1. Introduction
Lightning is one of the most powerful spectacles of nature, both for its danger and for its visual aspects. It also discharges a large amount of energy to Earth, produces significant chemical transformations (e.g., nitrogen oxides; Bond et al. 2002; Price and Rind 1994), and often has deadly effects (e.g., Holle et al. 2005; Ashley and Gilson 2009). Broadly speaking, lightning flashes may be grouped into two categories: those that strike the ground [cloud-to-ground (CG) lightning], and those that do not (cloud-to-cloud or intracloud lightning). CG lightning is usually the cause of interruptions in electric power transmissions and distribution, electromagnetic interference, and (quite often) extensive forest fires (e.g., Podur et al. 2003). Therefore, accurate information regarding CG lightning–prone areas is useful for safety purposes and for identifying the best locations for electric, telecommunications, and sensitive industrial infrastructures.
Such accurate information has become available with the introduction of the Lightning Location System (LLS), which occurred in the late 1980s (Krider et al. 1981, 1989). LLS has made it possible to compute the lightning densities, defined as the number of lightning flashes per a given unit of area and time. The most studied area is the United States (e.g., Fisher et al. 1993; Orville and Huffines 1999, 2001; Villarini and Smith 2013). Other frequently examined areas are Canada (Herodotou et al. 1993; Nash and Johnson 1996), Brazil (Pinto et al. 1999), Japan (Asakawa et al. 1997; Shindo and Yokoyama 1998), Australia (Kilinc and Beringer 2007), and, recently, some European countries such as Germany (Finke and Hauf 1996), the United Kingdom (Holt et al. 2001), Italy (e.g., Bernardi and Ferrari 2004), Spain (Rivas Soriano et al. 2005), Austria (Schulz et al. 2005), Romania (Antonescu and Burcea 2010), the Czech Republic (Novák and Kyznarová 2011), Estonia (Enno 2011), Finland (Mäkelä et al. 2011), and Portugal (Ramos et al. 2011; Santos et al. 2012).
Very recently, Anderson and Klugmann (2014) computed a lightning climatology across the full European continent, using only 5 years of lightning data observed by the new Arrival Time Differencing Network (ATDnet; Gaffard et al. 2008). In accord with that analysis (Anderson and Klugmann 2014, their Fig. 4) it appears that the main areas with more than 4 flashes km−2 yr−1 are the Western Alps and northeastern (NE) Italy. In particular, over all of Europe they found only two spots with lightning densities above 6.5 flashes km−2 yr−1 during the period of investigation. The largest area is located along the eastern Alps, while the second is located in the Ticino region, on the border between Switzerland and Italy, and is described in Nisi et al. (2014). That result is in very good agreement with the 1971–2008 precipitation climatology across the Alps, derived by a very dense network of rain gauges, recently published by Isotta et al. (2014) (cf. with their Fig. 6). Even in that case, the largest rain-density area in the whole domain is on the Julian Alps, along the border between Italy and Slovenia. From these results NE Italy appears to be a very active area from a meteorological point of view and deserves dedicated study.
By definition, a thunderstorm is a cumulonimbus associated with the presence of lightning and thunder, usually accompanied by strong gusts of wind and rain, sometimes also with hail (Glickman 2000). Storms characterized by intense lightning activity have been associated with strong convective events, like supercells, squall lines, mesoscale convective systems, and tropical cyclones (e.g., Goodman and MacGorman 1986; Branick and Doswell 1992; Rutledge et al. 1993; MacGorman and Burgess 1994). The complex interrelations between lightning and convective rainfall (e.g., Zipser 1994; Petersen and Rutledge 1998; Tapia et al. 1998; Rivas Soriano et al. 2001), flash floods (Petersen et al. 1999), hail (Changnon 1992; Solomon et al. 2004; Soula et al. 2004), and high winds and tornadoes (MacGorman and Burgess 1994; Carey and Rutledge 2003; van den Broeke et al. 2005) have been investigated in the past, given the severe consequences and damage that they can cause to society (Curran et al. 2000; Changnon 2001).
It would be very interesting to also study in detail these relations in NE Italy, but that is not easy because of the many different triggering mechanisms (e.g., synoptic cold fronts, convergence lines, thermal boundaries, orographic lifting) that can initiate and sustain storms in this area and because of the complex microphysics underlying cloud electrification and consequent lightning discharges (Williams et al. 2005; Rakov and Uman 2007). What has been learned from previous work is that the most intense convective events produce more CG lightning, together with heavy rain and strong winds. For example, Fig. 3 in Manzato (2003) presents a 3D scatterplot of CG lightning counts versus 10-m wind gusts versus accumulated rain-gauge observations every 6 h in the plain of the Friuli Venezia Giulia region (hereinafter FVG, northeastern Italy). Manzato’s Fig. 3 shows a strong clustering among heavy rain episodes, strong wind gusts, and very high CG lightning activity, while a similar kind of clustering occurs also for cases with weak rain, small maximum winds, and a low number of CG lightning events.
The target area of the present study is northeastern Italy, including the Po Valley, the central and eastern Alps, and parts of the neighboring countries of Switzerland, Austria, Slovenia, and Croatia. The Po “valley” is in reality a large plain basin oriented from west to east (for more than 650 km) and surrounded on the northern and western sides by the Alps, on the southern side by the Apennines, while on the eastern side it is limited by the Adriatic Sea. This region has been found to contain one of the areas in central Europe most highly affected by CG lightning (Feudale et al. 2013). The climatology of thunderstorms in the Po Valley has been studied by Cacciamani et al. (1995), while Manzato (2007) and Manzato (2012) focused, respectively, on the rain–thunderstorm and hail climatologies on the FVG plain subdomain. From these works and from the analysis of single-case studies (e.g., Rotunno and Ferretti 2003; Borga et al. 2007; Davolio et al. 2009; Manzato et al. 2014), the meteorological role of the north Adriatic Sea in providing moisture flux for feeding convection has been highlighted. For this reason, the distance from the sea is also a parameter that will be investigated in relation with the CG lightning spatial distribution. Moreover, given the role that complex topography often plays in storm initiation (Bourscheidt et al. 2009; Nisi et al. 2014), the relationships with the topographic characteristics, like orography height, altitude standard deviation, and its gradients [called directional slopes in Nisi et al. (2014)], will be considered.
In addition, following the results of Bell et al. (2008, 2009) and Rosenfeld and Bell (2011) regarding the possible connection between thunderstorms and anthropogenic aerosols (atmospheric pollution), this study also investigates possible relationships between lightning occurrence and anthropogenic emissions, carrying out analyses both on weekly (high frequency) and annual (low frequency) time scales.
The article is organized as follows. After a brief description of the data used, which is given in section 2, the spatial and temporal characteristics are described in section 3, with some factors that may influence the CG lightning distribution discussed. In section 4, evidence of a relationship between the CG lightning distribution and topography is presented, and in section 5 the possible relationship of the CG lightning distribution with the distribution of particulate diameter ≤ 10 μm (PM10) and instability indices is discussed. The conclusions follow in section 6.
2. Data
The domain of interest for this study is precisely defined as 44°–48°N and 9°–15° E. As seen in Fig. 1, this domain includes a large part of the Po Valley and the central-northeastern part of the Alps and Prealps on the northern side. The northern part of the Adriatic Sea, included on the southeast side of the domain, plays a role in mitigating the continental climate and is a source of moisture advected inland by southerly, southeasterly, or southwesterly flows. In fact, the whole Adriatic Sea is mostly parallel to the Apennines on the west side and to the Dinaric Alps on the east side, so that every flow with a southerly component can be “channelized” inside and pushed toward the Po Valley.
Target area with topography (grayscale; m), EUCLID lightning detector locations (gray triangles), the Udine WMO 16044 sounding station (gray square), and labels identifying the PM10 stations (listed in Table 1).
Citation: Journal of Applied Meteorology and Climatology 53, 12; 10.1175/JAMC-D-14-0037.1
The CG lightning data used in this analysis are provided by the Centro Elettrotecnico Sperimentale Italiano–Sistema Italiano Rilevamento Fulmini (CESI/SIRF; Bernardi and Ferrari 2004) within the framework of the European Cooperation for Lightning Detection (EUCLID) Network. The dataset consists of all CG lightning records, with both positive and negative polarities, observed between January 1995 and December 2012 (18 yr). These measurements are obtained through a network of Vaisala, Inc., “IMPACT ESP” sensors, which detect the return stroke (usually considered the largest discharge transfer during a lightning event) with a location accuracy of 500 m or better (e.g., http://www.hobeco.net/pdf/IMPACTESP_B210324en-A.pdf).
While cloud-to-cloud lightning accounts for the majority of the total lightning produced (e.g., Boccippio et al. 2001; de Souza et al. 2009), the Italian CESI/SIRF network measures only the CG lightning. Since the authors are not aware of other LLS networks also detecting cloud-to-cloud lightning and operating in Italy for at least 10 years, only the CG lightning data have been considered in the present work. Until 1999, the CESI/SIRF measurement network system was using up to 16 sensors located in Italy. Out of these, only 5 are inside the study area or are less than 100 km from its borders. In 2000, SIRF entered in the EUCLID Network and the lightning detection system was strengthened by the European collaboration. Another four sensors (near the study domain) were added, for a total of nine sensors monitoring the target area (their locations are shown by the nine gray triangles in Fig. 1). Bernardi and Ferrari (2004) and Feudale et al. (2013) found some small differences in the network efficiency before and after 1999 that, in any case, should not affect the total derived CG lightning climatology.
Following Feudale et al. (2013), the lightning dataset is first preprocessed to discard multiple strokes, considering as a single lightning event all the flash records reported in the same second and in the same area of 0.01° × 0.01°. After this preprocessing, the final number of useful records is about 7 million. Since Bernardi and Tommasini (2012) suggest that, given the network accuracy of 500 m in lightning location, a minimum grid of 2 km × 2 km should be considered, the data were regridded with a spatial resolution of 0.03° in longitude (~2.41 km) and 0.02° in latitude (~2.13 km), so that every grid box covers an area of about 5.13 km2 (the total domain area is about 205 300 km2). Finally, the gridded data were aggregated into hourly matrices.
In the second part of this study, the PM10 particulate matter data from 19 stations having time series of adequate length (from 2002 to 2010) are analyzed. The list of these stations is shown in Table 1, where the identification (ID) label used to show their location inside the domain (see Fig. 1) is defined. Data for the daily PM10 (usually accumulated from 0000 to 0000 local time) were retrieved through BRACE, an online air-quality database (http://www.brace.sinanet.apat.it/web/struttura.html) created by the Agency for Environmental Protection and Technical services [APAT, now known as the Superior Institute for Environmental Protection and Research (ISPRA)]. After this procedure, the correlation between the mean-domain value for the lightning density and the PM10 values is studied on annual and weekly time scales.
List of the 19 monitoring stations for PM10 analysis in the Po Valley area. Columns show the identification name used in Fig. 1 (ID), the town of the station location (city), the latitude of the station (lat), the longitude of the station (lon), and the altitude of the station (alt).
Finally, the high-vertical-resolution Udine–Campoformido [World Meteorological Organization (WMO) code 16044, latitude = 46.03°N, longitude = 13.18°E, altitude = 94 m; shown by a gray square in Fig. 1] radiosoundings made by the Italian Aeronautica Militare during the April–October 1995–2012 period were used to evaluate the climatological relation between the lightning probability and two indices related to the atmosphere instability. In addition, the weekly cycle for the probability of having severe weather in the FVG plain (derived from lightning and FVG surface station observations) was derived and compared with the other weekly cycles.
3. Temporal and spatial CG distribution
a. Mean spatial distribution
Figure 2 shows the CG lightning density (defined as lightning flashes h−1 km−2) in the whole study domain and period. As was already seen in a previous study on a slightly smaller domain and with a slightly shorter data time series (Feudale et al. 2013), the CG lightning in this case also has an arc-shaped maximum density in NE Italy. In particular, there are many grid boxes with maximum densities above 7 × 10−4 flashes h−1 km−2 on the southern flank of the Carnic and Julian Prealps.1 The core of this maximum (on the Julian Prealps) appears to be approximately 40 km to the southeast with respect to the maximum area shown in Fig. 4 of Anderson and Klugmann (2014). This shift may be due to their much shorter time series (5 vs 18 yr) or to the lower spatial resolution of the ATDnet network2 with respect to the EUCLID network. Also, the maximum density shown in Fig. 2 is much closer to the maximum in precipitation shown in Fig. 6 of Isotta et al. (2014), even though the latter seems to be shifted by a few kilometers eastward.
Spatial distribution of lightning density defined as the average number of CG flashes h−1 km−2 during the period 1995–2012 (18 yr) on a 0.03° × 0.02° grid. The scale is multiplied by a factor of 104 for convenience.
Citation: Journal of Applied Meteorology and Climatology 53, 12; 10.1175/JAMC-D-14-0037.1
Other areas with density above 5 × 10−4 flashes h−1 km−2 lie on the southern flank of the Prealps of the Veneto and Lombardia regions (just south of the 46° parallel) and on the southwestern flank of the Apennines in Tuscany (in the southwestern corner of the domain). Isolated spikes such as one in Austria along the border with Italy (46.58°N, 13.46°E) and another in Croatia (near the northern point of the island of Cres) correspond to transmitter antenna towers, which attract lightning discharges, as discussed in Feudale et al. (2013). The highest-density records in the domain are above 10 × 10−4 flashes h−1 km−2, corresponding to about 9 flashes yr−1 km−2.
On the other hand, there is an interesting minimum of lightning density in the region downstream of the Apennines chain, between the cities of Piacenza and Reggio Emilia (respectively, PC and RE in Fig. 1). The other large area of minimum density is between eastern Switzerland (Grisons region) and western Austria (Vorarlberg area and the southern part of Nordtirol), including the small state of Liechtenstein.
Comparing Fig. 2 with Fig. 1, it is possible to see a general correlation between orography and lightning density. At fine scales lightning density seems to show the “fingerprint” of the orography. In fact, many small alpine valleys have lower densities than the surrounding peaks. But in different parts of the domain this relation is more complex than what one can see at first glance. For example, the Italian Prealpine peaks have much higher densities than do the Alpine peaks or Austrian Prealps. More generally, the southern slopes of mountains have higher lightning densities than do the corresponding northern sides. Finally, there seems to be an increasing gradient of density going from the west toward the east. From this qualitative first analysis it appears that both topographic characteristics and the distance of peaks from the sea (which is a major source of moisture when southerly winds blow) are key factors in explaining the observed lightning density.
b. Mean temporal distribution
To evaluate the temporal characteristics of the lightning distribution and to see if there is some kind of periodic pattern of behavior in the lightning activity, and hence in the thunderstorm activity, the lightning density mean value in the study domain of Fig. 2 is computed and analyzed on different time scales.
Figure 3a displays the interannual mean value of CG lightning in the target domain, indicating large year-to-year variability, with a maximum in 2002 greater than 3.4 × 10−4 flashes h−1 km−2, followed by 2009, 2007, and 2008, with respective values greater than 3 × 10−4 flashes h−1 km−2. Even though there is a positive apparent trend toward a higher frequency of CG lightning activity, the result of a linear regression (straight line) shows that this positive trend is not statistically robust (R = 0.05 and p value = 0.39).
Domain means of the (a) CG lightning annual distribution and the 18-yr linear trend (straight line); (b) CG lightning monthly distribution (thin line) with the ±10-day moving average (thick line); (c) CG lightning weekly distribution, where the vertical lines are for the PM1 standard error; and (d) CG lightning daily cycle distribution. Units are number of flashes h−1 km−2 × 104.
Citation: Journal of Applied Meteorology and Climatology 53, 12; 10.1175/JAMC-D-14-0037.1
Figure 3b depicts the monthly mean value of CG lightning flashes on a yearday scale, and the thick line is the smoothed ±10-day moving average, which filters high-frequency values. Figure 3b shows that the active period starts approximately in April, reaches its maximum during the months of July and August (around 8 × 10−4 flashes h−1 km−2), and fades thereafter until October–November, when the lightning activity, though small, still persists. In the FVG plain, Manzato (2007) has found a similar pattern of behavior for the “strongest thunderstorms” (see the “CALCA6h> 0.79” histogram in his Fig. 4a) and has explained it as follows.
During the spring, convection is forced by frequent Atlantic cold fronts, while during the summer this pattern is often prevented by strong Mediterranean anticyclones and hence the convection is more due to stronger radiative warming, which produces larger potential instability (in this case triggering is often provided by valley or sea breezes, orographic lifting, and convergence lines). Finally, during autumn there are fewer anticyclone situations (with more cold fronts from the Atlantic Ocean), and while the potential instability is lower than in summer (Manzato 2007, his Fig. 5), the Adriatic Sea is much warmer than in spring, so that there is more available “fuel” for convection, sometimes embedded in larger precipitating structures. In addition, Costa et al. (2001) state that one of the most representative synoptic situations for producing heavy rainfall in northern Italy, in particular during the spring and autumn, is the presence of a trough in the western Mediterranean that brings midtropospheric moist southwesterly flow toward the Alps.
Decreasing the temporal scale of analysis, Fig. 3c shows the characteristic CG lightning activity on a day of the week time scale. Following work done by Bell et al. (2009) and Rosenfeld and Bell (2011), the horizontal weekly scale is replicated, to get a better picture of the transition during the weekend. Above the mean value computed for each day of the week using all of the available data from 1995 to 2012, a ± 1 standard error is also shown to give an idea of the data dispersion. Figure 3c shows that there is a gradual increase in the CG lightning activity starting from Sunday (the day with the minimum CG lightning activity) and reaching a maximum on Friday, followed by a rapid drop on Saturday. There is also a local minimum observed on Wednesday, but the error bars are large, so that it is not clear if it is statistically significant. Apart from that, the weekly periodicity seen in Fig. 3c could be potentially related to anthropic activities, like increased pollution during working days, as will be investigated in more detail in section 5.
Potential instability strongly depends on solar heating and hence varies with time of day. This is reflected in the thunderstorm activity as a function of the period of the day, as clearly evident in Fig. 3d, showing the mean daily distribution of CG lightning after averaging over the entire domain in hourly steps. The behavior is vaguely sinusoidal, with maximum activity during the afternoon, (peak greater than 4 × 10−4 flashes h−1 km−2 between 1600 and 1700 UTC, i.e., between 1800 and 1900 local time during summer) and a minimum during the morning (about 1 × 10−4 flashes h−1 km−2 between 0700 and 1000 UTC). The temporal interval separating the morning minimum from the afternoon maximum is only 7 h, indicating that the semidiurnal component is large.
These seasonal (Fig. 3b) and daily cycle (Fig. 3d) outcomes are expected since thunderstorm activity in Europe generally reaches its maximum during summer, with a peak in the afternoon due to the maximum radiative heating (e.g., Finke and Hauf 1996; Holt et al. 2001; Rivas Soriano et al. 2005; Manzato 2007; Antonescu and Burcea 2010; Novák and Kyznarová 2011; Gladich et al. 2011). It is not always so: for example, Santos et al. (2012) have found the maximum lightning activity in Portugal to be during the month of September, followed by April, with a bimodal distribution.
c. Spatial distribution for different months
In the previous section only the mean value of the whole domain was studied, but the CG lightning distribution presents different spatial patterns on a month-by-month basis. Starting gradually in May (not shown), the June CG lightning mainly affect the Prealps area, in particular across the eastern part of the domain (Carnic and Julian Prealps; Fig. 4a). They become more widespread (but predominantly confined in the 45.2° ≤ latitude ≤ 47.0° band) and reach their maximum frequency during July, as shown in Fig. 4b. In August, CG lightning also begins to affect quite frequently the coastal areas near the Adriatic Sea and the Ligurian Sea (the southwestern corner in Fig. 4c), while the activity in the inner central Alps is much lower than in July. Finally, in September (Fig. 4d) the CG lightning distribution is almost totally centered on the northern Adriatic Sea (around the Gulf of Trieste, near the TS label in Fig. 1, and the Kvarner Gulf, near Cres), with a secondary peak near the Ligurian Sea.
CG lightning spatial distribution during (a) June, (b) July, (c) August, and (d) September, calculated on an 18-yr basis (1995–2012). The scale shows the flashes h−1 km−2 × 104.
Citation: Journal of Applied Meteorology and Climatology 53, 12; 10.1175/JAMC-D-14-0037.1
This migration from the Alpine–Prealpine area during May–July toward the coastal area (in particular those that have steep reliefs in the vicinity, like the Apennines or the Veneto and FVG Prealps in Italy, or the northern part of the Dinaric Alps in Slovenia and Croatia) in August–September clearly indicates the importance of the interaction between the warm sea of late summer and initial autumn [e.g., sea surface temperature mean values shown in Fig. 4b of Manzato (2007)] and the orography.
4. CG relationship with topography
As shown at the end of the previous section and also found by previous studies (Rivas Soriano et al. 2005; de Souza et al. 2009; Antonescu and Burcea 2010), CG lightning density is also related to topography. To quantify this more objectively, in this section the analysis of the lightning distribution will be stratified for three different areas, characterized by their altitude range.
Figure 5c shows a “density plot” created to evaluate the domain grid-box distribution based on their number of CG lightning flashes and altitude ranges. Along the abscissa the altitude levels in 20-m steps are shown, while along the ordinate the mean number of lightning flashes is presented. The x–y colored field is the logarithm of the number of domain grid boxes that have CG lightning counts and an altitude in that given range of lightning frequency and orography altitude range. It is possible to see a general trend of decreasing density (which should not be confused with the lightning density field) with higher lightning frequency and with increasing altitude. The highest density is found for areas with altitudes less than 60 m and in particular in the first 0–20-m bar near the ordinate (orange/yellow points), while another relative maximum occurs above 400 m of altitude.
(a) Meridional variation of the orography standard deviation (south–north direction) (m). (b) Map of the distance from the coast (°). (c) Density of grid points in the corresponding range of CG lightning frequency (log) and orography altitude (m). (d) The target area divided into three main zones: the coastal–sea (cyan), the plains–foothills (green), and the mountain (maroon).
Citation: Journal of Applied Meteorology and Climatology 53, 12; 10.1175/JAMC-D-14-0037.1
a. CG lightning distribution across three topographic zones
Based on the “altitude–frequency” analysis in Fig. 5c, the lightning characteristics will be stratified into three altitude ranges, chosen for convenience using the 5- and 400-m thresholds. The first area is called the sea–coastal area, identified by grid boxes covering areas with altitudes below 5 m MSL (i.e., about 17% of the entire domain), including all of the seashore region and oversea zone (cyan area in Fig. 5d). The second area, corresponding to the “plains” and “foothills” region, is identified by grid boxes with altitudes in the 5–400-m range (green area in Fig. 5d, which covers about 24% of our domain); finally, the third area, called for convenience the mountains area even though mountains usually have higher elevations, is identified by grid boxes with altitudes above 400 m (maroon area in Fig. 5d, representing the remaining 59% of the domain). For each of these three areas, we computed the mean of the CG lightning flashes total number and evaluated its moving-average annual trend by yearday (thick line in Figs. 6a,c,e) and its moving-average daily cycle (Figs. 6b,d,f).
(a) Time series of index lightning per yearday and (b) its daily behavior during the 24 h in the coastal–sea area (area 1, height between 0 and 5 m). (c),(d) and (e),(f) As in (a) and (b), but for the internal–foothill area (area 2, height between 5 and 400 m) and the mountains area (area 3, height above 400 m), respectively. The thick curves in (a),(c), and (e) represent the ±10-day moving average.
Citation: Journal of Applied Meteorology and Climatology 53, 12; 10.1175/JAMC-D-14-0037.1
Figure 6 illustrates the different lightning characteristics in the three subareas. In particular, while in Fig. 6a the ±10-day moving average of the lightning probability from mid-June to July is always lower than 6 × 10−4 flashes h−1 km−2, the opposite occurs in Fig. 6e and, from late June, also in Fig. 6c. On the contrary, while the lightning mean density in the sea–coastal area (Fig. 6a) remains above 6 × 10−4 flashes h−1 km−2 until mid-September, that happens only until mid-August for the mountain area (Fig. 6e).
This analysis provides a clear example of how convection moves from the internal mountain areas in late spring to the coastal–sea regions in the late summer and autumn. Apart from the synoptic discussion already presented concerning the frequency of cold fronts (higher in spring and autumn than during summer), another important role can be played by the different triggering mechanisms acting during the storm season. In particular, during spring the orographic lifting of low-level moist air by fronts or valley breezes seems to prevail (more storms in the mountain region), whereas during late summer and autumn (more storms near the sea) other mechanisms are implicated, such as sea-breeze- or mesoscale-induced moisture fluxes coming from the sea, which during autumn is also still warm, because it has a higher heat capacity than does the inland territory.
Looking at the diurnal cycle of Figs. 6b,d,f, the situation is even more different in the different areas. While in the sea–coastal area (Fig. 6b) there is not a strong daily cycle of thunderstorm activity (just lower activity during late morning to early afternoon, between 0900 and 1500 UTC), in the plains–foothills area (Fig. 6d) and even more so in the mountain region (Fig. 6f) there is a much stronger daily cycle. Note that similar trends were found in Fig. 6 of Tuomi and Mäkelä (2008) for the land/sea areas of Finland.
To explain the different pattern of behavior in the diurnal cycle, a different triggering mechanism can be more appropriate than different synoptic patterns, because the latter usually have longer time scales than do the short variations during the period of the day. In particular, in the coastal–sea area, nighttime convection (Parker 2008; French and Parker 2010) seems to be much more frequent than in the two other areas, which provides further evidence that different triggering mechanisms, such as convergence lines, low-level moist jets, sea–land boundaries, bores, etc., play an important role in this area in particular during the night, when the low levels are more stable than during afternoon, and the convection can exist with a more elevated cloud base. On the other hand, over the plain and mountain areas, the classical theory based on potential instability (lifting the low levels with high equivalent potential temperature Θe) seems to be the main mechanism for thunderstorm development, because it follows the solar heating and reaches its maximum values during the afternoon hours.
The present analysis reveals that the different spatial distributions of convection between the coastal area (occurring primarily during the evening–nighttime in late summer) and the mountain area (occurring primarily in the afternoon during spring and summer) are probably related to different characteristics of the triggering mechanisms, which depend also on the distance from the Adriatic Sea and on its mean temperature.
b. CG lightning correlation with topographic characteristics of the area
To investigate in more detail the origin of the different timings of thunderstorm activity in the different target areas described in Fig. 6, relations with a few topographic characteristics have been studied. First, for each grid box of the domain we computed the standard deviation of the orography in the neighborhood (related to the slope of the orography and hence to the orographic lifting mechanism) with 5′ resolution, in order to capture the main structure on the larger scale, avoiding “noisy” effects due to a higher resolution. Subsequently, also following Nisi et al. (2014), we examined possible pointwise correlations of the CG lightning mean probability with the following quantities for all of the 5′ × 5′ grid boxes:
the simple orography height,
the standard deviation (std) of the orography,
the meridional (Fig. 5a) gradient of the orography std,
the zonal gradient of the orography std,
the sum of the meridional and the zonal gradients of the orography std,
the distance of each grid point from the sea (Fig. 5b), and
the combined correlation of the distance from the sea with one of the other quantities listed before (from points 1–5).
5. The CG relationship with anthropogenic emissions
In section 3 we discussed the general temporal characteristics of the lightning distribution and found that a weekly cycle seemed to arise from the analysis (see Fig. 3c). Evidence of weather weekly cycles has been reported previously in studies by Bell et al. (2008, 2009) in the United States, especially with respect to precipitation, lightning, and storm occurrence. However, for Europe, Laux and Kunstmann (2008) analyzed and identified a weekly cycle only with respect to temperature (which to a first approximation is similar to the trend shown in Fig. 3c), while they found no significant weekly trend in the precipitation field. The hypothesis proposed by Bell et al. (2008, 2009) and Rosenfeld and Bell (2011) (hereinafter called the Bell and Rosenfeld hypothesis) predicts that there is a dependence of thunderstorm occurrence on a particular day of the week (therefore, a weekly cycle) that is due to the variation of pollution with the day of the week. In fact, consistent with the weekly cycles of meteorological variables, a weekly periodicity of air pollutants and aerosols was also found (e.g., Gong et al. 2007).
Any such weekly cycle may be considered evidence of anthropogenic influences on the weather in that area, because, for the Bell and Rosenfeld hypothesis, the weekly cycle of working weekdays and resting Sunday should be associated with weekly varying levels of particulate air pollution. Pernigotti et al. (2012) found that the Po Valley is a hot-spot area in Europe for its high concentration of pollutants (especially during the winter season) and that the pollutant concentration levels in this area are expected to remain problematic even after applying antipollution strategies.
To see if the Bell and Rosenfeld hypothesis could be applied in this particular area, some representative monitoring stations distributed in the target domain were chosen. They are listed in Table 1 and their locations are shown on the map in Fig. 1. For these 19 stations the daily PM10 data were downloaded from the BRACE dataset. The data were available from 2002–03 to 2010, but for the stations of Padova and Cremona they were available only from 2006 (time series length reduced to 5 yr).
a. PM10 cluster analysis
Surface PM10 is a highly variable field in space and time and for this reason it is difficult to relate its value with local meteorological phenomena such as thunderstorms. The approach followed in this work is to look for an average value representative of the entire domain, to be related to the mean value of lightning probability in the whole domain. Both fields are studied only from April to October, avoiding the high values of PM10 recorded during the winter season. To extract a coherent mean PM10 signal from the 19 different surface stations, two analyses were performed. The first one is a k-means clustering analysis based on the mean and standard deviation of the annual PM10 value. From Fig. 7a it is possible to identify three groups: the high mean value group consists only of two stations (Verona and Mantova), while the low value group has mainly stations near foothills (Gorizia, Osoppo, Udine, and Pordenone) or in the mountains (Trento), located on the northeast side.
(a) Result of the cluster analysis on the monitoring stations by the k-means method with the choice of three clusters, using as a discretization, the mean and the standard deviation of the PM10 concentration; the plus sign, circle, and triangle represent the centroid of each cluster. (b) Dendogram of the results of the cluster analysis on the monitoring stations, using as the distance measure the correlation between their PM10 concentrations.
Citation: Journal of Applied Meteorology and Climatology 53, 12; 10.1175/JAMC-D-14-0037.1
The second analysis is a hierarchical clustering dendrogram (Fig. 7b) using as a metric (i.e., measure of distance between stations) the mutual linear correlation between the daily PM10 values of each pair of stations. While the first analysis is a low-frequency analysis, looking at the annual first two moments, the second one is more of a high-frequency analysis, looking for the daily coherence among different stations. In Fig. 7b, stations with high correlations are joined in the bottom part of the dendrogram and groups having larger distances apart (lower correlation) are joined in subsequent steps. The level with three groups (the three gray boxes) shows different results than those found with the k-means clustering. In particular, there are three stations that join at a very large distance: Trieste (coastal station) for the first group and Trento and Bolzano (stations inside the Alps region) for the third cluster.
To evaluate the strength of the correlation between the lightning probability and a spatially averaged PM10 signal, the result of the second clustering method is considered more appropriate. Based on the results of the hierarchical analysis, the stations of Trieste, Trento, and Bolzano were excluded from the following analysis, because they are considered to be outliers with respect to the features described by the other stations, which are mainly located on the plain or foothills. Also, it should be noted that inside the small valley with steep mountain surroundings (like the alpine valleys of Trento and Bolzano) the dispersion of pollution is more difficult than in the open plain. Moreover, in the mountainous part of the domain it is more likely that the emission of anthropogenic pollution is lower than in the plain area, where most of the industry and human population are concentrated. On the contrary, Trieste is located along the coast and is ventilated (e.g., frequent sea breezes or even strong bora winds), so that pollution can very easily be dispersed. These are additional reasons to remove Trieste, Trento, and Bolzano from the list of stations used to compute the “mean” PM10 value for our Po Valley domain.
b. CG lightning and PM10 annual trend
In this section the temporal characteristics of the mean PM10 concentration are analyzed in parallel with the temporal characteristics of the mean CG lightning distribution to highlight possible correlations. The analysis of daily values (24 h accumulated) is divided into two temporal scales: an annual scale, based on a low-pass-filtered time series with respect to the yearday (moving average of ±10 days), and a weekly scale, which also takes into consideration the high-frequency part of the signal.
Figure 8a shows the annual trend with a yearday resolution, after a low-pass filter of ±10 days, for the mean CG lightning distribution (black line) and the PM10 distribution, averaged on the selected 16 monitoring stations (gray line). As already described in section 3, the CG lightning distribution is more active between April (day 91) and October (day 304). The PM10 mean concentration for the same area shows a complementary trend: higher concentrations are recorded during the winter and the autumn. A drastic decrease occurs in mid-April, moving to low concentrations in spring and summer, and then increasing again in September. In fact, in autumn and winter high anthropogenic emissions (due to the building heating systems), in combination with a higher frequency of stagnant atmospheric conditions, cause very high PM10 concentrations.
(a) Yearday (“Julian day”) distribution for the domain mean of CG lightning (black line) and for the PM10 concentration averaged on the monitoring stations (gray line). (b) Weekly cycle of the domain-mean value for lightning (black) and PM10 (gray) during April–October 2002–10, after being detrended with respect to the yearday to filter the low-frequency variability. Bars show ±1 standard error. Units for lightning are number of flashes h−1 km−2 × 104. Units for PM10 are milligrams per meter cubed.
Citation: Journal of Applied Meteorology and Climatology 53, 12; 10.1175/JAMC-D-14-0037.1
Since we are interested in studying the relation with thunderstorms, we have analyzed only the “stormy season,” which is the April–October period. The correlation between the mean PM10 and lightning density in the April–October period (yeardays 91–304) is R = −0.64 (p value = <2 × 10−16). Even removing September and October (which have steep increases in PM10) and limiting the period to yeardays 91–250,3 the two variables are anticorrelated, with R = −0.40.
c. CG lightning and PM10 weekly cycle
The following analysis on a weekly time scale is also restricted to just April–October, but in this case the high-frequency part of the signal is retained, because a ±10-day moving average would remove the day-to-day variations. Figure 8b shows the mean CG lightning anomalies (detrended with respect to the mean value of each yearday, black line) and the mean (over 16 stations) PM10 anomalies (gray line) on a weekly time scale, together with their standard errors. Both trends are computed only during the April–October 2002–10 period, which explains why the lightning weekly cycle is slightly different from that shown in Fig. 3c (which covers the 1995–2012 period). As expected, the PM10 distribution shows a clear weekly signal, with lower PM10 concentrations during Sunday, rapidly increasing to the middle of the week (up to Thursday), and starting to decrease from Friday up to Sunday. In general, this can reflect the anthropic activity measured by the air pollution due to the industrial activity and daily commuting, even though Friday is still a working day. Note that the standard errors for PM10 are relatively small, so that the day-to-day variations seem statistically robust.
The CG lightning anomaly distribution seems to reflect a similar trend, except for Wednesday (which shows the lowest lightning activity) and for the shift in the maximum, which occurs on Friday instead of on Thursday. It is noted that a similar local minimum during Wednesday is also shown in Fig. 1 of Laux and Kunstmann (2008) for the temperature anomalies in Germany and in Fig. 1b of Sanchez-Lorenzo et al. (2008) for daily maximum temperature anomalies in Spain. Here, the standard errors are much larger than those for PM10, but the maximum on Friday is clearly outside the standard error bars of all the other days. The shift in the maximum could also be explained if one thinks of the thunderstorms as an efficient way to remove PM10, because of the rain and strong wind produced. Hence, on Friday the PM10 can be lower than on Thursday just because there is a maximum activity of thunderstorm during that day. The correlation between the seven pairs of this weekly cycle is R = 0.48, but it is not very significant because of the low number of points (p value = 0.28).
d. Discussion on the lightning and PM10 correlations
The two previous analyses of CG lightning distribution in comparison with the PM10 concentration distribution during the April–October period seem to give contrasting results: anticorrelation in the yearday analysis and correlation in the weekly analysis. But in the analysis of yeardays 91–304, averaging for each day of the year removes the day-of-the-week effect, since the same yearday has different days of the week in different years. Moreover, using the ±10-day moving average, the high-frequency signal has been removed in the analysis of the 214 yeardays and the day-of-the-week analysis (with periodicity of 7 days) can be considered a high-frequency signal. So, a question arises: why during the April–October period does the low-frequency signal anticorrelate, while the high-frequency signal seems to correlate?
Our hypothesis is that, from a low-frequency point of view, the mean emission of PM10 in the April–October period can be considered to be to a first approximation constant, and hence, when the surface stations measure low concentrations, there are, on average, more PM10 particles at higher elevations. That can lead to two possible scenarios. The first one is the Bell and Rosenfeld hypothesis, that is, that the PM10 transported at high elevations can act as—or be correlated with the presence of—cloud condensation nuclei (CCN, particles typically in the 0.02–0.2-μm range,4 on which water vapor may condense) and a higher concentration of CCN clearly increases the thunderstorm occurrence probability, for microphysical reasons. The second scenario is that the PM10 is transported to higher altitudes during the situations with more convection (more vertical transport in the lower planetary boundary layer, PBL) and hence during the situations that are already more prone to thunderstorm development. In this second case, the anticorrelation between the low-frequency PM10 and lightning would not be a direct relation but would be mediated by the PBL instability. Note that these two explanations are not mutually exclusive.
On the other hand, from a high-frequency point of view, the weekly cycle can track the true day-to-day high-frequency variation in the total PM10 emission: Sunday there is a lower production of anthropogenic PM10 and hence there are fewer CCN available at higher levels and a lower lightning probability, according to the Bell and Rosenfeld hypothesis. Within this context, the minimum probability of lightning on Wednesday shown in Fig. 8b is not well explained by the PM10 weekly cycle, apart from the fact that the standard error bars are very large.
e. Potential instability weekly cycle
To test the role of potential instability in the weekly cycle of lightning, another test has been done. Manzato (2003) defined a 6-h thunderstorm “intensity” (called CALCA6h) for the FVG plain, having an area of only 
Figure 9 shows the probability of having at least one 6-h period with CALCA6h ≥ 0.7 during the different days of the week (black line), together with the anomaly of two instability indices derived by the Udine–Campoformido soundings (WMO code 16044, latitude = 46.03°N, longitude = 13.18°E, altitude = 94 m) at 1200 UTC.5 The dark gray line with the plus signs is the equivalent potential temperature of the most unstable parcel (maximum Θe in the lowest 250 hPa), which is used as the initial parcel for pseudoadiabatic lifting when evaluating the sounding instability. The light gray line with triangles represents the potential instability. In fact, DT500 (Manzato 2003) is similar to the classical lifted index (Galway 1956), but uses as the initial parcel the 30-hPa-thick most unstable parcel, so it corresponds to the most unstable lifted index (MULI). Negative values of DT500 mean potential instability (the initial parcel lifted to 500 hPa is lighter than the environmental air), while positive values signify a potentially stable profile.
The weekly trends of the probability of having significant thunderstorms in the plain of the FVG region (defined as at least one 6-h period with CALCA ≥ 0.7; Manzato 2003) for the period April–October 1995–2012 (black line with circles), maximum Θe in the lowest 250 hPa (medium-gray line with plus signs), and the difference in temperature with the environment (DT500, light-gray line with triangles) for the initial parcel lifted pseudoadiabatically at 500 hPa. The last two indices have been derived from the April–October 1995–2012 Udine–Campoformido radiosoundings made at 1200 UTC and have been detrended to remove the mean value for each yearday before stratifying for the different days of the week. The unit for the Θe and DT500 anomalies is kelvins. On the right side, the scale indicates the probability of having a corresponding day of the week with at least one 6-h period with significant storm in the FVG plain.
Citation: Journal of Applied Meteorology and Climatology 53, 12; 10.1175/JAMC-D-14-0037.1
It is possible to see that during April–October 1995–2012 the maximum probability of having significant storms in the FVG plain is well peaked on Friday, while the minima are on Wednesday and Sunday. This trend is very similar to the April–October 2002–10 weekly lightning cycle seen in Fig. 8b, which is computed on a much larger domain, a period half as long, and also does not consider the observed rain and wind gust variables. Hence, the two similar weekly cycles can be considered to be at least partially independent.
The potential instability measured by DT500 shows a maximum (negative anomaly) on Tuesday and a minimum (positive anomaly) on Sunday, followed by Monday and Wednesday. The standard error bars are such that the maximum on Tuesday and minimum on Sunday do not overlap. The linear correlation with probability of significant thunderstorms is R = −0.58 (p value = 0.17). The weekly cycle of the most unstable parcel Θe shows that the minimum Θe anomaly occurs during Sunday and Monday, while the maximum Θe anomaly occurs on Thursday, followed by Friday. The linear correlation with probability of significant thunderstorms is slightly lower than for the potential instability: R = 0.53 (p value = 0.23).
In this case also, the decrease of Θe on Friday could be seen as a consequence of the maximum thunderstorm activity, if one considers that thunderstorms take their energy from the high-Θe air ingested and produce low-Θe output with their downdrafts.6 More generally, the weekly cycle of the maximum Θe in the lowest 250 hPa is quite similar to the PM10 weekly cycle seen in Fig. 8b (apart from the local minima on Wednesday) and could support the hypothesis that there is more PM10 at high altitudes (where it can act as CCN) during the days with more thermals (convective activity), but it does not explain if PM10 is a cause of thunderstorm or if there is just an indirect correlation with instability and PBL mixing. Nevertheless, we cannot imagine any natural effect able to produce a weekly trend in the maximum Θe in the lowest 250 hPa; therefore, it is possible that this thermal anomaly is also related to human activities and hence to anthropogenic emissions.
The local minimum in thunderstorm activity on Wednesday could be due to the maximum instability anomaly of the day before, because those thunderstorms remove the thermodynamic energy already accumulated. But in that case the standard error bars are quite large, so that is just a speculation.
6. Conclusions
The present study focuses on the climatological characteristics of CG lightning in northeastern Italy and the surrounding regions, observing the different patterns of behavior related to the topography and time scale. It is shown that CG lightning activity varies strongly with the period of the year and the time of the day and that it varies within three different areas having different types of topography. In the areas close to the coast, where the proximity with the sea is a key factor, the CG lightning activity peaks in the second half of the summer (August to mid-September) and much more often during the evening/nighttime. A different trend is observed in the mountain areas, where the CG lightning activity peaks during midsummer (from mid-June to mid-August), especially during the afternoon, when the surface heating produces the maximum potential instability.
The foregoing finding indicates that the main triggering mechanism is quite different in the different areas and periods: while in the mountains the afternoon low-level-based potential instability and the orographic lifting mechanism seem to play a major role, over the sea and coastal area it seems that other triggering mechanisms (possible candidates include convergence lines, low-level moist jets, sea–land boundaries, and bores) that are efficient also during the night and with less low-level potential instability acquire more importance, in particular during the end of the summer, when the sea reaches its maximum sea surface temperature values (Manzato 2007, his Fig. 4b). In the foothills area the CG lightning distribution behaves with a cycle that is about halfway between those of the coastal and the mountain regions, on both annual and daily time scales.
Considering the spatial distribution, the mountainous part of the FVG region (NE Italy) seems to be the area affected by the maximum CG lightning occurrence, reaching values of about 10 flashes km−2 yr−1. This distribution shows good correspondence with the orography and the land–sea surface type and in particular with the presence of steep mountains at only 70–80 km from the Marano–Grado Lagoon (a source of high-Θe air), followed by the Adriatic Sea. The CG lightning density linearly correlates well with the meridional gradient of the orography standard deviation and the distance from the sea used together. This result may be interpreted as a qualitative description of a typical situation favoring thunderstorm occurrence: convection is mainly fed by a strong meridional (south, southwesterly, or southeasterly) flow, advecting moist air from the sea inland, where the Carnic or Julian Prealps act to produce orographic lifting.
Cloud-to-ground lightning exhibits a weekly cycle, with a maximum on Friday and a minimum during Wednesday and Sunday. A similar pattern of behavior is found for different periods and areas and also when considering rain and wind gusts produced by thunderstorms, so it can be regarded as quite a robust signal. Also, the PM10 measured at the surface by 16 stations shows a weekly cycle, with a minimum during Sunday and Monday and a maximum on Thursday, so that there is a 1-day delay between the lightning and the PM10 maxima. A similar trend is found also for the most unstable parcel Θe in the lowest 250 hPa derived from the Udine 1200 UTC soundings, while potential instability (DT500) is highest on Tuesday, followed by Thursday, Friday, and Saturday. During April–October 2002–10 the mean PM10 and lightning weekly cycles in the Po Valley correlate with R = 0.48. During April–October 1995–2012 the probability of having significant storms in the FVG plain and potential instability or maximum Θe have correlations of −0.58 and 0.53, respectively.
Regarding the PM10 effect, no clear conclusions can be derived by the actual results, because of the many feedbacks relating PM10, instability (and PBL thermals activity), and thunderstorms. One possible interpretation of the results is that, on average, there is an anthropic effect that accumulates during the week (starting from Sunday), increasing anomalies of PM10 and of maximum Θe in the lowest 250 hPa until Thursday, while also increasing the probability of triggering thunderstorms on Friday, followed by Saturday. These thunderstorms remove this excess of PM10 (through rainfall and wind) and of Θe (because storms convert the environmental thermal energy into kinetic energy), restoring the initial conditions on Sunday. Of course, if this anthropic effect is real, it is not the primary cause of thunderstorm occurrence in the Po Valley, but rather it modulates the natural variability of these events. For example, the maximum variation in the probability of having a significant storm in the FVG plain during the different days of week is 3%, with respect to a mean probability of 12%.
Acknowledgments
This research was carried out under the support of Project 2CE120P3 INCA–CE, funded by the program ETC Central Europe FESR. We thank Dr. Marina Bernardi and the CESI/SIRF group for providing us with the CG lightning dataset on an extended domain. We thank Dr. Richard Rotunno (NCAR) for his help with a previous version of this manuscript. Finally, we thank the anonymous reviewers for their suggestions, which improved this work.
REFERENCES
Anderson, G., and D. Klugmann, 2014: A European lightning density analysis using 5 years of ATDnet data. Nat. Hazards Earth Syst. Sci., 14, 815–829, doi:10.5194/nhess-14-815-2014.
Antonescu, B., and S. Burcea, 2010: A cloud-to-ground lightning climatology for Romania. Mon. Wea. Rev., 138, 579–591, doi:10.1175/2009MWR2975.1.
Asakawa, A., K. Miyake, S. Yokoyama, T. Shindo, T. Yokota, and T. Sakai, 1997: Two types of lightning discharges to a high stack on the coast of the Sea of Japan in winter. Trans. IEEE Power Delivery,12, 1222–1231, doi:10.1109/61.636953.
Ashley, W. S., and C. W. Gilson, 2009: A reassessment of U.S. lightning mortalities. Bull. Amer. Meteor. Soc., 90, 1501–1518, doi:10.1175/2009BAMS2765.1.
Bell, T. L., D. Rosenfeld, K.-M. Kim, J.-M. Yoo, M.-I. Lee, and M. Hahnenberger, 2008: Midweek increase in us summer rain and storm heights suggests air pollution invigorates rainstorms. J. Geophys. Res.,113, D02209, doi:10.1029/2007JD008623.
Bell, T. L., D. Rosenfeld, and K. Kim, 2009: Weekly cycle of lightning: Evidence of storm invigoration by pollution. Geophys. Res. Lett.,36, L23805, doi:10.1029/2009GL040915.
Bernardi, M., and D. Ferrari, 2004: Evaluation of the LLS efficiency effects on the ground flash density, using the Italian lightning detection system SIRF. J. Electrost., 60, 131–140, doi:10.1016/j.elstat.2004.01.013.
Bernardi, M., and R. Tommasini, 2012: Number of lightning to earth in Italy. Int. Conf. on Lightning Protection, Vienna, Austria, IEEE, 1–5, doi:10.1109/ICLP.2012.6344368.
Boccippio, D. J., K. L. Cummins, H. J. Chiristian, and S. J. Goodman, 2001: Combined satellite and surface based estimation of the intracloud-cloud-to-ground lightning ratio over the continental United States. Mon. Wea. Rev., 129, 108–122, doi:10.1175/1520-0493(2001)129<0108:CSASBE>2.0.CO;2.
Bond, D. W., S. Steiger, R. Zhang, X. Tie, and R. E. Orville, 2002: The importance of NOx production by lightning in the tropics. Atmos. Environ., 36, 1509–1519, doi:10.1016/S1352-2310(01)00553-2.
Borga, M., P. Boscolo, F. Zanon, and M. Sangati, 2007: Hydrometeorological analysis of the 29 August 2003 flash flood in the eastern Italian Alps. J. Hydrometeor., 8, 1049–1067, doi:10.1175/JHM593.1.
Bourscheidt, V., O. Pinto, K. Naccarato, and I. Pinto, 2009: The influence of topography on the cloud-to-ground lightning density in south Brazil. Atmos. Res., 91, 508–513, doi:10.1016/j.atmosres.2008.06.010.
Branick, M. L., and C. A. Doswell III, 1992: An observation of the relationship between supercell structure and lightning ground-strike polarity. Wea. Forecasting, 7, 143–149, doi:10.1175/1520-0434(1992)007<0143:AOOTRB>2.0.CO;2.
Cacciamani, C., F. Battaglia, P. Patruno, L. Pomi, A. Selvini, and S. Tibaldi, 1995: A climatological study of thunderstorm activity in the Po Valley. Theor. Appl. Climatol., 50, 185–203, doi:10.1007/BF00866116.
Carey, L. D., and S. A. Rutledge, 2003: Characteristics of cloud-to-ground lightning in severe and nonsevere storms over the central United States from 1989–1998. J. Geophys. Res., 108, 4483, doi:10.1029/2002JD002951.
Changnon, S. A., 1992: Temporal and spatial relations between hail and lightning. J. Appl. Meteor., 31, 587–604, doi:10.1175/1520-0450(1992)031<0587:TASRBH>2.0.CO;2.
Changnon, S. A., 2001: Damaging thunderstorm activity in the United States. Bull. Amer. Meteor. Soc., 82, 597–608, doi:10.1175/1520-0477(2001)082<0597:DTAITU>2.3.CO;2.
Costa, S., P. Mezzasalma, V. Levizzani, P. P. Alberoni, and S. Nanni, 2001: Deep convection over northern Italy: Synoptic and thermodynamic analysis. Atmos. Res., 56, 73–88, doi:10.1016/S0169-8095(00)00091-0.
Curran, E. B., R. L. Holle, and R. E. Lopez, 2000: Lightning casualties and damages in the United States from 1950 to 1994. J. Climate, 13, 3448–3464, doi:10.1175/1520-0442(2000)013<3448:LCADIT>2.0.CO;2.
Davolio, S., D. Mastrangelo, M. M. Miglietta, O. Drofa, A. Buzzi, and P. Malguzzi, 2009: High resolution simulations of a flash flood near Venice. Nat. Hazards Earth Syst. Sci., 9, 1671–1678, doi:10.5194/nhess-9-1671-2009.
de Souza, P. E., O. Pinto Jr., I. R. C. A. Pinto, N. J. Ferreira, and A. F. dos Santos, 2009: The intracloud/cloud-to-ground lightning ratio in southeastern Brazil. Atmos. Res., 91, 491–499, doi:10.1016/j.atmosres.2008.06.011.
Enno, S., 2011: A climatology of cloud-to-ground lightning over Estonia, 2005–2009. Atmos. Res., 100, 310–317, doi:10.1016/j.atmosres.2010.08.024.
Feudale, L., A. Manzato, and S. Micheletti, 2013: A cloud-to-ground lightning climatology for north eastern Italy. Adv. Sci. Res., 10, 77–84, doi:10.5194/asr-10-77-2013.
Finke, U., and T. Hauf, 1996: The characteristics of lightning occurrence in southern Germany. Bietr. Phys. Atmos.,69, 361–374.
Fisher, R., G. Schnetzer, R. Thottappillil, V. Rakov, M. Uman, and J. Goldberg, 1993: Parameters of triggered-lightning flashes in Florida and Alabama. J. Geophys. Res.,98, 22 887–22 902, doi:10.1029/93JD02293.
French, A. J., and M. D. Parker, 2010: The response of simulated nocturnal convective systems to a developing low-level jet. J. Atmos. Sci., 67, 3384–3408, doi:10.1175/2010JAS3329.1.
Gaffard, C., J. Nash, N. Atkinson, A. Bennett, G. Callaghan, E. Hibbett, P. Taylor, M. Turp, and W. Schulz, 2008: Observing lightning from the ground over large areas of the globe. 20th Int. Lightning Detection Conf., Tuscon, AZ, WMO. [Available online https://www.wmo.int/pages/prog/www/IMOP/publications/IOM-96_TECO-2008/P1(16)_Gaffard_United_Kingdom.doc.]
Galway, J. G., 1956: The lifted index as a predictor of latent instability. Bull. Amer. Meteor. Soc., 37, 528–529.
Gladich, I., I. Gallai, D. Giaiotti, and F. Stel, 2011: On the diurnal cycle of deep moist convection in the southern side of the Alps analysed through cloud-to-ground lightning activity. Atmos. Res., 100, 371–376, doi:10.1016/j.atmosres.2010.08.026.
Glickman, T., Ed., 2000: Glossary of Meteorology. 2nd ed. Amer. Meteor. Soc., 855 pp. [Available online at http://glossary.ametsoc.org/.]
Gong, D.-Y., C.-H. Ho, D. Chen, Y. Qian, Y.-S. Choi, and J. Kim, 2007: Weekly cycle of aerosol-meteorology interaction over China. J. Geophys. Res.,112, D22202, doi:10.1029/2007JD008888.
Goodman, S. J., and D. R. MacGorman, 1986: Cloud-to-ground lightning activity in mesoscale convective complexes. Mon. Wea. Rev., 114, 2320–2328, doi:10.1175/1520-0493(1986)114<2320:CTGLAI>2.0.CO;2.
Herodotou, N., W. Chisholm, and W. Janischewskyj, 1993: Distribution of lightning peak stroke currents in Ontario using an LLP system. IEEE Trans. Power Delivery,8, 1331–1339, doi:10.1109/61.252659.
Holle, R. L., R. E. López, and B. C. Navarro, 2005: Deaths, injuries, and damages from lightning in the United States in the 1890s in comparison with the 1990s. J. Appl. Meteor., 44, 1563–1573, doi:10.1175/JAM2287.1.
Holt, M., P. Hardaker, and G. McLelland, 2001: A lightning climatology for Europe and the UK, 1990–99. Weather, 56, 290–296, doi:10.1002/j.1477-8696.2001.tb06598.x.
Isotta, F. A., and Coauthors, 2014: The climate of daily precipitation in the Alps: Development and analysis of a high-resolution grid dataset from pan-Alpine rain-gauge data. Int. J. Climatol., 34, 1657–1675, doi:10.1002/joc.3794.
Kilinc, M., and J. Beringer, 2007: The spatial and temporal distribution of lightning strikes and their relationship with vegetation type, elevation, and fire scars in the Northern Territory. J. Climate, 20, 1161–1173, doi:10.1175/JCLI4039.1.
Krider, E. P., R. C. Noggle, and M. A. Uman, 1981: Lightning detection system utilizing triangulation and field amplitude comparison techniques. U.S. Patent 4 245 190, filed 12 December 1978, issued 13 January 1981.
Krider, E. P., A. E. Pifer, and L. G. Byerley III, 1989: Thunderstorm sensor and method of identifying and locating thunderstorms. U.S. Patent 4 806 851, filed 3 September 1987, issued 21 February 1989.
Laux, P., and H. Kunstmann, 2008: Detection of regional weekly weather cycles across Europe. Environ. Res. Lett.,3, 044005, doi:10.1088/1748-9326/3/4/044005.
MacGorman, D. R., and D. W. Burgess, 1994: Positive cloud-to-ground lightning in tornadic storms and hailstorms. Mon. Wea. Rev., 122, 1671–1697, doi:10.1175/1520-0493(1994)122<1671:PCTGLI>2.0.CO;2.
Mäkelä, A., P. Rossi, and D. M. Schultz, 2011: The daily cloud-to-ground lightning flash density in the contiguous United States and Finland. Mon. Wea. Rev., 139, 1323–1337, doi:10.1175/2010MWR3517.1.
Manzato, A., 2003: A climatology of instability indices derived from Friuli Venezia Giulia soundings, using three different methods. Atmos. Res., 67–68, 417–454, doi:10.1016/S0169-8095(03)00058-9.
Manzato, A., 2007: The 6 h climatology of thunderstorms and rainfalls in the Friuli Venezia Giulia plain. Atmos. Res., 83, 336–348, doi:10.1016/j.atmosres.2005.08.013.
Manzato, A., 2012: Hail in northeast Italy: Climatology and bivariate analysis with the sounding-derived indices. J. Appl. Meteor. Climatol.,51, 449–467, doi:10.1175/JAMC-D-10-05012.1.
Manzato, A., S. Davolio, M. M. Miglietta, A. Pucillo, and M. Setvák 2014: 12 September 2012: A supercell outbreak in NE Italy? Atmos. Res.,153, 98–108, doi:10.1016/j.atmosres.2014.07.019.
Nash, C., and E. Johnson, 1996: Synoptic climatology of lightning-caused forest fires in subalpine and boreal forests. Can. J. For. Res., 26, 1859–1874, doi:10.1139/x26-211.
Nisi, L., P. Ambrosetti, and L. Clementi, 2014: Nowcasting severe convection in the Alpine region: The COALITION approach. Quart. J. Roy. Meteor. Soc., 140, 1684–1699, doi:10.1002/qj.2249.
Novák, P., and H. Kyznarová, 2011: Climatology of lightning in the Czech Republic. Atmos. Res., 100, 318–333, doi:10.1016/j.atmosres.2010.08.022.
Orville, R. E., and G. R. Huffines, 1999: Lightning ground flash measurements over the contiguous United States: 1995–97. Mon. Wea. Rev., 127, 2693–2703, doi:10.1175/1520-0493(1999)127<2693:LGFMOT>2.0.CO;2.
Orville, R. E., and G. R. Huffines, 2001: Cloud-to-ground lightning in the United States: NLDN results in the first decade, 1989–98. Mon. Wea. Rev., 129, 1179–1193, doi:10.1175/1520-0493(2001)129<1179:CTGLIT>2.0.CO;2.
Parker, M. D., 2008: Response of simulated squall lines to low-level cooling. J. Atmos. Sci., 65, 1323–1341, doi:10.1175/2007JAS2507.1.
Pernigotti, D., E. Georgieva, P. Thunis, C. Cuvelier, and A. Meij, 2012: The impact of meteorology on air quality simulations over the Po Valley in northern Italy. Air Pollution Modeling and its Application XXI, NATO Science for Peace and Security Series C: Environmental Security, D. G. Steyn and S. T. Castelli, Eds., Springer, 485–490.
Petersen, W. A., and S. A. Rutledge, 1998: On the relationship between cloud-to-ground lightning and convective rainfall. J. Geophys. Res.,103, 14 025–14 040, doi:10.1029/97JD02064.
Petersen, W. A., and Coauthors, 1999: Mesoscale and radar observations of the Fort Collins flash flood of 28 July 1997. Bull. Amer. Meteor. Soc., 80, 191–216, doi:10.1175/1520-0477(1999)080<0191:MAROOT>2.0.CO;2.
Pinto, O., Jr., I. Pinto, M. Gomes, I. Vitorello, A. Padilha, J. Dmiz, A. Carvalho, and A. Cazetta Filho, 1999: Cloud-to-ground lightning in southeastern Brazil in 1993. J. Geophys. Res., 104, 31 369–31 379, doi:10.1029/1999JD900800.
Podur, J., D. L. Martell, and F. Csillag, 2003: Spatial patterns of lightning-caused forest fires in Ontario, 1976–1998. Ecol. Modell., 164, 1–20, doi:10.1016/S0304-3800(02)00386-1.
Price, C., and D. Rind, 1994: The impact of a 2 × CO2 climate on lightning-caused fires. J. Climate, 7, 1484–1494, doi:10.1175/1520-0442(1994)007<1484:TIOACC>2.0.CO;2.
Rakov, V. A., and M. A. Uman, 2007: Lightning: Physics and Effects. Cambridge University Press, 700 pp.
Ramos, A., R. Ramos, P. Sousa, R. M. Trigo, M. Janeira, and V. Prior, 2011: Cloud to ground lightning activity over Portugal and its association with circulation weather types. Atmos. Res., 101, 84–101, doi:10.1016/j.atmosres.2011.01.014.
Rivas Soriano, L., F. de Pablo, and E. G. Díez, 2001: Relationship between convective precipitation and cloud-to-ground lightning in the Iberian Peninsula. Mon. Wea. Rev., 129, 2998–3003, doi:10.1175/1520-0493(2001)129<2998:RBCPAC>2.0.CO;2.
Rivas Soriano, L., F. De Pablo, and C. Tomas, 2005: Ten-year study of cloud-to-ground lightning activity in the Iberian Peninsula. J. Atmos. Sol. Terr. Phys., 67, 1632–1639, doi:10.1016/j.jastp.2005.08.019.
Rosenfeld, D., and T. L. Bell, 2011: Why do tornados and hailstorms rest on weekends? J. Geophys. Res.,116, D20211, doi:10.1029/2011JD016214.
Rotunno, R., and R. Ferretti, 2003: Orographic effects on rainfall in MAP cases IOP2b and IOP8. Quart. J. Roy. Meteor. Soc., 129, 373–390, doi:10.1256/qj.02.20.
Rutledge, S. A., E. R. Williams, and W. A. Petersen, 1993: Lightning and electrical structure of mesoscale convective systems. Atmos. Res., 29, 27–53, doi:10.1016/0169-8095(93)90036-N.
Sanchez-Lorenzo, A., J. Calbó, J. Martin-Vide, A. Garcia-Manuel, G. García-Soriano, and C. Beck, 2008: Winter “weekend effect” in southern Europe and its connections with periodicities in atmospheric dynamics. Geophys. Res. Lett., 35, L15711, doi:10.1029/2008gl034160.
Santos, J. A., M. A. Reis, J. Sousa, S. M. Leite, S. Correia, M. Janeira, and M. Fragoso, 2012: Cloud-to-ground lightning in Portugal: Patterns and dynamical forcing. Nat. Hazards Earth Syst. Sci., 12, 639–649, doi:10.5194/nhess-12-639-2012.
Schulz, W., K. Cummins, G. Diendorfer, and M. Dorninger, 2005: Cloud-to-ground lightning in Austria: A 10-year study using data from a lightning location system. J. Geophys. Res.,110, D09101, doi:10.1029/2004JD005332.
Shindo, T., and S. Yokoyama, 1998: Lightning occurrence data observed with lightning location systems in Japan: 1992-1995. IEEE Trans. Power Delivery,13, 1468–1474, doi:10.1109/61.714797.
Solomon, R., C. Medaglia, C. Adamo, S. Dietrich, and A. Mugnai, 2004: Relating cloud microphysical properties and lightning: Model analysis of an extreme hailstorm. Proc. Fifth Plinius Conf. on Mediterranean Storms, Ajaccio, Corsica, France, European Geosciences Union, 499–505.
Soula, S., Y. Seity, L. Feral, and H. Sauvageot, 2004: Cloud-to-ground lightning activity in hail-bearing storms. J. Geophys. Res., 109, D02101, doi:10.1029/2003JD003669.
Tapia, A., J. A. Smith, and M. Dixon, 1998: Estimation of convective rainfall from lightning observations. J. Appl. Meteor., 37, 1497–1509, doi:10.1175/1520-0450(1998)037<1497:EOCRFL>2.0.CO;2.
Tuomi, T., and A. Mäkelä, 2008: Thunderstorm climate of Finland 1998–2007. Geophysica, 44, 67–80.
van den Broeke, M. S., D. M. Schultz, R. H. Johns, J. S. Evans, and J. E. Hales, 2005: Cloud-to-ground lightning production in strongly forced, low-instability convective lines associated with damaging wind. Wea. Forecasting, 20, 517–530, doi:10.1175/WAF876.1.
Villarini, G., and J. A. Smith, 2013: Spatial and temporal variability of cloud-to-ground lightning over the continental U.S. during the period 1995-2010. Atmos. Res., 124, 137–148, doi:10.1016/j.atmosres.2012.12.017.
Williams, E., V. Mushtak, D. Rosenfeld, S. Goodman, and D. Boccippio, 2005: Thermodynamic conditions favorable to superlative thunderstorm updraft, mixed phase microphysics and lightning flash rate. Atmos. Res., 76, 288–306, doi:10.1016/j.atmosres.2004.11.009.
Zipser, E. J., 1994: Deep cumulonimbus cloud systems in the tropics with and without lightning. Mon. Wea. Rev., 122, 1837–1851, doi:10.1175/1520-0493(1994)122<1837:DCCSIT>2.0.CO;2.
The Julian Prealps are located along the border between Italy and Slovenia (north of the UD and GO labels in Fig. 1) and are WNW toward ESE oriented, while the Carnic Prealps are on their western side (north of the PN label in Fig. 1) and are SW–NE oriented, so that the two form a bow-shaped barrier.
Anderson and Klugmann (2014) report a location accuracy of better than 5 km, an aggregation of multiple lightning strikes within 20 km, and a regridding onto a 0.20° × 0.20° grid (about 22 km × 14 km), which means a much coarser resolution than that used in the present work.
Note that in Italy there is a law forbidding the use of building heating systems in the northeast region of the country from 15 April to 15 October.
Note that PM10 stations measure all particles with diameters ≤10 μm; therefore, particles that are much smaller than 10 μm are accumulated.
Since the PM10 concentration and lightning density describe daily mean values (0000–0000), only the midday sounding was selected.
Basically, a thunderstorm is an engine that converts thermal energy into kinetic energy and water vapor into condensed water.