The multi-scale nature and climate noise properties of teleconnection indices are
examined by using the Empirical Mode Decomposition (EMD) procedure. The EMD
procedure allows for the analysis of non-stationary time series to extract physically
meaningful intrinsic mode functions (IMF) and nonlinear trends. The climatologically
relevant monthly mean teleconnection indices of the North Atlantic Oscillation
(NAO), the North Pacific index (NP) and the Southern Annular Mode (SAM) are
analyzed.
The significance of IMFs and trends are tested against the null hypothesis of climate noise.
The analysis of surrogate monthly mean time series from a red noise process shows that the
EMD procedure is effectively a dyadic filter bank and the IMFs (except the first IMF) are
nearly Gaussian distributed. The distribution of the variance contained in IMFs of an
ensemble of AR(1) simulations is nearly χ2 distributed. To test the statistical significance
of the IMFs of the teleconnection indices and their nonlinear trends we utilize an
ensemble of corresponding monthly averaged AR(1) processes, which we refer
to as climate noise. Our results indicate that most of the interannual and decadal
variability of the analysed teleconnection indices cannot be distinguished from
climate noise. The NP and SAM indices have significant nonlinear trends, while
the NAO has no significant trend when tested against a climate noise hypothesis. |