1. Introduction

There is general consensus that the Antarctic sea ice responds to atmospheric variability on a variety of time-scales. Reference WalkerWalker (1923, Reference Walker1924) pioneered the observations on the relationships between the limit of the Antarctic sea ice and variations in sea-level pressure, storm tracks and the Southern Oscillation Index (SOI). Later, Reference Simmonds and JackaSimmonds and Jacka (1995) examined the links between SOI and sea ice in several sectors of Antarctica. They found strong relationships between ice extent in the southeast Indian Ocean in April– October and the SOI of the previous 12 months. More recently, Reference HarangozoHarangozo (2000) showed that the El Niño– Southern Oscillation (ENSO) teleconnection patterns and related changes in the meridional circulation can influence sea-ice extent and surface air temperature west of the Antarctica Peninsula (Bellingshausen Sea). In addition to ENSO, other atmospheric low-frequency modes can influence the interannual variability of the Amundsen and Bellingshausen sea-ice area (SIA). For instance, the Antarctic dipole, characterized by a quasi-stationary wave, has an out-of-phase relationship with sea-ice anomalies and temperature over the Bellingshausen and Weddell Seas (Reference Yuan and MartinsonYuan and Martinson, 2001; Reference RenwickRenwick, 2002).

Reference White and PetersonWhite and Peterson (1996), using data from a variety of observational techniques, identified significant interannual variations in the atmospheric pressure at sea-ice level, wind stress and sea-ice extent over the Southern Ocean. This system of coupled anomalies, known as the Antarctic Circumpolar Wave (ACW), propagates eastward, takes 8–10 years to encircle the pole and plays an important role in climate regulation and dynamics both within and beyond the Southern Ocean. The authors showed that the ACW can generate anomalies in the sea-ice extent, such as the intense ice retraction observed in the summers of 1988–91 in the Bellingshausen Sea.

Reference Baba and WakatsuchiBaba and Wakatsuchi (2001) found an eastward-propagating pattern on intra-seasonal timescales (5–30 days) with zonal wavenumber 2–4 in the marginal ice concentration around Antarctica. Reference Baba, Minobe, Kimura and WakatsuchiBaba and others (2006) applied complex empirical orthogonal functions (CEOFs) to examine the relationships between intra-seasonal (5–30 days) daily sea-ice concentration (1991–2001), ice-drifting velocity and meridional and zonal wind velocity at 10 m. They showed that the largest amplitudes in variability along the marginal sea-ice zone occur around West Antarctica, especially in the Amundsen and Bellingshausen Seas. The first CEOF mode indicated that the spatial phase of the meridional wind velocity precedes that of the sea-ice concentration by 90˚.

Our objective is to investigate extreme intra-seasonal anomalies in the Amundsen and Bellingshausen SIA, with focus on persistence and interannual variability of the events. The relationships between the Amundsen and Bellingshausen extreme SIA on intra-seasonal timescales and the propagation of wave trains in the subtropics and extratropics of the Southern Hemisphere (SH) are also examined. We investigate the vertical structure of the atmospheric wave trains and the relationships with changes in circulation near the surface and respective response on the Amundsen and Bellingshausen sea ice.

2. Data

Daily SIA of the Amundsen and Bellingshausen Seas is a satellite estimate from the scanning multichannel microwave radiometer (SMMR) and Special Sensor Microwave/ Imager (SSM/I F-8, SSM/I F-11e, SSM/I F-13) available from the US National Snow and Ice Data Center (NSIDC) during 1979–2004. Total ice-covered area is defined as the area of each pixel with at least 15% ice concentration, multiplied by the ice fraction (i.e. 0.15–1.00) in the pixel. Data from 1979 to 1987 were originally obtained every 2 days and were further interpolated to daily resolution.

Global daily US National Centers for Environmental Prediction/US National Center for Atmospheric Research (NCEP/NCAR) re-analysis data (Reference KalnayKalnay and others, 1996), with 2.5^˚ latitude and longitude, during 1979–2004 are investigated. The following variables are examined: 200 hPa geopotential height (gpm), zonal and meridional winds at 850 and 200 hPa (ms^–1) and skin temperature (K). The 200 hPa geopotential height and circulation at low and high levels are used to examine the main dynamical features associated with the atmospheric disturbances that propagate toward the Antarctic Peninsula and cause variations in SIA. Skin temperature is used here as a proxy for sea surface temperature anomalies (Reference KalnayKalnay and others, 1996).

3. Methods and Results

The Amundsen and Bellingshausen SIA is characterized by a very pronounced annual cycle. In addition, trends in SIA are observed in some seasons. Reference LimaLima (2007) showed a negative and statistically significant SIA trend (corresponding to about –371 875 km² during the 1979–2004 period) for the month with minimum SIA (February) in the Amundsen and Bellingshausen Seas. Nevertheless, no statistically significant trends were found during the month of maximum extent (September). Reference Zwally, Comiso, Parkinson, Cavalieri and GloersenZwally and others (2002) found negative and near-zero SIA trends during the winter and spring (1979–98) in all Antarctic oceanic sectors with the exception of the Ross Sea, which shows positive trends in all seasons. In order to identify the importance of other frequencies in the SIA spectrum, the annual cycle was removed from the data. The long-term linear trend was removed from daily data, and the SIA mean annual cycle was computed from the detrended SIA (1 January 1979 to 31 December 2004). The resulting mean annual cycle was further smoothed and removed from the detrended SIA.

The spectral density of daily Amundsen and Bellingshausen SIA anomalies indicates statistically significant peaks on intra-seasonal timescales of 20–100 days (Fig. 1). Based on this result, the Amundsen and Bellingshausen SIA were band-filtered such that only the 20–100 days band is retained. A fast Fourier transform filter was used for this purpose (e.g. Reference HelmsHelms, 1967; Reference Carvalho, Jones and AmbrizziCarvalho and others, 2005). Extreme intra-seasonal SIA anomalies (SIA_IS) were then determined based on the quartiles of the SIA_IS seasonal distribution. The focus of the present study is on the winter (June–September) because previous analyses (Reference LimaLima, 2007) indicated that the links between atmospheric intra-seasonal variability and SIA_IS are stronger during this season, compared with summer. Extreme positive, equivalent to expansion, (negative, equivalent to retraction) SIA_IS anomalies are defined here when SIA_IS ≥ upper quartile (SIA_IS ≤ lower quartile) of the distribution of anomalies observed during June–September.

Fig. 1. Spectral density of daily Amundsen and Bellingshausen SIA (km²). Thick (thin) solid (dashed) line represents the background red noise (95% confidence limit) spectrum. Only the intra-seasonal (20–100 days) band is shown for simplification. Hamming window (L = 5) was applied to the spectrum.

Persistence, defined as the number of consecutive days of extreme SIA_IS, is normally distributed with median ~12 days and interquartile range ~10 days for both positive and negative SIA_IS during the winter (June–September). The interannual variability of the mean persistence of winter extreme SIA_IS events (Fig. 2) is not stationary. During 1979–90 the lower quartile of the persistence distribution of negative (positive) SIA_IS extreme events is 8 days (9 days). The upper quartile of the persistence distribution of negative (positive) SIA_IS extreme events during 1979–90 is 16 days (15.5 days), with an interquartile range of 8 (6.5) days. During 1991– 2004 the lower quartile of negative (positive) SIA_IS extreme events was 7 days (7 days) and the upper quartile was 19 days (16 days), yielding an interquartile range of 12 days (9 days). The increase in the interquartile range for negative anomalies (i.e. retraction of SIA) in the winter of the last 14 years is essentially due to the increase in the upper quartile after 1991, which in turn is particularly influenced by large mean persistence observed in 1991, 1996 and 2001 (Fig. 2).

Fig. 2. Interannual variability of the mean persistence of extreme positive (black bars) and negative (grey bars) SIA_IS anomalies during the winter (June–September) in the Amundsen and Bellingshausen Seas.

The interannual variability of the number of independent extreme SIA_IS events during the winter with persistence ≥2 days is shown in Figure 3. Extreme SIA_IS negative and positive anomalies occur every winter, but not with the same proportion and variability. It is noteworthy that the variability in the number of intra-seasonal events increased after 1990. For instance, during 1979–90 there was no winter with more than three SIA_IS extreme events, either negative or positive. In the following 14 years, four SIA_IS extreme events occurred in approximately 42% of the winters. The increase in the number of SIA_IS extreme events and in the upper-quartile and interquartile range of the persistence of events may have implications for the sea-ice annual variability and trend observed in Reference Zwally, Comiso, Parkinson, Cavalieri and GloersenZwally and others (2002) and Reference LimaLima (2007).

Fig. 3. Interannual variability of the number of extreme positive (black bars) and negative (grey bars) SIA_IS anomalies during the winter (June–September) in the Amundsen and Bellingshausen Seas. Only events persisting at least 2 days are included.

The statistical significance in the difference between the number of extreme events observed in the decades 1979–90 and 1991–2004 was assessed by testing the difference in the winter averages in the two periods (e.g. Reference Spiegel, Schiller and SrinivasanSpiegel and others, 2000). During 1979–90 the average number of negative (positive) extreme SIA_IS events was 2.75 (2.25) events per winter, with standard deviation 0.45 (0.75) events per winter. During 1991–2004 the average number of negative (positive) extreme SIA_IS events increased to 2.93 (2.93) events per winter, with standard deviation 0.73 (0.73) events per winter. Statistical significance (at 5% significance level) was observed only for the increase in the number of extreme positive SIA_IS events. Reference Jacobs and ComisoJacobs and Comiso (1997) showed changes in the interannual variability of sea-ice extent and a decrease of about 20% in the Amundsen and Bellingshausen sea-ice extent during two decades after 1973, particularly in the summer. They also showed that the interannual decrease of sea-ice extent in the oceans westward of the Antarctic Peninsula was negatively correlated with the air temperature at the surface, which increased at a rate of about 0.5˚C (10 a)^–1 since 1940. In the present study, however, we show that the mean number of extreme SIA_IS anomalies increased during the winter in the 1991–2004 period compared with the 1979–90 period. Nevertheless, the winter average persistence of negative (i.e. retraction) SIA_IS events also increased in the 1991–2004 period, resulting in changes in the characteristics of the persistence distribution as suggested by changes in the upper- and lower-quartile and interquartile ranges. The decrease in sea-ice extent on interannual timescales discussed by Reference Jacobs and ComisoJacobs and Comiso (1997) during summer, and by Reference Zwally, Comiso, Parkinson, Cavalieri and GloersenZwally and others (2002) during winter, results from the interplay of phenomena occurring on several timescales. The present study suggests, however, that the persistence of the anomalous SIA retractions on intra-seasonal timescales in the winters of the last 14 years may have played a role for the reported seasonal trends.

Changes in the number of SIA_IS events after 1990 seem to be consistent with the study of Reference Fogt and BromwichFogt and Bromwich (2006). They showed that the decadal variability of ENSO influenced the intensity of the teleconnection patterns in the 1990s similarly to the 1980s, mainly in the austral spring. They showed that the 1980s September–November teleconnection is weak due to the relationships between the Pacific–South American pattern associated with ENSO and the Southern Annular Mode. The significantly positive correlation between these two modes during September– November (also existing in December–February) in the 1990s amplifies the geopotential height and pressure anomalies over the South Pacific, strengthening the tropical–extratropical teleconnection identified through empirical orthogonal function analyses. It is possible therefore that the decadal variations in the number and persistence of SIA_IS extreme events observed after 1990 are linked to the mechanisms discussed by Reference Fogt and BromwichFogt and Bromwich (2006). Moreover, Reference Jones and CarvalhoJones and Carvalho (2006) have shown that the number of Madden–Julian Oscillation events, the most significant intra-seasonal disturbance in the tropics (Reference Madden and JulianMadden and Julian, 1994), increased during the austral winter in the 1990s compared to the 1980s. The Rossby response of the Madden–Julian Oscillation plays an important role in modulating circulation in the extratropics, with subsequent influence on wave activity and phases of the Antarctic Oscillation (Reference Carvalho, Jones and AmbrizziCarvalho and others, 2005).

The association of the propagation of atmospheric wave trains on intra-seasonal timescales with the SIA_IS extreme events is illustrated in the lag composites of Figures 4–8. The intra-seasonal wave train can be analogously compared to the ACW on interannual timescales (Reference White and PetersonWhite and Peterson, 1996; Reference Venegas, Drinkwater and ShafferVenegas and others, 2001). Both modes represent a coupled ocean–atmosphere system that can cause anomalies in the Antarctic SIA and extent. Lag composites illustrate different phases of propagation of the intra-seasonal atmospheric disturbances and are performed for SIA_IS negative and positive events separately. In Figures 4–8, lag = 0 means that the atmospheric field is observed along with the occurrence of the SIA_IS. Negative (positive) lag indicates that the feature observed leads (lags) the events by the number of days indicated at the top of each frame. All atmospheric variables are band-filtered on intra-seasonal timescales (20–100 days). The statistical significance of the mean anomaly is accessed by performing at test at 5% significance level. We considered as independent events only those that are at least 2 days apart.

Fig. 4. Lag composites of the 200 hPa geopotential height (gpm) for dates with negative SIA_IS anomalies in the Amundsen and Bellingshausen Seas: (a) 10 days before the event; (b) 5 days before the event; (c) along with the event; and (d) 10 days after the event. Solid (dashed) lines indicate positive (negative) values starting with 200 gpm (–200 gpm) with interval 10 gpm. Shades show statistically significant anomalies at 5% significance level. Number of degrees of freedom (DOF) is 70, which corresponds to the number of independent events minus 2.

Fig. 5. Same as Figure 4, but for SIA_IS positive anomalies (DOF = 68).

Fig. 6. Same as Figure 4, but for the zonal wind intra-seasonal anomaly (m s^–1) at 850 hPa (U850_IS). Solid (dashed) lines indicate positive (negative) anomalies starting at 5 ms^–1 (–5ms^–1) with 1 ms^–1 interval. DOF = 70.

Fig. 7. Same as Figure 6, but for the intra-seasonal meridional wind anomaly (m s^–1) at 850 hPa (V850_IS).

Fig. 8. Same as Figure 7, but for skin temperature (K) intra-seasonal anomaly (SKT_IS). Solid (dashed) lines indicate positive (negative) anomalies starting at 4 K (–4 K) with 0.5 K interval. DOF = 70.

Figure 4 shows lag composites of the 200 hPa geopotential intra-seasonal anomalies (H200_IS) during extreme negative (ice retraction) SIA_IS events. A well-defined extratropical wave train can be observed 10–15 days before the events, lasting up to 10–15 days following the events. Around 20 days after (20 days before) the events, the wave pattern weakens (not shown). Reference Simmonds, Domack, Leventer, Burnett, Bindschadler, Convey and KirbySimmonds (2003) pointed out that the Antarctic Peninsula, which includes the Weddell and Bellingshausen Seas, is the region in the SH that is most subjected to the influence of atmospheric systems that develop in mid-latitudes, including those with longitudinal propagation. For lag = –10 days (Fig. 4a), a wavenumber-2 pattern in the H200_IS is clearly observed in the mid-latitudes of the South Pacific extending towards the South Atlantic. Positive geopotential anomalies are observed west of the Antarctic Peninsula and around New Zealand. For lag = –5 days (Fig. 4b), positive geopotential anomalies intensify over the Amundsen and Bellingshausen Seas, and on lag = 0 (Fig. 4c) this centre moves towards the eastern Antarctic Peninsula. It is interesting to notice also the influence of the wave train upon the subtropics, particularly over La Plata basin in South America and southern Africa on lags –5 and 0, when a zonal wavenumber 3 becomes evident. By lag = 10 days (Fig. 4d), negative H200_IS anomalies are observed over the Antarctic Peninsula as the wave train moves eastward, with spatial distribution of anomalies almost opposite to what is observed for lag = –10 days (Fig. 4a). The intra-seasonal wave train presents a coherent pattern of geopotential and circulation from low to high levels (not shown).

The occurrence of positive SIA_IS anomalies in the Amundsen and Bellingshausen Seas also depends on the phases of propagation of the mid-latitudes intra-seasonal wave train. Figure 5 illustrates this dependence by showing H200_IS lag composites obtained with respect to the occurrence of extreme intra-seasonal SIA_IS positive anomalies (i.e. extreme expansion of the sea ice on intra-seasonal timescales). For lag = –10 days (Fig. 5a) the wave pattern resembles that observed for lag = 10 days for negative SIA_IS anomalies (Fig. 4a). That is, negative geopotential anomalies are observed over the Antarctic Peninsula, particularly over the Amundsen and Bellingshausen Seas for lags of –5 to 0 days (Fig. 5a and c), being replaced by positive anomalies for lags of 5–10 days (Fig. 5d). These negative geopotential height anomalies are collocated with southerly V850_IS anomalies (not shown), which implies southerly advection of cold air and expansion of the ice sheet. It is important to notice that not all extreme positive (negative) SIA_IS anomalies are necessarily followed by extreme negative (positive) anomalies as the wave train moves eastward. The occurrence of extreme variations in the ice may depend on the interaction of several scales and local factors. Moreover, variations in the Southern Ocean currents, sea surface temperature and salinity (Reference Gordon and TaylorGordon and Taylor 1975; Reference GordonGordon 1981; Reference Lohmann and GerdesLohmann and Gerdes 1998), not explored in the present study, can play an important role in the sea-ice variability on intra-seasonal timescales.

Lag composites of zonal wind intra-seasonal anomalies at 850 hPa (U850_IS) associated with extreme negative SIA_IS anomalies are shown in Figure 6. The wavenumber-3 pattern is already evident 10 days before the event (Fig. 6a) and resembles the tropical–extratropical teleconnection patterns suggested in Reference Mo and PaegleMo and Paegle (2001). Westerly intra-seasonal wind anomalies are observed west of the Antarctic Peninsula for lag = –10 days (Fig. 6a) to lag = 0 (Fig. 6c). The strongest anomalies are observed preceding the events 10–5 days. For lag = 0 (Fig. 6c), U850_IS weakens near the Antarctic Peninsula, and for lag = 10 days (Fig. 6d) easterly anomalies replace the westerly anomalies, indicating a wave phase opposite to that observed for lag = –10 days (Fig. 6a).

Lag composites of intra-seasonal anomalies of the meridional wind component at 850 hPa (V850_IS) and skin temperature (SKT_IS) are shown in Figures 7 and 8, respectively. For lag = –10 days (Fig. 7a) northerly V850_IS anomalies are observed west of the Antarctic Peninsula, collocated with the Amundsen and Bellingshausen Seas. The low-level northerly components bring warm air to the west of the peninsula, and the net result is the increase of SKT_IS anomalies (Fig. 8a) due to the increase of sea surface temperature combined with the poleward retraction of the ice limits. Conversely, southerly V850_IS anomalies are observed along with a decrease in SKT_IS anomalies for all lags. Northerly V850_IS anomalies preceding negative SIA_IS anomalies persist and intensify from lag = –10 days to lag = –5 days (Fig. 7b), which is followed by an increase in SKT_IS anomalies (Fig. 8b) over the Amundsen and Bellingshausen Seas. V850_IS and SKT_IS anomalies weaken for lag = 0 (Figs 7c and 8c), and an opposite phase is observed for lag = 10 (Figs 7d and 8d). The eastward propagation of the SIA_IS anomalies has been identified also by Reference Baba, Minobe, Kimura and WakatsuchiBaba and others (2006).

Conclusions

The spectral density of the daily Amundsen and Bellingshausen SIA for 1979–2004 showed several statistically significant peaks on intra-seasonal timescales within the 20–100 day band. Some peaks occur at longer timescales than have been observed before (Reference Baba and WakatsuchiBaba and Wakatsuchi, 2001; Reference Baba, Minobe, Kimura and WakatsuchiBaba and others, 2006). Extreme variation in the band-filtered SIA_IS, defined according to the upper and lower quartiles of the SIA_IS, occurs every winter season. However, the proportion of events that cause extreme expansion or retraction of SIA varies from year to year. The total number of events causing extreme intra-seasonal variations in the SIA is about one to four per winter. There is some indication of an increase in the average number of events per year in the 1991–2004 period compared to 1979– 90. Statistical significance is found (at 5% significance level) for the increase in the average number of extreme positive (expansion) SIA_IS events per year. On the other hand, the upper quartile of the persistence (defined as consecutive days with extreme anomalies) of extreme negative (retraction) SIA_IS has also increased in the last 14 years. The combined effect of a larger number of SIA_IS events (either positive or negative) and the increase in the upper quartile of the persistence of ice retraction on intra-seasonal timescales, particularly during the 1991, 1996 and 2001 winters, may have played a role for the winter SIA trends in the last decade. The decadal variation of SIA_IS has observational support in the studies of Reference Baba and WakatsuchiFogt and Bromwich (2006) and Reference Jones and CarvalhoJones and Carvalho (2006).

Atmospheric disturbances on intra-seasonal timescales that propagate to the SH extratropics as wave trains have important impacts on the occurrence of extreme SIA_IS. We show that these disturbances appear as a wave 2–3 in the geopotential height and circulation. The origin of these disturbances is still debatable, but their spatio-temporal characteristics suggest the importance of tropical–extratropical teleconnections on intra-seasonal timescales. The occurrence of extreme retraction or expansion of the Amundsen and Bellingshausen SIA depends on the phase of propagation of the wave train and the persistence of the event. Nevertheless, the occurrence of extreme retraction in one phase of the wave does not always imply extreme expansion during the other phase, due to local factors, ocean currents and many other causes not explored in the present study.