Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Meteorol. Appl. 7, 261–279 (2000) The hurricane-like Mediterranean cyclone of January 1995 Ioannis Pytharoulis1, George C Craig1, Susan P Ballard2 1 Department of Meteorology, University of Reading, Earley Gate, PO BOX 243, Reading RG6 6BB, UK 2 Meteorological Office Unit, Joint Centre for Mesoscale Meteorology, University of Reading, UK (now at NWP Division, Met. Office, London Road, Bracknell RG12 2SZ, UK) The development of a hurricane-like vortex over the Mediterranean Sea was studied using (mainly) the UK Met. Office Unified Model. The Mediterranean cyclone formed in the morning of 15 January 1995 over the sea between Greece and Sicily. Strong convection was observed prior to its genesis. During the longest part of the cyclone’s lifetime, strong surface fluxes and, as a result, deep convection existed in its vicinity. Its track was influenced by the surface fluxes and the flow in the wider region. The forecast of the mesoscale and limited-area models reproduced the general characteristics of the actual system as they appeared at the surface and upper-air charts and at the satellite imagery. The investigation of the cyclone’s characteristics gave strong evidence (including an ‘eye’ and a warm core) to support the initial assertion that it was similar to tropical cyclones and some polar lows. Baroclinic instability does not seem particularly important, although the cyclone formed at the edge of a baroclinic zone. A numerical experiment showed the vortex did not develop in the absence of surface heat and moisture fluxes. Another experiment showed that sensible and latent heat fluxes were equally important in its development. 1. Introduction In mid-January 1995 a small hurricane-like cyclone was detected by the Meteosat and the polar-orbiting NOAA satellites over the central Mediterranean Sea. Generally, most of the cyclogenesis occurring in the Mediterranean is caused by the lee effect of the surrounding mountains (e.g. Buzzi & Tipaldi, 1978; Flocas, 1994). In addition, Prezerakos & Flocas (1996) have shown the importance of dynamically unstable upper-tropospheric ridges in inducing surface cyclogenesis in the region and particularly over the Aegean Sea (eastern Greece) where cyclogenesis is quite a rare phenomenon (Flocas & Karacostas, 1996). However, sometimes small cyclones with intense convection and looking remarkably like tropical cyclones develop over the Mediterranean Sea. At least nine times in the past forty years similar cyclones have been documented (Winstanley, 1970; Mayengon, 1983; Ernst & Matson, 1983; Mayengon, 1984; Rasmussen & Zick, 1987; Ziakopoulos & Marinaki, 1996; Malguzzi et al., 1998; Reale, 1998; Reale & Atlas, 1998). Some observational and numerical studies have already been carried out on the cyclone of January 1995 (Pytharoulis, 1995; Lagouvardos et al., 1996; Ziakopoulos & Marinaki, 1996; Blier & Ma, 1997). Vortices of much smaller size are observed more frequently; for example, see Alpert et al. (1994). However, their small size and short lifetime (a few hours) distinguish them from the class of cyclones discussed here. In this paper we undertake a close investigation of the characteristics of the Mediterranean cyclone of January 1995 and try to understand the factors that contributed to its genesis and development. Particular emphasis is given to the question of the extent to which the superficial resemblance of this system to a hurricane represents an underlying dynamical similarity. Our analysis is based on conventional and satellite data, and additionally makes use of simulations using the high-resolution mesoscale version of the UK Met. Office (UKMO) Unified Model. While conventional data are sparse over the sea, the availability of a successful numerical simulation makes an examination of the mesoscale structure of the system possible. Before presenting the results of our investigation, we consider first the evidence for the existence of such hurricane-like vortices from previous studies, and second the possible mechanisms that might contribute to their formation. 261 I Pytharoulis, G C Craig and S P Ballard 1.1. Review of similar cyclones in the Mediterranean Sea Winstanley (1970) describes a similar cyclone that developed over the central Mediterranean to the southeast of Malta on 23 September 1969 as the result of a Saharan depression and a mid-tropospheric cold cutoff cyclone. It dissipated over north-eastern Algeria after moving across north-western Libya and southern Tunisia. The system reached the intensity of a cyclonic storm and gale-force winds were recorded in its vicinity. The existence of a cloud-free area in the centre of the storm, the very strong surface winds and the heavy, widespread and prolonged rainfalls indicate that the system possessed the features of a tropical cyclone. Another vortex with tropical cyclone characteristics formed in the Mediterranean Sea in January 1982. This system appears to have had its genesis as an atypical Atlas Mountains lee depression on 23 January 1982 (Ernst & Matson, 1983). The disturbance was first detected over the sea north of Libya and dissipated over the far eastern Mediterranean, after passing near to Malta, Italy and Greece (Mayengon, 1983, 1984). Although the available observations in the vicinity of the cyclone were inadequate, they are consistent with the interpretation that the system was a small, intense, extra-tropical cyclone possessing some of the features observed in tropical cyclones. These were the spiral shape, the ‘eye’, the eyewall with the convectively driven cumulonimbus towers, the strong surface winds and the fact that the strongest winds corresponded to the eyewall surrounding the ‘eye’. Another similar cyclone was observed in the autumn of 1983 (Rasmussen & Zick, 1987; Mayengon, 1984). It was first detected on 27 September 1983 over the sea between Tunisia and Sicily and, after traversing a large circular path around Sardinia and Corsica, it finally decayed following landfall near Tunis in the morning of 2 October. During its lifetime the system made landfall twice: once on the south coast of Sardinia near Cagliari on 28 September and once in western Corsica, south of Ajaccio on 30 September. However, in each case the old low-pressure centre dissolved and a new one formed almost immediately over the sea. Baroclinic instability did not appear to have been important in the development of this cyclone. However, convection because of the high sea-surface temperature was decisive in its formation and evolution (Rasmussen & Zick, 1987). As in the previous cases, this cyclone exhibited tropical cyclone features, including the ‘eye’, the strong sustained winds near the centre of the storm, the strong convection with deep cumulonimbus towers in the eyewall and the existence of a warm core. The diameter of the cyclone (measured using the closed 1010 hPa isobar around the vortex center) was found to be between 200 262 and 300 km during most of its lifecycle, except after the second landfall (near Ajaccio) when it was reduced to about 100 km (Rasmussen & Zick, 1987). Furthermore, Rasmussen & Zick (1987) deduced that when the cyclone was better organised, regions of upper-level divergence and low-level convergence were almost above the location of the surface low, indicating that the vortex had had a vertical axis. Two other such cyclones were observed between 4 and 6 October 1996, and 7 and 9 October 1996 (Malguzzi et al., 1998; Reale & Atlas, 1998). The first formed over the sea between Sicily and Tunisia on 4 October 1996 and dissipated after landfall over Sicily and southern Italy. Floods were induced in Sicily and Calabria (southern Italy), and gusts up to 30 m s–1 were recorded in Calabria (Reale & Atlas, 1998). The second cyclone formed over the sea to the west of Sardinia on 7 October 1996. It weakened temporarily after landfall over Sardinia in the early morning of 8 October but regained its strength some hours later when it moved again over the sea, finally dissipating after landfall over Calabria. Damage caused by the strong winds associated with this cyclone was reported from the Aeolian islands (north of Sicily), and high precipitation amounts were recorded in both the Aeolian islands and Calabria (Reale & Atlas, 1998). Both systems exhibited the features of a tropical cyclone, such as a spiral cloud structure, strong convection, strong surface winds, and heavy precipitation. Additionally, an ‘eye’ appeared in the satellite images of the second system, and warm temperature anomalies were found in its core. Convection was important for the maintenance of both systems (Malguzzi et al., 1998; Reale & Atlas, 1998). Four more similar cyclones have been documented in the literature (but not in as much detail as those noted above). These cyclones appeared on 26 March 1983 (Ziakopoulos & Marinaki, 1996), between 29 and 30 December 1984 (Ziakopoulos & Marinaki, 1996), between 30 and 31 October 1997 (Reale, 1998) and between 5 and 8 December 1997 (Reale, 1998). All of them exhibited hurricane-like characteristics in the satellite imagery. 1.2. Mechanisms responsible for development Hurricanes intensify in response to release of latent heat in cumulus convection, which in turn is forced by surface fluxes of latent and sensible heat in a synoptic environment of low pressure, without or with weak baroclinicity. A number of theoretical descriptions of this process have been formulated, notably Conditional Instability of the Second Kind (CISK), introduced by Charney & Eliassen (1964), and Ooyama (1964), and Wind Induced Surface Heat Exchange instability (WISHE, but formerly referred to as Air–Sea Interaction Instability), described by Emanuel (1986). Mediterranean cyclone According to the Charney and Eliassen formulation of CISK, the latent heat release by the cumulus convection in a warm core vortex is proportional to the moisture convergence in the boundary layer of the storm. Therefore, a strengthening of the vortex will increase the low-level moisture convergence, which in turn will result in increased latent heat release and further strengthening of the system. Ooyama (1982) pointed out the importance of the sea surface fluxes for the intensification of the tropical cyclones, and Emanuel (1986) proposed (in the WISHE theory) that the latent heat release is governed by the wind dependent surface fluxes of heat and moisture from the underlying ocean. Recent evidence suggests that WISHE may be a more accurate theoretical model, but the structure of the weather systems they describe is similar (Craig & Gray, 1996). In mid-latitudes, however, most cyclones form as a result of baroclinic instability. Unless this mechanism can be ruled out, it cannot be established that the development of a Mediterranean cyclone is hurricane-like. A similar difficulty in establishing the mechanism of intensification has been noted for polar lows – smallscale cyclones that form during cold air outbreaks over the northern oceans (e.g. Rasmussen et al., 1992). It is generally believed that a spectrum of such systems can occur. At one extreme are small baroclinic disturbances that owe their small scale to low vertical stability and shallow depth. At the other extreme are hurricane-like vortices driven by surface fluxes, although sensible heat fluxes are as important as latent heat fluxes for polar lows (Rasmussen, 1979). In the case of Rasmussen et al. (1992), the cyclone showed both characteristics at different times in its development. In this study the data will be examined for the distinctive characteristics of both baroclinic instability and intensification driven by surface fluxes. In particular, baroclinic instability is the cooperative intensification of an upper-level potential vorticity anomaly, and an equivalent anomaly that develops on a low-level potential temperature gradient (Hoskins et al., 1985). Each of these features will be present in an intensifying baroclinic wave, typically with a westward tilt with height. On the other hand, strong surface fluxes will result in intense convection and a warm core vortex that shows little vertical tilt. The crucial dependence of a system on surface fluxes is difficult to establish from data but will be tested in numerical experiments. Description of the Mediterranean cyclone and of the synoptic situation in its vicinity using satellite images as well as surface and upper-air charts is the focus of Section 2. Section 3 gives an overview of the numerical simulation, while Section 4 presents a detailed analysis of the mesoscale structure of the cyclone. An analysis of the mechanisms contributing to the intensification of the disturbance is given in Section 5. 2. Overview of evolution of the Mediterranean cyclone 2.1. Synoptic environment prior to formation The cyclone that is the focus of this study formed in the morning of 15 January 1995 over the open sea west of Greece, close to the centre of a low of larger dimensions which had moved over Greece the previous day (Figure 1(a), (b)). This larger low had formed in the morning of 13 January over the central Mediterranean between Libya and Italy, and in the following days moved north-eastwards towards western Greece. Strong winds were associated with this low prior to the formation of the Mediterranean cyclone. The German research vessel Meteor reported sustained winds of 73 knots (37.6 m s–1) close to the centre of the low at 1400 UTC on 14 January (Blier & Ma, 1997). After the formation of the Mediterranean cyclone, the centre of the large-scale low continued moving eastwards along the coastline of Turkey towards Cyprus, decaying as it went (Figure 1(c), (d)). Throughout this period a trough was present in the mid-troposphere, extending southwards to northern Africa. On 13 January the trough axis lay over southern Italy and southern Tunisia (Figure 2(a)), whereas the next day the trough had moved eastwards and was aligned towards central Libya. There were two lowpressure regions in the mid-troposphere associated with this trough (Figure 2(b)): one over Ukraine (persisting there from the previous day) and the other, which formed on 14 January, located over the central Mediterranean. This latter low was probably connected with the low-level cyclone over western Greece at 1200 UTC (Figure 1(a)) that had intensified over the previous two days. The fact that the positive vorticity area (connecting the low-level cyclone and the mid-troposphere low) was almost vertical (as can be seen in Figures 1(a) and 2(b)) is a sign that the surface cyclone had reached the end of its intensifying phase. Indeed, 18 hours later the cyclone had filled 8 hPa and a new low-pressure centre had formed near the old one. This marked the appearance of the cyclone that is the subject of this study. 2.2. The formation of the Mediterranean cyclone A close examination of Meteosat infra-red (IR) imagery showed that the initial formation time of the Mediterranean cyclone was at about 0330 UTC on 15 January 1995 (not shown), over the sea between western Greece and Sicily. Formation time is considered to be the time that the cyclone first exhibited its hurricane-like features in the satellite imagery. About one hour earlier there was a hook in the cloud pattern marking the place where the vortex appeared in the 0330 UTC image. High clouds indicated that strong convection was present before the initial formation of 263 I Pytharoulis, G C Craig and S P Ballard Figure 1. The mean sea-level pressure analysis for Europe at (a) 1200 UTC on 14 January, (b) 0000 UTC on 15 January, (c) 0600 UTC on 15 January and (d) 1800 UTC on 15 January 1995. Isobars every 4 hPa. (UKMO Daily Weather Summary). the cyclone. Water vapour imagery sometimes gives early indications of the formation of a cyclone before it is visible in the cloud pattern (Weldon & Holmes, 1991). However, an investigation of the water vapour images for the day before the formation of the Mediterranean cyclone showed no evidence of the impending cyclogenesis. After its formation the vortex was clearly visible on the satellite images. It exhibited an axisymmetric pattern, a cloud-free vertical area in the centre corresponding to the ‘eye’ and spiral-shaped clouds around the ‘eye’. The first time that the cyclone was drawn on the UKMO six-hourly mean sea-level pressure subjective analysis charts was at 0600 UTC when its central pressure was 1002 hPa (Figure 1(c)). In the subjective three-hourly mesoscale mean sea-level pressure analyses of the Hellenic National Meteorological Centre (HNMC), the Mediterranean cyclone appeared at 0000 UTC on 15 January (Ziakopoulos & Marinaki, 1996). Comparisons 264 between the mean sea-level pressure analyses of the UKMO and the HNMC during the lifetime of the Mediterranean cyclone showed generally good agreement with regard to the location and central pressure of the cyclone. The two analyses usually agreed within 2 hPa, except for a few times near the last hours of the cyclone’s lifecycle. During the following hours the central pressure of the vortex increased and at 0000 UTC on 16 January it had reached 1013 hPa (in the UKMO six-hourly mean sealevel pressure analysis). The fact that the central pressure of the vortex was increasing (Table 1) is usually a sign that the surface winds are losing strength. However, this rule cannot be applied here since the system was a small-scale feature embedded in a large-scale environment of rising pressure. The only direct surface observations were those of some ships travelling in the vicinity of the cyclone. The reported winds near the Mediterranean cyclone Table 1. The actual central pressure of the Mediterranean cyclone (UKMO mean sea-level pressure subjective analysis) and its predicted values from 0600 UTC on 15 January to 0000 UTC on 18 January 1995 Forecast period and time/date T+6 T+12 T+18 T+24 T+30 T+36 T+42 T+48 — — — — Figure 2. 500 hPa geopotential heights (full line) and 1000–500 hPa thickness lines (dashed line) at: (a) 1200 UTC on 13 January and (b) 1200 UTC on 14 January 1995. The geopotential height contours and the thickness lines are in dam. Contour interval: 12 dam. (UKMO operational global analysis charts). vortex were generally not strong, being between 15 and 25 knots. The maximum value was 30 knots (15.5 m s–1), recorded by a ship located near 35° N, 20° E at 1200 UTC on 15 January. The threshold value of the maximum sustained surface winds for a tropical cyclone to be upgraded from tropical depression to tropical storm is 34 knots (Foley, 1995). Showers and thunderstorms were reported near the cyclone especially near the end of the day. Two ships in the vicinity of the cyclone reported winds of 17.5 m s–1 at 0000 UTC on 16 January (Lagouvardos et al., 1996). The satellite images of 15 January reveal a well-organised vortex with a clear ‘eye’ and cumulonimbi rotating anticlockwise around the core (Figure 3(a)). The existence of convective clouds near the ‘eye’, in the area corresponding to the eyewall, is a typical feature of tropical cyclones. Moreover, the visible band images show many low clouds to be present at a larger distance from the ‘eye’ (Figure 3(b)). After 1800 UTC the 0600 UTC 15/01/95 1200 1800 0000 UTC 16/01/95 0600 1200 1800 0000 UTC 17/01/95 0600 1200 1800 0000 UTC 18/01/95 Central pressure (hPa) Actual Predicted 1002 1004 1006 1013 1012 1010 1012 1012 1018 1016 1022 1022 — 1004 1006 1011 1014 1016 1019 1020 — — — — images exhibit a very tight vortex of smaller diameter than before (Figure 4), and which was becoming more axisymmetric. All that day, the convection was very intense, with the cloud-top temperatures of the highest cumulonimbus (as shown by the IR images) being below –50 °C (not shown), corresponding to very high cloud tops. At 0000 UTC on 16 January (Figure 4) the number of cumulonimbi had increased to four and were visible around the ‘eye’ (in the area corresponding to the eyewall), together with one additional big tower very close to the core of the storm. The cloud-top temperatures of these cumulonimbi were still colder than –50 °C (not shown), corresponding to deep convection. However, at that time, the ‘eye’ of the cyclone may have been tilting with height since the sea surface was not visible from above. 2.3. The track of the cyclone During the first day of its existence, the cyclone did not move significantly. From the time of its formation until the end of the first day (15 January) it remained positioned over open sea between Greece and Sicily but closer to Greece (Figure 5). This occurred even though the larger low-pressure area in which the cyclone was embedded was moving slowly eastwards. Probably the strong convection that was observed west of Greece, and which had favoured the development of the cyclone, prevented it from following the movement of the wider low-pressure area. Over the previous days a high pressure area had been building over central Europe, principally associated with a strong anticyclone that had moved across from the Bay of Biscay (between France and Spain). The interaction of the anticyclone with the low-pressure 265 I Pytharoulis, G C Craig and S P Ballard Figure 4. Infra-red satellite image from the NOAA polarorbiting satellites at 0000 UTC on 16 January 1995. The brighter grey-scales correspond to higher clouds. Figure 3. (a) Infra-red and (b) visible satellite images from the NOAA polar-orbiting satellites. Both at 1500 UTC on 15 January 1995. The brighter grey-scales correspond to higher clouds. area created strong pressure gradients and therefore strong northerly winds over the central Mediterranean. On 15 January the axis of the tightest isobars was lying north and east of the Mediterranean cyclone (Figure 1(c)) and the system experienced only a weak steering flow. However, after 1800 UTC on that day (Figure 1(d)) there was a strong north-easterly flow in the vicinity of the cyclone, forcing it to move south-westwards. 2.4. Landfall and decay On 16 January the cyclone first moved south-west towards Libya and then, after 1800 UTC when it was close to the Gulf of Sidra (in northern Libya), turned more southerly towards the middle of the gulf (Figure 266 Figure 5. The track of the Mediterranean cyclone between 0300 UTC on 15 January and 0600 UTC on 18 January 1995 derived from the satellite images. 5). At the end of the day, the cyclone was near the entrance of the gulf and its translation speed was reduced. That day showers and frequent thunderstorms were reported from northern Libya. At 1200 UTC on 16 January, a ship (9VYT) provided an observation from about 50 km north-northeast of the cyclone centre of winds blowing approximately eastsoutheast at 50 knots (25.75 m s–1) (Blier & Ma, 1997). At 0000 UTC on 17 January, the central pressure of the cyclone had increased only slightly (by 2 hPa) from its value 12 hours previously to 1012 hPa (Table 1). This pressure was almost the same as 24 hours previously. Mediterranean cyclone Later, the vortex continued moving slowly southwards (Figure 5) and at 0600 UTC the edge of its cloud pattern was over land. Finally, the ‘eye’ made landfall at about 1800 UTC on 17 January (Figure 5). By that time, the central pressure was 1022 hPa, significantly higher than it had been at midnight when the vortex was over the sea. Six hours later, at 0000 UTC on 18 January, a land station on the northern coast of Libya (inside the Gulf of Sidra) recorded winds of 30 knots. The satellite images of 16 January reveal that very strong convection was still present, at least at the beginning of the day. In the IR image of 0300 UTC (Figure 6), three cumulonimbi can be seen close to the ‘eye’, with cloud-top temperatures below –50 °C (not shown). Later, the convection was still strong but the cloud tops of the highest cumulonimbus were slightly warmer (–40 °C to –50 °C), corresponding to lower clouds than earlier. That day, the vortex appeared to be still tight but low clouds were apparent extending many miles to the north and north-west of the ‘eye’ and forming a solid layer (not shown). In the late evening of 16 January, the diameter of the high cloud region had increased and the vortex did not appear to be as tight as before. The next morning, when the cyclone was positioned well into the Gulf of Sidra, only a few high clouds were visible. The cyclone was mainly composed of low clouds, indicating reduced convection. Finally, some hours after landfall the hurricanelike cloud pattern began to disappear. Between 0600 and 1200 UTC on 17 January, when the cyclone reached the land, the pressure dropped (Table 1) as usually happens with hurricanes when they make landfall (Tuleya & Kurihara, 1978). The increased surface friction during landfall results in a decrease of the horizontal wind speed. This leads to an increase of the cross-isobar angle towards the low pressure. This enhanced inflow increases the mean mass convergence, increasing convection and resulting in a deeper low, perhaps accounting for the 30-knot winds observed earlier. However, without evaporation from the sea surface, the convection quickly dries out the boundary layer and the increased upward motion allows adiabatic cooling to dominate diabatic warming (Wakimoto & Black, 1994). Therefore, after the short intensification and deepening of the cyclone by 2 hPa during landfall, the central pressure began to increase significantly (Table 1). At 0600 UTC on 18 January, the cyclone had already begun to dissolve and by 1530 UTC no vortex could be identified in the Meteosat images. In conclusion, the Mediterranean cyclone formed at the centre of a synoptic scale depression. But as the larger low drifted eastwards, the small-scale vortex remained over water, eventually drifting southwards following the environmental flow, until it reached land and decayed. As with the examples cited in the introduction, the system showed a strong resemblance to a hurricane on satellite imagery. Figure 6. Satellite infra-red image from the NOAA polar-orbiting satellites at 0300 UTC on 16 January 1995. The brighter grey-scales correspond to higher clouds. 267 I Pytharoulis, G C Craig and S P Ballard 3. Overview of numerical simulation The discussion so far has raised a number of questions. In Section 1.2, the theories for the development of similar cyclones were described. A major question, then, is how this particular cyclone formed and what factors influenced its evolution. A further question concerns whether the cyclone had kinematic and thermodynamic structure akin to tropical cyclones. However, before examining the mesoscale structure of the simulated cyclone, it is necessary to verify as much as possible the accuracy of the model forecast. Since surface or upper air observations are not available on the scale of the system, the verification of the model forecast will be based on the comparison of analysed surface and mid-tropospheric charts and satellite images with the corresponding model results. A more detailed comparison of simulation and data is given by Pytharoulis (1995). 3.1. The model The model used in this study is the UK Met. Office (UKMO) Unified Model. A detailed description of the characteristics and configurations of the operational model is to be found in Cullen (1993), while a shorter description is given in the Appendix. In this study the global (1.25° × 0.83° horizontal resolution), limited-area (LAM) (0.4425° × 0.4425°) and mesoscale (0.15° × 0.15°) versions of the model were used, but the results were taken from the LAM and mesoscale forecasts, using as initial conditions the operational LAM analysis of 15 January 1995 at 0000 UTC (hereafter, this time will be referred to as T+0). At that time there was no indication of the presence of the small-scale Mediterranean cyclone. The global and LAM models use 19 levels in the vertical, while the mesoscale model has 31 levels. All have increasing resolution towards the surface. The sea surface temperature in all models was taken from a global (1.25° × 0.83°) resolution analysis. Also, a full set of parametrisations is included in the model. 3.2. Surface pressure and precipitation pattern In the previous section, the track and the central pressure of the Mediterranean cyclone were described. Here, the track of the system, according to the mesoscale model forecast, will be analysed and its central pressure and location at different times compared with the actual ones. In this section, references to the Mediterranean cyclone are to the system as it appeared in the model results. When the actual low is referred to, this will be stated explicitly. In the mesoscale model mean sea-level pressure forecast charts, which are extracted for every six hours, the vortex first appeared at 1200 UTC on 15 January (Figure 7(a)). Six hours previously at T+6 there was no 268 apparent vortex but the pattern of the inner closed isobar (not shown) indicated that the formation of a lowpressure centre over the sea between Greece and Sicily was probable. From its initial appearance until six hours later the system did not move significantly and was located over the sea west of Greece. However, at 1800 UTC on 15 January (T+18) the system began to move south-westwards. At 0600 UTC on 16 January (T+30, Figure 7(c)), it was located near the entrance of the Gulf of Sidra, and afterwards headed south towards the eastern side of the Gulf. It finally made landfall between 1800 UTC on 16 January (T+42, Figure 7(d)) and 0000 UTC on 17 January (T+48), after which it disappeared. The central pressure of the cyclone increased throughout its lifetime (Table 1). At T+12, when it initially appeared in the mesoscale model charts, its central pressure was approximately 1004 hPa. Twelve hours later at T+24 it had reached 1011 hPa, having increased by 5 hPa over the previous six hours. Later, it continued to increase and at T+42, just before landfall its pressure had risen to 1019 hPa. Over the next six hours, as the vortex made landfall, its central pressure increased by only 1 hPa. This reduction in the rate of the pressure rise may correspond to the transient intensification of the observed system (during landfall) described in Section 2.4. From a comparison of the location of the simulated vortex (Figure 7(a), (b)) with the location of the actual system it can be concluded that the model did very well for the first 24 hours. At 0000 UTC on 16 January (T+24) the separation between the forecast and the actual position of the centre of the vortex was less than 1 degree of latitude (about 111 km) (Figure 7(b)), and the predicted location of the cyclone was ahead of the actual one. Afterwards, the model moved the cyclone faster than it actually moved, although the error in position never became greater than 2 or 3 degrees of latitude (Figure 7(d)). A consequence of this error is that the model results are not useful from T+48 onwards since the cyclone had already made landfall (according to the model) whereas the actual system was still located over the sea and thus in a completely different environment. The forecast of the central pressure exhibited the same pattern as the separation. During the first hours (T+12 and T+18) there was no discrepancy between the actual and the predicted central pressure of the system (Table 1), and in the next 12 hours the discrepancy that appeared was no larger than 2 hPa. However, at later times this difference became significant since the central pressure in the modelled cyclone increased while the actual one remained almost the same. The pattern of the instantaneous rate of total precipitation (i.e. convective and dynamic precipitation), Mediterranean cyclone Figure 7. The mean sea-level pressure forecast and the instantaneous rate of total precipitation expressed in mm/hour, at (a) T+12, (b) T+24, (c) T+30 and (d) T+42 (from 0000 UTC on 15 January 1995). The isobars are drawn every 2 hPa. The scale below the panels corresponds to the rate of precipitation and is expressed in mm/hr. The actual location of the Mediterranean cyclone is indicated by the small cycle. expressed in mm per hour, is also depicted in Figure 7. Here, precipitation includes both rain and snow. The model almost always predicted a precipitation-free area at the centre of the cyclone while the heaviest precipitation was predicted to fall in the area of the cyclone corresponding to the eyewall. In the area close to the cyclone the greatest portion of the precipitation was produced by the convection scheme. However, the heaviest precipitation rates were sometimes due to the dynamic or ‘grid-scale’ precipitation scheme indicating that there is grid-scale saturation and/or that resolved scale motions are dominating the production of precip- itation in those areas. Unfortunately, there are no measurements of precipitation in the region of the vortex since it evolved over the sea. However, the precipitation pattern, as illustrated in Figure 7, seems to be very reasonable. With one exception, rainfall was always reported from ships in the vicinity of the cyclone. No rainfall was reported at 1200 UTC on 15 January by a ship located south of the vortex. From Figure 7 it is apparent that the model predicted no rainfall south of the cyclone at T+12, in agreement with this observation. 269 I Pytharoulis, G C Craig and S P Ballard 3.3. Conditional instability In Section 2 we mentioned that very deep convection was present on 15 January as well as in the first few hours of 16 January. In the following hours the IR images revealed that the convection was still strong but did not reach the same heights as previously. Here we will show that the observed cloud-top temperatures are consistent with the thermodynamic structure of the atmosphere predicted by the numerical model. This is achieved by constructing tephigrams based on a vertical cross-section of the temperature field taken in the area of the vortex (Figure 8). A location corresponding approximately to the area where deep convection would be observed (i.e. in the eyewall) was chosen and the temperature at each pressure level was established (starting at 1000 hPa and continuing for every 50 hPa to the pressure level of 200 hPa). Temperature profiles are shown in Figure 9 for 16 January at 0000 UTC, and for the same day at 1200 UTC when the convection was observed to have diminished. A prediction for maximum cloud-top height (and minimum cloud-top temperature) can be obtained by considering the pseudo-adiabatic ascent of an air parcel initially saturated at the sea surface temperature. The distribution of the sea surface temperature, which did not change significantly during the lifetime of the Mediterranean cyclone, is depicted in Figure 8. At the earlier times the sea surface temperature (SST) in the location of the vortex was determined to be about 16.4 °C whereas later it was approximately 17.5 °C. Figure 9. The two ‘ascents’ drawn on a tephigram. Full line: first ‘ascent’ (T+24 from 0000 UTC on 15 January). Dashed line: second ‘ascent’ (T+36 from 0000 UTC on 15 January). At 0000 UTC on 16 January the lapse rate was conditionally unstable up to the level of the tropopause at 350 hPa (Figure 9). Assuming a rising parcel of air to follow the saturated adiabat which is equal to the SST at this location, we find that the cloud top should be at about 280 hPa corresponding to a cloud-top temperature of –49 °C. The second ‘ascent’ (1200 UTC on 16 January, Figure 9) revealed the existence of conditionally unstable air at low levels as well as in the mid-troposphere with the temperatures at each level close to those of the previous ‘ascent’. However, the conditionally unstable region was bounded by a temperature inversion at approximately 450 hPa. Making the same assumption as before for a rising parcel of air, we find that the cloud top should reach the pressure levels of 300–310 hPa, with a temperature of about –41 °C. These results agree well with the IR images which revealed that the cloud-top temperatures of the highest cumulonimbus were below –50 °C, for the first case and between –40 and –50 °C for the second case. 4. Mesoscale structure of the Mediterranean cyclone Figure 8. The area of the first, 1 (at 0000 UTC on 16 January 1995) and second, 2 (at 1200 UTC on 16 January 1995) crosssections, the predicted position of the vortex (small cycle), the approximate locations of the ‘ascents’ (star ‘*’) and the distribution of the sea-surface temperature (in °C). Contour interval: 0.5 °C. The locations of the cross-sections in Figure 13 are depicted here by the lines (a) and (b). 270 The satellite images revealed that the Mediterranean cyclone had possessed features typical of tropical cyclones and polar lows. These included the cloud-free region in the centre corresponding to the ‘eye’, the non-symmetrically distributed deep convective clouds in the eyewall and the spiral bands of clouds around the ‘eye’. However, a detailed analysis of the structure of Mediterranean cyclone the simulated cyclone will help decide whether the Mediterranean cyclone was really hurricane-like. 4.1. Surface fluxes Two very important phenomena that occur in the region of the tropical cyclones and polar lows are the convection and the strong wind-induced surface fluxes of heat and moisture (Craig, 1995; Emanuel, 1986; Emanuel & Rotunno, 1989; Rasmussen, 1979; Rasmussen & Zick, 1987; Rotunno & Emanuel, 1987; and others). The hurricane-like distribution of convective precipitation has been noted previously. Strong surface fluxes of heat and moisture occurred close to the area of the vortex when it formed, and during its lifetime. The surface fluxes of sensible heat are presented in Figure 10 (the pattern of the latent heat is similar; not shown). The strongest fluxes were associated with the strongest winds on the right side of the direction of storm movement. In particular, just before landfall, at T+30 (Figure 10(b)) and T+36 the maximum sensible heat fluxes were observed south-west of the vortex (the vortex was moving southwards at these times). This is in agreement with the observations of hurricanes which show that the most strongly developed segment of the eyewall cloud and the strongest winds are usually found in the quadrant of the hurricane centred about 45° to the right of the direction of the storm movement in the Northern Hemisphere (Wallace & Hobbs, 1977). Generally, in the Northern Hemisphere, winds measured on the right side of a tropical cyclone tend to be stronger than those on the left because of the storm’s translation speed (Burpee, 1986). 4.2. Kinematic structure The vertical distribution of the field of the meridional component of the wind velocity (corresponding to the tangential velocity) for a cross-section in the area of the vortex is illustrated in Figure 11(a). The location of the cross-sections that appear in Figures 11 and 12 are marked in Figure 8, for the time T+36. The most apparent features of the azimuthal wind velocity are the decrease of the wind speed with height and the existence of much stronger winds in the western side of the vortex. The decrease of the wind speed with height indicates that the vortex had had a warm core. The change in the direction of the meridional wind from northerly to southerly (from west to east) at low levels in the area of the core, and thus a cyclonic circulation, is apparent in Figure 11(a). However, at upper levels (particularly near the 400 hPa level) anticyclonic circulation is apparent only through a weak horizontal shear of the meridional wind velocity. Probably the strong general cyclonic flow at upper levels did not allow the upper circulation of the Mediterranean cyclone to become clearly anticyclonic. Inspection of the diver- Figure 10. Surface fluxes of sensible heat for (a) T+12 and (b) T+30 (from 0000 UTC on 15 January 1995). Contour interval: 20 W m–2. Values larger than 200 W m–2 are not plotted. The scale below the panels is in W m–2. The predicted location of the cyclone is indicated by the letter H. gence field revealed that the upper-air flow was divergent only close to the vortex and the values of the divergence were usually near 5 × 10–5 sec–1. Tropical cyclones often reveal strong anticyclonic flow at upper levels; however, these are much more intense systems than the Mediterranean cyclone. Figure 11(a) also shows an increase of the absolute value of the meridional wind from the eastern side of the vortex to the western by about 9 m s–1. This is consistent with the stronger surface fluxes to the west of the cyclone, as discussed in the previous paragraph. 271 I Pytharoulis, G C Craig and S P Ballard Figure 11. Vertical cross-sections taken from the area marked in Figure 8 at T+36 (from 0000 UTC on 15 January 1995) for the fields: (a) meridional component of wind velocity (contours in m s–1), (b) zonal component of wind velocity (contours in m s–1), (c) vertical velocity, ω (=dp/dt) (contours in Pa/sec), and (d) wet bulb potential temperature, θw (contours in K). Values higher than –0.4, 0.4, –0.03, 281.9 are shaded in panels (a), (b), (c) and (d), respectively. Contour interval: 1.3, 0.7, 0.08, 0.25, respectively. Left edge: west; right edge: east. The approximate area of the core is indicated below each panel. The features of the tangential wind velocity shown in Figure 11(a), for the time T+36, appeared throughout the life of the cyclone. The wind speed decreased with height, revealing that the system had always had a warm core and the upper flow never became clearly anticyclonic. The difference in wind speed between the west and east sides of the system was generally about 10 m s–1, but early in its lifetime it was even larger. The vertical distribution of the zonal component of the wind velocity (corresponding to the radial velocity) is depicted in Figure 11(b). The features that appear for this component are also consistent with the observa- 272 tions of tropical cyclones. At low levels inflow towards the core of the system is observed. Convergence of the radial component of the wind would be expected under the eyewall (since ascent is expected there) with divergence under the area of the core since the air that descends (in the core) will be spread out horizontally at low levels. Indeed, these features are generally encountered in Figure 11(b), although the convergence of the air under the eastern side of the vortex is not clear. In the upper levels the outflow is apparent. The same features were also observed at the other times, but when the convection was more intense, the convergence of the air under the eyewall was much more pronounced. Mediterranean cyclone The pattern of vertical motions (Figure 11(c)) is consistent with the preceding discussion. Ascent due to strong convection was observed in the region around the core, corresponding to the eyewall, and descent in the area of the core. Moreover, the strongest convection and as a result the greatest vertical velocities were apparent on the western side of the vortex, where the strongest sea-surface fluxes were encountered. Though the same characteristics of the vertical motion can be seen at all times, the height that the vertical motions reached is not always the same. At the early times, when the convection was very strong, the upward motion in the eyewall reached very high levels. However, at later times this height was lower than before. At T+42 ascent was confined to the mid and lower troposphere. The descent of dry air in the core is illustrated in Figure 12, with the help of the field of the relative humidity. Low values of the relative humidity can be seen in the core whereas much larger values are found in the eyewall. ally observed in tropical cyclones (e.g. Jorgensen, 1984). Looking at north–south as well as west–east cross-sections of θw, we observed that early in its lifetime the system was slightly tilted. At T+12 it was tilted northwestwards with height (Figure 13; see Figure 8 for the location of the cross-sections) and at T+18 (not shown) it had had a slight southwards tilt. A possible explanation is that the system was not very strong at the early times and the ambient flow influenced its vertical struc- 4.3. Thermodynamic structure A very useful field that can be used in order to examine the region of the core and the potential for moist convection, is the wet bulb potential temperature, θw. The pattern of the vertical distribution of θw is depicted in Figure 11(d) for T+36 and is consistent with the observations that have been made in tropical cyclones for equivalent potential temperature, θe. The almost constant value of θw with height (near the core) indicates a lapse rate close to moist adiabatic. From the θw and ω (=dp/dt) fields, it was not possible to determine if there was an outward tilt of the eyewall with height as is usu- Figure 12. Vertical cross-section of the relative humidity (contours in %) at T+36 (from 0000 UTC on 15 January 1995). Values lower than 49% are shaded. Contour interval: 5%. The area of the cross-section can be seen in Figure 8. The approximate area of the core is indicated below the figure. Figure 13. Vertical cross-sections of the wet bulb potential temperature (contours in K) at T+12 (from 0000 UTC on 15 January 1995) for (a) orientation west–east and (b) orientation south–north. (The locations of the cross-sections are indicated by the lines (a) and (b) in Figure 8.) Values higher than 281.9 K in panel (a) and 282.1 K in panel (b) are shaded. Contour interval: 0.35, 0.45, respectively. The approximate area of the core is indicated below each panel. 273 I Pytharoulis, G C Craig and S P Ballard ture. Afterwards, when the system was much better organised, its core was almost vertical. 4.4. Summary After the presentation of the characteristics of the Mediterranean cyclone, it can be concluded that the system had many of the characteristics of the tropical cyclones. The warm core, the strong convection and the strong surface fluxes are the most important features of the environment of the hurricanes and they were encountered in the Mediterranean cyclone. Moreover, other significant features of the cyclone, such as the eyewall (with ascent into it), the fact that the strongest winds were found near the surface, the inflow at low levels and outflow at higher ones and finally the descent of relatively dry air in the core, support the initial assertion that the system was hurricane-like. However, some differences such as the weak upperlevel anticyclone, the small values of θw and the tilt of the ‘eye’ (the latter observed at the earlier stages of its lifetime) indicate that the Mediterranean cyclone was a weaker system. (Figure 14(a), (b)) shows very small values of θ for that part of the tropopause which was over the central and eastern Mediterranean. This is an indication of the trough seen in Figures 2(a), (b) (during the previous days) since the small values of θ correspond to lower levels of the troposphere. However, no small-scale anomaly of the size of the studied cyclone is observed on the iso-PV maps to be embedded in this trough. This is true not only for the first hours of its lifetime but also for later stages when the cyclone was more intense. While the large-scale trough determines the environment where the small-scale cyclone forms, it would not lead directly to cyclogenesis on that scale. In the case of subsynoptic cyclones, much smaller PV anomalies are required. Another way to investigate the upper-air flow in order to see if there is any anomaly that may interact with a low-level baroclinic zone is by looking for the existence of an upper-air jet. The existence of an upper-air jet in the region of a system may be decisive for its track 5. Mechanisms of intensification Section 1.2 presented the theories that have been formulated mainly to explain the development of the tropical cyclones and the polar lows but which might also be applied in hurricane-like systems such as the Mediterranean cyclone. Surface heat and moisture fluxes are believed to be responsible for the growth of tropical cyclones and to contribute strongly to the intensification of polar lows. However, in the extratropics where baroclinicity is important, the baroclinic theory may be also valid. Therefore, in the following investigation of the factors that influenced the development of the Mediterranean cyclone, both the air–sea interaction theory and the baroclinic theory must be examined. Moreover, the importance of the location of the subtropical jet will be discussed. The baroclinic theory states that the disturbances which arise due to baroclinic instability evolve through a cooperative interaction of an upper-level vorticity or potential vorticity anomaly with a low-level baroclinic zone. Hence, if the Mediterranean cyclone was a baroclinic disturbance two features should be found in its vicinity. The first is an upper-air anomaly over the location that the cyclone formed and the other is strong baroclinicity at low-levels. A complete and easily assimilated view of the upper tropospheric situation is provided by maps of potential temperature (θ) on the PV=2 PVU surface (1 PVU – potential vorticity unit – is equal to 10–6 m2s–1kg–1K) which follows the tropopause (Hoskins & Berrisford, 1988). The θ distribution on the PV=2 surface map 274 Figure 14. θ-distribution on the PV=2 surface at (a) T+6 and (b) T+12 (from 0000 UTC on 15 January 1995). Contour interval: 10 °C. The range below the panels corresponds to the potential temperature (in °C). The predicted location of the cyclone is indicated by the letter H. Mediterranean cyclone as well as for its intensity, mainly because of the vertical motions that are associated with the jet. Ascent occurs on the right of the jet entrance and on the left of the jet exit, while descent occurs on the other sides of the jet entrance and exit. Since such rising motions must imply vorticity stretching in the column below, cyclonic vorticity will tend to increase below the right flank of the jet entrance and the left flank of the jet exit (Uccelini & Johnson, 1977). When the Mediterranean cyclone formed, the subtropical jet was lying along the northern coastline of Africa, over northern Egypt and north-eastern Libya (Figure 15(a)). Afterwards, it moved slowly eastwards and its maximum wind speeds were reduced (Figure 15(b)). Since ascent is observed to the right of its entrance and to the left of its exit, and descent on the other sides, the presence of the jet-stream does not seem to have promoted the evolution of the Mediterranean cyclone. However, it may have played a role in inducing the eastward movement of the wider low-pressure area where the Mediterranean cyclone was embedded, since the low-pressure area was located below the left flank of the jet exit. In the first few hours of 15 January, although the parent low centre was decaying, it featured an occluded surface front that extended westwards, ending near the region where the vortex formed (Figure 1(b), (c)). Indeed, the model forecast for 0300 UTC on 15 January (T+3, Figure 16; obtained from the LAM) shows strong θ gradients at low levels over the northern Aegean Sea and Greece, extending north-east towards the Black Sea. The baroclinic zone extends with diminished intensity to the west of Greece. However, the Mediterranean cyclone formed at the southern edge of this zone and not inside it. It can be deduced that the Mediterranean cyclone was not primarily a baroclinic disturbance. The low-level baroclinicity does not appear to have been important since the cyclone formed at the edge of the weaker zone. Moreover, there was not any upper level feature capable of interacting cooperatively with the low-level baroclinicity. As a result, another mechanism was probably decisive in the development of the cyclone. The existence of strong wind-induced surface fluxes of heat and moisture is the most decisive part in the development of a cyclone according to the air–sea interaction theory. Even though we saw in Section 4 that the surface fluxes were important for this cyclone, stronger evidence is required in order to decide which one was the driving mechanism for the storm. An experiment was therefore carried out in which the surface latent and sensible heat and moisture fluxes were turned off and the evolution of the pressure and precipitation patterns with time was observed. If the cyclone developed then the surface fluxes of heat and moisture would not be decisive for the system. The results of this experiment are depicted in Figure 17. Although a low centre formed again (but at T+6) and moved southwards as did the actual system, it was less Figure 15. Isotachs of zonal wind (in m s–1) at the pressure level of 250 hPa for (a) T+6 and (b) T+24 (from 0000 UTC on 15 January 1995). Contour interval: 10 m s–1. Figure 16. θ-distribution (in °C) on the ‘η’ level eta = 0.8698 at T+3 (from 0000 UTC on 15 January 1995). Contour interval: 1 °C. The location of the actual cyclone is indicated by the hurricane symbol. 275 I Pytharoulis, G C Craig and S P Ballard intense than in the control experiment, and by T+36 (Figure 17(d)) no vortex could be identified. The precipitation was much weaker and at T+36 (1200 UTC on 16 January, Figure 17(d)) there was no precipitation at all. In a second experiment, the model was run with only the latent heating turned off (at all levels). The result (not shown) was that there was less development than in the control run but more than in the no surface heat and moisture fluxes run. This result points out the importance of the sensible heat fluxes in the development of the Mediterranean cyclone. This is in contrast to tropical cyclones, where latent heating is dominant, but consistent with polar lows where latent and sensible heating are thought to be comparable (Rasmussen, 1979). From the previous discussion and mainly from the ‘no fluxes’ experiment, it is obvious that the surface heat and moisture fluxes were extremely important for the development of the Mediterranean cyclone. Although the system formed initially when they were turned off, the lack of fluxes did not allow it to intensify, since the necessary energy did not exist. The expected structure of a baroclinic instability was not found even though Figure 17. The mean sea-level pressure forecast and the instantaneous rate of total precipitation expressed in mm/hour at (a) T+12, (b) T+24, (c) T+30 and (d) T+36 (from 0000 UTC on 15 January 1995) from the model run with no surface heat and moisture fluxes. The scale below the panels corresponds to the rate of precipitation and is expressed in mm/hr. 276 Mediterranean cyclone some baroclinicity was present at low levels. An interesting point from both experiments and especially from the ‘no fluxes’ run is that a vortex still formed initially. The strong pre-existing convection or the orography (Ziakopoulos & Marinaki, 1996) may be responsible for the initial development of the vortex. The strong north-easterly flow that existed over northern Greece prior to formation may have interacted with the high orography of western Greece in order to produce small-scale low-level vorticity anomalies on the lee side of the mountains (over the sea). 6. Conclusions The development of a hurricane-like cyclone over the Mediterranean Sea has been studied in this work. The major tool was the Unified Model of the UK Met. Office and particularly its limited-area and mesoscale versions. The Mediterranean cyclone formed on 15 January 1995, over the sea between Greece and Sicily but closer to Greece. Before the formation of the cyclone, a synoptic scale low was situated over the area and strong convection was observed. Afterwards, the larger low continued to move eastwards and decayed, but the vortex remained in a wider low-pressure area. Strong convective activity was associated with the Mediterranean cyclone during almost all of its lifetime. Its track was mainly influenced by the steering flow due to the parent low to the east and an anticyclone over central Europe. The system evolved for about three days and finally it made landfall in northern Libya at about 1800 UTC on 17 January 1995 and decayed. The satellite images and the model results gave strong evidence to support the idea that the Mediterranean cyclone was similar to tropical cyclones and the more convective polar lows. The most important features were the strong surface fluxes and, as a consequence, the strong convective activity, the warm core, the weak anticyclonic upper-level flow, the fact that the strongest cyclonic winds were blowing at the surface (consistent with the warm core structure of the cyclone) and the cloud pattern which was composed of spiral bands of clouds with an ‘eye’ in the centre. Finally, the theories which are valid for the development of tropical cyclones and polar lows were examined in order to find which factors influenced its evolution. Baroclinic processes were not found to be responsible for the development of the system even though it formed at the edge of a weak low-level baroclinic zone. The surface fluxes probably played the most important role in its development. The importance of the surface heat and moisture fluxes was verified by an experiment in which they were turned off and the system did not develop as much as in the control run. In a second experiment in which only the latent heating was switched off (at all levels), there was less development than in the control run but more than in the previous experiment. This result showed that the Mediterranean cyclone was similar to polar lows in which the sensible heat fluxes are comparable to the latent heat fluxes, in contrast to tropical cyclones, where latent heat fluxes dominate. The examination of the role of the nearby orography in the initial development of the cyclone is proposed for future work. The formation of hurricane-like systems in the Mediterranean Sea is not an unprecedented event. At least nine similar cyclones have been documented in the last few decades. The strong winds, heavy rain, very bad visibility and generally bad weather associated with these cyclones, as well as the fact that they form and evolve mainly over the sea, make them a great danger for ships travelling in their vicinity and for the contiguous coastal regions. These reasons, together with the fact that a large number of ships ply the Mediterranean Sea, support the idea that these systems should be studied in detail when they form. In particular, the meteorological offices of Mediterranean countries must use their fine mesh models to improve forecasts of the formation and the track of these cyclones. Indeed, such small-scale and rapidly evolving systems provide a very good opportunity for the meteorological offices of any country in the world to check the performance of their fine mesh models in difficult situations. Appendix. Description of the model The UK Met. Office Unified Model is a hydrostatic, grid-point model using spherical polar coordinates, which can be run in global or limited-area configurations on a regular latitude–longitude grid at any desired horizontal or vertical resolution. Hybrid sigma/pressure co-ordinates are used in the vertical (Simmons & Burridge, 1981). A conservative split-explicit integration scheme is used (Cullen & Davies, 1991) and cloud water and ice are included as prognostic variables, making allowance for fractional cloud cover, after Smith (1990). The co-ordinate pole can be rotated so that the equator runs through the area of interest, thus allowing more uniform resolution in limited-area configurations. The coordinates used for the pole of the LAM and the mesoscale models (of this study) are 30°N, 160°E and 52.5°N, 202.5°E respectively. The LAM forecast was a rerun of the operational forecast to provide extra diagnostics and to provide boundary condition data for a high resolution mesoscale forecast. The LAM uses a five-minute timestep, has resolution 0.4425° × 0.4425° (i.e. about 50 km × 50 km), has its own data assimilation sequence and uses lateral boundary conditions from a previous global forecast. The mesoscale forecast was run for an area over the 277 I Pytharoulis, G C Craig and S P Ballard central and eastern Mediterranean with 140 × 140 gridpoints and resolution of 0.15° × 0.15° using a timestep of 1.5 minutes. The initial conditions for the mesoscale forecast were taken from the LAM analysis interpolated onto the mesoscale grid. Both models were run to produce 48-hour forecasts. A full set of parametrisations is included in the model. The boundary-layer scheme is formulated in terms of conserved variables and so can represent both dry and cloudy boundary layers. It also includes a representation of non-local mixing (Smith 1994). Mixing coefficients are stability-dependent, using a moist Richardson number. Large-scale precipitation is calculated in terms of the liquid water or ice content of the cloud, and cooling of the atmosphere due to evaporation of precipitation is included (Smith, 1990; Gregory, 1995). The annual cycle of radiation is imposed on the model which employs an interactive radiation scheme. The cloud optical properties depend on the predicted cloud-water path and phase at both long wave (Senior & Mitchell, 1993) and short wave (Slingo, 1989). A mass-flux convection scheme with stability-dependent closure (Gregory & Rowntree, 1990) is included which incorporates a representation of deep convective down draughts (Gregory & Allen, 1991). Surface hydrology and a soil-temperature model are also included (Dolman & Gregory, 1992). The form drag of unresolved orography is represented via an effective roughness length for momentum which is calculated from the standard deviation and average slope of the unresolved orography in all versions of the model and a parametrisation of gravity wave drag is included in the LAM and global versions (Milton & Wilson, 1996). The LAM forecast used the same orography and surface characteristics as used operationally. The mesoscale forecast used values specially derived for its increased horizontal resolution. Soil and vegetation characteristics were derived from the 1° Wilson and Henderson-Sellers data, orography and its standard deviation from a 5′ global dataset, and surface roughness parameters from a 100 m dataset where available and 5′ data elsewhere. Acknowledgements We would like to thank G. Hutchinson for his help in extracting the results using the Hewlett-Packard workstation at the Joint Centre for Mesoscale Meteorology (JCMM) at the University of Reading, Nigel Roberts for his useful comments about the water-vapour satellite images, D. Ziakopoulos (head of forecasting at the Hellenic National Meteorological Centre) for providing us with the three-hourly surface subjective mesoscale analyses of the HNMC, O. Reale for providing some useful observations, and E. Rasmussen, R. Riddaway (editor of Meteorological Applications) and an anomymous reviewer for their helpful comments. The UK Met. Office provided us with the Unified Model and the Daily Weather Summary 278 charts. We are also grateful to the Graphics Department of the Met. Office for their assistance in the preparation of the figures. The main part of the work was done in the JCMM using the HewlettPackard workstation system. An animation of IR imagery of the studied Mediterranean cyclone can be accessed at the following Internet site: http://www.met.rdg.ac.uk/Data/Global/special.html References Alpert, P., Tsidulko, M. & Itziksohn, D. (1994). A shallow short-lived Meso-β scale cyclone over the Gulf of Antalia – a numerical study. In Preprints, 6th Conference on Mesoscale Processes, Am. Meteorol. Soc., 508–509. Blier, W. & Ma, Q. (1997). A Mediterranean Sea hurricane?. In Preprints, 22nd Conference on Hurricanes and Tropical Meteorology, Am. Meteorol. Soc., 592–595. Burpee, R. W. (1986). Mesoscale structure of hurricanes. In Mesoscale Meteorology and Forecasting (ed. Ray, P.S.), Am. Meteorol. Soc., 311–330. Buzzi, A. & Tibaldi, S. (1978). Cyclogenesis in the lee of the Alps: a case study. Q. J. R. Meteorol. Soc., 104: 271–287. Charney, J. G. & Eliassen, A. (1964). On the growth of the hurricane depression. J. Atmos. Sci., 21: 68–75. Craig, G. C. (1995). Radiation and polar lows. Q. J. R. Meteorol. Soc., 121: 79–94. Craig, G. C. & Gray, S. L. (1996). CISK or WISHE as the mechanism for tropical cyclone intensification. J. Atmos. Sci., 53: 3528–3540. Cullen, M. J. P. & Davies, T. (1991). A conservative splitexplicit integration scheme with fourth order horizontal advection. Q. J. R. Meteorol. Soc., 117: 993–1002. Cullen, M. J. P. (1993). The Unified Forecast/ Climate Model. Meteorol. Mag., 122: 81–94. Dolman, A. J. & Gregory, D. (1992). The parametrization of rainfall interception in GCMs. Q. J. R. Meteorol. Soc., 118: 455–467. Emanuel, K. A. (1986). An air–sea interaction theory for tropical cyclones. Part I: Steady-state maintenance. J. Atmos. Sci., 43: 585–604. Emanuel, K. A. & Rotunno, R. (1989). Polar lows as arctic hurricanes. Tellus, 41A: 1–17. Ernst, J. A. & Matson, M. (1983). A Mediterranean tropical storm?. Weather, 38: 332–337. Foley, G. R. (1995). Observations and analysis of tropical cyclones. Global Perspectives on Tropical Cyclones, WMO Technical Document No.TCP-38, World Meteorological Organization, 1–20. Flocas, A. A. (1994). A Course in Meteorology and Climatology. Zitis Press (Thessaloniki, Greece). 465 pp. (in Greek). Flocas, H. A. & Karacostas, T. S. (1996). Cyclogenesis over the Aegean Sea: identification and synoptic categories. Meteorol. Appl., 3: 53–61. Gregory, D. & Allen, S. (1991). The effect of convective scale down draughts upon NWP and climate simulations. In Preprints of the 9th Conf. on Numerical Weather Prediction, 14–18 October 1991, Denver, CO, USA. Gregory, D. & Rowntree, P. R. (1990). A mass flux convection scheme with representation of cloud ensemble characteristics and stability dependent closure. Mon. Wea. Rev., 118: 1483–1506. Gregory, D. (1995). A consistent treatment of the evapora- Mediterranean cyclone tion of rain and snow for use in large-scale models. Mon. Wea. Rev., 123: 2716–2732. Hoskins, B. J., McIntyre, M. E. & Robertson, A. W. (1985). On the use and significance of isentropic potential vorticity maps. Q. J. R. Meteorol. Soc., 111: 877–946. Hoskins, B. J. & Berrisford, P. (1988). A potential vorticity perspective of the storm of 15–16 October 1987. Weather, 43: 122–129. Jorgensen, D. P. (1984). Mesoscale and convective-scale characteristics of mature hurricanes. Part II: Inner core structure of Hurricane Allen (1980). J. Atmos. Sci., 41: 1287–1311. Lagouvardos, K., Kotroni, V., Nickovic, S. & Kallos, G. (1996). Evidence of a winter ‘Tropical Storm’ over southeastern Mediterranean: simulations with the Regional Atmospheric Modelling System (RAMS) and the ETA/NMC model. In Preprints, 7th Conference on Mesoscale Processes, Am. Meteorol. Soc., 53–55. Malguzzi, P., Chessa, P. & Buzzi, A. (1998). The role of surface heat fluxes in the development of a Mediterranean ‘Hurricane’. In Annales Geophysicae. Part II: Hydrology, Oceans & Atmosphere (Supplement II to Volume 16), EGS, C632. Mayengon, R. (1983). Cyclone in the Mediterranean, January 1982. Mariners Weather Log, 27: 141–143. Mayengon, R. (1984). Warm core cyclones in the Mediterranean. Mariners Weather Log, 28: 6–9. Milton, S. F. & Wilson, C. A. (1996). The impact of parametrized subgridscale orographic forcing on systematic errors in a global NWP model. Mon. Wea. Rev., 9: 2023–2045. Ooyama, K. V. (1964). A dynamical model for the study of tropical cyclone development. Geophys. Int., 4: 187–198. Ooyama, K. V. (1982). Conceptual evolution of the theory and modeling of the tropical cyclone. J. Meteorol. Soc. Japan, 60: 369–380. Prezerakos, N. G. & Flocas, H. A. (1996). The formation of a dynamically unstable ridge at 500 hPa as a precursor of surface cyclogenesis in the central Mediterranean. Meteorol. Appl., 3: 101–111. Pytharoulis, I. (1995). Simulation of a Mediterranean Cyclone. M.Sc. Dissertation. Department of Meteorology, University of Reading. Rasmussen, E. (1979). The polar low as an extratropical CISK disturbance. Q. J. R. Meteorol. Soc., 105: 531–549. Rasmussen, E. & Zick, C. (1987). A subsynoptic vortex over the Mediterranean with some resemblance to polar lows. Tellus, 39A: 408–425. Rasmussen, E. A., Pedersen, T. B., Pedersen, L. T. & Turner, J. (1992). Polar lows and arctic instability lows in the Bear Island region. Tellus, 44A: 133–154. Reale, O. (1998). Dynamics and classification of two sub-synoptic scale ‘Hurricane-like’ vortices over the Mediterranean Sea. In Annales Geophysicae. Part II: Hydrology, Oceans & Atmosphere (Supplement II to Volume 16), EGS, C634. Reale, O. & Atlas, R. (1998). A tropical-like cyclone in the extratropics. International Centre for Theoretical Physics preprint, Trieste, Italy. No. IC98007. Rotunno, R. & Emanuel, K. A. (1987). An air–sea interaction theory for tropical cyclones. Part II: Evolutionary study using a nonhydrostatic axisymmetric numerical model. J. Atmos. Sci., 44: 542–561. Senior, C. A. & Mitchell, J. F. B. (1993). Carbon dioxide and climate: the impact of cloud parametrization. J. Climate, 6: 393–418. Simmons, A. J. & Burridge, D. M. (1981). An energy and angular momentum conserving finite difference scheme and hybrid coordinates. Mon. Wea. Rev., 109: 758–766. Slingo, A. (1989). A GCM parametrization for the shortwave radiative properties of clouds. J. Atmos. Sci., 46: 1419–1427. Smith, R. N. B. (1990). A scheme for predicting layer clouds and their water content in a general circulation model. Q. J. R. Meteorol. Soc., 116 : 435–460. Smith, R. N. B. (1994). Experience and developments with the layer cloud and boundary layer mixing schemes in the UK Meteorological Office Unified Model. In Proceedings of ECMWF Workshop on Parametrization of the Cloud Topped Boundary Layer, 8–11 June 1993, ECMWF, Reading. Tuleya, R. E. & Kurihara, Y. (1978). A numerical simulation of the landfall of tropical cyclones. J. Atmos. Sci., 35: 242–257. Uccelini, L. W. & Johnson, D. R. (1977). The coupling of upper and lower tropospheric jet streaks and implications for the development of severe convective storms. Mon. Wea. Rev., 107: 682–703. Wakimoto, R. M. & Black, P. G. (1994). Damage survey of Hurricane Andrew and its relationship to the eyewall. Bull. Am. Meteorol. Soc., 75: 189–200. Wallace, J. M. & Hobbs, P. V. (1977). Atmospheric Science – An Introductory Survey. Academic Press, 467 pp. Weldon, R. B. & Holmes, S. J. (1991). Water vapour imagery: interpretation and applications to weather analysis and forecasting. NOAA Tech. Report NESDIS 57, 213 pp. Winstanley, D. (1970). The north African flood disaster, September 1969. Weather, 25: 390–403. Ziakopoulos, D. & Marinaki, A. (1996). Mesoscale Mediterranean vortices with characteristics of tropical cyclones. In Preprints, National Conference on Meteorology-Climatology-Physics of the Atmosphere, Athens, Greece, 154–159 (in Greek). 279