Attribution of Multidecadal Climate Trends in Observations and Models

Abstract:

We analyzed long-term trends and multidecadal variability in a network of well-known climate indices based on the sea-surface temperature and sea-level pressure fields, in observations and CMIP5 model simulations. Various ensemble-mean estimates of the forced variability were derived from 18 independent ensembles of these simulations, each using a single model with fixed physics package and an identical forcing history. It was shown that the residual intrinsic variability time series over the total of 116 simulations considered are statistically independent. The 18 estimates of the forced signal were further used for semi-empirical attribution of the observed climate variability. The model uncertainty results in a fairly broad range of the “intrinsic” multidecadal variability so estimated. Furthermore, the observed surface temperature “intrinsic” residuals exhibit the levels of multidecadal variance overwhelmingly exceeding those of the simulated intrinsic multidecadal variability. This may reflect the climate models’ consistently underestimating the true amplitude of the forced multidecadal variability, and/or be indicative of the lack of multidecadal intrinsic dynamics in these models.