Over 1979–2013 the annual mean SH SIE has increased at a rate of 195 × 103 km2 dec−1 (1.6 % dec−1), which is significant at p < 0.01 (Table 1). For the year as a whole SIE has increased by a small amount around the coast of East Antarctica and over the eastern Weddell Sea (Fig. 1). However, the largest increase has been in the Ross Sea sector (Comiso et al. 2011; Fan et al. 2014; Zwally et al. 2002) (160°E–130°W), where the annual mean SIE has increased at a rate of 119 × 103 km2 dec−1 (p < 0.05) (4.0 % dec−1). SIE has increased in the Ross Sea sector throughout the year, with the smallest absolute trend (although the largest percentage trend of 5.8 % dec−1) in summer (98 × 103 km2 dec−1), not significant as a result of the large inter-annual variability. The largest increase has been in spring (145 × 103 km2 dec−1) (3.6 % dec−1), with a trend significant at p < 0.05. In this sector of the Southern Ocean the SIC anomalies show a high degree of spatial consistency between the seasons, with the positive anomalies being carried northwards at the ice edge through the growth phase (Fig. 2). At the time of sea ice minimum in late summer/early autumn the positive SIC anomalies in the Ross Sea sector are just to the west of 180° off the coast of East Antarctica. However, through the winter and spring the positive SIC anomalies spread eastwards across 180° as the sea ice edge advances sufficiently northwards to enter the climatological westerly wind belt and starts to be advected eastwards. By the time of the SIE maximum in late winter/early spring the positive SIC anomalies extend eastwards from 110°E to 110°W, although the largest anomalies remain over the northern Ross Sea.
The annual mean SIE in the Ross Sea sector is significantly anti-correlated (p < 0.05) with the annual mean MSLP between the Antarctic Peninsula and the Ross Sea (Fig. 3a), which is the sector dominated by the ASL. A deeper (weaker) ASL is therefore associated with more (less) sea ice in the Ross Sea sector. In addition, the SIE is also positively correlated with MSLP to the north and northeast of the Ross Sea—a MSLP pattern that is associated with stronger winds over the southern South Pacific. The correlation of the annual mean SIE in the Ross Sea sector with the 200 hPa zonal wind (Fig. 3b) shows a p < 0.05 significant correlation (anticorrelation) with the upper tropospheric winds over the South Pacific in the latitude bands 45°–60°S (30°–45°S). A major feature of the upper tropospheric circulation of the southern Pacific Ocean is the split nature of the jet over the western Pacific Ocean (Bals-Elsholz et al. 2001). Here the circumpolar jet splits into the sub-tropical jet (STJ) and polar-front jet (PFJ), located to the north and south of New Zealand respectively. Figure 3b shows that more (less) SIE in the Ross Sea sector is associated with a stronger PFJ (STJ). Such a ‘flip–flop’ in the strength of the two jets is seen between the two phases of ENSO (Chen et al. 1996), indicating the role played by tropical Pacific climate variability in modulating the atmospheric circulation in this sector of the Antarctic. A stronger jet south of New Zealand results in a deeper ASL since the Amundsen Sea is located in the right exit area of the jet, which favours upper level divergence and greater cyclogenesis in this region.
As the sea ice off West Antarctica is strongly influenced by the near-surface wind field (Holland and Kwok 2012), it is not surprising that the annual mean SIE in the Ross Sea sector is significantly correlated with the meridional component of the 10 m wind (V10) across the Ross Sea (Fig. 3c) and especially the winds between 130°–180°W. However, the strong association between the SIE in the Ross Sea and the cyclonic circulation of the ASL means that the Ross Sea SIE is also anti-correlated with V10 over the Bellingshausen Sea; e.g. greater Ross SIE is associated with enhanced northerly flow west of the Antarctic Peninsula. Figure 3c also confirms that the annual mean SIE in the Ross Sea is influenced by the broadscale atmospheric circulation at mid- and high-latitude areas of the SH and beyond, but especially eastwards from Australia to the South Atlantic. Such a pattern of correlation indicates the linkages between Ross Sea SIE and the planetary waves around the Antarctic continent.
The ASL exhibits a marked annual cycle in its zonal location, being just to the west of the Antarctic Peninsula in summer and moving westward to the Ross Sea by winter (Hosking et al. 2013; Turner et al. 2012a). In contrast, its depth has a semi-annual form as a result of the Semi-Annual Oscillation, which gives the lowest MSLP values in spring and autumn (Turner et al. 2012a). The pattern of correlation between the Ross Sea SIE and MSLP varies over the year (Fig. 4). During the summer the Ross SIE is anti-correlated (correlated) with MSLP over the Antarctic continent (across 50–60°S), indicating the importance of the Southern Annular Mode (SAM) at this time of year in influencing SIE (Simpkins et al. 2012). Although the pattern of anti-correlation between Ross SIE and MSLP is very zonally symmetric around the continent the largest anticorrelation in summer is in the Amundsen Sea. In the other three seasons there is a clear and significant (p < 0.05) maximum in anticorrelation between Ross SIE and MSLP over the Amundsen Sea, with the largest area of significant correlation being in the spring.
Although the depth of the ASL is broadly related to the SIE in the Ross Sea, it is the meridional component of the near-surface wind over the Ross Sea that has the most direct influence on SIE through the advection of ice north- or south-wards. The seasonal correlation fields of Ross Sea SIE and V10 vary markedly over the year (Fig. 5). In autumn the SIE is only significantly correlated with V10 over the Ross Ice Shelf and the southern part of the Ross Sea, indicating that it is the strength of the katabatic winds flowing down onto the ice shelf and extending out over the Ross Sea that is important in giving a positive SIE anomaly over the ocean. In the winter the Ross Sea SIE is significantly correlated with V10 over parts of the eastern Ross Ice Shelf, but also large parts of Victoria Land, indicating that the ice extent is influenced by the katabatic winds in the coastal area at a time when they are strongest. However, the SIE is also significantly correlated with V10 over much of the Ross Sea, indicating the importance of the synoptic conditions over the ocean as the ice begins to extend further north. In spring the Ross SIE is correlated with V10 over a large part of the Ross and Amundsen Seas, with correlation values of >0.6, but there is no significant correlation over the Antarctic continent or the coastal regions, indicating that synoptic conditions over the northern part of the sea ice zone have the greatest influence on the SIE. During summer the Ross SIE is only significantly correlated with V10 over a small area of the coastal region of the Amundsen Sea.
For the year as a whole the MSLP between the Antarctic Peninsula and the Ross Sea has decreased by up to ~−0.70 hPa dec−1 (Fig. 3d), although the trend is not significant. However, this has strengthened the climatological southerly winds over the Ross Sea by up to ~0.15 m s−1 dec−1 (p < 0.05) (up to ~8 % over the 35 years). The trend in the depth of the ASL varies markedly over the year, with consequent impact on the changes in V10 over the Ross Sea and SIE in the Ross Sea. The deepening of the ASL of ~1.8 hPa dec−1 has been largest in the autumn, although the largest decrease in MSLP was over the Bellingshausen Sea, so that the impact on the sea ice of the Ross Sea was limited. However, in spring the MSLP over the northern Ross Sea has decreased by up to ~1.2 hPa dec−1, which is an optimal location to increase the strength of the southerly wind near the sea ice edge.
The changes in atmospheric circulation off West Antarctica in recent decades could be the result of alterations in one or more of the forcing factors that affect the area, such as changes in tropical SSTs or the loss of stratospheric ozone, or could be the result of intrinsic variability within the climate system. Estimating the impact on individual forcing factors is difficult, but it is instructive to examine the magnitude of the intrinsic variability of the key atmospheric drivers of sea ice variability in this area to determine if the recent observed changes are exceptional.
In order to investigate whether the changes in the atmospheric circulation off West Antarctica are within the bounds of intrinsic variability we have divided the ~20,000 years of output from the 51 CMIP5 model control runs into 35 year periods with a 1 year separation. The 35 year trend was selected since it corresponds to the length of the period for which we have reliable reanalysis data. We examined two quantities in the control runs. Firstly, the 35 year trends in the depth of the ASL, which we determined from the mean MSLP over 60°–75°S, 170°E–75°W. Secondly, the 35 year trends in V10 over the Ross Sea sector of 60°–75°S, 160°E–130°W.
Figure 6 illustrates the annual trends in the ASL depth and mean V10 over the Ross Sea sector in the CMIP5 model control runs using box and whisker plots. They show the trends since 1979 determined from the ERA-Interim data (indicated by red lines) and the distribution of all 35 year trends for each model run and across all model runs.
For the MSLP in the region of the ASL, the annual trend observed since 1979 is outside the inter-quartile range for all the models and also the multi-model mean. However, it is within the two standard deviation (SD) range of all the models. The drop in MSLP observed since 1979 is therefore not exceptional compared to the model control runs, even though these 35 years are unique in that it is the period over which there has been a loss of stratospheric ozone. The development of the ‘ozone hole’ has resulted in a higher frequency of the SAM being in its positive phase in austral summer and autumn (e.g. Thompson et al. 2011) and has been associated with a decrease of MSLP around Antarctica.
The strength of the meridional component of the near-surface wind over the Ross Sea is significantly correlated with the SIE in this sector of the Southern Ocean. Figure 6b shows that the observed trend in V10 since 1979 (0.0369 m s−1 dec−1) is close to the upper quartile of many of the model ensemble mean trends, but within 2 SD of all of the models.
We have also examined the 35 year trends in ASL depth and V10 over the Ross Sea in the historical runs of CMIP5. The range of trends over the period since 1850 has been indicated as an additional multi-model ‘whisker’ on Fig. 6a, b. The range of trends is very similar to that for the ‘all control runs’ range, suggesting that adding GHG and ozone hole forcing does not have a marked impact on the depth of the ASL or the Ross Sea near-surface winds, and that the changes are dominated by internal variability.
Considering the seasonal data (not shown), in winter the observed trends in ASL MSLP and Ross V10 are both very small and close to the model mean trends, and are within the 2 SD range of every CMIP5 model. The observed summer deepening of the ASL can be attributed largely to the loss of stratospheric ozone, which is not present in the pre-industrial control runs. The observed trend is therefore beyond the 2 SD range of trends in three of the models. In contrast, the observed trend in Ross Sea V10 is smaller and within the 2 SD range of all the models. The multi-model mean ASL depth in the historical runs has decreased since the 1970s at a rate similar to that observed, as the SAM has become more positive and summer MSLP decreased all around the Antarctic. However, the historical runs do not show any marked differences in Ross Sea V10 trends compared to the control runs.
The seasons of largest observed increase in Ross Sea V10 are autumn and spring. In the autumn the marked decrease of MSLP of 0.08 hPa dec−1 is beyond the 2 SD range of 14 of the models, however, the observed deepening has been close to the Antarctic Peninsula so that the observed increase of Ross V10 is relatively modest and within the 2 SD range of all 51 CMIP5 models. The Ross Sea V10 has increased most during spring, with the trend of 0.14 m s−1 dec−1 being a major factor in giving the largest seasonal trend in Ross Sea SIE. The observed trend in ASL depth was not particularly large, but the decrease in MSLP was over the northern Amundsen and Ross Seas, so that the stronger southerly flow advected the sea ice northwards in the Ross Sea. The observed increase in Ross V10 was beyond the 2 SD range of four CMIP5 models, although within the multi-model mean 2 SD range.