In this study, an updated and extended climatology of cyclonic tracks affecting the eastern Mediterranean region is presented, in order to better understand the Mediterranean climate and its changes. This climatology includes intermonthly variations, classification of tracks according to their origin domain, dynamic and kinematic characteristics, and trend analysis. The dataset used is the 1962–2001, 2.5° × 2.5°, 40-yr European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-40). The identification and tracking of the cyclones was performed with the aid of the Melbourne University algorithm. It was verified that considerable intermonthly variations of track density occur in the eastern Mediterranean, consistent with previous studies for the entire Mediterranean, while further interesting new features have been revealed. The classification of the tracks according to their origin domain reveals that the vast majority originate within the examined area itself, mainly in the Cyprus area and the southeastern Aegean Sea, while the tracks that originate elsewhere most frequently enter from the west. Deeper cyclones follow the southwest track originating from the area between Algeria and the Atlas Mountains. A greater size characterizes the westerly tracks (southwest, northwest, and west), while the northwest tracks propagate faster over the study area. A negative trend of the track frequency was found on an annual basis that can be mostly attributed to the winter months, being associated with variations in the baroclinicity. This negative trend is more prominent for the westerly and northeasterly tracks, as well as for those originating in the northern part of the examined area.
The source and path of extratropical cyclones, namely, the cyclonic tracks, strongly influence the climate in midlatitudes, especially the precipitation regime. Consequently, detailed knowledge of cyclonic tracks is essential in forecasting weather and understanding the atmospheric environment. Modifications of the cyclonic tracks caused either by anthropogenic effects or by long-term natural variability are important sources of information in understanding the atmospheric dynamics and determining the impact on regional climates in the future.
The eastern Mediterranean, extending between 20° and 38°E to include the Ionian, Aegean, and Levantine Seas (see Fig. 1), is an area of great interest with respect to cyclone behavior, because of its location between the subtropics and midlatitudes and also its complex topography (HMSO 1962). In addition, the Mediterranean Basin is considered to be particularly vulnerable to climate change (Solomon et al. 2007), and it is of value to diagnose the changes in cyclone behavior over an extended period.
Several studies have attempted to depict the characteristics of the cyclonic tracks in the Mediterranean by employing objective methods for cyclone detection and tracking (Alpert et al. 1990a,b; Trigo et al. 1999; Campins et al. 2000; Picornell et al. 2001; Bartholy et al. 2009). In particular, Alpert et al. (1990b) used European Centre for Medium-Range Weather Forecasts (ECMWF) operational analyses every 12 h for a short period of 6 yr (1982–87) and examined intermonthly variations of cyclonic tracks in the entire Mediterranean. Trigo et al. (1999), employing higher-resolution data (1.125° × 1.125°) for an 18-yr period (1979–96), examined cyclonic tracks as part of a complete climatological analysis of Mediterranean cyclones. Trigo (2006) performed a comparison of tracks in the Euro-Atlantic sector, including the Mediterranean, as derived by two different resolution reanalysis datasets: the 2.5° × 2.5° National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis and the 1.125° × 1.125° 40-yr ECMWF Re-Analysis (ERA-40). Furthermore, Maheras et al. (2001) examined cyclonic centers in the Mediterranean region with the aid of an objective method, without distinguishing between cyclones that form at a specific grid and those migrating at a subsequent development stage. Campins et al. (2000), Picornell et al. (2001), and Bartholy et al. (2009) focus on the western Mediterranean tracks.
Studies related to the cyclonic tracks in the eastern Mediterranean refer to specific types of cyclones, such as frontal (Flocas 1988), and in specific areas, such as the Aegean Sea and Cyprus (Maheras 1983; Nicolaides et al. 2004), employing synoptic analyses of mean sea level pressure (MSLP). Some of these studies (Maheras 1983; Flocas 1988; Alpert et al. 1990b; Trigo et al. 1999; Maheras et al. 2001) have demonstrated the importance of examining Mediterranean cyclonic tracks on a monthly basis.
Over the last 20 yr several numerical algorithms have been developed to objectively identify cyclones and their tracks, using different approximations in the definition of cyclonic centers and the strength of the cyclones detected (e.g., Le Treut and Kalnay 1990; Alpert et al. 1990a; Murray and Simmonds 1991a,b; König et al. 1993; Hodges 1994; Serezze 1995; Haak and Ulbrich 1996; Blender et al. 1997; Sinclair 1994; Lionello et al. 2002). Cyclone centers have been defined in terms of pressure minima at sea level or minima in 1000-hPa geopotential height (e.g., Le Treut and Kalnay 1990; Alpert et al. 1990a). Alternatively, cyclones can be defined in terms of maxima in low-level vorticity (e.g., Hodges 1994; Sinclair 1994). Methods searching for pressure minima tend to overestimate deep and mature cyclones, while they miss small-scale systems that are better identified from their local maxima in relative vorticity, for example, fast-moving systems or cyclones in the early and late stages of their life cycle (Hoskins and Hodges 2002). On the other hand, vorticity maxima are not always connected with local pressure minima. For this reason, Murray and Simmonds (1991a,b) and König et al. (1993) employed a combination of both criteria.
Despite the considerable research that has been performed on Mediterranean cyclonic tracks, there is a strong need to update their climatology for longer periods and further examine important cyclonic features. This is particularly so for the eastern Mediterranean, resulting from more limited research as compared to the western Mediterranean. In line with these considerations, the objectives of this study are (i) to investigate the intermonthly variations of cyclonic tracks affecting the eastern Mediterranean, (ii) to examine the tracks according to their origin domain, (iii) to investigate their fundamental dynamic/kinematic characteristics, and (iv) to explore possible trends in their frequency and intensity.
The dataset used in this study comprises 6-hourly analyses of MSLP on a 2.5° × 2.5° regular latitude–longitude grid for the period of 1962–2001, as derived from ERA-40. It should be noted that global analyses significantly improved after 1979, because of the major improvement of the overall observing system. However, the observational coverage in the Mediterranean basin was very good prior to that time, and hence the ERA-40 dataset there before 1979 can be used with confidence (Courtier et al. 1998; Uppala et al. 2005). The domain of study includes the eastern Mediterranean area, extending from 20° to 38°E and from 30° to 45°N (Fig. 1). The resolution of the input dataset means that the identified tracks do not include smaller-scale cyclones that form in the Mediterranean region, mainly during the warm period by thermal forcing (Trigo et al. 1999, 2002).
The cyclone identification and tracking was performed with the algorithm developed at Melbourne University, according to the Lagrangian perspective (MS algorithm; see Murray and Simmonds 1991a,b; Simmonds and Murray 1999). The MS algorithm “finds” a cyclone only if an open or closed depression can be associated with a vorticity maximum. This approach is considered to be crucial, because open lows are also incorporated into the storm life cycle, preventing possible time series breaks, if a temporary weakening to an open, low state occurs (Simmonds et al. 2008).
The performance of the MS algorithm has been assessed (Leonard et al. 1999; Pinto et al. 2005; Mesquita et al. 2009) and proved to be a very powerful tool, not only in the generation of cyclone climatologies, but also in the assessment of individual tracks. The algorithm was found to be capable of identifying cyclones in a range of locations and with different characteristics, including small-scale systems over secondary storm-track regions and fast-moving storms that produce extreme events in Europe (Pinto et al. 2005). Especially for the Mediterranean region, Pinto et al. (2005) demonstrated that their results are in good agreement with those of Trigo et al. (1999), while for individual cases there is good agreement with hand-analyzed synoptic weather maps. Leonard et al. (1999) pointed out that the MS algorithm is able to detect a larger number of cyclones in the Southern Hemisphere, as compared to other two algorithms developed by König et al. (1993) and Terry and Atlas (1996), and, moreover, it can track these cyclones and maintain their continuity over a long time. Furthermore, Leonard et al. (1999) concluded that the MS algorithm is most applicable for research purposes. Mesquita et al. (2009) demonstrated that the MS algorithm, since accounting for open, low systems, produces more reliable storm climatology results, as compared to the National Oceanic and Atmospheric Administration (NOAA)/Climate Prediction Center (CPC) current operational algorithm.
In our study, a bicubic spline interpolation was employed to a 121 × 121 polar stereographic grid with a resolution of 1.9° latitude at the northern pole, decreasing to half of this value (0.95°) at the equator. The control parameters of the MS algorithm have been modified, as compared to the application in the entire Northern Hemisphere (Lim and Simmonds 2002), in order to better capture the individual characteristics of cyclones in a closed basin with complex topography, such as the Mediterranean. In particular, (i) diffusive smoothing to the original data has not been applied; (ii) all cyclones identified at grid points with surface heights of more than 1500 m are excluded; (iii) an averaging radius equivalent to 4° in latitude is selected, considering that Mediterranean cyclones have a typical horizontal scale of 300 km (Trigo et al. 1999); (iv) during the detection stage of the algorithm, no strength criteria are set in the beginning of the life span of the cyclones, because many Mediterranean cyclones at the beginning of their life are weak and unimportant, but during the next steps they turn into strong systems (Trigo et al. 1999); and (v) a minimum life span of 24 h is imposed in order to exclude short-lived systems and to enable the calculation of time derivatives of the velocity and pressure tendency (Simmonds and Murray 1999).
Every track entering the eastern Mediterranean basin for at least one analysis time step was considered as an eastern Mediterranean track. The frequency of tracks entering the eastern Mediterranean was determined for each month of the year. Then, the frequency of tracks for each month was calculated according to their origin domain, which is represented by the first step of the track. More specifically, the following six sectors were distinguished (see Fig. 1): (i) the northwestern sector, (ii) the western sector, (iii) the southwestern sector, (iv) the northeastern sector, (v) the northern part of the examined area of eastern Mediterranean (EM1), and (vi) the southern part of the examined area (EM2). Based on this classification, our target area for the eastern Mediterranean is represented by the sectors EM1 and EM2, which hereafter will be referred to as EM12 area.
For each sector, the following dynamic and kinematic parameters of the cyclonic tracks are calculated for the period that they remain within the EM12 area: (i) the effective cyclone radius R [in degrees of latitude, (°lat)], which is defined as
where ri is the distance of the radial line from the cyclone center to the points at which the Laplacian of MSLP is zero around the edge of a cyclone and N is the number of the radial lines drawn at azimuthal spacings of 20° (Lim and Simmonds 2007), that is, N = 18, in this case; (ii) the Laplacian of the central pressure ∇2P [hPa (°lat)−2], representing an effective measure of cyclone intensity (Petterssen 1956); (iii) the cyclone depth D (hPa), which combines the cyclone size and intensity by the relationship and, in the case of an axisymmetric parabolic cyclone, can be written as
[this represents the “pressure deficit” of the cyclone, i.e., the difference between the pressure at the “edge” of a cyclone and at the center, and it is also related to the total kinetic energy of the cyclone (Simmonds et al. 1999; Simmonds and Keay 2000, 2009)]; and (iv) the average propagation velocity Uc (m s−1; Lim and Simmonds 2007).
3. Intermonthly variations
Figure 2 displays the tracks that enter the EM12 area for each month during the 40-yr period, covering their whole lifetime from their formation until their dissipation (viz., track density). It can be seen that there are considerable intermonthly variations of track density in the target area EM12, even for months belonging to the same season, which are results that are consistent with, or complementary to, previous studies (e.g., Alpert et al. 1990a,b).
More specifically, in December the track density is reduced, as compared to the other winter months, especially for cyclones originating in the western Mediterranean, as was found by Bartholy et al. (2009) for the winter period. In January, the maritime tracks increase substantially in the entire Mediterranean, suggesting the role of the enhanced sea–land temperature difference on cyclone formation and movement (Lolis et al. 2004). The tracks originating from southern Italy form mainly in January. Similarly, the track density over the Black Sea increases in January. In February, the tracks migrate northward along the northern Mediterranean coast, where the major topographic barriers are located, from the Gulf of Lions and Gulf of Genoa to the Adriatic, Ionian, and Aegean Seas, which are mainly associated with the lee origin of the cyclones affecting the examined area (Trigo et al. 2002; Bartholy et al. 2009).
In spring, the most notable feature is the increase of the North African tracks, forming in the south of the Atlas Mountains, reflecting the low-level baroclinicity increase (Trigo et al. 1999), as was also found by Alpert et al. (1990b). The tracks from North Africa differentiate after entering the eastern Mediterranean, in accordance with Prezerakos (1985). More specifically, in March they follow a northward/northeastward direction toward the Ionian and Adriatic Seas, while in April and May a more eastward route is prominent toward the Levantine Sea and Cyprus. In April, the northward cyclonic tracks become more common, contributing to the substantial density increase over the Balkan Peninsula and the Black Sea. In May the track density of the North African cyclones reduces, becoming very small in the following summer months. Furthermore, the density of the tracks originating in the western Mediterranean decreases, as compared to the other spring months, this being most likely related with the corresponding remarkable reduction of the number of cyclogenetic events in the Gulfs of Genoa and Lions, and the Ligurian and Adriatic Seas during this month (Trigo et al. 1999).
During boreal summer months, there are no substantial intraseasonal differences of the tracks passing over the eastern Mediterranean, in accordance with Alpert et al. (1990b), except that the track density is lower in July and August. The cyclonic tracks are mainly concentrated over the land, reflecting their thermal origins (Trigo et al. 2002). Moreover, the cyclones originate in the eastern Aegean and Cyprus (EM2), along the Turkish coast, because the subtropical high prevails over the whole Mediterranean region and the Pakistan low extends over the eastern Mediterranean (Ziv et al. 2004). Furthermore, the cyclones are not migratory, because of their thermal character and small life span (Trigo et al. 2002). It becomes evident that the tracks from the western Mediterranean are limited in summer, although the number of cyclogenetic events is increased in this region, because of the short distance that they cover (Picornell et al. 2001).
In September, similar features are depicted as for August, except that the number of tracks originating in the North African coastal region is slightly increased, because of the reduced influence of the subtropical high during this month. In October, the distribution of tracks changes considerably. The density of maritime tracks increases, especially for those starting from the Adriatic and Ionian Seas, which is probably related to sea surface temperature (SST) changes during this month, as was demonstrated by Marullo et al. (1999) and Skliris et al. (2010). Also, there are cyclones that form in the south of Crete. It should be noted that the thermal effect prevailing during summer months and September is almost eliminated during October and November. In November, the cyclonic routes increase in the western Mediterranean, mainly along the northern coast, where the low-level baroclinicity intensifies (Trigo et al. 2002).
In summary, as one can see from Fig. 3, the cyclonic tracks passing through EM12 are most numerous from December to April; their number decreases during the warm period and tends to increase again in October. The maximum number of cyclonic tracks over the target area is observed in January (11.2% of the annual total) and March (10.3%). The minimum number of tracks occurs in July (5.3%).
4. Track classification
The classification of the tracks passing through the EM12 area, according to their origin domain, during the period of 1962–2001 and for all months, reveals that the greatest number (36.8%) originate in the southern part of the examined area (EM2), that is, the Ionian Sea, the southeastern Aegean Sea, Crete, Cyprus, and the Levantine Sea (Fig. 4), representing a major cyclonic center in Mediterranean (Maheras et al. 2001). The number of these tracks is particularly high in the boreal summer and fall (Fig. 5), reflecting the influence of the thermal Pakistan low extension in the eastern Mediterranean during this period (Bitan and Saaroni 1992).
A considerable proportion of systems (26.8%) originate in the northern part of the examined area (EM1), including northern Greece, the north Aegean, the Balkans, and a major part of the Black Sea. The tracks originating in the Aegean Sea prevail during winter, and mainly in January (Fig. 5), consistent with the cyclogenesis frequency distribution (Flocas and Karacostas 1996). In the boreal summer, the number of cyclonic tracks originating in EM1 sector drops dramatically (Fig. 5), although the southeastern Black Sea is a cyclogenetic area throughout the entire year (Trigo et al. 1999).
Apart from these tracks that originate within the examined area, the next most common cyclonic tracks originate in the western sector (13.8%), which incorporates the Adriatic Sea and parts of the western Mediterranean and the Atlantic. Moreover, these tracks exhibit their maximum frequency in January and February (Fig. 5), complying with the consideration that the Gulf of Genoa is the greatest winter cyclogenetic area (Trigo et al. 1999) and that southern Italy is a major cyclonic center in winter (Maheras et al. 2001). It should be noted that the westerly track coincides with a common path of frontal depressions passing over Greece during winter (Flocas 1988) and of baroclinic depressions affecting Cyprus during the cold season (Nicolaides et al. 2004).
The tracks from North Africa and the Alboran Sea, being attributed to the southwest category, comprise a significant portion of the total track number (13%), in agreement with previous studies (Maheras 1983; Alpert et al. 1990b; Trigo et al. 1999). The tracks of Saharan depressions predominate during March and April (Fig. 5; see also Prezerakos 1985), while tracks from the Alboran Sea are evident during the boreal summer (Picornell et al. 2001).
A limited number of tracks (4.9%) originate in the northwest sector throughout the entire year, since the northwest and EM12 sectors share an infinitesimally short border. However, the northwest tracks are found to enter the target area mainly through the west sector. The northwest tracks predominate in December, with, however, no significant intermonthly variations (Fig. 5). This is the favored path for frontal depressions originating in the Atlantic near the British Isles from January to April (Flocas 1988). However, depressions originating in this sector are also apparent in the boreal summer (Fig. 5), as was also noted by Maheras (1983) for depressions passing over the Aegean Sea. This could be associated with the northward displacement of the subtropical jet, which does not allow the westerly or southwesterly tracks to enter the examined region (HMSO 1962).
The tracks originating in the northeast sector are also limited in number (4.7%), being related with cyclones, forming mainly in the northern part of the Balkan Peninsula, eastern Europe, and part of the Black Sea. These tracks do not exhibit any intra-annual variations (Fig. 5), in agreement with the cyclogenesis frequency in this area (Trigo et al. 1999).
From Fig. 5 it is evident that there are significant intermonthly variations of the track frequency originating from the same sector among the spring months and autumn months, suggesting corresponding variations in the atmospheric circulation responsible for cyclone formation during these months, as reflected in the distribution of the cyclogenesis events during these months (Trigo et al. 1999).
5. Trends of track frequency
The trends of the total frequency of cyclonic tracks entering the EM12 region are investigated for each month separately during the examined 40-yr period. The trends are calculated with the aid of linear regression analysis on a monthly basis, and the statistical significance of the trends at a = 0.10 and a = 0.05 was examined with the Student’s t test (Table 1).
Figure 6 outlines the interannual variation of the number of tracks entering the EM12 region. The total number of tracks passing over EM12 for the 40-yr period is 10 461, suggesting an average number of about 260 tracks per year. In the beginning of the period the number of the tracks is quite high during the cold months. The year 1969 is characterized by a peak (286) in the total annual number of tracks.
In the Mediterranean, the cyclone development and motion is greatly affected by low-level baroclinicity, topography, and sea surface fluxes, as well as by upper-level dynamics (Trigo et al. 2002; Maheras et al. 2002). Any observed changes in the frequency of Mediterranean cyclonic tracks are related principally with changes in baroclinicity and upper-level circulation. Enhanced sea surface fluxes associated with a warmer sea and mirroring increased SSTs play a more important role in the intensification of Mediterranean cyclones (Alpert et al. 1990a; Lionello et al. 2002; Solomon et al. 2007; Lagouvardos et al. 2007), and will be further examined in the trends of track intensity in section 6.
Taking into account that the low-level baroclinicity is proportional to the temperature gradient (Hoskins and Valdes 1990),
we examine the extent to which the cyclone changes can be seen as associated with changes in the 850-hPa temperature gradient magnitude (Wang et al 2009).
The isobaric level of 850 hPa is selected as being representative of the lower troposphere; it reflects the effect of the free-tropospheric flow and the surface characteristics, and thus best represents the position of the large-scale baroclinic zone. For each monthly field and for each year the magnitude of the temperature gradient over the greater Mediterranean region was calculated, following the approach presented in Simmonds and Lim (2009). Then, for the first (1962–81) and the second (1982–2001) half of the study period the average monthly magnitude is computed along with the difference between the average of the second half minus the corresponding average during the first half (see Fig. 7).
Changes in the upper-level circulation are not the main interest in the present study. However, an attempt is made to relate our results with the findings of previous studies, dealing with the seasonal trends of the anticyclonic and cyclonic types at 500 hPa over the Greek area during the period of 1958–2000 (Maheras et al. 2004, 2006) and the seasonal trends of the 500-hPa geopotential height over Europe for the period of 1958–2002 (Hatzaki and Flocas 2004).
For the entire 40-yr period, a negative trend (−0.2 decade−1) of the frequency of the cyclonic tracks was found, which is significant at a = 0.10. However, from Table 1 it becomes evident that the sign and the significance of the trends vary considerably, not only interseasonally, but also intraseasonally, following the intermonthly variations of frequency. In the following, we focus on the statistically significant monthly trends.
It can be seen that there is a negative trend of the frequency during all three winter months, from December to February. This is consistent with the overall negative trend of cyclonic tracks in winter over the Mediterranean, which is statistically significant at a = 0.1 over the northern Balkans and Turkey, found by Trigo (2006), and the decreasing trend of cyclonic occurrences in eastern Mediterranean during the rainy period found by Maheras et al. (2001). Furthermore, the negative trend during the winter months agrees well with the downward trend of surface cyclonic types prevailing over the greater Greek area in winter as well as the corresponding upward trend of surface anticyclonic types (Maheras et al. 2000). This result is further supported by the statistically significant positive trend values of the 500-hPa geopotential height over the EM12 target area in winter (Hatzaki and Flocas 2004), and the decreasing trend of cyclonic types at 500 hPa over the greater Greek area (Maheras et al. 2006).
More specifically on a monthly basis, the most significant negative trend of track frequency in winter was found in February (−1.4 tracks per decade). In December (Fig. 7a) and January (Fig. 7b), the negative trends are well related with a remarkable decrease of baroclinicity over the entire Mediterranean and European area, supporting the statement of Trigo (2006) for a northward shift of storm tracks in the Euro-Atlantic sector in winter. In both months, the peak of the change in the magnitude of the gradients is found in north Europe as well as northern Africa [−4 K (1000 km)−1]. By contrast, in February (Fig. 7c), the major part of the target area EM12 is characterized by a baroclinicity increase, which seems contradictory to the apparent track decrease. However, a careful inspection of the track origin during this month (Figs. 2 and 5) demonstrated that a significant proportion of the tracks originate in the west, southeast, and northeast sectors, where a substantial magnitude of the temperature gradient decrease was indeed found, with peak values smaller by 4 K (1000 km)−1. Therefore, it is appears that during February the remote baroclinic influences are more important than the local influences acting within the target area.
During spring, the trend values vary, being negative in March and May and positive in April, but they are not statistically significant and are not accompanied by any noticeable change of the baroclinicity (not shown). The trend differences among the spring months could be attributed to the profound intermonthly variations of track frequency (see section 3) and the significant intermonthly variations in the cyclogenetic mechanisms acting in this period (Trigo et al. 2002). This is also reflected in the different sign of trend among the cyclonic types affecting Greece in spring, following changes in the atmospheric circulation prevailing over the greater European area during this season (Maheras et al. 2000, 2004). It should be noted that the overall spring trend of 500-hPa geopotential height over EM12 is (significantly) positive, with, however, lower values, as compared to winter (Hatzaki and Flocas 2004).
Despite the overall increase of the frequency of surface anticyclonic types in summer (Maheras et al. 2000), the decrease of frequency of 500-hPa cyclonic types (Maheras et al. 2004) in Greece, and the significant positive trend values of 500-hPa geopotential height over EM12 (Hatzaki and Flocas 2004), only in August does a significant negative trend of cyclonic tracks become evident. This trend is consistent with the increasing trend of the frequency of the surface mixed-weather-type “Mb” (Maheras et al. 2000) during this month, which is characterized by a weak surface pressure gradient over the Aegean Sea and zonal circulation at the upper levels, resulting in calm weather conditions and surface warming (Good et al. 2008). This is also consistent with an overall decrease in the magnitude of the gradient of temperature (or baroclinicity) over the entire Mediterranean region, peaking [at −6 K (1000 km)−1] in the Iberia and southeastern Aegean Seas (Fig. 7d), representing strong cyclonegetic centers during August (Trigo et al. 1999).
In fall, the overall trend of frequency is positive, in general agreement with the positive trend values at 500 hPa over EM12, although these are not statistically significant in the northern sector EM1 (Hatzaki and Flocas 2004). On a monthly basis, a significant positive trend of cyclonic tracks is observed in September, which agrees well with a substantial baroclinicity increase in the major part of the Mediterranean and southern Europe (Fig. 7e), peaking in the area of Italy [6 K (1000 km)−1]. For November, there is also a statistically significant positive trend (a = 0.1) that is consistent with the increase of surface cyclonic types over the greater Greek area and the associated tropospheric cooling in the 1000–500-hPa layer during this month (Good et al. 2008). Furthermore, the track increase in the target area agrees with baroclinicity increase in the western Mediterranean, the Atlantic, and northwest Europe (Fig. 7f), suggesting the role of the baroclinic effects acting in these areas.
For each origin sector, the trends of track density were calculated at each grid point within the EM12 target area and for the entire examined period, using linear regression. The statistical significance of the trend at each grid point of the target area was examined at a = 0.05.
The spatial distribution of the density trends for the tracks originating in the west sector (Fig. 8a) shows a significant negative trend in the entire target area, except over the Black Sea. Greater values are found in western Greece and the Ionian Sea, as well as in Cyprus. A significant negative trend is also observed for the northeast tracks, but over the northern part of the EM12 area, namely, the Balkans, and the northern Aegean and Black Seas (Fig. 8b). On the contrary, a positive trend characterizes the southwest tracks, which are significant over Crete, Cyprus, and the Middle East (Fig. 8c). No significant trend was found for the northwest tracks (not shown). The tracks that originate within EM1 exhibit a negative trend that becomes significant over the southern Ionian Sea, the southern Turkish coast, and the eastern part of the Black Sea (Fig. 8d). The tracks originating within EM2 exhibit a significant increasing trend over Cyprus and the Middle East and a decreasing one in the remaining area, but mainly over the east Black Sea and the Ionian Sea (Fig. 8e).
6. Dynamic and kinematic parameters
In this section, in order to obtain a comprehensive picture of their overall behavior and impact on climate according to the different origin domains, insightful diagnostic dynamic and kinematic parameters are described that are associated with cyclones passing through the eastern Mediterranean. Furthermore, these parameters are very important for understanding changes in cyclonic activity in eastern Mediterranean.
The estimation of the following four parameters were performed only for the part of each relevant track within the EM12 target area and for each origin sector, as described in section 2: (i) the Laplacian of central cyclone pressure ∇2P, (ii) the cyclone depth D, (iii) the cyclone radius R, and (iv) the propagation velocity Uc. Table 2 summarizes the value range, the mean value, and the standard error of the four parameters.
It can be seen that the mean Laplacian ∇2P has the same mean value [0.3 hPa (°lat)−2] for the cyclones generating within the examined area (EM1 and EM2), as well as for those that originate in the northeast. The mean Laplacian value increases to 0.35–0.38 hPa (°lat)−2 when the cyclones originate in the northwest, southwest, and west. The maximum mean intensity of 0.38 hPa (°lat)−2 characterizes the southwesterly tracks.
The average cyclone radius exhibits significant spatial variation in the target area, with values increasing from the north to the south, particularly in winter (not shown), in agreement with Simmonds and Keay (2002), for cyclones in the Northern Hemisphere oceans. The radius is at least 2.5°–2.7°lat for all categories. Larger cyclones are associated with the southwest, west, and northwest categories, with average radii of 3°, 2.95°, and 2.94°lat, respectively.
The cyclone depth similarly exhibits larger values for the tracks originating from the southwest, northwest, and west, a fact that seems reasonable for systems developing outside the target area. For the northwest tracks, the cyclone depth changes significantly between the origin domain and the target area: 5.5 hPa in the Atlantic area reduces to 1.5–3.3 hPa over the EM12 area. Furthermore, the depth of the west cyclones exhibits substantial spatial variations: 4.50 hPa in the Atlantic and 2.50–2.90 hPa in the western Mediterranean, reducing to 1.7–2.5 hPa over EM12. The depth is almost uniform for the entire target area (2.5 hPa) for the southwest tracks.
The propagation velocity also differs substantially between the origin sectors, with the lowest values being associated, as one might have expected, with those originating within EM1 and EM2 (3.6 and 3.9 m s−1, respectively). By contrast, the northeast cyclones propagate comparatively faster, especially at the southern boundary. Larger velocities of 5 m s−1 in the whole area examined characterize the west and southwest categories. The northwest cyclones appear as the maximum velocities within the examined area at 7.5 m s−1, while their mean velocity in the Atlantic region was as high as 12 m s−1.
The trends of the above-mentioned parameters were calculated on a monthly basis for the whole target area and for the 40-yr period. Negative trends were found for ∇2P, D, and R, but these were statistically significant (at a = 0.05) only for radius (−0.01°lat decade−1). On a monthly basis (see Table 1), the trend signs of track intensity and depth indicate considerable changes through the year. Following the overall decreasing tendency of intensity, the trend is negative from December to April, but at a statistically significant level (at a = 0.1) only in April. Then, from May to November, the monthly trends of intensity are positive and are particularly noteworthy (and statistically significant at a = 0.05) during the summer months of June–August. On the contrary, the trend of depth is negative from January to April and positive from May to December, with statistically significant values in May, June, and November.
Considering the relationship of the SST increase with cyclone intensification (see section 5), we have explored the link between the monthly trends of the SST in the eastern Mediterranean region with corresponding trends of track intensity (Table 1). For this purpose, the Met Office Hadley Centre’s SST dataset (HadSST2) is employed, which is available on a monthly global field on a 5° latitude × 5° longitude grid (Rayner et al. 2006). In our case, the average of six grid boxes that cover the 30°–40°N, 20°–35°E region was calculated for the 1962–2001 period. It was found that the increasing trend of track intensity in June–August corresponds well with the SST increasing trend, which is significant in August. This is in accord with Alpert et al. (1990a), who demonstrated the major positive thermal effect of the sea on eastern Mediterranean cyclones, particularly in summer.
We next calculated the trend of each parameter by linear regression at each grid point of the target area for the examined period and for each sector separately. The spatial distribution of the intensity trend reveals that there are no statistically significant changes for any sector, except for a negative trend in the Middle East for the west track (Fig. 9a). The radius of the tracks exhibits an overall decreasing trend over the whole target area for all sectors, which is significant only over the Aegean Sea, Greece, and Turkey for the EM2 tracks (Fig. 9b), and for the Black Sea for the northeast tracks. Only for the northwest and southwest tracks, does a (non significant) positive trend appear in the eastern part of EM2, namely, around the area of Cyprus and the Middle East (not shown). Similarly, for propagation velocity, no significant trend was found in the E12 area.
With respect to the spatial distribution of the depth’s trend, it is demonstrated that the tracks originating in EM1 present a decreasing trend over western Greece and the Black Sea that is statistically significant only over the Ionian Sea (Fig. 10a). In the remaining area, the trend becomes positive, though not significant. The southwest and EM2 track depths exhibit the (not significant) positive trend of depth over southern Aegean Sea, Levantine Sea, and Cyprus, and the (not significant) negative trend over the Black Sea (not shown). On the contrary, the significant decreasing trend of the depth of the northwest (not shown) and west tracks (Fig. 10b) appears in the eastern sector of EM2, that is, Lebanon, Syria, and Israel.
In this study the main features of cyclonic tracks affecting the eastern Mediterranean region have been explored for a period of 40 yr (1962–2001) with the aid of the Melbourne University tracking algorithm. This long period of quality reanalysis has allowed us a new and more comprehensive and robust view of the eastern Mediterranean cyclone behavior.
It was verified that considerable intermonthly variations of track density occur over the eastern Mediterranean, consistent with the results of previous studies for the entire Mediterranean. As was expected for Mediterranean cyclones, the track frequency decreases after May and tends to increase in October. Peaks were found in January and March and a minimum was found in July. Moreover, this study has exposed interesting features of the intraseasonal track density variations. The track density is reduced in December, as compared to other winter months, especially for cyclones originating in the western Mediterranean. Similarly, in May the number of tracks originating in the western Mediterranean is reduced, as compared to the other spring months. In September, the distribution of the track density is similar to that of the summer months. The maritime tracks increase during January and October, while the tracks migrate northward to the northern Mediterranean coast in February and November, reflecting shifts in low-level baroclinicity. Although it is well known that the number of tracks originating in northern Africa increases during the summer months, it was demonstrated that those affecting the eastern Mediterranean follow a northward/northeastward route in March and an eastward one in the other spring months, April and May.
The classification of the tracks according to their origin domain has been very revealing, and has shown that the vast majority originate within the examined area itself, mainly in the Cyprus area and over the southeastern Aegean Sea, which is a very important finding for the cyclone climatology in the eastern Mediterranean region. The most common cyclonic tracks that do not originate in the examined area start from the western sector with a maximum frequency in January and February. The tracks from North Africa (the southwest sector) are most common in the spring months of March and April. The number of tracks from the northwestern sector is rather small throughout the whole year.
The study of the kinematic and dynamic parameters of tracks according to their origin demonstrated that deeper cyclones follow the southwest track. Greater size characterizes the westerly tracks (southwest, northwest, and west), while the northwest tracks propagate faster over the examined area.
The analysis of the frequency trends of the cyclonic tracks exhibit significant intermonthly variations of sign and statistical significance, following intermonthly variations of frequency. In general, there is a statistically significant negative trend on an annual basis. This is mostly attributed to the negative sign during the winter months that can be explained in terms of a decrease of the baroclinicity, while supported by geopotential height changes at the upper levels. On the contrary, significant positive trends were found in September and November, which could be associated with the increases in SST and tropospheric cooling in the 1000–500-hPa layer, respectively, as well as changes in the baroclinicity.
Furthermore, the negative trend of frequency characterizes the tracks originating in the western and northeastern sector and within the northern part of the examined area (EM1) that is statistically significant at specific subareas. On the contrary, positive trend characterizes the southwesterly tracks.
There is no statistically significant trend of intensity of tracks affecting the eastern Mediterranean for any sector, except for a negative trend in the Middle East for the west tracks. The size of the tracks tends to decrease over the whole target area for all of the sectors. Concerning the depth, a negative trend prevails over the examined area, which is significant for EM1, northwest, and west tracks but only at specific subregions.
We comment that the relatively coarse-resolution dataset employed in this study is not fully capable of representing smaller-scale cyclones in specific regions of the eastern Mediterranean, such as thermal lows developing during the warmer months around Cyprus and the Middle East (Trigo et al. 1999) and secondary centers within complex systems (Trigo 2006). Therefore, in our study the frequency of the tracks could be seen as being underestimated, especially during the period from June to September. Perhaps a more serious effect of the coarse resolution is related to an underestimate of cyclone intensity (Lionello et al. 2002; Trigo 2006). Nevertheless, the ERA-40 dataset seems to represent larger-scale cyclonic tracks that cover an extensive area encompassing the examined region and have considerable impact on the temperature and precipitation regime. In this sense, the small-scale features that our dataset cannot pick up may not be all that important in the overall picture of cyclonic influence in this region. The constraints imposed by the low resolution does not seem crucial for the trend analysis of the track frequency, because Trigo (2006) demonstrated the agreement between two different resolution datasets (2.5° × 2.5° and 1.125° × 1.125°) on the sign and location of trends over the Mediterranean region, with, however, some discrepancies regarding their strength and significance.
The MS algorithm has proven to be a very valuable tool for examining cyclonic tracks in a smaller-scale inland sea with complex topography, such as the Mediterranean. Moreover, it has been used to generate an extended climatology of cyclonic tracks that verifies results obtained in previous related studies based on other algorithms, and has revealed valuable additional insights related to cyclone frequency, kinematics, and dynamics. This is most likely related to the fact that the MS scheme accounts for both closed and open systems, and also performs well in maintaining cyclone temporal consistency (see, e.g., Pinto et al. 2005; Mesquita et al. 2009).
The study has expanded our knowledge of the structure and variability of Mediterranean cyclones and, in turn, contributes significantly to our understanding of physical processes associated with Mediterranean climate and its variability. The results postulated by this study highlight the changes in synoptic activity, which are not inconsistent with the vulnerability of the Mediterranean climate to global warming (Solomon et al. 2007).
This work was supported by the project KAPODISTRIAS 2009, which is funded by the Special Account for Research Grants of the University of Athens. Parts of the work were made possible by a grant from the Australian Research Council.
Corresponding author address: Helena A. Flocas, Building PHYS-5, Department of Environmental Physics and Meteorology, Faculty of Physics, University of Athens, University Campus, 157 84 Athens, Greece. Email: email@example.com