808 MONTHLY WEATHER REVIEW VOLUME 114Quasi'Stationary States in the Southern Hemisphere KINGTSE C. MOM/A-Corn Sigma Data, Inc., NASA/Goddard Space Flight Center, Greenbelt, MD 20771(Manuscript received 28 January 1985, in final form 27 September 1985) Pattern correlations between daily anomalies have been used to study the persistence of the Southern Hemisphere circulations. The dataset consists of daily Australian analyses of 500 mb heiF, hts and ~ level pressurefor the period from 1972 to 1983. Compared io the Northern Hemisphere, the pattern correlations are muchlower and more variable in the Southern Hemisphere. The mean one-day lag autocorrelatlon is only 0.57,compared to 0.81 in the Northern Hemisphere. The correlations increase significantly for the filtered anomalies,which consisl of the planetary wavenumbers from 0 to 4. Subjective criteria based on the pattern correlations are used to select quasi-stationary events. A series of 5or more daily maps is defined to be quasi-stationary if the pattern correlatiom between all pairs of five consecutivemaps in this time series are ia~er than or equal to 0.5. In winter, quasi-stationary events can be classified interms of wavenumbers. Waves 3 and 4 are by far the dominant waves. More than half of the events have largewave 3 amplitude with geographically fixed orientations.1. Introduction Persistence of anomalies of the atmospheric circulation is of great interest to meteorologists. Trenberth,(1985) documented the persistence of daily geopotential heights at 1000 mb and 500 mb levels over theSouthern Hemisphere (SH) using autocorreiations. Hehas shown that for most regions in midlatitudes, a firstorder autoregressive model fits the time lag autocorrelations very well, but over Antarctica, Australasia,and most of the tropics higher order autoregressivemodels are more appropriate: The question arises asto whether during some periods, the atmospheric circulation is more persistent than expected from a rednoise process. It is known that some atmospheric circulation patterns are more persistent from day to day and recurfrom time to time, therefore these patterns are morepredictable. To study the recurrence of patterns Lorenz(1969) and Gutzler and Shulda (1984) have used spatialcorrelations, or root-mean-square differences to identify "analogues" in the Northern Hemisphere (NH)winter. The identification of more persistent periods locallyhas been concerned mostly with large amplitudeblocking patterns. The two main elements of all criteriahave been the structure of the flow pattern and its persistence in time. The first objective way to identifyblocking events for 500 mb geopotential heights in theNH was suggested by Dole (1983). According to Dole,a blocking event is identified by persistence of ananomaly of one standard deviation or greater for 10days or more. In the SH; the first study of blocking wasdone by van Loon (1956). In general, blocking eventshave a much shorter lifetime (5-6 days) than in theNil (Trenberth and Mo, 1985). At longer times, thenumber of large persistent anomalies is much less inthe SH and this may be due to the strong westerliespresent in the middle and high latitudes. A comparisonof the climatology of blocking in the two hemisphereshas been done by Coughlan (1983). However, none ofthe definitions of blocking have taken into consideration of flow patterns at other distant locations. The multiple equilibria studies using simple modelsshow that for a given forcing, model atmospheres canhave several quasi-stationary patterns (Charney andDevore, 1979; Charney and Straus, 1980; Charney etat., 1981; and Reinhold and Pierrehumbert, 1982). Inthese models one equilibrium is associated with synoptically defined near zonal flow, and another withblocking (Legras and Ghil, 1985). The general circulation in the SH exhibits a large degree of zonal symmetry (Trenberth, 1979) so the criteria used to selectblocking events may exclude any such low amplitudenear zonal flow. Recently, Horel (1985) used objectively defined criteria based on pattern correlations to identify quasistationary regimes in the Northern Hemisphere winter.Some of the quasi-stationary regimes have a flow pattern similar to the climatological mean and the anomalies are weak. Some of the quasi-stationary regimesshow regionally persistent features similar to blockinghighs (Dole and Gordon 1983). However, Hotel dicated that the quasi-stationary regimes in the NH are not easily classified into just one or two categories since they exhibit considerable diversity in their spatial patc 1986 American Meteorological SocietyMAY I986 KINGTSE C. MO 809tems. In this paper, Horel's procedures will be followedto study the persistence of the planetary scale circulations in the Southern Hemisphere. We will identifythose most persistent patterns according to patterncorrelations in the SH and study their properties. InSection 2, a brief description of data and proceduresused is given. Pattern correlations are discussed in Section 3 and the quasi-stationary regimes identified andstudied in Section 4. Conclusions can be found in Section 5.2. Data and analysis techniques The dataset consists of the daily operational analysesfrom the World Meteorological Center in Melbourne,Australia. The data are interpolated to a 4- latitude by5- longitude grid from 10-S to 90-S and cover theperiod from June 1972 to July 1983. Some aspects ofthe quality of the analyses were described by Trenberth,(1979; 1981). In this study, 0000 GMT 500 mb analysesare used to identify quasi-stationary states and the sealevel pressure analyses are used to study their propertiesin more detail. Fourier analysis was used to remove the seasonalcycle from the data. The seasonal cycle was defined asthe sum of the I 1-year mean and the 1 lth and 22ndFourier components of the time series at each gridpoint. The anomalies were defined as the differencebetween the data and the seasonal cycle for each pointin space and time. The time series of anomalies werethen divided into 92 day winters (June, July and August) and 90 day summers (December, January andFebruary). Daily standard deviations were calculatedseparately for each season, using all eleven years. Theanomalies of each season were normalized by dailystandard deviations for that season, so that equal weightis given to all regions. Midlatitude regions, where variances are largest, will not be weighted more heavily. The pattern correlation p(t, r) between the maps onday t and day t + r is defined as the spatial correlationbetween themp(t, ~) = (z'(x, t)z'tx, t + r)) - (z'(x, t))(z'(x, t + o'[z'(x, t)l~r[z'(x, t + r)lwhere oa[z'(x, t)] = (z'2(x, t)) - [(z'(x, t))]2, z'(x, t) isthe normalized anomaly at time t, and angle bracketsindicate an area-weighted average over the grid points.The pattern correlation defined this way is more sensitive to the phase of the maps with relatively less sensitivity given to the magnitude of anomalies (Gutzlerand Shulda, 1984), compared to other measures ofpattern similarity.3. Pattern correlations In Fig. 1 the cumulative frequency of occurrence (in percentage) of the pattern correlations during the 11 -0.2 0 0.2 0.4 0.6 0.8 1.0 PATTERN CORRELATION FIG. 1. Cumulative frequency of occurrence (in percentage) of pattern correlations between 500 mb winter anomalies at lag I to 5 forthe total normalized anomalies.winters is plotted as a function of lag. These curvesrepresent the percentage of days during the 11 winterswhere the correlations at a given lag were less than aparticular value. For example, 20% of the hemisphereis covered by 1-day lag autocorrelations less than 0.4,and on 50% of the days the l-day (2-day) lag patterncorrelation exceeded 0.5 (0.3). In contrast for the NHwinter Horel (1985) showed that on 50% of the daysthe l-day (2-day) lag pattern correlation exceeded 0.8(0.56). The anomalies in the SH are far less persistentthan anomalies in the NH. This is expected since thepointwise autocorrelations of 500 mb height are, ingeneral, much lower in the SH (Trenberth, 1985; Mo,1983). Statistics can be compiled from the 11 year sampleof pattern correlations for both summer and winter.Table 1 lists the mean pattern correlations, and thestandard deviations as a function of lag for winter andsummer. The mean correlations for all lags are smallerthan those found in the NH by Horel (1985).'In general,the persistence after 1 day in the SH is equivalent tothat after 2 days in the NH. The correlations can be transformed by using theFisher Z-transformation (Gutzler and Mo, 1983; Horel,1985), Z(t, r) = 5 log ~ _ p(t, 'It is assumed that Z is normally distributed with thevariance (N- 3)-~ where N is the number of degreesof freedom. By knowing the x;ariance of Z, it is possibleto estimate the number of degrees of freedom per mapin the SH. As shown in Table 1, the standard deviationa(Z) varies from 0.22 to 0.18 in winter and 0.22 to0.15 in summer, thus the corresponding number ofindependent gridpoints is between 21 and 31 for winter,and between 21 and 41 for summer. In the NH, theFrO. 2. 500 mb height fields for (a) 6 June (b) 8 June lqG. 3. 500 mb height anomalies for (a) 6 June (b) 8 Juneand (c) 11 June 1982, (contour interval 120 m). and (c) 11 June 1982, (contour interval is 40 m).M^YI986 KINGTSE C. MO 811 TABLE 1. Means, standard deviations of the pattern correlations and the standard deviations of Z-transformation and the red noise correlations P,(Z) as a function of lag for the total anomalies and filtered anomalies.Total anomaliesFiltered anomaliesLag Mean a(p) a(Z) P~ Lag Mean a(p) a(Z) P~ WinterI 0.57 0.16 0.22 0.57 I 0.71 0. I 1 0.22 0.712 0.38 O. 18 0.22 0.32 2 0.53 O. 14 0.20 0.503 0.28 0.19 0.21 0.18 3 0.39 0.16 0.19 0.364 0.23 0.18 0.20 0.10 4 0.30 0.17 0.19 0.255 0.20 0.18 0.19 0.06 5 0.23 0.17 ' 0.19 0.1810 0.12 0.18 0.19 0.04 10 0.15 0.18 0.19 0.13 SummerI 0.59 O. 14 0.22 0.59 I 0.68 O. 15 0.26 0.682 0.38 0.15 0.18 0.34 2 0.49 0.16 0.22 0.463 0.28 0.15 0.16 0.20 3 0.38 0.17 0.21 0.314 0.21 0.14 0.15 0.12 4 0.31 0.18 0.20 0.215 0.19 0.15 0.16 0.07 5 0.26 0.19 0.21 0.1410 0.12 0.16 0.17 0.04 10 0.16 0.20 0.21 0.10number of independent gridpoints is between 30 and37 (Hotel, 1985). Hotel (1985) used objective criteriato identify quasi-stationary regimes for 500 mb geopotential heights in the NH. According to Horel, a series of 7 or more daily maps is quasi-stationary if thepattern correlations between all pairs of maps withinthis series remain larger than 0.5. In the SH, assuming20 independent points per map, a pattern correlationof 0.5 is significant at the 98% level. Consequently,daily maps in the SH having correlations with one another greater than 0.5 are assumed to be similar to oneanother. The daily maps for 6, 8 and 11 June 1982 aregiven in Fig. 2 to characterize this similarity. Figure 3shows the corresponding anomalies. The correlationbetween 6 and 8 June is 0.55 and for 6 and 11 June,100 O0 80 70 60 50 40 30 20 10-0.2 0 0.2 0.4 0.6 0.8 1.0 PATTERN CORRELATIONFIG. 4. As in Fig. 1 but for the filtered anomalies (zonal wave 0-4).0.52. During the whole period there are very strongwave 3 and wave 4 components. The blocking ridgenear the New Zealand coast persists through the period,the high in the Indian Ocean at 45 -S, 60-E is intensifiedat the end of the period. Since the daily maps are not very persistent due tosynoptic scale disturbances, we repeat the calculationusing filtered anomalies, which only contain the longwaves 0 to 4. Figure 4 is a plot of the cumulative frequency of occurrence of the pattern correlations usingfiltered anomalies for the winter season. It can now beseen that on 50% of the days the l-day (2-day) lag pattern correlation exceeds 0.68 (0,50). From Table 1, itis also apparent that the mean pattern correlation atlag 1 (lag 2) is enhanced considerably [from 0.57 (0.38)to 0.71 (0.53)] by using filtered anomalies. The increaseis larger than that for correlations in the NH. This suggests that transient scales smaller than wavenumber 4are relatively more important in the winter circulationof the SH than for the NH (Trenberth, 1981). The correlation at lag r expected from a red noisemodel has the approximate form pr(r) = [p(1)]~where p (1) is the mean correlation at lag 1. In the NH,Horel (1985) showed that the atmospheric circulationdecay time is shorter than that expected of a first-orderautoregressive process. However, in the SH, from Table1, the observed pattern correlations at lags 2 to 10 daysarc slightly more persistent than red noise. Trenberth(1985) indicated that a red noise model gives a best fitto the data locally in the midlatitudes but higher orderautoregressive models arc needed to fit the data forareas over Antarctica, and the tropical regions. The pattern correlations also show great intcrannualvariability. Figures 5a and 5b show the pattern corrc812 MONTHLY WEATHER REVIEW VOLUME 1141.00.8z 0.6I~ 0.4OOZ~ 0.2-0.210 20 1 10 20 1 10 20 31 JUNE JULY AUGUSTb1.00.80.6~ 0.2uJ-0.2-0.410 20 1 10 20 1 10 20 31 JUNE JULY AUGUSTFiG. 5. Pattern correlations at lag I (solid line) and 5 (dashed line) for the tohal anomalies for (a) winter 1976 and (b) winter 1982.lations at lag I (solid line) and lag 5 (dashed line) fortwo winters, 1976 and 1982 respectively. The correlations for the winter of 1982 are, in general, higherthan that of 1976. Correlations at lag I during 1982are above 0.5 over 90% of the time, while correlationsat lag 5 are less than 0.5 during the entire 1976 winter. Little difference in the mean correlations betweenwinter and summer can be discerned in Table 1.MAY 1986 KINGTSE C. MO 8134. Quasi-stationary statesa. Identification of quasi-stationary states Objective criteria based on p(t, r) are used to definethe quasi-stationary states: a series of 5 or more dailymaps is quasi-stationary if the pattern correlations between all pairs of 5 consecutive maps in this time seriesare greater than or equal to 0.5. For example, whenp(t, r) >~ 0.5 and p(t + 1, r) >~ 0.5 for r = 1 to 5, thenthe quasi-stationary event covers day t to day t + 6.p(t, 6) is not required to be larger than or equal to 0.5.Quasi-stationary states were selected for total anomaliesand filtered anomalies. Fourier decomposition wasperformed around each latitude circle. Only zonalwavenumbers 0 through 4 were retained for the filteredanomalies. In the SH, flow is less persistent, so we useless restrictive criteria than those proposed by Horel.The pattern correlations between all pairs within theperiod remain 0.5 is not required, so this allows moretemporal evolution of the maps over the periods. Thehemispheric correlations are used here, so if maps arepersistent on less than hemispheric scale, they will notbe selected as quasi-stationary states. For example,during a five day blocking period, if a blocking ridgein the Pacific Ocean persists but the flows in the IndianOcean are not similar from day to day, this will not bechosen as an event. Table 2 lists all quasi-stationar~ events selected usingthe total anomalies and the filtered anomalies. Thenumber and duration of events identified here dependon the criteria. If the limit is lowered to 0.4, then threemore events (26 June-4 July 1973, 1-6 June 1979 and1-6 June 1981) would be added to the list for the totalanomalies. The criteria used here are very conservativeand the occurrence of quasi-stationary events may beunderestimated. The only event on the list of quasi-stationary eventsselected using the total anomalies, but not on the listselected using the filtered anomalies, is 3-11 June 1973.The average map of this event is shown in Fig. 6, ithas a large wave 5 component. Quasi-stationary wave5 patterns have been observed during the SH summerseason (Hamilton, 1983), and the pentagonal featuresare particularly strong during the FGGE year (Salby1982). However, those features are not particularly stationary and have large traveling components, so thesesummer events do not appear in the list of quasi-stationary events. Some of the quasi-stationary statesidentified here show regionally persistent features,especially in the Australasian region. Owing to the definition of pattern correlation here, locally persistentanomalies do not always contribute enough to allowthe minimum criteria to be met. For example, largeamplitude stationary Rossby waves with zonal scale--, wavenumber 7 were present between 20 and 40-Sto the east of South America during January of theFGGE year (Kalnay and Paegle, 1983). These waves TABLE 2. List of blocking events and quasi-stationary states selectedfrom correlations for the total anomalies and filtered anomalies duringthe winter season.Quasi-stationary eventTotal anomalies Filtered anomaliesBlocking event 4-20 Jun 197222-28 Jun 197220-29 Aug 19723-10 Jun 197319-28 Aug 197329 Jun-4 Jul 197514-23 Jun 19794-11Jul 19796-17 Jun 198218-23 Jun 19829-19 Jan 197412-17Feb 19747-11 Jan 19781-8 Feb 1983 Winter11-21Jun 197222-27 Jun 197214-20 Jul 197221-26 Aug197211-18 Jul 197322-28 Jul 197318-28 Aug 197310-15 Jun 197416-21Jun 197529 Jun-4 Jul 197518-23 Jul 197715-23 Jun 1979 6-13 Jul 197924 Jul-I Aug 197922-29 Aug 197922-29 Aug198024-29 Jun 198124-29 Jul 1981 9-15 Aug 19816-15 Jun 19824-9 Jul 19825-12 Aug 198215-24 Aug 1982 Summer6-12Dec 197214-21Dec 19727-23 Jan 19741-10Feb 197412-16Feb 197411-17 Jan 197514-21 Jan 19767-19Dec 19767-14 Jan 197815-20Dec 197818-26Dec198215 Jan-ll Feb 19833-20 Jun 197230 Jul-5 Aug 19726-15 Aug 197228 Jun-7 Jul 19738-21 Jul 197321-27 Jun 197413-22 Jun 197611-27 Ju| 197711-21 Ju1197823 Jul-5 Aug 197913-19 Aug 197924-29 Jul 198010-19 Jun 198110-17 Jul 198124 Jul-4Aug 1981 9-17 Aug 19817-23 Jun 198211-19 Jul 198213-22Dec 19727-16Dec 197620-30 Jan 197828 Jan-5Feb 1980 5-19Feb 198021-30Dec 19809-18Dec 198222-29Dec 198228 Jan-5 Feb1983persisted through the month of January, but do notappear in our list of quasi-stationary events. In summary, there are 23 events in winter over 11seasons, and the average duration is 8 days, thus 18%814 MONTHLY WEATHER REVIEW VOLUME 114FiG. 6. Average map over the duration of the event for 3-11 June 1973.of a total of 1012 days fall into quasi-stationary regimes.For the NH, Horel (1985) used more restrictive criteriaand identified 58 events during 22 winter seasons. Theaverage duration of events in his study is 9 days, thus25% of a total of 2160 days are quasi-stationary.b. Quasi-stationary states and blocking events Table 2 also lists blocking events, chosen by requiring1) a positive anomaly greater than 150 m in winterand 100 m in summer at one location for more than6 days and 2) the presence of a blocking ridge. Thethreshold values are chosen as 150 m and 100 m, because they are the largest contour values on the mapof standard deviation of daily anomalies for winter andsummer respectively. Most blocking events do notmake the list of quasi-stationary events. This suggeststhat blocking on most occasions is a local phenomenonas concluded by Trenbe, rth and Mo (1985). There areseven events that appear on both lists. Six of them havelarge wave 3 components. Three of them (8-21 July1973, 23 July-5 August 1979 and 24 July-4 August1981) have large positive anomalies near 85 o E, 165 o Wand a secondary maximum at 30-W. These are preferred configurations where wave 3 is apt to persist.Another three events (3-20 June 1972, 11-27 July 1977and 7-23 June 1982) have the opposite phase. Examples are given in Fig. 7 where the total 500 mbheights averaged over the duration of blocking eventsis plotted for 23 July-5 August 1979 and 7-23 June1982.c. Classification of quasi-stationary events To assess the similarities between the quasi-stationary events, averages of the daily maps and averages ofthe total ar~omaly maps over the duration of each eventwere computed. It was found that in the SH eventscould be classified according to their dominant wavenumbers. 1) WINTER The amplitudes and phases for wave 1 to 4 werecomputed at each latitude for averaged daily maps.Table 3 is a list of the amplitudes and phases for wavesI to 4 at 50-S for each event. The amplitudes andphases of the 11 year winter mean at 50-S are alsogiven as a reference. Our climatology agrees with themean planetary waves documented by Trenberth(1980) using the data from 1972 to 1982. Table 3 gives a general indication of the dominantwavenumber in the midlatitudes for all events. Withthis as a guide, we visually segregated all events bysimilarities in the location of major troughs and ridges.There are only two events 18-28 August 1973 and 512 August 1982 where the average maps resemble theseasonal mean, so that the anomalies are small. ThereMAY 1986 KINGTSE C. MO 815abFIG. 7. The average map over the duration of the event for (a) 23 July-5 August 1979 and (b) 7-23 June 1982 (contour interval 120 m).816MONTHLY WEATHER REVIEW VOLUME 114TABLE 3. Amplitudes and phases of first four harmonics at 50-S for winter quasi-stationary events (total field).Amplitude Phase(m) (deg)Date Wave 1 Wave 2 Wave 3 Wave 4 Wave I Wave 2 Wave 3 Wave 411-21 Jun 1972 ' 3.3 '23.8 70.2 84.4 49W 172E 26E 43E22-27 Jun 1972 33.5 51.5 104.9 56.6 150E 43E 19E 75E14-20 Jul 1972 67.4 60.5 69.0 39.3 65W 45E 15E 81E21-26 Aug 1972 69.6 73.0 47.0 41.4 91W 152E 34E 79E11-18 Jul 1973 56.7 72.3 150.8 38.0 152W 164E 53E 62E22-28 Jul 1973 12.1 86.9 163.1 70.8 66E 170E 52E 70E18-28 Aug 1973 73.5 45.1 38.4 21.8 156W 3E 46E 55E10-15 Jun 1974 77.1 37.5 79.0 61.7 135E 48E 43E 52E16-21 Jun 1975 41.7 91.4 7.3.5 83.6 10E 124E 47E 53E29 Jun-4 Jul 1975 29.5 68.5 47.8 74.2 174E 162E 35E 54E18-23 Jul 1977 105.9 53.7 56.3 21.1 159W 150E 68E 64E15-23 Jun 1979 4.5 69.2 ~ 116.6 26.6 119W 163E 37E 34E 6-13 $ul 1979 77.6 78.7 112.0 43.4 39W 177E 45E 19E24 Jul-1 Aug 1979 98.8 42.5 107.4 11.4 140E 32E 56E 33E22-29 Aug 1979 20.0 129.2 36.3 21.8 149W 16E 57E 70E22-29 Aug 1980 73.3 48.1 87.5 33.5 102W 175E 50E 73E24-29 Jun 1981 103.4 67.8 37.8 85.2 153W 56E 57E 77E24-29 Jul 1981 5.2 34.9 162.5 49.3 21W 20E 53E 35E 9-15 Aug 1981 24.4 26.7 84.3 38.0 92W 6E 16E 73E6-15 Jun 1982 73.9 43.1 101.4 I01.1 103E 176E 39E 63E4-9 Jul 1982 65.4 68.6 89.4 28.0 147W 133E 29E 68E5-12 Aug 1982 86.7 34.3 42.8 19.8 178W 131E 46E 46E15-24 Aug 1982 48.8 29.2 72.9 15.9 150W 151E 36E 38EClimatology 67.0 32,0 33.8 10.7 142W 176E 40E 65Eare four cases, 11-21 June 1972, 16-21 June 1975, 29June-4 July 1975 and 6-15 June 1982, where wave 4is very large and has approximately the same phase.All four events have a ridge located to the south ofAustralia near where blocking highs are likely to occur.Figures 8a and 8b show the composites of the total 500mb heights and the total anomalies for these fourevents. Wave 4 is the dominant wave, but there alsois strong contribution to the pattern from wave 3. There are 14 cases where wave 3 is the most prom'inent. Within these 14 events, eight of them have approximately the same phase. These are 11-18 July1973, 22-28 July 1973, 10-15 June 1974, 15-23 June1979, 6-13 July 1979, 24 July-1 August 1979, 22-29August 1980 and 24-29 July 198 I. Three of these occurin the FGGE year, when an unusually large wave-3mean amplitude was observed in the winter. Wallaceand Hsu (1983) documented the wave-3 pattern for 6July-4 August 1981, but only a part of that periodenters our statistics. This may be due to the conservative criteria used here. Events in the 1973 winter,associated with blocking, have been reported by Trenberth and Mo (1985). Except for 15-23 June 1979 and22-29 August 1980, the remaining events show a ridgenear New Zealand or to the south of Australia, whereblocking highs frequently form (Wright, 1974). Composites of the total 500 mb heights and the totalanomalies for wavenumber 3 events are shown in Fig.9. There are three minima located in three oceans at50-S, 115-E, 60-S, 0- and 55-S, 140-W, respectively.The same strong wave-3 signal can be found in theteleconnection pattern calculated using monthly meandata (Mo and White, 1985). There are three events, 22-27 June 1972, 14-20 July1972, and 9-15 August 198l, which also have a largewave-3 amplitude but opposite phase. In Fig. 10 thecomposites of the total 500 mb heights and the totalanomalies of these three events are displayed. This orientation is not a favored position for large amplitudewave 3 to slow down and persist, according to Trenberth and Mo (1985).2) SUMMER Two events 14-21 December 1972 and 7-19 December 1976 coincide with blocking events, results ina ridge in the Pacific near New Zealand. Amplitudesand phases for.wave I to 4 fit 50-S for averaged mapsare listed in Table 4. Wave 1 is the most prominentwave for eight events. Only the event 12-16 February1974 has a large wave 3 component. Two events, 110 February 1974 and 18-26 December 1982 have theirMAY 1986 KINGTSE C. MO 817b FIG. 8. Composite map for 4 quasi-stationary events with a large wave-4 amplitude(a) 500 ml~ heights (contour interval 120 m) and (b) total anomalies (contour interval40 m).818 MONTHLY WEATHER REVIEW VOLUME 114ab FIG. 9. Composite map for eight quasi-stationary states with large wave-3 amplitude(a) 500 mb heights (contours interval 120 m) and (b) total anomalies (contour interval40 m).MArl986 KINGTSE C. MO 819abF~G. 10. As in Fig. 9 but for three events 12-27 June 1972, 14-20 July 1972 and 9-15 August 1981 (contour interval 40 m).820 MONTHLY WEATHER REVIEW VOLUME 114TABLE 4. Amplitudes and phases of first four harmonics at 50-S for summer quasi-stationary events (total field).Amplitude Phase(m) (deg)Date Wave I Wave 2 Wave 3 Wave 4 Wave I Wave 2 Wave 3 Wave 46-12Dec 1972 115 45 26 28 108W 171E 88E 38E14-21Dec 1972 94 26 23 56 I10W 63E 82E 54E7-23 Jan 1974 85 58 41 23 157W 178E 25E 73E1-10 Feb1974 38 24 49 85 171E 158E 55E 81E12-16Feb 1974 60 32 105 14 165W 154E 55E IE11-17 Jan 1975 136 31 6 72 157S 176E 20E 1E14-21 Jan 1976 128 60 19 17 179E 49E 40E 23E7-19Dec 1976 103 55 67 71 138W 8E 55E 14E7-14 Jan 1978 51 38 22 33 59W 8E 103E 9E15-20De- 1978 73 8 32 51 150W 43E 105E 33E18-26Dec 1982 51 52 33 59 161W II6E 62E 37E15 Jan-lfFeb 1983 85 38 30 51 123W 86E 63E 58EClimatology 74 20 26 7 134W 2E 59E 74Elargest contribution from wave 4. The longest event is15 January-1 1 February 1983. The differences betweenthe averaged map of that event and the summer meanfor sea level pressure and 500 mb heights are given inFig. 1 1. The structure is equivalent barotropic. Themaps show the features of a low/wet event with a highover the Australia regions and Antarctica and a low inthe Pacific Ocean.d. Vertical structure of quasi-stationary states Averages of the daily sea level pressure over the duration of each event were computed and compared withthe averaged maps of 500 mb heights. All of them showequivalent barotropic structure. As an example, thecase of 6-15 June 1982 is given in Fig. 12. Both 500mb and sea level pressure maps have a ridge near NewZealand and lows in the Indian Ocean, the PacificOcean and the Atlantic Ocean near South America.This is consistent with findings by Trenberth (1980)and van Loon and Jenne (1972) that the stationarywaves in the SH have an equivalent barotropic structureand do not transport sensible heat.5. Conclusions In this study, pattern correlations between dailyanomalies have been used to study the persistence ofSouthern Hemisphere circulations. Compared to theNH, the pattern correlations are much weaker andmore variable in the SH. In winter, the mean 1-daylag correlation is only 0.57. This indicates that the patterns in the SH are less persistent (and have large interannual variability). The correlations increase significantly for the filtered anomalies consisting of zonalwavenumbers from 0 to 4, showing that the daily variations are largely due to transient eddies smaller thanwave 4. Trenberth (198 I) showed that high frequencyeddies contribute significantly to the variances in themidlatitudes. However, there exist many time periodswhere the circulation remains quasi-stationary according to pattern correlations. The objective criteria have been used to identify thesequasi-stationary periods. There are nine events in thewinter and four in the summer over 11 years. If weconsider the planetary scales only, there are 23 eventsin winter and 12 events in summer. The criteria aremore relaxed than that used by Horel (1985) to identi~analogous periods in the NH. Nevertheless, fewerevents were found in the SH. There is considerableinterannual variation with several quasi-stationaryevents occurring during the winters of 1972, 73, 79and 82 but none during the winters of 1976 and 78.There are not many overlapping periods with blockingevents. This suggests that most blocking events are localphenomena. (Trenberth and Mo, 1985). In winter, unlike in the NH (Horel, 1985), quasistationary states can be classified in terms of wavenumbers. There are 14 events which have large wave3 amplitude, and eight of these have the same phase.The composite of these events shows minima at 50-S,115-E; 60-S, 0- and 55-S, 140-W. The same wave-3signal has been found in the winter teleconnection pattern. There are another three events with a similar pattern and amplitude, but opposite phase. Four otherevents are dominated by wave 4, and two events haveflow patterns similar to the seasonal mean. The resultssuggest that the quasi-stationary planetary scale statesoccur less frequently in the SH winter but when theydo occur, most of them are dominated by wave 3. In summer, wave I is the dominant wavenumberfor eight of 12 events. Except those events overlappingwith blocking periods, the averaged maps are similarto the mean of 11 summer seasons.'1986 KINGTSE C. MO 8218b FIG. 1 I. The difference of (a) 500 mb height (contour interval 40 m) and (b) sealevel pressure (contour interval 2 rob) between the averaged map of the 15 JanuaryI 1 February 1983 event and the summer mean of 11 years.822 MONTHLY WEATHER REVIEW VOLUME 114ab FIG. 12. The average of(a) 500 mb heights (contour interval 120 m) and (b) sealevel pressure (contour interval 4 mb) over the duration of the event 6-15 June 1982.MA-1986 KINGTSE C. MO 823 Acknowledgments. The author would like to thankDavid S. Gutzler, R. E. Livezey, Kevin Trenberth andan anonymous reviewer for making perceptive comments which strengthened the final manuscript. REFERENCESCharney, J. G., and J. G. Devore, 1979: Multiple flow equilibria in the atmosphere and blocking. J. Atmos. Sci., 36, 1205-1216. , and D. M. Straus, 1980: Form-drag instability, multiple equi libria and propagating planetary waves in baroclinic, orograph ically forced, planetary wave systems. J. Atmos~ Sci., 37, 1157 1166. - , J. Shukla and K. C. Mo, 1981: Comparison ofa barotropic blocking theory with observations. J. Atmos. Sci., 38, 762-779.Coughlan, M. J., 1983: A comparative climatology of blocking action in the two hemispheres. Aust. Meteor. Mag., 31, 3-13.Dole, R. M., 1983: Persistent anomalies of the extratropical Northern Hemisphere wintertime circulation. Large Scale Dynamical Processes in the Atmosphere. B. J. Hoskins, and R. P. Pearch, Eds., Academic Press, 95-111. , and N. D. Gordon, 1983: Persistent anomalies of the extra tropical Northern Hemisphere wintertime circulation. Geo graphical distribution and regional persistence characteristics. Mon. Wea. Rev, 111, 1567-1586.Gutzler, D. S., and K. C. Mo, 1983: Auto-correlation of Northern Hemisphere. geopotenfial heights. Mort. Wea. Rev., 111, 155 164. , and J. $hukla, 1984: Analogs in the winter time 500 mb height field. J. Atmos. Sci., 41, 177-189.Hamilton, K., 1983: Aspects of wave behavior in the mid and upper troposphere of the Southern Hemisphere. Atmos.-Ocean, 21, 40-54.Horel, J. D., 1985: Persistence of the 500 mb height field during Northern Hemisphere winter. Mort. Wea. Rev., 113, 2030-2042.Kalnay, E., and J. Paegle, 1983: Large amplitude stationary waves in the Southern Hemisphere observations and theory. Proc. First Int. Conf. on Southern Hemisphere Meteorology, Amer. Meteor. Soc., Boston, 89-92.Legras, B., and M. Ghil, 1985: Persistent anomalies, blocking and variations in atmospheric predictability. J. Atmos. Sci., 42, 433 471.Lorenz, E. N., 1969: Atmospheric predictability as revealed by nat urally occurring analogs. J. Atmos. Sci., 26, 636-646.Mo, K. C., 1983: Persistent anomalies of the Southern Hemisphere Circulation. Proc. First Int. Conf on Southern Hemisphere Me teorology, Amer. Meteor. Soc., Boston, 70-73.--, and G. H. White, 1985: Teleconnecfions in the SouthernHemisphere. Mon. Wea. Re,., 133, 22-37.Reinhold, B. B., and R. T. Pierrehumbert, 1982: Dynamics of weather regimes: Quasi-stationary waves and blocking. Mort. Wea. Rev., 110, 1105-1145.Salby, M. L., 1982: A Ubiquitous wavenumber 5 anomaly in the Southern Hemisphere during FGGE. Mort. Wea. Rev., 110, 1712-1720.Shukla, J., and K. C. Mo, 1983: Seasonal and geographical variations of blocking. Mon. Wea. Rev., 111, 388-402.Trenberth, K. E., 1979: Interannual variability of the 500 mb zonal mean flow in the Southern Hemisphere. Mort. Wea. Rev., 107, 1515-1524. ,1980: Planetary waves at 500 mb in the Southern Hemisphere. Mon. Wea. Rev., 108, 1378-1389. , 1981: Observed Southern Hemisphere eddy statistics at 500 mb: Frequency and spatial dependence. J. ~ltmos. Sci., 38, 2585 2605. ,1985: Persistence of daily geopotential heights over the Southern Hemisphere. Mon. Wea. Rev., 133, 38-53. , and K. C. Mo, 1985: Blocking in the Southern Hemisphere. Mon. Wea. Rev., 133, 3-21.van Loon, H., 1956: Blocking action in the Southern Hemisphere. Notos, 5, 171-177. , and R. L. Jenne, 1972: The zonal harmonic standing waves in the Southern Hemisphere. J. Geophys. Rev., 77, 992-1003.Wallace, T. M., and H. H. Hsu, 1983: Ultra long waves and two dimensional Rossby waves. J. Atmos. Sci., 40, 2211-2219.Wright, A. D. F., 1974: Blocking action in the Australian region. Austr. Bur. Meteor. Tech. Rep. No. 10, Bureau of Meteorology, Melbourne, 29 pp.