8 Electromagnetic Methods of Lightning Location

8 Electromagnetic methods of lightning location
8.1 Introduction
There are many individual physical processes in cloud and ground lightning flashes. Each
of these processes has associated with it electric and magnetic fields. Lightning is known to
emit significant electromagnetic energy in the frequency range from below 1 Hz to near
300 MHz, with a peak in the frequency spectrum near 5–10 kHz for lightning at distances
beyond 50 km or so. Further, electromagnetic radiation from lightning is detectable at even
higher frequencies, for example, in the microwave, 300 MHz to 300 GHz, and, obviously,
in visible light, roughly 1014–1015 Hz. At frequencies higher than those of the spectrum
peak, the field spectral amplitude varies roughly inversely proportionally to the frequency
up to 10 GHz or so (Pierce, 1977). Also, lightning is known to produce X-rays (up to
1020 Hz or more), although, at ground level, they are usually not detectable beyond a
kilometer or so from the source. In general, any observable electromagnetic signal from a
lightning source can be used to detect and locate the lightning process that produced it. Only
radio-frequency methods are considered in detail here.
8.2 Principles of lightning location
8.2.1 General
The three most common electromagnetic radio-frequency-locating techniques include the

magnetic direction finding (MDF), time of arrival (TOA), and interferometry. For each of
these techniques, the type of locating information obtained depends on the frequency f (or
equivalently on the wavelength λ = c/f, where c is the speed of light) of the radiation
detected (Rakov and Uman, 2003, Ch. 17). For detected signals whose wavelengths
are very short compared to the length of a radiating lightning channel, for example, the
very-high-frequency (VHF) range where f = 30–300 MHz and λ = 10–1 m, the whole
lightning channel can, in principle, be imaged in three dimensions. For wavelengths that
exceed or are a significant fraction of the lightning channel length, for example, the
very-low-frequency (VLF) range where f = 3–30 kHz and λ = 100–10 km and the
161
Downloaded from https:/www.cambridge.org/core. UCL, Institute of Education, on 11 Jan 2017 at 02:24:51, subject to the Cambridge Core terms of use, available at
https:/www.cambridge.org/core/terms. http://dx.doi.org/10.1017/CBO9781139680370.009
low-frequency (LF) range where f = 30–300 kHz and λ = 10–1 km, generally, only a
small number of locations can be usefully obtained. In the case of single location for a
cloud-to-ground discharge it is usually interpreted as some approximation to the ground
strike point. The mechanisms of radiation in the high-frequency (HF) region of the
spectrum and above are not fully understood. It is thought that this radiation (particularly
VHF, 30–300 MHz) is caused by numerous small sparks occurring during the formation of
new channels; that is, by the electrical breakdown of air rather than by high-current pulses
propagating in preexisting channels.
Accurate lightning locating systems, whether they image the whole lightning channel or
locate only the ground strike points or the cloud charge centers, necessarily employ multi-
ple sensors. Single-station ground-based sensors, such as the lightning flash counters,
detect the occurrence of lightning, but cannot be used to locate it on an individual flash
basis; nor are they designed to do so, because of the wide range of amplitudes and
waveshapes associated with individual events. Nevertheless, with single-station sensors,
one can assign groups of flashes to rough distance ranges if data are accumulated and
“averaged” for some period of time. There are many relatively simple commercially
available single-station devices that purport to locate lightning. Most operate like AM
radios with the amplitude of the radio static being used to gauge the distance to the
individual lightning flashes, a technique inherently characterized by large errors.
Single-station optical sensors on Earth-orbiting satellites detect the light scattered by the
volume of cloud that produces the lightning and hence cannot locate to an accuracy better than
about 10 km, about the diameter of a small cloud. Additionally, satellite-based sensors cannot
distinguish between cloud and ground discharges. The next-generation series of Geostationary
Operational Environmental Satellite (GOES-R) is planned to carry a Geostationary Lightning
Mapper (GLM), which will monitor lightning continuously over a wide field of view.
In the following, we will discuss how individual sensors measuring various properties of
the lightning electromagnetic radiation can be combined into systems to provide lightning
locations. More details can be found in the reviews by Cummins and Murphy (2009),
Rakov (2013), and Nag et al. (2015) and in references therein.
As noted above, there are three major lightning locating methods: magnetic direction
finding (MDF), time-of-arrival (TOA), and interferometry. The MDF and long-baseline
(hundreds of kilometers) TOA systems usually operate in the VLF/LF range and report one
location per lightning event. The interferometry and shorter-baseline (tens of kilometers or
less) TOA systems usually operate at VHF and provide multiple locations per lightning
event (VHF images of lightning channels). The MDF, TOA, and interferometric methods
are reviewed in Sections 8.2.2, 8.2.3, and 8.2.4, respectively. Lightning locating systems
operating on global scale utilize methods that are capable of extracting source information
from electromagnetic signals dominated by ionospheric reflections. These latter methods
are considered in Section 8.2.5.
8.2.2 Magnetic Direction Finding (MDF)
Two vertical and orthogonal loops with planes oriented N–S and E–W, each measuring the
magnetic field from a given vertical radiator, can be used to obtain the direction to the
163 8.2 Principles of lightning location
Fig. 8.1 Determination of lightning stroke location when only two direction finders (DFs) detect the stroke. The solid lines
represent the measured azimuths to the stroke; the broken lines represent the ±1° angular random error in the
azimuth measurements. The solid circle indicates the computed stroke location; the shaded region indicates the
uncertainty in location of the stroke. Adapted from Holle and Lopez (1993).
source. This is the case because the output voltage of a given loop, by Faraday’s law, is
proportional to the cosine of the angle between the magnetic field vector and the normal
vector to the plane of the loop (Chapter 7). For a vertical radiator, the magnetic field lines
are circles coaxial with respect to the source. Hence, for example, the loop whose plane is
oriented N–S receives a maximum signal if the source is north or south of the antenna, while
the orthogonal E–W loop receives no signal. In general, the ratio of the two signals from the
loops is proportional to the tangent of the angle between north and the source as viewed
from the antenna. Thus a pair of two orthogonal loops can be used as a direction finder (DF).
As illustrated in Fig. 8.1 for a two-DF system, the intersection of two direction (azimuth)
vectors from the DF to the apparent source, provides a stroke location, but a location
containing error because each azimuth vector has some random angular error and may have
some systematic error. If a three-DF system is employed, each pair of DFs provides a
location, so there are three locations, the distance between the locations providing some
measure of the system error, as illustrated in Fig. 8.2. For three or more DF responses to a
lightning return stroke, the optimal estimate of the location is best found using a χ2
minimization technique.
Crossed-loop magnetic direction finders (DFs) used for lightning detection can be
divided into two general types: narrow-band (tuned) DFs and gated wideband DFs. In
both cases, the direction-finding technique involves an implicit assumption that the radiated
electric field is oriented vertically and the associated magnetic field is oriented horizontally
and perpendicular to the propagation path.
Narrow-band DFs have been used to detect distant lightning since the 1920s (Horner,
1954, 1957). They generally operate in a narrow frequency band with the center frequency
in the 5–10 kHz range, where attenuation in the Earth-ionosphere waveguide is relatively
low and where the lightning signal energy is relatively high. Before the development of
Fig. 8.2 Determination of lightning stroke location when three DFs detect the stroke. The solid lines represent the measured
azimuths to the stroke. The open circles indicate the three possible locations defined by the three different
intersections of the azimuth vectors. The optimal stroke location (solid circle) is determined by minimizing the χ2
function. The broken lines show the azimuth vectors to the computed optimal location. Adapted from Holle and
Lopez (1993).
weather radars in the 1940s, lightning locating systems were the primary means of
identifying and mapping thunderstorms at medium and long ranges.
A major disadvantage of narrow-band DFs is that for lightning at distances less than
about 200 km, those DFs have inherent azimuthal errors, called polarization errors, of the
order of 10° (Nishino et al., 1973; Kidder, 1973). These errors are caused by the detection of
magnetic field components from non-vertical channel sections, whose magnetic field lines
form circles in a plane perpendicular to the non-vertical channel section, and by ionospheric
reflections, sky waves, whose magnetic fields are also improperly oriented for direction
finding of the ground strike point.
To overcome the problem of large polarization errors at short ranges inherent in the
operation of narrow-band DFs, gated wideband DFs were developed in the early 1970s
(Krider et al., 1976). Direction finding is accomplished by sampling (gating on) the N–S
and E–W components of the initial peak of the return-stroke magnetic field, that peak being
radiated from the bottom hundred meters or so of the channel in the first microseconds of
the return stroke. Since the bottom of the channel tends to be straight and vertical
(perpendicular to the ground), the magnetic field lines form circles in a horizontal plane.
Additionally, a gated DF does not record ionospheric reflections since those reflections
arrive long after the initial peak magnetic field is sampled. The operating bandwidth of the
gated wideband DF is typically from a few kilohertz to about 500 kHz. Interestingly,
although an upper frequency response of many megahertz is needed to assure accurate
reproduction of the incoming radiation field peak, particularly if the propagation is over salt
water, practical DFs only need an upper frequency response of a few hundred kilohertz in
order to obtain an azimuthal error of about 1°. This is the case because the ratio of the peak
signals in the two loops is insensitive to the identical distortion produced by the identical
associated electronic circuits of the two loops. Thus, the gated wideband DF can operate at
frequencies below the AM radio band and below the frequencies of some aircraft naviga-
tional transmitters, either of which could otherwise cause unwanted directional noise.
Gated wideband DFs, as well as narrow-band DFs, are susceptible to site errors. Site
errors are a systematic function of direction but generally are time-invariant. These errors
are caused by the presence of unwanted magnetic fields due to non-flat terrain and to nearby
conducting objects, such as underground and overhead power lines and structures, being
excited to radiate by the incoming lightning fields. In order to eliminate site errors
completely, the area surrounding a DF must be flat and uniform, without significant
conducting objects, including buried ones, nearby. These requirements are usually difficult
to satisfy, so it is often easier to measure the DF site errors and to compensate for any that
are found than to find a location characterized by tolerably small site errors. Once correc-
tions are made, the residual errors have been reported (using independent optical data) to be
usually less than two to three degrees (e.g. Mach et al., 1986).
Since it is not known a priori whether a stroke to ground lowers positive or negative
charge, there is an 180° ambiguity in stroke azimuth from the measurement of only the
orthogonal magnetic fields. That ambiguity is resolved in DF systems by the measurement
of the associated electric field (Chapter 7) whose polarity indicates the sign of the charge
transferred to the ground.
8.2.3 Time-of-Arrival (TOA) technique
A single time-of-arrival sensor provides the time at which some portion of the lightning
electromagnetic field signal arrives at the sensing antenna. Time-of-arrival systems for
locating lightning can be divided into three general types: (1) very short baseline (tens to
hundreds of meters), (2) short baseline (tens of kilometers), and (3) long baseline (hundreds
to thousands of kilometers). Very short and short baseline systems generally operate at
VHF – that is, at frequencies from 30 to about 300 MHz – while long baseline systems
generally operate at VLF and LF, 3–300 kHz. It is generally thought that VHF radiation is
associated with air breakdown processes, while VLF signals are due to current flow in
conducting lightning channels. Short baseline systems are usually intended to provide
images of lightning channels and to study the spatial and temporal development of
discharges. Long baseline systems are usually used to identify the ground strike point or
the “average” location of the flash.
A very short baseline system (tens to hundreds of meters; e.g. Cianos et al., 1972; Taylor,
1978; Ray et al., 1987) is composed of two or more VHF time-of-arrival (TOA) receivers
whose spacing is such that the time difference between the arrival of an individual VHF
pulse from lightning at those receivers is short compared to the time interval between
pulses, which is some microseconds to hundreds of microseconds. The locus of all source
points capable of producing a given time difference between two receivers is, in general, a
hyperboloid, but if the receivers are very closely spaced, the hyperboloid degenerates, in the
limit, into a plane on which the source is found. Two time differences from three very
closely spaced receivers yield two planes whose intersection gives the direction to the
source; that is, its azimuth and elevation. To find the source location, as opposed to
determination of the direction to the source, two or more sets of three closely spaced
receivers, the sets being separated by tens of kilometers or more, must be used. Each set of
receivers is basically a TOA direction finder, and the intersection of two or more direction
vectors yields the location.
Short-baseline TOA systems (e.g. Proctor, 1971; Lennon and Poehler, 1982; Rison et al.,
1999) are typically networks of 5–15 stations that make use of time-of-arrival information
for three-dimensional (3D) mapping of lightning channels. A portable version of such
system has been developed by researchers at the New Mexico Institute of Mining and
Technology. This system is presently referred to as the Lightning Mapping Array (LMA)
and has recently become a major tool for both lightning research and operational applica-
tions. The short-baseline VHF TOA systems provide electromagnetic images of the devel-
oping channels of any type of lightning flash. An example of LMA image of the upward
positive leader and initial continuous current of a flash triggered using the rocket-and-wire
technique at Camp Blanding, Florida, is shown in Fig. 8.3.
LMA Source Locations 081811 (UF 11-32, 20:37:29.870)
• • ••
5.5 • •••
Branch 5 ••
• •
• •• • • Branch 6
5 •••
•• •
•••••
• • ••
4.5 ••
• • ••• Time after 20:37:29.870
• •
••• •
• 180 < T < 200
Branch 4 • •••
4 •
Branch 3 • •• 160 < T < 180
•• •••
• •
••••• • •• 140 < T < 160
3.5 •
••• ••
• • •••
•• • •• •• 120 < T < 140
• •••• ••
•••
Altitude (km)
3 ••
• •• Branch 2 100 < T < 120
•• (ms)
•
• • ••• 80 < T < 100
•
2.5 • •••
• 60 < T < 80
• ••
• •••
2 Branch 1 • • 40 < T < 60
••
•• ••
• ••• 20 < T < 40
••
•• •••
1.5 •
• •• 0 < T < 20
•••
•• •
•••
1 •••••
•• •
• •
••
••••••
Initial UPL •••••••
••••••
••
0.5 ••••
•••
•• –4
Origin –3–3.5
Launcher –2 –2.5
0 –1–1.5
–2 –1.5 –1 0 –0.5 E/W (km)
–0.5 0 0.5 1 1.5 1 0.5
2
N/S (km)
Fig. 8.3 Three-dimensional view of the VHF source locations associated with the upward positive leader and initial
continuous current of flash UF 11–32 triggered at Camp Blanding, Florida, on August 18, 2011. The image was
obtained with LMA. The sources span 200 ms and are color coded in time in 20 ms time windows according to the
key at right. Some sources have been removed. Adapted from Hill et al. (2012).
One of the first long baseline (hundreds to thousands of kilometers) TOA systems
operated at VLF/LF (Lewis et al., 1960). It employed a pair of receiving stations in
Massachusetts with a bandwidth of 4–45 kHz, separated by over 100 km (the overall
network was composed of four stations), to compare differences in the times of arrival of
the signals at the two stations and hence to determine directions to the causative lightning
discharge in western Europe. The two-station system was basically a direction finder
similar to the very short baseline systems described above, but operating at lower frequen-
cies and longer baseline. The resultant “directions” compared favorably with the locations
reported by the British Meteorological Office’s narrow-band DF network. Spherical geo-
metry was used to account for propagation over the Earth’s surface in finding the locus of
points for a constant measured arrival time difference between receivers.
Another long-baseline TOA system, called the Lightning Positioning and Tracking
System (LPATS), was developed in the 1980s. The LPATS, operating at VLF/LF, used
electric field whip antennas at stations 200–400 km apart to determine locations via the
measured differences between signal arrival times at the stations. In the frequency band
used, return-stroke waveforms were generally the largest and hence most easily identified.
In general, responses from four stations (three time differences) are needed to produce a
unique stroke location, since the hyperbolas on the Earth’s surface from only two time
differences can, in general, intersect at two different points, as illustrated in Fig. 8.4. For
cloud-to-ground lightning near or within the network, there is often only one solution (see
Fig. 8.5), so in this case, the three-station solution suffices.
8.2.4 Interferometry
In addition to radiating isolated pulses, lightning also produces noise-like bursts of

electromagnetic radiation lasting for tens to hundreds of microseconds. These bursts
are difficult to locate using TOA techniques due to the difficulty in identifying the
individual pulses. In the case of interferometry, no identification of individual pulses is
needed, since the interferometer measures phase difference between narrow-band
signals corresponding to these noise-like bursts received by two or more closely spaced
sensors. The simplest lightning interferometer consists of two antennas some meters
apart, each antenna being connected via a narrow-band filter to a receiver. The
antennas, filters, and receivers are identical. The output signals of the two receivers
are sent to a phase detector that produces a voltage that is proportional to the difference
in phase between the two quasi-sinusoidal signals. The phase difference defines, as
does the time difference in very-short-baseline TOA systems, a plane on which the
source is located; that is, one direction angle to the VHF source. To find the azimuth
and elevation of a source, three receiving antennas with two orthogonal baselines are
needed, at minimum. To locate the source in three dimensions, two or more synchro-
nized interferometers, each effectively acting as a direction finder, separated by a
distance of the order of 10 km or more, are needed. The principles of interferometric
lightning location are described in detail by Lojou et al. (2009). Examples of inter-
ferometric images of lightning processes are found in Figs. 4.6 and 4.12.
Fig. 8.4 Determination of lightning stroke location by three TOA receivers when the solution is not unique. Shown are two
hyperbolas, defined by the TOA differences, that intersect at two points (open circles); one point corresponds to the
actual stroke position and the other is a false solution. Adapted from Holle and Lopez (1993).
Fig. 8.5 Determination of lightning stroke location by three TOA receivers when the solution is unique. Shown are two
hyperbolas, defined by the TOA differences, that intersect to define the unique location of the stroke (open circle).
Adapted from Holle and Lopez (1993).
Most interferometric systems operate over very narrow frequency bands (a few hundred
kilohertz to a few megahertz in the VHF/UHF bands, where UHF indicates the
ultra-high-frequency range, from 300 MHz to 3 GHz), since this allows the system to
have high sensitivity in a specific “quiet” band of operation. However, it also makes the
system performance subject to local broadband interference, may not provide the highest
possible signal-to-noise ratio (SNR), and places a specific limitation in the spacing of the
antenna array elements to avoid arrival time (phase) ambiguity. There is a recent trend
toward using broadband interferometry (e.g. Morimoto et al., 2004; Stock et al., 2014). This
169 8.3 Performance characteristics
trend is made possible by the advent of affordable broadband RF and digital signal
processing electronics.
8.2.5 Lightning location on global scale
Global lightning locating systems employ a relatively small (some tens or less) number of
sensors, so that the baselines are of the order of thousands of kilometers and signals are
dominated by ionospheric reflections. The VLF range (3–30 KHz) is used. Two location
methods employed by currently operating systems are outlined below.
Time-of-Group-Arrival (TOGA) method. This method is based on the fact that light-
ning VLF signals (sferics) propagating in the Earth-ionosphere waveguide experience
dispersion, in that the higher-frequency components arrive earlier than the lower-frequency
components (e.g. Dowden et al., 2002). The TOGA, a quantity that can be derived from the
measured sferic waveform, is related to the distance traveled by the sferic. The TOGA
lightning location method is implemented in the World Wide Lightning Location Network
(WWLLN).
MDF and TOA methods combined with a lightning waveform recognition
algorithm. Lightning locations are obtained using the MDF and long-baseline TOA
methods, similar to those described in Sections 8.2.2 and 8.2.3, respectively, in con-
junction with a lightning waveform recognition algorithm. The latter relies on a bank
of “canonical” waveforms corresponding to propagation distances of the order of
thousands of kilometers (Said et al., 2010). This method is implemented in the
Global Lightning Dataset (GLD360).
8.3 Performance characteristics
Generally, a modern VLF–MF (MF indicates the medium frequency range, from 300 kHz to
3 MHz) lightning locating system (LLS) is expected to record, in separate categories, and
locate over a certain area all or most cloud-to-ground strokes of either polarity, as well as
cloud discharges. Also expected for each discharge is a measure of its intensity, usually in
the form of peak current inferred from measured electric or magnetic field. Accordingly,
system’s performance can be evaluated using the following characteristics:
(a) cloud-to-ground (CG) flash detection efficiency;
(b) CG stroke detection efficiency;
(c) cloud discharge detection efficiency;
(d) percentage of misclassified events (particularly cloud discharges assigned to the
positive or negative CG stroke category);
(e) location accuracy (or location error);
(f) peak current estimation error.
Given below are definitions of LLS performance characteristics, followed by a discus-
sion of various approaches to their evaluation.
In general, the detection efficiency is the fraction (usually expressed in percent) of the
total events occurred that are detected by the system and is ideally equal to 100 percent.
While the CG stroke detection efficiency can be readily defined (since these strokes involve
a unique and observable feature – luminous channel to ground – and the total number of
occurred events can be practically determined), the cloud discharge detection efficiency
concept is rather uncertain. Indeed, there are many cloud discharge processes (some of them
poorly understood) occurring on different spatial- and time-scales and apparently exhibit-
ing no unique and readily observable features. As a result, the total number of occurred
events is generally unknown. In practice, if all cloud discharge events are accepted as
“counts,” the number of detected cloud discharges may be largely determined by the local
noise level and system’s signal transmission rate limit.
In defining the CG flash detection efficiency, probably the most important performance
characteristic for lightning locating systems used for the determination of ground flash
density, a flash is considered to be detected when at least one stroke of the flash was
detected. A similar approach could be applied to cloud flashes, although one would need
to decide whether a single “count” constitutes a flash and how to assign multiple “counts” to
individual flashes.
The location error is the distance between the actual location and that reported by the
system. In general, the location error consists of random and systematic components. The
latter in some cases can be accounted for (e.g. site errors in MDF systems).
The peak current estimation error is the difference between the actual peak current value
and that reported by the system, usually expressed as a percentage of the actual peak
current. Peak currents are estimated by lightning locating systems using either an empirical
or model-based field-to-current conversion equation. There are reasonable field-to-current
conversion equations for CG strokes, but not for cloud discharge processes.
In order to evaluate the performance characteristics listed above, independent
(ground-truth) data are needed. For example, discharges occurring at a precisely
known location equipped with a current-measuring device (tall tower or lightning-
triggering facility) can be used for estimating the location accuracy and peak current
estimation error. Detection efficiencies and percentage of misclassified events are
usually estimated based on time-resolved optical recordings. Sometimes, lightning-
related damage to various objects (buildings, trees, etc.) is used in estimating location
errors, although identification of causative lightning event in this approach is not
unique due to insufficient accuracy of timing information (usually not known within
better than a minute). Less definitive evaluations of lightning locating system’s per-
formance characteristics are possible via modeling or comparison with a more accurate
system operating in the same area. As of today, only a limited number of ground-truth
studies have been performed, particularly for first strokes in negative CG flashes,
positive CG flashes, and cloud discharges.
In some applications (e.g. tracking of thunderstorm cells), the tracking ability may be
more important than detection of individual lightning discharges. Performance of the
systems intended primarily for such applications is often tested against radar or infrared
satellite imagery, with a good correspondence between detected lightning and regions of
high radar reflectivity or low cloud-top temperature being viewed as indicative of system’s
171 8.4 Examples of modern lightning locating systems
output validity. For early warning, the ability to detect the first lightning is probably the
most important performance characteristic.
It is not clear how to define the performance characteristics for VHF lightning channel
imaging systems in terms of lightning events, as opposed to individual VHF sources.
Limitations in sensitivity usually prevent these systems from directly detecting and map-
ping positive leaders. Further, supplementary information about return strokes is usually
needed to reliably distinguish between cloud flashes (including attempted leaders develop-
ing toward the ground) and CG flashes, because the VHF radiation directly associated with
return strokes may be limited and difficult to detect. Also, no peak current estimates are
possible. Nevertheless, VHF lightning channel imaging systems represent a very valuable
tool for studying detailed lightning morphology and evolution, particularly inside the cloud,
and are often used in testing other types of lightning locating systems.
8.4 Examples of modern lightning locating systems
One VHF lightning channel imaging system (LMA) and six single location per lightning
event networks, three VLF/LF (NLDN, LINET, and USPLN), one ELF–HF (ENTLN), and
two VLF (WWLLN and GLD360) systems, are briefly reviewed here as representative
examples of modern lightning locating systems. Actually, LINET can report several loca-
tions per flash in 3D (e.g. Stolzenburg et al., 2012), which requires that the closest sensor is
within about 100 km of the lightning discharge, and hence can operate as a crude imaging
system. Information about other systems (including optical) can be found in Rakov and
Uman (2003, Ch. 17), Cummins and Murphy (2009), Betz et al. (2009), Nag et al. (2015)
and references therein. There are more than 60 lightning locating networks worldwide that
operate in the VLF/LF range.
Besides the general characterization of each system, the available information on its
performance characteristics is given with emphasis on those based on formal ground-truth
studies published in the peer-reviewed literature.
8.4.1 Lightning Mapping Array (LMA), 60–66 MHz
LMA networks typically consist of 10–15 stations separated by 15–20 km and connected by
wireless communication links to a central location (Thomas et al., 2004). Each station
receives the lightning signals (from both cloud and CG flashes) in a locally unused
television channel (usually TV channel 3, 60–66 MHz). Typical time resolution (measure-
ment time window) is 80–100 μs, which is sufficient for mapping relatively slow leader
processes. A larger time window, typically 400 μs, is used for real-time processing and
display.
The location accuracy of the New Mexico LMA has been investigated experimentally
using a sounding balloon carrying a VHF transmitter, airplane tracks, and observations of
distant storms (Thomas et al., 2004). Simple geometric models for estimating the location
uncertainty of sources both over and outside the network have also been developed. The
model results were found to be a good estimator of the observed errors. Sources over the
network at altitudes ranging from 6 to 12 km (well above the lower cloud boundary) were
located with an uncertainty of 6–12 m rms in the horizontal and 20–30 m rms in the vertical,
resulting in less than a 100 m 3D error for most located sources. Outside the network the
location uncertainties increase with distance.
8.4.2 US National Lightning Detection Network (NLDN), 400 Hz–400 kHz
The NLDN consists of more than 100 stations separated by typically 300–350 km and
mostly covering the contiguous USA. A combination of TOA and MDF locating
techniques is employed. Both cloud and CG lightning discharges are reported.
Classification is accomplished by applying field waveform criteria to individual mag-
netic field pulses. Generally, pulses wider than a certain threshold are interpreted as
being produced by return strokes, while narrower pulses are attributed to cloud
discharge activity. All positive events with estimated peak currents <15 kA are
classified by the NLDN as cloud pulses. Peak currents are estimated from measured
fields using an empirical formula based on rocket-triggered-lightning data, with the
field peaks being adjusted to account for expected propagation effects (stronger than
the inverse proportionality distance dependence). Further information on the evolution
of the NLDN, its enabling methodology, and applications of NLDN data can be found
in Rakov and Uman (2003, Ch. 17), Rakov (2005), Orville (2008), Cummins and
Murphy (2009), Nag et al. (2014) and references therein.
CG stroke and flash detection efficiencies have been investigated, using video cameras,
in Southern Arizona, Oklahoma, and Texas (Biagi et al., 2007). The stroke detection
efficiency in Southern Arizona was estimated to be 76 percent (N = 3620), and in Texas/
Oklahoma, it was 85 percent (N = 885). The corresponding flash detection efficiencies were
93 percent (N = 1097) and 92 percent (N = 367). Additionally, classification of lightning
events as cloud or CG discharges was examined in this study, as well as in a similar study
(but additionally using independent (LASA) electric field waveform measurements) in the
Colorado–Kansas–Nebraska region (Fleenor et al., 2009).
CG stroke and flash detection efficiencies have been also investigated, using as the
ground-truth rocket-triggered-lightning data, in the Florida region (Jerauld et al., 2005; Nag
et al., 2011; Mallick et al., 2014a). From the latest (2004–12) study, the CG stroke and flash
detection efficiencies were found to be 75 percent and 94 percent, respectively. Strokes in
rocket-triggered flashes are similar to regular subsequent strokes (following previously
formed channels) in natural lightning and, hence, the 75 percent stroke detection efficiency
value is applicable only to regular negative subsequent strokes in natural lightning. The
flash detection efficiency is expected to be an underestimate of the true value for natural
negative lightning flashes, since first strokes typically have larger peak currents than
subsequent ones. Zhu et al. (2016), using high-speed video and electric field data on 366
natural CG strokes in Florida, found the NLDN stroke detection efficiency to be 93 percent.
The classification accuracy in their study was 91 percent.
Nag and Rakov (2012) examined electric field waveforms produced by 45 positive
flashes containing 53 strokes. Out of these 53 strokes, the NLDN located 51 (96 percent),
of which 48 (91 percent) were correctly identified and three return strokes were misclassi-
fied as cloud discharges.
According to Cummins and Murphy (2009), the NLDN cloud-flash detection effi-
ciency (a flash was considered detected if at least one VLF/LF pulse produced by that
flash was detected) was in the range of 10–20 percent, depending on local differences
in distances between stations. From a more recent study based on using data from two
VHF lightning imaging systems (LMAs) as a reference, Murphy and Nag (2015)
reported the cloud-flash detection efficiency to be in the 50–60 percent range. Wilson
et al. (2013) stated that the NLDN typically reports one to three cloud pulses per flash.
Nag et al. (2010) examined wideband electric fields, electric and magnetic field
derivatives, and narrowband VHF (36 MHz) radiation bursts produced by 157 compact
intracloud discharges (CIDs; see Appendix 4). The NLDN located 150 (96 percent) of
those CIDs and correctly identified 149 (95 percent) of them as cloud discharges. Zhu
et al. (2016), using high-speed video and electric field data on cloud discharge activity
in both cloud and cloud-to-ground flashes in Florida, found the NLDN detection
efficiency for those events (sequences of cloud pulses) to be 37 percent (N = 95).
The classification accuracy in their study was 91 percent (N = 35). Out of ten complete
cloud flashes, five were detected and all of those five were correctly classified.
Mallick et al. (2014a) estimated, from comparison of NLDN-reported locations with the
precisely known locations of triggered-lightning ground attachment points, the median
location error to be 334 m, with the largest error being 8 km. Data acquired in 2004–12 were
used.
Peak current estimation errors have been estimated from comparison of NLDN-reported
peak currents with directly measured currents at the triggered-lightning channel base. In
2004–12, the median absolute value of current estimation error was 14 percent (Mallick
et al., 2014a). The current estimation errors never exceeded 127 percent in absolute value.
These results (also the location error results based on triggered-lightning data) are applic-
able only to regular negative subsequent strokes in natural lightning.
8.4.3 LIghtning detection NETwork (LINET), 1–200 kHz
The basic location method used in this system is TOA, although the magnetic field sensors
provide arrival-angle information that is employed as a “plausibility check” on computed
locations. Height information derived from the arrival time at the nearest reporting sensor is
employed to assist in classification of processes in cloud flashes and in-cloud processes (e.g.
preliminary breakdown) in CG flashes on the one hand and CG strokes on the other
(near-ground locations are assumed to be associated with CG strokes and elevated ones
with all the other processes). It is stated that the reliable separation of return strokes
and cloud pulses can be achieved as long as the closest sensor is within about 100 km of
the lightning discharge, which requires baselines of 200–250 km or less. Emphasis is placed
on detection of low-amplitude signals of both cloud and CG lightning and recognition of
thunderstorm cells for nowcasting purposes. Peak currents for processes in cloud
flashes, in-cloud processes in CG flashes, and CG strokes are estimated assuming direct
proportionality between the peak current and peak magnetic (or electric) field and
inverse distance dependence of field peak. More information about LINET can be found in
Betz et al. (2009) and references therein.
Similar to VHF channel imaging systems, it is not clear how to define the detection
efficiency for LINET, which, in a sense, also maps evolution of lightning channels,
although with a considerably smaller number of located sources per flash. Additionally,
in-cloud processes (e.g. preliminary breakdown) in CG flashes are assigned to the cloud
lightning category, which is apparently inconsistent with the traditional definitions of cloud
flash as a lightning discharge without CG strokes and CG flash as a lightning discharge that
consists of both in-cloud processes and CG strokes. This is probably immaterial for a
number of applications, such as cell tracking and detection of severe weather.
The random location error is claimed to be approximately 150 m, but the existence of
systematic errors is acknowledged. Betz et al. (2009) showed an example of 58 located
strokes apparently terminated on an instrumented tower with an average location error of
less than about 100 m, after compensating systematic errors that caused a location bias of
about 200 m. Peak current estimation errors for LINET are unknown (no comparison with
ground-truth data has been made to date).
8.4.4 US Precision Lightning Network (USPLN), 1.5–400 kHz
This network employs the VLF/LF TOA technique and consists of 100 electric field sensors
covering the continental USA and other parts of North America. No formal performance
testing studies regarding this system have been reported, but the operators of the system
claim, apparently from the network simulation analysis, 95 percent stroke detection effi-
ciency and 250 m typical location error throughout most of North America (>80 percent
detection efficiency and <1 km location error in “key deployment areas” elsewhere in the
world). Differentiation between cloud and CG processes is apparently accomplished by
examining the frequency content and amplitude of the received signals. The field-to-current
conversion procedure has not been formally described, nor is any information about testing
its validity available.
8.4.5 Earth Networks Total Lightning Network (ENTLN), 1 Hz–12 MHz
The ENTLN sensors operate in a frequency range from 1 Hz to 12 MHz (spanning the ELF,
VLF, LF, MF, and HF ranges, where ELF indicates the extremely low-frequency range,
from 3 to 30 Hz). The TOA method is employed. According to Heckman and Liu (2010),
the whole electric field waveforms are used in both locating the lightning events and
differentiating between cloud and CG processes. Strokes (or individual cloud events) are
clustered into a flash if they are within 700 ms and 10 km of the first detected stroke (or
cloud event). A flash that contains at least one return stroke is classified as a CG flash;
otherwise, it is classified as a cloud flash. In the cell tracking and thunderstorm alert
generation, only flashes (which are less likely than strokes to be missed by the system)
are used. The system operates in the USA and in a number of other countries.
The operators of the system claim 40–50 percent cloud flash detection efficiency across
much of the USA and up to 95 percent in the US Midwest and East (Heckman and Liu,
2010). Maximizing the detection efficiency for cloud flashes appears to be the primary
focus of this system. Peak currents for processes in cloud flashes, in-cloud processes (e.g.
preliminary breakdown) in CG flashes, and CG strokes are estimated assuming direct
proportionality between the peak current and peak electric field and inverse distance
dependence of field peak.
Mallick et al. (2015) have evaluated the performance characteristics of the ENTLN
using, as ground truth, data for 245 negative return strokes in 57 flashes triggered from June
2009 to August 2012 at Camp Blanding, Florida. The performance characteristics were
determined both for the ENTLN processor that had been in service at the time of acquiring
triggered-lightning data (June 2009 to August 2012) and for the new ENTLN processor,
introduced in November 2012. So, evaluation for the new processor simulates ENTLN
output as if the new processor were in service from June 2009 to August 2012. For the same
ground-truth dataset and the same evaluation methodology, different performance char-
acteristics for those two processors were obtained. For the old processor, flash detection
efficiency was 77 percent, stroke detection efficiency was 49 percent, fraction of misclas-
sified events was 61 percent, median location error was 631 m, and median absolute current
estimation error was 51 percent. For the new processor, flash detection efficiency was 89
percent, stroke detection efficiency was 67 percent, fraction of misclassified events was 54
percent, median location error was 760 m, and median absolute current estimation error was
19 percent.
8.4.6 World Wide Lightning Location Network (WWLLN), 6–18 kHz
The WWLLN utilizes the time-of-group-arrival (TOGA) method (see Section 8.2.5) to
locate lightning events. As of March 2012, WWLLN employed 57 sensors located on all
continents, although, according to Dowden et al. (2002), global coverage could be in
principle provided by as few as 10 sensors. Distances between the sensors are up to
thousands of kilometers. Presently, only those lightning events that triggered at least five
sensors and that had residuals (uncertainties in the stroke timing) less than or equal to 30 μs
are regarded as located with acceptable accuracy.
In their study of WWLLN performance characteristics, Abarca et al. (2010) used NLDN
data as the ground truth and found that the CG flash detection efficiency increased from
about 3.88 percent in 2006–7 to 10.3 percent in 2008–9, as the number of sensors increased
from 28 in 2006 to 38 in 2009. For events with NLDN-reported peak currents of 130 kA or
higher, the detection efficiency was reported to be 35 percent. The average location error
was estimated to be 4–5 km. Hutchins et al. (2012a) developed a model to compensate
for the uneven global coverage of the WWLLN. It is known that field peaks at
distances >700 km or so are due to ionospheric reflections rather than the ground wave.
Interaction of lightning signals with the ionosphere spectrally distorts the field waveform so
that it is not straightforward to infer the peak current and even polarity of lightning.
Nevertheless, Hutchins et al. (2012b) developed a method to convert the stroke-radiated
power in the 6–18 kHz band to peak current.
Mallick et al. (2014) evaluated the performance characteristics of the WWLLN using
rocket-triggered lightning data acquired at Camp Blanding, Florida, in 2008–13. The flash
and stroke detection efficiencies were 8.8 percent and 2.5 percent, respectively. The stroke
detection efficiency for strokes with peak current ≥25 kA was 29 percent. The median
location error was 2.1 km. The median absolute error in WWLLN peak currents estimated
from the empirical formula of Hutchins et al. (2012b) was 30 percent.
8.4.7 Global Lightning Dataset (GLD360), 300 Hz–48 kHz

The Global Lightning Dataset (GLD360), also referred to as the Global Lightning
Detection Network (GLDN), employs an unspecified number of VLF sensors strategically
placed around the world. As stated in Section 8.2.5, locations are obtained using both TOA
and MDF methods in conjunction with a lightning waveform recognition algorithm. A
lightning event must be detected by at least three sensors to be located. The system does not
distinguish between ground and cloud lightning events.
Demetriades et al. (2010) evaluated the GLD360 performance characteristics using
NLDN data as the ground truth and found that the CG flash detection efficiency was 86–
92 percent, and the median location error was 10.8 km. From a similar study, but using the
Brazilian lightning detection network, Naccarato et al. (2010) reported the CG flash
detection efficiency of 16 percent and the mean location error of 12.5 km. GLD360
performance in Europe in May–September 2011 was compared with that of the networks
participating in the European Cooperation for Lightning Detection (EUCLID) by Ponjola
and Makela (2013). Poelman et al. (2013), using electric field measurements in conjunction
with high-speed video recordings for 210 strokes in 57 negative CG flashes in Belgium,
estimated the flash and stroke detection efficiencies to be 96 and 70 percent, respectively.
They also estimated the median location error of 1.3 km (N = 134) relative to EUCLID
locations. Said et al. (2013), using NLDN data as the ground truth, estimated the ground
flash detection efficiency of 57 percent and median location error of 2.5 km.
The GLD360 also reports the peak current (inferred from the measured magnetic field
peak) and polarity. The latter is determined via the cross-correlation with the bank of
“canonical” waveforms. Said et al. (2013) estimated the mean and geometric mean errors in
peak current estimates relative to NLDN-reported peak currents to be 21 and 6 percent,
respectively. They also found that for 96 percent of matched events, GLD360 reported the
same polarity as the NLDN.
Mallick et al. (2014b) used rocket-triggered lightning data acquired in 2011–13 at Camp
Blanding, Florida, to evaluate the GLD360 performance characteristics. The flash and
stroke detection efficiencies were 67 and 37 percent, respectively. Out of 75 detected
strokes, one (1.3 percent) was reported with incorrect polarity. The median location error
was 2.0 km, and the median absolute current estimation error was 27 percent.
8.5 Summary
There exists a variety of lightning locating techniques that are based on the detection of
lightning radio-frequency electromagnetic signals, with accurate locating being possible
177 Further reading
only by using multiple-station systems. When a single location per cloud-to-ground light-
ning stroke, typically the ground strike point, is required, magnetic direction finding, the
long-baseline time-of-arrival technique, or a combination of the two can be employed.
Location accuracies of the order of hundreds of meters and flash detection efficiencies of
about 90 percent are possible. When electromagnetic imaging of the developing channels of
any type of lightning flash is required, the VHF time-of-arrival technique or VHF inter-
ferometry can be used. Lightning locating systems operating on global scale utilize
methods that are capable of extracting source information from electromagnetic signals
dominated by ionospheric reflections.
Although theoretical models can be used to evaluate the LLS performance character-
istics, ultimately ground-truth data are needed to verify model-predicted results. Such data
should include the measured time, position, and peak current of lightning discharges in a
region covered by LLS and can be obtained using instrumented towers or rocket-triggered
lightning. Video camera observations can be used for evaluating detection efficiency and
classification of lightning events (CG vs. non-CG).
Questions and problems
8.1. Explain the 180° ambiguity in magnetic direction finding systems.

8.2. Why are the time-of-arrival systems sometimes referred to as hyperbolic?
8.3. What is the minimum number of stations needed to obtain unique location with (a)
MDF technique and (b) TOA technique?
8.4. What is LMA?
8.5. Give examples of global lightning locating systems. In what frequency range do they
operate?
8.6. List the performance characteristics of lightning locating systems.
8.7. Discuss the use of rocket-triggered lightning for testing the performance characteristic
of lightning locating systems.
8.8. What is the detection efficiency? How can it be verified for (a) cloud-to-ground flashes
and (b) cloud flashes?
Further reading
Betz, H. D., Schumann, U., and Laroche, P., eds. (2009). Lightning: Principles, Instruments
and Applications, 691 pp., Springer.
Rakov, V. A. and Uman, M. A. (2003). Lightning: Physics and Effects, 687 pp., New York:
Cambridge.

8 Electromagnetic Methods of Lightning Location

Uploaded by

Copyright:

Available Formats

8 Electromagnetic Methods of Lightning Location

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

8 Electromagnetic Methods of Lightning Location

Uploaded by

Copyright:

Available Formats

8 Electromagnetic methods of lightning location

8.2 Principles of lightning location

The three most common electromagnetic radio-frequency-locating techniques include the

8.2.2 Magnetic Direction Finding (MDF)

8.2.3 Time-of-Arrival (TOA) technique

LMA Source Locations 081811 (UF 11-32, 20:37:29.870)

In addition to radiating isolated pulses, lightning also produces noise-like bursts of

8.2.5 Lightning location on global scale

8.3 Performance characteristics

8.4 Examples of modern lightning locating systems

8.4.1 Lightning Mapping Array (LMA), 60–66 MHz

8.4.2 US National Lightning Detection Network (NLDN), 400 Hz–400 kHz

8.4.3 LIghtning detection NETwork (LINET), 1–200 kHz

8.4.4 US Precision Lightning Network (USPLN), 1.5–400 kHz

8.4.5 Earth Networks Total Lightning Network (ENTLN), 1 Hz–12 MHz

8.4.6 World Wide Lightning Location Network (WWLLN), 6–18 kHz

8.4.7 Global Lightning Dataset (GLD360), 300 Hz–48 kHz

Questions and problems

8.1. Explain the 180° ambiguity in magnetic direction ﬁnding systems.

You might also like