FIGURE 2 Multitaper power spectra of the Cool Tongue (CT) and Global Residual (GR) climate indices. The inset is a plot of the corresponding power spectra of rotated Principal Component (PC) series: "tropical" PCs (solid curve) and "North Pacific" PCs (dashed curve). In the CT and GR spectra, at the lowest frequencies, powers > 80200 exceed 95% confidence levels—not shown here—in comparison to red noise (Mann and Lees 1996); at the highest frequencies, powers > 30 reach 95% confidence levels.

Frequency, in cycles/year

Frequency, in cycles/year a strong decadal character, much like the GR index (and is correlated with GR at r = +0.74), but has not had any ENSO variability explicitly removed; indeed, its correlation with the CT series is r = +0.38. The PDO was intended to reflect the decadal variability of the North Pacific in a simple index, but it has a substantial expression throughout the tropical and South Pacific (see Fig. 3d in this chapter or Fig. 2a of Mantua et al., 1997).

Tropical ENSO variability historically—e.g., Troup (1965) and most more recent references—has been indexed by the SOI and SST anomalies in selected equatorial regions. The SOI is usually defined as the difference between standardized monthly departures of sea level pressure (SLP) from their normal seasonal cycles at Tahiti and Darwin, Australia. When SLPs are low in the eastern tropical Pacific near Tahiti and high in the western tropical Pacific near Darwin (when SOI is negative), the tropical easterly winds become weak and the resulting slackening of equatorial upwelling of cool deep water in the eastern equatorial Pacific, together with a relaxation of warm waters from the western Pacific into the usually cool eastern Pacific, leads to warmer than normal SSTs in the eastern equatorial Pacific. These episodes of warm equatorial SSTs in the central and eastern Pacific have far-reaching climatic consequences and have traditionally been called El Niños. When the SLP anomalies are reversed (high at Tahiti and low at Darwin) and the SOI has a positive value, the tropical easterlies strengthen and the eastern equatorial Pacific cools anomalously in episodes that recently have been called La Niñas, to emphasize their contrasts with El Niños. Thus, the SOI index is an atmospheric measure of ENSO processes that complements, but largely reflects, the CT index, with which it is highly anticorrelated (r = —0.73). The PDO index is, in part, due to ENSO, as is indicated by the correlation between the PDO and SOI indices (r = —0.51).

Although the CT and GR indices of ZWB will be the focus of much of the remaining discussion, it was also useful to develop yet another pair of indices for inter-annual and decadal ENSO-like variations, a pair of indices that—by design—are entirely uncorrelated with each other but which are not derived from any explicit temporal filtering. In order to develop these new indices, three commonly used climate indices—SOI, CT, and PDO—were joined in a simple trivariate PC analysis. The two most influential of the resulting components were weighted heavily on (1) SOI and CT, the tropical indices, and (2) PDO, the extratropical index, respectively. The benefit of this PC analysis was that, by construction, it yielded statistically uncorrelated series that described the tropical and extratropical variations without arbitrary filtering of either. The two components were then rotated by the Varimax methodology

(Richman 1986), although this served only to reduce the amount of mixing between PDO or the tropical indices in the PCs (nudging the loadings toward even more one-sided weighting of one or the other set in each case). The two resulting, still uncorrelated, rotated PC series (dotted curves, Figs. 1a and 1b) will be called the tropical PC and the North Pacific PC, respectively, although it is clear from later results that the North Pacific PC reflects pan-Pacific climate variations that may yet have their roots in the tropics. The weights of SOI, CT, and PDO in each PC are given in Table 1. When the more commonly used Nino-3 SST average, from 5°S to 5°N, 150° to °W, is substituted for CT in this calculation, the amount of mixing between indices that is required to arrive at uncorrelated tropical and North Pacific climate indices is even smaller.

The power spectra of the resulting PC series are shown in Fig. 2 (inset) and are similar in character to the spectra of CT and GR, with the tropical PC naturally echoing CT's interannual character and the North Pacific PC following GR's strong decadal dominance. Although no temporal filtering was applied when these PCs were derived, the North Pacific PC has narrower spectral peaks than does the GR series, concentrated near the periods > 25 years and near 7 years, and the tropical PC has power concentrated in a more clearly defined interannual frequency range—0.16-0.3 cycles/ year (3-7 years)—than does CT. Indeed, the tropical PC is virtually identical to ZWB's CT*, the CT index filtered to remove variability longer than 6 years.

The oceanic and atmospheric patterns that accompany the interannual ENSO and the ENSO-like climate variations indexed by the CT and GR series and by the tropical and North Pacific PCs will be described in sections 1.3 and 1.4 by using regressions of various atmospheric fields, and correlations of SST fields, with the series. In Section 1.5, the climatic consequences of this variability at the surface of the Americas will be described by similar methods. SST variations are depicted from the HSSTD set used in developing the CT and GR indices. Atmospheric forms of the variations are depicted in terms of the recently released reanalyzed global atmospheric fields of 500-mbar height

TABLE 1 Principal Component (PC) Loadings for Varimax-Rotated Factors of October-September, 1915-88, Average Indices of Pacific Climate Variability


Tropical PC

North Pacific PC

0 0

Post a comment