Data And Definitions

Interannual and decadal climate variations in the Pacific basin can be traced through literally dozens of physical and biological indices (e.g., Ebbesmeyer et al., 1991). In this chapter, two sets of indices based on sea surface temperature (SST) patterns of global scale will be used to depict the North Pacific climate and its effects on the Americas. Two indices developed by ZWB—the Cool Tongue (CT) index and the Global Residual (GR) index—will be discussed for the most part. However, for completeness, a second pair of indices based on a linear combination of some more commonly available time series—the Southern Oscillation Index (SOI), the CT index, and the Pacific (in-ter)Decadal Oscillation (PDO) index—also will be considered. The origins and significance of these various indices are discussed later.

Global SSTs were analyzed by ZWB using linear regressions and principal component (PC) analyses to separate interannual ENSO-related SST variations from other interannual and decadal variations and to describe those other SST variations. Interannual ENSO-related SST variations were characterized by the CT index, which is the average of SST anomalies from 6°N to 6°S, 180° to 90°W, an index previously used by Deser and Wallace (1990). This region is the locus of cool SSTs in the eastern equatorial Pacific, where the El Niño process has its clearest surface expression. Monthly mean SST anomalies on a 5° X 5° grid, from an updated version of the global United Kingdom Meteorological Office Historical Sea Surface Temperature Dataset (HSSTD; Folland and Parker, 1990, 1995), were used to form this series (Fig. 1a), and the CT series was highpass filtered to remove variations with periods longer a

1910 1930 1950 1970 1990

1910 1930 1950 1970 1990

Year

FIGURE 1 (a) Unfiltered Cool Tongue (CT) index and rotated "tropical" Principal Component (PC) index of dominantly interan-nual El Niño / Southern Oscillation (ENSO) variability in the tropical Pacific, and (b) Global Residual (GR) index and rotated "North Pacific" PC index of dominantly decadal ENSO-like variability in the Pacific basin. The indices are described in the text. The CT and GR indices are identical to those of Zhang et al. (1997).

1910 1930 1950 1970 1990

Year

FIGURE 1 (a) Unfiltered Cool Tongue (CT) index and rotated "tropical" Principal Component (PC) index of dominantly interan-nual El Niño / Southern Oscillation (ENSO) variability in the tropical Pacific, and (b) Global Residual (GR) index and rotated "North Pacific" PC index of dominantly decadal ENSO-like variability in the Pacific basin. The indices are described in the text. The CT and GR indices are identical to those of Zhang et al. (1997).

than 6 years, forming a strictly interannual series called CT*. Then, the entire SST data set was regressed against the CT* time series, and the best linear fits at each grid point were subtracted from the grid-point SSTs to arrive at the time series of residuals. By construction, these residuals are uncorrelated with the CT* series. A PC analysis of these residual series yielded a leading mode of "non-ENSO" SST variation that was, surprisingly, very ENSO-like in its spatial patterns. This leading mode is called the GR index by ZWB and represents the decadal ENSO-like variations of the global and, especially, the Pacific SSTs (Fig. 1b). Although CT* and GR are not correlated, CT and GR are (r = —0.55). In this chapter, the interannual and decadal climate variations in the Pacific basin will be characterized in terms of these same CT and GR series; CT, rather than CT*, is used because it is more similar to such commonly used ENSO indices as the unfiltered SST indices along the equatorial Pacific in so-called Nino-3 and Nino-4 regions (e.g., Allan et al. 1996) than the band-pass-filtered CT*.

The differing timescales of the CT and GR series can be demonstrated by spectral analysis. Shown in Fig. 2 are multitaper power spectra (Percival and Walden, 1993; Lees and Park, 1995) of the two series. The CT

spectrum (solid curve) indicates variance spanning a broad range of frequencies centered near 0.2 cycles/ year, corresponding to a period of 5 years. In contrast, GR has a spectrum that increases rather steadily from high to low frequencies, with a local maximum near 0.16 cycles/year. Thus, GR is dominated by low-frequency variations with timescales between about 6 years and several decades; CT is dominated by inter-annual variations in the 3- to 7-year range. Recall that the CT series used here has not been temporally filtered; therefore, the interannual dominance of CT shown reflects the true mix of frequencies in the tropical Pacific ENSO variations. The GR index is derived as a residual from an interannually filtered CT* series (and its reflection in global SSTs), and thus, as a measurement of North Pacific climate variation, it is biased by design toward the more decadal parts of Pacific climate variability.

Alternative approaches and indices for characterizing the interannual and decadal modes of Pacific climate variation have been developed in the literature. Enfield and Mestas-Nunez (Chapter 2) used complex empirical orthogonal function (EOF) analyses of global SSTs to arrive at similar modes. Mantua et al. (1997) used the CT index and a Pacific Decadal Oscillation (PDO) index to describe essentially the same variability. The PDO index is defined by Mantua et al. (1997) as the first PC of monthly SSTs, from the HSSTD, poleward of 20°N in the Pacific basin, initially calculated from anomalies for the period 1900-93. This index has

0 0

Post a comment