The Pacific-North American (PNA) teleconnection pattern (Wallace and Gutzler, 1981), characterized by a wave train emanating and arching from the tropical Pacific through the North Pacific, and from Canada to North America, is the most prominent low-frequency mode and has been the subject of extensive investigations into understanding the mechanisms that affect its growth and maintenance (Frederiksen, 1982; Simmons et al., 1983; Nigam and Lindzen, 1989; Dole and Black, 1990; Feldstein, 2002). The linkage between the El Niño-Southern Oscillation (ENSO) and mid-latitude stationary wave anomalies is well established, but debate continues about the role of tropical sea surface temperature (SST) forcing in triggering and/or amplifying the internal modes of variability such as the PNA (Trenberth et al., 1998; Straus and Shukla, 2002). Nonetheless, most agree that the ENSO phenomenon does influence PNA conditions in some way (Hoskins and Karoly, 1981; Lau and Nath, 1994; Trenberth et al., 1998; Lau and Nath, 2001; Yu et al., 2009). The effects of the feedback of high-frequency transient eddy fluxes on low-frequency flow are also considered an important factor in PNA activity (Lau, 1988; Hall and Derome, 2000, Jin et al., 2006a, 2006b, Kug et al., 2009a).
Recent research has focused on the interdecadal climate shift that occurred in the late 1970s (Nitta and Yamada, 1989; Wallace et al., 1993; Trenberth and Hurrell, 1994; Jin, 1997; Gedalof and Smith, 2001; Deser et al., 2004;). It is a global-scaled variability and characterized by changes in the intensity of large-scale atmospheric circulations, including SST, air temperature, precipitation, and winds. The eastward shift of the Aleutian Low has been interpreted as a response in spatial structure to the interdecadal temperature change over the Northern Hemisphere and its association with tropical atmosphere–ocean interactions (Trenberth, 1990). Low-frequency climate variability is partially caused by the air–sea interactions between the subtropical gyre in the North Pacific Ocean and the Aleutian low-pressure system (Latif and Barnett, 1994). With respect to the interannual variability of the North Atlantic Oscillation (NAO), two centres shift eastward in the late 1970s (Jung et al., 2003). Similarly, some general characteristics of the PNA may have also experienced some changes at this time due to the interdecadal climate shift experienced by the PNA, dominant atmospheric circulation pattern in the North Pacific.
We examine structural changes in the PNA teleconnection pattern during the second half of the 1900s and their connection with interdecadal variability in tropical forcing. We demonstrate an intimate relationship between low-frequency variability and transient eddy activity on the interdecadal time scale and determine the extent of eddy feedback contribution to interdecadal structural changes in the PNA teleconnection pattern.
To identify the PNA teleconnection pattern, we use the November to February (NDJF) monthly mean geopotential height at 300 hPa (hereafter, Z300) during the years 1957–2002 derived from the European Centre for Medium-Range Weather Forecasts (ECMWF) 40-year reanalysis, ERA40 (Uppala et al., 2005). The horizontal resolution is 2.5° longitude × 2.5° latitude, and the PNA pattern is defined as the leading Empirical Orthogonal Function (EOF) over 150°E–60°W, 20–70°N. To detect an interdecadal pattern shift, we applied EOF analysis to the Z300 for the two separate time periods of 1958–1977 and 1983–2002, which we term the ‘pre-’ and ‘post-’ time periods, respectively. In addition to Z300, the 300 hPa zonal wind (U300), meridional wind (V300), and the vorticity (VO300) fields derived from ERA40 are analysed.
3. Changes in the PNA pattern
As a dominant climate signal of the Pacific Ocean, the PNA pattern may have undergone structural changes due to an abrupt climate shift over the Pacific Ocean during the late 1970s. Figure 1(a) shows the leading EOF of Z300 during the pre- and post-time periods. The first mode may represent the PNA pattern, when considering the similar spatial patterns of the horizontal shapes. For both time periods, the first mode displays three distinct cells over the central North Pacific, western Canada, and the southeastern US. There is a significant change in the PNA pattern. The post-period leading mode is displaced more to the east than is the pre-period mode. Therefore, the negative anomaly region over the central North Pacific is displaced clearly eastward. Also, a similar displacement is observed for the positive anomaly over Canada. The negative anomaly over the southeast portion of North America decreases in magnitude.
For the central North Pacific cell, the centre of action shifts northeastward, in addition to extending toward the west coast of North America, as denoted by a cross (pre-period) and a black dot (post-period) in Figure 1(a). The difference field between the pre- and post-periods shows three wave-like anomalies at boundaries of the PNA centres, which can represent expansions or shifts of the PNA pattern toward the eastern margins of each centre. In particular, negative anomalies over the northeastern Pacific indicate not only enhancement of the North Pacific branch but also weakening of the western Canadian branch over that area. The differences are significant at 90% confidence level (areas enclosed by a thick solid line). A test for significance is performed using a bootstrap method based on the EOF analysis of random samples. The most striking difference between the pre- and post-time periods is the eastward shift of the Canadian branch of the PNA. In addition, the centre of action of the central North Pacific branch also shifts eastward.
Though we showed that the leading EOF modes for two decades exhibit a significant pattern shift, one may argue whether the leading EOF modes represent the same physical phenomena because the EOF is just a statistical tool. We further check the second EOF, which exhibits a wave-like pattern (not shown). This second EOF shows interdecadal changes that also indicate a general eastward pattern-shift, implying that the basic state for the post-period offers favorable conditions for a systematic eastward shift of the dominant modes. From these results, we can conclude that the action centre of the low-frequency variability including PNA mode seems to be shifted eastward in the recent period over the Pacific-North America Region.
In order to further demonstrate such eastward displacement, we calculate the variances in the monthly Z300 anomalies during the pre- and post-periods (Figure 1(b)). It is evident that, during the post-period, the maximum variance region extends eastward and reaches the northeastern Pacific with a zonally elongated shape. The position of maximum variance shifts eastward during the late 1970s, but its displacement seems to be somewhat smaller than that of the leading mode. The displacement clearly demonstrates that the associated variance increased near the Gulf of Alaska, where the leading mode of Z300 shows a significant difference between the pre- and post-time periods.
In order to confirm the pattern change displayed in Figure 1, EOF analyses are performed 26 times using the monthly data from 20 winter seasons (NDJF). More specifically, the first EOF analysis is conducted for the 20 winters from 1957/1958 to 1976/1977, and the second for the period from 1958/1959 to 1977/1978. The final 26th analysis is carried out for the period from 1982/1983 to 2001/2002. Figure 2 depicts the peak longitudinal positions of the negative and positive anomalies at 45°N and 60°N, respectively. The action centres of both the central North Pacific and the Canadian cells gradually shift eastward, with the Canadian cell displaying faster movement than that of the central North Pacific cell (consistent with Figure 1(a)). These results show that the shift in the PNA pattern occurs gradually in the second half of the twentieth century and provide evidence that the eastward shift is not an artifact due to the changes in observational methods which occurred during the late 1970s.
To clarify the robustness of PNA pattern shift, we applied same procedure to NCEP-R1 dataset. The leading mode of each period shows consistent pattern with that of era40 dataset (not shown). PNA pattern during post-periods displaced to the eastward comparing to that of pre-period. The central North Pacific cell extends to North America and moves its action centre northeastward slightly. And the western Canada cell shows distinct displacement based on their positions during each period. Variance test also shows the eastward shift of PNA. In conclusion, the pattern change quite robust in independent datasets.
3.1. ENSO impact on the PNA pattern-shift
The characteristics of the ENSO (period, amplitude, spatial structure, temporal evolution, and onset) are changed following the interdecadal climate shift of the late 1970s (Wang, 1995; Wang and An, 2001). The most remarkable change occurs in the frequency of the ENSO cycle, as it moves from a high-frequency regime to a relatively low-frequency regime primarily due to a zonal shift in the wind stress anomaly that is an ENSO-coupled atmospheric mode (An and Wang, 2000). Recently, it has been reported on the changes in ENSO teleconnections over the North Pacific in a future warmer climate (Meehl et al., 2006; Muller and Roeckner, 2006; Meehl and Teng, 2007; Kug et al., 2010). The projected El Niño teleconnections are changed in the models with eastward- and northward-shifted anomalous lows in the North Pacific (Meehl and Teng, 2007). From the analysis of IPCC-AR4 CGCM simulations, the eastward shifts of not only El Niño but also La Niña teleconnection patterns over the North Pacific are caused by a systematic eastward migration of their associated tropical convection centre due to global warming (Kug et al., 2010). Our study concentrates on whether the interdecadal ENSO changes play a major role in the observed shift in the PNA pattern (Figure 1). If the ENSO teleconnection was changed, it might be reflected in the PNA pattern. In order to establish the teleconnection change induced by ENSO, the linear regression pattern of Z300 with respect to NINO3.4 SST is calculated during the pre- and post-periods. However, we found that the eastward shift of the ENSO teleconnection is not obvious over the mid-latitudes, unlike the zonal shift in wind stress anomalies over the Tropics. The action centre of the anomalous low does not move eastward, but is rather displaced slightly westward (not shown). From the above analyses, we conclude that the PNA pattern shift may have been caused by extratropical internal dynamics rather than by tropical forcing.
3.2. PNA and synoptic scale eddy activity
It is well known that energy transfers among waves of varying spatial scales. Generally, large-scale waves either maintain or decrease in strength by transferring their energy to smaller-scale waves, a phenomenon known as downscale energy cascade. On the other hand, high-frequency synoptic-scale waves can transmit their energy to larger horizontal-scale events such as the PNA, otherwise known as upscale energy cascade. The feedback of the high-frequency wave activity on the PNA pattern can be measured using the concept of eddy vorticity flux (EVF), which is defined as follows:
where u′, v′, and denote the 2–8 day band-pass filtered zonal wind, the meridional wind, and the vorticity, respectively, g is gravitational acceleration, f is the Coriolis parameter, and φ is geopotential height. The overbar indicates the monthly mean, the superscript, a, indicates anomalies from the monthly mean climate values. EVF represents how much synoptic scale eddy forcing directs to the low-frequency flow and its divergent component signifies the eddy-induced geopotential height tendency. Therefore, a convergence (divergence) of the EVF indicates a decreasing (increasing) tendency for the geopotential height, which translates into PNA enhancement (weakening).
The eddy-induced Z300 tendencies are regressed to each PNA index during the pre- and post-periods (Figure 3). Over the central North Pacific, negative geopotential height tendencies emerge following the PNA indices (Figure 3(a) and b). After the late 1970s, the pattern has shifted eastward with a zonally elongated shape, following the PNA pattern shift (as shown in Figure 1). The different distributions of the EVFs for the two periods is similar to those of the leading EOFs, with nearly negative values over the northeastern Pacific and positive values over the west and central North Pacific (Figure 3(c)). These changes imply that high-frequency eddy activity also follows the interdecadal climate change in the time mean concept. Many researchers have demonstrated the existence of positive feedback of high-frequency eddy activity to low-frequency large scale circulation (Nakamura and Wallace, 1990; Cai and van den Dool, 1991; Hoerling and Ting, 1994; Green, 1977; Lorenz and Hartmann, 2001, 2003; Feldstein, 2002, 2003; Orlanski, 2005; Jin et al., 2006a, 2006b; Meehl and Teng, 2007; Kug and Jin, 2009; Kug et al., 2009a; Ren et al., 2009). It seems somewhat trivial, therefore, that the negative EVF region shifts eastward following the change in PNA teleconnection pattern. Compared to the leading modes of Z300 (Figure 1), however, the movement of the EVF convergence zone is more striking than that of the cyclonic anomalies in EOF1, as shown by their action centre. These results indicate that synoptic eddy feedback may be related to the eastward shift of the PNA pattern.
In order to demonstrate the role of synoptic eddy feedback in the PNA pattern changes, we calculate the strength of the synoptic eddy feedback at each grid. The strength is defined by regressing the eddy-induced geopotential height tendency with respect to the monthly geopotential height at 300 hPa as follows:
This value may be regarded as a rough estimation of the eddy-induced local growth rate. Namely, a positive (negative) growth rate indicates that the eddy feedback locally enhances (weakens) the basic flow. A large value represents strong eddy feedback intensity at that grid point. As shown in Figure 4(a) and (b), the strong eddy-feedback area shifts eastward during the late 1970s. This indicates that low-frequency flow can have developed at the far east region during the post-period compared to the pre-period, consistent with the pattern change in the variance of the monthly Z300 (Figure 1(c).)
The strength of synoptic eddy feedback is determined by several factors, including synoptic eddy activity, eddy life time scale, and spatial length scale, among others (Jin et al., 2006a, 2006b). Among those factors, the intensity of the synoptic eddies is a critical component in determining the strength of synoptic eddy feedback (Jin, 2010). Figure 4(c) (contour) shows the difference between the 20-year averaged winter ‘storm tracks’ between the pre- and post-periods. Generally, the term, storm track, represents the region where the generation, movement and decay of synoptic scale eddy occur frequently near the upper troposphere. In this paper, we define synoptic eddy activity as a time mean of the square value of daily mean meridional winds at 300 hPa which are 2–8-day band-pass filtered every winter season (NDJF, 120 days). Compared to the climatology of the storm track (shading), the region of strong synoptic eddy activity extends southeastward after the late 1970s, with the maximum intensity locates near the northwest coast of North America. It is evident that the active EVF response zone co-locates with the region of anomalous synoptic eddy activity following its interdecadal pattern change. This observation is also consistent with the interdecadal shift in PNA pattern shown in Figure 1. It is plausible, therefore, that high-frequency eddy activity responding to interdecadal climate variation over the Pacific experiences a change in its pattern, causing the EVF to feed the corresponding low-frequency flow, the PNA teleconnection pattern, and its eastward displacement.
The temperature field of the North Pacific storm track is dominated by ripples on the baroclinic region across the oceans (Hoskins and Hodges, 2002). Interannual variation of storm track activity during mid-winter season can be modulated by high-latitude local baroclinicity induced by lower level temperature fields relating with East Asian winter monsoon phenomenon (Lee et al., 2010). Baroclinicity is affected by latitudinal temperature gradient, therefore, anomalously strong baroclinicity can develop at the northern part of the warm anomalies in the mid-latitude. Accordingly, the Pacific (inter) Decadal Oscillation (PDO) can contribute to the Pacific storm track as the baroclinicity changes its distribution at a certain low level temperature field relating with the PDO phase.
The PDO changes its pattern from a negative phase to a positive phase in the late 1970s, and the interdecadal change in synoptic eddy activity (storm track) is concurrent with this PDO phase shift. During the negative PDO phase, warm (cold) SST anomalies exist over the high latitude northwestern Pacific (over the subtropical eastern Pacific, including the western coast of North America), and vice versa during the positive PDO phase (Mantua et al., 1997; Mantua and Hare, 2002). Before the late 1970s, warm SST anomalies exists over the northwestern Pacific, synoptic eddies (storm tracks) activate near the northern part of the warm anomalies (high latitude of northwestern Pacific region) because larger temperature gradient enhances baroclinicity over that region. During the post-period, warm anomalies exists over the subtropical eastern Pacific, therefore anomalous baroclinicity can develop at the boundary between high latitude warm anomalies and low latitude cold anomalies, and formation of storm track is favourable at lower latitude comparing with that of the pre-period.
The typical PDO pattern in the SST fields has a zonal asymmetry, which means the boundary between the warm and cold anomalies slants to the northeast direction (Krishnan and Sugi, 2003; Peterson and Schwing, 2003; Mestas-Nu ez and Miller, 2006; Crueger et al., 2009). Under such conditions, it is conceived that baroclinicity changes depending on PDO phase are also zonally dependent. That is, during post-period baroclinicity can become relatively stronger at the eastern part of the cold anomalies (northeastern Pacific) due to the increased temperature gradient, compared to that during the pre-period. Such changes may give a favourable condition for eastward shift of the storm track, which result in downstream expansions of the PNA pattern with enhanced eddy vorticity feedback over the northeastern Pacific after the late 1970s. This hypothesis should be further investigated based on long-term observation and modelling works.
The PNA is basically driven by tropical forcing within the framework of the Rossby wave dynamics. Therefore, in relating with the mid-latitude pattern shift, simply the movement of tropical forcing and its amplitude can be considered as a key factor to drive PNA pattern change. Even so, the tropical forcing does not change its position remarkably in the interdecadal time scale. And the contribution of SST forcing over both Niño3, Niño4 regions to the mid-latitude have not changed enough for illuminating the PNA pattern shift because simple correlation test between the PNA index and Niño3, Niño4 indices do not show any worthy difference between pre- and post-periods. Therefore, we cannot expect the essential role of the tropical forcing for the PNA shift, even though more specific analysis remains. It might be that the eddy-induced forcing at the boundary of the low-frequency cells at the higher latitude is more efficient to redistribute the original circulation by feedback process and shift PNA pattern.
One may argue the role of eddy feedback effect because storm track shifted southeastward during the post-period, while the central North Pacific cell expanded to the northeast coast of Canada and its action centre shows some displacement northeastward. However, as we have shown in Figure 1, the northward shift of the PNA pattern is not significant compared to its eastward shift. Another point we have to recall to mind, is that the strong eddy feedback always appeared in the northern part of the storm track activity, which is defined by eddy kinetic energy. As we know that meridional length scale of the synoptic storm is quite long (∼several thousand kilometres), the latitudinal location of the effective eddy feedback can be different from that of the maximum eddy kinetic energy. However, this part will be further investigated.
We found a clear eastward shift of the PNA pattern after the late 1970s, which means the extension of the central North Pacific cell to the northwestern coast of Canada and displacement of action centre in the western Canada cell to the middle of the North America continent. The structural changes in the PNA are in good agreement with changes in the synoptic scale eddy feedback. It may seem trivial that the divergence of the EVF is well-matched with the PNA pattern shift because of the positive synoptic eddy feedback onto low-frequency flow. However, it is meaningful that the EVF, with respect to Z300, also shifts toward the northeastern Pacific after the late 1970s, because the eddy-induced local growth rate is independent of the pattern of low-frequency flow. Interdecadal changes in the ENSO mode cannot explain the PNA pattern shift. The anomalous low-frequency circulation over the North Pacific forced by the ENSO moves westward, not eastward, in observations. That changed our focus from the tropical forcing to the mid-latitude internal dynamics and found storm track shifts can be a key factor on the interdecadal shift in PNA pattern because frequent passage of eddies is a sufficient condition for strong feedback which influence on the PNA. We therefore propose the following mechanism. The Pacific storm track moves eastward corresponding to the interdecadal climate shift, and the active EVF response area is displaced eastward. Simultaneously, the PNA pattern also shifts eastward due to the positive feedback of EVF to the low-frequency flow. Accordingly, a structural change in the PNA teleconnection, which might have been resulted from an interdecadal change in the storm track not forced by the ENSO, also constitutes a part of the North Pacific climate shift in the late 1970s.
This work is supported by Korea Metrological Administration Research and Development Program under Grant RACS_2010-2007. YYL and GHL are supported by the Korea Foundation for International Cooperation of Science & Technology (KICOS) through a grant provided by the Korean Ministry of Education, Science Technology (MEST) in 2010 (No. K20903011170-10B0101-00500) and the second stage of the Brain Korea 21 Project.