You are on page 1of 15

Journal of Geophysical Research: Atmospheres

RESEARCH ARTICLE When does the lightning attachment process actually begin?
10.1002/2015JD023155
M. D. Tran1 and V. A. Rakov1
Key Points: 1 Department of Electrical and Computer Engineering, University of Florida, Gainesville, Florida, USA
• Negative downward leader was
intercepted by low-conductivity
positive streamers
• Break-through phase of the Abstract High-speed video and electric field records of 43 first and 7 new-ground-termination
attachment process was subsequent strokes in negative lightning flashes, obtained at the Lightning Observatory in Gainesville,
unexpectedly long
Florida, were examined. Eighteen (36%) of these strokes exhibited faintly luminous formations (FLFs) below
• The streamer zone was up to 200 m
in length the downward leader tip just prior to the return stroke. All the 18 FLFs were connected to the strike object
on the ground, with 14 of them being also in contact with the downward leader tip. For 11 located events
showing FLFs in contact with the downward leader tip, the FLF 2-D length ranged from 51 to 200 m and
Supporting Information:
• Text S1 the time interval between the end of exposure of the FLF frame and the return stroke onset ranged from 42
• Text S1 and Figures S1-S2 combined to 106 μs. We inferred that the FLFs mostly consisted of relatively low conductivity streamers, based on the
• Figure S1 following observations and estimates. (a) The rates at which the 11 FLFs were replaced by channels capable
• Figure S2
of guiding return stroke waves ranged from 0.7 × 106 m/s to 2.2 × 106 m/s, which is comparable to leader
speeds in virgin air (in the same records). (b) The overwhelming majority of downward leaders continued
Correspondence to: stepping while being connected to ground via FLFs. (c) The average electric field along FLFs is 1 to 2 orders
M. D. Tran,
manhtran@ufl.edu of magnitude higher than expected for hot, leader-like channels. Since in all cases the FLF determined the
strike point, we conclude that its inferred initiation up to a few hundreds of microseconds before the return
stroke onset signified the beginning of the lightning attachment process.
Citation:
Tran, M. D., and V. A. Rakov
(2015), When does the lightning
attachment process actually begin?, J.
Geophys. Res. Atmos., 120, 6922–6936, 1. Introduction
doi:10.1002/2015JD023155.
The lightning attachment process, which can be viewed as the transition from leader to return stroke, is one of
the most poorly understood lightning processes, primarily due to its short duration and frequent occurrence
Received 22 JAN 2015
in the same frame as the return stroke in high-speed video records. The attachment process in negative strokes
Accepted 16 JUN 2015
Accepted article online 20 JUN 2015
is generally assumed to include a positive upward connecting leader (UCL) and the so-called break-through
Published online 21 JUL 2015 phase [Rakov and Uman, 2003, chapter 4]. To date, both of these processes were observed in long sparks
[Lebedev et al., 2007; Shcherbakov et al., 2007] and in rocket-and-wire-triggered lightning [Biagi et al., 2009]
but not in natural lightning. The UCL is usually assumed to be initiated (via a sequence of processes includ-
ing glow corona, streamer, and streamer-to-leader transition [e.g., Aleksandrov et al., 2001; Becerra and Cooray,
2006; Bazelyan et al., 2009]) at the prospective strike object in response to the approaching downward leader
channel. Each of the leaders includes a hot, relatively high conductivity channel core surrounded by a rela-
tively low conductivity corona sheath and a relatively low conductivity streamer zone extending ahead of the
moving leader tip. The break-through phase starts when the relatively low conductivity streamer zones of
upward and downward leaders come in contact to form a common streamer zone. The subsequent acceler-
ated extension of the two relatively high conductivity plasma channels toward each other takes place inside
the common streamer zone. Once the two plasma channels make contact with each other, two return stroke
current waves are initiated from the junction point. One current wave moves upward, toward the cloud, and
the other moves downward, toward ground [Wang et al., 1999, 2013, 2014; Jerauld et al., 2007]. The down-
ward wave is short lived and, after its reflection from ground, probably catches up with the upward wave
front [e.g., Rakov, 2013]. A single upward moving return stroke wave is eventually formed. Examples of optical
images of UCL and the common streamer zone in rocket-triggered lightning and in laboratory sparks are
shown in Figure 1.
Wang et al. [1999], who employed the digital optical imaging system Automatic Lightning Progressing Fea-
ture Observation System, obtained highly resolved (3.6 m spatial and 100 ns time resolution) optical images
of the attachment process in two triggered lightning strokes, which are thought to be similar to subsequent
strokes in natural lightning. They observed a UCL in one of the strokes and inferred the existence of a UCL
©2015. American Geophysical Union.
in the other one. In both events, the return stroke was initiated as a bidirectional process that involved both
All Rights Reserved. upward and downward moving waves which originated at 7–11 m (in the event with the imaged UCL) and

TRAN AND RAKOV LIGHTNING ATTACHMENT PROCESS 6922


Journal of Geophysical Research: Atmospheres 10.1002/2015JD023155

Figure 1. Break-through phase of the attachment process in rocket-triggered lightning and in long laboratory sparks.
(a) The original and inverted-intensity versions of the high-speed video frame (20 μs integration time) just prior to the
return stroke frame of a negative triggered lightning stroke. (b and c) Single frames of two negative laboratory sparks
captured by image converter camera K008 with frame durations of 2 μs and 0.2 μs, respectively. Ls in Figure 1b is the
length of just-formed step (note the burst of negative streamers from the lower end of the step) of the downward
negative leader. In Figure 1c, light intensity is color coded, with the highest intensity shown in white. The common
streamer zone (a faint filament in Figure 1b, which is difficult to see in a reproduction) between the downward leader
channel and the upward connecting leader channel is imaged in all three cases. Images shown in Figures 1a–1c are
adapted from Biagi et al. [2009], Lebedev et al. [2007], and Shcherbakov et al. [2007], respectively.

4–7 m (in the event with no imaged UCL). The imaged UCL had duration of about 400 ns. It had a light inten-
sity 1 order of magnitude lower than its associated downward dart leader, and its propagation speed was
estimated to be about 2 × 107 m/s, similar to the typical speed of downward dart leaders. In similar studies
of 14 classical and 7 anomalous triggered lightning strokes, Wang et al. [2013, 2014] examined the junction
point height and the striking distance, each as a function of return stroke peak current. It is worth noting that
the UCL in natural lightning is often faint and, for this reason, is sometimes referred to as upward connecting
streamer [e.g., Berger, 1977], although streamers are usually considered to have relatively low conductivity, as
opposed to hot, leader-like channels. The triggered lightning UCL observed by Wang et al. [1999] was faint but
capable of supporting the propagation of a downward return stroke wave, which suggests that its channel
was hot.
Biagi et al. [2009], using a high-speed video camera (about 20 μs exposure time), recorded UCLs, ranging from
9 to 22 m in length, in eight strokes of a rocket-triggered lightning flash. In contrast with Wang et al. [1999]
results, the luminosity of those upward leaders was comparable to that of their corresponding downward
leaders. One of the UCLs reported by Biagi et al. [2009] is shown in Figure 1a, where a faint streamer fila-
ment is seen between the downward and upward connecting leaders, which is apparently indicative of the
break-through phase being in progress. A similar streamer filament between the downward and upward lead-
ers of a long spark can be seen in Figure 1b [Lebedev et al., 2007]. The common streamer zone is seen more
clearly in a different long spark shown in Figure 1c [Shcherbakov et al., 2007].
Tran et al. [2014], who used a high-speed video camera, reported on a 50 m long faintly luminous formation
(FLF) that was imaged in the pre-return-stroke frame between the downward leader tip of the first stroke in a
natural negative lightning flash and its ground termination point (apparently, similar formations can be seen
in Figure 1a of Jiang et al. [2014] and in Figure 2a of Gao et al. [2014]). This luminous formation was inferred to
be a poorly conducting streamer zone, because it was replaced by a hot, highly conducting plasma channel
at a rate of 7.0 × 105 m/s, which was comparable to the stepped leader speed in the preceding frames, in the
absence of FLF in those frames. No high-luminosity (hot) upward connecting leader channel was observed.
In this paper, we extend the study of Tran et al. [2014] using additional high-speed (HS) video records in con-
junction with electric field records of first and new-ground-termination (NGT) subsequent strokes in negative
natural lightning flashes.The overall sample size in this study is 50, with 18 events exhibiting FLFs. FLF record-
ings were made possible by our use of a relatively long dead time to prevent the occurrence of FLF in the
same frame with the bright return stroke and a relatively large aperture. If we eliminated the dead time in our
system, the number of recorded FLFs would reduce from 18 to 2.

TRAN AND RAKOV LIGHTNING ATTACHMENT PROCESS 6923


Journal of Geophysical Research: Atmospheres 10.1002/2015JD023155

Table 1. Characteristics of the 18 Events Exhibiting FLFsa

Exposure Dead Stroke Peak Current Distanceb H L ΔL Δt v rs1 rs2 Number of Average
Event ID Time (μs) Time (μs) f Number Order Group (kA) (km) (m) (m) (m) (μs) (106 m/s) (m) (m) Step Pulses Step Length (m)
2013-00155 30 370.0 f /2.8 1 A 30 6.2 50 51 0 71 0.72 102 51 0 -
2013-00698 30 247.8 f /2.8 1 A 93 7.9 150 152 0 72 2.1 301 151 2 76
2013-00700c 30 247.8 f /2.8 1 A 26 7.2 86 99 0 86 1.2 196 99 0 -
2013-00701 30 247.8 f /2.8 1 A 64 7.5 179 183 0 92 2.0 360 183 3 61
2013-00737 30 247.8 f /2.8 1 A 30 2.4 61 65 0 42 1.6 123 61 5 13
2013-01277 30 247.8 f /2.8 1 A 37 2.8 126 127 0 85 1.5 251 127 9 14
2014-00406 310 2.5 f /5.6 1 A 70 2.4 192 200 0 95 2.1 383 197 5 40
2014-00741 310 2.5 f /5.6 1 A 38 5.2 147 155 0 106 1.5 307 154 11 14
2014-01506-1 80 232.5 f /4 1 A 28 8.8 62 62 0 63 0.98 124 62 0 -
2014-01513-4 80 232.5 f /4 4 A 27 6.0 60 61 0 49 1.3 115 61 3 20
2014-01514-1 80 232.5 f /4 1 A 59 6.0 180 192 0 88 2.2 386 187 8 24
2013-01269d 30 247.8 f /2.8 1 B 22 0.76 144 40 130 126 1.4 175 88 7 24
2014-01277 100 212.5 f /4 1 B 99 2.0 276 122 175 115 2.6 395 195 4 74
2014-01514-2 80 232.5 f /4 2 B 25 3.4 96 34 64 139 0.71 130 65 10 10
2014-01514-3e 80 232.5 f /4 3 - 67 7.8 148 - - 58 - - - 5 -
2014-01533 80 232.5 f /4 1 C 17 2.0 - - - 63 - - - 6 -
2014-01547 80 232.5 f /4 1 A - - - - - 9 - - - 2 -
2014-01590 80 232.5 f /4 1 C 34 3.5 - - - 89 - - - 9 -
a H, L, ΔL, Δt, v , r , and r are the height of leader tip above the ground termination point, 2-D length of FLF, 2-D gap length, time interval between the end
s1 s2
of exposure of FLF frame and return stroke onset, FLF channel conditioning rate (including the gap for the group B events), and two estimates of striking distance
based on definitions illustrated in Figures 8c and 8d, respectively.
b For 14 events, the length of the semimajor axis of 50% NLDN location error ellipse was 0.2 km, and for two events (2013-00737 and 2014-01533) it was 0.3 km.
c This event struck a 152 m tall metallic tower, while all other events in groups A and B likely terminated on trees.
d Ground termination point of this event was out of the field of view of the HS camera. It was estimated via image extrapolation using three reference objects

in the field of view (FOV) and varying the FOV of the camera.
e It was not clear if the FLF in this stroke had a faint top portion connected to the downward leader tip or did not have connection to the downward leader tip,

which did not allow us to reliably attribute the event to either group A or group B.

2. Instrumentation
The data set was obtained at the Lightning Observatory in Gainesville (LOG), Florida [Rakov et al., 2014], from 4
June 2013 to 2 August 2014. The HS video records were acquired using a Phantom V310 HS video camera with
exposure times of 30 μs (18 records, with 370 μs or 248 μs dead time), 80 μs (17 records, with 232.5 μs dead
time), 100 μs (3 records, with 212 μs dead time), and 310 μs (12 records, with 2.5 μs dead time). Frame rates
ranged from 2500 to 3600 frames per second. The camera was coupled with a Sigma 20 mm lens. The focal
length was 20 mm. The size of each pixel on the HS camera sensor was 20 μm × 20 μm, and the resolution was
mostly 1280 × 800 pixels. The f number (f stop) varied from f /2.8 to f /5.6 (see Table 1), with the available range
being from f /1.8 to f /22 (the smaller the f number, the larger the light gathering area). The camera record
length was about 1.1 s with 100 or 200 ms pretrigger data. Additionally, data from the LOG electric field (low
gain and high gain) and electric field derivative (dE /dt) measuring systems were used. The −3 dB bandwidth
of the low-gain electric field measuring system was 16 Hz to 10 MHz, and for the high-gain system it was from
360 Hz to 10 MHz. The latter system has been available since 5 April 2014. The upper frequency response of
the dE /dt measuring system was 10 MHz. The field record length was 1 s with 100 or 200 ms of pretrigger data.
The instrumentation triggering scheme is described in the supporting information. The HS video camera was
GPS synchronized with the field measuring systems via Inter-Range Instrumentation Group time codes fed
to the camera from a GPS time card which was also used to produce the field-based trigger time stamp. The
synchronization accuracy was determined to be better than 2.4 μs (see supporting information). The lightning
strike locations reported by the National Lightning Detection Network (NLDN) were used to estimate heights,
2-D distances, and 2-D speeds presented in this paper. In the range of distances at which the strokes examined
here occurred (0.8 to 8.8 km) the spatial resolution of the camera varied from 0.8 to 8.8 m.

TRAN AND RAKOV LIGHTNING ATTACHMENT PROCESS 6924


Journal of Geophysical Research: Atmospheres 10.1002/2015JD023155

3. Data Presentation
3.1. Lightning Channel Termination Points
The HS video camera was installed in the glass cupola of LOG (on the roof of a five-story building) and was
pointing west. The terrain in its field of view (FOV) is essentially flat with a number of buildings in the fore-
ground and mostly trees at the distances at which the lightning strikes examined here occurred. It is likely
that most of the strike points were at or near treetops. Escobedo et al. [2012] studied the canopy structure
in the Gainesville area and found that the maximum heights for 10 most common (66% of all trees) species
ranged from 9 m to 30.5 m with the average maximum height being about 18.5 m. The average heights for
those species ranged from 6.1 to 26.5 m. There are also nine tall man-made towers located at distances rang-
ing from about 2 to 8 km in the FOV of the camera. Their heights are between 40 and 152 m. We recorded one
stroke without FLF, terminated on a 76 m tall tower, which is not included in the present study, and one stroke
with FLF terminated on a 152 m tall metallic tower. Characteristics of FLFs associated with the metallic tower
and with trees are similar (see Table 1). We also observed two FLFs developing from towers (see section 4) but
failing to attract/intercept a downward leader branch. Two strokes with FLFs had their ground strike points
obscured by one of the buildings in the camera FOV so that their FLF lengths could not be determined. For
these two events, we assumed that the FLFs were connected to grounded objects (likely trees) behind the
building (the reasons for this assumption will be discussed in section 4). For all other events, the lowest point
of the imaged channel was assumed to be the treetop strike point.
In summary, the majority of strike points were likely the tops of trees with heights less than 30 m, and in one
case it was a 152 m tall metallic tower.
3.2. Overview of Analyzed Events
In the following, we will refer to the frame immediately preceding the frame in which the return stroke is first
seen as the pre-return-stroke frame. The onset of the return stroke mostly occurred during the dead time
between the “pre-return-stroke frame” and the frame immediately after the pre-return-stroke frame. All the
FLFs were recorded in a single, pre-return-stroke frame. We selected 50 stepped leaders of first and NGT sub-
sequent strokes from our data set that satisfied the following three conditions: the NLDN-reported ground
termination point is within 12 km of the LOG (based on our experience, lightning channels at distances greater
12 km are not imaged clearly); the channel termination is visible and appears to be on the tree line (47 cases),
terminated on tower (one case), or obscured by nearby building, but its location is known from NLDN data
(two cases, both with FLFs); and the downward leader tip is visible in the pre-return-stroke frame.
A total of 18 out of the 50 first and NGT subsequent strokes exhibited FLFs below the downward leader tip.
The 18 events (see Table 1) are categorized into three types or groups based on their FLF appearance: group
A (12 events): continuous FLF between the downward leader tip and the prospective strike point; group B
(3 events): FLF apparently extending upward from the prospective strike point but not in contact with the
downward leader tip; and group C (2 events): FLF in contact with the downward leader tip, with the bottom
part of FLF being obscured by a building. Group C will be combined with group A for all the analyses not
requiring the knowledge of FLF length.
For one event, 2014-01514-3, the upper part of the FLF was too faint to decide whether it was in contact with
the downward leader tip or not. Hence, this event was not assigned to any of the three groups and could not
be used for analyses requiring the FLF length. One event in group A, 2014–01547, was missed by the NLDN
so that its FLF length could not be determined.
Examples of FLFs from groups A and B are presented in Figures 2a and 2b, respectively. Examples of electric
field and dE /dt records showing the signatures of processes occurring just prior to the return stroke onset are
found in Figures 3 and 4.
3.3. Characterization of the Observed Events
Characteristics of the 18 events with FLFs are summarized in Table 1. For comparison, we examined 32 events
without FLFs in the pre-return-stroke frame.
The following parameters were determined for each stroke:
1. The height, H, of the downward leader tip in the pre-return-stroke (FLF) frame above the prospective
strike point.

TRAN AND RAKOV LIGHTNING ATTACHMENT PROCESS 6925


Journal of Geophysical Research: Atmospheres 10.1002/2015JD023155

Figure 2. Examples of two types of events examined in this study: (a) group A (continuous FLF between the downward
leader tip and the strike object) and (b) group B (FLF extending from the strike object but not in contact with the
downward leader tip). The two images are enlarged portions of pre-return-stroke frames (exposure times of 80 μs and
100 μs for Figures 2a and 2b, respectively), which were background luminosity removed, intensity inverted, and contrast
enhanced for improved visualization. The images of downward stepped leader channels are saturated, while those of
FLFs are not.

2. The 2-D length, L, of FLF found by adding the lengths of approximating linear segments passing through
brighter pixels of the FLF image.
3. The 2-D gap length, ΔL, measured for group B events between the downward leader tip and the upper
end of FLF along the piecewise approximation of the channel illuminated during the return stroke stage.
For groups A and C, ΔL = 0. For strokes without FLF, ΔL was measured as the 2-D length of the channel
(illuminated by the return stroke) between the downward leader tip seen in the pre-return-stroke frame and
the ground termination point.

Figure 3. (a) Electric field and (b) electric field derivative (dE /dt) records showing the stepped leader to return stroke
transition (see dashed vertical line between 0 and −20 μs) in stroke 2014-01514-1. The slow front and fast transition of
the return stroke electric field waveform are marked SF and FT, respectively. Eight pronounced leader step pulses that
occurred after the end of exposure (see dashed vertical line between −80 and −100 μs) of the FLF (pre-return-stroke)
frame are each marked L.

TRAN AND RAKOV LIGHTNING ATTACHMENT PROCESS 6926


Journal of Geophysical Research: Atmospheres 10.1002/2015JD023155

Figure 4. (a) Electric field and (b) electric field derivative (dE /dt) records showing the stepped leader to return stroke
transition (see dashed vertical line between 0 and −20 μs) in stroke 2014-01277. The slow front and fast transition of the
return stroke electric field waveform are marked SF and FT, respectively. Out of six pronounced leader step pulses
labeled L, four occurred after the end of exposure of the FLF frame marked by the dashed vertical line between −140
and −120 μs.

4. The time interval, Δt, between the end of exposure of the pre-return-stroke (FLF) frame and the return stroke
onset, the latter being identified as the beginning of electric field slow front (see Figures 3a and 4a). For one
event, 2014-00406, an integrated dE /dt record was used to determine the beginning of slow front, since the
electric field record was saturated.
5. FLF channel conditioning rate, v , defined for group A as L∕Δt. For group B, v was defined as (L + ΔL)∕Δt so
that v accounted for conditioning of both the FLF and the gap between the FLF and the downward leader
tip. For group C, v could not be estimated, since L was unknown. For strokes without imaged FLF, v was
computed as ΔL∕Δt.
6. Striking distances, rs1 and rs2 , were estimated using two approaches (both based on FLF observations and,
hence, different from the traditional definitions of this parameter). The former was defined as the 2-D dis-
tance between the downward leader tip and the prospective strike point at the inferred instant of FLF
initiation at the strike point. We determined rs1 by “rolling back,” at the same rate, both the downward leader
channel and the corresponding FLF until the FLF length reduced to zero. Then rs1 was measured between
the bottom (lowest point) of the downward leader channel and the prospective strike point. For 11 events
in group A, rs2 was measured between the downward leader tip and the prospective strike point seen in
the FLF frame. For three events in group B, rs2 was determined by extending, at the same rate, the down-
ward leader and FLF toward each other until they met and then rs2 was measured between their junction
point and the strike point. Definitions of rs1 and rs2 , as well as other, more traditional definitions of striking
distance, are illustrated in Figure 8.
7. Number of pronounced step pulses in dE /dt record in the time window between the end of exposure of
pre-return-stroke (FLF) frame and the beginning of the slow front in the return stroke electric field waveform.
Only pronounced pulses whose peak-to-peak amplitudes exceeded both one third of that of the largest
pulse in the time window and twice the noise level were considered. These pulses (labeled L in Figures 3b
and 4b) were probably produced by steps in the downward negative leader channel. For three events in
group A and for three events without FLFs, no pulses indicative of leader stepping were found in dE /dt
records.
8. Average step length, found for group A as the ratio of L and the number of pronounced dE /dt pulses in the
time window between the end of exposure of pre-return-stroke (FLF) frame and the beginning of the slow
front in return stroke electric field waveform. This crude estimation was done based on the assumption that
all pronounced pulses were associated with steps in the main channel. For group B, L was replaced with
L + ΔL, and for events without FLFs, it was replaced with ΔL.

TRAN AND RAKOV LIGHTNING ATTACHMENT PROCESS 6927


Journal of Geophysical Research: Atmospheres 10.1002/2015JD023155

Figure 5. Four consecutive frame-to-frame downward leader speeds (v12 , v23 , v34 , and v45 ) and the rate of replacement
of FLF and/or virgin-air gap by hot channel (v ). Individual values are shown by circles. Straight lines connect the mean
values. Events with FLFs (groups A and B combined) are represented by red color (left group of circles) and those
without FLF by blue color (right group of circles).

9. We additionally found four consecutive frame-to-frame 2-D speeds, v12 , v23 , v34 , and v45 , of the downward
leader tip prior to the pre-return-stroke (FLF) frame. The subscript in each case indicates the pair of frames
used for speed estimation that are numbered consecutively, starting with the pre-return-stroke frame
(frame 1), in the reverse time direction. These speeds are presented (along with v ) in Figure 5.

4. Results and Discussion


It is likely that all 18 FLFs developed either entirely or largely upward from the strike object as explained below.
First, we never observed FLFs connected to the downward leader tip, but not to an object on ground, while
there were three events (group B) in which FLFs were grounded but not connected to the downward leader
tip. We did occasionally image streamer zones (as well as space stems/leaders) near the downward extending
negative leader branches (see Figures 2b and 6), but they were much shorter than FLFs. Further, the luminos-
ity of 6 (out of 15) FLFs in groups A and B decreased with increasing height, and for the other 9 FLFs there was
no clear trend. Additionally, we observed, in the same pre-return-stroke frame, two concurrent FLFs, involv-
ing separate grounded objects, one connected and the other unconnected to the downward leader channel.
There were three events, 2013-00698, 2014-00406, and 2014-01514-1, each exhibiting both connected and
unconnected FLFs. Two of the unconnected FLFs (in events 2014-00406 and 2014-01514-1) were, respectively,
43 and 71 m long and developed from towers 76 and 58 m in height, located 0.3 and 0.8 km from the cor-
responding strike objects (probably trees) with connected FLFs. One of these unconnected FLFs is shown in
Figure 6. The third unconnected FLF did not involve a tower so that its length and location could not be deter-
mined. Based on the above, we conclude that all FLFs developed upward from ground and assume that the
two events with partially obscured FLFs in group C were in contact with objects on ground, similar to events
in group A, although those two events could have involved significant bright UCLs. For this reason, we have
combined the events in groups A and C for any analysis which does not require a knowledge of FLF length or
leader tip height.
Brightness of 17 out of 18 FLFs was significantly lower than that of the downward stepped leaders, with most
of the leader images being saturated. Only in one event, shown in Figure 2b, the bottom section (11 m in
length) of 122 m long FLF appeared to be as bright as the downward leader channel, possibly because that

TRAN AND RAKOV LIGHTNING ATTACHMENT PROCESS 6928


Journal of Geophysical Research: Atmospheres 10.1002/2015JD023155

bottom section was actually a UCL. Even in


this event, none of the FLF pixels was satu-
rated. As noted above, UCLs could poten-
tially be present in the group C events.
Based on luminosity, in the overwhelming
majority of cases, there was no evidence
of bright (hot) UCL in group A events.
It could be present, but too short to be
seen in the FLF frame, developed during
the dead time and overwhelmed by the
return stroke in the next frame. Indeed,
as stated in section 2, the spatial reso-
lution of the camera varied from 0.8 to
8.8 m in the range of distances at which
the events examined here occurred. Dis-
tances for individual events are given in
Table 1. We will show in the following
Figure 6. Event 2014-00406 showing both (left) connected and (right)
unconnected FLFs in the same pre-return-stroke frame. The sections that FLFs were mostly composed
unconnected FLF developed from the 76 m tower, which was located of low-conductivity streamers, as opposed
2.7 km from the LOG. The distance between the connected and to hot, leader-like channels capable of
unconnected FLFs was 0.3 km. supporting return stroke waves.

4.1. Leader Dynamics Versus Presence of FLF


Shown in Figure 5 are ranges of variation of frame-to-frame speeds, v12 , v23 , v34 , and v45 , as well as those of
the rate of replacement, v , of FLF and/or virgin-air gap by a hot channel for events with (groups A and B are
combined) and without FLFs. There is an increasing trend with decreasing leader tip height for mean and
median values of speed (starting with v34 ). The mean values of v for events with and without FLFs are similar,
while the median value for events with FLF is larger (1.5 versus 0.93 ×106 m/s). The ranges of variation of v12
and v for events without FLFs are considerably larger. It is possible and even likely that in the events without
imaged FLFs such formations did occur but escaped recording due to either occurring during the camera
dead time and/or sharing frame with the much brighter return stroke. This can also explain why FLFs were
not observed in most previous HS video studies of lightning.
4.2. Return Stroke Onset Time
In this paper, we assume that the return stroke process starts at the beginning of slow front in its electric field
signature (see Figures 3 and 4). This assumption is based on the following. Jerauld et al. [2007], from their anal-
ysis of measured electric and magnetic fields of a triggered lightning stroke and modeling, suggested that the
slow front is produced by a pair of current waves propagating in opposite directions from the junction point
of downward and upward leaders, as opposed to being due to a UCL. For an anomalous triggered lightning
flash, Wang et al. [2014] found that the two return stroke current waves were initiated at a height of about
23 m above the strike object (at the junction point between the downward and upward connecting leaders),
and the initiation was associated with the beginning of the slow front. Wang et al. [2014] concluded that both
the slow front and fast transition “start at the return stroke initiation height.”
Based on the above, we assume that the beginning of slow front (identified by the initial deflection of electric
field from the preceding leader level; see Figures 3a and 4a) is associated with the initiation of bidirectional
return stroke process. Both return stroke current waves propagate along the hot leader channels and, hence,
are expected to move at speeds of the order of 108 m/s [Jerauld et al., 2007], although Wang et al. [2014]
reported a downward speed of as low as 2.2 × 107 m/s (considerably lower than the corresponding upward
speed) for the first return stroke in the anomalous triggered lightning flash, while for subsequent strokes those
two speeds were comparable. D. Wang (personal communication, 2014) found that downward and upward
speeds for one first stroke in natural lightning were similar, of the order of 108 m/s, and for another one a
lower, of the order of 107 m/s, downward speed was observed.
4.3. Replacement of FLF by Channel Capable of Guiding Return Stroke Wave
We measured the time intervals between the end of exposure of pre-return-stroke frame and the begin-
ning of slow front for all the 50 strokes examined here. Out of the 50 events, 18 (36%) exhibited FLFs in

TRAN AND RAKOV LIGHTNING ATTACHMENT PROCESS 6929


Journal of Geophysical Research: Atmospheres 10.1002/2015JD023155

pre-return-stroke frames. For the entire data set of 50 events, the time intervals ranged from 9 μs to 395 μs.
The upper and lower limits of time intervals for the 18 events with FLF were 9 and 139 μs, respectively. As
noted above, FLFs probably occurred in all the events but could not be imaged for all of them.
We now examine, using 11 group A events with known 2-D FLF lengths, the rate at which FLF is replaced by
a hot, high-conductivity channel. All these events exhibited FLFs connecting the downward leader tip and
the strike object in the pre-return-stroke frame, but the return stroke process started only tens of microsec-
onds after the end of exposure of that frame. The corresponding rate at which the FLF was “conditioned” to
become a channel capable of supporting propagation of return stroke waves ranged from 0.72 ×106 m/s to
2.2 ×106 m/s, with a mean value of 1.6 × 106 m/s. The latter value is only within a factor of 2 of the mean
stepped leader speeds determined from the preceding frames, which were not influenced by the presence of
FLF. This point is illustrated in Figure 5. Also, the FLF “conditioning” rates are at least an order of magnitude
lower than the speeds of downward return stroke waves traversing UCL channels [Wang et al., 2014; D. Wang,
personal communication, 2014], which argues against the notion that FLFs could be viewed as UCLs.
Further, for 9 out of 12 group A events, we found pronounced stepped leader pulses in dE /dt records occurring
during the time interval between the end of exposure of FLF frame and the return stroke onset. This implies
that the downward negative leader continued stepping while traversing the FLF. The number of pronounced
steps varies from 2 to 11. The average number of steps for groups A and C combined is 6. In fact, some of the
pronounced dE /dt pulses could have been produced by other (not involving FLF) branches of the stepped
leader. However, in view of (a) relatively regular interpulse intervals and (b) their similarity to typical interstep
intervals near ground, we believe that most of the pronounced pulses were produced by steps in the main
channel. Note that evidence of stepping (a space stem/leader) in the streamer zone is seen in Figure 1a.
The relatively low FLF conditioning rate and involvement of stepping process suggest that the FLFs mostly
consist of relatively low conductivity streamers, as opposed to being hot, leader-like channels. We cannot rule
out the possibility that FLFs were transient (disappear after the end of exposure of the FLF frame), but the fact
that they did determine the strike point many tens of microseconds prior to the return stroke onset argues
against this scenario.
It is interesting that FLFs are capable of guiding the stepped leaders to the FLF origin on ground while not
significantly influencing leader dynamics (propagation speed and stepping). Since a connection to ground is
already established (via FLF and possibly short UCL), the process can be also viewed as an abnormally long
duration (tens of microseconds) break-through phase during which the downward leader continues to move
in a stepwise fashion (along with a UCL) through the common streamer zone. As noted above, there is a space
stem (or space leader) in the common streamer zone in Figure 1a, which can be indicative of step formation
processes during the break-through phase.
Larsson [1998] and Larsson et al. [1998] studied “inhibited” laboratory discharges in rod-to-plane gaps (up to
1 m) subjected to a switching impulse and observed the unusually long duration break-through phase, which
they attributed to the large resistance (up to 4.7 MΩ) connected in series with the rod electrode. For a 1 m gap
and 1 MΩ series resistance, the break-through phase duration was about 55 μs versus a few microseconds for
“normal” discharges. Inhibited discharges began with streamers developing from the rod and bridging the
gap. As a result, the break-through (final-jump) condition was established and the leader (or leaders) crossed
the gap causing the voltage across the gap to drop. Larsson [1998] found that a large series resistance did not
have any effect on streamer development from the rod electrode but inhibited the formation of the leader,
which led to a considerably longer break-through phase. It is not clear whether their laboratory results are
applicable to lightning FLF observations reported here. One of the reviewers suggested that trees might have
presented megaohm range resistance values that did not influence the development of streamers (FLF) but
delayed, by tens of microseconds, the completion of break-through phase. However, our observation of FLF
from the 152 m tall metallic tower (see Figure 7), whose characteristics are similar to those initiated from
trees (see Tables 1 and 2), argues against this hypothesis. On the other hand, there might be other factors
making our events with FLFs kind of inhibited discharges, regardless of whether the strike object is a tree or a
metallic tower.

4.4. Average Electric Field Along FLF


We estimated the leader tip electric potential for the 11 events with known FLF length in group A,
using NLDN-reported peak currents and theoretically predicted current-potential relationships found in the

TRAN AND RAKOV LIGHTNING ATTACHMENT PROCESS 6930


Journal of Geophysical Research: Atmospheres 10.1002/2015JD023155

literature. Then we used the potential and


FLF length to estimate the average elec-
tric field along the FLF. Bazelyan et al.
[1978, Figure 4.11] showed the theoretically
predicted relations of return stroke peak
current versus charge density and charge
density versus leader tip potential, which
we combined to estimate potentials from
peak currents, the results being given in
Table 2. We also used the equation of
Cooray [2013], based on the Self-Consistent
Leader Inception and Propagation Model:
V = 3 × I 0.813 ,
where V is the leader tip potential in mega-
volts and I is the return stroke peak current
in kiloamperes. The latter equation pre-
dicts a steeper dependence of V on I than
the method of Bazelyan et al. [1978] (see
Figure 7. Event 2013-00700 showing FLF initiated from a 152 m Table 2).
tall metallic tower, which was located 7.2 km from LOG. Its FLF
characteristics are similar to those for the events terminating on For each event, the two approaches gave
trees (see Table 1). This event produced another termination reasonably close average electric field val-
(probably on a tree) several kilometers from the tower. ues ranging (for all results combined) from
0.29 to 0.98 MV/m. The mean values for the
two methods are 0.54 and 0.59 MV/m.
The average electric fields in negative and positive streamer zones are 1 and 0.5 MV/m, respectively [Bazelyan
and Raizer, 2000, p. 48], while for a hot leader channel it is expected to be less than 0.01 MV/m (10 kV/m)
[Bazelyan et al., 1978, p. 98]. Our estimated mean values essentially rule out the possibility that FLFs are
hot (UCL) channels. They are close to the average field for positive streamers, possibly suggesting that the
streamers comprising FLF are predominantly of positive polarity, which would be consistent with our assump-
tion that FLFs extended from ground toward the approaching downward negative leader. It is known from
long-spark experiments that positive streamers are considerably longer than negative ones.
As noted above, the average field values for individual events in Table 2 vary in a rather wide range. Lower
than 0.5 MV/m values could be indicative of the presence of UCL channel (although we found no clear optical

Table 2. Downward Leader Tip Potentials and Average Electric Fields Along FLF for 11 Events of Group A With Known FLF Length

Leader Tip Potential (MV) Average Electric Field (MV/m)

Event ID Peak Current (kA) Bazelyan et al. [1978] Cooray [2013] 2-D FLF Length (m) Bazelyan et al. [1978] Cooray [2013]
2013-00155 30 50 47 51 0.98 0.93
2013-00698 93 61 119 152 0.40 0.79
2013-00700c 26 49 42 99 0.49 0.43
2013-00701 64 58 88 183 0.32 0.48
2013-00737 30 50 47 65 0.76 0.72
2013-01277 37 53 57 127 0.41 0.45
2014-00406 70 59 95 200 0.29 0.48
2014-00741 38 53 57 155 0.34 0.37
2014-01506-1 28 50 45 62 0.80 0.72
2014-01513-4 27 49 44 61 0.80 0.71
2014-01514-1 59 57 83 192 0.30 0.43
Mean 46 53 66 123 0.54 0.59
a This event struck a 152 m tall metallic tower, while all others terminated on trees.

TRAN AND RAKOV LIGHTNING ATTACHMENT PROCESS 6931


Journal of Geophysical Research: Atmospheres 10.1002/2015JD023155

evidence of significant UCL in group A, possibly due to limited spatial resolution of the camera), and higher
(closer to 1 MV/m) values could suggest a significant contribution from negative streamers.
Perhaps the most important result of this section is that the estimated average electric fields along FLFs
are 1 to 2 orders of magnitude higher than expected for hot, leader-like channels. This finding supports the
inference made in section 4.3 that FLFs mostly consist of relatively low conductivity streamers.
4.5. Lightning Striking Distance
The attachment of the leader to the strike object is often described using the so-called electrogeometrical
model, the core of which is the concept of “striking distance.” This concept obscures some of the significant
physics but allows the development of relatively simple and useful techniques for designing lightning pro-
tection systems for various structures. In general, the striking distance is the distance from the descending
leader tip to the prospective strike point at the instant this strike point is uniquely determined. Depending
on the interpretation or assumed model, different specific definitions of the striking distance were proposed.
For example, it can be defined as the distance from the leader tip to the object to be struck at the instant
when an upward connecting leader (UCL) is initiated from this object [e.g., Golde, 1945]. Another definition
requires the establishment of the break-through condition (common streamer zone) between the tips of the
descending leader and the UCL. The latter definition is more appropriate for tall objects, when UCLs of sig-
nificant length are expected [Rakov and Lutz, 1990]. For objects shorter than 30 m or so, the presence of UCL
can be neglected in estimating the striking distance [Cooray et al., 2014]. The traditional definitions of rs are
illustrated in Figures 8a and 8b.
For all the events summarized in Table 1, the presence of FLF invariably determined the lightning strike point.
Therefore, FLFs can be used (similar to UCLs) for defining the striking distance. This approach differs from
the traditional ones noted above but is consistent with the general definition of striking distance. The two
definitions of striking distance based on our observations of FLFs are illustrated in Figures 8c and 8d. It is
worth noting that our definitions (although applicable only to events with FLFs) are based on experimen-
tal data for natural lightning, while the traditional definitions primarily rely on data for laboratory sparks
and/or modeling.
We labeled the striking distances based on the two definitions illustrated in Figures 8c and 8d as rs1 and
rs2 , respectively.
The first one, rs1 , was measured between the downward leader tip and the strike point at the inferred instant
of FLF initiation. That instant depends on the FLF extension speed which is unknown, since all FLFs were
each imaged in a single frame. We assume that the FLFs were primarily composed of positive streamers.
Positive streamers in laboratory were observed to extend at a speed of 105 m/s or so [Cabrera and Cooray,
1992]. For example, Larsson [1998] reported 2 × 105 m/s for the integrated mean electric field strength of
0.29–0.34 MV/m, which is about an order of magnitude lower than v12 , the downward negative leader speed
just before the FLF frame (see Figure 5). We assumed that the leader and FLF extension speeds are equal,
which probably resulted in a lower bound on rs1 . As noted in section 3.3, we determined rs1 by rolling back,
at the same rate, both the downward leader channel and the corresponding FLF until the FLF length reduced
to zero. Then rs1 was measured between the bottom (lowest point) of the downward leader channel and the
prospective strike point.
For 11 events in group A, rs2 was measured between the downward leader tip and the prospective strike point
seen in the FLF frame. For three events in group B, rs2 was determined by extending, at the same rate, the
downward leader and FLF toward each other until they met and then rs2 was measured between their junction
point and the strike point.
Figures 9a and 9b show rs1 and rs2 , respectively, each as a function of NLDN-reported peak current I. In both
figures, the blue curve labeled rs = 10×I0.65 , where I is in kiloamperes and rs is in meters, represents the striking
distance recommended by the International Lightning Protection Standard [International Electrotechnical
Commission (IEC), 2010] and is shown for reference. It is thought that the International Electrotechnical
Commission (IEC)-recommended equation corresponds to rs definition illustrated in Figure 8b, but with the
UCL having an insignificant (negligible) length [Cooray et al., 2014]. The median NLDN current estimation error
(absolute value) is 14% for subsequent strokes [Mallick et al., 2014] and is unknown but expected to be similar
for first strokes. This uncertainty is unlikely to alter our inferences based on Figures 9a and 9b.

TRAN AND RAKOV LIGHTNING ATTACHMENT PROCESS 6932


Journal of Geophysical Research: Atmospheres 10.1002/2015JD023155

Figure 8. Illustration of different definitions of striking distance. It is assumed that the downward leader tip is at striking
distance from the prospective strike point when (a) UCL is initiated at that point, (b) break-through phase begins, (c) FLF
is initiated, and (d) FLF comes in contact with the downward leader. Figure 8c is similar to Figure 8a with UCL replaced
by FLF. Figure 8d is similar to Figure 8b with an insignificant UCL. In Figure 8a streamers are not shown because they are
not part of this rs definition. FLFs in Figures 8c and 8d mostly consist of streamers but may contain a short UCL.

All values of rs1 and rs2 are likely to be underestimates, which is due to the following two reasons: (a) the
FLF extension speed is likely to be lower than the downward leader speed near ground and (b) our 2-D dis-
tances are shorter than the actual ones. For 11 group A events (see blue circles and triangles in Figures 9a
and 9b), there is an additional reason for underestimation of rs1 and rs2 : it is possible that the downward
leader has already traversed a portion of FLF (the length of imaged FLF is shorter than the FLF length at the
instant it first came in contact with the downward negative leader). It appears from Figures 9a and 9b that the
IEC-recommended striking distance is in agreement with rs2 and considerably smaller than rs1 . For the events
examined in this paper, the common streamer zone (FLF connected to the downward leader tip) was estab-
lished (see Figure 8d) tens of microseconds prior to the return stroke onset, and the FLF was launched from
the strike object (see Figure 8c) up to 300 μs or so prior to the return stroke onset (if the FLF extension speed
is assumed to be equal to v12 ). For comparison, the establishment of common streamer zone in Figure 8b is
expected (based on long-spark observations) to be followed by the return stroke within a few microseconds.
Wang et al. [2013] examined the striking distance as a function of return stroke peak current for 14 classical
triggered lightning strokes (from 2011 experiments at Camp Blanding) terminating on a 5 m high launch-
ing structure and found the resultant curve to be considerably lower than the one recommended by the IEC

TRAN AND RAKOV LIGHTNING ATTACHMENT PROCESS 6933


Journal of Geophysical Research: Atmospheres 10.1002/2015JD023155

Figure 9. Striking distances: (a) rs1 (see Figure 8c) and (b) rs2 (see Figure 8d), each as a function of peak current, for 14
events with FLFs. Blue and red colors mark events from groups A and B, respectively. Circles correspond to the events
striking trees, and the blue triangles correspond to the event striking a 152 m tall metallic tower. Upward directed
arrows indicate that the striking distances for group A could be actually larger, because the downward leader could have
already propagated along a portion of FLF. The curves (shown as reference) are the striking distance as a function of
peak current predicted by the equation (based on the definition shown in Figure 8b with UCL neglected), rs = 10 × I0.65 ,
recommended by the International Lightning Protection Standard [IEC, 2010]. R is the linear correlation coefficient.

Standard. In a follow-up study at Camp Blanding in 2012, Wang et al. [2014] obtained similar results for seven
strokes in the anomalous triggered lightning flash that terminated on a 10 m high pole. In the latter case, the
peak currents were estimated from the corresponding peaks of luminosity pulses based on the relationship
obtained by Wang et al. [2013] for classical triggered lightning strokes for which direct current measurements
were available. In both studies, a photoelectric measuring system Lightning Attachment Process Observation
System (LAPOS) was used and the speeds of the downward leader and UCL were assumed to be equal. On the
other hand, for the 11 classical triggered lightning strokes recorded with HS video cameras at Camp Bland-
ing in 2008 and 2009 (with launching facility heights of 17 and 15 m, respectively, and imaged UCL lengths
ranging from 5 to 74 m) [Biagi, 2011], 6 out of 11 values of similarly defined striking distance were above the
IEC Standard curve. Gamerota et al. [2015], for three triggered lightning strokes recorded during the 2013 and
2014 Camp Blanding campaigns, reported UCLs whose length in each case appeared to be less than 1 m. The
corresponding peak currents were 22, 17, and 12 kA, similar to those in 2008 and 2009. The launching facility
height was 5.7 m, similar to that in 2011, but lower than those in 2008 and 2009. The reasons for the discrepan-
cies among the different triggered lightning data sets are presently unknown. Note that triggered lightning
strokes usually develop in previously conditioned (warm-air) channels, as opposed to virgin (cold) air in the
case of first strokes which are the subject of this study.
The IEC standard formula is subject to large uncertainties related to a number of assumptions involved and a
large scatter in the experimental charge-current relation used to express the striking distance in terms of peak
current [e.g., Rakov, 2012]. In this study, we show, for the first time, that an upward extending formation mostly
consisting of relatively low conductivity positive streamers, up to 200 m in length, can in effect intercept a
descending negative leader with the resultant break-through phase having unexpectedly long duration of
some tens of microseconds. Interestingly, the size of this streamer formation is comparable to the striking
distance predicted by the IEC standard formula (see Figures 8d and 9b). Clearly, the lightning attachment
process is in need of further research.

5. Summary
1. We examined 50 first and new-ground-termination subsequent strokes and found that 18 (36%) of them
exhibited FLFs apparently extending from ground toward the approaching negative leaders. In all cases,
these formations determined the ground strike point well (typically tens of microseconds) before the return
stroke onset.

TRAN AND RAKOV LIGHTNING ATTACHMENT PROCESS 6934


Journal of Geophysical Research: Atmospheres 10.1002/2015JD023155

2. Eleven located group A FLFs, ranging from 51 to 200 m in length, were replaced by channels capable of
guiding return stroke waves at rates ranging from 0.72 ×106 m/s to 2.2 ×106 m/s, which are comparable to
the average frame-to-frame speed v12 , 0.74 × 106 m/s, immediately prior to the FLF frame and at least an
order of magnitude lower than the speeds of downward return stroke waves traversing upward connecting
leader channels.
3. The overwhelming majority of downward leaders continued stepping (exhibiting 2 to 11 pronounced steps)
in the presence of FLFs, which suggests that the latter were not hot, leader-like channels.
4. The estimated average electric fields along FLFs ranged from 0.29 to 0.98 MV/m. The mean values for the
two methods used to estimate the leader tip potential were 0.54 and 0.59 MV/m, 1 to 2 orders of magnitude
higher than expected for a hot leader channel.
5. Based on the findings listed in items 2, 3, and 4 above, we infer that the observed FLFs were composed
mostly of low-conductivity streamers. It appears that the streamers were of positive polarity and developed
from the strike object upward over distances ranging from tens to hundreds of meters.
6. Since the strike points were uniquely determined by FLFs in our data set, we proposed two new defini-
tions of striking distance, based on our FLF observations, and compared them with the striking distance
recommended by the IEC Standard.

Acknowledgments References
This work was supported in part by
NSF and DARPA. Shreeharsh Mallick Aleksandrov, N. L., E. M. Bazelyan, R. B. Carpenter Jr., M. M. Drabkin, and Y. P. Raizer (2001), The effect of coronae on leader initiation and
significantly contributed to the LOG development under thunderstorm conditions and in long air gaps, J. Phys. D Appl. Phys., 34(22), 3256–3266.
setup. The authors would like to thank Bazelyan, E. M., and Y. P. Raizer (2000), Lightning Physics and Lightning Protection, IOP Publ., Bristol, U. K.
Amitabh Nag and William Brooks of Bazelyan, E. M., B. N. Gorin, and V. I. Levitov (1978), Physical and Engineering Fundamentals of Lightning Protection, 223 pp., Gidrometeoizdat,
Vaisala Inc. for providing the NLDN Leningrad, Russia.
data and Daohong Wang for sharing Bazelyan, E. M., Y. P. Raizer, N. L. Aleksandrov, and F. D’Alessandro (2009), Corona processes and lightning attachment: The effect of wind
his unpublished LAPOS results. Helpful during thunderstorms, Atmos. Res., 94(3), 436–447, doi:10.1016/j.atmosres.2009.07.002.
comments of Douglas M. Jordan, Becerra, M., and V. Cooray (2006), A simplified physical model to determine the lightning upward connecting leader inception, IEEE Trans.
Danyal A. Petersen, and three Power Delivery, 21(2), 897–908, doi:10.1109/TPWRD.2005.859290.
anonymous reviewers are greatly Berger, K. (1977), The Earth flash, in Lightning, vol. 1, edited by R. H. Golde, chap. 5, pp. 119–190, Academic, Orlando, Fla.
appreciated. This work complies Biagi, C. J. (2011), Observations and modeling of processes in artificially initiated (triggered) lightning, PhD thesis, Electrical and Computer
with the AGU data policy; contact Engineering Department, Univ. of Florida, Gainesville, Fla.
the authors (manhtran@ufl.edu and Biagi, C. J., D. M. Jordan, M. A. Uman, J. D. Hill, W. H. Beasley, and J. Howard (2009), High-speed video observations of rocket-and-wire
rakov@ece.ufl.edu) for data availability. initiated lightning, Geophys. Res. Lett., 36, L15801, doi:10.1029/2009GL038525.
Cabrera, V. M., and V. Cooray (1992), On the mechanism of space charge generation and neutralization in a coaxial cylindrical configuration
in air, J. Electrost., 28(2), 187–196, doi:10.1016/0304-3886(92)90070-A.
Cooray, V. (2013), On the attachment of lightning flashes to grounded structures with special attention to the comparison of SLIM with CVM
and EGM, J. Electrost., 71(3), 577–581, doi:10.1016/j.elstat.2012.11.035.
Cooray, V., U. Kumar, F. Rachidi, and C. A. Nucci (2014), On the possible variation of the lightning striking distance as assumed in the IEC
lightning protection standard as a function of structure height, Electr. Power Syst. Res., 113, 79–87, doi:10.1016/j.epsr.2014.03.017.
Escobedo, F., J. A. Seitz, and W. Zipperer (2012), Gainesvilles Urban Forest Structure and Composition. School of Forest Resources and
Conservation, Florida Cooperative Extension Service, Institute of Food and Agricultural Sciences, University of Florida, Gainesville, Fla.
[Available at http://edis.ifas.ufl.edu/pdffiles/FR/FR27600.pdf, accessed April 2009.]
Gamerota, W. R., M. A. Uman, J. D. Hill, and D. M. Jordan (2015), Observations of corona in triggered dart-stepped leaders, Geophys. Res. Lett.,
42(6), 1977–1983, doi:10.1002/2014GL062911.
Gao, Y., W. Lu, Y. Ma, L. Chen, Y. Zhang, X. Yan, and Y. Zhang (2014), Three-dimensional propagation characteristics of the upward connecting
leaders in six negative tall-object flashes in Guangzhou, Atmos. Res., 149, 193–203, doi:10.1016/j.atmosres.2014.06.008.
Golde, R. H. (1945), The frequency of occurrence and the distribution of lightning flashes to transmission lines, Electr. Eng., 64(12), 902–910,
doi:10.1109/EE.1945.6441405.
International Electrotechnical Commission 62305-1 (2010), Protection Against Lightning—Part 1: General Principles, 1 ed., International
Electrotechnical Commission, Geneva, Switzerland.
Jerauld, J., M. A. Uman, V. A. Rakov, K. J. Rambo, and G. H. Schnetzer (2007), Insights into the ground attachment process of natural lightning
gained from an unusual triggered-lightning stroke, J. Geophys. Res., 112, D13113, doi:10.1029/2006JD007682.
Jiang, R., X. Qie, M. Liu, G. Lu, Z. Wang, Z. Sun, X. Wu, H. Zhang, K. Liu, and X. Li (2014), High speed video observation of lightning
attachment process and the associated leader behaviors. XV International Conference on Atmospheric Electricity, oral presentation
O-03-12, Norman, Okla. [Available at http://www.nssl.noaa.gov/users/mansell/icae2014/preprints/Jiang_118.pdf, accessed 15–20 June.]
Larsson, A. (1998), An experimental study of inhibited electrical discharges in air, J. Phys. D Appl. Phys., 31(15), 1823,
doi:10.1088/0022-3727/31/15/010.
Larsson, A., A. Bondiou-Clergerie, and I. Gallimberti (1998), Numerical modelling of inhibited electrical discharges in air, J. Phys. D Appl. Phys.,
31(15), 1831, doi:10.1088/0022-3727/31/15/011.
Lebedev, V. B., G. G. Feldman, B. N. Gorin, Y. V. Shcherbakov, V. S. Syssoev, V. A. Rakov, M. A. Uman, and R. C. Olsen (2007), Test of the image
converter cameras complex for research of discharges in long air gaps and lightning, paper presented at 13th International Conference
on Atmospheric Electricity, Beijing, China, 13–17 Aug.
Mallick, S., V. A. Rakov, J. D. Hill, T. Ngin, W. R. Gamerota, J. T. Pilkey, C. J. Biagi, D. M. Jordan, M. A. Uman, J. A. Cramer, and A. Nag (2014),
Performance characteristics of the NLDN for return strokes and pulses superimposed on steady currents, based on rocket-triggered
lightning data acquired in Florida in 2004–2012, J. Geophys. Res. Atmos., 119(7), 3825–3856, doi:10.1002/2013JD021401.
Rakov, V. A. (2012), Lightning discharge and fundamentals of lightning protection, J. Lightning Res., 4, 3–11.
Rakov, V. A. (2013), The physics of lightning, Surv. Geophys., 34(6), 701–729, doi:10.1007/s10712-013-9230-6.

TRAN AND RAKOV LIGHTNING ATTACHMENT PROCESS 6935


Journal of Geophysical Research: Atmospheres 10.1002/2015JD023155

Rakov, V. A., and A. O. Lutz (1990), A new technique for estimating equivalent attractive radius for downward lightning flashes, paper 2.2
presented at 20th International Conference on Lightning Protection, Swiss Electronical Assoc., Interlaken, Switzerland.
Rakov, V. A., and M. A. Uman (2003), Lightning: Physics and Effects, Cambridge Univ. Press, New York.
Rakov, V. A., S. Mallick, A. Nag, and V. B. Somu (2014), Lightning Observatory in Gainesville (LOG), Florida: A review of recent results, Electr.
Power Syst. Res., 113, 95–103, doi:10.1016/j.epsr.2014.02.037.
Shcherbakov, Y. V., V. B. Lebedev, V. A. Rakov, G. G. Feldman, B. N. Gorin, V. S. Syssoev, and M. A. Karpov (2007), High-speed optical studies of
the long sparks in very transient stages, in International Congress on High-Speed Photography and Photonics, vol. 6279, edited by X. Hou,
W. Zhao, and B. Yao, pp. 837–845, SPIE, Bellingham, Wash.
Tran, M. D., V. A. Rakov, and S. Mallick (2014), A negative cloud-to-ground flash showing a number of new and rarely observed features,
Geophys. Res. Lett., 41(18), 6523–6529, doi:10.1002/2014GL061169.
Wang, D., V. A. Rakov, M. A. Uman, N. Takagi, T. Watanabe, D. E. Crawford, K. J. Rambo, G. H. Schnetzer, R. J. Fisher, and Z. I. Kawasaki (1999),
Attachment process in rocket-triggered lightning strokes, J. Geophys. Res., 104, 2143–2150.
Wang, D., N. Takagi, W. R. Gamerota, M. A. Uman, J. D. Hill, and D. M. Jordan (2013), Initiation processes of return strokes in rocket-triggered
lightning, J. Geophys. Res. Atmos., 118(17), 9880–9888, doi:10.1002/jgrd.50766.
Wang, D., W. R. Gamerota, M. A. Uman, N. Takagi, J. D. Hill, J. Pilkey, T. Ngin, D. M. Jordan, S. Mallick, and V. A. Rakov (2014), Lightning
attachment processes of an anomalous triggered lightning discharge, J. Geophys. Res. Atmos., 119, 1524–1533,
doi:10.1002/2013JD020787.

TRAN AND RAKOV LIGHTNING ATTACHMENT PROCESS 6936

You might also like