Uncertainties in Predicted 21st Century Vegetation Carbon Storage By CMIP5 Earth System Models

Abstract:

Global averaged surface temperature has increased by 0.6 °C over the period 1986 to 2005; and will continue rising 1.0-3.7 °C during the last 30 years of this century. Land ecosystems can sequester approximately one third of annual anthropogenic carbon dioxide emission. Therefore, dynamics of land sink is one of the key components to determine the future atmospheric CO₂ concentration and accordingly surface temperature. The accuracy of predicted surface temperature will largely depend on the uncertainty of predicted land carbon uptake. Unfortunately, the uncertainties of future land sink predicted by Earth System Models (ESMs) involved in CMIP5 turned out to be very large. The spread of the land carbon uptake within a specific Representative Concentration Pathway (RCP) scenario was larger than those variation between the four scenarios. Moreover, predicted soil carbon stocks by the end of this century extended to a wide range. Quantifying the uncertainties in predicted vegetation carbon and identify the causes for the uncertainties will help improve ESMs’ performance and give the priorities for model development. In this study, we evaluated uncertainties in predictions of vegetation carbon by twelve CMIP5 ESMs during the twenty-first century and explored the sources of uncertainties across the models. We found that the predicted changes of vegetation carbon by the end of this century varied quite much across the ESMs, from declining of 250 Pg C to increasing of 440 Pg C under 8.5 scenario, and from losing of 130 Pg C to gaining of 410 Pg C under emission-driven 8.5 scenario. We further investigated differences in responses of vegetation carbon to temperature and CO₂ increasing among the ESMs and how much the initial state, net primary productivity and residence times of vegetation carbon contribute to the uncertainties in predictions of vegetation carbon across the ESMs. Our results have the potential to help improve CMIP5 ESMs for more reliable predictions.