Interdecadal Changes in El Nino Onset in the Last Four Decades

Bin Wang
Department of Meteorology, School of Ocean and Earth Science and
Technology, University of Hawaii
2525 Correa Road, Honolulu, HI 96822, USA

J. Climate, 8, 267-285

Abstract | Introduction | Summary


The characteristics of the onset of the Pacific basin-wide warming have experienced notable changes since the late 1970s. The changes are caused by a concurrent change in the background state on which El Nino evolves.

For the most significant warm episodes before the late 1970s (1957, 1965, and 1972) the atmospheric anomalies in the onset phase (November to December of the year preceding the El Nino) were characterized by a giant anomalous cyclone over east Australia whose eastward movement brought anomalous westerlies into the western equatorial Pacific, causing development of the basin-wide warming. Meanwhile, the trades in the southeastern Pacific (20oS-0o, 125oW-95oW) relaxed back to their weakest stage, resulting in a South American coastal warming which led the central Pacific warming by about three seasons. Conversely, in the warm episodes after the late 1970s (1982, 1986-87, and 1991), the onset phase was characterized by an anomalous cyclone over the Philippine Sea whose intensification established anomalous westerlies in the western equatorial Pacific. Concurrently, the trades were enhanced in the southeastern Pacific, so that the coastal warming off Ecuador occurred after the central Pacific warming.

It is found that the atmospheric anomalies occurring in the onset phase are controlled by background SSTs that exhibit a significant secular variation. In the late 1970s, the tropical Pacific between 20o S and 20oN experienced an abrupt interdecadal warming, concurrent with a cooling in the extratropical North Pacific and South Pacific and a deepening of the Aleutian Low. The interdecadal change of the background state affected El Nino onset by altering the formation of the onset cyclone and equatorial westerly anomalies and through changing the trades in the southeast Pacific, which determine whether a South American coastal warming leads or follows the warming at the central equatorial Pacific.


One of the fundamental questions concerning the nature and prediction of El Nino-Southern Oscillation (ENSO) is how the turnabout from a cold to a warm state (or vice versa) takes place. This question was originally raised by Bjerknes (1969) and has remained an outstanding issue to date (Cane 1993). A satisfactory answer to this question may also address other related important questions, such as what determines the preferred time scale and what causes the aperiodicity of the Southern Oscillation (SO), and why the transition between warm and cold states is intimately linked to annual cycles (Wyrtki 1982). The present analysis will focus on the transition from a cold to a warm state of the ENSO cycle.

Rasmusson and Carpenter (1982) made a comprehensive description of a composite ENSO scenario based on six events during 1950-1976. They found significant westerly anomalies occurring over the western equatorial Pacific in the onset (around the end of the year preceding El Nino) phase. These westerly anomalies occurring in the western equatorial Pacific can induce eastward propagating equatorial Kelvin waves and have been suggested as the trigger of the South American coastal warming (Wyrtki 1975; Philander 1981, 1985; Busalacchi and O'Brien 1981).

What is responsible for the initial changes in the western Pacific wind field? A variety of speculations have been made, including the twin cyclones which develop in the western-central Pacific (Keen 1982), the impacts of the cold surges from the Southeast Asia winter monsoon (Lau et al. 1983), the enhancement of the Australian summer monsoon (Hackert and Hastenrath 1986), and the persistent development of the intraseasonal oscillation (Lau and Chan 1986). In a series of studies, Barnett (1983, 1984, 1985) suggested that the surface wind and sea-level pressure (SLP) anomalies, which eventually cause most of the changes in SST in the equatorial Pacific, originate in the equatorial Indian Ocean and propagate slowly eastward into the Pacific. The eastward propagation of free tropospheric zonal wind anomalies from the South Asian monsoon region to the western Pacific seen by Yasunari (1985, 1990) and Gutzler and Harrison (1987) seem to support Barnett's finding. An interpretation of the monsoon-ENSO connection was offered by Meehl (1987) who emphasized the important impacts of the biennial variation of the monsoon circulation on ENSO. Trenberth and Shea's (1987) analysis of long record SLP data, on the other hand, revealed that the dominant feature of the SO is a standing seesaw, and the eastward propagation of SLP anomalies is not very regular and is not supported by the long-term record. They showed that changes over the South Pacific pole of the SO lead opposite changes in the Indonesian pole by one to two seasons. A possible South Pacific role in the ENSO onset was also discussed by van Loon and Shea (1985, 1987), Trenberth and Shea (1987) and Kiladis and van Loon (1988). The previous diverse and disputed views invite further investigations.

Three recent warm events (1982-83, 1986-87, 1991-92), which were documented by Gill and Rasmusson (1983), Kousky and Leetmaa (1989), Wang (1992), Janowiak (1993), and Kousky (1993), appear to behave quite differently from the canonical scenario of Rasmusson and Carpenter (1982) (RC composite hereafter). In the last three events the central Pacific warming was not preceded by South American coastal warming as described in RC composite. The change in the evolution of SST anomalies is not unique for the post-1950 period. Desser and Wallace (1987) noticed that the warming off Peru occurred in advance of the central Pacific warming (as inferred from the episode of positive Darwin SLP anomaly) in 1925 event, whereas it occurred subsequent to the central Pacific warming in the 1930-31 and 1940-41 events. They, however, did not explore the causes for the change.

Because each warm event has its own character, case studies are necessary to reveal differences from case to case. The derived information is valuable for understanding the cause of the irregularities of the ENSO cycle. The present study is based on a case analysis of the six most significant events during 1950-1992. The intent is to document differences between the Rasmusson-Carpenter canonical ENSO scenario and the latest three ENSO events, and to reveal the causes of the differences. The results indicate that there have been remarkable changes in the characteristics of the onset after the 1976 warm event. The changes are attributed to an abrupt interdecadal change which occurred around 1977.

In this paper, the term El Nino is taken to be synonymous with a warm (or ENSO) episode which means the major anomalous warmings of the eastern-central equatorial Pacific Ocean and along the coast of South America. Section 2 describes the data used in the present analysis. Following a brief discussion of the difference in the development of SST anomalies in section 3, a detailed documentation of the contrasting features of the atmospheric circulations and SST in the onset between pre-1977 and post-1977 events is given in section 4. Possible causes of the changes in the El Nino onset are discussed in section 5. The last section summarizes major results and discusses questions remaining open for future studies.


In the onset phase of El Nino the states of the atmosphere and ocean are close to their annual cycles. The coupling of the atmosphere and ocean is relatively weak, because the annual cycle, to a large extent, is determined by insolational forcing, especially in the western Pacific. Even in the tropical eastern Pacific where air-sea interaction is involved, the insolational forcing also fundamentally regulates the annual cycle (Mitchell and Wallace 1992; Wang 1994a). As such, in the onset phase the interdecadal variations of the background state may effectively influence the evolution of the ENSO cycle.

It is hypothesized that the interdecadal change of the backgroud state that occurred in late 1970s is responsible for the changes in the characteristic evolution of the El Nino onset. The hypothesis is based on the evidence that the composite SST anomalies in the antecedent and onset phases for the pre-and post-1977 events (Figs. 7a,b and 8a,b) resemble, respectively, the cold and warm stages of the interdecadal variation in the tropical Pacific (Fig. 9). The hypothesis is also based on the assertion that the warm events evolving from a warm state of the interdecadal mode display features of onset that differ from those of the warm events evolving from a cold state of the interdecadal mode.

The processes by which the interdecadal mode possibly affects the onset of El Nino episodes are schematically summarized in Fig. 11. The warm state of the interdecadal mode that occurred after the late 1970's may influence the ENSO onset in two ways. First, the warming in the equatorial central Pacific and off-equatorial tropical eastern Pacific can offset the La Nina (cold episode) condition of an ENSO cycle in the antecedent phase, yielding a near-normal Walker circulation over the western and central Pacific and enhanced trades in the southeastern Pacific. The persistence of the enhanced southeast trades in the onset phase may prohibit South American coastal warming in the development phase, resulting in an absence of the coastal warming in boreal spring of the ENSO year which otherwise could lead the central Pacific warming. Another role of the interdecadal warming near the date line is to raise the mean SST, so that the SST along the equator would reach a given temperature (say 28oC) further east. This would favor eastward movements of convection and associated westerly anomalies from the western to central Pacific, leading to a central Pacific warming followed by a coastal warming in the year subsequent to the ENSO year. Secondly, the cooling in the western North Pacific favors a southeastward extension of east Asian winter monsoons so that northeasterly anomalies prevail north of the Philippine Sea. The large SST gradients on the poleward side of the interdecadal warming in the North Pacific would induce anomalous southwest thermal winds which, together with the enhanced northeasterly monsoon, favor the formation of the Philippine Sea onset cyclone. After the late 1970s, the equatorial westerly anomalies, which eventually cause the Pacific warming, are associated with the development of this onset cyclone. Parallel arguments can be made as to how a cold state of an interdecadal variation could affect the background SST so that the coastal warming may lead the central Pacific warming and the formation of the Australia-SPCZ onset cyclone is favored.

The results derived from the present analysis suggest that secular changes in the background "mean" state may have profound impacts on the ENSO evolution. The anomaly models which are used to predict ENSO should take into account this factor in specification of the model's mean states.

In addition to the interdecadal mode, relatively high frequency (intraseasonal to annual time-scales) disturbances, perhaps primarily in the extratropical atmosphere, could also interfere with the ENSO mode and cause irregularities in the ENSO cycle. The occurrence of minimum pressure anomalies during the onset phase in the subtropics around the node line of the Southern Oscillation (15oN-30oN and 15oS-30oS, 160oE-155oW), especially in the North Pacific subtropics (Figs. 5b and 6b), may be related to abnormal seasonal variations in the extratropical atmosphere. This pressure fall favors the establishment of the onset cyclones and the equatorial westerly anomalies over the western Pacific which play a critical role in the initiation of the instability of the coupled ocean/atmosphere system in the western-central Pacific. The common features of the transition from a cold to a warm state of the ENSO cycle have been discussed in detail in an accompanying paper (Wang, 1994b). The exact cause of the pressure fall near the node line of the Southern Oscillation before and during the onset phase of El Nino is not clear and calls for further investigations.

There is a possible feedback from ENSO episodes to the interdecadal variations. However, the facts that the interdecadal mode has a different spatial pattern from ENSO mode and taht the interdecadal change of phase had occurred before the extremely strong 1982 ENSO appear to suggest that the interdecadal SST change is unlikely resulted from the excessive intensity of recent warm episodes. It is not known, however, what causes the relatively abrupt interdecadal change and what maintains the interdecadal SST anomalies. Further studies of the interaction between ENSO and interdecadal variation and the teleconnections between the tropics and midlatitudes of both hemispheres on both interdecadal and ENSO time scales are necessary to reveal the cause of the interdecadal variation and to further assess its impact on the evolution of ENSO.

Back to top | Back to menu