Possible feedbacks from terrestrial carbon cycling to atmospheric CO₂ were assessed using multiple models by Cramer et al. (2001). Six DGVMs (Figure 3.10a) (Foley et al., 1996; Brovkin et al., 1997; Friend et al., 1997; Woodward et al., 1998; Huntingford et al., 2000; Sitch, 2000) were driven first by CO₂ concentrations derived from the IS92a emissions scenario as in the SAR, and then with CO₂ changes plus climate changes derived from the HadCM2 coupled ocean-atmosphere general circulation model simulation including sulphate aerosol effects as described by Mitchell et al. (1995). Except for one empirical model (VECODE; Brovkin et al., 1997), the models included explicit representation of all the following processes: the CO₂ fertilisation effect on NPP (modelled explicitly in terms of photosynthesis, respiration, and feedbacks associated with carbon allocation); responses of NPP to climate specific to each plant functional type (PFT); competition among PFTs for light and water; dynamic shifts in vegetation structure due to climate and CO₂ effects; competitive limits to above-ground biomass; natural disturbance regimes and their interaction with PFT composition; soil temperature and moisture effects on heterotrophic respiration. Two models include an interactive N cycle. Land use and anthropogenic N deposition were not considered.

Driven by increases in CO₂ beyond the present day, the modelled sink due to CO₂ fertilisation continued to increase. By the middle of the 21st century the simulated land-atmosphere flux due to CO₂ was in the range -8.7 to -3.6 PgC/yr. Beyond mid-century the rate of increase became less, due to the declining photosynthetic response to CO₂. When the climate change scenario was included as well as the CO₂ increase, modelled uptake was reduced compared with the CO₂-only analysis. At mid-century, climate change reduced the uptake by 21 to 43%. A marked decline in terrestrial uptake after the mid-century was seen in two models, and one model had zero terrestrial uptake by 2100. By 2100 the range of model estimates of the land-atmosphere flux had widened to -6.7 to +0.4 PgC/yr. Increasing heterotrophic respiration in response to warming (Cao and Woodward, 1998a,b; Cramer et al., 2001) was a common factor (but not the only one) leading to reduced land uptake. The differences among the modelled climate responses were largely due to unresolved discrepancies in the response of global NPP to temperature. The balance of positive versus negative regional effects of climate change on NPP was estimated differently by these models, to the extent that the sign of the global response of NPP to climate change alone was not consistent. In addition, one model simulated a partial replacement of the Amazon rainforest by C₄ grassland. This response was not shown, or occurred on a much smaller scale, in the other models. The details of this modelling exercise are presumably dependent on sensitivity of the particular climate model, and regional aspects of the simulated climate change (Cramer et al., 2001).

IPCC Homepage