An Assessment of Reliability of Future Projections by Multi-Model Ensemble Based on a New Method Measuring Contribution of External Forcings to Total Variance

Abstract

This study estimates the effect of greenhouse gases (GHGs) on future climate projections produced by multi-model ensembles (MMEs) using twenty-one coupled Atmosphere-Ocean general circulation models. In this study, potential predictability (P) is applied to assess the reliability of GHGs effects on future climate projections. P is a ratio of external variance to total variance, which is used to measure the impact of external forcing. When the internal variance is very small, P will be close to 1, and the ensemble is highly predictable, indicating that the variation is largely controlled by external forcing. P reveals that future climate changes in the 21st century due to the GHGs effect are statistically more significant over higher latitudes than lower latitudes, and over land rather than ocean for the surface air temperature (SAT). For precipitation (PCP), most regions show that P is less than 0.5, except for the South African, North Eastern Pacific, North Atlantic, Central South American, and East Asian regions. It is clearly shown that the strong regions of correlation between MME and the observation where P is strong have a small inter-model difference. This study suggests that an analysis method such as P provides credibility to signals in projection of the SAT and PCP changes.