2.4                                   Synoptic–scale weather systems in the Antarctic  

One of the most remarkable features of the sub–Antarctic latitudes is the high frequency of cyclonic storms, many of which are intense. These systems have a large impact on weather in the region and particularly at coastal locations, where most Antarctic stations are located (Simmonds, 1998). It is clearly important for a number of reasons that the behaviour of these systems be understood and that they be forecast with acceptable accuracy. These transient systems are also significant in a climatological sense in that they effect most of the pole–ward energy transport in the region (van Loon, 1979; Peixoto and Oort, 1992).

The study of cyclones and cyclogenesis has experienced a remarkable renewal of interest in recent times due to the emergence of new theoretical problems and new approaches to diagnose these phenomena (Mass, 1991; Joly et al., 1997; Turner et al., 1998). There have been a number of significant papers that have examined the behaviour of cyclones over the Southern Hemisphere. Taken chronologically, these reflect a steady improvement in our fundamental understanding of these systems, the quality of the analyses from which they were obtained, and the techniques that have been used to identify systems. Amongst the literature we confine ourselves to citing the works of Lamb and Britton (1955), Taljaard (1967), Streten and Troup (1973), Carleton (1979, 1983), Kep (1984), Leighton (1992), Jones and Simmonds (1993a), Sinclair (1994, 1995, 1997), Simmonds and Murray (1999), and Simmonds et al. (1999).

In attempting to understand and document the characteristics of cyclones in the high southern latitudes, it is important to bear in mind that this is a region of great temporal variability of the synoptic environment (seen on a broad range of time scales) and is also presently displaying significant change (e.g., Jones and Simmonds, 1993b; King, 1994; Simmonds et al., 1998; Simmonds and Keay, 2000). Hence consideration of only short epochs may lead to an incomplete understanding of the behaviour of these important transient features.

2.4.1                                Depression occurrence

Among the most important developments to have taken place in recent decades in terms of our understanding of the behaviour of cyclonic systems has been the evolution of high quality digital analyses, and the development of automatic algorithms to find and track systems in them. One of these is the Melbourne University automatic cyclone–tracking scheme (Simmonds and Murray, 1999; Simmonds et al., 1999). This scheme has been applied to the six–hourly global re–analyses produced by the National Centers for Environmental Prediction (NCEP) (Kalnay et al., 1996)  (which cover the period 1958–97) and some of the results are shown below.

The density of systems (D) (the mean number per analysis found in a 103 (deg. lat.)2 area) in summer (December to February) and winter (June to August) over the 40–year period is presented in Figure In summer (panel a) in the Atlantic and Indian Ocean sectors the greatest cyclone densities exceed 4 × 10–3 cyclones (deg. lat.)–2 near 60o S, while in the Pacific the axis of the maxima lies somewhat further south. In winter (panel b) local maxima are also seen in the Bellingshausen Sea and in Prydz Bay. For the most part, the frequency is higher in this season, although this is not true everywhere (e.g., northern parts of the Weddell Sea).

( a )

( b )

( c )

( d )

( e )

Figure      (a) System density in DJF identified in the automatic algorithm for the period 1958–97; (b) as part (a) but for JJA; (c) July distribution of manually–derived cyclonicity for the period 1973–94; (d) system density in July identified in the automatic algorithm for the period 1973–94; and (e) when only systems with closed centres are considered. (The contour intervals in (a), (b), and (d) are 2×10–3 (deg. lat.)–2 (with an extra contour at 1×10–3 (deg. lat.)–2). In (c) the contour interval is 5 hours per month per 5×5 latitude‑longitude cell. In (e) the contour interval is 1×10–3 (deg. lat.)–2.)

It is important that these densities be compared with those obtained with more traditional techniques. Figure (c) displays the July distribution of ‘cyclonicity’ derived manually by Leighton from the Australian operational analyses for the 22–year period 1973‑94. Cyclonicity is defined as the total time (in hours) per month during which cyclone centres occupy a given 5 × 5 degree latitude–longitude cell. This method of counting can be understood to give values that are related to the ‘density’ measure used above, except for the fact that the counting base area becomes smaller as latitude increases. One can see that the overall structure of the plot is similar to that produced by the automatic scheme, with both showing maxima just to the south of 60o S in the Indian Ocean and to the south of Australia. However, it is important to be clear on the relationship between these measures of cyclone frequency.

To quantify this relationship in general let the cyclonicity in a box of longitudinal range dl and bordered by latitudes q1 and q2 (for an N–day period) be denoted by C. Then the mean number of cyclones to be found in that box per analysis is C/(24N). To allow comparison between the two measures of cyclone density we need to compare the two base areas. It can be easily shown that the area of the box considered above is given by Equation

where a is the radius of the Earth, dq = ( q1q2 ) / 2, and q = ( q1 + q2 ) / 2. The area of 103 (degree latitude)2 is given simply by Equation

It follows from the above that the general relationship for system density of cyclonicity is given by Equation

where the conversion factor, F, is a given by:

For the box size used by Leighton and taking N = 31 (the case for January and July) this reduces to D = 0.107 × C at 60o S, and D = 0.084 × C at 50o S.

It will be seen from Figures (b and c) that the cyclonicity in the manual analysis is about one third to one half that which would be implied by the automatic algorithm. Or, put another way, the computer scheme finds considerably more systems in the sub–Antarctic region, in agreement with the findings of Simmonds et al. (1999). To convince ourselves that this difference in cyclone numbers is not simply due to the different periods of analysis used in the two schemes, the mean July system density from the automatic scheme has been drawn for the same period used by Leighton (i.e., 1973–94). The mean system density for this period (Figure (d)) can be seen to be somewhat smaller than that for the 40–year period, but not by an amount to explain the discrepancy referred to above. The reason for the difference is resolved when we confine our system density compilation to consider only closed systems. The resulting system density distribution and values produced by the automatic scheme (Figure (e)) can be seen to be similar to that obtained by the manual process (once the correction above has been applied). That the difference can be explained in this way should not come as a surprise, because it will be appreciated that in regions of strong background pressure gradients and frequent systems the manual analyst will probably be biased toward the counting of ‘closed’ systems. It can be argued that open depressions are of similar importance to closed systems, and the automatic scheme will find these more easily that would the manual analyst. All the points raised here serve to indicate that when discussing cyclone occurrence and behaviour, there must be a clear definition of these features, and that the comparison of various compilations must be undertaken with the difference in definition in mind.                          Trends in depression occurrence

We saw above that the July sub–Antarctic system density during the period 1973–94 was somewhat less than that diagnosed over the 1958–97 epoch. For forecasting it is important to have an appreciation of the background level of cyclone activity, and the extent to which this may be changing. Prompted by the above observation, we have calculated the linear trend displayed by July cyclone density over the four–decade period of the reanalysis, and present this in Figure (a).





Figure     (a) Linear trend in July system density identified in the automatic algorithm for the period 1958–97; (b) linear trend in July system density identified in the automatic algorithm for the period 1973–94; (c) linear trend in July manually‑derived cyclonicity for the period 1973–94; (d) as for part (b) but when only systems with closed centres are considered. (Stippling denotes regions over which the trends differ significantly (95% confidence level) from zero. The contour interval is 0.2×10–3 (deg. lat.)–2decade–1, except in part (c) where it is 2 hours per month per 5 × 5 latitude–longitude cell per decade.)

There has been a general reduction, much of which is statistically significant, in cyclone numbers over most of the domain. The greatest reductions occur near the 60o S latitude circle. (Immediately on the coast there are several “bull's–eyes” of increases in the number of cyclones. Simmonds and Keay (1999a) noted similar features in their analysis of the annual totals of cyclones and offered reasons for the existence.) It is interesting to note that the reduction appears not to be dominated by changes in a given sub period. As support for this statement, we see that the July trends calculated over the 1973–94 epoch (Figure (b)) are very similar.

Leighton and Deslandes (1991), Leighton (1997), and Leighton et al. (1997) have also identified regional trends in manually–derived cyclonicity. When Leighton’s analysis of 22 years of July cyclonicity in the sub–Antarctic region is subject to trend analysis (Figure (c)) there are a number of features in common with those revealed by the automatic scheme over the period. Among these are the reductions in much of the eastern Pacific and to the north of the Weddell Sea, and the changes to the south of Australia. However, the sign of the change in the central Indian Ocean to the north of 60o S is opposite in the two compilations. In light of our discussion above it would seem fruitful to determine the trends for the automatic scheme when consideration is restricted to closed systems. Figure (d) shows that in this case positive trends are diagnosed in the central Indian Ocean, in broad agreement with the sign of the change revealed in the manual analysis. Following on from our point raised above, it is clear that the definition of a cyclone must be clearly expressed before we can speak sensibly of secular changes in their behaviour.

2.4.2                                Cyclogenesis

Here we consider the mean genesis of cyclonic systems in the Antarctic region as revealed in the automatic algorithm’s analysis of the 40 years of NCEP data. In summer (Figure (a)) most of the domain experiences genesis rates in excess of 0.5×10–3 cyclones (deg. lat.)–2 day–1. The axis of maxima is found on, or to the south of, 60o S and extrema are found over Graham Land, in the embayment to the north of Tierra del Fiego, off Victoria Land and the Siple Coast, as well as in the southern part of the Weddell Sea. Figure (b) shows that the pattern in winter is rather similar, but the level of cyclogenetic activity is increased. A minimum of cyclogenesis is apparent at about 45o S across much of the Pacific and Atlantic Oceans while a maximum is observed in the Indian Ocean.

2.4.3                                Depression tracks

To obtain a general appreciation of the behaviour of individual cyclones in the sub–Antarctic region we have plotted, as examples, the paths of all cyclones that appear in 1996 and 1997 in the months of January (Figure (a)) and July (Figure (b)). One can see that while these systems have a predominant eastward motion, there is a significant pole–ward translation of most systems, particularly in the region north of about 60o S. In both months there is a very dense network of tracks just off the Antarctic coast, particularly in the Indian and western Pacific Oceans. We had alluded to this aspect in our earlier discussion on the system density.



Figure     Cyclogenesis identified in the automatic

algorithm for the period 1958–97 for (a) DJF and (b) JJA.

                          (The contour interval is 0.25×10–3 cyclones (deg. lat.)–2 day–1.)



                   Figure     Tracks of all cyclones identified in the

                   automatic algorithm in 1996 and 1997 in (a) January and (b) July.

2.4.4                                Cyclolysis

Regarding the distribution of summer cyclolysis (i.e., termination of systems) Figure (a) shows very large values just north of the Antarctic coast, particularly off much of East Antarctica, and in the Ross and Bellingshausen Seas. While the pattern bears some resemblance to that of the cyclogenesis (Figure (a)) it can be seen that for the most part the rates of lysis exceed those of genesis. (Maxima are seen near 20o E and 120o E, and these are areas of only modest cyclogenesis.) Given that a significant proportion of the systems that are generated in the lower latitudes finish their days near the Antarctic coast (Figure (a)) this enhancement is not surprising. Overall, similar remarks may be made in connection with the cyclolysis in winter (Figure (b)) except that the overall level of cyclolytic activity is greater in the winter season.

2.4.5                                Difference between rates of cyclogenesis and cyclolysis

The relationship between the geographical distributions of cyclogenesis and cyclolysis can be appreciated more readily by considering the difference between the two. The patterns of difference are very similar in summer and winter (Figure although it will be noticed that the excess of lysis is greater in winter around most of West Antarctica, particularly in the Ross and Bellingshausen Seas. Over most of the domain of interest here cyclolysis rates exceed those of cyclogenesis. The only exceptions to this of note are that genesis greatly exceeds lysis off the east coast of the southern tip of South America, through the Drake Passage and down the eastern side of the Antarctic Peninsula. This indicates, among other things, that a significant portion of the systems that are born in these regions are mobile and end their days elsewhere. Genesis also exceeds lysis in both seasons in the region around Oates Land.

Hence, with these regional exceptions, over the high southern latitudes cyclolysis exceeds cyclogenesis and hence this domain is a cyclone “graveyards” in this mean sense. Having said this, it is important to bear in mind that there are significant levels of genesis in the sub–Antarctic regions.

2.4.6                                Weather systems over the interior

From what we have shown above it will be realised that few systems have been represented at mean sea level (MSL) over the main part of the Antarctic continent. Part of the reason for this is that the value of the MSLP chart in weather analysis is diminished when that level is far beneath the surface. In fact, to circumvent misleading deductions the automatic cyclone‑tracking scheme ignores any system identified at a location where the height of the orography exceeds 1 km (Simmonds and Murray, 1999). There are many theoretical reasons for believing that very few systems will penetrate inland (e.g., potential vorticity considerations, weakening of baroclinicity).

Notwithstanding these points, it is not unknown for weather systems to be detected over the interior of the Antarctic continent (e.g., Pook and Cowled, 1999). The difficulties associated with the identification of atmospheric pressure systems over the Antarctic interior are well known to analysts and have been discussed by, inter alia, Schwerdtfeger (1984) and Phillpot (1997). As well, identifiable weather systems moving inland from the Antarctic coast have proved very difficult to track. Apart from the obvious limitations of the observational network, cloud signatures of systems are difficult to identify over the underlying very cold ice surface. Furthermore, the elevation of the Antarctic Plateau which rises from approximately 2 km close to the coast to over 4 km at its highest point requires that synoptic and mesoscale systems moving inland have well defined vertical structures in order to survive.



                   Figure     Cyclolysis identified in the automatic algorithm

                   for the period 1958–97 for (a) DJF and (b) JJA.

                   (The contour interval is 0.25×10–3 cyclones (deg. lat.)–2 day–1.)



Figure     Difference between cyclogenesis and

cyclolysis identified in the automatic algorithm for the period

                   1958–97 for (a) DJF and (b) JJA.

                   (The contour interval is 0.2×10–3 cyclones (deg. lat.)2.day–1.)

Early inferences about pressure systems over Antarctica were influenced by charts of equivalent MSLP. However, the method of constructing MSLP charts requires the reduction of station barometric pressures to sea level by assuming that a column of air of known mean virtual temperature exists between the station and the datum. Errors introduced by this process make the practice of producing MSLP analyses over the Antarctic interior unreliable. This means that the regular appearance of very high pressure over Antarctica on MSLP synoptic charts cannot be given a physical significance, a point emphasized by Schwerdtfeger (1984).

Analysts are faced with the difficulty of selecting a pressure level which is not significantly affected by the orography of Antarctica but which is sufficiently close to the surface to be linked dynamically to the surface wind field. Phillpot (1991, 1997) selected the 500–hPa surface as the most suitable level for analysis over East Antarctica. The problem with working at this level is then to obtain sufficient data to complete the analysis. There has been a reduction of upper–air observations in the interior of the continent in recent years. At the time of writing (2002), Amundsen–Scott (South Pole) Station was the only inland station still conducting an upper–air programme of observations throughout the year. This contrasts with the situation during the IGY of 1957–58 when there were nine staffed scientific stations providing meteorological data from elevations of 1,500 m (~4,900 ft) or more (Dalrymple, 1966). This lack of rawinsonde stations makes conventional upper–air analysis impossible without the input of other data. At this stage remote sensing has not provided a satisfactory solution. Whereas Television Infra–red Observation Satellite (TIROS) Operational Vertical Sounder (TOVS) data provide good agreement with radiosonde data over the Southern Ocean their accuracy over the elevated Antarctic continent is problematic and evidence of strong seasonal bias has been detected (Adams et al., 1999). Improvements in the retrieval schemes employed to invert the TOVS data into geophysical quantities may ultimately lead to improvements in the quality of the data over the continent.

To some extent the problem has been eased by the gradual expansion in the network of surface observations that has been achieved in recent years by the installation of automatic weather stations (AWSs). Phillpot (1991) devised a method for estimating 500–hPa geopotential heights from station level observations of pressure and temperature at AWSs with elevations exceeding 2,500 m (~8000 ft). His analyses of the 500–hPa geopotential field over East Antarctica were incorporated in a set of analyses covering the region south of 50° S for the month of July 1994 (Phillpot, 1997; see his Figure 7.5). These analyses form part of a project known as the Antarctic First Regional Observing Study of the Troposphere (FROST) that provided an opportunity to investigate pressure systems over the interior of Antarctica. The FROST Project (Turner et al., 1996a) was designed to study the effects of all sources of “late” data on meteorological analyses over the Southern Ocean and Antarctic region and the probable impacts of these data on the performance of NWP models. A detailed description of the FROST analysis programme is given in Hutchinson et al. (1999).

Phillpot (1997) also contributed analyses for a period in January 1995. The analyses for this period (see his Figure 16) were constructed by modifying the Australian Global Assimilation Prediction model (GASP) analyses with the addition of estimated geopotential heights from AWS observations and some “late” observations from staffed stations. Monthly means for the months of July 1994 and January 1995 were modified by Phillpot (1997) in a similar manner (see his Figure 7.7).                          Anticyclones over continental Antarctica

The over–emphasized strength of the anticyclone over the Antarctic continent, that appears on some MSLP analyses, has little significance and arises from attempts to estimate equivalent MSLP from station level pressures at elevations generally exceeding 1 km. Saucier (1955, (1972 reprint), p. 56) discusses difficulties with pressure reduction to levels below the Earth's surface. For practical purposes, 450 m (~1,500 ft) is regarded as about the altitude limit for which accurate pressure reductions may be made to MSL (see, for example, Meteorological Office (1971, p. 13). However, the 500–hPa surface does not intersect the Antarctic terrain and can be regarded as a useful level at which to investigate the presence of significant anticyclonic systems in the free atmosphere above Antarctica (Schwerdtfeger, 1984; Phillpot, 1997).

Phillpot (1997) has demonstrated the relative complexity of the atmospheric circulation over Antarctica in a sequence of 500–hPa contour fields during the first Special Observing Period (SOP–1) of the FROST Project in July 1994. The analyses contrast the steady west to east movement of the major troughs and ridges north of the continent with the sluggish movement observed over Antarctica. It is notable that the range of pressure readings at surface stations over the Antarctic Plateau during periods of days or weeks are comparable to those observed at mid–latitude stations. As discussed in Turner et al. (1996a) significant variations of surface pressure were found to occur over East Antarctica during SOP–1. At 500 hPa the increase in geopotential over Victoria Land during the period 22 to 28 July 1994 was in excess of 250 m. Parish and Bromwich (1998) have demonstrated how a similar synoptic configuration in late June and early July of 1988 was associated with drainage of air from Antarctica through the Ross Sea, resulting in pressure decreases of 20 hPa or more across a large portion of Antarctica.

The FROST analysis exercise also revealed how a high over the continent can merge into an intense ridge extending from a blocking anticyclone in the southwest Pacific Ocean (Pook and Cowled, 1999). In this case the Pacific ridge appeared to propagate from the east across the plateau. Time series of pressure anomalies at AWSs at elevated locations showed pressure increasing at the easternmost stations first. Surface pressure variations for a selection of Australian AWSs on the Antarctic continent throughout SOP–1 have previously been shown by Turner et al. (1996a) in their Figure 10.

The monthly mean of the 500–hPa geopotential height over East Antarctica from SOP‑1 (Turner et al., 1996a; their Figure 11) identified three maxima (centres exceeding 5,000 gpm (~16,000 ft) over the high plateau of East Antarctica; near 50° E, 90° E and 120° E. This degree of detail was achieved by incorporating estimated geopotential heights from AWSs (see Section 5.4.3 and in particular Figure and contrasts with the published 500–hPa climatologies of Schwerdtfeger (1970) and Le Marshall et al. (1985).                          Cyclones over continental Antarctica

Broad regions of relatively low geopotential height are detectable over the Antarctic continent on the mean (e.g. 500–hPa) contour charts. Drawing on data from the IGY, Dalrymple (1966) identified four main features of the mean circulation in the middle troposphere over Antarctica. Regions of relatively high geopotential were found to occur over central East Antarctica and over Marie Byrd Land in West Antarctica with minima over the Ross and Weddell seas. Schwerdtfeger (1970) demonstrated that these features are preserved in the seasonal cycle for the most part but there is a decrease of total mass over Antarctica in winter and the region of lowest geopotential height tends to be more centred near the pole in summer.

Pook and Cowled (1999) produced a set of analyses of the 500–hPa surface south of 50° S for the period 22 to 28 July 1994 in the FROST SOP–1. These analyses followed the previous set of 500–hPa contour fields prepared for the period 1 to 15 July 1994 (Phillpot, 1997; his Figure 7.5) which incorporated AWS and “late” station data. For this “special” week of analyses, high quality visible and IR imagery from the US Defense Meteorological Satellite Program (DMSP) satellites (see Section was employed in addition to initial fields analysed by Phillpot and other members of the FROST analysis team. In their Figure 3, Pook and Cowled (1999) demonstrated that closed cyclonic circulations were in evidence over the continent in the daily analyses and underwent significant evolution with time. In the Indian Ocean sector the trough was oriented parallel to the coast throughout the period with individual centres over the ocean and, on some occasions, just inland.

Variations in station level pressure experienced at AWSs on the high plateau during the FROST experiment were similar to ranges experienced at lower latitudes of the Southern Hemisphere and suggest that the atmospheric circulation is complex, even at these high latitudes and elevations.                          Cyclones Migrating Inland from the Southern Ocean

According to published climatologies of cyclones over the Southern Hemisphere, lows over the Antarctic continent are extremely rare, particularly over East Antarctica south of 70°S (Sinclair, 1994; Jones and Simmonds, 1993a; Simmonds and Keay, 2000b). Similarly, studies of polar air vortices, or “polar lows” have shown an almost total absence of these systems over the continental interior (Carleton and Carpenter, 1990). While the effects of maritime lows can be felt inland (see, for example, Section 7.10.2) cyclones rarely move over East Antarctica as a discrete entity because of the “so–called” barrier effect of the elevated ice sheet (Bromwich, 1988). In any case, it is not a trivial problem to track cyclonic vortices over the Antarctic continent. Firstly, it is difficult to discriminate cloud from the ice surface on satellite imagery because of the similar brightness temperatures observed in the thermal infra–red channels and the similar albedos of the two surfaces in the visible channels (Turner and Row, 1995). Analysts normally find it necessary to compare and contrast Advanced Very High Resolution Radiometer (AVHRR) channels and to use techniques as basic as detecting cloud shadows on visible channel imagery. Secondly, the limited nature of the conventional observational network makes it very difficult to follow the progress of synoptic and mesoscale systems, although recent improvements in the network of AWSs over East Antarctica have helped analysts.

Although cyclonic systems generally remain offshore, moisture from these systems is regularly transported inland resulting in cloudiness and precipitation. Sinclair (1981) carried out a case study into a situation in December 1978 in which strong positive temperature anomalies and precipitation occurred over the Antarctic Plateau. Facquet (1982) used a compositing technique to develop a conceptual model of moist air advection across the Antarctic coast.

There are few published studies of the identification and tracking of weather systems over the Antarctic interior. Pook and Cowled (1999) have reported a situation in July 1994 during which two vortices were observed to move across the coast of Antarctica and migrate inland over Wilkes Land (see Figure These vortices were apparent in the DMSP satellite thermal IR cloud imagery and were steered across the Antarctic by an intense blocking anticyclone that developed in the southwest Pacific. An atmospheric pressure peak associated with the intensification of this ridge appeared to propagate across East Antarctica from the Pacific sector of the Southern Ocean. The easternmost vortex was detectable at the surface from observations reported by the AWS at Dôme C (74.5° S, 123° E, 3,280 m elevation above mean sea level (AMSL)). This station had indicated a steadily rising barometric pressure over several days but experienced a fall in pressure of approximately 3 hPa as the vortex passed the station and the anemometer recorded a rapid increase in wind speed from approximately 2 m s–1 at 0000 UTC to 7.1 m s–1 at 1200 UTC and back to zero m s–1 by 0000 UTC on the following day. The wind changed direction from easterly to northerly and back to easterly within 24 hours. It is important to note that the diameter of this system was probably never more than 2.5° latitude (275 km) and remained in the mesocyclone scale length throughout its lifetime.

Figure     A DMSP image of cyclonic vortices over East Antarctica at approximately 2300 UTC 26 July 1994. (From Pook and Cowled (1999) – their Fig. 7a.)

This study within FROST demonstrated that vortices originating over the Southern Ocean can penetrate the high plateau of East Antarctica and move well inland before decaying. The development of an intense blocking anticyclone in the Tasman Sea sector appears to have been a critical factor in this case. Although these events are relatively rare, occurrences of this type have the potential to influence precipitation events over the Antarctic interior in a significant way and further study is required to attempt to quantify the effects of these cyclones.

Numerical models currently in operational use do not have the necessary resolution to identify the relatively small cyclonic systems, with their comparatively short lifetimes, which penetrated inland during this study. Nor was it possible to locate them using traditional methods of manual analysis, as the density of observations in Antarctica is too low. (In contrast to the Phillpot method discussed previously where there was a clearly enough data to at least analyse the 500–hPa level). The new generation of mesoscale models nesting within global models may provide the possibility for modelling these systems over the Antarctic in future.

Now that good climatologies exist for many AWSs in Antarctica the technique employed in this paper to detect cyclonic systems over the Antarctic Plateau using AWS anomaly fields and high–resolution satellite imagery appears capable of adaptation to operational use. The possibility exists for analysis of AWS pressure anomalies from high frequency observations (possibly at 1–hourly intervals) with the addition of other parameters such as temperature, dew point and wind. In addition, the derivation of similar anomaly fields from numerical prognoses, especially in the next generation of high–resolution models, could provide a useful predictive capability.