Comparison of thunderstorm days in Poland based on SYNOP reports and PERUN lightning detection system

Thunderstorms, which are classified as extreme weather phenomena, are electrometeors (Bielec-Bąkowska 2003). According to the Meteorological Dictionary by Niedźwiedź (2003), thunderstorm days (TDs) are days on which atmospheric discharges related to the occurrence of cumulonimbus clouds are observed. Near and distant storms are distinguished: for near storms, the time between noting the lightning discharge and hearing the following thunder should not exceed 10 seconds (distance approx. 3 km). Also, according to the AMS Glossary of Meteorology, a thunderstorm is defined as ‘a local storm, invariably produced by a cumulonimbus cloud and always accompanied by lightning and thunder, usually with strong gusts of wind, heavy rain, and sometimes with hail’ (Byers & Braham 1949). Thus, it follows that lightning is inherent to thunderstorms, and must occur in order to speak of the occurrence of a thunderstorm.

Observing the occurrence of thunderstorms is burdened with the problem of the observer’s subjective judgment (Bielec & Kolendowicz 2001). With distant thunderstorms, some flashes are visible from miles away, especially at night (Wu et al. 2016). In addition, many weather stations are located in cities, where light pollution of the sky (Lechner & Arns 2013) masks atmospheric discharges, and the noise coming from the environment drowns out the thunder which follows the lightning. In mountainous regions, city centres, and wooded areas, the view of the entirety of the sky is reduced by hills, buildings, or trees. In Poland, difficult conditions for the observation of lightning flashes are found especially in mountainous areas in the south of the country (Fig. 1A). The systematic observation of thunderstorms in Poland are carried out at stations belonging to the Institute of Meteorology and Water Management of the National Research Institute (IMGW-PIB). The results of these observations have been published in many articles (e.g., Kolendowicz 1997, 2006; Bielec 2000; Lorenc 2005; Bielec-Bąkowska 2013; Bielec-Bąkowska et al. 2021).

Figure 1.

A - Hypsometric map of Poland based on the Shuttle Radar Topography Mission Global Coverage (SRTM3) including geographical regions (source: NASA 2021). B - Location of the meteorological stations used in the study. C - Locations of lightning detection sensors in the PERUN network, with buffer zones. D - Average CG lightning accuracy based on the PERUN database during 2002–2020

Source: own study

Lightning detection systems provide more objective data about thunderstorms. These have been introduced in many countries. A lightning detection system has been operating in Poland, since 2002. Initially, it was called SAFIR (Surveillance et Alerte Foundre par Interférométrie Radioélectrique), but it is now called PERUN, in reference to the god of thunder in Slavic mythology. The system now consists of 12 antennas and can detect up to 100 discharges per second with a location accuracy of up to 1 km in the central part of the country, while in border areas this accuracy decreases (Bodzak 2006; Czernecki et al. 2016). PERUN has been included into the European Blitzortung system (Gamracki 2015). On the basis of these data, climatological analyses of lightning discharges have been performed for Europe (e.g. Taszarek et al. 2019), Poland (Taszarek et al. 2015; Czernecki et al. 2016; Sulik 2021; Sulik 2022), and at a regional scale for the Kujawsko-Pomorskie Voivodeship (Sulik & Kejna 2022).

Thunderstorms are among the most dangerous meteorological phenomena, in Poland as in most regions around the world. This is due to the significant damage generated by thunderstorm events each year. The main danger caused by thunderstorms is the risk of lightning strikes, damaging cloud-to-ground flashes, large hail, or at the very least, severe convective wind gusts, which have often caused natural disasters in Poland, for example, on August 10 and 11, 2017 (Sulik & Kejna 2020). Of course, damage from severe weather phenomena has a direct relationship to material and economic losses (Mäkelä et al. 2013).

The progressive increase in air temperatures around the globe is affecting the frequency and strength of extreme phenomena. Even small changes in climate can cause a higher frequency of extreme weather events, including thunderstorms (Allen 2018). There is considerable regional variation in the occurrence of thunderstorms worldwide (e.g., Kuleshov et al. 2002; Lavigne et al. 2019; Taszarek et al. 2019; Koehler 2020).

There are also significant climatic changes occurring in Poland (ed. Falarz 2021). During the period 1966-2018, the air temperature increased at a rate of 0.33°C per 10 years. There were regional differences during the period 1951-2018: the warming was greatest in central-western Poland (0.3°C per 10 years) and slightly lower (below 0.2°C per 10 years) in the south-east (Kejna & Rudzki 2021). The greatest warming occurred in winter and summer (Ustrnul et al. 2021). Maximum air temperatures are also increasing; for example, in the summer seasons during the period 1951-2015 they rose at a rate of over 0.4°C per 10 years (Wypych et al. 2017). Despite this, there is no clear signal for an increase in the frequency of thunderstorms in Poland. For the period 1951–2018, an increase in the frequency of TDs was noted only in the eastern part of the country, while in the west the trend was not statistically significant, and at some stations there was even a decrease in frequency (Bielec-Bąkowska et al. 2021).

The purpose of this analysis is to compare the results of storm observations (SYNOP) and lightning detection data (PERUN). The spatial differentiation of the number of TDs in Poland was analysed, as was their variability throughout the year and from year to year during the period 2002–2020.

Reports on TDs from the IMGW database and PERUN network were used for this research. The study used data from 48 IMGW stations (Fig. 1B). As shown in research by Bielec-Bąkowska (2013), these data are mostly homogeneous. Unfortunately, in 2015, visual observations of atmospheric phenomena were interrupted at seven stations (Koło, Legnica, Leszno, Nowy Sącz, Płock, Sandomierz, and Tarnów), while in four stations (Olsztyn, Racibórz, Słubice, and Wieluń) there was a break in observations between 2015 and 2017. Although thunderstorms are a local phenomenon, there is a significant linear regression with nearby stations in the annual and monthly statistics of the number of TDs. The missing data from the above-mentioned stations were supplemented with data from 1966–2014 from neighbouring stations located in similar physical and geographic conditions.

The PERUN lightning detection system, introduced in Poland in 2002, was used to calculate the number of thunderstorm days. This system detects lightning flashes, which are divided into cloud-to-ground (CG) and cloud-to-cloud (CC) flashes using low-frequency electromagnetic waves. In addition to the division into the type of flashes the system allows values such as peak current estimate (kA), current charge time, and multiplicity, among other things, to be determined. The key parameter is the detection and location of the lightning flashes. The system detects flashes using an integrated measurement network consisting of, to date, nine stations in Poland: Gorzów Wielkopolski, Częstochowa, Kalisz, Toruń, Sandomierz, Warsaw, Olsztyn, Białystok, and Włodawa. In recent years, three more stations (Legnica, Chojnice and Kozienice) have been commissioned, and a station in Lesko was planned for 2022 (Gajda 2021) (Fig. 1C). Detected flashes are synchronised and saved using a GPS system. The coordinates of the detected lightning strikes are saved in decimal number format in the WorldGeodeticSystem84 coordinate system (latitude and longitude). In this analysis, to make the correct visualisation and further data calculations, it was necessary to convert the location data from the WGS84 system to the metric reference system PUWG92 (EPSG: 2180) which is applied in Poland.

The R software package (R Core Team 2014) was used to organise the lightning data. Importantly, only cloud-to-ground flashes were used for the study due to the inability of the system to correctly locate cloud-to-cloud flashes. As shown in studies by Cummins et al. (1998), some CG flashes with a charge below 10 kA can be detected by sensors as CC, so such discharges, too, were removed from the database. Importantly, in the case of CG flashes which have a negative current, these types of flashes tend to ‘strike’ several times in the same place during one flash (multistrikes). Therefore, only the first discharge was considered when processing data. According to Bodzak (2006), the location of the detection masts in various regions of the country ensures a location precision up to 1 kilometre, and detection coverage of 95% of the territory of Poland, which results in a discharge detection up to 100 km from the detector. We can therefore conclude that accuracy is best in the central part of the country. In recent years, most stations have replaced the SAFIR3000 system sensors, which have been operating since 2002, with Vaisala TLS200 sensors. In the future, IMGW plans to equip all stations with TLS200 receivers (Gajda 2021). The regular maintenance and updating of the PERUN system, along with the inclusion of three new stations in the network, improved the accuracy of the location of lightning flashes in the centre of the country from the previous 2 km to 0.5 km. And so, the data from the PERUN system is becoming ever more reliable. Analysing the data collected in the database, it can be concluded that the precision for about 88% of Poland is less than 1 km and for 12%, it is more than 1 km (Fig. 1D). This is certainly related to the most recent positioning of PERUN sensors across Poland.

This article analyses the spatial distributions and number of TDs based on SYNOP observations and PERUN atmospheric discharges. The ordinary kriging method, which is used to present discrete data, was used when interpolating the number of TDs using the data from stations. A 1×1-km grid was used to calculate the number of TDs on the basis of data from the PERUN system. A buffer with a radius of 15 km from the centre of each grid was assumed. A thunderstorm day was then counted if at least one lightning flash fell within the buffer circle. The number of thunderstorm days based on PERUN data was determined by taking the actual value from the grid cell from the exact location of the IMGW meteorological station. This provided results comparable with the SYNOP data. This method was carried out for the period 2002–2020 using the ArcPy 2.8 extension for ESRI ArcGIS software. All maps were computed in ArcGIS PRO 2.8.3. A similar methodology was used by Czernecki et al. (2016) using a radius of 17.5 km, and by Taszarek et al. (2019), who used a geographic grid with dimensions 0.5°×0.5°.

The key factor for further comparisons was the calculation of the number of days with lightning strikes in the vicinity of a station. As early as 1988, Changnon (1988) compared the observations of thunderstorms from meteorological stations in the USA with lightning strikes within a radius of 5, 10, and 20 km. Reap and Orville (1990) found that thunderstorms were detected by observers at a distance of 17 km during the day and 26 km at night. In Finland, Mäkelä et al. (2014) found that thunderstorms within a radius of 11.3 km correlated well with observations. Wapler (2013) suggested a radius of 15 km, while Enno et al. (2012) provided a figure of 14.7 km. Changnon (2001) found that the audibility of the thunder reached 8–24 km; while Koehler (2020) used a radius of 5 and 10 nautical miles (9.3 and 18.5 km) in his studies of thunderstorms in the United States.

In Poland, Czernecki et al. (2016) examined the distance at which the atmospheric discharges detected by the PERUN system were observed by humans at weather stations. On the basis of this data from stations in Poland, they obtained distances of 12 to 24 km (17.5 km on average). Also, Taszarek et al. (2019) considered that the appropriate distance for optimal determination of TDs fitted within the range of 15–18 km. This analysis compared the results from the SYNOP station against the PERUN network within a radius of 15 and 17.5 km. It was found that, taking the range of 17.5 km recommended by Czernecki et al. (2016), the average difference in recorded TDs for the 48 stations in Poland was 5.9 days and ranged from 0.8 in Łeba to 17.4 days in Rzeszów. Particularly large differences occurred at stations located in the mountains which had a limited observation horizon. Therefore, this radius was reduced to 15 km. The difference in relation to SYNOP decreased to 4.0 days, ranging from 0.4 days in Hel to 9.9 days in Sandomierz. So, the authors believe that the 15-km radius is the most reliable distance and corresponds best to human perception, and, consequently, better corresponds to the observations made at the stations.

The comparison of the number of TDs from the observations and the PERUN system showed that despite the buffer reduction to 15 km, the average difference for the 48 stations was 4.0 days (SYNOP average 25.9 days, PERUN average 29.9 days). The smallest differences occurred at stations with an open horizon, for example, at the seaside: Łeba (0.4 days), Hel (0.9 days), Świnoujście (1.9 days), and Elbląg and Terespol (1.9 days). The greatest differences occurred at stations located in mountain valleys: Zakopane (6.1 days), Lesko (7.3 days), and Rzeszów (7.9 days); and also in cities: Warsaw (6.7 days), Toruń (6.5 days), and Łódź (6.2 days). However, the biggest difference was in Sandomierz (9.9 days). This station is located outside the city and has favourable observation conditions, but these differences persist every year (Fig. 2), which may result from the subjective assessment of the occurrence or non-occurrence of thunderstorms in the vicinity of the station.

Figure 2.

The course of TDs and the correlation between the number of the TDs according to the SYNOP and PERUN data in Hel and Sandomierz during the period 2002–2020

Source: own study

There is significant linear regression between the observed TDs and their detection with the PERUN system, reaching a Pearson linear correlation coefficient of r=0.98 in the case of annual sums of TDs for stations at Gdańsk, Gorzów Wielkopolski, and Hel; and 0.97 at the Koło, Koszalin, and Kłodzko stations. The least correlated values are those at the Warsaw (0.32), Rzeszów (0.38), and Sandomierz (0.48) stations.

Thunderstorms are extreme weather phenomena. In Poland, thunderstorms most often occur along atmospheric fronts (cool and occluded) with air mass influx of sea origin from the west or north-west (Kolendowicz 2006; Sulik 2021, 2022).

Remote-sensing methods are increasingly used in the study of atmospheric phenomena, the results of which differ from traditional meteorological observations made by humans. In many stations, automatic measurements are introduced which do not ensure the detection of atmospheric phenomena. Hence, it is necessary to link the different series of observational and remote-sensing data on storm occurrence.

The frequency of TDs in Poland was analysed using parallel observational (SYNOP) and lightning discharge detection (PERUN) data from 2002–2020. For each station, the number of TDs was calculated for lightning discharges around the station in a buffer of 15-km radius. The buffer radius suggested by Czernecki et al. (2016) of 17.5 km caused excessive differences between the observed thunderstorms and the values from PERUN. Nevertheless, it was found that the observations (SYNOP) showed a lower number of TDs per year, with differences ranging from 1 to 8 days, compared to the PERUN lightning detection system. This may result from the subjectivity of observations and from significant light and noise pollution around stations located in cities, or due to the horizon being obscured (by hills, buildings, trees). The best compatibility between SYNOP and PERUN occurred for stations on the Baltic Sea coast, which was confirmed in an earlier analysis by Czernecki et al. (2016), who found that in Łeba, an observer could record discharges appearing within an average radius of 24 km. On the other hand, the greatest differences occurred in mountain valleys with limited horizons. The same authors stated that this radius was only 12 km in Bielsko-Biała in the Carpathian Mountains. In cities the conditions for observing thunderstorms are disrupted by the anthropogenic illumination of the sky.

Annual spatial distributions of TDs in Poland during 2002– 2020 obtained from SYNOP confirm the results of previous analyses (Kolendowicz 1997, 2006; Bielec 2000; Lorenc 2005; Bielec-Bąkowska 2013; Czernecki et al. 2016; Bielec-Bąkowska et al. 2021). The highest frequency of TDs was found in the south-eastern part of the country, especially in the mountains. The least frequent number of thunderstorms occurred on the coast of the Baltic Sea. Compared to the years 1951–2018, analysed by Bielec-Bąkowska et al. (2021), there were higher frequencies of TDs at some stations, for example, in Lesko, 37.8 days instead of 31.9 days; in Zakopane, 33.0 and 30.6 days, respectively; and in Gdańsk, 17.9 and 14.5 days, respectively. However, there are stations where the opposite situation occurred, for example, for Śnieżka, 20.6 days and 24.0 days, respectively. This is evidence of either climate change or the subjectivity of the lightning observations. The use of the PERUN network allowed these data to be refined, which is confirmed by a previous study by Taszarek et al. (2019) who analysed the distribution of thunderstorms in Europe. The spatial distributions of TDs according to SYNOP and PERUN were similar, but after using a 1×1 km grid, the spatial disposition of TDs was found to be more fragmented. The influence of local factors related to the orography of the area, and the presence of water bodies, forest complexes, and large agglomerations, is visible (Sulik & Kejna 2022). The obtained distribution of TDs confirms the legitimacy of separating the three storm regions in Poland, including the Baltic Sea coast, an area covering the lake districts of northern Poland, and the plains of western and south-eastern Poland (Kolendowicz 2006).

In the annual cycle, the maximum frequency of TDs in July is prominent (mean, 7.08 days for SYNOP and 7.62 days for PERUN). According to Kolendowicz (2006), the season of maximum thunder activity lasts from May 16 to August 13 on the Baltic Sea coast and until August 18 in the rest of the country. In May and the summer months, the thermodynamic instability of the atmosphere increases, favouring convective movements and the development of cumulonimbus clouds (Poręba et al. 2022). In the autumn (September–October) the frequency of TDs decreases sharply, but there is high frequency of thunderstorms along the Baltic coast. This is confirmed by the research of Taszarek et al. (2019), who found an increase in thunderstorm days during the cold half of the year in the Mediterranean coastal zone.

In many regions of the world, an increase in the frequency of extreme events, including thunderstorms, has been reported (AR6 Climate Change 2021). Studies concerning the variability in the occurrence of TDs have also been conducted in Poland. Research by Bielec-Bąkowska (2003), covering the years 1951 to 2000, did not show a statistically significant trend. More recent data from 1951 to 2018 (Bielec-Bakowska et al. 2021) distinguished western Poland as being characterised by a predominance of negative TD trends (up to 1 day per 10 years) and eastern Poland as having a predominance of stations registering an upward trend (up to 2 days per 10 years). In Kraków, as shown by a series of observations from 1901 to 2018, there was a significant variability in TDs from year to year: there were periods of increased storm activity and calmer periods. The beginning of the twenty-first century shows a downward trend (Bielec-Bakowska et al. 2021). However, in our analysis, spatially diversified trends were found. A statistically significant increase in the number of TDs was found at 14 stations for the period 2002–2020, based on observational data (SYNOP); these do not show any spatial regularity. PERUN data is more objective. They showed an increase in the number of TDs for all stations except Lublin. However, only at 12 stations was this increase statistically significant. Both methods confirm an increase in the number of TDs in many regions of the country. The reason for this may be the greater frequency of thunderstorms during the colder half of the year, especially on the Baltic coast (Bielec-Bakowska et al. 2021).

There are also varying trends in TDs and lightning flashes around the world. In a study by Lavigne et al. (2019), based on over 8,000 meteorological stations and satellite lightning detection systems (TRMM - Tropical Rainfall Measuring, LIS - Lightning Imaging Sensor), an increase in TDs was found for the Amazon, South-east Asia, India, Democratic Republic of the Congo, Central America, and Argentina, whereas, for example, China, Australia, and the Sahel exhibited decreases in the number of thunderstorm days. There was no significant TD trend in the United States during the 1993–2018 period. There were, however, periods with increased storm activity and large differences between individual parts of the country (Koehler 2020). In Europe, the mean annual number of TDs since 1979 has shown an increase over the Alps and Central, South-Eastern, and Eastern Europe, with a decrease over the south-west. (Taszarek et al. 2019). In Australia, there were also different TD trends across individual stations (Kuleshov et al. 2002).

Summing up; it should be stated that the results of the visual observations of thunderstorms and the detection of lightning flashes differ due to the perception and subjectivity of human assessment. Remote-sensing methods are objective and more accurate. Unfortunately, for now, these two storm investigation systems cannot be linked together. Also, the buffer which determines the possibility of observing the storm cannot be clearly determined; it is different for each station, and depends on the subjective opinion of the observer or the limitations of visibility related to the horizon and the light pollution of the sky. Further research in this matter is necessary to better monitor the course and spatial distribution of these phenomena.