- Abstract -
This paper presents a performance analysis of the European lightning detection network (EUCLID) with respect to warnings of cloud-to-ground lightning in western Europe.
These warnings allow to prevent accidents due to thunderstorms electrical activity, and can be employed in several domains such as industry, utility networks, leisure activities, transport, civil protection, …
In these sectors, a reliable and efficient warning is considered as vital, according to the risk of human and assets losses.
Based on a standard configuration, we have evaluated the EUCLID warning system’s efficiency and obtained some convincing results with 96% of probability of detection (POD) and with a 20 minutes’ lead time in more than 80% of the cases. Upon conclusion of the study, these results were compared to some previous studies which had evaluated the electric field mills’ efficiency.
- Abstract -
Severe storms exhibit a common pattern consisting in a rapid increase of the total lightning rate (i.e. Cloud-to-Ground and Cloud-to-Cloud flashes) few to tenths of minutes in advance to heavy precipitation, hail or tornado. This “lightning jump” is an interesting feature for now-casting weather applications since it can help predicting severe weather occurrence with a sufficient lead time in most cases [Williams et al, 1999; Murphy and Demetriades 2005; Schultz et al, 2009].
Several algorithms have been developed to monitor lightning rate trends and detect the onset of the lightning jump on the basis of VFH lightning data [Gatlin and Goodman 2010]. Out of those algorithms, the “2σ configuration” has been statically validated on various thunderstorm types and is likely to be the most effective to use for operational usage [Schultz et al. 2014].
Météorage has design and developed a cell identification method using the DBSCAN algorithm [Ester et al. 1996] to cluster VLF/LF total lightning data consisting in Cloud-to-Ground and Cloud-toCloud flashes collected by the French National Lightning Locating System. In this algorithm so called STORM, every individual cell is then tracked and its characteristics (eg. position, direction of propagation, speed, area and number of flashes) are monitored all long the lifecycle. The analysis of the evolution of the total lightning flash rate by the “2σ configuration” lightning jump algorithm helps predicting severe weather occurrences and triggering warning messages.
This study aimed at determining the overall performances of STORM by comparing computed lightning cell and severe weather alerts against ground truth hail observations. This dataset consists of 248 valid hail reports collect in 2014 across France by the ANELFA, the national association for hail risk prevention [Dessens et al 2006]. Preliminary results show a clear seasonal dependency since winter storms are less likely to be detected by STORM because they produce few lightning. However, they are encouraging since a Probability of Detection of 80% is obtained for severe hailstorms producing hailstones with a diameter equal to or greater than 2.5cm. In addition, the mean Warning Lead Time is found to be about 15 min and reach 18 min for severe thunderstorms. Those results being consistent with those from similar studies [Schultz et al. 2009] it turns out the usage of VLF/LF lightning data are relevant for severe storms tracking and alerts.
Further work shall be carried out to optimize STORM settings in order the cell identification algorithm is improved. Comparison of VLF/LF lightning cells with radar and Lightning Mapping Array should help in tuning the overall performances and better understanding strengths and weaknesses of such a tool.
- Abstract -
The location accuracy is one of the important parameters characterizing the performance of a lightning location system. It is also one of the most difficult to determine as the actual location of the discharge being located must be accurately known to achieve a reliable assessment of the real error. Among all the measurement techniques which can be used to collect such ground truth data, none can cover large area preventing the estimation of the location accuracy at a regional or national scale. Trying to get around this limitation, Météorage has developed a method based on lightning ground strike point data collected by the French national lightning locating system computing the separation distances of return strokes identified as using the same attachment point on the ground. As a result, statistics on the relative location accuracy over the last 10 years of operation at the national scale are produced. In order to determine whether this data could be a proxy for the absolute location accuracy they are compared against systematic errors estimated in the vicinity of high elevation towers well known to attract or trigger lightning. If the study shows some discrepancies between relative and absolute errors at the beginning of the period, mainly due to technological upgrades in the system, it turns out both parameters fit nicely since 2010. This tending to demonstrate the relative errors estimated based on the ground strike point can be used as a good proxy for the absolute location errors estimate.
- Abstract -
Lightning detection provides always higher performance in R&D context. But the applications of lightning data in Meteorology and industrial safety requires a better understanding of the quality of the measures provided by detection networks. We will highlight the key drivers for this quality and how they can be assessed.
- Abstract -
In the lightning protection domain, the vulnerability of structures to lightning is commonly estimated by using the rolling sphere method (RSM). Moreover, this method is recommended in the International Standard IEC 62305-3. This is an electrogeometric model (EGM) which consists in rolling over the structure an imaginary sphere the radius of which depends on the estimated peak current of the lightning flash return stroke. The sphere center is considered as the location of the negative downward leader tip which propagates vertically to the ground. Thus all the surface contact points are considered to require protection, whilst the unaffected volumes are assumed to be protected. However, the major drawback is that this model does not take into account some aspects like the ground influence (ground conductivity and local field reinforcement), the upward leaders development and other climatic and geographic parameters.
In order to evaluate the lightning impact probability over a structure, we propose in this paper a 3D method based on the electrogeometric model application by mainly taking into account the upward leaders development. The structure profile is enhanced at locations where the reinforcement of the electrical field at the ground is the most important. Then, applying the rolling sphere method on this new profile, the surface generated by the different trajectories of the center of the modified rolling sphere is deduced and lightning impact probabilities are evaluated.
This new approach has been applied to the case of the observatory of the Pic du Midi de Bigorre in south of France where a lightning station is installed. So the numerical results are compared to the observations on site, to any experimental measurements and to the lightning detection network data.
- Abstract -
Orange and Meteorage have unified their skills to set up an experiment related to the lightning storms. It consists in measuring the current values when direct strokes hit the air termination system on two radio telecommunication towers and in addition to check the way it is distributed within the structure.
For Orange the results will be precious to contribute to the ITUT Standardization in order to optimize the engineering rules for the radio communication towers electromagnetic protection. For Meteorage the interest is to strengthen and validate the algorithms of the stormy activity geo-localization, knowing that for a given lightning stroke Orange has this information. The two experimental sites are located in the Rhône-Durance (France) area. Each one is equipped with a field mill to register the electrostatic field in an area of 15km around the site, 24h a day, and with a digital oscilloscope to which are connected four current transformers dedicated to the direct lightning stroke current, to the current flowing through the surge arrestors, to the electric line and to the grounding system. The control of both instrumentations is performed through the 3G network. The paper focuses on the main objectives of this experiment, the difficulties encountered for the installation of the measuring equipment and the solution which have been implemented to fix them. Then the outcomes of the experiment will be discussed with regard to the ongoing work in standardization.
- Abstract -
Lightning locating systems (LLS) can provide flash data derived from individual return stroke based on a flash grouping algorithm. However the latter considers negative cloudto-ground (CG) flashes striking the ground in a unique point represented by the location of the first return stroke. According to video observations flashes have often different ground strike points. This can be a limitation in some engineering applications like the lightning risk assessment where the actual number of ground contacts is an important parameter. To get around this limitation Météorage has developed an algorithm allowing the identification of the location of the ground strike points (GSP) based on a statistical clustering (‘k-means’) method. The effectiveness of this algorithm, using operational LLS data, is tested on a total of 227 negative CG flashes observed with high speed video cameras in Austria and in France, in 2012 and 2013 respectively. The comparison between GSP computation and video observations reveals a GSP detection efficiency (DE) of about 95%. In addition the algorithm is able to discriminate between strokes creating a new ground contact (NGC) or using a pre-existing channel (PEC) in 83% out of the 767 observed strokes. The analysis shows that the limitation of the model is highly depending on the DE and location accuracy (LA) of the LLS collecting the data. Nevertheless, the fairly good results obtained with the GSP identification algorithm permits to build from existing VLF/LF LLS lightning data a hierarchical interlocked data structure composed of chronological events starting with the flash as the root event which is composed of GSPs being containing themselves strokes. This new dataset describes in a more complete way some lightning parameters related to a flash (e.g. flash multiplicity and number of ground strike points per flash) and their individual relationship, giving room to the improvement of engineering and research applications.
- Abstract -
This paper deals with performance evaluation of the European lightning location system EUCLID in France during the HyMeX  Special Observation Period 1 (SOP1) in 2012. Beside other instruments a Lightning Mapping Array (HyLMA) and a mobile Video and Field Recording System (VFRS) was deployed in the south of France. The data of those independent systems are used to determine the performance of the EUCLID lightning location system (LLS) in terms of detection efficiency (DE) and location accuracy (LA) for both CG and IC flashes.
Based on VFRS records of 161 flashes, we determined a negative/positive flash DE of 90/87% and negative/positive strokes DE of 87/84%. The negative flash DE is quite low compared to the usual performance noticed on similar LLS because during two days of the VFRS measurements, where a significant amount of the negative flashes was recorded, a nearby sensor was out of order. The positive flash DE is low because the criteria to determine if a stroke was correctly detected by the operational LLS are quite strict. In fact only one positive stroke out of 56 strokes was not detected by the LLS.
The HyLMA data gave us the opportunity to objectively determine for the first time an intra-cloud (IC) DE for a network in Europe. We analyzed the IC-DE for so called isolated ICs. Isolated ICs are IC discharges which are not related to any cloud to ground stroke. For one isolated storm we found a surprisingly good IC-DE of 47%. Unfortunately we also realized that the LA of the IC discharges is not as good as for CG strokes.
- Abstract -
This paper is intended to be an introduction to the capabilities of lightning detection networks and to provide guidelines to understand the type and quality of data that can be expected from those measurement systems. It first introduces the principle of lightning detection then covers the performances and limitations of the various technologies in use. A review of the knowledge on the lightning phenomenon gained through the analysis of the data provided by those networks in then presented. Typical values for lightning parameters are finally reviewed.
- Abstract -
Between May and October 2013 (period of sustained thunderstorm activity in France), several cloud-to-ground lightning flashes have been observed in Paris area with a high-speed video camera (14000 frames per second). The localization and the polarity of the recorded cloud-to-ground flashes have been obtained from the French lightning detection network Météorage which is equipped with the same low frequency sensors used by the US NLDN. In this paper we focused on 7 events (3 positive cloud-to-ground lightning flashes and 4 negative cloud-to-ground lightning flashes). The propagation velocity of the leaders and its temporal evolution have been estimated; the evolution of branching of the negative leaders have been observed during the propagation of the channel which get connected to ground and initiate the first return stroke. One aim of this preliminary study is to emphasize the differences between the characteristics of the positive and of the negative leaders.
- Abstract -
In this paper, we analyze the correlation between altitude and lightning stroke density in the area around Pic du Midi in France. The lightning stroke density was found to increase linearly over the altitude range 600-2700 m in the Pyrenees and remained constant in the plains. The relation between the lightning stroke amplitude and the altitude was also analyzed. The mean and maximum values of lightning return stroke peak currents observed over the altitude range 0-3000 m were found to remain constant within the area of study.
- Abstract -
In this paper we present a performance analysis of the European lightning location system EUCLID in terms of location accuracy (LA), detection efficiency (DE) and peak current estimation. The performance analysis is based on ground truth data from direct lightning current measurements at the Gaisberg Tower (GBT) and data from E-field and video recordings. The E-field and video recordings were taken in three different regions in Europe, in Austria, in Belgium and in France. The analysis shows a significant improvement of the LA over the past seven years. Currently the median LA is in the range of 100 m. The observed DE in Austria and Belgium is similar yet a slightly lower value is found in France because during the measurement period in France a nearby lightning location sensor was out of order. The accuracy of the lightning location system (LLS) peak current estimation for subsequent strokes is reasonable keeping in mind that the LLS estimated peak currents are determined from the radiated electromagnetic fields assuming a constant return stroke speed.
The results presented in this paper can be used to estimate the performance of the EUCLID network for regions with similar sensor baseline and sensor technology.
- Abstract -
Most of lightning statistics used by the lightning protection community are derived from lightning locating systems. Such systems collect flash data over large regions and during long period of time. In their life, those systems are likely to undergo several major changes as the state of the art in remote sensing techniques improves leading to some impacts on the data. In addition the flash data measured by lightning locating systems underestimate the actual lightning risk since it accounts only for one ground contact per flash that is well known nowadays not correct. Then the different evolutions in observing systems and underestimation introduced by flash data must be addressed and compensated in order to give a more accurate and relevant information for lightning risk assessment.
The French national lightning locating system operated by Météorage, has collected more than 20 years of lightning data all over the country. This system is no exception since from its inception several major changes, in either technology or system settings, have significantly modified the lightning detection performances affecting the homogeneity of the data and the relevancy of the GFD statistics.
The work presented in this paper is based on this long duration French dataset. It attempts to define a method which comes around the inhomogeneity introduced by lightning detection performance evolutions and suggests the use of the flash ground contacts multiplicity instead of flash data only. To achieve this goal the cumulative peak current distribution method developed by the CIGRE task force C404 is used to determine the compensation factors to correct the statistics for detection efficiency effect. In addition, it is suggested the use of ground contacts data instead of flash data for lightning risk statistics. This parameter is derived from the lightning data collected in France on 2011 with a program developed by Météorage based on a clustering algorithm so called ‘k-means’.
The method suffers from some necessary assumptions depending on the Météorage’s LLS history and operational background, but the final new ground contacts density parameter derived from the longest observation period available in France looks more realistic and reliable. This work is a first attempt that must be extended in the future to compute high spatial resolution statistics supporting new applications for lightning risk assessment.
- Abstract -
The lightning protection standards recommend the use of the lightning ground flash density (GFD) for the risk assessment. Nowadays the national lightning location systems (LLS) can provide the lightning protection community with high spatial resolution GFD data. However these systems generally underestimate the GFD since the multiple ground contacts in flashes are not discriminated and then not taken into account in the flash density calculation.
The recent improvement in the location accuracy of the French National LLS, so called Meteorage, in 2011 was estimated with the pre-existing channel method developed by Cummins et al (2010). The global location accuracy was found to be better than 150m in accordance with Honna et al (2011). Therefore it seems possible to identify the different ground contacts in flashes based on strokes separation distances, since several video records studies showed the distance between ground contacts in flashes ranges from 0.3 to 7.3 km (Thottappillil et al (1992), Stall et al (2009)).
A statistical clustering method based on the so called “k-means analysis”, was developed to spatially partition the closest subsequent strokes of a flash in one or several clusters whose centroïds spot the ground contacts. To check the efficiency of the clustering method, a qualitative analysis was run on two LLS flash dataset previously correlated with video records in USA (Az) and Austria. For each individual flash, the number of ground contacts given by the clustering analysis was compared to the corresponding flash video observations as truth reference: the correlation rate for flashes having the same number of ground contacts range from 72% to 82% depending on the LLS stroke locations accuracy.
This encouraging result shows the capacity of the clustering method to identify ground contacts information from the Météorage’s flash data observations. A dataset of 8081 cloud-to-ground negative lightning flashes collected by Météorage during the summer in 2011 was analyzed with this method, which shows 47% of negative flashes are made of two or more ground contacts with an average number of contacts per flash around 1.74. The median separation distance between contacts is found to be 2.2 km. All these results are in good agreement with several video observations analysis (Rakov et al (1990), Saba et al (2006)).
Finally, the identification of ground contacts in flashes provided a way to estimate the relative stroke location accuracy in a better way than the PEC method since it produces a larger strokes dataset. The previous results could be confirmed as the maximum relative location error is about 110 meters.
- Abstract -
An experimental site devoted to the study of direct lightning impacts to lightning rods has been selected at the top of the Pic du Midi in the French Pyrénées. This site offers the exceptional opportunity of investigating lightning strokes to a complex structure of limited extension, intermediate between a large building at moderate altitude and a high-elevated structure such like the well-known instrumented telecommunication towers (Gaisberg, Peissenberg, Säntis or CN Tower). At the Pic du Midi, the famous astronomical Observatory offers an unique scientific environment for lightning observations (sprites, Météorage) in collaboration with astronomers.
Among various experiments, we will devote this paper to that installed at the top of the so-called DIMM platform. Several scientific sensors are here installed, such like measurements of lightning current, electric field, highspeed video and related recordings. This will enable to characterize both lightning attachment to an instrumented rod and lightning detection network efficiency at high altitude (about 2,900 m).
- Abstract -
This paper presents existing lightning data services that can be use as part of a lightning prevention approach and the result of a study of the use of lightning data services by industrial users. The study takes into account the reasons for choosing a lightning prevention approach, the benefits of this approach and its success factors.
- Abstract - (French)
La première mesure du foudroiement sur un territoire a été le niveau kéraunique, « nombre de jour où l’on a entendu le tonnerre », relevé par des observateurs humains. L’apparition des réseaux de détection foudre, à partir des années 70, a permis d’obtenir une information continue sur la localisation et les caractéristiques des coups de foudre. Les informations fournies par ces réseaux permettent de calculer le nombre de jour où un orage a frappé un lieu donné. D’autres informations peuvent aussi être calculées à partir de ces réseaux comme la densité de foudroiement (Nombre d’impact par an et par km²) qui permet de comparer la violence des orages en divers points, ou le calcul de l’intensité électrique de chaque impact. Cette étude présentera ces informations, l’évolution de leur précision, et l’utilisation qui en est faite.
- Abstract -
It is well known that a lightning detection network is able to record lightning activity. Among such networks, those using IMPACT technology are supposed to detect about 90 to 95% of cloud-toground flashes with peak current greater than 15 kA, with an average location accuracy of 500 m. One important question is to understand why some of the flashes are apparently not detected and to make clear how does the network groups into flashes the detected and located strokes.
A long duration thunderstorm has been studied by means of video photography and results are compared to data collected by the French detection and location network, the so called « Météorage network». This will allow us to evaluate what is the flash detection efficiency of the network and what are the limitations of direct observations.
In practice, every cloud-to-ground flash has been detected by at least one sensor. However, as we need at least two sensors to compute the flash location, such CG flash are not taken into account in data provided by the network. Moreover, data exhibits CG flashes with multiple polarity restrikes and sometimes the same observed CG flash appears several times in data. At last, it appears that some CG flashes are detected by a very large number of sensors although their peak intensity are lower than 30 kA. It is one of the goals of the present paper to deal with these observations.
Discussion about the total flash duration as compared with apparently observed flash lifetime will be opened. In some rare cases, estimation of the lightning location accuracy will be given.
- Abstract - (French)
La détection des orages est devenue au cours des dernières années un enjeu considérable pour l'ensemble des pays industrialisés. Ainsi de nombreux systèmes ont été développés à des fins d'étude, de prévision et de protection. Bien que diverses techniques existent pour détecter les orages, les systèmes de localisation des éclairs sont les plus utilisés dans le monde. Ils reposent sur la détection des signaux électromagnétiques générés par les éclairs. Les progrès réalisés dans le domaine spatial permettent d'envisager sérieusement de surveiller les orages à grande échelle depuis l'espace.
Ce document présente les différentes techniques mises en œuvre dans les systèmes de détection des orages.