Upper-tropospheric bridging of wintertime surface climate variability in the Euro-Atlantic region and northern Asia

A remarkable feature of interannual climate variability is a robust link of wintertime anomalies of surface air temperature (SAT) in northern Asia to pan-Atlantic SAT variations associated with the North Atlantic Oscillation (NAO). Here statistical analyses of data from the era of satellite observations (1979–2017) are used to show that about 80% of the variance of the winter (December-March) mean area-averaged SAT anomalies in northern Asia can be explained by the anomalous surface circulation associated with an NAO-like mode of sea level pressure variability over extratropical Eurasia. These SAT anomalies are related equally strongly to the “Lake Baikal” vortex representing variations of the upper-tropospheric circulation over northern Asia. Support is given for the scenario that this vortex drives SAT anomalies in northern Asia via surface-reaching displacements of isentropic surfaces and that it is coupled to climate variability in the Euro-Atlantic sector via interactions between the North Atlantic storm track, quasi-stationary planetary waves, and zonal-mean zonal winds. The results underpin the importance of a lesser-known zonal wavenumber-3 structure of disturbances trapped over Eurasia by the polar front jet rather than the better-known zonal wavenumber-5 structure of disturbances trapped by the subtropical jet for NAO teleconnections.

. Relation of wintertime (DJFM mean) surface climate variability in the NH extratropics to the NAO in the ESO period. (a) Anomalies of the sea level pressure (thin contours and color shading) regressed onto the NAO index defined as the PC time series of the leading mode of SLP variability in the North Atlantic region (magenta box). (b) Anomalies of the surface air temperature (thin contours and color shading) and surface wind u s (arrows) regressed onto the NAO index. In (a and b) red (blue) contours represent positive (negative) anomalies. The contour interval (CI) is 0.4 hPa and 0.25 K per unit NAO index, respectively. The zero contour is omitted. Pink (aquamarine) shading denotes positive (negative) anomalies statistically significant at the 95% confidence level. In (a) thick contours show the wintertime climatology of SLP (in hPa). In (b) the anomalies of u s (subsampled and masked if both components are nonsignificant at the 95% confidence level) are in m s −1 per unit NAO index. The blue and magenta boxes delineate areas for which indices of SAT variability are constructed. The maps were generated by MathWorks MATLAB R2014a with M_Map (http://www.eoas.ubc. ca/~rich/map.html).
Mechanisms of the NAO/AO influence on Asian temperatures are complex and not fully understood. They may involve temperature advection by anomalous zonal-mean winds 25 as well as interactions of the NAO with the Ural blocking 27 , the Siberian High, and the East Asian monsoon (see refs 28,29 for reviews). Low-frequency teleconnections are often established via the waveguiding effect of the time-averaged upper-tropospheric jets 30 . This effect produces zonally oriented chains of perturbations governed by planetary (Rossby) waves dynamics. In winter, these perturbations have a circumglobal character 31,32 and contribute to the NAO 33,34 and AO 35,36 variability. Quasi-stationary planetary waves play a key role, for instance, in bridging the winter East Asian monsoon to the NAO/AO and the associated anomalies of the zonal-mean circulation in the upper troposphere/lower stratosphere 29,37,38 .
The studies on the NAO relation to the circumglobal wavetrains underscore the importance of patterns of disturbances with a zonal wavenumber-5 structure that is related to the waveguiding effect of the subtropical Asian jet [32][33][34]39 . However, this effect may not provide an adequate explanatory framework for the relation between the NAO and SAT anomalies in northern Asia. On subseasonal timescales, these anomalies are occasionally related to an anomalous upper-tropospheric circulation in the Lake Baikal area, which is linked to the NAO via planetary waves propagating through Europe from the North Atlantic region 40 . On the seasonal timescale, the wintertime circulation and air temperature anomalies in the Lake Baikal area are often related to the NAO via changes in the strength of the stratospheric polar night jet 41 . A quite robust relation of the NAO to the anomalous upper-tropospheric circulation in the Lake Baikal area, hereafter referred to as the "Lake Baikal" vortex, was reported in a recent study of the winter mean circulation in the ESO period 8 . Specifically, that study showed a strong association between the "Lake Baikal" vortex and the leading EOF mode of the wintertime storm track activity (STA) over Eurasia, hereafter referred to as the STA EA mode, that is closely related to the NAO. It also showed that the STA EA -related "Lake Baikal" vortex is embedded in a wavenumber-3 circumglobal waveguide pattern (CWP3) guided by the polar front jet. These results suggest that interactions between the North Atlantic storm track and high-latitude stationary waves are instrumental in the linkage of SAT anomalies in northern Asia to the NAO.
While high-latitude CWP3s may significantly contribute to subseasonal and multidecadal surface climate variability 31 , their role in bridging climate anomalies in Asia to the storm track/NAO variability in the Euro-Atlantic sector has not been explored yet. The present study expands upon the findings on the wintertime climate variability in Eurasia reported in the already mentioned related study 8 . The related study mainly focused on the predictability of the storm track/NAO variability from Arctic sea ice cover anomalies. The present study examines in greater detail the relation of air temperature anomalies in northern Asia to the concurrent variability of the large-scale atmospheric circulation and storm tracks, as well as the role of quasi-stationary planetary waves and zonal-mean zonal wind anomalies in maintaining this relation. Unless stated otherwise, the results are based on statistical analyses of linearly detrended data from the NCEP/NCAR reanalysis in the ESO period (see Supplementary Table S1 for a summary of acronyms used in the study).

Relation of Air temperature Anomalies in northern Asia to Variations in tropospheric circulation
Relation to surface circulation anomalies. Correlation coefficients between key indices of the wintertime climate variability over extratropical Eurasia during the ESO period are given in Table 1. Indices of air temperature variations are based on SAT anomalies at the latitudes (40°-70°N) of the broad northern lobe of significant NAO-covariant SAT anomalies extending across the Eurasian continent between its Atlantic and Pacific coasts (Fig. 1b). Two distinct cores appear in this lobe, one in northern Europe and one in northern Asia. Air temperature variations over Europe and northern Asia are represented by the SAT E and SAT A indices defined as standardised anomalies of SAT averaged over the blue and magenta boxes in Fig. 1b, respectively. The analogous measure of air temperature variations over entire northern Eurasia (the SAT E+A index) is based on SAT anomalies averaged over both boxes. Even though only the western part of Eurasia is under the direct influence of strong NAO-covariant surface wind anomalies (Fig. 1b), the NAO relation to air temperatures in northern Asia is as pronounced as its impact on air temperatures in Europe. Indeed, the correlations of the NAO index with the SAT A and SAT E indices are equally high ( = . r 0 78).  Fig. 1a), SLP EA index (PC1 of sea level pressure variations over extratropical Eurasia, EA box in Fig. 2b), GPH LB index (area-averaged geopotential height anomalies at 300 hPa over the Lake Baikal area, LB box in Fig. 2c) and STA EA index (PC1 of upper-tropospheric storm track activity ( ′ ′ v v 300 ) variations over extratropical Eurasia, EA box in Fig. 6b). The SAT E , SAT A , and SAT E+A indices represent the area-averaged temperature within the E, A, and E+A boxes in Fig. 1b, respectively. The PC1s are the first principal component time series from the EOF decomposition of the given variable in the given area. All indices are based on linearly detrended DJFM mean data in the ESO period (1980-2017, years of the January). All correlations are significant at the 99.9% confidence level.
Given the close relation of air temperatures in northern Asia to the NAO, the pattern of SLP anomalies regressed onto the SAT A index (thin contours and color shading in Fig. 2a) is similar to the NAO-covariant pattern of these anomalies (Fig. 1a). Higher-than-normal air temperatures in northern Asia are associated with reduced SLPs in the Arctic and increased SLPs in the middle latitudes of the Euro-Atlantic sector. However, compared to the NAO-covariant SLP dipole, the centers of action in the SAT A -covariant SLP dipole are shifted eastward. The northern (stronger) center appears over the Barents Sea, and the southern (weaker) center extends to southwestern Europe. Highly significant anomalies are found on the eastern side of the northern center. In the southeastern Barents Sea/southern Kara Sea region, the correlations of the SLP anomalies with the SAT A index exceed 0.85 (see the thick contours in Fig. 2a). These high correlations underscore the importance of processes that tend to extend the Icelandic Low along the Iceland-Barents Sea corridor for climate variability in northern Asia. .
{0 70, 0 75, 0 80, 0 85}). In (b) arrows show the SLP EA -covariant anomalies of the horizontal wind velocity (in m s −1 per unit SLP EA index, subsampled and masked if both components are nonsignificant at the 95% confidence level) at 300 hPa. (c) Anomalies of the geopotential height at 300 hPa (Z 300 , thin contours and color shading) regressed onto the SAT A index. The CI is 5 gpm per unit SAT A index. Black arrows depict the wintertime climatology of jet streams. (d) Anomalies of the surface air temperature (thin contours and color shading) and surface wind u s (arrows) regressed onto the GPH LB index. The CI is 0.25 K per unit GPH LB index. The anomalies of u s (subsampled and masked if both components are nonsignificant at the 95% confidence level) are in m s −1 per unit GPH LB index. The contour and shading colors used in (a-d) are explained in the caption to Fig. 1  In (c-f) the CI is 0.2 K day −1 per unit SLP EA index. In (a-f) dark shading masks approximately areas where the climatological pressure at the surface is lower than 925 hPa. The thin contour and other shading colors are explained in the caption to Fig. 1. The maps were generated by MathWorks MATLAB R2014a with M_Map (http://www.eoas.ubc.ca/~rich/map.html).
The pattern of the SAT A -covariant SLP anomalies in Fig. 2a bears a striking resemblance to the pattern of SLP anomalies regressed onto the PC time series of the leading mode of SLP variability over extratropical Eurasia (EA box in Fig. 2b), hereafter referred to as the SLP EA index (see the thin contours and color shading in Fig. 2b for the SLP EA -covariant SLP anomaly pattern). The SLP EA index correlates with the NAO index at 0.83 and the SAT A index at 0.91 (see Supplementary Fig. S1a,b for comparison of the time series). These correlations are consistent with a slightly stronger dependence of temperature variations in northern Eurasia (the SAT E+A index) on the SLP EA index ( = . r 0 89) than the NAO index ( = . r 0 82). To further examine the impact of anomalous circulation on lower-tropospheric temperatures in northern Eurasia, the middle and lower panels in Fig. 3 display patterns of the SLP EA -covariant terms of the thermodynamic equation (Eq. 2 in Methods) at 925 hPa. To facilitate interpretation of these terms, the arrows in the upper panels of Fig. 3 show the SLP EA -covariant wind anomalies superimposed on the climatological temperature ( Fig. 3a) and the climatological wind field superimposed on the SLP EA -covariant temperature anomalies (Fig. 3b) at the same level. Dark shading marks the approximate location of the mountain ranges transecting the 925 hPa level to emphasise their guiding effect on the anomalous near-surface circulation over Eurasia. Mean temperature advection by the anomalous wind is evidently a key driving agent for the near-surface NAO-like air temperature variations (compare the pattern of θ − ⋅ ∇ u a h m in Fig. 3c with the temperature anomalies in Fig. 3b). When the SLP EA index is positive, the Greenland-Labrador area is abnormally cooled by the enhanced transport of polar air by northerly and northwesterly wind anomalies while Europe is abnormally warmed by the reinforced transport of maritime air by anomalous westerlies. At the same time, Siberia is warmed due to the warm air advection by southwesterly wind anomalies. The distinct core of anomalous temperature north of Lake Baikal is induced by significant wind anomalies across a zone of strong baroclinicity (large horizontal gradients of the climatological temperature) in this area (Fig. 3a). The Asian and European cores of the anomalous temperature field are separated by the meridionally-oriented Ural Mountains at about 60°E (not seen in Fig. 3 because of their relatively small height), indicating that the topography of the Urals may play a role in forming these cores. The major mountain chains of the Eurasian continent should also be important, as suggested by the anomalous winds over eastern Europe and western Asia that follow a U-shaped route along the northern rim of these chains (Fig. 3a).
In northeastern Asia, some near-surface warming during the positive polarity of the SLP EA index results from anomalous descending motion across the climatological isentropic surfaces, as indicated by the sign of mean temperature advection by the anomalous pressure velocity ( ω θ − ∂ ∂p / a m ) at 925 hPa in this area (Fig. 3e). Farther south, along the Pacific coast, the SLP EA -covariant temperature anomalies are primarily due to anomalous temperature advection by the climatological wind (see Fig. 3d for the pattern of θ − ⋅ ∇ u m h a at 925 hPa). The combined forcing by anomalous eddy heat flux convergence (E a ) and diabatic heating (J a ) tends to destroy the www.nature.com/scientificreports www.nature.com/scientificreports/ SLP EA -covariant temperature anomalies in northern Europe as well as in northern Asia, as indicated by the sign of + E J a a at 925 hPa in these areas (Fig. 3f).

Relation to upper-tropospheric circulation anomalies. Above-normal air temperatures over northern
Asia are associated with a poleward shift of the North Atlantic jet stream. This shift is illustrated in Fig. 2c showing the anomaly pattern of GPH at 300 hPa (Z 300 ) regressed onto the SAT A index (thin contours and color shading) on the background of the climatological jets (black arrows). In the Euro-Atlantic sector, the GPH anomalies form a dipole. During the positive polarity of the SAT A index, this dipole consists of an Arctic trough (lobe of negative GPH anomalies) and a ridge (lobe of positive GPH anomalies) in middle latitudes, with centers of action over Greenland and western Europe, respectively. Since the GPH gradients drive geostrophic winds, the dipole represents cyclonic motion in the north (the reinforced upper-tropospheric polar vortex around Greenland) and anticyclonic motion in the south. These upper-tropospheric vortices correspond to the enhanced Icelandic Low and Azores High at the surface (Fig. 2a). The meridional GPH gradients across their common rim drive anomalous westerlies (see the arrows in Fig. 2b for the pattern of significant wind anomalies at 300 hPa regressed onto the related SLP EA index). These anomalous westerlies reflect the poleward migration of the North Atlantic jet. In Fig. 2c, this migration is recognised by a more southern location of the climatological jet than the boundary between the anomalous vortices. The upper-tropospheric wind anomalies in the Euro-Atlantic region are accompanied by an anomalous vortex over Asia with the center of action above Lake Baikal. In the positive phase of the SAT A /SLP EA index, this "Lake Baikal" vortex is anticyclonic and corresponds to a well-pronounced ridge straddling the climatological Eurasian polar front jet (Fig. 2b, arrows, and Fig. 2c). While the vortices in the Euro-Atlantic region are largely equivalent barotropic (anomalous winds near the surface tend to keep the direction of strong wind anomalies in the upper troposphere), the "Lake Baikal" vortex has a more baroclinic structure. It considerably weakens and migrates southward in the lower troposphere. At the surface, its remnants can be recognised as weak SLP anomalies in southern Asia (Fig. 2a,b). A baroclinic structure of anomalous vortices over eastern Eurasia was also noted in the context of intraseasonal amplification of the Siberian High 42 and intraseasonal variation of the strength of the East Asian trough 43 .
Following the related study 8 , the "Lake Baikal" vortex is characterised by the GPH LB index defined as standardised anomalies of Z 300 averaged around Lake Baikal (LB box in Fig. 2c). This index correlates quite highly with the NAO index ( = .
r 0 71) and even higher with the SLP EA index ( = . r 0 84; see Supplementary Fig. S1c for comparison of the time series) and the SAT A index ( = . r 0 89; see Table 1). These high correlations suggest that the "Lake Baikal" vortex controls thermal variability underneath and, consequently, contributes hand in hand with the anomalous surface westerlies in the Euro-Atlantic sector to coherent wintertime air temperature variations at www.nature.com/scientificreports www.nature.com/scientificreports/ the ground across the entire Eurasian continent. Such a scenario is consistent with the anomaly patterns of SAT and surface wind regressed onto the GPH LB index (Fig. 2d), which strongly resemble the corresponding NAO-covariant patterns (Fig. 1b). The anomaly pattern of SAT regressed onto the GPH LB index after using regression to remove the signal associated with the NAO index exhibits a significant lobe in northern Asia but not in either Europe or Greenland ( Supplementary Fig. S2a). Conversely, the anomaly pattern of SAT regressed onto the NAO index after removing the GPH LB -covariant signal from the time series shows the canonical Europe-Greenland seesaw in temperature anomalies but lacks their Asian lobe ( Supplementary Fig. S2b), further supporting the scenario that the Asian lobe is controlled by the "Lake Baikal" vortex.
Vertical structure of temperature anomalies over Asia. It was previously suggested from an analysis of downstream NAO influences on subseasonal timescales 40 that upper-tropospheric circulation anomalies in  the Lake Baikal area may drive air temperature variations in northern Asia via surface-reaching displacements of isentropic surfaces. This mechanism should be even more relevant on the seasonal timescale. Indeed, the link between the anomalies of the winter mean SAT and upper-tropospheric circulation over northern Asia is much stronger than the corresponding link between the month-to-month anomalies in winter. While the correlations of the SAT A index with the local GPH anomalies at 300 hPa in the Lake Baikal area exceed 0.8 for the winter mean data, the corresponding correlations drop below 0.4 for the month-to-month anomalies ( Supplementary  Fig. S3a).
To further support the scenario of an upper-tropospheric regulation of the wintertime SAT variability in northern Asia, Fig. 4a-c displays the latitude-pressure distributions of the GPH LB -covariant anomalies of the relative vorticity ζ, air temperature T and buoyancy frequency N averaged from 90° to 125°E (thin contours and color shading) together with the corresponding correlations (thick contours). In the Lake Baikal area, the GPH LB -covariant anomalies of the relative (and hence absolute) vorticity extend throughout the atmosphere but have the largest amplitude in the upper troposphere/lower stratosphere (Fig. 4a). During the anticyclonic polarity of the "Lake Baikal" vortex (positive GPH anomalies and negative anomalies of ζ), anomalously cold temperatures in the polar lower stratosphere coexist with anomalously warm lower-stratospheric temperatures in low latitudes and, therefore, strengthen the north-to-south temperature gradient above the tropopause (Fig. 4b). At the same time, the thermal contrast between the low and moderately-high latitudes is weakened below the tropopause, as indicated by a tropospheric dipole of temperature anomalies that is out-of-phase with the dipole in the lower stratosphere. Significant anomalies in the tropospheric lobes reach the surface. In contrast to the southern lobe where the largest tropospheric temperature anomalies appear at upper levels, in the northern lobe, the largest temperature anomalies are found at the surface, on the northern edge of the "Lake Baikal" vortex. However, the temperature anomalies within the vortex (between 50° and 60°N) are related to the GPH LB index very tightly throughout the troposphere, with correlations exceeding 0.95 between 850 and 400 hPa. These anomalies change sign at the core level (300-250 hPa) of the vortex (compare the thin contours in Fig. 4a,b). In the positive phase of the GPH LB index, such a structure of temperature anomalies corresponds to decreased static stability throughout the troposphere, with an extreme magnitude at 300 hPa (Fig. 4c). Therefore, the anomalous anticyclonic absolute vorticity and reduced static stability in the "Lake Baikal" vortex are manifestations of a negative upper-tropospheric potential vorticity anomaly. Such an anomaly can produce surface warming by pushing isentropic surfaces downward 44 . Conversely, in the negative phase of the GPH LB index, a positive upper-tropospheric potential vorticity anomaly corresponding to the anomalous cyclonic absolute vorticity and increased static stability in the "Lake Baikal" vortex can produce surface cooling by pulling isentropic surfaces upward. While these processes do not require teleconnections to the North Atlantic region (see the discussion of Supplementary Fig. S2a above), they are often associated with such teleconnections, as indicated by significant NAO-covariant upper-tropospheric GPH anomalies in the Lake Baikal area (Fig. 4d).
To explain the surface amplification of temperature anomalies in the "Lake Baikal" vortex form the perspective of the heat balance, Fig. 5 shows the latitude-pressure distribution of the GPH LB -covariant terms of the thermodynamic equation averaged zonally over the 90°-125°E longitudes (only anomalies in the layer 1000-100 hPa are plotted). On the northern side of the vortex, the tropospheric and lower-stratospheric temperature anomalies are maintained by mean temperature advection by the anomalous wind, as indicated by the sign of the respective lobes in the distribution of θ − ⋅ ∇ u a h m (Fig. 5a). This advection has nearly the same magnitude throughout its tropospheric lobe. Anomalous temperature advection by the mean wind displaces the tropospheric temperature anomalies towards the southern side of the vortex, as indicated by a tropospheric dipole in the distribution of θ − ⋅ ∇ u m h a (Fig. 5b). This dipole is confined to the free troposphere so that the total anomalous horizontal advec- u a h a is negligible) exhibits a maximum at the surface on the northern side of the vortex. This maximum maintains the corresponding temperature maximum (Fig. 4b). Anomalous temperature advection by the total (mean plus anomalous) pressure velocity is negligible (not shown) while mean temperature advection by the anomalous pressure velocity ( ω θ − ∂ ∂p / a m ) maintains tropospheric temperature anomalies around the central latitude of the vortex and destroys them on the southern and northern sides of the vortex (Fig. 5c). In the lower stratosphere, mean temperature advection by the anomalous pressure velocity counteracts mean temperature advection by the anomalous wind.
The combined forcing by anomalous eddy heat flux convergence and diabatic heating ( + E J a a ) tends to destroy the temperature anomalies in the "Lake Baikal" vortex throughout the troposphere. During the anticyclonic phase of the GPH LB index, this forcing represents a heat sink having an extreme magnitude at or near the surface (Fig. 5d). A localised heat sink should produce positive potential vorticity above the sink that is apportioned between increased absolute vorticity (cyclonic circulation) and increased static stability 45 . Such changes are of opposite sign to the changes of absolute vorticity and static stability associated with the GPH LB index (Fig. 4a,c). Therefore, the + E J a a forcing should also exert negative dynamic feedback on the "Lake Baikal" vortex. Two-way interactions between upper-tropospheric circulation anomalies and near-surface processes also contribute to intraseasonal climate variations over Asia 42,43 .

Relation to Variations in Storm track Activity
Displacements of the North Atlantic storm track. Higher-than-normal air temperatures over north- www.nature.com/scientificreports www.nature.com/scientificreports/ EOF mode of the variability in ′ ′ v v 300 over extratropical Eurasia (EA box in Fig. 6b). Both patterns (see the thin contours and color shading in Fig. 6b for the STA EA -covariant anomalies of ′ ′ v v 300 ) exhibit a zonally-elongated lobe of significant STA anomalies north of about 45° that extends from eastern North America across the North Atlantic and Europe to Asia and, over eastern Europe, from the Black Sea to the Barents Sea. This northern lobe is accompanied by a southern lobe of weaker and far less significant out-of-phase STA anomalies with the center of action over the Iberian Peninsula. In the northern lobe, the largest anomalies appear over the North Sea region, but the most significant anomalies are found slightly downstream, over northern Europe at about 60°N (see the thick cyan contours in Fig. 6a,b for the isolines of high correlations). This feature underscores the importance of processes that tend to extend the North Atlantic storm track along the Eurasian jet stream (see the arrows in Fig. 6a for the NH jets climatology) for climate variability in northern Asia.
The STA EA index correlates highly with the SAT A index ( = . r 0 80; see Table 1) and the SLP EA index ( = . r 0 88; see Supplementary Fig. S1d for comparison of the time series), as well as the NAO ( = . r 0 83) and GPH LB ( = . r 0 76) indices 8 . Consistent with these tight relations, the STA EA -covariant anomaly patterns exhibit all major features associated with the other indices, including the upper-tropospheric "Lake Baikal" vortex and large static stability anomalies in the vortex area (see Fig. 6c; arrows for the wind anomalies and thin contours and color shading for the buoyancy frequency anomalies at 300 hPa). They show the equivalent barotropic structure of the zonal wind anomalies in the Euro-Atlantic sector (compare the arrows in Fig. 6c with the arrows in Fig. 6d representing the wind anomalies at 925 hPa). They also exhibit all NAO-like thermal features in the lower troposphere, including the European and Asian cores in the broad lobe of air temperature anomalies over northern Eurasia (see the thin contours and color shading in Fig. 6d for the temperature anomalies at 925 hPa). Taken together, the anomaly patterns in Fig. 6 strongly suggest that the displacements of the North Atlantic storm track play a key role in the tight relation of air temperatures over Eurasia to the large-scale circulation anomalies. This scenario is further supported by a much stronger link of air temperatures in northern Asia to the STA and wind anomalies in the Euro-Atlantic sector for the winter mean data than the month-to-month anomalies. Indeed, whereas the correlations of the SAT A index with the anomalies of ′ ′ v v 300 over Europe exceed 0.7 for the winter mean data, the corresponding correlations drop below 0.3 for the month-to-month anomalies ( Supplementary Fig. S3b). Similarly, the winter mean and month-to-month SAT A indices correlate with the corresponding upper-tropospheric vortices in the Euro-Atlantic region at levels above 0.7 and below 0.3, respectively ( Supplementary Fig. S3a).

Synoptic eddy forcing. Transient eddy forcing of the mean flow can be expressed by the extended
Eliassen-Palm vectors E u and E v (see Eqs 6 and 7 in Methods). Divergences of these vectors represent local eddy-induced accelerations of the zonal and meridional winds, respectively 46 . In the lower troposphere, these accelerations occur mainly through baroclinic processes described by the vertical components , where ′ ′ u T is the eastward eddy heat flux. Assuming that the eddy heat flux vanishes at the surface, a positive anomaly of the poleward (resp. eastward) eddy heat flux at a lower-tropospheric level should drive a westerly (resp. northerly) wind anomaly at that level. The relation between the winter mean wind anomalies and their baroclinic forcing due to synoptic eddies at 850 hPa associated with the leading mode of variability in the surface circulation over Eurasia is summarised in Fig. 7a, in which the arrows show the anomaly pattern of significant SLP EA -covariant vectors ) provide full information on the barotropic eddy forcing of both wind components. Arrows in Fig. 7b show the anomaly pattern of the SLP EA -covariant E u h ( ) vectors due to synoptic eddies at 300 hPa on the background of the corresponding anomalies of the zonal wind (u 300 ; thin contours and shading) and the correlations of the meridional wind at the same level (v 300 ) with the SLP EA index (thick contours).  (Fig. 7b) mirrors the corresponding pattern associated with the leading mode of storm track activity variations over Eurasia (Fig. 7c), reflecting the already noted quasi-equivalence of these modes. In the Euro-Atlantic sector, both patterns are similar to the corresponding pattern of the E u h ( ) vectors associated with the month-to-month AO mode derived from a numerical model 35 . However, in contrast to the latter, they show a significant propagation of eddy activity from Scandinavia into Asia and lack a significant connection between the Pacific and Atlantic storm tracks across North America. Still, the link of the winter mean upper-tropospheric circulation anomalies over Asia to the synoptic eddy forcing in the ESO period should mainly reflect nonlocal connections. The anomalous E u h ( ) vectors associated with the SLP EA /STA EA mode are less significant over Asia than Europe and, over Asia, they do not exhibit a clear pattern able to explain the "Lake Baikal" vortex (Fig. 7b,c). Moreover, the NAO-covariant E u h ( ) vectors are generally not significant over Asia (Fig. 7d) despite the significant relation of the "Lake Baikal" vortex to the NAO (Fig. 4d).

Relation to Variations in Stationary Planetary Waves and Zonal Mean Winds
Anomalous wavenumber-3 pattern. Stationary planetary waves are driven by zonally-asymmetric forcings, such as land orography, sources/sinks of diabatic heating, and transient eddies 47,48 . Therefore, changes in the eddy momentum and heat fluxes by storm track displacements as well as the associated reorganisation of diabatic heating sources or interactions of the induced circulation anomalies with land orography may produce anomalous planetary waves, promoting non-local responses. Anomaly patterns of the meridional wind at 300 hPa indicate that quasi-stationary planetary waves are important for connecting the upper-tropospheric winds and lower-tropospheric temperatures over Asia to the circulation and storm track anomalies in the Euro-Atlantic region. The anomalies of v 300 regressed onto the SAT A and GPH LB indices (see the thin contours and color shading in Fig. 8a,b) exhibit a quasi-zonal wavenumber-3 circumglobal structure in high latitudes (approximately   50°N and 75°N). A similar structure is associated with the SLP EA , STA EA , and NAO indices (see the thick contours in Fig. 7b-d). The six lobes of this high-latitude CWP3 are denoted by letters A-F in Fig. 8a. Lobe A (over the Canadian Arctic Archipelago) and lobe B (over the northern North Atlantic) correspond to, respectively, the western and eastern limbs of the anomalous Greenland vortex (Fig. 2b,c). Lobes D and E over northern Asia correspond to, respectively, the western and eastern limbs of the "Lake Baikal" vortex (marked by the black box in Fig. 8b). Lobe D is separated from lobe B by a strong lobe located over central Europe (lobe C). Lobe E is separated from lobe A by a weak and much less significant lobe F that is confined to the high Arctic (to the area north of the Bering Strait). Lobes C-E are aligned with the climatological Eurasian polar front jet (see the arrows in Fig. 8a). In the south, they are accompanied by lobes (marked as G and H in Fig. 8a) that are located on the climatological axis of the northern Africa-southern Asia subtropical jet. These lobes seem to be part of a less significant low-latitude wavetrain having a distorted wavenumber-5 structure. Their relation to transient eddies in the Euro-Atlantic region is consistent with a recent analysis of month-to-month circulation variability along the wintertime subtropical Asian jet 49 .
The anomalous high-latitude CWP3 represents a shift of a quasi-zonal wavenumber-3 climatological planetary wave. This shift is shown in Fig. 8b by superimposing the climatological meridional wind at 300 hPa (thick contours) on the GPH LB -covariant anomalies of this wind. The shift has zonal and meridional components. The anomalous wavetrain appears on the average about 10° of latitude poleward of the climatological one. In the zonal direction, it is approximately in quadrature with the climatological wavetrain. In particular, the Asian lobes (D and E) of the anomalous wavetrain straddle one of the three strongest climatological lobes, that is, the one co-located with the center of the "Lake Baikal" vortex.
While the circumglobal character of the high-latitude wavetrain is well pronounced in the meridional wind anomalies, it is less evident in the corresponding anomalies of the geopotential height (Fig. 2c) since, because of the geostrophic balance, v is proportional to the zonal gradient of Z and thus has almost no zonal mean. In contrast, the geopotential height anomalies include a significant zonally-symmetric component corresponding to the anomalies of the zonal-mean zonal wind {u} discussed below. This component somewhat masks the high-latitude wavetrain that is present in the zonally-asymmetric component Z* of the geopotential height anomalies (see the thin contours and color shading in Fig. 8c for the STA EA -covariant anomalies of Z* at 300 hPa). From all lobes of the ⁎ Z 300 wavetrain, the Asian one (marked as DE in Fig. 8c) is linked most significantly to the STA EA index. This feature is shown by the isolines of high correlations of the anomalies of ⁎ Z 300 with the STA EA index in Fig. 8c (thick contours). The highest correlation (0.80) is found in the center of the Asian lobe located north of Lake Baikal at 60°N, supporting the scenario that the "Lake Baikal" vortex is linked to the storm track variability in the Euro-Atlantic sector via planetary waves.
The vertical structure of the high-latitude CWP3 is illustrated in Fig. 9a,b showing the longitude-pressure distributions of the meridional wind anomalies (thin contours and color shading) averaged from 50° to 70°N and www.nature.com/scientificreports www.nature.com/scientificreports/ regressed onto the GPH LB and STA EA indices, respectively, together with the corresponding correlations (thick contours). Generally, the CWP3 wavetrain has an equivalent barotropic structure in the free troposphere, below the cores of the meridional velocity anomalies at 300-250 hPa. An exception is lobe F, which is a stratospheric feature. A significant lower-stratospheric signature (the highest correlations at 200-100 hPa) is also found in the Greenland vortex lobes (A and B). From the perspective of the surface climate variability, the most outstanding lobe is the one over northwestern Asia (lobe D). This lobe has only a weak signature in the lower stratosphere but reaches the surface. In fact, it is most significant just at the surface where it correlates very highly ( = .
r 0 86) with the GPH LB index and also highly ( = . r 0 76) with the STA EA index. Moreover, at the surface, lobe D spreads significantly eastward underneath lobe E of the upper-tropospheric wavetrain. Therefore, in agreement with the patterns of surface and near-surface wind anomalies (Figs 2d and 6d), most of northern Asia is under the influence of anomalous meridional winds that strongly affect local air temperatures (Fig. 3a-c).
Anomalous zonal-mean zonal winds. In the lower stratosphere, the six lobes of the high-latitude CWP3 are anchored to an anomalous quasi-annular flow corresponding, in the positive phase of the GPH LB and STA EA indices, to the strengthening of the cyclonic polar vortex. This feature is illustrated in Fig. 8d displaying the STA EA -covariant anomaly patterns of the wind velocity (arrows) and its meridional component (thin contours and color shading) at 150 hPa. These patterns also exhibit significant lower-stratospheric extensions of the Asian lobes of the low-latitude wavetrain in the upper troposphere (lobes G and H in Fig. 8a). During the positive polarity of the STA EA index, these extensions represent a southward diversion of an easterly flow anomaly from the North Pacific (lobe H) and a northward diversion of a westerly subtropical flow from the North Atlantic (lobe G). A cyclonic loop from the latter feeds anomalous easterlies over the North Atlantic found at the latitudes (30°-45°N) of the anomalous easterlies over eastern Asia. Consequently, it contributes to the southern lobe of a dipolar structure of the zonal-mean zonal wind anomalies in the NH extratropics. This structure is demonstrated explicitly in the upper panels of Fig. 10 showing the latitude-pressure distributions of the anomalies of {u} associated with the GPH LB and STA EA indices (thin contours and color shading) together with the corresponding correlations (thick contours).
The southern lobe of the anomalous {u}-dipole, centered at 37.5°N, has a core at 200 hPa (Fig. 10a,b). The northern lobe exhibits most significant anomalies at about the same level and also near the surface (at 925-850 hPa). However, the correlations within this lobe are quite uniform throughout the troposphere. In the upper troposphere, the lobe is centered at a latitude (60°N) corresponding to the "central" latitude of the high-latitude CWP3 defined as the average latitude of the centers of action in its five most significant lobes (A-E) at 300 hPa. At any level, the northern lobe of the anomalous {u}-dipole is more significant than the southern one. Correlations in the northern lobe reach 0.84 at 60°N and 200 hPa for the STA EA index (Fig. 10b) and 0.76 at 65°N and 150 hPa for the GPH LB index (Fig. 10a). These high correlations and the "co-location" of the high-latitude CWP3 with the northern lobe of the anomalous {u}-dipole strongly suggest that the circumglobal teleconnectivity in high latitudes involves feedbacks between synoptic eddies, stationary waves, and zonal-mean zonal wind perturbations.
Anomalous planetary wave activity fluxes. Theoretically, the anticyclonic phase of the "Lake Baikal" vortex could be maintained by eastward propagating quasi-stationary Rossby waves exited by anomalous upper-level wind divergences/convergences in the Euro-Atlantic sector and trapped by the Asian jets waveguide. Regardless of whether such divergences/convergences are driven by the synoptic eddies or vice versa, the steering of the "Lake Baikal" vortex by planetary wave sources in the Euro-Atlantic sector would be consistent with an analysis of downstream NAO influences on subseasonal timescales during warm El-Niño-Southern Oscillation winters 40 . Significant out-of-phase lobes are indeed found in the STA EA -covariant anomaly pattern of the vertical velocity ω at 500 hPa (ω 500 ; see Fig. 11a). In this pattern, the lobe of negative anomalies of ω 500 centered over the Nordic Seas corresponds to an anomalous upper-level wind divergence while the lobe of positive anomalies of ω 500 centered over the Mediterranean Sea corresponds to an anomalous upper-level wind convergence. In the northern region, the wind divergence at 300 hPa averaged over the Norwegian Sea and Scandinavia (red box in Fig. 11a) correlates with the STA EA index at 0.86. The STA EA index is also linked tightly ( = . r 0 80) to the wind convergence at 300 hPa averaged over the Mediterranean Sea (blue box in Fig. 11a). Anomalous convergences in this area were suggested to be drivers of local precipitation anomalies 50 and downstream NAO influences on subseasonal timescales via planetary waves trapped by the subtropical jet 39 .
To examine how the winter mean displacements of the North Atlantic storm track are related to sources and sinks of quasi-stationary Rossby waves, the horizontal (F s h ( ) ) and vertical (F s z ( ) ) components of the Plumb wave activity flux F s are calculated (see Eqs 3 and 4 in Methods) and regressed onto the STA EA index. The F s vector is approximately parallel to the direction of the wave energy propagation 51 . The anomaly pattern of the STA EA -related F s h ( ) vectors at 300 hPa is shown in Fig. 11b (arrows) on the background of the corresponding anomalies of the zonal wind (contours and color shading). In the positive phase of the STA EA index, the anomalous wave activity emanates from the subpolar latitudes of the North Atlantic region where its divergence entails anomalous Rossby wave generation by barotropic processes. This "Atlantic" stream propagates southeastward and piles up wave activity (F s h ( ) anomalies converge) in low latitudes but does not enter into Asia. Over eastern Europe, it confluences with a stream that emanates from the Arctic regions of Asia. This "Asian" stream is most significant in the Lake Baikal area where it propagates mainly equatorward. Its divergence indicates a source of wave activity in the lobe of the westerly wind anomalies on the northern side of the "Lake Baikal" vortex while its convergence indicates a sink of wave activity in the lobe of the easterly wind anomalies on the southern side of the vortex. While differences are found in the Pacific sector, over the Atlantic-Eurasian region, the pattern of the STA EA -covariant F s h ( ) vectors is reminiscent of the corresponding pattern associated with the month-to-month AO variations in a numerical model 35 .
( ) at 500 hPa). The zonally-averaged meridional and vertical components of the F s vector are proportional to the zonally-averaged equatorward momentum flux {− ⁎ ⁎ u v } and poleward heat flux {v*T*} by the stationary waves, respectively. Their divergence D F (see Eq. 5 in Methods) corresponds to the divergence of the conventional Eliassen-Palm flux, which is a driving agent for the zonal-mean westerlies 52 . An analogous driving agent due to the synoptic eddies (D E ) is obtained by zonal averaging of the divergence of the E u vector (see Eq. 8 in Methods). The latitude-pressure distributions of the STA EA -covariant anomalies of D F and D E (thin contours and color shading) together with the corresponding correlations (thick contours) are shown in the bottom panels of Fig. 10. The middle panels of this figure display the distributions of the barotropic contribution to these anomalies. The quasi-stationary momentum fluxes (Fig. 10c) conform with the synoptic momentum fluxes (Fig. 10d) in forcing the STA EA -covariant meridional displacement of the westerlies (Fig. 10b). At the tropopause level (about 250-300 hPa), the forcing by the quasi-stationary momentum fluxes dominates in the high-latitude lobe of the anomalous westerlies while the forcing by the synoptic momentum fluxes dominates in the mid-latitude lobe. The forcing by the synoptic momentum fluxes tends to be more significant in the lower stratosphere while the forcing by the quasi-stationary momentum fluxes is more significant near the surface. The total forcing by either the quasi-stationary waves (Fig. 10e) or synoptic eddies (Fig. 10f) has a more complicated structure, which is characterised by a vertical tilt in the troposphere imposed by the baroclinic component. Below the tropopause, a weaker forcing by the synoptic heat fluxes generally tends to counteract a stronger forcing by the quasi-stationary heat fluxes. In the lower stratosphere, the quasi-stationary heat fluxes drive the poleward extension of the high-latitude lobe of the anomalous westerlies. At the tropopause, the total forcing by the synoptic eddies is more significant than the total forcing by the quasi-stationary waves not only in the mid-latitude core but also in the high-latitude core of the anomalous westerlies. This feature is not evident in the forcing of the zonal-mean zonal wind anomalies associated with the month-to-month AO variability 35 .

Discussion
A remarkable feature of climate variability in the North Atlantic-Eurasian region is a robust NAO-related recurrence of a lobe of coherent wintertime air temperature anomalies extending from the Atlantic to the Pacific coast of northern Eurasia. Results from the related study 8 show that the same wintertime NAO index as employed here accounts for 74% of the variance ( = . r 0 86) of the leading mode of the concurrent SAT variability in extratropical Eurasia during the ESO period. Here it is shown that the wintertime SAT anomalies in northern Asia are related to the NAO as strongly as the corresponding SAT anomalies in northern Europe. It is also shown that the SAT A index representing the wintertime area-averaged SAT anomalies in Asia north of 40°N is strongly coupled (83% of the variance explained) to the anomalous surface circulation represented by the leading mode of sea level pressure variability over extratropical Eurasia (the SLP EA mode). The SLP EA mode is an NAO-like mode of variability characterised by anomalous surface westerlies in the Euro-Atlantic region and a northern center of action in the SLP anomaly field moved into the Barents Sea region. A high fraction (79%) of the SAT A variance is also explained by upper-tropospheric circulation anomalies over northern Asia encapsulated in the GPH LB index representing the "Lake Baikal" vortex (an anomalous regional-scale ridge/trough centered over Lake Baikal). Therefore, this vortex and the SLP EA mode of the surface circulation variability are often manifestations of the same large-scale phenomenon. The vertical structure of the GPH LB -covariant anomalies of the absolute vorticity and static stability indicates that the "Lake Baikal" vortex regulates the tropospheric temperature variability in northern Asia via displacements of isentropic surfaces in a surface-reaching potential vorticity anomaly. Together with the concurrent heat transport by the anomalous surface westerlies in the Euro-Atlantic sector, it contributes to coherent SAT variations across entire northern Eurasia. When the "Lake Baikal" vortex appears independently from the NAO-related westerly wind anomalies in the Euro-Atlantic sector, it still drives significant SAT anomalies over northern Asia. Without the forcing by the "Lake Baikal" vortex, the NAO-related SAT anomalies do not extend far into Asia. As deduced from other studies 8,40 and demonstrated here, the impact of the "Lake Baikal" vortex on the SAT variability in northern Asia is much more robust in the case of the winter mean anomalies than the month-to-month anomalies in winter.
It is also found that the linkage of the SAT variability in northern Asia to storm track changes is much more robust for the winter mean than month-to-month anomalies. Moreover, the winter mean SAT anomalies in northern Asia are related more tightly to anomalous upper-tropospheric storm track activity over Europe than Asia. The large-scale anomaly pattern of this activity associated with the SAT A index is reminiscent of the leading mode of storm track activity variations over extratropical Eurasia (the STA EA mode). The latter shares 77% of its variance with the SLP EA mode of the surface circulation variability in that region. All these features suggest that displacements of the North Atlantic storm track play a key role in bridging the surface climate variability in the Euro-Atlantic region and northern Asia. This scenario is also supported by a common wavenumber-3 circumglobal structure of upper-tropospheric/lower-stratospheric anomalies of the meridional wind in high-latitudes associated with the STA EA and GPH LB indices. The anomalous CWP3 wavetrain corresponds to a zonal and meridional shift of an analogous climatological wavetrain. The anomaly patterns of the STA EA / SLP EA -covariant extended Eliassen-Palm vectors indicate that this shift could be triggered or maintained by the synoptic eddy forcing of the mean flow in the Euro-Atlantic sector. This forcing is not only consistent with the meridional shift of the North Atlantic jet but also supports strong meridional wind anomalies over Europe in the area between two relatively weak lobes of the climatological wavetrain.
The main centers of action in the anomalous CWP3 wavetrain appear in the latitude band of the northern lobe of the STA EA /GPH LB -covariant extratropical dipole in the zonal-mean zonal wind anomalies. The conventional Eliassen-Palm diagnostics indicate that the synoptic eddies, as well as the quasi-stationary waves, are important drivers of these anomalies. The anomaly patterns of the STA EA -covariant Plumb wave activity flux suggest that the linkage of the "Lake Baikal" vortex to the North Atlantic storm track displacements is not maintained via anomalous propagation of quasi-stationary Rossby waves from the North Atlantic region to Asia. These patterns exhibit local barotropic wave activity sources not only in the North Atlantic region but also over northern Asia, where also a strong baroclinic source of wave activity appears at mid-tropospheric levels in the Lake Baikal area.
Conceptually, the relationships between changes in the circulation, storm tracks, and quasi-stationary waves in the Atlantic-Eurasian region summarised above are consistent with the previously demonstrated ability of synoptic eddies to drive NAO-like circulation anomalies in the Euro-Atlantic region 11,53,54 and zonal-mean zonal wind anomalies associated with the NAO variability 36,55 . They also conform with the ability of quasi-stationary waves to drive the NAO/AO-related storm track variations 14 and zonal-mean zonal wind anomalies 35,56 . They are also consistent with the ability of the perturbed zonal-mean zonal winds to drive anomalous stationary waves through interactions with the climatological stationary waves 55,57 and the two-way coupling between the anomalous stationary waves and zonal-mean winds 58 . A novel result, suggested in the related study 8 and elaborated here, is the likely importance of the anomalous CWP3 wavetrain for the robust relationship between wintertime displacements of the North Atlantic storm track and air temperature anomalies over northern Asia. This feature has remained unnoticed by others probably because the studies on the NAO/AO linkages to planetary waves and storm tracks are usually focused on subseasonal timescales at which relations between climatic anomalies in the Euro-Atlantic sector and northern Asia are not robust. Another reason is that these studies often include data from pre-ESO decades. Before the ESO period, the circulation anomalies over northern Asia might not have been robustly linked to the NAO even on the seasonal timescale 41 .
The vertical structure of the GPH LB -covariant anomalies of diabatic heating and transient eddy heat flux convergence indicates that these anomalies exert a local negative dynamic feedback on the "Lake Baikal" vortex. Since only concurrent relations are investigated here, one cannot rule out the scenario in which local anomalies in northern Asia are, at least for some forced events, precursors of large-scale feedbacks leading to the NAO/ AO like wintertime circulation changes. Such precursors may be related, for instance, to autumnal snow cover  64 . The anomalies are computed by subtracting the local linear trend from the wintertime data at each grid point. The convergence of meridians is taken into account by weighting the anomalies by the factor φ cos 1/2 . The principal component (PC) time series from the EOF decomposition are standardised to have zero mean and a standard deviation of one. The anomalies of selected fields are then regressed onto the PC of the first leading mode of a given variable.
The first modes of extratropical variability in sea level pressure are computed separately for a North Atlantic domain (Fig. 1a, magenta box), extending from 20°N to 80°N and from 90°W to 40°E, and an Eurasian domain (Fig. 2b, magenta box) extending from 30°N to 80°N and from 10°E to 140°E. These modes are referred to as the NAO mode and the SLP EA mode, respectively. The NAO mode is a version of the standard EOF-based NAO mode 4 , calculated over the same domain as in the related study 8 . The first mode of extratropical variability in storm track activity ( ′ ′ v v 300 ) is computed for the Eurasian domain. It is the STA EA mode introduced in the related study 8 . The NAO, SLP EA and STA EA modes explain 52.4%, 58.6% and 28.2% of the variance in the data, respectively. All these modes are statistically reliable according to North's "rule of thumb" employed to assess their uniqueness from uncertainty on the eigenvalues of the covariance matrix 65 .
The PC time series of the NAO, SLP EA and STA EA modes are referred to as the NAO, SLP EA and STA EA indices, respectively. Additional indices of interannual variability are constructed by area-averaging of selected variables. The GPH LB index, introduced in the related study 8 , is obtained by averaging the geopotential height at 300 hPa in the Lake Baikal area (Fig. 2c, blue box) extending from 45°N to 60°N and from 90°E to 125°E. The SAT E and SAT A indices are obtained by averaging the surface air temperature over northern Europe (Fig. 1b, blue box) and northern Asia (Fig. 1b, magenta box). The European domain extends from 40°N to 70°N and from 0°E to 60°E. The Asian domain extends from 40°N to 70°N and from 60°E to 140°E (approximately the same area as used for computation of the corresponding SAT A index in the related study 8 ). The SAT E+A index is obtained by averaging the surface air temperature over the combined European and Eurasian domains. The linearly detrended and standardised time series of indices used in regressions are shown in Supplementary Fig. S1.
Relations between time series are quantified using the sample correlation coefficient r. A two-sided Student's t-test 64 is employed to assess statistical significance of r. The test is performed using an effective sample size 66  , where N 0 is the length of the series while r a and r b are the lag-one autocorrelations of the correlated series a and b. Correlation coefficients between the "thermodynamic" indices (SAT E , SAT A and SAT E+A ) and "dynamic" indices (NAO, SLP EA , GPH LB and STA EA ) are given in Table 1. In regression plots (Figs 1-11), the anomalies of scalar fields significant at the 95% confidence level are shaded and the anomalies of vector fields are plotted at points at which the anomaly of any vector component is significant at the 95% confidence level.