Several ways of extracting a large-scale, eastward-propagating "MJO" are explored and compared. Input fields were pentad OLR (1979-98) and ECMWF 10 m winds (1985-99).
The fields were bandpassed using complex demodulation, with central periods of 40 and 60 days. The half-power points were 28 to 72 days and 42 to 108 days, respectively.
SVD analysis (Bretherton et al 1992) was used to decompose bandpassed fields reconstructed from complex demodulation. Using an idea of Mike Wallace's, a second time series was formed by shifting the dates of the original by 15 days (1/4 period). The original and shifted time series were decomposed by SVD. The results give eigenvectors in quadrature pairs that represent propagating variability.
From another point of view, the eastward-propagating variability was extracted by successively time-shifting the band-passed fields (in x,t) so the eastward propagation became simultaneous. Then the fields were smoothed in x, then the shifting was reversed. The results are not restricted to EOF modes, but show ALL the eastward-propagating large-scale intraseasonal signals.
Figures on this page compare results from these analyses.
This page is old. The paper has been published (J.Clim. 2001)
New page for figures made during revision of this paper
Correlations (some are old):
Comparisons with different speeds and smoothings: