-
We first investigate the relationship between Arctic SIC and summertime extratropical temperature fluctuations over Central and East Asia (Caspian Sea–Balkhash Lake–Baikal Lake–Northern China). As the atmospheric states may relate to the Arctic SIC in earlier seasons, the SVD analysis is conducted between the summertime (the JJA and each monthly mean) temperature fluctuations (including components in different timescales) and each monthly SIC from January to the summer months. All results of the conducted SVD analysis are summarized in Table 1, which shows that the most significant covariation occurs between the 10–30 day August temperature fluctuations and the Arctic SIC in earlier seasons, especially in January and February. To intuitively understand the interannual variations of August temperature fluctuations, the time series of
${T{'}}^{2}$ ,$ {T{'}}_{h}^{2} $ and$ {T{'}}_{l}^{2} $ averaged over mid-latitude Eurasia are displayed in Fig. 1. The 10–30 day temperature fluctuations account for about one-third of the total$ {T{'}}^{2} $ and are slightly weaker than the 2–10 day temperature fluctuations$ {T{'}}_{h}^{2} $ . Note that, in addition to the data from 1991–2020, the same SVD analysis is also conducted using the data of the previous 30 years (1981–2010). However, only the 1991–2020 data provides all of the statistics (SC, R, and SCF) of the SVD mode that exceeds the threshold confidence level, suggesting that the relationship between Arctic sea ice and summertime mid-latitude 10–30 day temperature fluctuations may exhibit decadal variations. This premise is consistent with previous studies (Xu et al., 2019; Liu et al., 2020) suggesting a strengthened Arctic-midlatitude linkage after the late 1990s.$ {\mathit{T}\mathit{\text{'}}}^{2} $ JJA mean Jun Jul Aug Total × R=0.88(May) × × 2–10 days × × R=0.77(Mar)
R=0.82(Apr)
R=0.85(Jul)R=0.81(Jun) 10–30 days SC=3.8(Feb)
SCF=42%(Feb)R=0.84(Feb) × Jan, Feb
[see Table 2]Table 1. Overview of statistics of the first SVD mode between summertime
${T'}^{2}$ (including its synoptic and low-frequency component by Lanczos filter) and each monthly mean SIC from January to August from 1991–2020. Statistics include the squared covariance (SC, 105 m2 s–2), the temporal correlation (R), and the squared covariance fraction (SCF, %). The cross sign indicates that the statistic is not significant. Values significant at the 90% (black) and 95% (blue) confidence levels according to a Monte Carlo test are listed.Figure 1. Time series of August temperature fluctuations (time scale of 2–90 days) averaged over the mid-latitude Eurasia region (30°–80°N, 30°–150°E) and its 10–30 day, 2–10 day components by Lanczos filter with detrended data.
Detailed statistics of the first SVD mode between the August 10–30 day temperature fluctuations and the monthly Arctic SIC in earlier seasons are listed in Table 2. To test the robustness of the result, we also use three different time filters to extract the 10–30 day temperature fluctuations for the SVD analysis, whose results are all included in Table 2 for comparison. As shown in the table, the August temperature fluctuations on the time scale of 10–30 days are significantly correlated with the variation in SIC in late winter and early spring (from January to April). For January and February, the squared covariance and temporal correlation are both significant. The squared covariance passes the significance test at the 99% confidence level, with the temporal correlation between the two fields stronger for February and higher than 0.7. The first SVD mode for February and January SIC accounts for about 60% of the squared covariance, with the percentage for February slightly higher, which is larger than that in other months. The above results are similar no matter which filter extracts the 10–30 day temperature fluctuations. Therefore, the results are not sensitive to the details of the timescale filter, and hereafter we only show the results using the Lanczos filter.
Jan Feb Mar Apr May Jun Jul Aug Harmonic SC 20.0 18.7 10.8 11.5 10.5 13.7 21.8 20.4 R 0.761 0.763 0.703 0.692 0.648 0.754 0.768 0.811 SCF 58.5 59.1 47.0 49.4 39.0 29.7 31.2 26.6 Lanczos SC 18.7 17.7 11.3 10.5 8.5 14.1 22.0 17.7 R 0.747 0.753 0.716 0.671 0.613 0.747 0.764 0.780 SCF 60.4 61.0 51.4 49.9 36.4 33.3 34.9 26.6 Butterworth SC 17.1 16.1 10.2 9.1 7.6 13.3 20.7 16.3 R 0.747 0.753 0.714 0.662 0.608 0.753 0.769 0.781 SCF 59.5 60.1 50.3 47.8 35.1 33.4 34.8 26.1 Table 2. Statistics of the first SVD mode between August
${T'}^{2}$ on a timescale of 10–30 days and each monthly mean SIC from January to August using harmonic, Lanczos, and Butterworth filters. Statistics include the squared covariance (SC, 105 m2 s–2), the temporal correlation (R), and the squared covariance fraction (SCF, %). The values in blue (bold) indicate statistical significance exceeding the 90% (95%) confidence level according to the Monte Carlo test.The spatial patterns of the first SVD mode between the February Arctic SIC and subsequent August 10–30 day temperature fluctuations over mid-to-east Asia and their time series are displayed in Fig. 2. The first SVD mode shows that when the February SIC in Barents-Kara Sea is above normal, a widespread region of mid-west Asia extending into northwest China experience considerably higher-than-normal temperature fluctuations, with commensurate lower-than-normal temperature fluctuations over and to the east of Baikal Lake including East China, showing a zonal dipolar pattern (Figs. 2a, b). This study mainly focuses on the enhanced temperature fluctuations from mid-west Asia into northwest China because the eastern negative anomaly, characterized by weak temperature fluctuations, does not pass the 95% significance test.
Figure 2. The first SVD mode between August 10–30 day temperature fluctuations and February Arctic SIC with detrended data. Spatial patterns (heterogeneous maps) of (a) February SIC, (b) August temperature fluctuations on the 10–30 days timescale, and (c) the normalized time series of the two fields. The dotted regions denote values statistically significant above the 95% confidence level according to a Student’s t-test.
Note that, in Fig. 2a, the significant sea-ice anomalies appear not only over the Barents-Kara Sea but also over the western and eastern coastlines of Greenland, which may be forced by means of a large-scale atmospheric circulation such as the NAO (Strong, 2012) or oceanic heat transport (e.g., Germe et al., 2011). However, as shown in previous studies and in our analysis, the Barents-Kara Sea is the key region affecting atmospheric circulation in mid-west Asia and northern China region. The SIC anomalies along the coast of Greenland (i.e., the Greenland Sea and Baffin Bay) are instead most related to Greenland blocking as well as weather patterns in North America and Europe (e.g., Chen and Luo, 2021). Furthermore, the Barents-Kara Sea exhibits the most predominant sea-ice variability in the Arctic in boreal winter and spring. The correlation between the time series of February SIC in the SVD mode and the SIC anomalies averaged over the Barents-Kara Sea is 0.91. Therefore, our main focus is the February SIC anomaly over the Barents-Kara Sea region.
Therefore, the SVD analysis suggests a significant linkage between the August 10–30 day temperature fluctuations over the mid-west Asia–northwest China region and the preceding February Arctic sea ice in the Barents-Kara Sea. When the Barents-Kara Sea SIC in late winter is above (below) normal, the 10–30 day temperature fluctuations over mid-west Asia into Northwest China in August increase (decrease).
-
The changes in the atmospheric circulation associated with this coupled mode are further investigated. Linear regressions of the August 500-hPa geopotential height (Z500), sea level pressure (SLP), and 300-hPa zonal wind (U300) upon the time series of the August temperature fluctuations are performed. As shown in Fig. 3, the atmospheric circulation changes most significantly in the mid-to-high latitudes east of 60°E. In strong temperature fluctuation years, the 500-hPa geopotential height shows a distinct dipolar pattern, which increases in the Arctic, especially around the Barents-Kara Sea, with accompanying decreases to the east of Ural Mountains (Fig. 3a). The composites of Z500 and high- and low-temperature fluctuation years in Fig. 3b further illustrate the differences in the atmospheric circulations. In strong temperature fluctuation years, the atmospheric circulation exhibits more meridional meanders coincident with a weak ridge in northern Europe. These configurations are conducive to the formation of blocking, guiding cold surges from the Norwegian Sea into the Ural region through the Barents-Kara Sea.
Figure 3. (left) Regression maps of 500-hPa geopotential height (Z500, units: dagpm), sea level pressure (SLP, units: hPa), and the 300 hPa zonal wind (U300, units: m s–1) in August based on the time series of temperature fluctuations in the SVD mode in Fig. 2c (color shading). The dotted regions (contours) denote statistical significance at a greater than 95% (90%) confidence level according to an F-test. (right) Composites of Z500 (intervals: 8 dagpm), SLP (intervals: 5 hPa), and U300 (intervals: 4 m s–1) in strong (contours) and weak (shading) temperature fluctuation years. The strong (weak) years are defined as the years whose variations are stronger than 0.8 (weaker than –0.8) standard deviation of the temperature fluctuations time series (strong years: 1996, 1997,1998, 2011, and 2019; weak years: 1993, 1994, 1995, 2000, 2016, 2017, and 2018). The green frame denotes the area of temperature fluctuations from the SVD analysis.
The change of SLP associated with the temperature fluctuations is consistent with Z500 (Figs. 3c, d), suggesting that the change of atmospheric circulation is approximately barotropic. Compared with the upper layer, the decreased SLP area in Eurasia is more significant. As shown in Fig. 3e, the changes of zonal wind at 300 hPa display a meridional dipolar structure over the coast of the Barents-Kara Sea and mid-latitude Eurasia. This suggests that the westerly winds along the Arctic coast weaken, especially over the Barents-Kara Sea, with a commensurate and significant enhancement of the mid-latitude westerlies to the north of Lake Barkhash, as shown in the composite analysis in Fig. 3f.
The wave energy propagation associated with summer temperature fluctuations is further diagnosed by wave activity flux. Figure 4 shows the composite of WAF on the timescales of 10–30 days in the upper- troposphere in typical strong and weak temperature fluctuation years. In strong temperature fluctuation years (Fig. 4a), two branches of wave trains emerge from the Atlantic sector to the Eurasian continent in the mid-to-high latitudes. One propagates from the North Atlantic to the Barents-Kara Sea, then meanders southward through the Ural region before finally reaching the northern side of the Tibetan Plateau. The significant southeastward wave activity flux from the Barents-Kara Sea to Eurasia is in accordance with the increased temperature fluctuations along and to the northeast of the Caspian Sea. The other branch of the wave train propagates eastward from the North Atlantic across Europe into the Ural region before joining the northern branch. In weak temperature fluctuation years (Fig. 4b), the eastward wave energy propagation still exists but with weaker strength. However, the one passing through the Barents-Kara Sea propagating southward to the Ural region disappears. The difference in 10–30 day WAF between strong and weak temperature fluctuations years is also displayed in Fig. 4c. This figure again confirms that the main differences between WAF lie in the significant southward propagation of wave energy from the Barents-Kara Sea to the northern side of the Tibetan Plateau, which is consistent with the increase of the 10–30 day temperature fluctuations in the region.
Figure 4. Composite 300-hPa horizontal wave activity flux on the time scale of 10–30 days (arrows, units: m2 s–2) and perturbation stream function (shading, units: 105 m2 s–2) in (a) strong temperature fluctuation years, (b) weak temperature fluctuation years, and (c) their differences, in August based on the time series of 10–30 day August temperature fluctuations, in the SVD mode. Green arrows denote that the wave activity flux in either its x- or y- components is statistically significant above the 95% confidence level as determined by a Student’s t-test. The panel headings WAF_TN01 indicate wave activity flux, as defined by Takaya and Nakamura (2001).
In summary, when the August 10–30 day temperature fluctuations over and to the south of the Ural Mountains are active and coupled with the preceding February SIC in the Barents-Kara Sea, the atmospheric background circulation in the Arctic–Eurasian continent (especially to the east of 60°E) features more meridional meanders causing the westerly wind in the region to slows down, which favors the exchange of air masses between Arctic and midlatitudes and the consequent occurrence of cold surges and blocking. Within such a background circulation, strong meridional energy propagating waves with timescales of 10–30 days are observed from the Barents-Kara Sea to the Ural Mountains, coincident with the observed enhanced 10–30 day temperature fluctuations to the east of 60°E. These results support the view that a weaker polar jet stream in high latitudes and slower Rossby waves lead to more persistent and extreme atmospheric conditions (Tang et al., 2014; Coumou et al., 2018; Li et al., 2019).
-
How can the sea ice anomaly in February affect the temperature fluctuations one-half a year later? As atmospheric circulation is a relatively fast process and contains little long-term memory, we first apply regression analysis to explore the evolution of sea surface temperature and sea ice anomalies from February to August (Figs. 5, 6) with the coupled mode. Significant SST anomalies (Fig. 5a) exist in both the North Atlantic and North Pacific in February. In the North Atlantic, a significant meridional tripolar structure is observed. Meanwhile, the SST anomaly in the Atlantic subpolar region shows a zonal dipolar pattern with warm anomalies near southern Greenland and cold anomalies in the Nordic Seas in strong temperature fluctuation years. In the North Pacific, a warm anomaly in the Bering Sea is observed. Consistent with the SST anomaly, the SIC increases in the Barents-Kara Sea, while decreases are noted to the west of Greenland. From late winter to summer (Figs. 5a–d), the SST warm anomaly in the Bering Sea becomes weak, and the tripolar pattern in the Northern Atlantic gradually converts to uniform warming. However, the significant western warm and eastern cold SST anomalies in the Atlantic subpolar region maintain themselves until summer, suggesting their continuous influence on atmospheric circulation as well temperature fluctuations. Meanwhile, in strong temperature fluctuation years, the increased sea ice in the Barents-Kara Sea is prominent in the entire Arctic in the winter and spring (Figs. 6a–d). Although the Barents-Kara Sea ice melts in summer, there is still a positive sea ice anomaly in the marginal melting zone.
Figure 5. Regression maps of sea surface temperature (SST, units: °C) in February, April, June, and August (a–d, respectively) based on the time series of temperature fluctuations in the SVD mode in Fig. 2c (shading). The crossed regions are statistically significant, above the 95% confidence level according to an F-test. The contours represent the climatology (intervals: 5°C).
Figure 6. Same as in Fig. 5., but for sea-ice concentration (SIC). The black curve denotes the ice edge (20% sea-ice area fraction) for the climatology.
Wu et al. (2013) investigated the relationship between winter sea ice and summer atmospheric circulation over Eurasia and showed that persistent winter-spring SIC and a horseshoe-like triple pattern of SST anomalies in the North Atlantic act as bridges. In addition, Guo et al. (2014) suggested that the close Arctic-midlatitude relationship is explained by persistent SST anomalies in the Pacific. Although these two anomalous SST signals are also observed in this study, they become less obvious in summer.
If the SST and SIC anomalies in subpolar and Arctic regions last from late winter to summer, why are the significant temperature fluctuations only limited to August? To address this question, we first investigate the change in the background atmospheric circulation from July to August. Figure 7 shows the composites of background circulation in the middle and upper troposphere from July to August in strong and weak temperature fluctuation years. In strong temperature fluctuation years, the weak European ridge and trough in high latitudes amplify rapidly and extend southward from July to August, bridging the mid-to-high latitudes with the Caspian Sea area upstream of the deep trough. In contrast, during weak temperature fluctuation years, the circulation in the midlatitude Eurasian continent becomes more zonal in August than in July (Figs. 7a, b). Therefore, the evolution of atmospheric circulation between strong and weak temperature fluctuations years significantly differ. The circulation in the upper troposphere is consistent with the lower layers but exhibits a stronger trough in the eastern Ural region in strong temperature fluctuation years (Figs. 7c, d). The deepening of troughs and stagnation of ridges are also consistent with the weakening of the jet stream in August. Such changes in atmospheric circulation from July to August favor the meridional propagation of 10-to-30-day wave energy (Fig. 4), as suggested by Gu et al. (2018).
Figure 7. Composites of 500-hPa (a–b, interval: 4 dagpm, 580 dagpm bold) and 250-hPa (c–d, interval: 8 dagpm, 1072 dagpm bold) geopotential height in July (black) and August (red) in strong and weak (left and right, respectively) temperature fluctuation years.
The evolution of wave activity flux on timescales of 10–30 days in strong and weak temperature fluctuation years also exhibits evident differences in July and August (Fig. 8). In strong temperature fluctuation years, the July WAF is dominated by eastward propagation. In August, the most prominent feature is the evident southeastward wave energy propagation from the Barents-Kara Sea through the southeast Ural region into the Tibetan Plateau, along with the evident meridional meanders contained within the atmospheric circulation. The differences between strong and weak temperature fluctuation years are small in July, existing mainly over the North Atlantic and Eurasian regions, while they become strong and evident in August, which is consistent with the differences between the atmospheric background circulation. In weak fluctuation years, the August WAF is still dominated by zonal eastward propagation and exhibits only slight southeastward propagation in midlatitudes, despite the weak and subtle differences between the strong and weak temperature fluctuation years in July. In this way, the changes in WAF from July to August are consistent with the background circulation. Therefore, compared with subtle difference between strong and weak temperature fluctuation years in July, only in August does the atmospheric circulation and the WAF on timescales of 10–30 days exhibit evident changes and support the Arctic-midlatitude linkage, through which the anomalous thermal state in subpolar and Arctic regions (i.e., the Barents and Kara Seas) is able to affect the mid-latitude temperature fluctuations (Zhao et al., 2004; Cohen et al., 2018; Lee et al., 2019).
Figure 8. Same as in Fig. 7, but for 300-hPa wave activity flux on a timescale of 10–30 days (magnitude ≥ 3 m2 s–2).
In summary, when the SIC in the Barents-Kara Sea increases in February, the SST anomaly in the mid-and-high latitudes shows a dipolar pattern with warm anomalies near southern Greenland and a cold anomaly in the Nordic Seas. These SST anomalies maintain themselves until summer, accompanied by the persistence of excessive SIC in the Barents-Kara Sea. With the distinct adjustment of the atmospheric background circulation from July to August, the European ridge and trough to the east of the Ural region evidently amplify in August, within which the low-frequency (10–30 days) wave energy stored in southern Greenland propagates through the Barents-Kara Sea southeast to the lower latitudes into the Ural region and Tibetan Plateau. Meanwhile, another branch of WAF propagates eastward along the lower latitudes and converges with the northern branch in the Ural region, especially near the Caspian Sea, making this a region with strong wave energy propagation. The above-mentioned background circulation and the WAF strongly promote the invasion of high-latitude cold surges and the formation of blocking. As a result, August temperature fluctuations on timescales of 10–30 days significantly increase in the Ural region, especially around the Caspian Sea.
$ {\mathit{T}\mathit{\text{'}}}^{2} $ | JJA mean | Jun | Jul | Aug |
Total | × | R=0.88(May) | × | × |
2–10 days | × | × | R=0.77(Mar) R=0.82(Apr) R=0.85(Jul) | R=0.81(Jun) |
10–30 days | SC=3.8(Feb) SCF=42%(Feb) | R=0.84(Feb) | × | Jan, Feb [see Table 2] |