Abstract

Through analysis of observational data for the period of 1973–94, the late-winter formation of an interannual seesaw between the surface Aleutian and Icelandic lows (AL and IL, respectively) is shown to significantly impact the covariance structure of the leading mode of the interannual variability in the geopotential height field over the extratropical Northern Hemisphere. The tropospheric leading mode for early winter (November–January) is characterized by a polar–midlatitude dipole over the Euro–Atlantic sector with a high degree of the annularity, coupled with the anomalous lower-stratospheric polar vortex. Over the North Pacific, no significant anomalies are associated with this mode. After the formation of the AL–IL seesaw, however, the dipole no longer dominates in the upper-tropospheric variability. The dipole signature is masked in late winter (February–April) by the predominant combined signature of the so-called Pacific–North American pattern and a meridional dipole over the northwestern Atlantic as an upper-level manifestation of the seesaw. Though somewhat less pronounced, the leading mode of the near-surface variability is modified accordingly in late winter by the superposition of the distinct signature of the AL–IL seesaw. The annularity of the leading mode of the tropospheric variability is thus reduced in late winter, particularly at the upper levels. Nevertheless, because of the particular geographical alignment between the anomalous AL and IL, their seesaw changes the zonal wind coherently between the two ocean basins, yielding a strong projection on the meridional plane whose latitudinal profile is almost indistinguishable from the counterpart of the Arctic–midlatitude dipole.

It is argued that what is called the Arctic oscillation in some recent literature, defined as the leading mode of the sea level pressure variability for the entire cold season, may be interpreted as a superposition of the AL–IL seesaw upon a dominant signal of the Arctic–midlatitude dipole. The corresponding leading mode for the upper troposphere primarily represents the variability associated with the seesaw. It is also argued that the late-winter tropospheric variability over the North Atlantic may not necessarily be associated with the Arctic–midlatitude dipole. The remote influence of the North Pacific variability accounts for as much as 30%–50% of the variance in the vicinity of the IL for the data period considered.

1. Introduction

Recent studies have revealed that a seesawlike oscillation is dominant in the wintertime Northern Hemisphere lower stratosphere between the polar vortex and its surroundings, which is strongly coupled with the mode of tropospheric variability reminiscent of the well-known North Atlantic oscillation (NAO; Baldwin et al. 1994; Cheng and Dunkerton 1995; Perlwitz and Graf 1995; Kitoh et al. 1996; Kodera et al. 1996; Thompson and Wallace 1998, 2000). Such variability with a deep equivalent barotropic structure is pronounced particularly in the winter half of the year. Its tropospheric manifestation has indeed been identified in the leading empirical orthogonal function (EOF) of wintertime monthly sea level pressure (SLP) anomalies (Kutzbach 1970; Wallace and Gutzler 1981; Trenberth and Paolino 1981; Thompson and Wallace 1998). This pattern, characterized by significant pressure anomalies with a particular sign that cover the entire Arctic region, and exhibits a high degree of zonal symmetry. Introducing the nomenclature “Arctic oscillation (AO),” Thompson and Wallace (1998) distinguished the EOF pattern from the somewhat more regional pattern of the NAO derived from its station-based index. Thompson and Wallace (2000), Thompson et al. (2000), and Wallace (2000) postulated the concept of “annular mode,” to describe essential dynamics common in the AO and its counterpart in the Southern Hemisphere. Owing to their particular zonal symmetry, the annular modes in the two hemispheres are well represented in the zonally averaged zonal wind anomalies on the meridional plane. Deser (2000) showed that the annularity of the AO is due to the dominance of the associated Arctic anomalies rather than to the midlatitude interbasin link between the North Atlantic and North Pacific. Wallace (2000) and Thompson et al. (2000) further argued that the NAO may be regarded as a regional expression of the Northern Hemisphere annular mode or AO. Yet, Ambaum et al. (2001) argued that unlike the AO, the NAO can be extracted from the data in a physically consistent manner and the hemispheric configuration of the AO may be an artifact of an EOF analysis. Many of the latest studies mentioned above nevertheless emphasized the strong linkage between the atmospheric anomalies over the Arctic and North Atlantic as a dynamically unified entity, which accounts for a larger fraction of the interannual variance within the extratropical Northern Hemisphere than any other anomaly pattern.

Bjerknes (1966) showed, however, that a seesawlike interannual oscillation occurred in some of the winters during the late 1950s between the North Atlantic and North Pacific. Based on observed SLP and lower-tropospheric anomalies, van Loon and Rogers (1978),1 Rogers and van Loon (1979), Wallace and Gutzler (1981), and van Loon and Madden (1983) demonstrated the existence of a significant anticorrelation between the surface Icelandic low (IL) and Aleutian low (AL), suggesting that such a seesaw between the two permanent low pressure cells (hereafter referred to as the AL–IL seesaw) may be an additional important characteristic of the NAO. They pointed out that the seesaw signature was apparently imprinted on the leading EOF pattern of monthly SLP anomalies as obtained by Kutzbach (1970). Angell and Korshover (1974) found the seesaw signature in the quasi-biennial timescale. The seesaw signature is hinted at even in the leading EOF of wintertime monthly SLP anomalies obtained by Thompson and Wallace (1998), in which they identified the AO. The seesaw signature is also apparent in composite and/or correlation maps of the lower-tropospheric circulation obtained, for example, by Dickson and Namias (1976), who studied a remote influence of the anomalous surface baroclinicity along the U.S. east coast upon the IL intensity and by Overland et al. (1999), who examined the AL anomalies associated with the decadal climate variability. One may recognize a hint of a surface signature of the seesaw in statistics presented in such recent studies as Walsh and Chapman (1990), Watanabe and Nitta (1998, 1999), Xie et al. (1999), and Deser et al. (2000). The signature of the AL–IL seesaw can be recognized more clearly in the pattern of the variability prevailing in the wintertime midtropospheric circulation as identified through an EOF or singular value decomposition (SVD) and an cluster analysis (e.g., Cheng and Wallace 1993; Wallace et al. 1993; Cheng and Dunkerton 1995; Koide and Kodera 1999). It is also fairly clear in the correlation and/or composite maps of the midtropospheric anomalies as presented in Perlwitz and Graf (1995), Ting et al. (1996), Overland et al. (1999), and Xie et al. (1999). As pointed out by Wallace and Gutzler (1981), the anomalous AL and IL are closely related to the Pacific–North American (PNA) pattern and NAO, respectively, which are the two most prominent teleconnection patterns in the wintertime Northern Hemisphere midtropospheric circulation (Kushnir and Wallace 1989). The significant anticorrelation between the anomalous IL and AL is thus reflected in the leading EOF (or SVD) pattern of the midtropospheric circulation, as obtained in the aforementioned studies, as a combination of the NAO and PNA pattern. In fact, Cheng and Dunkerton (1995) were successful in decomposing the tropospheric leading SVD pattern into the NAO and PNA pattern through a rotation technique.

In our recent paper (Honda et al. 2001; hereinafter referred to as Part I), we investigated the seasonal dependence of the AL–IL seesaw and the processes involved in its formation on the basis of observational data for 22 recent years. We found that the seesaw tends to be particularly apparent during late winter (February–mid-March), and it exhibits an equivalent barotropic structure extending from the surface to the upper troposphere (Fig. 1). As shown in Fig. 6 of Part I, the seesaw formation is initiated by midwinter as the development of the AL anomalies. Then, in midwinter, the influence of the anomalous AL spreads downstream, forming a PNA-like pattern over North America and another wavy pattern across the North Atlantic that includes the incipient IL anomalies. The seesaw formation is completed by late February after the IL anomalies amplify forming into a dipolelike structure over the northwestern Atlantic, while the AL anomalies remain strong. It was shown in Part I that the feedback forcing from synoptic-scale eddies migrating along the Pacific and Atlantic storm tracks significantly contributes to the development and maintenance of the AL and IL anomalies, respectively. An important characteristic of the seesaw is that the anomalous IL forms through the downstream influence of the North Pacific variability. This mechanism has already been discussed by Bjerknes (1966), who attempted to explain the cold winter over Northern Europe as a remote influence of the 1957/58 El Niño event. Recently, Lau and Nath (2001) and Honda et al. (2001) confirmed that the remote influence of the El Niño–Southern Oscillation can reach into the North Atlantic, which can also contribute to the formation of the AL–IL seesaw. The extended influence of the PNA pattern onto the North Atlantic was also found in a recent study by Ambaum et al. (2001). One may regard the AL–IL seesaw as merely an extension of the so-called PNA pattern, but it was shown in Part I that how the influence of the anomalous AL extends into the North Atlantic is somewhat different over North America and over the North Atlantic. Our analysis in Part I based on the wave activity flux of Takaya and Nakamura (1997, 2001) indeed suggests that the stationary anomaly centers over North America and the North Atlantic are associated with stationary Rossby wave trains. Yet, the interaction with the Atlantic storm track is of additional importance in the development of the IL anomalies. Moreover, as shown in Part I, the extension of the PNA pattern into the Atlantic does not always occur. The tendency does not become apparent until midwinter.

Fig. 1.

Maps of (a) 250-hPa geopotential height (Z250) and (b) SLP for its peak period (31 Jan–16 Mar), linearly regressed upon the index of the AL–IL seesaw index (AII), representing local changes in (a) Z250 (m) and (b) SLP (hPa) when the AII increases by its unit standard deviation. Light and dark shading represent anomalies that are correlated with the AII at the 90% and 95% confidence levels, respectively, based upon the t statistic with 11 degrees of freedom (one-half of 22 yr) assumed for a conservative measure of the statistical significance. After Honda et al. (2001) 

Fig. 1.

Maps of (a) 250-hPa geopotential height (Z250) and (b) SLP for its peak period (31 Jan–16 Mar), linearly regressed upon the index of the AL–IL seesaw index (AII), representing local changes in (a) Z250 (m) and (b) SLP (hPa) when the AII increases by its unit standard deviation. Light and dark shading represent anomalies that are correlated with the AII at the 90% and 95% confidence levels, respectively, based upon the t statistic with 11 degrees of freedom (one-half of 22 yr) assumed for a conservative measure of the statistical significance. After Honda et al. (2001) 

Fig. 6.

As in Fig. 2, but for the first EOFs of monthly (from left to right) SLP, Z250, and 50-hPa geopotential height (Z50) for the NDJ period (upper panels) and FMA period (lower panels). Contour intervals: 1 hPa (a) and (d), 20 m (b) and (e), and 30 m (c) and (f). Negative contours are dashed

Fig. 6.

As in Fig. 2, but for the first EOFs of monthly (from left to right) SLP, Z250, and 50-hPa geopotential height (Z50) for the NDJ period (upper panels) and FMA period (lower panels). Contour intervals: 1 hPa (a) and (d), 20 m (b) and (e), and 30 m (c) and (f). Negative contours are dashed

Given the particular seasonal dependence of the seesaw formation between the IL and AL, both of which reside where the interannual variability tends to be strongest, one may wonder what impact the seesaw formation might possibly exert upon the structure of the leading mode of the tropospheric variability over the extratropical Northern Hemisphere especially during late winter. Should the impact be substantial, identifying the AO-like zonally symmetric structure or the annular mode suffers from the contamination by the zonally asymmetric signal of the AL–IL seesaw. Furthermore, part of the tropospheric fluctuations in the North Atlantic especially in the vicinity of the IL might be influenced by the North Pacific anomalies rather than in conjunction with the Arctic anomalies.

Bearing these issues in mind, we perform the following statistical analyses in this study with operational analysis data of the atmospheric circulation over 22 recent years as in Part I. It should be stressed that arguing the reality of the AO or the dynamical relevance of the concept of the annular mode is beyond the scope of our study. We nevertheless acknowledge the fact that what is called the AO in some recent literature, an anomaly pattern representing the Arctic–midlatitude seesaw with a high degree of zonal symmetry, is captured in the leading EOF of the wintertime SLP anomalies. We assume the dominance of the AO at the first place as a sort of null hypothesis for the following examinations. First, in section 3, we apply an EOF analysis separately to nine overlapping subseasonal periods from early December to early April, in order to examine (sub) seasonal dependence, if any, of the leading mode of the tropospheric variability. We are particularly interested in any structural change in the first EOF that may occur in the course of the formation of the AL–IL seesaw in February. In section 4, we investigate the coupling between the tropospheric and stratospheric anomalies in the course of the entire winter season and how significantly the coupling tends to be influenced by the seesaw formation between the IL and AL. Then, in section 5, we compare the first EOF of the wintertime tropospheric anomalies over the extratropical Northern Hemisphere with the first EOF of the corresponding anomalies from which the signal of the AL–IL seesaw or what is called the annular mode in some literature has been statistically removed. In this manner we attempt to assess how much the seesaw-related interannual variability contributes to the leading mode of variability over the extratropical Northern Hemisphere. Then, in section 6, we compare the zonally averaged zonal wind anomalies accompanied by the seesaw with the counterpart related to another mode of the dominant variability. In section 7, summarizing our findings, we refer to the significance of a contribution from the North Pacific variability to the interannual variability in late winter over the northern part of the North Atlantic, which is through an atmospheric bridge in the form of the AL–IL seesaw.

2. Data and analysis procedures

The primary dataset used for our analysis is the National Meteorological Center [currently the National Centers for Environmental Prediction (NCEP)] operational analyses obtained from the National Center for Atmospheric Research (NCAR) data library. We used twice-daily fields of sea level pressure (SLP), 250- and 500-hPa geopotential height (Z250 and Z500, respectively) and 250-hPa wind (U250 and V250), which are archived on a 1977-point octagonal grid that covers the entire area north of 20°N. We also use monthly-mean and twice-daily 50-hPa geopotential height (Z50) fields based on the NCEP–NCAR reanalyses, which are available on a 2.5° × 2.5° regular latitude–longitude grid (Kalnay et al. 1996). They have been obtained also from the NCAR data library.

As in Part I, we confined ourselves to the 22-yr period from 1973 to 1994, because the negative correlation between the AL and IL intensities was stronger in that period than any other period over the last five decades.2 Then, we can assess the impact of the AL–IL seesaw on the leading mode of the interannual wintertime variability in the Northern Hemisphere in the clearest manner as possible. At the same time, the impact as revealed through the following analysis may be the greatest impact the seesaw can possibly exert. The daily AL and IL intensities were defined as the respective minima in a 31-day moving-average SLP field within given areas over the North Pacific and North Atlantic (Fig. 1 of Part I), respectively. In recognition of the strongest negative correlation coefficient between their intensities in February through mid-March (Fig. 2 of Part I), we defined the AL–IL index (AII) as the normalized IL-central pressure subtracted from the normalized AL-central pressure, both of which had been averaged over the “peak period,” that is, the 45-day period from 31 January to 16 March (Fig. 4a of Part I). In order to investigate the seasonal evolution of the seesaw, we prepared circulation anomaly data averaged within each of nine overlapping 45-day periods that are staggered equally by 15 days (17 Nov–31 Dec, 2 Dec–15 Jan, 17 Dec–30 Jan, … , 17 Mar–30 Apr), as in Part I. These nine periods are centered in early December, late December, early January, … , and early April, and the peak period corresponds to the late-February period.

Fig. 2.

The first EOFs of Z250 anomalies for the nine overlapped 45-day periods that roughly correspond to (a) early Dec, (b) late Dec, (c) early Jan, … , and (i) early Apr, showing the (sub) seasonal modification in the structure of the dominant interannual variability in Z250 over the domain poleward of 20°N. Plotted for each of the periods is the linear regression coefficient between a local Z250 anomaly and the leading PC, which corresponds to a local change in Z250 (every 20 m; dashed lines for negative values) when a given PC increases by its unit standard deviation. Light and dark shading represent anomalies that are correlated with the PC at the 90% and 95% confidence levels, respectively, as in Fig. 1 

Fig. 2.

The first EOFs of Z250 anomalies for the nine overlapped 45-day periods that roughly correspond to (a) early Dec, (b) late Dec, (c) early Jan, … , and (i) early Apr, showing the (sub) seasonal modification in the structure of the dominant interannual variability in Z250 over the domain poleward of 20°N. Plotted for each of the periods is the linear regression coefficient between a local Z250 anomaly and the leading PC, which corresponds to a local change in Z250 (every 20 m; dashed lines for negative values) when a given PC increases by its unit standard deviation. Light and dark shading represent anomalies that are correlated with the PC at the 90% and 95% confidence levels, respectively, as in Fig. 1 

Fig. 4.

Seasonal evolution of the lag correlation coefficient between the AII defined for the peak period (late Feb) and the leading PC of each of anomalous Z250 (closed circles) and SLP (open circles) averaged over each of the nine 45-day periods from early Dec to early Apr. The 95% confidence level with 11 degrees of freedom is indicated for a conservative measure of the statistical significance. All based on statistics for the period of 1973–94

Fig. 4.

Seasonal evolution of the lag correlation coefficient between the AII defined for the peak period (late Feb) and the leading PC of each of anomalous Z250 (closed circles) and SLP (open circles) averaged over each of the nine 45-day periods from early Dec to early Apr. The 95% confidence level with 11 degrees of freedom is indicated for a conservative measure of the statistical significance. All based on statistics for the period of 1973–94

3. Seasonal dependence of the dominant interannual variability in the troposphere

In this section, we investigate the seasonal evolution of the dominant interannual variability in the tropospheric circulation during the course of the winter season, focusing on how the leading EOFs for the upper and lower troposphere are modified under the influence of the formation of the AL–IL seesaw in late winter. For this purpose, we first performed an EOF analysis on the nine 45-day mean anomalies of Z250 and SLP, separately, within the entire domain poleward of 20°N. Figures 2 and 3 show maps of the linear regression coefficients between the mean Z250 and SLP anomalies, respectively, and their leading principal components (PCs) for the same nine periods as we defined in Part I to illustrate a typical evolution of the seesaw. Those maps represent typical anomaly patterns associated with the prevailing interannual variability in Z250 and SLP for the individual subseasonal periods.

During early winter (Figs. 2a–c), the leading EOF of Z250 represents an Arctic–midlatitude seesaw with a high degree of annularity. The associated midlatitude anomalies consist of four cells. In late January the associated anomalies over the North Pacific become more apparent (Fig. 2d), and then in early February, significant anomalies over the Arctic suddenly disappear (Fig. 2e). At this stage, the leading EOF has lost its annularity and it is characterized by the PNA pattern and a meridional dipole over the northwestern Atlantic as the upper-tropospheric manifestation of the AL–IL seesaw (Fig. 1a). Nearly the same pattern is also obtained as the leading EOF for late February (i.e., the peak period of the seesaw; Fig. 2f) and early March (Fig. 2g). This profound structural change in the leading mode occurs concomitantly with the formation of the seesaw shown in Part I. After these periods the two anomaly patterns that constitute the seesaw become less apparent, while significant signals reemerge over the Arctic and remain significant until early spring (Figs. 2h and 2i).

Though less pronounced, some significant changes in the structure of the dominant variability are also found in the SLP field between the early and late winter periods. In late December through late January (Figs. 3b–d), the leading EOF is characterized by a prominent polar–midlatitude dipole between the Arctic and North Atlantic that resembles the NAO with no significant anomalies over the North Pacific. In early February (Fig. 3e), the North Pacific anomalies start becoming recognizable and the Arctic anomaly center starts weakening, while the center near Iceland remains strong. As these tendencies continue with the enhancement of anomalies over the midlatitude northwestern Atlantic, the characteristic structure in the leading mode has changed within the month of February from the NAO-like polar–midlatitude dipole to a more hemispheric pattern similar to the AL–IL seesaw (Fig. 3f; cf. Fig. 1b). After this period the midlatitude anomalies over the two ocean basins become less apparent gradually, while the Arctic anomaly center reemerges and remains strong until early April (Figs. 3g–i).

Fig. 3.

As in Fig. 2, but for the leading EOFs of SLP (contour intervals: 1 hPa)

Fig. 3.

As in Fig. 2, but for the leading EOFs of SLP (contour intervals: 1 hPa)

Those rather sudden changes in the leading EOF structures evident in Figs. 2 and 3 appear to be related to the formation of the AL–IL seesaw. To confirm this point, we computed the lag correlation coefficient between the PC that corresponds to each of the EOFs shown in Figs. 2 and 3 and the AII defined for the peak period of the seesaw (Fig. 4). For the upper troposphere, the correlation is low in December but rapidly increases in January. After reaching a peak in late February, it starts decreasing gradually but it remains significant until early spring. The correlation exceeds 0.80 in February through mid-March, which confirms the dominance of the AL–IL seesaw in the leading mode of the late-winter interannual variability as shown in Fig. 2. A similar tendency is evident in the corresponding correlation between the AII and leading PC for SLP. Interestingly, the rapid increase in the PC–AII correlation for the upper level precedes the corresponding increase for the lower level by half a month, which will be discussed in the final section.

The structural transition in the leading mode of the tropospheric variability that occurs in late January and February is further confirmed by examining the “persistence” of the leading PC over a single winter season (Fig. 5). Here, the seasonal persistence is measured as the correlation coefficients between the leading PC for a particular 45-day period and those for the eight other 45-day periods. If a particular anomaly pattern remains dominant throughout the winter, the leading PC for an early-winter period should be significantly correlated with those for late-winter periods. The correlation of the leading PC of Z250 for each of the late-December and early-January periods, on one hand, drops abruptly in February (Fig. 5a). On the other hand, the corresponding correlation for each of the four periods in late February through early April increases rapidly in January and early February and maintains the significance afterward. The leading PC of the SLP anomalies for each of the periods exhibits almost the same seasonal dependence in its persistence (Fig. 5b) as the counterpart for of Z250. The loss of the persistence of the early-winter PC in February is even more marked for SLP. The overall tendency shown in Fig. 5 is in good agreement with those in Fig. 4, as the leading EOF for late winter represents the AL–IL seesaw. This result, combined with Figs. 2 and 3, suggests that the dominant anomaly patterns in the troposphere are fundamentally different between the early- and late-winter seasons.

Fig. 5.

(a) As in Fig. 4, but for correlation coefficients between the leading PC of Z250 for a given 45-day period and those for the eight other 45-day periods. (b) As in (a) but for the leading PC of SLP

Fig. 5.

(a) As in Fig. 4, but for correlation coefficients between the leading PC of Z250 for a given 45-day period and those for the eight other 45-day periods. (b) As in (a) but for the leading PC of SLP

In order to show the significant influence of the AL–IL seesaw upon the seasonality of the leading mode of interannual variability in a more straightforward manner, we applied an EOF analysis to the monthly extratropical SLP anomalies for the November–January (NDJ) and February–April (FMA) periods, separately. The leading EOF for the NDJ period represents an Arctic–North Atlantic dipole with a strong projection upon the NAO (Fig. 6a). The EOF includes no significant anomalies over the North Pacific. The NAO-like pattern is still apparent in the first EOF for the FMA period (Fig. 6d), but one can recognize a significant contribution from the AL–IL seesaw as well with significant anomalies in the North Pacific. In fact, the correlation coefficient of the AII with the leading PC is only 0.11 in NDJ, but it climbs up to 0.85 in FMA. The same analysis as above was applied to monthly Z250 anomalies, where an even more pronounced difference was found between the early- and late-winter periods. The leading EOF for the NDJ period is characterized by a seesaw between the Arctic and midlatitude (Fig. 6b). The pattern exhibits a high degree of zonal symmetry, although the Arctic anomaly center is shifted to the Atlantic–North American sector and the midlatitude anomalies are split into four centers. In contrast, the leading EOF for the FMA period includes no significant anomalies over the Arctic (Fig. 6e). Rather, the EOF represents an upper-tropospheric manifestation of the AL–IL seesaw that consists of the PNA pattern and a dipole over the northwestern Atlantic. Actually, the correlation coefficient between the AII and the leading PC of Z250 is only 0.19 in NDJ, but it jumps up to 0.92 in FMA, indicating the dominance of the seesaw in the upper-tropospheric interannual variability in late winter. The overall features of Fig. 6 are in good accordance with those in Figs. 2 and 3. It is again suggested that the formation of the AL–IL seesaw leads to notable structural changes in the leading tropospheric EOF between early and late winter, with significant decline in the annularity of the anomaly pattern.

4. Vertical coupling of the dominant anomaly patterns

In this section, we first investigate the coupling of the leading mode of the interannual variability within the troposphere. For this purpose, we prepared the maps of Z250 anomalies linearly regressed upon the leading PCs of SLP for each of the nine 45-day periods, as shown in Figs. 7a–c for selected periods. These regression maps overall exhibit striking resemblance to the leading EOFs of Z250 in Fig. 2 for the corresponding periods. The corresponding maps of SLP regressed upon the leading PCs of Z250 as shown in Figs. 7d–f also exhibit striking similarity to the leading EOFs of SLP in Fig. 3. These results indicate the tendency of the strong coupling between the leading EOFs of Z250 and SLP throughout the winter. In fact, the simultaneous correlation between their leading PCs tends to be significantly positive exceeding 0.8 throughout the winter except in early February (Fig. 8a). The weak minimum (0.75) in the correlation observed in early February corresponds to the transition from the dominance of the NAO-like pattern to that of the AL–IL seesaw, whose occurrence in the upper troposphere tends to precede the occurrence near the surface by half a month (Figs. 2–4). This slight weakening of the vertical coupling within the troposphere can be detected even in the late-January period, during which the North Pacific anomalies and their downstream extension start developing. In fact, for that period, the significance of the SLP anomalies over the North Pacific is much higher in the map regressed upon the leading PC of Z250 (Fig. 7e) than in the leading EOF of SLP itself (Fig. 3d). Likewise, Z250 anomalies over the North Pacific are more significant in the leading EOF itself (Fig. 2d) than in the map regressed upon the leading PC of SLP (Fig. 7b). The regression patterns for early February are almost the same as those in Figs. 7b and 7e.

Fig. 7.

(a)–(c): As in Fig. 1a, but maps for mean Z250 anomalies for the 45-day periods of (a) late Dec, (b) late Jan, and (c) late Feb, linearly regressed upon the leading PCs of SLP anomalies for the corresponding periods. (d)–(f): As in (a)–(c), but maps for SLP anomalies for those three periods linearly regressed upon the leading PCs of Z250 anomalies for the corresponding periods

Fig. 7.

(a)–(c): As in Fig. 1a, but maps for mean Z250 anomalies for the 45-day periods of (a) late Dec, (b) late Jan, and (c) late Feb, linearly regressed upon the leading PCs of SLP anomalies for the corresponding periods. (d)–(f): As in (a)–(c), but maps for SLP anomalies for those three periods linearly regressed upon the leading PCs of Z250 anomalies for the corresponding periods

Fig. 8.

(a) Simultaneous correlation coefficients among the leading PCs of the mean SLP, Z250, and Z50 anomalies for each of the nine 45-day periods from early Dec to early Apr. The 95% confidence level with 11 degrees of freedom is indicated for a conservative measure of the statistical significance. (b) As in (a), but for 45-day mean “polar cap anomalies” of SLP, Z250, and Z50 defined as their averages over the area to the north of 80°N. (c) Standard deviations of the 45-day mean polar cap anomalies of SLP, Z250, and Z50 for the nine periods. Standard deviation of a given variable for a particular period is normalized by its largest value (shown in parentheses) among the nine periods (shown in percentage), so as to emphasize the (sub) seasonal dependence of the interannual variability within the Arctic polar cap at each of those vertical levels. All based upon the 22-yr period from 1973 to 1994

Fig. 8.

(a) Simultaneous correlation coefficients among the leading PCs of the mean SLP, Z250, and Z50 anomalies for each of the nine 45-day periods from early Dec to early Apr. The 95% confidence level with 11 degrees of freedom is indicated for a conservative measure of the statistical significance. (b) As in (a), but for 45-day mean “polar cap anomalies” of SLP, Z250, and Z50 defined as their averages over the area to the north of 80°N. (c) Standard deviations of the 45-day mean polar cap anomalies of SLP, Z250, and Z50 for the nine periods. Standard deviation of a given variable for a particular period is normalized by its largest value (shown in parentheses) among the nine periods (shown in percentage), so as to emphasize the (sub) seasonal dependence of the interannual variability within the Arctic polar cap at each of those vertical levels. All based upon the 22-yr period from 1973 to 1994

We extend our investigation into the troposphere–stratosphere coupling, performing an EOF analysis on the monthly Z50 anomalies for the NDJ and FMA periods, separately (Figs. 6c and 6f, respectively). Unlike in the tropospheric EOFs, the leading EOF of Z50 for each of the two periods includes a common characteristic structure,3 which is the well-known seesaw between the Arctic and the surrounding midlatitude belt. This structure with an extremely high degree of zonal symmetry represents the variability in the strength of the stratospheric polar vortex (Perlwitz and Graf 1995; Kitoh et al. 1996; Kodera et al. 1996; Thompson and Wallace 1998). In fact, this particular structure is evident in each of the leading EOF of the Z50 anomaly of the nine overlapping 45-day periods as defined for Figs. 2 and 3 (not shown but for the early-February period in Fig. 9a).

Fig. 9.

(a) As in Fig. 2, but for the first EOF of Z50 anomalies for early Feb (16 Jan–1 Mar; every 30 m; dashed lines for negative values). (b)–(c): As in (a), but for the anomaly maps of (b) Z250 (every 20 m) and (c) SLP (every 1 hPa) linearly regressed upon the leading PC of Z50 anomalies for the same period

Fig. 9.

(a) As in Fig. 2, but for the first EOF of Z50 anomalies for early Feb (16 Jan–1 Mar; every 30 m; dashed lines for negative values). (b)–(c): As in (a), but for the anomaly maps of (b) Z250 (every 20 m) and (c) SLP (every 1 hPa) linearly regressed upon the leading PC of Z50 anomalies for the same period

With the prevalence of that particular pattern in the Z50 anomaly field throughout the winter, the corresponding leading PCs for the nine subseasonal periods can be regarded as an indicator of the polar vortex variability in the lower stratosphere. The simultaneous correlation coefficient between the leading PCs of Z50 and each of Z250 and SLP is also plotted in Fig. 8a as a function of the nine periods. Seasonal dependence is more apparent in the coupling between the leading tropospheric and stratospheric EOFs than in the coupling between the leading tropospheric EOFs. The correlation between the leading PCs of Z50 and SLP rapidly increases since early December until it reaches a peak (0.80) in early January. Then, the correlation coefficient declines to 0.42 by late February as it loses its significance. Similar seasonal dependence is found in the correlation between the leading PCs for Z50 and Z250. It is suggested that the coherence between the leading tropospheric and stratospheric modes tends to be reduced in February and early March, during which the AL–IL seesaw substantially modifies the characteristics of the dominant tropospheric variability. Nevertheless, the seesaw formation exerts no dramatic impact upon the leading mode of the lower-stratospheric variability, since no substantial differences are found between the leading EOFs of Z50 between for the NDJ and FMA periods (Fig. 6). It is hence suggested that the AL–IL seesaw is essentially a tropospheric phenomenon.

The significant correlation between the tropospheric and stratospheric leading PCs in early winter (Fig. 8a) indicates the existence of a prevailing mode of the interannual variability that extends from the surface up to the lower stratosphere representing the anomalous polar vortex intensity. With this strong coupling among the leading EOFs of Z50, Z250, and SLP, the patterns in Figs. 6a–c approximately represent its three dimensional structure. In good agreement with previous studies (e.g., Baldwin et al. 1994; Perlwitz and Graf 1995; Kitoh et al. 1996; Kodera et al. 1996; Deser 2000), the anomalous stratospheric polar vortex tends to be coupled with an NAO-like polar–midlatitude dipole in the troposphere with no significant anomalies over the North Pacific. In fact, the leading PC of Z50 anomalies for the November–April period exhibits a higher correlation with the NAO index (0.61) than with the PNA index (0.34), both of which are defined for the DJF period. The robustness of this particular structure of the troposphere–stratosphere coupling throughout the winter can be confirmed in the maps of Z250 and SLP that have been regressed upon the leading PC of Z50 for each of the nine 45-day periods. The vertically coupled structure is present even in the early-February period (Fig. 9), during which the formation of the AL–IL seesaw tends to mask the tropospheric component of that structure in the leading EOFs. The robustness of the deep polar-vortex anomalies can be further confirmed by examining the correlation among the anomalies of SLP, Z250, and Z50 averaged over the Arctic region to the north of 80°N. As shown in Fig. 8b, the simultaneous correlation between any pair of the three anomalies within the polar cap is significantly high in any of the 45-day periods since late December. The vertical coupling within the polar cap remains strong in the later half of the winter, despite the substantial weakening of the tropospheric interannual fluctuations in the polar cap (Fig. 8c). This vertically coupled structure associated with the anomalous polar vortex may reflect the characteristics of the annular mode indicated by Thompson and Wallace (1998, 2000) and Deser (2000). Our analysis indicates that the structure is unambiguously identified in the leading tropospheric EOF only in early winter, whereas in late winter, its tropospheric component is masked by the AL–IL seesaw.

5. Dominance of the AL–IL seesaw in the tropospheric variability over the wintertime Northern Hemisphere

Thus far we have shown the apparent structural changes in the leading tropospheric EOF between the early- and late-winter seasons associated with the late-winter formation of the AL–IL seesaw. In this section, we examine how significantly the AL–IL seesaw contributes to the dominant interannual variability in the extratropical Northern Hemisphere circulation defined for the entire winter season. We also examine whether the AO-like pattern with a high degree of annularity (or called the annular mode in some literature) is really identified in the leading interannual EOF for the entire winter at any tropospheric level. For this purpose, we first attempted to remove the signal of the seesaw from atmospheric fields based on a linear lag regression method, as in Part I. We first computed the linear regression coefficient between the AII defined for the peak period over the 22-yr period and a monthly anomaly of a particular variable at each grid point during the cold season from November to April. The signal of the AL–IL seesaw for a given month of a particular winter was defined as the product of the local regression coefficient for the month and an AII value for that winter. The seesaw-related signal thus defined was then subtracted from the observed anomaly field for a particular year and month, to form an anomaly field for each winter that is statistically independent of the signal of the AL–IL seesaw (AIS). This residual field is hereafter called the “AIS-removed” field. The dominance of the AL–IL seesaw is suggested, if the structure of the leading EOF of the AIS-removed field differs substantially from that of the original field. Likewise, local anomalies that are significant in the leading EOF for the original field but are missing or lose the significance in the EOF for the AIS-removed field are likely to be associated with the seesaw. In the following, the leading EOF was obtained for the 6-month winter from November to April over the domain poleward of 20°N.

The leading EOF for the AIS-removed field of SLP, which accounts for 20.8% of the total variance within the domain, is characterized by an NAO-like dipole between the polar and midlatitude regions (Fig. 10a). The pattern exhibits certain similarity to the leading EOF of the original SLP field (Fig. 10b), which is the definition of the AO by Thompson and Wallace (1998).4 Yet, the North Pacific anomalies that are barely significant in the original EOF almost disappear and lose the significance in the AIS-removed EOF. In this EOF, the significance is also lost of the midlatitude Atlantic anomalies just off the North American coast. These two regions correspond to the surface anomaly centers associated with the PNA pattern, which is an important element of the AL–IL seesaw (Part I). In fact, the correlation of the DJF-mean PNA index, based on the definition of Wallace and Gutzler (1981),5 weakens markedly from 0.58 with the PC for the original EOF to 0.15 with the PC for the AIS-removed EOF. Though may not be of particular significance, the signal near Iceland somewhat weakens after the removal of the seesaw signal. In summary, the AIS-removed EOF consists only of a polar–midlatitude dipole over the Euro–Atlantic sector with a strong projection upon the NAO and a high degree of annularity as well. In fact, the correlation of the DJF-mean NAO index based on Hurrell's (1995) definition increases slightly from 0.65 with the leading PC for the original anomalies to 0.71 with the leading PC of the AIS-removed anomalies.

Fig. 10.

As in Fig. 2, but for the maps of wintertime (Nov–Apr) monthly SLP anomalies (every 1 hPa) linearly regressed upon the leading PCs of the following SLP anomalies; (a) anomalies without AL–IL seesaw (AIS removed), (b) the original anomalies, and (c) anomalies without the variability associated with the leading EOF for the NDJ period (EOF1-NDJ removed). Signals of the AL–IL seesaw and EOF1-NDJ have been removed through a linear regression technique. Details are described in the text

Fig. 10.

As in Fig. 2, but for the maps of wintertime (Nov–Apr) monthly SLP anomalies (every 1 hPa) linearly regressed upon the leading PCs of the following SLP anomalies; (a) anomalies without AL–IL seesaw (AIS removed), (b) the original anomalies, and (c) anomalies without the variability associated with the leading EOF for the NDJ period (EOF1-NDJ removed). Signals of the AL–IL seesaw and EOF1-NDJ have been removed through a linear regression technique. Details are described in the text

The same EOF analysis as above but applied to the AIS-removed and original Z250 anomalies separately yields more pronounced differences between their leading modes. A striking fact is that the leading EOF pattern of the original Z250 is characterized by a combined signature of the PNA pattern and a dipole over the northwestern Atlantic (Fig. 11b). The pattern indeed represents an upper-tropospheric manifestation of the AL–IL seesaw (cf. Fig. 1a). In fact, the correlation coefficient between the corresponding PC and AII is 0.77, which is indeed higher than the counterpart (0.58) between the AII and the leading PC for the original SLP (or the AO index of Thompson and Wallace). The leading EOF of the AIS-removed field, in contrast, is characterized by an upper-tropospheric manifestation of the NAO-like polar–midlatitude dipole with no significant anomalies over the North Pacific (Fig. 11a). The dipole over the North Atlantic in Fig. 11b differs somewhat from the upper-tropospheric manifestation of the NAO-like dipole shown in Fig. 11a 6 in the sense that the node of the former is located slightly to south of the node of the latter. Furthermore, the southern component of the upper-level NAO-like dipole is much more elongated in the zonal direction than the counterpart associated with the AL–IL seesaw. Consistent with this argument, the correlation coefficients of the corresponding PCs for the EOFs of the original and AIS-removed Z250 anomalies are 0.45 and 0.72, respectively, with the DJF-mean NAO index, and 0.84 and 0.32, respectively, with the DJF-mean PNA index.

Fig. 11.

As in Fig. 10, but for Z250 anomalies (contour intervals: 20 m)

Fig. 11.

As in Fig. 10, but for Z250 anomalies (contour intervals: 20 m)

The NAO-like annular mode dominant in the AIS-removed anomalies over the entire winter season7 (Figs. 10a and 11a) appears to be related to the deep anomaly structure associated with the anomalous polar vortex as identified in the leading EOFs of the tropospheric and stratospheric anomalies for early winter (Figs. 7a–c). Actually, the correlation coefficient is as high as 0.90 between the leading PC of the wintertime (November–April) AIS-removed SLP and that of the original SLP for the NDJ period. The correlation is also quite high (0.89) between the corresponding PCs for the two kinds of the Z250 anomalies. Furthermore, as shown in Fig. 12, the leading PC for the AIS-removed SLP field is highly correlated with the leading PC of each of SLP and Z250 over three of the overlapping 45-day periods in early winter. The correlation suddenly drops in February, when the AL–IL seesaw becomes dominant. These results again suggest that the seesaw may be independent of the NAO-like annular mode, since the latter can be extracted as the leading mode of the variability for the entire winter season once the signal of the former has been removed from the data.

Fig. 12.

As in Fig. 4, but for (a) correlation coefficients of the leading PC of Z250 anomalies with the following time indices as indicated: the so-called AO indices based on the original and AIS-removed SLP anomalies, and the AII defined for the peak period (late Feb). (b) As in (a), but for the leading PC of SLP anomalies. The two AO indices are labeled as AOI and AOI (rmAIS), respectively. They are defined as the wintertime (Nov–Apr) mean values of the leading PCs for the original and AIS-removed SLP anomalies, respectively

Fig. 12.

As in Fig. 4, but for (a) correlation coefficients of the leading PC of Z250 anomalies with the following time indices as indicated: the so-called AO indices based on the original and AIS-removed SLP anomalies, and the AII defined for the peak period (late Feb). (b) As in (a), but for the leading PC of SLP anomalies. The two AO indices are labeled as AOI and AOI (rmAIS), respectively. They are defined as the wintertime (Nov–Apr) mean values of the leading PCs for the original and AIS-removed SLP anomalies, respectively

Last, we try to confirm that the AL–IL seesaw can be identified as the leading mode of the tropospheric variability for the entire winter season once the signal of the NAO-like annular dipole has been removed from the data in advance. For the removal, we applied the same linear regression technique as we used to remove the AL–IL signal, but this time the PC time series for the leading EOF of SLP or Z250 for the NDJ period was used in place of the AII. After removing the signal of the NAO-like dipole based on this particular PC from monthly anomaly fields, we applied an EOF analysis to the residual anomaly fields of SLP and Z250 separately. In the leading EOF of the residual SLP anomalies (Fig. 10c), the AL and IL anomalies both become much more apparent and significant than in the original EOF (Fig. 10b). Moreover, the removal of the signal of NAO-like polar–midlatitude dipole caused the shrink of the Arctic anomalies in the leading EOF, particularly along the Siberian and Alaskan coasts. The leading EOF of the residual SLP anomalies thus exhibits striking resemblance to the AL–IL seesaw as shown in Fig. 1b. The corresponding PC is indeed highly correlated with the AII. The correlation (0.85) is substantially higher than that (0.58) between the AII and AO index. The leading EOF of the residual Z250 anomalies shown in Fig. 11c also represents the upper-level manifestation of the AL–IL seesaw with no significant anomalies over the Arctic. The corresponding PC is also highly correlated (0.84) with the AII. Since the signal of the AL–IL seesaw dominates over the signal of the NAO-like polar–midlatitude dipole in the upper troposphere, the removal of the latter signal from the data left no substantial impact on the structure of the leading EOF of Z250.

It is becoming apparent through our analysis that what Thompson and Wallace (1998) called the (surface) AO (Fig. 10b) is more or less a mixture of the two dominant anomaly patterns in the wintertime troposphere over the extratropical Northern Hemisphere; that is, the NAO-like polar–midlatitude (annular) dipole (Fig. 10a) and AL–IL seesaw (Fig. 10c), although the contribution from the former appears to dominate. Accordingly, the correlation coefficient between the AO index based on their definition and each of the leading PCs of Z250 and SLP peaks in late January or in early February (Fig. 12), during which the leading EOF structure is changing from the NAO-like annular dipole to the AL–IL seesaw. The correlation remains significant almost through the entire winter season. Because of these combined contributions from the NAO-like annular dipole and the AL–IL seesaw to the first EOF of SLP over the entire winter season, these two dominant modes of the tropospheric interannual variability cannot be isolated from one the other through conventional EOF analysis for the Northern Hemisphere. Cheng and Dunkerton (1995) showed that a rotated EOF analysis tends to force such a hemispheric pattern as the AL–IL seesaw to break up into more localized patterns over the two ocean basins. This is why we needed to apply the linear regression method as mentioned above.

6. Zonal mean structure of the AL–IL seesaw

Very recently, DeWeaver and Nigam (2000a) pointed out that the meridional profiles of the anomalies in the zonally averaged zonal wind ([U]) accompanied by the NAO and PNA pattern are mutually very similar. One may thus imagine that the AL–IL seesaw, which includes contributions from those two anomaly patterns, may exhibit a strong projection on the meridional plane. Figure 13a shows that is indeed the case. Near the surface, the northern flank of the anomalous AL and southern flank of the anomalous IL are both located at ∼55°N (Fig. 1b). Hence, the AL–IL seesaw acts to vary the near-surface zonal wind around that latitude coherently between the two ocean basins. Furthermore, since the Azores high and AL tend to vary in phase associated with the seesaw, zonal wind anomalies also tend to be in phase over the two ocean basins along their southern flanks at ∼35°N. Likewise, reflecting the particular geographical alignments of the PNA pattern and meridional dipole over the northwestern Atlantic, the upper-tropospheric zonal wind also varies coherently over the two ocean basins in association with the AL–IL seesaw (Ambaum et al. 2001). Thus, the AL–IL seesaw imposes a significant projection on the meridional plane, so as to vary the upper- and lower-tropospheric [U] in seesaw between ∼55° and ∼35°N (Fig. 13a). In fact, the 250-hPa [U] anomalies at 55° and 35°N in late February (i.e., the peak period of the AL–IL seesaw) are both highly correlated (0.81 and −0.82, respectively) with the AII.

Fig. 13.

(a) Typical meridional profiles of zonally averaged zonal wind [U] anomalies (m s−1) at the 1000-hPa (open circles) and 250-hPa (closed circles) levels associated with the AL–IL seesaw. Based on the maps of SLP and Z250 linearly regressed on the AII as shown in Fig. 1. (b) As in (a), but for the [U] anomalies associated with the leading EOFs of the wintertime SLP and Z250 from which the seesaw signal has been statistically removed so as to extract the variability associated with the Arctic–midlatitude dipole. Based on their linear regression maps shown in Figs. 10a and 11a, respectively. (c) As in (a), but for the [U] anomalies associated with the leading EOFs of the wintertime SLP and Z250 anomalies as observed. Based on their linear regression maps shown in Figs. 10b and 11b, respectively. The former correspond to the so-called AO, while the latter primarily represents the upper-tropospheric signal of the AL–IL seesaw. Scaling for anomalous [U] is arbitrary

Fig. 13.

(a) Typical meridional profiles of zonally averaged zonal wind [U] anomalies (m s−1) at the 1000-hPa (open circles) and 250-hPa (closed circles) levels associated with the AL–IL seesaw. Based on the maps of SLP and Z250 linearly regressed on the AII as shown in Fig. 1. (b) As in (a), but for the [U] anomalies associated with the leading EOFs of the wintertime SLP and Z250 from which the seesaw signal has been statistically removed so as to extract the variability associated with the Arctic–midlatitude dipole. Based on their linear regression maps shown in Figs. 10a and 11a, respectively. (c) As in (a), but for the [U] anomalies associated with the leading EOFs of the wintertime SLP and Z250 anomalies as observed. Based on their linear regression maps shown in Figs. 10b and 11b, respectively. The former correspond to the so-called AO, while the latter primarily represents the upper-tropospheric signal of the AL–IL seesaw. Scaling for anomalous [U] is arbitrary

The profile of the anomalous [U] associated with the AL–IL seesaw bears striking resemblance to the counterpart of the NAO-like polar–midlatitude dipole, despite the profound difference in annularity between the two patterns. The latter is shown in Fig. 13b, based on the first EOFs of wintertime (November–April), SLP, and Z250 anomalies from which the AL–IL seesaw signature has been removed statistically (Figs. 10a and 11a). The only noticeable difference is that the entire profile for the AL–IL seesaw is shifted slightly southward relative to that for the polar–midlatitude dipole. Yet, this difference is so slight that nearly the same profile of anomalous [U] emerges when signals of the two patterns are merged together as in the first EOFs of the observed wintertime (November–April), SLP, and Z250 anomalies shown in Figs. 10b and 11b, respectively. The corresponding [U] profiles are plotted in Fig. 13c. It is suggested from our analysis in the preceding section that the 250-hPa [U] anomalies in Fig. 13c manifest the prime contribution from the AL–IL seesaw. The AII indeed exhibits a significant correlation (0.63) with the anomalous 250-hPa [U] at 55°N averaged over the entire winter season (November–April). In the 1000-hPa [U] anomalies in Fig. 13c, a contribution from the polar–midlatitude dipole must be larger. The anomalous [U] in Fig. 13c exhibits a marked resemblance to its counterpart for the AO shown by Thompson and Wallace (2000). Moreover, the SLP anomalies for the DJF period in extreme events of the [U] seesaw between 35° and 55°N shown by Ting et al. (1996) include not only a signature of what may be called the annular mode or AO but also the AL–IL seesaw. The latter signal appears to dominate over the former in the corresponding anomalies at the 500-hPa level.

As pointed out by Wallace (2000), the anomalous [U] profiles presented in Fig. 13c are not in direct correspondence to the traditional measure of the so-called zonal index (Rossby 1939; Namias 1947, 1950; Willett 1948). Nor are the profiles in Figs. 13a and 13b. All the profiles in Fig. 13 are nearly orthogonal to the traditional index defined as the anomalous [U] averaged between 35° and 55°N. Lorenz (1951) found that the anomalous [U] at 55°N best describes the dominant variability in the zonally symmetric circulation. Consistent with Wallace (2000), Lorenz's index well represents the [U] anomalies associated with the NAO-like polar–midlatitude dipole (Fig. 13b). Figure 13 suggests that his index also includes a significant contribution from the interannual seesaw between the IL and AL, the two prominent low pressure systems in the Northern Hemisphere. At the same time, Fig. 13 also suggests that zonal averaging of the wintertime tropospheric anomalies does not necessarily act to extract the contribution from the NAO-like polar–midlatitude dipole or annular mode, because of the significant contribution from the AL–IL seesaw, especially in late winter. Therefore, any dynamical interpretation of the former pattern (or AO) based on zonal averaging of the observed anomalies requires some caution if the analysis period includes late winter.

Our statistics presented in Part I indicate that, in the course of the AL–IL seesaw formation, the development of the North Atlantic anomalies tends to lag the development of the NP anomalies by about a month. It is implied that interpreting the AL–IL seesaw formation in such a simple framework as a planetary wave model of Charney and Eliassen (1949) may not be appropriate, where the resonant response of the planetary wave components to [U] is assumed. The lack of any significant anomalies over Siberia associated with the AL–IL seesaw is consistent with the above implication. The concept of a Green's function, as discussed in Held (1983), based on the zonal dispersion of a stationary Rossby wave packet emanating from a localized wave source appears to be more appropriate. One may infer that the PNA pattern and NAO-like dipole over the northwestern Atlantic that appear in the course of the seesaw formation constitute the planetary wave response to anomalous [U] associated with the seesaw. Yet, as shown by DeWeaver and Nigam (2000a), the NAO generates a barotropic feedback on [U] mainly through the interactions between the climatological-mean and anomalous wave components, whereas the PNA pattern does so primarily through the self-interaction among the anomalous wave components. Therefore, how these two anomaly patterns interact with [U] differs fundamentally from one another. It is thus further implied that interpreting the AL–IL seesaw formation based on the separation between the zonally symmetric flow and eddies may not necessarily be appropriate. Our dynamical analysis presented in Part I was based on the separation between the time-mean flow and anomalies, where the two anomaly patterns are assumed to “feel” the regional time-mean flows as their basic states. As shown in Part I, the anomalous Pacific and Atlantic storm tracks impose continual feedback forcing on the quasi-stationary AL and IL anomalies throughout the course of the AL–IL seesaw formation, which must also be taken into account in the interpretation.

7. Summary and discussion

It became apparent through our analysis of the operational analyses for 22 recent years that the formation of the AL–IL seesaw in February exerts a substantial impact on the structure of the leading EOFs of the SLP and Z250 anomalies over the wintertime extratropical Northern Hemisphere. At these levels, the leading mode in the first half of the cold season (November–January) is characterized by a distinct meridional dipole between the Arctic polar cap and midlatitude Atlantic with a large contribution from the NAO (Figs. 6a and 6b). This pattern that exhibits a high degree of annularity may be viewed as the tropospheric manifestation of the so-called Northern Hemisphere annular mode, which is strongly coupled with the anomalous polar vortex in the stratosphere (Thompson and Wallace 2000; Wallace 2000). In contrast, the upper-tropospheric leading mode in the second half of the cold season (February–April) is characterized by a combination of the PNA pattern and a meridional dipole over the northwestern Atlantic as the upper-level manifestation of the AL–IL seesaw (Fig. 6e). Unlike the annular mode, each of these patterns consists of a regional meridional dipole and/or a stationary wave train downstream of it (cf. Nakamura et al. 1987). After the formation of the AL–IL seesaw, therefore, the leading mode of tropospheric variability becomes less annular, especially in the upper troposphere. Near the surface, the AL–IL seesaw formation is manifested in the leading interannual EOF as an emergence of strong anomalies over the North Pacific and a strengthening of the meridional seesaw over the western half of the North Atlantic (Fig. 6d). The decline of the annularity in this leading EOF of the late-winter SLP anomalies is associated with a shift of the primary center of action from the Arctic Sea to the region near Iceland (Fig. 3). This decline near the surface is, however, substantially less pronounced than at the upper levels. Stationary anomalies associated with midlatitude teleconnection patterns including the PNA pattern whose wave activity propagates dominantly eastward are regarded generally as an external mode and exhibit equivalent barotropic structure with maximum amplitude at the tropopause level (e.g., Held 1983). Hence, the effect of masking the deep NAO-like dipole by the AL–IL seesaw should be greatest at that level and much weaker near the surface. This tendency may account for the slight delay in the structural change in the leading EOF from the zonally symmetric pattern to the AL–IL seesaw near the surface relative to the corresponding change at the upper levels (e.g., Figs. 2 and 3). That tendency may also account for the structural distinction between the leading EOFs for the upper and lower troposphere defined for the entire winter season (Fig. 10b vs Fig. 11b). We found that reflecting the particular geographical alignment of the IL and AL, their seesaw changes the zonal wind in phase between the North Atlantic and Pacific. The significant zonally symmetric component in the anomalous zonal wind thus yielded is distributed with a latitudinal profile that is almost indistinguishable from the counterpart of the NAO-like Arctic–midlatitude dipole. Presumably, these two dominant modes of the interannual tropospheric circulation over the wintertime Northern Hemisphere are of comparable importance in the zonal index cycle, but the separation of their contributions may not necessarily be straightforward.

An implication of our result is that the leading EOF of monthly extratropical SLP anomalies over the entire cold season, that is, the definition of the AO by Thompson and Wallace (1998), does not necessarily represent the deep polar-vortex anomalies (the annular mode in their terminology) in the purest form. This is because the EOF is subject to some contamination by the AL–IL seesaw signal. Thompson and Wallace (2000) argued that enhanced zonal asymmetries of the AO signature in the mid- to upper troposphere may be a manifestation of a tropospherically confined baroclinic structure that forms through anomalous zonal advection of the climatological-mean land–sea temperature contrasts. An analysis by DeWeaver and Nigam (2000b) suggests that the effect of anomalous diabatic heating associated with the meridionally displaced Atlantic storm track may also contribute substantially to the zonal asymmetries in the upper-tropospheric AO signature. Based on our analysis, we argue that the dominance of the PNA pattern and meridional dipole over the northwestern Atlantic associated with the AL–IL seesaw may also be an important contributor to those zonal asymmetries whenever the analysis period includes the late-winter season.

Our analysis indicates that a significant part of the late-winter variability over the North Atlantic is generated as a remote influence of the North Pacific variability accompanied by the anomalous AL. In fact, in late February the covariability accounts for 30%∼50% of the interannual variance in the upper and lower troposphere over the northern part of the North Atlantic (Fig. 14). It was shown in Part I that the development of the IL anomalies tends to lag the development of the anomalous AL by a month or so in the course of the formation of their seesaw. It is hence implied that the AL–IL seesaw might provide a basis of potential predictability in the late-winter mean weather condition over the Euro–Atlantic sector. Interestingly, the correlation coefficient between the station-based NAO index defined by Hurrell (1995) and the leading PC for the wintertime hemispheric SLP anomalies increases from 0.51 in the NDJ period to 0.71 in the FMA period (Table 1). An increase is also observed in the correlation between the NAO index and the leading PC for the North Atlantic SLP anomalies; it is enhanced in the 3-month means. It is conjectured that the “localized” North Atlantic anomalies associated with the AL–IL seesaw render the correlation even higher in late winter. In fact, when regressed upon the NAO index for late February, the sequences of the lag regression maps of the SLP and Z250 anomalies both exhibit a strong similarity to the sequence of the seesaw formation, showing no apparent linkage with the Arctic anomalies (not shown).

Fig. 14.

(a) Map of the simultaneous correlation coefficient between local SLP anomaly and the anomalous AL intensity for the 45-day period from 31 Jan to 16 Mar. (b) As in (a) but for the mean Z250 anomaly. Light and dark shading represent correlations stronger than 0.7 and 0.5, respectively; i.e., more than 49% and 25% of the interannual variance, respectively, is explained by the remote influence from the North Pacific

Fig. 14.

(a) Map of the simultaneous correlation coefficient between local SLP anomaly and the anomalous AL intensity for the 45-day period from 31 Jan to 16 Mar. (b) As in (a) but for the mean Z250 anomaly. Light and dark shading represent correlations stronger than 0.7 and 0.5, respectively; i.e., more than 49% and 25% of the interannual variance, respectively, is explained by the remote influence from the North Pacific

Table 1.

Simultaneous correlation coefficients between the monthly station-based NAOa index and first PCs of the monthly SLP anomalies over the extratropical Northern Hemisphereb and only over the North Atlanticc for each of the NDJ and FMA periods for 1973–94. Values in parentheses denote the correlation coefficients based on the 3-month mean anomalies

Simultaneous correlation coefficients between the monthly station-based NAOa index and first PCs of the monthly SLP anomalies over the extratropical Northern Hemisphereb and only over the North Atlanticc for each of the NDJ and FMA periods for 1973–94. Values in parentheses denote the correlation coefficients based on the 3-month mean anomalies
Simultaneous correlation coefficients between the monthly station-based NAOa index and first PCs of the monthly SLP anomalies over the extratropical Northern Hemisphereb and only over the North Atlanticc for each of the NDJ and FMA periods for 1973–94. Values in parentheses denote the correlation coefficients based on the 3-month mean anomalies

Our present analysis has demonstrated the particular significance of the AL–IL seesaw in the tropospheric interannual variability over the extratropical Northern Hemisphere. The seesaw indeed becomes the dominant mode of the upper-tropospheric variability in late winter. The AL–IL seesaw is a distinct circulation anomaly pattern that accompanies a strong interbasin link in the zonal direction between the North Pacific and Atlantic. This particular linkage is in contrast, at least in its appearance, to the other dominant mode of the interannual variability that represents the meridional linkage between the Arctic and North Atlantic regions (i.e., the annular mode in some literature). It appears as if the AL–IL seesaw were competing in late winter with the other pattern for the position of the leading mode of the tropospheric variability within the extratropical Northern Hemisphere.

Acknowledgments

We would like to express our special thanks to Dr. J. M. Wallace for his careful reading of the earlier version of this paper and giving us a number of comments that led to the substantial improvement of this paper. Particularly, discussion with him was very helpful in interpreting the relationship between the AL–IL seesaw and the Arctic oscillation (or annular mode). We also thank the two anonymous referees whose sound criticism and suggestions also contributed to the improvement. Our thanks are extended to Drs. J. Ukita, K. Kodera, Y. Tanimoto, S. Yamane, M. Watanabe, and U. Bhatt for their useful comments and encouragement. The Grid Analysis and Display System (GrADS) was used for drawing the figures.

REFERENCES

REFERENCES
Ambaum
,
M. H. P.
,
B.
Hoskins
, and
D. B.
Stephenson
,
2001
:
Arctic oscillation or North Atlantic oscillation?
J. Climate
,
14
,
3495
3507
.
Angell
,
J. K.
, and
J.
Korshover
,
1974
:
Quasi-biennial and long-term fluctuations in the centers of action.
Mon. Wea. Rev
,
102
,
669
678
.
Baldwin
,
M. P.
, and
T. J.
Dunkerton
,
1999
:
Propagation of the Arctic oscillation from the stratosphere to the troposphere.
J. Geophys. Res
,
104
,
30937
30946
.
Baldwin
,
M. P.
,
X.
Cheng
, and
T. J.
Dunkerton
,
1994
:
Observed correlations between winter-mean tropospheric and stratospheric circulation anomalies.
Geophys. Res. Lett
,
21
,
1141
1144
.
Bjerknes
,
J.
,
1966
:
A possible response of the atmospheric Hadley circulation to equatorial anomalies of ocean temperature.
Tellus
,
18
,
820
828
.
Brooks
,
C. E. P.
, and
W. A.
Quennell
,
1926
:
Classification of monthly charts of pressure anomaly over the Northern Hemisphere.
Geophys. Mem., 31, 12 pp. plus 36 figures
.
Charney
,
J. G.
, and
A.
Eliassen
,
1949
:
A numerical method for predicting the perturbations of the middle latitude westerlies.
Tellus
,
1
,
38
54
.
Cheng
,
X.
, and
J. M.
Wallace
,
1993
:
Cluster analysis of the Northern Hemisphere wintertime 500-hPa height field: Spatial patterns.
J. Atmos. Sci
,
50
,
2674
2696
.
Cheng
,
X.
, and
T. J.
Dunkerton
,
1995
:
Orthogonal rotation of spatial patterns derived from singular value decomposition analysis.
J. Climate
,
8
,
2631
2643
.
Deser
,
C.
,
2000
:
On the teleconnectivity of the “Arctic oscillation.”.
Geophys. Res. Lett
,
27
,
779
782
.
Deser
,
C.
,
J. E.
Walsh
, and
M. S.
Timlin
,
2000
:
Arctic sea ice variability in the context of recent wintertime atmospheric circulation trends.
J. Climate
,
13
,
617
633
.
DeWeaver
,
E.
, and
S.
Nigam
,
2000a
:
Do stationary waves drive the zonal-mean jet anomalies of the Northern Hemisphere?
J. Climate
,
13
,
2160
2176
.
DeWeaver
,
E.
, and
S.
Nigam
,
2000b
:
Zonal-eddy dynamics of the North Atlantic oscillation.
J. Climate
,
13
,
3893
3914
.
Dickson
,
R. R.
, and
J.
Namias
,
1976
:
North American influences on the circulation and climate of the North Atlantic sector.
Mon. Wea. Rev
,
104
,
1255
1265
.
Held
,
I. M.
,
1983
:
Stationary and quasi-stationary eddies in the extratropical troposphere: Theory.
Large-Scale Dynamical Processes in the Atmosphere, B. J. Hoskins and R. P. Pearce, Eds., Academic Press, 127–168
.
Honda
,
M.
,
H.
Nakamura
,
J.
Ukita
,
I.
Kousaka
, and
K.
Takeuchi
,
2001
:
Interannual seesaw between the Aleutian and Icelandic lows. Part I: Seasonal dependence and life cycle.
J. Climate
,
14
,
1029
1041
.
Hurrell
,
J. W.
,
1995
:
Decadal trends in the North Atlantic oscillation: Regional temperatures and precipitation.
Science
,
269
,
677
679
.
Kalnay
,
M. E.
, and
and Coauthors
,
1996
:
The NCEP/NCAR 40-Year Reanalysis Project.
Bull. Amer. Meteor. Soc
,
77
,
437
471
.
Kitoh
,
A.
,
H.
Koide
,
K.
Kodera
,
S.
Yukimoto
, and
A.
Noda
,
1996
:
Interannual variability in the winter stratospheric-tropospheric circulation in a coupled ocean–atmosphere GCM.
Geophys. Res. Lett
,
23
,
543
546
.
Kodera
,
K.
,
M.
Chiba
,
H.
Koide
,
A.
Kitoh
, and
Y.
Nikaidou
,
1996
:
Interannual variability of the winter stratosphere and troposphere in the Northern Hemisphere.
J. Meteor. Soc. Japan
,
74
,
365
382
.
Koide
,
H.
, and
K.
Kodera
,
1999
:
A SVD analysis between the winter NH 500-hPa height and surface temperature fields.
J. Meteor. Soc. Japan
,
77
,
47
61
.
Kuroda
,
Y.
, and
K.
Kodera
,
2001
:
Interannual variability of the Polar-Night Jets in the Northern and Southern Hemispheres.
J. Geophys. Res., in press
.
Kushnir
,
Y.
, and
J. M.
Wallace
,
1989
:
Low frequency variability in the Northern Hemisphere winter: Geographical distribution, structure and time-scale dependence.
J. Atmos. Sci
,
46
,
3122
3142
.
Kutzbach
,
J. E.
,
1970
:
Large-scale features of monthly mean Northern Hemisphere anomaly maps of sea-level pressure.
Mon. Wea. Rev
,
98
,
708
716
.
Lau
,
N-C.
, and
M. J.
Nath
,
2001
:
Impact of ENSO on SST variability in the North Pacific and North Atlantic: Seasonal dependence and role of extratropical sea–air coupling.
J. Climate
,
14
,
2846
2866
.
Lorenz
,
E. N.
,
1951
:
Seasonal and irregular variations of the Northern Hemisphere sea-level pressure profile.
J. Meteor
,
8
,
52
59
.
Nakamura
,
H.
,
M.
Tanaka
, and
J. M.
Wallace
,
1987
:
Horizontal structure and energetics of the Northern Hemisphere wintertime teleconnection patterns.
J. Atmos. Sci
,
44
,
3377
3391
.
Namias
,
J.
,
1947
:
Physical nature of some fluctuations in the speed of the zonal circulation.
J. Meteor
,
4
,
125
133
.
Namias
,
J.
,
1950
:
The index cycle and its role in the general circulation.
J. Meteor
,
7
,
130
139
.
Overland
,
J. E.
,
J. M.
Miletta
, and
N. A.
Bond
,
1999
:
Decadal variability of the Aleutian low and its relation to high-latitude circulation.
J. Climate
,
12
,
1542
1548
.
Perlwitz
,
J.
, and
H-F.
Graf
,
1995
:
The statistical connection between tropospheric and stratospheric circulation of the Northern Hemisphere in winter.
J. Climate
,
8
,
2281
2295
.
Rogers
,
J. C.
, and
H.
van Loon
,
1979
:
The seesaw in winter temperatures between Greenland and northern Europe. Part II: Some oceanic and atmospheric effects in middle and high latitudes.
Mon. Wea. Rev
,
107
,
509
519
.
Rossby
,
C-G.
,
1939
:
Relation between variations in the intensity of zonal circulation and displacements of the semi-permanent centers of action.
J. Mar. Res
,
2
,
38
55
.
Takaya
,
K.
, and
H.
Nakamura
,
1997
:
A formulation of a wave-activity flux for stationary Rossby waves on a zonally varying basic flow.
Geophys. Res. Lett
,
24
,
2985
2988
.
Takaya
,
K.
, and
H.
Nakamura
,
2001
:
A formulation of a phase-independent wave-activity flux for stationary and migratory quasigeostrophic eddies on a zonally varying basic flow.
J. Atmos. Sci
,
58
,
608
627
.
Thompson
,
D. W. J.
, and
J. M.
Wallace
,
1998
:
The Arctic oscillation signature in the wintertime geopotential height and temperature fields.
Geophys. Res. Lett
,
25
,
1297
1300
.
Thompson
,
D. W. J.
, and
J. M.
Wallace
,
2000
:
Annular modes in the extratropical circulation. Part I: Month-to-month variability.
J. Climate
,
13
,
1000
1016
.
Thompson
,
D. W. J.
,
J. M.
Wallace
, and
G. C.
Hegerl
,
2000
:
Annular modes in the extratropical circulation. Part II: Trends.
J. Climate
,
13
,
1018
1036
.
Ting
,
M.
,
M. P.
Hoerling
,
T.
Xu
, and
A.
Kumar
,
1996
:
Northern Hemisphere teleconnection patterns during extreme phases of the zonal-mean circulation.
J. Climate
,
9
,
2614
2633
.
Trenberth
,
K. E.
, and
D. A.
Paolino
,
1981
:
Characteristic patterns of variability of sea level pressure in the Northern Hemisphere.
Mon. Wea. Rev
,
109
,
1169
1189
.
van Loon
,
H.
, and
J. C.
Rogers
,
1978
:
The seesaw in winter temperatures between Greenland and northern Europe. Part I: General description.
Mon. Wea. Rev
,
106
,
296
310
.
van Loon
,
H.
, and
R. A.
Madden
,
1983
:
Interannual variations of mean monthly sea-level pressure in January.
J. Climate Appl. Meteor
,
22
,
687
692
.
Wallace
,
J. M.
,
2000
:
North Atlantic oscillation/annular mode: Two paradigms—one phenomenon.
Quart. J. Roy. Meteor. Soc
,
126
,
791
805
.
Wallace
,
J. M.
, and
D. S.
Gutzler
,
1981
:
Teleconnections in the geopotential height field during the Northern Hemisphere winter.
Mon. Wea. Rev
,
109
,
784
812
.
Wallace
,
J. M.
,
Y.
Zhang
, and
K-H.
Lau
,
1993
:
Structure and seasonality of interannual and interdecadal variability of the geopotential height and temperature fields in the Northern Hemisphere.
J. Climate
,
6
,
2063
2082
.
Walsh
,
J. E.
, and
W. M.
Chapman
,
1990
:
Short-term climate variability of the Arctic.
J. Climate
,
3
,
237
250
.
Watanabe
,
M.
, and
T.
Nitta
,
1998
:
Relative impacts of snow and sea surface temperature anomalies on an extreme phase in the winter atmospheric circulation.
J. Climate
,
11
,
2837
2857
.
Watanabe
,
M.
, and
T.
Nitta
,
1999
:
Decadal changes in the atmospheric circulation and associated surface climate variations in the Northern Hemisphere winter.
J. Climate
,
12
,
494
510
.
Willett
,
H. C.
,
1948
:
Patterns of world weather changes.
Trans. Amer. Geophys. Union
,
29
,
803
809
.
Xie
,
S-P.
,
H.
Noguchi
, and
S.
Matsumura
,
1999
:
A hemispheric-scale quasi-decadal oscillation and its signature in northern Japan.
J. Meteor. Soc. Japan
,
77
,
573
582
.

Footnotes

Corresponding author address: Dr. Hisashi Nakamura, Department of Earth and Planetary Science, Graduate School of Science, University of Tokyo, Tokyo, 113-0033 Japan. Email: hisashi@eps.s.u-tokyo.ac.jp

1

As they pointed out, a hint of the AL–IL seesaw is found in two of the SLP anomaly patterns based on the classification by Brooks and Quennel (1926).

2

A. Kaplan (2001, personal communication) recently confirmed the significant negative correlation between the AL and IL intensities based on the ship measured data over the last 140 yr.

3

In spite of the marked resemblance of the leading EOF patterns between the NDJ and FMA periods, the correlation between the corresponding PCs for the two periods is very low (0.22) in the presence of the ∼90-day fluctuations in the intensity of the stratospheric polar vortex (Baldwin and Dunkerton 1999; Kuroda and Kodera 2001).

4

The pattern in Fig. 10b based on monthly SLP over the 22-yr period (1973–94) is almost identical to the AO pattern in the definition of Thompson and Wallace (1998) based on monthly SLP over the 40-yr period (1958–97).

5

We have reversed the sign of the PNA index from their original definition.

6

The dipole over the northwestern Atlantic as a part of the AL–IL seesaw looks similar to the Western Atlantic (WA) pattern defined by Wallace and Gutzler (1981), but the southern center of action of the dipole is somewhat to the west of the counterpart of the WA pattern, and the node of the former is shifted slightly northward.

7

No significant differences are found between the leading EOFs of the original and AIS-removed Z50 anomalies, both of which are strikingly similar to the corresponding EOFs for the NDJ and FMA periods (Figs. 6c and 6f, respectively).