Variability of the polar night jet in the Northern and Southern Hemispheres

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 106, NO. D18, PAGES 20,703-20,713, SEPTEMBER 27, 2001 Variability of the polar night jet in the Northern and Southern Hemispheres Yuhji Kuroda and Kunihiko Kodera Meteorological Research Institute, Tsukuba, Japan Abstract. The spatial and temporal characteristics of the month-to-month variability of the polar night jet and its relationship with troposphericirculation is investigated for both the Northern and Southern Hemispheres. The variabilities of the hemispheres have many common characteristics of the Polar Night Jet Oscillation (PJO). These common characteristics include the following: (1) the anomalous zonal-mean zonal winds shift poleward and downward; (2) the anomalous polar temperatures propagate downward from the stratopause to the upper troposphere; and (3) they are made through a wave-mean flow interaction with mainly the planetary wave of zonal wave number one. Annular modes associated with the PJOs appear in both hemispheres when the zones of maximum polar temperature anomaly descend to the lowermost stratosphere and upper troposphere. The major difference in the PJOs of the two hemispheres is found in their temporal characteristics. In the Southern Hemisphere, the phase of the PJO is closely locked to the annual cycle, while in the Northern Hemisphere it exhibits quasiperiodic variability with its envelope controlled by the annual cycle. The origin of the differences between the PJOs is discussed based on the theory of the wave-mean flow interaction 1. Introduction in the two hemispheres. For example, in the SH, the phase Stratosphericirculation exhibits large month-to-month of the PJO is closely locked to the annual cycle, while in the variabilities in accordance with the occurrence of stratospher- NH, only the envelope of the PJO is controlled by the annual ic warming in the Northern Hemisphere (NH) winter [e.g., cycle. Labitzke, 1982]. However, this intraseasonal variability does The aim of the present study is to clarify, through a denot occur randomly but has a temporal and spatial structure tailed comparison, whether the phenomenological similarity [e.g., Perlwitz and Graf, 1995; Perlwitz et al., 2000; Baldwin between the PJOs in both hemispheres is produced from et al., 1994]. In particular, zonal-mean zonal wind anomaessentially similar processes and whether their different aplies propagate poleward and downward to the troposphere pearances are due to conflicting circumstances, such as a in turn from the stratopause region [Kodera, 1995; Kuroda difference in the planetary wave forcing. Another interesting and Kodera, 1999] (hereinafter referred to as KK99). The aspect of the PJO is that in the NH the Arctic Oscillation periodic nature of this variation can be more clearly seen (AO) (or the NH annular mode) [Thompson and Wallace, in general circulation model (GCM) simulations under per- 1998, 2000] preferentially occurs at a certain period of the petual winter conditions [Christiansen, 1999; Yamazaki and PJO (KK99). In the present study, we also investigate the Shinya, 1999]. relationship between the PJO and the Antarctic Oscillation In contrast, stationary planetary wave activity is very weak (AAO), a counterpart of the AO. in the Southern Hemisphere (SH), and major stratospheric This paper is organized as follows. The data set and the warming does not occur during midwinter, but minor warmprincipal method of analysis are described in section 2. Secings generally take place in spring at an altitude that descends tion 3 provides the results of analyses in both hemispheres over time [Farrara and Mechoso, 1986]. In spite of the large on (1) the spatial structure of the PJO, (2) the relationship differences in stratospheric variability between the two hemiwith the troposphere, and (3) the temporal characteristics of spheres [Shiotani and Hirota, 1985], a similar poleward and the stratospheric variability. After a discussion in section 4, downward propagation of zonal wind anomalies is found in a conclusion is offered in section 5. the SH [Kuroda and Kodera, 1998] (hereinafter referred to as KK98). For convenience, we designate such structured slow variability in the polar night jet in both hemispheres as 2. Data and Method of Principal Analysis the "Polar Night Jet Oscillation" (PJO). However, as will be The stratosphere and troposphere data used in the present,.,., are also large, fterences n btuuy are upuatcu n. from vwoo...,,u. A cover oc years, from January 1979 to December The original stratosphere Copyright 2001 by the American Geophysical Union. data were analyzed by U.S. National Centers for Environ- Paper number 2001 JD /01/2001 JD ,703 mental Prediction (NCEP)/Climate Prediction Center (CPC) (formerly NMC/CAC). The stratospheric winds were calcu-

2 20,704 KURODA AND KODERA: VARIABILITY OF POLAR NIGHT JET lated from a satellite-derived geopotential height analyzed by CPC using the nonlinear balanced wind relation [Randel, 1992]. The stratospheric data of the NH up to 10-hPa levels from April 1996 to April 1997 were completely missing at NCEP. The missing data for this period are compensated by those derived from the Met Office, UK geopotential height data [Bailey et al., 1993], after being adjusted for being continuously connected at the equator to the SH counterpart of the NCEP data. Operational data for the troposphere at 100 hpa and below are replaced by reanalysis data of the NCEP/National Center for Atmospheric Research (NCAR) [Kalnay et al., 1996]. All missing data, except for the above mentioned substantial missing period, were linearly interpolated in time, and the monthly mean data were then calculated as 30-day mean data, with the exception of July, which is 35-day mean data. Note that the Eliassen-Palm (E-P) flux, residual velocity, and wave-activity flux by Plumb [1985] were calculated from 5-day mean data and the monthly average was taken thereafter. The monthly mean E-P flux calculated from the daily mean NCEP/NCAR data below 10 hpa was also used as a supplemento investigate the high-frequency transient eddy component. Different analysis methods were applied to each hemisphere in previous studies. These methods included the multiple empirical orthogonal function (M-EOF) analysis to the SH (KK98) and the extended singular value decomposition cross-covariance matrix due to extended components can be efficiently performed by the method of Kuroda [ 1998]. Most of the analyses in the present study were conducted based on the monthly mean data. The analyses of the zonalmean zonal wind and vertical component of the E-P flux were directly performed by the E-SVD. Other variables were regressed onto the "developing coefficients of the zonal wind of the leading E-SVD" (hereinafter denoted as the PJO index) to study related variability. Similarly, lagging components outside of the original definition of the E-SVD, such as lag -1 month, were also calculated by a regression onto the PJO index. 3. Results 3.1. Spatial Structure Figures 1 a and 1 b show the leading E-SVD in the Northern and Southern Hemispheres. They are displayed as heterogeneous regression maps of the zonal-mean zonal wind and the E-P flux. Note that the meridional component of the E-P flux is calculated by regression of the zonal-mean zonal wind expansion coefficient. The lag -1 month component was also calculated by regression (hereinafter, the words lag-1 month are denoted by lag -1 for simplicity). The name of the month corresponding to each lagged component is indicated by initials. Regions where the heterogeneous correlation coefficients exceeded the 95% significance level (0.32 for the NH for 19 winters and 0.31 for the SH for 20 winters) are shaded. (E-SVD) analysis to the NH (KK99). A recent study [Kodera et al., 2000] revealed that the stratospheric variability in the NH can endure for a very long time during the cold season, The leading mode in the NH explains 28% of the total as in the SH. Therefore the same E-SVD analysis is applied squared covariance fractions and is comparable to the second for both hemispheres to extract the variabilities that extend one (explaining 23%). However, the spatial patterns of the over a 5-month period in the present study. second mode are very similar to those of the 1-month-delayed For this purpose, a SVD analysis [Bretherton et al., 1992] components of the leading mode. This condition reflects the is applied to the extended vectors, which incorporate lagged periodic nature of the PJO in the NH, as will be shown later. vectors into the original one, to extract a propagating phe- In contrast, the leading mode (explaining 57%) in the SH nomenon with time. This technique is a direct extension of dominates the second one (explaining 10%). E-EOF analysis [Weare and Nasstrom, 1982]. The E-SVD The present results eproduce major features of previous analysis is calculated based on extended vectors x, i and zt i as studies; in the NH, three consecutive months, lag 0 to lag 2, correspond to the results in KK99. In the SH, lag-1 to i x, - [u': (m0 + t), u (m0 + t + 1 ),... lag 3 components correspond to the 2-month average of the, u i (m0 + t + 4)] May to October pattern in KK98, due to the use of two-set i - (,-,o + t), (,-,o + t + ),... vectors in the present analysis. This ensures that the result is, ei(mo + t + 4)], (1) not sensitive to a small change in a parameter or the method of analysis. where ui(rn,) and ei(m) are the anomalous monthly mean In the NH (Figure l a), the negative zonal wind anomaly zonal-mean zonal wind and the vertical component of the first appears in the subtropical upper stratosphere in associa- E-P flux, respectively, at each height and latitude for the ruth tion with an enhanced upward propagation of anomalous E-P month of the ith year. The field at the month of rn,0 + t + k is flux from the high-latitude troposphere (lag- 1 to 0). The negcalled the lag k month component. To treat the stratosphere ative wind anomalieshift toward high latitude while develand tropospherequally, all vectors are normalized by their oping (lag 1) and then move poleward and downward to the standardeviations of the year-to-year variabilities of respec- lower stratosphere, giving their place to positive anomalies tive months prior to the calculation of the cross-covariance in the subtropical stratosphere (lag 2). Positive anomalies matrix to perform SVD. Two sets of vectors corresponding also shift poleward and downward, following the negative to t = 0 and 1 are used in the present analysis for each anomalies (lags 3 to 4). A close relationship between the cold season; November (June) was selected as rn,0 for the anomalous E-P fluxes and zonal winds was observeduring NH (SH), and so the vector for t = 0 corresponds to the the whole sequence; the anomalous E-P flux was directed five consecutive months from November to March (June to toward the region of negative anomalies of zonal wind. This October), and that for t = 1, to December to April (July to happens because the region of negative wind anomaly coin- November). The present calculation of the SVD of the large cides approximately with the region of anomalous E-P flux

3 KURODA AND KODERA: VARIABILITY OF POLAR NIGHT JET 20,705 Figure 1. Heterogeneous regression maps of the leading E-SVD between the zonal-mean zonal wind (contour) and the (vertical component of) E-P flux (vector) in the (a) Northern and (b) Southern Hemispheres. The meridional component of the E-P flux was constructed by regression. The numbers above the panels indicate the lag in months, and the initials represent the corresponding months. Shading denotes those regions where the heterogeneous correlations of zonal winds are significant at the 95% level. The contour interval is 2 m s- ], and the dashed lines indicate negative values. The zero contour line is plotted as a thin solid line. The E-P flux is scaled by the inverse square root of the pressure, and small arrows are omitted. convergence for this slow variability, as will be shown later. dencies, except for the accelerations of the zonal winds in The meridional propagation of the E-P flux in the tropo- the high latitudes of the SH at lag 0. This overall similarity sphere is also enhanced at the stage when the meridional between wave forcing and the zonal wind tendency indicates dipole structure is formed in the anomalous zonal wind field that the variability of PJO is made through wave-mean flow (lags 1 and 2). interaction. The zonal wind acceleration in the high latitude In the SH (Figure lb), negative zonal wind anomalies of SH at lag 0 is considered to be due to Coriolis forcing, appear in the subtropics in association with an enhanced although other possibilitiesuch as unresolved gravity wave upward propagation of anomalous E-P flux, similar to the exist. The wave forcings and the E-P flux due to zonal wave NH (lags-1 to 0). However, the negative anomalies do not number 1 and 2 components are shown in Figure 3. In the shift poleward. This dipole pattern persists during the winter NH, the main contribution comes from the zonal wave num- (lag 1 to 2). In spring, negative wind anomalieslowly shift ber 1 component, as documented by Kodera et al. [2000]. poleward and downward and positive anomalies gradually The zonal wave number 2 component contributes little to disappear from the stratosphere (lags 3 to 4). the total wave forcing in the upper stratosphere, but it is rel- Positive zonal wind anomalies appear in the subtropical atively significant in the middle stratosphere and below. For upper stratosphere in the NH when the vertical component of example, the poleward propagating wave in the troposphere anomalous E-P flux change sign from positive to negative in lag 2 of the NH comes mainly from the zonal wave number (lag 2). In contrast, vertical component of anomalous E- 2 component. In the SH, the zonal wave number 1 compo- P fluxes in the SH are positive throughout the sequence. nent is dominant, and the zonal wave number 2 component However, it must be noted that the latitudes of enhanced contributes to the lower part of the area of total wave forcing, vertical propagation in the lower stratosphere gradually shift but it is relatively small. poleward in association with the enlargement of the negative Figure 4 presents regression maps of the anomalous residwind anomalies. Changes in meridional propagation of the ual circulation and zonal mean temperature. The residual tropospheric E-P flux are not evident in the SH, except for the velocities were calculated following the method by Soel and high-frequency component, which will be described below. Yamazaki [ 1999]; transformed Eulerian equations are itera- To gain an insight into the origin of the zonal-mean zon- tively solved with the upper boundary condition of no vertical al wind variability, associated accelerations of zonal-mean motion at the 0.5-hPa level. This upper boundary condition zonal wind due to the divergence of E-P flux (wave forcings) is appropriate for all the seasons in both hemispheres except were calculated by regression onto the PJO index (Figure 2). for the polar area in the SH in June and July [Haynes et It can be seen that the positive and negative anomalies of al,, 1991]. The areas of downward and upward motion of the wave forcing in the Northern and Southern Hemispheres anomalous residual circulation in both hemispheres correcorrespond well to those of the zonal-mean zonal wind ten- sponded well to the high- and low-temperature anomalies.

4 20,706 KURODA AND KODERA: VARIABILITY OF POLAR NIGHT JET NH 30 1 O0 300 (0 ) N//D 2 i 2 :::::::::::::::::::::::::::::::::::i::11iiii,. i ;:. :g f::: i:::::::::::::::::::::::::::::::::::::::::::: o... :???:::?: :????? o... : ' ':::: : :'"' ========================================================================= ':... "... :::::::::::::::::::::::... 75::.':?: : :: : :: ; ::: ' :'"!... < < :::::::? ::::::? ::?' ': :: ::::::: ::::::::: : N (b) (-1) M/J.,',0) J/J. (1 J/A, (2) Ap/S., (3) S/O (4) O/N ============================ 2 :::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::... I 2 / ::i':'::'::': : :: ::::::..:::: I... i 1... :... : '::'::' o : :; : : ' - - F : od??. ;. ;.:/ : ;?] o... ::::::::::::::::::::::::::::::::::::: 30 :... : ': ß 1... ::'74 l:::? 1 -:4 ::::::::::::::::::::::::::::::::::::::::::::::::::::::: / :::::::::::::::::::::::::::::: ================================================================================== 300 ::::' ::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: 20s,os 60s 80s 20s s 60s 80s 26s 46s 66s 86s 20s,os 60s 80s 20s s 60s 80s 26s s 66s 80s Figure 2. Same as Figure 1, except for the regression map of the wave forcing due to the E-P flux divergence (contour) and the E-P flux (vector) assoc ated with the PJO. Contour interval is 0.5 m s-] d -], and negative values are shaded. Zero contour line is omitted. The largest temperature anomalies in the NH were found in N H, ;0) N/D, (1) D/J, (2) J/F the polar region. Positive and negative anomalies descend in turn (Figure 4a). There is a tendency for temperature 2... : : :.., ':'" ' i'::"'":'4 ' ' : '... ::/.: '... :; 2 ' ': : : : t3; : :... :.' ' :: - 2 anomalies in the subtropics to be opposite to those in the polar region. (In particular, a quadrupole structure of the temperature anomaly was noted at lag 2 of the NH.) :::: ================================================ :::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: / :...? " : ':': :i: ' :... 4" : :: : : : 'a 1 v v. In the SH, high-temperature anomalies are first created. ::::::::::::::::::::::::?::::::::::::::::::::::::::::::::::::::::::::: :: ::..: ;:: ::::::::::::::::::::::::.. I ].. in the midlatitudes of the middle stratosphere in early win- : :....:,:.4 : - :; :: : :.,X::.d.: i: ::' : :' : :' " :... :: : : 8 ; :::':::: b:::::?:::::?:?: f )' 8 / :: :: :?:;?::""'"' '"'" -. 8 ter. In contrast with the NH, downward velocity is larger ON N 66N 86N 26N N 66N 86N ON N N 80N in the midlatitudes than in the polar region owing to the (b), o).....,...,,).?...,,,( ) 4:[:,... stronger convergence of E-P flux in the lower latitudes (lag WN2 :: :;i?::::f::.l:: :::: :Z : 5 :::::::::::::::::::::::::::::::::::::::::::: 5 ===================== -1 to 0). The warm area gradually shifts poleward (lags 1 : ::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: :':::::::::::::::::::::::::::::::: f::::: and 2) and downward (lags 3 and 4) in accordance with the poleward shift of the convergence zone of E-P flux. Lowtemperature anomalies can be seen at the equatorial side of :::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::, t?: ' ::?:?:?: :: ' :?: :::?: ::?::'::::4 ' t...,' the high-temperature anomalies, similar to the NH; however, the overlaying lower-temperature anomalies in the polar 26N N N ß 80N 20N N N 80N 20N N 60N 80N region at 1 hpa descend only a short distance. SH 0 JJ 1 JA 2 AS A time-height cross section of the anomalous polar temperatures (averaged poleward of 80 ø) illustrates well the differ- ent characteristics of the temperature variations associated 10 : :. :. P ' g :.'..: ;::7..:;: :,... with the PJOs (Figure 5, top panels). In the NH, positive o... ::f:] ] ::]:: :: ::! ]!i ::; :::: :: ':]]' i::: ]'Ji:} :'": i i :: 0 )[ ]::... '... :½:.:: :.:.:-: :.: 0 and negative anomalies propagate downward in turn, while interannual variability is very large and positive or negative anomaly propagates downward only once a year in the SH. This structure of temperature variabilities can be ob- IOS S S 20S 60S 80S 20S S 60S 80S tained from a simple height-time cross section of the polar I temperature anomalies of individual winters. For example, ::! 5?:f:f:?:?:L:::.- :.???: ::. ::.:: :: :: ::::::::::::::::::::::::::::: ::y::::: :?:??' Figures 6a and 7a show 30-day running mean northern and southern polar temperature anomalies for three winters from 1986 to Successive downward propagation of positive and negative anomalies are found in NH winters, while this occurs only once per year in the SH Relation With the Troposphere Two types of interactions between the troposphere and stratosphere are noted in KK99. One is the response of the stratosphere to changes in the vertical propagation of tropo- 0!-... ':: :;: '"? t",, o!'??: !, 0 r : ::? :::5':... 20S S 60S S 20S S 60S 80S 20S S 60S Figure 3. Same as Figure 2, except for zonal wave number (a) one and (b) two components of the Northern Hemisphere and zonal wave number (c) one and (d) two components of the Southern Hemisphere. Contour interval is 0.2 m s-] d -].

5 KURODA AND KODERA: VARIABILITY OF POLAR NIGHT JET 20,707 Figure 4. Regression map of the zonal-mean temperature (contour) and the residual velocity (vector) associated with the PJO in the (a) Northern and (b) Southern Hemispheres. The numbers above the panels indicate the lag in months, and the initials representhe corresponding months. Shading denotes regions where the correlations of the temperatures are significant at the 95% level. Contour interval is 1 K. spheric waves. This situation was found in both hemispheres by Thompson and Wallace [1998, 2000]. The correlation at similar periods of lags -1 and 0 in the present analysis coefficients with the annular mode indices become largely (Figure 1). negative when the warming signal reaches the lower strato- The other interaction is a formation of the annular mode sphere in the hemispheres (lag 1 for NH and lag 4 for SH). (or the AO in the NH) in the troposphere in association with This period (hereinafter denoted as the annular period) cora change in the stratospheric polar vortex. Figure 5 (bottom responds in both hemispheres to the period of formation of panels) gives the correlation coefficients between the PJO dipole anomalies in the tropospheric zonal-mean zonal windindices and time coefficients of the annular mode (annular s, with a node around 45 ø latitude (Figure 1). In the NH, mode indices). The annular modes are defined here as the changes in meridional propagation of E-P flux occur conleading EOF of the month-to-month variability of the 850- currently with the formation of the dipole structure in the hpa levels in both hemispheres in the present data set for troposphere. However, this change is not found in the SH in 1979 to 1998; they are essentially the same as those obtained lag 4, shown in Figure 1. However, it must be remembered ) 12 North Pole "emperature (b) l South Pole Tem: erature 10 : '"" ':... ::./ ; :: :.. I:': ' X 10 X " : :: : : :, x,. : ::: : :: F:. i... : : :::: ::::::::::::::::::::::: 50 ' :: : '{:: :::: F 50 1 O0 x 1 O K do,ao) K do, o) o ::::;... o Figure 5. Same as Figure 4, except for the regression map of the polar temperatures (upper panels), displayed as a function of time and height. Contour interval is 1 K. Correlation coefficients between the PJO and the AO or AAO indices are displayed in the bottom panels.

6 20,708 KURODA AND KODERA: VARIABILITY OF POLAR NIGHT JET NH : :: :: :: :: :: ::?. - m - -[ 'rr ' ' ' ' '.'-'.' OUL loo 3oo 850,,,,,,,,,, JUL OCT JAN APR JOL OCT JAN APR JUL OCT JAN APR JL Figure 6. (a) Thirty-day running averaged anomalous polar temperature in the Northern Hemisphere from 1987 to (b) Interannual standardeviation of monthly mean polar temperature in the Northern Hemisphere. Contour interval is 5 K in Figure 6a an,l 1 K in Figure 6b. Shading indicates negative values in Figure 6a, and values larger than 4 K in Figure 6b. that the E-P flux depicted in Figure 1 was calculated from display only three components of lags - 1,0, and 1 for the NH 5-day mean data. and- 1, 3, and 4 for the SH. It can be seen that in the SH the Figure 8 shows a regression map of the high-frequency anomalous E-P flux in the midlatitudes of the troposphere transient component of an anomalous monthly mean E-P is directed toward the region of negative anomalies of the flux, defined as the difference between those calculated from zonal-mean zonal wind. In the NH, meridional propagation the daily mean and from the 5-day mean data. The figures is also found in the high-frequency transient component at SH :: i i :i :: i! i ::. i :: :.!.: i :: :: :.; 5... :... :i Zi::i::Siii. i ::! :: :: :: :: i i :: i. :::: :.! : :i:... : : : ::: :... /: iii!i::i::. ii:: :: ::iiii i::ili:: ::::::iiii::iiiii::::ii ::? :!!!!!!!!!!' i:::::::: iiiii:::: ===================================== 3o ] loo... :.Y X :...: -.:.: JUL dan JUL dan J IL '"'"'" JAN '... ' O IO0 3OO 85O APR JUL OCT dan APR JUL OCT dan APR dul OCT dan APR Figure 7. Same as Figure 6 except for the Southern Hemisphere from 1986 to 1988, and the contour interval is 3 K for Figure 7a.

7 KURODA AND KODERA: VARIABILITY OF POLAR NIGHT JET 20,709 ; o-i h I ; o ; o [, : ^ i ß. ::?:!::!::i:: :. / I lag 1, but the contribution of the stationary component is dominant (compare lag! of Figure 1 and Figure 8). A regression map of 500-hPa geopotential height was calculated to investigate more closely the relationship between the PJO and the troposphericirculation (Figure 9). The associated wave activity flux of Plumb [ 1985] is also shown by arrows. Since our focus here is on planetary waves, the wave-activity flux was calculated after truncation at zonal wave number 5. Note that zonal averaging of the wave activity flux reproduces the E-P flux meridional section (Figure 1). The numbers below the respective panels indicate the correlation coefficients between the PJO and the annular mode indices. The geopotential height fields in both hemispheres exhibited wave trains at lag -1 when the upward anomalous E-P flux was still confined mainly in the troposphere and lower stratosphere. It can be seen that height anomalies around Scandinavia or the Ross and Amundsen Seas accom- 850/i?!i i' 20S 40S, ':::i:::;:::!i i 60S... 80S, S 40S 60S 80S' 85 80S pany an enhanced upward propagation of waves (Figure 10). It should be noted that the areas of enhanced upward prop- Figure 8. Same as Figure 1, except for the regression maps agation of the waves are located at the areas of significant of the E-P flux (vector) of high-frequency transient waves for heat transport (i.e., east of the trough and west of the ridge at selected lag components (see text). lag -1 in Figure 9). The contribution of the stationary eddy (O) (-1) O/N (1) D/J (2) J/F (b) SH (-1) M/J (4) O/N (5) N/D 90W Figure 9. Regression map of the geopotential height of the 500-hPa level (contour) and the horizontal component of the wave activity flux by Plumb [1985](vector) associated with the PJO for selected lag components. Contour interval is 10 m, and the dashed lines indicate negative values. Shading denotes regions where the correlations of geopotential heights are significant at the 95% level. Numbers below each panel indicate the correlation coefficient between the PJO indices and the AO or AAO indices.

8 20,710 KURODA AND KODERA: VARIABILITY OF POLAR NIGHT JET NH'(-1) O/N SH'(-1) M/J lated from anomalous zonal-mean zonal winds of all months. About 65% of the total variability is explained in each hemisphere by the two leading EOFs. Figure 11 shows the distribution of the time coefficients of EOF 1 and 2 for December and March of each year for the NH (top) and for July and October for the SH (bottom). The distribution of go goe the coefficients has no clear structure in the NH, whereas the coefficients in the SH are clustered along a line that changes direction with the season. This indicates thathe spatial pattern of the interannual variability changes with the seasons in the SH; i.e., it is a dipole type pattern in July, but a monopole type in October, as is confirmed by Figure 1. Figure 10. Regression map of the vertical component of the The lack of structure in the NH as depicted in Figure wave activity flux by Plumb [ 1985] at the 500-hPa level asso- 11 does not indicate that there is no organized variability, ciated with the PJO at lag -1 in the (left) Northern and (righ0 but that the phase of the phenomenon is not locked to the Southern Hemispheres. Contour interval is 5 x 104 Kg s -2, annual cycle. In fact, Kodera et al [2000] demonstrated and the dashed lines indicate negative values. that the trajectory of a phase vector tends to trace a circle on the plane, suggesting a periodic nature. Multichannel singular spectrum analysis (MSSA)[Vautard and Ghil, 1989] component is evidently important, since the upward wave was conducted for the anomalous zonal-mean zonal wind to propagation was observed even with the E-P flux calculated extract the periodic structure more clearly. MSSA is the from the monthly mean data (not shown). same operation as E-EOF excepthat the attention is paid A seesaw of geopotential heights between the polar re- to the spectrum structure of the time coefficients. Therefore gion and midlatitudes, i.e,, AO or AAO (or annular mode), the operation is very similar to the E-SVD, excepthat all appears at lag 1 for the NH and 4 for the SH in Figure 9. the months of data are used in MSSA, whereas the E-SVD A significant correlation between the PJO and the annular mode was observed in this stage as for the NH and for the SH. As expected from the meridional cross sec- NH tions in Figure 1, the height anomaly in the NH accompanied (Dec) EOF2 (Mar) EOF2 a change in the meridional propagation of waves, particularly over the Atlantic and eastern Asian sector. Such change in 4- wave propagation was not found in the SH. In the following 4- O0 period (lags 2 for NH and 5 for SH), the annular modes e- 2- volved into more regional north-south dipole patterns over the North Atlantic sector in the NH and the central southern Pacific sector in the SH Temporal Structure The standar deviation of interannual variability of the polar temperatures in each month of the year repeated for three cycles is shown in Figures 6b and 7b for the NH and SH. Considerable variability occurs over the whole depth of the stratosphere during the cold season from November to April in the NH. In contrast, the zone of large variability in the SH is confined to a relatively small height range, which gradually descends from August to December. A comparison of the seasonal structures of the standard deviation with the anoma- -2- ß O0 -'4-2 o ' (b) SH lous patterns of individual years revealed that those anomalous patterns fit well into the standard deviation structure in the SH, whereas in the NH, the "envelope" of anomalous patterns may correspond to the patterns of standard deviation. This demonstrates that intraseasonal variability is more ' / closely locked to the annual cycle in the SH. -4' -4' / The above conclusion was drawn from the temperature variability of the polar region. Although polar temperature -4 c2 0 4 c2 0 is a good indicator of the wave-mean flow interaction in the Figure 11. Distribution of the expansion coefficients of EOF midlatitudes of the stratosphere, it is relatively insensitive to 1 and 2 of the zonal-mean zonal wind anomaly displayed for subtropical wind variability in early winter. To investigate selected months. (a) No hern Hemisphere; December and the variability of the entire stratosphere, analysis is made in March. (b) Southern Hemisphere; July and October. Dashed low-dimensional space spanned by two leading EOFs ca!cu- lines in Figure 11 b indicate least squares fit. EOF1 o -2- EOF1, -4 :2 o,, (JuO EOF2 (Oct) EOF I I i I

9 KURODA AND KODERA: VARIABILITY OF POLAR NIGHT JET 20,711 (a) NH POWER SPECTRUM Frequency (cycle/yr) (b) SH POWER SPECTRUM,, 0 ß Frequency (cycle/yr) winter in the SH, but positive anomalies are formed in the polar region (lags 0 and 1 ), creating a meridional dipole pattern in the stratosphere. The differences of the two hemispheres can be explained by the differences in the convergence zones of the E-P flux in both hemispheres (Figure 2). Convergence of the E-P flux occurs mainly in the lower latitudes in early winter in the SH. The induced downward and poleward motion of the residual circulation produces warming in the midlatitudes by adiabatic heating and acceleration of the zonal wind in the polar region through Coriolis forcing, respectively. Since wave forcing is smaller in the high latitudes in early winter in the SH, positive wind anomalies are produced in the polar region, in contrast with the negative anomalies in the lower Figure 12. Power spectrum of the leading mode of the MSSA latitudes. When the convergence zone shifts to the polar for the anomalous zonal-mean zonal wind for all months in region in spring (e.g., lag 3 of Figure 2b), the upward and the (a) Northern and (b) Southern Hemispheres as a function downward flows of residual circulation produce adiabatic of frequency (cycle/year) (see text for details). cooling and warming in the subtropics and the polar region, respectively. The poleward flow accelerates the zonal flow in the subtropics, where wave forcing is small. However, calculated here only uses data from the cold season. The band leogth was set to 7 months (a lag component of 0 to 6 months) to cover the entire cold season. The leading two modes explain almost the same total variance (12%) and form a pair. The time evolution of the spatial pattern of the leading mode was very similar to that shown in Figure 1 a. The power spectrum of the normalized time coefficient of the leading mode is shown in Figure 12a. The power this effect is quite small. In the NH, on the other hand, the convergence of anomalous E-P flux occurs in high latitudes from early winter; so the dipole structure observed in the SH is not found in this period. The growth of positive anomalies in zonal winds in the NH at lag 2 is mainly explained by the change in wave forcing in the upper stratosphere. The reason why the annular modes preferably appear when the zones of maximum polar temperature anomaly descend to spectrum exhibited a sharp peak at the period of 5.1 months. the lowermost stratosphere can be explained as follows: If the The lateral peaks around 3.5 months and 7 months may be attributed to amplitude modulation of the 5-month signal by the annual cycle. The same analysis was conducted for the SH. In that case, the leading mode explained 18% of the total variance and differed from the second one (10%). Therefore the leading two modes do not form a pair, which indicates that there is polar temperature at the lowermost stratosphere is warmer, the zonal wind at high-latitude stratosphere becomes weaker owing to the thermal wind relation. As a result, the refractive index in the high latitude of the lowermost stratosphere becomes larger, and the tropospheric waves tend to propagate more polewardly, which makes a negative phase of the annular mode. no apparent periodic signal with the intraseasonal timescale. The vertical propagation of anomalous E-P flux in the The power spectrum of the normalized time coefficient of the NH changes with the development of the opposite polarity leading mode in the SH is shown in Figure 12b. It is clear of an anomalous zonal wind in the subtropical stratosphere that the power associated with the intraseasonal timescale is (Figure 1 a). It is quite interesting that in the SH the vertical very small and that most is associated with the interannual propagation of the waves is maintained over 6 months with timescale (i.e., a period longer than 1 year) in the SH. This the development of the wind in the stratosphere; the central indicates that there exists only some periodicity with the latitude of upward propagation shifts to a higher latitude with interannual timescale in the SH. We found that the peaks the increase in negative wind anomalies (Figure 1 b). near the origin (note that the power at the origin is zero owing The successive downward propagation of negative and tu ti c u c ui' a,,u,,,aiuu d,m ) and at the period of 2 years positive anomalies in the zonal wind anomalies in the NH correspond to a long-term trend and the biennial oscillation, suggests a periodic nature, which can be more clearly seen in respectively, as noted by KK98. GCM simulations under perpetual winter conditions [Christiansen, 1999; Yamazaki and Shinya, 1999]. Such variability 4. Discussion The spatial and temporal characteristics of the PJO were investigated for both the Northern and Southern Hemiis identified as stratospheric vacillation in a mechanistic model [Holton and Mass, 1976; Scaife and James, 2000; Kodera and Kuroda, 2000a]. Stratospheric vacillation cannot be expected during winter in the SH, where stationary planetary spheres. Negative zonal wind anomalies formed in the sub- wave activity is much weaker. However, when the annual tropical stratopause lags -1 and 0 propagated poleward cycle is taken into account, planetary waves can propagate and downward in both hemispheres (Figure i). However, in _'._.1_!....J... : J ß rnl... HltO tilt: Stl-tltOSpHt lt uurlng autulilil allcl spring Itctums, a I 8 L the NH, negative anomalies quickly shifted poleward, giving Yoden, 1990] and produce a wave-mean flow interaction in way to the positive anomalies in the subtropics and forming the stratosphere. In the real world, planetary waves always a deep meridional dipole structure (lags 1 and 2). In contrast, propagate into the stratosphereven during the winter but negative anomalieshift poleward only minimally during the are deflected from the polar region by the strong polar night

10 20,712 KURODAND KODERA: VARIABILITY OF POLAR NIGHT JET jet. Hence no important wave-mean flow interactions are produced in the polaregion until spring. This difference in climatological conditions during winter may explain why no poleward shift of the anomalous wind occurs during midwinter in the SH. Therefore better similarity is found with the early winter part of the NH PJO if the midwinter period is omitted from SH PJO, as illustrated in Figure 8. This suggests that the PJO in both hemispheres essentially the same phenomenon produced through the interaction between the planetary waves and zonal-mean flow. The similarity between the NH and SH PJOs is also found in association with the tropospheric circulation pattern in early winter. This may be a coincidence, but wave trains appear in 500-hPa height fields (lag-1 of Figure 9) in both hemispheres, with enhanced vertical propagation of the planetary waves (lag- 1 of Figure 1 ). In addition, annular modes appear in the troposphere at the stage when the stratospheric anomalies propagate down into the lower stratosphere (Figures 5 signal can be traced back to before the formation of the AO in the troposphere at the annular period. This indicates that the AO-related variability could be understood better in the framework of the PJO. 5. Conclusion The month-to-month variability of the stratosphere and troposphere during the cold season was investigated by E- SVD analyses for both Northern and Southern Hemispheres. The extracted variability in both hemispheres can be considered to be the same variability mode, i.e., the PJO. It is characterized by a poleward and downward propagation of zonal-mean zonal wind anomalies from the subtropical stratopause to the troposphere through interaction with planetary waves (particularly the wave number 1 component). An association of the annular mode with the PJO was also com- monly found in both hemispheres. The major difference is their temporal characteristics. The NH PJO is essentially an intraseasonal variability for which the envelope is controlled by an annual cycle, while the SH PJO has an interannual variability with an intraseasonal structure. This difference is thought to originate from a weak wavemean flow interaction during the SH winter. In short, the and 9). SH PJO is similar to an NH PJO for which evolution is The AO index changes its polarity from negative to posfrozen during the winter. It would be interesting to conduct itive in late winter to spring in the NH, according to the GCM experiments using perpetual autumn or spring runs to successive downward propagation of the positive and neginvestigate the structure of the SH PJO and its association ative temperature anomalies with the PJO (Figure 5). This with the annular mode. result shows good agreement with the analysis of the origin In the present study, we focused our attention on the spatial of the month-to-month variability of the AO by Kodera and and temporal structures of the intraseasonal timescale. It is Kuroda [2000b]. also importanto know the origin of the interannual variations Downward propagation of the temperature anomaly seen in the PJOs, particularly the biennial oscillation and trends in Figure 5 is reminiscent of the downward propagation of in the SH PJO (KK98). the AO signal by Baldwin and Dunkerton [ 1999]. However, it must be noted that the downward signal shown in this figure Acknowledgment. The authors are grateful to T Hirooka for is the polar temperature due to the PJO. Therefore the AO providing Met Office, UK data. References Bailey, M. J., A. O'Neill, and V. D. Poge, Stratospheric analysis Changes in the meridional propagation of tropospheric produced by the United Kingdom Meteorological Office, J. Appl. waves with the formation of' the annular mode were rec- Meteoro!., 32, , ognized in both hemispheres. However, they occmted in Baldwin, M.P., and T J. Dunkerton, Propagation of the Arctic stationary waves in the NH and could be seen only for tran- Oscillation from the stl atosphere the troposphere, J. Geophys. sient waves in the SH, which is consistent with the annular Res., 104, 30,937-30,946, mode differences in the two hemispheres [Limpasuvan and Baldwin, M.P., X. Cheng, and T J. Dunkerton, Observed correlation between winter-mean tropospheric and stratospheric anoma- Hartmann, 2000]. lies, Geophys. Res. Lett., 21, , The annular mode evolves into a less zonally symmetric Bretherton, C. S., C. Smith, and J. M. Wallace, An intercomparison structure with time, as seen in Figure 9. This could also be of method for finding coupled patterns in climate data, J. Clim., coincidental, but it is interesting to note that these sectors 5, , where dipole anomalies develop correspond in both hemi- Christiansen, B., Stratospheric vacillations in a general circulation spheres to the sector where double-jet or blockings occur model, J. Atmos. Sci., 56, , Fartara, J. D., and C. R. Mechoso, An observational study of the fimore frequently [Lejenas, 1984]. This suggests further excinal warming in the Southern Hemisphere stratosphere, Geophys. tation of the regional mode. Res. Lett., 13, , Thompson and Wallace [2000] noted an out-of-phase rela- Haynes, P. H., C. J. Marks, M. E. Mcintyre, T G. Shepherd, and tionship in the annular mode between the polar and equatorial K. P. Shine, On the "downward control" of extratropical diabatic temperatures in the lower stratosphere. A similar tempera- circulations by eddy-induced mean zonal forces, J. Atmos. Sci., ture anomaly pattern can be seen in Figure 4 (lag 1 for NH 48, , and 4 for SH). In fact, these temperature anomalies are pro- Holton, J. R., and C. Mass, Stratospheric vacillation cycles, J. Atmos. Sci., 33, , duced through changes in residual circulation associated with Kalnay, M. E. et al., The NCEP/NCAR reanalysis project, Bull. the stratospheric process (PJO). As the annular mode and the Am. Meteorol. Soc., 77, , PJO are well correlated at the annular period, the structure of Kodera, K., On the origin and nature of the interannual variability of the annular mode reflects that of the PJO. Note that the cor- the winter stratosphere circulation in the Northern Hemisphere, relation does not mean one-to-one correspondence [Kodera J. Geophys. Res., 100, 14,077-14,087, and Kuroda, 2000b]. Kodera, K., and Y. Kuroda, A mechanistic model study of slow-

11 KURODA AND KODERA: VARIABILITY OF POLAR NIGHT JET 20,713 ly propagating coupled stratosphere-troposphere variability, J. Scaife, A. A., and I. N. James, Response of the stratosphere to Geophys. Res., 105, 12,361-12,370, 2000a. interannual variability of tropospheric planetary waves, Q. J. R. Kodera, K., and Y. Kuroda, Stratospheric and tropospheric aspects Meteorol. Soc., 126, , of the Arctic Oscillation, Geophys. Res. Lett., 27, , Shiotani, M., and I. Hirota, Planetary wave-mean flow interaction in 2000b. the stratosphere: A comparison between Northern and Southern Kodera, K., Y. Kuroda, and S. Pawson, Stratospheric sudden warm- Hemispheres, Q. J. R. Meteorol. Soc., 111, , ings and slowly propagating zonal-mean zonal wind anomalies, Soel, D.-I., and K. Yamazaki, Residual mean circulation in the J. Geophys. Res., 105, 12,351-12,360, stratosphere and upper troposphere: Climatological aspects, J. Kuroda, Y., An effective SVD calculation method for climate anal- Meteowl. Soc. Jpn., 77, , ysis, J. Meteorol. Soc. Jpn., 76, , Thompson, D. W. J., and J. M. Wallace, The Arctic Oscillation Kuroda, Y., and K. Kodera, Interannual variability in the tropo- signature in the wintertime geopotential height and temperature sphere and stratosphere of the Southern Hemisphere winter, J. fields, Geophys. Res. Lett., 25, , Geophys. Res., 103, 13,787-13,799, Thompson, D. W. J. and J. M. Wallace, Annular modes in extrat- Kuroda, Y., and K. Kodera, Role of planetary waves in the ropical circulation, part 1, Month-to-month variability, J. Cli n., stratosphere-troposphere coupled variability in the Northern Hemi- 13, , sphere winter, Geophys. Res. Lett., 26, , Vautard, R., and M. Ghil, Singular spectrum analysis in nonlinear Labitzke, K., On the interannual variability of the middle strato- dynamics, with application to paleoclimatic time series, Physica sphere during the northern winters, J. Meteorol. $oc. Jpn., 60, D, 35, , , Weare, B.C., and J. S. Nasstrom, Example of extended empirical Lejenas, H., Characteristics of Southern Hemisphere blocking as orthogonal function analysis, Mon. Weather Rev., 110, , determined from a time series of observational data, Q. J. R Meteorol. Soc., 110, , Yamazaki, K., and Y. Shinya, Analysis of the Arctic Oscillation Limpasuvan, V., and D. L. Hartmann, Wave-maintained annular simulated by AGCM, J. Meteorol. Soc. Jpn., 77, , modes of climate variability, J. Clim., 13, , Perlwitz, J., and H.-E Graf, The statistical connection between Yoden, S., An illustrative model of seasonal and interannual varitropospheric and stratospheric circulation of the Northern Hemi- ation of the stratospheric circulation, J. Atmos. Sci., 47, sphere in winter, J. Clim., 8, , , Perlwitz, J., H.-E Graf, and R. Voss, The leading variability mode of the coupled troposphere-stratosphere winter circulation in different climate regimes, J. Geophys. Res., 105, , Plumb, R. A., On the three-dimensional propagation of stationary waves, J. Atmos. Sci., 42, , Y. Kuroda and K. Kodera, Meteorological Research Institt'te, Plumb, R. A., On the seasonal cycle of stratospheric planetary 1-1 Nagamine, Tsukuba, Ibaraki, , Japan. ( waves, Pure Appl. Geophys., 130, , kuroda@ mri-jma.go.jp; kodera@ mri-jma.go.j p) Randel, W. J., Global Atmospheric Circulation Statistics, mb, NCAR Tech. Note 366+STR, 256 pp., Natl. Cent. for A :mot,. (Received September 18, 2000; revised April 20, 2001; Res., Boulder, Colo., accepted April 30, 2001.)