Mechanisms for decadal variability of the tropical Pacific ocean-atmosphere system have been discussed extensively in the last decade since the seminal paper of Nitta and Yamada (1989) opened the door to this important topic. The spatial pattern associated with the decadal variability in the tropical Pacific is amazingly similar to what is observed during the ENSO events.

Efforts devoted to understanding the tropical decadal variability or ENSO-like variation may be classifed into four categories. The first claims that it is due to tropical-extratropical interaction (Gu and Philander 1997; Wang and Weisberg 1998; Lau 1997). The second suggests that it is merely due to the influence of extratropical decadal variability (Kleeman et al. 1999; Barnett et al. 1999). The third states that decadal tropical variability is due solely to tropical processes (Knutson and Manabe 1998; Schneider et al. 1999). The fourth holds that stochastic atmospheric forcing may contribute to generating decadal variation in the tropics (e.g., Kirtman and Schopf 1998).

One drawback of those previous studies is that they were mostly based on model outputs or simple hypotheses. Another is that possible in uences from the South Pacific were completely neglected. Therefore, we have examined the existing hypotheses using observed upper-ocean temperature and atmospheric data collected in the past four decades 1958-97 (see White 1995; Kalnay et al. 1996; Parker et al. 1995). Based on the detailed analysis, we have proposed a new mechanism for the ENSO-like decadal variability with special emphasis on the South Pacific (Luo and Yamagata 2001). Here we describe briefly the outline of our hypothesis.

2 Decadal ENSO-like variation

After filtering out the interannual and trend components with the complex Morlet wavelet transform (Torrence and Compo 1998), we applied the joint complex EOF (CEOF) method to study the propagation feature of sea level pressure (SLP) (not shown here) and subsurface temperature anomalies averaged between  =23.2 and =26.3. These isopycnals correspond to the equatorial main thermocline. The first joint CEOF mode clearly captures the ENSO-like decadal variation; this mode accounts for 40% of total decadal variance and the complete cycle takes about 14 years.

Figure 1a shows the eastward movement of this decadal signal along the equator. It takes about 3 years to propagate from the central Pacific (near 180°W) to the eastern Pacific (near 100°W). The time history of this ENSO-like decadal mode shows that it strengthens around 1970, but starts weakening in the early 1990s. We note that the slow eastward movement of the temperature anomaly along the equatorial thermocline seems to be associated with slow eastward propagation of a zonal pseudo-wind stress anomaly after the early 1970s (Figure 1b). For example, the region of anomalous easterlies moved eastward during 1985-90, together with the eastward propagation of the negative temperature anomaly. The former lags the latter by 1-2 years for the period after 1970. This evolution pattern of the thermal and zonal wind anomalies along the equator is similar to that of ENSO, which also appears to be consistent with the model result of Knutson and Manabe (1998). It reminds us of the classical coupled equatorial Kelvin wave (Philander et al. 1984) although the time scale is different. Prior to 1970, the zonal pseudo-wind stress anomaly appears to move westward; poor coupling with the ocean for this period may explain why the decadal subsurface temperature variations were weak.

3 A role of the South Pacific

Our analysis also shows that the subsurface signal is not from the North Pacific but from the South Pacific. Figure 2a shows subsurface temperature variations averaged between 180° and 160°W for the first joint CEOF mode. Temperature anomalies vary coherently between 12°S and 5°N; we note that the signal in the southern ocean precedes that at the equator by about 1-2 years (Figure 2b). Once the temperature anomalies reach the equator from the south, they turn to move eastward gradually (see Figure 1a). Since the subducted water in the central North Pacific cannot reach the equator, as seen clearly in Figure 2a, the signal originating in the tropical South Pacific seems to be a more plausible candidate for inducing the decadal ENSO-like variation.

Figure 3 shows a 0.5-year lag-regressed pattern of pseudo-wind stress referenced to Niño3 subsurfacetemperature variations derived from the first joint CEOF mode. In the South Pacific, anomalous westerlies extend southeastward up to the west coast of South America. Equatorward anomalous winds occur in the western tropical South Pacific. This wind field in the South Pacific may be related to the southward shift of the Subtropical High; it is reminiscent of the atmospheric teleconnection pattern in the Southern Hemisphere (see the AGCM result of Lau and Nath 1994). Associated with this teleconnection pattern, negative wind curl anomalies are distributed in the southeast-northwest direction as shown by the shading. Forced by this anomalous wind field, the thermocline may shoal, leading to a cold subsurface temperature anomaly. This subsurface signal then moves northwestward and reaches the equatorial region (see Figure 2).

4 A new scenario for the origin of ENSO-like decadal variability

Based on the analysis of observational data, we have suggested a new scenario for the origin of ENSO-like decadal variation as shown schematically in Figure 4. When we observe a positive temperature anomaly in the eastern tropical Pacific, we expect the southeastward wind anomaly excited in the South Pacific owing to the atmospheric teleconnection (Lau and Nath 1994). This atmosphericresponse needs to be discussed in more detail using AGCMs. Because of the large meridional shear associated with this wind anomaly, a negative wind stress curl pattern may appear in the tropical South Pacific. This negative wind stress curl then causes upwelling and thus induces a negative subsurface temperature anomaly there owing to oceanic processes. The subsurface signal subsequently moves northwestward and reaches the western and central equatorial Pacific. This movement could be attributed to Rossby wave propagation (Jin 2000, personal communication) and/or mean low advection (McCreary and Lu 1994; Liu et al. 1994). The mechanism, however, needs to be examined more carefully. After reaching the equatorial region, the thermal signal moves eastward along the equator; it is associated with eastward movement of anomalous easterlies. This suggests the existence of air-sea interaction, explaining the slow eastward propagation along the equator. The negative temperature signal nally reaches the eastern equatorial Pacific after about 3 years and replaces the original positive temperature anomaly there. The negative anomaly then may grow via tropical air-sea interaction processes.

Corresponding to the reversed temperature anomaly in the Niño3 region, the wind stress curl anomaly in the tropical South Pacific must change the sign owing to the atmospheric teleconnection in the Southern Hemisphere. The anomalous wind field may deepen the thermocline there, leading to an increase of the subsurface temperature. Thus, a similar but warm phase starts in the cycle. This completes the whole scenario for the ENSO-like decadal variation. Our estimate shows that the decadal phenomenon takes about 14 years to complete the cycle.

In summary, the present note stresses the importance of the role of the tropical South Pacific hitherto discarded in order to explain the turnabout of the tropical Pacific decadal phenomenon.

Figure 1: The eastward movement of the decadal signal: (a) decadal subsurface temperature variations along the equator from the first joint CEOF mode; (b) decadal zonal pseudo-wind anomaly along the equator (unit: m 2 / s 2 ). (Click to Enlarge)

Figure 2: The equatorward propagation of the decadal signal: (a) same as in Figure 1a, but along the meridional plane averaged between 180° and 160°W; (b) time evolution of the subsurface temperature anomalies averaged at 14°-10°S (solid line) and at 4°S-4°N (dashed line). (Click to Enlarge)

Figure 3: 0.5-year lag-regression coefficients of pseudo-winds (in m / s ) based on the Niño3 subsurface temperature (150°W-90°W, 4°S-4°N) from the rst joint CEOF mode. Contours denote the related wind curl pattern. Contour interval is 10 (Click to Enlarge)

Figure 4: Schematic diagram for the dynamics of ENSO-like decadal variation. (Click to Enlarge)

• Barnett, T. P., D. W. Pierce, M. Latif, D. Dommenget, and R. Saravanan, Geophys. Res. Lett., 26 ,
  615-618 (1999).
• Gu, D. F., and S. H. Philander, Science, 275 , 805-807 (1997).
• Kalnay, E., and Coauthors, Bull. Amer. Meteor. Soc., 77 , 437-471 (1996).
• Kleeman, R., J. P. McCreary, and B. A. Klinger, Geophys. Res. Lett., 26 , 1743-1746 (1999).
• Knutson, T. R., and S. Manabe, J. Climate, 11 , 2273-2296 (1998).
• Lau, N. C., and M. J. Nath, J. Climate, 7 , 1184-1207 (1994).
• Lau, N. C., Bull. Amer. Meteor. Soc., 78 , 21-33 (1997).
• Liu, Z., S. H. G. Philander, and R. C. Pacanowski, J. Phys. Oceanogr., 24 , 2606-2623 (1994).
• Luo, J.-J., and T. Yamagata, Long-term ENSO-like variation with special emphasis on the South Pacific. To appear in J. Geophys. Res. (2001).
• McCreary, J.P., and Lu P., J. Phys. Oceanogr., 24 , 466-497 (1994).
• Nitta, T., and S. Yamada, J. Meteor. Soc. Jpn, 67 , 375-383 (1989).
• Parker, D. E., C. K. Folland, and M. Jackson, Climate Change, 31 , 559-600 (1995).
• Philander, S. G. H., T. Yamagata, and R. C. Pacanowski, J. Atmos. Sci., 41 , 604-613 (1984).
• Schneider, N., A. J. Miller, M. A. Alexander, and C. Deser, J. Phys. Oceanogr., 29 , 1056-1070 (1999).
• Torrence, C., and G. P. Compo,Bull. Amer. Meteor. Soc., 79 , 61-78 (1998).
• Wang, C., and R. H. Weisberg, Geophys. Res. Lett., 25 , 3979-3982 (1998).
• White, W. B., Prog. Oceanogr., 36 , 169-217 (1995).