Sea ice variability over the Barents Sea with its resultant atmospheric response has been considered one of the triggers of unexpected downstream climate change. For example, East Asia has experienced several major cold events while the underlying temperature over the Arctic has risen steadily. To understand the influence of sea ice in the Barents Sea on atmospheric circulation during winter from a synoptic perspective, this study evaluated the downstream response in cyclone activities with respect to the underlying sea ice variability. The composite analysis, including all cyclone events over the Nordic seas, revealed that an anticyclonic anomaly prevailed along the Siberian coast during light ice years over the Barents Sea. This likely caused anomalous warm advection over the Barents Sea and cold advection over eastern Siberia. The difference in cyclone paths between heavy and light ice years was expressed as a warm-Arctic cold-Siberian (WACS) anomaly. The lower baroclinicity over the Barents Sea during the light ice years, which resulted from a weak gradient in sea surface temperature, prevented cyclones from traveling eastward. This could lead to fewer cyclones and hence to an anticyclonic anomaly over the Siberian coast.
The decline in Arctic sea ice during summer has had a leading role in temperature amplification during autumn and winter, partly through air–sea heat transfer (Graversen et al. 2008; Screen and Simmonds 2010). One of the important heat transfer processes is the release of ocean heat associated with autumn cyclone activity along the marginal ice zone (Inoue and Hori 2011). Frequent meridional heat transport, as well as air–sea heat exchange during autumn, is vital to the freezing of the Arctic Ocean until early winter. Winter temperature anomalies are especially large over the Barents Sea (Serreze et al. 2011). In this region, warm advection by anomalous winds helps to keep the ocean ice free and the overlying atmosphere warm.
Remote responses to the warm Arctic at midlatitudes have been found in recent years and are receiving increased research attention (Overland et al. 2010). Francis et al. (2009) showed that low values of summer ice extent are related to higher winter temperatures not only over the Arctic but also throughout the Northern Hemisphere. One exceptional area is northern Siberia, which exhibits a cooling anomaly when the ice extent is low. Recent radical shifts of atmospheric circulations were responsible for the cold winter anomaly over the Eurasian continent during the winters of 2001/02 to 2005/06 (Zhang et al. 2008). The impacts of Siberian coldness during winter 2005/06 were widespread from Europe to East Asia. Using numerical experiments, Honda et al. (2009, hereafter H09) concluded that the reduced ice cover in the Barents and Kara Seas in summer 2005 accounted for the cold anomalies in East Asia the following winter. This was due to a stationary Rossby wave induced by anomalous turbulent heat fluxes, which in turn amplified the Siberian high. Petoukhov and Semenov (2010) found the same response in a limited situation with 80%–40% ice reduction over the Barents and Kara Seas. The sea ice anomaly in 2007 also intensified surface anticyclones over the Eurasian and American continents in association with anomalous advection of cold polar air on their eastern sides, bringing colder temperatures along the Pacific coast of Asia and northeastern North America (Orsolini et al. 2012). The effects of record persistence of the negative phase of the North Atlantic Oscillation (NAO) and the Arctic Oscillation (AO) in winter 2009/10 on the United States, Europe, and East Asia were investigated (Jung et al. 2011; Cattiaux et al. 2010; Cohen et al. 2010). While the NAO/AO was helpful for determining the hemispheric tendency of cold air flowing in and out of the Arctic region and into the midlatitudes, the downstream effect of a blocking high over the Nordic seas gave a more deterministic and predictable view of the cold-air advection from the Arctic (Hori et al. 2011). Croci-Maspoli and Davies (2009) also found that the anomalous cold European winter in 2005/06 was not related to a negative phase of the NAO but to a pattern with a blocking high located over the North Atlantic Ocean.
A strong blocking high over the North Atlantic is closely related to the generation of polar lows over the Barents Sea during winter (Blechschmidt et al. 2009). In addition, a cyclonic anomaly over northern Norway is known to be associated with an anticyclonic anomaly along the west coast of Greenland, as much as 3 days prior to the outbreak of polar lows (Businger 1985). Although not detailed in the literature, a response downstream of an anticyclonic anomaly is also visible following the mature stage of cyclones (Fig. 7 of Businger 1985). This anticyclonic anomaly should induce cold advection over the Eurasian continent. Although the temporal and spatial scales differ between polar lows and synoptic cyclones, baroclinic instability seems to be a common mechanism for the generation of both types of cyclones. In addition, analyses of synoptic activity (e.g., cyclone tracking) sometimes provide good explanations of the physical mechanisms behind statistically observed relationships (Zhang et al. 2004; Finnis et al. 2007; Stroeve et al. 2011). Therefore, the cyclone activity over the Barents Sea during winter might be a good indicator for the interpretation and prediction of cold events over the downstream region.
The climatology of Arctic cyclone activity shows a high cyclone count over the North Atlantic sector and from the Iceland/Greenland Sea to the Barents Sea (Zhang et al. 2004). The position of a sea ice edge likely affects cyclones, particularly their development and track. The sea ice distribution over the Barents Sea has a large year-to-year variability with a strong air–ice–sea coupled system. Although sea ice retreat over the Barents Sea was hypothesized to enhance the westerly wind-driven oceanic inflow via frequent local cyclogenesis (Ikeda 1990; Bengtsson et al. 2004), the cyclone density over the Nordic seas was found to be only weakly correlated with the Barents Sea ice extent during winter (Sorteberg and Kvingedal 2006). Therefore, the dependence of cyclone behavior over the Barents Sea on the variability of sea ice cover and cyclone impact on the downstream climate system has not been fully clarified.
Here, we focus on how each cyclone generated over the Nordic seas is influenced by the ice edge over the Barents Sea during winter and impacts the warm-Arctic and cold-continental pattern. By comparing the cyclone tracks between light and heavy ice years over the Barents Sea, we assess the linkage between cyclone characteristics and downstream impact.
2. Data and methods
We obtained atmospheric data from the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis (Kalnay et al. 1996) for mean sea level pressure (SLP), surface air temperature (SAT), geopotential height, and wind fields. The data have a spatial resolution of 2.5° × 2.5° on a regular latitude–longitude grid. Six-hourly data from December 1979 to March 2011 were used.
The Met Office Hadley Centre Sea Ice and Sea Surface Temperature (SST) dataset version 1 (HadISST1) (Rayner et al. 2003) was also used in this study. These data consist of monthly globally complete fields on a 1.0° × 1.0° regular latitude–longitude grid.
b. Cyclone identification and tracking
An algorithm for cyclone identification and tracking, developed by the University of Melbourne (for details, see Simmonds and Murray 1999), was applied to the NCEP–NCAR reanalysis. To find possible cyclones, the Laplacian of pressure (∇2p) at each grid point was compared with values at neighboring grid points for the whole Northern Hemisphere during winter [December–February (DJF)] during the period 1979–2011. When a potential cyclone was identified, the position of the associated pressure minimum was determined by iteration to the center of the ellipsoid best fit to the pressure surface. Identified cyclones were tested by a concavity criterion, which required that the average value of the Laplacian exceed 0.2 hPa (°lat)−2 over a radius of 2° latitude. After identifying potential cyclones with this algorithm for each time step, a tracking algorithm estimated each cyclone track by scanning the connection from a point detected in the previous time step based on cyclone characteristics (e.g., cyclone steering velocity). Matching between each old and new cyclone was evaluated as possible combinations. The greatest probability gave the matching for the cyclone track. To remove noise, we only included cyclones that lasted more than 1 day. We focused on cyclones generated over the Nordic seas, including the Barents Sea (65°–85°N, 30°W–60°E), to reveal the atmospheric response to the variability in sea ice distribution.
3. Warm-Arctic cold-Siberian (WACS) anomaly
To understand the effect of sea ice variability on the cyclones and their synoptic environment, we selected typical cases, defined as heavy and light ice cover years over the Barents Sea. Figure 1a shows the standard deviation of ice concentration in December from 1979 to 2010. The northern Barents Sea had the maximum variability (box area in Fig. 1a). Using the time series of the anomaly field over this area, we selected heavy and light ice cover years (Fig. 1b). We confirmed that the anomaly of ice concentration in December statistically persists during the whole winter. As heavy ice years, 1980, 1981, 1988, 1997, and 2003 were selected for analysis, while as light ice years 2004–07 and 2009 were used.
For the heavy and light ice cases, 205 and 207 cyclones were detected, respectively, in winter (including the following January and February). Although the numbers are nearly equal between the cases, the mean central minimum SLP of cyclones in the heavy ice case (982.9 hPa) was lower than that in the light ice case (986.9 hPa). Furthermore, the average position shifted northward by about 2° in the light ice case (dots in Fig. 1a). This suggested that the sea ice retreat might have affected the cyclone tracks. Figure 2 shows all the cyclone tracks generated over the Nordic seas. In the heavy ice case (Fig. 2a), the tracks and the positions where the SLP reached a minimum (green crosses in Fig. 2) were concentrated over the Norwegian coast and Barents Sea, whereas in the light ice years (Fig. 2b), the tracks tended to be spread out, with some heading directly toward the North Pole.
To show the atmospheric environment in the heavy and light ice cases, the composite SLP fields during DJF when each cyclone reached the minimum SLP are also shown in Fig. 2 (shading). The SLP was 5 hPa deeper over the Norwegian Sea in the heavy ice case (Fig. 2a) than in the light ice case (Fig. 2b), supporting the cyclone statistics mentioned before. In the light ice case, the Siberian high expanded northward up to 70°N. To highlight the atmospheric responses to the sea ice anomalies, we created a difference field by subtracting the heavy ice SLP from the light ice SLP (Fig. 3a). An anticyclonic anomaly was visible along the coastal area of Siberia near the Taymyr Peninsula (75°N, 90°E) as well as in Scandinavian regions. This anticyclonic anomaly seemed to bring anomalous warm air from the North Atlantic sector and cold air from northeastern Siberia (Fig. 3b), creating the WACS anomaly, which is likely a precursor to severe weather in the downstream East Asian region.
Serreze et al. (2011) showed that the recent enhanced warm anomaly over the Barents Sea is influenced by the enhanced warm advection under declining sea ice. The anomalous warm advection might lead to reduced surface heat fluxes because of the low air–sea temperature difference, preventing the air mass from cooling. The warm southwesterly wind anomaly likely prevents the sea ice from forming and advecting southward. This would help explain why the warm anomaly over the northern Barents Sea extended to 85°N (Fig. 3b). To confirm this notion from an atmospheric point of view, we calculated the baroclinicity, which we defined as the vertical zonal wind shear between 500 and 925 hPa during cyclogenesis (i.e., the time when each cyclone was initially detected). Figure 4a shows the anomaly field of baroclinicity between heavy and light cases. A remarkable weakening zone was observed from the east Greenland coast to the Kara Sea. Over the Barents–Kara Seas, the reduced baroclinic zone was limited to a narrow region. Therefore, the zonal wind at the steering level of cyclones (e.g., 500 hPa) became weaker in the light ice years, hindering the eastward propagation of cyclones from this region. As expected, the system density (i.e., number and strength of cyclones) along the Barents–Kara Seas also decreased (Fig. 4b) because of its northward shift (Fig. 2b), hinting that the sea ice edge and SST distribution play important roles in synoptic development (Adakudlu and Barstad 2011).
To better understand the role of SST in the WACS anomaly, we focused on the SST field over the Barents Sea. Figures 5a,b show the horizontal distributions of SST in December 2003 (heavy ice year) and December 2005 (light ice year). In the light ice year, the area of open water in the northern Barents Sea expanded northward with a SST below 1°C (Fig. 5b). Compositing 5 yr for both cases, we found a robust difference in the SST gradient over the area (Fig. 5c). Because SSTs in the coastal region did not vary greatly from 2°C, the meridional gradient basically depended on the southernmost ice edge in the Barents Sea. The distance from the coastline at which the SST became lower than −1°C differed by as much as 300 km, signifying a large difference in baroclinicity (Fig. 4a).
4. Summary and discussion
To elucidate the mechanisms of the recent severe cold winters in East Asia, the effect of the upstream atmospheric circulation triggered by sea ice variability over the Barents Sea was investigated. The synoptic characteristics of cyclones generated near the Nordic seas showed that the sea ice variability over the Barents Sea very likely controls the cyclone tracks through changes in baroclinicity. The relationship between the warm anomaly over the northern Barents Sea and the positive system density around Svalbard under a light ice situation is linked to anomalous warm horizontal advection by northward-moving cyclones. The northward shift of cyclone tracks creates an anomalous anticyclonic circulation over the Siberian coast, triggering a distinct advection of cold air over the northern Siberia sector. Therefore, our findings support the idea that cyclone paths under a light ice situation over the Barents Sea are the driving mechanism for generating the WACS anomaly.
To date, the large-scale response to reduced ice extent over the Barents Sea sector has been discussed using general circulation models. Alexander et al. (2004) showed a significant anticyclonic SLP anomaly over eastern Siberia under a case of reduced ice concentration (winter 1995/96). The horizontal distribution is partly the same as in the WACS anomaly (Fig. 3a), although the amplitude is significantly stronger in our case. This suggests that each synoptic event is important for generating the WACS anomaly. Magnusdottir et al. (2004) confirmed the remote response to SST and sea ice anomalies over the North Atlantic sector. Although each anomaly is responsible for weakening storm activity over the North Atlantic basin, a cold anomaly over eastern Siberia was only found in the reduced-ice case. This result also supports the idea that the difference in local baroclinicity over the Barents Sea influences the continental cold anomaly. A significant large-scale atmospheric circulation response was also found in projected Arctic sea ice loss at the end of the twenty-first century (Deser et al. 2010). The WACS-like and baroclinic vertical structure anomaly was seen in early winter; however, this response was modified to the equivalent barotropic pattern in late winter, suggesting a difference in cyclone activity in association with projected sea ice anomalies between early and late winter. Regarding cyclone intensity and frequency, Finnis et al. (2007) found that the cyclone intensity over the Barents Sea and northern Siberian sector was slightly weakened during winter under a reduced-ice situation (i.e., twenty-first-century run), although the frequency did not change between the twentieth and twenty-first centuries; this tendency is very similar to our result. They speculated that the loss of autumn ice cover greatly reduces meridional temperature gradients in the lower troposphere.
The WACS anomaly is also very similar to those found by H09 and Petoukhov and Semenov (2010), who concluded that the cold anomaly over east Siberia is triggered by a stationary Rossby wave emanating from anomalous turbulent heat fluxes as a result of anomalous ice cover over the Barents–Kara Seas. The study of H09 was based on a seasonal time scale, at which the effect of the ice anomaly during October persists into late winter. While the results of H09 were statistically significant, the source of this persistence was not well discussed. Our analysis offers a more detailed view of the sea ice influence on the downstream anticyclonic anomaly on intraseasonal time scales as manifested in the changing cyclone tracks.
To further elucidate this point, we also analyzed the 250-hPa geopotential height (Z250) response described by H09 between light and heavy ice years (Fig. 6). Notably, the upper-tropospheric pattern shifted upstream by about a quarter wavelength relative to the SLP pattern (Fig. 3a), reflecting the baroclinic nature of the response found by H09. Significant wavelike anomalies occur across Eurasia, which are associated with the propagation of wave activity flux (WAF), as indicated by arrows (Takaya and Nakamura 2001) in Fig. 6. This suggests that this wave train was probably associated with a stationary Rossby wave excited by anomalous turbulent heat fluxes around the Barents Sea as described by H09. Therefore, the dynamical remote response from the anomalous ice cover over the Barents Sea toward the upper atmosphere in east Siberia reported by H09 also exists on intraseasonal time scales. Because the remote response proposed by H09 is triggered by the turbulent heat flux over the ice-free ocean, it strongly corresponds to the case in which cold-air advection is present. In our study, the mean position of the cyclone center is located over the Barents Sea opening regardless of whether it is a heavy or light ice year (dots in Fig. 1a), which creates warm advection over the Barents Sea. Thus, both the cold advection near the ice margin and the warm advection brought by the cyclone systems are responsible for the creation of the WACS anomaly.
To demonstrate that the existence of fewer cyclones over northern Siberia enhances the anticyclonic anomaly (i.e., northward expansion of the Siberian high), we created composite anomaly maps by simply averaging SLP and SAT anomaly fields during five winters between cases of heavy ice and light ice years (Fig. 7). Compared to the cyclone composite field (Fig. 3), the anticyclonic anomaly is weakened by 2 hPa along the Siberian coast (Fig. 7a); accordingly, the cold anomaly is also reduced over central Siberia (Fig. 7b). This fact suggests that cyclones developing over the Barents Sea have a leading role in the emergence of the WACS anomaly. From the viewpoint of larger atmospheric circulation change, however, the northward shift of cyclone centers and tracks might also be related to the northeastward shift of the NAO/AO center of action over the past decade (Zhang et al. 2008).
Although the midlatitude climate is also influenced by other teleconnections, for example, the El Ñino–Southern Oscillation (Sakai and Kawamura 2009), each cyclone path over the Arctic should be worth monitoring for cold-air accumulation over Siberia in short-term forecasts (weekly or less). Furthermore, the variability in Barents ice cover has the potential predictability for long-term forecasts (seasonal and monthly scales). However, the atmospheric circulation leading up to the WACS anomaly might change in the near future as sea ice is significantly diminished (e.g., Petoukhov and Semenov 2010). Therefore, we must pay careful attention to the transitional phase of the Arctic system and its changing impact on the midlatitude climate system.
We thank Prof. I. Simmonds at the University of Melbourne for providing the cyclone tracking algorithm. JI is partly supported by the Japan–Norway Researcher mobility programme (Norwegian Research Council Project 211932/F11). The authors also thank three anonymous reviewers for their helpful comments.