4.4. El Niño variations
Although the episodic warming off the Peru coast had been known and named El Niño since the 1890s (Philander, 1990), its connection to the planetary-scale "Southern Oscillation" was only beginning to be understood with the pioneering work of Bjerknes (1961, 1972). The mechanism of coastal warming continued to be assumed local until Wyrtki (1974a, 1975a) used island and coastal sea level time series to establish that mid-Pacific westerly wind anomalies generated downwelling Kelvin waves that deepened the eastern thermocline and allowed warming to occur, independent of the local winds. The key to this insight was the realization that coastal winds during El Niño events were usually upwelling-favorable, and thus could not produce the observed warming (Fig. 11). Noting the strong alongshore winds during the 1964–1965 El Niño, Wyrtki (1975a) commented "Consequently, one must abandon the idea that winds off Peru are abnormally weak during the periods of El Niño, and one must search for other ways to explain this phenomenon". The extreme SST, wind and rainfall anomalies produced by the El Niño of 1982–1983 spurred a major effort that continues to the present to observe and understand the ENSO (El Niño-Southern Oscillation) phenomenon (see, e.g., McPhaden et al., 1998; Wallace et al., 1998; and Neelin et al., 1998, for reviews of this effort, and Wang and Fiedler, 2006, for a review of recent theories). Today we know that El Niño events begin in the western Pacific and spread eastward, with the Peru coastal warming one of the final signatures. During most El Niño events, winds over the eastern Pacific change late in the event, if at all, and the SST warming that occurs in the east is primarily remotely forced (Huyer et al., 1991; Cronin and Kessler, 2002; Carr et al., 2002). As Wyrtki realized, the drastic changes in thermocline depth and currents occur as the result of wind-forced variability carried to the east by equatorial Kelvin waves (Spillane et al., 1987; Enfield, 1987; Kessler and McPhaden, 1995b). Westerly wind anomalies over the western and central Pacific flatten the equatorial thermocline across the basin, and at the same time greatly weaken the SEC (Fiedler et al., 1992; Kessler and McPhaden, 1995a; Lagerloef et al., 1999; McPhaden, 1999; Wang and McPhaden, 2000; Bonjean and Lagerloef, 2002). Although the SEC is partly frictionally driven by the trade winds, it can weaken substantially even in the absence of local wind changes because the deepening equatorial thermocline depresses the upward bulge that in normal conditions produces westward geostrophic shear above the EUC (e.g., Fig. 6). Thus eastward current anomalies span the equator, and advection appears to be a significant part of the El Niño warming in the west-central Pacific (Kessler and McPhaden, 1995a; Ralph et al., 1997), although it is probably not the case that warm west Pacific water is carried all the way across the basin (Picaut et al., 1996; Bonjean, 2001). In the east, the most important warming influence appears to be that the deeper thermocline allows local surface fluxes to heat the surface layer (Kessler and McPhaden, 1995a; Vialard et al., 2001; Cronin and Kessler, 2002) as the source of cool upwelled water is cut off (although upwelling within the thick warm layer may still be occurring under near-normal easterly trade winds). A positive feedback occurs because warmer SST tends to inhibit the formation of the usual east Pacific stratus decks, thereby allowing further warming. Although warmest actual SSTs tend to occur in January–March of the year following the largest El Niño wind anomalies, warm SST and eastward zonal current anomalies have usually peaked in the previous September; thus El Niño appears in the east as a suppression of the cold phase of the annual cycle. It is not known why El Niño events are phase-locked to the annual cycle in this way (Wang and Fiedler, 2006).
Fig. 11. The Southern Oscillation Index (top) and interannual wind stress anomalies along the coast of Peru (bottom). Interannual anomalies are defined as the 11-month running mean of anomalies from the average annual cycle. Shading shows stress anomaly magnitude, with shading levels every 0.1 N m; lighter shades show weaker than normal winds, with a contour drawn at zero. The scale vector is at lower left. Negative SOI corresponds to El Nino conditions.
Most studies of ENSO have concentrated on the equatorial strip; there is far less information to diagnose interannual variability in the eastern Pacific warm pool region, except for SST time series, for which instrumental records with sufficient spatial resolution to examine detailed features extend back to 1981 (Appendix C.2). The XBT data used in this study (Appendix A.1) were taken during the period 1979–1996, but contain few places with sampling sufficient to produce interannual timeseries. Because of the paucity of subsurface data, Fiedler (2002) cited the results of a numerical model forced with observed winds to comment briefly on interannual variations of the Costa Rica Dome during 1980–2000 (he noted that it was weak or absent in El Niño years and was apparently stronger in the one La Niña year simulated by the model). Barberan et al. (1984) compared the results of two CTD surveys across the dome during relatively normal conditions in 1979 versus early in the El Niño event of 1982, finding a large increase in isotherm depths and implied poleward speeds along the coast but relatively small anomalies in the center of the dome. Other data sets that have been used to study interannual variability in the tropical Pacific include island sea level records (Wyrtki, 1974b, 1979), but these provide less detail on the oceanic conditions in the east because the only long records are along the coast. For these reasons, some studies of interannual variations in the northeastern tropical Pacific have relied on correlating basin-scale modes with coastal time series (Baumgartner and Christensen, 1985). However, diagnosing the mechanisms of variability requires analysis of the offshore spatial patterns of currents and thermal structure.
The lengthening time series from the Topex altimeter (Appendix C.3) have been used in several comprehensive descriptions of eastern Pacific annual and interannual variability (Strub and James, 2002b,c; Carr et al., 2002). The Topex instrument gives a good picture of the evolution of the region since its launch in 1992, sampling at least the El Niño of 1997–1998 and the change to La Niña conditions following that event, and we will use its spatial variability pattern to interpret interannual SST variability. Of course it must be kept in mind that statistics of the Topex variations are dominated by the very large signals of 1997–1998 (Strub and James, 2002a), and are not necessarily typical of El Niño events over the long term.
A description of the spatial pattern of interannual variability of both the Topex sea level and SST is constructed by correlating interannual variability across the region with that at 0°W, 95°W (Fig. 12), which is a good index for ENSO. For SST, correlations among the interannually smoothed 20-year time series have about 15 degrees of freedom, which means that correlation values greater than about 0.5 are significantly different from zero at the 95% level, indicated by gray shading in Fig. 12. For sea level, there are fewer degrees of freedom, and correlations are significant above about 0.7. In either case, significant correlations indicate a close correspondence of interannual variations all along the equator and spreading poleward in a narrow region along the American coast. This is a classic signature of El Niño, in which deep thermocline anomalies (equivalently high sea level), accompanied by warm SSTs, propagate eastward along the equator as Kelvin waves, then poleward along the Americas. One might consider these two maps to represent anomalies of SST and sea level at the height of a warm event. It is noted that a correlation map similar to that in Fig. 12 (top) was found from the Kaplan et al. (1998) SST record that spans more than 100 years.
Fig. 12. Correlations of interannually smoothed quantities with the same quantity at 0°W, 95°W. Top: SST from the Reynolds SST product (1981–2005). Bottom: sea level from the Topex/Jason altimeter (1993–2005). Interannual anomalies are defined as the 11-month running mean of anomalies from the average annual cycle. Shading indicates correlations significant at the 95% level (see text).
High interannual SST correlations extend to 8–10°N east of 120°W, while sea level correlations are high only to 6–7°N (Fig. 12). As mentioned above, SST changes within the eastern equatorial region are primarily due to vertical processes associated with the deepening thermocline, however the correlation patterns suggest that the meridional height gradient on the northern flank of the sea level anomaly drives eastward geostrophic currents that warm the extra-equatorial region by advection. This is consistent with previously observed increases in the flow of the NECC, and corresponding weakness or reversal of the SEC at 5–10°S, in the central Pacific during El Niños (Kessler and Taft, 1987; Kessler and McPhaden, 1995a), which are also associated with the increased depth of the cross-equatorial thermocline at the height of the warm event. Indeed, when sea level anomalies from Topex during late 1997 are added to the seasonal XBT dynamic heights (Fig. 7), the surface height gradient across the NECC is approximately doubled, all the way up around the southern limb of the Costa Rica Dome, as is suggested by Fig. 12 (bottom). Further, since the sea level anomalies follow the coast narrowly, the Costa Rica Coastal Current is also greatly enhanced. However, this flow continues north along the coast rather than circling around the dome, and thus should not be seen as a strengthening of the dome itself, but due instead to narrow thermocline deepening along the coast, with its offshore scale determined principally by the width of coastal Kelvin waves. Thermocline anomalies in the dome are poorly correlated with equatorial anomalies (note the 0.9 contour looping around the dome Fig. 12, bottom), and the center of the Costa Rica Dome deepens only slightly during El Niño events (Barberan et al., 1984; Fiedler, 2002).
The coastal El Niño signals are frequently noted as surges of poleward flow that can continue all the way to Alaska and Chile at timescales from intraseasonal to interannual, most commonly during the boreal winter season (Chelton and Davis, 1982; Spillane et al., 1987; Hormazabal et al., 2001; Strub and James, 2002c,a; Carr et al., 2002; Hormazabal et al., 2002). As noted in Section 4.1, several investigators have shown evidence of increased penetration of tropical water to the mouth of the Gulf of California associated with El Niño events (Baumgartner and Christensen, 1985; Filonov and Tereshchenko, 1999; Lavín et al., 2003), consistent with Fig. 12, associated with these surges. Comparable low-frequency pulses of the Peru-Chile Undercurrent driven by El Niño-generated Kelvin waves were described by Pizarro et al. (2002).
We noted above that the occurrence of upwelling-favorable winds along the Peru coast during El Niño events (Fig. 11) led Wyrtki (1975a) to realize that the quintessential signature of coastal warming must be remotely forced and hence that El Niño was a connected basinwide phenomenon. While the coupled dynamics that produce the coastal wind anomalies has not been diagnosed in the literature, it is consistent with an atmospheric boundary layer response to heating by SST. El Niño SST anomalies are largest from the equator to about 5°S along the coast, which makes the anomalous SST gradient northward during warm events, fostering equatorward winds at Peruvian latitudes as observed (Fig. 11). This would be an example of winds forcing the ocean in one place, oceanic Kelvin waves carrying the signal in thermocline depth across thousands of kilometers, thus changing the SST remotely, which then feeds back to modify the winds in the distant location.
Return to previous section or go to next section