The present examination of Antarctic ice drift and its response to wind and ocean currents pertains to interests in the recent trends in sea ice coverage and the formation of dense water in the Southern Ocean; both topics are relevant to understanding variability and changes in the climate system. The overall increase in Antarctic sea ice extent over the satellite record is the sum of opposing trends in different sectors of the Southern Ocean (Comiso and Nishio, 2008; Comiso et al., 2011); currently, there is no consensus on the causal mechanisms. Advances and retreats of the ice edge have been linked to trends in wind-driven ice drift, due to large-scale intensification in surface winds associated with the circumpolar lows around Antarctica (e.g., Holland and Kwok, 2012; Haumann et al., 2014; Zhang, 2014; Kwok et al., 2016). Hence, the variability in wind and drift patterns is of particular geophysical interest. From the perspective of dense water formation, ice growth in leads and polynyas dominates the heat flux into the atmosphere and thus brine flux into the ocean. On a regional scale, the redistribution of freshwater via the transport and subsequent melt of relatively fresh sea ice modifies ocean buoyancy forcing. Of particular climatic interest is the dense/high salinity water that is produced in Antarctic polynyas; these water masses sink along the continental slope after mixing with surrounding waters to form the Antarctic Bottom Water (AABW) of the deep ocean (Foster and Carmack, 1976; Gordon, 1991). The best-documented sources of AABW lie along the margins of the Weddell Sea. Sources of dense bottom water have also been identified at other locations around the perimeter of Antarctica, such as the Adelie Coast, and off Enderby Land. The Ross Sea, in the Pacific sector, is also thought to be a significant source of this cold high salinity water. Hence, understanding the changes in ice drift over a broad-length time scale is of substantial interest.
Since the late 1990s, the availability of the moderate resolution ice drift from satellite passive microwave observations (e.g., Agnew et al., 1997; Emery et al., 1997; Kwok et al., 1998) has allowed large-scale studies of drift patterns over Antarctic sea ice. The great strengths of this dataset are its spatial coverage and the length of the data record, which is more than three decades (at this writing), for the combination of SMMR, SSM/I and AMSR-E instruments. The limitation is that the fairly coarse spatial resolution of the imagery produces uncertainties of several kilometers for individual displacement vectors. These datasets are better suited for understanding synoptic and longer-term drift patterns rather than the detailed characteristics of daily ice motion. Previous investigations using these data sets have focused broadly on: the relationship between wind and ice drift (e.g., Kwok et al., 1998; Kimura, 2004); the export of sea ice from the Weddell and Ross seas (e.g., Martin et al., 2007; Drucker et al., 2011); and the variability and trends in ice drift (e.g., Comiso et al., 2011; Kimura and Wakatsuchi, 2011; Holland and Kwok, 2012; Kwok et al., 2016). With a 34-year record (1982–2015) of Antarctic satellite ice drift, the aim of this paper is to examine the following five topics: 1) the mean and trends of the drift; 2) the variability in the location of the drift patterns; 3) the relationship between wind and drift; 4) the relationship between the trends in ice edge and ice drift; and 5) ice export from the Weddell and Ross seas.
We begin by describing the derivation of the ice motion fields and ancillary data sets used in our analysis, also providing an assessment of the quality of drift estimates in the satellite passive microwave record using ice drift from high-resolution Synthetic Aperture Radar (SAR) imagery (Section 2). We then present the three distinct drift patterns seen in monthly mean fields over the 34-year record and investigate the variability in their spatial location and the correlations in their behavior (Section 3). After examining the relationship between wind and ice drift in the Southern Ocean sea ice cover (Section 4), we present computations of the spatial trends in wind and ice drift, and explore their relationships to trends in ice edge and to large-scale atmospheric modes (Section 5). We extend the time series of ice area export from the Weddell and Ross seas as discussed elsewhere (Section 6), and end with a summary of our findings and conclusions.
In this section, we describe the ice drift and other data sets used in this paper. In particular, an assessment of the gridded satellite drift estimates using ice drift derived from high-resolution SAR data is provided.
Ice drift data used here are retrievals from successive satellite brightness temperature fields (Kwok et al., 1998). We used only the March through November ice drift estimates as the ice tracking results are unreliable during the Antarctic summer and transitional months. This lack of reliability is due to spatial variability in the brightness-temperature fields associated with water vapor, cloud liquid content, and surface wetness during these months. The gridded fields of ice drift (100 km spacing) – on a polar stereographic projection – were constructed by blending ice drift derived from two satellite radiometer channels (37 GHz and 85 GHz – 91 GHz since 2009), viz.
The weighting coefficients α and β were determined by an optimal interpolation scheme (Kwok et al., 2013); the indices i and j are the available observations from individual radiometer channels. Motion and drift are used interchangeably throughout this paper. A spatial correlation length scale of 300 km was used to create the interpolated motion field. This length scale was selected at an intermediate scale based on the density of satellite observations, but short enough that the expressions of coastal effects and drift are not noticeably degraded. A consistent and updated time series of passive microwave brightness temperature and ice concentration fields (Maslanik and Stroeve, 2004; Gloersen, 2006) were used to produce the satellite ice drifts. Uncertainties in drift estimates from the SMMR (1982–1987) and SSM/I (1988–present) datasets are between 3 and 6 km (depending on spatial resolution of the passive microwave channel) for individual displacement vectors. Ice motion fields from multiple channels on the same instrument (e.g., 37 GHz and 85 GHz on SSM/I) were used when they were available. Together, the length of the ice drift record provided by the combination of sensors spans more than three decades. Our records start in 1982 to avoid gaps in the earlier (1978–1981) brightness temperature fields. Based on the number of observations and expected uncertainties in the passive microwave (PMW) ice motion estimates, the procedure above provides an analyzed error of each motion estimate. An expected average uncertainty of 3–4 km day–1 in the individual interpolated estimates is typical (addressed in Section 3), although the uncertainty varies with the density of measurements available within the neighborhood of each estimate. Henceforth, these passive microwave estimates will be referred to as PMW ice drift. The ice motion data set used here can be found at: https://rkwok.jpl.nasa.gov/antarc_icemotion/index.html.
Available near-daily drift estimates from Envisat SAR imagery (resolution: ~150 m) between 2007 and 2010 were used to assess the quality of the passive microwave ice drift. These SAR-derived ice drifts are primarily in the southern Ross and Weddell seas and are sampled on a uniform 10-km grid; uncertainties in the estimates are ~300 m day–1 (Lindsay and Stern, 2003) and thus of significantly better quality than those derived from the passive microwave fields (several kilometers per day). They are considered as truth in the following analysis. The tracking of common ice features in a sequence of SAR images has been described by Kwok et al. (1990).
A comparison of 4 years (2007 through 2010) of PMW and SAR ice drift estimates is provided in Figure 1; each drift vector has been decomposed into its two orthogonal components with the orientation defined by the SSM/I polar stereographic projection (x-axis: 90°E, y-axis: 0°E). To match the space-scale of the PMW drift estimates, the comparisons are to averages of high-resolution Envisat drift vectors within a radius of 50 km of each grid point or each 100-km gridded PMW estimate. As mentioned earlier, available SAR-derived ice drifts are primarily of the southern Ross and Weddell seas (mainly south of 65°S); spacecraft resources and space agency (European Space Agency) programming determine the repeat SAR coverage. Thus, faster ice drifts closer to the ice edge and ice margins are not sampled.
The difference statistics for the Ross and Weddell seas are consistent with each other. Over the 4 years, the correlations between the PMW and SAR ice drifts are between 0.73 and 0.86. The mean difference varies between 1.5 and –0.5 km day–1, with standard deviations that range between 3.3 and 6.8 km day–1 for individual vector components. These differences are comparable to an earlier assessment of PMW ice drift in the Ross Sea using RADARSAT ice drift (Kwok, 2005) of ~4.5 km day–1, and show relative consistency in the quality of derived PMW motion estimates over different periods. The differences are somewhat higher than the analyzed uncertainties (i.e., 3–4 km day–1) provided by the optimal interpolation procedure. This finding is not unexpected, but worthwhile noting here is that the correlations and differences fall between two estimated quantities, and errors/noise in these quantities would lower the true correlation and increase the differences. In addition, the following factors contribute to the variance of the observed differences: (1) the drift estimates are not contemporaneous in that there is an offset between the start time of each displacement vector (sometimes up to 12 hours, as determined by satellite revisit times), and thus the drift estimates are often decorrelated in time; (2) the PMW drift represents an average over a larger space scale (~300 km) compared to the averaged Envisat-derived drifts, which also depend on the density and availability of measurements in the neighborhood of each grid point as determined by the Envisat image overlaps; and (3) the drift estimates are from PMW imagery that are daily composites with ill-defined time stamps. Thus, it is not unexpected that the assessed differences are higher than that expected by the interpolation procedure.
The locations of the time-varying ice edge used here were sampled at longitudinal increments of one degree (i.e., 360 increments are used to define the circumpolar ice edge). Ice edge is defined as the latitudinal location where the ice concentration first exceeds 15% in the transition from open-ocean to the consolidated ice cover. Gridded maps of ice concentration (1982–2015) were derived from the Scanning Multi-channel Microwave Radiometer (SMMR) and Special Sensor Microwave Imager (SSM/I) data using the Bootstrap Algorithm (Comiso and Nishio, 2008).
Sea-level pressures (SLP) are from the ERA-Interim atmospheric reanalysis project (http://apps.ecmwf.int/datasets/data/interim-full-daily/levtype=sfc/) and the Southern Oscillation (SO) and Southern Annular Mode (SAM) indices are from NCDC (www.ncdc.noaa.gov/teleconnections/enso/indicators). The geostrophic wind fields ( : f = Coriolis parameter; ρ = air density; P = sea-level pressure; and is the surface normal) used here were derived from SLP fields.
Figure 2 shows the location of circumpolar seas and the three drift patterns within the Antarctic sea ice zone in the monthly motion fields of the 34-year record. These wind-driven drift patterns can be seen as linked to distinct atmospheric low-pressure centers (at sea level) within the circumpolar trough around Antarctica, as discussed in more detail in Section 4. The first pattern is associated with the Amundsen Sea Low (ASL) centered roughly over the Amundsen Sea; it spans the Ross, Amundsen, and Bellingshausen seas (roughly between 150°E and 60°W). Due to influence of the ASL on the West Antarctic climate and its link to the climate of the tropical Pacific, this climatic pattern has received more attention (Turner et al., 2013; Raphael et al., 2016) than the following two. The second, centered over the Riiser-Larsen Sea (~30°E), spans the Weddell, Lazarev, Riiser-Larsen, and Cosmonaut seas (between 60°W and 60°E). A third, centered over the Davis Sea (~90°E), spans the Cooperation, Davis, Mawson, D’Urville, and Somov seas (between 60°E and 150°E). For convenience here, we refer to the second and third centers of low pressure as the Riiser-Larsen Sea Low (RLSL) and the Davis Sea Low (DSL), respectively, based on their location. The first two drift patterns are dominant in all winter months, while the third drift pattern is only visible (in Figure 2) between August and November due to the absence of ice drift (in our low-resolution fields) to trace the wind-driven circulation during the other months. Below, we summarize the features of these three drift patterns separately. As ice drift is largely wind-driven, we examined the variability of the zonal and meridional location of the three low-pressure centers and the covariance of the sea-level pressure between these centers of action. The location of the centers impacts ice export (discussed in Section 6), the occurrence of coastal polynyas, and the location of the ice edge as discussed by Kwok et al. (2016).
The cyclonic (clockwise) drift pattern, associated with the ASL centered over the northeast Ross Sea, is evident in all months between March and November (Figure 2). This oceanic circulation pattern is known as the Ross Sea Gyre. Along the Antarctic coast, the average ice drift is westward towards the Ross Sea. The drift pattern shows a coastal inflow of sea ice into the Ross Sea from the Amundsen Sea in the east, then southward along the coast of the Ross Sea embayment, with a considerably stronger northward outflow in the west. This imbalance in the overall circulation points to significant divergence and production of open water and areas of rapid ice growth in the Ross Sea polynyas off the Ross Ice Shelf (RIS) that are frequent occurrences during all winter months (Martin et al., 2007; Comiso et al., 2011). Some recirculation of the outflow in the eastern Ross Sea is also evident. North of Cape Adare, the northward ice drift splits into two branches with one that moves westward into the Somov Sea and the other northeastward. Farther north, the prevailing motion is eastward as the sea ice becomes entrained in the fast moving Antarctic Circumpolar Current during the late winter.
Atmospheric forcing plays a significant role in the enhancement of the sea ice outflow in the western Ross Sea (Figure 2). Importantly, this outflow influences the ice production in polynyas at the ice front of the RIS (the Ross Sea Polynya, sometimes referred to as the Ross Ice Shelf Polynya) and Terra Nova Bay (the Terra Nova Bay Polynya). Records from automatic weather stations (AWS) deployed over the RIS reveal that the dominant near-surface airflow over the western RIS is northward, passes to the east of Ross Island (Savage and Stearns, 1985) and appears to be the primary atmospheric forcing for development of the Ross Sea Polynya (Bromwich et al., 1993). Bromwich et al. (1998) found that synoptic cyclones near Roosevelt Island induce SLP distribution over the RIS with isobars oriented parallel to the Transantarctic Mountains. This setup results in the intensification and northward propagation of the katabatic winds across the ice shelf with associated low-level warming. An immediate impact of katabatic surges is the development of polynyas where heat and salt fluxes associated with new ice growth are intense: about 60% of the polynya events are linked to katabatic surge events; and 40% from katabatic drainage winds (from glaciers) and barrier winds (winds that flow northward along the Transantaractic Mountains and are deflected eastward by topographic barriers along the Scott Coast) (Bromwich et al., 1998). The interannual variability in ice export from the Ross Sea is discussed in Section 5.
This cyclonic drift pattern (also an expression of the oceanic Weddell Gyre) dominates the ice drift over the months of March through November (Figure 2). A feature of this pattern is the band of strong westward coastal drift just south of the center of the RLSL (seen to span the Lazarev, Riiser-Larsen, and Cosmonaut seas). To the west, the ice drift between May and September is westward with a slight northward component. In October and November, the turn southward into the Weddell Sea (near Cap Norwegia) is especially apparent; this branch flows towards the Filchner and Ronne Ice shelves before turning westward and northward away from the ice shelves and out of the Weddell Sea. The large northward drift in the west exports sea ice produced in the Southern Weddell Sea and the coastal polynyas along the Ronne and Brunt ice shelves (Drucker et al., 2011). North of the Antarctic Peninsula, similar to that in the Ross Sea, prevailing motion is faster and towards the east as the ice merges with the southern-most reaches of the Antarctic Circumpolar Current, and is especially dominant during the late winter.
The cyclonic DSL drift pattern (Figure 2), as mentioned earlier, is not as prominent as that of the ASL and RLSL patterns. Even though the low-pressure system is evident in the SLP distribution, the narrow band of sea ice in this region, which extends only ~400 km from the coast at maximum extent, does not allow the pattern to be traced out by the low-resolution drift fields (used here) until later in the winter (i.e., after August). A band of strong westward coastal drift just south of the center of the DSL low is also seen to span the Cooperation, Davis, Mawson, D’Urville, and Somov seas.
As ice drift is largely wind-driven (see Section 4), our interest in the variability in the meridional/zonal location and strength of these low-pressure centers is related to their impact on ice export, polynya activity, and on the extent of the ice edge as discussed by Kwok et al. (2016). The monthly mean locations (March through November) of the three atmospheric lows (ASL, RLSL, and DSL) over the 34-year record are shown in Figure 2, while the variability in their seasonal locations over the record is shown in Figure 3a. Locations with the lowest monthly sea-level pressure anomalies within each of the three sectors (described above) are designated as the centers of the seasonal lows in the monthly fields; there are 192 such locations (in the 34-year record) within each sector.
Three distinct clusters, well separated from each other, can be seen in the spatial distribution of the centers of the seasonal lows (Figure 3a). In all three sectors, the zonal spread in the lows is higher than that of the meridional spread. When the geographic variability of the seasonal centers are separated into the four seasons (as depicted by four ellipses), the following features are apparent: 1) in all three sectors, the spatial spread of the centers is much smaller during the winter (light blue ellipses in Figure 3a) compared to the other seasons, suggesting seasonality in the expected location of the lows and hence variability in the circulation pattern; and 2) the zonal locations of the ASL (red ellipses) are farther to the east during the summer than those seen in the other sectors. In fact, Fogt et al. (2012) and Turner et al. (2013) noted a well-defined annual cycle in the average zonal location of the ASL, with the low being found immediately west of the Antarctic Peninsula in austral summer (December–February) and shifting westward to the Ross Sea by winter (June–August). A similar annual cycle in zonal location of the DSL (i.e., farther east during the summer) is also noticeable, but not for the RLSL.
Correlations (ρ) between the time series of monthly SLP anomalies at ASL, RLSL, and DSL are of interest as they reveal the linked variability of these circumpolar lows, and hence the linked behavior in the sea ice circulation patterns and their associated impact on ice export, polynya activity, and the ice extent (discussed in later sections). The seasonal correlations are seen in Figure 3b. All three time-series are positively correlated. These circumpolar teleconnections can be seen in the vertical banding of the SLP anomalies (red: positive, blue: negative) in the longitude-time diagrams (Figures 3c–e). In three latitude ranges (60°S–65°S, 65°S–70°S, 70°S–78°S), these vertical bands of the nearly same color indicate that anomalies in the circumpolar SLP co-vary and, on average, in phase with each other. In summary, because the wind explains a large fraction of the variance in ice drift (discussed next), the behavior of the circumpolar ice cover is connected.
In this section, we examine the relationship between wind and ice drift in the 34-year record. We quantify the response of daily ice drift to geostrophic wind and compare our results with earlier work on this topic.
To examine the response of daily ice drift to geostrophic wind over the Southern Ocean sea ice cover, we followed Thorndike and Colony (1982) (henceforth TC82) and used a linear model to relate the observed ice motion (u) to geostrophic wind (G) and mean ocean current ( ), viz.
A is a complex constant and ε represents that part of the ice velocity that is neither a constant nor a linear function of the geostrophic wind. To compute the complex coefficient (A) and intercept (), the components of the motion (u) and geostrophic wind (G) vectors are thought of as the real and imaginary parts of a complex number. The following parameters: |A|, θ, and ρ2 are obtained by regression analysis of the time series of daily wind and ice drift at each grid point. With the 34-year record, we constructed gridded spatial fields of these parameters. The magnitude of the scaling factor (i.e., |A|) tells us about the coupling between wind and ice and the variability in the internal ice stresses that tend to oppose ice motion. θ is the turning angle from the geostrophic wind direction to the direction of ice motion (positive is to the left of the wind), and the squared correlation coefficient (ρ2) is that fraction of the ice motion explained by the geostrophic wind.
The fields of |A| and ρ2 are shown in Figure 4. We first discuss their spatial variability over the record. Geographically, |A| is higher (>0.014) towards the ice edge away from the Antarctic coasts, while the lowest values for |A| can be found in the southern Weddell, western Weddell next to the Antarctic Peninsula, and western Ross seas. Between March and November, the average complex scaling factor was 0.014 ± 0.002exp(i5 ± 2.3°). This result can be compared to a scaling factor that ranged between 0.005 and 0.015 in the Weddell Sea (Kottmeier and Sellmann, 1996), and was 0.017exp(i4°) in the Ross Sea (Kwok, 2005). It can also be contrasted with the scaling factor for the winter Arctic: 0.0077exp(–i5°) from buoy drift (Thorndike and Colony, 1982) and 0.009 ± 0.0015exp(–i1.9 ± 2.6°) from satellite ice drift (Kwok et al., 2013). Due to Coriolis, the ice drift direction in the Southern Hemisphere is turned to the left of the wind instead of the right. Of note is that the scaling factor is nearly 50% higher than that of the Arctic – even with the observed thinning. In fact, this scaling factor is closer to that obtained during the Arctic summer (June through September) of 0.011 exp(–i18) (Thorndike and Colony, 1982) and 0.01 ± 0.001 exp(–i7.1 ± 3.6°) from recent satellite ice drift (Kwok et al., 2013).
The contrast between the Arctic and Antarctic suggests that the predominantly seasonal Antarctic sea ice cover is closer to free drift conditions than that observed in the Arctic, where the thicker ice is geographically confined within a basin. Generally, to achieve the same ice velocity, higher wind stresses are required to oppose ice stress where ice is thick or more compact (i.e., higher ice strength). The generally higher |A| over the Antarctic ice cover also suggests a much thinner and less compact ice cover compared to that of the Arctic.
That fraction of the variance of ice motion, which is explained by geostrophic wind, is given by the squared correlation coefficient, ρ2. At distances of 200–400 km from the Antarctic coast, ρ2 is about 0.65 ± 0.05 and the values of ρ2 are generally lower within 400 km of the coast, although this zone varies spatially. Coastal geometry, or mechanical constraints on ice drift, tends to reduce the correlation between wind and motion. Based on the regression analysis, the linear model – on average – explains all but about 40% of the variance of the ice motion.
Recent studies have reported large and statistically significant trends in Antarctic ice drift in most sectors that are associated with the intensification of surface winds (Holland and Kwok, 2012; Zhang, 2014), suggesting that regional wind-driven changes may be one of the drivers of ice extent around much of Antarctica. Here, we examine in more detail these trends in our longer 34-year drift record and relate these to trends in winds and the latitudinal extent of the ice edge. Ice drift and wind vectors were decomposed into their respective zonal and meridional components. Figure 5 shows trend maps of the following five parameters in six 2-month periods: meridional wind and drift (Figure 5a and b), ice edge (Figure 5c), and zonal wind and drift (Figure 5d and e). In the following, we first look at the trends in wind and ice drift, and then their correlations to trends in ice edge.
The mean circumpolar trends in ice drift for April to October for a 19-year period between 1992 and 2010 were discussed by Holland and Kwok (2012). The trends in Figure 5 extend the span of the record to 34 years, which cover nearly the entire satellite era. Further, the partitioning of the record into 2-month maps of the zonal and meridional components allows for a more detailed view of the regional and intra-seasonal variability of the drift trends, and their relationship to trends in the wind field.
Visual inspection reveals that there are substantial large-scale correlations of the spatial pattern of the polarity of trends in wind and ice drift in all 2-month sections (Figure 5a, b, d and e). But, discrepancies between the trend patterns can be seen. Even though the wind explains ~60% of the variance in ice drift (Section 3), there are limitations in the accuracy of current reanalysis fields, especially in representing their trends. Two areas of differences are of note. One standout is that the positive drift trends in the Ross Sea outflow, especially striking in the meridional drift fields (Figure 2) between July and October, is noticeably absent in the corresponding wind field. In coastal Antarctica, the absence of a corresponding trend in the wind field may be due to the inability of the ERA-interim reanalysis to capture the local intensification and northward propagation of the katabatic and drainage winds – boundary layer effects – linked to the development of polynyas in the Ross Sea (Sanz Rodrigo et al., 2012). This absence can be contrasted to the better correspondence of trends in the large-scale (synoptic) wind field with ice drift away from the continent. Also notable is the lack of correspondence between the August–September meridional wind and drift trends. We do not have a suitably satisfying explanation for this discrepancy.
Opposing trends in Antarctic sea ice extent can be observed in different sectors of the Southern Ocean (Comiso and Nishio, 2008; Comiso et al., 2011). The predominantly two-wavenumber pattern in the ice-edge trends (see Figure 5) suggests that they are associated with trends in the large-scale (synoptic) wind field connected to the atmospheric lows discussed in Section 3. A recent study by Kwok et al. (2016) showed that on average two-thirds of the winter ice edge trend in the Pacific sector can be explained by ice drift and meridional winds, linked to extratropical atmospheric anomalies associated with the Southern Oscillation. Here we first examine the circumpolar trends in ice edge and their relationship to the trends in the meridional and zonal wind and ice drift.
As seen in Figure 5, it is apparent that the expansion of the ice cover or advances in the latitudinal extent of the ice edge can be linked, to a certain degree, to positive trends in the local meridional winds. For trends in zonal winds or ice drift, Figure 6 illustrates how zonal ice drift trends could affect local trends in ice edge. Note that the left-turning tendencies of ice drift due to Coriolis (relative to the geostrophic wind direction) is already part of the observed drift; hence, the discussion here is not associated with this component of ice drift. Rather, due to the zonal asymmetry of the Antarctic ice cover, it can be seen that a coherent zonal drift trend or rotation of the mean ice edge (from red to blue in Figure 6) would create local ice edge trends: off-ice zonal winds would advect into areas that are typically open ocean or away from areas of ice coverage and vice versa for on-ice winds. The decomposition into zonal and meridional directions seems unnecessary, but this step is useful for examining circumpolar westerlies (for example) in the context of large-scale changes in the wind field (e.g., due to the Southern Annular Mode). Both the strengthening and weakening of the westerlies impact the location of the ice edge.
While there is correspondence between the trends in the ice edge and trends in drift/wind, direct correlations are not seen at all times or everywhere (Figure 5). In May and June, the positive ice edge trends in Sector-1 (S1) and the western part of S4 are clearly associated with the trends of similar polarity in the zonal rather than meridional wind and ice drift. However, opposite meridional and zonal wind trends in S5 have little effect on the ice edge trend. For June–July, there is general correspondence of the polarities of the ice edge trend with the ice drift/wind trends. Thus, the relationship between ice edge and drift trends is not seen everywhere, although some of the issues may be due to the limitations in the reanalysis products in the Southern Ocean seen in the discrepancies in wind and drift discussed earlier.
The plots in Figure 7 provide a seasonal (March–April–May: MAM, June–July–August: JJA, and September–October–November: SON) perspective of the circumpolar trends in ice edge and in meridional/zonal winds at different latitude bands. Again, the two-wavenumber pattern in the ice edge trends is evident in all three seasons, although the location of the peaks varies zonally. Except for the zonal wind trends during the fall (MAM), the meridional/zonal wind trends have a two- to three-wavenumber pattern. The highest correlations are observed between the circumpolar trends in ice edge and meridional/zonal wind during winter (JJA), when the mean ice edge is farthest north and extends into the Antarctic Circumpolar Current. The correlations between circumpolar ice edge and zonal wind trends are higher (0.74 between 55°S and 60°S) in winter than in other seasons, and in fact higher than the correlation with meridional winds (0.34 between 55°S and 60°S). During the spring (SON), the ice-edge/meridional-wind trend correlations are weaker (0.46 between 65°S and 70°S) and ice-edge/zonal-wind trend correlations are negative (–0.42 between 65°S and 70°S). In the fall (MAM), the location of the peaks just east of Victoria Land and the Antarctic Peninsula suggest that the trends may be controlled by the coastal boundaries (see Figure 7a and b) when the developing ice cover is generally within the embayments of the Ross and Weddell seas. The results suggest that ice edge trends are linked to zonal wind trends, and that – within the context of ice drift and wind – coastal constraints may have a role in the observed trends in ice extent.
Previous work has shown that anomalies in circumpolar ice edge location and surface climate in the Southern Ocean are linked to these two dominant atmospheric modes: the Southern Oscillation (SO) and Southern Annular Mode (SAM) (e.g., Kwok and Comiso, 2002a, 2002b; Stammerjohn et al., 2008; Yuan and Li, 2008; Simpkins et al., 2012; Kwok et al., 2016). Here, we summarize broadly the large-scale anomalies in our 34-year record of ice drift, together with those of wind and ice edge, associated with extremes in the SOI and SAMI (I denotes index). The composite maps in Figure 8 show these anomalies in their positive (SOI+: SOI > 1.0; SAMI+: SAMI > 1.0) and negative (SOI–: SOI < –1.0; SAMI–: SAMI < –1.0) phases for two periods: May–August and August–November. As in Figure 5, ice drift and winds have been decomposed into their meridional and zonal components.
The SOI+ and SOI– composite maps (Figure 8a and b) show that the meridional drift/wind/ice edge anomalies are organized in distinct spatial patterns with broadly opposing polarities but with different magnitudes. These characteristics are less pronounced in the zonal drift/wind composites, and of note is that the zonal wind anomalies do not exhibit opposing patterns seaward of the ice-covered ocean. The responses to the SO are different and stronger in August–November. For both periods, the anomalies in meridional ice drift during SOI– are stronger, suggesting potential asymmetry in the response to the two extremes of the SO. In the Pacific Sector (between ~180°E and 60°W), the positive ice edge anomalies in the eastern Ross/Amundsen seas and the negative ice edge anomalies in the Bellingshausen Sea in the SOI+ composite resemble the pattern in the observed time trends (Figure 5), and coincide with the positive/negative anomalies in meridional ice motion and winds. A recent study (Kwok et al., 2016) reported that the trend in SOI over a 32-year period is able to quantitatively explain the July–November trends in sea ice edge, drift, and surface winds in this sector.
The anomalies in the SAM+ and SAM– composite maps (Figure 8c and d) also show distinct spatial patterns with broadly opposing polarities, but with contrasting magnitudes that are more pronounced in the meridional composites. In both periods, the strengthening and weakening of the westerlies in the zonal drift/wind composites – a characteristic of the SAM – are unmistakable and dominate the entire spatial pattern. Also, there is less asymmetry and a more balanced (i.e., the two periods look similar) zonal response to the SAM. The meridional responses to the SAM are different in May–August and August–November, with the strongest response in August–November.
In the composites of ice edge anomalies, whether in extremes of the SO or SAM, it is interesting to note that the location of the 2-wavenumber ice edge anomalies remain relatively stationary, even though they are of different magnitudes. It should also be pointed out that these composites are not pure responses to given atmospheric modes, as they are embedded responses in a coupled climate system.
The interest in ice export from the Ross and Weddell seas pertains to ice production in their coastal polynyas. Coastal polynyas are large, persistent regions of open water and thin ice that form adjacent to a lee shore within the pack ice. When winter winds blow the pack ice away from the coast, seawater at near freezing temperatures is exposed to a large negative heat flux, resulting in rapid formation of new ice and brine rejection. As Jacobs (2004) describes, the polynyas and ice export at these two Antarctic seas contribute to the formation of dense shelf water and subsequently Antarctic bottom water. In both of these seas, polynya activity and ice drift are driven by the depth and zonal location of low-pressure systems to the east (e.g., ASL and the Ross Sea polynyas).
Because of the importance of these seas to shelf- and deep-water production and the similarity of their atmospheric forcing, we computed the annual, inter-annual, and decadal variability in ice area export over the 34-year record. To extend earlier records (Kwok, 2005; Drucker et al., 2011), flux gates were placed parallel to the 1000-m isobaths for calculating ice export (Figure 9). The daily area flux, F, was estimated by integrating the cross-gate motion profile using the simple trapezoidal rule, , where u is the magnitude of the motion perpendicular to the flux gate, Δx is the spacing between the motion estimates, Ci is the ice concentration (between ui and ui+1) from PMW analyses, and n is the number of motion samples along the gate.
At these flux gates, we calculated the winter inflow, outflow and net outflow of ice area at the flux gate as the total of the daily area flux from the beginning of March until the end of November. The net flux is the difference between outflow and inflow. A positive net outflow would be a measure of the ice area produced in the Ross Sea if there were no melt, deformation or sea ice advected into the area, and the net production would be zero if export equals import. It is therefore of interest to examine the two contributions to the net outflow because it provides a rough estimate of the ice that is advected in, from the east, and the total that is exported to the west; that is, the ice area produced in the Ross or Weddell seas and the re-export of the eastern inflow.
The eastern and western termini of the Ross Sea flux gate are located at Land Bay and Cape Adare, respectively. The gate spans a length of ~1400 km and encloses an area of ~490 × 103 km2 to the south, where ice production occurs in the Ross Shelf Polynya and the Terra Nova Bay and McMurdo Sound polynyas. Previous studies have examined the variability of Ross ice area exports between 1992 and 2008 (Comiso et al., 2011; Drucker et al., 2011). Here, the area flux record (Figure 9) has been extended to cover a 34-year PMW record starting in 1982.
The average net (out minus in) outflow for the 34-year record is 0.75 × 106 km2 and ranges from a low of 0.35 × 106 km2 in 1986 to a peak of 1.36 × 106 km2 in 2001, a more than fourfold variability over the timespan. At the peak in 2001, the conditions that are favorable for large net outflow are SLP isobars that are nearly perpendicular to the edge of the RIS associated with the the location of the center of the ASL low. In this case, the highest outflow (1.56 × 106 km2) is accompanied by the lowest inflow of 0.22 × 106 km2 on record. The Ross Sea, with an area of ~490 × 103 km2 south of the flux gate, exports more than twice its area of sea ice every 9 months. The standard deviation of the net flux, at 0.27 × 106 km2, is high. A more moderate positive trend of ~77 × 103 km2 decade–1 (statistically insignificant) can be seen in the 34-year record, and can be compared to the higher ~300 × 103 km2 decade–1 trend reported in the shorter 17-year record (1982–2008) (Comiso et al., 2011). The net positive trend during the 34 years appears to be the result of a positive trend in the outflow and a smaller positive trend in the inflow.
The gate that connects the two termini of the Weddell Sea flux gate has length of ~1100 km and encloses an area of ~283 × 103 km2. Polynya ice production occurs primarily in two regions within this area: around the Brunt Ice Shelf, which we call the Eastern Weddell Polynya (EWP), at the Ronne Ice Shelf. For ice area export, Drucker et al. (2011) estimated an average area flux of 0.52 × 106 km2 (April and November) for 2003–2008.
The average net (out minus in) outflow for the 34-year record is 0.32 × 106 km2 and ranges from a minimum of 0.02 × 106 km2 in 1990 to a peak of 0.70 × 106 km2 in 1986, a more than tenfold variability over the period. Similar to the Ross Sea, the high flux can be attributed to the depth and the zonal location of the RLSL. The Weddell Sea, with an area of ~283 × 103 km2 within the flux gate, exports approximately its own area of sea ice every 9 months. The standard deviation of the net flux, at 0.16 × 106 km2, is high. A negative trend of ~45 × 103 km2 decade–1, which is statistically insignificant, can be seen in the 34-year record.
In this paper, we have examined patterns, trends and variability in sea ice drift and circulation in the Southern Ocean in a 34-year data set of satellite-derived ice drift. Antarctic sea ice drift and circulation are of relevance to understanding changes in the climate system, as observed trends in ice drift, export and circulation patterns are related to large-scale changes in atmospheric circulation, dense water formation, and also connected to trends in sea ice coverage. Here, we highlight our results and conclusions:
The ice motion data set used here can be found at: https://rkwok.jpl.nasa.gov/antarc_icemotion/index.html.
The ice motion data set used here can be found at the following URL: https://rkwok.jpl.nasa.gov/antarc_icemotion/index.html. RK, SSP, and SK carried out this work at the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration.
RK, SSP, and SK carried out this work at the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration.
The authors have no competing interests to declare.
Agnew T, Le H and Hirose T 1997. Estimation of large-scale sea-ice motion from SSM/I 85.5 GHz imagery. Ann Glaciol 25: 305–311, DOI: http://dx.doi.org/10.1017/S0260305500014191
Bromwich DH, Carrasco JF, Liu Z and Tzeng RY 1993. Hemispheric atmospheric variations and oceanographic impacts associated with katabatic surges across the Ross Ice Shelf, Antarctica. J Geophys Res-Atmos 98(D7): 13045–13062, DOI: http://dx.doi.org/10.1029/93JD00562
Bromwich DH, Liu Z, Rogers AN and Van Woert M 1998. Winter atmospheric forcing of the Ross Sea Polynya In: Jacobs, S and Weiss, RF eds. Ocean, Ice, and Atmosphere: Interactions at the Antarctic Continental Margins. AGU, pp. 101–133.
Comiso JC, Kwok R, Martin S and Gordon AL 2011. Variability and trends in sea ice extent and ice production in the Ross Sea. J Geophys Res 116(C04): 021.DOI: http://dx.doi.org/10.1029/2010JC006391
Comiso JC and Nishio F 2008. Trends in the sea ice cover using enhanced and compatible AMSR-E, SSM/I, and SMMR data. J Geophys Res 113(C2)DOI: http://dx.doi.org/10.1029/2007JC004257
Drucker R, Martin S and Kwok R 2011. Sea ice production and export from coastal polynyas in the Weddell and Ross Seas. Geophys Res Lett 38(17)DOI: http://dx.doi.org/10.1029/2011GL048668
Emery WJ, Fowler CW and Maslanik JA 1997. Satellite-derived maps of Arctic and Antarctic sea ice motion: 1988 to 1994. Geophys Res Lett 24(8): 897–900, DOI: http://dx.doi.org/10.1029/97GL00755
Fogt RL, Wovrosh AJ, Langen RA and Simmonds I 2012. The characteristic variability and connection to the underlying synoptic activity of the Amundsen-Bellingshausen Seas Low. J Geophys Res 117(D7)DOI: http://dx.doi.org/10.1029/2011JD017337
Foster TD and Carmack EC 1976. Temperature and salinity structure in Weddell Sea. J Phys Oceanogr 6(1): 36–44, DOI: http://dx.doi.org/10.1175/1520-0485(1976)006<0036:TASSIT>2.0.CO;2
Gloersen P 2006. Nimbus-7 SMMR polar gridded radiances and sea ice concentrations, Version 1, Boulder, Colorado USA. NASA National Snow and Ice Data Center Distributed Active Archive Center, DOI: http://dx.doi.org/10.5067/QOZIVYV3V9JP
Haumann FA, Notz D and Schmidt H 2014. Anthropogenic influence on recent circulation-driven Antarctic sea ice changes. Geophys Res Lett 41(23): 8429–8437, DOI: http://dx.doi.org/10.1002/2014GL061659
Holland PR and Kwok R 2012. Wind-driven trends in Antarctic sea-ice drift. Nat Geosci 5(12): 872–875, DOI: http://dx.doi.org/10.1038/ngeo1627
Jacobs SS 2004. Bottom water production and its links with the thermohaline circulation. Antarct Sci 16(4): 427–437, DOI: http://dx.doi.org/10.1017/S095410200400224X
Kimura N 2004. Sea ice motion in response to surface wind and ocean current in the Southern Ocean. J Meteorol Soc Jpn 82(4): 1223–1231, DOI: http://dx.doi.org/10.2151/jmsj.2004.1223
Kimura N and Wakatsuchi M 2011. Large-scale processes governing the seasonal variability of the Antarctic sea ice. Tellus A 63(4): 828–840, DOI: http://dx.doi.org/10.1111/j.1600-0870.2011.00526.x
Kottmeier C and Sellmann L 1996. Atmospheric and oceanic forcing of Weddell Sea ice motion. J Geophys Res 101(C9): 20809–20824, DOI: http://dx.doi.org/10.1029/96JC01293
Kwok R 2005. Ross Sea ice motion, area flux, and deformation. J Climate 18: 18.DOI: http://dx.doi.org/10.1175/JCLI3507.1
Kwok R and Comiso JC 2002a. Southern Ocean climate and sea ice anomalies associated with the Southern Oscillation. J Clim 15(5): 487–501, DOI: http://dx.doi.org/10.1175/1520-0442(2002)015<0487:SOCASI>2.0.CO;2
Kwok R and Comiso JC 2002b. Spatial patterns of variability in antarctic surface temperature: Connections to the Southern Hemisphere Annular Mode and the Southern Oscillation. Geophys Res Lett 29(14): 1705.DOI: http://dx.doi.org/10.1029/2002GL015415
Kwok R, Comiso JC, Lee T and Holland PR 2016. Linked trends in the South Pacific sea ice edge and Southern Oscillation Index. Geophys Res Lett 43(19): 10,295–10,302, DOI: http://dx.doi.org/10.1002/2016GL070655
Kwok R, Curlander JC, Mcconnell R and Pang SS 1990. An ice-motion tracking system at the Alaska Sar Facility. IEEE J Ocean Eng 15(1): 44–54, DOI: http://dx.doi.org/10.1109/48.46835
Kwok R, Schweiger A, Rothrock DA, Pang S and Kottmeier C 1998. Sea ice motion from satellite passive microwave imagery assessed with ERS SAR and buoy motions. J Geophys Res 103(C4): 8191–8214, DOI: http://dx.doi.org/10.1029/97JC03334
Kwok R, Spreen G and Pang S 2013. Arctic sea ice circulation and drift speed: Decadal trends and ocean currents. J Geophys Res 118(5): 2408–2425, DOI: http://dx.doi.org/10.1002/jgrc.20191
Lindsay RW and Stern HL 2003. The RADARSAT geophysical processor system: Quality of sea ice trajectory and deformation estimates. J Atmos Ocean Technol 20(9): 1333–1347, DOI: http://dx.doi.org/10.1175/1520-0426(2003)020<1333:TRGPSQ>2.0.CO;2
Martin S, Drucker RS and Kwok R 2007. The areas and ice production of the western and central Ross Sea polynyas, 1992–2002, and their relation to the B-15 and C-19 iceberg events of 2000 and 2002. J Mar Syst 68(1–2): 201–214, DOI: http://dx.doi.org/10.1016/j.jmarsys.2006.11.008
Raphael MN Marshall GJ Turner J Fogt RL Schneider D et al. 2016. The Amundsen Sea Low variability, change, and impact on Antarctic climate. Bull Am Meteorol Soc 97(1): 111–121, DOI: http://dx.doi.org/10.1175/BAMS-D-14-00018.1
Sanz Rodrigo J, Buchlin J-M, van Beeck J, Lenaerts JTM and van den Broeke MR 2012. Evaluation of the antarctic surface wind climate from ERA reanalyses and RACMO2/ANT simulations based on automatic weather stations. Clim Dyn 40(1–2): 353–376, DOI: http://dx.doi.org/10.1007/s00382-012-1396-y
Simpkins GR, Ciasto LM, Thompson DWJ and England MH 2012. Seasonal relationships between large-scale climate variability and Antarctic sea ice concentration. J Clim 25(16): 5451–5469, DOI: http://dx.doi.org/10.1175/JCLI-D-11-00367.1
Stammerjohn SE, Martinson DG, Smith RC, Yuan X and Rind D 2008. Trends in Antarctic annual sea ice retreat and advance and their relation to El Niño–Southern Oscillation and Southern Annular Mode variability. J Geophys Res 113(C3)DOI: http://dx.doi.org/10.1029/2007JC004269
Thorndike AS and Colony R 1982. Sea ice motion in response to geostrophic winds. J Geophys Res 87(NC8): 5845–5852, DOI: http://dx.doi.org/10.1029/JC087iC08p05845
Turner J, Phillips T, Hosking JS, Marshall GJ and Orr A 2013. The Amundsen Sea low. Int J Climatol 33(7): 1818–1829, DOI: http://dx.doi.org/10.1002/joc.3558
Yuan X and Li C 2008. Climate modes in southern high latitudes and their impacts on Antarctic sea ice. J Geophys Res 113(C6)DOI: http://dx.doi.org/10.1029/2006JC004067
Zhang J 2014. Modeling the impact of wind intensification on Antarctic sea ice volume. J Clim 27(1): 202–214, DOI: http://dx.doi.org/10.1175/JCLI-D-12-00139.1