In this article two subsynoptic-scale cyclones that developed between 3 and 10 October 1996 over the western-central Mediterranean, causing floods, strong winds, and severe damage, are analyzed. Surface observations reveal that the accumulated rainfall at Santuario di Polsi (southern Calabria, Italy) is more than 480 mm for the first event (cyclone 9610A). The second cyclone (9610B) was characterized by a storm track predominantly over the sea, thus causing less recorded precipitation, but stronger wind. Satellite imagery shows two intensely convective vortices with a scale of 200–400 km and a spiral structure, with the cyclone 9610B displaying a well-defined eyelike feature.
The corresponding National Centers for Environmental Prediction analyses, although limited by 1° resolution, confirm the cyclones’ positions and intensities, as they can be inferred from satellite imagery, SSM/I data, and observations, but display also the “signature” of two tropical cyclone–like vortices, including a perfect alignment between the cutoffs at all levels with the surface center, and a warm core. The wind speed cross sections in the meridional and zonal directions through the eyelike feature reveal a virtually motionless column of air. A comparison with the cross sections taken in the same analyses across a named tropical storm in the Atlantic show a strong analogy between the gridded representation of these events.
Other remarkable features include very strong horizontal shear in the midtroposphere, and simultaneous lack of vertical shear; increasing low-level vorticity at the expenses of upper-level vorticity; creation of a low-level vorticity maximum; and finally strong low-level convergence and upper-level divergence during the onset and development of each cyclone.
Two anomalous cyclones observed over the Mediterranean Sea at approximately 40°N in October 1996 are analyzed. These events, which had no resemblance with any typical midlatitude event (baroclinic or orographic cyclones), acquired some of the features of a tropical storm. They also seemed to be different from polar lows, not only because of the much higher sea surface temperatures (SSTs) involved, but also because of the cyclogenetic mechanism. In fact, according to Sardie and Warner (1985) and Businger and Reed (1989), barotropic instability cannot be considered a cause for the development of polar lows. Conversely, in our two cases, barotropic instability seems to contribute to the cyclonic development.
Tropical cyclones need a large-scale low-level convergent, and upper-level divergent, environment, small amounts of vertical wind shear, high values of low-level vorticity that may be associated with large horizontal shear, and conditions favorable for barotropic instability (McBride and Zehr 1981). They depend strongly upon latent and sensible heat fluxes from the ocean (Charney and Eliassen 1964), since their main source of potential energy is the thermodynamic disequilibrium between the atmosphere and the underlying ocean (Ooyama 1969; Emanuel 1988). Latent heat release through cumulus convection (Kuo 1965) is their driving force, and the mechanism of air–sea interaction (Emanuel 1986; Rotunno and Emanuel 1987) is a crucial requirement for their intensification. The scale of tropical cyclones is of the order of hundreds of kilometers; satellite imagery reveals an eyelike feature and an axisymmetric shape. The horizontal structure is characterized by concentric isotachs around a calm eye such that wind speeds increase outward to a maximum then decrease beyond that point. The thermodynamic structure in the vertical has a warm core located over the surface pressure minimum with the warmest anomaly located in the midlevels (Anthes 1982).
In contrast, midlatitude cyclones grow mainly because of baroclinic instability (Charney 1947; Holton 1992), which requires horizontal temperature gradients and vertical wind shear. The scale of midlatitude cyclones is of the order of thousands of kilometers. Satellite imagery of baroclinic cyclones reveals an asymmetric cloud pattern with no eyelike feature. The vertical structure of midlatitude cyclones reveals a “tilt” between the upper-level troughs or cutoffs and the surface low. The thermal structure is asymmetrical, with cold advection generally to the north and to the west of the surface low, and warm advection to the south and to the east.
There is no sharp dividing line between tropical and extratropical cyclones. A third generic type of cyclone is the polar low (PL) referred to as an Arctic hurricane (Emanuel and Rotunno 1989). These intensely convective cyclones cannot be explained by dry baroclinic instability alone: latent heat release (Reed 1979; Sardie and Warner 1983; Businger and Reed 1989) and air–sea interaction (Emanuel and Rotunno 1989) play a major role in their development. Differences and similarities between polar lows and tropical cyclones are examined by Rasmussen (1989).
The first task of this study is to analyze the intense subsynoptic-scale cyclones that occurred on 3–6 October 1996 (hereafter, cyclone 9610A) and 7–10 October 1996 (hereafter, 9610B), to demonstrate their profound differences from baroclinic cyclones, and observe some similarities to tropical cyclones. We compare surface observations, satellite imagery, Special Sensor Microwave/Imager (SSM/I) data, and show the consistency with 1° National Centers for Environmental Prediction (NCEP) operational analyses. Consequently, the second task is to demonstrate that the NCEP analyses, although limited by the 1° resolution, are valid tools to detect the“signature” of some tropical cyclone–like disturbances.
2. Mediterranean cyclones
a. Baroclinic lee cyclones
The conception of cyclogenesis over the Mediterranean Sea has been dominated primarily by the baroclinic lee-cyclogenesis theory. The interaction of a large-scale baroclinic wave with an orographic obstacle like the Alps, by virtue of the conservation of potential vorticity (cyclonic vorticity is produced whenever a rotating column of air is stretched downstream of an obstacle), is the cause of a smaller-scale, orographically induced, baroclinic lee cyclone that is generated to the lee of the Alps, in response to a larger-scale cyclone over the Atlantic. The lee cyclone often moves southeastward and increases its scale from 100–300 to 500–1000 km (Buzzi and Tibaldi 1978). This is the so-called Alpine–Mediterranean cyclogenesis. A high incidence of baroclinic lee cyclones (BLCs) is found in the Genoa Gulf (hence the name of “Genoa” cyclones); however, Mediterranean lee cyclones are detected also in the lee of the Pyrenees, of the Atlas Range, of the Taurus Mountains, etc. Mediterranean BLCs have been studied from observational (Buzzi and Tibaldi 1978; McGinley 1982), theoretical (Speranza et al. 1985; Buzzi and Speranza 1986), and modeling perspectives (Tibaldi et al. 1980; Tibaldi and Buzzi 1983; Buzzi et al. 1994). Classifications have been made by Alpert et al. (1990) and by Alpert and Neeman (1992).
Furthermore, the roles played by latent and sensible heat fluxes from the Mediterranean Sea surface in amplifying cyclone growth have been examined by several authors and, in general, have been considered marginal or even negligible with respect to the role played by the orography and by baroclinic instability (Buzzi and Tibaldi 1978). More recently, Alpert et al. (1996) investigated the roles of convection, and sensible and latent heat fluxes in Alpine cyclogenesis, by separating their contributions. Although the authors distinguish between various phases of the typical life cycle of the cyclone, they agree that topographical blocking is the dominant factor in at least the first and most rapid cyclogenetic phase.
The above theory of Mediterranean BLCs, although applicable to a large fraction of Mediterranean cyclones, does not completely cover all of the different types of events that occur over the Mediterranean basin. In fact, at almost the same time as the Mediterranean BLC theory was being formalized, a different kind of subsynoptic cyclone, with similarities to tropical cyclones and polar lows, was observed.
b. Subsynoptic-scale vortices
Subsynoptic-scale vortices over the Mediterranean were first discussed in the early 1980s. The terms hurricane-like and Mediterranean tropical storms were used, respectively, by Billing et al. (1983) and Ernst and Matson (1983). Later, Mayengon (1984) spoke more generically about “warm core cyclones in the Mediterranean,” and Rasmussen and Zick (1987) pointed out their similarity with polar lows. Businger and Reed (1989) considered the Mediterranean warm-core cyclones as a particular case of PL: the so-called cold-low type. Analyzing the 14–15 January 1995 cyclone Blier and Ma (1996) refer to a Mediterranean Sea tropical storm and Pytharoulis et al. (1999) to a hurricane-like Mediterranean cyclone. The event that these authors studied is noteworthy for its exceptionally well-defined eyelike feature (as it appears in the satellite imagery of 15 Jan 1985) and for the surface winds recorded (up to 38 m s−1), measured by the oceanographic ship Meteor on 14 January 1995, at the time of maximum proximity of the ship to the storm’s center (data from Deutsche Wetterdienst, Germany).
The crucial point in common among all the above studies on Mediterranean subsynoptic-scale vortices is that baroclinic instability plays a very marginal role, in direct contrast to the theory of Mediterranean BLCs. Rasmussen and Zick (1987) concluded their case study of the 27 September–1 October 1983 event by stating:“It is unlikely that baroclinic instability plays any role for the formation and continued existence of the vortex. Convection on the other hand is an important factor and the large amount of convective available potential energy (CAPE) indicates that some kind of conditionally instability of second kind (CISK) may be the driving mechanism.” Conversely, Buzzi and Tibaldi’s (1978) study of the 3 April 1973 BLC concluded with the following: “latent heat plays an unimportant role in the initiation of the cyclone, although it may modify its characteristics in subsequent stages.” More recently, Craig and Gray (1996), in their article on CISK or wind induced surface heat exchange as the mechanism for tropical cyclone intensification, refer to “intensely convective polar lows and Mediterranean cyclones, which resemble tropical cyclones and are believed to intensify by the same process.”
In spite of the marginal role of baroclinicity observed in the subsynoptic-scale vortices, however, the cyclogenetic process takes place in an environment in which baroclinicity might have been present at an earlier stage. This is suggestive of some similarity with the so-called subtropical cyclones (Hebert 1973; Hebert and Poteat 1975), which are described as initially baroclinic cyclones eventually evolving into tropical systems.
3. Synoptic situation and observational evidence
a. The cyclone 9610A
In the past, the lack of observational data was the main reason why warm-core mesocyclones went undetected in the Mediterranean. Even in the 1980s, when they were discovered, the case studies were based almost entirely on few and sparse surface and upper air data, subjective weather maps, and satellite imagery. The amount of dynamic evaluations with this kind of information was of course quite limited. However, the increased resolution of current gridded data allows detection and analysis of these systems in a way that was not possible before. Gridded analyses represent the best and most coherent way of synthesizing the information coming from different and scattered instruments.
In contrast with earlier studies, this investigation will also use, beyond the standard observations, the NCEP Medium-Range Forecast model gridded initial conditions, at 1° resolution. It will be shown that the surface observation and the satellite imagery confirm what can be seen in the gridded analyses, but the vertical structure and causative mechanism of the cyclone can be much better investigated with the gridded data.
The sea level pressure (slp) map (Fig. 1a) and the geopotential height at 500 hPa (Fig. 1b) at 1200 UTC 2 October 1996 display a baroclinic wave from the Atlantic crossing the Alps. At this time a cyclone, similar to that described by Buzzi and Tibaldi (1978), is present in the lee of the Alps. At 0000 UTC 3 October, the cyclone is dissipating (Fig. 1c) and the 500-hPa geopotential height has increased over the Alps with respect to 12 h before, although a 500-hPa cutoff low is isolated over northern Italy (Fig. 1d). At 1200 UTC 3 October, the cyclone apparently dissipated: Fig. 1e shows a large surface trough over the south of Italy, while Fig. 1f still displays a filling 500-hPa cutoff, since the height in its center is about 40 m higher than 12 h before.
At that point, the BLC was almost at its end, but a change occurred with respect to the expected evolution. The cyclone described by Buzzi and Tibaldi (1978) occurred in the month of April, a time of the year in which the latent heat flux from the sea surface is relatively small. In this case however, the observed sea surface temperature was higher, ranging from 22° to 26°C, as it will be shown later. So, fluxes from the Mediterranean Sea might have played a role in rejuvenating the system through latent heat release and cumulus convection. Moreover, a strong cyclonically curved upper-tropospheric jet appears over northeastern Algeria and Tunisia (not shown).
Figure 1g shows a small-scale cyclone close to the Lybian coast (at about 15°E), at 0000 UTC 4 October 1996, and a 500-hPa cutoff low (Fig. 1h), tilted with respect to the surface low. Normally, if the tilt is maintained, a similar pattern leads to a BLC-type development, with an increase in scale. But 12 h later, at 1200 UTC 4 October (Fig. 1i), the cyclone is tracking northward with no increase in size; moreover, the minimum in slp is vertically aligned with the 500-hPa minimum (Fig. 1j). At 1726 4 October 1996 the National Ocean and Atmosphere Administration (NOAA) infrared polar orbiter satellite provides a picture of a mesocyclonic vortex, with a spiral cloud structure, between Sicily and Tunisia (Fig. 2). The cloud distribution differs substantially from the one characteristic of BLCs, much more frequently detected in the Mediterranean: BLC satellite signatures have a much larger scale and usually provide clear evidence of fronts (e.g., Buzzi and Tibaldi 1978).
Surface reports allow the detection of a small-scale cyclonic circulation between Sicily and Tunisia, at 1800 UTC 4 October 1996 (Fig. 3a), but observations were not available within 50 km of the storm center. This can be estimated from the satellite picture to be at about 37.5°N and 12°E.
However, a powerful contribution toward the understanding of the storm’s structure comes from the SSM/I. In Fig. 3b the liquid cloud water is displayed at 1718 UTC 4 October 1996. A very clear spiral structure is detected between Sicily and Tunisia, in agreement with the satellite image (Fig. 2). However, in the satellite picture there is some cloudiness also to the east of the storm: a wide band, partly covered by a cirrus shield, elongated in the northwest–southeast direction across Calabria and over the Ionian Sea. This cloudiness belongs to the same cyclonic circulation, and the SSM/I reveals that, below the cirrus shield, there are very high values of liquid water (Fig. 3b, between 18° and 19°E). It is reasonable to extrapolate those high values also over land, across Calabria (where SSM/I data are obviously not available), by virtue of the agreement with the NOAA infrared image (Fig. 2).
This feature resembles some “rainbands” often observed in tropical systems. The spiral structure and the rainband feature are indirectly confirmed also by the SSM/I based wind speed (Fig. 3c), because the data are“shadowed” wherever heavy precipitation occurs.
Of enormous interest is the SSM/I-derived water vapor for the same time (Fig. 3d). A broad area of high values of water vapor, in agreement with the satellite picture, and with the other fields obtained from SSM/I, appears to the southeast of Calabria, inserted in the same cyclonic circulation, the center of which is between Tunisia and Sicily.
At 1200 UTC 5 October 1996 the cyclone’s storm is between Calabria (southernmost part of Italy) and Sicily, according to the NCEP analyses (Figs. 4a,b). The cyclone is still vertically aligned. The gradual increase in symmetry, the alignment of the center of the 500-hPa cutoff with the surface minimum center, occurring between 1200 UTC 3 October and 1200 UTC 5 October 1996, is associated with intensification (evident from the fall in sea level pressure at the storm center) and scale reduction, unlike occluding, mature, baroclinic cyclones, in which the disappearing of the tilt between the minimum values of geopotential at the various levels leads to weakening and eventual dissipation.
The landfall occurred over Sicily in the first hours of 5 October is confirmed by the NOAA infrared satellite image of 0226 UTC 5 October 1996 (Fig. 5), which displays a spiral structure and a compact area of convection over those regions. Surface data at 0000, 0600, and 1200 UTC 5 October 1996 (Figs. 6a–d) confirm an east-northeastward track, with landfall over the island of Sicily and the Calabrian peninsula.
The sparsity of synoptic observations over southern Italy is a serious limitation for categorizing the cyclone:in fact, with little surface wind data available, the mesolow would barely enter the category of tropical storm intensity. However, precipitation was extremely intense, surely comparable with landfall hurricanes, and caused severe floods in Sicily and over the southern part of Calabria. A fairly dense network of rain gauges of the Istituto Idrografico e Mareografico di Catanzaro reveals that the accumulated precipitation for 3–5 October 1996 exceeds 300 mm at four locations and the highest value is 480.4 mm (of which 284.8 fell on 4 Oct only) at Santuario di Polsi (Figs. 7a and 7b).
b. Cyclone 9610B
On 6 October at 0000 UTC, another baroclinic wave crossed the Alps (Figs. 8a,b) and a 500-hPa cutoff generated a Genoa cyclone over northwestern Italy in agreement with the orographic–baroclinic theory.
On the sea level pressure map (Fig. 8a), the remnant of the mesocyclonic vortex 9610A over southeastern Italy is at a dissipating stage. It is interesting to note two very different cyclones at a close distance to each other: the lee cyclone over northwestern Italy and the 500-hPa cutoff over France reveal a typical baroclinic structure (slp minimum placed in the area of positive vorticity advection of a 500-hPa wave); the 9610A mesocyclone over southeastern Italy displays a perfect alignment of the surface center with the 500-hPa minimum. Both cyclones are nested into a larger-scale, weak, 500-hPa trough.
Within the framework of Alpine–Mediterranean cyclogenesis, it would be reasonable to expect an increase in scale and a baroclinic development, but this is not the case. At 1200 UTC 6 October 1996 the lee cyclone has already dissipated (Fig. 8c). The broad and weak trough over most of Italy is associated with the last stage of existence of the less than fully developed lee cyclone that 12 h earlier was placed over the Genoa Gulf. A very broad and weak 500-hPa trough (Fig. 8d) and a weak baroclinic environment, left by the occluded cyclone, dominate most of the central-northern Mediterranean. At that point, another subsynoptic-scale cyclone begins to develop (Fig. 8c) close to the Algerian coast.
At 0000 UTC 7 October (Fig. 8e), the small-scale cyclone is located to the west of Sardinia. We defined this event as 9610B. The diameter of the 1008-hPa closed isobar is smaller than 12 h earlier, while the cyclone is deeper. The 500-hPa cutoff appears elongated in the direction of the surface low (Fig. 8f), but the 500-hPa geopotential field is very flat over the developing low. Conversely, a steep 500-hPa geopotential gradient, indicative of a strong westerly jet that will be discussed later, can be seen immediately to the south of the developing storm (similarly to 9610A).
The process of reduction of scale and intensification continued: at 1200 UTC 7 October, the diameter of the 1008-hPA isobar is unchanged, but the central pressure in the gridded analyses is 999 hPa (Fig. 8g). As it will be shown from the available surface observations the cyclonic circulation is confirmed, but no data are recorded between Sardinia and the Balearic Islands where the storm center is located. As such, it is likely that the actual central pressure is lower than the value of 999 hPa in the gridded analyses at 1° resolution. The surface cyclone is aligned perfectly with the 500-hPa cutoff (Fig. 8h) and with the geopotential minima at all levels, from the surface up to the tropopause. It will be shown later that a warm anomaly is centered at the axis of the system in the lower levels, nested within a larger-scale cold anomaly.
According to the NCEP analyses the cyclone moves eastward: at 0000 UTC 8 October the system reaches the southeastern coast of Sardinia (Figs. 8i,j) and at 1200 UTC 8 October the cyclone is again on the sea, still tracking eastward (Figs. 8k,l).
The NCEP analyses are confirmed by the satellite imagery. In fact, at 1130 UTC 7 October 1996, the European geosynchronous meteorological satellite Meteosat provides a picture, in the infrared bandwidth, of a spiral-shaped mesocyclonic vortex, with an evident eyelike feature (Fig. 9a), to the west of Sardinia (Italy). The sequence of previous and later pictures (not shown) reveals that the angular rotation rate of the small vortex, whose diameter does not exceed 300 km, is much faster than the cloudiness over the northern part of Italy, associated with the previous baroclinic lee cyclone. In the following hours, the cyclone crosses the island, temporarily loosing its spiral structure and eye, possibly because of increased frictional drag (Sardinia’s highest peak reaches 1800 m) and lack of the energy source provided by the warm sea. But on 8 October the system regains its spiral structure; a smaller (10–30 km) and better-defined eye reappears (Fig. 9b). At this stage, the different speed of the convective clouds, associated with the small vortex, and the predominantly stratiform cloudiness on a larger scale (over northern Italy), allows a separation of the two systems, that is, a rotating mesocyclonic vortex, nested into a slower cyclonically rotating synoptic-scale cloud system.
The SSM/I data at 1015 7 October 1996 confirm the spiral structure to the west of Sardinia. This is evident in both the liquid cloud water and in the wind speed (Figs. 10a,b), although the latter is partly “masked” by the heavy precipitation occurring.
The surface observations also confirm the storm location and track, as it appears in the NCEP analyses. The storm’s center is clearly between the Balearic Islands and Sardinina at 1200 UTC 7 October (Fig. 10c) and landfall appears to be taking place on the southeastern part of Sardinia (Fig. 10d) at 0000 UTC 8 October 1996 (note one slp value of 996 hPa, lower than the gridded analyses).
The observations confirm the farther eastward progression (Figs. 10e,f), and the placement to the east of Sardinia at 1200 UTC 8 October. No data are recorded in the most intense part of the storm but one ship located approximately 100 km to the east of the cyclone eye (easily seen in the satellite picture of Fig. 9b) records at 1200 UTC 8 October 1996 a sustained southerly wind of 25 m s−1 (Fig. 10e).
On 9 October, the cyclone intensifies and becomes smaller in horizontal extent, but this process is not captured well in the NCEP analyses. In fact, the gridded analyses between 0000 UTC 9 October and 0000 UTC 10 October (Fig. 11) still display the alignment between surface and upper-level minima, but erroneously portray a weakening with respect to 0000 and 1200 UTC 8 October (Fig. 8). In reality, the satellite imagery reveals that the vortex is acquiring a smaller but more distinct eyelike feature and becoming so compact (Figs. 12a,b;1319 and 1717 UTC 9 Oct 1996) that the 1° resolutionof the NCEP analyses becomes inadequate to resolve it.
Observations at 0000, 0600, and 1200 UTC 9 October (Figs. 10f and 13a,b) indicate a cyclone in the Tyrrhenian Sea. No observation is available in the proximity of the storm’s center. However, the intensity of the storm during the afternoon of 9 October is verified by two surface reports. At 1500 and 1800 UTC sustained northwesterly wind of 22.5 m s−1 (Fig. 13c,d) is reported over the island of Ustica (38.7°N, 13.2°E). As seen from the satellite picture taken at 1717 UTC 9 October 1996 (Fig. 12b), the storm eye was located approximatively 100 km to the east-northeast from the reporting station. Although affected by the proximity to land and by the heavy precipitation rates occurring, the presence of the vortex to the north of Sicily is not contradicted by the SSM/I data at 1624 UTC 9 October (Figs. 13e,f).
The maximum damage due to wind occurred over the Aeolian Islands, at 38.5°N, 15°E, to the northeast of Sicily: assistance for disaster relief was required. Unfortunately, no weather station data were available, but the media reported sheds, roofs and harbor devices destroyed, and houses and electric lines damaged, due to“extremely strong westerly wind.” The perfect agreement between the observations at Ustica (Figs. 13c,d), the storm scale, the eyelike feature position (Fig. 12b), and the damages over the Aeolian Island reasonably suggest that the hurricane-level intensity of 32 m s−1 was reached over the Aeolian Islands.
Then the cyclone, in its southeastward motion, crosses Calabria. However, precipitation recorded by the Calabrian rain gauge network is much more moderate than for 9610A: most of the gauges report values between 40 and 100 mm of rain, and no measurement exceeds 150 mm (not shown). On 10 October, the surface center moves toward to east-southeast, starting a dissipation phase (Figs. 11e and 11f).
4. Vertical structure
The presence of an eye in the satellite picture does not allow for unique inferences about the three-dimensional structure of the cyclone. Very intense baroclinic oceanic cyclones can have an eyelike feature, and might look similar to tropical cyclones (Bluestein 1992) in the visible satellite imagery, because of the seclusion of clear, dry air. The famous Presidents’ Day snowstorm (Bosart 1981; Atlas 1987) is a good example of this feature. In general, the eyelike features typical of nontropical lows are surrounded by low stratus or stratocumulus clouds, whereas hurricanes have tall cumulonimbus clouds near the center. Both systems 9610A and 9610B seem to have deep convective clouds around the eye. However, these features cannot be considered peculiar for a tropical cyclone.
Therefore, the investigation of the three-dimensional structure becomes necessary. In the absence of adequate observations like the ones that are routinely performed by aircraft across the tropical systems, the NCEP operational analyses will be used in this study as an approximation to the atmospheric structure. Although limited by a horizontal resolution of 1°, a number of significant features, which we believe are the “gridded signature” of a tropical cyclone–like vortex, are evident. These include the following:
negligible baroclinicity and very small or negative vertical shear on the scale of the vortex;
perfect alignment between the geopotential minima at all levels and the sea level pressure minimum;
a virtually windless column of air centered over the sea level pressure minimum, extending from the lower levels up to the tropopause; and
a warm anomaly in the low troposphere, centered over the corresponding sea level pressure minimum, and nested into a larger-scale cold anomaly.
Figure 14 displays the wind velocity and temperature structure across the center of the cyclone at 1200 UTC on 7 October. The cross sections are taken at 40°N and 6°E, corresponding to the surface minimum (Fig. 8g) position. An almost vertical column of wind speed smaller than 5 m s−1 is evident between 800 and 300 hPa. On the southern side of the meridional cross section we can see an upper-level jet that might play a role in the creation of the vortex, as discussed later. The two cross sections indicate an almost cylindrical velocity structure. The 10 m s−1 isotach around the eye is vertical, surrounding a column of no wind, and extends from the surface minimum to the tropopause. Although almost no vertical shear is present on the scale of the vortex, a very strong horizontal shear can be seen around the eye. Moreover, the isotherms plotted across the axis of the system give a hint of a warm core from the surface level up to 500 hPa. The warm core was not present 24 h before (not shown). In analogy with Hurricane Diana (Bosart and Bartlo 1991), which developed in a baroclinic environment, the warm core gradually replaces the original cold anomaly, or, using Bosart and Bartlo’s words, “a previously existing cold-dome is destroyed by the creation of a warm-core.”
In Fig. 15, the zonal and meridional cross sections relative to the position of the minimum on 8 October at 1200 UTC (40°N and 11°E) can be seen. The vertical structure is quite similar to that in Fig. 14. However, in the satellite picture (recall Fig. 9b), the scale appears to be much smaller. As a result, the coarse resolution of the gridded analyses did not capture the true intensity of the vortex (recall Fig. 10e) by “spreading” it on a larger scale.
However, the NCEP gridded representation of the cyclone 9610B compares remarkably with the Atlantic Tropical Storm Josephine (1996). This is particularly evident in the cross sections across the center. Josephine was named a tropical storm at 1800 UTC 6 October 1996, when the position of its center was 25.1°N, 91.8°W and the central pressure was 1001 hPa. By 0000 UTC, the position was 25.5°N, 90.4°W and the central pressure 996 hPa. On 7 October, Josephine strengthened and almost reached hurricane intensity (30 m s−1). However, on 8 October, the storm weakened because of increased baroclinicity and Josephine was declared extratropical by 0600 UTC 8 October 1996 (Pasch 1997). We chose Josephine for a comparison because it developed in proximity of or within a baroclinic environment, and its observed intensity is similar to the storm 9610B.
The low-level gridded wind speed of cyclone 9610B reaches 20 m s−1, as it appears in the cross sections, and is comparable to the values seen in the cross sections across Josephine (Fig. 16) at 26°N and 90°W, 0000 UTC 7 October 1996. In fact, the low-level wind speed maxima range from 15 to 20 m s−1. Cross sections across the cyclone 9610A at 1200 UTC 5 October are also comparable (not shown).
The above-described vertical structure and the scale reveal clear similarities among cyclones 9610A, 9610B, and Josephine, and differences with baroclinic cyclones, and, in particular, with the BLCs more common in the Mediterranean (Buzzi and Tibaldi 1978).
However, the vertical sections also differ substantially from sections across tropical cyclones that are completely developed in a tropical environment (which is not the case with Josephine). In fact, tropical cyclones normally display the maximum wind at the low levels, whereas the meridional cross sections across cyclones 9610A, 9610B, and Josephine, all surrounded by a baroclinic environment, show evidence of a jet at the tropopause. Thus, events 9610A and 9610B are somehow similar to the westward moving tropical systems that are occasionally observed in the midlatitudes (sometimes placed to the north of an upper-tropospheric westerly jet), rather than to eastward tracking tropical systems, that are fully developed in a tropical environment and enbedded in an easterly flow.
5. Other similarities with tropical cyclones
a. Divergence and vorticity
The onset of the 9610B storm reveals some other interesting analogies with the tropical cyclogenesis. Figure 17 provides a clear picture of an intensifying low-level convergent (panels a and c) and upper-level divergent environment (panels c and d), at the onset of the storm. Between 0000 and 1200 UTC 6 October 1996, the low-level convergent area becomes narrower and more intense, while the upper-level divergent environment becomes broader and more intense as well.
In good correspondence with the maximum low-level convergence and with the position of the surface low, a small-scale 850-hPa vorticity maximum at 1200 UTC 6 October can be seen in Fig. 18a. Another vorticity maximum operates from 500 hPa to the tropopause; the 300-hPa vorticity at 1200 UTC 6 October can be seen in Fig. 18b. At 1200 UTC 7 October, the low-level vorticity maximum keeps intensifying and appears centered and symmetric with the position of the slp minimum (Fig. 18c). The 300-hPa vorticity maximum is advected toward the area in which the storm is developing (Fig. 18d). The storm appears to be the result of a cooperative interaction between upper-level and low-level vorticity, with some similarities to Hurricane Diana (Bosart and Bartlo 1991).
By 0000 UTC 8 October 1996, the vorticity in the lower levels has strongly increased, whereas at the upper levels it has been reduced (Figs. 18e and 18f). Moreover, the 850-hPa vorticity maximum is symmetric with the surface low minimum (Fig. 8i). At this time, the 300-hPa vorticity maximum has taken the shape of a broad ring surrounding the cyclone with a 2° radius.
Vorticity cross sections, centered over the developing surface low, reveal even more clearly the existence of two vorticity maxima. At 1200 UTC 6 October 1996, an upper-level 300-hPa maximum appears on the zonal vertical section at 40°N (Fig. 19a). The meridional cross section at 5°E reveals an intense low-level vorticity maximum at 38°N (Fig. 19b) where the surface low first appears. The upper-level vorticity maximum that is seen in Fig. 19b is linked with the same maximum that appears in the zonal section, as revealed by Fig. 19a. This maximum belongs to the inner flank of a powerful 300-hPa cyclonically curved jet, which will be discussed shortly.
By 1200 UTC 7 October 1996, the gradual creation of a cylindrical feature is evident, but the most noticeable fact is the increase of low-level vorticity and the decrease of the upper-level vorticity (Figs. 19c,d). At 0000 UTC 8 October 1996, the similarity between the 9610B event and a tropical cyclone is evident in the two cross sections (Figs. 19e,f) centered over the surface low at 40°N and 8°E (recall Fig. 8i). The vorticity is stretched along the vertical axis of the cyclone, mostly between the surface and 500 hPa.
These findings reveal a substantial difference with any typical baroclinic cyclone. However, there is an important difference with tropical cyclones as well: tropical cyclones usually display an anticyclone, and thus negative vorticity, between 200 and 100 hPa; we could not find a convincing evidence of such feature in cyclones 9610A and 9610B.
b. Horizontal and vertical shear
The main source of vorticity for the developing storm is from the very intense horizontal shear in the mid- and upper troposphere. Figure 20 illustrates the positions of a 50–60 m s−1 300-hPa jet from 0000 UTC 6 October 1996 to 1200 UTC 7 October 1996. The abrupt eastward turn of the jet between Spain and the Atlas range is evident in the figure. This is a significant feature as it results in generating an impressive horizontal shear of about 60 m s−1 over 200 km (3 × 10−4 s−1) on its cyclonic shear side.
The anomalous horizontal shear is also observed in the midtroposphere. At 500 hPa in the vicinity of the jet, the wind speed decreases from 50 m s−1 to almost zero in about 200 km in a line perpendicular to the jet axis (Figs. 21a,b). This produces a horizontal shear of 2.5 × 10−4 s−1. A virtually stagnant area of no wind (less than 5 m s−1) can be seen at 500 hPa on October 6 at 1200 UTC and October 7 at 0000 UTC directly over the incipient surface low (Figs. 21a and 21b).
The analysis of the vertical wind shear from 850 to 300 hPa, used by Alpert and Neeman (1992) to investigate the baroclinicity of the so-called cold small-scale cyclones over the eastern Mediterranean, provide an interesting insight. The authors found that these cyclones were characterized with a vertical shear of about 25 m s−1. But the same quantity plotted in our case reveals that the entire area affected by the cyclogenetic process has a vertical shear smaller than 5 m s−1 (Fig. 21c). At 0000 UTC 8 October 1996 the contour lines of 5 m s−1, and even −5 m s−1 in the inner part, perfectly delineate the scale and the location of an almost barotropic cyclone (Fig. 21d). The scale of the event (about 300 km) and the slightly negative vertical wind shear would place this cyclone completely apart from all the cyclones examined by Alpert and Neeman (1992, their Fig. 7).
After 0000 UTC 8 October, the jet was far to the east and weakened significantly (not shown). The weakening of the jet and the strengthening of the vortex support the idea that the disturbance has developed by extracting kinetic energy from a barotropically unstable mid- and upper-tropospheric flow, with the contribution of latent heat release through intense cumulus convection.
Kuo (1949) investigated the concept of instability of a zonal flow varying with latitude and time U(y, t) and having a nonzero meridional velocity shear ∂U/∂y, symmetrical with respect to a certain latitude. By using a linearized theory and a barotropic atmosphere (without vertical shear), he defined a critical latitude yk as the latitude where the absolute vorticity of the zonal flow has a maximum or a minimum. The necessary condition is, therefore,
Kuo (1949) concluded that a wave can grow at the expense of the kinetic energy of the zonal current.
Nitta and Yanai (1969) analyzed the barotropic instability of two profiles of a westerly and an easterly jet:
where b can be conceived as the half-width of the jet. For each flow, they produced the critical curve that gives the growth rate for a wave extracting energy from that flow, as a function of L/b, where L is the wavelength involved.
Based on this work, general agreement has been reached that horizontal shear, and therefore barotropic instability, cannot be considered a cause for development of polar lows. This is mostly because in all of the areas where polar lows occur, jets do not get narrow enough to develop a shear adequate for the small wavelengths typical of polar lows. Sardie and Warner (1985) and Businger and Reed (1989) concluded that, given a “strong” shear for a westerly jet (i.e., 80 m s−1 over 500 km) the corresponding growth rate would be very small. Therefore, Sardie and Warner (1985) point to moist baroclinicity and CISK working together as a causative mechanism for the northern Atlantic polar lows, where a noticeable vertical shear into the lower levels has to be present.
These conclusions do not seem to apply to our case for two reasons. First, during the entire time of the storm’s development, the vertical shear is virtually zero right over the developing low (Figs. 21c,d). Second, the jet observed in our case has indeed a much stronger horizontal shear. At 300 hPa, we can observe a decrease in speed of more than 60 m s−1 over only 200 km (Fig. 20c), which is about twice the shear hypothesized from the above authors as strong. Even at lower levels, the horizontal shear is very high for the entire time in which the storm develops.
The verification of Kuo’s barotropic instability condition cannot be rigorously tested for our case, since the atmosphere is not barotropic and the flow is not zonal. However, the quantity
can be conceived as an indicator of where barotropic instability is not impossible, in response to the shear of the zonal component of the wind.
In Fig. 22, the quantity K(y) calculated at 700 hPa is plotted against the sea level pressure. It appears that a line K(y) = 0 crosses the storm center at all times after 0000 UTC 7 October. The situation does not change at different levels (not shown). These findings are suggestive that barotropic instability might play some role in the development. Recalling Sardie and Warner (1985) this would be an important difference with polar lows.
Conversely, there are some interesting similarities with the vortex described by Krishnamurti et al. (1981). In that study, the authors analyze the dynamics of the so-called onset vortex of the summer monsoon over India. The vortex has the appearance of a tropical storm and forms on the cyclonic (northern) side of a westerly jet. The authors demonstrate that the jet satisfies the necessary condition for barotropic instability and show how the disturbance could grow extracting energy from the flow. In addition, the authors acknowledge the possible contributions of other sources of energy, and hypothesize that there could be a cooperative interaction between barotropic–baroclinic instability and convection.
6. Temperature analysis
a. Sea surface temperature and fluxes
It has been generally accepted that 26.5°C is the SST threshold for hurricane development. This is however a very empirical threshold, since several named hurricanes were observed over much cooler waters (e.g., cyclones Ivan and Karl during the 1980 season).
Specifically for the two cases in this study, the analyzed SSTs range from about 22.5° to 25°C over the area where the 9610A cyclone develops, respectively, to the southeast of Calabria and to the southwest of Sicily, between 1200 UTC 1 October and 1200 UTC 4 October 1996 (Figs. 23a and 23b). Cyclone 9610B develops over even cooler waters: mostly about 21.5°C, as can be seen from Figs. 23c and 23d, relative to 1200 UTC 6 October and 1200 UTC 8 October 1996.
However, the observed range of 21.5°–25°C for storms 9610A and 9610B is much closer to the tropical SST values than the values normally detected during polar low development. Conversely, air temperatures recorded at stations at sea level or over ships are lower than the air temperature recorded during tropical storms.
In Fig. 24, we compare the analyzed SST with the small amount of observed marine data (buoys and ships), including also the actual air and dewpoint temperatures, wind, pressure, and observed weather. In Fig. 24a the fields are displayed at 1200 UTC 4 October 1996, during 9610A’s development (cf. Figs. 1i and 2). It is remarkable to observe that two reports at 35°–36°N and 18°–19°E show SSTs of 26°C. This warm anomaly is not captured in the analyses. For the northernmost of these two reports, we calculate the sensible Hs and latent HL heat with the usual “bulk formulas”:
where ρ is the air density, CD the drag coefficient (1.5 × 10−3), cp the specific heat at constant pressure, Lυ the latent heat of vaporization, V is the surface wind speed, Ta is the air temperature, qs is the saturated specific humidity, and qa is the actual air specific humidity. The calculation provides approximately 480 W m−2 of latent heat and 85 W m−2 of sensible heat.
In Fig. 24b the reports relative to storm 9610B at 1200 UTC 8 October 1996 (cf. Figs. 8k and 9b) show a reasonable agreement of the observed SSTs with the analyzed ones. The calculation of the fluxes for the ship located at 37.4°N and 11°E, in spite of the relatively low SST (22°C), provides even stronger fluxes: 640 W m−2 of latent heat and 180 W m−2 of sensible heat, for a total of about 820 W m−2. The calculation for the ship located at 40°N and 13°E, leads to comparable values: the noticeable wind speed of 25 m s−1 allows values of 480 W m−2 and 85 W m−2 in spite of the relatively small air–sea temperature difference. These values are remarkable, but lower than the ones expected in tropical storms environment (total flux of the order of 1000 W m−2). However, none of these reports are taken in the most intense part of the storm.
The remarkable aspect is that latent heat flux is much stronger than sensible heat. This is one of the most important differences with polar lows, in which sensible and latent heat are normally of comparable magnitude.
b. Low-level temperature advection
In polar lows substantial cold and warm advection takes place in the lower levels (Sardie and Warner 1985). In Fig. 25 the 1000-hPa temperature and wind from the NCEP analyses are displayed during two stages of 9610B’s storm evolution. At 1200 UTC 7 October 1996 the storm is in its deepening phase, and the alignment of the surface center with the 500-hPa cutoff has just taken place (Figs. 8g,h). The storm center can be estimated at approximately 40°N and 7°E (Fig. 25a), in agreement with the satellite image (Fig. 9a) and the observations (Fig. 10c). There is cold advection from the south to the east of the center, and warm advection from the east, to the north of the center. This pattern leads to an almost isothermal cyclone. For a baroclinic cyclone, the consequence would be an occlusion and rapid dissipation. At 1200 UTC 8 October 1996 (Fig. 25b) the storm center can be seen at approximately 40°N and 11°E, in agreement with satellite (Fig. 9b) and observations (Figs. 10e and 24d). The vortex is completely contained within the 19° isotherm, and a hint of a warm core can be inferred. Some cold (warm) advection is present on a larger scale, to the southwest of Sardinia (to the southwest of Sicily), but not on the scale of the vortex, which is almost isothermal.
Yet, from section 3b, we know that the storm went through further strengthening on 9 October. Therefore, the almost isothermal structure of the storm, given the good agreement between the analyzed and observed temperature fields (Figs. 24b and 25b), provides us with further confidence that storm 9610B is not a baroclinic cyclone.
7. Tentative mechanism
The purpose of the present paper is limited to the observational evidence of the anomalous nature of cyclones 9610A and 9610B. An investigation of the causative mechanism will be the subject of a further study. Moreover, the poor data coverage and the absence of aircraft data in the storms’ eye are some implicit limitations.
However, given these caveats, we tentatively summarize a plausible development mechanism for cyclone 9610B.
A mid- and upper-tropospheric cold trough approaches the western Mediterranean, over an area with some very weak baroclinicity and preexisting convection. Positive vorticity advection in the midlevels triggers ascent, a decrease in sea level pressure, and an increase of low-level convergence over a broad marine area, where fluxes rapidly increase due to increasing winds over a relatively warm sea surface temperature. This baroclinic startup is similar to Hurricane Diana (Bosart and Bartlo 1991). Simultaneously, on the southern flank of the trough, a very strong and narrow barotropically unstable mid- and upper-tropospheric jet generates strong horizontal shear over a marine area where vertical shear is very small. This area coincides with the developing surface low. By extracting kinetic energy from the barotropically unstable jet the cyclone starts to develop in a fashion similar to the vortex described in Krishnamurti et al. (1981). Convection intensifies and thereby reduces the cold anomaly. Again, the destruction of the cold anomaly by convection is similar to Bosart and Bartlo’s (1981) results.
Overall, there is some evidence of a plausible cooperation between large-scale baroclinic instability and smaller-scale barotropic instability over an area with strong fluxes and with the development of deep convection. This cooperation was already hypothesized for that fraction of tropical cyclones [one in 20, according to Frank (1977)] that normally develop in a baroclinic environment.
It is significant to observe that cyclone 9610B, during the entire day of 9 October, is almost “locked” in the southeastern Tyrrheanian Sea. The almost perfectly circular shape of the sea surrounded by Calabria and Sicily, where the storm was placed with very little motion over the relatively warm water, suggests an important channeling role of orography. We can speculate that, in the absence of strong external forcings, the system was self-sustaining during 9 October, in a sort of equilibrium, and that air–sea interaction and CISK may have played an important role.
The question that arises naturally is: why did these, and other similar cyclones develop in the Mediterranean Sea? At this stage, we do not have sufficient evidence to claim that storms 9610A and 9610B are similar to other Mediterranean vortices observed in wintertime. However, for these two particular cases, it can be said that the observed SSTs are higher than the ones observed at the same latitude, in October, outside the Mediterranean.
In this study we attempted to demonstrate that two anomalous Mediterranean mesocyclones detected on 3–6 and 7–10 October 1996 differ significantly from the common baroclinic lee cyclones. Conversely, we notice that some similarities exist with those tropical cyclones that develop outside the tropical atmosphere, in proximity of a baroclinic environment. The similarities are not only in their scale, satellite imagery, precipitation rates, and wind strength, but also in the vertical structure and the cyclogenetic process: importance of surface fluxes; absence of baroclinicity and therefore of vertical shear during the storm development, even when the surrounding environment is baroclinic; and presence of very strong horizontal shear in the mid- and upper troposphere. This study suggests that the contribution of barotropic instability to the storm development is plausible.
By having compared the surface observations, SSM/I data, and the satellite imagery with the NCEP operational analyses at 1° resolution, it was demonstrated that the analyses are a valid tool to detect the signature of tropical cyclone–like vortices, even if observations in the storm center are lacking. However, we realize that the analyses show some inconsistencies, the most important of which is the apparent lack of an upper-level anticyclone. We can speculate at this point that a more extensive data coverage in the Mediterranean region might contribute to eliminating, or clarifying, some of these aspects.
The study started under CNR Grant 970080PF42 and Contract NAS-5-32484. It was then supported by the Centro Interuniversitario di Ricerca in Monitoraggio Ambientale, Genoa, Italy, under a cooperative agreement for the improvement of precipitation forecasts in the western Mediterranean region. We deeply thank Dr. J. C. Jusem for his encouragement and suggestions. We also wish to thank Drs. P. Alpert and L. Bosart for helpful comments. The NOAA satellite imagery is a courtesy of the Dundee Satellite Receiving Station, Dundee University, Scotland. The Meteosat images (courtesy EUMETSAT) are provided by the University of Nottingham, United Kingdom. Precipitation data from Calabria were kindly provided by Dr. Niccoli from the Ufficio Idrografico e Mareografico di Catanzaro, Dipartimento dei Servizi Tecnici Nazionali, Presidenza del Consiglio dei Ministri. We thank J. Ardizzone, L. Ferraris, and J. Terry for their aid in acquiring data. We also thank three anonymous reviewers for helpful comments.
Corresponding author address: Dr. Oreste Reale, Center for Ocean–Land–Atmosphere Studies, 4041 Powder Mill Road, Suite 302, Calverton, MD 20705.