How does the SST variability over the western North Atlantic Ocean control Arctic warming over the Barents–Kara Seas?

The rapid increase in Arctic temperature and retreat of sea ice have been reported and widely discussed in the scientific literatures (Comiso et al 2008, Stroeve et al 2012, Vihma 2014). The increase in Arctic temperature is most pronounced during early winter (October–December) and is not spatially uniform, but exhibits several particular regional warm cores (Screen and Simmonds 2010) including the Barents–Kara Seas, East Siberian–Chukchi Seas, and northeast Canada and Greenland. Interestingly, the atmospheric warming over each location in the Arctic is known to lead to mid-latitude cooling, but with quite different spatial patterns (Mosley-Thompson et al 2005, Cohen et al 2012, Francis and Vavrus 2012, Hanna et al 2014, Kim et al 2014, Mori et al 2014, Kug et al 2015, Nakanowatari et al 2015, Lim et al 2016). Therefore, the peculiar recent phenomena called 'Warm Arctic-Cold Continents' (Overland and Wang 2010, Overland et al 2015) can be effectively categorized by the regional warm cores in the Arctic.

Although there are many studies on how the above-mentioned regional Arctic warming and reduced sea ice cover over those regions could induce cold winter extremes in mid-latitudes, relatively few studies have been devoted to finding the driving mechanism for those regional Arctic warming events. Recently, a linkage between the oceanic thermal condition of the North Atlantic Ocean and Arctic surface temperature has been suggested (Zhang et al 2013, Nakanowatari et al 2014, Sato et al 2014, Luo et al 2016), which is supported by other findings that show both the surface air temperature over the Barents–Kara Seas (BKSAT) and sea surface temperature (SST) over the western North Atlantic Ocean (WNAO) have rapidly increased in recent decades (Wu et al 2012, Pershing et al 2015, Saba et al 2016). It is also found that the warming over the WNAO is in association with the northward shift of SST front over the Gulf Stream (Minobe et al 2008, Wu et al 2012).

Among these studies, we revisit Sato et al (2014) which provides a close observational link between the Barents–Kara Seas and the WNAO, over which the northern part of the Gulf Stream passes. Using linear baroclinic model experiments, Sato et al (2014) suggested that the changes in the local diabatic heating in association with the poleward shift of the Gulf Stream can induce a large-scale circulation pattern travelling into the Arctic inducing significant Arctic warming. However, the linear response shown in figure 5(d) of their paper was quite weak and more importantly, missed a possible contribution from the large baroclinic eddy activities over the region, which is amply noted by other studies (Sampe et al 2010, Frankignoul et al 2011, Sung et al 2014). As the transient eddy forcing in the North Atlantic tends to induce the large-scale teleconnection pattern, called the Scandinavian pattern (SCAND), travelling over the north Atlantic and Arctic (Bueh and Nakamura 2007), it is important to take into account baroclinic eddy activities.

In this regard, Sato et al (2014)'s study is incomplete, although their finding casts considerable light on the divergent perspectives about 'Warm Arctic-Cold Continents' by revealing that apparent links between the Barents Sea ice cover and cold Eurasian winters form just a sector of a teleconnection pattern that originates remotely in the North Atlantic Gulf Stream region (Simmonds and Govekar 2014). Therefore, it is worthwhile evaluating whether the warming over the WNAO induces a sufficient transient eddy forcing for the large-scale teleconnection pattern over the North Atlantic and Arctic region.

In this study, we aim to provide a more plausible explanation on how the warm SST anomaly in the WNAO sector modulates the Eurasian teleconnections and affects warming over the Arctic, and in particular, the Barents–Kara Sea in early winter. Special attention will be devoted to the role of transient eddy forcing, which was not studied by Sato et al (2014). The relative importance of transient eddy forcing to the thermal forcing was assessed by a simple model specially designed to treat each forcing separately. General circulation model experiments were also conducted to support observational findings and simple model results.

The primary observational dataset used in this study includes Hadley Centre Sea Surface Temperature (HadISST) data with 1° × 1° horizontal resolution (Rayner et al 2003) and the re-analysis dataset obtained from the U.S. National Centers for Environmental Prediction (NCEP)/National Center for Atmospheric Research (NCAR), which has a 2.5° × 2.5° horizontal resolution (Kalnay et al 1996). Both the daily and monthly mean dataset for the 1979–2013 period were utilized in this study.

In order to investigate distinguishable influences from several independent SST modes of the North Atlantic Ocean separately, Empirical Orthogonal Function (EOF) analysis was applied for early winter (October–December) mean SST anomalies over the North Atlantic Ocean domain (95°W∼15°E, 20.5°N∼88°N). Latitude weighting was applied by multiplying the square root of the cosine prior to the EOF analysis. North's rule of thumb (North et al 1982) was used to test the significance of EOF modes. Regression analysis was conducted using the obtained EOF principal component (PC) time series to retrieve the associated circulation patterns.

In this study, interannual variability of surface air temperature over the Atlantic sector of the Arctic region in early winter is represented by the detrended time series of area-averaged BKSAT. The boxed area indicated in figure 1(a) was used as the area average.

Zoom In Zoom Out Reset image size

The stationary wave model (hereafter SWM, Ting and Yu (1998)) was employed to examine the dominant forcing mechanism of stationary Rossby waves. This SWM is the dry dynamical core of a fully nonlinear baroclinic model. The prognostic variables include vorticity, divergence, temperature and log-surface pressure with R30 truncation in the horizontal and L14 vertical levels on sigma coordinates. The main forcings in this model were diabatic heating, convergence of transient eddy vorticity fluxes and transient eddy heat fluxes. The forcing terms can be tested using idealized distribution or diagnosed forcing fields derived from observations. In this study, the latter approach was used (see supplementary information available at stacks.iop.org/ERL/12/034021/mmedia). The three forcing terms can be defined as:

$$\begin{equation} TF_{\text{vor}} = - \Delta (\overline{V^\prime, \xi^\prime} ) \end{equation} \tag{ 1 }$$

$$\begin{equation} TF_{\text{temp}} = - \left( {\frac{P}{{P_0 }}} \right)^{{\raise0.7ex\hbox{$R$} \!\mathord{\left/{\vphantom {R {C_p }}}\right.\kern-\nulldelimiterspace}\!\lower0.7ex\hbox{${C_p }$}}} \left[ {\nabla \cdot (\overline {V^\prime\theta^\prime} ) + \frac{{\partial (\overline {\omega^\prime\theta^\prime})}}{{\partial p}}} \right] \end{equation} \tag{ 2 }$$

$$\begin{equation} \text{Q}_\text{1} = \frac{{\partial \overline T }}{{\partial t}} + \overline V \;\Delta \;\overline T + \overline \omega \left( {\frac{{\partial \overline T }}{{\partial p}} - \frac{{R\overline T }}{{C_p p}}} \right) - TF_{\text{temp}} \end{equation} \tag{ 3 }$$

where ξ is the vorticity, V is the horizontal wind, p is pressure, ω is the pressure vertical velocity, and TF_vor and TF_temp indicate the non-linear transient eddy vorticity flux convergence and transient eddy heat flux convergence, respectively. Q₁ indicates the monthly mean diabatic heating. Note that Q₁ used in this study is different from that in Sato et al (2014) because of the existence of TF_temp in (3). The bar represents the monthly mean and prime shows the deviation from the monthly mean. Further details of the model equations or information can be found in Ting and Yu (1998) and Wang and Ting (1999).

To investigate the impact of SST warming over the WNAO in a more realistic modelling framework, we used a fully coupled general circulation model (GCM), Climate Model Version 2.1 (CM2.1) developed by the Geophysical Fluid Dynamical Laboratory (GFDL) (Delworth et al 2006). As a control run, we conducted climatological equilibrium simulations with 400 ppm CO₂ for 100 years. In a forced simulation, SST over the WNAO region (the box in figure 5(a), i.e. 38^oN–48°N, 55°W–75°W) was restored toward the prescribed warm SST conditions with 5 day restoring time scale. According to Pershing et al (2015), the WNAO region is the highest warming place on the earth and, in the last decade, there was 2 °C increase of SST. Accordingly, we prepared the warm SST condition over the WNAO region by adding the observed SST trend of the recent 11 years (2004–2014) to the climatological SST fields of control run. Note that the model freely evolves except for the boxed region in figure 5(a) in the forced run. To estimate the response to the SST forcing over the WNAO, we will analyze differences between the results of the forced run and the control run.

3.1. Warming over Barents–Kara Seas and SCAND teleconnection pattern

3.2. EOF analysis on North Atlantic SST variabilities

Prior to verifying the above hypotheses, we conducted EOF analysis to determine whether there exists an identifiable North Atlantic SST variability linked to the Arctic warming over the Barents–Kara Seas. The early winter averaged SST anomalies during the 1979–2013 period were decomposed into three dominant modes: the first mode (EOF1) explains 36.6% of the total variance and exhibits a strong linear trend. The spatial pattern of EOF1 shows apparent warming over the entire North Atlantic basin. Although the pattern contains significant SST warming over the Barents–Kara Seas, the correlation between the PC1 and BKSAT is low (0.07). Note that BKSAT is a detrended index.

The second mode explains 14.5% of the variance, and has three centres of action which are located over the western North Atlantic Ocean, the northern North Atlantic Ocean, and the eastern North Atlantic (figure 2(c)). The temporal correlation coefficient between the second PC and BKSAT time series is very low (0.03) indicating no significant relationship, as with EOF2 showing no anomalies in the Arctic Sea region. The most similar pattern to the regressed pattern depicted in figure 1(a) is described in EOF3, which shows a tri-polar pattern with warm SST anomaly over the WNAO; cold over the south of Greenland and Labrador Seas, and warm over the Barents–Kara Seas. The similarity is quite remarkable. As expected by the warm centre over the Barents–Kara Seas in figure 2(e), the PC3 time series shows a significant correlation with the BKSAT time series (corr. = 0.4) at 99% confidence level (figure 2(f)). The PC3 timeseries also has a high correlation coefficient with the SCAND index (corr. = 0.57) (table 1). According to North's rule of thumb, the three EOF modes are well separated (North et al 1982).

Zoom In Zoom Out Reset image size

Table 1. Correlation coefficients among the atmospheric teleconnection modes and the PC time series.

	SCAND	EAWR	NAO
PC1	0.1	0.29	0.37
PC2	0.1	0.48a	0.02
PC3	0.57a	0.10	0.4

^aStatistically significant at p < 0.01.

It is notable that the SST anomaly over the WNAO lies over the northern edge of the Gulf Stream, which shows a strong SST gradient (see isotherms in figure 2(e)). The warm SST anomaly over this region may represent the poleward shift of the Gulf Stream and intense baroclinic zone. Since it is well-known that the SST gradient associated with the western boundary current is known to be a great source of baroclinicity (Minobe et al 2008), it is the source of available potential energy for the growth of transient eddies. This leads us to investigate the role of transient eddies in the large-scale teleconnection pattern that links the North Atlantic and the Arctic regions.

3.3. Physical mechanism of Atlantic origin of Arctic warming

Atmospheric circulation features related with the EOF3 of SST variability are depicted in figure 3. Geopotential height anomaly at 250 hPa representing the upper-level circulation features a wave train pattern emanating from the WNAO toward Eurasia across the north-eastern Atlantic and the Barents–Kara Seas (contour in figure 3(a)). As expected by the similarity between the SST anomaly regressed on to the BKSAT (figure 1(a)) and the EOF third mode (figure 2(e)), this upper-level circulation pattern is similar to the SCAND pattern in figure 1(b). The response is equivalent barotropic (contour in figures 3(a) and (b)) and, therefore, the regressed surface air temperature anomaly (shaded in figure 3(a)) is in general in phase with the upper-level geopotential height anomaly. The significant warming over the Barents–Kara Seas can partly be explained by the enhanced warm advection along the western edge of the anticyclonic anomaly over western Europe induced by this barotropic large-scale anomaly at lower-levels (figure 3(b)).

Zoom In Zoom Out Reset image size

In association with the downstream propagation of SCAND toward East Asia, cold temperature anomalies appear primarily over Central and East Asia, where upper-level cyclonic response dominates (figure 1(b)). In this case, the upper-level cyclonic response reduces the thickness of the air column over East Asia and therefore, the column average temperature drops. Combined with the climatologically strong northerly flow in this region, strong cold advection is induced. The warm and cold anomalies explained above resemble 'Warm Arctic-Cold Continents' or 'Warm Arctic-Cold Siberia' pattern (Overland et al 2011, Inoue et al 2012, Kim et al 2014, Mori et al 2014, Kug et al 2015).

Returning to the North Atlantic, the source region of the wave train seems to lie in the WNAO region (box in figure 2(e)). Compared with the EOF3 in figure 2(e), this wave activity source region coincides with the warm SST anomaly over the WNAO. Sato et al (2014) examined the possible role of the diabatic heating over the WNAO by calculating the apparent heat source and resultant linear stationary eddy response. In this work, we investigated another possibility. The warm SST anomaly over the WNAO can be interpreted as the northward extension of the Gulf Stream (Wu et al 2012) indicating northward shift of the ocean front. Since the WNAO region exhibits strong SST gradients as shown in figure 2(e), we expect that the warm SST anomaly could alter the activities of synoptic-scale eddies which are sensitive to the temperature gradient and diabatic heat sources (Brayshaw et al 2008, Nakamura et al 2008). Indeed, the transient eddy activities estimated by the variance of the 300 hPa daily meridional wind anomaly regressed to EOF3 also shifted eastward compared with its climatological position (figure 3(c)) and the northward shift of Atlantic sub-polar jet occurred at the same time (figure 3(d)). These results are consistent with the previous studies that addressed the importance of the SST gradient in the alteration of transient eddy activities (Sampe et al 2010, Frankignoul et al 2011, Sung et al 2014).

Combined changes in the transient eddy activities and Atlantic sub-polar jet in association with the SST variability over the WNAO hint at the possible role of transient eddy activities on large-scale teleconnection patterns (Bueh and Nakamura 2007, Lim and Kim 2016). To investigate the relative role of transient eddies linked to SST variability over the WNAO, we used the SWM alternatively forced by diabatic heating or transient eddy forcing and estimated the relative importance of each forcing term by comparing the SWM responses forced by each forcing term separately (see supplementary information).

Forcings and associated responses of SWM experiments are represented in figure 4. Within the boxed region of the WNAO, the negative TF_vor in figure 4(a) was consistent with significant high anomalies shown in figure 1(b). Interestingly, the diabatic forcing in figure 4(c) and convergence of transient eddy heat flux compensated each other, meaning that the diabatic heating was largely balanced by eddy heat transport.

Zoom In Zoom Out Reset image size

As noted previously, the WNAO region is a key region of strong SST variability and is associated with the large changes in SST gradient and in storm track. We examined the relative importance of these three forcings in the excitation of large-scale circulation. As shown in figure 4, a major response was obtained with transient eddy vorticity forcing and this forcing reproduced the SCAND wave structure remarkably (compare figure 3(a) and figure 4(a)). In addition, the wave-like feature in the model response had a high pattern correlation with observed SCAND pattern (0.62). A relatively weaker contribution was obtained from the transient eddy temperature forcing and total diabatic heating forcing. Considering that we only applied the forcing in the restricted region (black box in figure 4), the result is rather surprising and confirms the important role of the storm activities in large-scale teleconnection patterns. These results provide evidence that transient vorticity flux related to the SST interannual variability over the WNAO is a key factor for the SCAND teleconnection pattern.

The last evidence of the importance of SST over the WNAO for the Arctic warming comes from fully coupled model experiments (figure 5). In general, model experiments successfully capture various features depicted in the observational analysis results: Model SST response in figure 5(a) shows the warm-cold-warm SST pattern similar to the EOF3 pattern (figure 2(e)). Considering that the SST nudging was only applied to the boxed region in figure 5(a) in the model simulation. Therefore, the warm SST anomaly over the Barents–Kara Seas was internally generated by the fully coupled model as a response. The upper-level response was also reproduced reasonably well (figure 5(b)). Therefore, the results from the regression analysis (figures 1 and 3) are supported by the fully coupled model experiments.

Zoom In Zoom Out Reset image size

Sato et al (2014) showed that the poleward shift of the Gulf Stream influences the increase (decrease) of temperature (sea ice extent) over the Barents–Kara Seas and cooling over Eurasia through planetary waves triggered over the Gulf Stream region. In this study, the origins of the planetary waves are investigated in detail.

First, we show that the variability in the surface air temperature over the Barents–Kara Seas is largely controlled by two dominant SST modes in the domain including the North Atlantic Ocean and the Atlantic sector of Arctic Ocean. The warming trends in both the Atlantic Ocean and the Barents–Kara Seas are largely depicted by EOF first mode and this pattern resembles the basin-wide warming pattern. On the other hand, interannual variability is controlled by the tri-polar SST pattern depicted as EOF third mode in this study. The third SST mode represents the poleward shift of the Gulf Stream and accompanying changes in storm track as indicated by Sato et al (2014).

Through a simple modelling study using SWM, we concluded that the altered upper-level transient eddy vorticity forcing in association with the changes in the storm track plays a major role in the generation of the SCAND pattern and therefore, plays a bridging role between the North Atlantic Ocean and the Atlantic sector of Arctic in early winter at the interannual time-scale. We could reproduce an upper-level circulation pattern that was very similar to SCAND only with altered transient eddy vorticity forcing in the upper-level.

The surface warm advection along the high pressure center of SCAND at the Barents–Kara Seas, which is essentially barotropic, is an important source of warming of BKSAT. The direct influence of the diabatic heating over the WNAO sector was relatively minor compared to the transient forcing.

Although this study emphasizes the importance of the enhanced transient eddy forcing during the warm period of the WNAO, it should be noted that a large portion of the warming is also contributed by the subsequent reduction of sea ice concentration over the Barents–Kara Seas through the enhanced energy fluxes from the Arctic Ocean (figure A1 in supplementary information). However, in this study, we did not conduct any quantitative assessments on that part since we are only interested in the Atlantic origin of the warming.

It is still unknown why the transient eddy activities show those systematic behaviors responding to the specific SST patterns over the North Atlantic Ocean. To deal with this issue, we need to understand how individual Atlantic storms respond to warm SST over the WNAO by tracking storm intensity and its passage (storm track) along the storm. Both systematic changes in storm intensity and track in association with the particular SST pattern over the North Atlantic should collectively contribute to the monthly-timescale transient eddy forcing. We are currently investigating this problem by tracking individual Atlantic storms.

We thank the two reviewers for their helpful comments. This work is conducted to fulfill the task assigned to the project of Korea Polar Research Institute titled 'Development and Application of the Korea Polar Prediction System (KPOPS) for Climate Change and Disasterous Weather Events'. First author, Jung Ok, is partly supported by 'Development of cloud-precipitation Algorithms' project, funded by ETRI, which is a subproject of 'Development of Geostationary Meteorological Satellite Ground Segment (NMSC-2016-01)' program funded by NMSC (National Meteorological Satellite Center) of KMA (Korea Meteorological Administration). Jee-Hoon Jeong is supported by the Korea Meteorological Administration Research and Development Program (KMIPA2015–2091).