1 Introduction

The Arctic surface has warmed at a rate approximately twice the Northern Hemisphere average in recent decades, which is termed “Arctic amplification” (Cohen et al. 2014; Feng and Wu 2015; Francis and Vavrus 2012; Gao et al. 2015). It is generally acknowledged that the recent Arctic warming is strongly related to declining Arctic sea ice (Luo et al. 2016; Onarheim and Årthun 2017; Onarheim et al. 2015; Outten and Esau 2012; Screen et al. 2012; Smedsrud et al. 2013) and retreating snow cover (Screen and Simmonds 2010; Xu et al. 2018a). Atmospheric poleward heat and moisture transport (Chen et al. 2017; Luo et al. 2017; Screen et al. 2012; Woods and Caballero 2016), and local Arctic processes such as changes in water vapour and cloud cover (Graversen and Wang 2009) as well as sulphate aerosol and black carbon concentrations (Shindell and Faluvegi 2009), are other factors contributing to the recent Arctic warming.

Arctic surface warming has been suggested to influence winter temperature at midlatitudes, especially over Eurasia, forming the “warm Arctic-cold Eurasia” pattern (Cohen et al. 2014; Inoue et al. 2012; Kim et al. 2014; Liu et al. 2012; Luo et al. 2016; Wang and Liu 2016; Xu et al. 2018a; Zhu et al. 2018). Overland et al. (2011) found a strengthened linkage between the warm Arctic and the midlatitude climate, manifested in the form of higher geopotential heights in the Arctic accompanied by enhanced meridional winds. A weakening of the midlatitude zonal mean flow due to reduced poleward temperature gradients associated with surface warming in the Barents-Kara Seas, contributes to the observed cooling trends across the midlatitude Eurasia (Luo et al. 2016; Outten and Esau 2012). Three possible mechanisms through which the Arctic warming could induce midlatitude weather extremes are changes in storm track, jet stream, and planetary waves (Cohen et al. 2014). In terms of the East Asian winter monsoon (EAWM) system, several studies have shown that a warm Arctic increases the geopotential height aloft, enhancing the upper-level ridge near the Urals and thus intensifying the Siberian high (Cheung et al. 2013; Cohen et al. 2014; Kug et al. 2015; Overland et al. 2015; Takaya and Nakamura 2005). However, other studies have disputed this, and attributed the winter climate change over Eurasia to internal atmospheric variability rather than to Arctic sea ice loss or Arctic warming (McCusker et al. 2016; Ogawa et al. 2018; Sorokina et al. 2016; Sun et al. 2016).

The Siberian high is one of the most distinct climate features in the Northern Hemisphere winter (He and Wang 2012; Jeong et al. 2011; Li et al. 2015). The expansion of the Siberian high is closely related to a strengthened EAWM, bringing more cold-air to East Asia (Bueh et al. 2011; He 2015; He and Wang 2013; Hu et al. 2015; Jin et al. 2016; Li and Wang 2013; Li and Yang 2010; Li et al. 2018; Wang et al. 2010; Xu et al. 2018b). The strong pressure gradient between the Siberian high and the Aleutian low gives rise to intensified northerly winds along the east flank of the Siberian high and hence causes severe winters over East Asia (Overland et al. 1999).

Planetary-scale downward motion and strong radiative cooling are thought to promote the formation and development of the Siberian high (Ding 1990; Ding and Krishnamurti 1987). More autumn Eurasian snow cover and subsequent enhanced winter surface cooling seem to have strengthened the Siberian high in recent decades, reversing a declining trend prior to the 1990s (Jeong et al. 2011; Li and Gao 2015; Panagiotopoulos et al. 2005). Much research effort has been spent in understanding the relationship between the Siberian high and changes in the Arctic properties (He and Wang 2013; Jeong et al. 2011; Jhun and Lee 2004; Li et al. 2014). For example, the negative phase of the Arctic Oscillation (AO) is linked to stronger descending motion over Siberia, exerting significant impacts on the intensity of the Siberian high (Wu and Wang 2002). Kug et al. (2015) found a relationship between surface warming in the Barents-Kara region and the strength of the Siberian high in the period 1979/1980–2013/2014, and He et al. (2016) attributed cold conditions over East Asia in January 2016 to the extreme warming in the Arctic and the concurrency of the enhanced the Siberian high.

In recent decades, there has been a striking warming trend over the Arctic (Francis and Vavrus 2015; Kug et al. 2015). As suggested by previous studies (Kumar et al. 1999; Li et al. 2014; Screen and Francis 2016), interdecadal change in background state may lead to shifts in teleconnection patterns. Notable surface warming has occurred over the Barents-Kara Seas since the late-1990s (Fig. 1a), a result of delayed ice freezing due to weaker ocean stratification in the area (Lind et al. 2018). At the same time, the Siberian high has strengthened significantly (Fig. 1b).

Fig. 1
figure 1

a The climatology of winter SAT during 1979–2017 (contours) and the differences in winter SAT between 1997–2017 and 1979–1996 (shading). b Same as a, but for SLP. The boxes in a and b mark the regions where the ARTI_2m and SHI are defined, respectively

Full size image

Motivated by the analysis of Kug et al. (2015), this study investigates the time-varying linkage between the Eurasian midlatitude circulation and Arctic climate variability in order to gain new insights into the part of the climate system that is relevant for Arctic and midlatitude climate predictions.

2 Data and methods

The atmospheric datasets employed in this study include monthly surface air temperature (SAT), sea level pressure (SLP), air temperature, winds and vertical velocity at all levels obtained from the European Centre for Medium-Range Weather Forecasts (ECMWF) Interim Reanalysis (ERA-I) with a resolution of 2.5° × 2.5° from 1979 to 2018 (Dee et al. 2011). SLP with a resolution of 5° × 5° from the Met Office’s Hadley Center (version of HadSLP2r) is employed to support the results derived from ERA-I reanalysis (Allan and Ansell 2006).

The Arctic surface temperature index (ARTI_2m) used in this study is defined as the area-averaged SAT over the Barents-Kara Seas (30°–70°E, 70°–85°N; the black frame in Fig. 1a), and the Arctic mid-tropospheric temperature index (ARTI_500) is defined as the area-averaged temperature at 500 hPa over the same area. The Siberian high index (SHI) is defined as the area-averaged SLP in the region of 80°–120°E, 40°–65°N (the black frame in Fig. 1b) (Jeong et al. 2011). A winter of a particular year refers to December that year and January and February next year. To focus on the interannual variability, all data are linearly de-trended before the regression and correlation analyses.

We utilize four uncoupled (atmosphere-only) climate models, each forced by daily-varying SST and sea ice in the period of 1982–2014 (Ogawa et al. 2018). The models are the Community Atmosphere Model, version 4 (CAM4, 0.9° × 1.25° with 26 vertical levels up to 3 hPa) (Neale et al. 2013), the Whole Atmosphere Community Climate Model (WACCM, 0.9° × 1.25° with 66 vertical levels up to 0.000006 hPa) (Marsh et al. 2013), and the Integrated Forecast System (IFS, T255 with 91 vertical levels up to 0.01 hPa) (Balsamo et al. 2009) with twenty independent runs each, and the fourth generation atmospheric general circulation model developed at the Institute of Atmospheric Physics, Chinese Academy of Science (IAP4, 1.4° × 1.4° with 26 vertical levels up to 10 hPa) (Dong et al. 2012) with ten independent runs. All experiments are forced with the Coupled Model Intercomparison Project Phase 5 (CMIP 5) historical forcing for 1982–2005 and Representative Concentration Pathway 8.5 scenarios (RCP 8.5) for 2006–2014. The surface boundary forcing involves observed daily SST and sea ice taken from NOAA OISST (Reynolds et al. 2007).

3 Interdecadal variation in the relationship between midlatitude atmospheric circulation anomalies and Arctic temperature variability

Figure 2a shows the temporal evolutions of the undetrended winter ARTI_2m and its 7-year running mean during 1979–2017. One of the most striking features is the interdecadal transition of Arctic surface temperature anomaly, with a steep warming trend after the late-1990s as pointed out by Kug et al. (2015) and also found in the water below (Lind et al. 2018). The linkage between Eurasian midlatitude climate and Arctic surface temperature variability (Kug et al. 2015), with an enhanced Siberian high and an anomalous anticyclonic circulation in western Russia associated with warm anomalies in the Barents-Kara Seas, is indicated by the significant positive correlations between the ARTI_2m and Siberian high pressure (Fig. 3a, d). The strengthened linkage between Arctic surface warming and midlatitude atmospheric circulation around the late-1990s (Fig. 2d) is investigated in more detail in the following analysis.

Fig. 2
figure 2

Normalized time series of a undetrended winter ARTI_2m (bars) and its 7-year moving averages (solid lines) and detrended winter b ARTI_2m and c SHI during 1979–2017. d The 15-year sliding correlation coefficients between the detrended ARTI_2m and SHI during 1979–2017. The solid lines indicate the values significant at the 95% and 99% confidence levels

Full size image
Fig. 3
figure 3

a Regressions of winter SLP (shading) and UV850 (vectors) with respect to the ARTI_2m derived from ERA-I reanalysis data during a 1979–2017, b 1979–1996, and c 1997–2017, respectively. d–f Same as a–c, but for SLP derived from Met Office’s Hadley Centre data during d 1979–2016, e 1979–1996, and f 1997–2016, respectively. Areas with correlations exceeding the 95% confidence level are dotted. The vector winds are above the 95% confidence level

Full size image

Winter SLP and UV850 are regressed onto the ARTI_2m for the two periods 1979–1996 and 1997–2017 (Fig. 3b, c). During 1979–1996, there are positive but mostly insignificant SLP anomalies largely confined to western Siberia reaching 1.5 hPa (Fig. 3b; shading). Accordingly, at 850 hPa, the anomalous anticyclonic flow over western Siberia-eastern Europe for 1979–2017 (Fig. 3a; vectors) is replaced by anomalous southeast and east airflows on the west of the Lake Baikal (Fig. 3b; vectors). It seems that changes in Arctic surface temperature have a weaker and less significant relationship with the midlatitude atmospheric circulation during 1979–1996, relative to that during 1979–2017. For the period 1997–2017, a significant anomalous high pressure is located over the east of the Ural Mountains reaching 3 hPa (Fig. 3c; shading). Also, a significant anticyclonic anomaly appears over northern Eurasia (Fig. 3c; vectors). Comparison of circulation anomalies between the two periods demonstrates the strengthening of the linkage between the Arctic surface temperature and the Siberian high in the latter period. The results based on ERA-I reanalysis (Fig. 3a–c) resemble closely those based on observations from Hadley Center (Fig. 3d–f), supporting the robustness of the strengthened midlatitude-Arctic linkage.

Generally, the Siberian high is closely linked to upper-tropospheric circulation (Jhun and Lee 2004; Panagiotopoulos et al. 2005; Yang et al. 2002). The 300 hPa zonal wind (shading) and wave activity flux (vectors) (computed according to Takaya and Nakamura 2001) are therefore regressed onto the ARTI_2m during 1979–2017, 1979–1996, and 1997–2017 (Fig. 4a–c). The prevailing westerly winds encircling the Arctic show a remarkable weakening during the Arctic surface warm years for all three periods (Fig. 4a–c; shading), which might be related to the decreased meridional temperature gradient associated with Arctic warming through the thermal wind relation (Francis and Vavrus 2015; Overland et al. 2015). The magnitude of the anomalous easterly winds during 1997–2017 is slightly larger than that for 1979–1996 (Fig. 4b, c; shading). Meanwhile, there is an accelerated and westward-extended East Asian jet stream after the late-1990s (Fig. 4c; shading). The “negative-positive” anomaly structure of the 300 hPa zonal wind oriented north-south over East Asia-western Pacific sector is closely connected with the amplifying of the Siberian high (Jhun and Lee 2004). Downstream-propagating wave trains, which are initiated by the augmented ridge related to Arctic amplification and can be expressed by the wave activity flux, could induce cold spells over East Asia (Kug et al. 2015; Mori et al. 2014). Apparently, upward-directed wave activity flux associated with warm conditions in the Arctic surface is observed during all three periods (Fig. 4d–f; vectors). The eastward propagation of upper-level planetary-scale waves only extends to high-latitude Eurasia during 1979–1996 while reaches East Asia after the late-1990s (Fig. 4a–c; vectors), promoting the linkage between the midlatitudes and the Arctic. The presence of the enhanced and upstream-extended East Asian jet stream coincides with the southeastward-propagating wave trains after the late-1990s (Li et al. 2014).

Fig. 4
figure 4

Regressions of winter 300 hPa zonal wind (shading) and wave activity flux (vectors; m2 s−2) with respect to the ARTI_2m during a 1979–2017, b 1979–1996, and c 1997–2017, respectively. d–f Same as a–c, but for quasi-geostrophic streamfunction (shading; 106 m2 s−1) and wave activity flux (vectors; m2 s−2, vertical component is multiplied by 100) averaged along 65°–90°N. Areas with correlations exceeding the 95% confidence level are dotted

Full size image

In the conventional description, the intensity of the Siberian high strongly depends on the behavior of the descending motion in the Siberian region (Ding 1990). In order to investigate the changes in the dynamic processes associated with the Arctic temperature-Siberian high relationship, the vertical circulation is regressed upon the ARTI_2m (Fig. 5). Associated with the surface warm conditions in the Barents-Kara Seas during 1979–2017, there is an anomalous ascending motion tilting eastward with altitude (75°–85°N) (Fig. 5a) and an anomalous descending motion over Siberia (55°–65°N) (Fig. 5d). The averaging range for vertical sections is selected according to the significant vertical velocity values at 500 hPa (figure not shown). In latitude-height plane, the southward airflow to the east of the Barents-Kara Seas might link the anomalous ascending motions in the Arctic and descending motion over Siberia due to the Coriolis force (70°–90°E) (Fig. 5g). Therefore, we speculate that it forms a plausible atmospheric bridge between the midlatitudes and the Arctic, through which the Arctic surface temperature change is linked to the intensity of the Siberian high. Comparing 1979–1996 with 1997–2017, striking differences are seen in Siberia and the Arctic. The anomalous ascending motion over the Arctic and the subsidence over Siberia are much stronger in the latter period (Fig. 5c vs. b, f vs. e). Furthermore, the anomalous northerly from the Arctic to Siberia is more significant after the late-1990s (Fig. 5i vs. h), concurrent with pronounced anomalous subsidence at midlatitudes (around 60°N) where the Siberian high is located (Fig. 5i). The notable difference in the dynamic processes associated with Arctic surface warm anomalies in the two periods means that the atmospheric bridge is more likely to be present after the late-1990s.

Fig. 5
figure 5

Regressions of winter vertical circulation (vectors; m s−1) and vertical velocity (shading; 10−3 Pa s−1) anomalies averaged along 75°–85°N with respect to the ARTI_2m during a 1979–2017, b 1979–1996, and c 1997–2017, respectively. d–f And g–i same as a–c, but averaged along 55°–65°N and 70°–90°E, respectively. Areas with correlations exceeding the 95% confidence level are dotted. The vector winds are above the 95% confidence level

Full size image

To further explore this, we regress the divergence of the wind field at both 1000 hPa and 300 hPa with respect to the ARTI_2m during the three periods (Fig. 6). A dynamically consistent structure associated with Arctic surface warm anomalies is observed during 1979–2017, with significant anomalous convergence (divergence) in the lower troposphere and anomalous divergence (convergence) in the upper troposphere over the Arctic (Eurasia) (Fig. 6a, d). This leads to anomalous downward motion in Siberia and the development of the Siberian high (Ding 1990). Accordingly, the atmospheric bridge can be physically understood in the following way. Significant Arctic surface warm anomalies lead to stretching of air column, higher pressure and divergence aloft, which in turn causes lower pressure and convergence close to surface. The anomalous divergence aloft favors a large-scale upper-tropospheric mass convergence to the east and descending motion over Siberia, leading to lower-level divergent airflow and enhancement of the Siberian high. The lower-level divergence in turn favors the Arctic mass convergence, thereby causing an atmospheric bridge between the Barents-Kara Seas and Siberia. It is apparent that the lower-level convergence (divergence) and upper-level divergence (convergence) anomalies over the Arctic (Eurasia) associated with Arctic surface warm anomalies strengthen after the late-1990s (Fig. 6c vs. b, f vs. e), illustrating the intensified midlatitude-Arctic linkage.

Fig. 6
figure 6

Regressions of winter 1000 hPa divergence (shading; 10−7 s−1) and divergent winds (vectors; m s−1) with respect to the ARTI_2m during a 1979–2017, b 1979–1996, and c 1997–2017, respectively. d–f Same as a–c, but for 300 hPa. Areas with correlations exceeding the 95% confidence level are dotted. The vector winds are above the 95% confidence level

Full size image

4 Discussion on the interdecadal change in the midlatitude-Arctic linkage

Arctic warming has a maximum at the surface and extends to the upper troposphere, decreasing with height (Screen and Simmonds 2010). Cohen et al. (2018) pointed out that Arctic tropospheric warming is associated with more extreme winter weather in the United States. Amplified Arctic warming throughout the troposphere after the late-1990s (Figure not shown) motivates us to consider the relationship between the variability of the Arctic mid-tropospheric temperature and change in the midlatitude-Arctic linkage. Figure 7a shows the normalized time series of the detrended ARTI_500. Amplitudes are enhanced after the late-1990s, with standard deviation increasing from 0.87 to 1.12 between 1979–1996 and 1997–2017. Also the sliding correlation between the ARTI_500 and SHI becomes stronger after the late-1990s (Fig. 7b), which is even more robust than change in the relationship between the ARTI_2m and SHI (Fig. 2d).

Fig. 7
figure 7

a Normalized time series of detrended winter ARTI_500 during 1979–2017. b The 15-year sliding correlation coefficients between the detrended ARTI_500 and SHI during 1979–2017. The solid lines indicate the values significant at the 95% and 99% confidence levels. c Same as Fig. 2d, but for the residual winter ARTI_2m after linearly removing the ARTI_500

Full size image

To investigate the potential role of Arctic mid-tropospheric temperature, we calculate the sliding correlation between the residual ARTI_2m (where the ARTI_500 effect has been removed using linear regression) and SHI. It turns out that the Arctic temperature-Siberian high relationship becomes statistically insignificant when the mid-tropospheric temperature variability is removed (Fig. 7c). Previous study revealed that the anomalous Siberian high is concurrent with significant positive anomalies of the surface temperature over the Barents-Kara Seas (Kug et al. 2015), as shown in Fig. 3a. Note that the surface temperature anomalies over the Barents-Kara Seas are significantly positive (Fig. 8a) while there is no significant intensified Siberian high (Fig. 8b). Thus it seems that the relationship between the Arctic temperature and the Siberian high might be determined by the Arctic mid-tropospheric temperature and not the surface temperatures per se.

Fig. 8
figure 8

a Regressions of winter SAT with respect to the residual ARTI_2m during 1979–2017. b Same as a, but for SLP (shading) and UV850 (vectors). Areas with correlations exceeding the 95% confidence level are dotted. The vector winds are above the 95% confidence level

Full size image

Then, is it possible that the Arctic surface warm anomalies are accompanied with stronger Arctic mid-tropospheric warm anomalies after the late-1990s? In Fig. 9, the zonally averaged winter air temperature and zonal wind anomalies over the region of 30°–70°E in the warmer Arctic surface years (ARTI_2m ≥ 0.8) during 1979–1996 and 1997–2017 and their difference (1997–2017 minus 1979–1996) are presented. Apparently, there are significant upper tropospheric warm anomalies in the warmer Arctic surface years during 1997–2017 (Fig. 9b; shading), while there are only significant lower tropospheric warm anomalies during 1979–1996 (Fig. 9a; shading). That is, Arctic surface warm anomalies are significantly associated with mid-tropospheric warm anomalies after the late-1990s. Also, Arctic mid-tropospheric temperature anomaly in the latter period can reach up to 1.8 °C (Fig. 9b; shading), approximately twice the values during 1979–1996 (Fig. 9a; shading). In other words, even if the detrended Arctic surface temperature anomalies are comparable between 1979–1996 and 1997–2017 (around 2.4 °C), the associated Arctic mid-tropospheric warm anomaly after the late-1990s is larger by ~ 1.2 °C (Fig. 9c; shading). Meanwhile, there is a significant weakening of the midlatitude jet stream in the warmer Arctic surface years after the late 1990s with larger amplitude (~ − 2.5 m s−1) than during 1979–1996 (~ − 1.5 m s−1) (Fig. 9; contours). We therefore hypothesize that Arctic mid-tropospheric warming after the late 1990s might be associated with the interdecadal strengthened midlatitude-Arctic linkage.

Fig. 9
figure 9

Winter temperature (shading) and zonal wind (black contours) anomalies averaged along 30°–70°E in the warmer Arctic surface years (ARTI_2m ≥ 0.8) during a 1979–1996 and b 1997–2017 and c their difference (1997–2017 minus 1979–1996); the anomalies are relative to the climatology of 1979–2017. a, b Regions with white dots (green contours) indicate that the temperature (zonal wind) anomalies are significant at the 95% confidence level

Full size image

It is difficult to determine it is the Arctic mid-tropospheric warming causes the interdecadal shift in the midlatitude-Arctic linkage or otherwise. Actually, there is no consensus among scientific communities about the causality between midlatitude circulations and Arctic warming (Chen et al. 2017; Liu et al. 2012; Mori et al. 2014; Woods and Caballero 2016). They could be caused by internal atmospheric variability or driven by some external forcing. We therefore apply multi-model simulations (Ogawa et al. 2018) to search for the cause for the interdecadal change in the Arctic tropospheric temperature and the midlatitude-Arctic linkage.

Arctic warming in the ensemble-mean of climate model simulations is mainly confined to near-surface (Ogawa et al. 2018; Screen et al. 2018). However, in the individual ensembles there are members having shallow Arctic warming and others where the warming spans from surface to mid- and even to upper troposphere. Specifically, eleven members from CAM4, fourteen members from WACCM, fifteen members from IFS, and two members from IAP4 successfully simulate Arctic mid-tropospheric warming (identified by the threshold above 0.2 °C per decade in 500 hPa warming trend over the Barents-Kara Seas). These members are selected. For these selected ensembles’ mean in each model, it is apparent that the simulated Arctic warming extends across the mid-troposphere with maximum values at the surface (Fig. 10a–d) (referred as the category of Arctic mid-tropospheric warming); while in the rest ensembles, the simulated Arctic warming is confined to near-surface (Fig. 10e–h) (referred as the category of Arctic surface warming). All models are forced with the same SST and sea ice but different initial conditions. The different vertical structures of the warming in the same model suggest that internal stochastic atmospheric variability might play an important role in the observed Arctic mid-tropospheric warming. Since the ensemble members have different warming structures, we next test the models for the midlatitude-Arctic linkage.

Fig. 10
figure 10

a–d The simulated winter temperature trend averaged along 30°–70°E during 1982–2013 from the ensembles that simulate the Arctic middle tropospheric warming trend in a CAM4, b WACCM, c IFS, and d IAP4, respectively. e–h Same as a–d, but from the rest ensembles that simulate the Arctic warming trend confined to near-surface

Full size image

Figure 11 displays the winter SLP anomalies regressed onto the ARTI_2m in the two categories of Arctic warming. Compared with the ensemble members showing shallow Arctic warming (Fig. 11e–h), the winter Siberian SLP anomalies associated with ARTI_2m are uniformly stronger and more significant in the ensemble members where the Arctic warming reaches mid-troposphere (Fig. 11a–d). Quantitatively, the Siberian area-averaged SLP anomalies (unit: hPa; black frame in Fig. 11) in the category of Arctic mid-tropospheric warming are 0.32, 0.31, 0.16, and 0.49 in each model, which are correspondingly stronger than those (0.19, 0.27, − 0.02, and 0.25) associated with Arctic surface warming. Thus, the multi-model simulations support our hypothesis that the interdecadal strengthened midlatitude-Arctic linkage is associated with Arctic mid-tropospheric warming and suggest that the internal atmospheric variability plays an important role in the interdecadal change. This is also consistent with the conclusion by Sung et al. (2018) that the interdecadal variability of the warm Arctic-cold Eurasian pattern in the twentieth century might be attributed to the internal atmospheric variability.

Fig. 11
figure 11

a–d The simulated regressions of winter SLP with respect to the ARTI_2m during 1982–2013 from the ensembles that simulate the Arctic middle tropospheric warming trend in a CAM4, b WACCM, c IFS, and d IAP4, respectively. e–h Same as a–d, but from the rest ensembles that simulate the Arctic warming trend confined to near-surface. Areas with correlations exceeding the 90% confidence level are dotted. The box marks the region where the SHI is defined

Full size image

With respect to the contentious topic, some discussion is further given. It might be worth noting that the change of the poleward temperature gradient and the midlatitude zonal wind is not apparent when Arctic warming is confined to near-surface, which is much more dominant when Arctic warming extends from surface to the mid-troposphere (figure not shown). The warm Arctic mid-troposphere after the late-1990s, which might be due to the internal atmospheric variability as suggested by our model simulations and previous studies (Graversen et al. 2008; Lee et al. 2017), may contribute to the weakening of the zonal wind based on the thermal wind balance and the intensified linkage between Arctic warming and the midlatitude climate (Liu et al. 2012; Vavrus 2018; Yao et al. 2017). The possible mechanisms implied here are currently under investigation in a companion study.

5 Conclusions

Previous study has shown that surface warming over the Barents-Kara Seas in boreal winter is generally associated with an intensified Siberian high (Kug et al. 2015). Based on the reanalysis data, this study explores the strength of the midlatitude-Arctic linkage. The correlations between the Arctic surface temperature and the Siberian high show an apparent intensification around late-1990s. After this, the warmer-than-normal Arctic is generally concurrent with significant positive SLP anomalies stretching from the Barents-Kara Seas to East Asia, indicating a stronger-than-normal Siberian high. Moreover, the westward-extended East Asian jet stream occurs in conjunction with the southeastward propagation of Rossby waves from the Arctic to East Asia, displaying an apparent midlatitude-Arctic linkage. Anomalous ascending/descending motions and upper tropospheric divergence/convergence anomalies over the Arctic-Eurasian sector indicate a potential dynamical atmospheric bridge that favors the midlatitude-Arctic linkage after the late-1990s. By contrast, Eurasian midlatitude atmospheric circulation anomalies barely correlate with Arctic surface warm anomalies in the decades before late-1990s.

There has been a rapid tropospheric warming trend over the Arctic after the late-1990s, with increased Arctic mid-tropospheric temperature variability. Further analysis shows that no significant intensified Siberian high is observed when the Arctic warm anomalies are confined to near-surface without significant warm anomalies in the mid-troposphere. That is, the interannual relationship between the Arctic surface temperature and the Siberian high might be determined by the variability of the Arctic mid-tropospheric temperature. Moreover, the Arctic surface warm anomalies are accompanied with significantly stronger Arctic mid-tropospheric warm anomalies after the late-1990s. We therefore hypothesize that the interdecadal Arctic mid-tropospheric warming might be associated with the interdecadal intensified midlatitude-Arctic linkage.

Simulations from multi-model ensembles forced with observed SST and sea ice also support the concurrency of Arctic mid-tropospheric warming and the strengthened midlatitude-Arctic linkage. While the multi-model ensemble mean displays a shallow Arctic warming and no apparent anomalous atmospheric circulation at midlatitudes (Ogawa et al. 2018), some individual ensemble members reproduce the observed Arctic warming across the mid-troposphere. It is found that the category of Arctic tropospheric warming does display a significant Arctic temperature-Siberian high relationship, while it is much weaker in the category of Arctic surface warming. The analysis from the reanalysis dataset and the multi-model simulations does not emphasize the effect of the Arctic mid-tropospheric warming or the Siberian high on the interdecadal shift in the midlatitude-Arctic linkage, but shows robust evidence for the concurrency of the Arctic mid-tropospheric warming and the strengthened midlatitude-Arctic linkage. The causal relationship is currently under investigation in a companion study. It should also be pointed out that all the models in this study are forced with the same observed SST and sea ice but different initial conditions, suggesting the potential contribution of the internal atmospheric variability to Arctic mid-tropospheric warming and the interdecadal shift in the midlatitude-Arctic linkage.