The Relationship of Lightning Activity with Microwave Brightness Temperatures and Spaceborne Radar Reflectivity Profiles in the Central and Eastern Mediterranean

ABSTRACT
In this paper, the relationship of lightning activity in the central and eastern Mediterranean with the 85-GHz polarization-corrected temperature (PCT) and radar reflectivity provided by (he Tropical Rainfall Measuring Mission (TRMM) satellite is investigated. Lightning observations were mainly provided by the Met Office's Arrival Time Difference system as well as by the TRMM Lightning Imaging Sensor. The studied period spans from September 2003 to April 2004 and focuses on the events with the most important lightning activity, It was found that 50% of the eases with flashes have PCTs lower than 225 K, while only 3% of the "no lightning" cases have PCTs below this value. Further, if PCT is used as a proxy for the presence of lightning. the value of 217 K gives the best statistical scores for the presence of at least one observed flash. In addition, the ratio of cloud-to-ground lightning to total lightning activity has higher values for the "colder" PCT values and decreases as PCT increases. In addition, the mean and maximum reflectivety profiles with collocated lightning are from 3 to 10 dB and from 6 to 15 dB. respectively, higher than that without lightning. Further, a reflectivity profile with values greater than 53 dBZ in the low levels (below 3 km), of ˜ 45 dBZ at 5 km and 40 dBZ at 7 km is associated with a probability of 80% for lightning occurrence.

1. INTRODUCTION
    The Mediterranean Sea is a relatively small and warm body of water surrounded by throe major continents, which according to Orville (1981) and Christian et al. (1999a,b) is one of the major centers of electrical activity during the Northern Hemisphere winter. Altaratz et al, (2003) found that the annual distribution of lightning in the eastern Mediterranean presents a clear maximum in January, and that during midwinter there is a higher lightning density over the sea. In a recent study Katsanos et al. (2007) have shown that during the wet period of the year the lightning activity occurs over the maritime area and near the coasts, almost delineating the Mediterranean coastline, with no clear preferred hour for the lightning occurrence.
    In addition to the study of the spatiotemporal variability of lightning over various regions, including the Mediterranean, many studies in the recent literature have been devoted to the examination of the relationship of lightning activity with microphysical properties of mesoscale systems, because these are indirectly manifested in passive microwave or radar measurements.
    The relationship of lightning activity with brightness temperature at 85 GHz has been addressed in a number of studies that mainly focused on mesoscale convective systems (MCSs) over the United States and the Tropics (Mohr et al. 1996: Toracinta and Zipser 2001; Toracinta et al. 2002). Indeed, Mohr et al. (1996) analyzed the 85-GHz polarization-corrected temperature (PCT) calculated from the Special Sensor Microwave Imager (SSM/I) and the U.S. National Lightning Detection Network data for nine MCSs over Texas during the period from April to July 1993. The authors found that the flash density was inversely correlated to PCT and that the amount and distribution of lightning within a MCS is related to the presence of large ice particles in the mixed-phase region of these systems. In their study. Toracinta and Zipser (2001) used brightness temperature data at 85-GHz from the SSM/I and lightning data from the Optical Transient Detector, and presented a systematic comparison of the distributions of MCSs and lightning for 19 geographical regions classified as land. ocean, or a mixture of land and ocean between 35°N and 35°S, over four 3-month periods beginning in June 1995. Their results showed that a land MCS is more likely to produce lightning than an oceanic MCS. implying that there are differences in the ice microphysi-cal processes between land and ocean convective storms.
    Regarding the investigation of the relationship between lightning activity and radar reflectivity, a number of previous studies are found in the literature. These studies mainly focus on summertime thunderstorms. Carey and Rutledge (2000), in the framework of the Maritime Continent Thunderstorm Experiment, studied convective cells within a tropical convective complex based on a C-band radar and surface lightning network observations. They found that both cloud-to-ground and total lightning were highly correlated to the radar-inferred mixed-phase ice mass in time and space. Toracinta et al. (2002) used Tropical Rainfall Measuring Mission (TRMM) precipitation radar (PR), Microwave Imager (TMI), and Lightning Imaging Sensor (LIS) data in order to study precipitating systems in the Tropics for a 3-month period (August-October 1998). They found that lightning is much more likely in tropical continental features than tropical oceanic features with similar brightness temperatures or similar reflectivity heights. Moreover, they found that features with lightning have greater magnitudes of reflectivity and smaller decreases of reflectivity with height above the freezing level than systems without detected lightning. Adamo et al. (2003), using 1-yr TRMM data for the region of the Mediterranean Sea, correlated average PR reflectivity profiles with various values of flash rates for both convective and stratiform regions of storms and concluded that higher flash rates are correlated with higher-reflectivity profiles during all seasons. They also concluded that lightning-producing storms are associated with a significant percentage of precipitation during all seasons, but mostly during summer.
    Most of the aforementioned studies on the relationship of lightning with radar and passive microwave observations were devoted to the characterization of mesoscale convective systems either in the United States or the Tropics and mainly during summer. In a recent study by Defer el al. (2005), a winter case over Greece was presented where the presence of lightning (as sensed by LIS) was associated with brightness temperature depressions as low as 110 K. In the same paper, the authors have tried to characterize winter-type lightning activity (20 days), emphasizing the study of the flash development, duration, and extension.
    This work has been motivated by the scarcity of studies that relate lightning with radar and passive microwave observations in the Mediterranean and also for precipitating systems associated with frontal activity. For that reason, the events associated with important lightning activity that occurred in the central and eastern Mediterranean Sea during autumn and winter 2003/04 have been identified, with the aim to study the relationship of this lightning activity with the brightness temperature depression as measured by TMI as well as the corresponding reflectivity profiles provided by the PR. The lightning data were provided mainly by the Arrival Time Difference (ATD) system operated by the Met Office, but also by the TRMM LIS.
    The rest of the paper is organized as follows: section 2 presents the data and the methodology used, while section 3 discusses the comparison between lightning and TMI brightness temperatures. Section 4 is devoted to the analysis of PR reflectivity profiles in relation with the occurrence and intensity of lightning activity. Section 5 is devoted to the concluding remarks and prospects of this study.

2. DATA AND METHOD

A. DATA
    The lightning database that will be used in the analysis is based on the observations from the Met Office long-range sferics ATD system and the TMI, PR, and LIS onboard the TRMM satellite. A brief description of the instruments is given below, while for detailed discussion one should refer to the referenced literature.
    The Met Office long-range sfercs ATD system records cloud-to-ground (CG) lightning activity. This system consists of seven stations in total that sense the very low frequency (VLF) signal radiated during the connection to the ground of the flashes and records the time of the lightning event. The basis of the technique is the assumption that electromagnetic radiation emitted from a lightning discharge propagates (as a VLF radio wave) at a uniform speed in all directions. At (east four stations are required in order to remove the ambiguity of the flash location (I lull et al. 2001). The ATD data files consist of a list of flashes based on the time they occur, their location, the error of their location. and some parameters for quality control. Keogh el al. (2006) report that the detection efficiency of the ATD system, in sensing cloud-to ground lightning, ranges from 70% to 90% during the winter period over the United Kingdom and Europe. The ATD system covers the entire Mediterranean Sea.
    The National Aeronautics and Space Administration (NASA) TRMM satellite has flown at an altitude of ˜ 402.5 km since August 2001. All instruments on TRMM are described thoroughly by Kummerow et al. (1998). One of these instruments is the TRMM Microwave Imager. Its antenna scans conically, with a viewing angle 49° off nadir (52.8° incidence angle). TM1 has nine channels in total [10.7-GHz vertical (V) and horizontal (H), 19.4-GHz V and H, 21.3-GHz V, 37-GHz V and H. and 85.5-GHz V and H polarization]. At low frequencies (<37 GHz), brightness temperature responds mostly to emission from rain and cloud liquid. At high frequencies, brightness temperature responds mainly to scattering from cloud ice and hail. In the frame of this study, the 85-Gllz channels of TMI are used, because this frequency is sensitive to the presence of ice in clouds, which is related to both intracloud (IC) and cloud-to-ground lightning activity, because collisions between graupel (or hail) and ice crystals in the presence of supercooled water are required for cloud electrification (Blyth et al. 2001). The 85-GHz channel has a swath of ˜ 878 km and a ground resolution of 5.1 km.
    Moreover, TRMM is the first satellite that carries a precipitation radar on board that measures reflectivity profiles. The Precipitation Radar is an active phased-array. 2-km-long radar system, which operates near 13.8 GHz. It scans across a 247-km swath, with 80 vertical levels extending to 20 km above the earth's surface. The horizontal resolution is 5 km and the vertical is 250 m, while the minimum detectable signal is approximately ˜ 18 dBZ (Kummerow et al. 1998).
    Last, the LIS on board the TRMM satellite platform detects total lightning activity, because cloud-to-ground, intracloud, and cloud-to-cloud discharges all produce optical pulses that are visible from space (Christian et al. 2000). At 777.4 nm, LIS senses the radiated light from lightning flashes. Its sensor can monitor individual storms or storm systems for almost 90 s as it passes overhead, with an approximate field of view of almost 500 × 500 km[sup2], a spatial resolution of 10 km, and a very high temporal resolution of 2 ms. Boc-cippio et al. (2002) have assessed the performance of LIS, and they concluded that the detection efficiency of the system is 93% +/- 4% during nighttime and 73% +/- 11% during daytime.

B. METHOD
    The inspection of datasets, as described above, has shown that the areas with low brightness temperatures present a high spatial coherence with the areas of vigorous lightning activity. An example of such a case is the one that occurred in 13 December 2003. During that day a shallow low pressure center, located north of the Sidra Gulf, was associated with a cold front extending northwest-southeast from south of continental Greece toward the island of Crete. There was a TRMM overpass over the area of interest at ˜ 2000 UTC 13 December 2003. The position of the cold front over the area is well denoted by the distribution of the brightness temperature field measured by TMI at the 85-GHz (vertical) channel. Indeed the brighter colors denote the colder brightness temperatures that mark the frontal discontinuity well (Fig. la). A focus in the brightness temperature field over Crete is given in Fig. lb, while the reflectivity field from PR at 3-km height in the same area is given in Fig. 1c. The frontal discontinuity is well marked by high reflectivity values exceeding the value of 45 dBZ at distinct reflectivity cores. A vertical cross section of the reflectivity field near the northern edge of the PR swath is given in Fig. 1d, where the vertical extent of the frontal convection is depicted. The convection reaches 9-10 km while the reflectivity exceeds 45 dBZ at the low levels. The frontal discontinuity that is characterized by cold brightness temperatures and high reflectivity values is also associated with significant lightning activity as can be seen in Fig, 1e, where total lightning activity sensed by LIS during this TRMM overpass is reported, and also in Fig. If, where cloud-to-ground lightning activity recorded by the ATD system during the time window of the LIS observation is reported. As expected, the number of ATD flashes is smaller than that of the LIS flashes, but both instruments detect the line convection associated with the frontal discontinuity.
    Such cases as the aforementioned motivated the study over a longer time period (September 2003-April 2004). The study area covers the central and eastern Mediterranean Sea, bounded by 30° and 40°N and 10° and 35°E (Fig. la). Because special interest is given to days associated with significant lightning activity over the study area, all days within the 8-month period were sorted by their number of cloud-to-ground lightning recorded by ATD, and the 60 days with the higher values were selected. Among these 60 days, 15 distinct weather systems that lasted 2, 3, or even 4 days have been identified, while 6 days corresponded to single-day activity. The majority of the analyzed events were low pressure systems associated with frontal activity in the study area.
    For all of these cases, the corresponding TMI orbital files were collected. The total number of TMI orbital files that were found to adequately cover the study area was 121, covering 41 out of the 60 days. These orbital data were converted to gridded files with 0.25° × 0.25° resolution over the study area. Each grid box is characterized by its latitude, its longitude, and the minimum of all PCT measurements at 85 GHz that lie inside this grid box. The polarization-corrected temperature is calculated using the formula suggested by Spencer et al. (1989).
    PCT[sub85] = 1.82T[subb85(V)] - 0.82T[subb85(H)].
    The ATD data have been matched to the same grid, and therefore to each grid box, in addition to its latitude, longitude, and PCT value, the number of recorded flashes that lie inside it for a specific time window has been added. A +/-10-min time window around the time of the TRMM overpass over the area was chosen. The +/-10-min window is assumed to adequately represent the electrical activity that is associated with the mesoscale features sensed during the satellite overpass. The resulting database of grid boxes will be statistically analyzed in the following section.
    Further, the ATD data will be studied along with the LIS and TMI data to investigate the relationship between the cloud-to-ground-to-total-lightning ratio with the PCT. The comparison is again performed over the central and eastern Mediterranean Sea for 118 out of the 121 aforementioned orbits (for 1 day with intense solar activity the LIS instrument was switched off to prevent possible damage), The 118 LIS and TMI orbital files and the ATD measurements were matched to a common 0.25° × 0.25° grid. For each grid box, the number of LIS flashes within it has been counted, along with The number of ATD flashes that were recorded within the grid box during a lime interval that starts at the time of the first LIS flash and ends 1 s after the last LIS flash in the grid box. Then, for each grid box with at least one LIS flash counted, the ratio of the ATD/LIS number of flashes is calculated. Thus, the database that will be investigated contains (for each grid box) its latitude, its longitude, the minimum of all PCT measurements at 85 GHz that lie inside this grid box, and the ATD/LIS ratio. Taking into account the diurnal variability of the detection efficiency of the LIS system, the analysis of the relation of PCT to the ATD/LIS ratio is also performed separately for day and night TRMM overpasses.
    Last, for the analysis of the relation between lightning and reflectivity profiles, the PR data along with the ATD will be used. Because the PR swath is much narrower than the TMI swath (247 km as compared with 878 km), the number of the PR orbital files that had a fair coverage of the study area is much smaller than the
    [Graph or Chart Omitted]
    121 TMI orbital files. Indeed, only 29 PR orbital files were used, covering 20 days out of the 41. These orbital files were also converted to 0.25° × 0.25° grid boxes by averaging at each PR vertical level the reflectivity profiles within each grid box, and also by identifying the maximum reflectivity within it. Thus, a second database of grid boxes has been constructed containing the latitude, the longitude, and the mean and maximum reflectivity for each PR vertical level, as well as the number of recorded ATD flashes.

3. COMPARISON OF TMI 85-GHZ PCT WITH LIGHTNING ACTIVITY
    This section is devoted to the comparison of the PCT at 85 GHz for the 121 TMI orbits with the corresponding lightning activity. Various statistical tools have been applied with the aim to investigate the possibility of the use of PCT as an indicator for the presence of lightning activity.

A. CUMULATIVE FREQUENCY DISTRIBUTION OF PCT
    From the database of grid boxes described in the previous section, the distribution of the cumulative frequency of PCT values for the grid boxes with and without recorded CG lightning activity has been constructed (Fig. 2). It is evident that the grid boxes with flashes correspond to colder (in general) PCT values, and as PCT decreases the likelihood that lightning is produced rapidly increases. The colder 25% of the grid boxes with recorded lightning activity have PCTs lower than 190 K, while the colder 25% of the grid boxes without lightning have PCTs in the 200-270-K range (note also that only the coldest 5% of those PCTs lie in the 200-250-K range, and the remaining 95% are warmer than 250 K, up to 270 K). Furthermore, 50% of
    [Graph or Chart Omitted]
    the grid boxes with flashes have PCTs lower than 225 K, while only 3% of the "no lightning" grid boxes have PCTs below this value. Last, 90% of the grid boxes with flashes have PCTs lower than 265 K, which can be compared with a respective percentage of only 10% for the no-lightning grid boxes. In their recent study based on the analysis of a 3-yr period of TRMM data, Cecil et al. (2005) have also identified that with decreasing PCT values the probability of lightning increases. In their Fig. 3b they reported that almost 60% of their precipitation features with lightning have PCTs lower that 200 K, This percentage (although larger than the 30% in our study) compares favorably to our results only qualitatively because the two studies have followed different methods. Namely, Cecil et al. (2005) have analyzed only precipitation features (we used all grid boxes/features in the study area for each event) and also used the minimum PCT anywhere in the contiguous precipitation feature (whereas we used the minimum of each grid box).

B. STATISTICAL SCORES
    The database was also used to construct a 2 × 2 contingency table for various PCT threshold values and the occurrence or not of lightning activity (Table 1). The aim is to find the PCI threshold value below which the probability of occurrence of lightning activity is maximized. In other words, the aim is to investigate if the PCT value can be used as a proxy tool for the presence of lightning activity.
    Based on the Table 1 notation of a, b, c, and d, the following statistical scores are calculated. The probability of detection (POD) is a/(a + c), which equals the number of hits (grid boxes with lightning and PCT below the selected threshold) divided by the total number of grid boxes with lightning, and thus it gives a simple proportion of the grid boxes with electrical activity that have a PCT colder than the threshold. The POD ranges from 0 to 1, with the perfect score being 1. The false-alarm ratio (FAR) is b/(b + a), which is equal to the number of false alarms (number of grid boxes for which the PCT is lower than the threshold and there are no recorded flashes) divided by the total number of grid boxes for which PCT is below the threshold. The FAR also ranges from 0 to 1 and the perfect score is equal to 0. The Heidke skill score is
    HSS = (a + d) - [(a + b)(a + c) + (b + d)(c + d)]/n/n - [[(a + b)(a + c) + (b + d)(c + d)]/n
    where n = a + b + c + d, or, as an alternative,
    HSS = 2(ad - cd)/c[sup2] + b[sup2] + 2ad + (b + c)(a + d)
    The HSS ranges between -1 and +1 and is considered to be a measure of correlation (Doswell et al. 1990): if the PCT threshold is assumed as a proxy for the occurrence of lightning, then HSS measures the fraction of the correct "forecast" of lightning occurrence, after elimination of the correct "forecasts." which are due to random chance.
    All scores were calculated for various PCT thresholds (from 160 to 280 K, with 1-K steps). The maximum value of HSS indicates the best relation between POD and FAR (Doswell et al. 1990) and identifies the corresponding PCT threshold that provides the best scores for the occurrence of lightning. In addition, all score calculations have been made for various numbers of flashes for each PCT threshold. The results of the aforementioned calculations are given in Table 2 and Fig. 3.
    If the criterion for a proxy ("yes" in the contingency table) is at least one recorded CG flash, then the PCT threshold corresponding to the maximum value of HSS (0.48) is 217 K (Table 2). The corresponding scores for POD and FAR are 0.44 and 0.46, respectively. This means that below that value of PCT there is a 44% probability that at least one flash has occurred. The use of one single recorded flash as a criterion may not be appropriate because in the database there are single lightning records in "warm" (in terms of PCT) grid boxes that could have been initiated from adjacent grid boxes with lower PCT. Inspection of Table 2 shows that the best HSS score (0.50) corresponds to a PCT threshold of 202 K and to a number of recorded flashes greater than or equal to 3. Last, Fig. 3 shows how the PCT threshold decreases for the increasing number of recorded flashes, indicating thus that significant lightning activity is correlated with low (cold) PCT values.

C. COMPARISON OF PCT WITH THE ATD/LIS RATIO
    In this section, the relation of PCT with the ratio of cloud-to-ground (as measured by ATD) to total (as measured by LIS) lightning activity will be investigated using the database of 0.25° × 0.25° grid boxes that has been constructed as described in section 2. The analysis showed that almost 96% of the grid boxes with LIS measurements have an ATD/LIS ratio lower than or equal to 1. and thus only less than 4% of the grid boxes have been excluded from the analysis that obviously have been subject to instrument error (either LIS or ATD). Furthermore, in 79% of the grid boxes the ATD/LIS ratio was equal to zero, meaning that, in most of the cases, LIS detected flashes while ATD did not, indicating the absence of cloud-to-ground lightning activity, However, this feature could be also attributed to the higher detection efficiency of LIS during nighttime.
    The ratio of cloud-to-ground lightning to total lightning within storms has received the interest of many authors. Price and Rind (1993) suggested that the proportion of CG lightning within storms is better correlated with the thickness of the cold part of the cloud, Boccippio et al. (2000) used 4-yr data from NASA's Optical Transient Detector and from the National Lightning Detection Network to investigate the geographic distribution of the intracloud-to-cloud-to-ground lightning ratio, over the continental United States. The authors did not find clear evidence to support latitudinal dependence of this ratio and suggested that storm type, morphology, and level of organization
    [Graph or Chart Omitted]
    may dominate over environmental cofactors in the local determination of this ratio.
    With the aim to study the relationship of the CG-to-total-lightning ratio with various ranges of PCT, the ATD/LIS ratio will be used as an indicator of the CG/total lightning ratio. These ratios are not equal because both ATD and LIS suffer from imperfect detection efficiencies (as mentioned in section 2). Additionally, LIS presents a diurnal variability of its detection efficiency. Based on this latter comment the percent distribution for various thresholds of ATD/LIS ratios over all eases, but also separately for nighttime and daytime cases, has been calculated (Fig. 4). Inspection of Fig. 4 reveals that for 80% of the cases the ATD/LIS ratio is zero; thus, there is an absence of CG activity sensed by ATD. This percentage is valid either when taking into account all cases or when treating the night and day cases separately. In the presence of CG lightning, for ATD/LIS ratio values <0.2 the day cases percentage is lower that that of the night cases, while for the ranges 0.4-0.6 and 0.6-0.8 the opposite is true. Although a larger sample should he available in order to derive conclusions, the increased percentages of larger ATD/LIS ratios during the day could he partly attributed to the decreased detection efficiency by LIS during daytime.
    Figure 5 shows the relative frequency of the ATD/LIS ratio within the range of 0-1 (at bins of 0,2), and for various ranges of PCT (from less than 120 K to more than 280 K, with a step of 20 K). It is obvious that the 0 bin (corresponding to only intracloud and cloud-to-cloud activity) has the largest percentage for all PCT ranges. This percentage increases for higher (warmer) PCT values. Indeed, the 0 bin is higher than 90% for PCT values warmer than 200 K. Further, for decreasing values of PCT the percentage of nonzero ATD/LIS
    [Graph or Chart Omitted]
    ratios increases, indicating that colder PCTs are associated with more significant cloud-to-ground lightning activity. More specifically, the 0 bin drops to 45% for PCT values colder than 120 K and to 35% for PCT values ranging from 121 to 140 K. On the other hand, the higher values of the ratio (0.8-1) never exceed 5%-6%. The increase of the number of CG lightning sensed by ATD for decreasing cold PCTs that correspond to increased ice concentration could be attributed to a more
    [Graph or Chart Omitted]
    efficient gravitational charge separation within the clouds, as explained in Williams (2001).
    Further, the cumulative frequency of flash number as recorded by both ATD and LIS for the same ranges of PCT (from less than 120 to more than 280 K, with a step of 20 K) is given in Fig. 6. Inspection of this figure shows that 85% of the ATD flashes correspond to PCT values colder than 200 K, while the cumulative frequency of LIS for PCT < 200 K is 64%. This is an indication that CG lightning as sensed by ATD is strongly related to the colder PCT bins.

4. COMPARISON OF PR REFLECTIVITY PROFILES WITH LIGHTNING ACTIVITY
    In this section the relationship between radar reflectivity profiles and lightning activity is investigated. The analysis is based on the database of reflectivity profiles from 29 orbital PR data and the lightning records from ATD that has been constructed following the method described in section 2b.
    Figure 7a shows the mean radar reflectivity profile calculated for the grid boxes with and without recorded flashes by ATD, The height range of the analysis spans from 1 up to 13 km. This is because over water the lowest detectable bin above the surface echo is at least 0.5-km altitude at nadir, ranging to nearly 1.5 km at the limb, while above 13 km the amount of measurements was very small, because either the systems examined were not so deep or the reflectivity was lower than the lowest value PR can measure (˜ 18 dBZ). It is obvious that the average reflectivity in the presence of lightning has higher values at all vertical levels than that without lightning. The difference of the two profiles ranges from 3 to almost 10 dB, and it is more pronounced at lower altitudes and mainly up to 3.5 km. Note that, climatologically in the study area, the position of the
    [Graph or Char Omitted]
    0°C isotherm is located at a height ranging from 2.5 up to 3.5 km. In the presence of lightning, the rate of decrease of reflectivity is ˜ 2.5 dB km[sup-1] below 5 km while above this height the decrease rate is ˜ 1.7 dB km[sup-1].
    Figure 7b shows the mean reflectivity profile for various numbers of recorded lightning within each grid box. It is evident that reflectivity increases at all altitudes with an increasing amount of lightning, indicating that strong lightning activity is associated with higher values of reflectivity. The mean reflectivity profile associated with one occurrence of lightning is about 5 dB Lower than that associated with more than 10 occurrences of lightning for the heights above 2 km, while this difference exceeds 10 dB at the low levels.
    Figure 7c shows the average maximum reflectivity at each PR level. The difference of the average maximum reflectivity with and without lightning is more pronounced than that of the mean reflectivity (Fig. 7a), ranging from 15 dB at low levels to 6 dB. In presence of lightning the average profile has a reflectivity of about
    [Graph or Chart Omitted]
    44 dBZ near the surface, which drops to about 30 dBZ above 8 km. These results are consistent with the findings of Cecil et al. (2005, presented in their Fig, 9). Allaratz et al. (2003) was one of the few studies devoted to winter thunderstorms and their characterization in the eastern Mediterranean through the analysis of lightning and weather radar data. The authors concluded that in winter rain clouds become thunderclouds if their maximum reflectivity exceeds 45 dBZ and the reflectivity at the level of the -10°C isotherm is larger than 35 dBZ. These findings are consistent with the maximum reflectivity profile in the presence of lightning shown in Fig. 7c.
    Taking into account the fact that the number of grid boxes without lightning is much larger than that with lightning (as recorded by ATD), the percentage distribution of average reflectivity measurements, both per vertical level and per reflectivity unit, has been also calculated, More precisely, the cumulative frequency has been calculated for altitudes ranging from 1 up to 13 km (in 0.25-km steps) and reflectivity values ranging from 15 up to 60 dBZ (in 1-dB steps); The results are shown in Fig. 8. It is obvious that the distribution of cumulative frequency is wider and the reflectivity values are larger for the grid boxes with recorded flashes (Fig. 8a) than those without recorded flashes (Fig. 8b). Indeed, for the grid boxes with recorded flashes, at 3-km height 50% (the median) of the reflectivity values
    [Graph or Chart Omitted]
    are greater than 30 dBZ, while in the absence of lightning only 10% of the grid boxes have reflectivities greater than 30 dBZ. At 6-km height 50% of the reflectivity values are greater than 23 dBZ, while in the absence of lightning only 20% of the grid boxes have reflectivities greater than this value.
    Additionally, the probability that a grid box has at least one occurrence of lightning given its maximum reflectivity has been calculated, and the results are given in Fig. 9. [t is evident that with a maximum reflectivity of ˜ 53dBZ in the low levels (below 3 km) there is a probability of 80% for lightning occurrence. For weaker reflectivities, when 45 dBZ reaches 5 km or 40 dBZ reaches 7 km. then again the probability of lightning occurrence is 80%, indicating that high-reflectivity values aloft are a prerequisite for lightning to occur.

5. CONCLUDING REMARKS -- PROSPECTS
    In this study, lightning observations have been analyzed along with spaceborne passive microwave and radar observations in the central and eastern Mediterranean. The initial motivation was to explore the possibility of defining indicators for the occurrence of lightning activity, because inspection of individual case studies has shown a strong relationship between the regions of cold PCTs and significant lightning activity. This feature has been already addressed in the literature, but the analyzed observations were mainly from over tropical and subtropical areas (Toracinta and Zipser 2001; Toracinta et al. 2002).
    In this paper, the analysis is based on the use of ATD lightning observations, and also the TRMM Microwave Imager brightness temperature at 85 GHz. the precipitation radar reflectivity profiles, and the Lightning Imaging Sensor data over the central and eastern Mediterranean Sea (from 30° to 40°N, and from 10° to 35°E). Namely, the days with the largest observed lightning activity in the period of September 2003-April 2004 have been selected. During these days, low pressure systems associated with important frontal activity have been observed in the study area.
    The first indicator of lightning activity that has been explored is the polarization-corrected temperature at 85 GHz because brightness temperature at this frequency is sensitive to the presence of ice particles, which in turn are necessary for the electrification processes within clouds. It was found that 50% of the grid boxes with flashes have PCTs lower than 225 K, while only 3% of the "no lightning" grid boxes have PCTs below this value. The relation of various PCT thresholds with lightning has been examined based on the calculation of statistical scores. The PCT threshold value that gives the highest HSS score for at least one observed flash is 217 K. This threshold becomes lower as the amount of observed lightning rises, dropping to 170 K for a number of flashes exceeding 10.
    The availability of TRMM/LIS data, along with the ATD data, permitted calculation of the ATD/LIS ratio, which provides an indication of the ratio of cloud-to-ground lightning to total lightning. The ATD/LIS ratio is then related to the brightness temperature measurements at 85 GHz, The results showed that this ratio has higher values for the "colder" PCT values, and decreases as PCT increases. In presence of lightning, at PCTs warmer than 200 K, in 90% of the grid boxes there was only intra-cloud and cloud-to-cloud lightning activity, while below 200 K for at least 20% of the cases cloud-lo-ground lightning has been also detected. This percentage increased for decreasing PCT. The aforementioned results are subject to the fact that the ATD/LIS ratio does not equal the cloud-to-ground/total lightning ratio because both ATD and LIS suffer from imperfect detection efficiencies; in addition, LIS detection efficiency presents a diurnal variability.
    Further, the lightning activity has been correlated with the radar reflectivity measured by the TRMM PR, It was found that in the presence of lightning the average reflectivity profile has higher values than that in absence of lightning, with the difference ranging from 3 up to 7 dBZ. The respective difference with and without lightning of the mean maximum reflectivity profile ranges from 6 to 15 dBZ. Further, a reflectivity profile with values greater than 53 dBZ in the low levels (below 3 km), of ˜ 45 dBZ at 5 km and 40 dBZ at 7 km, is associated with a probability of 80% for lightning occurrence.
    In this study the lightning activity has been related indirectly to the microphysical properties of the studied mesoscale systems in the central and castern Mediterranean during the cold season of the year. It can be considered complementary to numerous studies referenced in this paper that have been performed mainly over the United States and the Tropics. It is in the authors' plan to further investigate the relationship of the presence of lightning activity with dynamical properties of the same systems, such as thermal advection, vorticity advection, or potential vorticity structure. In addition, lightning activity will be related directly to microphysical parameters as provided by cloud-resolving numerical models.
ADDED MATERIAL
    D. K. KATSANOS Institute for Environmental Research and Sustainable Development, National Observatory of Athens, Athens, and Department of Physics, University of Patras, Patras, Greece
    K. LAGOUVARDOS AND V. KOTRONI Institute for Environmental Research and Sustainable Development, National Observatory of Athens, Athens, Greece
    A. A. ARGIRIOU Department of Physics, University of Patras, Rio, Greece
    Corresponding author address: D. Katsanos, National Observatory of Athens, Institute of Environmental Research and Sustainable Development, Lofos Koufou, P. Penteli, 15236 Athens, Grecce.
    E-mail: katsanos@meteo.noa.gr
    Acknowledgments. This work has been jointly financed by the European Union (75%) and the Greek Ministry of Development (25%) in the framework of the program "Competitiveness -- Promotion of Excellence in Technological Development and Research -- Excellence in Research Centers, Action 3.3.1" (M1S64563), as well as by the Greek-NonEU countries Cooperation Program 012 financed by the Greek General Secretariat for Research and Technology. The authors acknowledge useful comments on this work provided by Dr. E. Defer. All TRMM data are downloaded from NASA's corresponding data severs: PR and TMI data are taken from http://lake.nascom.nasa.gov/data/dataset/TRMM/, and LIS data are downloaded from http://thunder.msfc.nasa.gov. The Met Office is thanked for the ATD data.
    TABLE 1. General form of the 2 × 2 contingency table used for various brightness temperature thresholds.

                        No. of flashes observed
PCT [less or equal] threshold      Yes     No            Total
Yes                    a       b            a + b
No                     C       d            c + d
Total                a + c   b + d    a + b + c + d = n

    TABLE 2. Statistical scores for various categories of number of ATD flashes, giving the PCT at 85 GHz that maximizes the HSS, the maximum HSS, the POD, and the FAR.

No. of flashes  PCT at HSS max  HSS max   POD     FAR
  1                  217             0.48      0.44    0.46
  2                  208             0.49      0.46    0.48
  3                  202             0.50      0.50    0.50
  4                  193             0.49      0.47    0.50
  5                  188             0.49      0.48    0.50
  6                  188             0.48      0.51    0.54
  7                  188             0.48      0.56    0.58
  8                  188             0.47      0.59    0.61
  9                  181             0.46      0.54    0.59
 10                  170             0.46      0.46    0.55
>10                  170             0.45      0.48    0.58

FIG. 1. TRMM overpass at 2000 UTC 13 Dec 2003: (a) brightness temperature at 85 GHz (vertical polarization; K); (b) as in (a) with the same grayscale, but a zoom within the rectangle in (a); (c) horizontal cross section of PR reflectivity at 3 km (dBZ); (d) vertical cross section of radar reflectivity [dBZ; with the same grayscale as in (c), across the dashed line in (c)J; (e) total lightning observed by LIS; and (0 cloud-to-ground lightning observed by ATD.
FIG. 2. Cumulative density function of PCT at 85 GHz for grid boxes with and without recorded cloud-to-ground flashes sensed by ATD.
    [Graph or Chart Omitted]
FIG. 3. PCT at 85 GHz, which maximizes the HSS score as a function of the number of ATD flashes.
    [Graph or Chart Omitted]
FIG. 4. Percent distribution of grid boxes (or various thresholds of ATD/LIS ratios: all (white bar), nighttime (gray bar), and daytime (black bar) grid boxes.
    [Graph or Chart Omitted]
FIG. 5. Relative frequency of values of the ATD/LIS ratio (at 0.2 step), for various ranges of PCT.
    [Graph or Chart Omitted]
FIG. 6. Cumulative frequency of ATD (thin line) and LIS (thick line) flashes, for various ranges of PCT
    [Graph or Chart Omitted]
FIG. 7. (a) Profile of mean reflectivity for the grid hoses with (thick line) and without (thin line) lightning. The standard error of the mean is also plotted. The gray area denotes the climatological range of the height of the 0°C isotherm, (b) As in (a), but for various numbers of ATD flashes, (c) As in (a), but for the mean maximum reflectivity.
    [Graph or Chart Omitted]
FIG. 8. (a) Cumulative frequency of reflectivity for the grid boxes with recorded ATD flashes (thick black lines refer to 10%, 50%, and 90%). (b) As in (a), but for the grid boxes without recorded ATD flashes.
    [Graph or Chart Omitted]
FIG. 9. Probability of lightning occurrence given the maximum reflectivity value.
    [Graph or Chart Omitted]

REFERENCES
    Adamo, C., R. Solomon, S. Goodman, D. Cecil, S. Dietrich, and A. Mugnai, 2003: Lightning and precipitation: Observational analysis or LIS and PR. Proc. Fifth ECU Plinius Conf., Ajac-cio, France.
    Altaratz, O., Z. Levin, Y. Yair, and B. Ziv, 2003: Lightning activity over land and sea on the eastern coast of the Mediterranean. Mon. Wea. Rev., 131, 2060-2070.
    Blyth, A. M., H. J. Christian, K. Driscoll, A. M. Gadian, and J. Latham, 2001: Determination of ice precipitation rates and thunderstorm anvil ice contents from satellite observations of lightning. Atmos. Res., 59-60, 217-229.
    Boccippio, D. J., S. J. Goodman, and S. Hcckman, 2000: Regional differences in tropical lightning distributions. J. Appl. Meteor., 39, 2231-2248.
    Boccippio, W.J. Koshak, and R. J. Blakeslee, 2002: Performance assessment of the Optical Transient Detector and Lightning Imaging Sensor, fart 1: Predicted diurnal variability. J. At-mns. Oceanic Technol., 19, 1318-1332.
    Carey, L. D., and S. A. Rutledge, 2000: The relationship between precipitation and lightning in tropical island convection: A C-band polarimetric radar study. Mon. Wea. Rev., 128, 2687-2710.
    Cecil, D. J., S. J. Goodman, D. J. Boccippio, E. J. Zipser, and S. W. Nesbitt, 2005: Three years of TRMM precipitation features. Part I: Radar, radiometric, and lightning characteriscs, Mon. Wea. Rev., 133, 543-566.
    Christian, H. J., and Coauthors, 1999a: The lightning imaging sensor. Proc. 11th Int. Conf. on Atmospheric Electricity, Hunts-ville, AL, ICAE, 746-749.
    Christian, H. J., and Coauthors, 1999b: Global frequency and distribution of lightning as observed by optical transient detector (OTD). Proc. 11th Int. Conf. on Atmospheric Electricity, Huntsville, AL, ICAE, 726-729.
    Christian, R. J. Blakeslee, S. J. Goodman, and D. A. Mach, 2000: Algorithm theoretical basis document (ATBD) for the lightning imaging sensor (LIS). NASA, 53 pp. [Available online at http://eospso.gsfc.nasa.gov/eos_homepage/for_scienlists/atbd/docs/LIS/atbd-lis-01.pdf.]
    Defer, E., K. Lagouvardos, and V. Kotroni, 2005: Lightning activity in the eastern Mediterranean region. J. Geophys. Res., 110, D24210. doi:10.1029/2004JD005710.
    Doswell, C. A., III, R. Davies-Jones, and D. L. Keller, 1990: On summary measures of skill in rare event forecasting based on contingency tables. Wea. Forecasting, 5, 576-585.
    Holt, M. A., P. J. Hardaker, and G. P. McClelland, 2001: A lightning climatology for Europe and the UK. 1990-99. Weather, 56, 290-296.
    Katsanos, D., K. Lagouvardos, V. Kotroni, and A. Argiriou, 2007: Combined analysis of rainfall and lightning data produced by mesoscale systems in the central and eastern Mediterranean. Atntos. Res., 83, 55-63.
    Keogh, S. J., E. Hibbert, J. Nash, and J. Eyre, 2006: The Met Office Arrival Time Difference (ATD) system for thunderstorm detection and lightning location. Met Office Forecasting Research Tech. Rep. 488, 22 pp. [Available online at http://www.metoffice.com/research/nwp/publications/papers/technical_reports/2006/FRTR48S/FRTR488.pdf]
    Kummerow, C., W. Barnes, T. Kozu, J. Shiue, and J. Simpson, 1998: The Tropical Rainfall Measuring Mission (TRMM) sensor package. J. Atmos. Oceanic Technol., 15, 808-816.
    Mohr, K. I., E. R. Toracinta, E. J. Zipser, and R. E. Orville, 1996: A comparison of WSR-88D reflectivities. SSM/I brightness temperatures, and lightning tor mesoscale convective systems in Texas. Part II: SSM/1 brightness temperatures and lightning. J Appl. Meteor., 35, 919-931.
    Orville, R. E., 1981; Global distribution of midnight lightning September to November 1977. Mon. Wea. Rev., 109, 391-395.
    Price, C., and D. Rind, 1993: What determines the cloud-to-ground lightning fraction in thunderstorms? Ceophys. Res. Lett., 20, 463-466.
    Spencer, R., H. M. Goodman, and R. E. Hood, 1989: Precipitation retrieval over land and ocean with the SSM/I: Identification and characteristics of the scattering signal. J. Atmos. Oceanic Technol., 6, 254-273.
    Toracinta, E. R., and E. J. Zipser, 2001: Lightning and SSM/I-ice-scattering mesoscale convective systems in the global Tropics. J. Appl. Meteor., 40, 983-1002.
    Toracinta, D.J. Cecil, E.J. Zipser, and S. W. Nesbitt, 2002: Radar, passive microwave, and lightning characteristics of precipitating systems in the Tropics. Mon. Wea. Rev., 130, 802-824.
    Williams, E. R., 2001: The electrification of severe storms. Severe Convective Storms, Meteor. Monogr., No. 50, Amer. Meteor. Soc., 527-561