814 MONTHLY WEATHER REVIEW VOLUME122Observations of the 40-50-Day Tropical Oscillafion--A Review ROLAND A. MADDENNational Center for Atmospheric Research, * Boulder, Colorado PAUL R. JULIANNational Meteorological Center, Washington, D.C.(Manuscript received 4 January 1993, in final form 25 August 1993)ABSTRACT Observational aspects of the 40-50-day oscillation are reviewed. The oscillation is the result of large-scalecirculation cells oriented in the equatorial plane that move eastward from at least the Indian Ocean to the centralPacific. Anomalies in zonal winds and the velocity potential in the upper troposphere often propagate the fullcircumference of the globe. Related, complex convective regions also show an eastward movement. There is azonally symmetric component to the oscillation. It is manifest in changes in surface pressure and in the relativeatmospheric angular momentum. The oscillation is an important factor in the timing of active and break phasesof the Indian and Australian monsoons. It affects ocean waves, currents, and air-sea interaction. The oscillationwas particularly active during the First GARP (Global Atmospheric Research Program) Global Experiment year,and some features that were evident during the Monsoon Experiment are described.1o Introduction Reviewing the literature on the phenomenon that we(Madden and Julian 1971) called the 40-50-day oscillation (hereafter referred to as the oscillation) is notan easy task. It is not possible for us to identify forcertain which of the many aspects of the oscillation thathave been reported will prove to be most important.There has been considerable literature on the subject inthe period since 1979. It was then that Yasunari's work(Yasunari 1979) related the eastward-propagating,near-equatorial oscillation that we had described to active and break periods of the Indian summer monsoon.The First GARP (Global Atmospheric Research Program) Global Experiment (FGGE) also took place during that year and provided the most complete globaldataset available to that time. Lorenc (1984) used thedataset to show a regular eastward propagation of theupper-tropospheric velocity potential X and its closerelation to the active and break monsoon periods. Manyother authors studied these data and revealed previously unknown features of the oscillation. Here, we * The National Center for Atmospheric Research is sponsored bythe National Science Foundation. Corresponding author address: Dr. Roland A. Madden, NCAR/Climate Analysis Section, P.O. Box 3000, Boulder, CO 80307-3000.describe what we judge to be some of the most important findings from studies of these and of other data. The discussion concentrates on eastward-movingcloud complexes that are associated with the oscillation, the northward-propagating cloud zones related tothe break and active phases during the northern summermonsoon, some possible effects in the extratropics, atmospheric angular momentum changes, and effects inthe ocean. Active workers in the field will find nothingnew in this review. We hope that it will bring thosewho want a general knowledge of the oscillation up todate. The review deals only with observational studies.We have, for the most part, neglected a considerablebody of theoretical and modeling work aimed at explaining the oscillation.2. Back~und In the late 1960s tropical data were sparse, as theystill are in many areas. Spectral analysis was beginningto be used effectively to extract information from rawinsonde data collected at widely scattered observingstations. Maruyama (1967, 1968) had used it to describe the mixed Rossby-gravity waves in the equatorial stratosphere that Yanai and he had discovered(Yanai and Mamyama 1966). Wallace and Kousky(1968) used it to identify stratospheric Kelvin waves.These papers influenced us along with ones by Yanaiet al. (1968) and Wallace and Chang (1969). In them,the authors again used spectrum analysis to describe-c 1994 American Meteorological SocietyMAY 1994 MADDEN AND JULIAN 815this time--tropospheric, synoptic-scale features in thetropics. In those days, long time series were not readily available from tropical rawinsonde stations, nor were theyeasy to process with the computers of the time. Theabove authors were constrained to use time series thatranged from only three months to two years long. Wethought that we could add to their findings because theNational Center for Atmospheric Research's (NCAR)Data Support Section had begun to collect rawinsondedata from around the world and had data from at leastone tropical station, Canton Island~ (2.8-S, 171.7-W),for nearly a ten-year period. We also had access atNCAR to the biggest computer available to atmospheric scientists (Control Data Corporation 6600). After computing spectra and cross spectra of variablesfrom the Canton rawinsondes, we noticed large coherence between surface pressure, zonal winds, and temperatures at various levels over a broad period range thatmaximized between 41 and 53 days (Madden and Julian1971). The spectra of these variables also had relativemaxima, or peaks, at these periods. The coherences exceeded zero and the spectral peaks exceeded a smoothbackground continuum at levels passing stringent a posteriori tests for statistical significance. The a posterioritests were required because we had no beforehand theoretical evidence that extra variance or coherence shouldoccur in this period range. Surface pressure was coherentand in phase with the 850-hPa zonal (u) wind, and coherent and nearly out of phase (within a tenth of a cycle)with temperatures from 700 to 150 hPa. The 850-hPa uwas coherent and out of phase with the u wind from 300to 100 hPa. We found little evidence that the meridional(v) wind played a role in the variations. These resultsare consistent with an oscillation that is--at 850 hPa atleast--in approximate geostrophic balance with varyingpressure maxima and minima centered on the equator.A low-level low pressure anomaly at the equator is accompanied by low-level anomalous easterly winds atCanton. Spectra and cross spectra with data from other stations revealed spectral peaks and also marked maximain coherence in the 40-50-day range between stationsfar removed from one another (Madden and Julian1972a). Figure 1 is an example of coherence betweensurface pressure at Canton Island and Balboa in Panama (9.0-N, 79.6-W). There are two frequency bandsthat are highly coherent. One is at 5-6-day periods andthe other extends from 12- to 100-day periods with themaximum at 40-50 days. There is an interesting difference between the phase angles in these two periodranges. At 5-6 days Balboa leads Canton (westwardpropagation), and at the longer periods Canton leadsNow Kanton.50 5 DAYS 0.4 Conl'on 0.2 Leads 0 -0.2 Balboa Leads -0.4 I.O el' , ~ ~ ~ ii C-h2 BW 0.8 ~ STATION PRESSURE. Ir CANTON 171~43'W 0.6 '-'./~ BALBOA 79055'W l',l . Q4 ~ ~[ [ 0.2 ~ .J~ I 4 , o,i ~ v~~ 95~ 0 0.1 0.~ 0.5 0.~ 0.5 FREQUENCY (Cycles/Day) FIG. l. Coherence squar~ (bottom) and phase ~ fractions of acycle be~een the su~ace pressures at Canton (2.8-S, 171.7-W) andBalboa (9.0ON, 79.6~W) for the period 16 December 1959-24 March1967. The bandwidth ~d the 95~ limit for a null hypothesis of zcrocoherence are indicated.Balboa (eastward propagation). The longitudinal separation between these two stations is 92- and the phaseat 5-6-day periods is about 90- (0.2-0.3 cycles), consistent with a propagating disturbance of zonal wavenumber 1 scale. This westward-propagating disturbance appears to be the global scale, 5-day Rossbywave (Madden and Julian 1972b). The nearly linear increase in phase with increasingfrequency for the longer-period variations is consistentwith disturbances that reach Balboa three days after theyaffect Canton. That is a longitudinal propagation speedof 39 m s-~. There were peaks and coherence in surfacepressure cross spectra between stations from the eastcoast of Africa to the western half of South America.Phase angles indicated a 20 m s -~ eastward propagationfrom Gan Island (0.7-S, 73.2-E) in the Indian Ocean toCanton Island in the Pacific (Fig. 2). Evidence of suchrapid propagation suggested that the disturbance must inaddition have characteristics of a standing oscillation inorder that it may be reflected in a period on the order of45 days in a single station time series.816 MONTHLY WEATHER REVIEW VOLUME 122 FtG. 2. Mean phase angles (deg), coherence squares, and background coherence squares for approximately the 36-50-day period rangeof cross spectra between surface pressures at all stations and those at Canton. The plotting model is given in the lower right-hand corner.Positive phase angle means Canton time series leads. Stars indicate stations where coherence squares exceed a smooth background at the95% level. Mean coherence squares at Shemya (52.8-N, 174.1-E) and Campbell Island (52.6-S, 169.2-E) (not shown) are 0.08 and 0:02,respectively. Both are below their average background coherence squares. Values at Dares Salaam (0.8-S, 39.3-E) are from a cross spectrumwith Nauru. The arrows indicate propagation direction (adapted from Madden and Julian 1972). There was a slower poleward propagation of about2 m s -~ over the Pacific that we did not study. We hadno data over Asia so we were unaware of a northwardpropagation of 1-2 m s -a that is very important for theIndian summer monsoon. This feature is discussed insection 6. The 40-50-day surface pressure spectralpeaks were approximately confined to _+10- of theequator, but relative maxima in coherence extendednorth from Canton to Midway (28.2-N, 177.4-W) andsouth to Raoul Island (29.2-S, 177.9-W). Lower-tropospheric u winds from Singapore ( 1.4-N, 103.9-E) to Canton exhibited spectral peaks and co herence with the Canton pressure in the 40-50-day range. Variations in upper-tropospheric u winds from near-equatorial stations resulted in spectral peaks and in peaks in their coherence with Canton pressure the full circumference of the globe. Phases with Canton pressure indicated an eastward propagation at a rate slower than that of the pressure. The sum of the evidence suggested that the oscilla tions were the result of an eastward movement of large scale circulation cells oriented in the equatorial plane as depicted schematically in Fig. 3. Because we found no evidence of the oscillation in the lower troposphere over the Atlantic or western Africa, we assumed that it originated somewhere in the Indian Ocean. The indi cated convection was supported by the convergence (divergence) in the lower-(upper) level u winds and by mixing ratios and temperatures. From Fig. 3'it is clear that the near-geostrophic bal ance with equatorial-centered pressure oscillations that was suggested at 850 hPa at Canton (Madden and Jul ian 1971) is not found at all longitudes. Even at Truk2Now Chuuk.(7.4-N, 151.8-E) only 4000 km to the west the u windat 850 hPa led the surface pressure perturbation byabout 0.25 cycles (Madden and Julian 1972a, Fig. 8).Recently, Nishi (1989) quantified the relation betweengeopotential height z and the u wind with cross-spectralanalyses. He found that in the lower troposphere 3060-day variations in z and u are in phase only near thedate line, while u tends to lead z by 0.33 cycles overIndia and Indonesia. We made similar calculations ofphase angles between z and u at 45-day periods to present in Table 1. The lower-tropospheric in-phase relation is seen only near the date line at Canton and Majuro (7.1-N, 171.4-E), as pointed out by Nishi. Evidence for an in-phase relation is also present in theupper troposphere at Gan. For geostrophic balance thein-phase relationship is consistent with maximum pressure variations at the equator, while the more nearlyout-of-phase relationship suggests maximum pressurevariations off the equator. The fact that the pressure anomalies do not add tozero along the equator (see times A and E in Fig. 3)indicates that mass flows in and out of the equatorialbelt with the oscillation. Later we will see that a meanzonal or wavenumber zero component is manifest inan oscillation in the atmospheric angular momentum. We pointed out that the oscillation was a broadbandone and that spectra from different periods differed indetail. For example, Canton surface pressure data from1 June 1957 to 27 April 1962 had a peak at 46 days,while that from 28 April 1962 to 24 March 1967 hada peak at 33 days. For pressure data at Balboa fromroughly the same periods the spectral peaks were at 48and 33 days (Madden and Julian 1972a). Figure 4 illustrates a similar difference in the spectra of Truk Island pressure data for two additional periods. Nevertheless, the overall character and the presence of theMA-1994 MADDEN AND JULIAN 817 EAST LONGITUDE WEST LONGITUDE20- 60- IO0~ 140o 180- 140- IO0- 60- 20- I~ III IlIiAFRICA INDONESIA S. AMERICA Fro. 3. Schematic depiction of the time and space (zonal plane)variations of the disturbance associated with the 40-50-day oscillation. Dates are indicated symbolically by the letters at the left of eachchart and correspond to dates associated with the oscillation in Canton's station pressure. The letter A refers to the time of low pressureat Canton and E is the time of high pressure there. The other lettersrepresent intermediate times. The mean pressure disturbance is plotted at the bottom of each chart with negative anomalies shaded. Thecirculation cells are based on the mean zonal wind disturbance.Regions of enhanced large-scale convection are indicated schematically by the cumulus and cumulonimbus clouds. The relative tropopause height is indicated at the top of each chart (taken from Maddenand Julian 1972a).oscillation has some stationarity in time. This was demonstrated by cross spectra of data recorded during the1890s at Nauru Island (0.4-S, 161.0-E) and at DaresSalaam (0.8-S, 39.3-E). Figure 5 is the spectrum ofpressure at Nauru. It is very similar tO that for CantonIsland pressure observed from 1957 to 1967 and to thatfor Truk pressure from 1967 to 1979. We also foundthe coherence and phase based on these older data(shown in Fig. 2) to be consistent with eastward propagation.3. Observations during the 1970s In 1973 Parker reported results of a study of uppertropospheric and lower-stratospheric equatorial winds(Parker 1973). He found the oscillation in the 100-hPau winds and temperatures. Parker looked at severalyears of data during the 1960s but concentrated on theJanuary-May 1966 period when the oscillation wasparticularly marked. He found it at equatorial stationsall the way around the earth. The u wind did not appearto play a role. In the eastern hemisphere the 100-hPatemperature tended to lead the 100-hPa u wind by 0.25cycles [parallel result for Truk and Canton in Maddenand Julian (1972a, Fig. 8)], and over Gan Island atleast, the 100-hPa height and u wind were in phase(seen also at 150 hPa in Table 1), indicating a geostrophic balance if the maximum pressure variationswere at the equator. The oscillation also fell off in amplitude away from the equator. Its progression was irregular though, moving relatively fast from 140-W tothe Greenwich meridian, moving more slowly near thedate line. The average speed was near 13 m s-~. Thesefindings were in reasonably good agreement with oursfor the very high troposphere. Parker considered theoscillation to be sufficiently like a Kelvin wave to beconsidered as such. Dakshinamurti and Keshavamurty (1976) computedspectral analyses of the 850-hPa u and v winds overIndia and found 30-day peaks. Dakshinamurti and Keshavamurty suggested that these variations were associated with north-south oscillations of the monsoontrough. This is the northward propagation of the troughthat has been studied extensively in recent years and isdiscussed here in section 6. To our knowledge therewere no other published analyses of local winds thatrelated to the oscillation during the decade. The first published evidence based on actual clouddata that there might be eastward-propagating cloudsystems similar to those of Fig. 3 came from Gruber(1974). He computed space-time spectra of satellitecloud brightness data from May through October of1967. He found relative maxima for eastward propagation at a period of 50 days, for zonal wave 1 at theequator, 5-N, and 10-N. The 50-day eastward maximum was not evident in data from 15-N. Zangvil(1975) showed the eastward-moving clouds with atime-longitude diagram at the equator. His analysis ofthe wavelength, however, suggested that it was closerto wave 2 than wave 1. Zangvil also presented spacetime spectra that had evidence of wave i and 2, 40818 MONTHLY WEATHER REVIEW VOLUME122day, eastward variance. During northern summer(1967) the maximum activity was along 5- and 10-N,and during northern winter (1967/68) it was at theequator and 5-S. This is consistent with maximum 4050-day cloud activity residing in the vicinity of theintertropical convergence zone (ITCZ). We consider 1979 to be a breaking point in the studyof the oscillation and that is why we have treated observations during the 1970s as a separate section. Inwhat follows we summarize some of the features of theoscillation that have been described primarily duringthe 1980s and early 1990s. The summary is broken upinto sections on the time scale of the oscillation, relatedcloud anomalies, the role of the oscillation in the Indianand Australian monsoons, effects in the extratropics,anomalies in the atmospheric angular momentum, seasonal variations, anomalies in the oceans, and finally asurvey of the literature dealing with the oscillation during the Monsoon Experiment (MONEX) of FGGE.4. On the time scale of the oscillation In our first paper, we referred to the oscillation asthe "40-50-day oscillation" (Madden and Julian1971). It was in this range (actually 41-53 days) thatspectra and cross spectra of the 10-yr record of upperair data from Canton Island showed peaks or relativemaxima. In the case of the coherence between variablesthere were more often than not absolute maxima in thisperiod range. We did acknowledge that the oscillationwas a relatively broadband phenomenon and not ahighly tuned periodicity. After examining data frommore stations and for different time periods we pointedout that the 40-50-day bounds were only approximateones for the period of the responsible physical phenomenon (Madden and Julian.1972a). To further empha3OZO~0PRESSURE SPECTRA PERIOD (DAYS) I00 50 40 25 20 15 f f I ~ I ~ - bw \ i i \ 67-79 " .... 80-85 [ I I I [ I .01 ,02 .03 .o4 .05 .o6 FREQUENCY (CYCLES/DAY) FIG. 4. Spectra of Truk (7.4-N, 151.8-E) pressure for two periods(1967-79 and 1980-85). Ordinate is variance per unit frequency.The bandwidth (bw) is indicated. The spectral estimate at 3/365day-~ for the 1980-85 data is off the scale and equal to 58 mb2 day(taken from Madden 1987).size the broadband nature of the oscillation it is oftenreferred to as the 30-50- and the 30-60-day oscillation(Krishnamurti and Subrahmanyam 1982; Weickmannet al. 1985; and many others). The broadband natureof the oscillation is evident in Figs. 1 and 5, and thechanging frequency of the associated peak in the spectra is demonstrated in Fig. 4. The shift of the spectral TABLE 1. Phase in fractions of a cycle between z and u at a 45-day period (0.02214 day-l). Results are based on 2304-day time seriesbeginning on the first of the indicated month. Associated coherence squares have approximately 46 degrees of freedom, so the 95% limit fora null hypothesis of zero coherence is 0,13, Only phase angles whose coherence square exceeds 0.13 are given. Entries of M denote levelswhere no cross spectra were computed. Positive phase angles mean u leads z.Level (hPa)Station Latitude Longitude Start 1000 850 700 500 400 300 200 150 100 80Canton 3-S 172-W 7/54 0.06Majuro 7-N 171-E 7/61Trek 7-N 152-E 7/65Koror 7-N 134-E 7/61 0.34Singapore i-N 104-E ,1/65' MCocos 12-S 97-E 7/57 0.37Gan l-S 73-E 1/65Nairobi l-S 37-E 7/59 MAscension 8-S 14-W 7/59Dakar 15-N 18-W 7/55Piarco 1 i-N 61 -W 7/72Lima 12-S 77-W 7/58Balboa 9-N 80-W 7/52 M0.09 0.46 0.330.06 -0.49 0.44 0.34 0.28 0.26 0.470.35 0.360.20 -0.48 -0.47 -0.480.33 -0,280,34 0.32 -0.27 -0.06M M M 0.48 0.45 M 0.44 M 0.26 0,23 0.170.41M MM MM MM M0.23 0.31M MMAY 1994 MADDEN AND JULIAN 819 NAURU I. Jan. 1894 - Jan. 18982O4IIIIII 2 I ~ I00 50 25 14 PERIOD (days) FIG. 5. Variance spectrum for station pressures at Nauru Island,0.4-S, 161.0-E. Ordinate (variance/frequency) is logarithmic and abscissa (frequency) is linear. The 40-50-day period range is indicatedby the dashed vertical lines. Prior 95% confidence limits and thebandwidth of the analysis (0.008 day-1) are indicated by the cross(taken from Madden and Julian 1972).peak to a 26-day period during the 1980-85 period andthe strong 1982/83 warm event are consistent with suggestions that the oscillation tends to have a higher frequency during warm water or El Nifio-Southern Oscillation (ENSO) years (Gray 1988; Kuhnel 1989). The oscillation manifests itself in many variables. Tostudy the period, we choose the u winds at Truk Islandto serve as an index. In Fig. 6 are histograms of theperiods of observed oscillations in Truk's u wind at 150hPa. These periods were estimated subjectively by examining the time series of the wind. The average periodis 45 days (48 days for September, October, and November) with a wide spread from the 22 to 79 days.There is no obvious change in the average period withseason. Anderson et al. (1984) and Cadet and Daniel(1988) reached a similar conclusion after their moreobjective analyses of the period. The analysis uponwhich Fig. 6 is based suggested that the oscillation waspresent 58% of the time. Analyses of different variablescan give slightly different results. For example, Knutson et al. (1986) presented a histogram generally similar to those of Fig. 6 but based on eastward-propagating anomalies in outgoing longwave radiation (OLR).However, they found a larger spread than in Fig. 6 withtwo cases having periods less than 20 days and twogreater than 79 days, and concluded that the oscillationwas present more than 75% of the time. The period of the oscillation spans a wide range butit is separated from synoptic-scale variations (less than10 days) and from seasonal variations. Its most frequent occurrence is near 45 days.5. Clouds We have seen that Gruber (1974) and Zangvil(1975) used space-time spectral analysis to show thatthere was evidence of eastward-propagating cloudy areas in the equatorial region with a wave 1 to wave 2zonal scale. Zangvil also showed the eastward-movingclouds with a time-longitude section, and he demonstrated that they were most marked on the summer sideof the equator. Based on OLR measurements, Lau and Chan (1983)reported on a 2-3-month oscillation in clouds betweenIndonesia and the equatorial region at the date line,which they referred to as a dipole. In a subsequent paper, Lau and Chan (1986a, their Fig. 4) presented amap of the percent OLR variance contained in the 4050-day period range based on seven 6-month segmentsfrom May to October. It showed that 12% of the variance was in the 40-50-day band over nearly the entireIndian Ocean north of the equator and also over thewestern Pacific from about 10- to 20-N. For a whitenoise process we would expect only 2% in this range.Weickmann et al. (1985) showed that spectra of OLRhas significantly (95% level) more variance in the 2872-day range than appropriate red noise spectra overregions in the central Pacific, equatorial eastern SouthAmerica, and Africa, in addition to those over the Indian Ocean and western Pacific. Cross-spectral analyses (Weickmann et al. 1985) and lag correlation between points (Lau and Chan 1986a) revealed the eastward propagation of the OLR anomalies from about 60-to 160-E on these time scales. Nakazawa (1988) studied the structure of the eastward-moving cloud masses with 3-h geostationaryOLR data. He found that the eastward-moving cloudsystems were composed of several eastward-movingsuper cloud clusters (SCC). Each SCC was, in turn,composed of smaller cloud clusters (CC) that movedwestward. Each SCC has a scale on the order of 103km and consists of a few of the CC whose scales are10~ kin. The lifetime of the CC is only 1-2 days. NewCC tend to form east of a fully developed CC. Figure7 is a schematic taken from Nakazawa's paper. Thisbehavior is confined to within 15- of the equator.820 MONTHLY WEATHER REVIEW VOLUME122MAM Mean =45Median=44J J A Mean --45 ;~ Median =43 _ I-I 'rlSON IL Mean =48/ Median --45DO F Mean = 45Median =46 _ 6 4 2 6~o3tu 4o3,,~('> 20n..'bJ03 6~ 4 2 ~ 4 20 :50 40 50 '60 TO 80 DAYS BETWEEN SUCCESSIVE MINIMA OF THE TRUK 150rob ZONAL WIND FIo. 6. Histograms of the observed periods b~sed on subjectiveestimates of the time between successive minima of the 150-mb u atTruk (7.4-N, 151.8-E). Number of occurrences of 49- and 50~dayperiods is indicated at the point marked' 50 and similarly at otherpoints. The season for an occurrence is determined by the beginningdate (from Madden 1986).At 15- the cloud movement on all spatial scales tends'to be westward. But even near the equator there is additional complexity in the cloud systems b6yond the SCC and CC.Further subtleties and variations in the cloud hierarchies were pointed out by Lau et al. (1991) and Suiand Lau (1992). Also, there'are two regions wherethe eastward-traveling SCC are accompanied by localflareups of convection. They are near 100- and 150-E(Weickmann'and Khalsa 1990). Khalsa and Steiner(1988) have studied the low-level (1000-700 hPa)precipitable water from the TIROS (TelevisionInfrared Observation Satellite) Operational Vertical Sounder (TOVS) satellite instrument and they show stationary maxima in that quantity from October 1981 through December 1985 near 100- and 150-E. They point out that almost every occurrence of the precip itable water anomalies that exceed 38 mm in those two regions is accompanied by eastward-moving upper level divergence as measured by the 250-hPa X. There are other interesting aspects of the large-scalecloud complexes associated with the oscillation besidesthe eastward propagation. We will see in a later discussion of monsoons that they are often related to theonset of the Indian and Australian summer monsoons.There are also the northward-propagating branches ofthe cloud systems over India and Southeast Asia. However, it should be noted that all large cloud complexesnear the equator' in the Indian and Pacific Oceans arenot related to the eastward-propagating oscillation.Wang and Rui (1990) studied OLR data and have summarized various behaviors of large-scale tropical cloudcomplexes. They defined tropical intraseasonal convection anomalies (TICA) to have a minimum lifetime(four pentads), a minimum scale (30- of longitude),and a threshold anomaly strength or intensity (-15W m-2). At their strongest stage the TICA must belarger than 50- of longitude and their anomalies mustbe less than -50 W m-2. Wang and Rui found 122TICA during a 10-yr period ( 1975-85 with 1978 missing). Of those, the majority moved eastward (77), 27moved northward with no connection to an eastwardmoving TICA, and 18 moved westward. These last 18were relatively weak as measured by their coldest OLRanomalies. This further emphasizes that while the schematic of Fig. 3 is representative of the oscilJation nearthe equator, there are large-scale cloud 'complexes thatbehave differently. We point out that Cadet (1~983) andMurakami (1984) also found westward-propagatingcloud complexes, mostly west of 140-E. We presume that the 77 eastward-propagating TICAare like'Nakazawa's SCC and rela~ed. ~to the o. scillation.Wang and Rui divide them into three groups. The first,32 cases, are equatorially trapped in that their centersstay within _+15- of the equator during [heir, lifetime.- Wang and Rui call them the EE mode. A second group,25"cases, moves eastward algng the equato.r but near100-E begins a northeast (13 case) or southeast (10cases) movement, or in two cases splits with one partmoving northeast and one mo-ing southeast'.' These arecalled the NE and SE modes, respectively. The thirdgroup, 20 cases, combines eastward movement withnorthward movement over India and/or the westernPacific and are called the EN mode. Wang and Ruisummarize the behavior of these three groups by determining the frequency of occurrence of each in 2-x 2- latitude-longitude squ~res, and their figure is reproduced here as Fig. 8. The different behavior in theTICA is dearly shown. Wang and Rui found that 78%(25 cases) of the equatorial eastward TICA (EE) ocMAY 1994 MADDEN AND JULIAN 821HIERARCHY OF INTRASEASONAL VARIATIONSlO203o40~07Od*10000 20000 km West ~ > East 7" ................. .~,,,~,. Super Cluster~ ~'~.~";~ ....... % ........1 ~day decaying Cloud Cluster Fla. 7. Schematic describing the details of the large-scale eastward-propagating cloud complexes [slanting ellipses marked 1SV (intraseasonal variability) on the left-hand side]. Slantingheavy lines represent super cloud clusters (SCC) within the larger complexes or ISV. The righthand side illustrates the fine structure of the SCC with smaller westward-moving cloud clustersthat develop, grow to maturity, and decay in a few days (from Nakazawa 1988).curred during the 6-month period from December toMay. Members of the second group that move to thesoutheast near 100-E (SE) occurred exclusively fromNovember through April, indicating a relation to theITCZ and Australian monsoon. Those that movednortheast (NE) occurred primarily, but not exclusively,in the northern summer, again associated with the ITCZand the East Asian monsoon. The group that combinedboth northward and eastward (EN) movement tookplace most frequently in May and October. None wasrecorded during June through August, a time when independent northward-moving TICA were frequent.However, this is probably not always the case. We willsee eastward- and northward-moving TICA, ENmodes, during June and August of 1979 [Fig. 11 tofollow, which is taken from Lau and Chan (1986a)].We will also discuss Yasunari's work (1979) that documents EN-like clouds occurring during June to August1973. This apparent discrepancy may be due in part tothe fact that since TICA often span more than i month,Wang and Rui set the time of occurrence during themonth when the TICA were at their maximum intensity. Wang and Rui (1990) also documented the areaswhere the eastward-moving TICA formed. Their figureshowing the results is presented here as Fig. 9a. Themajor formation region is the west-central equatorialIndian Ocean as predicted in Fig. 3. There is a secondary source just west of equatorial Africa. They alsodetermined that the TICA tended to have the lowestOLR values in the region of the east-central IndianOcean with a weak secondary region of low OLR alongthe equator near 160-E (Fig. 9b). These correspond tothe regions of quasi-stationary pulsations discussed byWeickmann and Khalsa (1990). Wang and Rui (1990)showed that the maximum intensification rates of theTICA were in the central equatorial Indian Ocean, withsecondary areas of intensification near 160-E north ofthe equator, and one extending from Australia to thedate line south of the equator. Intensification occurs inthe first secondary area during boreal summer and inthe second during boreal winter, which is again consistent with these regions of intensification being coincident with the ITCZ (Wang and Rui 1990).6. Monsoonsa. The Indian summer monsoon Raghavan et al. (1975) showed a remarkable variation in precipitation over west coastal Indian stationsfrom Vengurla (16-N) to Dahanu (20-N). There werelarge maxima 33 days apart on 7 July and 9 August1962 with near zero precipitation in between. The lowlevel, northward-flowing wind over Kenya had corresponding fluctuations. Although there was some controversy over whether or not the Kenyan winds couldbe used to predict the Indian rainfall, Findlater (1969)argued that these same precipitation changes were inresponse to the changing Somali jet. The fluctuating822 MONTHLY WEATHER REVIEW VOLUME 1224~ffOIE4O8BIOtatOuIOdf Ir8JM,fO~B~64~O,S/O.O80~ )l*O~ JO0 IIeu b4~11IONjog [lllosIrOI,BOI 40W~ IfJ~U4 - F~G. 8. Contour plot of the total number of occurrences of strictly eastward-moving cloud complexes (a); eastward complexes that split either to the north or to the south over the eastern IndianOcean (b); and eastward-moving complexes that are connected with cloud systems that movenorthward into southern Asia (c); in each 2- x 2- box for a 10-yr period (1975-85, 1978 missing).The contour interval is 0.8 in (a) and (c), and 0.6 in (b). The heavy lines indicate the central paths(from Wang and Rui 1990).MAY 1994 MADDEN AND JULIAN 82340#~ON2ON$0#-0I05205SOS405 0S0.40/0.0 0,40/0.0~o~FO. ~. Contou~ plots of (a) the total numbe~ of ~o~ations (solid) ~d i~inationsthe cas~a~d-movin~ ~loud ~o~plex~s, and (b) of the occu~uncu of the conics, o~ locationlo~si OLR valuc~ at th~ st~on~sz phase. Couiou~ intc~al is 0.4. ~ spatial s~ooth~ has~pplicd to the basic counts (~o~ ~an~ and RuiIndian rains are called the "break" and "active"phases of the monsoon. Figure 10 is an example of theeffects on Indian west coast precipitation of break andactive phases during MONEX (taken from Cadet1986). The break and active monsoon phases shownin Fig. 10 were accompanied by correspondingly smalland large water vapor flux across the Arabian Sea (Cadet and Greco 1987). Wylie and Hinton (1982) computed 10-day-averaged surface wind stress over the Indian Ocean duringMONEX. They showed maximum values of 0.35N m-2 in the region of the Somali jet over the westernArabian Sea in mid-June (their Fig. 7) before the onsetof heavy rainfall over western India, and 0.60 N m-2in late June coincident with it (their Fig. 9). Subsequently, the stress fell to 0.40 N m-: through the firsttwo-thirds of July during the monsoon break and thenincreased to 0.55 N m-~ in late July coinciding with thesecond active period shown by Cadet's figure. This isconsistent with Findlater's (1969) and Raghavan et.al.'s (1975) suggested link between variations in largescale surface winds and the Indian monsoon. Krishnamurti and Ardanuy (1980) also studied monsoonbreaks and found associated large-scale pressurechanges in addition to fluctuations in the Somali jet. Inparticular, space-time spectra, indicated 30-40-day,eastward-propagating zonal wave 1 and 2 surface pressure oscillations, and that 30-40-day filtered pressureridges passed India between 20- and 30-N about 5 daysafter a break. It happens that the active monsoon periods are oftenconnected with northward-moving cloud zones similar824 MONTHLY WEATHER REVIEW VOLUME122 PREEIPITABLL WA'rER ~R~R~AN SEA5O10/,031)20 , , I , I - I i I] I 15 1 15 I 15 1 15 1 15 MAY JUNE JULY AUG SEP Fla. 10. Time series of the precipitable water from the surface to700 hPa over the Arabian Sea (thin line) from TIROS-N, and theprecipitation along the west coast of India during MONEX (adaptedfrom Cadet 1986).to those documented by Wang and Rui (1990). Murakami (1976), studying cloud data from eight JuneSeptember periods, used lag correlations to show thatclouds associated with the active phases propagatednorthward through the Indian Ocean and Indian subcontinent at about 1- latitude per day. These northwardpropagating clouds were oriented as northwest-southeast bands, The first paper to directly relate eastward propagation of clouds near the equator and the northward propagating bands was that of Yasunari (1979). He presented a longitude-time diagram for the northern summer of 1973 that shows clouds propagating eastwardnear the equator from 60-E to 150-W. The longitudinalrange is very similar to that suggested by the schematicof Fig. 3 and by the frequencies of occurrence documented by Wang and Rui (1990, and Fig. 8 here). Healso presented a latitude-time diagram of cloudinessalong the 72--84-E longitude zone in which northwardmovement of the clouds is clear. From these two figuresone could see that as the cloudy zone propagates eastward from 60- to 80-E, a part moves northward to30-N. This is an example of one of Wang and Rui'sEN modes. The northward propagation was confirmed' throughstudies of several Indian monsoons by Sikka andGadgil (1980), Yasunari (1980, 1981), Krishnamurtiand Subrahmanyam (1982), and Lau and Chart(1986a). Figure 11 is adapted from Lau and Chan(1986a) and shows the phenomena during MONEX.The two active periods shown in Fig. 10 correspond tothe negative OLR anomalies shown in the left-handpanel of Fig. 11 that reach 15-N in mid-June and lateJuly. The mid-June episode is clearly linked with eastward-moving clouds at 80-E and the equator (righthand panel of Fig. 11). The late July episode is notclearly linked to eastward-moving clouds at 80-E andthe equator but it does develop about the time of thepassage of an eastward-moving upper-level velocitypotential maximum and divergence as shown by Lorenc (1984), Krishnamurti et al. (1985), and Chert etal. (1988) and may be related to the eastward-movingclouds that are evident at 120-E in Fig. 11. The northward-propagating cloud zones often occur at approximately 30-40-day intervals. Indeed, Murakami's(1976) spatial correlations reach large negative valuesat lags of 16-20 days suggesting a period range of 3240 days. To stress again the variability of the oscillationwe point out that Mehta and Krishnamurti (1988)found regular northward propagation during the northern summers of 1979, 1982, and 1983 but not duringthe summers of 1980, 1981, and 1984. Interestingly,during the northern summers of 1980 and 1981, Knutson et al., (1986) still found regular eastward propagation of 250-hPa u-wind anomalies averaged from 0-to 10-S. The relevance of the 40-50-day equatorial oscillations to the break-active phases of the Indian monsoondemonstrated by Yasunari renewed our interest. To further substantiate Yasunari's suggestion that the northward movement of cloudiness and the break and activephases of the monsoon might be related to the equatorial 40-50-day oscillation, we published compositesof rainfall and cloudiness (Julian and Madden 1981)that showed the oscillation at Car Nicobar (9.2-N,92.8-E) and Port Blair (11.7-N, 92.7-E) in the Bay ofBengal and at Minicoy (8.3-N, 73.0-E) just south ofIndia (Fig. 12). The composites of Fig. 12 were basedon the 150-hPa zonal winds at Truk Island over 6000km to the east. When the 150:hPa zonal wind is a maximum (minimum), the precipitation and cloudiness isMAYJUNJUL %:x:._AUG '~2:-~OCT t.- ,~ 50S 25 0 ~5 50N FIo, 11. Longitude-time section of anomalies in outgoing longwave radiation along the equator (5-S-5-N ) during MONEX (righthand side). Latitude-time section along 80-E (75--85-E) for thesame period (left-hand side). Contours are watts per square meterwith negative values dashed and the zero contour suppressed. Timesof active monsoon phases from Fig. 10 are indicated on the left-handside by the dark bars (adapted from Lau and Chan 1986a).M~Y1994 MADDEN AND JULIAN 825.~5.5g5.0 4.5 4.0I ~ I I I I ! I II I I ! I i ! tI 2 3 4 5 6 7 8 ID E F G H A B C D Fla. 12. Composited clouds and rainfall for Minicoy (8.3-N,73.0-E) (dashed) and for Car Nicobar (9.2-N, 92.8-E) and Port Blair(11.7-N, 92.7-E) (solid). Letter phases are as in Fig. 3. Clouds arefor the 5- x 5- latitude-longitude square over Minicoy and 10- x 5-latitude-longitude square over Car Nicobar and Port Blair. The numbers at the bottom correspond to times of specific 150-hPa u-windanomalies at Truk (7.4-N, 151.8-E). Times 1 and 5 are those ofminimum and maximum u wind, respectively (from Julian and Madden 1981).a near maximum (minimum) over the Indian regionconsistent with a strengthening (weakening) of the circulation in the equatorial plane. The letters at the bottom of Fig. 12 approximately correspond to those ofFig. 3. It appears that during the Indian summer monsoonclouds arrive from Africa or form in the equatorial Indian Ocean and then move both northward at about 1-latitude per day and eastward at 5- longitude per dayon the 40-50-day time scale. Judging from Figs. 9aand 11, most frequently the clouds do not form veryfar south of the equator, but whether or not there areimportant influences originating in southern midlatitudes is an interesting question. Based on evidencefrom Tananarive (18.8-S, 47.5-W), Yasunari (1981)speculates that the origin of the clouds might be coldair outbreaks from Southern Hemisphere midlatitudes.Murakami (1987) noted that when there was clear eastward propagation of disturbances along the equatorthere was also a strong cold surge from the southernmidlatitude Indian Ocean at 850 hPa. While this evidence is intriguing, the role of southern midlatitudes inthe active periods of the Indian summer monsoon remains unknown. Further details of the behavior of these northwardmoving cloud zones were presented by Hartmann andMichelsen (1989) in their study of rainfall data from3700 Indian stations during 70 summer monsoons.They averaged data from all available stations in 1-squares and computed spectra. They found spectralpeaks near 40 days over most of the southern half ofIndia. A break in the monsoon over central India onthis time scale usually occurred during a time of maximum near-equatorial precipitation indicating a meridional scale of the cloud zones of I x 103-2 x 103 km.These were times of equatorial westerlies and easterliesover southern India. As the equatorial 40-50-day lowmoves eastward, a trough progresses northward overIndia and westerlies and rain return initiating an activeperiod. There are some smaller-scaled features that differ from this simple picture. For example, during abreak, Hartmann and Michelsen found a maximum inprecipitation in southeast India as easterlies north ofthe equatorial low release their moisture as they rise upthe Ghat Mountains.b. The Australian summer monsoon Holland (1986) noted an average of 40 days betweenactive bursts of the Australian summer monsoon. Undoubtedly some of these variations are related to theSE modes documented by Wang and Rui (1990). Hendon and Liebmann (1990b) showed that there was apronounced 30-50-day modulation of monsoonalwesterlies and that 27 out of 30 monsoon onsets from1957 to 1987 coincided with the arrival of clouds associated with an eastward-propagating 40-50-day oscillation (Hendon and Liebmann 1990a). However, thepoleward propagation of clouds that accompanies themajor active periods of the Indian monsoon was notfound. Hendon and Liebmann (1990a) pointed to anapparent poleward expansion of the clouds during active phases of the Australian monsoon rather than regular poleward progression. McBride (1983) reached asimilar conclusion based on his analysis of the 1978/79 monsoon. Certainly, the SE modes of Fig. 8b areinvolved.7. Effects in the extratroplcs Some have argued that there may be a separate, midlatitude 40-day oscillation (Dickey et al. 1991; Ghiland Mo 1991a,b). Indeed, modeling studies indicatethat there is an unstable global, barotropic mode witha 40-day period (Simmons et al. 1983). Also, Legrasand Ghil (1985) have proposed that there may be instability caused by the interaction of the jet stream andmountain ranges at midlatitudes whose dominant period is near 40 days. Supporting the idea that therecould be a 40-day midlatitude oscillation separate fromthe tropical one is the fact that the University of California, Los Angeles general circulation model is knownto exhibit a 50-day variation in the relative atmosphericangular momentum without having the tropical 40-50826 MONTHLY WEATHER REVIEW VOLUM-122day oscillation (Marcus et al. 1990). In addition, Magafia (1993) has observed small oscillations in the atmospheric angular momentum in the extratropics of 40(Northern Hemisphere) and 50 days (Southern Hemisphere) that are apparently independent of any tropicalactivity. Our purpose here is to describe only the tropical oscillation and any midlatitude manifestations that it mayhave. A disturbance that is of global scale longitudinally and that has such a profound effect on tropicalconvection, pressures, and circulation may also affectmidlatitudes. Anderson and Rosen (1983) have shownthat some of the 40-50-day variations in relative atmospheric angular.momentum propagates up and outof the tropics to midlatitudes. Nevertheless, it is ourcontention that robust midlatitude responses are hardto find. That it is likely that the "average" responsedoes not regularly occur due to the complexities of theever-changing background flows of extratropical latitudes, This section describes some work on midlatitudevariations thought to be related to the tropical oscillation, but we stress that any single event may not affectthe midlatitudes in the described manner. Weickmann (1983), Weickmann et al. (1985),Knutson et al. (1986), Knutson and Weickmann(1987), and Kiladis and Weickmann (1992) have described global aspects of the oscillation as they occurred in OLR and 250-hPa streamfunction and velocity potential. They show regular eastward propagationin the equatorial region of OLR and of velocity potential. In particular, the OLR anomalies follow the upperlevel divergence and are strongest over the Indian andwestern Pacific Oceans, negligible over the cooler waters of the eastern Pacific and the Atlantic, and weakbut present over South America and Africa. The OLRanomalies are also found to be strongest on the summerside of the equator. The zonal winds at 250 hPa propagate about 6 m s -~ from 40- to 160-E and then at 15m s-~ from 160-E to the Greenwich meridian. Whenthe clouds are in the Indonesian region the upper tropospheric flow tends to be characterized by twin anticyclones near the longitudes of the convection. To theeast, twin cyclones were found. These anticyclones andcyclones extend roughly +_.40- from the equator. Figure13 is a schematic taken from Weickmann et al. (1985)for the 250-hPa circulation. [A more detailed pictureis contained in Kiladis and Weickmann (1992), theirFigs. 5 and 6. ] This general picture is favored whenthe cloudiness is between 100- and 140-E and corresponds approximately to phase H or A of Fig. 3. Weickmann et al. point out that when the clouds push to thedate line the upper-level circulation is approximatelyreversed from that of Fig. 13. This is also a time whenan anomously clear region propagates eastward from60- to 140-E and corresponds to phase D or E of Fig.3. There is an expansion of the circumpolar vortex inregions of equatorial cloudiness and subtropical anticyclones and a contraction in regions of suppressedequatorial cloudiness and subtropical cyclones. Krishnamurti and Gadgil (1985) and Weickmann et al.(1985) showed that equatorial circulation anomaliesare out of phase in the vertical but that poleward ofabout 20- they are in phase or equivalent barotropic. During the May-October period, anomalous westerlies tend to occur at 850 and 250 hPa over southernAustralia 5-10 days after the peak convective activityoccurs in the western tropical Pacific (Knutson andWeickmann 1987). Graves and Stanford (1987) reported 45-53-day variations in the upper-troposphericzonal winds over Easter Island (27.1-S, 109.3-W).These variations were strongest in June-August. Theymay be related to the changes in the 250-hPa westerliesover Australia, although they have not as yet been directly related to them nor to the equatorial oscillation. Chen and Murakami (1988) have shown an interesting 40-day north to south oscillation of clouds along140-E (longitude near Japan) north of 20-N during thenorthern summer of 1979. These are undoubtedly related to fluctuations in the east Asian monsoon orMeiyu over China (Baiu over Japan) documented byLau and Chan (1986a) and Lau et al. (1988). Therehave also been published reports of 40-50-day spectralpeaks in the winds south of Japan (Matsuo 1984; Kai1985). Kai found these spectral peaks in surface windsover the Nansei Islands (25--30-N). They are predominant in the northern summer and thought to reflect typhoon activity. There is some evidence that tropicalcyclone formation may be related to the oscillation(Nakazawa 1986). A possible mechanism might beEkman pumping in the boundary layer poleward ofwesterly surges associated with the oscillation, whichwould provide large-scale convergence favorable fortyphoon and tropical cyclone formation. Kousky (1985) shows variations in troposphericthickness over the eastern United States. During the1984/85 northern winter, positive thickness anomalies(ridging) over the eastern United States occurred whenthe 40-50-day convection shifted from the IndianOcean to Indonesia. The opposite occurred when theconvection approached the central Pacific. This seemsto be consistent with the contracted and expanded circumpolar vortex of Fig. 13 over the United States (andits opposite) pointed to by Weickmann et al. (1985).Kousky's observations are based on only one season,so they must be considered with care. If such relationships stand up, they could offer some help in makinglong-range predictions for the United States.8. Atmospheric angular momentum Because the angular momentum of the earth-atmosphere-ocean system remains constant except for slowchanges due to tidal influences, a change in the angularmomentum of the atmosphere is reflected in oppositeMAY 1994MADDEN AND JULIAN'~ ~ ~. ~]~ ~ I ..... 1 ..... ] I~ ~~'~: ~--~$ , / "--1--~-" :~ ~ .... -'~ ,,,..e ~,~.,~'~. "~ ~- ~" ...... ,~ .~1 ' ' ' CONTRACTED~% ~t ~~ ~LI ~ - ..,~ ~A~t,-- m -~-=~=~ -% - ~ 1 ..... .......I/L~7~ 7ZLC~._1 ..... J -~ .~ ~] ~ ~, .... ,: 1~ -I~-!o~-ll~-I~-II~~l~o ~t~ l~ 2_.BO 80 ~.0 ~ ~ -~0 -4,0 -CO 080 Fro, 13. Schematic of relationship between outgoing longwave radiation as signified by "cloud" and"clear" regions and the 250-hPa circulation at a time when maximum cloudiness is in the eastern Indianand extreme western Pacific Oceans (from Weickmann et al. 1985 ).827changes in the oceans or the solid earth. Munk andMiller (1950) and Frostman et al. (1967) showed thatseasonal changes in the length of day (LOD), or rotation of the solid earth, could be explained by similarbut opposite changes in the relative angular momentum(AAM) of the atmosphere. These changes are about 1ms (10-3 s), with the longest days occurring duringnorthern winter when the AAM exceeds that of thenorthern summer by about 6 x 1025 kg m2 s-t. For anup-to-date review, see Rosen (1993). In 1974 Lambeck and Cazenave compared the LODand AAM and found that changes in LOD on timescales less than six cycles per year were caused bychanges in the AAM. They showed an oscillation in theAAM with a 50-day period from 20 February to 9 April1968 (Lambeck and Cazenave 1974, their Fig. 5).Feissel and Gambis (1980) reported a 50-day period inthe LOD during 1979, and Langley et al. (1981) documented the 50-day period in both LOD and AAM ina 4-yr period from 1976 through 1979. Spectra basedon time series (1976-81) from Rosen and Salstein'scontinuing monitoring of the AAM have peaks near 50day periods (Rosen and Salstein 1983). Gutzler andMadden (1992) showed that these peaks were the resultof variations that occurred primarily from mid-Januarythrough September. An example of the AAM variationthat occurred during MONEX is shown in Fig. 14. Theamplitude of the oscillations about the smoothly varying seasonal trend is on the order of 10% of the totalAAM and a few tenths of a millisecond change inthe LOD. Anderson and Rosen (1983) were the first to directlyrelate the 50-day AAM oscillations to the tropical oscillation. They showed that the zonal mean wind wascorrelated with the AAM on 45-day time scales fromabout 40-S to 60-N. The variations in the zonal windseemed to originate in the upper troposphere of theequatorial region and move poleward and downward.When the upper-tropospheric zonal wind maximamoved poleward of 20-N and 20-S, the AAM reachedits maximum (Anderson and Rosen 1983, their Fig. 8).The poleward propagation is shown nicely for the MONEX period by Magafia and Yanai (1991, their Fig.13). Although there is coherence between zonal meanwinds and AAM from 40-S to 60-N, Benedict and Haney (1988) and Gutzler and Madden (1993) showedthe magnitude of the oscillations are such that, on average, nearly all of the contribution to the 50-day oscillation in the AAM comes from variations in the zonalwinds between 20-N and 20-S. By correlating the LOD with equatorial western Pacific surface pressures we were able to link indirectlythe LOD and presumably AAM with the oscillation asit is depicted in Fig. 3 (Madden 1987). We concludedthat the LOD and AAM reach their maxima when theconvection associated with the oscillation is weakeningover the central Pacific (between phase B and D of Fig.3). During MONEX the relative maxima in AAM during late July and mid-August (Fig. 14) correspond withthe time that convective anomalies have progressed atleast to the date line (Fig. 11). This is also approximately the time of opposite phase or. slightly before the828 MONTHLY WEATHER REVIEW VOLUME122165 Observed151)~ ~55 MAY dUN JUL AUG 1979 F[o. 14. Obsemed an~lar momentum of the atmosphere daringMONEX taken from Rosen and Salstein (1983). Values at 0~0 and1200 UTC are plotted. Gaps reflect missing data. ~e dotted cu~erepresents an approximate seasonal variation. ~e amplitude of aco~esponding 0.l-ms change in the LOD is indicated. Sloping linesin lower left are the expected seasonal chang~ in ~ in units of10~8 kg m2 s-~ based on torques computed ~om longer data records(W~ and Oo~ 1984; Newton 1971) (from Madden 1988).opposite phase of Fig. 13 (Weickmann et al. 1985). Inaddition, we showed that when convection was strongover the eastern Indian and western Pacific Oceans(phase H and A of Fig. 3) surface wind stress associated with a strengthening of the easterlies over the central Pacific could serve as the main mechanism increasing the AAM (Madden 1987, 1988; Kang and Lau1990). Weickmann et al. (1992) studied the AAM cycleduring an oscillation. They found that minima in AAMoccur when the convection is enhanced over the IndianOcean and that a maximum increasing tendency ofAAM occurs when the convection moves from the Indian to the Pacific Ocean. This agrees with our conclusions. However, their work indicated that the singularrole of surface frictional stresses that we proposed isnot correct during northern winter. At that time, theyshowed that surface wind stresses over the central Pacific are nearly out-of-phase with AAM. If they werethe only driving force, the wind stresses would lead theAAM by 0.25 cycles. We computed the coherence andphase between AAM and the torque due to surfacewind stresses over the equatorial belt from 28.9-S to28.9-N for variations with periods of about 30-70 daysover a recent 4-yr period (Madden 1992b)~. We foundthat during n6rthern winter the wind stress ~orque leadsthe AAM by 0.4 cycles in good agreement with theout-of-phase (0.5 cycles) result of Weickmann et al.But from late July through October the wind stresstorque leads AAM by 0.25 cycles, which is at leastconsistent with the possibility that it has the major influence on the AAM. Weickmann et al. (1992) alsofound the 0.25-cycle relationship during the May-September period. Thus, fluctuations in tropical wind stresses associated with the oscillation are large and likely play animportant role in forcing the oscillations in AAM, butit is necessary to study global frictional stresses andmountain or pressure torques to evaluate the oscillation's AAM budget. Weickmann and Sardeshmukh(1994) computed frictional and mountain torques forDecember 1984 and January 1985. Their sum was ingood agreement with the AAM tendency during a 45day oscillation. Weickmann and Sardeshmukh (1994)showed that during this oscillation the frictional torqueand mountain torques contributed about equally. Thefrictional torque reached positive anomaly values first,about the time of largest negative anomaly AAM, inconcert with an out-of-phase relationship during thatseason. About 10 days later, anomaly mountain torqueswere at their maximum. In this connection, Ghil (1987)has argued that extratropical mountain torques may bean important contributor to the oscillation. The fact thatAnderson and Rosen (1983) have shown that anomalies in AAM propagate poleward also points to a rolefor extratropical torques. Mountain torques and frictional stresses over land are difficult to estimate, but itis clear that they will have to be considered in order toreach a full understanding of changes in AAM duringthe oscillation.9. The seasonal variation . There is some seasonal variation in the oscillationassociated with its roles in the Indian and Australianmonsoons. Also, researchers have found that the oscillation's cloud activity favors the summer hemisphereand the general location of the ITCZ (Zangvil 1975;Wang and Rui 1990; Wcickmann et aL 1985; Knutsonet al. 1986), a result that is also at least partly relatedto the monsoons. We have seen that the average periodof the oscillation shows no large change with season.There is, however, some change in strength as measured by the vertical shear in the u winds at Truk (Anderson et al. 1983; Madden 1986). Figure 15 is preMAY 1994MADDEN AND JULIAN829A f NA0 -Africe 0DKLH L pSONGI c . ~ I I~ II 12 II - I 1 e n DJF ~,* I44 ~, 8 ~ ~ I ffi In~nesia S. America I I I ~ E 180 90 W 0 Fto. 15. Summary of the filtered 150-hPa u-wind variance at 19equatorial stations. Longitudes of the stations are indicated by theirinitial and by the triangles along the top line. All stations are within14- of the equator. Initials and triangles above (below) the line arefor stations north (south) of the equator. The four lines below containthe results for northern spring (March, April, and May--IMAM),summer (June, July, and August--JJA), autumn (September, October, and November--SON), and winter (December, January, andFebraary--DJF). Vertical bars are included during a season for eachstation whose 47-day filtered variance exceeds that of its 31- and 99day filtered variances. This situation reflects a spectral peak at 47days. The number below a bar is the value of the 47-day filteredvariance, and the length of the bar indicates how much that varianceexceeds the average of the adjacent bands (left-hand scale). At Ascension Island (8.0-S, 14.4-W) and Recife (8.1-S, 34.9-W), variancein the 31-day band sometimes exceeded that in the other bands. Thosecases are indicated and there the bar and number represents 31-dayvariances. The longitudes where land masses intersect the equator areindicated at the bottom.sented to further illustrate seasonal changes in the oscillation as measured by the 150-hPa u wind at tropicalstations. It is taken from Table 2 in Madden (1986).The horizontal axis represents a zonal cross sectionalong the equator. The line at the top indicates the longitudinal location of the 19 stations that were studied.The u winds from these stations were bandpass filteredwith three slightly overlapping filters centered at 1/99,1/47, and 1/31 day. The half-power bandwidth of thefilters was about 1/100 day. The variance in each bandwas computed separately for each season. A verticalbar is present at a station if the 47-day variance exceedsthat in the other two bands. The numbers below thebars indicate the total variance in the 47-day band, andthe length of the bar indicates how much it exceeds theaverage variance of the other two bands. By these measures the oscillation is largest during December-February and smallest during June-August. It is smallestat stations in the western Pacific and largest at stationsin the Indian Ocean and at Canton and Balboa. Similarconclusions were reached by Gutzler and Madden(1989). It should be noted that the 47-day variance thatexceeds the average variance of the other two bands(i.e., that which exceeds a smoothly varying spectrum)ranges from only a few percent to 15% (at Canton andBalboa during December, January, and February) ofthe total daily variance [ deduced from Table 2 of Madden (1986)1. A feature of the oscillation specifically related to thevertical shear is the tendency for the u wind to be coherent and out-of-phase between the lower and uppertroposphere over the tropical Pacific and IndianOceans. We found that that feature had an interestingseasonal variation (Madden 1986). Figure 16 showsthe average coherence at 47-day periods between the850- and 150-hPa u wind at Koror (7.3-N, 134.5-E)and at Darwin (12.0-S, 130.9-E). The phase angles arenot shown but they indicate an out-of-phase relationship all year at Koror and during the time of maximumcoherence at Darwin. We interpret seasonal variations1.00 0.58134.5EDarwin12.0S 130.9 E 0 0 N D J F M AM J J A S FIG. 16. Seasonally varying coherence, squared between the 850and 150-hPa u winds at Koror (7.3-N, 134.5-E) (top) and at Darwin(12.0-S, 130.9-E) (bottom) at 47-day periods. Time runs from October through September (from Madden 1986).830MONTHLY WEATHER REVIEWVO~.V~E 122such as these that occur at stations in the Indian andwestern Pacific Oceans to result from the seasonal migration of convection associated with the ITCZ. TheITCZ is farthest south during northern winter, and thevertical coherence between the u winds is biggest atDarwin and other Southern Hemisphere stations at thattime. The opposite is true during northern summer.[See Fig. 7 of Madden (1986) for coherence at otherstations.] We originally thought the lack of coherence betweenthe u and v winds was a characteristic of the oscillation(Madden and Julian 1971). In fact, upper-troposphericu and. v are very coherent in the 40-50-day range atstations from at least Gan eastward to Balboa. This coherence, however, is revealed only by seasonal statistics. The reason is that u and v tend to be in phase duringnorthern summer and out of phase during northern winter. Figure 17 contains the coherence at 47-day periodsbetween u and v at 150 hPa over Singapore as a function of time of year. The maximum in coherence fromDecember through April is associated with an Out-ofphase relation, while that from June through Octoberreflects a very nearly in-phase relation. Because of thisseasonal phase change, cospectra (and coherence)based on time series that extend across seasons tend tobe near zero. This seasonal change in the phase relationship between the u and v winds is related' to the seasonalchange in the climatological winds. During northernwinter the upper flow over the equatorial Indian andwestern Pacific Oceans is primarily from the southeast,while it is from the northeast during northern summer.The eastward-moving clouds associated with the oscillation cause a strengthening and weakening of the meanflow as they pass. As a result, there are fluctuations inthe upper-level climatological northward transport ofmomentum (u and v in phase) across the equator duringnorthern summer and in the southward transport there(u and v out of phase) during northern winter.10. Oceans Along with the surface Wind stress, variations overthe Arabian Sea (Wylie and Hinton 1982) and the Pacific Ocean (Madden 1988), we should expect the effects of the oscillation to be manifest in the underlyingseas. Lau and Chan (1985, 1986b) proposed an atmosphere-ocean interaction as a possible link between theoscillation and the onset of the E1 Nifio. Krishnamurtiet al. (1988)found sea surface temperature, variationson 30-50-day time scales on the order of 0.5--1-C,with the strongest variations occurring over the equatorial western Pacific Ocean and the Bay of Bengal.They also studied sensible and latent heat fluxes between atmosphere andocean. In particular, during MONEX, the surface u wind over the Bay of Bengal (1 i-N,90-E) has relative maxima in late May~ late June, and1.00.5Singupore150-rob U-V47-d b=nd180 ~ 0.5 O ON D J F M AM J J A S FlG. 17. Seasonally varying coherence squared (bottom) and phase(top) between 150-hPa Singapore u and v winds at 47-day periods.Phase between 0.0 and 0.5 cycles means that u leads v (from Madden1986).early August (Krishnamurti et al. 1988, their Fig. 11).The latter two dates are ones of heavy rainfall overwestern India (Cadet 1986), strong winds over theArabian Sea (Wylie and Hinten 1982), and strongwinds over the entire tropical Pacific (Madden 1988).There are fluctuations in the latent heat flux at the surface over the Bay of Bengal too. Their amplitudes areroughly 40 W m-2 or about 20% of the average of 200W m-2 there (Krishnamurti et al. 1988, their Fig. 11).It is clear that the exchange of heat between ocean andatmosphere is affected significantly. In addition, thereis evidence for dynamic responses in the ocean to thewinds of the oscillation and some of those follow.a. Locally forced responses McPhaden noted that low-level zonal winds at GanIsland were coherent and in phase with ocean currentsto a depth of 100 m in the vicinity at 30-60-day periods(McPhaden 1982). The mixed-layer depth and the upper-thermocline temperature are similarly related to thewind. This is an example of the oscillation's wind variMAY1994 MADDEN AND JULIAN 831ations driving similar variations in the upper ocean. Another one is the Somali Current off Kenya. Mysak andMertz (1984) present evidence of a 40-60-day oscillation in the long-shore current and temperature off theKenyan coast. They conclude that this oscillation isdriven by the wind stress or the wind stress curl there.They offer further evidence to support this conclusionwith an examination of the surface wind stress over thewestern Indian Ocean during January-October 1976and January-September 1979 (Mertz and Mysak1984). At the equator and 48-E, a spectrum of the vwind stress has no excess power in the 40-60-day bandduring the 1976 period but the u wind stress does. Thewind-stress curl also has a spectral peak near 40 days.During 1979 both the u and v wind stresses have broadspectral peaks in the 40-50-day range. Another possible example of locally forced ocean responses to the40-50-day oscillation are 50-80-day fluctuations inthe current in the Arabian Sea (16-N, 60-E) during theMay-September period of 1986 (Shetye et al. 1991).Composites of low-level winds associated with the 4050-day oscillation indicate a relatively strong signalthere, so local forcing is possible (e.g., Murakami1984). More evidence for local forcing is found in Cadet (1986), who found 50-day peaks in spectra for the850-hPa u wind over the Arabian Sea east of 50-E.b. Remotely forced responses Not all of the ocean responses to the low-level windsof the oscillation are local. The remarkable 40-60-dayvariations in sea level height (SLH) that Enfield(1987) has shown to exist from at least Callao, Peru(12-S), in the south to San Francisco in the north during the 1979-84 period is the leading example. Luther(1980) reported 35-80-day spectra peaks in SLH fromCanton to the Galapagos Islands, and Enfield (1987)shows relatively high coherence and a nearly linearchange in phase across a broad frequency range between the SLH at Nauru in the west-central Pacific andTalara (81-W) on the coast of Peru. The linear changeof phase with frequency that he found is characteristicof a constant time delay between the two stations anda constant phase velocity. The phase velocity was eastward at about 2.9 m s-2 along the equator, which isconsistent with that expected for the first internal-modeKelvin wave (Enfield 1987). Apparently these Kelvin waves are excited west ofthe date line where the low-level winds have considerable power in the 40-50-day range. They propagateat 3 m s-~ eastward and then poleward along the westcoast of the Americas. Erickson et al. (1983) showeda coherence peak at 40-day periods between winds atBoru Island (1.3-S, 171.0-E) and an ocean pressuregauge at Isabela Island (0.0-S, 91.5-W) with a lag of43 days, which corresponds to a propagation velocityof 2.9 m s-~. Also, Enfield and Lukas (1984) foundthat when winds at 170-E are offset by 50 days, mostof the prominent features of the winds are aligned withthe SLH at Callao consistent with a propagation justunder 3 m s-2. Peaks in the spectra of SLH and bottompressure that may relate to these Kelvin waves havebeen found during various time periods and at severallongitudes in the equatorial Pacific (Mitchum and Lukas 1987; Chiswell et al. 1988).c. Other ocean variations There have been a number of reported 40-50-dayvariations in the ocean that cannot be clearly linked tothe atmosphere oscillation. Quadfasel and Swallow(1986) found a dominant 50-day period in currents justnorth of Madagascar during the first half of 1975.Schott et al. (1988) reported that 41% of the variancein the transport was accomplished by 40-55-day timescales in a nearby region from October 1984 to September 1985. They pointed out that there were no similar local wind variations on that time scale, and attributed the ocean variability to shear instabilities in themean flow. This conclusion is supported by the fact thatocean models of the region generate 50-day oscillationswhen forced only by monthly mean winds (Kindle andThompson 1989; Woodberry et al. 1989). Kindle andThompson attributed these variations to barotropic instability associated with the East African Coastal Current as an alternative to direct wind forcing. In the Gulfof Guinea (West Africa), Picaut and Verstraete (1976)found 40-50-day variations in sea surface temperatureand height, which they tentatively ascribed to a resonant mode responding to local atmospheric variations.11. The oscillation during MONEX The oscillation was active during the entire FGGEyear (e.g., Lorenc 1984). In particular, during theMay-September period we have seen rainfall variations on the west coast of India (Cadet 1986, Fig. 10here), associated clouds (Lau and Chan 1986a, Fig. 11here), and AAM (Madden 1988, Fig. 14 here). Therehave been numerous studies of the oscillation duringthat time and some of their results for the MONEXunderscore the very large scale effects of the oscillation. The northward-propagating cloud zones that moveover India (Fig. 11) were accompanied by similarlypropagating cyclonic vorticity zones, low pressure andconvergence in the lower troposphere, and anticyclonicvorticity and divergence aloft (Krishnamurti et al.1985; Mehta and Ahlquist 1986; Chen et al. 1988). Theupper-level divergence was related to the large-scale(zonal wave 1-2), eastward-propagating X field (Lorenc 1984; Krishnamurti et al. 1985; Nogues-Paegle andMo 1987; Chen et al. 1988). During the period 21-27 June, when the Indianmonsoon was active (Fig. 10), Murakami et al. (1984)832 MONTHLY WEATHER REVIEW VOLUME122show at least three major cyclonic synoptic disturbances in the cyclonic vorticity zone that extendedwest-northwest to east-southeast from the Arabian Seato approximately 135-E. This is consistent with thenorthward-propagating clouds along 80-E and the eastward-propagating ones along the equator documentedby Lau and Chan (1986a) (and Fig. 11 here). Murakami et al. (1984) noted that the synoptic disturbancesmoved toward the northwest along the trough linewhile the line itself moved northward and eastward.(These off-equator, westward-moving disturbanceshave a longer time scale and somewhat bigger spacescale than the CCs reported by Nakazawa.) This caseis a good example of the equatorial oscillation's relation to active and monsoon break periods. It is an EN(eastward and northward moving) cloud mode definedby Wang and Rui (1990). Another manifestation of asimilar trough line associated with an EN mode wasthat during 29 August-2 September of 1976. Wang andRui (1990) showed that a narrow band (1-2 x 103kin) of OLR anomalies less than -15 W m-2 extendedfrom India east-southeastward to the equator at 150-E,during that 5-day period. Krishnamurti et al. (1985)found a similar orientation in the 30-50-day filtered200-hPa X field during active monsoon phases on 24June and 29 July 1979 (their Fig. 5). Figure 18 shows the surface wind stress over thePacific during an active (20 June) and a break (10 July)period of the Indian monsoon. Positive wind stressmeans a transfer of eastward momentum from the earthto the atmosphere and occurs where the surface windsare easterly (u ~ 0). During the active phase (20 June)the wind stress is greater than that during the break (10July), particularly in the 10--30- latitude bands bothnorth and south of the equator. During the active phase,the subtropical anticyclones were strong over the Pacific in each hemisphere resulting in a broad band ofrelatively strong easterlies from 30-S to 30-N (Fig.18a). During the break phase, the subtropical anticyclones were weaker and midlatitude troughs extendednorth of 20-S east of Australia and near 130-W in theSouth Pacific, and south of 30-N at about 140-E and140-W in the North Pacific (Fig. 18b). Krishnamurtiand Subrahamanyan (1982) have remarked on the circulation change near 20-N, 140-E, south of Japan, withanticyclonic low-level flow present there on 19 June(active phase) and cyclonic flow there on 10 July(break phase), which is reflected in the change fromwesterly to easterly winds there (upper left-hand corners of Figs. 18a,b). ~The overall strengthening andweakening of the northern subtropical high with aroughly 45-day period is revealed by the remarkable,unfiltered (in time) time series of the 850-hPa heightaveraged over the region 25--45-N and 180--130-Wdiscussed by Chen (1987). It has minima in the heightbefore the monsoon, during the break, and after its retreat in early June, near 10 July, and late August, respectively. Maxima occur in late June and early Augustduring the monsoon active phases. We have already seen that extrema in the surfacewind stress vary from 0.60 N m-2 to 0.40 and back to0.55 N m-~ over the Arabian Sea during these sameactive and break phases (Wylie and Hinton 1982). Itshould be noted that these latter wind stress variationshave opposite sign of those over the Pacific, Madden(1988) ~hows that it is the stresses over the Pacific thatdominate and explain nearly all of the 40-50-dayanomalies in the tropical (32-S-32-N) frictionaltorques. It seems that the entire Pacific-Indian Ocean circulation pulsates with the oscillation. This pulsation isseen in the composites of the 850-hPa winds of Murakami (1984) during the period. Also, Chen (1985)studied the energetics in the tropics during MONEXand found large 45-day oscillations in eddy availablepotential and kinetic energy. Both had maxima during,or a few days after, the most active monsoon phases inlate June and early August. These maxima were nearlya factor of 2 larger than the minima before monsoononset, during the July break, and after monsoon retreat.Chen also showed a very large variation in the largescale (zonal wave 1) momentum transport along 10-Sin the upper troposphere with premonsoon (1-10June) values near zero, active phase values of 31 and58 m~ s-z (21-30 June and 1-10 August, respectively), and break (13-20 July) and postmonsoon retreat (22-31 August) values of 9 and 0 m: s-2. Undoubtedly the in-phase oscillations of 150-hPa u and vduring northern summer at Singapore .revealed in Fig.17 are a part of these large-scale variations. Figure 18a (20 June) corresponds to phases H or Ain Fig. 3 and to the upper-air schematic of Weickmannet al, (1985, Fig. 7) (Fig. 13 here). Comparing Fig.18a and Fig. 13 we can see the baroclinic structure ofthe wind anomalies over the equatorial Pacific (strongeasterlies near the surface and westerlies aloft) and thebarotropic structure poleward of about +20- (predominately strong easterlies at the surface and easterliesaloft). Comparing Fig. 18b corresponding to phases Dor E of Fig. 3 and the opposite upper-air pattern of Fig.13 reveals a similar equatorial baroclinic structure andextr~tropical barotropic structure. That is, the surfaceeasterlies are now (10 July) less strong everywhere andare overlaid by anomalous easterlies near the equatorand westerlies poleward of +20-. Table 2 is presented to relate the approximate timingof some of the phenomena observed during' MONEXwith the schematic of Fig. 3 for the oscillation in theequatorial plane and with that of Weickmann et al.(1985 Fig. 7) for the upper-level extratropics.12. Discussion Initial observational studies pointed out similarities in the oscillation's structure to that of an atmoMAY 1994 MADDEN AND JULIAN 833 2Oz IOa) 20 JUNE 1979 I ~ %: ,, ,: ~50 160 ~70 180 ~?0 ~60 150 ~40 ~30 ~20 ,0 I00 90 East West Longitude I~ ~0 IILJo ~ (~ :i?~i:i!::~iii~i!ii::f:i::iliiii~i~i?I II I I I I I ~50 160 170 ~80 ~70 ~60 East ~,...:~;;;;~::!i::'''~ '~- -, -:'-- .. - .' f:f:i I I I I I I I ~50 ~40 130 120 ,o ~oo 90West Longifude30IOI03::D 200~0 C) ZONAL AVERAGES-6 -4 -2 0 2 4 6 $ ~0 ~2! I I I I I eo- 20 June ,o ~-~.o .. ~oTime Avera( FiG. 18. Zonal wind stress over the tropical Pacific in unitsof 0.1 N m-2 during an (a) active monsoon phase, 20 June1972, and (b) break phase, 10 July 1972. Stippling signifiesnegative (u :> O) values. (c) Zonal averages of wind stressover the Pacific in units of 0.01 N m-2 fox 20 June 1979 (dotted), 10 July 1979 (dashed), and the May-August 1979 average (solid) (from Madden 1988).spheric Kelvin wave (Madden and Julian 1971; Parker 1973). There were some early modeling andtheoretical work that further suggested the possiblerole of the theoretically predicted Kelvin wave. Forexample, Holton (1973) found a maximum Kelvinwave response to white-noise forcing in the tropicalupper troposphere at periods longer than 30 days inhis linear, primitive equation model. There was aproblem, however, because the Kelvin wave speed isproportional to its vertical scale, and the relativelylarge vertical scale of the observed oscillation wouldresult in a propagation speed larger than observed.On the other hand, Lindzen (1974a,b) pointed outthat wave-CISK (conditional instability of the second kind) indicated a 10-m equivalent depth for thetropical troposphere, which would be consistent withthe observed slow speed. Another possible reconciliation between the observed, large vertical scale andthe slow phase speed could be provided through theinclusion of viscous damping as was shown by834 MONTHLY WEATHER REVIEW VOLUME 122 TABLE 2. Approximate timing of oscillation phenomena during MONEX with respect to that of the schematics of Fig. 3 and Fig. 7 in Weickman et al. (1985) (Fig. 13 here); AE and Ke denote eddy available potential energy and eddy kinetic energy, respectively.Symbolic dates from Fig. 3.F G H A B C D EIndian MonsoonTropical AE and KeLow-level wind and water vapor flux over Arabian SeaSubtropical low-level anficylcones over PacificAtmospheric angular momentumFrictional torqueExtratropics from schematic of Weickman et al. (1985) active maximum maximum strongminimum maximum maximum Fig. 13breakminimumminimumweakminimumopposite Fig. 13Chang (1977). In any case, exploration of the Kelvinwave was promising. Yamagata and Hayashi (1984) examined the possibility that the structure of the oscillation is that of theresponse of the tropical troposphere to localized heating. They forced the model of Gill (1980) with periodic, 40-day heating. The solution consists of a Kelvinmode and Rossby modes to the east and west of theheating, respectively (Yamagata and Hayashi 1984;Gill 1980). Yamagata and Hayashi noted that, near theequator, the u wind and the pressure perturbations ofthe Kelvin mode are nearly in phase, while those of theRossby mode response are nearly out of phase. SinceCanton and Majuro are nearly always east of the mainconvection, the observed relations between u and pressure in the lower troposphere are consistent with thismodel (Nishi 1989, and Table 1 here). The same cannot be said about the more nearly out-of-phase character of u and z above 500 hPa at Canton and Majuro.'Madden (1986) interpreted the high coherence between u and v shown in Fig. 17 as reflecting the presence of asymmetric Rossby waves near, or west of, themajor convection. While we think that the coupled Kelvin-Rossby mode model serves as an adequate firstorder model, it leaves many aspects of the oscillationunexplained. Besides the out-of-phase character of uand z in the upper troposphere, the schematic of thestructure east of the convection shown in Weickmannet al. (1985) and here as Fig. 13 differs from that ofan equatorial Kelvin wave. We have surveyed most of the more recent theoretical and modeling papers addressing the oscillation.They all agree on the critical importance of moist processes to explain it, and from the empirical evidencethis agreement is totally warranted. How the moist processes enter into any theoretical model is an importantaspect but that is beyond any critical comment wemight make. There is a feature of the descriptive material that we feel has not been given enough attentionor comment, possibly for good reason. It is the presenceof the mean zonal manifestation of the oscillation.However the zonal and meridional scale of the phenomenon is decomposed in theory (e.g., in zonal wavenumber), it should be remembered that the mean zonalstructure (wavenumber 0) is present in several variables. This would seem to be an important feature tobe explained. It also would seem to be the simplestnonphysical explanation of the coherence with theAAM and LOD quantities. Anderson and Stevens(1987) did study mean zonal modes in a zonally symmetric model. They suggested that the oscillation mayresult from the combined effects of zonally symmetricmodes and asymmetric traveling modes. The 40-50-day oscillation has provided fertileground for research by empiricists, theoreticians, andmodelers. Much is still to be learned. The oscillation isan important part of the general circulation, and because of its relatively long time scale, may help longrange forecasts (Cadet and Daniel 1988; Krishnamurtiet al. 1990; Ferranti et al. 1990; yon Storch and Xu1990). It is clear that as understanding of the oscillationcontinues to improve so will the understanding andlong-range prediction of weather and climate. Acknowledgments. We thank T. N. Krishnamurti forencouraging us to write this review. D. Cadet, K.-M.Lau, H. Lejen~is, H. van Loon, B. Wang, and K. Weickmann provided helpful comments on an earlier version.J. Martin patiently and skillfully typed several drafts.REFERENCESAnderson, J. R., and R. D. Rosen, 1983: The latitude-height structure of 40-50 day variations in atmospheric angular-momentum. J. Atmos. Sci., 40, 1584-1591.--., and D. E, Stevens, 1987: The presence of linear wavelike modes in a zonally symetric model of the tropical atmosphere. J. Atmos. Sci., 44, 2115-2127. , and P. R. Julian, 1984: Temporal variations of the trop ical 40-50 day oscillation. Mon. Wea. Rev., 112, 2431-2438.Benedict, W..L., and R. L. Haney, 1988: Contribution of tropical winds to subseasonal fluctuations in atmospheric angular mo mentum and length of day. J. Geophys, Res., 93, 15 973 15 978.Cadet, D. L., 1983: The monsoon over the Indian Ocean during sum mer 1975. Part II: Breaks and active monsoons. Mon. Wea. Rev., 111, 95-108.M~-1994 MADDEN AND J.ULIAN 835 ,1986: Fluctuations of precipitable water over the Indian ocean during the 1979 summer monsoon. Tellus, 38A, 170-177.--, and S. Greco, 1987: Water vapor transport over the Indian Ocean during the 1979 summer monsoon. Mon. Wea. Rev., 115, 653-663. , and P. Daniel, 1988: Long-range forecast of the break and active summer monsoons. Tellus, 40A, 133-150.Chang, C-P., 1977: Viscous internal gravity waves and low-fre quency oscillations in the tropics. J. Atmos. Sci., 34, 901-910.Chert, T-C., 1985: On the time-variation in the tropical energetics of large-scale motions during the FGGE Summer. Tellus, 37, 258 275. , 1987:30-50 day oscillation of 200-mb temperature and 850 mb height during the 1979 northern monsoon. Mon. Wea. Rev., 115, 1589-1605. , and M. Murakami, 1988: The 30-40 day variation of convec tive activity over the western Pacific Ocean with emphasis on the northwestern region. Mon. Wea. Rev., 116, 892-906. , R. Y. Tzeng, and M. C. Yen, 1988: Development and life cycle of the Indian monsoon--Effect of the 30-50 day oscil lation. Mon. Wea. Rev., 116, 2183-2199.Chiswell, S. M., M. Wimbush, and R. Lukas, 1988: Comparison of dynamic height measurements from an inverted echo sounder and an island tide-gauge in the central Pacific. J. Geophys. Res., 93, 2277-2283.Dakshinarmuti, J., and R. N. Keshavamurty, 1976: On oscillations of period around one month in the Indian summer monsoon. Indian J. Meteor. Hydrol. Geophys. 27, 201-203.Dickey, J. O., M. Ghil, and S. L. Marcus, 1991: Extratropical aspects of the 40-50 day oscillation in length-of-day and atmospheric angular momentum. J. Geophys. Res., 96, 22 643-22 658.Enfield, D. B., 1987: The intraseasonal oscillation in eastern Pacific sea levels--How is it forced. J. Phys. Oceanogr., 17, 1860 1876. , and R. B. Lukas, 1984: Low-frequency sea level variability along the South American coast in 1982-83. Trop. Ocean-At mos. Newslett., 28, 2-4.Erickson, C. C., M. B. Blumenthal, S. P. Hayes, and P. Ripa, 1983: Wind-generated equatorial Kelvin waves observed across the Pacific Ocean. J. Phys. Oceanogr., 13, 1622-1640.Feissel, M., and D. Gambis, 1980: La raise en 6vidence de variations rapides de la Dur6e de Jour. C. R. Hebd. Seances Acad. Sci., Set. B, 291, 271-273.Ferranti, L., T. N. Palmer, F. Molteni, and E. Klinker, 1990: Tropi cal-extratropical interaction associated with the 30-60 day os cillation and its impact on medium and extended range predic tion. J. Atmos. Sci., 47, 2177-2199.Findlater, J., 1969: A major low-level air current near the Indian Ocean during the northern summer. Quart. J. Roy. Meteor. Soc., 95, 362-380.Frostman, T. O., D. W. Martin, and W. Schwerdtfeger, 1967: Annual and semiannual variations in the length of day, related to geo physical effects. J. Geophys. Res., 72, 5065-5073.Ghil, M., 1987: Dynamics, statistics and predictability of planetary flow regimes. Irreversible Phenomena and Dynamical Systems Analysis in Geosciences. C. Nicolis and G. Nicolis, Eds., D. Reidel, 241-283.--, and K. Mo, 1991a: Intraseasonal oscillations in the global at mosphere. Part I: Northern Hemisphere and tropics. J. Atmos. Sci., 48, 752-779. , and --., 1991b: Intraseasonal oscillations in the global at mosphere. Part II: Southern Hemisphere. J. Atmos. Sci., 48, 780-790.Gill, A. E., 1980: Some simple solutions for heat-induced tropical circulation. Quart. J. Roy. Meteor. Soc., 106, 447-463.Graves, C. E., and J. L. Stanford, 1987: Low-frequency atmosphericoscillations over the southeastern Pacific. J. Atmos. Sci., 44,260-264.Gray, B. M., 1988: Seasonal frequency variations in the 40-50 day oscillation. J. Climatol. 8, 511-519.Gruber, A., 1974: Wavenumber-fyequency spectra of satellite measured brightness in tropics. J. Atmos. Sci., 31, 1675 1680.Gutzler, D. S., and R. A. Madden, 1989: Seasonal variations in the spatial structure of intraseasonal tropical wind fluctuations. J. Atmos. Sci., 46, 641-660. , and ,1993: Seasonal variations of the 40-50 day oscil lation in atmospheric angular momentum. J. Atmos. Sci., 50, 850-860.Hartmann, D. L., and M. L. Michelsen, 1989: Intraseasonal periodic ities in Indian rainfall. J. Atmos. Sci., 46, 2838-2862.Hendon, H. H., and B. Liebmann, 1990a: A composite study of onset of the Australian summer monsoon. J. Atmos. Sci., 47, 2227 2240. , and , 1990b: The intraseasonal (30-50 day) oscillation of the Australian summer monsoon. J. Atmos. Sci., 47, 2909 2923.Holland, G. J., 1986: Interannual variability of the Australian summer monsoon at Darwin--1952-82. Mort. Wea. Rev., 114, 594 604.Holton, J. R., 1973: On the frequency distribution of atmospheric Kelvin waves. J. Atmos. Sci, 30, 499-501.Julian, P. R., and R. A. Madden, 1981: Comments on a paper by T. Yasunari, a quasistationary appearance of 30 to 40 day period in the cloudiness fluctuations during the summer monsoon over India. J. Meteor. Soc. Japan, 59, 435-437.Kai, K., 1985: Spectrum climatology of the surface winds in Japan. 1: The 40-60 day fluctuations. J. Meteor. Soc. Japan, 63, 873 882.Kang, I.-S., and K.-M. Lau, 1990: Evolution of tropical circulationanomalies associated with 30-60 day oscillation of globally averaged angular momentum during northern summer. J. Meteor.Soc. Japan, 68, 237-249.Khalsa, S. J. S., and E. J. Steiner, 1988: A TOVS dataset for study of the tropical atmosphere. J. Appl. Meteor., 27, 851-862.Kiladis, G., and K. M. Weickmann, 1992: Circulation anomalies as sociated with tropical correlation during northern winter. Mon. Wea. Rev, 120, 1900-1923.Kindle, J. C., and J. D. Thompson, 1989: The 26-day and 50-day oscillations in the western Indian Ocean--Model results. J. Geophys. Res., 94, 4721-4736.Knutson, T. R., and K. M. Weickmann, 1987:30-60 day atmo spheric oscillations: Composite life cycles of convection and circulation anomalies. Mon. Wea. Rev., 115, 1407-1436. , and J. E. Kutzbach, 1986: Global-scale intraseasonal oscillations of outgoing longwave radiation and 250 mb zonal wind during northern hemisphere summer. Mon. Wea. Rev., 114, 605-623.Kousky, V. E., 1985: The global climate for December 1984-Feb ruary 1985--A case of strong intraseasonal oscillations. Mon. Wea. Rev., 113, 2158-2172.Krishnamurti, T. N., and P. A. Ardanuy, 1980: The 10 to 20-day westward propagating mode and "breaks in the monsoons." Tellus, 32, 15-26.---, and D. Subrahmanyam, 1982: The 30-50 day mode at 850 mb during MONEX. $. Atmos. Sci., 39, 2088-2095. - - , and S. Gadgil, 1985: On the structure of the 30 to 50 day mode over the globe during FGGE, Tellus, 37A, 336-360. , D. K. Jayalcumar, J. Sheng, N. Surgi, and A. Kumar, 1985: Divergent circulation on the 30 to 50 day time scale. J. Atmos. Sc-, 42, 364-375. , D. K. Oosterhof, and A. V. Mehta, 1988: Air-sea interaction on the time scale of 30 'to 50 days. J. Atmos. Sci., 45, 1304 1322.---, M. Subrahmanyam, D. K. Osterhof, and G. Daughenbaugh, 1990: Predictability of low frequency modes. Meteor. Atmos. Phys., 44, 63-83.Kuhnel, I., 1989: Spatial and temporal variation in Australia-Indo nesian region cloudiness. Int. J. Climatol, 9, 395-405.836 MONTHLY WEATHER REVIEW VOLttM- 122Lambeck, K., and A. Cazenave, 1974: The earth's rotation and at mospheric circulation--ii. The continuum. Geophys. J. Roy. As tron. Soc., 38, 49-61.Langley, R. G., R. W. King, I. I. Shapiro, R. D. Rosen, and D. A. Salstein, 1981: Atmospheric angular momentum and the length of day: A common fluctuation with a period near 50 days. Na ture, 294, 730-732.Lau, K.-M., and P. H. Chan, 1983: Short-term climate variability and atmospheric teleconnections from satellite-observed outgoing longwave radiation simultaneous relationships. J. Atmos. Sci., 40, 2735-2750. , and ,1985: Aspects of the 40-50 day oscillation during the northern winter as inferred from outgoing longwave radia tion. Mon. Wea. Rev., 113, 1889-1909.--, and ,1986a: Aspects of the 40-50 day oscillation during the northern summer as inferred from outgoing longwave radi ation. Mon. Wea. Rev., 114, 1354-1367.--, and ,1986b: The 40-50 day oscillation and the El Nino/ Southern Oscillation--A new perspective. Bull. Amer. Meteor. Soc., 67, 533-534.--, G. J. Yang, and S. H. Shen, 1988: Seasonal and'intraseasonal climatology of summer monsoon rainfall over East Asia. Mon. Wea. Rev., 116, 18-37.--, T. Nakazawa, and C. H. Sui, 1991: Observations of cloud clus ter hierarchies over the tropical western Pacific. J: Geophys. Res., 96, 3197-3208.Legras, B., and M. Ghil, 1985: Persistent anomalies, blocking and variations in atmospheric predictability. J. Atmos. Sci., 42, 433 471.Lindzen, R. S., 1974a: Wave-CISK in the Tropics. J. Atmos. Sci., 31, 156-179. , 1974b: Wave-CISK and tropical spectra. J. Atmos. Sci., 31, 1447-1449.Lorenc, A. C., 1984: The evolution of planetary-scale 200-mb diver gent flow during the FGGE year. Quart. J. Roy. Meteor. Soc., 110, 427-441.Luther, D. S., 1980: Observations of long period waves in the tropical oceans and atmosphere. PhD. thesis, Massachusetts Institute of Technology-Woods Hole Oceanographic Institution, 210 pp.McBride, J. L., 1983: Satellite-observations of the Southern Hemi sphere monsoon during winter MONEX. Tellus, 35a, 189-197.McPhaden, M. J., 1982: Variability in the central equatorial Indian Ocean: Ocean dynamics. J. Mar. Res., 40, 157-176.Madden, R. A, 1986: Seasonal variations of the 40-50 day oscilla tion in the Tropics. J. Atmos. Sci., 43, 3138-3158.--, 1987: Relationships between changes in the length of day and the 40 to 50 day oscillation in the tropics. J. Geophys. Res., 92, 8391-8399.--, 1992a: Large intraseasonal variations in wind stress over the tropical Pacific. J. Geophys. Res., 93, 5333-5340.--, 1992b: Changes in atmospheric angular momentum associated with the intraseasonal oscillation. Trends Geophys. Res. Council Sci. Res. Integration Pubs., 1, 263-272.--, and P. R. Julian, 1971: Description of a 40-50 day oscillation in the zonal wind in the tropical Pacific. J. Atmos. Sci., 28, 702 708.--, and ,1972a: Description of global-scale circulation cells in the tropics with a 40-50 day period. J. Atmos. Sci., 29, 1109 - 1123.--, and ., 1972b: Further evidence of global-scale 5-day pres sure waves. J. Atrnos. Sci., 29, 1464-1469.Magafia, V., 1993: The 40- and 50-day oscillations in atmospheric angular momentum at various latitudes. J. Geophys. Res., 98, 10 441-10 450.--, and M. Yanai, 1991: Tropical-midlatitude interaction on th~ time scale of 30 to 60 days during the northern summer of 1979. J. Climate, 4, 180-201.Markus, S. L., M. Ghil, J. O. Dickey, and T. M. Eubanks, 1990: Origin of the 30-60 day oscillation in the length of day and atmospheric angular momentum: New findings from the UCLA general circulation model. The Earth's Rotation and Reference Frames for Geodesy and Geodynarnics, G. A. Wilkins, Ed., Springer-Verlag.Maruyama, T., 1967: Large-scale disturbances in the equatorial lower ' stratosphere. J. Meteor. Soc. Japan, 45, 391-408. ,1968: Time sequence of power spectra of disturbances in the equatorial lower stratosphere in relation to the quasi-biennial oscillation. J. Meteor. Sob. Japan, 46, 327-342.Matsuo, T., 1984: About 30-day, 40-day, and 50-day period oscil lations, emerging from time variations of meteorological ele ments around Japan. Pap. Meteor. Geophys., 35, 181-197.Mehta, A. V., and T. N. Krishnamurti, 1988: InterannUal variability of the 30-50 day wave motions. J. Meteor. Soc. Japan, 66, 535-548.Mehta, V. M., and J. E. Ahiquist, 1986: Interannual variability of the 30-50 day activity in the Indian-summer monsoon. Meteor. At mos. Phys., 35, 166-176.Mertz, G. J., and L. A. Mysak,' 1984: E~,idence ~or a 40-60 day oscillation over the western Indian ocean during 1976 and 1979. Mon. Wea. Rev., 112, 383-386.Mitchum, G. T., and R. Lukas, 1987: The latitude frequency structure of Pacific sea-level variance. J. Phys. Oceanogr., 17, 2362 2365.Munk, W. H., and R. L. Miller, 1950: Variation in the earth's angular velocity resulting from fluctuations in atmospheric and oceanic circulation. Tellus, 2, 93-101.Murakami, M., 1984: Analysis of the deep convective activity over the western Pacific and Southeast Asia 2. Seasonal and intra seasonal variations during northern summer. J. Meteor. Soc. Ja pan, 62, 88-108.Murakami, T., 1976: Cloudiness fluctuations during the summer mon~ooh. J. Meteor. Soc. Japan, 54, '175-181.---, 1987: Intraseasonal atmospheric teleconnection patterns during the northern hemisphere summer. Mon. Wea. Rev., 115, 2133 2154.---, T. Nakazawa, and J. He, 1984: On the 40-50 day oscillations during the 1979 Northern Hemisphere summer: 1: Phase prop agation. J. Meteor. Soc. Japan, 62, 440-468.Mysak, L. A., and G. J. Mertz, 1984: A 40-day to 60-day oscillation in the source region of the Somali Current during 1976. J. Geo phys. Res., 89, 711-715.Nakazawa, T., 1986: Intraseasonal variations in OLR in the tropics during the FGGE year. J. Meteor. Soc. Japan, 64, 17-34. , 1988: Tropical super clusters within intraseasonal variations Over the western Pacific. J. Meteor. Soc. Japan, 66, 823-839.Newton, C. W., i971: Global angular momentum balance: Earth torques and atmospheric fluxes. J. Atmos. Sci., 28, 1329-1341.Nishi, N., 1989: Observational stady on the 30-60 day variations in geopotential and temperature fields in the equatorial region. J. Meteor. Soc. Japan, 67, 187-203.Nogues-Paegle, J., and K. Mo, 1987: Spring-to-summer transitions of global circulations during May-July 1979. Mort. Wea. Rev., 115, 2088-2102.Parker, D. E., 1973: Equatorial Kelvin waves at 100 millibars. Quart. J. Roy. Meteor. Soc., 99, 116-129.Picaut, J., and J.-M. Verstraete, 1976: Discovery of a 40-50 day frequency current affecting coasts of Gulf of Guinea. Cah. OR STOM Ocean., 14, 3-14.Quadfasel, D. R., and J. C. Swallow, 1986: Evidence for 50-day period planetary-waves in the south equatorial current of the Indian Ocean. Deep Sea Res., 33, 1307-1312.Raghavan, K., D. R. Sikka, and S. V. Gujar, 1975: The influence of cross-equatorial flow over Kenya on the rainfall of western dia. Quart. J. Roy. Meteor. S. oc., 101, 1003-1004.Rosen, R. D., 1993: The axial momentum balance of earth and its fluid envelope. Surv. Geophys., 14, in press. , and D. A. Salstein, 1983: Variations in atmospheric angular momentum on global and regional scales and the length of day. J. Geophys. Res., 88, 5451-5470.MAY 1994 MADDEN AND JULIAN 837Schott, F., M. Fieux, J. Kindle, J. Swallow, and R. Zantopp, 1988: The boundary currents east and north of Madagascar. 2: Direct measurements and model comparisons. J. Geophys. Res., 93, 4963-4974.Shetye, S. R., S. C. Shenoi, and D. Sundar, 1991: Observed low frequency currents in the deep mid-Arabian Sea. Deep Sea Res., 38, 57-65.Sikka, D. R., and S. Gadgil, 1980: On the maximum cloud zone and the ITCZ over Indian longitudes during the southwest monsoon. Mon. Wea. Rev., 108, 1840-1853.Simmons, A. J., J. M. Wallace, and G. W. Branstator, 1983: Baro tropic wave propagation and instability and atmospheric tele connection patterns. J. Atmos. Sci., 40, 1363-1392.Sui, C.-H., and K.-M. Lau, 1992: Multiscale phenomena in the trop ical atmosphere over the western Pacific. Mon. Wea. Rev., 120, 407-430.yon Storch, and J. Xu, 1990: Principal oscillation pattern analysis ofthe 30- to 60-day oscillation in the tropical troposphere. ClimateDyn., 4, 175-190.Wahr, J. M., and A. H. Oort, 1984: Friction and mountain torques and atmospheric fluxes. J. Atmos. Sci., 41, 190-204.Wallace, J. M., and C.-P. Chang, 1969: Spectrum analysis of large scale wave disturbances in the tropical lower t~oposphere. J. Atmos. Sci., 26, 1010-1025. , and V. E. Kousky, 1968: Observational evidence of Kelvin waves in the tropical stratosphere. J. Atmos. Sc-, 25, 900-907.Wang, B., and H. Rui, 1990: Synoptic climatology of transient trop ical intraseasonal convection anomalies. Meteor. Atmos. Phys., 44, 43-61.Weickmann, K. M., 1983: Intraseasonal circulation and outgoing longwave radiation modes during Northern Hemisphere winter. Mon. Wea. Rev., 111, 1838-1858.--, and S. J. S. Khalsa, 1990: The shift of convection from the Indian Ocean to the western Pacific Ocean during a 30-60 day oscillation. Mon. Wea. Rev., 118, 964-978.---, and P. D. Sardeshmukh, 1994: The atmospheric angular mo mentum cycle associated with the Madden-Julian oscillation. J. Atmos. Sci., in press. , G. R. Lussky, and J. E. Kutzbach, 1985: Intraseasonal (30-60 day) fluctuations of outgoing longwave radiation and 250 mb stream function during northern winter. Mon. Wea. Rev., 113, 941-961. , S. J. S. Khalsa, and J. Eischeid, 1992: The atmospheric angular momentum cycle during the tropical Madden-Julian oscillation. Mon. Wea. Rev., 120, 2252-2263.Woodberry, K. E., M. E. Luther, and J. J. O'Brien, 1989: The wind driven seasonal circulation in the southern tropical Indian Ocean. J. Geophys. Res., 94, 17 985-18 002.Wylie, D. P., and B. B. Hinton, 1982: The wind stress patterns over the Indian Ocean during the summer monsoon of 1979. J. Phys. Oceanogr., 12, 186-199.Yamagata, T., and Y. Hayashi, 1984: A simple diagnostic model for the 30-50 day oscillation in the tropics. J. Meteor. Soc. Japan, 62, 709-717.Yanai, M., and T. Maruyama, 1966: Stratospheric wave disturbances propagating over the equatorial Pacific. J. Meteor. Soc. Japan, 44, 291-294. , --, T. Nitta, and Y. Hayashi, 1968: Power spectra of large scale disturbances over the tropical Pacific. J. Meteor. Soc. Ja pan, 46, 308-323.Yasunari, T., 1979: Cloudiness fluctuations associated with the Northern Hemisphere summer monsoon. J. Meteor. Soc. Japan, 57, 227-242.--, 1980: A quasi-stationary appearance of the 30-40 day period in the cloudiness fluctuations during the summer monsoon over India, J. Meteor. Soc. Japan, 58, 225-229. , 1981: Structure of an Indian summer monsoon system with around 40-day period. J. Meteor. Soc. Japan, 59, 336-354.Zangvil, A., 1975: Temporal and spatial behavior of large-scale dis turbances in tropical cloudiness deduced from satellite bright ness data. Mon. Wea. Rev., 103, 904-920.