1

To appear as a chapter in

Present and Past Interhemispheric Climate Linkages in the Americas and Their Societal Effects(V. Markgraf, ed., Academic Pess)

----------------------------------

INTERHEMISPHERIC EFFECTS OF INTERANNUAL AND DECADALENSO-LIKE CLIMATE VARIATIONS ON THE AMERICAS

Michael D. Dettinger

1

, David S. Battisti

2

, McCabe, G.J.

3

, Cecilia M. Bitz

4

,

and Rene D. Garreaud

2

1

U.S. Geological Survey, Scripps Institution of Oceanography, 9500 Gilman Drive, La Jolla, CA 92093-0224 USA

2

Department of Atmospheric Sciences, University of Washington, Seattle, WA 98195-1640 USA

3

U.S. Geological Survey, Lakewood, Colorado USA

4

Quaternary Research Center, University of Washington, Seattle, Washington USA

Abstract

Interannual El Niño/Southern Oscillation (ENSO) and decadal ENSO-like climate variations of the PacificOcean basin are important contributors to the year-to-year (and longer) variations of the climate of Northand South America. Analysis of historical observations of global sea-surface temperatures, global 500 mbpressure surfaces, and western hemisphere hydroclimatic variations that are linearly associated with theENSO-like climate variations yields striking cross-equatorial symmetries as well as qualitative similaritiesbetween the climatic expressions of the interannual and decadal processes. These similarities areimpressive because, at present, the mechanisms that are believed to drive the two time scales of ENSO-likevariability include several candidates that are quite dissimilar in terms of physical processes andlocations. Despite potentially different source mechanisms, both interannual and decadal ENSO-likeclimate variations yield wetter subtropics (when the ENSO-like indices are in positive, El Niño-likephases) and drier midlatitudes and tropics (overall) over the Americas, in response to equatorward shiftsin westerly winds and storm tracks in both hemispheres. The similarities of their continental-surfaceclimate expressions may impede separation of the two ENSO-like processes in paleoclimaticreconstructions.

Resumen

Variaciones asociadas con El Niño/Oscilación Sur (ENSO) a escala de tiempo de caracter interanual ydecadal en el Océano Pacífico, resultan contribudores importantes del clima de Norte y Sur America.Analisis de los registros historicos de la temperatura de superficie del mar (SSTs), del campo global denivel de la superficie a 500 mb, y de variaciones hidroclimáticas en el Hemisferio Occidental asociadas enforma lineal con variaciones de origen ENSO, exhiben simetrias marcadas en ambos hemisferios en lamayoria de sus expresiones de variabilidad climatica, en ambas escalas—interanual y decadal.

Estas similaridades son notables, porque, al presente, los mecanismos que se creen responsablespor las variaciones climaticas en estas dos escalas de variación temporal, vinculadas con fluctuacionesgeneradas por el fenomeno ENSO, incluyen varios candidatos, los cuales son asociados con distintosprocesos físicos y zonas geográficas de acción. No obstante estos mecanismos potentialmente distintos,variaciones climaticas de escala interanual y decadal resultan en el desarrollo de condiciones maslluviosas que lo normal en las regiones subtropicales (durante la fase calida del ENSO), y condiciones masseca que lo normal en las latitudes medias y en los tropicos en general sobre el continente Americano.Esto ocurre en repuesta a los cambios de vientos de superficie en la zona ecuadorial de Pacifico, y cambiosde las trajectorias de tormentas en ambos hemisferios. La similaridad de las expresiones climaticas ensuperficies en ambos hemisferios y escalas temporales, si acaso hará mas dificil la separación de lasseñales de indole paleoclimaticas de estas de variación del sistema ENSO.

2

Introduction

A climatologist who specializes in modernclimate variations was immediately struck,upon attending a workshop of the Pole-Equator-Pole (PEP 1) Paleoclimate of theAmericas Program, by the need to reconcilethe various past climate reconstructions withclimatic processes and conditions that we canrecognize from instrumental climatevariations. Some past-climate variations arerecognizable; many are not. Indeed, oninterannual and decadal timescales, thenumber of “typical” patterns of historicalclimate variation is limited. In the presentclimate, three climatic phenomena dominateclimate variations in the Americas oninterannual to decadal timescales. Best knownis the El Niño/Southern Oscillation (ENSO)phenomenon, which dominates global climatevariations on interannual timescales ranging(mostly) from 3 to 6 years; many climatestudies have described this quasi-regularphenomenon of tropical air–sea interactionscentered in the equatorial Pacific and itsteleconnections to many parts of the globe (seereviews in Diaz and Markgraf 1992; Allan etal. 1996; Wallace et al. 1997). Less wellcharacterized are decadal (10- to 50-yeartimescales) climate processes centered in thePacific and in the Atlantic Ocean basins. Ontimescales longer than interannual, thedominant climate phenomenon in the PacificOcean and overlying atmosphere has anENSO-like spatial distribution (Zhang et al.1997, hereafter ZWB) of surface temperaturesand atmospheric circulations and, as will beshown here, interhemispheric climate effectson the Americas. The physical processesresponsible for this decadal ENSO-likevariability remain uncertain but are tied towell-documented pan-Pacific changes in theatmosphere and ocean in the 1950s and mid-1970s (Ebbesmeyer et al. 1991). Finally,decadal variability in the climate of theAtlantic basin also has a basin-wide signature

(Deser and Blackmon 1993; Hurrell 1995;Enfield and Mestas-Nuñez, 2000, this volume),but is somewhat distant from the cordillera ofthe Americas and thus will not be addressedhere.

In this chapter, modern interannual anddecadal climate variations of the Pacific basinwill be outlined and their interhemisphericclimatic effects on the Pacific Ocean and theAmericas will be compared. Several indicesmay be used to describe the historicalvariations of the ENSO and decadal ENSO-like phenomena, but often it is difficult to besure that the two phenomena have beenseparated so that none of the climatic effectsare attributed twice. Thus, a new set ofuncorrelated indices for the two processes willbe developed and used, alongside a pair ofindices that have been used in the literature, tocharacterize oceanic, atmospheric, and land-surface climate variations associated with theENSO-like phenomena. The construction ofindices, and datasets that will be compared tothose indices, will be described in the nextsection of this chapter. In subsequent sections,oceanic, atmospheric, and continentalconditions that are linearly related to theinterannual and decadal phenomena will bedelineated by simple correlation andregression analyses. In the final two sections,the results will be discussed and conclusionssummarized.

Data and definitions

Interannual and decadal climate variationsin the Pacific Ocean basin can be tracedthrough any of literally dozens of physical andbiological indices (e.g., Ebbesmeyer et al.1991). In this chapter, two sets of indices basedon sea-surface temperature (SST) patterns ofglobal scale will be used to depict the NorthPacific climate and its effects on the Americas.Two indices developed by ZWB—the CoolTongue (CT) index and the Global Residual(GR) index—will be discussed for the most

3

part but, for completeness, a second pair ofindices based on a linear combination of somemore commonly available time series—Southern Oscillation Index (SOI), the CTindex, and the Pacific (inter)DecadalOscillation (PDO) index—also will beconsidered. The origins and significance ofthese various indices are discussed below.

Global SSTs were analyzed by ZWB usinglinear regressions and principal component(PC) analyses in order to separate interannualENSO-related SST variations from otherinterannual and decadal variations and todescribe those other SST variations.

Interannual ENSO-related SST variations werecharacterized by the CT index, which is theaverage of SST anomalies from 6

°

N to 6

°

S,180

°

to 90

°

W, an index previously used byDeser and Wallace (1990). This region is thelocus of cool SSTs in the eastern equatorialPacific, where the El Niño process has itsclearest surface expression. Monthly meanSST anomalies on a 5

°

x 5

°

grid, from anupdated version of the global United KingdomMeteorological Office Historical Sea SurfaceTemperature Dataset (HSSTD; Folland andParker 1990, 1995), were used to form thisseries (Fig. 1a), and then the CT series was

Figure 1. (a) Unfiltered Cool-Tongue (CT) index and rotated “tropical” PC index, of dominantly interannual El Niño-Southern Oscillation (ENSO) variability in the tropical Pacific, and (b) Global Residual (GR) index and rotated “North Pacific” PC index, of dominantly decadal ENSO-like variability in the Pacific Ocean basin. The

indices are described in the text. CT and GR indices are identical to those of Zhang et al. (1997).

1910 1930 1950 1970 1990Y

−4

−2

0

2

4

Dim

ensi

onle

ss

(b) Decadal, extratropical indices

GR INDEXN. PACIFIC PC INDEX

1910 1930 1950 1970 1990−4

−2

0

2

4

Dim

ensi

onle

ss

(a) Interannual, tropical indices

CT INDEXTROPICAL PC INDEX

4

high-pass filtered to remove variations withperiods longer than 6 years, forming a strictlyinterannual series called CT*. The entire SSTdataset then was regressed against the CT*time series, and the best linear fits at each gridpoint were subtracted from the grid-point SSTsto arrive at time series of residuals. Byconstruction, these residuals are uncorrelatedwith the CT* series. A PC analysis of theseresidual series yielded a leading mode of “non-ENSO” SST variation that was, surprisingly,very ENSO-like in its spatial patterns. Thisleading mode is called the GR index by ZWBand represents the decadal ENSO-likevariations of the global and, especially, thePacific SSTs (Fig. 1b). Although CT* and GR

are not correlated, CT and GR are (r = -0.55).In this chapter, the interannual and decadalclimate variations in the Pacific basin will becharacterized in terms of these same CT andGR series; CT, rather than CT*, is usedbecause it is more similar to such commonlyused ENSO indices as the unfiltered SSTindices along the equatorial Pacific in so-calledNiño-3 and Niño-4 regions (e.g., Allan et al.1996) than is the band-pass filtered CT*.

The differing timescales of the CT and GRseries can be demonstrated by spectralanalysis. Shown in Figure 2 are multitaperpower spectra (Percival and Walden 1993;Lees and Park 1995) of the two series. The CTspectrum (solid curve) indicates variance

Figure 2. Multitaper power spectra of the Cool Tongue (CT) and Global Residual (GR) climate indices. The inset is a plot of the corresponding power spectra of rotated PC series: “tropical” PCs (solid curve) and “North Pacific” PCs (dashed curve). In the CT and GR spectra, at lowest frequencies, powers greater than about 80 to 200 exceed

95% confidence levels—not shown here—in comparisons to red noise (Mann and Lees 1996); at highest frequencies, power greater than about 30 reaches 95% confidence levels.

0.0 0.2 0.410

100

1000

Pow

er

PC indices

0.0 0.1 0.2 0.3 0.4Frequency, in cycles/year

1

10

100

1000

Pow

er

Multitaper power spectra of CT and GR indices

CTGR

5

spanning a broad range of frequencies centerednear 0.2 cycles/yr, corresponding to a period of5 years. In contrast, GR has a spectrum thatincreases rather steadily from high frequenciesto low, with a local maximum near 0.16 cycles/yr. Thus GR is dominated by low-frequencyvariations with timescales between about 6years and several decades; CT is dominated byinterannual variations in the 3- to 7-year range.Recall that the CT series used here has notbeen temporally filtered; therefore, theinterannual dominance of CT shown reflectsthe true mix of frequencies in the tropicalPacific ENSO variations. The GR index isderived as a residual from an interannuallyfiltered CT* series (and its reflection in globalSSTs) and thus, as a measurement of NorthPacific climate variation, is biased by designtoward the more decadal parts of Pacificclimate variability.

Alternative approaches and indices forcharacterizing the interannual and decadalmodes of Pacific climate variation have beendeveloped in the literature. Enfield andMestas-Nunez (2000, this volume) usedcomplex empirical-orthogonal function (EOF)analyses of global SSTs to arrive at similarmodes. Mantua et al. (1997) used the CT indexand a PDO index to describe essentially thesame variability. The PDO index is defined, byMantua et al. (1997), as the first PC of monthlySSTs, from the HSSTD, poleward of 20

°

N inthe Pacific basin, initially calculated fromanomalies for the period 1900-93. This indexhas a strong decadal character much like theGR index (and is correlated with GR at r =+0.74) but has not had any ENSO variabilityexplicitly removed; indeed, its correlation withthe CT series is r = +0.38. The PDO wasintended to reflect the decadal variability of theNorth Pacific in a simple index but has asubstantial expression throughout the tropicaland South Pacific (see Fig. 3d here or Fig. 2aof Mantua et al. 1997).

Tropical ENSO variability historically—e.g., Troup (1965) and most more recent

references—has been indexed by the SOI andSST anomalies in selected equatorial regions.The SOI is usually defined as the differencebetween standardized monthly departures ofsea-level pressure (SLP) from their normalseasonal cycles at Tahiti and Darwin,Australia. When SLPs are low in the easterntropical Pacific near Tahiti and high in thewestern tropical Pacific near Darwin (whenSOI is negative), the tropical easterly windsbecome weak and the resulting slackening ofequatorial upwelling of cool deep water in theeastern equatorial Pacific, together with arelaxation of warm waters from the westernPacific into the usually cool eastern Pacific,leads to warmer than normal SSTs in theeastern equatorial Pacific. These episodes ofwarm equatorial SSTs in the central andeastern Pacific have far-reaching climaticconsequences and have traditionally beencalled El Niños. When the SLP anomalies arereversed (high at Tahiti and low at Darwin) andthe SOI has a positive value, the tropicaleasterlies strengthen and the eastern equatorialPacific cools anomalously in episodes thatrecently have been called La Niñas, toemphasize their contrasts with El Niños. Thus,the SOI index is an atmospheric measure ofENSO processes that complements, but largelyreflects, the CT index, with which it is highlyanticorrelated (r = -0.73). The PDO index is, inpart, due to ENSO, as indicated by thecorrelation between PDO and SOI indices (r =-0.51).

Although the CT and GR indices of ZWBwill be the focus of much of the remainingdiscussion, it was also useful to develop yetanother pair of indices for interannual anddecadal ENSO-like variations, a pair of indicesthat—by design—are entirely uncorrelatedwith each other but which are not derived fromany explicit temporal filtering. In order todevelop these new indices, three commonlyused climate indices—SOI, CT, and PDO—were joined in a simple trivariate PC analysis.The two most influential of the resulting

6

components were weighted heavily on (a) SOIand CT, the tropical indices, and (b) the PDO,the extratropical index, respectively. Thebenefit of this PC analysis was that, byconstruction, it yielded statisticallyuncorrelated series that described the tropicaland extratropical variations, without arbitraryfiltering of either. The two components werethen rotated by the Varimax methodology(Richman 1986), although this only served toreduce the amount of mixing between PDO orthe tropical indices in the PCs (nudging theloadings toward even more one-sidedweighting of one or the other sets in eachcase). The two resulting, still uncorrelated,rotated PC series (dotted curves, Figs. 1a and1b) will be called the tropical PC and the NorthPacific PC here, respectively, although it isclear from later results that the “North Pacific”PC reflects pan-Pacific climate variations thatmay yet have their roots in the tropics. Theweights of SOI, CT, and PDO in each PC aregiven in Table 1. When the more commonlyused Niño-3 SST average, from 5

°

S-5

°

N,150

°

W-90

°

W, is substituted for CT in thiscalculation, the amount of mixing betweenindices that is required to arrive at uncorrelatedtropical and North Pacific climate indices iseven smaller.

The power spectra of the resulting PCseries are shown as an inset to Figure 2, andare similar in character to the spectra of CTand GR, with the tropical PC naturally echoingCT’s interannual character and the NorthPacific PC following GR’s strong decadal

dominance. Indeed, although no temporalfiltering was applied when deriving these PCs,the North Pacific PC has narrower spectralpeaks than does the GR series, concentratednear the periods greater than 25 years and near7 years, and the tropical PC has powerconcentrated in a more clearly definedinterannual frequency range—0.16 to 0.3cycles/yr (3- to 7-years)—than does CT.Indeed, the tropical PC is virtually identical toZWB’s CT*, the Cool Tongue index filtered toremove variability longer than 6 years.

The oceanic and atmospheric patterns thataccompany the interannual ENSO and theENSO-like climate variations indexed by theCT and GR series, and by the tropical andNorth Pacific PCs, will be described in the nexttwo sections by regressions of variousatmospheric fields, and correlations of SSTfields, with the series. In a subsequent section,the climatic consequences of the variability atthe surface of the Americas will be describedby similar methods. Sea-surface temperaturevariations are depicted from the HSSTD setused in developing the CT and GR indices.Atmospheric forms of the variations aredepicted in terms of the recently releasedreanalyzed global-atmospheric fields of 500mbar height anomalies from 1958-96 on a2.5

°

x2.5

°

grid (Kalnay et al. 1996), from theNational Centers for Environmental Predictionand National Center for Atmospheric Research(NCAR). The reanalyzed fields provide globalcoverage but are of questionable reliability athigh latitudes where observations are

Table 1: Principal component loadings for Varimax-rotated factors of October-September, 1915-88, average indices of Pacific climate variability.

Index Tropical PC North Pacific PC

SOI -0.85 -0.38

CT +0.95 +0.10

PDO +0.21 +0.97

7

particularly sparse and can be problematicalong major coastlines where there areparticularly steep “gradients” in the density ofobservations; overall, however, the fieldsprovide the best available, global depiction ofatmospheric conditions in the last 40 years.Also considered is a transformation of the 500mbar fields that represents monthly variationsin high-frequency storminess. As described byBitz and Battisti (1999), daily 500 mbar heightanomalies at each grid point were band-passfiltered to isolate variability with timescalesbetween 2 and 8 days; the monthly root meansquare of these high-frequency fluctuationswere then used to map changes in monthlymean storminess. In the Northern Hemisphere,these analyses will be extended back to thefirst half of the Twentieth Century by applyingthe same analysis of high-frequency variationsto 5

°

x5

°

-gridded daily Northern Hemispheresea-level pressure fields from NCAR(Trenberth and Paolino 1980).

The continental effects of the ENSO-likeclimate variations are depicted in terms ofsurface air temperatures, precipitation, andstreamflow in North and South America.Temperatures from an updated version of themonthly, 5

°

x5

°

-gridded temperature anomalyset of Jones et al. (1986a,b; Carbon DioxideAnalysis Center [CDIAC] at Oak RidgeNational Laboratory product NDP020), from1904–90, and land-precipitation anomalies ona similar grid from Eischeid et al. (1991,1995), for the same time period, are analyzed.Attendant streamflow variations are analyzedin the monthly records catalogued in a WesternHemisphere subset of the global streamflowset compiled by Dettinger and Diaz(submitted); this subset includes long-termmonthly streamflow records at 248 sites in theAmericas. The series were drawn from amongthe U.S. Geological Survey’s HydroclimaticData Network for United States streamflow(Slack and Landwehr, 1992; updatedsubsequently to 1995 by L. Riddle, ScrippsInstitution of Oceanography) and from several

public-domain international sources, and, asmuch as possible, are free from overwhelminghuman influences. Also included in the presentanalyses are 13 proprietary Brazilian flowseries provided by Jose Marengo, InstitutoNacional de Pesquisas Espacias, courtesy ofEletrobras and Eletronorte in Brazil.

The results presented here are comparisonsbetween climate index, SST, atmospheric-circulation pattern, precipitation, surface-airtemperature, and streamflow anomaliesaveraged over “water years” that beginOctober 1 and end September 30. The decisionto everywhere analyze the (northern wateryear) October-September averages is arbitrarybecause seasons reverse across the Equator andthus are not strictly comparable. No singledefinition of annual averages will be bestsuited everywhere; however, Dettinger et al.(2000) have shown that persistence ofinterannual teleconnections across theAmericas is sufficiently long to allow usefulinterpretation of climatic and hydrologicresponses within such a “water year.” Decadalvariations also are expected to yield slowlyvarying teleconnections that can be studiedwithin this “water year.”

Sea-surface temperature variations

Both pairs of climate indices used here tocharacterize interannual and decadal ENSO-like variability—CT/GR and the tropical/NorthPacific PCs—are derived from SSTs. Thus,initially, the global SST variations associatedwith the indices will be outlined. The patternsof SST associated with the CT and GR seriesare indicated by correlations mapped inFigures 3a-b; positive correlations indicateregions where the ocean is, on average,warmer when the index is more positive andcooler when the index is more negative. Asexpected from the preceding descriptions ofthe indices, the CT index is more (positively)correlated with SSTs in the tropical Pacificthan is the GR index. Conversely, the GR

8

index is more (negatively) correlated withSSTs of the North Pacific. Extratropical SSTsin the southern Pacific are about equallycorrelated to the two indices. Both indices aremodestly correlated with SSTs in the tropicalIndian Ocean. Despite the differences inemphasis between the CT and GR correlations,the two most remarkable aspects of the SSTpatterns associated with the two indices are (1)the symmetry of each correlation pattern aboutthe Equator in the Pacific basin, and (2) theoverall similarity between the CT and GRpatterns. This similarity led ZWB to describethe decadal variations that dominate the GRindex as “decadal ENSO-like variability” andmotivated this study. The strong cross-equatorial symmetries of the SST patterns arereflected in strong symmetries of atmosphericcirculations and hydroclimatic responsesanalyzed in subsequent sections of this

chapter.The rotated tropical and North Pacific PC

indices are correlated to SSTs in patterns(Figs. 3c-d) that are similar to CT and GR,respectively. However, SST correlations withthe tropical PCs emphasize the tropical PacificSSTs (relative to extratropical SSTs) evenmore than do the CT SST correlations. Sea-surface temperature correlations with theNorth Pacific PCs emphasize the North Pacifictemperatures even more than does GR. Onceagain, southern SSTs are about equallycorrelated with the two PC series.

Thus the two pairs of Pacific climateindices correspond to similar patterns of SSTvariation. The CT corresponds largely toENSO variations by its very definition; thetropical PC emphasizes those variations as anoutgrowth of the parsimony provided by all PCanalyses. As representations of ENSO

Figure 3. Correlation coefficients between annual-averaged sea surface temperatures (SSTs) and (a) Cool Tongue (CT) index (1903-90), (b) Global Residual (GR) index (1903-90), (c) rotated tropical PC index (1914-90), and (d) North Pacific PC index (1914-90). The contour interval is 0.2, dashed where negative. Correlations greater

than +0.2, or less than -0.2, pass a two-tail student-t test of being different from zero at 95% significance levels.

0

0

0

0

0

0

20

20

20

20

20

40404060

60

-20

-20

-20

0

0

0

0

0

0

07 607E 1207E 1807 1207W 607W 07607S

307S

07

307N

607N

00

0

0

0

0

0

0

20

20

20

2020

20

4040

40

40

-40

-20-20

-20

-20

-20

0

0

0

0

0

0

0

0

07 607E 1207E 1807 1207W 607W 07607S

307S

07

307N

607N

0

0

0

0

0

0

20

20 20

20

20

20

40

404060

60

-20

-20

-20

0

0

0

0

0

0

07 607E 1207E 1807 1207W 607W 07607S

307S

07

307N

607N

00 0

00

0

0

20

20

20

20

20

20

20

40

40

-60

-40

-20

-20

-20

-20 0

0 0

0

0

0

0

07 607E 1207E 1807 1207W 607W 07607S

307S

07

307N

607N

(a) CT versus SSTs

(b) GR versus SSTs

(c) Tropical PC versus SSTs

(d) North Pacific PC versus SSTs

9

variations in the tropical Pacific, the CT andtropical PC series have dominantly interannualtimescales (Fig. 2). The GR and, even more so,the North Pacific PC series capture variationsthat are not correlated with CT variations; thatare dominantly decadal (although neither hasbeen limited to that timescale; Fig. 2); and thatare similar to SST patterns in the world oceansassociated with ENSO. How the interannualand decadal climate indices can be related tosuch similar SST patterns, while having suchdifferent (and, in the case of the PCs, entirelyuncorrelated) temporal variations, issomething of a mystery, but these similaritieswill be echoed in the atmospheric expressionsof the time series as well as in theirhydroclimatic expressions over the Americas.

Miller and Schneider (1998) recently listedsix categories of mechanisms that maycontribute to the decadal timescales of SSTvariations in the North Pacific:

• Stochastic atmospheric forcing with a low-frequency SST response (e.g., Barsugli and Battisti 1998)

• Decadal tropical forcing of midlatitude SSTs (e.g., Trenberth 1990, Graham 1994)

• Decadal midlatitude ocean–atmosphere interactions of ocean gyre strength and wind stress curls (e.g., Latif and Barnett 1994, and, as a North Atlantic analog, Deser and Blackmon 1993)

• Tropical–extratropical interactions through subduction of midlatitude (atmospherically forced) SST anomalies into the subsurface ocean to provide source waters of upwelling in the El Niño region of the tropical Pacific (e.g., Gu and Philander 1997)

• Slow oceanic-wave teleconnections from the tropics to the extratropical oceans (e.g., Jacobs et al. 1994)

• Intrinsic decadal vacillation of

midlatitude ocean currents (e.g., Jiang et al. 1995)

To this list, we would add the slow vestigesof irregular interannual ocean–atmosphereclimate variations. Aperiodic variations oninterannual scales will yield decadalcomponents upon averaging to decadaltimescales, and those “decadal” variations willnecessarily be similar in appearance to theinterannual climate processes. Somecombination of these mechanisms, then,presumably generates decadal Pacific SSTvariations that, in turn, influence and evendrive parts of the global climate system withclear, if irregular, decadal timescales.

Notably, the present analysis demonstratesremarkable interhemispheric symmetries inexpressions of the decadal climate variationsassociated with the “North Pacific” modes.This interhemispheric symmetry may provideclues as to which of the processes listed aboveis most likely to be dominant; e.g., those thatwould be expected to yield strong symmetriesabout the equator. Resolution of the questionof which mechanisms dominate is, however,beyond the scope of the present analysis andwill require a careful combination of ocean–atmosphere modeling with careful analyses ofthe details of global atmospheric and oceanicobservations over the last 50 years.

Atmospheric associations

Superposed on these SST patterns, andinteracting with them, are the atmosphericcounterparts (and consequences) of the CT andGR SST variations. Regression coefficientsrelating 500 mb height anomalies (deviationsof the heights of 500 mbar pressure surfaces inthe atmosphere from their long-term seasonalaverages) to the CT and GR series are shownin Figures 4a and 4c. Note that regressioncoefficients are related to correlationcoefficients by

10

β

=

r

σ

500

/

σ

clim

,

where

β

is a regression coefficient,

r

is thecorresponding correlation coefficient,

σ

500

isthe local standard deviation of 500 mb heightanomalies, and

σ

clim

is the standard deviationof the climate index. Hence, these maps can beinterpreted as the typical atmospheric anomalythat accompanies a modest excursion (1standard deviation) of the climate index.Negative coefficients relate the 500 mb heightanomalies and CT over vast Siberian andNorth Pacific regions, whereas a northwest-southeast slanted dipole of positive andnegative coefficients is indicated over theSouth Pacific (Fig. 4a). The negativecoefficients correspond to lower than normal500 mb heights when CT is positive (El

Niños). The broad region of negativecoefficients over the North Pacific correspondsto an average 500 mb height decline of asmuch as 20 m for a 1 standard deviation rise inCT. Positive CTs (El Niños) are associatedwith deeper than normal Aleutian Lows, andthe resulting steep north-south coefficientgradients near 35

°

N 160

°

W in Figure 4acorrespond to anomalously strong 500 mbheight gradients, and anomalously strongwesterly winds, when CT is positive.Anomalies in atmospheric circulationsassociated with positive CT values (El Niños)thus bring storms and precipitation to thesouthwestern and southeastern parts of NorthAmerica. The circulation pattern associatedwith positive CTs also results in enhancedmidlatitude blocking and the diversion of

Figure 4. Regression coefficients relating (a) 500 mb height anomalies to Cool Tongue (CT) index, (b) 500 mb storminess to CT index, (c) 500 mb height anomalies to Global Residual (GR) index, and (d) 500 mb storminess to GR index. The contour intervals in (a) and (c) are 5 m per climate-index standard deviation, and in (b) and (d) are 0.5 m per climate-index standard deviation. Storminess in panels (b) and (d) is measured by monthly standard

deviation of band-pass (3-10 days) filtered 500 mb height anomalies.

07 607E 1207E 1807 1207W 607W 07

607S

307S

07

307N

607N

00

0

00

5

5

55

5

5

10-15 -10

-10

-5

-5

-5

-5-5

00

0

00

07 607E 1207E 1807 1207W 607W 07

607S

307S

07

307N

607N

00

00

0

0

0

0

00

0.50.5

0.50.5

1

1

-2-1.5

-1

-1

-0.5

-0.5

-0.5-0.5

00

00

0

0

0

0

0

0

07 607E 1207E 1807 1207W 607W 07

607S

307S

07

307N

607N

0

0

0

0

0

55

55

5

5

10

-15-10

-5

-5

-5

0

0

0

0

0

07 607E 1207E 1807 1207W 607W 07

607S

307S

07

307N

607N

0

00

0

0

0

0

0

0

0

00

0.5

0.5

0.5

1

1

-1

-1

-0.5 -0.5

-0.5-0.5

-0.5

0

0

0

0

0

0

0

0

0

0

00

(a) CT versus 500-mb heights

(b) CT versus 500-mb storms

(c) GR versus 500-mb heights

(d) GR versus 500-mb storms

11

storms away from northwestern NorthAmerica. The effects of these circulationchanges on North American climate andhydrology will be discussed in the nextsection.

The dipole of positive and negativecoefficients over the southeast Pacific in Figure4b corresponds to diversions of low pressuresystems and attendant storms toward thesubtropical belt of South America when CT ispositive (El Niños), in a rough parallel to thechanges in circulation in the northernhemisphere. Observations of 500 mb heights inthe far south are rare enough, however, so thatthe coefficients may reflect dynamicconsistency requirements of the reanalysisprocess (Kalnay et al. 1996) more than anobserved feature of atmosphere. Anomalouslystrong westerly winds are associated withpositive CT along a belt extending from thesouthern ocean near the international date lineto the southern Archipielago de los Chonos ofChile. These patterns have been described asthe Pacific South American pattern, a primarymode of Southern Hemisphere atmosphericvariability (Mo and Ghil 1987, Kidson 1988,Mo and Higgins 1998, Garreaud and Battisti,1999).

Together, the low-pressure bands (ofnegative coefficients) associated with positiveCT in the Northern and Southern Hemispheresreflect the extratropical extensions of the SOIseesaw of pressures. The Southern Oscillationpattern is not well represented at 500 mb in thetropics (Fig. 4a), but along 30

°

N and 30

°

S itappears as the equatorward increase incoefficients (toward positive values in thetropics). In the tropical eastern Pacific, theincrease in height at 500 mb is mirrored as adecrease in pressure in the lower troposphereand a slackening of easterly trade winds duringEl Niños.

The relations between CT and midlevelwinds and low-pressure systems are inferredfrom regression coefficients mapped in Figure4a. Corresponding relations between CT and

storminess can be inferred from regressioncoefficients relating monthly standarddeviations of high-frequency 500 mb heightanomalies to CT, as in Figure 4b. Positivevalues in Figure 4b correspond to regions inwhich 500 mb heights are more variable, on 3-to 10-day timescales, than normal when CT ispositive; regions with negative values are lessvariable and less stormy. Positive CTs (ElNiños) are associated with less than normalstorminess across Siberia, Alaska,northwestern Canada, and the northwesternUnited States. By this measure, thesouthwestern United States is more stormythan normal during El Niños, but the EasternPacific storm track, near 35

°

N 160

°

W, isclearly enhanced, is extended farther eastwardover the southern United States, and isdisplaced farther south than normal.Storminess may also be enhanced during ElNiños over northeastern Canada. A similaranalysis (not shown here) of monthly standarddeviations of high-frequency SLP variability inthe Northern Hemisphere (where daily SLPrecords span the entire twentieth century)indicates that the associations of SLP-basedstorminess with CT closely parallels those inthe northern half of Figure 4b both before andafter the 1950s (the period encompassing thebulk of available 500 mb height observations).

In the Southern Hemisphere, positive CTs(El Niños) are associated with greaterstorminess over the central Pacific along theband from 30

°

to 50

°

S. Between SouthAmerica and the date line, storminess appearsto be diverted northward from midlatitudes(40

°

to 60

°

S) towards about 30

°

S. Lessstorminess is indicated over southernArgentina.

The atmospheric expressions of GR arebroadly similar to CT’s influences, but—aswith SSTs—with different spatial emphases.In Figure 4c, notice that regression coefficientsrelating 500 mb height anomalies to GR aredominated by negative coefficients over theNorth Pacific and a weaker north-south dipole

12

of negative and positive coefficients over theSouth Pacific. The negative coefficients overthe North Pacific are very similar tocoefficients associated with CT. However, theless similar circulation anomalies over NorthAmerica indicate that the circulationsassociated with CT are more zonal than thoseassociated with GR. In the SouthernHemisphere, the dipole of coefficients is muchless intense, implying relatively lessexpression of GR than CT in SouthernHemisphere 500 mb heights and winds.

Storminess patterns associated with GRloosely parallel those associated with CT, butvariations of GR (Fig. 4d) seem to yieldstorminess anomalies that are more focusedover the North Pacific and the PacificNorthwest, with a much weaker expressionover the southern United States than isassociated with CT. In the midlatitude Pacificof both hemispheres, the relationship betweenstorminess anomalies and the average 500 mbheight anomalies during ENSO variations issimilar to the relationship between storminessanomalies and the average 500 mb heightsduring the decadal ENSO-like variability: Adecrease (increase) in storminess is found onthe poleward (equatorward) flank of thenegative midlatitude height anomaly in thecentral Pacific (e.g., on the flanks of theAleutian Low) when the indices are positive.The Northern Hemisphere storminess patternsassociated with GR (Fig. 4d) also are reflectedin correlations of pre-1950s GR values withSLP-based storminess indices from that earlierperiod (not shown).

Taken together, the panels of Figure 4suggest qualitative similarities in theatmospheric expression of interannual ENSOvariability (CT) and decadal ENSO-likevariability (GR). These atmosphericsimilarities, and the SST similarities discussedpreviously, will be related to their similarhydroclimatic effects in the Americas in thenext section. Overall, the atmosphericexpressions of both CT and GR (as well as

tropical and North Pacific PC indices—notshown here) are remarkably symmetric aboutthe Equator, especially over the Pacific Ocean:positive CT and positive GR variations areassociated with equatorward diversions ofwesterlies, low-pressure systems, and stormsfrom the midlatitude Pacific basins towardsubtropical latitudes. These cross-equatorialsymmetries may reflect tropical roots for bothCT and GR (and for the tropical and NorthPacific PCs). Thus strong interhemisphericsymmetries of ENSO-like variations of Pacificclimate are likely to produce stronginterhemispheric climate variations along theentire length of the North and South Americancordillera on both interannual and decadaltimescales.

Climatic effects in the Americas

Most of the paleoclimatic records ofclimate variability studied as parts of PEP-I areproxies for either near-surface temperaturevariations or precipitation/soil-moisturevariations over the Americas. The interannualand decadal ENSO-like climate variationsdepicted in the previous sections playimportant roles in determining thosevariations, roles that will be delineated below.

Annual-scale temperature, precipitation,and streamflow variations associated with CTand GR are illustrated in Figure 5, whichshows regression coefficients relatingdeviations of October-September averages ofprecipitation and temperatures from normal toCT (Figs. 5a and 5c) and GR (Figs. 5d and 5f),respectively. Also shown are correlationcoefficients between streamflow and CT (Fig.5b) and GR (Fig. 5e). Correlations areindicated by circles with radii proportional tothe coefficients. Sizing the circles with radiiproportional to correlation (and regression)coefficients makes the area of the circlesproportional to the fraction of variance that isassociated with the climate index at each site.

13

Precipitation and streamflow

Strong spatial similarities between CT- andGR-forced patterns of precipitation areimmediately obvious from a comparison ofFigures 5a and 5d. Streamflow variations withCT and GR also are similar (Figs. 5b and 5e)and mostly reflect, and even amplify(Dettinger et al. 2000), basin-scale

precipitation variations. Both CT and GR arepositively correlated with precipitation andstreamflow in the southwestern United States,Central America, and in Paraguay, Uruguay,and northern Argentina. Both CT and GR arenegatively correlated with precipitation andstreamflow in much of northwestern NorthAmerica, in the tropical parts of SouthAmerica, east of the Andes, and possibly—

Figure 5. Regression coefficients—B—relating Cool Tongue (CT) index to October-September (a) precipitation, 1904-90, and (c) surface-air temperatures, 1904-90; and correlation coefficients—r—between CT and (b)

streamflows, periods of record ranging as long as 1904-1990 but commonly on order of 40 years. Panels (d)-(f) are same as (a)-(c) but for Global Residual (GR) index. Radii of circles are proportional to magnitude of

regression and correlation coefficients; black for positive relations, white for negative. Circles inset near 30

°

S, 135

°

W, indicate scale of influences: B is regression coefficient and r is correlation coefficient (r=0.3 is significantly different from zero at 95% confidence level for average length of streamflow records).

1357W 907W 457W607S

307S

07

307N

607N

B = -10mm/monDry

Cool-Tongue index, 1904-90

(a) Precipitation

1357W 907W 457W607S

307S

07

307N

607N

B = +10mm/monWet

Global-Residual index, 1904-90

(d) Precipitation

1357W 907W 457W607S

307S

07

307N

607N

B = -0.2CCool

(c) Temperature

1357W 907W 457W607S

307S

07

307N

607N

B = +0.2CWarm

(f) Temperature

1357W 907W 457W

r = -0.3Dry

(b) Streamflow

1357W 907W 457W

r = +0.3Wet

(e) Streamflow

14

although data are particularly sparse there—insouthernmost South America. Compositeaverages (not shown here) of precipitation andstreamflow during negative and positive phasesof CT and GR indicate that drier than normalconditions in tropical South America and in thenorthwestern United States are more extremeduring positive phases of both CT and GR, andcontribute more to the regression relationsshown in Figure 5 than do wet conditionsduring negative CTs and GRs. Positive valuesof both CT and GR also contribute more(wetness) to the regression relations for thesouthwestern United States and Paraguay-Uruguay-Argentina than do negative phases.Negative phases may be more importantcontributors to the regression relations in thenortheastern United States. Dettinger et al.(1998) and Cayan et al. (1998) also found clearsimilarities between ENSO and decadalprecipitation patterns over western NorthAmerica.

In the midlatitudes of the NorthernHemisphere, the precipitation and streamflowpatterns shown in Figure 5 reflect theequatorward shifts in midlatitude westerliesand storm track associated with positive valuesof CT and GR that were indicated in Figure 4and discussed in the preceding section. WhenCT is positive, more storms, precipitation, andstreamflow occur in the southern United Statesand, across the equator, in Paraguay-Uruguay;the northwestern United States and much ofwestern Canada States experiences lessstormy, drier and warmer conditions.

The patterns of precipitation, streamflow,SST, and circulation anomalies over theAmericas associated with CT and GR areremarkably similar. However, there are notabledifferences in spatial emphasis. Precipitationand streamflow anomalies in the tropical partsof South America and in the subtropical bandsacross North and South America are larger andmore consistent (spatially) in the CT responsesthan in the GR responses (Figs. 5a-b and Figs.5d-e). Precipitation and, especially, streamflow

responses to GR are larger in northwesternNorth America, Mexico, and Central Americathan are the corresponding CT responses.Again, we are struck by the remarkableinterhemispheric symmetries of ENSO-likeclimate variations in the western Americas.

As expected from their close ties to CT andGR, regression coefficients relating the rotatedPC series (tropical PC and North Pacific PC) toprecipitation and streamflow (Figs. 6a-b and6d-e) are similar to the corresponding patternsfor CT and GR. The precipitation andstreamflow patterns associated with the twoPC series are also remarkably similar to eachother, even though the two PCs areuncorrelated time series; the patterncorrelation between the precipitationcoefficients in Figures 6a and 6d is +0.41 andthe pattern correlation between streamflowcorrelations in Figures 6b and 6e is +0.42.Furthermore, three-fourths of the correlationshave the same signs in Figure 6b and 6e. Aswith CT and GR, positive tropical PC andNorth Pacific PC excursions are associatedwith wetter than normal southwestern UnitedStates and Paraguay-Uruguay-Patagonia, anddrier than normal tropical South America,northwestern North America, andsouthernmost South America. As with CT andGR, the decadal North Pacific PC is morestrongly expressed in northwestern NorthAmerican precipitation and streamflowvariations (and perhaps in southernmost SouthAmerica) than is the tropical PC, whereasprecipitation and streamflow in most otherregions more strongly express the tropical PCseries. These parallels and connections canalso be identified by computing the leadingPCs of streamflow and precipitation in theAmericas and then correlating the variations ofthose leading modes with global SSTs(Dettinger et al. 2000) to obtain SSTcorrelation patterns that closely resemble thecombined ENSO and decadal ENSO-like SSTpatterns studied here.

15

Surface air temperature

Positive CT and GR indices are associatedwith warmer than normal surface conditions inmuch of the Americas. The patterns indicatedfor CT temperatures (Fig. 5c) and GRtemperatures (Fig. 5f), however, are not assimilar as are the precipitation patternsdiscussed previously. Positive CTs (El Niños)are associated with warm temperatures inCanada, the western United States, CentralAmerica, much of tropical South America (onboth sides of the Andes), Paraguay-Uruguay-

Patagonia, and southeasternmost Brazil (seealso Diaz and Kiladis 1992). El Niños bringcool temperatures in the southeastern UnitedStates and in the eastern Amazon basin.Positive GRs are associated with warmer thannormal conditions in northwestern NorthAmerica and cooler than normal conditionsacross southeastern and eastern NorthAmerica. Temperature conditions elsewhereare not as closely related to the GR index asthey are to CT. These temperature relations arelargely echoed in regression coefficients forthe tropical and North Pacific PCs (Figs. 6c, f).

Figure 6. Same as Figure 5, but for the tropical PC index (a-c) and North Pacific PC index (d-f), 1915-88, respectively.

1357W 907W 457W607S

307S

07

307N

607N

B = -10mm/monDry

Tropical PC index, 1915-88

(a) Precipitation

1357W 907W 457W607S

307S

07

307N

607N

B = +10mm/monWet

North-Pacific PC index, 1915-88

(d) Precipitation

1357W 907W 457W607S

307S

07

307N

607N

B = -0.2CCool

(c) Temperature

1357W 907W 457W607S

307S

07

307N

607N

B = +0.2CWarm

(f) Temperature

1357W 907W 457W

r = -0.3Dry

(b) Streamflow

1357W 907W 457W

r = +0.3Wet

(e) Streamflow

16

As was the case with CT and GR, the spatialpatterns of temperature relations with the twoPC series are not as similar as are the relationsfor precipitation and streamflow; the patterncorrelation between Figures 6c and 6f is only+0.16.

These temperature responses to CT andGR are as expected from the atmosphericcirculation changes associated with the twoindices. When positive, both CT and GR tendto route anomalously southerly winds overwestern North America as circulations becomemore zonal over much of the North Pacific.The winds then tend to be anomalouslynorthward near the west coast of NorthAmerica from northern Mexico to Alaska (Fig.4a). Warmer than normal American tropicsderive from warmer SSTs and drier conditionsin tropical South America (Diaz and Kiladis1992). In subtropical South America, polewarddeviations of the westerlies also result inwarmer temperatures overall. Coolertemperatures in the southeastern United Stateswith positive CTs and along the entire easternUnited States are consistent with morestorminess and anomalously northerly windsthere, respectively, as indicated by the detailsof the 500 mb height regression coefficientsshown in Figures 4a and 4b.

Discussion

The interannual and decadal ENSO-likevariations of the Pacific climate systemdescribed here make important contributionsto modern climatic and hydrologic variationsof the Americas. For example, the numbers oftemperature, precipitation, and streamflowseries in which various amounts of varianceare attributable to a combination of the tropicaland North Pacific PCs are summarized in Table2. Because the PC series are uncorrelated, thepercentage of a series variance attributable tothem is simply the sum of the square of theirrespective correlations with the series. Morethan 30% of the variance is explained byvarious combinations of the two PC series inover a quarter of all the weather seriesconsidered here; almost a third of thetemperature series and a fifth of the streamflowseries owe half or more of their year-to-yearvariance to the interannual and decadal ENSO-like climate variations.

Perhaps the most important implication ofthese strong dependences upon the interannualand decadal ENSO-like climate variations isthat strong interhemispheric symmetries are tobe expected in, at least, temperature, preci-pitation, and streamflow variations. Otherforms of climate variation (e.g., North Atlantic

.

Table 2: Percentages of temperature, precipitation, and streamflow series, from sites indicated in Figure 5, with more than threshold fractions of variance explained by the combination of the

tropical and North Pacific PC series.

Fractions of variance

Variance threshold fraction

> 0.2 > 0.3 > 0.4 > 0.5

Temperature

58 42 38 30

Precipitation

35 27 25 14

Streamflow

39 35 30 21

17

climate oscillations) may be distinguishablefrom the ENSO-like variations simply by theirequatorially asymmetric influences on theAmericas.

Given the important role of these ENSO-like climate variations, a capacity to recognizetheir variations in paleoclimatic proxies isdesirable; reconstructions of these ENSO-likeclimate variations could form a groundworkfor assessing the stability and long-termcontributions of the Pacific climate system tothe modern climate of the Americas. Given thepossibly quite different mechanisms thatunderlie the interannual and decadal ENSO-like variations, it also would be desirable toreconstruct their separate contributions inpaleoclimatic proxies. Separate recon-structions would help us to understand howstable and reliable these important contributorsare to modern climate within their differingtimescales, in the face of natural and human-induced climate changes.

The interannual and decadal ENSO-likeclimate variations are influential over largeareas of the PEP 1 region, and both affecttemperatures, precipitation, streamflow, and(presumably) soil moisture—all of which maybe amenable to a variety of paleoclimaticreconstructions. At the same time, however,the interannual and decadal ENSO-likevariations may yield spatial patterns ofvariation of proxies that are difficult toseparate from each other. If the analyses ofinstrumental records presented in previoussections are any indication, it will be difficultto use paleoclimate proxies to distinguishbetween the two forms of Pacific climatevariability from patterns of temperature and,especially, precipitation proxies alone. Indeed,PC analyses of precipitation and streamflow inthe Americas (Dettinger et al. 2000) similarlyfound that interannual and decadalhydroclimatic variations associated withENSO-like atmospheric forcings could not beseparated orthogonally, whereas other, lessinfluential combinations of ENSO-like and

North Atlantic climate variability could be.Perhaps, whether ENSO related or decadal innature, the similar SST patterns associatedwith the ENSO-like variations cause suchrapid and similar responses from theatmosphere that the particular processes thatdetermine those SST variations matter little totheir atmospheric and continental outcomes(e.g., Deser and Blackmon 1993). In that case,it may be exceedingly difficult to be certainwhich of the ENSO-like climate forces areproviding the interannual or decadalfluctuations in North and South Americanproxies. We have limited this analysis toinstrumental records, which span roughly 100years; careful scrutiny of multiple-series,spatial networks of paleoclimate proxies todetermine how commonly theinterhemispheric ENSO-like pattern recurs andover what timescales it is present clearly iswarranted.

The changes in atmospheric circulationsand continental climate associated with theENSO-like climate variations representtemporary reorganizations of distributions ofheat and water on global scales. Because ofthis large scale and because of the strongsimilarities between the influences of ENSOand decadal ENSO-like climate variations onthe Americas, these climate variations alsohave important effects on the geochemicalmakeup of the atmosphere. Overall, thepositive (El Niño-like) CT and GR forcings areassociated with drier tropical conditions in theAmericas and mostly wetter extratropicalconditions (Figs. 5a-b and 5d-e), along withmostly warmer temperatures (Figs. 5c and 5f);these responses are part of a global pattern ofdrier tropical lands during El Niño-like periods(Dettinger et al. 2000). On interannualtimescales, these influences affect vegetationand surface fluxes of CO

2

in ways that areobservable on global scales (Siegenthaler1990; Keeling et al. 1995; Dettinger and Ghil1998). The strong similarities of effects ofinterannual ENSO and decadal ENSO-like

18

climate influences in the Americas (andelsewhere) suggest that similar ENSO-likevariations of global CO

2

concentrations mayoccur on decadal timescales.

Finally, the similarities between theinfluences of ENSO and decadal ENSO-likeclimate variations on the Americas, togetherwith their differing timescales, ensures thatsometimes CT and GR—and their respectiveprecipitation, streamflow, and temperaturepatterns—share the same signs and sometimethey are in opposition. When the signs of theindices agree, the precipitation, streamflow,and temperature patterns in Figures 5 and 6can be expected to reinforce each other; whenthe indices disagree, the patterns may interfereand reduce the overall precipitation,streamflow, and temperature responses.Constructive and destructive influences such asthese have been documented by Gershunovand Barnett (1998) and Gershunov et al.(1999) over the United States and resultspresented here indicate that similar influencescan be expected in many other regions ofNorth and South America.

Conclusions

During the twentieth century, threephenomena have dominated climate variationsin the Americas on interannual and decadaltimescales. The ENSO phenomenon dominatesglobal climate variations on interannualtimescales, and ENSO-like variations over thePacific Ocean basin and North Atlantic climatevariations dominate climate variations ondecadal timescales. In this chapter, historicaloceanic and atmospheric patterns of climatevariability associated with ENSO and ENSO-like variations of the Pacific basin are analyzedand their influences on the temperatures,precipitation, and hydrology of the Americasare compared.

Several climate indices that reflect theENSO interannual and ENSO-like decadalclimate variations have been developed in the

literature, including the SOI, Niño-3temperatures, and the CT index for the tropicalENSO variations, and the PDO index and so-called GR index for decadal ENSO-likevariability. In this chapter, another pair ofindices is developed on the basis of a PCanalysis of SOI, CT, and PDO; the resultingPC series are closely related to CT and GR, butthey were not formed by application of explicittemporal filters to separate ENSO and decadalENSO-like variations (as was GR). Byconstruction, the PC series are completelyuncorrelated with each other, to prevent doublecounting of climate variations when theirrespective contributions to regional climatevariations are compared. The CT and tropicalPC indices are two expressions of thetropically based, dominantly interannualclimate variations associated with ENSO.Compared to the interannual ENSO signal, theGR and North Pacific PC indices aredominantly decadal and are expressed mostfully in the subtropics and midlatitudes of bothhemispheres in the Pacific basin. A number ofdifferent mechanisms have been described inthe literature to explain the decadal characterof the extratropical Pacific SST variations—some directly related to ENSO and othersnot—and no single one of these explanations ispreferred by the scientific community at thistime; see the section on Sea-SurfaceTemperatures earlier in this chapter for someexamples of the alternatives underconsideration.

At the 500 mbar level in the atmosphere(about 5 km above sea level), positive phasesof the CT and GR, and their counterpart PCseries, are associated with anomalously lowheights (pressures) over the North Pacific atthe climatological position of the AleutianLow, stronger westerly winds, and southwarddiversion of westerlies and storminess acrossthe North Pacific basin. In the tropics, CT isassociated with the entire suite of ENSOteleconnections and atmospheric changes; GRis not expressed so clearly in the tropics (as

19

gauged by SSTs there). In the SouthernHemisphere, positive CT and GR, and PCs, areassociated with a dipole of 500 mb heightanomalies that also shifts westerlies andstorminess equatorward. The spatial patternsof 500 mb height anomalies associated withCT and GR are remarkably similar, reflectingthe qualitative similarities in SST patternsassociated with the indices, despite theirdiffering time series. The patterns ofatmospheric circulation changes associatedwith the indices also are remarkablysymmetric about the Equator in the Pacificsector.

Over North and South America, thesimilarities of the spatial distributions ofclimatic forcings associated with theinterannual and decadal ENSO-like variationsof the Pacific basin result in similarprecipitation and streamflow patterns that arebasically symmetric about the Equator in thewestern Americas, on both interannual anddecadal timescales. The interannual anddecadal ENSO-like patterns of precipitationand streamflow also are similar to each otheroverall, although more differences arise whenthe precipitation associations with the tropicalPC and North Pacific PC (which are by designcompletely uncorrelated) rather than with CTand GR (which are share some variability) areconsidered. Surface air temperature patternsassociated with the ENSO-like climateforcings may be somewhat less spatiallysimilar from timescale to timescale. Wetterthan normal conditions accompany positivephases of CT (El Niños) and GR in thesubtropical belts of both continents; drier thannormal conditions accompany them in much oftropical South America (east of the Andes) andin the midlatitudes of the northwestern UnitedStates and perhaps in southernmost SouthAmerica. At the hemispheric scale, positivephases of the CT (El Niños) and GR indiceshistorically have resulted in more consistentand stronger precipitation and streamflowresponses in the Americas than have negative

phases, except in the same midlatitude belts ofthe two continents. Temperatures in theAmericas have tended to be warmer overallduring positive phases of both CT and GR,except across the southeastern United States,eastern United States (for GR), and parts of theAmazon basin.

Together the interannual and decadalENSO-like climate variations are related tomore than half the year-to-year variance inone-third to one-fifth of the temperature andstreamflow series analyzed here. Presentunderstanding of the processes leaves open thepossibility that the interannual and decadalSST variations derive from very differentphysical processes with different timescalesand different vulnerabilities to long-termnatural and human-induced changes. Theatmospheric variations and climatic con-sequences for the Americas that are associatedwith the interannual and decadal SSTvariations may derive from quick physicalresponses to overall SST variation that dependrelatively little upon which SST process isdominant. Even if the atmosphere andcontinents prove to be slavish in theirresponses to SST variations, regardless of thesource of those variations, the climaticconsequences still would reflect the differentvulnerabilities to change of the various SSTprocesses. Thus, improved understanding ofthe processes represented by CT and GRshould be a high priority for present-climate,paleoclimate, and climate-change researchers.It is unfortunate, then, that the strong spatialsimilarities between the hydroclimaticoutcomes of historical ENSO-like climatevariations will probably impede most efforts toreconstruct, separately, these two dominantmodes of climate variation in the Pacific basinby paleoclimatic proxies from the Americas.The results presented here, however, compareENSO-like SST and climate variations; othermodes of SST and climate variation may beimportant in many locales and may be easier todistinguish from ENSO and ENSO-like

20

behavior of the global climate system. Finally,because this analysis is restricted to variationsduring the last century, the amplitude of theSST variations at work is only a few degreeslarge. Extrapolation to the larger SSTanomalies of past climates would bespeculative and, indeed, even the close spatialsimilarities of ENSO and decadal ENSO-likeclimate variations found here may break downunder larger SST forcings or under changingmean temperature conditions of the distantpast.

Acknowledgments

We are very grateful to Vera Markgraf fororganizing and shepherding the PEP-I Workshop inMerida, Venezuela; it was a remarkablyinformative and useful week. Discussions withDave Enfield, of the National Oceanic andAtmospheric Administration (NOAA) and theAtlantic Oceanographic and MeteorologicalLaboratory; Jose Marengo, Instituto Nacional dePesquisas Espacias; and Henry Diaz, NOAAClimate Diagnostics Center, contributed to theideas developed here. Comments by Dan Cayan,Scripps Institution of Oceanography, and MichaelEvans, Lamont-Doherty Earth Observatory, greatlyimproved this chapter. Access to streamflow,precipitation, and temperature data provided byDan Cayan and Larry Riddle at Scripps Institutionof Oceanography; to Brazilian streamflow seriesmade available by Jose Marengo, courtesy ofEletrobras and Eletronorte; and to 1948-1958 daily500 mbar height fields by Wayne Higgins of theNOAA Climate Prediction Center made parts ofthis analysis possible. In addition to NationalScience Foundation’s funding for the workshop asa whole, MDD’s efforts were supported by the U.S.Geological Survey’s Global Change HydrologyProgram and DSB’s efforts were supported bygrants from NOAA’s Office of Global Programs.

References

Allan, R., J. Lindesay, and D. Parker, 1996:

El Niño,Southern Oscillation and Climate Variability

.Collingwood, Australia: CSIRO Publishing, 405 p.

Barsugli, J.J., and D.S. Battisti, 1998: The basic effectsof atmospheric-ocean thermal coupling on midlatitudevariability.

Journal of the Atmospheric Science

s,

55

:477-493.Bitz, C.M., and D.S. Battisti, 1999: Interannual to

decadal variability and the glacier mass balance inWashington, Western Canada, and Alaska.

Journal ofClimate

,

12

:3181-3196.Cayan, D.R., M.D. Dettinger, H.F. Diaz, and N.

Graham, 1998: Decadal variability of precipitationover western North America.

Journal of Climate

,

11

:3148-3166.Deser, C., and M. Blackmon, 1993: Surface climate

variations over the North Atlantic Ocean duringwinter: 1900-1989.

Journal of Climate

,

6

:1743-1753.Deser, C., and J.M. Wallace, 1990: Large-scale

atmospheric circulation features of warm and coldepisodes in the tropical Pacific.

Journal of Climate

,

3

:1254-1281.Dettinger, M.D., and H.F. Diaz, submitted: Global

characteristics of streamflow seasonality andvariability.

Journal of Hydrometeorology

.Dettinger, M.D., and M. Ghil, 1998: Seasonal and

interannual variations of atmospheric CO

2

andclimate.

Tellus

,

50B:1-24.Dettinger, M.D., D.R. Cayan, H.F. Diaz, and D. Meko,

1998: North-south precipitation patterns in westernNorth America on interannual-to-decadal time scales:Journal of Climate, 11:3095-3111.

Dettinger, M.D., D.R. Cayan, G.J. McCabe, and J.A.Marengo, 2000: Multiscale hydrologic variabilityassociated with El Niño/Southern Oscillation. In Diaz,H.F., and V. Markgraf (eds.), El Niño and the SouthernOscillation--Multiscale Variability, Global andRegional Impacts. Cambridge: Cambridge UniversityPress, 113-146.

Diaz, H.F., and G.N. Kiladis, 1992: Atmosphericteleconnections associated with the extreme phases ofthe Southern Oscillation. In Diaz, H.F. and V.Markgraf (eds.), El Niño: Historical andpaleoclimatic aspects of the Southern Oscillation.Cambridge: Cambridge University Press, 7-28.

Diaz, H.F., and V. Markgraf (eds.), 1992: El Niño:Historical and paleoclimatic aspects of the SouthernOscillation. Cambridge: Cambridge University Press,476 p.

Ebbesmeyer, C.C., D.R. Cayan, D.R. McClain, F.H.Nichols, D.H. Peterson, and K.T. Redmond, 1991:1976 step in the Pacific climate: Forty environmentalchanges between 1968-75 and 1977-84. Proceedingsof the 7th Annual Pacific Climate (PACLIM)Workshop, California Department of Water Resources,115-126.

Eischeid, J.K., C.B. Baker, T.R. Karl, and H.F. Diaz,1995: The quality control of long-term climatologicaldata using objective data analysis. Journal of AppliedMeteorology, 34:2787-2795.

21

Eischeid, J.K., H.F. Diaz, R.S. Bradley, and J.D. Jones,1991: A comprehensive precipitation data set forglobal land areas. DOE/ER-69017T-H1, TR051, 81 p.

Enfield, D.B., and A.M. Mestas-Nunez, 2000:Interannual to multidecadal climate variability and itsrelationship to global sea-surface temperatures. InMarkgraf, V. (ed.), Present and Past InterhemisphericClimate Linkages in the Americas and Their SocietalEffects. San Diego: Academic Press, this volume.

Folland, C.K., and D.E. Parker, 1990: Observedvariations of sea surface temperature. In Schlesinger,M.E. (ed.), Climate-Ocean Interaction. New York:Kluwer, 21-52.

Folland, C.K., and D.E. Parker, 1995: Correction ofinstrumental biases in historical sea surfacetemperature data. Quarterly Journal of the RoyalMeteorological Society, 121:319-367.

Garreaud, R.D., and D.S. Battisti, 1999: Interannual andinterdecadal variability in the southern Hemisphere.Journal of Climate, 12:2113-2123.

Gershunov, A., and T.P. Barnett, 1998: Interdecadalmodulation of ENSO teleconnections. Bulletin of theAmerican Meteorological Society, 79:2715-2725.

Gershunov, A., T.P. Barnett, and D.R. Cayan, 1999:North Pacific interdecadal oscillation seen as factor inENSO-related North American climate anomalies. Eos(Transactions of the American Geophysical Union),80:25, 29, 30.

Graham, N.E., 1994: Decadal scale variability in the1970’s and 1980’s: Observations and model results.Climate Dynamics, 10:135-162.

Gu, D.F., and S.G.H. Philander, 1997: Interdecadalclimate fluctuations that depend on exchangesbetween the tropics and extratropics. Science,275:805-807.

Hurrell, J. W., 1995: Decadal trends in the NorthAtlantic Oscillation: Regional temperatures andprecipitation. Science, 269:676-679.

Jacobs, G.A., H.E. Hurlburt, J.C. Kindle, E.J. Metzger,J.L. Mitchell, W.J. Teague, and A.J. Wallcraft, 1994:Decadal-scale trans-Pacific propagation and warmingeffects of an El Niño anomaly. Nature, 370:360-363.

Jiang, S., F.F. Jin, and M. Ghil, 1995: Multipleequilibria, periodic and aperiodic solutions in a wind-driven, double-gyre, shallow-water model. Journal ofPhysical Oceanography, 25:764-786.

Jones, P.D., S.C.B. Raper, R.S. Bradley, H.F. Diaz, P.M.Kelly, and T.M.L. Wigley, 1986a: NorthernHemisphere surface air temperature variations: 1851-1984. Journal of Climate and Applied Meteorology,25:161-179.

Jones, P.D., S.C.B. Raper, and T.M.L. Wigley, 1986b:Southern Hemisphere surface air temperaturevariations: 1851-1984. Journal of Climate and Applied

Meteorology, 25:1213-1230.Kalnay, E., M. Kanamitzu, R. Kistler, W. Collins, D.

Deaven, L. Gandin, M. Iredelli, S. Saha, G. White, J.Woolen, Y. Zhu, M. Chelliah, W. Ebisuzaki, W.Higgins, J. Janowiak, K. Mo, C. Ropelewski, J. Wang,A. Leetmaa, R. Reynolds, R. Jenne, and D. Joseph,1996: The NCEP/NCAR Reanalysis Project. Bulletinof the American Meteorological Society, 77:437-471.

Keeling, C.D., T.P. Whorf, M. Wahlen, and J. van derPlicht, 1995: Interannual extremes in the rate of rise ofatmospheric carbon dioxide since 1980. Nature,375:666-670.

Kidson, J.W., 1988: Interannual variations in theSouthern Hemisphere circulation. Journal of Climate,1:1177-1198.

Latif, M., and T.P. Barnett, 1994: Causes of decadalclimate variability over the North Pacific and NorthAmerica. Science, 266:634-637.

Lees, J., and J. Park, 1995: Multiple-taper spectralanalysis: A stand-alone C subroutine. Computers andGeosciences, 21:199-236.

Mann, M.E., and J.M. Lees, 1996: Robust estimation ofbackground noise and signal detection in climatic timeseries. Climatic Change, 33:409-445.

Mantua, N.J., S.R. Hare, Y. Zhang, J.M. Wallace, andR.C. Francis, 1997: A Pacific interdecadal climateoscillation with impacts on salmon production.Bulletin of the American Meteorological Society,78:1069-1079.

Miller, A.J., and N. Schneider, 1998: Interpreting theobserved patterns of Pacific Ocean decadal variations.In G. Holloway, P. Muller, and D. Henderson (eds.),Biotic Impacts of Extratropical Climate Variability inthe Pacific. Proceedings of the Aha Huliko’a HawaiianWinter Workshop, University of Hawaii at Manoa, 19-27.

Mo, K.C., and M. Ghil, 1987: Statistics and dynamics ofpersistent anomalies. Journal of the AtmosphericSciences, 44:877-901.

Mo, K.C., and R.W. Higgins, 1998: The Pacific SouthAmerican modes and the tropical intraseasonaloscillation. Monthly Weather Review, 126:1581-1596.

Percival, D.B., and A.T. Walden, 1993: Spectral analysisfor physical applications—Multitaper andconventional univariate techniques. Cambridge:Cambridge University Press, 580 p.

Richman, M.B., 1986: Rotation of principalcomponents. International Journal of Climatology,6:293-335.

Siegnthaler, U., 1990: El Niño and atmospheric CO2.Nature, 345:295-296.

Slack, J. R., and J.M. Landwehr, 1992: Hydro-climaticdata network (HCDN): A U.S. Geological Surveystreamflow data set for the United States for the study

22

of climate variations, 1874-1988. U.S. GeologicalSurvey Open-File Report 92-129, 193 p.

Trenberth, K.E., 1990: Recent observed interdecadalclimate changes in the Northern Hemisphere. Bulletinof the American Meteorological Society, 71:988-993.

Trenberth, K.E., and D.A. Paolino, 1980: The northernhemisphere sea-level pressure data set trends, errors,and discontinuities. Monthly Weather Review,108:855-872.

Troup, A. J., 1965: The ‘southern oscillation’. QuarterlyJournal of the Royal Meteorological Society, 91: 490-506.

Wallace, J.M., E.M. Rasmusson, T.P. Mitchell, V.E.Kousky, E.S. Sarachik, and H. Storch, 1997: On thestructure and evolution of ENSO-related climatevariability in the tropical Pacific: Lessons fromTOGA. Journal of Geophysical Research, 103:14241-14259.

Zhang, Y., J.M. Wallace, and D.S. Battisti, 1997: ENSO-like interdecadal variability: 1900-93. Journal ofClimate, 10:1004-1020.