Tornadoes are the most violent hazardous phenomenon associated with convective storms. They have been historically observed in many places of the world, but the most frequent and best documented events occur in North America. Fujita (1973) presents a general distribution of tornado occurrences around the world. Maximum densities in his map are found in the United States, but other areas show also significant incidence (Bangla Desh, Japan, Australia, Argentina, South Africa, Western Europe), including some Mediterranean countries, as Italy. However, no references of tornadoes in Spain appear, in spite of existing ancient chronicles and some citations in newspapers about devastating damages produced by wind. In fact, tornadoes were considered a very rare phenomenon in this country because official observatories only archive those phenomena affecting them directly. Even today, tornadoes are not included in any standard Spanish climatological bulletin. As an illustrative example, Huerta (1984) presents a complete compendium of studies on Spanish meteorological aspects during this century, and only five references to tornadoes or waterspouts appear.
However, since 1975 evidences of more than 40 tornadoes in Spain have been collected by one of the authors of this paper (Gayá), who has also filed reports of waterspouts, funnel clouds, and downbursts. This appreciable tornado incidence has suscited [not an English word! I do not know what your intended meaning is.] some recent studies. Gayá and Soli[[section]]o (1996) present the characteristics (track, damage etc.) of three well-documented severe storms producing tornadoes and a downburst in the Balearic Islands, and an analysis of the thermodynamic atmospheric vertical structures, as well as meteorological synoptic situation in which they developed. Martin et al. (1996) focuses on a tornado that occurred in central Spain, with a detailed description of the meteorological situation and the convective system using remote sensing data. Ramis et al.(1997) present an observational study (based on satellite, radar and lightning data) of the evolution of a convective system producing a tornado in the northeast of Spain.
More specifically, 22 of the 40 tornadoes were registered in the Balearic Islands, where this study is focused. The islands are located in the Western Mediterranean (Fig. 1), and have an approximate area of 5000 km2. The first objective of this work is to present a preliminary climatology of tornadoes in the Islands (section 2). Further, we select 10 tornado days in which 16 well-documented tornadoes occurred. Working with soundings launched in Palma (see Fig. 1), we investigate the atmospheric vertical structures of those days (section 3). In section 4, we characterize the synoptic patterns in which the tornadic thunderstorms developed. Focusing on representative meteorological situations, we investigate secondary circulations and meso-[[alpha]] ingredients to determine the zones where supercells could have been favoured. Finally, section 5 contains the conclusions.
During 1989-1996, 22 tornadoes have been recorded in the Balearic Islands on 15 different days. Sixteen tornadoes were surveyed from land and four of them also from the air. They are listed in Table 1. Note that more than two tornadoes occurred on two dates. The database includes touchdown date and time (~20 minutes error), touchdown and liftoff points in terms of latitude and longitude or U.T.M. coordinates (~0.5 km error), and estimated intensity following the F-scale (Fujita 1981). Such information for the other 6 tornadoes is incomplete. Moreover, other phenomena also have been included in the database. In particular, 33 waterspouts, 13 funnel clouds, and several downbursts and severe gust fronts.
Table 1. Closest urban area, date, intensity, and path length for 16 tornadoes in Balearic Islands.
Most tornadoes run from SW to NE, three from SE and other three from NW (Fig. 1). Almost all tornadoes began the coast; therefore it is possible they started as waterspouts. Observe that the mountainous northwest and northeast of Mallorca, its southern part, and central Ibiza have no reported tornadoes. However, in thinly populated areas or with scarce forest, tornadoes are more difficult to recognize. Urbanized areas have more elements to infer presence and intensity of tornadoes, and therefore tornadoes frequency and intensities increase (Grazulis 1993). Figure 2A shows that tornadoes had a path length of 3-9 km, with only one exceeding 15 km. The average path length is 6.2 km and the standard deviation 3.9 km. Such path length is similar to the findings by Turner and Snow (1993) in Indiana (USA), but it is longer than that calculated by Livingston and Schaefer (1993) for the whole United States.
In order to assess the occurrence of tornado-producing thunderstorms as a function of diurnal heating, touchdown times have been converted to normalized solar time relative to Palma. Figure 2B displays its distribution. A maximum in early afternoon and two minima in the morning and late afternoon appear. This configuration agrees with the findings by Richard (1988) for Southeastern United States, and by Blier and Batten (1993) for California. A secondary maximum appears also in our case, but the distribution is dominated by the two multiple episodes "E" and "O" that occurred during the dark. It is likely that our database is too small yet to get a reliable indication of the climatological distributions.
There are two maxima in the monthly relative distribution (Fig. 2C), one in the late summer and autumn, and another one in spring. Tornadoes, waterspouts and funnel clouds distributions (same figure) show exhibit maxima in the fall season. Waterspouts and funnel clouds exhibit a secondary maximum in winter, but tornadoes have not been observed in winter.
Figure 2C also shows the monthly relative distribution of the 1019 thunderstorm days observed in the Balearic Islands during the 1989-1996 period. A thunderstorm day is when at least one of the 180 meteorological daily reports includes thunder. Although human observations of thunder have a number of problems, Gonz+lez (1996) found good agreement between observed thunderstorms and cloud to ground lightning data.
As revealed in Fig. 2D, violent tornadoes have not been observed, and only one was rated strong (F-3). This intensity distribution is consistent with that in North America (Livingston and Schaefer, 1993), where around 70% of all tornadoes are classified as weak. The F-0 tornado frequency in Balearic Islands seems to be notably lower than F-1 and F-2 frequency. The infrequency of observed F0 tornadoes is a good indication that many brief, weak events are not being reported as tornadoes.
In spite of the limited period considered, the overall tornado occurrence relative to the area seems to be as high as in Oklahoma or Texas (Grazulis et al., 1993), and much higher than in other Mediterranean countries (Dessens and Snow, 1993). Fortunately, however, there have been no fatalities, but damages can occasionally reach several millions U.S. dollars, as on 12 September 1996 when the outbreak "O" case affected touristic and industrial zones. The F-3 tornado on 8 October 1992, only produced devastating damages in a forested zone.
In this section the thermodynamic vertical structures observed when tornadoes occurred are analyzed using soundings taken in Palma. We considered the sounding just prior to the tornado touchdown. Given the reduced spatial dimensions of the Balearics, these soundings are considered representative of the existing air mass. However, since we are limited by 12 h sounding frequency (00 and 12 UTC), it is possible that the proximity sounding conditions (Brooks et al., 1994) are not reached in all cases.
First of all, individual soundings have been compared against with corresponding daily mean sounding (distinguishing between 00 and 12 UTC). These climatological soundings were calculated by linear interpolation between the monthly mean soundings (assumed valid on day 15), following the method used by Ramis (1977). Figure 3 shows temperature anomalies and dew point depressions of the 10 tornado days. Half of the cases are appreciably colder than climatological values at all levels. The other are only slightly warmer, except case "H", which is very warm between 1000 and 600 hPa. This result is somewhat surprising; we expected to find thermodynamic environments with warm lower tropospheres and cold temperatures at middle and upper tropospheric levels. Nevertheless, an overall minimum of the temperatures is seen about 600 hPa. The same figure shows that the majority of cases exhibit high humidities in the lower troposphere, and that only three of the cases show a notable midtropospheric dry layer.
In addition, we calculated some stability indices: lifted parcel (LP), lifted index (LI), Showalter (SH), potential wet bulb index (PI), K index (K), total totals (TT), WINDEX (WI) (see Tudurdeg. and Ramis (1997) for a review). Figures 4a and b summarize, using boxplots, the values obtained. Except TT, which concentrates about 48, all indices show strong dispersion among the cases.
Davies (1993) found a useful relation between Convective Available Potential Energy (CAPE) and Storm Relative Helicity (SRH) to identify supercell environments: those for which the Energy Helicity Index (EHI = CAPE x SRH /160000) is greater than 1 are considered favorable for supercells. As can be observed in Fig. 5, only three of the cases lie over the EHI=1 curve (the outbreak event "O", the F3 case "D", and the longest path tornado "B"). It should be noted that for the computation of SRH, the storm velocity has been estimated using the 30R75 rule (Davies and Johns, 1993). No radar is available at present in the Balearic area, and therefore thunderstorms velocities have to be estimated. It appears that many of the tornadoes in the Balearics may not be associated with supercells, but the most noteworthy events seem to occur with supercell thunderstorms.
Figure 6 shows SRH versus ground-relative helicity (H). H can be considered as an integrated value of temperature advection (in this case over the lowest 3000 m) (Tudurí and Ramis, 1997). SRH expresses the same concept but with respect to a storm-relative coordinate system. As shown in Fig. 6, seven of the cases are characterized by positive integrated thermal advections (H>0), although not very intense in general. Observe also that the general effect of including the storm motion is to increase the helicity.
Finally, a classification of the thermodynamic environments was attempted using cluster analysis (the hierarchical Ward's method with Euclidean distances (Everitt, 1980)). As variables, we used temperature anomalies and dew point depressions at 50 hPa intervals for 1000-100 and 1000-500 hPa respectively, precipitable water (1000-500 hPa), CAPE, CAPN (convective inhibition energy), SRH and H. These variables were standardized to avoid different weighting in the clustering process. Two well-defined clusters emerged. One of the groups contains those cases ("B", "D", "H" and "O") which contrast with the rest because, on the average, they have warmer tropospheres, are more humid at low and upper levels, and have smaller values of CAPE and CAPN but larger values of H and precipitable water.
In this section, synoptic patterns associated with the 10 selected tornadic days (Table 1) are examined. We have used ECMWF gridded analysis, given at 00, 06, 12 and 18 UTC with a resolution of 0.75 deg of latitude . The discussion focuses on the closest previous analysis to the tornado touchdown time. First of all, considering a limited area around Balearic Islands, we have diagnosed the presence and intensity of some ingredients favouring convection: presence of a low-level jet (LLJ) and jet streak (JS) aloft; quasi-geostrophic forcing at 925 and 500 hPa (QGF925, QGF500); water vapour flux convergence in the 1000-850 hPa layer (WVFC) (McNulty 1995). In addition, regarding each grid point as a sounding, we have calculated the spatial distribution of CAPE and SRH in order to identify broad areas where supercells could have been likely.
In all 10 cases we could clearly identify a JS over the western Mediterranean. In general, the islands appeared located in any of the JS-relative quadrants with associated ascending motion (the left exit and right entrance regions). Four cases presented a well defined LLJ (speeds greater than 20 kt at 925 hPa), leading to the strongest values of WVFC (200-400 g m-2 s-1) over the Islands. These cases are "B", "D", "H", and "O" which in fact were also identified as a cluster in section 3. In addition, our most relevant cases (F3 case "D", and outbreak case "O") presented also the most significant warm advection by their southeasterly jets. As a consequence, these cases have the maximum positive values of QGF925 (15 and 25 x 10-18 m kg-1 s-1, respectively). At 500 hPa, quasi-geostrophic forcing was positive, although marginal (less than 20 x 10-18 m kg-1 s-1) for all cases. Most cases presented CAPE greater than 1000 J kg-1, except "A", "C", "F" and "K" that gave less than 500 J kg-1. Observe from Table 1 that these four cases were associated with weak tornadoes. Again, "D" and "O" were characterized by the most significant CAPE values, reaching 2000 J kg-1.
The next step was to attempt a classification of the meteorological situations, based on the geopotential wave patterns at 500 hPa. The classification criterion was the correlation between maps within an elliptic area centered at (40 N, 0 E), with 15 and 10 [15 and 10 what? I don't know what this means!] for W-E and S-N semiaxes, respectively. We selected that area shape to avoid undesirable contributions to the correlation coefficient from distant grid points. In addition, the selected window extends further along the W-E direction in order better to encompass the geopotential waves. Following the technique described by Esteve (1984), and starting with a 0.8 correlation level, four groups emerge. The first group ("A", "D", "E", "K", and "O",) is characterized by a positively tilted trough with the low located in the southern half of the Iberian Peninsula, and strong southwesterly flow over the Balearic Islands (Fig. 7a). In the second group ("F" and "G"), the low is located at higher latitudes, and the flow over the islands is essentially from the west (Fig. 7b); "C" and "B" form the third group, and their waves are characterized by slightly negative tilting. In this case the flow is southwesterly as in the first group (Fig. 7c). Finally, "H" defines a distinct group owing to the shortness of the wave (Fig. 7d). Despite the above classification, we are aware that each case has its own particularities. However, we have selected "O", "G", "C" and "H" (Fig. 7) as group representatives to be discussed in more detail:
At low levels (figure not shown), the "O" case exhibits a closed cyclonic circulation occupying all the western Mediterranean, with a warm air tongue extending from Tunisia toward the Gulf of Lyons. The system intensity at upper levels is also quite substantial. Figures 8a, b show isotachs and potential vorticity (PV) at 300 hPa, as well as vertical velocity at 500 hPa. A JS with peak values exceeding 90 kts can be identified to the southeast of an isolated PV maximum (more than 2 PV units). Vertical velocity at 500 hPa is strongly upward in the left exit region over the Mediterranean coasts of Spain, precisely where the PV advection is strongest. In this area, deep convection developed as early as the afternoon of the previous day, with extraordinary rainfalls (up to 600 mm daily amounts) in the Valencia region. This episode is also noteworthy because during its final stage on 12 September, a small cyclone with tropical characteristics developed near Ibiza and crossed Mallorca (Gili et al,. 1997). SRH and CAPE for this case (Fig. 9a) are both quite appreciable. The maxima are found east of the Balearic Islands (150 m2 s-2 and 4000 J kg-1 respectively), but significant values overlap in the vicinity of the Balearics.
All the above described structures for the "O" case are also observed with the "D" case (not shown). Recall that these cases are the strongest ones in the dataset. In addition, some other common aspects distinguish them from the remainder: in particular, satellite pictures reveal long-lived MCSs over the western Mediterranean in both cases, whereas convection was localized and short-lived for the other episodes.
"G" case does not show a low-level cyclone over the western Mediterranean. Rather, a weak low is located over southeastern France, producing northerly winds over the Balearic Islands. The advection is slightly cold in this case, and satellite pictures show that the convection is embedded in a cold front moving from NW to SE (figures not included). An isolated PV maximum (more than 4 PV units) is observed over southeastern France, but the JS (centered in the Gulf of Lyons) does not show appreciable upward motion over the Balearics (lying in its right entrance quadrant). Important SRH and CAPE values overlap over the Islands, but their maxima are found to southeast of the Balearics (see Fig. 9b).
For the "C" case, again, no cyclone is observed at low levels in the area of the western Mediterranean. In fact, relative vorticity is slightly negative over the Islands. Winds are weak from the west, advecting slightly warmer air from the heated Iberian Peninsula. Isotherms at low levels indicate a foehn effect warming the surface air over the eastern coasts of Spain (figures not included). In addition, both CAPE and SRH exhibit low values (less than 500 J kg-1 and 30 m2 s-2 respectively; Fig. 9c). It seems, therefore, that low-levels synoptic ingredients are rather marginal in this case. However, in the middle and upper troposphere, the conditions are quite favorable, as Figs. 8c, d show. Positive vertical velocities at 500 hPa, associated with a well marked JS and accompanying tropopause folding are shown in the ECMWF data close to the Islands.
As commented before, the "H" case was characterized by a short baroclinic wave that experienced rapid development as it moved from northern Spain toward the western Mediterranean. At 300 hPa, PV exceeds 5 PV units in this case, and the JS is located to the south of the western Mediterranean with the Balearic Islands in its left exit region. As a consequence, intense upward motion is induced (see the sequence of Figs. 8e and f). A well defined low-level cyclone develops over the western Mediterranean as in "O" case. In addition, a warm air tongue extending from north Africa toward the Gulf of Lyons is also present. As a result of that configuration (figure not shown), thermal advection at low levels is highly positive east of the Balearics, but negative in the west. Such a dipolar structure is sufficiently strong to appear reflected in the SRH field (Fig. 9d). Nevertheless, over the Islands the SRH is positive (about 60 m2 s-2), and CAPE attains 1500 J kg-1.
Note that there are discrepancies between ECMWF-derived CAPE and SRH values over the Islands, and those calculated from the soundings in last section. This is attributable to several reasons: the analyses respond for the large scale distribution of CAPE and SRH, whereas soundings capture the local conditions (including earlier convection effects); CAPE and SRH are characterized by sharp gradients (see Fig. 9), and therefore their values become very sensitive to spatial and temporal shifts. These problems were already discussed by Brooks et al. (1994).
In spite of the small existing tornado database for the Balearic Islands (8 years), this area is reveals to be a real "tornado alley." The most violent observed tornado has attained F3 intensity, but the most frequent intensity is F1. All reported tornadoes occur between June and November, but there is a clear preference for the autumn season, in which rainfalls also are the most substantial and generally of convective origin. Typical path lengths are 2-8 km but an exceptional 15 km path has been observed. The normalized touchdown time shows that there is a preference for early afternoon; however, the number of cases is too limited to extract clear conclusions.
The analysis of the thermodynamic vertical structures of the air masses in which the tornadoes develop in the Balearics shows that most of them occur in cold air relative to the mean climatological temperature. Humidity is very high at the lowest levels but in most of the cases there was a dry layer somewhere from 800 - 600 hPa. Stability indices, except TT, exhibit large dispersion and may not be very useful. For many of the events, the instability of the air masses, represented by CAPE, generally is not very high, and SRH was also low (perhaps negative for some cases). This suggests that many of the tornadoes in the Balearics are non-supercell events. Only three cases had EHI greater than 1, and therefore supercell environments can be inferred.
A global view of the meteorological situations in which tornadoes occurred (as given by ECMWF) shows that, in general, the tornadic thunderstorms developed as the cold air advection appeared over the region, rather than within a warm airmass. However, the advection relative to the storm typically is positive, as shown by the SRH values. In the middle troposphere, a low affecting the western Mediterranean is generally present. Influence from upper levels seems to be very important, since a JS affecting the Balearic Islands can be identified in all cases. In addition a tropopause folding is clearly identified on the spatial structure of PV at 300 hPa.
Acknowledgments . This work has been partially supported by DGICYT grant PB94/1169-CO2-C2.The authors acknowledge the support provided by J. A. Guijarro and M. A. Heredia in data analysis tasks.
Blier W. and K. A. Batten (1993). Climatological analysis of California tornadoes. Preprints, 17th Conf. on Severe Local Storms, St. Louis. MO. Amer. Meteor. Soc., 15-19.
Brooks, H. E., C. A. Doswell III and J. Cooper (1994). On the environments of tornadic and non tornadic mesocyclones. Wea. Forecasting, 9, 606-618.
Davies, J. M.(1993). Hourly helicity, instability and EHI in forecasting supercell tornadoes. Preprints, 17th Conf. on Severe Local Storms. Saint Louis. MO. Amer. Meteor. Soc., 107-111.
Davies, J. M. and R. H. Johns (1993): Some wind and instability parameters associated with strong and violent tornadoes. 1. Wind shear and helicity. The Tornado : Its Structure, Dynamics, Prediction and Hazards. Geophys. Monogr. 79, Amer. Geophys. Union. 573-582.
Dessens J. and J. T. Snow (1993): Comparative description of tornadoes in France and United States. The Tornado : Its Structure, Dynamics, Prediction and Hazards . Geophys. Monogr. 79, Amer. Geophys. Union. 427-434
Esteve, J. (1984): Contribucicentsn al estudio de las situaciones de gran transparencia en Barcelona. Ph. D. Thesis. Univ. Of Barcelona.
Everitt, B. (1980): Cluster Analysis . Halsted Press, 136 pp.
Fujita T. T. (1973): Tornadoes around the World. Weatherwise, 27, 56-62.
Fujita T. T. (1981): Tornadoes and downbursts in the context of generalized planetary scales. J. Atmos. Sci., 38, 1511-1534.
Gayá, M. and A. Soli[[section]]o (1996): Caps de fiblcents (trombas y tornados). Algunas observaciones recientes. Preprints III Simposio Nacional de Prediccicentsn. Instituto Nacional de Meteorologdeg.a, Madrid. 19-25.
Gili M., A. Jansa, J. Riesco and J. A. García-Moya (1997): Quasi-tropical cyclone on 12th Setember 1996 in the Balearics. INM/WMO International Symposium on Cyclones and Hazardous Weather in the Mediterraneean. April 14-17, 1997 (elsewhere in this volume).
Grazulis (1993). A 110-Year perspective of significant tornadoes. The Tornado: Its Structure, Dynamics, and Hazards . C. Church, Ed. Amer. Geophys. Union. 467-478.
Grazulis T. P., J. T. Shaefer, and Robert F. Abbey (1993): Advances in tornado climatology, hazards, and risk assement since Tornado Symposium II. The Tornado : Its Structure, Dynamics, Prediction and Hazards. Geophys. Monogr. 79, Amer. Geophys. Union. 409-426
Gonzalez J. (1996): Distribucicentsn de las tormentas en Baleares. Boletdeg.n Climatolcentsgico Islas Baleares. Vol. 53 Nß13. Centro Meteorolcentsgico Territorial de Baleares.
Huerta F. (1984): Bibliografdeg.a meteorolcentsgica espanla. Instituto Nacional de Meteorologdeg.a. Madrid. pp 458.
Livingston, R. L. and J. T. Shaefer (1993): County-by-county data on strong and violent tornadoes. Preprints 17th Conf. on Severe Local Storms. St. Louis, MO., Amer. Meteor. Soc., 6-9.
Martin, F., L. de Esteban and R. Riosalido (1995): Estudio del tornado de SigÅenza. STAP, nota tÇcnica nß 25. Instituto Nacional de Meteorologia. Madrid.
McNulty, R. P. (1995): Severe and convective weather: A Central Region forecasting challenge. Wea. Forecasting, 10, 187-202.
Ramis C. (1977): Contribucicents en a la climatologia de aire superior en Mallorca. Universidad de Barcelona. 55 pp.
Ramis C., J. Ar[[sterling]]s, J.M. Lcentspez, and A. Mestres (1997): Two cases of severe weather in Catalonia (Spain). An observational study. Meteor. Appl . (in press).
Richard A. (1988): Tornado/severe thunderstorm climatology for the Southeastern United States. Preprints, 15th Conf. on Local Severe Storms, Baltimore, MD., Amer. Meteor. Soc., 511-516.
Tuduri. E. and C. Ramis (1997): On the environments of significant convective events in the western Mediterranean. Wea. Forecasting . (in press).
Turner K. D. and J. T. Snow (1993): Indiana tornadoes revised: analysis of path data. Preprints, 17th Conf. on Severe Local Storms . St. Louis, Mo. Amer. Meteor. Soc., 42-45.
Fig. 1. The Balearic Islands. The sites mentioned in the text are indicated. The paths of the tornadoes referenced in Table 1 have been drawn (the tornado reference has been put at the liftoff point). Other tornadoes with unknown path have been also indicated by the "][" symbol at approximated touchdown point. The triangle indicates the radiosonde station in Palma.
Fig. 2. (A) Relative distribution of path lengths for the 16 tornadoes listed in Table 1. (B) Relative distribution of their normalized touchdown times. (C) Monthly relative distribution of thunderstorm days (1019), tornado days (15), and waterspout and funnel cloud days (44), during the period 1989-1996. (D) F-Scale distribution of the 16 tornadoes listed in Table 1.
Fig.3. Temperature anomalies (from 1000 to 100 hPa), and dew-point depressions (from 1000 to 500 hPa), for the 10 tornado days listed in Table 1.
Fig. 4. Boxplots of (a) LP, LI, SH, PI, and (b) K, TT, WI stability indices, calculated from the soundings of the 10 tornadic days (Table 1).
Fig. 5. Convective Available Potential Energy (CAPE - J kg-1), and Storm Relative Helicity (SRH - m2 s-2) for the 10 tornado cases (Table 1). Curves EHI=1 and EHI=2 have been included.
Fig. 6. Storm Relative Helicity (SRH) versus Helicity (H), for the 10 tornado cases (Table 1).
Fig. 7. Synoptic patterns at 500 hPa for (a) "O" case (12 September 1996 at 00 UTC), (b) "G" case (8 August 1995 at 12 UTC), (c) "C" case (3 June 1992 at 12 UTC), and (d) "H" case (22 November 1995 at 06 UTC). Solid line indicates geopotential (gpm), and dashed line temperature ([[macron]]C).
Fig. 8. Evolution of PV (thick solid line, in PV units) and isotachs (thick dashed line, in m s-1) at 300 hPa, and vertical velocity at 500 hPa (positive values contoured as thin solid line, and negative as thin dashed line, in cm s-1). (a) and (b) for "O" case (at 00 and 06 UTC respectively). (c) and (d) for "C" case (at 12 and 18 UTC). (e) and (f) for "H" case (at 06 and 12 UTC).
Fig. 9. As in Fig. 7 but for CAPE (solid line) and SRH (dashed line).