OCTOBER 1982 JEFFERY C. ROGERS AND HARRY VAN LOON 1375Spatial Variability of Sea Level Pressure and 500 mb Height Anomalies over the Southern Hemisphere JEFFERY C. ROGERSDepartment of Geography, The Ohio State University, Columbus 43210 HARRY VAN LOONNational Center for Atmospheric Research? Boulder, CO 80307(Manuscript received 2 December 1981, in final form 22 June 1982) ABSTRACT The spatial variability of mean sea level pressure (SLP) and 500 mb height anomalies over the SouthernHemisphere during summer (DJF) and winter (JJA) is determined using eigenvector analysis based on dallysynoptic maps from 1972 to 1979. The patterns of spatial distribution of pressure and height anomalies arefurther verified and examined by means of station data, and the eigenvectors are compared between theseasons and to those found for the Northern Hemisphere. The first eigenvector shows that midlatitude anomalies of SLP and 500 mb height are of an opposite signto those found over and around Antarctica. The pattern is highly barotropic and suggests strengthening andweakening of the zonal wind in alternating latitude belts. The 500 mb height differences are calculated forfive midlatitude to Antarctic station pairs using data from the late 1950's onward. These latitudinal heightdifferences are also used to describe the association between the hemispheric westerlies and the SouthernOscillation. They also suggest that the summertime westerlies strengthen and weaken simultaneously aroundthe hemisphere, whereas in winter this zonal symmetry is interrupted in the South America-AntarcticPeninsula region. In this region, wintertime variations in the westerlies are associated with the secondeigenvector, and large interannual variations in the SLP and latitudinal variations in storm tracks occur overthe Drake Passage. The second eigenvector is shown to be associated with the large standard deviations ofwinter mean temperature along the Antarctic Peninsula, to variations in sea ice duration at Laurie Islandin the South Orkneys, and to the Trans-Polar Index of Pittock, which describes the tendency for the surfacepolar vortex to be displaced either toward Tasmania or the Falkland-South Sandwich Islands region. It is concluded that temperature and circulation teleconnections, similar to those of the Northern Hemisphere, also occur in the Southern Hemisphere, and are associated with standing waves in the atmosphere.The small amount of land south of 35 -S probably accounts for the uniform variations in the westerlies acrossthe Southern Hemisphere, whereas, in the Northern Hemisphere, variations in the westerlies shown by theeigenvectors are largely confined to the oceans.1. Introduction The mean distribution of sea level pressure (SLP)and 500 mb height in the Southern Hemisphere hasbeen known for some time and is shown, for example,in the climatological atlas of the Southern Hemisphere by Taljaard et al. (1969). Many earlier studiesdescribe specific features of the circulation of the atmosphere over particular regions of the SouthernHemisphere, such as the semiannual wave in surfacepressure first pointed out by Reuter (1936) and Wahl(1942). This pressure wave is out-of-phase betweenmiddle and high latitudes (see, e.g., Vowinckel, 1955),and its areal distribution has been outlined bySchwerdtfeger and Prohaska (1956), Schwerdtfeger(1960, 1967) and van Loon (1967, 1971). A feature ~ The National Center for Atmospheric Research is sponsoredby the National Science Foundation.0027-0644/82/101375- ! 8508.50c 1982 American Meteorological Societyof the circulation at lower latitudes, the SouthernOscillation, has often been the subject of study beginning with Walker (1923). Meinardus (1938) andLamb (1959) paid special attention to the circulationat middle and high southern latitudes. In recent years, the number of papers dealing withthese and other features of the circulation and climates on the Southern Hemisphere has steadily increased. Kidson (1975a,b) showed the eigenvectorsof monthly mean pressure based on station data forthe period 1951-60. His analysis revealed a Walkertype of circulatio.n over the equatorial Atlantic Oceanand he found that the Southern Oscillation can berepresented by the first components of not only pressure but also temperature and rainfall. Pittock (1973,1980) and Trenberth (1976a) have devised regionalindices of the circulation. Trenberth's indices wereindicators of the zonal and meridional componentsof the flow in the Australia-New Zealand region and1376 MONTHLY WEATHER REVIEW VOLUM-110 - -o .. .o 4- 4- 4 ~,F ' .o.,~- ".,.......,-' '~ ..... : "~"'":'/'"' "'"'""'";,.;";"'""::':ii"..4- ....... .4- :. 4- 4- : '".,...,-" "'., ,.,"'- "., '.~ "! ..... t - ~. '~. .. ~,-.., d- -., ;- % '%~,,o', . z ,4, -. ,9~,~- '. 4. .,., 4. ', 4. ;, 4.. / o..~.o ,. .o- .. .. .. - ........ .. ...... ~ ~ '~c" " '' '., ;o ,- '.,, ./ ~ ,..~,~' . , -~. ., .. ,o .. . .,~ .. . . :' .... ". ."' '"'::" '"'.. ..... "... '-.-.' i ...... ! ..... ~1 ; 4-"./ 4..' '..4./,"'..,~ ..' .~ ~ : ' /: .... 2 '" "' " " '~" " ' ' 'S,..: .. ~.,. ~,~o'?*** ...,..,.... ... .:. .......... .............. /"i ........ :......-_q ..+ :' ........... ..' .... '...+ / '-... ~.'.. '..~ .~. .......... .~ : ,~-,. "C, >. ',~.'* - ",L. ? :' '"* '*d.......... 26 / ...... *"'. / '*'.. ' '. "'.,' ,.. ~ /+ i ~ i .... ~-*:" ....... ~**.. 23o~ i :+2-----~ ...... % i !4. i .4- : + ""~ i:'.~( ~ ,' ~ .' : ..,o. ; ... '~, -,~,oO4' ~4-,. .... .~ ...... ,""i""~". ..'" .:' :' i ~', "..,."... '::,...~'" ~' '..... .,' .-'"""'~'": .~-...,~ .... ~..:~,.'",,. .. ...",'"",. ". .... .,. - '; ' '- -. ~ .I.,' : ~ ... - ' '~- ~, .-' 4.- :. ..: ..... ; ............. -..-i;-.. ~.. .... .. : 22 .. 'i' "o--O o."~ '.. ;' ?. '. ' "~. . i . .-'--' - % .,' ... ~,.. ?. i'.. /' ~' :"--/ ..... ;'... ? / .. -. -4.':"-.... +,,, >,, .; '.../ - ?. "....' "....,...."' ~ ... z-% o..+ 7'-.. 4.,,,' ,.~ +". .. ..., ,, .., ,o,- '.., .,.' -~,: ,..:..'.. '.. .....-,.,:. .,' -... ..... ~.'.; - .. + "...*:i: '. '-. / ~.~' ...o,-' .. ,~, .-' '. .:e ' .. o ..- -.:,~.-'... + .'~ ..... * + '~ +.... ....~.; .~ ............ ! ....... .... ,.~?.-~ ...... :..~. .... ~ ~- !.,, ' -~H .:. .... ~ ! ; ~ ! ! + - ";'"-~--4- ...... :4. .: . .: -?.... : , 4-..' 4. :' "-?,.., --,.. - " '~r~s ,' ~ - .,.,.. .., ~ - --1'. ;' ~v': "~-. .'" 0''-... 4-..' 4- ..~ '.,-* ~ - -d" ' ~"" '"' ' "*" ' ~"' '""t~../ :: ......... f ................... :~ .......... ' ......... "~,,:, ,,~ ...... :'4. ; ~'-~ .~ ~-.. ....... .~.. .~'/ ' ....... -i.._; ........ ............... , ............. ..,. ........... ,. . . . 4. ............... ... ..// lqo. 1. Locations of the 112 grid points (Crosses) where daily SLP and'500 mb height data Wereavailable for the eigenvector analysis. The numbers are used in Table 1 to identify the stations wheredata were also used in subsequent analyses.they were also shown to be related at certain monthlylags to the Southern Oscillation (Trenberth, 1976b)as represented by SLP at Darwin and Pacific islandstations. The Tram-Polar Index (TPI) (Pittock, 1980)was devised because Kidson's (1975a) fourth eigenvector suggested an opposition in pressure betweensouthern South America and eastern Australia. Pitrock computed values of this index by subtracting themean monthly SLP at Stanley (52-S, 58-W) fromthat at Hobart (43-S, 147-E) and the numerical values represent the tendency for the polar vortex to bedisplaced toward one or the other of these stations-The L index (Pittock, 1973) is the monthly meanlatitude of the subtropical high pressure cell east ofAustralia and is highly correlated to the TPI. Studies of the southern circulation have usuallybeen impeded by the lack of data over the southernoceans. This drawback was not overcome, but wassomewhat reduced in the last decade with the production of daily synoptic maps by the AustralianBureau of Meteorology, starting in April 1972, whichincorporated satellite data to produce relatively consistent fields over the oceans- These synoptic data.have been used to study for example, the interannualvariability of the circulation (Trenberth, 1979, 1980b)and the sensible heat transport (van Loon, 1980) overthe Southern Hemisphere. It is the purpose of this study to show the Spatialvariability of SLP and 500 mb height over the Southern Hemisphere by means of the Australian data set,adding station data to substantiate and elaborateupon our results. The SLP and 500 mb height variability in space is shown by the eigenvectors of dailyanomalies during the extreme summer (DJF) andwinter (JJA) seasons. We will show how the principaleigenvectors are similar or different at the surface and500 mb, and in the two seasons. We will compare theSouthern Hemisphere eigenvectors to those of theNorthern Hemisphere (Wallace and Gutzler, 1981;Rogers, 1981b). We will also place the eigenvectorsOCTOBER 1982 JEFFERY C. ROGERS AND HARRY VAN LOON 1377in the context of observed climate and circulationvariability of the Southern Hemisphere over the last2-3 decades, to see, for example, how air temperaturechanges with changes in the atmospheric circulation.The eigenvector approach used here differs from thatused by Kidson (1975a,b), who combined 120 monthsof station data (1951-60) and presented the eigenvectors in which the annual cycle was not removed,as well as those in which the annual cycle was removed. Trenberth (1980a) performed an eigenvector analysis of the monthly mean 500 mb heights in the Australian data set until January 1978 (69 months). Hedid not present the eigenvectors, but discussed therelationship of the fourth and fifth patterns to othercirculation indices which he derived. Trenberth(198 l a) further explained that the empirical orthogonal function (EOF) approach was limited withSouthern Hemisphere monthly data, because thestandard EOF analysis could not detect progressiveoscillations in the flow and depict them in one eigenvector. Progressive waves predominate in theSouthern Hemisphere circulation and a combinationof several eigenvectors is necessary which then depicta sequence of events.2. Method The data used for the eigenvector analysis consistof the Australian 2300 GMT Southern Hemisphereanalyses, which are available in grid format from theNational Center for Atmospheric Research (NCAR).The daily data were from the summers 1972-73through 1978-79 and the winters 1972-79. Trenberth(1979, 1980b) and van Loon (1980) used these dataand found they were adequate for describing largescale features in space and time but that they are notsuited for computations of quantities which rely, fortheir accuracy, on the correctness of daily data atindividual points, such as eddy-flux calculations. Daily mean values of SLP and 500 mb hright werecalculated for the seven summers or eight winters foreach point on a 12 x 12 grid which is a subset of the47 x 47 grid of points originally available from theAustralian Bureau of Meteorology. Every fourth gridpoint in each direction was used. Pressure and heightdata were not available for eight tropical points ineach corner of the grid, reducing the. final networkto 112 variables. Daily departures from the seven oreight year daily means were then computed for eachgrid point. The daily departures of both seasons werecombined to produce one large set of 620 summercases (10 days missing) and one of 728 winter cases(eight days missing). Missing days were due to unreported data at some or at all grid points. Becausethere is only an eight-year sample, there is some dayto-day noise in the daily means which has not been TABLE 1. Name and location of the stations numbered in Fig.1, and the span of years for which data were used. The text indicatesthat at some stations only SLP data were used while at other stations 500 mb height or air temperature data were also used. 1. Amundsen-Scott (90-S) 1957-19792. McMurdo (78-S, 167-E) 1960-19793. Halley Bay (76-S, 27-W) 1956-19784. $ANAE (71-S, 3-W) 1958-19795. Mawson (68-S, 63-E) 1959-19796. Mimy (67-S, 93-E) 1958-19757. Argentine Island (65-S, 64-W) 1951-19788. Hope/Esperanza (63-S, 57-W) 1952-19789. Orcadas (61-S, 45-W) 1951-197810. Ushuaia (55-S, 68-W) 1951-1978'11. Grytviken (54-S, 37-W) 1951-197812. Macquarie Island (54-S, 159-E) 1951-197813. Campbell Island (52-S, 169-E) 1951-197814. Rio Gailegos (52-S, 69-W) 1952-197815. Stanley (52-S, 58-W) 1951-197816. Marion Island (47-S, 38-E) 1957-197917. Christchurch (44-S, 177-E) 1951-197818. Chatham Island (44-S, 176-W) 1951-197919. Hobart (43-S, 147-E) 1951-197820. Puerto Montt (41-S, 73-W) 1957-197921. Gough Island (40-S, 10-W) 1957-197822. New Amsterdam Island (38-S, 78-E) 1959-197923. Raoul Island (29-S, 178-W) 1957-197924. Rapa (28-S, 144-W) 1969-197925. Tahiti (18-S, 150-W) 1957-197926. Pago Pago (14-S, 171-W) 1966-197927. Cocos Island (12-S, 96-E) 1957-197928. Atuona (10-S, 139-W) 1969-1979filtered out of the data. The grid is shown in Fig. 1which, together with Table 1, also shows the locationsof stations used in the analysis and their length ofdata record. The eigenvectors are determined for the 112 x 112covariance matrix of the daily pressure or height departures, and they emphasize the variability overmiddle latitudes where variance of SLP and 500 mbheight is greatest (Trenberth, 1981 b). Kutzbach (1970)describes the analysis method used here and alsoshows that the eigenvectors produced from a covariance matrix are analogous to anomaly maps fromwhich it is possible to infer anomalous mean pressuregradients and winds. Associated with each eigenvector is a time series of coefficients, the sign of whichindicates whether the eigenvector pattern shown inthe figures below, or the pattern with the oppositesign (negative coefficient), is occurring on a given day.The eigenvectors shown below were contoured bycomputer; in some of the figures unrealistic acuteangles appear. The percentage of cumulative-explained variancefor the first six eigenvectors of SLP and 500 mb heightdepartures is given in Table 2. The first five eigenvectors explain more of the total variance in summerthan in winter, and the first eigenvector explains substantially more variance than the others.' The percentage of explained variance is small, due to the use1378 MONTHLY WEATHER REVIEW VOLUME 110 TABLE 2. Percentage of cumulative explained variance for thefirst six eigenvectors of daily SLP and 500 mb height departuresin summer and winter.Summer (Dec-Feb)Winter (Jun-Aug)SLP 500 mb SLP 500 mb1. 11.5% (11.5) 10.2% (10.2) 8.8% (8.8) 8.2% (8.2)2. 17.4% (5.9) 16.3% (6.1) 15.0% (6.2) 14.2% (6.0)3. 22.5% (5.1) 22.2% (5.9) 20.6% (5.6) 19.5% (5.3)4. 27.4% (4.9) 27.5% (5.3) 25.6% (5.0) 24.5% (5.0)5. 31.6% (4.2) 32.3% (4.8) 30.2% (4.6) 29.1% (4.7)6. 35.6% (4.0) 36.8% (4.5) 34.6%'(4.4) 33.4% (4.3)of daily data, which is characterized by the presenceof traveling synoptic-scale systems, and amounts toonly about one-half of the monthly variance explained by eigenvectors based on monthly means(Trenberth, 1980a). -The significance of the eigenvectors was examinedusing the method described by/Barnett and Preisendorfer (1978) in which the eigenvalues, associatedwith the eigenvectors, are tested to see if they differsignificantly from those generated by a matrix of random numbers with zero mean and unit variance.Such an analysis of our data suggests that, in bothseasons, at least six eigenvectors can be conside?edsignificantly different from those obtained from analyses of random number fields. The station data of SLP and 500 mb height consistsof monthly means available from Monthly ClimaticData for the World or from the World Weather Records, which are available on tape at NCAR.3. The first eigenvectors of SLP and 500 mb heighta. Results The first SLP eigenvectors based on seven summersof daily data is shown in Fig. 2a. The pattern is predominantly zonal and is one in which departures,around Antarctica, are opposite in sign to those ofthe latitudes equatorward of ~50-S. The area in theSouth Pacific, centered at 35-S, 120-W, where anomalies would be large, does not include any island stations, but the belt of large anomalies spanning theIndian Ocean includes several islands. The wavelikestructure of the zero line resembles the pattern ofridges and troughs at 50-S observed in the mean SLPdistribufon (Taljaard et al., 1969). The first 500 mb height eigenvector (Fig. 2b) ofsummer resembles its SLP counterpart in middle andhigh latitudes. At lower latitudes of the South Pacificand South Atlantic Ocean, the eigenvector loadingshave the same sign as those over Antarctica. Theselow latitude areas closely correspond to regions of 500mb eastedies which generally lie equatorward of 20-Sas shown by van Loon et al. (1971) and by Trenberth(1979) for Januaries. The pattern in Fig. 2b suggeststhat the westerlies tend to weaken north of 40-S,while they strengthen south of 40-S and the tropicaleasterlies increase, and vice versa (Trenberth, 1979). Table 3 gives the mean seasonal eigenvector timecoefficients computed from the daily data for the firsteigenvectors. The mean coefficients for SLP and 500mb heights for the summer eigenvectors are correlated at r --- +0.90 (N = 7). This indicates, in conjunction with Fig. 2, the barotropic nature of pressureand height anomalies shown by the first eigenvectorsthroughout the middle and high southern latitudes.Closer to the equator, the eigenvectors in Fig. 2 suggest that anomalies of SLP and 500 mb height areweak. The first eigenvectors of winter SLP and 500 mbheights based on eight years of daily data are shownin Fig. 3. The 500 mb height eigenvectors (Fig. 3b)resemble that of SLP (Fig. 3a) except over a portionof the central South Pacific where negative values areindicated. Both eigenvectors in Fig. 3 resemble thoseof summer, in that the ,pressure anomalies at middleand high latitudes are out-of-phase, and at 500 mbover the South Pacific Ocean those of midlatitudesare out-of-phase with tropical anomalies. In winter,the greatest midlatitude anomalies are found overChatham Island, rather than over the Indian Oceanas in Fig. 2. During the southern winter, the largepositive values across the Tasman Sea and New Zealand in Fig. 3b'coincide with the region of cut-offlows in blocking situations, which suggests that blocking there tends to occur with a weak polar vortex andstrong subtropical westerlies at ~25-30-S, east ofAustralia and north of New Zealand (see the meanwind distribution in van Loon et al., 1971; Trenberth,1979). The time coefficients of the winter first eigenvectors of SLP and 500 mb height (Table 3) are correlated at r = +0.93 (N = 8), but they are not highlycorrelated to the time coefficients of the summer firsteigenvectors. The first eigenvectors of SLP and 500 mb heightin Figs. 2 and 3 change little from winter to summer.The large degree of zonal symmetry of land and seain middle and high southern latitudes and the factthat Antarctica remains a heat sink and the ocean at TABLE 3. The seasonal means and standard deviations (in parentheses) of the daily time coefficients of the first eigenvectors ofSLP and 500 mb height.Summer (Dec-Feb)Winter (Jun-Aug)Year SLP 500 mb Year SLP 500 mb1972-73 -12 (35) 129 (372) 1972 -34 (39) -341 (437)1973-74 48 (41) 358 (466) 1973 25 (37) 131 (322)1974-75 -21 (31) -246 (392) 1974 -22 (48) -232 (505)1975-76 13 (31) -18 (390) 1975 -1 (40) -49 (492)1976-77 -49 (43) -585 (490) 1976 13 (28) 43 (289)1977-78 10 (40) 218 (410) 1977 -12 (37) -197 (476)1978-79 16 (31) 249 (367) 1978 -6 (55) 52 (517) 1979 40 (39) 624 (399)OCTOBER 1982 JEFFERY C. ROGERS AND HARRY'VAN LOON 1379 .o .- oo.OO'-% .o.--- "\. oO '"". -o'"' oO-.. - ,,. oo-.[,~. :~g -... o.. -..- '.....~..oO''--- 'x .?- --... -;::' '~'......... t...Eig.1 ,,,OJE..--.. ..,,. -.... p,--; ..... :...% .; -.......... ...o.~.t-o- '-- ., ..~. '" ..... Eig,1 DJF - " m__,."-~ -. o'2.. FIG. 2. Top (a): Eigenvector I of daily SLP departures during summers (DJF). Bottom (b): Eigenvector 1 of daily 500 mb height departures during summers. Solid lines indicate positive numerical values, dashed lines negative numerical.values. :;'%. "~::: ......1380MONTHLY wEATHER REVIEW- ! ....,?' /z '.% o., ..*.~. *,. ~ ~ - : , : ; ....... i ....... ~a0~" ', :,-' - ., ..,." ".,,. ." ...... 7 ......i : '~. ....,:'.; 90 \ ..........- **. ....... ~ ........... ~ ........... - ~SLP Eig.*l \~.~-~"~ ..........14:a-.-! ...... %" ~: ~ ',, ..,.o. : .~- ~ : ~.,, : .~ ~o \ .... :~'" ..... '"'. - ..........~o. 3. ~ in ~g- 2, but for ~nte~ (33A).0,:%'-.-;~/ ",,.. 3*:-*:: ,./ /3d/k.VOLUME 1 I0 l .:/OCTOBER 1982 JEFFERY C. ROGERS AND HARRY VAN LOON 1381 TABLE 4. The correlation coefficients (X 100) of 500 mb heights between six midlatitude and six Antarctic stations for data in summerand winter (in parentheses) since the late 1950's. Underlined coefficients are significant at the 95% and double underlined at the 99%confidence level. See Table I for the number of years available.McMurdo Amundsen-Scott Mawson Mirnyi SANAE Argentine IslandChatham Island -64 (-57) -55 (-46) -62 (-59) -74 (-68) -7__~0 (-50) -6~2 (+32)Invercargill -57 (-47) -33 (-05) -29 (-4~) -40 (-66) -57 (-66) -56 (+62)New Amsterdam Island -71 (-21) -65 (-64) -54 (-51) -63 (-88) -59 (-37) -86 (+17)Marion Island -33 (-01) -54 (+27) -32 (-26) -12 (-26) -34 (-21) -4~ (+30)Gough -27 (+07) -45 (-34) -61 (-35) -46 (-66) -41 (+04) -41 (-21)Puerto Montt -7___[1 (-21) -43 (-15) -35 (-52) -50 (-09) -47 (+03) -7.__[1 (-39)middle latitudes a heat source, in all seasons, are responsible for maintaining the characteristic circulation anomalies through the year. Trenberth and Paolino (1981) show that in the Northern Hemisphere,the first SLP eigenvectors in winter and summer bothcorrespond to a basic zonal circulation at high latitudes with anomalies of opposite sign found everywhere over the middle and high latitudes except overAsia. At 500 mb in the Northern Hemisphere, thereis much less zonal symmetry to the first eigenvectorsin winter and summer (Wallace and Gutzler, 1981;Rogers, 198 lb); cells of opposite sign of anomaly alternate across the hemisphere at low, middle and highlatitudes. In the Northern Hemisphere the eigenvectors of SLP and 500 mb height show more differencesbetween the extreme seasons, related to the largemonsoonal circulation changes which are associatedwith the seasonal switch in heat sources and sinks(Rogers, 198 lb). The opposition in pressure anomalies betweenmiddle and high southern latitudes is confirmed, fora longer period, by monthly mean 500 mb heightsfrom stations available since the late 1950's. Table4 shows the correlations of monthly mean 500 mbheights between six midlatitude and six Antarctic stations for summer and winter. In summer, all correlations are negative and many are statistically significant. However, in winter only four negative correlations are significant and some correlation coefficientsare positive, particularly those to Argentine Island.In view of the barotropic nature of SLP and 500 mbheight anomalies, results similar to those in Table 4are obtained for SLP station data (not shown). ForSLP, however, some correlations are positive in summer but more are positive in winter, particularly thoseto Argentine Island. These weaker correlations inwinter are due to the circulation anomalies associatedwith eigenvector 2 and are discussed below. The opposition in pressure and height anomaliesbetween middle and high latitudes is closely tied tofluctuations in the strength of the Southern circumpolar westerlies south of 40-S. When pressure ishigher than normal in midlatitudes, and lower thannormal around or over Antarctica (positive eigenvector time coefficients), the westerlies will be stronger than normal, and vice versa. Since changes in thestrength of the southern westerlies in middle and highlatitudes are closely linked to the half-yearly oscillations south of the subtropical high (van Loon, 1967,1972), the first eigenvector and its time coefficientspresumably play a role in determining fluctuationsin the amplitude of the semi-annual oscillation overa series of years. The alternation of positive and negative seasonalmean eigenvector coefficients in Table 3 representsthe interannual variability discussed by Trenberth(1979, 1980b). The standard deviations of the dailyeigenvector coefficients, also given in Table 3, are insome cases smaller than the absolute value of theseasonal mean coefficient. This indicates that a majority of the daily eigenvector coefficients had thesame sign as that of the mean, and it represents a'high degree of persistence in the anomalies of thegeneral circulation during the given season. For example, the eigenvector coefficients indicate that during summer 1973-74 for SLP, and winter 1979 for500 mb, stronger than normal westerlies persistedthrough the season, whereas during summer 197677 weak westerlies persisted at both sea level and 500mb. The persistent blocking south of New Zealandassociated with weak westerlies during summer 197677 was discussed by Trenberth (1979), and van Loonand Rogers (198 l b) described the large circulationdepartures of 1979. Other highly persistent SLP regimes existed during the winters of 1972, 1973, and1979. The results in Table 3 thus show some detailof the intraseasonal variability of the westerlies andeigenvector 1.b. The southern westerlies, air temperature varia tions, and the Southern Oscillation Using station data, we further examined the spatialand temporal variability of the southern circumpolarwesterlies. Latitudinal differences at 500 mb werecomputed by subtracting the mean seasonal 500 mbheights at Antarctic stations from heights at midlatitude stations at approximately the same longitudes.The long-term mean height differences for winter andsummer for these station pairs were then subtracted1382 MONTHLY WEATHER REVIEW VOLUME 110 TABLE 5. Coefficients of the detrended correlations of 500 mb height differences between five midlatitude/Antarctic station pairsduring summers (lower left) since 1957-58 and winters (upper right) since 1958. Underlined coefficients are significant at the 95%confidence interval and double underlined are significant at 99%. New Amsterdam/ Puerto Montt/Chatham/McMurdo Gough/SANAE Mirny Marion/Mawson Argentine IslandChatham/McMurdo 1Gough/SANAE 0.73New Amsterdam/Mirny 0.84Marion/Mawson 0.73Puerto Montt/Argentine - 0.62 Winter0.59 0.7~4 0.67 0.15I 0.7.~4 0.47 0.190.64 ! 0.78 0.250.74.. 0.8___[ ~ 0.240.51 0.7.~3 0.57 1 Summerfrom individual yearly values to create departuresfrom normal which represent the relative strength ofthe westerlies between the stations. The station painare Chatham Island/McMurdo, Gough Island/SANAE, New Amsterdam Island/Mirny,.Marion Island/Mawson, and Puerto Montt/Argentine Island.If a latitudinal difference could not be computed dueto missing data at the given Antarctic station, thenAmundsen-Scott data were used instead of those ofthe station; specifically, the departure from the meanheight difference between the midlatitude station andAmundsen-Scott was used. In other years, when dataat all Antarctic stations are available, 500 mb heightsat Amundsen-Scott are very highly correlated tothose at the relatively nearby coastal stations. Table 5 shows the correlations between 500 mbheight difference departures computed among the fivestation pairs during summer and winter after all,DEPARTURES FROM THE MEAN DjK 500 MB HEIGHTDIFFERENCES, CHATHAM I5. TO MCMURDO, ANTARCTICA-5O~La~x[udznat hexgh~ d~fference$ Trend tzne of he~gh~ differencesFIG. 4. Summer 500 mb height gradient departures from normal between Chatham Island and McMurdo, Antarctica, for 1957-58 through 1978-79.I YEAROCTOBER I982 JEFFERY C. ROGERS AND HARRY VAN LOON 1383trends had been removed from the time series. Thecorrelation coefficients are statistically significant forall but one station pair combination in summer,while in winter the coefficients are lower, particularlyfor the Puerto Montt/Argentine Island pair. The basicfeatures of the correlations in Table 4 reappear inTable 5; however, a high coefficient in Table 5 canoccur, despite a low correlation between a midlatitudeand an Antarctic station. The time series of the Chatham Island/McMurdo500 mb difference departures in summer, shown inFig. 4 for 1957-58 through 1978-79, is representativeof the four other station pairs in Table 5, because ofthe high correlations of the height differences betweendifferent station pairs. The mean Chatham Island/McMurdo summer gradient is 566 gpm (monthly ~= 67 gpm). The 500 mb departures from 1972-73through 1978-79 in Fig. 4 are correlated at r = +0.87to the mean time coefficients of the summer 500 mbfirst eigenvector (Table 3). The downward trend ofthe linear regression line through the data in Fig. 4is 3 gpm y-~ but is not statistically significant. Theslope of the regression line of the seasonal 500 mbheight different in Fig. 4 is typical of the slope for theother station pair height differences, and the negativegradient trend is associated with upward trends inheights at the Antarctic stations (Kraus, 1977; Trenberth, 1979) and height decreases at the midlatitudestations. The relationship between the strength of thewesterlies and Antarctic temperatures is shown inTable 6. The Amundsen-Scott base mean summersurface air temperature is given for six summers withdistinctly weak westerlies from normal, and eightsummers with distinctly strong westerlies from normal (see Fig. 4).2 Temperatures at Amundsen-Scottare noteworthy because they are characterized in eachseason by a bi-modal dist-ribution, which, as Table6 suggests, is associated with the strength of the circulation. Correlations of temperature levels withstrengths of the westerlies are characteristic of mostAntarctic stations. Fig. 5 shows the mean vertical temperature distribution over Amundsen-Scott during periods of weakand strong westerlies at the 500 mb level in ~ummerand winter. In summer, statistically significant meantemperature differences, such as those at the surfacebetween the strong and weak westerlies cases (Table6), occur vertically to the 250 mb level. Above the150 mb level in summer, mean temperatures becomewarmer with stronger westerlies at the 500 mb levelalthough mean heights (not shown) at AmundsenScott, extending to the 20 mb level, continue to be 2 Summers with weak westefl~s are 1957-58, 1964-65, 196667, 1971-72, 1974-75 and 1976-77. Strong we~efliesoccur in1958-59, 1959-60, 1961-62, 1962-63, 1969-70, 1973-74, 197778 and 1978-79. TABLE 6. The mean Amundsen-Scott surface air temperaturesfor summers 1957-58 through 1978-79, and winters 1958-1979,when the Southern Hemisphere 500 mb westerlies determined bystation pair height differences were relatively strong and relativelyweak.Amundsen-Scotttemperatures (-C)Summer Winter .Years with weak westerlies -30.6 -57.7Years with strong westerlies -33.0 -60.2Significance level of temperaturedifference 99.8% 99.8%statistically significantly higher during weak westerlycases than during strong flow cases. The mean summer temperature difference of 2.6-C in the inversionat the 650 mb level is slightly larger than the difference in surface air temperature (Table 6), and theinversion has only slightly greater intensity during theperiod of weak westerlies (4.0-C vs. 3.8-C). An association between the southern westerlies andthe Southern Oscillation (SO) is suggested by the results of Figs. 2 and 3, and has been suggested byKidson (1975a), Trenberth (1976b), van Loon andMadden (1981) and van Loon and Rogers (1981a).The spectrum of the detrended Chatham Island/McMurdo 500 mb height differences (Fig. 4) for the22 summers is shown in Fig. 6. The lag period was15 years with spectral information available at frequency intervals of 0.033 cycle y-I. The spectral estimate at 0.267 cycle y-~ (period of 3.7 years) wasfound to be statistically significant at the 95% confidence interval, when tested against both red andwhite noise continuums. It is possible that one of the15 spectral estimates might be expected to be significant by chance, when testing at the 95% confidenceinterval, but this periodicity lies within the 3-6 yearrange suggested by Trenberth (1976b), Wright (1977)and Julian and Chervin (1978) as being characteristicof the SO. The mean summer 500 mb height differencesacross the five midlatitude/Antarctic station pairs allhad low negative correlations to SLP at Cocos Island(an Indonesian center of the SO) and positive correlations to SLP at Tahiti (a Pacific center of the SO)during the occurrences of the High/Dry and Low/Wet years of the SO (defined by van Loon and Madden, 1981). This suggests some association betweenthe strength of the high latitude hemispheric westerlies and the extreme events of the SO, although onlytwo of the 10 correlations were statistically significant(Gough/SANAE, r - +0.65, and Marion/Mawson,r = +0.68, both to Tahiti) and all correlation coefficients were non-significant when non-extreme SOevents are included in the analysis. However, the results of van Loon and Madden (198 l) and van Loonand Rogers (198 l a) indicate that the westerlies rep1384MONTHLY WEATHERI I I IWinter and Summer Mean Air Temperaturesat Scott Base (90-S) During Periods ofRelatively Strong and Weak SouthernWesterliesREVIEWWinter (JJA) Summer (DJF)Stro~/Veak-80 -~0 -60 -50 -40 ,15203O50'608015025O~ 30035O40045O$0055060O650700 (-C)FIG. 5. Winter and summer mean air temperatures at Amundsen-Scott during years with relatively strong and relatively weak southern westerlies at the 500 mb level. 'VOLUME 110resented in this study by station pair height differences span too many parallels of latitude to adequately assess the relationship between the SO andcenters of high variability within the southern westedies; the question of control of one by the othercannot be made here. Station pair 500 mb height differences were alsocalculated for several low latitude stations includingPago Pago/Raoul, Tahiti/Rapa and Atuona/Rapa,and they were compared to those of higher latitudes.They showed, ~as do the eigenvectors in Figs. 2b and3b, that the South Pacific low latitude westerliesweaken, while the midlatitude westerlies stiengthen,and that in winter the strong westerlies east of Australia weaken when the Chatham/McMurdo gradientintensifies; this is also a characteristic of the changesbetween the extremes of the SO (van Loon and Rogers, 1981a). During winter the Marion/Mawson station pair hasthe fewest missing data and, as Table 5 indicates, itis fairly representative of the temporal variations ofthe height difference departures between the otherstation pairs, excluding Puerto Montt/Argentine Island. Fig. 7 shows the 500 mb height gradient departures for Marion/Mawson from the winter meanof 498 gpm (a = 64 gpm). A slight upward trendOCTOBER 1982->JEFFERY C. ROGERS AND HARRY VAN LOON SPECTRUM O; DJ; MEaN See mb HEIGHT DIFFERENCES BETUEEN CH~?H~ 25. and ~C~URDO,1958-19791385E$TIHTE3eoo2000 Snec~r.t e~m~. ','7'i'~ 9S~ ~gnificance teuet of red no~e continuum-'- ~ed noise con~inuu~I I I I I I 0 0.! 0.2 0.3 0.4 0.5 FR-(iUENC- -c~ctes per year)FIG. 6. Spectrum analysis of the Chatham Island/McMurdo 500 mb height gradient departures oF Fig. 4. Units along the ordinate arc squared gcopotcntial rectors.(Table 6) exists in the regression line through the datain Fig. 7, mainly due to the first and last value in theseries. This non-significant trend is noted at all stationpairs, except Chatham Island/McMurdo where thetrend is downward and where data from McMurdowere not available after 1974. The departures in Fig.7 for 1972-79 are correlated at r = +0.83 to the winter500 mb eigenvector coefficients (Table 3). Table 6shows Amundsen-Scott surface air temperatures arealso a significant indicator of the 500 mb height vortex strength in winter? At upper levels over Amundsen-Scott (Fig. 5), winter mean temperatures remainhigher during periods of weak westerlies up to the 50mb level, although mean temperature differences arenot significant above 350 mb.4. The second eigenvectorsa. SLP in summer and winter The second eigenvector of summer SLP (Fig. 8a)indicates that an opposition in pressure exists between 3 Based on all station pairs but Puerto Montt/Argentine Island.The winters of weak westerlies were 1959, 1962, 1963, 1964, 1974,1977 and strong westerlies occurred in 1960, .1966, 1969, 1970,1971, 1979.the Falkland/South Sandwich Island region and thearea south of New Zealand. With the exception ofthe former area, SLP anomalies around Antarcticaare of the same sign as those east and south of NewZealand. Using station data since 1948, 16 correlationcoefficients were calculated for summer SLP betweenour New Zealand area stations (Chatham, Mcquafie,and Campbell Islands, and Hobart) and four SouthAmerican sector stations (Ushuaia, Rio Gallegos,Grytviken and Stanley). All monthly correlation coefficients were negative, as the eigenvector based ondaily data suggests, but only Campbell/Grytviken (r-- -0.39) was statistically significant. Summer SLP eigenvector 2 is associated with variations in the direction of meridional flow over southern South America, and with variations in the direction of zonal flow across the Antarctic Peninsula, NewZealand and Ross Sea area. Lower pressure in the leeof the Andes and over the Antarctic Peninsula wouldtend to be associated with higher than usual pressureover and south of New Zealand. The seasonal meantime coefficients of eigenvector 2 are shown in Table7. The coefficient means are never as large as are theirstandard deviations suggesting only weak persistencein the daily pressure distributions. Table 8 shows the correlation between mean sum1386 MONTHLY WEATHER REVIEW~EPAR?UR-5 FROM THE ~-~N JJA S00 MB HEIGHTDIFFERENCES, ~ARION 15. TO MAUSON, ANTARCTICAVOLUME 1 10ETER0-SO-100~Letitudinat height differences Trend fine of height differences I ) I I 1960 1965 1970 197S YEAR~G. 7. Winter 500 mb heist ~adient-depanures ~om norm~ ~tween Marion Island and Mawson, Anta~tica, ~r 1958-1979.Imer surface air temperature at stations near theanomaly centers associated with the second summerSLP eigenvector. These large-scale teleconnections ofair temperature result because, according to Fig. 8a,more southerly flow over Hobart is associated withmore northerly flow over Chatham Island and Ushuaiabut with southerly flow over Grytviken, and viceversa. The second summer eigenvector is associated withthe Trans-Polar Index (TPI) defined by Pittock(1980). The TPI is the SLP anomaly difference between Hobart and Stanley. During summer, the negative correlation in SLP between these stations since1948 (r = -0.25) is not the highest.among the eightstations cited above. The second eigenvector of SLP during winter (Fig.8b) shows two large cells of the same anomaly signcentered over an area southeast of Chatham Islandand over the Scotia Sea, with an area of opposite signpartially wedged between them in the southeasternSouth Pacific Ocean. Together with the negative values over the Indian Ocean, these patterns outlinelarge-scale meridional flow components in contrastto the more zonal eigenvector 1. To a large degree,this holds true for eigenvector 2 in summer as well.The anomaly over the Scotia sea, suggestive of largeinterannual variability there, is of interest in Fig. 8b.Cyclones traveling from the South Pacific into theSouth Atlantic Ocean frequently pass through theDrake Passage (Taljaard, 1967), although they sometimes stagnate west of the Antarctic Peninsula. Taljaard (1967) and Trenberth (198 lb) report evidenceof considerable interannual variability of cyclonetracks in this area; Taljaard shows that there werefrequent cyclones across the Passage in July 1957 butduring July 1958 no cyclones passed between 56 and74-S. Table 7 indicates that, in recent years, morelows-occurred over the Antarctic Peninsula in thewinter of 1976 than during the winter of 1973. AtArgentine Island (65-S) and at Orcadas base (61 -S)on Laurie Island, the winter mean SLP in 1976 was1.9a and 3.6tr below the long-term means, respectively, but 0.7a and 1.0a above the mean in 1973.The SLP at Chatham Island and Christchurch was0.ga below the mean at both locations in 1976 but2.0~ and 1.7tr, respectively, above the mean in 1973.The standard deviations of the time coefficients of thesecond SLP eigenvectors (Table 7) are larger in winter,than in summer. The source region for cyclones moving through theDrake Passage in winter during the IGY was theSouth Pacific Ocean east of New Zealand (TaljaardOCTOBER 1982 JEFFERY C. ROGERS AND HARRY VAN LOON 1387Eig.2 ..... DJF ! t; ..'97'-.,....: -*' ',o,,J ',. ';, ',. .~0 ~... '.'~,.- ' o~ ~, \\ ....... i - : ~, ~. ..... q .; ........ ~ .......... ~ ' 'i--4 4 .......... [---~,-?--~ ........ ~ i j/.FIG. 8. (a) Eigenvector 2 of daily SLP departures during summers. (b) Eigenvector 2 of daily SLP departures during winters.;'....1388MONTHLY .WEATHER REVIEWVOLUME 1 10 TABLE 7. The seasonal means and standard deviations (in parentheses) of the daily time coe~cients of the second SLP andsecond 500 mb height eigenvectors.Summer WinterYear SLP 500 mb Year SLP 500 mb1972-73 2 (32) -55 (283) 1972 -5 (34) 2 (430)1973-74 -17 (24) 118 (205) 1973 -24 (33) -179 (360)1974-75 -7 (27) 70 (244i 1974 0 (32) 44 (274)1975-76 -9 (24) -I (233) 1975 6 (25) 21 (300)1976-77 9 (22) -37 (236) 1976 24 (36) 92 (356)1977-78 11 (35) 29 (250) 1977 -14 (31) -15 (315)1978-79 11 (28) -127 (218) 1978 5 (38) 80 (376) 1979 6 (44) -26 (372)and van Loon, 1962), where Fig. 8b suggests thatanomalies of the same sign as over the Drake Passagewill occur. Longer-term SLP data in winter show positive but small correlations between stations. Longer term station SLP data suggest a greaterdegree of correlation between Hobart/Stanley (r= -0.32) and Hobart/Grytviken (r = -0.47; for 28cases each) than is suggested by Fig. 8b. The stationdata indicate that although the TPI, to a large extent,is represented by the second SLP eigenvectors, therelationship between them is not as apparent in winter (Fig. 8b) as in summer (Fig. 8a). Table 8 showsthe teleconnections of-mean winter surface air temperature across the Southern Hemisphere which areassociated with the anomalous flow implied by thesecond winter SLP eigenvector. In winter, the largest standard deviations of meansurface air temperature in the Southern Hemisphereoccur around the Antarctic Peningula region andWeddell Sea, with a = 3.4-C at Argentine Island(1951-78) being the largest at a reporting station. Thepressure variations associated with SLP eigenvector2 (Fig. 8b), and particularly the TPI, are the majordeterminants of these large temperature fluctuations.Numerical values of the TPI were calculated for winters 1951.- 1978 and the mean winter air temperatureswere calculated for the two sets of years when indexvalues were either >+1 mb or <-1 rob.4 Mean temperatures for the positive cases (relatively lower pressure anomalies at Stanley), and negative cases (relatively lower pressure anomalies at Hobart) were compared using a two-tailed Student t-test. The resultsare shown in Table'9. Table 9 shows that the largest winter temperaturedifference associated with opposite phases of the TPIoccurs at Argentine Island (3.22-C) where the largeststandard deviation of mean temperature is found. 4 Index values greater than +1 mb occurred in the winters of1951, 1957, 1960, 1965, 1966, 1969, 1975, 1976, 1977 while thoseless than-1 mb occurred in 1952, 1953, 1955, 1956, 1962, 1963,1968, 1970.North of 60-S, where winter temperature standarddeviations decrease sharply, the temperature differences are smaller, but they are still statistically significant at Ushuaia and Stanley. The cause of these large winter air temperaturefluctuations in the Antarctic Peninsula/southern SouthAmerica region is partly deduced by examining Fig.8b. Wind direction reversals associated with the opposite modes of the eigenvector would lead to flowcoming either from Antarctica and the ice-coveredWeddell Sea or from the more open water to thenorth, depending on the relative SLP anomaly nearthe Antarctic Peninsula. Higher relative pressure nearArgentine Island, where the eigenvector loadings area maximum in Fig. 8b, is associated with lower temperatures as cold air is swept northward over ice surfaces. Near the Falkland islands, the SLP anomalyis weaker or even opposite in sign because of the proximity to the wave cyclones along the equatorwardedge of the polar outbreaks. Sea ice conditions, asreported by Schwerdtfeger et al. (1959; and subsequently updated by Schwerdtfeger) at Scotia Bay onLaurie Island in the Orkneys, are also highly variable.The average duration of ice at Scotia Bay (i.e., notonly during winter months) was 106 days (e = 49)for negative TPI winters, and 158 days (a = 57) forpositive TPI winters. Although the difference of 52days for 14 available cases was only significant at the90% confidence level, we note that more severe iceconditions tend to accompany' lower pressure at Stanley and a positive TPI. In summary, SLP eigenvector 2 in winter and summer represents both interannual variability of pressure in both seasons (Table 7) and variability withina season of the polar highs and storm tracks near theDrake Passage.b. The second eigenvectors of SO0 mb height The second eigenvector of summer 500 mb height(Fig. 9a) shows a pattern of anomalies in which thedepartures to the southwest and southeast of Australiaare out of phase with those south of Australia. Thetime coefficients of the eigenvector (Table 7), andseasonal mean height anomaly maps indicate thatduring 1973-74, heights were relatively high to thesoutheast and southwest of Australia, while during1976-77 negative height anomalies occurred in thoseregions. Fig. 9a suggests a three-wave pattern withdomination of wavenumber one. The seasonal average summer 500 mb second eigenvector coe~cients(Table 7) are correlated to those of SLP at r = -0.73(n ~ 7), suggesting some relationship between thepatterns in Figs. 8a and 9a, although this correlationcoefficient is not significant. The second eigenvector of 500 mb height in winter(Fig. 9b) is also a multicellular pattern of anomaliesof opposite sign extending across the hemisphere beOCTOBER I982 JEFFERY C. ROGERS AND HARRY VAN LOON 1389 TABLE 8. Correlations between the mean summer and winter surface air temperatures at places near the anomaly centers associatedwith the second eigenvector. Underlined and double underlined correlation coefficients are significant at the 95% and 99% confidenceintervals, respectively.Summer: Hobart Adelaide Chatham Island Oreadas Grytviken UshuaiaHobart 1.00 +0.67 -0.48 +0.28 +0.38Maccluarie Island -0.15 -0.52 +0.64 +0.14 +0.01 Tristan da Winter: Hobart Adelaide Chatham Island Orcadas Cunha -0.61 +0.03Amundsen-Scott SANAEHobart 1.00 +0.58 +0.04 -0.42 +0.54Campbell Island -0.18 -0.08 +0.45 +0.65 -0.52+0.16 +0.34-0.37 -0.30tween 30 and 70-S. This eigenvector shows a threewave pattern in the 500 mb circulation. Standingwave 3 reaches its greatest amplitude between latitudes 40-60-S during Southern winter (van Loon andJenne, 1972; Trenberth, 1980b). The longitude of thefirst ridge lies at ~50-S, 45-E. The implied meridional flow in Fig. 9b, when we treat the eigenvectorsas height anomalies, is very similar to the meridionalflow shown in van Loon (1972) for 50-S in July. Flowof one direction between longitudes 0-45-E, 80160-E, and 145-70-W occurs simultaneously whenflow from the opposite direction takes place elsewhere. The pattern in Fig. 9b thus suggests the interannual (Table 7).and day-to-day fluctuations ofstanding wave 3. The coefficients of the second eigenvector in the winter 500 mb heights (Table 7) aresignificantly correlated to those of the second winterSLP eigenvector (Fig. 8b) at r = +0.83. Argentine Island mean winter 500 mb heights arenegatively correlated to heights at other Antarctic stations since the late 1950's. The most significant correlation, r = -0.73 (99% confidence level), occursbetween mean winter 500 mb heights at ArgentineIsland and SANAE, and this out-of-phase height relationship appears in the second 500 mb eigenvector(Fig. 9b).5. Summary The first and second eigenvectors of daily SLP and500 mb height anomalies on the Southern Hemisphere for summers and winters between June 1972and August 1979 have been shown. The first eigenvector of SLP and 500 mb height on the SouthernHemisphere indicates a tendency for west winds toweaken (easterlies to strengthen) when the westerliesstrengthen (easterlies weaken) in neighboring, zonally-oriented latitude belts. This is different from whathappens on the Northern Hemisphere, where the alternation of strengthening and weakening winds associated with the primary eigenvectors is confinedmainly to the North Atlantic and North PacificOceans. The difference is associated with the difference in zonal symmetry (stationary waves) on the twohemispheres, which is related to the existence of uninterrupted ocean on the Southern Hemisphere in thelatitude of greatest west-east alternation of land andsea on the Northern Hemisphere. With the available station data, it is possible toshow some evidence of uniform hemisphere-wide interannual fluctions in the 500 mb westerlies (Table5), as represented by height differences between station pairs. One exception to this occurs over theeastern Pacific-South America region, where, in winter, the strength of the westerlies is poorly Correlatedto the 500 mb westerlies between the other stationpairs. This exception seems related to another standing oscillation of the southern atmosphere (Fig. 8b)in which SLP anomalies east of New Zealand havethe same sign as those over the Drake Passage. Thisoscillation is associated with latitudinal shifts in thestrongest westerlies in the longitudes of southernSouth America and the Antarctic Peninsula, andsince it is a major feature of the second eigenvectorin SLP and 500 mb height (Fig. 9b), it is presumablynot related to the westerlies elsewhere to the extentthat they are dominated by the first eigenvector. Inthe southern summer, three longitudinal oscillationsin atmospheric mass are thus known to occur predominantly over the North and South Pacific Oceans,between the Bering Sea and Antarctica: the variationin the mean position of the mean Aleutian Low associated with the North Pacific Oscillation (Rogers,' TABLE 9. Mean winter air temperature (-C) differences betweenyears when the Trans-Polar Index was positive and years when itwas negative from 1951 to 1978 (positive minus negative).Temperature Significancedifference level of(-C) differences (%)Rio Gallegos (52-S, 69-W) -0.7Stanley (52%, 58-W) -0.5Ushuaia (55-S, 68-W) - 1.4Grytviken (54-S, 37-W) -0.5Orcadas (61-S, 45-W) -2.3Hope/Esperanza (63-S, 57-W) -2.6Argentine Island (65-S, 64-W) -3.2Halley Bay (76-S, 27-W)+1.6Amundsen-Scott (90-S) - 1.4909599.8909599998O8O1390 MONTHLY WEATHER REVIEW VOLUME 110 /.t' ,o \ '~ ..... ':" .........../500mb Eicj.... _2. _D~'3_'.' t-~ ~ -- o~.. ,, ,, ,~ / ~, '%".... ',.~. ~'.. ~/i 9O ~........ .~ ........... - ........... ~, ......... :, /500rob Eig~_2...F~G. 9. As in Fig. 8, but for 500 mb height departures..,2 / t'?oaOCTOBER 1982 $EFFERY C. ROGERS AND HARRY VAN LOON 13911981 a); the Southern Oscillation; and the alternationin mass over the South Pacific Ocean associated withthe second SLP eigenvector (Fig. 8a). In the Northern Hemisphere, air temperatures between stations which are distant from one anotherare often highly correlated in association with recurring patterns of pressure and height anomalies (vanLoon and Rogers, 1978; Rogers 198 la). Tables 4 and8 indicate that similar correlations in seasonal meantemperatures occur between widely separated stationsin the Southern Hemisphere. Tables 6 and 9 alsoshow that significant differences occur in seasonalmean surface air temperatures between the periodsof opposite anomalies in the southern circulation,associated with both eigenvectors 1 and 2. The seasonal mean temperature differences, indicated in Table 6 and 9, are relatively small when compared tothose which occur in association with the periods ofwintertime extremes of the North Atlantic Oscillation(6-C differences over Greenland and northern Europe) and the North Pacific Oscillation (15-C differences over western Canada). The eigenvectors described in this paper bear someresemblance to those of Kidson (1975a), which wereobtained by analysis of station SLP data. Except overparts of Australia, Africa and the southeastern IndianOcean, Kidson's second eigenvector shows an opposition in pressure between middle and high latitudes similar to the first eigenvector in the presentanalysis. Kidson's fourth eigenvector bears some resemblance to the eigenvectors in Fig. 8, although heused consecutive monthly data over a 10 year period,and there is some suggestion here that the secondeigenvector pattern may vary between seasons. Comparing the two sets of eigenvectors, it appears that thewavelike structure of the anomalies is better resolvedwith the Australian Bureau of Meteorology datawhere satellite coverage is incorporated over theoceans, than with eigenvectors based on the scantobservational network. Acknowledgments. We thank Margaret Eccles ofMet Tech Enterprises, Boulder, CO, for programmingassistance. The National Center for AtmosphericResearch, which is sponsored by the National ScienceFoundation, provided computer time and support forJCR in this research. In particular, Greg Walters ofNCAR helped with the tapes. This material is basedupon work supported by the Division of AtmosphericSciences, National Science Foundation under GrantATM-8006330. W. Schwerdtfeger of the Universityof Wisconsin, Department of Meteorology providedupdated Scotia Bay sea ice data.REFERENCESBarnett, T. P., and R. W. Preisendorfer, 1978: Multifield analog prediction of short-term climate fluctuations using a climate state vector. J. Atmos. Sci., 35, 1771-1787.Julian, P. R., and R. M. Chervin, 1978: A study of the Southern Oscillation and Walker Circulation phenomenon. Mon. Wea. Rev., 106, 1433-1451.Kidson, J. W., 1975a: Eigenvector analysis of monthly mean sur face data. Mon. Wea. Rev., 103, 177-186.--., 1975b: Tropical eigenvector analysis and the Southern Os cillation. Mon. Wea. Rev., 103, 187-196.Kraus, E., 1977: Subtropical droughts and cross-equatorial energy transports. Mon. Wea. Rev., 105, 1009-1018.Kutzbach, J. E., 1970: Large-scale features of monthly mean Northern Hemisphere anomaly maps of sea-level pressure. Mon. Wea. Rev., 98, 708-716.Lamb, H. H., 1959: The southern westerlies: A preliminary survey; main characteristics and apparent associations, Quart. J. Roy. Meteor. Soc., 85, 1-22,Meinardus, W., 1938: Klimakunde der Antarktis. Handbuch der Klimatologie, W. KiSppen and R. Geiger, Eds., Gebriider Borntr/~ger, 133 pp.Pittock, A. B., 1973: Global meridional interactions in the strato sphere and troposphere. Quart. J. Roy, Meteor. Soc., 99, 424 437.---, 1980: Patterns of climatic variation in Argentina and Chile- I. Precipitation, 1931-60. Mon. Wea. Rev., 108, 1347-1361.Reuter, F., 1936: Die synoptische Darstellung der l/2-j'~hrigen Druckwelle. Verrff Geophys. Inst., Univ. Leipzig, 7(5), 257 295.Rogers, J. C., 1981a: The North Pacific Oscillation. J. Climatol., 1, 39-57.--, 198 lb: Spatial variability of seasonal sea level pressure and 500 mb height anomalies. Mon. Wea, Rev., 109, 2093-2106.Schwerdtfeger, W., 1960: The seasonal variation of the strength of the southern circumpolar vortex. Mon. Wea. Rev., 88, 203 208. -, 1967: Annual and semi-annual changes of atmospheric mass over Antarctica. J. Geophys. Res., 72, 3543-3547. , L. M. de la Canal and J. Scholten, 1959: Meteorologia des criptiva del sector Antartico $udamericano. Inst. Antard. Ar gent. Publ., No. 7,425 pp. [Available from Instituto Antarctico Argentino, Cerrito 1248, Buenos Aires].--, and F. Prohaska, 1956: Der Jahresgang des Luftdrucks auf der Erde und seine halbj'fihrige Komponente. Meteor. Rundsch., 9, 33-43.Taljaard, J. J., 1967: Development, distribution and movement of cyclones and anticyclones in the Southern Hemisphere during the IGY. J. Appl. Meteor., 6, 973-987. , and H. van Loon, 1962: Cyclogenesis, cyclones and anti cyclones in the Southern Hemisphere during the winter and spring of 1957. Notos, 11, 3-20. , H. L. Crutcher and R. L. Jenne, 1969: Temperatures, dew points and heights at selected pressure levels. Climate of the upper air: Southern Hemisphere. Vol. I, NAVAIR 50-1C 55, Chief of Naval Operations, Washington, DC, 135 pp. [Available from Naval Weather Service Command, Washing ton Navy Yard, Bldg. 200, Washington, DC 20309].Trenberth, K. E., 1976a: Fluctuations and trends in indices of the southern hemispheric circulation. Quart. J. Roy. Meteor. Soc., 102, 65-75. , 1976b: Spatial and temporal variations of the Southern Os cillation. Quart. J. Roy. Meteor. Soc., 102, 639-653. , 1979: Interannual variability of the 500 mb zonal mean flow in the Southern Hemisphere. Mon. Wea. Rev., 107, 1515 1524. , 1980a: Atmospheric quasi-biennial oscillations. Mon. Wea. Rev., 108, 1370-1377. , 1980b: Planetary waves at 500 mb in the Southern Hemi sphere. Mon. Wea. Rev., 108, 1378-1389. , 198 l a: Interannual variability of the Southern Hemisphere 500 mb flow: Regional characteristics. Mon. Wea. Rev., 109, 127-136.1392 MONTHLY WEATHER REVIEW VOLUME 110--, 1981 b: Observed Southern Hemisphere eddy statistics at 500 mr: Frequency and spatial dependence. J. Atmos. Sci., 38, 2585-2605. , and D. A. Paolino, Jr., 1981: Characteristic patterns of vari ability of sea level pressure in the Northern Hemisphere. Mon. Wea. Rev., 109, 1169-1189.van Loon, H., 1967: The half yearly oscillations in middle and high southern latitudes and the coreless winter. J. Atmos. Sci., 24, 472-486. , 1971: On the interaction between Antamtica and middle latitudes. Research in the Antarctic, L. Quam, Ed., Amer. Assoc. Adv. Sci., 447-487. , 1972: Wind in the Southern Hemisphere. Meteorology of the Southern Hemisphere, C. Newton, Ed., Meteor. Monogr., No. 35, Amer. Meteor. Soc., 87-100.--, 1980: Transfer of sensible heat by transient eddies in the atmosphere on the Southern Hemisphere: An appraisal of the data before and during FGGE. Mon. Wea. Rev., 108, 1774 1781. , and R. L. Jenne, 1972: The zonal harmonic standing waves in the Southern Hemisphere. J. Geophys. Res., 77, 992-1003.--, and R. A. Madden, 1981: The Southern Oscillation. Part I: Global associations with pressure and temperature in northern winter. Mon. Wea. Rev., 109, 1150-1162.--, and J. C. Rogers, 1981a: The Southern Oscillation. Part II: Association with changes in the middle troposphere in the northern winter. Mort. Wea. Rev., 109, 1163-1168.--, and ,198 lb: Remarks on the circulation over the South ern Hemisphere in FGGE and on its relation to the phases of the Southern Hemisphere in FGGE and on its relation to the phases of the Southern Oscillation. Mort, Wea. Rev., 109, 2255-2259.--, J. J. Taljaard, R. L. Jenne and H. L. Grntcher, 1971: Zonal geostrophic winds. Climate of the upper air: Southern Hemi sphere. Vol. 2, NCAR TN/STR-57 and NAVAIR 50-1C-56, 43 pp.Vowinckel, E., 1955: Southern Hemisphere weather map analysis: Five year mean pressures. Notos, 4, 17-50 and 204-216.Wahl, E., 1942: Untersuchungen iiber den j'ahfiichen Luftdnick gang. Ver6ff Meteor. Inst., Univ. Berlin, 4, 3-71.Walker, G. T., 1923: Correlation in seasonal variations of weather VIII. Mere. Indian Meteor. Dept., 24, 75-131.Wallace, J. M., and D. S. Gutzler, 1981: Teleconnections in the geopotential height field during the Northern Hemisphere winter. Mon. Wea. Rev., 109, 784-812.Wright, P. B., 1977: The Southern Oscillation--patterns and mech anisms of the teleconnections and the persistence. Hawaii Inst. Geophys., Rep. HIG-77-13, 107 pp. [Available from Hawaii Inst. Geophys., University of Hawaii, Publications, Room 262, 2525 Correa Road, Honolulu 96822].