Modes of Interannual Variability in the Southern Hemisphere Atmospheric Circulation: Assessment and Projected Changes in CMIP5 Models
In this paper, we use recently developed methodologies (Zheng and Frederiksen, 2004 (Climate Dynamics); Zheng et al., 2009(Climate Dynamics)) that allow the separation of the ‘signal' from the ‘noise'. For observations or reanalysis data, it is assumed that the seasonal mean of a climate variable can be thought of as consisting of two components: (1) an “intraseasonal weather noise” component related to internal dynamics predominantly on intraseasonal time scales and (2) a “slow” component related to internal dynamics and external forcing on slowly varying (interannual or longer) time scales. It is possible then to estimate the covariance matrix of each component and derive modes of variability in each component. For an ensemble of coupled models, it is possible to identify separately the modes for both the slow internal and slow externally forced parts of the slow component.
An assessment is first made of the Coupled Model Intercomparison Project Phase 5 (CMIP5) models in simulating the modes of variability in the 500hPa geopotential height reanalysis data for the Southern Hemisphere summer and winter seasons in the 20th century. We find that the leading modes of variability in the intraseasonal component are generally well reproduced in the CMIP5 models in both seasons. There are clear differences between models in their reproduction of the leading modes in the slow component (related to SAM and ENSO variability). On this basis, the models have been ranked and an ensemble of “good” models has been used to estimate the structures of the modes in the slow internal and slow external components. The slow internal modes reflect the structure of SAM and ENSO modes of variability. The leading slow external mode has a statistically significant trend and a spatial structure that is consistent with changes in greenhouse concentrations.
This same ensemble of models shows that there are only small changes in the leading modes in the intraseasonal component under the different RCP scenarios in the twenty-first century. Changes in the variance and percentage explained are larger for the modes in the slow component, and there are subtle regional scale changes in their spatial structure. By far, the largest changes are in the leading mode in the slow external component in both the spatial structure and variance explained.