Spectral comparison of ENSO and stratospheric zonal winds

INTERNATIONAL JOURNAL OF CLIMATOLOGY

Int. J. Climatol. 18: 1195–1208 (1998)

SPECTRAL COMPARISON OF ENSO AND STRATOSPHERIC ZONALWINDSR.P. KANE*

Instituto Nacional de Pesquisas Espaciais—INPE, Caixa Postal 515, 12201-9709, Sao Jose dos Campos, SP, Brazil

Recei6ed 11 No6ember 1996Re6ised 27 February 1998Accepted 11 March 1998

ABSTRACT

The 12-monthly running means of stratospheric low latitude zonal winds at 10, 15, 20, 30, 40, 50, 70 mb, SouthernOscillation Index represented by Tahiti−Darwin atmospheric mean sea level pressure difference (T−D) andequatorial eastern Pacific sea surface temperature (SST) were subjected to maximum entropy spectral analysis for thewhole period 1953–1995 as also for its portions. The SST and (T−D) had similar periodicities, as expected, themajormost at ca. 4.5 years and other smaller ones in the quasi-biennial and quasi-triennial regions. The stratosphericwinds had the most prominent periodicity at ca. 2.17 years (26 months) in the early years, which changed to ca. 2.22years (26.5 months) in the middle and ca. 2.50 years (30 months) in recent years, and had other minor periodicities.Some of the wind periodicities matched some of the SST, (T−D) periodicities during some intervals, indicatingpossible partially common origins and/or mutual influences. © 1998 Royal Meteorological Society.

KEY WORDS: stratospheric wind; maximum entropy spectrum analysis (MESA); sea surface temperature; Southern Oscillation;quasi-biennial oscillation; periodicities

1. INTRODUCTION

A quasi-biennial oscillation in low latitude stratospheric zonal winds was discovered several decades ago(Reed et al., 1961; Veryard and Ebdon, 1961; Angell and Korshover, 1962). Since then, detailedcharacteristics have been reported by several workers (Naujokat 1986, and references therein). Thewesterlies and easterlies remain so for several months but change from one to the other within a fewmonths. The maxima occur later at lower altitudes, by ca. 10 months from 10 mb to 50 mb. A theoreticalexplanation was given by Lindzen and Holton (1968) and Holton and Lindzen (1972) in terms ofabsorption in the stratosphere of vertically propagating Kelvin and Rossby-gravity waves generated in thetroposphere. Plumb and Bell (1982) gave a numerical model which reproduces many of these features.

Another phenomenon known to have quasi-biennial and quasi-triennial oscillations (QBO and QTO) isEl Nino/Southern Oscillation (ENSO). In some years, warm water episodes (El Nino) occur at thePeru–Ecuador coast. In the same years, the Southern Oscillation Index (SOI), a see-saw between theatmospheric pressures in different parts of the Pacific Ocean (e.g. Tahiti−Darwin, T−D) attains aminimum and the equatorial eastern Pacific sea surface temperature (SST) attains a maximum. Thefrequency of such events is irregular (2–7 years).

The relationship between stratospheric zonal winds and ENSO is controversial. The winds have a QBOspacing of ca. 2.25 years, while El Ninos have an average spacing of ca. 4.5 years. Angell (1992) comparedthe time series of these two phenomena and found a certain pattern, namely an El Nino tended to beassociated with a QBO east-wind maximum except when the El Nino associated with the previous

* Correspondence to: Instituto Nacional de Pesquisas Espaciais—INPE, Caixa Postal 515, 12201-9709, Sao Jose dos Campos, SP,Brazil; e-mail: [email protected]

Contract grant sponsor: FNDCT Brazil; Contract grant number: FINEP-537/CT

CCC 0899–8418/98/111195–14$17.50© 1998 Royal Meteorological Society

R.P. KANE1196

east-wind maximum was a major one and/or the SST maximum followed the previous east-windmaximum by a few seasons. He also cautioned that because of the intermittency of this relationship,conventional lagged-correlation or co-spectral analyses will not indicate a significant relationship betweenQBO and SST. In this paper the spectral characteristics of QBO, (T−D) and SST are compared.

2. DATA

The stratospheric zonal wind data series were obtained from Pawson et al. (1993) and further data fromDr. Barbara Naujokat, (personal communication) and were compiled from data at three stations:

Station Location Time period

Canton Island 2°46%S, 171°43°W January 1953–September 1967Gan, Maldive Island 0°41%S, 73°09%E September 1967–December 1975Singapore 1°22%N, 103°55%E January 1976–June 1996

Thus, the latter half of the data was from Singapore alone. Data for Singapore, Balboa and AscensionIsland were also obtained privately from Dr. Angell. Data for the Southern Oscillation Index (SOI)represented by Tahiti−Darwin pressure difference (T−D) were obtained from Parker (1983) and wereupdated from Meteorological Data Reports of the World. SST data were obtained from Angell (1981)and further personal communication.

3. METHOD OF ANALYSIS

The method used was maximum entropy spectral analysis (MESA, Burg, 1967; Ulrych and Bishop, 1975),known to be much more accurate than the conventional Blackman and Tukey (1958) (BT) method. As anillustration, Figure 1 shows the spectra for an artificial sample of 160 data points obtained by superposingsix sinusoids of periodicities 8.0, 8.20, 8.40 and 12.0, 12.4, 12.8. Thus, the spectra should show only twoclosely spaced triplets. In MESA, there is a parameter called length of the prediction error filter (LPEF).For low LPEF, only low periodicities are resolved. For higher LPEF, higher and higher periodicities areresolved, even those near the data length; but lower periodicities show peak splitting. Experience showsthat LPEF ca. 50% of the data length (in the present case, LPEF=80) shows reasonably good resolutionfor a wide range of periodicities. In Figure 1, the LPEF for the upper plot is 30%. If the 160 data pointsare considered as 40 years data with four seasonal points (trimester) per year, the input periodicities 8.0,8.2, 8.4 and 12.0, 12.4, 12.8 would be equivalent to 2.00, 2.05, 2.10 and 3.09, 3.10, 3.20 years. As can beseen from Figure 1, these are resolved reasonably well even at 30% LPEF. At 50% LPEF (the lower plot)the resolution is still better. Thus, in the QBO and QTO region, the peak detection would be veryaccurate. In contrast, the BT spectra for this sample (with smoothing by Jenkins and Watts method) weretwo broad humps as shown by the dashed lines in Figure 1, thus bringing out the superiority of MESAover BT method. Since the data used in the present paper are from 1953 up to 1996 (ca. 42 years) andwe propose to use four values per year, the wind data are very similar in length to the artificial sampleillustrated above and the peak detection should be accurate to about 90.02 years (95% confidence level).Results of analysis of many other artificial samples have been reported by Chen and Stegen (1974) andKane (1977, 1979) and show that MESA can detect periodicities in a very wide range; but, errors increasefor larger periodicities.

In spite of the high accuracy of peak detection, MESA has a major defect, namely the power estimatesare not reliable. In the above artificial sample, all the sinusoids were allotted the same amplitude. But in

© 1998 Royal Meteorological Society Int. J. Climatol. 18: 1195–1208 (1998)

ENSO AND STRATOSPHERIC ZONAL WINDS 1197

Figure 1, the peaks have very different heights (note that the ordinate scale is logarithm of power) andthe relative proportions change with LPEF. According to Lacoss (1971), in MESA, the heights of theoutput peaks are expected to be proportional to the square of the power of the signal and the area underthe peak is expected to be proportional to the power of the signal. However, as shown in Kane andTrivedi (1982), this is not invariably true. Hence, an alternative procedure was adopted, 6iz. MESA wasused only for locating all possible peaks Tk (k=1 to n) (e.g. T1 . . . T6 in the above sample) and then, theseTk were used in the expression:

f(t)=Ao+ %n

k=1

[ak sin(2pt/Tk)+bk cos(2pt/Tk)]+E=Ao+ %n

k=1

rk sin(2pt/Tk+fk)+E (1)

where f(t) is the observed series and E the error factor. A MRA (multiple regression analysis, Bevington1969) was then carried out to obtain the best estimates of Ao, (ak, bk) and their standard errors. (TheMRA is not related to MESA and the estimates are obtained by a least-square fit). From these, rk andtheir standard error sr (common for all rk in this methodology, which assumes white noise) can beestimated and rk exceeding 2sr may be considered as significant at a 95% (a priori ) confidence level.

4. CHECKING LONGITUDE INDEPENDENCE

The wind data (Pawson et al., 1993) used in this analysis are a mixture of data from three low latitudelocations at widely different longitudes (171°W, 73°E, 104°E). Several workers have mentioned that datafrom all these show similar characteristics (Naujokat 1986; Barnston and Livezey, 1989). Dunkerton(1990) compared the data for Singapore (1°N, 104°E), Balboa (9°N, 80°W) and Ascension Island (8°S,14°W) and showed good agreement, except that amplitudes of QBO were somewhat larger at Singapore(near the equator). The locations Canton and Gan are also very near equator. Angell (1986) reportedsome differences between Balboa (9°N) and Ascension Island (8°S), related to the effect of the eruptionof the volcano (Agung) in 1963. Since we are obtaining periodicities to an accuracy of 0.01 years, it isnecessary to check that changes of location have not caused differences.

Figure 1. Maximum entropy spectral analysis of an artificial sample of 160 data points having input sinusoids of periodicities 8.0,8.2, 8.4 and 12.0, 12.4, 12.8 which, if considered as four seasonal values per year, are equivalent to periodicities 2.00, 2.05, 2.10 and3.0, 3.1, 3.2 years. Upper plot for LPEF=53 (30% of data length) and lower plot for LPEF=80 (50% of data length). The dashed

line is the Blackman and Tukey spectrum for a lag m=40°


R.P. KANE1198

Figure 2. Monthly means series of the 50 mb zonal wind for: (a) Canton for 1953–1967 and Gan for 1967–1975; (b) Singapore for1953–1975; (c) 3-monthly means for Canton and Gan (full line) and Singapore (crosses); and (d) 4 season (12-monthly) running

means for Canton and Gan (full line) and Singapore (crosses)

Figure 2 shows plots of the monthly means of the 50 mb zonal wind for (a) Canton Island, 2°46%S,171°43%W, Janaury. 1953–September 1967 and Gan, Maldive Island, 0°41%S, 73°09%E, September 1967–December 1975 and (b) Singapore alone, 1°22%N, 103°55%E. Data in Figure 2(a) were obtained fromPawson et al. (1993) and are in units of (m/s). Data in Figure 2(b) were obtained from Dr. Angellprivately and are in units of knots (1 knot=ca. 0.4 m/s). Some well-known features are visible such aslonger durations of westerlies (positive values) as compared to easterlies (negative values) and sharpreversals (easterlies to westerlies and 6ice-6ersa). There are minor differences between (a) and (b) plots;but the major easterlies (minima in the plots) tally exactly, as shown by the vertical lines. Figure 2(c) (fulllines) shows the 3-monthly means (JFM, AMJ, JAS, OND) for the Pawson data and crosses representSingapore data alone. Whereas minima tally quite well, some differences persist. Figure 2(d) shows the12-monthly (4 season) running means. The match is reasonably good. Hence, for analysis, only12-monthly running means centered 3 months apart (4 values per year) were used.

From daily radiosonde observations, Pawson et al. (1993) obtained and reported monthly mean zonalwinds at 70, 50, 30, 20 and 10 mb, with values at 40 and 15 mb interpolated linearly in pressure scaleheights, for 1953–1993. Dr. Barbara Naujokat kindly supplied data up to mid-1996. Figure 3(a) showsthe plots of 12-monthly running means of zonal winds at 10, 20, 30, 50, 70 mb, while Figure 3(b) showssimilar plots for equatorial eastern pacific SST and SO index (T−D). The maxima of SST and theminima of (T−D) are shown black and coincide with the occurrences of El Nino shown by rectangles(black=strong S, hatched=moderate M, open=weak W, Quinn et al., 1978, 1987).

5. RESULTS OF SPECTRAL ANALYSIS

Figure 4 shows the amplitudes of the periodicities obtained by MESA for the various series. The hatchedmarkings indicate the 2s (a priori ) limits and significant periodicities (in years) are marked. Since running



means over four consecutive seasonal values are used for analysis, the degrees of freedom are appropriatefor only 1/4 of the number of values and hence, the appropriate standard error level should be 4s insteadof 2s and is shown by additional dashed lines. For zonal winds at all levels, the most prominentperiodicity (significant at a better than 4s level) is ca. 2.35 years, i.e. ca. 28 months. The next prominentperiodicity (about 1/2 size) is ca. 2.7 years, i.e. ca. 32 months. Other periodicities are at ca. 2.10, ca. 3.00,ca. 3.60, ca. 4.8, ca. 7.0 and ca. 20 years; but some of these are not significant at a 4s level at someatmospheric levels. The bottom part of Figure 4 shows spectra for EEP (Equatorial Eastern Pacific) SSTand (T−D). Both have prominent periodicities at ca. 4.8 years, matching with the ca. 4.8 yearsperiodicity of the wind.

The SST and (T−D) have also a QTO at ca. 3.5 years, matching the small QTO of wind at ca. 3.6years. In the QBO region, SST and (T−D) have peaks at ca. 2.7 years, matching with ca. 2.7 years ofwind. On the other hand, (T−D) has peaks at 2.45 and 2.16 years which are different from the windpeaks at ca. 2.35 and ca. 2.10 years. As illustrated earlier, MESA is very accurate (90.02 years) in thisregion and these differences are significant.

In the high periodicity region, SST and (T−D) have peaks at ca. 6.4 years and 12–14 years. Wind hassome peaks in these regions, but not fully significant. SST has a peak at 21.2 years (Hale magnetic solarcycle?) and 20 mb wind has a similar peak (20.7 years).

Concluding, most of the QTO and larger periodicities of SST and (T−D) have almost equivalentperiodicities in wind while in the QBO region, periodicity at ca. 2.70 years is also common while themajormost wind periodicity at ca. 2.36 years is exclusively in wind. Incidentally, the 2–3 yearsperiodicities in wind at all levels are very similar (even in relative proportions). But ca. 3.6 years appearsignificant only at 40, 50 mb and (ca. 4.8) years only at 15, 20, 50 mb.

Similarity of periodicities is not necessarily a conclusive proof of common origin. But, here is probablysome indication of a mutual influence between stratospheric winds and ENSO.

Figure 3. Time series of the 4 season (12-monthly) running means of (a) Stratospheric equatorial zonal winds (westerly, positive;easterly, negative) at 10, 20, 30, 50, 70 mb and (b) equatorial eastern Pacific SST, SO index Tahiti (T)−Darwin (D) atmospheric

pressure difference (T−D) and El Ninos (rectangles, full=strong, hatched=moderate, blank=weak)


R.P. KANE1200

Figure 4. Amplitudes of the periodicities (numbers indicate years) obtained by MESA of the time series for 1953–1995 forstratospheric winds (10, 15, 20, 30, 40, 50, 70 mb) and EEP-SST and SO index (T−D). The hatched portions indicate 2s (a priori)

limits and the dashed lines 4s (see text)

The spectra refer to the 43 year period 1953–1995. However, Pawson et al. (1993) mention that whereasthe QBO was reported as a ‘26-month oscillation’ in radiosonde data (Reed et al., 1961; Veryard andEbdon, 1961), the period increased later to 27.7 months (Naujokat 1986) and to 28.1 months in recentyears. In our analysis, there is a major periodicitiy at 28 months and a minor but significant periodicityat 32 months. It could be that these periodicities and/or their relative proportions changed with time. Tocheck this, the data sets were divided into two almost equal parts, namely 1953–1974 and 1974–1995.Figure 5 shows the results of spectral analysis for (a) 1953–1974 and (b) 1975–1996. The results aredifferent from those for the whole period 1953–1995, indicating considerable time variation. Thefollowing may be noted:

(i) in Figure 5(a) for 1953–1974, the majormost periodicity for wind at all levels is 2.38 years, similar toca. 2.35 years for the whole period 1953–1996 (Figure 4) and missing in SST and T−D. (Theperiodicity is different, namely 2.44 years for 10 mb but these data started later in 1956). However,the wind peak at ca. 2.70 years seen in Figure 4 has disappeared and instead, the peak at ca. 3.00years has strengthened considerably in both, wind and ENSO (i.e. SST and (T−D)). Also, there is



a strong biennial peak(ca. 2.00 years) in the wind but not in SST, (T−D). The winds also have minorpeaks at 5.00–5.71 years (insignificant at a 4s level), not seen in SST, (T−D). On the other hand,SST and (T−D) have significant QTO at ca. 3.6 years, which is missing in the wind data;

(ii) in Figure 5(b) for 1974–1995, the majormost peak in wind is at ca. 2.48 years, considerably differentfrom 2.38 years of Figure 5(a) and the amplitudes have almost doubled. (Note the scale differencebetween (a) and (b)). Other wind peaks are very small, insignificant at a 4s level. Surprisingly, SSTand (T−D) now do have peaks (at ca. 2.45 years) matching with the major wind peak. In contrast,the prominent QTO (ca. 3.5 years) in SST, (T−D) is very weakly reflected in wind. Similarly, themost prominent peak (ca. 4.5 years) in SST, (T−D) is poorly reflected in wind.

Concluding, the wind and SST, (T−D) characteristics changed considerably from 1953–1974 to1974–1995 and there were similarities between wind and ENSO but at very different periodicities in thetwo intervals (ca. 2.90 in 1953–74, ca. 2.48 in 1974–1995).

Since the change from the first 21 years to the next 21 years was very large, analysis for a finer timeinterval was warranted. Hence the data were divided into 3 parts, namely 1953–1967, 1967–1981 and1981–1995. Figure 6 shows the results of spectral analysis. In the first 14 years (1953–1967, Figure 6(a)),the winds had a prominent periodicity at (ca. 2.15 years) i.e. ca. 26 months and another of slightly smalleramplitude at ca. 3.0 years i.e. ca. 36 months. The SST and (T−D) also had QBO but at ca. 2.05 and ca.2.80 years, near but not exactly at (slightly smaller than) the wind periodicities. However, the mostprominent periodicity for SST and (T−D) was at ca. 4.5 years, not reflected in winds.

Figure 5. Same as Figure 4, with (a) 1953–1974 and (b) 1974–1995. (Note change in amplitude scale in (a) and (b).)


R.P. KANE1202

Figure 6. Same as Figure 4, with (a) 1953–1967 and (b) 1967–1981 and (c) 1981–1995

In the 14 years 1967–1981 (Figure 6(b)), the most prominent peak had slightly larger period (2.21–2.22years, ca. 26.5 months), as compared to that for 1953–1967 (26 months) but the next prominent peak wasat ca. 3.0 years, ca. 36 months, almost the same as for 1953–1967. The SST, (T−D) had one peak at ca.2.40 years, not equal to any of the wind peaks but in between 2.21 and 3.00 years. Again, the mostprominent peak for SST, (T−D) was at ca. 4.1 years, not reflected in wind.

In the last 14 years 1981–1995 (Figure 6(c)), wind had one prominent peak at ca. 2.5 years, ca. 30months and no other significant peak. Pawson et al. (1993) attribute the longer QBO period partly to the1991 eruption of Mt. Pinatubo in 1991. However, the spacings during 1975–1992 were 29, 29, 29, 31, 31,30, 28 months, indicating larger spacings even before the Mt. Pinatubo eruption. The SST and (T−D)had a small QBO (2.33–2.35 years), not matching with the wind QBO (2.50 years) and SST, (T−D) hadthe most prominent peak at ca. 4.3 years, not reflected in the wind.

In conclusion, for data in these three consecutive 14 years intervals, considerable changes occurred inwind as well as SST, T−D characteristics. Whereas SST, (T−D) matched each other well (even in thechanges), their match with wind was not good. This is very surprising (and discouraging) in view of thegood match obtained (Figure 5) when data were divided into two equal parts. Part of this could bebecause of larger errors in these shorter data sets.

Besides the Pawson et al. (1993) data, Dr. Angell had kindly supplied data for three locations—Singapore (1°N, 104°E, SIN), Balboa (9°N, 80°W, BAL) and Ascension Island (8°S, 14°W, ASC). Thespectra for these are shown in Figure 7 for (a) 1951–1967, (b) 1967–1981, (c) 1981–1995 (or less) for 30mb (upper half) and 50 mb (lower half) and are very similar to those of Figure 6, indicating that thesecharacteristics are genuine and are shown not only very near the equator, but also within 910°.



In the analysis, the basic time series used for spectral analysis are 12-monthly means. Could this haveresulted into some distortions? MESA does not need filtering and all periodicities are revealed, indepen-dently of each other. Figure 8 shows the spectra for 50 mb wind for Canton-Gan for 1953–1974. The topplot, (a), is for monthly mean series and shows prominent peaks in the QBO region (2.00, 2.37, 2.96 years)and smaller peaks in the 6–18 months range, indicating a small annual wave of ca. 2 m/s amplitude.When 3-monthly running means are used, plot (b) shows the same peaks in the QBO region and slightlyreduced peak amplitudes in the 6–18 month range. When 12-monthly running means are used, plot (c)shows only QBO but at almost exactly the same periodicities as in (a) and (b). Thus, in spite of warningsgiven by some specialists (e.g. Mitchell et al., 1966) about potential problems, the procedure of using12-monthly running means does not seem to have caused any distortions in the present case, perhapsbecause the seasonal variation in stratospheric winds has an amplitude smaller than the QBO amplitude.

Fraedrich et al. (1993) made an EOF analysis of the vertical-time delay structure of wind QBO andreported a well-defined peak near 28 months for 1956–1991. In view of the changes with time reportedin this paper, it would be interesting to repeat their analysis for shorter intervals.

6. PHASE DIFERENCES

The zonal wind maxima and minima at different altitudes do not occur simultaneously. The maxima arereported to occur later by ca. 10 months, from 10 mb to 50 mb (Naujokat, 1986). The observed QBOoften exhibits a delay in the onset of the easterly winds when their downward propagation is interrupted.Dunkerton (1990) showed that the easterly wind onset at 50 mb occurred preferentially in the Northernhemisphere’s spring and summer, suggesting some annual modulation of the easterly phasing. Since wehave used 12-monthly running means, these effects should be obliterated. However, since the QBOcharacteristics changed considerably in the successive 14-year intervals, the phase differences also could bedifferent. In this analysis, different altitudes showed very similar periodicities in the same 14-year intervaland hence, even though amplitudes changed slightly (maximum at 15–20 mb), the phases could becompared meaningfully, for the significant periodicities ca. 2.17 and ca. 3.03 years in 1953–1967, ca. 2.22

Figure 7. Same as Figure 4, with (a) 1951–1967, (b) 1967–1981 and (c) 1981–1995 for 30 mb (upper half) and 50 mb (lower half)for the three locations Ascension Island (ASC), Balboa (BAL) and Singapore (SIN)


R.P. KANE1204

Figure 8. Spectra of 50 mb zonal wind over Canton-Gan for 1953–1974, for (a) monthly values, (b) 3-monthly values, (c)12-monthly running means

and ca. 2.95 years in 1967–1981 and ca. 2.51 years in 1981–1995. Figure 9 shows the plots. Thus, inFigure 9(a) for 1953–1967, the periodicities were 2.13, 2.15, 2.17, 2.17, 2.20, 2.18 years at 15, 20, 30, 40,50, 70 (mb) (data for 10 mb started later and hence phases were not directly comparable) giving anaverage of ca. 2.17 years. The phases charged from −0.56 radians at 15 mb to +1.87 radians at 70 mb.For 2.17 years (ca. 26 months), 2p radians (6.283) correspond to 26 months, giving 4.14 m/radian. Thescale to the left is in radians and that to the right in months. From 15 mb to 70 mb, the phase shift was2.43 radians or ca. 10 months (marked circles). From 15 mb to 50 mb, the phase shift was 6.7 months(marked circles).

The lower half of Figure 9(a) shows the plot for periodicities 2.96, 2.96, 3.08, 3.29, 3.07, 3.10 years for15, 20, 30, 40, 50, 70 mb. Here, 3.29 years for 40 mb is different from the rest and is omitted (cross). Theothers have an average of ca. 3.03 years and the phases changed from −2.94 radians at 15 mb to −0.96radians for 70 mb. Thus, from 15 mb to 70 mb, the phase shift was 1.98 radians. With 2p radians(=6.283 radians) equivalent to 3.03 years (36.4 months) i.e. 5.79 months/radian, the phase shift was 11.5months. From 15 mb to 50 mb, the phase shift was 1.51 radians or 8.7 months.

Figure 9(b) refers to 1967–1981. The upper half shows periodicities 2.22, 2.22, 2.21, 2.22, 2.21, 2.22,2.22 years for 10, 15, 20, 30, 40, 50, 70 mb with an average of ca. 2.22 years, and the lower half refersto periodicities 2.96, 2.93, 2.98, 3.01, 2.90, 2.96 years for 10, 15, 20, 30, 40, 50 mb. For 70 mb, theperiodicity was 2.76 (very different) and was omitted.

Figure 9(c) refers to periodicities 2.48, 2.49, 2.52, 2.52, 2.52, 2.52, 2.49 years for 10, 15, 20, 30, 40, 50,70 mb, with an average of ca. 2.51 years.

For these five periodicities, the phase shifts are summarised in Table I (see also Figure 9).Since the periodicities have amplitudes often significant at a level better than 10s, the standard error

of the phase should be ca. 0.1 radians and, for differences, ca. 0.15 radians or roughly 1 month. Thenumbers in any column seem to differ by more than one month, indicating that the phase shift withaltitude may be different for different periodicities in different intervals.



7. CONCLUSIONS AND DISCUSSION

The 12-monthly running means of the stratospheric low latitude zonal winds at several levels (10, 15, 20,30, 40, 50, 70 mb) and ENSO parameters—Southern Oscillation Index represented by Tahiti−Darwinatmospheric pressure difference and equatorial eastern Pacific sea surface temperature were subjected tomaximum entropy spectral analysis to detect periodicities and to multiple regression analysis to estimatetheir amplitudes. The results were compared for the 43 years, 1953–1995 as a whole as also for twoportions 1953–1974 and 1974–1995 and three portions 1953–1967, 1967–1981, 1981–1995. The follow-ing was noted:

(i) For the whole period, winds at all levels had the most prominent periodicity at ca. 2.35 years i.e. ca.28 months, a smaller periodicity (ca. 1/2 size) at ca. 2.68 years i.e. ca. 32 months and a still smallerperiodicity (ca. 1/3 size) at ca. 3.0 years i.e. ca. 36 months, besides other smaller periodicities at ca. 2.10and ca. 3.65 years, (at 10, 30, 40, 50 mb only) and ca. 4.8 years.

For the same interval, SST and (T−D) had common periodicities at ca. 2.15, ca. 2.45, ca. 2.75, ca. 3.5,ca. 4.8 and ca. 6.4 years from which ca. 4.8 years had the largest amplitude. The SST had a periodicityat 3.10 years which was missing in (T−D).

The majormost periodicity of ca. 2.35 years in stratospheric winds did not have a match in SST,(T−D). But the majormost periodicity at ca. 4.8 years and the smaller periodicities at ca. 3.5 years and2.75 years of SST, (T−D) were reflected in stratospheric winds also. In addition, 3.10 years of SST (notseen in T−D), was also reflected in stratospheric winds.

(ii) For the two 21 year intervals 1953–1974 and 1974–1995, the spectra were dissimilar, indicatingconsiderable time variation. In 1953–1974, stratospheric winds at all levels had the most prominentperiodicity at 2.38 years (28.5 months), and a slightly smaller periodicity at ca. 2.95 years (35 months) anda still smaller (but highly significant) periodicity at ca. 2.00 years (24 months). In 1974–1995, all thesedisappeared and only one strong periodicity appeared at ca. 2.48 years (30 months).

In SST and (T−D), 1953–1974 had a major periodicity at ca. 3.65 years and a minor (but significant)periodicity at ca. 2.85 years. However, in 1974–1995, a new majormost periodicity appeared at ca. 4.55years, the ca. 3.65 year periodicity changed slightly to ca. 3.45 years and a new periodicity appeared atca. 2.47 years.

In the earlier interval (1953–1974), wind, SST and (T−D) matched at ca. 2.90 years. In 1974–1995,there was a match at ca. 2.47 years.

(iii) For the three 14 year intervals 1953–1967, 1967–1981 and 1981–1995, the variations from oneinterval to the next were clearly seen. The majormost wind peak changed from ca. 2.17 years (26 months)in the first interval to 2.22 years (26.5 months) in the second interval and to ca. 2.50 years (30 months)in the third interval. A peak at ca. 3 years (36 months) was substantial in the first interval, reduced in thesecond interval and disappeared in the third interval. Thus, in recent years (1980s and 1990s), the windQBO has been almost monochromatic, at ca. 30 months.

The SST and (T−D) peaks compared well with each other. The most prominent peak was at ca. 4.55years in the first interval, ca. 4.15 years in the second interval and ca. 4.30 years in the third interval. But

Table I. Phase shifts (in months) from one level (10 or 15 mb) to another (50 or 70 mb)for different periodicities

PeriodicityInterval Phase shifts (months)

15–50 mb 15–70 mb10–70 mb10–50 mb

1953–1967 2.17 — — 6.7 10.11953–1967 3.03 — — 8.7 11.5

11.89.113.811.02.221967–1981—2.961967–1981 16.2 — 12.313.59.414.810.82.511981–1995


R.P. KANE1206

Figure 9. Phases (left scale, radians; right scale, months) at 10, 15, 20, 30, 40, 50, 70 mb levels of the zonal wind significantprominent periodicities (a) ca. 2.17 years and ca. 3.03 years during 1957–1967, (b) ca. 2.22 years and ca. 2.49 years during1967–1981 and (c) ca. 2.51 years during 1981–1995. The total phase shifts (months m) between 10 and 50 mb, and/or 10 and 70

mb and/or 15 and 50 mb and/or 15 and 70 mb are indicated

the minor peaks at ca. 2.05 years and ca. 2.80 years in the first interval disappeared and were replaced bypeaks at ca. 2.43 years in the second interval and ca. 2.34 years in the third interval. Surprisingly, the ca.3.5 years peak in SST, (T−D) in the whole period 1953–1995, as also in the two halves 1953–1974 and1974–1995, does not appear at all in the three intervals.

In the three intervals, the matching between wind and SST (T−D) was not good. Thus, in the firstinterval, wind had 2.17 and 3.00 years while SST, (T−D) had 2.05 and 2.80 years. The strongest SST,(T−D) peak at ca. 4.55 years was not reflected in wind at all. In the second interval, wind had 2.22 yearsand 3.00 years while SST, (T−D) had only ca. 2.43 years. Again, the strongest SST, (T−D) peak at ca.4.15 years was not reflected in wind at all. In the third interval, wind had ca. 2.50 years while SST,(T−D) had ca. 2.35 years, not a good match. Again, the strongest SST, (T−D) peak at ca. 4.30 yearswas not reflected in the wind at all.

Thus, for a shorter interval of 14 years, wind and SST, (T−D) (i.e. ENSO) peaks did not match well;but for a longer interval of 21 years, matching for some bands was very good and was reflected in theoverall interval of 42 years also.



(iv) The phase of stratospheric zonal winds changes with altitude, with the westerly maxima occurringearlier at higher levels (see Figure 3). For the significant periodicities at ca. 2.2, 2.5 and 3.0 years, thephase changes (shifts of the positions of the maxima, in months) were studied separately for 1953–1967,1967–1981, 1981–1995. Significant differences in the total phase changes from 10 mb to 70 mb wereobserved for the three intervals, indicating long-term changes in the propagation characteristics.

Matching of peaks does not necessarily imply links or common origin, but could be suggestive of thesame. Surface signs of a biennial atmospheric pulse have been investigated by many workers (Landsberget al., 1963; Landsberg and Kaylor, 1976). Relationships with winds at other levels have also beenreported (Trenberth, 1980; Arkin, 1982) and mostly associations with ENSO are claimed. Gray (1984)studied the relationship between Atlantic seasonal hurricanes, El Nino and 30 mb wind QBO. Yasunari(1989) gave some evidence for a possible link between the QBOs of stratosphere, troposphere andsea-surface temperature. Barnett (1989) wondered whether El Nino and the stratospheric wind QBO are‘the same creature’, though Barnett (1991) and Ropelewski et al. (1992) seem to feel that troposphericQBO is mainly related to ENSO. Angell (1992) presented evidence of a relationship between El Nino andQBO. Gray et al. (1992) described hypothetical mechanisms by which QBO of lower stratospheric zonalwinds alters the distribution of intense deep convective activity throughout the tropical west Pacific.Geller and Zhang (1991) and Geller et al. (1997) explored a mechanism by which SST variations modulatetropical wave activity which may force stratospheric zonal flow with the same period as the QBO of SST.The matching of some periodicities of stratospheric winds and ENSO for some intervals as illustrated inthis paper could be a partial evidence of a linkage between these two phenomena.

ACKNOWLEDGEMENTS

Thanks are due to Dr. Barbara Naujokat and to Dr J.K. Angell for providing data privately. This workwas partially supported by FNDCT Brazil under contract FINEP-537/CT.

REFERENCES

Angell, J.K. 1981. ‘Comparison of variations in atmospheric quantities with sea surface temperature variations in the equatorialeastern Pacific’, Mon. Wea. Re6., 109, 230–243.

Angell, J.K. 1986. ‘On the variation in period and amplitude of the quasi-biennial oscillation in the equatorial stratosphere,1951–85’, Mon. Wea. Re6., 114, 2272–2278.

Angell, J.K. 1992. ‘Evidence of a relation between El Nino and QBO, and for an El Nino in 1991–1992’, Geophys. Res. Lett., 19,285–288.

Angell, J.K. and Korshover, J. 1962. ‘The biennial wind and temperature oscillations of the equatorial stratosphere and theirpossible extension to higher latitudes’, Mon. Wea. Re6., 90, 122–132.

Arkin, P. A. 1982. ‘The relationship between interannual variability in the 200 mb tropical wind field and the Southern Oscillation’,Mon. Wea. Re6., 110, 1393–1404.

Barnett, T.P. 1989. ‘A solar-ocean relation: Fact or fiction’, Geophys. Res. Lett., 16, 803–806.Barnett, T.P. 1991. ‘The interaction of multiple time scales in the tropical climate system’, J. Climate, 4, 269–285.Barnston, A.G. and Livezey, R.E. 1989. ‘A closer look at the effect of the 11-year solar cycle and the quasi-biennial oscillation on

northern hemisphere 700 mb height and extratropical North American surface temperature’, J. Climate, 2, 1295–1213.Bevington, P.R. 1969. Data Reduction and Error Analysis for the Physical Sciences, McGraw-Hill, New York, pp. 164–176.Blackman, R.B. and Tukey, J.W. 1958. The Measurements of Power Spectra, Dover, New York. pp. 190.Burg, J.P. 1967. Maximum Entropy Spectral Analysis, Paper presented at the 37th Meetine, Society of Exploration Geophysics,

Oklahoma City, October.Chen, W.Y. and Stegen, G.R. 1974. ‘Experiments with maximum entropy power spectra of sinusoids’, J. Geophys. Res., 79,

3019–3022.Dunkerton, T.J. 1990. ‘Annual variation of deseasonalized mean flow acceleration in the equatorial lower stratosphere’, J. Meteorol.

Soc. Jpn., 68, 499–508.Fraedrich, K., Pawson, S. and Wang, R. 1993. ‘An EOF analysis of the vertical-time delay structure of the quasi-biennial

oscillation’, J. Atmos. Sci., 50, 3357–3365.Geller, M.A., Shen, W., Zhang, M. and Tan, W-W. 1997. ‘Calculation of the stratospheric quasi-biennial oscillation for time-varying

wave forcing’, J. Atmos. Sci., 54, 1 April 1997.Geller, M.A. and Zhang, M. 1991. Sea Surface Temperatures, Equatorial Wa6es, and the Quasibiennial Oscillation, Abstract QB-5,

Symposium on Middle Atmospheric Sciences, Kyoto.Gray, W.M. 1984. ‘Atlantic seasonal hurricane frequency, Part I: El Nino and 30 mb QBO influences’, Mon.Wea. Re6., 112,

1649–1668.


R.P. KANE1208

Gray, W.M., Sheaffer, J.D. and Knaff, J.A. 1992. ‘Hypothesized mechanism for stratospheric QBO influence on ENSO variability’,Geophys. Res. Lett., 19, 107–110.

Holton, J.R. and Lindzen, R.S. 1972. ‘An updated theory for the quasi-biennial cycle of the tropical stratosphere’, J. Atmos. Sci.,29, 1076–1080.

Kane, R.P. 1977. ‘Power spectrum analysis of solar and geophysical parameters’, J. Geomag. Geoelect., 29, 471–495.Kane, R.P. 1979. ‘Maximum entropy spectral analysis of some artificial samples’, J. Geophys. Res., 84, 965–966.Kane, R.P. and Trivedi, N.B. 1982. ‘Comparison of Maximum Entropy Spectral Analysis (MESA) and Least-Squares Linear

Prediction (LSLP) methods for some artificial samples’, Geophysics, 47, 1731–1736.Lacoss, R.T. 1971. ‘Data adaptive spectral analysis methods’, Geophysics, 36, 661–675.Landsberg, H.E. and Kaylor, R.E. 1976. ‘Spectral analysis of long meteorological series’, J. Interdisciplinary Cycle Res., 7, 237–243.Landsberg, H.E., Mitchell, J.M., Crutcher, H.L. and Quinter, F.T. 1963. ‘Surface signs of the biennial atmospheric pulse’, Mon.

Wea. Re6., 91, 549–556.Lindzen, R.S. and Holton, J.R. 1968. ‘A theory of the quasi-biennial oscillation’, J. Atmos. Sci., 25, 1095–1107.Mitchell, J.M., Dzerdzeevskii, B., Flohn, H., Hofmeyr, W.L., Lamb, H.H., Rao, K.N. and Wallen, C.C. 1966. Climatic Change,

Technical Note 79, 5–6 pp., World Meteorological Organisation, Geneva.Naujokat, B. 1986. ‘An update of the observed Quasi-biennial oscillation of the stratospheric winds over the tropics’, J. Atmos. Sci.,

43, 1873–1877.Parker, D.E. 1983. ‘Documentation of a Southern Oscillation index’, Meteorol. Mag., 112, 184–188.Pawson, S., Labitzke K., Lenschow, R., Naujokat, B., Rajewski, B., Wiesner, M. and Wohlfart, R.-N. 1993. Climatology of the

Northern Hemisphere Stratosphere deri6ed from Berlin Analyses-Part 1: Monthly Means, Meteorologische Abdandlungen desInstituts fur Meteorologie der Freien Universitat Berlin, Neue Folge Serie A Monographien Band 7/Heft 3.

Plumb, R.A. and Bell, R.C. 1982. ‘A model of the quasi-biennial oscillation on an equatorial beta-plane’, Q. J. R. Meteorol. Soc.,108, 335–352.

Quinn, W.H., Zopf, D.O., Short, K.S. and Kuo Yang, R.T.W. 1978. ‘Historical trends and statistics of the Southern Oscillation,El Nino and Indonesian Droughts’, Fish. Bull., 76, 663–678.

Quinn, W.H., Neal, V.T. and Antunez de Mayolo, S.E. 1987. ‘El Nino occurrences over the past four and a half centuries’, J.Geophys. Res., 92, 14449–14461.

Reed, R.J., Campbell, W.J., Rasmusson, L.A. and Rogers, D.J. 1961. ‘Evidence of a downward propagating annual wind reversalin the equatorial stratosphere’, J. Geophys. Res., 66, 813–818.

Ropelewski, C.F., Halpert, M.S. and Wang, X. 1992. ‘Observed tropospheric biennial variability and its relationship to SouthernOscillation’, J. Climate, 5, 594–614.

Trenberth, K.E. 1980. ‘Atmospheric Quasi-biennial oscillations’, Mon. Wea. Re6., 108, 1370–1377.Ulrych, T.J. and Bishop, T.N. 1975. ‘Maximum Entropy Spectral Analysis and autoregressive decomposition’, Re6. Geophys., 13,

183–200.Veryard, R.G. and Ebdon, R.A. 1961. ‘Fluctuations in tropical stratospheric winds’, Meteorol. Mag., 90, 125–143.Yasunari, T. 1989. ‘A possible link of the QBOs between the stratosphere, troposphere and the surface temperature in the tropics’,

J. Meteorol. Soc. Jpn., 67, 483–493.


Spectral comparison of ENSO and stratospheric zonal winds

Documents

Transcript of Spectral comparison of ENSO and stratospheric zonal winds