Are recent global mean temperature trends anomalous relative to the CMIP5 ensemble?

Abstract:

Recent studies have described a slow-down in the rise in global mean temperature over the past 15 years, noting that these trends are anomalous with respect to the trends predicted by the CMIP5 ensemble. This result has been shown to be sensitive to the choice of starting year when assessing the hiatus with a simple linear regression. It is also sensitive to assumptions about the independence of CMIP5 ensemble members when estimating distributions of the CMIP5 ensemble. We address these interdependences of ensemble members by exploring two end-member scenarios of the CMIP5 ensemble: one in which each ensemble member is assumed independent and a second in which modeling centers are assumed independent. A more stable metric of trend, the $\delta$-slope, is employed to measure the recent divergence of a temperature time-series from the average CMIP5 projection. When treating each CMIP5 ensemble member as independent, we find that observed recent trends are anomalous for all hiatus periods starting after 1997 at the 95$\%$ confidence level. However, for the scenario assuming modeling centers to be independent, we find that observed recent trends are consistent with the CMIP5 ensemble at the 95$\%$ confidence level. This result holds across all plausible hiatus starting years tested (1990-2005). Using the Ansari-Bradley rank-sum test, seven modeling centers comprising 58 ensemble members are each shown to have lower spread in $\delta$-slope values than the distribution of $\delta$ slope values of CMIP5 ensemble members outside each modeling center at the 95$\%$ confidence level. This supports the end-member scenario that treats each modeling center as independent. In addition, examination of spatial patterns of $\delta$-slope in each CMIP5 ensemble member reveals that the simulation that is closest to observations in global mean temperature $\delta$-slope value also has one of the highest $\delta$-slope spatial pattern correlations with observations. This pattern reflects a strong negative phase of the Pacific Decadal Oscillation and further supports the notion that the recent period of warming is captured by the CMIP5 ensemble.