1 Introduction

Over the past two decades, Arctic warming, Arctic sea ice loss, and their impacts have been the focus issues (Walsh 2014; Barnes and Screen 2015; Kug et al. 2015; Wu 2017; Cohen et al. 2014, 2018, 2020; Dai and Song 2020; AMAP 2021; Overland et al. 2021; Wu et al. 2022). These variations in the Arctic not only have an important impact on the Arctic ecological environment, but also influence weather events, climate variability, and the ecological environment outside the Arctic through complex interactions and feedback processes (Deser et al. 2004; Honda et al. 2009; Petoukhov and Semenov 2010; Francis and Vavrus 2012; Vihma 2014; Nakamura et al. 2015; Wu et al. 2017; Screen et al. 2018; Chen et al. 2018; Siew et al. 2020; Cohen et al. 2021; AMAP 2021; Ding et al. 2021; Ding and Wu 2021; Wu et al. 2022). Through analyzing winter cases (the winter of 2009/2010 and December 2010), Overland et al. (2011) proposed the concept of the warm Arctic-cold continents. The subsequent study demonstrated that the Barents-Kara Seas and Eastern Siberian-Chukchi Seas are the two crucial sea regions where winter atmospheric warm anomalies are closely associated with cold anomalies in Eurasia and North America, respectively (Kug et al. 2015). Warm Arctic-cold Eurasia generally corresponds to strong cold air activities, which even affect the tropical region of East Asia such as the winter of 2015/2016 (Wu 2017; He et al. 2020; Labe et al. 2020). Previous works generally discussed this specific pattern in terms of warm Arctic favors midlatitudes cooling, while the motivation of the present study is to investigate the cause of winter warm Arctic-cold Eurasia from a new perspective.

Many studies have explored the possible reasons for Arctic warming through multi-spheric interactions and Arctic-midlatitude interactions. It is well known that the North Atlantic warm ocean current brings huge heat to high latitudes and makes the Barents Sea the northernmost open water in the Northern Hemisphere during winter, which may be one possible reason for Arctic sea ice loss and increasing ocean temperatures across the Nordic Seas and the Arctic Ocean (Walsh 2014; Polyakov et al. 2017; Lind et al. 2018; Sung et al. 2018; Tsubouchi et al. 2021). Observational analyses have shown that Arctic sea ice loss, particularly in the Barents-Kara Seas, could enhance East Asian winter monsoon and decrease winter cyclone activities over the Siberian marginal seas, which favors the strengthening of the winter Siberian high and frequent occurrence of cold waves over Eurasia (Wu et al. 1999; Wu and Zhang 2010; Wu et al. 2011, 2013; Inoue et al. 2012; Tang et al. 2013; Chen et al. 2016). The heavy (light) sea ice winter (DJF) in the Barents-Kara Seas generally corresponds to a weakened (strengthened) East Asian winter monsoon, and February cold air activities in China tend to be inactive (active) (Wu et al. 1999). The December light sea ice in the Barents Sea is associated with winter anticyclonic anomaly that prevailed along the Siberian marginal seas, producing anomalous warm advection over the Barents Sea and cold advection over eastern Siberia (Inoue et al. 2012). Arctic sea ice loss in summer and autumn is significantly correlated with a negative Arctic Oscillation (AO) phase, which reduces zonal westerlies over the mid- and high-latitudes of Eurasia, producing the warm Arctic-cold Eurasia pattern (Wu and Zhang 2010).

Some simulation experiments further confirmed that Arctic sea ice loss could give rise to the warm Arctic-cold Eurasia pattern (Newson 1973; Royer et al. 1990; Honda et al. 2009; Petoukhov and Semenov 2010; Wu et al. 2013; Kim et al. 2014; Mori et al. 2014, 2019; Peings and Magnusdottir 2014; Semenov and Latif 2015). In the early 1970s, Newson (1973) investigated the model atmospheric response to the ice-free Arctic and found that the absence of Arctic sea ice would be conducive to the formation of blocking high in the mid- and high-latitudes. Royer et al. (1990) found that the removal of Arctic sea ice in a sensitivity experiment produces moderate cooling over Eurasia. Honda et al. (2009) showed that winter cold anomalies over Eurasia are associated with Arctic sea ice loss in the Siberian marginal seas from September to December. Petoukhov and Semenov (2010) suggested that winter sea ice decrease in the Barents-Kara Seas could cause extreme cold events in Eurasia and further indicated the associated atmospheric response is highly nonlinear. Peings and Magnusdottir (2014) showed that anomalous Rossby waves induced by negative sea ice anomalies penetrate the stratosphere in February, weakening the stratospheric polar vortex, and the corresponding anomalies propagate into the troposphere in the following weeks, leading to cold anomalies over Eurasia. Mori et al. (2019) suggested that sea ice loss in the Barents-Kara Seas significantly enhances the probability of severe winter occurrence over Eurasia, and ~ 44% of the central Eurasian cooling trend for 1995–2014 is attributable to sea ice loss in the Barents-Kara Seas.

Certainly, the Arctic-midlatitude association exhibits big uncertainties, and some studies have shown that their associations are mainly affected by the internal variability of the atmospheric circulation and the tropics (Screen and Simmonds 2011; Wallace et al. 2014; Walsh 2014; Overland et al. 2015, 2021; McCusker et al. 2016; Sun et al. 2016; Warner et al. 2019; Sato et al. 2014; Screen and Francis 2016; Smith et al. 2019; Blackport et al. 2019; Blackport and Screen 2020; Chen et al. 2021; Peings et al. 2021; Rudeva and Simmonds 2021). Additionally, the Arctic-midlatitude linkage also displays strong phasic fluctuations, and Arctic sea ice loss may be conducive to the alternative occurrence of warm Arctic-cold Eurasia (2007/2008–2012/2013) and warm Arctic-warm Eurasia (2013/2014–2018/2019) (Wu et al. 2022).

Although Arctic sea ice loss is the largest contributor to the Arctic surface warming trend (Screen and Simmonds 2010), it accounts for only about 20% of the magnitude of 1000-500-hPa warming in autumn and early winter (Perlwitz et al. 2015). They suggested that the main factors responsible for Arctic tropospheric warming are recent decadal fluctuations and long-term changes in sea surface temperatures (SSTs) outside the Arctic (Perlwitz et al. 2015). Negative SST anomalies in the tropical central and eastern Pacific and the negative phase of the Pacific decadal oscillation (PDO) are favorable for Arctic warming (Ding et al. 2014; Screen and Francis 2016). In addition, atmospheric norward energy transport plays a crucial role in causing Arctic warming (Graversen 2006; Graversen et al. 2008).

Winter warm anomalies around the Barents-Kara Seas are closely associated with the Ural atmospheric blocking anomaly, which has been regarded as the main reason for strong low temperatures and strong cold air activities in Asia (Wu et al. 2013; Mori et al. 2014; Feng and Wu 2015; Kug et al. 2015; Wu 2017; Yao et al. 2017; Peings 2019). However, it is unclear whether winter cold-Eurasia, in turn, contributes to Arctic warm anomalies. None of simulation experiments to date have explored the possible impact of winter cold-Eurasia on Arctic warm anomalies. Investigating this issue will help us to comprehensively understand the association between the Arctic and the mid-latitudes of Eurasia. This study shows for the first time that winter cold-Eurasia is one reason for winter Arctic warm anomalies.

2 Data and methodology

Datasets used in this study include the monthly mean sea level pressure (SLP), 1000-hPa and 500-hPa geopotential height from January 1979 to December 2019, obtained from the NCEP/NCAR reanalysis I. The intensity index of the winter (DJF) Siberian high is defined as the regionally averaged SLP over 80°-120°E and 40°-60°N (Wu and Wang 2002a). The monthly mean vertical integrations of the atmospheric northward and eastward total energy fluxes are obtained from the ERA-Interim reanalysis data from January 1979 to August 2019 (Graversen et al. 2006). The monthly mean North Atlantic Oscillation (NAO) index and Niño3.4 SST index for the period of 1979–2019 are taken from the Climate Prediction Center of the United States. These re-analysis data are used to analyze the observational Arctic-midlatitude association and compare it to the results of simulation experiments. Additionally, monthly Eurasian snow cover area data for the period of 1979–2019 are also used (https://climate.rutgers.edu/snowcover).

To investigate the possible impacts of the strengthened winter Eurasian cold anomalies on winter atmospheric circulation, particularly the impact on Arctic warm anomalies, the Community Atmosphere Model version 4.0 (CAM4), its version 5.4 (CAM5.4), and the Specified Chemistry Whole Atmosphere Community Climate Model version 4.0 (SC-WACCM4) are used in this study. Three models have the identical horizontal resolution of 1.9° in latitude and 2.5° in longitude, and their vertical resolutions have 26, 30, and 66 levels, respectively. More details on three models can be found in Neale et al. (2010), Neale et al. (2012), and Smith et al. (2014). The control runs for CAM4, CAM5.4 and SC-WACCM4 were forced by the climatological monthly SSTs and sea ice concentrations (SICs) averaged over 1982–2001 (model-derived data) (Hurrell et al. 2008). The greenhouse gas (GHG) emissions and aerosol forcing hold constant at the year 2000 level. The integration times of the control runs are 60 years for each model.

The sensitivity experiments for CAM4 were performed with the same boundary condition (SSTs and SICs) as the control runs, but with an enhanced ground albedo in the Community Land Model (CLM) over the region of 50°-60°N and 50°-100°E from December to February of the next year. More specifically, the ground albedo is artificially raised to 0.8 in the sensitivity experiments, which directly generates Eurasian cooling mainly through reducing net solar radiation at the surface. The sensitivity and control experiments for CAM4 were carried out using the same initial conditions, the model-derived January 1 datasets. The sensitivity experiments for CAM5.4 and SC-WACCM4 are the same as that in CAM4, but with different/same ground albedo values over different/same regions from December to February. Similarly, the CAM5 and WACCM-SC experiments employed the respective model-derived initial conditions of January 1 for integration. Table 1 summarizes all sensitive experiments. The purpose of setting up these simulation experiments is to demonstrate that simulated winter Arctic warm anomalies do not depend on a specific model. The atmospheric circulation anomalies associated with the perturbed winter ground albedo are defined as the difference in the ensemble means between the sensitivity experiments and control runs since the 11th year (simulation experiments for the first 10 years are as spin up). The signal-to-noise ratio (SNR) is defined as the variability induced by the prescribed albedo forcing divided by the model internal variability, more details on the calculation method refer to our study (Wu et al. 2022).

Table 1 Description for sensitivity experiments with different models
Full size table

3 Dominant features of warm anomalies over the Barents-Kara Seas

Winter mean atmospheric thickness (1000-500-hPa) over the Barents-Kara Seas displays a strong interannual variability superposed on a phasic shift (Fig. 1a). Negative thickness anomalies were dominant before the winter of 2003/04 and were replaced by frequently positive anomalies, consistent with the phase change in winter Arctic warming shown in Fig. 1b of Wu (2017). This phasic shift has been detected in Arctic winter surface air temperature (SAT) variability (Feng and Wu 2015). It is seen that strong warming anomalies (σ > 1.0) occurred in the winters of 1994/95, 2004/05, 2007/08, 2011/12, 2015/16, and 2017/18. Spatially, positive thickness anomalies occupy the most of the Arctic and high-latitudes of Eurasia, with negative anomalies in the mid- and low-latitude Asian continent and northern parts of the African continent (Fig. 1b). In contrast to East Asia, positive thickness anomalies also appear over the most of the northern North Atlantic and North America. Positive SLP anomalies occupy the mid- and high-latitude Eurasia and most of the North Atlantic and the North Pacific. In the northern North Atlantic and the entire Arctic Ocean, SLP anomalies are negative with a center between Iceland and Svalbard Islands, suggesting the strengthened Icelandic low moves to the northeast (Fig. 1c). SLP anomalies reflect the Arctic dipole pattern between the Arctic and Eurasia to a great extent. Consequently, accompanied by warm anomalies over the Barents-Kara Seas, both the Siberian high and the positive phase of the NAO are strengthened simultaneously. Figure 1c closely resembles SLP anomalies in the winter of 2011/2012 (see Fig. 6a of Wu et al. 2017) when negative SAT anomalies appeared in the mid- and low-latitude Eurasia with positive SAT anomalies centered on the Barents-Kara Seas (not shown).

Fig. 1
figure 1

(a) Normalized time series of the regionally (60°-80°N and 30°-90°E, simply indicated as the Barents-Kara Seas) averaged winter (DJF) mean 1000-500-hPa thickness (blue line), red dashed lines represent the means averaged over 1979/80-2003/04 and 2004/05-2018/19, respectively. (b) Winter 1000-500-hPa thickness anomalies (gpm), derived from a linear regression on the time series shown in Fig. 1a. Green contours denote thickness anomalies at the 95% confidence level. (c) Same as in (b), but for winter mean SLP anomalies (hPa)

Full size image

The correlation coefficient of the winter atmospheric thickness over the Barents-Kara Seas with the Siberian high intensity (Fig. 2a) (the NAO, not shown) is 0.59 (0.57). Thus, the tropospheric temperature variability over the Barents-Kara Seas is directly related to the atmospheric circulation anomalies in the mid- and high-latitudes of the Northern Hemisphere. The winters with the Siberian high intensity > 1σ included 1983/84, 1985/86, 2004/05, 2005/06, 2007/08, 2011/12, 2015/16, 2017/18, and 2018/19, and five of which (2004/05, 2007/08, 2011/12, 2015/16, and 2017/18) corresponded to winter warm anomalies (> 1σ) (Fig. 1a). The correlation coefficient between the atmospheric thickness and the Siberian high intensity (the NAO) becomes 0.29 (0.45) after removing the winters with the Siberian high intensity > 1σ. Similarly, after removing the seven winters with the NAO intensity > 1σ (1988/89, 1994/95, 1999/00, 2011/12, 2014/15, 2015/16, 2017/18) (not shown), the correlation coefficient between the atmospheric thickness and the Siberian high intensity (the NAO) becomes 0.44 (0.46). Consequently, the NAO displays a closer linear association with warm anomalies over the Barents-Kara Seas relative to the Siberian high, while the influence of the Siberian high mainly emerges in its strengthened phase, corresponding to warm anomalies over the Barents-Kara Seas. This study focuses on the association between the Siberian high and winter warm anomalies over the Barents-Kara Seas.

Fig. 2
figure 2

(a) Normalized time series of the winter Siberian high intensity averaged over 40°-60°N and 80°-120°E (blue line), and the red dashed line represents one standard deviation. (b) Winter SAT anomalies (°C), derived from the mean averaged over nine winters with the Siberian high intensity > 1 σ (1983/84, 1985/86, 2004/05, 2005/06, 2007/08, 2011/12, 2015/16, 2017/18, and 2018/19) minus the winter mean averaged over 1979 to 2019. The green contours denote SAT anomalies at the 95% confidence level. (c)-(e) As in (b), but for 1000-500-hPa thickness (gpm), SLP (hPa), and 500-hPa geopotential height (gpm), respectively. In (c) and (d), the black dashed (solid) contours denote the winter means averaged over 1979 to 2019 (nine winters with the Siberian high intensity > 1 σ)

Full size image

Figure 2a shows that the strengthened Siberian high frequently occurred since the winter of 2004/05, consistent with enhanced winter Arctic warming (Feng and Wu 2015; Wu 2017). Spatially, SAT and 1000-500-hPa thickness anomalies consistently show a warm Arctic-cold Eurasia pattern, and anomalous warm centers are located around the Barents-Kara Seas (Fig. 2b and c). The strengthened Siberian high corresponds to a systematically northward shifting of the 1000-500-hPa iso-thickness lines in most of the domain, producing positive thickness anomalies (Fig. 2c). A striking northward migration occurs in northern Eurasia and the Siberian marginal seas. Over nearly the whole North American continent, most of the North Atlantic, and the Arctic Ocean, a systematic northward extension of positive thickness anomalies is concurrent with active cold mass into Eurasia, featured by a southward retreat of the iso-thickness lines and negative thickness anomalies over Eurasia. The strengthened Siberian high (> 1σ) corresponds to positive SLP anomalies occupying the most of the mid- and high-latitudes and a systematic northward shift of SLP over the North Pacific, Eurasia, and the North Atlantic sector (Fig. 2d). The strengthening of the Siberian high is closely associated with an enhancement of positive height anomalies around the Ural Mountains (Fig. 2e).

Figure 3 shows the vertical integrated atmospheric total energy flux anomalies associated with the Siberian high. The strengthened Siberian high favors increases in the atmospheric norward total energy flux over the most of Europe, the Nordic Seas, and the Barents-Kara Seas (Fig. 3a). Simultaneously, significant increases in the atmospheric eastward total energy flux also occur over the Kara Sea (Fig. 3b). The estimated contribution from the Siberian high to the standard deviation of the atmospheric total energy flux over the Barents-Kara Seas exceeds 40% (Fig. 3c and d). Thus, the mid- and high-latitudes of Eurasia may be a source contributing to Arctic warm anomalies. It is very likely that cold-Eurasia contributes to Arctic warm anomalies through the strengthening of the Siberian high.

Fig. 3
figure 3

(a) Winter atmospheric northward total energy flux anomalies (108 W m− 1), derived from a linear regression on the normalized Siberian high intensity index. The green contours represent anomalies at the 95% confidence level. (b) As in (a), but for atmospheric eastward total energy flux. (c) The standard deviations of the atmospheric northward total energy flux variability induced by the Siberian high divided by the standard deviations of the atmospheric northward total energy flux varibility. (d) As in (c), but for the atmospheric eastward total energy flux

Full size image

4 Simulated impacts of cold-Eurasia on the Arctic

In this section, simulation experiments are conducted to demonstrate the impact of cold-Eurasia on the Arctic, which has been ignored in previous studies. Figure 4 shows differences in the ensemble means between the ground albedo forcing experiments and the control runs, derived from CAM4 experiments. Increases in winter regional ground albedo directly cause significant surface cold anomalies in most of the Asian continent and significant warm anomalies in the Arctic, including parts of the Barents-Kara Seas, the Arctic Ocean, and northern and northeastern Canada (Fig. 4a). Additionally, warm anomalies also are seen over Northern Africa and the mid-to-lower latitudes of Eurasia. Arctic surface warm anomalies can extend upward into the mid-troposphere (Fig. 4b). Compared to the surface warm anomalies, positive 1000-500-hPa thickness anomalies even are more robust over the Arctic, while the center of negative thickness anomalies migrates eastward to Northeastern Asia. Winter Eurasian regional cold anomalies also significantly strengthen the Siberian high and the Aleutian low (Fig. 4c). The spatial distribution of SLP anomalies resembles the negative phase of the Arctic Oscillation (AO), but the positive SLP center appears over around the Barents-Kara Seas. At 500 hPa, height anomalies display a tripole structure over Eurasia and the Arctic, and the strengthened positive height anomalies over the Ural Mountains are dynamically consistent with the enhanced Siberian high (Fig. 4d). Consequently, CAM4 experiments clearly demonstrate that winter Eurasian cold anomalies directly cause warm anomalies over the Arctic and the strengthening of both the Siberian high and positive height anomalies over the Ural Mountains. Additionally, winter Eurasian cold anomalies indirectly cause warm anomalies in the high-latitudes of North America.

Fig. 4
figure 4

(a) Differences in ensemble means of winter SATs (°C) between CAM4 sensitivity experiments forced by prescribed Eurasian regional (red box: 50°-60°N and 50°-100°E) ground albedo during wintertime and the control runs, and green contours represent differences at the 95% confidence levels. (b)-(d) As in (a), but for 1000-500-hPa thickness (gpm), SLP (hPa), and 500 hPa geopotential height (gpm)

Full size image

The result that cold-Eurasia causes Arctic warm anomalies is independent of the used models, as shown in Fig. 5. Distinct sensitivity experiments consistently support this conclusion, but with different locations and amplitudes of Arctic warm anomalies (Fig. 5a, e, i, b, f, and j). Simulated Arctic warm anomalies in three simulation experiments consistently occur from eastern Greenland counterclockwise to the Far East rather than northern North America, differing from CAM4 results (Fig. 4a and b). Among these three experiments, SLP and 500-hPa geopotential height anomalies also exhibit great differences over Eurasia and the Arctic, except for the apparent common feature with positive SLP anomalies occupying most of Eurasia (Fig. 5c, g, k, d,  h, and l). WACCM experiments clearly indicate that although a cold-Eurasia enhances the Siberian high (Fig. 5k), it fails to produce positive height anomalies around the Ural Mountains. Consequently, locations and amplitudes of simulated winter Arctic warm anomalies depend on the used models, the prescribed albedo forcing regions, and albedo values.

Fig. 5
figure 5

(a) Differences in ensemble means of winter SATs (°C) between CAM5_1 sensitivity experiments forced by prescribed Eurasian regional (red box: 50°-60°N and 60°-110°E) ground albedo during wintertime and the control runs, and green contours represent differences at the 95% confidence levels. (b)-(d) As in (a), but for 1000-500-hPa thickness (gpm), SLP (hPa), and 500 hPa geopotential height (gpm). (e)-(h) and (i)-(l) As in (a)-(d), but for derived from CAM5_2 and WACCM simulation experiments, respectively. In Fig. 5e (5i), the red box is the prescribed ground albedo forcing region in 40°-50°N and 60°-110°E (50°-60°N and 50°-100°E)

Full size image

Compared to observational associations (Fig. 2), all simulation experiments well reproduce some major features in the observational associations, including Arctic warm anomalies as shown in the surface and 1000-500-hPa thickness (Figs. 4 and 5). Particularly, CAM4 experiments capture Arctic warm anomalies centered around the Barents-Kara Seas, the strengthening of the Siberian high, positive height anomalies around the Barents-Kara Seas and the Ural Mountains, and the deepening of the Aleutian low (Fig. 4). This implies that winter warm anomalies over the Barents-Kara Seas cannot be simply regarded as the reason for winter Eurasian cold anomalies. Additionally, observed Arctic warm anomalies mainly appear over the Barents-Kara Seas instead of northeastern and northern North America (Fig. 2b and c) and well coincide with simulation results in Fig. 5, but differ from that in Fig. 4. Although four simulation experiments consistently show positive SLP anomalies in most of the mid- and high-latitude of Eurasia, they fail to produce observed SLP anomalies over the northern North Atlantic and the Greenland Sea, which resemble the positive NAO (Fig. 2d vs. Figures 4c and 5c, g, and k). Observed two separated positive centers over northern Eurasia and the North Atlantic shown in SLP and 500-hPa geopotential height (Fig. 2d and e) are replaced by one monopole center in CAM4 simulation experiments (Fig. 1c and d). Besides the defects of the model itself, the lack of Arctic sea ice change and the associated air-sea ice interaction may be another important reason for some simulation deviations and differences between simulations and observations.

Model results confirm the possible association from cold-Eurasia to Arctic warm anomalies as shown in Figs. 4 and 5, however, possible mechanisms need further exploration. Figure 6 shows the pressure-longitude cross section along 52.10°N, including temperature, pressure vertical velocity, and geopotential height anomalies superposed on the wave activity flux. It is seen that cooling dominates most of the troposphere below 200 hPa over Eurasia, and the maximum negative anomalies appear in the lower-troposphere over Asia, reflecting the impact of increased ground albedo (Fig. 6a). Influenced by the cooling, downward motion anomalies nearly occupy most of the troposphere over Asia with a center in the mid- and lower-troposphere over the Ural Mountains (Fig. 6b) and dynamically contribute to the enhancement of the Siberian high (Fig. 6c), consistent with the results of Ding and Krishnamurti (1987) and our previous study (see Fig. 9 of Wu and Wang2002b). Meanwhile, the strengthening of the Siberian high triggers upward atmospheric energy propagation in the troposphere and the stratosphere over Asia (Fig. 6c). Strong horizontal energy propagation in the mid- and high-troposphere spreads the impact of cold-Eurasia to remote regions. The enhancement of the Siberian high is generally concurrent with positive height anomalies around the Ural Mountains through downward motion anomalies and atmospheric energy propagations. The strengthening of the Siberian high and its northward migration transport more warm and moist air into the Arctic, leading to warm anomalies over the Barents-Kara Seas (Fig. 3).

Fig. 6
figure 6

(a) The pressure-longitude cross section along 52.10°N of differences in ensemble means of temperatures (°C) between CAM4 sensitivity experiments and the control runs. The green contours denote differences at the 95% confidence level. (b) Same as in (a), but for pressure vertical velocity (pa s-1; multiplied by 1000.0). (c) Same as in (a), but for wave activity flux (m2 s-2, the vertical component of wave activity flux was multiplied by 100.0 ) superposed on geopotential height anomalies (gpm).

Full size image

It is necessary to know the relative contributions induced by the prescribed ground albedo forcing to the model internal variability. We still focus on CAM4 experiments. Figure 7 shows the spatial distribution of the simulated SNR by CAM4 experiments. For SAT variability, regions with large SNR (> 0.3) mainly emerge in the Arctic and Asia (Fig. 7a). For 1000-500-hPa thickness variability, large SNR appears over East Asia to the northern North Pacific, northeastern and northern Canada to the Barents-Kara Seas, and the Gulf of Alaska (Fig. 7b), showing more organized spatial distribution relative to SAT variability. For SLP and 500-hPa geopotential height variability, large SNR occupies northern Eurasia, the Nordic Seas, the Barents-Kara Seas, the northern North Pacific, and eastern North America (Fig. 7c and d), indicating that cold-Eurasia can even affect atmospheric circulation over North America rather than merely dominating Eurasia and the neighboring area.

Fig. 7
figure 7

(a) SNR of simulated winter SAT variability by CAM4, derived from the variability induced by prescribed ground albedo forcing divided by the model atmospheric internal variability. (b)-(d) As in (a), but for 1000-500-hPa thickness, SLP, and 500 hPa geopotential height

Full size image

5 Discussions

To further confirm the stability of CAM4 simulation experiments, the integration times are extended from 60 years (see Table 1) to 120 years for both the sensitivity experiments and control runs, and the simulated differences are shown in Fig. 8. The extended simulation experiments again reproduce Arctic warm anomalies in SAT and 1000-500-hPa thickness (Fig. 8a and b). Two separated warm centers are respectively located over the Barents-Kara Seas and northwestern North America. Compared to Fig. 4a and b, warm anomalies are obviously weaker in northeastern and northern North America (Fig. 8a and b). Meanwhile, cold anomalies emerge in the mid to high-latitudes of Eurasia. An enhanced Siberian high and deepened Aleutian Low jointly reappear (Fig. 8c), closely resembling that shown in Fig. 4c. Simulated winter 500-hPa geopotential height anomalies (Fig. 8d) also support the conclusion in the 60 years simulation experiments (Fig. 4d). Consequently, for CAM4 simulation experiments, 60 years of integrations are long enough.

Fig. 8
figure 8

Same as in Fig. 4, but derived from 120 years simulation experiments by CAM4. The red box refers to 50°-60°N and 50°-100°E

Full size image
Fig. 9
figure 9

Same as in Fig. 8, but derived from CAM5 120 years simulation experiments

Full size image

We also used CAM5.4 to repeat the CAM4 120 years simulation experiments, and similar results were reached (Fig. 9). Although dominant features of the SLP pattern (Fig. 9c) are similar to Fig. 8c, the CAM5 experiments fail to produce a stronger Siberian high relative to the control runs. Compared to Fig. 4, amplitudes of anomalies in Figs. 8 and 9 are obviously weaker, reflecting the impacts of the model internal variability.

Winter SLP anomalies shown in Fig. 1c, to a great extent, resemble winter SLP patterns in Wu (2017) that were derived from the second EOFs of winter mean SLP variability over different domains, and these patterns consistently correspond to an Arctic warm pattern identified by the leading EOF of winter 1000-500-hPa atmospheric thickness variability north of 60°N. Similar to Wu (2017), this study extracts the first two leading patterns of winter mean SLP variability, derived from EOF analysis over the domain north of 20°N for the period 1979–2019, and they respectively account for 32% and 13% of the variance. The leading pattern represents the AO (not shown), and the second pattern dominantly reflects an Arctic warm pattern, as shown in Fig. 10. Negative phases frequently occurred before the winter of 2004/2005 and were then replaced by frequent positive phases afterwards (Fig. 10a). Although SLP anomalies in Figs. 1c and 10b show a similar spatial distribution, their differences are also apparent. The most significant positive center emerges over the Northern North Pacific (Fig. 10b) rather than over the mid-latitude North Atlantic or northern Eurasia (Figs. 1c and 2d). Amplitudes of negative SLP anomalies over the Arctic in Fig. 10b are obviously weaker relative to that in Fig. 1c. At 1000-500-hPa thickness fields, differences in thickness anomalies are dominant over the Northern North Pacific, the northern and northeastern North America (Figs. 1b and 10c). Since the correlation coefficient between the two time series in Figs. 1a and 10a is 0.38, winter warm anomalies over the Barents-Kara Seas (Fig. 1) cannot be replaced by the second SLP pattern (Fig. 10). The second SLP pattern dominantly reflects atmospheric internal variability and may be modified by ENSO events due to a high correlation between the second SLP pattern and the Niño3.4 SST index (r=–0.76). Consequently, from the large-scale dynamic perspective, winter Arctic warm anomalies, to a great extent, may be attributed to the jointed influences of both atmospheric internal variability and the tropic SST forcing, consistent with previous studies (Wallace et al. 2014; Ding et al. 2014; Screen and Francis 2016; Warner et al. 2019). On the other hand, from the thermal dynamic and local-scale viewpoints, it is difficult to attribute Arctic warm anomalies over the Barents-Kara Seas to the influence of the tropic SST forcing because the correlation coefficient between Niño3.4 SST index and winter atmospheric thickness over the Barents-Kara Seas is -0.07.

Fig. 10
figure 10

(a) Normalized PC2 time series of the second EOF of winter mean SLP variability north of 20°N. (b) Winter SLP anomalies (hPa), derived from a linear regression on the normalized PC2 time series, and green contours denote SLP anomalies at the 95% confidence level. (c) Same as in (b), but for 1000-500-hPa thickness anomalies (gpm)

Full size image

Eurasian snow cover has shown increasing trends over the past decades (Fig. 11) and is likely to cause Eurasian cold anomalies due to its high albedo. Autumn snow cover shows apparent phasic fluctuations superposed on a linear trend, and more snow cover frequently occurred after 2008 (Fig. 11a). In winter, more snow cover dominantly emerged in winters of 2002–2013 (Fig. 11b), which well corresponds to the strengthening of the winter Siberian high (Fig. 2a) and prevailed warm Arctic-cold Eurasian pattern (Wu et al. 2022). Winter SLP and SAT anomalies are closely associated with the winter Eurasian snow cover area (Fig. 11c and d). It is seen that positive SLP anomalies occupy northern Eurasia and the Siberian marginal seas, leading to strengthening of the Siberian high. The correlation coefficient between the Siberian high intensity and winter Eurasian snow cover area is 0.32 at the 0.05 significance level. Winter SAT anomalies display a warm Arctic-cold Eurasian pattern, with the maximum positive center over the Barents-Kara Seas. Consequently, winter Eurasian snow cover is a significant factor in producing the warm Arctic-cold Eurasia pattern.

Fig. 11
figure 11

Normalized (a) autumn and (b) winter Eurasian snow cover area time series (blue bar) and its linear trend (dashed purple line). Winter (c) SLP (hPa) and (d) SAT (°C) anomalies, derived from a linear regression on normalized winter Eurasian snow cover area time series, green contours denote anomalies at the 95% confidence level

Full size image

This study focuses on atmospheric circulation variability in the troposphere induced by prescribed regional ground albedo forcing, thus, it remains unclear whether the stratosphere-troposphere interaction enhances the impacts of the ground albedo on Arctic warm anomalies. Additionally, it is also urgent to know whether a Eurasian cold wave process also induces Arctic warm anomalies on the intraseasonal-time scales. These questions need to be investigated in the future. It should be pointed out that the observational associations are independent of the NCEP/NCAR reanalysis I, and very similar results are also derived from ERA-Interim reanalysis data.

6 Conclusions

This study explored the reason for winter warm anomalies around the Barents-Kara Seas from a new perspective. Observational analyses show that winter warm anomalies in the mid- and lower-troposphere over the Barents-Kara Seas display strong interannual variability superposed on a phase shift in the beginning 2000s, indicating strengthening of the Arctic local warm anomalies afterward.

Arctic warm anomalies around these sea regions are closely associated with the strengthening of both the Siberian high and the NAO. The strengthened Siberian high well corresponds to a systematic northward shift of winter atmospheric circulation over northern Eurasia, the northern North Atlantic, and northern North America, transporting more atmospheric total energy into the area over the Barents-Kara Seas.

Simulation experiments forced by prescribed winter regional ground albedo in Eurasia demonstrate that at interannual time scales, a regional cooling directly causes Arctic warm anomalies, and spatial distributions and amplitudes of Arctic warm anomalies depend on the used models and the locations and strength of albedo forcing. The cooling sometimes strengthens the Siberian high and generates positive height anomalies over the Ural Mountains (or Ural blocking high anomalies) through large-scale atmospheric subsidence anomalies and energy propagations, causing warm anomalies around the Barents-Kara Seas. The increased winter Eurasian snow cover is one of the important factors in strengthening the regional ground albedo, which produces the warm Arctic-cold Eurasian pattern. Thus, winter warm anomalies over the Barents-Kara Seas cannot simply be regarded as the reasons for cold Eurasia.