Southern Hemisphere Winter Extratropical Cyclone Characteristics and Vertical Organization Observed with the ERA-40 Data in

Transcription

1 1JUNE 2007 L I M A N D S I MMONDS 2675 Southern Hemisphere Winter Extratropical Cyclone Characteristics and Vertical Organization Observed with the ERA-40 Data in EUN-PA LIM* AND IAN SIMMONDS School of Earth Sciences, The University of Melbourne, Melbourne, Victoria, Australia (Manuscript received 4 January 2006, in final form 11 September 2006) ABSTRACT The mean characteristics and trends of Southern Hemisphere (SH) winter extratropical cyclones occurring at six levels of the troposphere over the period have been investigated using the 40-yr ECMWF Re-Analysis (ERA-40) data. Cyclonic systems were identified with the Melbourne University cyclone finding and tracking scheme. This study shows that mean sea level pressure (MSLP) cyclones are more numerous, more intense, smaller, deeper, and slower moving than higher-level cyclones. The novel vertical tracing scheme devised for this research revealed that about 52% of SH winter MSLP cyclones have a vertically well organized structure, extending through to the 500-hPa level. About 80% of these vertically coherent SH cyclones keep their westward tilt until the surface cyclones reach their maximum depths, and the mean distance is 300 km between the surface and the 500-hPa level cyclone centers when the surface cyclones obtain their maturity. According to the authors definition of vertical organization, explosively developing cyclones are vertically very well organized systems, whose surface development is antecedent to their 500-hPa level counterpart. Over cyclones have increased in their system density, intensity, and translational velocity but decreased in their scale at almost all levels. However, some of the trends are not statistically significant. The proportion of vertically well organized systems in the entire population of SH winter extratropical cyclones has considerably increased over the last 23 yr, and the mean distance between the surface and the 500-hPalevel cyclone centers has decreased. Such changes in vertical organization of extratropical cyclones are statistically significant at the 95% confidence level. 1. Introduction * Current affiliation: Bureau of Meteorology Research Centre, Melbourne, Victoria, Australia. Corresponding author address: Dr. Eun-Pa Lim, Bureau of Meteorology Research Centre, GPO Box 1289K, Melbourne 3001, Victoria, Australia. e.lim@bom.gov.au Extratropical synoptic systems comprise daily weather patterns in the mid- and high latitudes. In particular, cyclones are often accompanied by rainfall and strong winds and can impact safety and the economy (Hennessy 2004). Moreover, they contribute to maintaining the atmospheric budgets of energy, momentum, and moisture by redistributing them across the globe and thus influencing the larger-scale circulation in which they are embedded. With this in mind, it is important to understand the responses of extratropical cyclones to climate change. There is a general agreement in the atmospheric research community that the atmosphere and ocean have experienced an unprecedented rate of increasing temperature since the late nineteenth century. Houghton et al. (2001) reported that the global atmospheric temperature has increased about 0.6 C over the last 100 yr. Also, a significant rate of warming was found through much of the world oceans with a volume mean warming of 0.06 C, and the warming in the top layer of m was about 0.31 C from 1948 to 1998 (Levitus et al. 2000). An increase in atmospheric carbon dioxide (CO 2 ) is considered to be the leading causative factor for such warming along with other greenhouse gases (Houghton et al. 2001; Folland et al. 2001; Karoly et al. 2003). According to observations at Mauna Loa in Hawaii, atmospheric CO 2 has increased up to ppmv in 2004 from ppmv in 1959 (data from the Carbon Dioxide Information Analysis Center, available online DOI: /JCLI American Meteorological Society

2 2676 J O U R N A L O F C L I M A T E VOLUME 20 at and Karoly et al. (2003) found that observed changes in North American temperature from 1950 to 1999 were consistent with simulations with increasing atmospheric greenhouse gases and sulfate aerosols. While greenhouse gas concentration and global temperature have been increasing, more intense cyclones and extreme events of precipitation and winds have been observed, particularly in the Southern Hemisphere (SH), which is the region of our interest in this study (Key and Chan 1999; Simmonds and Keay 2000a; Hougthon et al. 2001; Lim and Simmonds 2002). Key and Chan (1999) analyzed the National Centers for Environmental Prediction National Center for Atmospheric Research (NCEP NCAR) reanalysis 40-yr data (Kalnay et al. 1996) and showed a decreasing trend in the 1000-hPa level cyclone frequency but an increasing trend in their 500-hPa-level counterpart in the SH winter. Key and Chan suggested that the increasing trend of 500-hPa cyclone frequency could mean an increase in longer-lasting cutoff lows and a long-term change in the amplitude of the jet stream. Sinclair et al. (1997) detected an increase in the number of cyclones defined with vorticity maximum in compared to , using the 15-yr European Centre for Medium- Range Weather Forecasts (ECMWF) Re-Analysis data. They attributed this increase to the sensitivity of small-scale cyclones (defined with vorticity maximum) to the improved data coverage over the Southern Ocean, having considered that mean sea level pressure (MSLP) had only small changes in the same period. Simmonds and Keay (2000a), who utilized an updated version of the Melbourne University (MU) cyclone finding and tracking scheme (Simmonds et al. 1999), found that the annual mean number of MSLP cyclones had decreased over most of the SH except for the periphery of Antarctica since the early 1970s, and this negative trend was obvious in all seasons except summer. Although the decreases in cyclone numbers were similar in the S and S latitude bands (their Figs. 4b and 4c, respectively), Simmonds and Keay showed that their year-to-year changes were negatively correlated, implying that there is an interannual compensation of cyclone numbers between the mid- and high latitudes. However, MSLP cyclone intensity, radius, and depth all showed significant increases in the second-half period of the reanalysis (Simmonds and Keay 2000a). Likewise, trends to fewer but more intense, larger, and deeper MSLP cyclones were detected over the Southern Ocean in (Simmonds et al. 2003). Additionally, Keable et al. (2002) discovered that the physical properties of SH 500-hPa-level cyclones had similar changes in to their MSLP counterparts presented in Simmonds and Keay (2000a). On the other hand, the system density of 500-hPa-level cyclones decreased during the 1980s as did that of MSLP cyclones, but there was no clear change in the 1990s. Fyfe (2003), who analyzed extratropical MSLP cyclones with the NCEP NCAR dataset and an automated cyclone finding scheme, confirmed the decreasing number of cyclones in S and the increasing number poleward of 60 S. He found that very similar changes occurred in a climate simulation undertaken with conditions of increasing CO 2, raising the possibility of the role of human activity influencing the recent changes in SH cyclones. In this paper, we examine a new climatology and recent changes of SH cyclone features in various levels of the troposphere using the 40-yr ECMWF Re- Analysis (ERA-40) data, which were produced at a higher resolution of T159 and supplied with more observations and improvements from ECMWF 15-yr analysis. We also report on an algorithm we have devised to diagnose the vertical structure of extratropical systems with the rationale that the three-dimensional structure of cyclones influences greatly their surface development (Petterssen 1956; Sanders and Gyakum 1980). Hence, we will quantify the extent to which low pressure systems can be defined and traced up through the troposphere, and how the cyclone vertical structure has changed in In section 2, details of the datasets and our methodology are presented. Section 3 will consist of three parts: the climatology of SH winter cyclone features, mean features of cyclone vertical structure, and recent trends of cyclone properties and vertical organization. Finally, conclusions will be drawn in section Data and methodology The ERA-40 dataset was made in a sophisticated manner with data from different countries and various observational sources, making more use of satellite data (Uppala et al. 2005). The reanalysis data were completed through quality controlling and assimilating collected data with a frozen state-of-the-art threedimensional variational global data assimilation system. In addition, this reanalysis was performed at a horizontal resolution of about 125 km with 60 levels of vertical resolution. As comprehensive meteorological satellite data have been available since 1979, and the data quality and quantity were radically improved by its use, we focus our analysis on the period of Our data

3 1JUNE 2007 L I M A N D S I MMONDS 2677 included 6-hourly global analyses of MSLP, and 925-, 850-, 700-, 600-, 500-, and 300-hPa geopotential heights (hereafter, Zheight) of June August (JJA) Cyclones were detected with the Melbourne University cyclone finding and tracking scheme (hereafter MU cyclone scheme; Murray and Simmonds 1991; Simmonds and Murray 1999; Simmonds et al. 1999), which was developed according to the Lagrangian perspective as follows. First, the reanalysis data on a latitude longitude grid were interpolated by bicubic spline interpolation to a polar stereographic array centered on the South Pole. Then the finding routine started searching for local maxima of the Laplacian of the pressure or geopotential height (GPH) by comparing it to those at the neighboring four grid points. From these points the associated pressure (or GPH) minimum was found by ellipsoidal minimization techniques. Each cyclone found was categorized as open or closed according to whether or not it had a closed contour around its pressure (or GPH) minimum. To select only meteorologically significant systems the lows had to satisfy a concavity criterion, which was 0.2 hpa (1.7 m for GPH systems) over 2 ( latitude) 2 (which was set by experimentation) over the cyclone centers (Simmonds et al. 1999). This process was applied to identify cyclones at each of the selected levels up to the 500-hPa level. These cyclones were then tracked by an algorithm that made an estimate of the new position of each cyclone, calculated the probability of associations between the predicted and actual new positions by estimating a decreasing function of the separation and central pressure differences between the two positions, and finally, found the most probable combination of associations in groups where each system belonged. Systems that did not last at least 1 day were not considered in the analysis (Murray and Simmonds 1991). From the cyclone tracks, statistics were obtained over S for cyclone system density, cyclone intensity [as measured by 2 P (or 2 Z) in the vicinity of the center of the cyclone], cyclone radius (R, taken as the weighted mean distance from the cyclone center to the points at which 2 P or 2 Z is zero around the edge of a cyclone), and depth [D, which represents overall influence of a cyclone including its intensity and scale (Simmonds et al. 1999)], which is determined by D P or Z R 2. Figure 1 displays these main physical features of a cyclone and our method to determine mean radius of a cyclone. 1 FIG. 1. Illustration of (a) cyclone parameters and (b) their determination. To define the vertical structure of a cyclone, we devised a numerical scheme to trace low pressure and GPH centers vertically as follows. First, we found the location and time of a surface cyclone when it reached its maximum depth. In a circle of 4 latitude radius ( 444 km) centered on the location at that time, we searched for the vertical extension of the cyclone at the next data level (444 km was chosen based on the consideration of the mean cyclone radius in the extratropics in both hemispheres and the tests with synoptic charts in the six vertical levels). If a partner cyclone was found, this tracing was continued to the 500-hPa level. This vertical tracing method is illustrated in Fig Results a. A climatology of baroclinicity and cyclone features in Before examining the mean characteristics of cyclone properties, it is useful to review some fundamental concepts of baroclinicity, which is known as a major driving force for the development of extratropical low pressure

4 2678 J O U R N A L O F C L I M A T E VOLUME 20 FIG. 2. Schematic diagram of our vertical tracing method: (a) vertically well organized case and (b) an example of a shallow case finishing its vertical identification at the 850-hPa level. The square represents the location of a surface cyclone when it reaches its maximum depth, and the asterisks represent the locations of low GPH centers. systems. Although other processes such as a horizontal barotropic shear and diabatic heating are important for a full understanding of cyclone development (James 1987; Emanuel et al. 1987; Hoskins 1990; Nakamura 1992), baroclinic instability accounts for the characteristics of extratropical cyclone development reasonably well (Berbery and Vera 1996; Kageyama and Valdes 2000). Here, we examine baroclinicity using the measures proposed in Charney (1947) and Eady (1949). First, Charney s model includes the beta effect, thereby, the interior potential vorticity gradient of the fluid is nonzero. Equation (2) defines the baroclinic parameter (BP), a measure of baroclinicity used in the study of Green (1960), BP f 2 2 U z fm 2 HN HN h 2 H, 2 where f is the Coriolis parameter, U z is the vertical shear, (du/dz) (where U is the eastward component of the wind and z is a vertical distance), is the latitudinal variation of the Coriolis parameter, H is the density scale height, N is the Brunt Väisälä frequency, and M 2 is a measure of the meridional gradient of potential temperature (hereafter, for simplicity referred to as the meridional temperature gradient ) defined as ( gd / dy) (where is potential temperature and y is a meridional distance). Here, the meridional temperature gradient is also equal to fu z when the zonal thermal wind balance is obtained and h is the vertical scale of a disturbance, ( f 2 U z /N 2 ). According to Pedlosky (1987), when N is set to be constant, if the vertical shear in the environment is small (so that the effect becomes relatively more significant, i.e., BP 1), the vertical growth of a disturbance is small, and the horizontal scale of it is also small. By contrast, if the vertical shear is strong enough to dominate the effect (BP 1), the vertical and horizontal growth of a disturbance can be large. The maximum Eady growth rate is also widely used as a measure of baroclinicity. Although Eady s model ignores the effects of the earth s sphericity and makes the assumption of a rigid upper boundary, previous studies have shown that the basic dynamics of baroclinic instability and the predictions of the features of baroclinic waves are well represented by Eady s model (Pedlosky 1987; Lindzen 1993; Berbery and Vera 1996). The maximum Eady growth rate is formulated as Eady 0.31 fu 2 z 0.31M N N. 3 Figure 3 presents the winter average BP and Eady calculated at the 850-hPa level (Figs. 3a,b) and of the atmospheric variables comprising them, namely, the meridional temperature gradient ( M 2 ) (proportional to vertical shear; Fig. 3c) and the square of the Brunt Väisälä frequency (N 2 ) between the 700- and 1000-hPa levels as a measure of the static stability of the environment (Fig. 3d). Greater BP is found in higher latitudes in general (Fig. 3a). In Eady strong baroclinicity is less concentrated over the Circumpolar Trough (CPT) and detected over the spiral arm stretching from the Atlantic Ocean to the Drake Passage (Fig. 3b). As diagnosed by both BP and Eady the Atlantic and Indian Oceans tend to host stronger baroclinicity than the Pacific Ocean despite the presence of the intense subtropical jet over the Pacific Ocean in the 30 S latitude band in winter. The strong vertical shear associated with the subtropical jet over the Pacific Ocean tends to be offset by the weaker Coriolis effect [smaller f and larger in Eqs. (2) and (3)] than the Coriolis effect in

5 1JUNE 2007 L I M A N D S I MMONDS 2679 FIG SH winter averages of (a) the BP, (b) Eady, (c) M 2, and (d) N 2 at the 850-hPa level. The contour interval is (a) 0.5, (b) 0.2 day 1 (c) s 2, and (d) s 2. The blank section poleward of 80 S in (a) is the screened area due to extreme values over the Pole. the high latitudes hosting the subpolar jet. Also, Nakamura and Shimpo (2004) showed that the baroclinicity associated with the subpolar jet mainly over the Atlantic and Indian Oceans was strong through the mid- and lower troposphere, whereas the baroclinicity below the subtropical jet over the Pacific Ocean in the S latitudes became considerably weaker at lower heights. In Fig. 3c the strongest meridional temperature gradient at the 850-hPa level is seen at the Antarctic coast and over the eastern Atlantic Ocean and the Indian Ocean in the S latitude band. A local maximum is over the Pacific Ocean in the midlatitudes, reflecting the strong vertical shear associated with the subtropical jet. In regard to the static stability, the atmosphere is vertically the least stable in the Southern Ocean in the entrance of the Drake Passage and off the sea ice edge in the East Antarctic section ( E), in part due to the katabatic winds from the high elevation of East Antarctica (Fig. 3d). As a result of applying the MU cyclone scheme to the six levels of our consideration, using the same selection criterion for cyclones at each level, we obtained the averages of cyclone frequency and physical properties over the extratropics between 25 and 90 S (Table 1). For comparisons of cyclone features between MSLP and other GPH fields, MSLP cyclone intensity and

6 2680 J O U R N A L O F C L I M A T E VOLUME 20 Level hpa TABLE 1. The 23-yr time means of SH winter cyclone properties. System density Intensity Radius Depth Translational velocity No. of systems in 10 3 ( lat) 2 m( lat) 2 lat m m s MSL * * 9.9 * The converted values from pressure to GPH by the hydrostatic equation. depth in hectopascals were converted to those in meters by the hydrostatic equation. According to Table 1, more cyclones are detected at the mean sea level (MSL) and in the midtroposphere than in the lower troposphere. Note that averages of cyclone system density and intensity decrease from the surface to the 700- hpa level, but increase thereafter. Such minima of cyclone frequency and intensity at the 700- and 850-hPa levels seem to be related to the baroclinicity minimum in those levels shown in Trenberth (1991), Berbery and Vera (1996), and Nakamura and Shimpo (2004, their Figs. 3d f). Also, the development of cyclones at the 700- and 850-hPa levels is less influenced by strong vortices along well-developed upper-level troughs and by the lower-level orographical effects. The cyclone scale is smallest at the surface and largest at the 700-hPa level, showing an opposite vertical pattern to that of system density and intensity. Cyclone depth, a combined product of cyclone intensity and scale, increases as the height increases. Cyclones tend to move faster in the lower troposphere from the surface to the 850-hPa level than in the midtroposphere of the 700- to 500-hPa level. The climatology of MSLP cyclone properties averaged for the 23 yr of JJA is presented in Fig. 4. As described in previous studies (Jones and Simmonds 1993; Sinclair 1994; Simmonds and Murray 1999; Simmonds and Keay 2000b; Wernli and Schwierz 2006), the maximum of cyclone system density is found along the CPT particularly off East Antarctica (Fig. 4a). Also, the distribution of cyclone system density exhibits spiral arms that stretch from the east of Australia and Uruguay. Inatsu and Hoskins (2004) investigated the sensitivity of winter storm track to sea surface temperature (SST) and orography in their general circulation model and suggested that the spiral structure of the lowerlevel cyclone activity was associated with the stronger SST gradients over the Atlantic and Indian Oceans and the topography of South America and South Africa. In winter, cyclone activity rapidly increases in the midlatitudes in the Pacific Ocean, starting from the Tasman Sea with the development of the winter subtropical jet. However, the cyclone system density associated with the subtropical jet is not as high as that associated with the subpolar jet although the subtropical jet is stronger. This seems to be due to the moderate baroclinicity there in the lower troposphere as we mentioned earlier. Furthermore, Nakamura (1992) and Nakamura and Shimpo (2004) pointed out that very strong westerlies could actually act against eddy growth at the mid- and upper troposphere, driving eddies rapidly away from the baroclinic zone. Nakamura and Shimpo (2004) showed that, in the presence of the strong westerlies, the coherence tended to be lower between subweekly fluctuations in temperature and the meridional or vertical wind component, leading to the less efficient energy conversion from the mean flow to the eddies. On the other hand, there is a split of system density south of the Tasman New Zealand (NZ) sector related to the double jet structure in the upper troposphere (Taljaard 1972; Trenberth 1991; Sinclair 1994; Hurrell et al. 1998; Simmonds and Keay 2000b). More intense, larger, and deeper cyclones tend to occur in the higher latitudes (Figs. 4b d). While the Pacific Ocean and the Tasman Sea are more favored for relatively more intense, larger, and deeper cyclones than other regions in the midlatitudes of S, the Pacific cyclones are less intense, smaller, and less deep in the high latitudes. The Drake Passage, the Antarctic Peninsula, and NZ regions are also hosts to local minima of cyclone intensity, radius, and depth. In regard to larger cyclones in the high latitudes with strong baroclinicity, Welch and Tung (1998) and Kageyama and Valdes (2000) showed that longer waves would be favored by strong baroclinicity as they can carry more heat. Although the maxima of system density are found in the latitude band between 55 and 65 S, the maxima of cyclone intensity are shown in the 50 S latitude band in the Atlantic and Indian Oceans. These locations of intense systems seem compatible with those of the maximum frequency of cyclones selected with vorticity maxima in the works of Sinclair (1994, 1995, 1997). Figure 4e shows that cyclones over the Atlantic and Indian Oceans move most rapidly, and this finding is consistent with the results of Hoskins and Hodges (2005). Furthermore, the distribution of translational velocity has similarities with that of M 2 (which is proportional to vertical shear) shown in Fig. 3c, as would be expected. Finally, it is interesting to note that the overall geographical patterns of cyclone system density, inten-

7 1JUNE 2007 L I M A N D S I MMONDS 2681 FIG. 4. JJA averages of MSLP cyclone properties: (a) system density, (b) intensity, (c) radius, (d) depth, and (e) translational velocity. The contour interval is (a) ( lat) 2, (b) 0.2 hpa ( lat) 2, (c) 0.5 lat, (d) 2 hpa, and (e) 3 m s 1.

8 2682 J O U R N A L O F C L I M A T E VOLUME 20 sity, and depth appear to be more compatible with those of the winter baroclinicity measured with the BP (Fig. 3a) rather than Eady (Fig. 3b) in that they are zonally symmetrical and increase with latitude. However, the features associated with the development of the subtropical jet across the Pacific Ocean and the spatial pattern of cyclone translational velocity seem better explained with Eady. b. Mean characteristics of vertical structure of extratropical cyclones The importance of the interaction with upper-level systems in the development of surface cyclones, in particular, explosively developing cyclones, has been addressed by a number of researchers. Petterssen (1956) observed that in most cases cyclone development tended to commence either in a preexisting upper trough with strong vorticity or in a region of strong cyclonic shear poleward of the axis of a strong zonal current. These cyclones have been known as type B. Sanders and Gyakum (1980) showed that Z500 cyclones tended to catch up to the surface cyclones from the west when the surface systems developed rapidly. Sanders (1986) found that in the western North Atlantic Ocean rapidly deepening cyclones occurred when a 500-hPa absolute vorticity maximum, which had great initial separation distance, approached with a great speed. This 500-hPa absolute vorticity maximum passed by the surface cyclone center during the period of 24 h before maximum deepening and 24 h afterward. Sanders suggested that the explosive maritime cyclones were basically baroclinic disturbances in which the low-level response to a given upper-level forcing was large. Gyakum et al. (1992) added that the 500-hPa upper-level forcing was conditioned by the lower-level antecedent vorticity development. With this background, we now go on to present a climatology of the vertical features of cyclones, based on the method explained in detail in section 2. We define a vertically well organized system as a cyclone whose cyclonic identification can be traced all the way to the 500-hPa level. On average, surface cyclone features are distinctly different according to whether or not surface cyclones can be traced all the way up to the 500-hPa level. We compare the surface cyclone properties when the surface cyclones have a vertically well organized structure connecting to the 500-hPa level and the counterparts when cyclones lose their cyclonic identification below the 700-hPa level. Table 2 shows that the winter surface cyclones having a Z500 cyclone partner are more intense, larger, and deeper than those ending their connection below the 700-hPa level. Moreover, the MSLP Z500 TABLE 2. Comparison of the properties of SH winter MSLP cyclones when the MSLP cyclones have well-organized structure up to the 500-hPa level (MSL 500) and their counterparts that end their connection at the 850-hPa level or lower levels (MSL 850). MSL 500 MSL 850 Intensity [hpa ( lat) 2 ] Radius ( lat) Depth (hpa) Life span (days) Speed (m s 1 ) coupled cyclones have about twice the life span. The mean translational speed of surface cyclones is slightly slower in the case of vertically better organized cyclones than shallow structured systems at the maximum development of surface cyclones. One reason for the slower movement of vertically well developed cyclones could be found from the fact that these vertically well developed cyclones include cutoff lows that become nearly barotropic at their maturity and do have low translational speed. Furthermore, shallow structured cyclones tend to have more zonally symmetrical steering winds than their vertically well developed counterparts that have stronger cyclonic circulation up to the 500-hPa level often considered as a steering level for surface systems (Gill 1982; Murray and Simmonds 1995; Simmonds et al. 1999). When surface cyclones experience their deepest stages, the winter average distance between Z500 and MSLP cyclone centers is about 300 km. To determine whether there are preferred positions for Z500 cyclones to interact with surface cyclones, we counted the number of couples of surface and Z500 cyclones in each 15 latitude longitude box and distributed them according to the direction of the Z500 cyclone relative to the surface feature. The directions were distributed into the eight points of the compass (i.e., 45 sectors). For convenience, we call these directions, clockwise, NNE, ENE, ESE, SSE, SSW, WSW, WNW, and NNW hereafter (Fig. 5). Then the number in each direction was normalized by the total number of cyclones in the 15 latitude longitude area. With this normalization we can obtain the proportion of MSLP Z500 paired cyclones to the entire population of extratropical cyclones and exclude a bias caused by decreasing longitudinal distance toward the Pole. In Figs. 5 and 6 the arrows indicate the directions, and their length represents the ratio of the number of pairs in the given direction to the entire MSLP cyclone population in the area. According to our investigation, about 80% of MSLP Z500 winter cyclone pairs have westward tilts at the time when the surface cyclones reach their maximum

9 1JUNE 2007 L I M A N D S I MMONDS 2683 FIG. 5. Diagram of the method to determine the relative direction of Z500 cyclones to the MSLP cyclones. depths. About 43% of SH MSLP cyclones are linked with Z500 cyclones positioned southwestward (WSW and SSW). Figure 6 demonstrates the geographical distribution of the relative directions of Z500 cyclones to the surface cyclones. A number of cyclones tend to have a strong component of northward tilt in the S latitude band from the eastern Indian Ocean to the western Pacific Ocean, whereas the cyclones over the entire Atlantic Ocean and the Pacific Ocean in the S latitude band are coupled with southwardpositioned Z500 cyclones. The cyclones appearing off the southern coast of Australia are also seen to be actively interacting with Z500 cyclones to the WNW and WSW in the cold season. This feature is likely contributed by the development of the subtropical jet, which amplifies across Australia and the western Pacific Ocean on the 30 S latitude band. In comparison, most of the surface cyclones occurring off the Antarctic coasts have the WSW tilt toward their Z500 cyclone partners. WSW tilts are particularly dominant for the cyclones occurring near the coast of East Antarctica ( E), the Ross Sea, and the Bellingshausen Sea. To explore further connections between Z500 and MSLP cyclones, we have determined the distribution of the relative timing of the maximum depth of cyclones at the 500-hPa level to the maximum depth of cyclones at MSL. In Fig. 7 the peak of the distribution is found at zero lag [i.e., when the timing of the maximum depth of the surface feature is almost synchronous (to within 6 h) with that of the Z500 cyclone]. This accounts for 9% of the cyclones having Z500 partners in the SH winter. Except for these synchronously deepening cyclone couples the histogram shows that slightly more of SH surface cyclones experience the deepest stages after the Z500 ones do (surface before Z500 versus surface after Z500: 39% versus 52%), suggesting that upper-level processes play a relatively more important role in the SH cyclone development than lower-level processes. These vertically well organized cyclones account for 52% of the total number of SH winter surface extratropical cyclones. Figure 8 presents the distribution of the percentage of system density of MSLP cyclones that are vertically well organized in JJA. The plot shows that the systems having well-organized vertical structure account for more than 40% of the entire population of extratropical cyclones over the SH extratropics. In fact, over most of the extratropics the percentages are greater than 52% (the mean percentage of the number of vertically well organized systems) because the system density is contributed by the whole length of a cyclone track. And it is shown in Table 2 that vertically FIG. 6. Relative positions of Z500 cyclones to their partners at the MSL in JJA. The arrows point to 45 sectors of the compass, and the length of the arrows represents the number of pairs in each direction in a 15 lat lon, normalized by the total number of cyclones occurring in the area.

10 2684 J O U R N A L O F C L I M A T E VOLUME 20 FIG. 7. Winter average of relative timing of Z500 cyclone maximum depth to MSLP cyclone maximum depth. Negative time lag shows that surface cyclones reach their maximum depths earlier than the Z500 counterparts, while positive time lag indicates the opposite. Each bar represents a 6-h interval. well organized cyclones tend to have a longer life span. Consequently, it is reasonable to see higher ratios in Fig. 8 than the average percentage. The Great Australian Bight is host to the highest frequency of vertically well organized cyclones (more than 80% of the entire population), and the Tasman Sea NZ and the Atlantic and Indian Oceans are favored for vertically well developed cyclones. The maximum over the Atlantic and Indian Oceans over S is located over the most favored places for Z500 cyclones to reach their maximum depths (not shown), which agrees with a strong preference for Z500 cyclones to deepen before their surface cyclone partners in the SH shown in Fig. 7. Next, we examine the properties of vertical structure of explosive cyclones (often called meteorological bombs ). In our analysis explosive cyclones were defined as cyclones whose relative central pressure (actual central pressure minus the climatological pressure in the region) drops at least 24 hpa in 24 h at the 60 latitude (Sanders and Gyakum 1980; Lim and Simmonds 2002). This was expressed by the normalized deepening rate, NDR r 1 (Lim and Simmonds 2002), where NDR r P r 24h sin60 sin, where P r is the change of the relative central pressure of a system over 24 h, and is the latitude of the 4 system. The use of relative central pressure was adopted to exclude cyclones that obtain a rapid development rate due to the strong meridional climatological pressure gradient in the SH. Using the NCEP Department of Energy Reanalysis II (NCEP2) data (Kanamitsu et al. 2002) of along with the MU cyclone scheme, Lim and Simmonds (2002) found that the SH had about 11 explosive cyclones on average in JJA, and the east coast of Australia and the Southern Ocean south of Australia were identified as regions of frequent occurrence. In regard to the vertical structure of explosive cyclones detected with the ERA-40 data, about 76% of SH winter explosive cyclones (whose number averages 10.4 per winter) are found to have vertically well organized structure reaching to the 500-hPa level at their maximum depths. The mean distance between the MSLP and Z500 winter cyclones is 321 km when the surface explosive cyclones reach their maximum depths. This distance is 21 km greater than that of all extratropical cyclones. The greater distance tilt between surface explosive cyclones and their 500-hPa level cyclones seems to confirm the importance of strong baroclinicity, one of the main mechanisms driving the explosive development (Sanders and Gyakum 1980; Sanders 1986; Huo et al. 1996; Yoshida and Asuma 2004). According to the baroclinic instability theory (Holton 1992), an important factor for the baroclinic energy to be released is the phase tilt with height for which the scale of waves needs to be considered. In our investigation SH winter explosive cyclones have about 19 km smaller radius on average than the mean radius of extratropical cyclones in the SH JJA, and this would increase the angle of the axis of vertical tilt. Therefore, in this particular case the greater distance tilt of explosive cyclones could mean the greater phase tilt, which implies stronger baroclinicity for the rapid development of these cyclones. In addition, about 10% of the surface explosive cyclones and their 500-hPa level cyclone partners appear to attain their maximum depths almost at the same time. For the rest of the explosive cyclones, surface systems tend to reach their maximum depths before their Z500 cyclone partners do (surface before Z500 versus surface after Z500: 71% versus 19%). This preference toward antecedent surface cyclone development implies the importance of surface processes such as sensible and latent heat fluxes, positive anomalies of potential vorticity due to latent heat release, and lower static stability in the development of rapidly developing cyclones. These statistics also support the conclusions of Sanders (1986) and Gyakum et al. (1992) emphasizing the importance of lower-level response in the de-

11 1JUNE 2007 L I M A N D S I MMONDS 2685 FIG. 8. Percentage of vertically well organized MSLP cyclone system density over the extratropics in JJA. The contour interval is 10%. velopment of explosive cyclones. In terms of the direction of tilt from the surface cyclones to the Z500 counterparts, almost all (92%) of the explosive cyclone couples have westward tilt, and 63% have southwest tilt (WSW and SSW) at the time of maximum depths of surface explosive cyclones. Therefore, it can be concluded that explosive cyclones are the systems having a vertically very well organized structure. c. Trends As mentioned in the introduction, the studies using 40-yr NCEP NCAR reanalysis data suggested a statistically significant reduction in MSLP cyclone frequency (Key and Chan 1999; Simmonds and Keay 2000a; Fyfe 2003). According to Lim (2005) NCEP2 data from also showed a decreasing trend in MSLP cyclone frequency. In regard to using the ERA-40 dataset for examining a signal of climate change, Bengtsson et al. (2004) investigated trends in global mean temperature in the lower troposphere, integrated water vapor, and kinetic energy. Their work showed that there have been positive trends in the lowertroposphere temperature and integrated water vapor but no significant change in kinetic energy in In the intercomparison of storm tracks using ECMWF, NCEP NCAR, NCEP2, and National Aeronautics and Space Administration (NASA) Geostationary Operational Environmental Satellite 1 (GOES-1) reanalysis data, Hodges et al. (2003) found that there was an overall good agreement in cyclone track density and cyclone intensity in the NH lower troposphere whereas the agreement was less consistent in the SH lower troposphere. As no work has been undertaken with respect to changes in SH cyclone properties with the ERA-40 dataset, here we analyze trends of cyclone features in the ERA-40 dataset, comparing them to the counterparts detected in the NCEP2 dataset (Tables 3 and 4). According to Table 3, cyclones have increased in their number from the surface to the 500-hPa level, and the statistically significant increases are found from the 850- to 500-hPa levels. It is shown that the increase of

12 2686 J O U R N A L O F C L I M A T E VOLUME 20 TABLE 3. Trends in ERA-40 winter cyclone properties in The units are the same as in Table 1, but per decade. The bold numbers are the trends statistically significant at the 95% CL. Level System density Cyclogenesis Life span Intensity Radius Depth Translational velocity MSL * * 0.06 * The converted values from pressure to GPH by the hydrostatic equation. cyclone system density at the surface and the lower troposphere seems to be influenced by the increase in cyclone life span whereas the increase of system density at the 700- to 500-hPa levels is due to the increase in the frequency of cyclone generation. The absence of a statistically significant trend in MSLP cyclone frequency in the ERA-40 data is different from the statistically significant negative change shown in Table 4 with the NCEP2 data and in the previous studies that used the NCEP NCAR data (Key and Chan 1999; Simmonds and Keay 2000a; Fyfe 2003; Pezza and Ambrizzi 2003). Figure 9a presents the time series of winter MSLP cyclone system density obtained from the ERA-40 and NCEP2 datasets. It is interesting to note that there is a better agreement between them with time although they show opposite trends over the entire period. However, cyclones occurring in the other levels exhibit positive trends in both datasets, and the increase in cyclone frequency is more significant in the ERA-40 dataset (Table 4 and Fig. 9b). Table 3 also shows that intensity has statistically significantly increased from the surface to the 700-hPa level [at the 95% confidence level (hereafter, CL)]. The trends in Z600 and Z500 cyclone intensity are positive as well and are at least as large, but they do not achieve TABLE 4. Trends in NCEP2 winter cyclone properties in The units are the same as in Table 1 but per decade. The bold numbers are the trends statistically significant at the 95% CL. Level System density Intensity Radius Depth Translational velocity MSL * * 0.1 * The converted values from pressure to GPH by the hydrostatic equation. statistical significance because of larger temporal variances. On the other hand, cyclones have been smaller over , and as a result, cyclone depth does not show statistically significant changes in most levels. Last, there is no significant change detected in cyclone translational velocity. In comparison, cyclones have been more intense, larger, deeper, and faster moving in almost all levels up to the 500-hPa level in in NCEP2 (Table 4). In particular, the positive trends in cyclone intensity and depth are statistically significant in all levels from the surface to the 500-hPa level. In regard to recent changes in the characteristics of cyclone vertical structure, Fig. 10 shows that the percentage of SH vertically well organized cyclones increased from 1979 to 2001 at a rate of 2.2% decade 1 (statistically significant at the 95% CL). The increase of these kinds of cyclones seems to be correlated with the increase of Z500 cyclone system density with the correlation coefficient 0.45 at the 95% CL. The other interesting change in cyclone vertical structure is that the mean distance between Z500 and MSLP cyclone centers when the MSLP cyclones reach their maximum depths reduced by 12.4 km decade 1 over the last 23 yr (99% CL). It is noteworthy that NCEP2 data also suggest an enhancement in cyclone vertical consistency in the same period, showing a 2.6% increase decade 1 in the proportion of vertically well organized cyclones and 21.5 km decade 1 decrease in the mean distance between MSLP and Z500 cyclones at the maturity of MSLP cyclones. Further details of cyclone features observed in the two reanalyses can be found in the appendix. Finally, explosive cyclones detected with the ERA-40 data do not have any considerable change in their frequency and vertical organization in Discussion and conclusions We have assembled a new climatology of SH winter extratropical cyclones in , using the ERA-40

13 1JUNE 2007 L I M A N D S I MMONDS 2687 FIG. 9. Time series of cyclone system density detected from the ERA-40 and NCEP2 datasets in JJA (a) MSLP cyclones and (b) Z500 cyclones. reanalysis data. Our analyses have shown that MSLP cyclones have greater intensity but a smaller scale than higher level systems. Geographically, MSLP cyclone frequency, intensity, radius, and depth tend to increase as the latitude increases particularly in the Southern Ocean off East Antarctica due to the strong baroclinicity there. We have presented an automatic and simple method for diagnosing the vertical structure of cyclones. Application of this shows that about 52% of SH winter extratropical cyclones are vertically well organized, reaching up to the 500-hPa level. The mean distance between surface cyclones and their Z500 cyclone partners is about 300 km when the surface cyclones reach their FIG. 10. Time series of the proportion of vertically well organized cyclones to all extratropical cyclones in the SH winter in

14 2688 J O U R N A L O F C L I M A T E VOLUME 20 maximum maturity. In addition, SH surface cyclones tend to reach their maximum depth after their Z500 cyclone partners. Based on these mean characteristics of all SH extratropical cyclone vertical structure, explosively developing cyclones (called meteorological bombs) are vertically very well organized systems, whose lower-level processes are more important. For the 23 yr from 1979 to 2001, cyclones tend to have increased in their system density, intensity, and translational velocity but decreased in their scale at almost all levels, but some of the trends are not statistically significant at the 95% CL. In regard to recent changes in the vertical structure of SH winter extratropical cyclones, the proportion of the vertically well organized cyclones has statistically significantly increased in Given that the surface systems having a Z500 cyclone partner are more intense, larger, and deeper, the trend to more vertical organization of cyclones might be reflected in the positive trends in the mean intensity, radius, depth, and system density of SH surface cyclones shown in Table 3. Another very interesting change is that the average distance between surface and their Z500 cyclones has statistically significantly decreased in this period, which might suggest that cyclones have become more barotropic (or symmetric) as they fully develop. This result leaves a few intriguing questions for future study, including whether this implies a positive trend toward warm-cored extratropical cyclones (e.g., by extratropical transition and the development of warm seclusions within extratropical cyclones), which was reported as more intense systems than the cyclones remaining cold-cored through their lifetimes (Hart 2003), and whether this more symmetrical structure affects the intensity and longevity of fronts and rainfall accompanied with them, which are characterized with asymmetric extratropical cyclones. Last, we have seen that there are some discrepancies in the trends of winter cyclone frequency, scale, and depth between the ERA-40 and NCEP2 datasets. However, both datasets show that there have been increases in cyclone intensity and the proportion of vertically well organized cyclones in the entire population of SH extratropical cyclones. To understand better recent changes in the weather systems presented in this study further investigation for the intercomparison of largescale atmospheric features affecting extratropical cyclones between different reanalyses needs to be followed. Acknowledgments. Parts of this research were made possible by grants from the Australian Research Council, the Antarctic Science Advisory Committee, and the David Lachlan Hay Memorial Fund at the University of Level hpa TABLE A1. The 23-yr time means of SH winter cyclone properties detected with the NCEP2 data in System density Intensity Radius Depth Melbourne. The authors thank ECMWF and NCEP for providing the data. We are grateful for useful discussion over cyclone results with Dr. Ross Murray and the data preparation and other assistance from Mr. Kevin Keay. Last, two anonymous reviewers are greatly appreciated for their constructive and thorough review, which enabled us to improve the quality of this work. APPENDIX Translational velocity No. of systems in 10 3 ( lat) 2 m( lat) 2 lat m m s MSL * * 9.9 * The converted values from pressure to GPH by the hydrostatic equation. The Mean Characteristics of SH Winter Cyclones Detected in the NCEP2 Reanalysis Since the ERA-40 data became freely available to the atmospheric research community, there has been an increasing interest in the comparisons of atmospheric phenomena using different reanalyses datasets. In the same context, Hodges et al. (2003) and Wang et al. (2006) compared the frequency and intensity of extratropical cyclones found in the ERA-40, NCEP NCAR, and NCEP2 reanalyses. Both studies concluded that the ERA-40 and NCEP NCAR/NCEP2 reanalyses are in a good agreement with each other in capturing the behavior of NH extratropical cyclones whereas they are less consistent with the SH counterpart. Furthermore, ERA-40 tended to produce more cyclones with stronger intensity than the other two reanalyses. According to our analysis on the ERA-40 and the NCEP2 datasets in , the mean characteristics of SH winter cyclones in different levels of the troposphere detected in the NCEP2 reanalysis (Table A1) are consistent with those in ERA-40 presented in Table 1. The cyclones found with the NCEP2 data are fewer in number, less intense, and slightly slower in their movement velocity but larger in their radius and deeper than their counterparts with the ERA-40 data. On the

15 1JUNE 2007 L I M A N D S I MMONDS 2689 TABLE A2. Comparison of the properties of NCEP2 SH winter MSLP cyclones when the MSLP cyclones have well-organized structure up to the 500-hPa level (MSL 500) and their counterparts that end their connection at the 850-hPa level or lower levels (MSL 850). MSL 500 other hand, the vertical profile of the number and intensity of cyclones showing their minima at the 850- and 700-hPa levels is confirmed in the NCEP2 reanalysis. In regard to the mean features of vertically well organized cyclones, the NCEP2 reanalysis also presents vertically well organized cyclones to be more intense, larger, deeper, and longer lasting than the vertically shallow structured ones (Table A2). Moreover, the slower movement of vertically well organized cyclones compared to the shallow structured ones is found with the NCEP2 data as well. Consequently, it is encouraging to see such consistent physical features of SH extratropical cyclones in the two reanalyses despite the discrepancies shown in the trends of cyclone features discussed earlier. REFERENCES MSL 850 Intensity [hpa ( lat) 2 ] Radius ( lat) Depth (hpa) Life span (days) Speed (m s 1 ) Bengtsson, L., S. Hagemann, and K. I. Hodges, 2004: Can climate trends be calculated from reanalysis data? J. Geophys. Res., 109, D11111, doi: /2004jd Berbery, E. H., and C. S. Vera, 1996: Characteristics of the Southern Hemisphere winter storm track with filtered and unfiltered data. J. Atmos. Sci., 53, Charney, J. G., 1947: The dynamics of long waves in a baroclinic westerly current. J. Meteor., 4, Eady, E. T., 1949: Long waves and cyclone waves. Tellus, 1, Emanuel, K. A., M. Fantini, and A. J. Thorpe, 1987: Baroclinic instability in an environment of small stability to slantwise moist convection. Part I: Two-dimensional models. J. Atmos. Sci., 44, Folland, C. K., and Coauthors, 2001: Global temperature change and its uncertainties since Geophys. Res. Lett., 28, Fyfe, J. C., 2003: Extratropical Southern Hemisphere cyclones: Harbingers of climate change? J. Climate, 16, Gill, A. E., 1982: Atmosphere Ocean Dynamics. Academic Press, 662 pp. Green, J. S. A., 1960: A problem in baroclinic stability. Quart. J. Roy. Meteor. Soc., 86, Gyakum, J., P. Roebber, and T. Bullock, 1992: The role of antecedent surface vorticity development as a conditioning process in explosive cyclone intensification. Mon. Wea. Rev., 120, Hart, R. E., 2003: A cyclone phase space derived from thermal wind and thermal asymmetry. Mon. Wea. Rev., 131, Hennessy, K., 2004: Storms and climate change in Australia. Proc. Int. Conf. on Storms, Brisbane, Australia, Australian Meteorological and Oceanographic Society, Hodges, K. I., B. J. Hoskins, J. Boyle, and C. Thorncroft, 2003: A comparison of recent reanalysis datasets using objective feature tracking: Storm tracks and tropical easterly waves. Mon. Wea. Rev., 131, Holton, J. R., 1992: An Introduction to Dynamic Meteorology. Academic Press, 507 pp. Hoskins, B. J., 1990: Theory of extratropical cyclones. Extratropical Cyclones: The Erik Palmén Memorial Volume, C. W. Newton and E. O. Holopainen, Eds., Amer. Meteor. Soc., , and K. I. Hodges, 2005: A new perspective on Southern Hemisphere storm tracks. J. Climate, 18, Houghton, J. T., Y. Ding, D. J. Griggs, M. Noguer, P. J. van der Linden, X. Dai, K. Maskell, and C. A. Johnson, Eds., 2001: Climate Change 2001: The Scientific Basis. Cambridge University Press, 881 pp. Huo, Z., D.-L. Zhang, and J. R. Gyakum, 1996: The life cyclone of the intense IOP-14 storm during CASP II. Part I: Analysis and simulations. Atmos. Ocean, 34, Hurrell, J. W., H. van Loon, and D. J. Shea, 1998: The mean state of the troposphere. Meteorology of the Southern Hemisphere, Meteor. Monogr., No. 49, Amer. Meteor. Soc., Inatsu, M., and B. J. Hoskins, 2004: The zonal asymmetry of the Southern Hemisphere winter storm track. J. Climate, 17, James, I. N., 1987: Suppression of baroclinic instability in horizontally sheared flows. J. Atmos. Sci., 44, Jones, D. A., and I. Simmonds, 1993: A climatology of Southern Hemisphere extra-tropical cyclones. Climate Dyn., 9, Kageyama, M., and P. J. Valdes, 2000: Synoptic-scale perturbations in AGCM simulations of the present and Last Glacial Maximum climates. Climate Dyn., 16, Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77, Kanamitsu, M., W. Ebisuzaki, J. Woollen, S.-K. Yang, J. Hnilo, M. Fiorino, and G. L. Potter, 2002: NCEP DOE AMIP-II Reanalysis (R-2). Bull. Amer. Meteor. Soc., 83, Karoly, D., K. Braganza, P. Stott, J. Arblaster, G. Meehl, A. Broccoli, and K. Dixon, 2003: Detection of a human influence on North American climate. Science, 302, Keable, M., I. Simmonds, and K. Keay, 2002: Distribution and temporal variability of 500 hpa cyclone characteristics in the Southern Hemisphere. Int. J. Climatol., 22, Key, J. R., and A. C. Chan, 1999: Multidecadal global and regional trends in 1000 mb and 500 mb cyclone frequencies. Geophys. Res. Lett., 26, Levitus, S., J. I. Antonov, T. P. Boyer, and C. Stephens, 2000: Warming of the world ocean. Science, 287, Lim, E.-P., 2005: Global changes in synoptic activity with increasing CO 2. Ph.D. thesis, The University of Melbourne, Victoria, Australia, 381 pp., and I. Simmonds, 2002: Explosive cyclone development in the Southern Hemisphere and a comparison with Northern Hemisphere events. Mon. Wea. Rev., 130, Lindzen, R. S., 1993: Baroclinic neutrality and the tropopause. J. Atmos. Sci., 50, Murray, R. J., and I. Simmonds, 1991: A numerical scheme for