing a lower bound on uncertainty since they do not admit possible behaviours outside the range considered in recent modelling studies. In the Bern-CC model, "high-CO2" and "low-CO2" alternatives are calculated by making bounding assumptions about carbon cycle processes (for example, in the high-CO2 parametrization CO2 fertilisation is capped at year 2000; in the low-CO2 parametrization Rh does not increase with warming). This approach yields generally larger ranges of projected CO2 concentrations than the ISAM approach. The ranges cited from the Bern-CC model can be regarded as approaching an upper bound on uncertainty, since the true system response is likely to be less extreme than the bounding assumptions, and because the combination of "best" and "worst" case assumptions for every process is intrinsically unlikely.
Other reports in this collection