The Southern Oscillation: Historical Origins

Donald R. Mock
former NOAA-Skaggs Laboratory Director.



From the exhaustive correlation studies of Walker in the 1920's to the 1980 Kelvin wave-front theories of Wyrtki and others, the phenomenon known as the Southern Oscillation has attracted the considerable interest of many capable meteorologists and oceanographers, yet a comprehensive explanation of the observed characteristics has not yet been obtainable. The present study, as distilled from numerous original and review articles, presents an outline of the historical development of the observations and theories pertaining to the Southern Oscillation, with an emphasis on the evolution in thought on this fascinating subject.


The earliest work on the Southern Oscillation (SO) was detailed in a presidential address to the Royal Meteorological Society by G.T. Walker in 1928 (contained in Walker, 1928). The rudiments of the SO were first suggested in 1897 by H.H. Hildebrandsson who took ten years of pressure data from a world-wide network of 68 stations and noted certain relationships between the interannual trends of some to the plotted time series. Again using purely graphical techniques, Norman and W.J.S. Lockyer in 1902 confirmed Hildebrandsson's discovery of an apparent "seesaw" in pressure between South America and the Indonesian region. Hildebrandsson felt that these interannual variations, affecting such a large portion of the globe, could not be caused by anything of tropical or even temperate origin, as he believed these regions were dominated by "circumstances" of too regular a nature. Instead he felt that the variations must be explainable by changes in the ice conditions of the polar seas.


In the 1920's and 30's, Sir Gilbert Walker and his collaborator, E.W. Bliss, were interested in the development of a seasonal forecast scheme for the prediction of the strength of the Indian monsoon. Walker's efforts, which have been thoroughly summarized by Montgomery (1940a), were largely based on the compilation of voluminous tables of contemporaneous and serial (lag) correlation coefficients. Walker obtained pressure, temperature, and rainfall records, usually of around 40 years duration, from a much more extensive station network than had been available to Hildebrandsson. Each record was averaged by season within each year and then grouped by season to form four time series for each variable at each station. Walker also included time series of such miscellaneous variables as river flood stages, mountain snowpack depths, lake levels, and sunspot activity in an effort to detect cause and effect relationships between these and the standard meteorological parameters. In an examination of any coefficient that was "greater than the probable largest" he noticed that in large regions the seasonal parameters tended to behave uniformly and that there were pairs of regions in which the parameters tended to act in unison but in an opposite sense to each other. Furthermore, these characteristics were found in each season around the same "centers of action" with only minor variation in their position or in the shape of the areas under their influence. Since a major portion of the Southern Hemisphere was affected by the phenomenon (and there was already a North Pacific Oscillation and a North Atlantic Oscillation), Walker decided to call it the "Southern Oscillation" and described it as "the tendency of pressure at stations in the Pacific ... to increase, while pressure in the region of the Indian Ocean decreases."


To explain as much as possible of the variance of the meteorological parameters in the regions under the influence of the SO, Walker devised a complicated set of equations using weighted station data to generate what later become known as the Southern Oscillation Index (SOI). The appropriate equation for the SOI varies from season to season, but each formula utilizes some data, like Nile River flood or South American rainfall (incorporating some 21 different stations), which make it difficult at best to keep Walker's version of the index up-to-date. Walker noted the high lag correlation (+.83) between the June-August (J-A) SOI and the following December-February (D-F) SOI, which led him to believe that the index would be a useful tool in long-range forecasting. Many of Walker's forecast formulae for the Indian monsoon incorporate a basic part of the SO in terms of the seasonally averaged pressure for a collection of South American stations. Verification of the formulae by Montgomery (1940b) in subsequent years revealed that only the South American pressure term consistently maintained its original correlation with the strength of the monsoon as measured by Indian rainfall.


While dismissing any direct influence of solar activity on the phase of the SO, Walker, like Hildebrandsson, did take quite an interest in a possible connection with the polar circulation. Since pressure tendencies over South America usually preceded any changes in pressure or temperature elsewhere, in the interannual sense, Walker reasoned that the root cause of the variations in the SOI must originate in or near that continent. He hypothesized that variations in the amount of ice thrown off of the antarctic coastline by its northward extension at 60 W must depend on the strength of the high-latitude easterlies. Any resultant changes in the sea-ice content of the South Atlantic would change the sea and air temperatures in the vicinity of Cape Horn and this in turn could affect the pressure over a wide area near South America. In this manner low winter pressures near the cape could produce high summer pressures in the same area with a subsequent shift of weather patterns northward and westward. Further correlation studies by Walker did not tend to support this mechanism and he eventually and reluctantly abandoned it. He continued to believe, however, that a linkage between the SO and the polar circulation must exist.


Concurrent with some of Walker's early work on the SO, Brooks and Braby (1921) investigated an important aspect of the phenomenon in great detail. Working with island station data from the equatorial Pacific, they found a number of consistent relationships between rainfall, wind direction, air temperature and humidity. Periods of extensive anomalous rainfall were clearly associated with a shift, cessation, or even reversal of the normal trade wind regime at the stations in the normally arid zone east of the dateline. Brooks and Braby surmised that the pressure distribution along the equatorial trough was displaced zonally during the anomalous periods, but they were unable to sort out the underlying cause for the shift. Some of the observed westerlies, they felt, could be caused by the formation of large eddies along the "equatorial front" where the trades from one hemisphere are lifted over the trades from the other hemisphere due to relatively minor differences in their densities. In any case the inter-annual nature of the anomalous periods and the widespread coherent response of the atmosphere clearly related these feature to the work of Walker, but this was not realized until 1933 when J.B. Leighly saw the connection (Julian and Chervin, 1978).


Little further work was done on the SO until the early 1950's when Willet and Bodurtha (1952) suggested an abbreviated version of Walker's formulae for the J-A and D-F SOI's. Their formulae retained only the terms that were formed from easily obtainable data and yet the new form of the SOI maintained a high correlation (+.94) with the original version. They emphasized the potential usefulness of the SOI in long-range forecasting and as an example they pointed out the remarkably high correlation (-.57) between the J-A SOI and the following D-F average temperature over southwestern Canada.


Schell (1956) reviewed the known characteristics of the SO and sought to offer a physical explanation as to their cause. He surmised that the strength of the South Pacific anticyclone was maintained by the advection of cold water by the northward flowing current along the west coast of South America. Schell conjectured that the "oscillatory" nature of the SO could be the result of a type of cyclic feedback process: a strengthening of the South Pacific High would drive more cold water and air north and then west along the equator, eventually leading to colder temperatures and a weakening of the low pressure in the Indian Ocean. The diminished surface low would in turn provide less outflow at high levels back to the South Pacific, thus weakening the High, diminishing the advection of cold air and water northward, and so on. Schell proposed two mechanisms involving variations in solar activity which might trigger the sequence. 1) A perturbation of the Antarctic circulation, somehow caused by an infusion of corpuscular radiation over the pole, might lead to a variation in the amount of cold water maintaining the South Pacific High. 2) A preferential expansion of the airmass over the tropical Indian Ocean due to an increase in ultraviolet radiation could increase the high level flow of air into the South Pacific High. Schell believed that the observed correlation between sunspot activity and temperature over the tropical Indian Ocean was credible evidence in support of the second mechanism. The lack of an eleven year cycle for the SO was explained by the existence of some breakdown threshold inherent in the nature of the general circulation that would lead to a quasi-periodic readjustment to the slowly varying external forcing provided by the sun. Why the region around the Indian Ocean should be especially susceptible to solar forcing was not dealt with by Schell.


Ichiye and Petersen (1963) updated some of the work of Brooks and Braby and found much the same conditions in the equatorial Pacific. They did note that the anomalous rainfall of the dry zone occurred primarily as the result of very heavy rains rather than periods of prolonged rain. They proposed that a decrease in the strength and thickness of the tradewinds along the equator would diminish the amount of cold water advected from the east. A subsequent warming of the surface and the onset of convective activity would then completely disrupt the trades. The upper level westerlies would then move down to the surface and stop the upwelling caused by the Ekman drift. Thus the surface heating would be further enhanced and the westerlies would act to perpetuate themselves.


The El Nino was first directly related to variations in the strength of the South Pacific anticyclone and therefore to the SO by Schell (1965). He took southern hemisphere pressure analyses for the period 1951-57 and composited them into four groups by classifying the charts by time of year (March-November and December-February) and by years of below or above average sea surface temperature (SST) along the west coast of South America. He estimated the strength of the westerlies in the far southeast Pacific and the southerlies and southeasterlies along the coast of Chile by taking pressure differences between selected map grid points. Schell found that warm coastal water during the El Nino season (December-February) was preceded in the composite period of March-November by a weakening and increased convergence of the westerlies in the region 35-50 S and 90-135 W and weakened southerlies and southeasterlies along and inland of the coast. Offshore southeasterlies exhibited no clear change in intensity. Thus it was concluded that El Nino was caused by decreased advection of cold water due to the weakening of the westerlies and by decreased upwelling along the coast due to the slackened coastal southeasterlies. If the composite was limited to March-May, the same pattern was present as in the March-November composite. This led Schell to believe that it should be possible to predict the occurrence of an El Nino.


Many of Walker's observations were reconfirmed by Troup (1965). However, Troup found that many of the original relationships had weakened significantly, mainly due to a reduction in the persistence and variability in the pressure data since 1920. He used regression techniques to generate pressure anomaly maps corresponding to a value of the SOI one standard deviation away from its mean. Troup postulated that variations in the SO originate as perturbations in the tradewinds of the South Pacific. He thought of the tradewinds as an easterly jet the strength of which is controlled by the two subtropical highs. When the trades are strong, the formation of vortices is displaced to the west and the number and/or intensity of the vortices is increased. These conditions lead to enhanced upward motion and subsequent latent heat release in the western Pacific. The resulting intensification of the longitudinal temperature gradient alters the upper wind pattern and so on. The availability of surface heating along the equator east of the dateline is also largely controlled by the southeast trades by the Ekman-drift mechanism of equatorial upwelling. Therefore, a weakening of the southeast trades would result in a shift in rainfall toward the east along the equator and a decrease in the longitudinal temperature gradient throughout the tropical troposphere across the Pacific.


In a discussion based on data collected by many investigators during the 1957-58 El Nino and equatorial warming, Bjerknes (1966) suggested a theory of air-sea interaction which would have a large scale impact on the general circulation. He hypothesized that anomalously warm equatorial SST in the Pacific would act as a heat source to cause a spin-up of the Pacific sector of the Hadley circulation, with a subsequent increase in the transport of angular momentum into the subtropical jet and a strengthening of the mid-latitude westerlies. He assumed that the comparatively cool waters that normally lie along the equator east of the dateline are largely maintained by upwelling. The upwelling is caused by the surface divergence due to the Ekman-drift driven by the surface easterlies. The easterlies also maintain a sea-level difference across the Pacific basin. Cessation of the equatorial easterlies would stop or reverse the surface upwelling and encourage the flow of warm water from the western to the eastern Pacific. In support of his theories Bjerknes noted evidence that the average pressure distribution from the 1957-58 equatorial SST warming was associated with the southward displacement of the subtropical high in the North Pacific. Because the equatorial easterlies are associated with the southern Hadley cell, the cessation of winds is blamed on a circulation that has weakened more than usual for the southern hemisphere summer.


Doberitz (1968) performed a thorough analysis of rainfall (35 stations) and SST in the Pacific equatorial dry belt using cross spectrum techniques. He found much the same coherent pattern as Brooks and Braby did in 1921. Doberitz determined that the anomalous rainfall in the dry belt was largely due to a southward expansion of the Intertropical Convergence Zone east of the dateline. He found that the expansion generally occurs with a time scale of two to three months. Some evidence of a similar northward expansion of the South Pacific Convergence Zone was also present. Because the rainfall anomalies tended to precede increases in the SST over the equatorial region, Doberitz concluded that the atmosphere must play the dominant role in the phenomenon.


In an elaboration on his previous work, Bjerknes (1969) used the observations taken during the 1963-64 and 1965-66 equatorial warmings to lend additional support to this hypotheses. He found that Canton Island, which is in the normally arid equatorial zone, has rainfall excesses which appear to be closely correlated to a warming of the surrounding SST to a value greater than the ambient air temperature. Since Canton Island conditions are likely to apply equally well to a large part of the equatorial dry zone, the increase in surface latent and sensible heat fluxes from ocean to atmosphere, coupled with the abnormal rainfall, produce a shift in the heat source for the northern Hadley circulation to a position further south than is usual during the northern winter. From pressure analyses it was found that a strengthening of the northeast trades and the mid-latitude westerlies occurred at the same time that the equatorial easterlies were very weak (January 1964 and January 1966). Bjerknes also described a zonal circulation pattern extending across the equatorial Pacific which he called the Walker Circulation (WC). The WC is of thermal origin with an ascending branch over the warm waters of the western Pacific near 165 E and a descending branch over the cold waters of the eastern Pacific at 90 W. When the longitudinal temperature gradient and the related pressure gradient are disrupted by the cessation of equatorial upwelling the axis of the WC shifts eastward allowing the ascending branch and its associated cloudiness and rain to occur over normally suppressed regions. Thus the WC is a part of the overall mechanism of the SO. Changes in the intensity and/or longitudinal displacement of the WC directly reflect the distribution of pressure, temperature, and rainfall over a vast area of the Pacific-Indonesian region.


In his final work on the subject, Bjerknes (1972) presented 200 mb height analyses for November 1964 and 1965 which indicate an eastward enlargement of the upper level Indonesian equatorial high during the period of warm SST anomalies of 1965-66. This is in general agreement with the concept of an eastward shift in the ascending branch of the WC. The analyses also indicated a general warming of the entire tropical troposphere. A comparison of winds and SST at Canton Island for the period 1962-67 reinforce the correlation between the slackening of the equatorial easterlies and the warming of the surface waters. The deepening of the Aleutian Low and a weakening of the Icelandic Low and other high-latitude effects of the equatorial warming were also pointed out by Bjerknes. A picture of the SO as a world-wide phenomenon of great importance to long-range weather prediction was now clearly taking shape.


The complex ocean-atmosphere interactions associated with the SO by now obviously required an intimate knowledge of the dynamics of the ocean basins. After Bjerknes, the role of the oceanographer in deciphering the puzzle of the SO was to become increasingly important. The contributions made by the investigators reviewed above provided the foundation for a profusion of research that came forth from where Bjerknes left off.


-Donald R. Mock, Seattle, May 1981
(minor changes, 12/97 -DRM)


  • Bjerknes, J., 1966: A possible response of the atmospheric Hadley circulation to equatorial anomalies of ocean temperature. Tellus, 18, 820-829.
  • Bjerknes, J., 1969: Atmospheric teleconnections from the equatorial Pacific. Monthly Weather Review, 97, 163-172.
  • Bjerknes, J., 1972: Large-scale atmospheric response to the 1964-65 Pacific equatorial warming. Journal of Physical Oceanography, 2, 212-217.
  • Brooks, C.E.P., and H.W. Braby, 1921: The clash of the trades in the Pacific. Quarterly Journal of the Royal Meteorological Society, 47, 1-13.
  • Doberitz, R., 1968: Cross spectrum analysis of rainfall and sea temperature at the equatorial Pacific Ocean. Bonner Meteorologische Abhandlungen, 8, 1-61.
  • Ichiye, T., and J.R. Petersen, 1963: The anomalous rainfall of the 1957-58 winter in the equatorial central Pacific arid area. Journal of the Meteorological Society of Japan, 172-182.
  • Julian, P.R., and R.M. Chervin, 1978: A study of the southern oscillation and Walker circulation phenomenon. Monthly Weather Review, 106, 1433-1451.
  • Montgomery, R.B., 1940a: Report on the work of G.T. Walker. Monthly Weather Review, Supplement no. 39, 1-22.
  • Montgomery, R.B., 1940b: Verification of three of Walker's seasonal forecasting formulae for India monsoon rain. Monthly Weather Review, Supplement no. 39, 23-24.
  • Schell, I.I., 1956: On the nature of the southern oscillation. Journal of Meteorology, 13, 592-596.
  • Schell, I.I., 1965: The origin and possible prediction of the fluctuations in the Peru Current and upwelling. Journal of Geophysical Research, 70, 5529-5540.
  • Troup, A.J., 1965: The "southern oscillation." Quarterly Journal of the Royal Meteorological Society, 102, 490-506.
  • Walker, G.T., 1928: World Weather. Monthly Weather Review, 56, 167-170.
  • Willet, H.C., and F.T. Bodurtha, Jr., 1952: An abbreviated southern oscillation. Bulletin of the American Meteorological Society, 33, 429-430.

Mock, D.R., 1981: The Southern Oscillation: Historical Origins. Unpublished term paper, University of Washington, Seattle, WA, 9 pp.