16 September 2013

Contributions of Radiative Forcing and Climate Feedbacks to Range of CMIP5 Global Warming Responses


In climate model projections of global warming, carried out as part of the Coupled Model Intercomparison Project (CMIP5), the range of responses to prescribed changes in greenhouse gases and aerosols can be attributed to differences in the radiative impacts of these evolving constituents as well as model differences in the strengths of various climate feedbacks.  A diagnostic experiment in which a large ensemble of GCMs was subjected to an abrupt quadrupling of CO2 concentration has now been analyzed to determine the extent to which differences in forcing and feedbacks can account for the range of CMIP5 model responses to various future greenhouse emission scenarios.  Results from this analysis have been published in Geophysical Research Letters.


To determine why CMIP5 models differ in response to common forcing scenarios, a linear forcing-feedback regression analysis of Gregory et al. (2004) was applied to the abrupt quadrupling of CO2 experiments.  The equilibrium climate sensitivities (for a doubling of CO2) were found to range from 2.1 to 4.7 K for the 15 models analyzed (see figure). As in earlier studies, differences in cloud feedbacks were found to be important contributors to this range.  In addition, however, significant differences in “adjusted” radiative forcing were found that account for both the instantaneous radiative impact of greenhouse gases, but also the “fast adjustments,” now included in the radiative forcing, which are not a consequence of subsequent surface temperature changes. The clear-sky CO2 radiative forcing in the CMIP5 models was found to vary by about 25%, but the range was enhanced by differences in cloud (and water vapor) masking and “fast adjustments” that ultimately resulted in differences of about 40%.


The range of climate sensitivity of CMIP5 models is similar to the range found in previous generation models. Uncertainty in future projections will only be reduced if cloud responses induced directly by radiative forcing as well as feedbacks linked to mean global temperature can be accurately represented.