Evaluation of the terrestrial carbon cycle, future plant geography and climate-carbon cycle feedbacks using five Dynamic Global Vegetation Models (DGVMs)

摘要

This study tests the ability of five Dynamic Global Vegetation Models (DGVMs), forced with observed climatology and atmospheric CO, to model the contemporary global carbon cycle. The DGVMs are also coupled to a fast 'climate analogue model', based on the Hadley Centre General Circulation Model (GCM), and run into the future for four Special Report Emission Scenarios (SRES): A1FI, A2, B1, B2. Results show that all DGVMs are consistent with the contemporary global land carbon budget. Under the more extreme projections of future environmental change, the responses of the DGVMs diverge markedly. In particular, large uncertainties are associated with the response of tropical vegetation to drought and boreal ecosystems to elevated temperatures and changing soil moisture status. The DGVMs show more divergence in their response to regional changes in climate than to increases in atmospheric CO content. All models simulate a release of land carbon in response to climate, when physiological effects of elevated atmospheric CO on plant production are not considered, implying a positive terrestrial climate-carbon cycle feedback. All DGVMs simulate a reduction in global net primary production (NPP) and a decrease in soil residence time in the tropics and extra-tropics in response to future climate. When both counteracting effects of climate and atmospheric CO on ecosystem function are considered, all the DGVMs simulate cumulative net land carbon uptake over the 21st century for the four SRES emission scenarios. However, for the most extreme A1FI emissions scenario, three out of five DGVMs simulate an annual net source of CO from the land to the atmosphere in the final decades of the 21st century. For this scenario, cumulative land uptake differs by 494 Pg C among DGVMs over the 21st century. This uncertainty is equivalent to over 50 years of anthropogenic emissions at current levels.