The data collected during the three special observing periods (SOPs) of the Antarctic First Regional Observing Study of the Troposphere project provide an excellent base upon which to study the behavior of cyclonic systems in winter, spring, and summer in the Southern Hemisphere. This paper provides a report on the behavior of these cyclonic systems during the three SOPs as revealed in the twice-daily ECMWF operational analyses.
The study has been undertaken with an objective cyclone tracking algorithm applied to the digital analyses. The results revealed cyclone behavior generally in accord with long-term climatologies developed with this scheme. In the SOPs the authors observed many systems to be generated in the western part of the ocean basins and then to move east and, to a lesser extent, south. In the three periods they found a concentration of tracks just to the north of the Antarctic continent, being particularly noticeable in the Indian Ocean. At the same time, they found significant differences in cyclone behavior between the climatology and the SOPs in specific regions. The monthly mean sea level pressure (MSLP) anomalies during the SOPs were quite large (and exceeded 10 hPa in places), particularly in the Pacific and in the region to the south of Australia. It appears that the anomalous cyclone structure during the SOPs could be related to the anomalies of the MSLP. The authors suggest that the three SOPs cannot be regarded as typical of their time of year, but it could be argued that no specific period could be so regarded.
The results obtained with these high quality analyses during the SOPs have confirmed the Antarctic coast as a region of high cyclone density and of very active cyclogenesis. The identification of these high levels of coastal cyclogenesis appears to differ from early studies that suggested the greatest (winter) cyclogenetic activity to be much farther north in the 40°–50°S region. The results presented here, however, concur with recent studies undertaken with high-resolution satellite data and four-dimensional data analyses, and the theoretical consequences of the baroclinic structure of the Antarctic coastal region.
One of the most remarkable features of the atmosphere in the mid- to high latitudes of the Southern Hemisphere (SH) is the presence of large numbers of intense, mobile cyclonic systems. These are linked with daily weather over a significant fraction of the hemisphere and, particularly, on the coastal regions of Antarctica. Less obvious, but of no lesser importance, is the fact that these systems are associated with considerable southward fluxes of heat and moisture, and hence they play an important role in the maintenance of the global energy and hydrological cycles. The study of these systems is an important component of understanding the workings of SH weather and climate and, in particular, the Antarctic environment.
One of the specific goals of the Antarctic First Regional Observing Study of the Troposphere (FROST) project (Turner et al. 1996) was to “gain insight into the hemispheric-, synoptic-, and mesoscale atmospheric processes important in the Antarctic and over the Southern Ocean. . . .” This paper is a contribution to the examination of the data collected and analyzed during FROST. There were three special observing periods (SOPs) in the project (July 1994, 16 October–15 November 1994, and January 1995; we shall refer to these as SOP-1, SOP-2, and SOP-3, respectively). Our aim here is to use the operational analyses of the European Centre for Medium-Range Weather Forecasts (ECMWF) during the SOPs to document the behavior of SH extratropical cyclones. We undertake this task with a completely objective and automated scheme that, from digital analyses of mean sea level pressure (MSLP), is designed to “find” and “track” cyclones. The scheme also compiles many statistics pertaining to cyclone behavior; here we will focus on the geographical distribution of the density of cyclones, as well as studying the regions of cyclone formation and decay. We will also comment on whether the SOPs can be thought of as typical of their time of year.
2. Data and techniques
We have based our investigation on the 0000 and 1200 UTC ECMWF analyses on a 2.5° × 2.5° latitude–longitude grid. Trenberth (1992) has performed a fairly recent evaluation of this dataset and documented changes that have occurred in the assimilation system that produced them. During the FROST SOPs considerable effort was made to ensure that the additional data collected were passed on via the Global Telecommunication System to operational centers for use in the construction of the analyses.
In this work the automatic scheme we implement finds and tracks cyclones from the digital analyses. The algorithm, based on that of Murray and Simmonds (1991a, hereafter MS), is an objective system, which has been applied to a variety of problems associated with synoptic behavior in extratropical regions (Murray and Simmonds 1991b, 1995; Simmonds and Wu 1993;Jones and Simmonds 1993, 1994; Simmonds and Law 1995; Godfred-Spenning and Simmonds 1996). The cyclone finding and tracking routines have undergone some development over that used in the above studies, and we outline these refinements below.
Previously the scheme had been applied to low- to medium-resolution analyses or model output and one could have expected the need for some modification when applied to high-resolution synoptic charts. Figure 1a shows the MSLP field for the first day of SOP-1 (0000 UTC 1 July 1994) from the ECMWF analyses; it is seen to possess considerable structure and we have indicated on the analysis points at which our original low finding system identifies cyclones. In the algorithm, the Laplacian of the MSLP, ∇2p, is taken as a measure of the strength of a low. In this study it is used not only to exclude insignificant depressions but to discriminate between two categories of depression. Systems with strength greater than 0.7 hPa (° lat)−2 are classified as strong systems and those with strength between this and 0.2 hPa (° lat)−2 weak. These classes are used in the tracking, which at successive times preferentially matches strong with strong and weak with weak. We have indicated on the analysis where the system has detected these lows. Solid symbols indicate closed depressions and hollow symbols open depressions (where closed and open systems are as defined by MS). In addition, circles denote the systems classified as strong and stars those classified as weak. It will be appreciated that not all the features identified have significance for the large-scale circulation.
To try and eliminate these features, we have applied a multipass smoother to the analyses. This consisted of performing corrections (on the polar stereographic grid) of Δp = (r2/4n)∇2p, where r is the diffusive radius (2° latitude), and the number of passes (n) is set to 7. The effect of this is similar to the Cressman smoother applied by Sinclair (1997) who used a pass radius of 5°. Figure 1b shows the analysis after it has been subject to the smoother and also displays the low pressure systems that have survived this procedure. In July 1994 the average number of systems per unsmoothed analysis identified by our scheme south of 10°S is 142, and this drops to 63 when the fields are smoothed as outlined above. Figure 1b shows there to be a number of small-scale features still present over regions of elevated topography and in areas where the concavity of the topography becomes large. Cyclones over elevated terrain are simply dealt with by ignoring any system identified at a location where the height of the topography exceeds 1 km. (This is done because of the concept of MSLP has limited meaning where the surface elevations exceed this amount.) Cyclones over concave topography are filtered out by applying a penalty to the pressure Laplacian before comparing it to the criterion for a cyclone to be regarded as a strong or weak depression. We require that, for a system to be classified as a cyclone, ∇2p − α|∇2zs| exceed the values specified above, where zs denotes the height of the topography. This is only applied as a final test; the actual finding and identifying cyclonic systems essentially follows the scheme described in MS and makes use of the pressure field only. Figure 1c shows the lows identified under these requirements (with α set to 0.005 hPa m−1). The distribution of systems now appears realistic, and an average of 40 per analysis are identified in July 1994. This value for α was chosen to remove the maximum number of spurious lows without also removing what were thought to be genuine systems. The investigations discussed below have been performed with these settings.
The tracking procedure is based on the projection of cyclone positions from one analysis time to the next and the comparisons of the projected positions with those of the cyclone analysis at the new time. In earlier applications of the scheme, the predicted displacements were based on a weighting of the previous movement of each cyclone and the climatological average cyclone displacement at the location. The climatological input has now been replaced by a “steering” component based on the idea of vorticity advection by the midtropospheric winds; this method gives displacements that are in accord with the instantaneous state of the atmosphere. The steering velocity is taken to be the surface geostrophic velocity averaged over the region of a low, multiplied by an appropriate factor, in the present case 2.0. [For more complete details see Simmonds et al. (1999).] This averaged geostrophic velocity is taken to be that implied by the pressure distribution around a circle of radius 5° latitude centered on the cyclone. The weighting adopted is 0.6 for the steering component and 0.4 on persistence of the speed and direction between the previous analyses. We finally comment that our present study only considers systems that last at least 24 h.
In order to set the mean synoptic background against which the fluctuations seen in our periods of interest occur it is important to be aware of the MSLP anomalies observed during these times. The anomalous pressure patterns for the three SOPs are presented in Fig. 2. These anomalies are taken from the ECMWF 0000 UTC analyses and are calculated as the deviations from the means for the period 1980–95 (the January calculation includes the data for 1996). Figure 2a presents the data for the month of July, Fig. 2c for January, while Fig. 2b exhibits the deviation in SOP-2 from the average of the October and November long-term means. In SOP-1 the dominant anomalous features are the anticyclone in the Amundsen Sea, and modest pressure decreases between 40° and 60°S. The pattern is very similar to that in the U.K. Meteorological Office analyses (Fig. 14 in Turner et al. 1996) except that the Amundsen Sea system is somewhat weaker. The anomalies in SOP-2 (Fig. 2b) exhibit large positive values in the Weddell, Bellingshausen, and Amundsen Seas as well as in the eastern Pacific. A strong anomalous low is found to the south of New Zealand. (We should mention that the nature of the MSLP anomalies for the entire months of October and November 1994 were very different from each other over much of the region of interest. For example, October exhibited a large anomalous anticyclone covering much of the eastern Pacific extratropics with a maximum value in excess of 16 hPa near 60°S; in November much of this region was under the influence of anomalous cyclonic circulation, with the monthly MSLP being up to 9 hPa below that month’s long-term mean. It would seem that SOP-2 covers a period during which the anomalous circulation changed considerably.) By contrast, in January 1995 (Fig. 2c) the region over New Zealand and to the south of Australia is host to anomalously high pressures, while pressures are lower almost everywhere south of 55°S.
To put the magnitude of these anomalies in context we need to quantify the climatological variability of the MSLP at high southern latitudes. Figure 3 shows the standard deviation of the monthly means of the MSLP for the month of July (Fig. 3a) and January (Fig. 3c), and Fig. 3b exhibits the mean of the October and November variabilities, all derived from the 16 years of ECMWF analyses. Each of these shows an increase of interannual variability with latitude, with (ignoring the maxima over Antarctica) the largest values being found in the region south of 60°S. There is also considerable longitudinal structure, with very large variability in the Bellingshausen and Amundsen Seas region and smaller values off Terre Adélie in the winter and spring. The variabilities in summer are smaller.
It is convenient to deal with the normalized MSLP anomalies (i.e., the anomalies divided by the standard deviations presented above) for the three SOPs as this ratio indicates the magnitude of the anomalies with respect to the background interannual variability. The distribution of this quantity for July 1994 is presented in Fig. 4a. In the plot we have stippled regions over which the magnitude of these normalized anomalies exceed unity (assuming normality, one would expect to observe deviations greater than 1.0 on about 16% of occasions, and values less than −1.0 with a similar frequency). The values exceed unity in the Bellingshausen and Amundsen Seas, but this is the only region south of 50°S where this occurs. The mean for the period 16 October–15 November has been normalized in this fashion (with the mean of the October and November variabilities) (Fig. 4b). Its magnitude exceeds unity in the Weddell, Bellingshausen, and Amundsen Seas as well as in the region 60°–90°E. Large positive anomalies are found in the eastern Pacific south of 30°S, and an area of negative values is situated over the western Pacific to the south of New Zealand.
The normalized anomalies appear strongest for SOP-3 (January 1995). This is accentuated to some extent by the more modest variability displayed in summer. Figure 4c shows that the normalized anomalies are less than −1.0 over most of the area south of 60°S (with the exception of the Weddell Sea), while positive anomalies exceeding this magnitude occur in a broad region from New Zealand to immediately south of Australia.
These plots demonstrate that the three observing periods have a level of anomalous behavior that is consistent with the large background variability observed in these high southern latitudes. It will be clear that great care must be exercised in interpreting the results as those typical of the month under consideration. With this background, we now examine the statistics of the cyclone behavior observed in the three SOPs.
The tracks of all cyclones identified by our automated algorithm in July 1994 are presented in Fig. 5a. The general structure is consistent with that identified over a longer sample by Jones and Simmonds (1993, hereafter JS), with systems being generated in the western part of the ocean basins and, for the most part, exhibiting an eastward and (more modest but still marked) southward motion. There is a concentration of tracks just to the north of the Antarctic continent, being particularly noticeable in the Indian Ocean. By contrast, the density of tracks in the Bellingshausen and Amundsen Seas area appear considerably less. Even though the sample size (of one month) is small, it is still useful to calculate the system density. The display of this (Fig. 5b) supports the impressions gained from the distribution of tracks. The density is at a maximum just off the Antarctic coast, with values exceeding 6 × 10−3 cyclones (° lat)−2 off Dronning Maud Land, to the south of Australia, and in the Ross Sea. These features are similar to those displayed in the JS climatology. By contrast, the number of cyclones to the west of the Antarctic Peninsula and upstream of the Drake Passage in SOP-1 is much less than observed in their climatology. This would appear consistent with the presence of the blocking anticyclone that was situated in the region during that period (Fig. 2a).
The distribution of cyclogenesis (Fig. 5c) shows a fairly noisy pattern as one would expect, but a number of features stand out. Considerable genesis is evident over the Indian Ocean south of 50°S, around New Zealand and into the central Pacific. It is seen that the Antarctic coastal regions are home to much cyclogenetic activity in this month, particularly from about 110°E around to the Peninsula. In Fig. 5d one can see that the main regions of cyclolysis tend to be displaced south of those for the cyclogenesis, and that the pattern shows more of a concentration on the Antarctic coastal region, as one would expect. While there is more longitudinal symmetry in the distribution of cyclolysis, regions such as the Indian Ocean and south of Australia and New Zealand appeared to be preferred terminating regions in SOP-1. The figure shows some interesting similarities as well as differences from the long term mean of JS (their Fig. 5b). For example, both show very high levels of cyclolysis immediately off the Antarctic coast. On the other hand, the region immediately to the north of Oates Land (72°S, 160°E) is found to be one of significant cyclolysis in July 1994, while the climatology of JS shows a minimum in that area.
We finally mention that the cyclone track behavior identified in Fig. 5a differs from that shown for the same month in Fig. 15 of Turner et al. (1996). There are a few reasons for this. First, the cyclone tracking scheme has been enhanced in a number of respects (as detailed in section 2) from that used in the Turner et al. study. Second, our present compilation is based on the ECMWF analyses, whereas the earlier work used the lower-resolution Australian Global Assimilation and Prediction System (GASP) analyses (Seaman et al. 1995). Perhaps of most importance is the fact that the GASP analyses showed major inconsistencies with analyses from three other operational centers in the first week of July 1994 (as documented by Lachlan-Cope et al. 1997).
The tracks recorded during the spring SOP shown in Fig. 6a tend to be somewhat south of their winter counterparts. This is consistent with the behavior of the semiannual oscillation (SAO) (see, e.g., Simmonds and Jones 1998). The southward shift is particularly evident in the eastern Pacific where very few tracks are seen north of about 55°S. This is concordant with the fact that the normalized anomalies exceeded +1.0 in much of this region (Fig. 4b). Similarly, few systems are present to the south of Western Australia, a broad area over which the pressure was seen to be above average. By contrast, an extensive region to the south of New Zealand experienced significantly lower pressures during the 30-day period and is seen to be a domain of organized cyclone behavior. These perceptions are supported by the distribution of system density shown in Fig. 6b. The large values are mostly confined to the region between 60°S and the coast with the largest values being found to the south of Australia and New Zealand.
Figure 6c identifies the area to the south of New Zealand as one of considerable cyclogenetic activity in this particular 30-day period, and Fig. 2b shows that this is a region of anomalously low pressure. The bulk of the cyclogenesis in this period is seen also to take place at high latitudes and particularly to the south of Australia. Very few cyclogenetic events are observed in the Bellingshausen Sea and off Princess Elizabeth Land; both of these localities experience considerably higher pressures during SOP-2. The pattern of cyclolysis (Fig. 6d) has substantial similarities to that in July 1994, but there are interesting differences. In this spring, termination of systems is more likely to occur in the Ross Sea and less likely in the region south of Australia than was found to be the case for winter. This is presumably associated with the fact that the spring cyclones in that region appear more mobile than their winter counterparts. The reader is reminded, however, that the background against which these cyclones are propagating shows large interannual variability, and comments of the sort made above should really be seen only as applicable to the particular years discussed.
The characteristics of the cyclonic systems during January 1995 (Fig. 7a) differ significantly from the winter and spring cases presented above. The tracks assume their most southerly positions and the density around the Antarctic continent is again high. We remarked above that the fact that the spring tracks were south of their winter counterparts was consistent with the behavior of the SAO. An association of this kind would imply that in summer the storm tracks should spread to the north. This is indeed what the climatology of JS shows (their Figs. 3 and 9), but the opposite is the case in SOP-3. It should be remembered, however, that the SAO shows considerable interannual variability (e.g., Simmonds and Jones 1996). In addition, it had been observed in the other SOPs that significant numbers of cyclone tracks start from the western side of the ocean basins. In January 1995 this appears to be true only in the Atlantic. Reference to Fig. 4c indicates that the midlatitudes to the east of Africa and Australia were regions where the normalized pressure anomalies exceeded +1.0 (and indeed +2.0 in an area around New Zealand, a condition that would be expected on only 2% of occasions). The small number of systems in the western Pacific is in marked contrast to those apparent in the climatology of Jones and Simmonds (1993); it is clear that January 1995 is atypical of this month. As we remarked earlier, most of the MSLPs south of 60°S were at least one standard deviation below the mean, and this would be expected to be associated with a greater concentration of cyclonic systems. The cyclone density (Fig. 7b) exhibits high values around the Antarctic coast and only modest numbers north of 60°S. Comparison with JS indicates that the strong summer climatological maximum observed off Wilkes Land was not present during January 1995. The sector from the Ross Sea to the Peninsula was host to the greatest number of systems.
Figure 7c, the plot of cyclogenesis density, is perhaps even more noisy than those of the other two seasons. There is a tendency for greater genesis off the Antarctic coast but it is not particularly striking. The cyclolysis distribution (Fig. 7d) shows more of a propensity for large values at those latitudes, and the number of systems that terminate in the Amundsen Sea in this SOP is particularly high.
In interpreting the cyclone behavior in the three SOPs and comparing such behavior against climatologies we must always bear in mind that these periods are very short. In addition to this, the behavior within a given epoch may be heavily influenced by large-scale, low-frequency indices of the general circulation. The number of cyclones appears to undergo long period variations over many parts of the globe. For example, Changnon et al. (1995) found that in the North American sector (60°–140°W) annual cyclone counts decreased from the early 1950s to the mid-1980s, before recovering and increasing into the mid-1990s. The work of Chen and Kuo (1994) indicates that during the period 1958–87 there was an increase in the frequency of cyclogenesis in the northwest Pacific and a decrease over the east Asian continent. The Southern Oscillation index (SOI) is a proxy for one example of low-frequency atmospheric oscillations. It is becoming clear that there is an SOI signal in many climate parameters in the Antarctic regions, including synoptic and cyclone behavior (e.g., Smith and Stearns 1993; Jones and Simmonds 1994; Simmonds and Jacka 1995; Gloersen 1995; Cullather et al. 1996). During the period of the SOPs the SOI was persistently negative (Fig. 3.16 of Houghton et al. 1996), and Cullather et al. (1996), in particular, have pointed to the complex relationships between cyclonic activity in the Amundsen Sea and the SOI.
Notwithstanding these comments, the characteristics of high-latitude cyclone numbers and motion appear consistent with those revealed in earlier studies and, in particular, with the climatology of Jones and Simmonds (1993). Of great interest is the fact that with these high quality analyses we have confirmed the high frequency of cyclonic systems around the Antarctic coast and reinforced findings that show the coastal region to be a domain of considerable cyclogenesis. It should be mentioned that this feature has not been found in all climatologies. Sinclair (1994, 1995) compiled a climatology of cyclones present in seven years (1980–86) of ECMWF analyses. In marked contrast to our results, he found a winter maxima of cyclogenesis in the 40–50°S belt and a minimum in the Bellingshausen Sea. Sinclair suggested that this overall picture was consistent with the satellite climatologies of the 1970s and with “general synoptic experience.” In a later study, Sinclair (1997) spoke of the assessment of these products with previous results obtained with “the old, time-honored manual approach”and he pointed out (Sinclair 1994) that the early studies revealed that most developing vortices occurred north of 55°S and that those “early results have withstood the test of time.”
That Sinclair has obtained such different results from ours may at first appear puzzling. He reminds us that the term cyclone is usually used to denote closed pressure centers while the method he used for identifying and tracking cyclones is based on determining maxima in the geostrophic vorticity field and opined that these centers correlate better with vortices seen in the satellite imagery, rather than pressure minima. Sinclair suggested that the use of a pressure minimum criterion tends to bias the collection to stronger and slower moving systems. (We remind the reader, however, that the scheme used here also finds “open depressions,” i.e., those that do not possess regions of closed isobars.) Sinclair chose to delete from consideration any system that moves a total distance less than 10° of latitude, with a view to eliminating cyclones caused by local orographic effects because “their contribution to the weather and climate is likely to be small.”
A recent comprehensive study by Turner et al. (1998) has cast considerable light on the issue. They developed a satellite-derived compilation of synoptic-scale low pressure systems within the Antarctic Peninsula sector of the circumpolar trough and found that more than half the systems identified there had their genesis within the region. The level of genesis they found in this coastal region actually exceeded that indicated by Jones and Simmonds (1993). However, when one takes into account the different resolutions and frequency of sampling of their respective datasets, the levels of cyclogenetic activity seem quite comparable. Turner et al. (1998) pointed out that the satellite imagery available to modern studies is far superior, particularly as a result of higher radiometric and horizontal resolution, to that available to researchers in the 1970s when a number of studies reached conclusions that are still held as truth today. Turner et al. used very high quality satellite data as well as drawing on the experience of weather forecasters in the Peninsula region; this “ground truthing” is seen as an important factor in clarifying the nature and role of synoptic systems in these high southern latitudes. They have also made the point that in the Antarctic zone weather systems have a much less clear frontal structure than at lower latitudes. Hence the interpretation of satellite imagery in the past may have been biased against the high southern latitudes.
Strong support for the concept of the Antarctic coast being a region for cyclogenesis in the SOPs and the climatology comes from a number of different sources. Mechoso (1980) showed that “the region around Antarctica far from being a place where all baroclinic processes are damped out by topographic slopes, is baroclinically very active . . .” and Kottmeier (1986) found that the relationships between wind and temperature could only be explained when the strong baroclinicity at the Antarctic coast was taken into account. Endorsement for the reality of the cyclogenesis around the coast of Antarctic has also come in the study of Berbery and Vera (1996). They showed that the winter 850–700-hPa Eady growth rates (their Fig. 1b) indicate strong baroclinicity in the environs of the Antarctic coast, particularly in the Bellingshausen Sea. It will be recalled that Sinclair (1994) questioned the reality of systems that exhibit little mobility. Berbery and Vera suggest that many of the coastal systems are “real,” and not artificially tied to the topography, because their locations are consistent with the structure of the baroclinicity field.
In the material given above we have presented statistics on all (i.e., both strong and weak) cyclones. One may ask whether the characteristics of cyclones that form in the general region of the subantarctic trough are different from those that come from farther north and, in particular, whether they show comparable strength. To address this we have formed from the catalog of cyclone tracks a subset for which the systems were classified as strong at at least one point on their track. The distribution of cyclogenesis of these for the three SOPs is shown in Fig. 8. The patterns are very similar to those when no such screening was performed (Figs. 5c, 6c, and 7c), and the magnitudes are only slightly smaller. There is certainly no suggestion that this screening has resulted in a greater reduction of cyclogenetic activity in the Antarctic coastal regions compared to the midlatitudes. (The greatest changes occur in connection with continental heat lows and to the north of Madagascar.)
We have mentioned that the analyses produced during FROST SOPs could be regarded as the best available for the time of year, given the special effort to include additional data south of 50°S during these periods. The quantification and understanding of cyclone behavior obviously improves with more data and higher-resolution analyses. Sinclair (1994) found that a sudden increase in the total number of cyclones present in the ECMWF analyses in 1983 coincided with the time that the ECMWF model was changed from a gridpoint to a spectral model with envelope topography (April 1983). He found that the increases were due mostly to systems that moved less distance than 20° latitude. Further analysis on this by Sinclair et al. (1997) indicated that the second half of the 1980–94 period had many more cyclones over the expanses of the Southern Ocean devoid of conventional data and that most of the increase was accounted for by these “new” systems. Their Fig. 5a indicates the greatest increases occurred at the high subantarctic latitudes, with the axis of greatest increase at about 60°S. These results indicate that the improvement in analysis techniques and data quantity and quality appear to be showing the immediate subantarctic latitudes to be an even more active region than might have been suggested by earlier studies. Cullather et al. (1996) performed an analysis of moisture transport onto the Antarctic continent using the ECMWF analysis. Their Fig. 4 suggests that the eddy component of the West Antarctica moisture convergence has increased significantly since 1980. While this increase may be real and part of the general variability at these latitudes, it could be suggested that the more recent analyses are representing the actual atmospheric structure more accurately and that the extra structure is important, at least for the meridional transport of moisture at the Antarctic coast.
5. Concluding remarks
We have applied an objective cyclone tracking algorithm to the digital analyses produced by the ECMWF during the three FROST special observing periods. These high-resolution analyses displayed considerable structure and it was thought desirable to apply a weak spatial smoother to them before the cyclone analysis was undertaken.
The overall behavior of cyclones in the SOPs was consistent with that identified in the climatology of Jones and Simmonds (1993); many systems were generated in the western part of the ocean basins and moved to the east and, to a lesser extent, to the south. In the three periods we found a concentration of tracks just to the north of the Antarctic continent, being particularly noticeable in the Indian Ocean. Notwithstanding these similarities, marked regional differences between the climatology and the SOPs are apparent. Clearly the 1-month periods considered are far too short for strong conclusions to be drawn. The deviations of the monthly MSLP during the SOPs from their long-term average were considerable during the SOPs, even after taking account of the high interannual variability in the high southern latitudes. This seemed marked in the Pacific and in the region to the south of Australia. To a great extent regions over which cyclone behavior shows considerable differences from climatology could be related to the MSLP anomalies in the month. In many respects, the three SOPs cannot be considered typical of their time of year, but one might ask whether any specific period could be so regarded.
Our analysis with these high quality analyses during the SOPs has confirmed the Antarctic coast as a region of high cyclone density and of very active cyclogenesis. This latter feature appears to be at variance with early studies that suggested the greatest (winter) cyclogenesis in the 40°–50°S belt. However, it is consistent with recent studies undertaken with high-resolution satellite data and four-dimensional data analyses, as well as the implications of baroclinic theory.
The authors are grateful for the support of the Antarctic Science Advisory Committee in undertaking this research.
Corresponding author address: Dr. Ian Simmonds, School of Earth Sciences, The University of Melbourne, Parkville, Victoria 3052, Australia.