1. Introduction
From the late 1970s until the mid-2010s, sea ice extent in the two polar regions was trending in opposite directions. There was a rapid decline in the Arctic (Cavalieri and Parkinson 2012) while a slow expansion in most of the Antarctic until about 2014 when the expansion suddenly ceased followed by rapid contraction (Parkinson 2014, 2019; Eayrs et al. 2019). Numerous studies have examined sea ice trends and their underlying mechanisms, but the majority were concentrated on either the Arctic or the Antarctic sea ice. Reviews of these studies have been provided by Stroeve et al. (2012), Meier et al. (2014), Stroeve and Notz (2018), and Richter-Menge and Druckenmiller (2020) for the Arctic and by Hobbs et al. (2016), Eayrs et al. (2019), and Scambos and Stammerjohn (2020) for the Antarctic.
Nevertheless, several studies have compared and contrasted Arctic and Antarctic sea ice variability from historical and future perspectives and offered explanations for the differences in seasonal and interannual changes and long term trends. Turner and Overland (2009) noted that the geographical differences in topography and land/sea distributions can result in large differences in climate change, including sea ice change, between the two polar regions. Stammerjohn et al. (2012) observed different spatiotemporal patterns in the timing of annual sea ice advance or retreat in the Arctic and Antarctic, which they attributed to the differences in net ocean heat changes. Simmonds (2015) showed opposite trends in seasonal and annual mean sea ice extent from 1979 to 2013 between the Arctic and Antarctic, with the rate of increase in the Antarctic approximately one-third the rate of the decrease in the Arctic. Holland and Kimura (2016) derived a climatology of the sea ice concentration budget at both poles and illustrated how sea ice advection, divergence and mechanical redistribution help maintain or melt sea ice in different regions of the Arctic and the Antarctic. Ordoñez et al. (2018) compared the predictability of sea ice anomalies between the Arctic and the Antarctic by an Earth system model. They showed substantially larger regional variations of seasonal sea ice predictability in the Arctic than the Antarctic. They also illustrated that inclusion of ocean dynamics in the model led to an increase in predictability in the Arctic but a decrease in the Antarctic. These and some other differences in the Arctic and Antarctic sea ice variability and predictability are discussed in two recent reviews by Serreze and Meier (2019) and Maksym (2019).
The aforementioned comparative studies helped elucidate the differences in sea ice variability and predictability between the Arctic and Antarctic. They, however, fell short of exploring possible teleconnection between sea ice anomalies at the opposite polar regions. Based on the indication that there exists a teleconnection between the surface air temperature at the two polar regions from the perspectives of sea surface temperature (SST) (Chylek et al. 2010), atmospheric processes (Wang et al. 2015), and surface energy budget (Gao et al. 2019), Yu et al. (2017) put forward a hypothesis that the opposite sea ice trends in the Arctic and the Antarctic in recent decades may be related and linked to known modes of climate variability. They evaluated this hypothesis by analyzing seasonal mean sea ice concentration anomalies in the Arctic and Antarctic jointly. Their results showed that the Arctic and Antarctic sea ice concentration variations are connected on interdecadal time scales and the leading mode of variability is positively correlated with the Atlantic multidecadal oscillation (AMO) and negatively correlated with the Pacific decadal oscillation (PDO). Particularly, two wave trains related to the PDO and the AMO appear to produce anomalous surface-air temperature and low-level wind fields in both poles that contribute to the opposite long-term trends in sea ice concentration between the Arctic and Antarctic regions.
The current study will explore the Arctic and Antarctic sea ice anomalies holistically. In contrast to Yu et al. (2017) that used seasonal sea ice data to investigate primarily concurrent variability modes of sea ice at two poles and their contributions to seasonal sea ice trends for each season, the current study uses the monthly global sea ice data that not only capture variability at interannual and interdecadal time scale, but also intraseasonal and seasonal time scales. In addition, the current analyses utilize the self-organizing map (SOM) pattern extracting method, which has been shown to be more robust in extracting modes of variability in climate data compared to the more traditional empirical orthogonal function (EOF) method used in Yu et al. (2017) (Reusch et al. 2005). Specifically, SOM is used to extract simultaneously the main modes of variability in the monthly sea ice anomalies at both poles, revealing synchronous variability patterns of sea ice in the Arctic and the Antarctic and the variations of these patterns on different time scales ranging from intraseasonal to interdecadal. Finally, the time period of analyses is extended from the recent four decades (1979–2018), as in most other sea ice studies, to 180 years (1836–2015) so that decadal- to multidecadal-scale sea ice variability can be explored, with the caveat that sea ice observations prior to the satellite era were severely limited.
The study will address the following questions:
-
What are the major patterns of variability in the monthly sea ice concentration anomalies in the Arctic and Antarctic when considered jointly? For given anomalous sea ice patterns in the Arctic, what are the corresponding patterns in the Antarctic?
-
How are these patterns distributed over the annual cycle?
-
How closely are the changes in the occurrence of these patterns on the interannual and interdecadal scales related to the known variability modes in the SST and large-scale atmospheric circulations on similar time scales? Can the spatial patterns be explained by anomalous atmospheric circulations?
-
Do these patterns exhibit any significant trends, and if they do, how much can they explain the observed sea ice trends in the Arctic and Antarctic?
2. Datasets and methods
The primary period of analysis is 1979–2018, and three different datasets are utilized for the analysis. The first is global monthly sea ice concentration data (spatial resolution of 25 km) from the U.S. National Snow and Ice Data Center (NSIDC) (http://nsidc.org/data/NSIDC-0051) (Cavalieri et al. 1996). The second is the monthly sea surface temperature (SST) data (spatial resolution of 2°) from the U.S. National Oceanic and Atmospheric Administration (NOAA) Extended Reconstructed Sea Surface Temperature (ERSST) version 5 (ERSST v5; Huang et al. 2017). (https://www.ncei.noaa.gov/products/extended-reconstructed-sst). The third is the monthly atmospheric data (spatial resolution of 0.75°) from the European Centre for Medium-Range Weather Forecasts (ECMWF) Interim Reanalysis data (ERA-Interim; Dee et al. 2011) (https://apps.ecmwf.int/datasets/data/interim-full-moda/levtype=sfc/). The ERA-Interim reanalysis dataset has the best performance in depicting recent Antarctic climate except for the ERA-5 (Hersbach et al. 2020) (Bracegirdle and Marshall 2012). ERA-Interim reanalysis data contain some biases in the Antarctic (Jones and Lister 2015) and in the Arctic (Lindsay et al. 2014), with magnitudes depending on variables. However, as the current analysis uses anomalies of several atmospheric variables, rather than their absolute magnitudes, the influence of the biases on the results is expected to be small.
The global sea ice refers to sea ice in the regions between 50° and 90° latitude in the two hemispheres. In these regions, grid points on land or ice-free ocean surfaces and those that are not covered by satellite observations are not included in the analysis. At each grid point included in the analysis, a monthly anomaly is calculated as a departure from the 40-yr (1979–2018) averaged value for the month. By analyzing anomalies for each month, rather than the magnitude of the monthly values, seasonal and intraseasonal cycles are effectively removed from the time series.
The main patterns of variability in the global monthly sea ice concentration anomalies over the 40-yr period are identified using the SOM method (Kohonen 2001). For each pattern of the global sea ice concentration anomalies, composite analysis is utilized to obtain the corresponding patterns of the atmospheric circulation and SST anomalies. Similar to the calculation of monthly sea ice anomalies, the monthly anomalies of atmospheric variables and SST at each grid point in the gridded dataset are calculated by subtracting the 40-yr averaged value for the month. Statistical significance of the difference is tested using the Welch’s t test.
The 40-yr time series is not detrended and thus the results of the analyses contain possible trends in monthly sea ice concentration anomalies. For the 40-yr time period, detrending removes not only long-term trends due possibly to global climate change associated with increasing greenhouse gas emissions, but also the influences from multidecadal climate variability modes such as PDO and AMO. To remove the long-term trend due to global climate change while retaining decadal to multidecadal internal climate system variability, time series much longer than 40 years are necessary. This can be achieved by repeating the aforementioned analyses using data from the Twentieth Century Reanalysis version 3 (20CR3) (Slivinski et al. 2019). Note that the sea ice concentration in 20CR3 was prescribed using the HadISST 2 (Rayner et al. 2003) and the data cover the globe with a 1° latitude and 1° longitude grid. For the current analysis, the period from 1836 to 2015 is utilized. Monthly sea ice anomalies are calculated and the linear trend in the 180-yr sea ice anomaly time series are removed before the SOM method is applied.
The SOM method was first introduced by Kohonen (1982) and has been widely used for mining large datasets in Earth and atmospheric sciences including sea ice datasets (Skific and Francis 2012). As a type of artificial neural network models, the SOM method utilizes the unsupervised learning technique with the goal to map high dimensional data onto a plane. Specifically, input vectors that are close in multidimensional space are mapped onto nearby nodes, or neurons, in the two-dimensional (2D) array. Each node in the 2D array has a spatial pattern reflecting continuous variability of the input data. The similarity of the spatial patterns of any two nodes is reflected through the Euclidean distance between the two nodes. Sammon maps (Sammon 1969) are usually utilized to visualize the relative distances between nodes. Details regarding SOM training are available in Hewitson and Crane (2002) and Reusch et al. (2005). For this application, all 40 years (1979–2018) of monthly global sea ice anomaly data are used to train SOM. The process of unsupervised training cycles through all 480 anomalous sea ice fields and maps each of them based on minimum Euclidean distance onto a 2D array with predefined number of rows and columns, which will be discussed later. Each node in the 2D array represents a typical pattern in the 480 monthly anomalous sea ice patterns over the 40 years.
The SOM method has been widely utilized in the analysis of sea ice concentration changes and their interactions with atmospheric circulations in the Arctic or the Antarctic (Reusch and Alley 2007; Higgins and Cassano 2009; Yu and Zhong 2018; Yu et al. 2018, 2019). For example, Higgins and Cassano (2009) used the SOM technique to investigate the response of synoptic-scale atmospheric circulation, temperature and precipitation in Arctic winters to projected depletion of sea ice as simulated by a community atmospheric model. Mills and Walsh (2014) used the SOM method to identify the main spatial patterns of the Arctic sea ice variability on the synoptic time scales. Lynch et al. (2016) examined Arctic circulation regimes related to different September sea ice modes in the Atlantic and Pacific sectors of the Arctic using SOM. Yu and Zhong (2018) and Yu et al. (2019) utilized the SOM technique to identify the main modes of variability in the Arctic autumn monthly sea ice concentration and summer daily sea ice concentration and linked the sea ice modes to known modes of atmospheric circulations. The SOM method was also applied to study Antarctic sea ice variability including the annual cycle of sea ice edge (Reusch and Alley 2007) and summer sea ice concentration variability and the underlying drivers (Yu et al. 2018). However, previous work on sea ice variability has centered either on the Arctic or the Antarctic. This study emphasizes on the concurrent changes in monthly Arctic and Antarctic sea ice concentration for the whole year to assess whether the main modes of sea ice in the Arctic and Antarctic are connected.
As mentioned earlier, the number of SOM nodes in a grid used to classify the input data is predetermined. Previous studies have suggested rules to choose an optimal SOM grid that captures the main modes of variability in the input data but avoids making the analysis and interpretation overly cumbersome (Lee and Feldstein 2013). One rule is based on changes in spatial correlations between the patterns in the SOM nodes and in the input data. For the current application, the spatial correlations between the monthly global sea ice concentration anomalies and the best matching SOM pattern are calculated and the averages of these correlations are shown in Table 1. For the 10 SOM grids tested with total node number ranging from 3 (3 × 1) to 16 (4 × 4), a relatively steep increase in the average correlation followed by steady values appears twice: once from 0.22 (3 nodes, 3 × 1 grid) to 0.29 (4 nodes, 2 × 2 grid) and another from 0.31 (10 nodes, 5 × 2 grid) to 0.34 (12 nodes, 4 × 3 grid). Considering a 4 node or a 2 × 2 SOM grid with a correlation of 0.29 may be too small to capture the modes of variability in the sea ice anomaly data, the 12 node (4 × 3) SOM grid is chosen for this application. Notice that an additional analysis using a 20 node (5 × 4) grid yielded similar results (Fig. S1 in the online supplemental material).
Spatial correlations (Corr.) between the monthly year-round Arctic sea ice concentration and the corresponding SOM pattern for each month from 1979 to 2018.
The contribution from a given SOM node to the trends in the annual sea ice concentration anomalies is calculated by the product of the sea ice concentration pattern depicted by the given node and the trend in the time series of the occurrence of that node normalized by 12, the total number of months in a year (Higgins and Cassano 2009; Lee and Feldstein 2013). The sum of the contributions over all 12 nodes represents the trend explained by all SOM nodes. Residual trends are obtained by subtracting the SOM-explained trends from the total trends. Statistical significance is determined using the Welch’s t test.
3. Results
a. Sea ice variability depicted by SOM nodes
The monthly Arctic and Antarctic sea ice concentration anomaly patterns depicted by the 4 × 3 or 12 node SOM grid are show in Fig. 1. The percentages explained by the nodes range from a minimum of 4% (nodes 2 and 7) to a maximum of 15.2% (node 1) and more than 50% are accounted for by only two pairs of nodes with opposite spatial patterns (nodes 1, 12 and nodes 9, 4). The most frequent patterns are represented by node 1 (15.2%) and its opposite node 12 (13.5%), showing monthly sea ice concentration fluctuating in phase across the Arctic (except for a small area in the Bering Sea) with spatially varying amplitude that peaks in the Barents Sea. However, a dipole pattern is depicted in Antarctic where sea ice concentration fluctuates out of phase between the Bellingshausen/Amundsen and eastern Ross Seas and the rest of the Southern Ocean.
The SOM patterns of the anomalous monthly sea ice concentration on a 4 × 3 grid for the 1979–2018 period. The percentages at the top left of each Arctic panel indicate the frequency of occurrences of the pattern. The number at the top left of each Antarctic panel denotes the SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
The SOM patterns of the anomalous monthly sea ice concentration on a 4 × 3 grid for the 1979–2018 period. The percentages at the top left of each Arctic panel indicate the frequency of occurrences of the pattern. The number at the top left of each Antarctic panel denotes the SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
The SOM patterns of the anomalous monthly sea ice concentration on a 4 × 3 grid for the 1979–2018 period. The percentages at the top left of each Arctic panel indicate the frequency of occurrences of the pattern. The number at the top left of each Antarctic panel denotes the SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
The second most frequent anomalous sea ice patterns are denoted by the opposite node pair, 4 (12.9%) and 9 (10.2%). Similar to node 1 and 12, the spatial pattern in the Arctic also displays in-phase fluctuations everywhere. Higher amplitudes are found not only in the Barents Sea, as in nodes 1 and 12, but also in the Beaufort and Eastern Siberian Seas. However, different from the dipole pattern seen in nodes 1 and 12, the spatial pattern depicted by nodes 4 and 9 in the Antarctic exhibits a zonal wavenumber-3 structure, suggesting that there may be a strong connection between the variability of sea ice captured by nodes 4 and 9 and the variability in the zonal wavenumber 3.
The other eight nodes, occurring at frequencies less than 10%, capture the transition between the opposite patterns discussed above. Despite the spatial variations in amplitudes, the monthly sea ice concentration tends to fluctuate in phase across the Arctic, with nodes in the left (right) two columns capturing a state of below (above) average sea ice concentration. However, for Antarctica, the fluctuations of the monthly sea ice not only vary in amplitude, but also in sign. The Antarctic patterns vary considerably among the 12 nodes. For example, nodes 1 and 5 show overall above-normal sea ice concentration, but for nodes 6 and 10 the extent and magnitude of positive and negative sea ice anomalies are comparable.
The frequency of occurrence of each node appears to vary considerably with month (Fig. 2, Table 2). Opposite nodes tend to occur during the same months, although the frequency is not the same. For example, the opposite nodes 1 and 12 appear most frequently from December through May, the cold (warm) seasons of the Northern (Southern) Hemisphere. In contrast, nodes 4 and 9 and, to some degree nodes 5 and 8, are much more common from June through November or during the warm (cold) seasons of the Northern (Southern) Hemisphere. The other nodes appear to spread across the year.
The number of occurrences of each SOM node for each month for the 1979–2018 period. The number on the left denotes the SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
The number of occurrences of each SOM node for each month for the 1979–2018 period. The number on the left denotes the SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
The number of occurrences of each SOM node for each month for the 1979–2018 period. The number on the left denotes the SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
Total number of occurrences for each SOM node during boreal winter–spring (DJFMAM) and summer–autumn (JJAASO) months for the 1979–2018 period. Italics denote that the differences between DJFMAM and JJAASO are at the above 95% confidence level.
Apart from the seasonal dependence, the occurrence of the nodes also exhibits decadal variation over the four decades (Fig. 3). Different from the seasonal dependence where the opposite nodes tend to occur in the same seasons, the two opposite nodes appear in separate decades. For example, nodes 1, 9 and 5, depicting negative sea ice concentration anomalies everywhere in the Arctic and overall positive anomalies in Antarctic, appear after 2005, whereas their counterparts, nodes 12, 4, and 8, tend to occur from the 1980s to the early 1990s. Such decadal separation in the occurrence between the patterns of below- and above-average sea ice concentrations contributes to a significant downward trend across the Arctic and a weak upward trend in most regions of the Antarctic. Other nodes tend to appear mostly in the middle period from the late 1990s to the early 2000s.
The number of occurrences of each SOM node from 1979 to 2018. The number on the left denotes the SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
The number of occurrences of each SOM node from 1979 to 2018. The number on the left denotes the SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
The number of occurrences of each SOM node from 1979 to 2018. The number on the left denotes the SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
b. Mechanisms behind the sea ice variability depicted by the SOM nodes
The spatial pattern and time variation of the nodes may be explained from the perspective of anomalous SST and large-scale atmospheric circulations. For each node, the corresponding anomalous SST pattern is obtained by averaging the monthly mean SST over all the months within the 40 years when that node occurs. Similarly, composite maps are also produced for each node for selected atmospheric variables including 500-hPa geopotential height, mean sea level pressure (MSLP), 2-m air temperature, 10-m wind, accumulated surface downward longwave radiation, and low-level specific humidity (Figs. 4–13).
Composite of anomalous sea surface temperature (SST) (°C) for each node. Dotted regions denote the above 95% confidence level.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
Composite of anomalous sea surface temperature (SST) (°C) for each node. Dotted regions denote the above 95% confidence level.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
Composite of anomalous sea surface temperature (SST) (°C) for each node. Dotted regions denote the above 95% confidence level.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
1) Nodes 1, 5, and 9
The average anomalous SST patterns corresponding to nodes 1, 5, and 9 show a similar spatial distribution (Fig. 4) but different magnitude. The SST spatial distributions resemble several known modes of low-frequency oscillations including the positive phase Atlantic multidecadal oscillation (AMO) (Enfield et al. 2001), negative phase Pacific decadal oscillation (PDO) (Mantua et al. 1997) and positive phase Indian Ocean basin mode (IOBM). These three nodes appear after 2000, consistent with the phase of the AMO and PDO (Yu et al. 2017). In the Arctic Ocean, the SST anomalies are all positive, consistent with the negative sea ice anomalies (Yu and Zhong 2018; Yu et al. 2019). In the Southern Ocean, SST anomalies change sign in a coherent way to the observed sea ice changes in the region. Yu et al. (2017) suggested that SST anomalies related to the AMO and PDO generate sea ice concentration anomalies in the polar regions by exciting Rossby wave trains propagating toward the extratropics.
The corresponding 500-hPa height and MSLP anomalies (Figs. 5 and 6a) differ somewhat among the three nodes. For nodes 1 and 9, positive SST anomalies over the tropical western Pacific Ocean generate two Rossby wave trains, which propagate northeastward and southeastward, respectively (Fig. 5). Similar wave trains were also noted to influence the Arctic (Ding et al. 2014) and Antarctic (Ding et al. 2011; Meehl et al. 2019). Over the Arctic, the two nodes show negative phase of the Arctic Oscillation (AO). Over the Antarctic, nodes 1 and 9 feature large 500-hPa height anomalies over the southeastern Pacific and the southern Atlantic Oceans. For node 9, in particular, there is a remarkable response of 500-hPa height to La Niña events, which is characterized with a strengthened Amundsen Sea low (ASL) (Turner 2004; Yuan and Li 2008). The pattern for node 5 displays a positive–negative–positive sandwich structure from midlatitude to the Antarctic continent, with a large positive center over southern Australia and a negative center over the Ross Sea. The positive SST anomalies over the North Pacific generate northeastward propagating Rossby wave trains, leading to a weak dipole anomaly in the Arctic.
Composite of anomalous 500-hPa geopotential height (gpm) for each node. Dotted regions denote the above 95% confidence level.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
Composite of anomalous 500-hPa geopotential height (gpm) for each node. Dotted regions denote the above 95% confidence level.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
Composite of anomalous 500-hPa geopotential height (gpm) for each node. Dotted regions denote the above 95% confidence level.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
Composite of (a) anomalous mean sea level pressure (Pascal), (b) 10-m wind vector, (c) surface air temperature (°C), (d) accumulated surface downward longwave radiation (105 W m−2), and (e) surface–750-hPa specific humidity (g kg−1) in the Northern Hemisphere north of 50°N for nodes 1, 5, and 9 (to the left of the figures in the top row). Dotted and shaded regions denote the above 95% confidence level.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
Composite of (a) anomalous mean sea level pressure (Pascal), (b) 10-m wind vector, (c) surface air temperature (°C), (d) accumulated surface downward longwave radiation (105 W m−2), and (e) surface–750-hPa specific humidity (g kg−1) in the Northern Hemisphere north of 50°N for nodes 1, 5, and 9 (to the left of the figures in the top row). Dotted and shaded regions denote the above 95% confidence level.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
Composite of (a) anomalous mean sea level pressure (Pascal), (b) 10-m wind vector, (c) surface air temperature (°C), (d) accumulated surface downward longwave radiation (105 W m−2), and (e) surface–750-hPa specific humidity (g kg−1) in the Northern Hemisphere north of 50°N for nodes 1, 5, and 9 (to the left of the figures in the top row). Dotted and shaded regions denote the above 95% confidence level.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
The pressure gradients between the western Eurasian continent and the Arctic Ocean generate anomalous southerly winds west of the anomalous high for node 1 (Figs. 6a,b). An anomalous high over Northern Europe induces anomalous southwesterly winds over the Norwegian Sea and the Barents Sea for node 5 (Figs. 6a,b). The anomalous winds advect warm and moist air into the Barents and Kara Seas (Fig. 6c), in alignment with the large negative sea ice concentration anomalies there. The positive water vapor–longwave radiation feedback (Figs. 6d,e) further increases surface air temperature and reduces sea ice concentration. The significant relationship among the surface air temperature, downward longwave radiation and specific humidity was also observed in previous studies (Ding et al. 2017; Gong et al. 2017; Yu et al. 2018, 2019). The anomalous pressure gradients give rise to anomalous southerly winds over the Pacific sector of the Arctic Ocean, reducing the sea ice concentration. For node 9, the MSLP pattern depicts a negative phase Arctic dipole (Wang et al. 2009). The corresponding anticyclonic circulation around the anomalous high south of the Barents Sea, the southerly winds over Canadian Arctic Archipelago, and the southerly winds over the Bering Strait transport warm and moisture-laden air into the Arctic Ocean, leading to negative sea ice concentration in the Arctic Ocean with small regional differences in the marginal seas.
The aforementioned mechanisms in the Arctic Ocean can also be invoked to explain the anomalous sea ice patterns depicted by these nodes in the Southern Ocean. The composite maps for the anomalous atmospheric variables are shown in Fig. 7. For nodes 1 and 5, the anomalous northerly winds related to the anomalous MSLP patterns transport warm and moist air southward into the Amundsen Sea, contributing to sea ice melt there (Figs. 7a,b). Positive feedback between lower-level water vapor and downward longwave radiation increases surface air temperature and decreases sea ice concentration (Figs. 7c–e). The anomalous northerly winds also mechanically redistribute sea ice onto the coast. The processes are reversed over the Weddell and Ross Seas as southerly anomalous winds prevail in these regions. In the Bellingshausen Sea, the dynamic effects of southerly anomalous winds result in negative and positive sea ice concentration anomalies in its southern and northern parts. Similarly, the dynamic and thermodynamic effects expected of the anomalous atmospheric circulation patterns corroborate the anomalous sea ice patterns captured by node 9. The patterns of all atmospheric variables analyzed here consistently indicate the structure of zonal wavenumber 3. The anomalous meridional winds related to the anomalous 500-hPa height and MSLP largely explain the anomalous sea ice concentration patterns.
As in Fig. 6, but for the Southern Hemisphere south of 50°S. The number to the left of the figures in the top row denotes the SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
As in Fig. 6, but for the Southern Hemisphere south of 50°S. The number to the left of the figures in the top row denotes the SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
As in Fig. 6, but for the Southern Hemisphere south of 50°S. The number to the left of the figures in the top row denotes the SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
2) Nodes 4, 8, and 12
For nodes 12, 8, and 4, which have near opposite spatial patterns to those of nodes 1, 5, and 9 and appear primarily before 2000, the corresponding SST composite patterns are complete reversals of their counterparts (Fig. 4). In other words, the SST patterns resemble negative phase AMO, positive phase PDO, and negative phase IOBM, with negative SST anomalies over the Arctic Ocean (Fig. 4). Similar wave trains also exist in the anomalous 500-hPa height fields (Fig. 5), which is opposite to those in the study of Yu et al. (2017). The spatial pattern of 500-hPa height anomalies for node 4 features a positive phase AO. Anomalous northerly surface winds over the Atlantic sector of the Arctic Ocean (Fig. 8b) are conducive to decreased surface air temperature (Fig. 8c) and increased Arctic sea ice extent (Fig. 1). The process is further enhanced by drier air and lower downwelling longwave radiation (Figs. 8d,e). The Antarctic 500-hPa height and MSLP anomalies for node 4 show negative phase SAM with a weakened ASL (Figs. 5 and 9a). The anomalous southerly (northerly) surface winds and the associated negative (positive) surface air temperature and offshore (onshore) ice flow corroborate the positive (negative) sea ice anomalies over the western Weddell and Bellingshausen (Amundsen and Ross) Seas (Figs. 9b–e).
As in Fig. 6, but for nodes 4, 8, and 12. The number to the left of the figures in the top row denotes the SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
As in Fig. 6, but for nodes 4, 8, and 12. The number to the left of the figures in the top row denotes the SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
As in Fig. 6, but for nodes 4, 8, and 12. The number to the left of the figures in the top row denotes the SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
As in Fig. 7, but for nodes 4, 8, and 12. The number to the left of the figures in the top row denotes the SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
As in Fig. 7, but for nodes 4, 8, and 12. The number to the left of the figures in the top row denotes the SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
As in Fig. 7, but for nodes 4, 8, and 12. The number to the left of the figures in the top row denotes the SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
For node 8, the anomalous 500-hPa height and MSLP patterns are characterized by negative (positive) anomalies over the western (eastern) Arctic, resembling the negative phase of the Arctic dipole (AD) (Figs. 5 and 8a). The anomalous southerly winds over the Barents Sea/Frame Straight sectors (Fig. 7) act to reduce sea ice export from the Arctic Ocean (Wang et al. 2009). In addition, the fact that node 8 appears primarily in Northern Hemisphere summer (Fig. 2) helps explain the positive sea ice anomalies in the Barents Sea region (Fig. 1), because transport of the Atlantic water in summer is found to be less effective at melting Barents Sea ice (Årthun et al. 2012). The anomalous cyclonic circulation over Greenland decreases heat and moisture import from northern Canada into the Arctic Ocean and the related water vapor–longwave radiation feedback over the Arctic (Ding et al. 2017), thus increasing Arctic sea ice extent. In contrast, the anomalous northerly winds from the North Pole (Fig. 8b) transport cold and dry air to the Pacific sector of the Arctic Ocean (Fig. 8c), increasing the sea ice concentration in the region. The decreases of water vapor over most of the Arctic Ocean also favor smaller downward longwave radiation, contributing to the Arctic sea ice increase (Figs. 8d,e). The corresponding anomalous 500-hPa height and MSLP over the Antarctic for node 8 is characterized by the zonal wavenumber-3 structure (Fig. 5). The associated anomalous patterns of surface wind, temperature, low-level moisture and downward longwave radiation (Fig. 9) give support to the anomalous sea ice distribution in the Southern Ocean captured by node 8 (Fig. 1).
For node 12, which occur primarily in Northern Hemisphere cold season (Fig. 2), the 500-hPa height anomalies are negative, opposite to those of MSLP, over the Arctic Ocean (Fig. 5). The anomalous anticyclonic surface wind vectors around the anomalous MSLP high over the Atlantic sector of the Arctic Ocean produce sea ice outflow from the Arctic Ocean. These anomalous winds also transport cold and dry air to the East Greenland and the Barents Seas, increasing sea ice concentration in these regions. The water vapor–longwave radiation feedback further contributes to the sea ice increase. In the Southern Hemisphere warm season, fewer areas in the Antarctic have significant 500-hPa height, MSLP, and 10-m wind field anomalies (Figs. 6 and 9). But the patterns of water vapor, downward longwave radiation, and surface air temperature anomalies (Fig. 9) are closely related to the patterns of the sea ice concentration anomalies (Fig. 1).
3) Other nodes
For the other six nodes, the anomalous SST patterns are somewhat similar to the central Pacific El Niño (nodes 2, 3, 6, and 7) or La Niña (nodes 10 and 11) (Fig. 4). However, the magnitudes of the SST anomalies for these nodes are smaller than those for the aforementioned nodes. Apart from nodes 7 and 11 with relatively uniform negative anomalies in the Indian Ocean, the SST anomalies in the Indian and Atlantic Oceans exhibit considerable spatial variability. The SST anomalies over the central Pacific Ocean influence the atmospheric variability through Rossby wave trains propagating northeastward (Fig. 5). Despite the smaller SST anomalies, some of these nodes exhibit distinct atmospheric circulation patterns. The anomalous wind fields related to pressure anomalies and water vapor–longwave radiation feedback (Figs. 10 and 11) largely support the anomalous Arctic sea ice concentration (Fig. 1). The SST anomalies over the central Pacific Ocean trigger a Rossby wave train propagating into the Antarctic (Fig. 5), a feature also shown in the study of Ciasto et al. (2015). Over the Southern Ocean the height anomalies feature a zonal wavenumber-3 structure (Fig. 5). The anomalous MSLP patterns for the six nodes show different phases of the SAM mode (positive phase for nodes 7, 10, and 11; negative phase for nodes 2 and 3) with the exception of node 6 (Figs. 12 and 13). The asymmetric structure of the SAM mode produces anomalous surface wind and temperature fields (Figs. 12 and 13) that are consistent with the anomalous Antarctic sea ice patterns (Figs. 1).
As in Fig. 6, but for nodes 2, 6, and 10. The number to the left of the figures in the top row denotes the SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
As in Fig. 6, but for nodes 2, 6, and 10. The number to the left of the figures in the top row denotes the SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
As in Fig. 6, but for nodes 2, 6, and 10. The number to the left of the figures in the top row denotes the SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
As in Fig. 6, but for nodes 3, 7, and 11. The number to the left of the figures in the top row denotes the SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
As in Fig. 6, but for nodes 3, 7, and 11. The number to the left of the figures in the top row denotes the SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
As in Fig. 6, but for nodes 3, 7, and 11. The number to the left of the figures in the top row denotes the SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
As in Fig. 7, but for nodes 2, 6, and 10. The number to the left of the figures in the top row denotes the SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
As in Fig. 7, but for nodes 2, 6, and 10. The number to the left of the figures in the top row denotes the SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
As in Fig. 7, but for nodes 2, 6, and 10. The number to the left of the figures in the top row denotes the SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
As in Fig. 7, but for nodes 3, 7, and 11. The number to the left of the figures in the top row denotes the SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
As in Fig. 7, but for nodes 3, 7, and 11. The number to the left of the figures in the top row denotes the SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
As in Fig. 7, but for nodes 3, 7, and 11. The number to the left of the figures in the top row denotes the SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
In short, the decadal variability of the occurrences for the three paired nodes, (1, 12), (4, 9) and (5, 8), appears to be related to the interdecadal variability of the AMO, PDO and IOBM indices. The variability of the occurrences for the other nodes seems to be associated with the central Pacific ENSO and spatially heterogeneous variability in the Indian and Atlantic Oceans on interannual time scales. The polar sea ice anomalies for these nodes may also be associated with atmospheric variability in mid and high latitudes, such as AO, AD, and SAM.
c. Sea ice trends explained by the SOM nodes
As shown in Fig. 14, trends in sea ice concentration over the 40 years are consistently negative across the Arctic. The negative trends peak in the Barents Sea and are also significant in the Sea of Okhotsk, Baltic Sea, Hudson Bay, and Labrador Sea. In contrast, trends in the Antarctic sea ice concentration vary from significantly negative in the Bellingshausen and Amundsen Seas and the northern portions of the Southern Atlantic and Indian Oceans to positive in the rest of the Southern Ocean. The overall sea ice extent displays a significant decreasing (increasing) trend in the Arctic (Antarctic), which was also observed by Parkinson (2019).
Trends in annual mean sea ice concentration (yr−1) for the 1979–2018 period. Dotted regions denote the above 95% confidence level.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
Trends in annual mean sea ice concentration (yr−1) for the 1979–2018 period. Dotted regions denote the above 95% confidence level.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
Trends in annual mean sea ice concentration (yr−1) for the 1979–2018 period. Dotted regions denote the above 95% confidence level.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
The trends explained by each node are shown in Fig. 15 and Table 3. The majority of these trends are explained by the three dominant pairs, namely, (1, 12), (9, 4), and (5, 8). The spatial pattern of the trends explained by each of these six nodes (Fig. 15) closely resembles, or is opposite to, that of the sea ice anomalies depicted by the node (Fig. 1). Averaging over the Arctic and the Antarctic separately, the sum of the domain-averaged trend amounts explained by the two opposite nodes is the same for the Arctic and the Antarctic, ranging from 30% for nodes (1, 12), to 16% for nodes (9, 4), and to 11% for nodes (5, 8). In total, the 12 nodes explain 60% of the sea ice trends in the Arctic and in the Antarctic (Table 4).
Trends in annual Arctic and Antarctic sea ice concentration explained by each SOM node (yr−1) for the 1979–2018 period. The number to the left of the panels denotes the SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
Trends in annual Arctic and Antarctic sea ice concentration explained by each SOM node (yr−1) for the 1979–2018 period. The number to the left of the panels denotes the SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
Trends in annual Arctic and Antarctic sea ice concentration explained by each SOM node (yr−1) for the 1979–2018 period. The number to the left of the panels denotes the SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
Trends in the frequency of occurrence for each SOM node (month yr−1). An asterisk indicates a significant trend at the above 95% confidence level.
The fraction (%) of the total sea ice trend represented by each node.
The spatial pattern of sea ice trends explained by all 12 nodes (Fig. 16a) is similar to the pattern of the total trends in the Arctic and Antarctic (Fig. 14). The pattern of the residual trends (Fig. 16b) is also similar to that of the total trends (Fig. 14). There are, however, considerable spatial variations in the ratios of the SOM-explained sea ice trends to the total trends (Fig. 16c). In the Arctic, large ratios are found in the Greenland Sea, Sea of Okhotsk, and Hudson Bay and the ratios are generally larger in the eastern than western Arctic Ocean (Fig. 16c). In the Southern Ocean, larger ratios are found in the Weddell Sea and the Amundsen Sea.
(a) SOM-explained and (b) residual annual sea ice concentration trends (yr−1), and (c) the ratios of the SOM-explained to the total sea ice concentration trends for the 1979–2018 period. The blue–red color bar is for (a) and (b) and the yellow–red color bar is for (c).
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
(a) SOM-explained and (b) residual annual sea ice concentration trends (yr−1), and (c) the ratios of the SOM-explained to the total sea ice concentration trends for the 1979–2018 period. The blue–red color bar is for (a) and (b) and the yellow–red color bar is for (c).
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
(a) SOM-explained and (b) residual annual sea ice concentration trends (yr−1), and (c) the ratios of the SOM-explained to the total sea ice concentration trends for the 1979–2018 period. The blue–red color bar is for (a) and (b) and the yellow–red color bar is for (c).
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
d. Sea ice variability on century time scale
Extending the SOM analysis of sea ice anomalies from the period of the most recent 40 years to 180 years yielded broadly similar spatial patterns (Figs. 17 and 18) in both the Arctic and the Antarctic, although the corresponding node numbers differ. For example, the opposite patterns depicted by the node pair (1, 12) in Fig. 1 are now captured by (12, 1) in Fig. 17. A careful comparison, however, reveal some differences in magnitude, extent and locations of maximum anomalies between the short- and long-term datasets, which for some nodes are clearly noticeable. For example, node 1 in Fig. 17 shows negative sea ice anomalies in the Laptev Sea, while sea ice anomalies are consistently positive across the Arctic Ocean in Fig. 1. Nodes 5 and 8 in Fig. 17 show a dipole structure of sea ice anomalies with the opposite variability between the Pacific and Atlantic sectors of the Arctic Ocean, which are not observed in Fig. 1. In general, the magnitude of sea ice anomalies in the Pacific sector of the Arctic Ocean for the 40-yr period is larger than that for the 180-yr period.
The SOM patterns of the detrended anomalous monthly sea ice concentration on a 4 × 3 grid for the 1836–2015 period. The numbers at the top-left and top-right corners of each panel indicate the SOM node and the frequency of occurrences of the node, respectively.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
The SOM patterns of the detrended anomalous monthly sea ice concentration on a 4 × 3 grid for the 1836–2015 period. The numbers at the top-left and top-right corners of each panel indicate the SOM node and the frequency of occurrences of the node, respectively.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
The SOM patterns of the detrended anomalous monthly sea ice concentration on a 4 × 3 grid for the 1836–2015 period. The numbers at the top-left and top-right corners of each panel indicate the SOM node and the frequency of occurrences of the node, respectively.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
The number of occurrences of each SOM node in 180 years from 1836 to 2015. The number to the left indicates SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
The number of occurrences of each SOM node in 180 years from 1836 to 2015. The number to the left indicates SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
The number of occurrences of each SOM node in 180 years from 1836 to 2015. The number to the left indicates SOM node.
Citation: Journal of Climate 35, 9; 10.1175/JCLI-D-21-0756.1
In Antarctica, the dipole structure identified in the 40-yr time series (nodes 1, 12 in Fig. 1) is retained in the 180-yr data (nodes 4, 5 and nodes 8, 12 in Fig. 14), but the extent of sea ice concentration anomalies is somewhat larger for the 180-yr period than that for the 40-yr period, especially in the Indian sector of the Southern Ocean. The wavenumber-3 structure in the 40-yr time series (nodes 4, 9 in Fig. 1) is replaced with the wavenumber-2 structure in the 180-yr time series, where there is no dipole structure in the Indian sector of the Southern Ocean. Several nodes (6, 9, and 10) show a nearly uniform patterns in Fig. 17, which is not seen in Fig. 1.
The frequency of occurrences also exhibits an interdecadal variability (Fig. 18), which is significantly correlated with the annual AMO and PDO indices (Table 5). Specifically, the occurrences of nodes 6 and 11 (4, 5, and 9) show significant positive (negative) correlations with the annual AMO time series, while nodes 8 and 9 (1 and 2) display significant positive (negative) correlations with the annual PDO time series.
Correlation coefficients between the annual PDO and AMO indices and the annual number of occurrences for each SOM node. The monthly sea ice concentration data are detrended prior to being decomposed using SOM. An asterisk indicates significant correlation at the above 95% confidence level.
Finally, the mean and standard deviation of the global sea ice concentration for the 40-yr and 180-yr periods are compared (Figs. S2 and S3). The spatial distributions are similar between the two periods in both the Arctic and the Antarctic, but the magnitudes are smaller for the 40-yr period compared to the 180-yr period. The 40-yr period exhibits larger variance in the Arctic, but smaller variance in the Antarctic. The 180-yr time series of Arctic and Antarctic sea ice extent (Fig. S4) shows generally smaller variability prior to 1900. It is worth noting that the sea ice data from 1836 to 2015 came from three sources (Slivinski et al. 2019): HadISST2.2 (Titchner and Rayner 2014) from 1972 onward, HadISST2.3 from 1850 to 1971, and the 1860–91 HadISST2.3 climatology prior to 1850, which explains the lack of variability from 1836 to 1849.
4. Summary and discussion
The main modes of variability in the global monthly sea ice concentration data over the recent four decades (1979–2018) are extracted using the SOM clustering method. Depicted by a 4 × 3 or 12 node SOM grid, the monthly sea ice concentration is dominated by in-phase variability across the Arctic, but out-of-phase variability between regions of the Antarctic. Specifically, the 12 nodes can be divided equally into two groups of broadly similar patterns: one group (3, 4, 6, 7, 9, 10) show relatively uniform anomaly values in the Arctic and a zonal wavenumber-3 structure in the Antarctic, and another group (1, 2, 5, 8, 11, 12) display a pronounced anomaly peak in the Barents Sea in the Arctic and a dipole pattern in the Antarctic with out-of-phase variations between the Bellingshausen and Amundsen Seas and the rest of the Southern Ocean.
The occurrences of some patterns exhibit seasonal variability. Nodes 1 and 12 appear predominantly in boreal winter (austral summer), in opposite to nodes 4 and 9 that occur mostly in boreal summer (austral winter). The former corresponds to the spatial patterns of leading EOF modes of anomalous sea ice concentration in boreal winter (DJF) and spring (MAM), and the latter is consistent with the spatial patterns of EOF modes in boreal summer (JJA) and autumn (SON) (Yu et al. 2017). There is a pronounced decadal variability in the occurrences of the three dominant node pairs: (1, 12), (9, 4), and (5, 8). In particularly, nodes 12, 4, 8, which have positive anomalies across the Arctic and overall negative anomalies in the Antarctic, occur mainly in the 1980s and 1990s, while their counterparts with opposite spatial patterns, nodes 1, 9, and 5, appear predominantly after 2000. Such decadal separation in the occurrences of the nodes with overall positive anomalies from those of overall negative anomalies contributes to the long-term negative (positive) overall trends in sea ice concentration in the Arctic (Antarctic).
The occurrences of some nodes can be linked to known modes of decadal variability of SST and atmospheric circulations. Specifically, the occurrences of nodes 1, 5, and 9 are connected to the positive phase of the annual AMO and IOBM indices and the negative phase PDO index; their opposite nodes, 12, 8, and 4, are related to the opposite phase of the same indices. The other six nodes, which tend to occur more frequently in the central part of 1979–2018, appear to have some connection to the central Pacific El Niño. Large magnitudes of MSLP anomalies for the other six nodes also suggest a potential connection of sea ice variability with atmospheric internal variability modes, such as AO (Rigor et al. 2002), AD (Wang et al. 2009) and the other modes (Lynch et al. 2016) for the Arctic, and SAM (Hall and Visbeck 2002; Sen Gupta and England 2006) and the Pacific–South American (PSA) pattern (Irving and Simmonds 2016; O’Kane et al. 2017) for the Antarctic. Although the other six nodes make small contributions to the long-term trends in the polar sea ice concentration, they play an important role in sea ice variability on interannual time scales. This result on interannual time scales was not reflected in the study of Yu et al. (2017). The spatial patterns of the leading variability modes can be largely explained by mechanical redistribution of sea ice and the thermodynamic feedback processes associated with the anomalous atmospheric circulations induced by SST and MSLP anomalies.
The 12 nodes explain about 60% of the total sea ice trend in the Arctic and in the Antarctic over the past four decades (1979–2018), among which 57% is associated with the trends of occurrences of the two dominant pairs (nodes 1, 2 and nodes 4, 9). With record length of 40 years, it is nearly impossible to fully distinguish trends arising from multidecadal oscillations, such as PDO and AMO, from those caused by external forcing, such as greenhouse gas emission and ozone depletion. Hemisphere-specific variability may also contribute to the changes in sea ice concentration, which is not reflected in a joint SOM analysis of sea ice in the two poles. An extension of the SOM analyses from four decades (1979–2018) to 180 years (1836–2015) did not change the dominant patterns, but some differences exist. The relatively uniform sea ice anomaly pattern in the Arctic Ocean in the 40-yr sea ice data is not observed in the long-term data, while a new dipole structure of opposite variability between the Pacific and Atlantic sectors emerged in the 180-yr data. In the Antarctic, the zonal wavenumber-3 structure is replaced by the wavenumber-2 structure, in addition to a new nearly uniform sea ice variability pattern in the long-term data. The linkage between the leading sea ice variability modes and the AMO and PDO indices identified in the 40-yr time series is confirmed in the 180-yr time series. The results from the 180-yr time series need to be taken with caution, as sea ice data, particularly in the Antarctic, are unreliable prior to the 1960s or the satellite era.
The current study is, to our knowledge, the first to apply the SOM clustering algorithm to extract the main modes of variability in the global monthly anomalies of sea ice concentration data. The method is superior to the EOF method due to no requirement of orthogonality. The joint hemispheric analyses here have not only confirmed the results from several previous studies (e.g., Yu and Zhong 2018; Yu et al. 2018, 2019) that focused on either the Arctic or the Antarctic sea ice variability, but also reveal the anomalous sea ice patterns in the opposite polar regions that tend to appear concurrently, along with the types of large-scale oceanic and atmospheric conditions under which they occur.
Acknowledgments.
This study is financially supported by the National Natural Science Foundation of China (41941009) and the National Key R&D Program of China (2018YFA0605701, 2019YFC1509102).
Data availability statement.
Monthly sea ice concentration data are produced through the National Aeronautics and Space Administration (NASA) Team sea ice algorithm (https://nsidc.org/data/pm/nasateam-index) from the U.S. National Snow and Ice Data Center (ftp://sidads.colorado.edu/DATASETS/nsidc0051_gsfc_nasateam_seaice/final-gsfc). The ERA-Interim data are available from https://apps.ecmwf.int/datasets/data/interim-full-daily/levtype=sfc/. The monthly sea surface temperature (SST) data from the U.S. National Oceanic and Atmospheric Administration (NOAA) Extended Reconstructed Sea Surface Temperature (ERSST) version 5 (ERSST v5) are available online (https://www.ncdc.noaa.gov/data-access/marineocean-data/extended-reconstructed-sea-surface-temperature-ersst-v5). Monthly sea ice concentration data from 20th Reanalysis V3 are derived from the following website (https://psl.noaa.gov/data/gridded/data.20thC_ReanV3.monolevel.html). The monthly PDO indices were obtained from NOAA (https://www.ncdc.noaa.gov/teleconnections/pdo/) and the monthly AMO indices were downloaded from NOAA (http://www.psl.noaa.gov/data/timeseries/AMO/). The code of SOM analysis is available from the following website (http://www.cis.hut.fi/projects/somtoolbox/). Other computer code used to generate results is available upon request.
REFERENCES
Abrahamsen, E. P., and Coauthors, 2020: Antarctica and the Southern Ocean [in “State of the Climate in 2019”]. Bull. Amer. Meteor. Soc., 101 (8), S287–S320, https://doi.org/10.1175/BAMS-D-20-0090.1.
Andersen, J. K., and Coauthors, 2020: The Arctic [in “State of the Climate in 2019”]. Bull. Amer. Meteor. Soc., 101 (8), S239–S286, https://doi.org/10.1175/BAMS-D-20-0086.1.
Årthun, M., T. Eldevik, L. H. Smedsrud, Ø. Skagseth, and R. B. Ingvaldsen, 2012: Quantifying the influence of Atlantic heat on Barents Sea ice variability and retreat. J. Climate, 25, 4736–4743, https://doi.org/10.1175/JCLI-D-11-00466.1.
Bracegirdle, T. J., and G. J. Marshall, 2012: The reliability of Antarctic tropospheric pressure and temperature in the latest global reanalyses. J. Climate, 25, 7138–7146, https://doi.org/10.1175/JCLI-D-11-00685.1.
Cavalieri, D. J., and C. L. Parkinson, 2012: Arctic sea ice variability and trends, 1979–2010. The Cryosphere, 6, 881–889, https://doi.org/10.5194/tc-6-881-2012.
Cavalieri, D. J., C. L. Parkinson, P. Gloersen, and H. J. Zwally, 1996: Sea ice concentrations from Nimbus-7 SMMR and DMSP SSM/I-SSMIS Passive Microwave Data, Version 1. NASA National Snow and Ice Data Center Distributed Active Archive Center, accessed 17 March 2021, https://doi.org/10.5067/8GQ8LZQVL0VL.
Chylek, P. C., K. Folland, G. Lesins, and M. K. Dubey, 2010: Twentieth century bipolar seesaw of the Arctic and Antarctic surface air temperature. Geophys. Res. Lett., 37, L08703, https://doi.org/10.1029/2010GL042793.
Ciasto, L. M., G. R. Simpkins, and M. H. England, 2015: Teleconnections between tropical Pacific SST anomalies and extratropical Southern Hemisphere climate. J. Climate, 28, 56–65, https://doi.org/10.1175/JCLI-D-14-00438.1.
Dee, D. P., and Coauthors, 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553–597, https://doi.org/10.1002/qj.828.
Ding, Q., E. J. Steig, D. S. Battisti, and M. Küttel, 2011: Winter warming in West Antarctica caused by central tropical Pacific warming. Nat. Geosci., 4, 398–403, https://doi.org/10.1038/ngeo1129.
Ding, Q., J. M. Wallace, D. S. Battisti, E. J. Steig, A. J. E. Gallant, H.-J. Kim, and L. Geng, 2014: Tropical forcing of the recent rapid Arctic warming in northeastern Canada and Greenland. Nature, 509, 209–212, https://doi.org/10.1038/nature13260.
Ding, Q., J. M. Wallace, D. S. Battisti, E. J. Steig, A. J. E. Gallant, H.-J. Kim, and L. Geng, 2017: Influence of high-latitude atmospheric circulation changes on summertime Arctic Sea ice. Nat. Climate Change, 7, 289–295, https://doi.org/10.1038/nclimate3241.
Eayrs, C., D. Holland, D. Francis, T. Wagner, R. Kumar, and X. Li, 2019: Understanding the seasonal cycle of Antarctic sea ice extent in the context of longer-term variability. Rev. Geophys., 57, 1037–1064, https://doi.org/10.1029/2018RG000631.
Enfield, D. B., A. M. Mestas-Nunez, and P. J. Trimble, 2001: The Atlantic Multidecadal Oscillation and its relationship to rainfall and river flows in the continental U.S. Geophys. Res. Lett., 28, 2077–2080, https://doi.org/10.1029/2000GL012745.
Gao, K., A. Duan, D. Chen, and G. Wu, 2019: Surface energy budget diagnosis reveals possible mechanism for the different warming rate among Earth’s three poles in recent decades. Sci. Bull., 64, 1140–1143, https://doi.org/10.1016/j.scib.2019.06.023.
Gong, T., S. Feldstein, and S. Lee, 2017: The role of downward infrared radiation in the recent Arctic winter warming trend. J. Climate, 30, 4937–4949, https://doi.org/10.1175/JCLI-D-16-0180.1.
Hall, A., and M. Visbeck, 2002: Synchronous variability in the Southern Hemisphere atmosphere, sea ice, and ocean resulting from the annular mode. J. Climate, 15, 3043–3057, https://doi.org/10.1175/1520-0442(2002)015<3043:SVITSH>2.0.CO;2.
Hersbach, H., and Coauthors, 2020: The ERA5 global reanalysis. Quart. J. Roy. Meteor. Soc., 146, 1999–2049, https://doi.org/10.1002/qj.3803.
Hewitson, B. C., and R. G. Crane, 2002: Self-organizing maps: Applications to synoptic climatology. Climate Res., 22, 13–26, https://doi.org/10.3354/cr022013.
Higgins, M. E., and J. J. Cassano, 2009: Impacts of reduced sea ice on winter Arctic atmospheric circulation, precipitation, and temperature. J. Geophys. Res., 114, D16107, https://doi.org/10.1029/2009JD011884.
Hobbs, W. R., R. Massom, S. Stammerjohn, P. Reid, G. Williams, and W. Meier, 2016: A review of recent changes in Southern Ocean sea ice, their drivers and forcings. Global Planet. Change, 143, 228–250, https://doi.org/10.1016/j.gloplacha.2016.06.008.
Holland, P. R., and N. Kimura, 2016: Observed concentration budgets of Arctic and Antarctic sea ice. J. Climate, 29, 5241–5249, https://doi.org/10.1175/JCLI-D-16-0121.1.
Huang, B., and Coauthors, 2017: Extended Reconstructed Sea Surface Temperature version 5 (ERSSTv5), Upgrades, validations, and intercomparisons. J. Climate, 30, 8179–8205, https://doi.org/10.1175/JCLI-D-16-0836.1.
Irving, D., and I. Simmonds, 2016: A new method for identifying the Pacific–South American pattern and its influence on regional climate variability. J. Climate, 29, 6109–6125, https://doi.org/10.1175/JCLI-D-15-0843.1.
Jones, P. D., and D. H. Lister, 2015: Antarctic near‐surface air temperatures compared with ERA-Interim values since 1979. Int. J. Climatol., 35, 1354–1366, https://doi.org/10.1002/joc.4061.
Kohonen, T., 1982: Self-organized formation of topologically correct feature maps. Biol. Cybern., 43, 59–69, https://doi.org/10.1007/BF00337288.
Kohonen, T., 2001: Self-Organizing Maps. 3rd ed. Springer, 501 pp.
Lee, S., and S. B. Feldstein, 2013: Detecting ozone- and greenhouse gas–driven wind trends with observational data. Science, 339, 563–567, https://doi.org/10.1126/science.1225154.
Lindsay, R., M. Wensnahan, A. Schweiger, and J. Zhang, 2014: Evaluation of seven different atmospheric reanalysis products in the Arctic. J. Climate, 27, 2588–2606, https://doi.org/10.1175/JCLI-D-13-00014.1.
Lynch, A. H., M. C. Serreze, E. N. Cassano, A. D. Crawford, and J. Stroeve, 2016: Linkages between Arctic summer circulation regimes and regional sea ice anomalies. J. Geophys. Res. Atmos., 121, 7868–7880, https://doi.org/10.1002/2016JD025164.
Maksym, T., 2019: Arctic and Antarctic sea ice change: Contrasts, commonalities, and causes. Annu. Rev. Mar. Sci., 11, 7868–7880, https://doi.org/10.1146/annurev-marine-010816-060610.
Mantua, N. J., S. R. Hare, Y. Zhang, J. M. Wallace, and R. C. Francis, 1997: A Pacific interdecadal climate oscillation with impacts on salmon production. Bull. Amer. Meteor. Soc., 78, 1069–1079, https://doi.org/10.1175/1520-0477(1997)078<1069:APICOW>2.0.CO;2.
Meehl, G. A., J. M. Arblaster, C. T. Y. Chung, M. M. Holland, A. DuVivier, L. Thompson, D. Yang, and C. M. Bitz, 2019: Sustained ocean changes contributed to sudden Antarctic Sea ice retreat in late 2016. Nat. Commun., 10, 14, https://doi.org/10.1038/s41467-018-07865-9.
Meier, W. N., and Coauthors, 2014: Arctic sea ice in transformation: A review of recent changes and impacts on biology and human activity. Rev. Geophys., 52, 185–217, https://doi.org/10.1002/2013RG000431.
Mills, C. M., and J. E. Walsh, 2014: Synoptic activity associated with sea ice variability in the Arctic. J. Geophys. Res. Atmos., 119, 12 117–12 131, https://doi.org/10.1002/2014JD021604.
O’Kane, T. J., D. P. Monselesan, and J. S. Rishey, 2017: A multiscale reexamination of the Pacific–South American pattern. Mon. Wea. Rev., 145, 379–402, https://doi.org/10.1175/MWR-D-16-0291.1.
Ordoñez, A. C., C. M. Bitz, and E. Blanchard-Wrigglesworth, 2018: Processes controlling Arctic and Antarctic sea ice predictability in the Community Earth System Model. J. Climate, 31, 9771–9786, https://doi.org/10.1175/JCLI-D-18-0348.1.
Parkinson, C. L., 2014: Global sea ice coverage from satellite data: Annual cycle and 35-yr trends. J. Climate, 27, 9377–9382, https://doi.org/10.1175/JCLI-D-14-00605.1.
Parkinson, C. L., 2019: A 40-y record reveals gradual Antarctic sea ice increases followed by decreases at rates far exceeding the rates seen in the Arctic. Proc. Natl. Acad. Sci. USA, 116, 14 414–14 423, https://doi.org/10.1073/pnas.1906556116.
Rayner, N. A., D. E. Parker, E. B. Horton, C. K. Folland, L. V. Alexander, D. P. Rowell, E. C. Kent, and A. Kaplan, 2003: Global analyses of sea surface temperature, sea ice, and night marine air temperature since the late nineteenth century. J. Geophys. Res., 108, 4407, https://doi.org/10.1029/2002JD002670.
Reusch, D. B., and R. B. Alley, 2007: Antarctic sea ice: A self-organizing map-based perspective. Ann. Glaciol., 46, 391–396, https://doi.org/10.3189/172756407782871549.
Reusch, D. B., R. B. Alley, and B. C. Hawitson, 2005: Relative performance of self organizing maps and principle component analysis in pattern extraction from synthetic climatological data. Polar Geogr., 29, 188–212, https://doi.org/10.1080/789610199.
Rigor, I. G., J. M. Wallace, and R. L. Colony, 2002: Response of sea ice to the Arctic Oscillation. J. Climate, 15, 2648–2663, https://doi.org/10.1175/1520-0442(2002)015<2648:ROSITT>2.0.CO;2.
Sammon, J. W., 1969: A non-linear mapping for data structure analysis. IEEE Trans. Comput., C-18, 401–409, https://doi.org/10.1109/T-C.1969.222678.
Sen Gupta, A., and M. H. England, 2006: Coupled ocean–atmosphere–ice response to variations in the southern annular mode. J. Climate, 19, 4457–4486, https://doi.org/10.1175/JCLI3843.1.
Serreze, M. C., and W. N. Meier, 2019: The Arctic’s sea ice cover: Trends, variability, predictability, and comparisons to Antarctic. Ann. N.Y. Acad. Sci., 1436, 36–53, https://doi.org/10.1111/nyas.13856.
Simmonds, I., 2015: Comparing and contrasting the behavior of Arctic and Antarctic sea ice over the 35 year period 1979–2013. Ann. Glaciol., 56, 18–28, https://doi.org/10.3189/2015AoG69A909.
Skific, N., and J. Francis, 2012: Self-organizing maps: A powerful tool for the atmospheric sciences. Applications of Self-Organizing Maps, M. Johnsson, Ed., IntechOpen, 251–268.
Slivinski, L. C., and Coauthors, 2019: Towards a more reliable historical reanalysis: Improvements for version 3 of the Twentieth Century Reanalysis system. Quart. J. Roy. Meteor. Soc., 145, 2876–2908, https://doi.org/10.1002/qj.3598.
Stammerjohn, S., R. Massom, D. Rind, and D. Martinson, 2012: Regions of rapid sea ice change: An inter-hemispheric seasonal comparison. Geophys. Res. Lett., 39, L06501, https://doi.org/10.1029/2012GL050874.
Stroeve, J., and D. Notz, 2018: Changing state of Arctic sea ice across all seasons. Environ. Res. Lett., 13, 103001, https://doi.org/10.1088/1748-9326/aade56.
Stroeve, J., M. C. Serreze, M. M. Holland, J. E. Jay, J. Malanik, and A. P. Barrett, 2012: The Arctic’s rapidly shrinking sea ice cover: a research synthesis. Climatic Change, 110, 1005–1027, https://doi.org/10.1007/s10584-011-0101-1.
Titchner, H. A., and N. A. Rayner, 2014: The Met Office Hadley Centre sea ice and sea surface temperature data set, version 2: 1. Sea ice concentrations. J. Geophys. Res. Atmos., 119, 2864–2889, https://doi.org/10.1002/2013JD020316.
Turner, J., 2004: The El Niño–Southern Oscillation and Antarctica. Int. J. Climatol., 24, 1–31, https://doi.org/10.1002/joc.965.
Turner, J., and J. Overland, 2009: Contrasting climate changes in the two polar regions. Polar Res., 28, 146–164, https://doi.org/10.1111/j.1751-8369.2009.00128.x.
Wang, J., J. Zhang, E. Watanabe, M. Ikeda, K. Mizobata, J. E. Walsh, X. Bai, and B. Wu, 2009: Is the dipole anomaly a major driver to record lows in Arctic summer sea ice extent? Geophys. Res. Lett., 36, L05706, https://doi.org/10.1029/2008GL036706.
Wang, Z., X. Zhang, Z. Guan, B. Sun, X. Yang, and C. Liu, 2015: An atmospheric origin of the multi-decadal bipolar seesaw. Sci. Rep., 5, 8909, https://doi.org/10.1038/srep08909.
Yu, L., and S. Zhong, 2018: Changes in sea-surface temperature and atmospheric circulation patterns associated with reductions in Arctic sea ice cover in recent decades. Atmos. Chem. Phys., 18, 14 149–14 159, https://doi.org/10.5194/acp-18-14149-2018.
Yu, L., S. Zhong, J. A. Winkler, M. Zhou, D. H. Lenschow, B. Li, X. Wang, and Q. Yang, 2017: Possible connections of the opposite trends in Arctic and Antarctic sea-ice cover. Sci. Rep., 7, 45804, https://doi.org/10.1038/srep45804.
Yu, L., S. Zhong, M. Zhou, and B. Sun, 2018: Antarctic summer sea ice trend in the context of high-latitude atmospheric circulation changes. J. Climate, 31, 3909–3920, https://doi.org/10.1175/JCLI-D-17-0739.1.
Yu, L., S. Zhong, M. Zhou, D. H. Lenschow, and B. Sun, 2019: Revisiting the linkages between the variability of atmospheric circulations and Arctic melt-season sea ice cover at multiple time scales. J. Climate, 32, 1461–1482, https://doi.org/10.1175/JCLI-D-18-0301.1.
Yuan, X. J., and C. H. Li, 2008: Climate modes in southern high latitudes and their impacts on Antarctic sea ice. J. Geophys. Res. Oceans, 113, C06S91, https://doi.org/10.1029/2006JC004067.
