-
Figure 1a shows the linear trend of summer (June–July–August, JJA) Arctic SIC over the 42-year period 1979–2020. The largest decline in sea ice has occurred in the Arctic marginal seas, and especially in the coastal regions of Siberia and Alaska, at a rate of more than 5% SIC per decade (black contours). Because the sea ice edges are the most fragile, sea ice loss over these Arctic marginal seas largely determines the overall decline in Arctic sea ice extent (ASIE). The summer (averaged for the months of June, July, and August) ASIE shows a rapid decline at a rate of 0.65 million km2 decade–1, reaching its second lowest value in 2012 (7.69 million km2) and its lowest in 2020 (7.65 million km2) (Fig. 1b, red line). The sea ice decline in September closely follows that of summer at 0.83 million km2 decade–1, with its lowest level in 2012 (3.41 million km2) and its second lowest in 2020 (3.74 million km2) (Fig. 1b, red line).
Figure 1. (a) Linear trend of summer (JJA) Arctic sea ice concentration (SIC) over the period 1979–2020. Stippling denotes that the linear trends are significant at the 95% confidence level. Black contours denote the SIC trends of 10% (10 yr)–1 and 5% (10 yr)–1. (b) Time series of the summer Greenland high index (GL-high) (blue), summer and September (Sep) Arctic sea ice extent (ASIE) (red) (ordinate reversed) over the period 1979–2020. After detrending, the correlations between the GL-high and JJA ASIE and September ASIE are –0.52 and –0.50, respectively. The Siberian (60°–180°E and 70°–80°N) and Alaskan coastal boxes (180°–270°E and 70°–80°N) used for analysis are marked by the dashed-red and dashed-blue lines, respectively, in (a).
Numerous studies (Ogi and Wallace, 2007; Overland and Wang, 2010; Ding et al., 2017) have suggested that in addition to global warming, high-latitude atmospheric circulation anomalies contribute significantly to interannual as well as long-term changes in the summer ASIE. The GL-high is one of the most important circulation patterns in the high latitudes. As shown in Fig. 1b, the summer GL-high shows a significant positive trend, which is overlaid by a strong interannual variability. After detrending, the GL-high is highly correlated with the summer ASIE and September ASIE with correlations of about –0.52 and –0.50, respectively (Table 1). The interannual variability of the GL-high accounts for about 30% of the interannual variability of the summer ASIE. However, the GL-high only accounts for part of the variance. The remaining fraction cannot be explained by the GL-high.
To understand the summer sea ice variability, we divide the time series of summer ASIE into that part that is linearly dependent on the GL-high and the remainder, the latter being linearly independent of the GL-high, both times series (ASIE and GL-high) being first linearly detrended. The correlation coefficients between the summer ASIE, these two components of the summer ASIE, and the summer sea ice concentration are shown in Figs. 2a–c. Clearly, the interannual variability of the summer ASIE consists of significant marginal sea ice changes: largely over the Barents, Kara, Laptev, and East Siberian Seas of the Siberian Arctic (Fig. 2a, red box), and over the Chukchi and Beaufort Seas (Fig. 2a, blue box). These are also consistent with the regions of largest sea ice decline (Fig. 2a vs. Fig. 1a). The intensified GL-high is strongly correlated with reduced sea ice along the Alaskan coast with correlations greater than 0.40, significant at the 99% confidence level (Fig. 2b). There is also a significant sea ice decrease (increase) over the Baffin and Hudson Bays (western Greenland), which is due to the anticyclonic anomaly over Greenland that transports warm and moist (cold and dry) air masses northward (southward) over western (eastern) Greenland (Wang et al., 2020). The component linearly unrelated to the GL-high is correlated with substantial sea ice loss along the Siberian coast and around Svalbard Island. Therefore, summer Arctic sea ice variability shows different patterns: sea ice changes along the Alaskan coast, which are significantly affected by the GL-high, and sea ice changes along the Siberian coast, which are unrelated to the GL-high. In the following (see Fig. 2), the Siberian coast is defined as the area encompassing the Barents–Kara Seas, the Laptev Sea, and the East Siberian Sea; and the Alaskan coast is defined as the area encompassing the Chukchi Sea (part of it) and the Beaufort Sea. We then calculated the SIE along the Alaskan coast (Ala-SIE) and along the Siberian coast (Sib-SIE) (defined as areas with at least 15% SIC).
Figure 2. Upper panel: partition of summer Arctic sea ice changes according to their relationship to the GL-high. Correlations over the period 1979–2020 between the linearly detrended summer Arctic SIC and ASIE (a), the GL-high component (b), and the component independent of the GL-high (c). Stippling denotes that the correlations are significant at the 95% confidence level (R > 0.30). The Siberian and Alaskan coastal boxes are marked by the red and blue lines, respectively. Lower panel: (d) timeseries of anomalous summer (JJA) ASIE (black line), Sib-SIE (blue line), and Ala-SIE (red line) over the period 1979–2020. (e) is the same as (d) but for the detrended time series of ASIE (black), Sib-SIE (blue), and Ala-SIE (red).
JJA ASIE Sep ASIE JJA GL-high JJA Ala-SIE JJA Sib-SIE JJA ASIE 1 − − − − Sep ASIE 0.86 1 − − − JJA GL-high –0.52 –0.50 1 − − JJA Ala-SIE 0.45 0.53 –0.62 1 − JJA Sib-SIE 0.81 0.69 –0.31 0.22 1 Table 1. Correlations between summer ASIE, September (Sep) ASIE, summer GL-high, summer Ala-SIE, and summer Sib-SIE over the period 1979–2020. The linear trends of the time series have been removed prior to correlation analysis. The correlation coefficients that are statistically significant at the 95% confidence level are printed in bold.
Figures 2d–e show the time series of the summer ASIE, Ala-SIE, and Sib-SIE over the period 1979–2020. The Sib-SIE displays a declining trend at a rate of 0.22 million km2 (10yr)−1 (Fig. 2d, red line); the Ala-SIE at a rate of 0.11 million km2 (10yr)−1 (Fig. 2e, blue line). In 2012, the summer Ala-SIE reaches its lowest level, corresponding to the lowest annual minimum ASIE (contributed by both); in 2020, the summer Sib-SIE reaches its lowest level, corresponding to the second-lowest annual minimum ASIE (contributed primarily by Sib-SIE). After detrending, the correlation coefficients among JJA (and Sep) ASIE, Ala-SIE, and Sib-SIE and the GL-high are shown in Table 1. In terms of interannual variability, the Sib-SIE variability accounts about 65% of the summer ASIE variability, and the Ala-SIE accounts for about 21%. Note that the correlation between the Sib-SIE and Ala-SIE is relatively low (R = 0.22), suggesting that their interannual variability is largely independent of each other. Therefore, the sea ice over the two regions contributes separately and also highly to the variability of the summer ASIE and Sep ASIE. In the following section, we investigate the interannual connections between the summer SIE, Ala-SIE, and Sib-SIE and the atmospheric circulation.
JJA ASIE | Sep ASIE | JJA GL-high | JJA Ala-SIE | JJA Sib-SIE | |
JJA ASIE | 1 | − | − | − | − |
Sep ASIE | 0.86 | 1 | − | − | − |
JJA GL-high | –0.52 | –0.50 | 1 | − | − |
JJA Ala-SIE | 0.45 | 0.53 | –0.62 | 1 | − |
JJA Sib-SIE | 0.81 | 0.69 | –0.31 | 0.22 | 1 |