See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/328687814

Introduction to the IEC 62858: lightning density based on Lightning Locating

Systems

Conference Paper · October 2018

CITATIONS READS

8 1,275

1 author:

Stéphane Pédeboy
Météorage
48 PUBLICATIONS 621 CITATIONS

SEE PROFILE

All content following this page was uploaded by Stéphane Pédeboy on 02 November 2018.

The user has requested enhancement of the downloaded file.

 ILPS 2018 - International Lightning Protection Symposium
October 24-27, 2018
Shenzhen – China

Introduction to the IEC 62858: lightning density

based on Lightning Locating Systems

Stéphane Pédeboy
Vice President of Technology
Météorage
Pau, France
[email protected]

Abstract— The lightning data collected by Lightning Locating isokeraunic level (Nk), based on human observations. Nowadays
Systems (LLS) operating in the Low Frequency (LF) range is of the flash data which provides a far more accurate information
great interest to compute the lightning density (Ng) in a given area about the lightning occurrence is used to determine the Ng
of interest for the lightning protection community. The parameter.
effectiveness of the recommendations for lightning protection
derived from the lightning risk analysis are sensitive to the Since the Ng is derived from a calculation based on data
reliability of the Ng parameter used in the models. To guarantee collected by remote sensing systems (e.g. LLS) and not direct
the reliability and homogeneity of this input data, the IEC 62858 measurements, the quality of the lightning data is crucial to
requires on one hand a minimum level of performance for an LLS produce reliable results. Therefore, the performance of the LLS
and on another hand recommends a computational method for the delivering the Ng is a key parameter that must be considered by
Ng aiming at guaranteeing a maximum error on the Ng less than the engineer in charge of the risk assessment analysis prior
20%. This document reviews the main points required by the IEC validating the results of the study. Furthermore, since several
62858 standard before introducing the different lightning data LLS may cover the same given area of interest, it is possible to
types collected by an LLS. More particularly, a focus is made on a use Ng coming from different operators. Here again, the
new lightning data dealing with branched flashes that can identify comparison could be almost impossible when big differences in
and locate all the Ground Strike Points (GSP) in a branched flash terms of either performances exist on both LLS. Another issue
and therefore better address the real threat on the ground.
that may affect the quality of the Ng is the way it is computed.
Keywords—Lightning Locating Systems (LLS); Lightning
Indeed, several parameters like the size of the sample of
Density (Ng/Nsg); Ground Strike Point data; IEC 62858 lightning flashes, the size of the area of concern and the period
of lightning data collection may dramatically influence the
I. INTRODUCTION result.
Lightning data collected by Lightning Locating Systems An LLS is a complex system whose performance is sensitive
(LLS) operating in the Low Frequency (LF) range has become to quite a lot of factors either related to the technology, the
an important dataset for the lightning risk assessment [1]. sensors network design or the operations [2]. Because of the
Indeed, such systems can locate the different return strokes increasing number of LLS in the world, the variety of
occurring in Cloud-to-Ground (CG) lightning flashes over wide technologies used in such systems and the different level of
areas, as large as a country up to a continent, during long period operation quality it is almost impossible for an engineer using a
of time. Out of this data, so called “stroke data”, some LLS can Ng value to fully rely on the Ng data and therefore to guarantee
identify individual CG flash using a flash grouping algorithm the lightning risk analysis results. Thus, it became necessary to
running on the lightning processor. These algorithms are usually release an international standard giving a common framework to
based on spatial-temporal criteria to identify the strokes that are LLS operators for the computation of the Ng parameter [5]. As
related to the same flash regarding the inter-stroke delay and a result, the IEC 62858 Edition 1 [6] is the first version of a
distance leading to the so-called “flash data” [2][3]. standard introducing and requiring all the necessary measures to
make reliable and homogeneous the values of Ng obtained from
The IEC 62305-2 [4] international standard is applicable to
LLS in various countries.
risk assessment for a structure due to lightning flashes to earth.
Its purpose is to provide a procedure for the evaluation of such a This paper intends to introduce the IEC 62858 Edition 2 [7]
risk based on a probabilistic approach using the lightning flash standard that is currently under maintenance with the MT16
density (Ng) parameter defined as the number of CG flashes per working group. The main points are reviewed and commented
square kilometer and per year. Decades ago, the Ng parameter for a better understanding of the complex notions addressed in
used to be derived from the number of thunderstorm days, the this standard. Then, a detailed overview of the lightning data
structure is given with the objective to explain the different Determining the detection efficiency is quite challenging since
levels of abstraction characterizing the lightning phenomenon as it is really difficult not to say impossible to account for the exact
observed by an LLS. Finally, the reader is made sensitive to a number of events occurring in nature. The minimum flash DE
new dataset that can identify and locate individual Ground Strike requiered by the standard is 80%.
Point (GSP) that may occur in branched downward flashes.
• Location Accuracy (LA), defined in meters as the
Indeed, this datum might better access the real lightning threat
in delivering a more accurate information resulting in an distance between the located return stroke and its real position
improvement of the effectiveness of the lightning protection on the ground. Again, this parameter is not easy to assess as the
systems. ground truth data used as reference must be accurately
timestamped for an efficient stroke time correlation. The
II. OVERVIEW OF THE IEC 62858 STANDARD maximum error committed on the flash location is 500 meters.
The main goal of the IEC 62858 standard is to deliver a • Classification Accuracy (CA), is a new parameter that
common framework to the LLS operators for the Ng appeared with the recent capability of LLS to detect Intra-Cloud
computation. However, the reliability and the homogeneity of (IC) discharges. It is defined as the percentage of IC or CG
the Ng values are not only depending on the computational classification by the LLS against the real class of the event. of
method as it is also extremely sensitive to the performance of the correct events classification in terms of CG or IC. The issue
LLS. One can easily understand that such a system that can only with a bad IC/CG classification is that the number of CG flashes
detect a fraction of the CG lightning flashes is not able to deliver taken into account in the Ng calculation can either be
a reliable Ng although the method used for the computation is overestimated because IC flashes are wrongly considered or on
correct. the opposite underestimated when the LLS classifiy CG in IC
As a result, the IEC 62858 defines specifications and flashes. The minimum CA requiered by the standard is 85%.
recommendations for LLS operators to a) assess the real
performances of their system and b) give a common method to The minimum requirements for detection efficiency, location
compute the Ng. The objective of the standard is to guarantee and classification accuracies are respectively 80%, 500 meters
the engineers dealing with lightning risk analysis the Ng value and 85%. These quality indicators are detailed in the Section 4
they use are not affected with an error larger than 20%.
of the IEC 62858 document that also gives an overview of the
different techniques used in LLS performance validation.
A. LLS performances and lightning data quality
B. The method recommended for the Ng calculation
There is no doubt that lightning data collected by high
resolution LLS operating in LF are of great interest to determine
In addition to the validation of the LLS performance and data
the Ng. Such systems are designed to detect and locate the return
quality the IEC 62858 standard describes the procedure that
strokes occurring in lightning flashes, that represent the real
must be used to compute the Ng. Indeed, in addition the quality
threat for human being and assets. Nowadays, the most modern
of the input data the result of the Ng computation is highly
and up-to-date systems can detect 90% and more of the return
sensitive to the size of the dataset because it influences the
strokes with a median location error in a range of 100 meters
statistical error committed on the result. The procedure
[8][9][10]. The LLS performance depends on several parameters
recommended by the standard is based on the Poisson law and
like the sensor technology, the design of the network of sensors,
the rare events theory. According to this statistical law, the
namely the sensor baseline distances, and the location
uncertainty of a result at a given confidence level depends on the
processing algorithm. However, the quality of the system
size of the sample used to compute the statistics, namely the
operation is one of the most important factors as it can
number of flashes for the Ng calculation. The choice made in the
dramatically drop down the overall quality of the lightning
IEC 62858 is to use a sample data size not smaller than 80
dataset. Thus, system and sensors failures, power and
events. With this criterion, the uncertainty in the result is about
telecommunication breakdowns, sensors and location algorithm
20% at a confidence level of 90%. Of course, the uncertainty
settings or the limited specifications of the hardware architecture
decreases as the sample size increases meaning that the biggest
(i.e. CPU, RAM, LAN…) will affect the detection efficiency.
event samples produce the more accurate statistics. Dealing with
Because it exists different LLS technologies, sensors lightning data, the period of observation and the grid cell size
network design and operation modes it may exists large used to compute and build the Ng maps are key parameters that
discrepancies in the quality of the raw data used to compute the directly affect the number of events. The standard gives the
Ng, resulting in information level homogeneity and reliability formula in (1) that must be fulfilled for the Ng calculation to
issues. To prevent such problems, the IEC 62858 defines LLS guarantee an uncertainty smaller than 20% at a confidence level
performance parameters and requires a minimum level of quality of 90% [11].
on those parameters. Generally, the performance of an LLS is
defined by the combination of at least three parameters that are
the: Ng*Tobs*Acell >= 80 (1)
where:
• Detection Efficiency (DE), defined as the percentage Ng is the ground flash density, in km-2 year-1;
of number of lightning event (stroke or flash) that are observed Tobs is the observation period, in years;
by the LLS in respect to the actual number of real events. Acell is the area of each single cell, in km2.
 Interesting to note that the period of observation not only
permits to increase the number of data but also consider the
annual variability of the lightning activity. The IEC 62858
specifies that at least 10 years of data is mandatory for the
calculation, with the most recent year not older than 5 years.

C. LLS Lightning data structure

The IEC 62858 states that the Ng must be derived from the
“flash data”. This paragraph intends to introduce the reader
with more details about the global lightning dataset and present
the different data abstraction layers that can be derived from
LLS measurements.

The collection of the return strokes detected and located by

the system is so called the “stroke data” can be considered as a
raw data since this is the less structured information delivered
by the LLS. This data comes from the measurements of the
electromagnetic signals generated by the return strokes and Fig. 1. A photography showing a forked flash where one can see two different
paths for return strokes hitting the ground in different places.
therefore the observation is fully consistent with the bnatural
phenomenon. Based on the stroke data a flash grouping Because forked flashes are not rare, some LLS operators
algorithm may be used to agregate individual return stroke in a like Météorage have developed a GSP identification algorithm
mother entity so called the flash [12]. Thus, this algorithm that intends to detect the different downward channels inside a
creates a more syntetic dataset that is called the “flash data” flash [18][19][20]. Nowadays, it exists two main algorithms,
aiming at describing the lightning flash phenomenon that is the first based on the individual stroke electrical parameters and
observed by human being. The human eye cannot distinguish the like the rise time, the peak current amplitude and the second
the individual return strokes occurring in a channel because of that uses the separation distance between strokes [19][20].
the very small interstroke delays but it rather integrates the light These algorithms are presented in the annex B of the IEC
produced by all strokes in one common discharge. In general, 62858. As an example, the figure 2 shows the results of the GSP
LLS operators all use the same algorithm which groups strokes clustering algorithm developped by Météorage using the
according to their separation distance and delay. In general, all separation distance between strokes.
the strokes located at a maximum distance of 10 km from each
other within 500 ms delay are considered to be part of the same
flash. The time difference between the last and the first stroke 1st stroke = Flash
in the same flash cannot exeed 1 s as it is specified in the Section 1st GSP
4 of the IEC 62858. It is important to note that by convention,
3rd stroke
LLS operators have decided that a flash is defined by the
characteristics (date, time, location, peak current amplitude...) 4th stroke
of its first stroke. As a result, the flash dataset is a subset of the
stroke dataset only made of the first return strokes after the flash
2nd stroke
grouping algorithm processing. Of course, this works nicely for
and 2nd GSP
lightning flashes exhibiting only one ionized channel to the GSP location
ground but a problem arise when considering the so-called Flash location
Stroke location
“branched” or “forked lightning flashes”. In most of the cases
the dart learder that prepares the occurrence of a subsequent Fig. 2. An example derived from lightning data collected by Météorage that
stroke uses the existing conducting path created by the stepped corresponds to a flash exhibiting 2 GSP and 4 strokes. One can clearly see the
leader/first stroke sequence. Sometimes the dart leader only two different clusters of strokes in the upper-left and lower-right corners of the
partly follows the preceding return stroke channel and creates image. The 1st stroke intiated the 1st GSP and by definition gives the location of
the flash. The 2nd stroke occurred in a 2nd GSP at 1.2 km from the previous one.
its own path to the ground producing a multiple GSP flash (see Then the 3rd and 4th strokes used the ionized channel created by the 1st stroke in
figure 1) [13]. In average, between a third to half the multiple the 1st GSP. The scattering of the stroke locations I the 1st GSP is due to the
strokes downward -CG flashes exhibit more than one contact to location errors committed by the LLS. The position of the GSP is computed as
the ground depending on the terrain and the season. According the average of the stroke locations.
to some reports based on high speed video observations found
As a consequence of the GSP algorithm, it is possible to
in literature, the mean separation distance between the different
obtain from a LLS three dataset with a different level of
GSP in such flashes is 1.8 km [14][15][16][17][18]. This result
abstraction. The lowest level corresponding to the raw data
is consistent with the statistics computed based on the results of
observed by the LLS is the stroke data. Then, strokes are
the GSP algorithm that will be presented hereafter [19].
classified in clusters representing the ionized channel or
downward branches in a flash leading to a GSP. The GSP laboratory to perform the evaluation of the performance of their
location is computed as the average of the position of the system. In practice, the setup of LLS validation techniques is
strokes that used the ionized channel. Finally, the flash is the rather difficult to achieve because it requires specific skills and
top level in the hierachical lightning data structure. material. This is also true for the operators themselves and only
few of them can run an operational monitoring of their system
performances. Most of the time, the sensor technology is
validated based on test networks deployed in small areas by the
manufacturer and the LLS operator use these results to prove
their performances. Of course, this is not valid for the IEC 62858
that states that the operational LLS must be validated and not
only the technology.
Finally, the IEC 62858 specifies that the flash data must be
used for the Ng calculation. This is already a good point because
so far it was not so clear that some operators delivered the Ng
based on the stroke or flash data. As a result, in the former case
the risk was likely to be overestimated because all the return
strokes using the same attachment point to the ground instead of
Fig. 3. This example shows the corresponding lightning data structure derived accounting for the first stroke only. Nevertheless, the flash data
from the real data presented in figure 2. One can see the flash, GSP and stroke gets also some limitations as it does not consider all the ground
datasets are organized in horizontal layers each one corresponfoing to a new contacts in branched flashes which are not rare in nature (about
level of abstraction.
30% to 50% of the -CG flashes). Thus, as only the first stroke of
a flash is accounted in the Ng, the risk is underestimated as the
It is interesting to note that the suggested lightning data
different GSP in a flash are not all considered. The annex A
structure presented here below can be completed with an upper stipulates that the GSP data is to be used when available because
level corresponding to the “lightning cell dataset”. This new it better reflects the real lightning threat as the considering all
object results from the grouping of all the flashes in consitent the contacts of a flash to the ground. Today, only few not to say
clusters whose countours and characteristics describe the only one LLS operator (Météorage) can deliver the GSP data in
thundercloud as observed from a LLS. agreement with the IEC 62858 requirements and therefore
compute the Nsg that is the lightning density based on GSP data.
III. DISCUSSION Because this dataset seems to be promising and several methods
The IEC 62858 is a great improvement in terms of lightning are available, the maintenance team MT16 has started a work on
data quality management for both LLS operators and engineers the comparison and validation of the GSP identification
in charge of lightning risk analysis. The former has now a algorithms [21]. When this work is done, the LLS operators can
common framework to evaluate the performances of their choose the best algorithm and compute the GSP data. However,
systems and the latter benefit from a more reliable information the annex also considers that in some areas there is either no LLS
to use in their analysis because they know the maximum error at all or the national LLS only deliver flash data. In this case, it
(20%) committed on the Ng values. However, some comments gives a clue to derive the Ng from satellite data. Of course, the
can be done on the status of the document. corresponding result suffers from large errors and this mode of
computation is to be prohibited when higher resolution lightning
First, in some areas where the lightning activity is weak, it data is available. As a last comment, the annex A specifies that
may be difficult to reach the 80 events per cell, even during a 10 it is possible to transform the Ng to Nsg using a factor of 2 [1].
years period. In this case, when it is not possible to anymore It turns out that this method overestimated the risk as in average
increase the period, the recommendation is to increase the size the number of GSP per flash varies between 1.3 and 1.7 and only
of the cell till the minimum number of required flashes is some mountainous regions can exhibit an average of 2 GSP per
reached. It must be noted that this might result in a loss of spatial flash [18].
resolution and therefore a possible underestimation of the risk.
For instance, if we consider a high tower on a flat area, it is likely
that most of the flashes occur on the tower itself and few of them
around. Expanding the area surrounding the tower to account for REFERENCES
the distant flashes in the objective of reaching the minimum of
80 events will lead to a loss of accuracy on the real Ng [1] C. Bouquegneau, A. Kern, and A. Rousseau, “Flash Density Applied to
concerning the tower only. In the worst case, despite the area Lightning Protection Standards,” in International Conference on
Grounding and Earthing and 5th International Conference on Lightning
increasing, the LLS operator cannot provide any Ng value… In Physics and Effects (GROUND/LPE), 2012.
such cases, the IEC document should consider providing the Ng
[2] “Cloud-to-Ground Lightning Parameters Derived from Lightning
result along with the corresponding uncertainty even if it is Location Systems - The Effects of System Performance”, CIGRE Report
larger than the 20%. 376, 2010.
[3] Rakov V. A.,”Lightning phenomenology and parameters important for
Second, the test and validation of a manufactured product is lightning protection”, IX SIPDA 2007, Iguaçu – Brasil
performed by a third-party laboratory authorized to certify the
[4] “International Standard IEC 62305-2 Edition 2, Protection against
product in accordance to a given standard requirements. It must lightning – Part 2: Risk management”, International Electrotechnical
be noted that today LLS operators cannot rely on such a Commission, 2010, ISBN 978-2-88912-281-3
[5] C. Bouquegneau, "The Need for an International Standard on Lightning [14] Fleenor S.A., C. J. Biagi, K.L. Cummins, E.P. Krider,”Characteristics of
Location Systems", in International Lightning Detection Conference and cloud-to-ground lightning in warm-season thunderstorms in the great
International Lightning Meteorology Conference (ILDC/ILMC), 2014. plains”, ILDC2008.
[6] “International Standard IEC 62858 Edition 1, Lightning density based on [15] Saba M. F., M. G. Ballarotti and O. Pinto Jr, 2006: Negative cloud-to-
lightning location systems (LLS) – General principles”, International ground lightning properties from high speed video observations, JGR, vol
Electrotechnical Commission, 2015, ISBN 978-2-8322-2820-3 111, D03101.
[7] “International Standard Draft IEC 62858 Edition 2, Lightning density [16] Stall C, K.L. Cummins, E.P. Krider, J. Cramer (2009), Detecting Multiple
based on lightning location systems (LLS) – General principles”, Ground Contacts in Cloud-to-Ground Lightning Flashes. Journal of
International Electrotechnical Commission, 2018 Atmospheric & Oceanic Technology, vol. 26 (11), pp. 2392-2402,
[8] A. Nag et al., “Evaluation of U.S. National Lightning Detection Network November 2009.
performance characteristics using rocket-triggered lightning data acquired [17] Thottappillil R., V. A. Rakov, M. Uman, W. Beasley, M. Master and D.
in 2004–2009,” J. Geophys. Res. Atmos., vol. 116, no. D2, p. D02123, Shelukhin, 1992: Lightning subsequent stroke electric field peak greater
2011. than the first stroke peak and multiple ground terminations, JGR, 97(D7),
[9] W. Schulz, G. Diendorfer, S. Pedeboy, and D. R. Poelman, “The European 7503-7509.
lightning location system EUCLID - Part 1: Performance analysis and [18] K. L. Cummins, “Analysis of Multiple Ground Contacts in Cloud-to-
validation,” Nat. Hazards Earth Syst. Sci., vol. 16, no. 2, pp. 595–605, ground Flashes Using LLS Data: The Impact of Complex Terrain,” in
2016. International Lightning Detection Conference and International Lightning
[10] W. Schulz, S. Pedeboy, C. Vergeiner, E. Defer, and W. Rison, “Validation Meteorology Conference (ILDC/ILMC), 2012
of the EUCLID LLS during HyMeX SOP1,” in International Lightning [19] S. Pedeboy, “Identification of the multiple ground contacts flashes with
Detection Conference and International Lightning Meteorology lightning location systems,” in 22nd International Lightning Detection
Conference (ILDC/ILMC), 2014. Conference and 4th International Lightning Meteorology Conference
[11] G. Diendorfer, “Some Comments on the Achievable Accuracy of Local (ILDC/ILMC), 2012.
Ground Flash Density Values,” in Lightning Protection (ICLP), 2008 [20] L. Z. S. Campos, K. L. Cummins, and O. J. Pinto, “An algorithm for
International Conference on, 2008, no. June, pp. 1–6. identifying ground strike points from return stroke data provided by
[12] Cummins, K. L., M. J. Murphy, E. A. Bardo, W. L. Hiscox, R. B. Pyle, Lightning Location Systems,” in Asia-Pacific Conference on Lightning
and A. E. Pifer (1998), A Combined TOA/MDF Technology Upgrade of (APL), 2015, pp. 475–478
the U.S. National Lightning Detection Network, J. Geophys. [21] S. Pédeboy and W. Schulz, “Validation of a ground strike point
Res., 103(D8), 9035–9044, doi:10.1029/98JD00153. identification algorithm based on ground truth data,” in International
[13] L. Z. S. Campos, “On the mechanisms that lead to multiple ground Lightning Detection Conference and International Lightning Meteorology
contacts in lightning,” PhD Thesis, INPE, Brazil, 2016. Conference (ILDC/ILMC), 2014.
[22] W. Schulz, S. Pedeboy, and M. M. F. Saba, “LLS Detection Efficiency of
Ground Strike Points,” in Lightning Protection (ICLP), 2014 International
Conference on, 2014, no. 2, pp. 381–384.

View publication stats