Austrian Electrotechnical Association

Forecasting lightning is critical due to its significance for public safety, infrastructure resilience, and disaster response. Traditional numerical weather models are computationally intensive and typically designed for long-term forecasting. In contrast, machine learning models enable faster predictions, but require training on large datasets.The use of Machine Learning (ML) methods for lightning forecasting has shown promising results. They can be used for a shorter predictive horizon, in the range of 15 to 60 minutes, but also present some limitations as they require large amounts of historical data for training. Additionally, they can be difficult to interpret, making it challenging to understand the underlying mechanisms driving the predictions. This paper presents a new approach to lightning forecasting that addresses such limitations. The proposed approach identifies key lightning activity patterns through the use of Clustering and Support Vector Machines, bypassing the need for training on historical data. The proposed model was applied to satellite-based lightning data, and its performance was found to be comparable to that of machine learning models, while being computationally more efficient.

<strong class="journal-contentHeaderColor">Abstract.</strong> Lightning location system (LLS) sensors, which detect and locate atmospheric discharges, are typically powered by cables buried up to one meter underground. Within the LLS community, it is well known that these cables can create spurious magnetic fields, which can in turn adversely impact the sensor measurements and the resulting data. This issue arises from currents induced in the cable shield by the lightning electromagnetic fields that penetrate the ground. The magnetic field generated by these currents lead to "site errors," causing inaccuracies in estimating the angle of incidence and the peak current of lightning strokes. Although these sensor-specific errors can be partially corrected, a better understanding of the coupling mechanism between the lightning electromagnetic field and the cable could help in minimizing the site errors. This study presents an analysis of the lightning electromagnetic field interaction with cables and examines the influence of various ground and cable properties on this interaction. This work represents a first step toward understanding the physical mechanism leading to LLS sensor site errors. Considering simplified scenarios involving a single insulated or bare conductor, this work provides practical insights that LLS operators can use to estimate worst-case site errors for a provisioned sensor site. Additionally, we show that some site errors observed in operational sensors can be successfully reproduced with good agreement using the proposed approach.

<strong class="journal-contentHeaderColor">Abstract.</strong> Evaluating the risk of lightning strikes to a particular structure typically involves adhering to the guidance outlined in IEC 62305-2. Among the multitude of factors influencing the overall risk, flash density emerges as a crucial parameter. According to its definition, each flash is assigned only one contact point to ground. Nevertheless, it is well known that, on average, flashes exhibit multiple ground termination points as shown by high-speed camera observations. In this research, lightning data collected by the European Cooperation for Lightning Detection (EUCLID) network is utilized in combination with a ground strike point (GSP) algorithm that aggregates individual strokes within a flash into ground strike points. This approach enables the examination of spatio-temporal patterns of GSPs across Europe throughout a decade, spanning from 2013 to 2022. Average GSP densities exhibit variations across the European continent, mirroring the observed patterns in flash densities. The highest densities are concentrated along the Adriatic Sea and the western Balkan region, reaching peak values of up to 8.5 GSPs km^-2 yr^-1. The spatial distribution of the mean number of ground strike points per flash reveals a noticeable increase in the Mediterranean, Adriatic, and Baltic Sea regions compared to inland areas. Moreover, it has been determined that the average number of GSPs per flash reaches its peak between September and November. Additionally, a daily pattern is discernible, with the lowest number of GSPs per flash occurring between 12 and 18 UTC (Universal Time Coordinated). It is found that 95 % of the separation distances between distinct GSPs are less than 6.7 km. Lastly, it is worth noting that the presence of the Alps has an impact on GSP behaviour, resulting in lower GSP counts in comparison to the surrounding areas, along with the shortest average distances between different GSPs.

In the early days, the risk of lightning strikes was described by the average number of thunderstorm days or thunderstorm hours, where a thunderstorm day is defined as an "Observational day during which thunder is heard at the station" [1]. On the basis of long-term records of thunderstorm days by the meteorological services, maps showing the so-called isoceraunic level were produced for the individual countries, from which the regional thunderstorm hazard could be obtained [2]. The first attempts to locate lightning discharges date back to the 1920s. W. Watt in [3] describes the considerations and experimental observations made at that time, which made it possible to identify thunderstorms and lightning discharges as the main cause of the observed electromagnetic disturbances in the early days of long-range radio communication.In the beginning, the problem of electromagnetic disturbances in radio communication was the driving force for research activities related to lightning electromagnetic fields and their source location. Only since the middle of the 1980s lightning detection has been applied in the field of thunderstorm observation for meteorological applications, in the field of lightning risk assessment and in fault analysis for power utilities. These new applications also placed significantly higher demands on the performance of the lightning location systems (LLS), especially with regard to detection efficiency and location accuracy. Lightning detection has thus evolved from a real-time thunderstorm observation tool to the most precise detection of each individual electrical discharge in lightning both in cloud-to-ground (CG) flashes and discharges within the thundercloud, often referred to as intracloud lightning (IC).In a typical thunderstorm, the number of IC discharges exceeds the number of CG discharges many times with an average ratio of IC/CG discharges being in the range of 5-10, but this ratio can be much higher in individual thunderstorms. As CG lightning poses the greatest danger to humans and property, at the beginning of the development of LLS, the focus was on the best possible and reliable detection of CG lightning. Only in recent years, the trend is increasingly moving towards the detection of total lightning. New lightning data have recently become available through the detection of lightning discharges by the Geostationary Lightning Mapper (GLM) on the satellites GOES West and GOES East [4]. Compared to land-based LLS, optical detection from space has the advantage that lightning activity is monitored over continents and oceans with approximately the same detection quality. The main limitation of lightning detection from space is the lack of differentiation between CG and IC discharges and the comparatively limited detection accuracy due to the spatial resolution of the optical sensor in the range of 4-5 km looking to earth surface from a distance of about 36,000 km. In this chapter, we will focus on ground-based systems only which are employing Magnetic Direction Finding (MDF) and/or Time of Arrival (TOA) technique, as data from these systems are used for many applications from lightning risk management to severe storm forecast.

Three distinct data sets are analyzed in terms of probability of flashes having subsequent strokes with higher peak currents than the first stroke in the flash. This type of flashes may have some implication in the shielding failure rate of transmission lines. For three distinct data sets we obtained about 30 % of flashes with a ratio I <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">sub</inf> /I <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</inf> >1. For about 10 % of the flashes the peak current of a subsequent stroke is more the two times the peak of the first stroke, and for about 3 to 4 % of the flashes we obtained a ratio I <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">sub</inf> /I <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">l</inf> is grater then 3.

Abstract. Two long-lasting Thunderstorm Ground Enhancement (TGE) events were registered at the Milešovka meteorological observatory in Czechia (50.55N, 13.93E, altitude 837 m) on 23 April 2018, during linearly organized thunderstorms. Two intervals of increased photon counts were detected by a plastic scintillator, respectively lasting 70 and 25 minutes, and reaching 31 % and 48 % above the background radiation levels. Using numerical simulations, we verified that the observed increases of count rates are consistent with the energy spectrum of previously observed TGEs. We investigated the relevant data from a suite of meteorological instruments, a Ka-band cloud radar, an electric field mill, and a broadband electromagnetic receiver, all placed at the Milešovka observatory, in order to analyze the context in which these unique continental TGEs occurred at an exceptionally low altitude. The onset of the TGEs preceded the onset of precipitation by 10 and 3 minutes, respectively, for the two events. Both this delayed rain arrival and a lower energy threshold of 6.5 MeV for registered particles clearly exclude the detection the decay products of the radon progeny washout during the TGE intervals. At the same time, the European lightning detection network EUCLID detected numerous predominantly negative intracloud lightning discharges at distances closer than 5 km from the particle detector, while the occurrence of cloud-to-ground discharges was suppressed. The cloud radar recorded presence of graupel below the melting level and the composition of hydrometeors suggested good conditions for cloud electrification. The observed variations of the near surface electric field were unusual, with very brief negative electric field excursions reaching -20 kV in a quick succession. At the same time, sub-microsecond unipolar pulses emitted by close corona discharges saturated the broadband magnetic loop antenna. All these measurements indicate that a strong lower positive charge region was present inside the thundercloud. The bottom thundercloud dipole was probably responsible for acceleration of the seed electrons in the air. These seed electrons might originate not only in the secondary cosmic ray particles but could also come from a high concentration of radon in the air collected during the propagation of the convective system above the uranium-rich soils before the thunderstorms overpassed the Milešovka observatory.

Lightning causes significant damage and casualties globally by directly striking humans and livestock, by igniting forest fires, and by inducing electrical surges in electronic infrastructure, airplanes, rockets, etc. Monitoring the evolution of thunderstorms by tracking lightning events using lightning locating systems can help prepare for and mitigate these disasters. In this work, we propose to use Benford’s law to assess the quality of the data provided by lightning locating systems. The Jensen–Shannon and Wasserstein distances between the recorded data distribution and Benford’s distribution are used as metrics for measuring the performance of the lightning locating systems. The data are provided by the European lightning detection network (EUCLID) for the years from 2000 to 2020. The two decades consist of three time windows between which the lightning locating system underwent several upgrades to improve the detection of both positive and negative strokes. The analysis shows that the agreement with Benford’s law is consistent with the expected behavior caused by the applied upgrades to the system throughout the years. The study suggests that the proposed approach can be used to test the success of software and hardware upgrades and to monitor the performance of lightning locating systems.

Abstract. Information about lightning properties is important in order to advance the current understanding of lightning, whereby the characteristics of ground strike points (GSPs) are in particular helpful to improving the risk estimation for lightning protection. Lightning properties of a total of 1174 negative downward lightning flashes are analyzed. The high-speed video recordings are taken in different regions, including Austria, Brazil, South Africa and the USA, and are analyzed in terms of flash multiplicity, duration, interstroke intervals and ground strike point properties. According to our knowledge this is the first simultaneous analysis of GSP properties in different regions of the world applying a common methodology. Although the results vary among the data sets, the analysis reveals that a third of the flashes are single-stroke events, while the overall mean number of strokes per flash equals 3.67. From the video imagery an average of 1.56 GSPs per flash is derived, with about 60 % of the multiple-stroke flashes striking the ground in more than one place. It follows that a ground contact point is struck 2.35 times on average. Multiple-stroke flashes last on average 371 ms, whereas the geometric mean (GM) interstroke interval value preceding strokes producing a new GSP is about 18 % greater than the GM value preceding subsequent strokes following a pre-existing lightning channel. In addition, a positive correlation between the duration and multiplicity of the flash is presented. The characteristics of the subset of flashes exhibiting multiple GSPs is further examined. It follows that strokes with a stroke order of 2 create a new GSP in 60 % of the cases, while this percentage quickly drops for higher-order strokes. Further, the possibility of forming a new lightning channel to ground in terms of the number of strokes that conditioned the previous lightning channel shows that approximately 88 % developed after the occurrence of only one stroke. Investigating the time intervals in the other 12 % of the cases when two or more strokes re-used the previous lightning channel showed that the average interstroke time interval preceding a new lightning channel is found to be more than twice the time difference between strokes that follow the previous lightning channel.