21. M. A. Cane (J. Phys. Oceanogr., in press)questions the adequacy of the linear theory.

22. The ocean circulation can be represented as thesum of an external (barotropic) mode, in whichthe entire vertical column responds as a unit,and an infinite number of internal (baroclinic)modes. Only the baroclinic response is impor-tant at the long time scales characteristic of ElNifio.

23. R. A. Knox and D. Halpern, J. Mar. Res. 40(Suppl.), 329 (1982); C. C. Eriksen et al., J.Phys. Oceanogr., in press.

24. D. Enfield, J. Geophys. Res. 86, 2005 (1981).25. Wind fields are composites from E. M. Rasmus-

son and T. H. Carpenter (15).26. C. Ramage and A. M. Hori, Mon. Weather Rev.

110, 587 (1982).27. A. Leetmaa, J. Phys. Oceanogr. 13, 467 (1983).28. The currents slow in response to a weakening of

the southeast trade winds.

29. P. Schopf, J. Phys. Oceanogr., in press.30. J. Sadler and B. Kilonsky, Trop. Ocean-Atmos.

Newsl. 16, 3 (1983).31. G. Meyers and J. R. Donguy, ibid., p. 8.32. Available climatologies differ by more than 1°C

in this area due to different averaging periods,data bases, and analysis techniques.

33. K. Wyrtki, Trop. Ocean-Atmos. Newsl. 16, 6(1983).

34. G. Meyers, private communication.35. E. Firing, R. Lukas, J. Sadler, K. Wyrtki,

Science 222, 1121 (1983).36. The Equatorial Undercurrent is a permanent

(with the exception noted) feature in the Atlanticand Pacific. It is a subsurface, eastward-movingcurrent at the equator with speeds often inexcess of 1 n/sec.

37. In May 1983 a westward jet was also found atnormal undercurrent depth at 95°W (S. Hayes,private communication).

38. R. L. Smith, Science 221, 1397 (1983).39. Spec. ClGm. Diagn. Bull. (15 July 1983) (avail-

able from Climate Analysis Center, NOAA,Washington, D.C. 20233).

40. The significance of these excursions was firstnoted by K. Wyrtki [Mar. Technol. Soc. J. 16, 3(1982)].

41. Courtesy of R. Reynolds.42. Island station data are courtesy of K. Wyrtki;

Callao data are from D. Enfield and S. P. Hayes[Trop. Ocean-Atmos. Newsl. 21, 13 (1983)].

43. A. Leetma, D. Behringer, J. Toole, R. Smith,ibid., p. 11.

44. I thank the many colleagues who reviewed anearly version of the manuscript. Special thanksare extended to those who generously contribut-ed unpublished data and to Lenny Martin forassistance in preparing the manuscript. Thiswork was supported by grant OCE-8214771from the National Science Foundation.

and named by Sir Gilbert Walker more

than a half-century ago (1). The primarymanifestation of the Southern Oscillationis a seesaw in atmospheric pressure atsea level between the southeast Pacific

Summary. The single most prominent signal in year-to-year climate variability is theSouthern Oscillation, which is associated with fluctuations in atmospheric pressure atsea level in the tropics, monsoon rainfall, and wintertime circulation over NorthAmerica and other parts of the extratropics. Although meteorologists have knownabout the Southern Oscillation for more than a half-century, its relation to the oceanicEl Niino phenomenon was not recognized until the late 1960's, and a theoreticalunderstanding of these relations has begun to emerge only during the past few years.The past 18 months have been characterized by what is probably the mostpronounced and certainly the best-documented El Nin'o/Southern Oscillation episodeof the past century. In this review meteorological aspects of the time history of the1982-1983 episode are described and compared with a composite based on sixprevious events between 1950 and 1975, and the impact of these new observationson theoretical interpretations of the event is discussed.

characterized by large, remarkably co-

herent climate anomalies over much ofthe globe. The pattern inherent in theseanomalies has been recognized gradual-ly, over a period of decades, as a resultof the collection and analysis of manydifferent climatic records; the recogni-tion process has been somewhat like theassembly of a global-scale jigsaw puzzle.Some of the pieces of this puzzle are

implicit in the Southern Oscillation, a

coherent pattern of pressure, tempera-ture, and rainfall fluctuations discovered

16 DECEMBER 1983

subtropical high and the region of lowpressure stretching across the IndianOcean from Africa to northern Australia.Other manifestations involve surfacetemperatures throughout the tropics andmonsoon rainfall in southern Africa, In-dia, Indonesia, and northern Australia(2, 3). When Walker's scientific contem-poraries expressed doubts concerningthese statistical relations because of thelack of a physically plausible mechanismfor linking climate anomalies in far-flungregions of the globe, he replied, "I think

the relationships of world weather are socomplex that our only chance of explain-ing them is to accumulate the facts em-pirically .. . there is a strong presump-tion that when we have data of thepressure and temperature at 10 and 20km, we shall find a number of newrelations that are of vital importance"(4).

Descriptive studies of the 1957-1958El Nifio event, based in part on routinemerchant ship data from the tropicalPacific, were instrumental in revealingthe link between El Ninlo and the South-ern Oscillation. The large-scale interac-tion between atmosphere and ocean wasconfirmed by retrospective statisticalstudies of past episodes (5). The emerg-ing unified view of the El Nifio/SouthernOscillation (ENSO) phenomenon is ex-emplified by Bjerknes's investigations(6) of the 1957-1958, 1963-1964, and1965-1966 ENSO episodes. These stud-ies were among the first in which satelliteimagery was used to define the region ofanomalously heavy rainfall over the dryzone of the equatorial central and east-ern Pacific during episodes of warm sea-surface temperature (SST), an aspect ofthe phenomenon that Walker apparentlywas unaware of. Bjerknes showed thatthese fluctuations in SST and rainfall areassociated with large-scale variations inthe equatorial trade wind systems, whichin turn reflect the major variations of theSouthern Oscillation pressure pattern.The linking of El Ninlo with the South-

ern Oscillation was viewed as evidencethat ocean circulation plays the role of aflywheel in the climate system and isresponsible for the extraordinary persist-ence of the atmospheric anomalies frommonth to month and sometimes even

Eugene M. Rasmusson is chief, DiagnosticsBranch, Climate Analysis Center, National Meteo-rological Center/National Weather Service, Wash-ington, D.C. 20233. John M. Wallace is a professorin the Department of Atmospheric Sciences anddirector of the Joint Institute for the Study of theAtmosphere and Ocean, University of Washington,Seattle 98195.

1195

Meteorological Aspects of theEl Ninfo/Southern Oscillation

Eugene M. Rasmusson and John M. Wallace

Each year various parts of the globeexperience regional climate anomaliessuch as droughts, record cold winters,and unusual numbers of storms. Butsome years, such as 1982 and 1983, are

on

Janu

ary

5, 2

010

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

from season to season. The host of stud-ies in the years since Bjerknes's workprovide a much more complete descrip-tion of the ENSO phenomenon, whichemerges as the dominant global climatesignal Qh time scales of a few months to afew years (7-10).Somd 6f the relations identified and

documented in these studies are implicitin the time series in Fig. 1. The verticallines correspond to ENSO episodes asdefined by equatorial SST, rainfall atisland stations in the equatorial centralPacific, and sea-level pressure at Dar-win, Aust,lia (Fig. 1, A to C). Note thealmost peifect correspondence amongthese three series, whose mutual correla-tion coefficients all exceed 0.8. The spac-ing of the individual episodes rangesfrom 2 years to about a decade. In a fewcases, such as 1976 to 1978, a singleepisode spans several consecutive years.The negative correlation between seriesC and D in Fig. 1 illustrates the east-westseesaw in sea-level pressure noted byWalker and earlier investigators (11). Se-lected rblations involving rainfall in otherparts of the tropics and subtropics areindicated by series E, F, and G, and

Fig. 1.phenonwhereEquatorainfallpressur(170S, IMay; (1NovemnorthwiMexiccepisodeavailab

relations involving regional climateanomalies at extratropical latitudes areillustrated by the remaining time series.The climate variables indicated by aster-isks were considered by Walker to beinvolved in the Southern Oscillation;hence the segments of those time seriesfrom the late 1920's on may be regardedas confirmation of his findings.The time series in Fig. 1 do not resolve

the spatial structure and time history ofindividual ENSO episodes and theirstrong seasonal dependence. We will re-turn to that subject presently, but firstwe will describe a conceptual frameworkfor interpreting some of the atmosphericphenomena associated with these epi-sodes.

A Conceptual Framework

A meteorologist may think of theSouthern Oscillation as the atmosphericresponse to a prescribed SST anomaly,just as an oceanographer may treat ElNihlo as a response to a prescribed windstress at the sea surface. Both approach-es are limited but nevertheless valid in

the sense that they lead to well-posed,relatively tractable scientific questionswhose answers are prerequisites for ef-fectively addressing the much more diffi-cult problem of coupled atmosphere-ocean interactions in the time-dependentclimate system. Throughout this sectionwe will adopt the parochial perspectiveof the meteorologist.The mechanisms through which SST

anomalies in the equatorial Pacific mightproduce the worldwide climate anoma-lies described in Fig. 1 have been asubject of intense debate in the recentmeteorological literature. Althoughmany aspects of the workings of theglobal climate system have yet to beresolved, the broad outlines of a theoryexplaining the linkages between equato-rial SST, tropical precipitation, globalcirculation patterns at the jet stream lev-el (10 km), and climate anomalies at theearth's surface are beginning to emerge.On time scales of weeks and longer

and on space scales of more than 1000km, the primary way in which the tropi-cal atmosphere responds to temperaturecontrasts at the earth's surface isthrough thermally direct circulations. Insuch circulations the warmest regions atthe earth's surface are characterized byascent of moisture-laden air from theplanetary boundary layer that condensesto form widespread cloudiness and pre-cipitation. Elsewhere there is subsidenceof dry air from the upper troposphere,forming a lid on the planetary boundarylayer and preventing small trade-wind

\--j 1_4 v v -NJ, V \JVI \vjF 'V \j V\ cumulus clouds from growing to a sizethat can produce substantial rainfall. Ex-

[ [ ] | | [ i | 4 D 9 >T / \ amples of such thermally direct circula-PacificI s~iblrdpical ~aihfall* tions are the monsoons, which bring

Itv_~\ / \ 9 br icl |heavy rainfall to the subtropical conti-l l l l l l l l l lIndia r*i,1fall* nents during summertime, when they are

warmer than the surrounding ocean.Over the equatorial Pacific there is

ll lll l ll,Souhatern Afric tmpratre usually a strong SST gradient, with rela-tively cool waters in the east and warmer

South*eMtern Canailai temper4ture (by 3 to 6 K) waters in the west towardNew Guinea. This east-west gradient is

/:X 41/ W iA tX SZ @ uassociated with a thermally direct circu-Gul\ O lt.io raina H i sinkingmolationin the equatorial plane, with (i)

sinking motion in the eastern Pacific, (ii)I westward low-level flow (the equatorial

1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 extension of the trade winds), (iii) risingTime series of selected variables whose interannual variability is related to the ENSO motion and deep cumulus convectionnenon. Each data point represents an average over I year or a season, as indicated, over the extreme western Pacific andthe years on the horizontal axis refer to the beginning of the averaging periods. (A) Indonesia, and (iv) eastward return flow)rial eastern Pacific (South American coast to the date line) SST, April to March; (B) at the cirrus cloud level (10 to 15 km).index fot central equatorial Pacific island stations, April to March; (C) sea-level

re at Darwin, Australia (12°S, 131°E), April to March; (D) sea-level pressure at Tahiti Bjerknes (12) referred to this circulation150°W), April to March; (E) rainfall index for the subtropical North Pacific, November to cell as the Walker circulation in memoryF) rainfall index for India, June to September; (G) rainfall index for southeastern Africa, of Sir Gilbert Walker. Under normaliber to May; (H) temperature index for stations in southwestern Canada and the conditions rainfall at equatorial stations,estern United States, November to April; and (I) rainfall index for stations in northern cnditions, rfat ate leq ia stuatio) and the U.S. Gulf Coast, November to February. Vertical lines indicate ENSO nearoreastofthedatelineisvirtuallynas as inferred from the top three series. Further details and quantitative values are because of the large-scale subsidencele from E.M.R. associated with the Walker circulation.

1196 SCIENCE, VOL. 222

on

Janu

ary

5, 2

010

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

During ENSO episodes the equatorialwaters in the eastern half of the Pacificare warmer than normal while SST'swest of the date line are near or slightlybelow normal, so that the east-west tem-perature gradient is diminished and wa-ters near the date line may be as warm asthose anywhere to the west (- 29°C).The region of heavy rainfall shifts east-ward so that Indonesia and adjacent re-gions experience drought while the des-ert islands in the equatorial central Pacif-ic experience month after month of tor-rential rainfall. Near and to the west ofthe date line the usual easterly surfacewinds along the equator weaken or shiftto westerly (which has implications forthe ocean dynamics), while anomalouslystrong easterlies are observed at the cir-rus cloud level.The effects of the equatorial SST

anomalies would be confined to theequatorial belt were it not for the earth'srotation, which makes it possible forchanges in the tropical circulation toexcite a large-scale quasi-stationarywave pattern, best defined in the uppertroposphere, that can give rise to sub-stantial anomalies in the extratropicalcirculation. The pattern of upper tropo-spheric circulation anomalies associatedwith ENSO episodes during the North-ern Hemisphere winter is illustrated inFig. 2A, which is based on a compositeof past events (8). Note the anticycloniccirculation anomalies straddling the re-gion of enhanced rainfall in the equatori-al Pacific, with one center near Hawaiiand the other near 20°S. These anticy-clonic gyres can be understood as adirect response to the upper-level diver-gence from the region of heavy rainfall,which, in the presence of the earth'srotation, leads to the production of anti-cyclonic vorticity along its polewardflanks. Poleward and eastward of thecenter near Hawaii in Fig. 2A is anapparent wave train with a negative cen-ter over the North Pacific, a positivecenter over western Canada, and anoth-er negative center over the southeasternUnited States. The "centers of action"associated with this wave train lie alonga broad "great circle" arc. The disper-sion of these so-called Rossby waves isnow widely believed to be the dynamicmechanism responsible for the strongtemporal correlations between tropicaland extratropical climate anomalies suchas those shown in Fig. 1. Although Walk-er postulated the existence of such a

problem was fully recognized (8, 14).The dynamics of Rossby wave trains

emanating from regions of enhancedtropical rainfall have been intensivelyinvestigated during the past few yearswith a variety of simple dynamic models.Among the insights derived are the fol-lowing:

1) The magnitude of the extratropicalresponse appears to be sensitive to thebackground flow on which the equatorialrainfall anomaly is superimposed. In par-ticular, the extratropical response in themodels is enhanced by the presence ofthe climatological mean stationary wavepattern forced by the major mountainranges and land-sea heating contrasts(15).

2) In the Northern Hemisphere duringwintertime, when the stationary wavesare very strong, the response in themodels takes the form of a geographical-ly fixed normal mode, which resemblesthe pattern in Fig. 2. That is, when theregion of enhanced rainfall is movedeastward or westward along the equator,the shape and geographical location ofthe simulated extratropical response re-mains as in Fig. 2, although the ampli-tude and perhaps even the polarity

Fig. 2. Schematic il-lustration of atmo-spheric conditionsduring winier in theNorthern Hemisphereafter the peak of atypical ENSO epi-sode, based on theunderstanding of thephenomenon beforethe 1982-1983 epi-sode. (A) Region ofenhanced precipita-tion (cloud outline)and circulation anom-alies at the jet stream(- 10-kmi) level. (B)SST and surface windanomalies (contourinterval, 0.5 K; zerocontour is heavier).The longest arrowscorrespond to windanomalies of - 1.5 m/sec (8, 9).

40°N

00mechanism long ago, the basic theoreti-cal concepts did not appear in the mete-orological literature until the late 1970's(13), and it was not until the early1980's that their relevance to the ENSO16 DECEMBER 1983

400S L900E

change. This normal-mode behavior isassociated with barotropic instability ofthe climatological mean flow pattern andmay be excited in many different ways:forcing by equatorial rainfall anomaliesis just one possibility (16).

3) Nonlinear interactions between thebackground flow and the circulationanomalies forced by the perturbations inequatorial rainfall may be important,particularly in the tropics. Hence theatmospheric response at a given timemay depend on the time history of theatmospheric circulation during previousmonths (17).

It is possible to model the atmosphericresponse to a prescribed SST anomalywithout referring to any of the dynamicconcepts mentioned above, but usinginstead state-of-the-art general circula-tion models (GCM's) (18), which arecapable of realistic simulations of globalclimate, including the normal annual cy-cle. A number ofGCM experiments havebeen conducted over the past 10 years atvarious institutions. Most of these ex-periments consist of a pair of simula-tions, one with a prescribed SST distri-bution based on climatological meanconditions and the other with a distribu-

North Pole

1800 90°W1197

on

Janu

ary

5, 2

010

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

tion based on typical conditions duringENSO episodes. The results of theseexperiments show a remarkably consist-ent pattern that is in general agreementwith observations. The models correctlysimulate the observed eastward shift ofthe region of heavy rainfall from theextreme western Pacific toward the cen-tral Pacific, the upper level anticyclonepair straddling the region of enhancedprecipitation, and most of the experi-ments based on wintertime conditions in

2 .ii , * .I...rA

SST-index area

82/o

0)

0-o SST-Puerto Chicama0e L

0

3 82/83

Composite/2~~~~~~\~

0

Jul. (-1 ) Jan. (O) Jul. (0) Jan. (1)

1198

the Northern Hemisphere show evi-dence of teleconnections to extratropicallatitudes that are reminiscent of the pat-tern in Fig. 2 (19), and consistent withthe relations in Fig. 1, H and I. The oneexperiment that has been conducted forsummertime conditions showed evi-dence of a suppression of the Indianmonsoon rainfall, in agreement with Fig.1 (20).Both GCM's and much simpler models

are capable of simulating the diminished

4

2

0 Fig. 3. (A) Monthlyrunning mean SST

d anomalies for the re-@ gion from 4°N to 4°S

and 150°W to 160°E.0 (B) Three-month run-0 ning mean SST anom-

8 alies for Puerto Chi-c cama (7.7°S, 79.3°W).

Continuous and dot-n ted lines refer to the

6 1982-1983 and 1940-I 1941 episodes, re-O spectively. For these0 two curves, year 0 ont the abscissa refers to

4 - 1982 and 1941, re-spectively. Thedashed curve refers to

E a composite of six0

2 ° warm episodes be-2tween 1950 and 1975

co (9). The ordinateco scales for the 1982-

1983 and 1940-1941o events (right) repre-

sent SST anomaliestwice as large as thoserepresented by thecorresponding scale

2 (left) for the compos-ite event. (C) Sea-lev-el pressure anomaliesat Tahiti (170S,150°W) and Darwin(12°S, 131°E) duringthe 1982-1983 epi-

2 sode. The difference2< between the two pres-E sures (Tahiti minus> Darwin), indicated by

shading, is often usedas an index of the

X Southern Oscillation.g

X,

Jul. (1)

easterly flow along the equator near andjust west of the region of enhanced rain-fall (Fig. 2B) and the related changes insea-level pressure (Fig. 1, C and D)observed during episodes ofwarm SST'sin the equatorial Pacific. It is these samechanges in surface wind over the tropicalPacific that physical oceanographers usein modeling the time variability of equa-torial SST's. Hence it appears that theatmospheric models contain many of thebasic ingredients required to simulate theatmosphere-ocean system. However,thus far the atmospheric modeling ex-periments have not addressed the time-dependent aspects of the ENSO problemin a way that takes into account the largeobserved mean annual cycle in SST's inthe eastern equatorial Pacific and theevolution of SST anomalies relative tothat seasonally varying climatologicalmean.The 1982-1983 ENSO episode provid-

ed scientists with an opportunity to testthis conceptual framework on the basisof a new, independent set of data.

Delayed Onset of the 1982-1983 Episode

Scientists were caught off guard by theunusual timing of the onset of the 1982-1983 episode relative to the climatologi-cal mean annual cycle. A composite de-scription, based on the six major eventsthat occurred between 1950 and 1975 (9),is dominated by the positive SST anoma-lies, approaching 2 K, appearing firstalong the South American coast early inthe calendar year and spreading west-ward to cover the entire equatorial re-gion east of 170°E by July. Most of theseevents were preceded by a period ofabnormally strong easterly surface windsalong the equator to west of the date line.The winds weakened abruptly aroundNovember, coincident with a swing inthe Southern Oscillation associated withthe partial collapse of the surface anticy-clone in the southeast subtropical Pacif-ic. The resulting change in surface windstress along the equator was believed tobe responsible for the onset of warmSST's (21). These precursors in the sur-face wind and sea-level pressure fieldswere not observed during 1981 or early1982. Furthermore, it is evident fromFig. 3B that there was no dramaticwarming of surface waters along theSouth American coast during early 1982;SST at Puerto Chicama on the coast ofPeru did not begin to rise sharply untilSeptember and October 1982. Similartiming with respect to the climatologicalmean annual cycle was observed in the

SCIENCE, VOL. 222

on

Janu

ary

5, 2

010

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

1940-1941 event, which, by some mea-

sures at least, was the strongest event onrecord before the recent one (22).

Farther to the west, the contrast be-tween the timing of the composite EN-SO warming and the 1982-1983 episodeis much less striking. In most of the sixevents in the composite, small but possi-bly significant increases in SST were

observed along the equator in the vicini-ty of the date line as early as Decemberor January. In the 1982-1983 episode,mean equatorial SST between 150°W and160°E (Fig. 3A) began rising at the usualtime and, apart from the curious dipduring July, followed a course very simi-lar to that in the composite.Although the initial anomalies were

generally less than 1 K, they were largeenough to shift the location of the warm-est surface waters eastward by as muchas 300. Whether these subtle changes inSST in the central Pacific were instru-mental in setting the stage for the more

dramatic atmospheric changes that tookplace a few months later is unknown.The results shown in Fig. 3 deserve

comment in relation to the widely publi-cized suggestion that the eruption of ElChich6n played an essential role in trig-gering the 1982-1983 ENSO episode(23). The evidence put forth in support ofthis suggestion-that the recent episodewas unprecedented in its strength andtiming relative to the annual cycle andthat its onset followed the volcanic erup-tion closely-is largely circumstantial. Inour view, the apparent similarities be-tween the 1940 and 1982 episodes withrespect to intensity and timing, togetherwith the evidence of a rising trend in SSTanomalies over much of the equatorialPacific before the eruption, cast doubton these arguments.

A Swing in the Southern Oscillation

In the composite ENSO episode thedecline in sea-level pressure (relative tothe climatological mean annual cycle)that begins in the southeast subtropicalPacific around November continuesthrough the following August, whilepressure in the broad region that formsthe western end of the seesaw rises.Meteorologists refer to these pressurechanges as a swing of the Southern Os-cillation. In the composite event the sea-

level pressure difference, Darwin minusTahiti, which is often used as an index ofthe oscillation, drops from about I mbarabove normal in November at the onsetof the event, to near normal in January,to almost 2 mbar below normal in Au-

16 DECEMBER 1983

Fig. 4 (A to C) Anom-alies in satellite-sensed outgoing long-wave radiation (OLR)(contours) and windat 850 mbar (arrows)for three seasons dur-ing the 1982-1983 epi-sode. Negative anom-alies in OLR, indicat-ed by the solid con-tours and labeled Wfor "wet," corre-spond to regions ofenhanced precipita-tion, and vice versa(D, "dry"). Contourinterval, 20 W/m2(where contours cor-respond to ± 10, ±

30, and so forth). Thelongest arrows corre-spond to wind anoma-lies on the order of 10m/sec.

20°N

00

200S

20°N

00

20°S

20°N

200S

900E

gust. Toward the end of this period thereis usually a more pronounced weakeningor even a reversal of the surface easter-lies along the equator in the vicinity ofthe date line, accompanied by a furthereastward shift of the rain area from thewestern Pacific to near the date line.The 1982-1983 episode was character-

ized by an unusually abrupt swing of theSouthern Oscillation, which began latebut quickly overtook its counterpart inthe composite. Between April and Au-gust 1982 the sea-level pressure differ-ence between Tahiti and Darwindropped from near normal to about 3mbar below normal (Fig. 3C). In June theeasterly surface winds along the equatorto the west of the date line shiftedabruptly to fitful westerlies (24). Thiswind shift marked the onset of an ex-

tended period of extremely heavy pre-cipitation in the equatorial central Pacif-ic. Farther to the west, Indonesia, east-ern Australia, Melanesia, Southern In-dia, Sri Lanka, and southern Africa wereexperiencing the beginning of a severe,and in some places record, drought, withattendant crop losses and the threat offamine (25, 26).Anomalies in wind and precipitation

for the period June to August 1982 are

illustrated in Fig. 4A. The arrows repre-sent wind anomalies at the 850-mbar(- 1.5-km) level and the contours repre-sent anomalies in outgoing infrared radi-ation sensed from satellites. Regions ofenhanced mid-altitude and high clouds,

1800 90°W

which normally accompany convectiveprecipitation in the tropics, show up as

negative anomalies because clouds in thecold middle and upper troposphere emitless radiation to space than the warmer

earth's surface. Note the strong corre-

spondence between the regions of en-

hanced precipitation, convergence, andwesterly wind anomalies in the equatori-al belt.

A Sharp Rise in SST

September and October 1982 were

marked by dramatic rises in SST and sea

level along the South American coastand throughout the eastern equatorialPacific. These changes followed, byabout 3 months, the abrupt shift in low-level winds west of the date line fromeasterly to westerly and the correspond-ing east-west seesaw in sea-level pres-sure. A similar phase lag is observed inthe composite ENSO episode, in whichthe events in question begin 6 monthsearlier. Hence the new observations sup-port the idea that the rising SST in theeastern Pacific is a response to abruptchanges in surface wind stress west ofthe date line.

Coincident with the rises in SST in theeastern Pacific, the intertropical conver-

gence zone, a narrow east-west band ofheavy precipitation that, at this time ofyear, normally lies near 10°N, shiftedsouthward toward the equator, and a

1199

Ao .- D - - .nJ.Aug. 1982

a.) -.::

6~~~~ ~ ~~~ Dec4--Fb-r-

to - --D-'''''---''L .X

* Fg . . - s ' ' ' ' ' v ' ' ' wWA-0. *'ss .8 \\ \ ,,....... v m \ >va

D~~~~~\''-

C cZ IIT4 NLxM y ,E&.,-..w -- > e.4F 8

e X\ oo st* NF KsWw .\sXXoS,

f . .. , \ \ sS-\ r; :: s_ cs..c! -t _ 4 =~~_ /Is S\ f

on

Janu

ary

5, 2

010

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

40°N

400N

400S40°N

40 0S

400N

00

400S - -

900E 1800

prolonged period of heavy precipitationbegan in Ecuador and northwesternPeru, several months in advance of thenormal wet season. Rainfall recordswere shattered month after month fromNovember through June. For example,Guayaquil, Ecuador, recorded a total of2636 mm in the 6-month interval fromNovemberAto April, compared with theprevious record of 1670 mm set duringthe 1972-1973 ENSO episode (27). Ironi-cally, while the desert regions of north-west Peru received record rainfall, majoragricultural areas in southern Peru andnorthern and central Bolivia were expe-

riencing serious drought (28).

Peak of the 1982-1983 Episode

December 1982 to February 1983 saw

further intensification and extension ofthe anomaly patterns of the previousmonths. Drought in southern Africa,southern India and Sri Lanka, and theAustralian-Indonesian region continuedand, as in past ENSO episodes, expand-ed to cover the boreal subtropical Pacificfrom the Philippines eastward throughthe Hawaiian Islands. The deluge contin-ued over Ecuador and northern Peru.

Figure 5 shows the observed SST pat-tern for December 1982 to February1983, the climatological mean pattern forthat season, and the corresponding SSTanomaly pattern, which is simply the

1200

Fig. 5. Sea-surfacetemperature patterns(contour interval, IK). (A) December1982 to February1983. (B) December-February climatolo-gy. The heaviestshading correspondsto SST's > 29°C. (C)Anomalies for De-cember 1982 to Feb-ruary 1983. The heav-iest shading denotesanomalies > 3 K.

90ow

difference between the two fields. TheSST anomaly pattern resembles the com-posite pattern based on previous ENSOepisodes for December to February (ap-proximately 1 year after the onset ofwarm SST's along the South Americancoast), except that the magnitude of thewarm anomalies is more than twice as

large as in the composite pattern (9). Theanomalies for 1982 to 1983 were so largethat it was nearly as warm along theSouth American coast as in the westernPacific, and the warmest SST's in theequatorial belt crossed the date line andextended considerably farther to the eastthan in the composite of past episodes.The Southern Oscillation index (sea levelpressure at Tahiti minus that at Darwin)took another sharp downward swingfrom 4 mbar below normal in November1982 to an unprecedented 6 mbar belownormal in February 1983 (Fig. 3C).The anomaly charts for wind at 850

mbar and outgoing infrared radiation forDecember 1982 to February 1983 are

shown in Fig. 4B. The drought areas

mentioned above show up as positiveanomalies in outgoing infrared radiation.As in Fig. 4A, there is a strong corre-

spondence between the region of en-

hanced cloudiness and precipitation(negative anomalies in outgoing infraredradiation) and the region of westerlywind anomalies along the equator. Com-pared to the June-August pattern shownin Fig. 4A, the region of enhanced pre-

cipitation lies farther to the east. Thisgradual eastward shift of the rain area,surface westerly wind anomalies, andmaximum SST is a striking characteristicof the time history of the 1982-1983episode. A comparison of Figs. 2 and 4shows that at this season the rain areaalong and south of the equator was shift-ed farther east than in the composite.With it, the region of tropical stormgenesis in the South Pacific shifted far tothe east and, as a result, French Polyne-sia, noted for its benign climate, wasbattered by five full-blown hurricanes(29, 30).

Figure 6 shows the circulation anoma-lies at the jet stream level (200 mbar)during December 1982 to February 1983.As in Fig. 2, the pattern is characterizedby anticyclonic gyres straddling theequator, with easterly wind anomaliescovering the region of enhanced rainfallover the equatorial central Pacific andwesterly wind anomalies at subtropicallatitudes of both hemispheres. Cycloniccirculation anomalies covered the NorthPacific from the ocean surface to thetropopause. In this region, mean sea-level pressures and upper level geopo-tential heights exhibited record low val-ues and numerous storms exhibited cen-tral pressures below 950 mbar (31). Thisobservation is in agreement with Fig. 2and with the results ofGCM simulations.The anticyclonic circulation anomaly

over south-central Canada is slightly far-ther east than its counterpart in Fig. 2.This feature and the southerly flow alongits western flank was associated with abelt of unusually mild wintertime tem-peratures (up to 6 K above normal) cen-tered along the border between the Unit-ed States and Canada from Lake Superi-or westward to the Pacific. This region ofabove-normal temperatures correspondsclosely to that observed during pastENSO episodes (Fig. IH).The most striking climate anomalies

observed over the United States duringthis period were those associated withwesterly wind anomalies extendingacross the subtropical Pacific, the south-ern border of the United States, and theGulf of Mexico (Fig. 6). Such windanomalies indicate a pronounced intensi-fication and southward displacement ofthe normal westerly jet stream thatsweeps across this region. Associatedwith this intensified jet stream were anunusual number of winter storms thatbattered the California coast with highwinds, brought flooding to parts of Cali-fornia and the Gulf states, and buriedmountain regions of the Southwest withheavy snow in late winter and spring(32). Abnormally wet winters in the Gulf

SCIENCE, VOL. 222

on

Janu

ary

5, 2

010

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

states have been observed in associationwith past ENSO episodes (Fig. 1I) and insome of the GCM simulations.Not all of the major climate anomalies

of the 1982-1983 winter can be unambig-uously interpreted as interlocking piecesof the ENSOjigsaw puzzle. Wet, stormywinters over California and other partsof the Southwest have occurred duringwarm episodes in the equatorial Pacific(such as 1977-1978), but there have alsobeen abnormally dry winters duringwarm episodes (such as 1976-1977). Sowhen the entire historical record is takeninto account the correlations betweenvarious indices of the ENSO phenome-non and California rainfall are not veryimpressive. Despite some dramatic ex-amples of cold winters over the easternUnited States during past ENSO epi-sodes (for example, 1957-1958, 1969-1970, 1976-1977, and 1977-1978), thecorrelations involving winter tempera-tures for this region as a whole are notvery strong either, and the 1982-1983episode turned out to be somewhatwarmer than normal throughout most ofthe region. One of the most remarkableanomalies of the winter of 1982-1983 wasthe abnormal warmth across not onlyNorth America but also most of Eurasiaat latitudes near 50°N (31). There is noevidence of a systematic pattern ofanomalies over Eurasia during ENSOepisodes. Likewise, over the Atlanticsector there were large anomalies in thepressure and wind fields. Note, for ex-ample, the easterly wind anomalies overBrazil (Fig. 4, A to C) (3, 33). Theseunexplained anomalies serve as a re-minder that only a part of the year-to-year climate variability can be regardedas part of the ENSO signal, even in yearswhen that signal is very strong.

Return to Normal Conditions

Many aspects of the pattern describedin the previous section persisted throughMarch, April, and May 1983. The anom-alous wind pattern and equatorial rainfallcontinued but weakened through May(Fig. 4C). Pressure remained well belownormal in the North Pacific, and thestorm track across the southern UnitedStates remained active through much ofApril.The decay phase, like the onset phase,

appeared to be preceded by a dramaticswing in the Southern Oscillation back tonear-normal conditions from February toMay 1983 (Fig. 3C). While the atmo-spheric circulation was returning to nor-mal, SST along the South Americancoast remained nearly constant, then be-

16 DECEMBER 1983

North Pole

Fig. 6. Atmosphericconditions for De-cember 1982 to Feb-ruary 1983 (compare -Fig. 2A).

gan to fall. But the decrease was lessthan the normal seasonal change, sothere was actually a second increase inthe SST anomalies. Peak anomaly val-ues, in excess of 8 K near South Ameri-ca, were reached in May and June. Thissecond maximum was largely confinedto the eastern equatorial Pacific. Fartherwest, anomalies continued to decreaseslowly from their peak in November andDecember (Fig. 3, A and B). The secondrise near South America might have beena response to the second sharp down-ward swing of the Southern Oscillationindex that took place 3 to 4 monthsearlier (Fig. 3C).

Sea-surface temperature anomaliesnear the South American coast under-went an abrupt decrease during earlyJuly, followed by a more gradual de-crease through September 1983. Thesechanges may have been a response to theswing of the Southern Oscillation backtoward normal from February to May.Although the normal low-level wind pat-tern has been reestablished along theequator at this writing, SST east of140°W is still well above normal, and theassociated area of heavy rainfall, whichmigrated northward from Equador andnorthwest Peru during June and July, isstill in evidence north of the equator.

Discussion

The peculiar time history of the 1982-1983 ENSO episode invites speculationas to whether all such episodes might beviewed as a superposition of two, inter-

related kinds of events, which havemuch in common and usually occur incombination about 6 months apart:

1) An enhancement of the mean annu-al cycle in SST, largely confined to theeastern third of the Pacific, with maxi-mum amplitude near the South Americancoast. Positive SST anomalies develop-ing rather abruptly along the coast duringJanuary or February and spreadipg west-ward, peaking in May or June, and disap-pearing by September or October. Aresponse to the swing of the SouthernOscillation, associated with the suddenweakening of the southeast Pacific anti-cyclone near Tahiti starting around No-vember of the previous year.

2) A broader scale warming of theequatorial Pacific from near the date lineeastward, with maximum amplitudefrom 900 to 150°W, starting around themiddle of the calendar year, peakingnear the end of the year, and disappear-ing a few months later. A response to aswing in the Southern Oscillation involv-ing sea-level pressure rises at the west-ern end of the seesaw (near Darwin).

Events of the first kind ronform tooceanographers' notion of El Niiio in itspurest form, while events of the secondkind are more important from a meteoro-logical perspective because of theirmuch broader longitudinal extent. Onemight think of a typical ;ENSO episode,as exemplified by the compositte (9), asconsisting of an event of the first kindfollowed, with some overlap, by anevent of the second kind. The 1982-1983episode can be interpreted in terms ofthe reverse sequence, conmmencing in

'12W01

on

Janu

ary

5, 2

010

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

1982 with an unusually intense broad-scale warming and followed, in the firsthalf of 1983, by a more localized warm-ing of the first kind that is at this writingin its decaying phase.Whether this interpretation is valid or

useful remains to be seen. In any case,the 1982-1983 episode brings into sharp-er focus a series of probing "chicken andegg" type questions concerning the na-ture of atmosphere-ocean interactions inthe tropics.

1) What are the processes responsiblefor the pronounced and unexpectedswings of the Southern Oscillation dur-ing this and other ENSO episodes? Is theatmosphere somehow responding, in anexaggerated way, to subtle changes inthe tropical Pacific SST pattern, or is itswitching between different circulationregimes for other reasons?

2) Why did the heaviest rainfall, thesurface westerlies, and the warmestSST's gradually migrate eastward acrossthe equatorial Pacific during the epi-sode?

3) What is the role of the annual cyclein the initiation, evolution, and decay ofthis and other episodes?The recent episode also raises ques-

tions about the relation between theENSO phenomenon and climate anoma-lies throughout the world:

1) Were the unusually strong westerlyflow and storminess over the the south-western United States and the unusualwarmth over Eurasia related to the EN-SO episode or did they occur for otherreasons?

2) Is it possible to distinguish betweenepisodes like the most recent one, char-acterized by mild conditions over muchof the northern United States, and thosesuch as the 1976-1977 episode, in whichthe East was significantly colder thannormal?Such questions will be the focus of an

extensive national research program onthe ENSO phenomenon, now in theplanning stage.

References and Notes1. G. T. Walker, Mem. Indian Meteorol. Dep. 24,

275 (1924).2. __ and E. W. Bliss, Mem. R. Meteorol.

Soc. 4, 53 (1932).3. H. P. Berlage, K. Ned. Meteorol. Verh. 88

(1966), entire volume.4. R. B. Montgomery, Mon. Weather Rev. 39

(Suppl.), 1 (1940).5. J. Namias and D. R. Cayan, Science 214, 869

(1981); T. Ichiye and J. Petersen, J. Meteorol.Soc. Jpn. 41, 172 (1963); R. Doberitz, Bonn.Meteorol. Abh. 8, 1 (1968).

6. J. Bjerknes, Tellus 18, 820 (1966); Mon. WeatherRev. 97, 163 (1969); J. Phys. Oceanogr. 2, 212(1972).

7. K. E. Trenberth, Q. J. R. Meteorol. Soc. 102,639 (1976); B. C. Leare, A. R. Navato, R. E.Newell, J. Phys. Oceanogr. 6, 671 (1976); T. P.Barnett, ibid. 7, 633 (1977); E. R. Reiter, Mon.Weather Rev. 106, 324 (1978); W. H. Quinn, D.0. Zopf, K. S. Short, R. T. W. K. Yang, Fish.Res. Bull. 76, 663 (1978); P. R. Julian and R. M.Chervin, Mon. Weather Rev. 106, 813 (1978); C.S. Ramage and A. M. Hori, ibid. 109, 1827(1981); P. A. Arkin, ibid. 110, 1393 (1982); Y. H.Pan and A. H. Oort, ibid. 111, 1244 (1983); S. G.H. Philander, Nature (London) 302, 295 (1983).

8. J. D. Horel and J. M. Wallace, Mon. WeatherRev. 109, 813 (1981).

9. E. M. Rasmusson and T. H. Carpenter, ibid.110, 354 (1982).

10. H. van Loon and R. A. Madden, ibid. 109, 1150(1981).

11. H. H. Hildebrandsson, K. Svenska Vetensk.-Akad. Handl. 29, 1 (1897); N. Lockyer and W. J.S. Lockyer, Proc. R. Soc. London Ser. A 73,457 (1904).

12. J. Bjerknes, Mon. Weather Rev. 97, 163 (1969).13. B. J. Hoskins, A. J. Simmons, D. G. Andrews,

Q. J. R. Meteorol. Soc. 103, 553 (1977).14. A. E. Gill, ibid. 106, 447 (1980); B. J. Hoskins

and D. J. Karoly, ibid. 107, 1179 (1981); J. D.Opsteegh and H. M. Van den Dool, J. Atmos.Sci. 37, 2169 (1980); P. J. Webster, ibid. 38, 554(1981); ibid. 39, 41 (1982).

15. A. J. Simmons, Q. J. R. Meteorol. Soc. 108, 502(1982); P. J. Webster and J. R. Holton, J.Atmos. Sci. 39, 722 (1982); G. W. Branstator,ibid. 40, 1689 (1983).

16. A. J. Simmons, J. M. Wallace, G. W. Bransta-tor, J. Atmos. Sci. 40, 1363 (1983).

17. K.-M. Lau and H. Lim, ibid., in press.18. These numerical models are based on the gov-

erning equations for large-scale atmosphericmotions in a global domain, and are similar tothose used in weather forecasting but integratedfor I month of simulated time or longer togenerate climate statistics.

19. P. R. Rountree, Q. J. R. Meteorol. Soc. 98, 290(1972); J. Shukla and J. M. Wallace, J. Atmos.Sci. 40, 1613 (1983); M. L. Blackmon, J. E.Geisler, E. J. Pitcher, ibid., p. 1410.

20. R. N. Keshavamurty, J. Atmos. Sci. 39, 1241(1982).

21. K. Wyrtki, J. Phys. Oceanogr. 5, 572 (1975); A.J. Busalacchi and J. J. O'Brien, J. Geophys.Res. 86, 10,901 (1981).

22. Little is known about SST patterns farther to thewest over the open ocean during 1940 to 1941because the retrieval of merchant ship observa-tions was disrupted during World War II.

23. R. A. Kerr, Science 219, 157 (1983).24. In a letter to J. Rasmussen (director, NOAA

Climate Analysis Center) dated 17 Janaury 1983,B. Onorio (chief fisheries officer, Tarawa, Re-

public of Kiribati) noted, "Our area has beengetting constant waves and westerly winds dis-rupting fisheries operations. More recentlywinds have been unusually strong easterly."(The return to strong easterly flow occurred latein 1982.)

25. E. M. Rasmusson and J. M. Hall, Weatherwise36, 164 (1983).

26. Washington Post, 4 February 1983; J. L. Rowe,Jr., and T. B. Edsall, ibid., 18 February 1983, p.Al; R. Hoyle, Time (Asian edition), 28 March1983, p. 6; M. von Dijk, D. Mercer, J. Peterson,New Sci., 7 April 1983, p. 30; U.S. News WorldRep., 11 Apnl 1983, p. 48; J. Kapstein, Bus.Week, 25 April 1983, p. 6; J. Lelyveld, NewYork Times, 18 May 1983, p. 6.

27. Special El Nifio report prepared by the Ecuador-ian Naval Oceanographic Institute, Guayaquil,Ecuador (April 1983).

28. P. Bennett, Washington Post, 28 May 1983, p.A2; P. J. Hilts, ibid., 11 June 1983, p. Al; J.Diehl, ibid., 31 July 1983, p. A17; M. deC.Hinds, New York Times, 2 June 1983, p. A8; E.Schumacher, ibid., 12 June 1983, p. Al.

29. D. DeAngelis, Mar. Weather Log 27, 106 (1983).30. In a letter to J. Rasmussen dated 15 February

1983, J. Couchard (chief, FAAA MeteorologicalCenter, Tahiti) noted, "We experienced a tropi-cal hurricane in January and have to go back to1906 for recording so strong a tropical cyclone inthe Tuamotu archipelago. Marquesas Islandsreported rain quantities as never recorded be-fore. From November, 1982, the total is aboutfive times in excess of the mean value." Subse-quently the Marquesas station at Atuona record-ed 2952 mm of rainfall from January throughApril (compared to a mean of 398 mm), and themost devastating hurricane of modern timesstruck Tahiti in April 1983. In the NorthernHemisphere an unusual hurricane (Iwa), movingnorthward from the region of warm water,caused extensive damage in the Hawaiian Is-lands in late November 1982, almost 25 years tothe day since a similar storm struck the islandsduring the warm episode of 1957 [H. E. Rosen-dal, Mar. Weather Log 27, 63 (1983)].

31. R. S. Quiroz, Mon. Weather Rev., in press.32. A. Gullenhaal, Miami Herald, 19 February 1983,

p. 1; B. Siegol, Los Angeles Times, 17 August1983, p. 1.

33. Outgoing longwave radiation data show below-normal rainfall over parts of northeastern SouthAmerica during late 1982 and early 1983 (Fig. 4).Such anomalies are common at this stage of awarm episode [G. T. Walker, Q. J. R. Meteorol.Soc. 54, 79 (1982); C. N. Cavides, Proc. Assoc.Am. Geogr. 5, 44 (1973)]. On the other side ofthe Atlantic, D. C. Duffy (Beneguela EcologyProgramme, University of Cape Town), in aletter to the NOAA Climate Analysis Centerdated 15 March 1983, reported reduced upwell-ing in the southern African area during the 1982-1983 event. L. V. Shannon (unpublished manu-script, Sea Fisheries Research Institute, CapeTown, South Africa) stated, "Inspection of ourpast records of warm and El Nifio type eventsshows that similar perturbations [reduced up-welling along the Cape west coast] were evi-dent." The eastern North Atlantic-western Eu-ropean sector has not shown a consistent ENSOanomaly pattern (10).

34. We thank P. A. Arkin and R. W. Reynolds forpreparing the analysis products for the 1982event and P. B. Wright and T. H. Carpenter forconstructing the historical indices. Recent SSTdata for Puerto Chicama were provided by D. V.Hansen. The work of J.M.W. was supported byNSF grant 81-06099.

SCIENCE, VOL. 2221202

on

Janu

ary

5, 2

010

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om