1. Introduction
A cloud-to-ground (CG) lightning flash is on average registered on 150–160 days per year in Poland (Taszarek et al. 2015). Thunderstorms are most frequent during the summer (i.e., June–August) and over the southeastern part of the country (Bielec-Bąkowska 2003). The same period is also linked to the peak occurrence of severe weather outbreaks such as derechos and tornadic supercells that are responsible for considerable material losses in Poland (Celiński-Mysław and Matuszko 2014; Pilorz 2015; Widawski and Pilorz 2018; Pilguj et al. 2019; Poręba and Ustrnul 2020; Surowiecki and Taszarek 2020). A similar distribution of severe weather events can be observed based on the records from the European Severe Weather Database (ESWD; Dotzek et al. 2009), which also indicate that the most common convective threat across central Europe is severe wind, followed by excessive rainfall, large hail, damaging lightning, and tornadoes (Groenemeijer et al. 2017; Taszarek et al. 2019).
Although thunderstorm intensity can be measured in various ways, it can be generally assumed that the presence of severe weather reports, high CG lightning flash rates, and/or high radar reflectivity depends largely on the vertical velocity and size of the convective updraft (Apke et al. 2018). Regional research studies that focus on evaluating thunderstorm intensity and its relation to atmospheric environments are highly important to operational forecasters, as these studies provide forecasters with guidance on the metrics and corresponding values that are useful in predicting severe convective storms over specific areas, times of year, and times of day. For example, an increase in the availability of low-level moisture, midtropospheric lapse rates, and the degree of convective organization (e.g., supercells, squall lines) governed by sufficient vertical wind shear leads to stronger updrafts (Doswell 2001; Markowski and Richardson 2010; Coffer and Parker 2015; Dennis and Kumjian 2017; Lin and Kumjian 2021). Despite growing numbers of studies on thunderstorm environments, there are still some limitations associated with challenging identification and recognition of thunderstorm intensity (further described in section 2b).
Thus, in this study we address this issue by using the peak 1-min CG lightning flash rate as a proxy to categorize thunderstorms according to their lightning intensity and then intercompare them with collocated atmospheric environments. Although studies on thunderstorm environments in central Europe were conducted in the past (Púčik et al. 2015; Westermayer et al. 2017; Taszarek et al. 2020), none of them evaluated specific CG lightning flash rates, which in our approach constitutes a new, not previously applied in Europe, thunderstorm intensity proxy. Similar approaches have been tested for the tropics and subtropics (Liu et al. 2020). A preliminary comparison of CG lightning flash rate intensity categories defined in this study with 14 427 ESWD large hail, severe wind, heavy rain, and tornado reports from the territory of Poland indicates an increasing relative number of reports along with increasing CG lightning flash rates (Fig. 1).
Mean number of ESWD severe weather reports per grid with specific thunderstorm intensity category. PERUN CG lightning data and ESWD reports were collocated on 0.25° grids within 1-h intervals. For each thunderstorm intensity category grids (defined based on a peak 1-min CG lightning flash rate), proximal ESWD reports were counted and then divided by the total number of grids representing each category.
Citation: Weather and Forecasting 37, 8; 10.1175/WAF-D-21-0099.1
By evaluating 18 years (2002–19) of CG lightning flash rate data from the PERUN lightning detection network, we aim to better understand the climatological aspects of severe thunderstorms and their accompanying environments across Poland. An improved resolution of the fifth generation of ECMWF reanalysis (ERA5) allows us to study lightning environments in the way that was not possible with prior generations of reanalyses. We use hourly intervals (contrary to 3–6-h steps in previous studies) to evaluate the diurnal, annual, and spatial aspects of convective environments collocated with specific CG lightning flash rates—an aspect rarely explored before. An auxiliary goal constituting the background of the research is also to present an updated 18-yr climatology of CG lightning flashes for Poland.
2. Prior research
a. Lightning climatologies
Lightning climatologies are based on data from lightning detectors, which can be ground-based and satellite-based, global or regional. Measurements of global lightning activity are applied with the ground-based lightning sensors systems, such as World Wide Lightning Location Network (WWLLN; http://wwlln.net) and Vaisala GLD360 (Said et al. 2011) or satellite-based instruments like Lightning Imaging Sensor (LIS; Christian et al. 1999). Both sources confirm that the highest frequency of lightning occurs over continents in tropical areas (Virts et al. 2013; DiGangi et al. 2021). Global ground-based systems make it possible to monitor lightning over a broader area including higher latitudes, but their detection efficiency is lower than regional or satellite-based systems (Bürgesser 2017; DiGangi et al. 2021).
Regionally, a large number of lightning climatologies have been performed for the United States with the National Lightning Detection Network (NLDN; Orville 1991; Orville and Huffines 2001; Orville et al. 2002; Zajac and Rutledge 2001; Cummins and Murphy 2009; Holle 2014; Holle et al. 2016; Kingfield et al. 2017; Koehler 2020). These studies show that the highest annual mean CG lightning flash density occurs over Florida, reaching 16 CG lightning flashes km−2 yr−1 and over the coast of Gulf of Mexico with 8–12 CG lightning flashes km−2 yr−1. In the same region the number of thunderstorm days is the highest with more than 110 thunderstorm days per year (Koehler 2020).
In Europe, research on lighting climatologies has been conducted using the Advanced Technology Demonstration Network (ATDnet), European Cooperation for Lightning Detection (EUCLID), and ZEUS networks (Anderson and Klugmann 2014; Poelman et al. 2016; Taszarek et al. 2019, 2020; Enno et al. 2020). These studies showed that peak lightning flash rate in Europe is in the afternoon (1400–1600 UTC) during summer. The highest annual lightning flash rates are in the Alps and over northeastern Italy reaching 7–8 km−2 yr−1 (Anderson and Klugmann 2014; Enno et al. 2020). The highest annual mean number of thunderstorm days is associated with mountainous regions and Mediterranean boundary layer moisture, especially along the Alps, Dinaric Alps, and Apennines with values more than 60 (Taszarek et al. 2019). In regional studies for selected countries, the results are consistent with the previously mentioned distributions with mean annual peak CG lightning flash rates reaching 15 flashes km−2 yr−1 over the border between Italy and Slovenia (Schulz et al. 2005), 9 flashes km−2 yr−1 in Italy (Feudale et al. 2013), 7.43 flashes km−2 yr−1 in the Czech Republic (Novák and Kyznarová 2011), 3.06 flashes km−2 yr−1 over the Carpathians in Romania (Antonescu and Burcea 2010), 1.01 flashes km−2 yr−1 in Estonia (Enno 2011), and around 0.9 flashes km−2 yr−1 in Finland (Mäkelä et al. 2011). However, these numbers should be interpreted with caution when making comparisons between regions as each lightning detection network features different detection efficiency and spatial inhomogeneities (Price 2008).
In Poland, research on the temporal and spatial variability of thunderstorms was limited for many years to observations from manned surface synoptic stations (Stopa 1962; Grabowska 2001; Bielec-Bąkowska 2002, 2003, 2013; Kolendowicz 2006; Ustrnul and Czekierda 2009). These studies showed that thunderstorms in Poland occur most often over southeastern Poland (30–35 thunderstorm days per year), and least frequently along the coast of the Baltic Sea (10–15 days). However, such data have limitations (associated with human perception) in thunderstorm identification (Czernecki et al. 2016; Koehler 2020; DiGangi et al. 2021), spatial biases, and no method to determine storm intensity. The development of the PERUN lightning detection system in the early twenty-first century (Łoboda et al. 2009) has allowed for the monitoring of thunderstorms with a time lag of less than 1 min and enhanced understanding of the spatial and temporal distribution of convective storms in Poland. An 11-yr period (2002–13) of CG lightning flashes data from the PERUN was analyzed by Taszarek et al. (2015), who confirmed that the largest number of thunderstorm days is over southeastern Poland (in accordance with estimates from manned surface observations), but indicated a peak lightning flash rate over eastern part reaching 3.1 flashes km−2 yr−1. Taszarek et al. (2015) also concluded that the highest fraction of nocturnal lightning is over western Poland.
b. Thunderstorm intensity
The determination of thunderstorm intensity, which can be based on the evaluation of lightning frequency, radar reflectivity, or the occurrence of severe weather, is a considerable methodological challenge. One of the methods uses severe weather reports collected from weather observers and media sources (Dotzek et al. 2009; Edwards et al. 2013; Elmore et al. 2014; Seimon et al. 2016; Krennert et al. 2018). However, the determination of thunderstorm intensity based on these reports is quite arbitrary (Doswell 1985), and depends, inter alia, on the population density and local reporting efficiency inducing spatial and temporal inhomogeneities (Doswell 1985; Verbout et al. 2006; Allen and Tippett 2015; Blair et al. 2017; Groenemeijer et al. 2017; Edwards et al. 2018; Taszarek et al. 2019). The ESWD database for the area of Poland consists of more than 22 000 large hail, tornado, and severe wind reports for the period 2008–20.
Teledetection data such as radar, lightning detectors, or satellites provide more homogenous information and offer an attractive alternative for the determination of thunderstorm intensity (Punge et al. 2017; Enno et al. 2020; Fluck et al. 2021). Well-organized thunderstorms with broad and strong updrafts have larger CG lightning flash rates compared to ordinary cells with weaker updrafts, and large flash rates are typically associated with the occurrence of tornadoes and large hail (Goodman et al. 1988; Macgorman et al. 1989; Williams et al. 1989, 1999; Wiens et al. 2005; Steiger et al. 2007; Deierling and Petersen 2008; Gatlin and Goodman 2010; Schultz et al. 2016). Moreover, severe thunderstorms are often associated with lightning jumps, which are good predecessors of severe weather (Perez et al. 1997; Schultz et al. 2009, 2011, 2017; Farnell et al. 2017, 2018; Farnell and Rigo 2020). A prominent example might be strong, long-lived supercell thunderstorms that feature powerful updrafts speeds and are well known for producing severe weather (Smith et al. 2012). These thunderstorms may have lightning flash rates exceeding 200 per minute (Lang et al. 2004; Markowski and Richardson 2010). On the other hand, there are reports of tornadoes and severe convective wind gusts associated with weaker and shallower convection, which rarely produce any lightning (Pacey et al. 2021). In central Europe, such events typically occur in cold seasons and narrow cold-frontal rainbands (NCFR; Gatzen 2011; Surowiecki and Taszarek 2020). For these events, the estimation of thunderstorm intensity based on flash rates may be not possible and represents a limitation of using lightning data for identification of severe convective storms.
c. Convective parameters
Severe convective storms can be predicted and studied with convective parameters that, when applied in the NWP models, may serve as a proxy of expected thunderstorm intensity. Studies confirmed that thunderstorm severity is driven mainly by increasing convective instability and vertical wind shear, and that these metrics can be used to discriminate between severe and nonsevere convection (Brooks et al. 2003; Thompson et al. 2012; Allen et al. 2011; Púčik et al. 2015; Taszarek et al. 2020). However, despite the growing usage of convective indices in operational forecasting, they feature certain limitations. Some indices, especially composite, merge meteorological elements such as convective instability and shear without physical basis (Doswell and Schultz 2006). Calculations and especially the scaling of parameters are in many ways arbitrary and vary depending on the geographical location. Additionally, convective parameters demonstrate only possible atmospheric environments, but do not predict if convection will be initiated. Due to these factors, one should interpret the values of convective indices with caution (Doswell and Schultz 2006). Moreover, most convective indices were developed in the United States, where convective environments feature higher moisture and convective instability, while European environments are much drier with steeper lapse rates in the lower troposphere (Brooks et al. 2007; Riemann-Campe et al. 2009; Taszarek et al. 2020). Additionally, the obtained results will also depend on the choice of the parcel used for CAPE calculations and application of the virtual temperature correction (Doswell and Rasmussen 1994). A better agreement between continents is observed for wind shear. Low-level lapse rates, wind shear, and synoptic-scale lift [which are included in parameters such as SHERB (severe hazards in environments with reduced buoyancy) or MOSH (modified SHERB)] have overall better skill in discriminating between severe and nonsevere convection in low CAPE and high shear environments that are common in Europe (Hanstrum et al. 2002; Sherburn and Parker 2014; Sherburn et al. 2016; Celiński-Mysław et al. 2020; Rodríguez and Bech 2021; Pacey et al. 2021).
Many factors influencing thunderstorm development such as vertical wind profile, moisture availability, convective mode, and the regional climatological aspects of atmospheric environments require extensive research aimed at assessing specific values of convective parameters (e.g., Gensini and Ashley 2011; Allen and Karoly 2014; Anderson-Frey et al. 2016; Chernokulsky et al. 2019; Ingrosso et al. 2020). Such studies demonstrate the pronounced utility of convective indices in forecasting different types of thunderstorms. Thunderstorms in Europe occur most often in CAPE below 500 J kg−1, and higher convective instability is typically associated with severe weather such as large hail and heavy precipitation. Given an unstable environment, wind shear is a good discriminator between nonsevere and extremely severe storms (∼20 m s−1 can be used as a proxy for very large hail and significant tornadoes; Púčik et al. 2015; Taszarek et al. 2020). However, one should be aware that it is difficult to compare results between studies using different reanalysis datasets, as each of them represents a different underlying climatology of convective environments, and the values of certain parameters may change along with changing resolution.
Although many of the previously mentioned studies focused on comparing severe wind, large hail, heavy rainfall, and tornadoes with accompanying ambient environments, only few evaluated different classes of CG lightning flash rates, which is the main focus of this work. Although CG lightning flash rates are not typically predicted by operational forecasters, a correlation between specific CG lightning flash rates and the relative number of ESWD reports can be observed (Fig. 1).
3. Data and methods
a. Lightning data
In this study we used all the available CG lightning flashes detected by the PERUN system on the territory of Poland between 2002 and 2019 (18 years). PERUN is a Polish lightning detection network working operationally since 2002. Its name corresponds to the Slavic god of lightning and thunder (Gieysztor 2006). At the beginning, the system consisted of 9 sensors, which after 2014 were expanded to 12 (Fig. 2). At the same time, the sensors began upgrading from SAFIR3000 (Surveillance et Alerte Foudre par Interférometrie Radioélectrique) to TLS200 (Total Lightning Sensor). As of 2021, the PERUN system consists of 8 TLS200 and 4 SAFIR3000 sensors and is supported by the data from sensors in the neighboring countries. CG lightning flash detection efficiency estimated in 2006 reached 95% within the area of Poland (Bodzak 2006; Bodzak et al. 2006). Since the sensors are not distributed evenly throughout the country, only 38.3% of the country’s coverage has lightning location accuracy finer than 1 km. Accuracy finer than 2 km is found for 76.6% of the coverage, with the lowest detection accuracy over northwestern Poland (Taszarek et al. 2015). As new sensors have been added to the network, the detection accuracy has increased since 2014, especially in the southwestern part of the country. Each CG lightning flash detection in the database included the time, location, and a peak current of the discharge. To provide better homogeneity of the data we took into account only situations when a flash had a peak current of at least 15 kA, as lower values are typically linked to intracloud detections (Wacker and Orville 1999; Cummins and Murphy 2009; Koehler 2020). After this filter, a total of 8 251 273 CG lightning flashes (limited to the area of Poland) were used in this study.
Hypsometric map of Poland based on the Digital Terrain Model (USGS/SRTM; Farr et al. 2007). Points indicate the location of the PERUN lightning detection sensors (red points indicate sensors added to the PERUN network after 2014).
Citation: Weather and Forecasting 37, 8; 10.1175/WAF-D-21-0099.1
The total number of CG lightning flashes within a specific thunderstorm depends not only on the evolution of updraft intensity, but also on the duration and spatial coverage of the entire convective system. Thus, the total number of CG lightning flashes over a broader area (e.g., associated with mesoscale convective systems; Houze 2018) may not represent thunderstorm intensity well in comparison to local severe convective cells with a strong updraft (e.g., supercells), which can produce a comparable number of CG lightning flashes but over a much smaller area (Wiens et al. 2005; Steiger et al. 2007; Calhoun et al. 2013). Thus, to avoid biases associated with increasing the thunderstorm area but not intensity, we decided to use a peak of 1-min flash rate per 100 km2.
To obtain information about the CG lightning flash rates, all detections were gridded to 10 km × 10 km boxes—a metric commonly used in prior work (e.g., Biron 2009; Enno 2011; Sulik 2021). Then, within these boxes, we investigated the peak 1-min flash count. In this way, a rate of CG lightning flashes per 100 km2 per minute was obtained, and then subsequently used to determine thunderstorm lightning intensity (Table 1). Intensity classes of weak, moderate, intense, and extreme thunderstorms were defined based on the 50th, 75th, 90th, and 95th percentile thresholds of CG lightning flash rate distribution. A limitation of this approach is that due to the gridding process the thunderstorm intensity data refer to the peak flash rates, which is not always associated with one thunderstorm (several thunderstorms could have occurred in the temporal and spatial scope of the reanalysis). Nocturnal thunderstorms were classified based on the criterion of the solar angle below −12° (computed for the location and time of CG lightning flash occurrence). This metric is known as the nautical dawn (NOAA/NWS Glossary 2021).
Thunderstorm intensity classes determined on the basis of 1-min peak CG lightning flashes per 10 km × 10 km grid (considered in hour steps) with corresponding percentile values, number of collocated ERA5 grids and a mean annual number of days given a specific category.
b. ERA5 reanalysis
To investigate thunderstorm environments, we used the fifth generation of ECMWF atmospheric reanalysis (ERA5; Hersbach et al. 2020). ERA5 has a horizontal grid spacing of 0.25°, a 1-h temporal step, and 137 terrain-following hybrid-sigma model levels (Table 2), which made it possible to explore convective environments and their corresponding climatologies in a way that was not possible with prior reanalyses, especially considering diurnal cycles. The 1-h time step allowed us to assign lighting flash events to specific convective environments with higher precision compared to previous reanalyses. In this case, the maximum time difference between the CG lightning flash and the corresponding ERA5 grid was less than 30 min instead of 3 h for reanalyses consisting of 6-h steps (e.g., ERA-Interim).
Characteristics of ERA5 reanalysis domain used in this study.
For each grid from the previously described CG lightning flash rate database, we assigned a proximal grid from ERA5. For each ERA5 grid, we assigned a thunderstorm category based on the highest peak 1-min CG lightning flash rate within a specific hour (to match the ERA5 temporal step). That procedure allowed us to merge two datasets with different resolutions. Weak thunderstorms contributed to the largest number of ERA5 grids, while those of extreme intensity were the least frequent, representing an annual mean of 161 and 55 days, respectively (Table 1). With the increase in the thunderstorm intensity category, the number of grids decreases, especially in winter—in the case of extreme thunderstorms in January, only three grids were used (Table 3).
Number of unique ERA5 grids in a given thunderstorm CG lightning flash intensity category over each month.
For each reanalysis profile, temperature, humidity, pressure, geopotential, U and V were interpolated vertically. A mixed-layer (ML) parcel was defined by mixing a layer 0–500 m above ground level (AGL), while a most unstable (MU) parcel was based on the highest equivalent potential temperature (Θe) in the 0–3 km AGL; both versions used the virtual temperature correction (Doswell and Rasmussen 1994). For the computations of 0–3 km AGL storm-relative helicity (SRH03), the internal dynamics method was applied for right-moving supercells (Bunkers et al. 2000). Vertical wind shear was computed as a magnitude of vector difference between the surface and a specific height. The indices used in the study are associated with the concept of the ingredient-based forecasting (Johns and Doswell 1992; Doswell et al. 1996) focusing on the assessment of humidity characteristics, convective instability, and a vertical profile of the wind, which governs the convective mode and storm organization (Weisman and Klemp 1982; Thompson et al. 2012). In addition, we also used composite indices, which turned out to be useful in assessing thunderstorm intensity across central Europe: WMAXSHEAR (a square root of 2 times CAPE multiplied by 0–6-km wind shear; Taszarek et al. 2020) dedicated to forecasting severe storms, SCP (supercell composite parameter; Thompson et al. 2003) with an updated formula from Gropp and Davenport (2018), and HSI (hail size index; Czernecki et al. 2019) dedicated for forecasting large hail. As supercells and large hail are typically associated with strong updrafts that also favor intense lightning, we believe these metrics are worth comparing with other convective parameters and among CG lightning flash rate categories. A complete list of variables used in this study is presented in Table 4 with additional metrics provided in the appendix.
List of convective parameters used in this study (see the appendix for additional metrics).
c. ESWD reports
A total of 14 427 ESWD reports from the area of Poland were used for comparison with lightning intensity categories (Fig. 1). Only ESWD reports with Q1 and Q2 quality level classes (meaning reports from reliable sources or confirmed by scientific study; Dotzek et al. 2009), time lag of less than 1 h, and spatial proximity of 0.25° (from the detected specific CG lightning density) were considered. For the higher lightning intensity categories, the number of available cases decreases. We calculate a mean number of ESWD reports that are within the aforementioned proximity of the grid with specific CG lightning flash rate. In the end, we divide the number of ESWD reports assigned to a specific lightning intensity category by the number of grids for this specific category.
4. Results
a. Climatological aspects of CG lightning flashes in Poland
Over the period 2002–19, we identified a total number of 2926 days with at least two CG lightning flashes, which represented a mean of 161 thunderstorm days per year. The highest annual mean number of hours with CG lightning flashes occurred over southeastern and central Poland, reaching around 80 h (Fig. 3a), consistent with results from Earth Networks Global Lightning Detection Network (ENGLN; DiGangi et al. 2021). The regions with the most frequent CG lightning flashes included mainly higher elevation areas such as the Carpathian Mountains, the Lublin Upland, and the Kraków–Częstochowa Upland, as well as the lowland areas across central Poland (geographical regions of Poland are provided in Fig. 2). Conversely, the lowest frequency of thunderstorm hours occurred across northwestern Poland, with an annual mean of less than 20 h. The highest mean annual CG lightning flash rate reaching 3.0 CG km−2 yr−1 was recorded over the central part of the country and in the region of the Kraków–Częstochowa Upland (Fig. 3b).
The annual cycle of CG lightning flashes in Poland is consistent with typical climatological distributions of temperature and precipitation in this part of Europe (Lorenc 2005). Thunderstorms are the most frequent in mid-July, but the timeframe with an enhanced mean number of CG lightning flashes starts in mid-April with more than 7000, and ends in late September with 14 700 (Figs. 4a and 5a). During that period, the annual mean number of CG lightning flashes is greatest, especially in July, reaching 155 474 (Fig. 4a). June has a slightly higher number compared to August with around 15 000 more flashes. Thunderstorms are clearly less frequent during the autumn, winter, and early spring. From October to March, the mean number of CG lightning flashes typically does not exceed 1300 (Fig. 4a). Considering the fractional distribution of CG lightning flashes in each hour, convection in Poland has a well-defined diurnal cycle with the highest probability for CG lightning flashes around 1400–1500 UTC, and the lowest around 0700–0800 UTC. This pattern is evidently visible during the summer and spring, whereas from October to March it is less clear (Figs. 4b and 5a). The spring features the highest fraction (12%) of CG lightning flashes occurring in the afternoon (1200–1500 UTC), while during the summer the peak fraction occurs in the same hours, but with lower values of around 8%–9%. At night, CG lightning flashes are considerably less frequent, with the lowest fractions in spring. During winter and autumn, the fraction of nocturnal CG lightning flashes is slightly higher than during summer and spring.
(a) Mean annual number of hours with at least one CG lightning flash, and (b) mean annual number of CG lightning flashes per 1 km2. Calculations are based on data from the PERUN lightning detection system for 2002–19 (10 km × 10 km grid). Spatial distribution of median of (c) MUCAPE, (d) MUCIN, (e) DLS, and (f) 90th percentile of MUWMAXSHEAR. Convective parameters are derived from ERA5 reanalysis for a period 2002–19 (0.25° × 0.25° grid). Please note that only situations with detected CG lightning flashes are considered.
Citation: Weather and Forecasting 37, 8; 10.1175/WAF-D-21-0099.1
(a) Annual mean number of CG lightning flashes per month, (b) hourly fraction of CG lightning flashes per meteorological season, (c) year-to-year variability of CG lightning flashes per meteorological season, and (d) fraction of nocturnal CG lightning flashes over the years. Calculations were based on the data from the PERUN lightning detection system for the years 2002–19. Please note the logarithmic scale in (c). Nocturnal lightning in (d) is determined based on the solar angle (below −12°).
Citation: Weather and Forecasting 37, 8; 10.1175/WAF-D-21-0099.1
(a) Annual and daily variability of grids with at least one CG lightning flash, (b) median of MUMIXR, (c) LR85, (d) MUCAPE, (e) DLS, and (f) 90th percentile of MUWMAXSHEAR. Convective parameters are derived from ERA5 reanalysis for a period 2002–19. Vertical axis has hourly resolution while horizontal weekly. Labels indicate the first week of each month. Please note that only situations with detected CG lightning flashes are considered.
Citation: Weather and Forecasting 37, 8; 10.1175/WAF-D-21-0099.1
Considering the interannual variability (Fig. 4c), the highest overall number of CG lightning flashes occur every year during the summer (in 2017 almost 1 000 000) and the lowest in the winter (in 2008 only 10). Although the spring typically has a much higher number of CG lightning flashes compared to the autumn, in some years, the number of detections can be comparable (e.g., in 2006, the autumn had a higher number of detections compared to spring). Slight increases in the number of annual CG lightning flashes are probably due to improvements in the PERUN system over time. Strong year-to-year variability is also observed for the annual fraction of nocturnal CG lightning flashes with values from 7% in 2006 to as much as 22% in 2009 (Fig. 4d). However, no significant long-term trend can be defined.
The highest daily number of CG lightning flashes was recorded on 10 August 2017, reaching 141 628 (Table 5), which is 57 981 more detections than the second ranked 28 June 2017. On 10 August 2017, the coverage of thunderstorms was also the largest of all analyzed cases and reached 55 791 km2 (17.8% of the area of Poland). Considering other days, the area covered by grids with CG lightning flashes ranged typically from 10.7% to 12.3% of the area of Poland (Table 5). Among the 10 days with the highest daily number of CG lightning flashes, four cases occurred both in June and July, with two in August. The most intense thunderstorms in terms of 1-min peak CG lightning flash rate occurred on 21 June 2013, 18 June 2013, and 10 August 2017, reaching 108, 80, and 74 CG lightning flashes per 100 km2 per minute, respectively (Table 6).
Top 10 days with the highest number of CG lightning flashes (left), and the highest spatial coverage of 10 km × 10 km grids with at least one detection (right).
Time and location of top 10 grid boxes (10 km × 10 km) with the highest CG lightning flash rates.
b. Annual and diurnal variability of CG lightning flash environments
In this section we present the climatological temporal and spatial distribution of ERA5 convective environments for thunderstorms in Poland. The statistics presented in this section refer only to situations when a CG lightning flash was detected, leading to a smaller sample size during winter and larger in summer (Table 3). Thus, results should be interpreted with caution as nonelectrified convection is not considered.
The components of atmospheric convective instability: MUMIXR (mixing ratio of the most unstable parcel) and LR85 (800–500-hPa temperature lapse rate) are characterized by differing patterns in temporal distributions. Peak values of MUMIXR reach 12 g kg−1 in July and early August, typically around 1800 UTC (Fig. 5b), while LR85 of 7 K km−1 is observed between 1000 and 1600 UTC during spring from March to May (Fig. 5c). MUMIXR has a better relationship with the diurnal cycle of CG lightning flashes but is delayed by about 3 h. The best overlap of MUMIXR and LR85 is in late July and early August between 1200 and 1700 UTC, with the resulting median most unstable convective available potential energy (MUCAPE) reaching around 850 J kg−1 (Fig. 5d). Conversely, when no insolation is available during the night, MUCAPE has its minimum. Only for the period from early July to late August does nocturnal MUCAPE have a median of around 400 J kg−1. During the spring and autumn, peak values of MUCAPE occur around noon, driven by the surface heating and development of steep lapse rates. Climatological patterns in MUCAPE are well correlated with peak CG lightning flash activity and have a well-defined diurnal cycle during the summer in opposition to a poor diurnal cycle during winter when thunderstorms in Poland are typically rare (Figs. 5a,d).
In comparison to MUCAPE, the distribution of deep-layer shear (DLS) features a reversed pattern (Fig. 5e). From April to October, the median of DLS generally does not exceed 10 m s−1, with the lowest values in the afternoon hours (1200–1600 UTC). However, shear notably increases during nighttime hours by 5–7 m s−1, which is particularly visible in the summer months. In the period with lesser convective occurrence from October to April, DLS has a typically less defined diurnal cycle, but much higher overall values (a median of 20–30 m s−1).
A combination of thermodynamic convective instability and vertical wind shear is a useful environmental proxy for assessing thunderstorm severity (Brooks et al. 2003; Púčik et al. 2015; Taszarek et al. 2020). MUWMAXSHEAR, which combines both MUCAPE and DLS, indicates peak severe thunderstorm potential from June to August, with the 90th percentile exceeding 800 m2 s−2 during this period (Fig. 5f). The highest values typically occur between 1600 and 2000 UTC in late July and early August. Although DLS is relatively low in July (median 10–14 m s−1), the high MUCAPE in that period is the main driver of the overall MUWMAXSHEAR value. A similar effect occurs in the diurnal cycle. Although MUCAPE drops during nighttime, MUWMAXSHEAR is still high, as it is compensated by increasing DLS. This reflects an enhanced nocturnal potential for severe thunderstorms, that as a result of increasing shear, often evolve into large and well-organized mesoscale convective systems (MCS) with the potential to produce severe wind and heavy rain (Geerts et al. 2017; Reif and Bluestein 2017; Haberlie and Ashley 2019; Surowiecki and Taszarek 2020). Noteworthy also are the increased values of MUWMAXSHEAR (around 500 m2 s−2 of 90th percentile) during spring and autumn, which are mainly driven by large DLS and rather marginal MUCAPE. These conditions are so-called high-shear low-CAPE (HSLC) environments, and are often associated with an enhanced potential of squall lines and low-topped supercells capable of producing damaging winds and tornadoes (Sherburn and Parker 2014; Sherburn et al. 2016; Anderson-Frey et al. 2019; Mathias et al. 2019; Gatzen et al. 2020).
c. Spatial variability of CG lightning flash environments
Spatially, median MUCAPE (only for CG lightning flash events) increases from the northwest toward the southeast (Fig. 3c), and thus the lowest values are along the Baltic Sea coast (450 J kg−1) and the highest in the Bieszczady Mountains over the southeast (700 J kg−1). This pattern is consistent with the frequency of thunderstorm days based on SYNOP reports (Bielec-Bąkowska 2003; Czernecki et al. 2016).
The median of the MUCIN (convective inhibition) has a spatial pattern reversed to MUCAPE with the strongest inhibition over the western and northwestern parts of the country reaching −30 J kg−1 and weakest over the south (−10 J kg−1), indicating the area with the lowest resistance to convection initiation (Fig. 3d). We hypothesize that the enhanced inhibition occurring along the coast of the Baltic Sea is associated with low sea surface temperatures that lead to temperature inversions in typically marginal MUCAPE situations. The regions with the strongest inhibition over western Poland experience the lowest average annual rainfall in Poland (Lorenc 2005), which also seems to be consistent with the reduced convective frequency over that area (Fig. 3a).
Similarly to MUCIN, the spatial patterns in DLS are also reversed with respect to MUCAPE, with the lowest values over southeastern Poland (∼10 m s−1) and the highest over the northwest (∼14 m s−1; Fig. 3e). This pattern indicates more frequent HSLC environments over northwestern Poland and in turn more common low-shear, high-CAPE (LSHC) situations over the southeast. However, noteworthy is the northeastern part of the country, where a combination of climatologically enhanced MUCAPE (∼600 J kg−1) and DLS (∼13 m s−1) indicates increased conditional potential for severe thunderstorms in this region, despite their limited frequency compared to the southeast (Fig. 3a). This is confirmed by the spatial distribution of the 90th percentile of MUWMAXSHEAR, which in this region reaches even 850 m2 s−2— the highest values among the entire country (Fig. 3f).
d. Variability of convective environments in relation to CG lightning flash rates
The two-dimensional distributions of MUCAPE and DLS for different CG lightning flash rate intensity categories (weak, moderate, intense, extreme; Table 1, Fig. 6) indicate that the conditional probability for larger CG lightning flash rates increases as convective instability and vertical wind shear increase. The median for MUWMAXSHEAR increases for each category with 330, 375, 441, and 519 m2 s−2, respectively (Fig. 6). This result is broadly consistent with Liu et al. (2020).
Number of situations with specific MUCAPE and DLS values for weak, moderate, intense, and extreme thunderstorms (divided based on a peak 1-min CG lightning flash rate, as in Table 1). Constant values of MUWMAXSHEAR (a product of MUCAPE and DLS; Taszarek et al. 2020) are indicated by dashed lines. White points indicate the median value for each category. Convective parameters are derived from ERA5 reanalysis for a period 2002–19. Please note that only situations with detected CG lightning flashes are considered.
Citation: Weather and Forecasting 37, 8; 10.1175/WAF-D-21-0099.1
Both weak (Fig. 6a) and moderate (Fig. 6b) thunderstorm categories feature similar kinematic environments but a slightly differing thermodynamic convective instability (a difference of 129 J kg−1 in the medians of MUCAPE). One can also see a dependence that thunderstorms developing in an environment with increased MUCAPE require lower DLS and conversely, a higher DLS is typically accompanied by weaker convective instability. This pattern is well depicted by the constant values of MUWMAXSHEAR on the scatterplots (dashed lines on Fig. 6). Thunderstorms with both high CAPE and high DLS are very rare in Poland. For intense and extreme thunderstorms (Figs. 6c,d), a shift in the number of cases toward higher MUCAPE and a gradual increase in DLS can be recognized. In the extreme thunderstorm category (Fig. 6d), a median of 834 J kg−1 MUCAPE and 12.7 m s−1 DLS is notably higher compared to other classes. However, the largest increase of 153 J kg−1 in the medians of MUCAPE is between the moderate and intense thunderstorm categories.
In summary, although the probability for higher CG lightning flash rates increases with both DLS and MUCAPE, the increase in the latter is generally more important. A median MUWMAXSHEAR of 441 m2 s−2 for the intense thunderstorm category is similar to the value of 450 m2 s−2 from Taszarek et al. (2019), which was considered as a discriminator between severe and nonsevere thunderstorms across central Europe (conditional on successful convective initiation). On the other hand, a different pattern of DLS and CAPE dependency was found in Europe for the relative frequency of lightning (Westermayer et al. 2017). High relative frequency was found in low (<10 m s−1) and high (>20 m s−1) DLS environments, while CAPE values below approximately 200–400 J kg−1 demonstrated low relative frequency. This suggests that both enhanced CAPE and DLS are necessary for storms producing high CG lightning flash rates, but only a small amount of CAPE is needed for occurrence of any lightning (given successful convective initiation). However, as thunderstorms with high CG lightning flash rates may not always be associated with severe weather and vice versa, caution needs to be taken when interpreting this result.
To provide additional statistical tests for MUCAPE, DLS and MUWMAXSHEAR, we calculated the Heidke skill score (HSS). Peak HSS values were found for MUCAPE of 1000 J kg−1 (HSS of 0.13), DLS of 12.5 m s−1 (HSS of 0.05), and MUWMAXSHEAR of 600 m2 s−2 (HSS of 0.16) for extreme versus weak CG lightning flash rate category (Fig. 7). A shift of peak HSS values among specific CG lightning flash rate categories was observed with increasing values of MUCAPE and MUWMAXSHEAR, while for DLS, differences were marginal. The biggest differences among categories and the highest HSS were observed for MUWMAXSHEAR, which seems to better discriminate among CG lightning flash rate categories compared to MUCAPE or DLS alone. Overall low values of HSS indicate that predicting CG lightning flash rate with convective parameters is a very challenging task, and its application in operational forecasting is limited. However, forecasting any type of convective hazards across Europe is a difficult task (Brooks et al. 2011).
Heidke skill score for MUCAPE, DLS, and MUWMAXSHEAR calculated by comparing the weak thunderstorm intensity category with at least moderate (orange), at least intense (red), and extreme (magenta).
Citation: Weather and Forecasting 37, 8; 10.1175/WAF-D-21-0099.1
The increase of MUMIXR and MUCAPE corresponds with increasing conditional probability for a higher CG lightning flash rate category. The annual and daily distributions of these indices are similar (Figs. 8a,c), reaching the highest values (medians from 10 g kg−1 and 500 J kg−1 to 12 g kg−1 and 1000 J kg−1 for weak to extreme thunderstorms) in summer (June–August), except for the 90th percentile of MUCAPE for extreme thunderstorm category, which is notably highest (∼2700 J kg−1) in June (compared to July for MUMIXR). Diurnally, for both indices, the highest values are in the afternoon (Figs. 8b,d). MUMIXR median oscillates around 10–13 g kg−1 during both day and night (Fig. 8b). The highest median in MUCAPE for weak and moderate thunderstorms occurs between 1000 and 1400 UTC, for strong between 1200 and 1400 UTC, and for extreme between 1400 and 1800 UTC. This pattern indicates that storms developing in a high MUCAPE environment during late afternoon are more likely to become severe, which is possibly linked to increasing vertical wind shear toward late evening and night.
Box-and-whisker plots presenting the (left) annual and (right) diurnal distributions of thunderstorm intensity classes (weak, moderate, intense, extreme; as in Table 1) for (a),(b) MUMIXR, (c),(d) MUCAPE, and (e),(f) MUCIN. Convective parameters are derived from ERA5 reanalysis for a period 2002–19. Please note that only situations with detected CG lightning flashes are considered. The horizontal line inside the box demonstrates the median, the extent of the box is from the 25th to 75th percentile, and whiskers represent the 10th and 90th percentiles.
Citation: Weather and Forecasting 37, 8; 10.1175/WAF-D-21-0099.1
Analysis of MUCIN demonstrated that environments for severe and extreme thunderstorms are associated not only with higher CAPE and shear but also with stronger inhibition (Figs. 8e,f). The strongest inhibition (the most negative values of MUCIN) occurs in the summer, but in contrast to MUCAPE, value distributions are relatively even from June to September (median for all categories from −5 to −15 J kg−1). Diurnally, the weakest inhibition (highest values) occurs during strong surface heating between 1000 and 1600 UTC, while outside this period a median of MUCIN varies from −5 to −20 J kg−1 (Fig. 8f). Convective inhibition is typically stronger in environments favorable for higher CG lightning flash rates. This is confirmed by calculating the conditional probability for each thunderstorm category and MUCIN intervals from 0 to −50 J kg−1 in 5 J kg−1 steps (Table 7). The results indicate that for each thunderstorm category, probability for the occurrence is highest for MUCIN values above −10 J kg−1 (weaker inhibition) and lowest below −40 J kg−1. While in higher MUCIN (weaker inhibition), conditional probability increases along with decreasing thunderstorm intensity, an opposite relationship can be observed for lower MUCIN (stronger inhibition). Delayed convection initiation during the day driven by stronger inhibition subsequently promotes stronger moisture pooling, higher convective instability, better storm organization (shear is higher toward evening hours; Fig. 9b), and a more sudden development of convective updrafts promoting higher CG lightning flash rates. Moreover, environments favorable for extreme thunderstorms are often associated with advection of very warm air with steep lapse rates and an elevated mixed layer, a situation commonly referred to as a “loaded gun” profile. These environments in Europe are often associated with “Spanish plume” synoptic setups (Carlson and Ludlam 1968; Van Delden 2001; Mathias et al. 2017). On the other hand, stronger convective inhibition during extreme thunderstorms is also pronounced at night (medians around −20 J kg−1), which may correspond to well organized, elevated thunderstorms (Colman 1990; Grant 1995). These thunderstorms initiate above a stable layer without being influenced by strong mixed layer or surface-based convective inhibition.
As in Fig. 7, but for (a),(b) DLS, (c),(d) LLS, and (e),(f) SRH03.
Citation: Weather and Forecasting 37, 8; 10.1175/WAF-D-21-0099.1
Conditional probability distribution for MUCIN intervals and thunderstorm category.
As previously discussed, months with the highest thermodynamic convective instability feature the lowest values of kinematic parameters (Fig. 9). DLS, 0–1-km wind shear [low-level wind shear (LLS)], and SRH03 from April to September has a similar range of values (∼11 m s−1, ∼6 m s−1, and ∼70 m2 s−2, respectively) among thunderstorm intensity classes (Figs. 9a,c,e). Only for the extreme thunderstorms, this relationship is better pronounced with overall slightly higher values of all three metrics. The diurnal distributions of DLS, LLS, and SRH03 (Figs. 9b,d,f) are characterized by the lowest values in the hours with the highest thermodynamic convective instability (from 1200 to 1600 UTC), which is especially evident for LLS that has a stronger diurnal amplitude compared to DLS and SRH03. At night and in the morning, the extreme thunderstorm category features higher median values compared to other categories, reaching 17 m s−1, 7 m s−1, and 140 m2 s−2 for DLS, LLS, and SHR03, respectively. Low differences between thunderstorm categories, demonstrates limited utility in forecasting specific flash rates using only kinematic parameters.
Composite indices combining multiple parameters into one product aimed at forecasting specific phenomena offers an attractive approach for forecasters, and thus we also evaluated them in our study. MUWMAXSHEAR reaches the highest median during summer months (300–600 m2 s−2) with a 90th percentile of extreme thunderstorms exceeding 1250 m2 s−2 (Fig. 10a). A notion of increasing MUWMAXSHEAR with increasing CG lightning flash rates is valid from April to October, while from November to March, such a dependence is not present. Diurnally (Fig. 10b), the biggest difference between the weak and extreme category is between 0400 and 0800 UTC (morning hours of local time in Poland)—the period with the lowest frequency of thunderstorms and lowest convective instability. It is also worth highlighting that, contrary to previously analyzed variables (LR85, MUCAPE, MUCIN, and LLS), MUWMAXSHEAR has no clearly defined diurnal cycle. This is because of the distribution of its components—convective instability, which has a clear diurnal cycle with the maximum in the afternoon, and shear, which has an inverted diurnal cycle with minimum values in the afternoon. The combination of these two results in a flattened diurnal distribution.
As in Fig. 7, but for (a),(b) MUWMAXSHEAR, (c),(d) SCP, and (e),(f) HSI.
Citation: Weather and Forecasting 37, 8; 10.1175/WAF-D-21-0099.1
While we are aware that SCP and HSI are parameters designed to forecast supercells and large hail, the strength of the updraft in the supercells and conditions favorable for large hail formation are also supportive for lightning generation. Thus, we evaluate these parameters in our study as well and note that their increasing values are consistent with more frequent lightning. The highest values of both SCP and HSI are observed between May and August. While values of SCP during other months rarely exceed 0, the HSI timeframe with elevated values is from April to October. A rapid increase in the HSI in April is maintained at a similar level until August with the 90th percentile of every storm category exceeding a value of 2 (Fig. 10e), which highlights an elevated risk of large hail in this period. For the SCP, 90th percentile for intense and extreme thunderstorms during the summer exceeds a value of 2 (Fig. 10c), which is a good discriminator between supercell and nonsupercell thunderstorms based on a calibration derived from environments across the United States (Thompson et al. 2003, 2007; Gropp and Davenport 2018). This may lead to the conclusion that conditions supportive for supercell development in Poland are relatively uncommon and occur typically between May and August. Diurnally, SCP has no clear peak, but its minima (with the 90th percentile below 0.7) can be explicitly depicted between 0000 and 0200 UTC (Fig. 10d). In contrast, HSI has a clear diurnal cycle (Fig. 10f) with its peak of ∼1.2 (medians) between 1200 and 1400 UTC (hours with the highest lapse rates). For both parameters, higher values are characteristic in more intense thunderstorm categories, but for SCP, the difference between categories is even higher. The distribution of the 90th percentile for HSI demonstrates rather similar values (3.0–3.4 during summer) for each thunderstorm category, while for SCP, extreme thunderstorms have clearly higher values compared to other categories. Based on that, we can conclude that supercells in Poland can occur at any time of the aforementioned period and that CG lightning flash rates increase with increasing chances for supercells (i.e., values of SCP). This result is unsurprising, as powerful updrafts in supercell thunderstorms may reach vertical velocities of even 50 m s−1 (Lehmiller et al. 2001) and drive generation of very large CG lightning flash rates.
In the distributions presented, considerable overlap occurred between thunderstorm categories. For convective instability and composite parameters, this overlap was less pronounced, suggesting that CG lightning flash rates are more sensitive to convective instability parameters than shear. Improving the analysis by the relation of convective parameters with the radar data would probably reduce this overlap, but is out of the scope of this study. In addition to the parameters evaluated in this study, in the appendix, we provide median, 10th, and 90th percentiles for specific CG lightning flash rate categories.
e. Importance of relative humidity on convection initiation
The last analyzed meteorological aspect is the relative humidity at the isobaric levels of 850, 700, and 500 hPa, and a mean in the 0–4 km AGL layer (Fig. 11). Relative humidity, although it does not influence thunderstorm intensity, is a crucial factor for the development of any thunderstorms (Westermayer et al. 2017). The lower the relative humidity is, the stronger the dry air entrainment can be, which may strongly limit the development of convective updrafts. On the other hand, relatively dry mid tropospheric air but in environments with strong synoptic-scale forcing and sufficient humidity at low levels, can result in releasing potential instability and successful convection development. This effect is demonstrated on a rather flat distribution of relative humidity at 500 hPa compared to lower levels.
Histograms of relative humidity at 850-, 700-, and 500-hPa isobaric levels and a mean layer of 0–4 km AGL for all CG lightning flashes in Poland. Yellow lines indicate the interquartile range, i.e., position of the 25th percentile (Q1) and 75th percentile (Q3), and the red line indicates the median (Q2). Relative humidity values are derived from ERA5 reanalysis for a period 2002–19.
Citation: Weather and Forecasting 37, 8; 10.1175/WAF-D-21-0099.1
Among the previously mentioned pressure levels, the highest values of the interquartile range and the median of relative humidity occurred at 850 hPa (approximately 1.5 km MSL; Fig. 11b). At this level, 3/4 of all analyzed lightning cases had a relative humidity between 70% and 85%. Equally high values of the interquartile range occurred also in a mean 0–4-km relative humidity (Fig. 11a), stretching from 66% to 80%, while at the level of 700 hPa (Fig. 11c) this spread varied from 65% to 83%. The relative humidity at the level of 500 hPa turned out to be less important as thunderstorms occurred in a wide spectrum of values and the median was 56% (Fig. 11d). A mean 0–4-km relative humidity turned out to be the most consistent in terms of value distribution, and had a well-defined peak at 75%. These results are consistent with Westermayer et al. (2017), who also highlighted the importance of relative humidity with respect to convection initiation across Europe. That study concluded that the probability for a thunderstorm development drops significantly when the relative humidity is below 50%, which is very similar to the distributions provided in our work, but was for a longer research period.
5. Concluding remarks
In this study we used over 8 million CG lightning flashes detected across Poland over the course of 18 years (2002–19) and intercompared specific CG lightning flash rates with accompanying convective environments from ERA5. The hourly steps of ERA5 and a large sample size of the PERUN lightning database provided the opportunity to explore the diurnal cycles of thunderstorm environments and their relation to CG lightning flash rates, which was not possible with a lower resolution of prior reanalyses. We also evaluated if convective parameters can serve as predictors of specific CG lightning flash rate intensity categories (an aspect rarely studied in prior research) and investigated their spatiotemporal climatological variability. We proposed a new definition for thunderstorm intensity categories by using 1-min peak CG lightning flash rates. The proposed methodology features some limitations: severe, shallow convection may have poor lightning activity, the choice of timeframe and grid to define flash rates is arbitrary, and intracloud flashes, which are important in lightning jump calculations (Farnell et al. 2018; Farnell and Rigo 2020), were omitted in our study (due to low quality of this data). Synthesis of PERUN lightning data with ERA5 reanalysis yielded numerous findings, among which the most important are listed below:
-
A mean of 161 thunderstorm days occurs each year in Poland, most frequently over the southeast. However, the highest frequency of CG lightning flashes is in eastern Poland with up to 3 flashes km−2 yr−1.
-
July is the month with the largest number of CG lightning flashes (around 150 000 per year), reaching peak frequency typically between 1400 and 1600 UTC. Nocturnal CG lightning flashes share an annual fraction of 4%–24% for the specific years.
-
The greatest convective instability occurs from June to August between 1400 and 1600 UTC, with the highest low-level moisture at 1800 UTC. Thunderstorms occurring from March to May feature the highest midtropospheric lapse rates. Patterns in wind shear are reversed to MUCAPE and have the highest values during the winter and at night. The most conducive conditions for convection initiation in Poland are between 1000 and 1400 UTC as evidenced by the weakest convective inhibition and steepest lapse rates.
-
The best overlap of convective instability and wind shear is in July and August, typically between 1400 and 2000 UTC (based on the 90th percentile of MUWMAXSHEAR). However, values exceeding 500 m2 s−2 can occur from March to November and at any time of the day.
-
Thunderstorms in Poland are the most frequent in MUCAPE below 1000 J kg−1 and DLS between 8 and 15 m s−1. When greater convective instability is available, DLS is typically lower (warm-season thunderstorms). Conversely, in highly sheared environments, convective instability is typically marginal (cold-season thunderstorms). Situations with both high MUCAPE and DLS are rare in Poland. LSHC environments are the most frequent over southeastern Poland, while HSLC are the most common over the northwest.
-
Along with increasing MUCAPE and DLS, peak CG lightning flash rates increase as well. However, a combination of these two (MUWMAXSHEAR) demonstrated the highest (but still limited) skill compared to MUCAPE or DLS alone.
-
The vast majority of thunderstorms had a low to midlevel relative humidity higher than 60%.
-
Proposed thunderstorm intensity proxy based on CG lightning flash rates demonstrated a correlation with the number of ESWD severe weather reports.
The results listed above are broadly consistent with prior studies concerning the climatological aspects of lightning (Anderson and Klugmann 2014; Poelman et al. 2016; Taszarek et al. 2019; Enno et al. 2020), convective environments accompanying central European thunderstorms (Púčik et al. 2015; Taszarek et al. 2020; Walawender et al. 2017; Westermayer et al. 2017), and the comparison of satellite observations of thunderstorms with ERA-Interim environments over tropical and subtropical regions (Liu et al. 2020).
Given a complex European orography and coastline that lead to high variability in local thunderstorm climatologies, focusing on smaller areas (e.g., area of Poland) can provide more precise information with respect to what is happening on the regional scale. This is confirmed by comparison to prior research that showed maximum thunderstorm activity in Poland being shifted by more than a month compared to neighboring parts of Europe (Taszarek et al. 2020). Concerning spatiotemporal distributions of convective environments, the area of Poland seems to not differ substantially from the area of eastern and central Europe. However, significant differences can be found if Poland is compared to regions of the Mediterranean or northwestern Europe that are more heavily influenced by their proximity to the sea surface and cyclonic activity (Tudurí and Ramis 1997; Cohuet et al. 2011; Holley et al. 2014). Mediterranean storms also typically feature much higher instability with a peak of the season shifted toward late summer and autumn (Riemann-Campe et al. 2009). Comparing our results to previous research, we noted that while enhanced CAPE and DLS are typically necessary for storms producing high CG lightning flash rates, only a small amount of CAPE (∼200 J kg−1) is needed for the occurrence of any lightning, given successful convective initiation (Westermayer et al. 2017).
We also found that median MUCAPE and MUWMAXSHEAR values for the extreme lightning category during summer correspond to large hail environments in central Europe (Púčik et al. 2015; Taszarek et al. 2020). On the other hand, a median DLS of 12.5 m s−1 for the extreme lightning category was much lower than 20 m s−1 typically required for significant tornadoes and very large hail. This indicates that while there is some relationship between the relative number of ESWD reports and CG lightning flash rates, the use of lightning data as a proxy for convective hazards features certain limitations. As we showed in this study, predicting CG lightning flash rates with convective parameters is a very challenging task and its application in operational forecasting is limited. However, it is worth highlighting that forecasting convective hazards such as large hail, tornadoes, and severe wind has always been a very challenging task for operational forecasters across Europe (Brooks et al. 2011). The use of CG lightning flash rates for studying storm intensity offers a much larger sample size and spatial homogeneity compared to severe weather reports that are biased by population density. With improvements in ground-based lightning detection networks, we expect similar studies to be developed in future for other regions of the world, thus contributing to better understanding of local thunderstorm climatologies and their accompanying environments.
Acknowledgments.
This research was funded by the Institute of Meteorology and Water Management - Polish Research Institute (project S7), the Priority Research Area Anthropocene under the program “Excellence Initiative—Research University” at the Jagiellonian University in Kraków, and supported by a grant from the Polish National Science Centre (project 2020/39/D/ST10/00768). The reanalysis computations were performed in Poznan Supercomputing and Networking Center (project 448).
Data availability statement.
ERA5 data (temperature, specific humidity, geopotential, pressure, U, and V) were downloaded from the European Centre for Medium-Range Weather Forecasts (ECMWF), Copernicus Climate Change Service (C3S) at Climate Data Store (https://cds.climate.copernicus.eu/). PERUN lightning data was provided by the Polish Institute of Meteorology and Water Management - National Research Institute and due to the proprietary nature of the data, cannot be made openly available. Contact rafal.lewandowski@imgw.pl for usage information.
APPENDIX
Statistical Distribution of Convective Indices in Thunderstorm Intensity Classes
This appendix contains statistical distribution (10th, 50th, and 90th percentiles) of all convective indices that were used in this study in function of thunderstorm intensity Table A1.
The 10th, 50th, and 90th percentiles for selected convective parameters associated with weak (wea), moderate (mod), intense (int), and extreme (ext) CG lightning flash rate categories. EL = equilibrium level; MUEL = MU equilibrium level; WBZ = wet bulb zero; STP = significant tornado parameter; CPTP = cloud physics thunder parameter.
REFERENCES
Allen, J. T., and D. J. Karoly, 2014: A climatology of Australian severe thunderstorm environments 1979–2011: Inter-annual variability and ENSO influence. Int. J. Climatol., 34, 81–97, https://doi.org/10.1002/joc.3667.
Allen, J. T., and M. K. Tippett, 2015: The characteristics of United States hail reports: 1955–2014. Electron. J. Severe Storms Meteor., 10 (3), 1–31.
Allen, J. T., D. J. Karoly, and G. Mills, 2011: A severe thunderstorm climatology for Australia and associated thunderstorm environments. Aust. Meteor. Oceanogr. J., 61, 143–158, https://doi.org/10.22499/2.6103.001.
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, https://doi.org/10.5194/nhess-14-815-2014.
Anderson-Frey, A. K., Y. P. Richardson, A. R. Dean, R. L. Thompson, and B. T. Smith, 2016: Investigation of near-storm environments for tornado events and warnings. Wea. Forecasting, 31, 1771–1790, https://doi.org/10.1175/WAF-D-16-0046.1.
Anderson-Frey, A. K., Y. P. Richardson, A. R. Dean, R. L. Thompson, and B. T. Smith, 2019: Characteristics of tornado events and warnings in the southeastern United States. Wea. Forecasting, 34, 1017–1034, https://doi.org/10.1175/WAF-D-18-0211.1.
Antonescu, B., and S. Burcea, 2010: A cloud-to-ground lightning climatology for Romania. Mon. Wea. Rev., 138, 579–591, https://doi.org/10.1175/2009MWR2975.1.
Apke, J. M., J. R. Mecikalski, K. Bedka, E. W. McCaul Jr., C. R. Homeyer, and C. P. Jewett, 2018: Relationships between deep convection updraft characteristics and satellite-based super rapid scan mesoscale atmospheric motion vector–derived flow. Mon. Wea. Rev., 146, 3461–3480, https://doi.org/10.1175/MWR-D-18-0119.1.
Bielec-Bąkowska, Z., 2002: Zróżnicowanie przestrzenne i zmienność wieloletnia występowania burz w Polsce (1949–1998). Wydawnictwo Uniwersytetu Śląskiego, 159 pp.
Bielec-Bąkowska, Z., 2003: Long-term variability of thunderstorms occurrence in Poland in the 20th century. Atmos. Res., 67–68, 35–52, https://doi.org/10.1016/S0169-8095(03)00082-6.
Bielec-Bąkowska, Z., 2013: Burze i grady w Polsce (Thunderstorms and hails in Poland). Prz. Geogr., 132, 99–132, https://doi.org/10.4467/20833113PG.13.005.1095.
Biron, D., 2009: LAMPINET-Lightning detection in Italy. Lightning: Principles, Instruments and Applications: Review of Modern Lightning Research, H. D. Betz, U. Schumann, and E. Laroche, Eds., Springer, 141–159.
Blair, S. F., and Coauthors, 2017: High-resolution hail observations: Implications for NWS warning operations. Wea. Forecasting, 32, 1101–1119, https://doi.org/10.1175/WAF-D-16-0203.1.
Bodzak, P., 2006: Detekcja i lokalizacja wyładowań atmosferycznych (Lightning Detection and Localization). Instytut Meteorologii i Gospodarki Wodnej, 135 pp.
Bodzak, P., Z. Dziewit, and G. Wojciech, 2006: The ratio of cloud to cloud-to-ground flashes in summer seasons for Poland territory on the basis of data from PERUN system. 19th Int. Lightning Detection Conf. (ILDC)/1st Int. Lightning Meteorology Conf. (ILMC), Tucson, AR, Vaisala, https://my.vaisala.net/en/events/ildcilmc/archive/Pages/ILDC-2006-Archive.aspx.
Bright, D. R., M. S. Wandishin, R. E. Jewell, and S. J. Weiss, 2004: A physically based parameter for lightning prediction and its calibration in ensemble forecasts . 22nd Conf. on Severe Local Storms, Hyannis, MA, Amer. Meteor. Soc., 4.3., http://ams.confex.com/ams/pdfpapers/84173.pdf.
Brooks, H. E., J. W. Lee, and J. P. Craven, 2003: The spatial distribution of severe thunderstorm and tornado environments from global reanalysis data. Atmos. Res., 67–68, 73–94, https://doi.org/10.1016/S0169-8095(03)00045-0.
Brooks, H. E., A. R. Anderson, K. Riemann, I. Ebbers, and H. Flachs, 2007: Climatological aspects of convective parameters from the NCAR/NCEP reanalysis. Atmos. Res., 83, 294–305, https://doi.org/10.1016/j.atmosres.2005.08.005.
Brooks, H. E., and Coauthors, 2011: Evaluation of European Storm Forecast Experiment (ESTOFEX) forecasts. Atmos. Res., 100, 538–546, https://doi.org/10.1016/j.atmosres.2010.09.004.
Bunkers, M. J., B. A. Klimowski, J. W. Zeitler, R. L. Thompson, and M. L. Weisman, 2000: Predicting supercell motion using a new hodograph technique. Wea. Forecasting, 15, 61–79, https://doi.org/10.1175/1520-0434(2000)015<0061:PSMUAN>2.0.CO;2.
Bürgesser, R. E., 2017: Assessment of the World Wide Lightning Location Network (WWLLN) detection efficiency by comparison to the Lightning Imaging Sensor (LIS). Quart. J. Roy. Meteor. Soc., 143, 2809–2817, https://doi.org/10.1002/qj.3129.
Calhoun, K. M., D. R. MacGorman, C. L. Ziegler, and M. I. Biggerstaff, 2013: Evolution of lightning activity and storm charge relative to dual-doppler analysis of a high-precipitation supercell storm. Mon. Wea. Rev., 141, 2199–2223, https://doi.org/10.1175/MWR-D-12-00258.1.
Carlson, T. N., and F. H. Ludlam, 1968: Conditions for the occurrence of severe local storms. Tellus, 20, 203–226, https://doi.org/10.3402/tellusa.v20i2.10002.
Celiński-Mysław, D., and D. Matuszko, 2014: An analysis of selected cases of derecho in Poland. Atmos. Res., 149, 263–281, https://doi.org/10.1016/j.atmosres.2014.06.016.
Celiński-Mysław, D., A. Palarz, and M. Taszarek, 2020: Climatology and atmospheric conditions associated with cool season bow echo storms in Poland. Atmos. Res., 240, 104944, https://doi.org/10.1016/j.atmosres.2020.104944.
Chernokulsky, A. V., M. V. Kurgansky, and I. I. Mokhov, 2019: On characteristic reanalysis-based values of convective instability indices for Northern Eurasia tornadoes. IOP Conf. Ser. Earth Environ. Sci., 231, 012012, https://doi.org/10.1088/1755-1315/231/1/012012.
Christian, H. J., and Coauthors, 1999: The lightning imaging sensor. 11th Int. Conf. on Atmospheric Electricity, H. J. Christian, Ed., NASA Conf. Publ. NASA/CP-1999-209261, 746–749.
Coffer, B. E., and M. D. Parker, 2015: Impacts of increasing low-level shear on supercells during the early evening transition. Mon. Wea. Rev., 143, 1945–1969, https://doi.org/10.1175/MWR-D-14-00328.1.
Cohuet, J. B., R. Romero, V. Homar, V. Ducrocq, and C. Ramis, 2011: Initiation of a severe thunderstorm over the Mediterranean Sea. Atmos. Res., 100, 603–620, https://doi.org/10.1016/j.atmosres.2010.11.002.
Colman, B. R., 1990: Thunderstorms above frontal surfaces in environments without positive CAPE. Part II: Organization and instability mechanisms. Mon. Wea. Rev., 118, 1123–1144, https://doi.org/10.1175/1520-0493(1990)118<1123:TAFSIE>2.0.CO;2.
Cummins, K. L., and M. J. Murphy, 2009: An overview of lightning locating systems: History, techniques, and data uses, with an in-depth look at the U.S. NLDN. IEEE Trans. Electromagn. Compat., 51, 499–518, https://doi.org/10.1109/TEMC.2009.2023450.
Czernecki, B., M. Taszarek, L. Kolendowicz, and J. Konarski, 2016: Relationship between human observations of thunderstorms and the PERUN lightning detection network in Poland. Atmos. Res., 167, 118–128, https://doi.org/10.1016/j.atmosres.2015.08.003.
Czernecki, B., M. Taszarek, M. Marosz, M. Półrolniczak, L. Kolendowicz, A. Wyszogrodzki, and J. Szturc, 2019: Application of machine learning to large hail prediction - The importance of radar reflectivity, lightning occurrence and convective parameters derived from ERA5. Atmos. Res., 227, 249–262, https://doi.org/10.1016/j.atmosres.2019.05.010.
Deierling, W., and W. A. Petersen, 2008: Total lightning activity as an indicator of updraft characteristics. J. Geophys. Res., 113, D16210, https://doi.org/10.1029/2007JD009598.
Dennis, E. J., and M. R. Kumjian, 2017: The impact of vertical wind shear on hail growth in simulated supercells. J. Atmos. Sci., 74, 641–663, https://doi.org/10.1175/JAS-D-16-0066.1.