ROCKET-TRIGGERED LIGHTNING
By
VINOD JAYAKUMAR
Copyright 2004
by
Vinod Jayakumar
ACKNOWLEDGMENTS
I thank Dr. Vladimir Rakov for his infinite patience, guidance, and support
throughout my graduate studies. I would like to thank Dr. Martin Uman and Dr. Doug
Jordan for their valuable suggestions during the weekly lightning conference. I thank Dr.
Megumu Miki for responding to all my questions. I sincerely thank Jason Jerauld, Jens
Schoene, Rob Olsen, Venkateshwararao Kodali, and Brian De Carlo for helping me with
the data and software, and for other innumerable favors (without which I would not have
been able to complete my thesis). Research in my thesis was funded in part by the
National Science Foundation.
iii
TABLE OF CONTENTS
page
ACKNOWLEDGMENTS ................................................................................................. iii
LIST OF TABLES............................................................................................................. vi
LIST OF FIGURES .......................................................................................................... vii
ABSTRACT..................................................................................................................... xiii
1
INTRODUCTION ........................................................................................................1
LIST OF REFERENCES.................................................................................................120
BIOGRAPHICAL SKETCH ...........................................................................................123
LIST OF TABLES
Table
page
vi
LIST OF FIGURES
Figure
page
3-5: Comparison of magnitudes of the vertical electric field peaks measured with
Pockels sensors and a flat-plate antenna, both at 5 m ..............................................35
3-6: Experimental setup .....................................................................................................36
3-7: V-shaped electric field signatures ..............................................................................37
3-8: Stroke S0013-1. ..........................................................................................................40
3-9: Scatter plot of screen current, IS vs. strike rod current, IR, for 2000 ..........................41
3-10: Time variation of electric field, current, power, and energy for stroke S006-4. .....42
3-11: Same as Figure. 3-10, but for Stroke S008-4 ...........................................................43
3-12: Same as Figure. 3-10, but for Stroke S0013-1. ........................................................43
3-13: Same as Figure. 3-10, but for Stroke S0013-4 .........................................................44
3-14: Same as Figure. 3-10, but for Stroke S0015-2 .........................................................44
3-15: Same as Figure. 3-10, but for Stroke S0015-4. ........................................................45
3-16: Same as Figure. 3-10, but for Stroke S0015-6 .........................................................46
3-17: Same as Figure. 3-10, but for Stroke S0023-3. ........................................................46
3-18: Histogram of peak current for strokes characterized by V- shaped electric field
signatures with ERS = EL. .....................................................................................48
3-19: Histogram of EL for strokes characterized by V- shaped electric field signatures
with ERS = EL. ......................................................................................................49
3-20: Histogram of peak power for strokes characterized by V- shaped electric field
signatures with ERS = EL. .....................................................................................50
3-21: Histogram for input energy for strokes characterized by V- shaped electric field
signatures with ERS = EL. .....................................................................................51
3-22: Histogram for action integral for strokes characterized by V- shaped electric field
signatures with ERS = EL. .....................................................................................52
3-23: Histogram of the risetime of current for strokes characterized by V-shaped
electric field signatures with ERS = EL.................................................................53
3-24: Histogram of the 0-100 % risetime of power per unit length for strokes
characterized by V-shaped electric field signatures with ERS = EL. ....................53
viii
3-25: Peak power vs. peak current for strokes characterized by V- shaped electric field
signatures with ERS = EL ......................................................................................54
3-26: Energy vs. peak current for strokes characterized by V- shaped electric field
signatures with ERS = EL ......................................................................................55
3-27: Peak power vs. EL for strokes characterized by V- shaped electric field signature
with ERS = EL. ......................................................................................................56
3-28: Energy vs. EL for strokes characterized by V- shaped electric field signature
with ERS = EL. ......................................................................................................56
3-29: Energy vs. Action Integral for strokes characterized by V- shaped electric field
signature with ERS = EL........................................................................................57
3-30: Flash 0013, stroke 1; Correction factor of 1.6 is applied at the instant of negative
E-field peak ..............................................................................................................59
3-31: Flash 0013, stroke 1; Correction factor of 1.6 is applied at the instant of negative
E-field peak, which is shifted by 0.24 s to the left in order to partially account
for the 0.5 s uncertainty in the position of the peak............................................60
3-32: Flash 0013, stroke 1; Correction factor of 1.6 is applied at the instant of negative
E-field peak, which is shifted by 0.24 s to the right in order to partially account
for the 0.5 s uncertainty in the position of the peak............................................61
3-33: Evolution of the various quantities for the first 0.58 s for Flash S006, stroke 4....64
3-34: Evolution of the various quantities for the first 0.4 s for Flash S008, stroke 4......65
3-35: Evolution of the various quantities for the first 1.4 s for Flash S0013, stroke 1....66
3-36: Evolution of the various quantities for the first 1.2 s for Flash S0013, stroke 4....67
3-37: Evolution of the various quantities for the first 2.1 s for Flash S0015, stroke 2....68
3-38: Evolution of the various quantities for the first 1.5 s for Flash S0014, stroke 4....69
3-39: Evolution of the various quantities for the first 1.3 s for Flash S0015, stroke 6....70
3-40: Evolution of the various quantities for the first 5 s for Flash S0023, stroke 3.......71
3-41: Evolution of channel radius for S0008-4..................................................................72
3-42: V-shaped signature with EL> ERS. ERS represents the average electric field
between 44 s to 50 s after the beginning of the return stroke (at 50 s)..............76
ix
3-43: Histogram of peak current for strokes characterized by V-shaped electric field
signatures with ERS < EL and flattening within 20 s. .........................................76
3-44: Histogram of EL for strokes characterized by V-shaped electric field signatures
with ERS < EL and flattening within 20 s. ..........................................................77
3-45: Histogram of peak power per unit length for strokes characterized by V-shaped
electric field signatures with ERS < EL and flattening within 20 s.....................77
3-46: Histogram of energy per unit length for strokes characterized by V-shaped
electric field signatures with ERS < EL and flattening within 20 s.....................78
3-47: Histogram of action integral for strokes characterized by V-shaped electric field
signatures with ERS < EL and flattening within 20 s. .........................................78
3-48: Peak power vs. peak current for strokes characterized by V- shaped electric field
signatures with ERS < EL and flattening within 20 s. .........................................79
3-49: Energy vs. peak current for strokes characterized by V- shaped electric field
signatures with ERS < EL and flattening within 20 s. .........................................80
3-50: Energy vs. EL for strokes characterized by V- shaped electric field signatures
with ERS < EL and flattening within 20 s............................................................80
3-51: Energy vs. EL for strokes characterized by V- shaped electric field signatures
with EL< ERS and flattening within 20 s............................................................81
3-52: Energy vs. Action integral for strokes characterized by V- shaped electric field
signatures with EL< ERS and flattening within 20 s...........................................81
3-53: Histogram of EL for strokes characterized by V- shaped electric field signatures
with ERS (t) < EL and no field flattening within 20 s. ........................................83
3-54: Histogram of EL for strokes characterized by V- shaped electric field signatures
with ERS (t) < EL and no field flattening within 20 s. ........................................84
3-55: Histogram of peak power for strokes characterized by V- shaped electric field
signatures with ERS (t) < EL and no field flattening within 20 s........................85
3-56: Histogram of action integral for strokes characterized by electric field signatures
with ERS (t) < EL and no field flattening within 20 s. ........................................86
3-57: Peak power vs. peak current for strokes characterized by electric field signatures
with ERS (t) < EL and no field flattening within 20 s .........................................87
3-58: Peak power vs. EL for strokes characterized by electric field signatures with
ERS (t) < EL and no field flattening within 20 s .................................................88
4-1: Flash 03-31, bipolar flash ...........................................................................................90
x
xii
xiii
current (ICC pulses) and similar pulses superimposed on the continuing current that
follows the return stroke process (M-component pulses) were analyzed for years 2002
and 2003, and compared with statistics found in the literature. The following comparisons
were made: (a) ICC pulses in triggered lightning recorded at the ICLRT in 2002 and 2003
(relatively high sampling rate) vs. their counterparts recorded earlier (relatively low
sampling rate), (b) ICC pulses in triggered lightning vs. those in object-initiated
lightning, and (c) ICC pulses in triggered lightning vs. M-component pulses in triggered
lightning. Duration, risetime, and half-peak width of ICC pulses were somewhat greater
than those of M-component pulses. Current peak, charge, and action integral of Mcomponents and ICC pulses were similar.
xiv
CHAPTER 1
INTRODUCTION
Lightning strikes are the cause of many deaths and injuries. Electromagnetic fields
and currents associated with lightning also can have deleterious effects on nearby
electronic devices. The energy of lightning is a fundamental quantity required, for
example, in estimating nitrogen oxide (NO) produced by lightning; which, in turn, is
needed in global climate-change studies. Trace gases produced by atmospheric electric
discharges are important to the ozone balance of the upper troposphere and lower
stratosphere. Atmospheric electric discharges might have played an important role in
generating the organic compounds that made life possible on Earth. Currently, there is no
consensus on lightning input energy. Estimates found in literature differ by one to two
orders of magnitude. Our study measured electric fields using Pockels sensors in the
immediate vicinity of the lightning channel, along with the channel base currents, to
estimate the energy and peak power in triggered lightning. We also analyzed the action
integral (energy per unit resistance at the strike point) and other parameters of return
strokes and pulses superimposed on steady currents in triggered lightning.
CHAPTER 2
LITERATURE REVIEW
Three approaches used to estimate lightning peak power and energy input were
found in the literature.
The second approach was described by Krider et al. (1968). Radiant power and
energy emitted within a given spectral region from a single-stroke lightning flash
are compared with those of a long spark whose electrical power and energy inputs
are known with fair accuracy. The value of lightning input energy per unit
channel length estimated by Krider et al. (1968) is 2.3 105 J/m.
The third approach involves the use of gas dynamic models proposed by a number of
researchers (Plooster. 1971; Paxton et al. 1986, 1990; Dubovoy et al. 1968, 1991; Hill
1977; Strawe 1979; Bizjaev et al. 1990). A short segment of a cylindrical plasma
2
3
column is driven by the resistive (joule) heating caused by a specified flow of electric
current as a function of time. Lightning input energy predicted by these models is one
to two orders of magnitude lower than that of Krider et al. (1968).
2.1 Cloud Formation and Electrification
The primary source of lightning is the cloud type, cumulonimbus (commonly
known as thundercloud). The process of charge generation and separation is called
electrification. Apart from the cumulonimbus, electrification can also take place in a
number of other cloud types (in stratiform clouds, for example; and in clouds produced
by forest fires, volcanic eruptions, and atmospheric charge separation in nuclear blasts.
According to Henry et al. (1994), eight types of thunderstorms are known. Among them,
some common in Florida are sea/land-breeze thunderstorms, oceanic thunderstorms, airmass thunderstorms, and frontal thunderstorms. Portier and Coin (1994) give other
classifications. The formation of air-mass clouds is explained next.
On a sunny day, Earth absorbs heat from the sun, causing both water vapor and air
to rise to higher atmospheric levels, forming clouds. The energy of the water vapor
decides the intensity of the thunderstorm; the hotter the air, the more water vapor it can
hold, and the more powerful the thunderstorm can be. When water vapor condenses, it
releases the same amount of energy required for heating water, to produce water vapor.
Convection causes warm, humid air to reach higher altitudes. The released energy heats
the surrounding atmosphere, which raises the cloud to higher altitudes, pulling the humid
air from below (setting a chain reaction); with an updraft velocity of round 30 m/s
forming cells. A cell is said to be in mature stage (actually this stage is related to the
clouds ability to generate lightning) when it reaches higher altitudes, and its top flattens,
forming an anvil. This kind of cloud formation can be divided into three stages
Developing stage
-Starts with warm, rising air
-The updraft velocity increases with height
-Super-cooled water droplets are far above freezing level
-Small-scale process that electrifies individual hydrometeors takes place in this stage
Mature stage
-The heaviest rains occur
-The downdraft occurs, due to frictional drag of the raindrops
-Evaporative cooling leads to negative buoyancy
-The top of the cloud forms an anvil
-The graupel-ice mechanism and the larger convective mechanism take place, leading
to electrical activity.
Dissipating stage
-The downdraft takes over the entire cloud
-The storm deprives itself of supersaturated updraft air
-Precipitation decreases
-The cloud evaporates
Various measurements were made to estimate the distribution of charge within the
cloud. Initially, from ground-based measurements, it was assumed that the charge within
the cloud forms an electric dipole (positive charge region above negative charge region).
Simpson and Robinson (1941) made in-cloud measurements with balloons, and suggested
a tripole model with an additional positive charge at the base of the cloud. There is still
no consensus on the detailed distribution of charge within the cloud.
Figure 2-1: Electrical structure of a cumulonimbus. [Simpson, G. and Scrase, F.J.; The
Distribution of Electricity in the Thunderclouds, Proc. R. Soc. London Ser.
A, 161: Figure. No.4. pp.315, 1937]
Precipitation model: Heavy soft hail (graupel) with a fall speed > 0.3 m/s interacts
with lighter particles (ice crystals) in the presence of small water droplets. As a result,
heavy particles in cold regions (T<-15 C) acquire negative charge; heavy particles in
warm regions (T>-15 C) acquire positive charge. The second process (gravitational
force) separates the heavier and lighter charged particles, forming an electric tripole.
Convection model: Charges are supplied by external sources such as corona and
cosmic rays. Separation of charges is accomplished by organized convection.
2.2 Natural Lightning Discharges
Cloud discharges
-
Intracloud discharges
Cloud-to-cloud discharges
Cloud-to-air discharges
Cloud-to-ground discharges
6
-
Bipolar discharges
7
ground. Such bipolar discharges are usually of upward type and constitute probably less
than 10% of all cloud-to-ground discharges.
Figure 2-3: Four types of discharges between cloud and ground.1. Downward negative 2.
Upward negative 3.Downward positive 4. Upward positive. [M. A. Uman, The
Lightning Discharge; Dover Publications, Minneola, New York; Figure.. 1-3,
pp.9, 1987]
At t=0 ms, the thundercloud has a tripolar charge structure with positive charge in
the upper region, negative charge in the lower region, and a small pocket of positive
charge at the cloud base. Between t=0 and t=1 ms, preliminary breakdown occurs within
the cloud due to the local discharge between the pocket of positive charge at the base and
the primary negative charge. The local discharge neutralizes the positive charge at the
base and continues towards the ground as a stepped leader.
Figure 2-4: Downward negative cloud-to-ground lightning [V.A. Rakov, Uman, M.A,
Lightning: Physics and Effects; Cambridge University Press, New York;
2003].
This leader consists of a narrow current-carrying core and a much wider radial
corona sheath. Between 1.10 ms to 19 ms, the stepped leader moves towards the ground
with an average speed of 105 to 106 m/s, exhibiting steps of some tens of meters in length
and separated by some tens of microseconds. At t= 20 ms, the stepped leader approaching
the ground causes the electric field near ground to exceed the breakdown value for air,
which in turn results in an upward positive leader extending from ground towards the
9
descending stepped leader. Between 20.0 and 20.1 ms, the downward leader attaches to
one of the upward leader branches and a return stroke is initiated with a typical peak
current of 30 kA. From t= 20.10 ms to 20.20 ms, the first return stroke propagates
upward towards the cloud in the ionized channel left behind by the stepped leader with a
speed of around 108 m/s. The return stroke neutralizes the negative charge (typically 5 C)
deposited by the leader (lowers negative charge to ground). At t= 40 ms, some in-cloud
processes called K and J processes occur inside the cloud. At t= 60 ms to 62 ms, a dart
leader propagates downward along the channel left by the first return stroke with an
average speed of 107 m/s [Uman, 1987]. The dart leader deposits a negative charge of the
order of 1C onto the channel.
When the dart leader reaches the ground, a second return stroke is initiated which
travels upward with an average speed of 108 m/s. A sequence of leader and return stroke
is called a stroke, with the average number of strokes per flash being 3 to 5 [Rakov and
Uman, 2003].
Positive cloud-to-ground discharges can originate from the upper positive charge
region or positive charge pocket at the cloud base (assuming the tripolar model of cloud
charge distribution). It starts with a downward propagating positive leader and connects
to a negative upward leader launched from the ground. Then an upward return stroke is
initiated which transfers positive leader charge to ground. Typically there are no
subsequent strokes in positive cloud-to-ground discharges. The typical values of first
stroke peak currents measured at ground for positive cloud-to-ground discharges is
35 kA, not much different from 30 kA for negative cloud to ground lightning. On the
other hand, larger currents are more probable in positive strokes that in negative ones.
10
2.3 Mechanism of NO Production by Lightning
During a return stroke, the lightning channel attains a peak temperature of 30,000 K
in a few microseconds. If the cooling of the channel takes place slowly, equilibrium
composition at a given temperature is established, i.e., the final constituents of the cold
air would be the same as the constituents prior to the return stroke. It has been shown by
Uman and Voshall (1968) and Picone et al (1981) that the residual hot channel cools
from around 10,000 K to 3,000 K in a few milliseconds. The time required by NO to
attain equilibrium concentration increases rapidly with decreasing temperature. Hence, as
the channel cools down to the freeze out temperature, the temperature at which the
reactions that produce and destroy NO become too slow to keep NO in equilibrium
concentration, and hence NO remains at the density characteristic of the freeze out
temperature. Chemical reactions, which characterize the production of NO, are
O2 O + O
O + N2 NO + N
O2 + N NO + O
The reactions that compete with the NO producing reactions are shown below.
NO + N O + N2
NO + N O + N2
NO N + O
NO + NO N2O + O
These equations assume importance as NO produced by natural processes decreases
the ozone (O3) concentration in the stratosphere via the dominant equation
NO + O3 NO2 + O2
11
Ozone in the stratosphere is important to life because it shields the Earth from Suns
harmful ultraviolet radiation. Ozone in the troposphere acts as a greenhouse gas by
absorbing the infrared radiation.
2.4 Rocket-Triggered Lightning
The study of natural lightning is extremely difficult since it is impossible to
accurately predict its occurrence in space and time. For this reason, in order to study the
lightning properties a method to produce lightning artificially from natural thunderclouds
has been developed. The rocket and trailing wire technique is used to initiate lightning
that is referred to as rocket-triggered lightning.
Currently, rocket-triggered lightning (Rakov, 1999) can be produced in two
different ways:
12
13
ground. The latter leader/return stroke type sequence serves to re-establish the interrupted
current flow to ground.
14
(forming a bi-directional leader) are initiated, each propagating at a speed of 105 m/s. The
electric field produced by the downward negative leader initiates an upward connecting
positive leader from the grounded 50-m wire, which connects to the downward negative
leader. Finally, a return stroke is initiated which travels with a speed of 107-108 m/s and
catches up with the upward positive leader tip. After this stage, the processes are similar
to those of the classical rocket- triggered lightning.
2.5 Estimates of Peak Power and Input Energy in a Lightning Flash
There are various methods to estimate the peak power and energy dissipated in
lightning discharges. Some of them are described in the following sections.
2.5.1 Optical Measurements and Long Spark Experiments
Krider et al. (1968) estimated the average energy per unit length and peak power
per unit length to be 2.3105 J/m and 1.2109 W/m. Their optical measurements were
similar to those performed by Krider (1966) and are described below. A calibrated silicon
photodiode and an oscilloscope were used as a fast-response lightning photometer
covering the visible and near-infrared regions of the spectrum from 0.4 to 1.1 m.
Simultaneous still photographs of the discharge channels were taken to determine the
dependence of the photoelectric pulse profile on the type of lightning and the geometry of
its channel. The photodiode detector consisted of an Edgerton, Germeshausen and Grier
model 560-561 lite-mike and detector head (Krider, 1966). The photodiode and
associated circuitry were linear over a wide range of incident light levels (within 5% over
7 decades) and had a response time of less than 1 s. The calibrated relative response of
the detector is shown in Figure. 2-7. Photographs of the lightning channels were taken
with an Ansco Memar 35-mm camera, which had a focal length of 45 mm. Using the
15
same experimental setup optical measurements have been performed on a 4-meter air
spark produced by the Westinghouse 6.4 MV impulse generator.
Figure 2-7: Relative spectral response versus wavelength for the photodiode detector
used by Krider (1966) and Krider et al. (1968).
The basic principle of power and energy estimation is as follows. The spark current
and voltage were recorded as functions of time, enabling the calculation of electrical
power and total energy input. The radiant power reaching the detector is proportional to
the voltage measured at the output of the optical detector. Considering the spark channel
to be straight (to avoid taking the dependence of the radiant power on the azimuth of the
channel), the radiant power output from the light source is determined from equation (1).
It has been demonstrated experimentally using the long spark that the distance
dependence of the radiant flux is 1/R2. Hence, the total radiant power emitted in all
directions within the detector bandwidth is given by
P= (V/K) (4R2/ A)
(1)
where, V is the optical detector output voltage, K is the pulse calibration factor, R is the
distance from the light source to the detector, and A is the sensitive area of the detector.
The radiant power vs. time curve can be integrated to obtain the total radiant energy
emitted in a given bandwidth. The radiative efficiency is calculated by comparing the
value of total radiant energy to that of measured electrical energy input. Making a
16
simplifying assumption that the radiative efficiencies for laboratory spark and lightning
are the same, power and input energy can be found for a lightning stroke whose total
radiant energy is known from measurements. In this experiment, knowing the
approximate location of the lightning, cloud base height was obtained from the U.S.
Weather Bureau. Using the cloud base height, which determined the length of the
lightning channel that was visible, the size of the photographic image, and knowledge of
the camera focal length one can estimate the distance to the channel. This method
assumes that the channel is vertical. Figure 2-8 b shows the photoelectric voltage pulse
corresponding to the lightning whose still picture is shown in Figure 2-8a.
17
from the lightning stroke is calculated to be 1.11010 W, or dividing by the channel
length, to be as 6.2106 W/m. The corresponding measurements for the spark were made
at a distance of 23 m, and the peak radiant power within the detector bandwidth was
obtained to be 1.01010 W/m. The peak electrical power dissipated in the spark was
obtained accurately from the direct traces of current recorded as a function of voltage. At
the time of peak power, the current was 4.2103 A and the voltage
1.8106 V, yielding a peak electrical power input of 7.6109 W, or 1.9109 W/m
Comparison of the radiant and electrical peak powers for the long spark yields a radiative
efficiency of 0.52%. Assuming the same radiative efficiency for the lightning at the
instant of peak radiant power, the peak electrical power dissipated in the lightning stroke
is 2.11012 W. Dividing this value by the channel length, peak electrical power dissipated
per unit length is obtained as 1.2109 W/m.
The electrical energy per unit length dissipated in the long spark is obtained by
integrating over time the product of the current and voltage values obtained from the
traces taken during the experiment. Comparing the radiant and electrical energy values,
the average radiative efficiency of 0.38% is obtained for the spark. Applying the same
radiative efficiency to lightning, the total average energy dissipation per unit length is
estimated to be 2.3105 J/m. This value is in agreement with the thunder theory data of
Few et al. (1969), but is one to two orders of magnitude larger than the values predicted
by gas-dynamic models (Section 2.5.3).
2.5.2 Electrodynamic Model
We present here the electrodynamic model proposed by Borovsky (1995), which
describes dart leaders and return strokes as electromagnetic waves that are guided along
18
conducting lightning channels. The downward propagating dart leader deposits negative
charge onto the channel and deposits electrostatic energy around the channel. The
subsequent upward-propagating return stroke drains the negative charge off the channel
and heats the channel by expending the stored electrostatic energy. The net result is that
the negative charge is lowered from the cloud to the ground and the energy is transferred
from the cloud to the channel. This electrodynamic model also accounts for the flow of
energy associated with the flow of charge. In this model energy dissipated per unit length
in lightning channels is calculated as a relation to the linear charge density on the channel
and not to the cloud-to-ground electrostatic potential difference.
This model serves as a tool to visualize the dynamics of lightning during the dartleader and return stroke phases. Figure 2-9 illustrates the concept of the model.
The amount of energy deposited on the lightning channel can be estimated based on
electrostatic considerations, from the following expression (Uman, 1984, 1987)
W/L = (Qlow tot)/ Lmain
(2)
Where, Qlow is the amount of charge lowered from the cloud to ground, tot is the
electrostatic potential difference between the thundercloud charge region and the ground,
and Lmain is the length of the main channel.
19
Figure 2-9: Conceptual flow of charge and energy. a) Flow of charge during a dart leader
and return stroke. b) Flow of energy during the dart leader and return stroke.
[J. E. Borovsky, Lightning Energetics: Estimates of energy dissipation in
channels, channel radii, and channel-heating risetimes, J.Geophys. Res.,
vol.103, Figure.1, pp.11538, May. 1998]
The above relation for W/L is unreliable due to the following reasons,
As seen in Figure 2-10, which gives a sketch of the structure of lightning channel,
the energy expended in creating the feeder channels in the cloud and branches
20
(for first strokes only) has to be included. The total length of the channel network is
very difficult to estimate.
Cloud base
Figure 2-10: Channel structure of lightning depicting the main channel, the branches
(feeder channels) in the thundercloud, and branches below the thundercloud.
Borovsky (1995) gives a more accurate estimate of the energy dissipation by considering
the stored electrostatic energy density around the channel. Electrostatic energy density
(ED) is given by (3).
ED = 0 E2/2
(3)
Where, 0 = 8.85 10-12 F/m and E is the electric field at a distance r from the channel. E
is given by (4).
E = L / (2 0 r)
Where, L is the charge per unit channel length.
(4)
21
Note that equations (3) and (4), employ SI units, while the corresponding equations given
by Borovsky (1995) are in CGS metric unit system. The original equations and
conversion can be found in the Appendix. The stored energy per unit length equation is
derived as follows. If all the charge resides on the channel, the electric field very close to
the channel exceeds the air-breakdown limit Ebreak. Ebreak is approximately equal to 2106
V/m (Cobine, 1941). At locations along the lightning channel where E exceeds Ebreak,
conductivity increases rapidly, facilitating the movement of free charge, thus reducing the
electric field E. So, around the channel the electric field will be approximately equal to
Ebreak, up to a radius of rbreak.
rbreak= L / (2 0 Ebreak )
(5)
Beyond rbreak, the electric field falls off as 1/r. Thus, the radial dependence of the
electrostatic energy density residing around the channel is given by,
ED = 0 Ebreak 2/2
if
ED = L 2/82 r2 0
if
r rbreak
r rbreak
(6)
(7)
Total amount of electrostatic energy per unit length stored around the channel is given by
Wstored /L = (ED) 2r dr
(8)
0
Because of the radial dependence of electrostatic energy, the above integral can be
broken into
rbreak
(9)
(10)
Where rcut is the cutoff radius that is introduced to prevent the integral from
logarithmically diverging as r . The physically reasonable choice for rcut is the radius
22
at which the electric field of the channel equals the background electric field. Therefore,
rcut is given by the expression
rcut = L/ (2 0 Ecloud )
(11)
(12)
23
parameters chosen are disso =9.8 eV, ioniz= 14.5 eV, Tinit= 30000 K, Tatmos= 300 K,
where Tinit is the temperature of the channel before expansion, Tatmos is the temperature of
the ambient air outside the channel, disso and ioniz are the dissociation and ionization
constants. The final channel radius is estimated to be about 4.7 cm. Similarly, for the
dart-leader channel, the initial and final radii are found to be 0.26 cm and 3.8 cm. In this
latter case, the values chosen for L and atomic are 110-4 C/m and 5.01018 cm- 3.
(a)
(b)
Figure 2-11: Electrodynamic Model. (a) Downward propagating dart leader that loads
charge and electrostatic energy onto a lightning channel, (b) upwardpropagating return stroke that drains charge off the channel and uses up the
stored electrostatic energy [J. E. Borovsky, An electrodynamic description of
lightning return strokes and dart leaders: Guided wave propagation along
conducting cylindrical channels, J.Geophys. Res., Figure. 10, pp. 2717, Feb.
1995].
According to this model, the vertical (longitudinal) electric field outside the channel
decreases with increasing the distance r as loge (0.9 out r), which is a slowly varying
24
function of r, where out is the external wave number (Borovsky, 1995). out is chosen to
be (8.8 + i 9.4)10-5 cm-1 for the return stroke break-down pulse used in the illustration of
the variation of the horizontal and vertical components of the electric field shown in
Figure 2-12.
Figure 2-12: The peak values of electric and magnetic fields produced by the returnstroke breakdown pulse for the case of rch = 0.15 cm, T=15,000 K, t = 500
ns, and Imax= 20 kA, plotted as functions of the radial distance from channel
axis. rch is the channel radius, T is the channel temperature, t is the rise time
of the wave e-iwt, where w is the angular frequency( = 1/t, i e. t is around
one sixth of the time period of the sine wave), and Imax is the peak current . [J.
E. Borovsky, An electrodynamic description of lightning return strokes and
dart leaders: Guided wave propagation along conducting cylindrical
channels, J.Geophys. Res., Figure. 8, pp.2712, Feb. 1995]
2.5.3 Gas Dynamic Models
Gas dynamic models consider a short segment of a cylindrical plasma column
driven by the resistive (Joule) heating caused by a specified flow of electric current as a
function of time. Rakov and Uman (1998) review essentially all the gas dynamic models
25
found in the literature. The basic assumptions in the most recent models are: 1) the
plasma column is straight and cylindrical; 2) the algebraic sum of all the charges is zero;
3) local thermodynamic equilibrium exists at all times. The initial conditions of the
lightning channel are temperature of the order of 10000K, channel radius of the order of
1 mm, and pressure equal to ambient (1 atm) or mass density equal to ambient (of the
order of 10-3 g/cm3), the latter two conditions representing, respectively, the older and
newly created channel sections. The initial condition assuming the ambient pressure best
represents the upper part of the of the leader channel, since that part had sufficient time to
expand and attain equilibrium with the surrounding air. The initial condition of ambient
density is most suitable for the recently created, bottom part of the leader channel. At
each time step: 1) electrical energy sources; 2) the radiation energy sources; 3) Lorentz
force are computed and the gas dynamic equations are solved for the thermodynamic and
flow parameters of the plasma.
The energy input is determined as follows. The plasma channel is visualized as a
set of concentric annular zones, in which the gas properties are assumed constant. For a
known temperature and mass density, plasma composition can be obtained from the Saha
equation (Paxton et al. (1986, 1990), Plooster (1971)) or from tables of precompiled
properties of air in thermodynamic equilibrium (Hill (1971), Dubovoy et al. (1991,
1995)). The plasma conductivity can be computed from the plasma composition,
temperature and mass density. The current is distributed among the annular zones as if
they were a set of resistors connected in parallel. Using the cross-sectional distribution of
current and plasma conductivity, the amount of electrical energy input can be computed
for each of the annular zones.
26
The energy is deposited at the center of the channel in the form of heat, which is
transported to cooler outer regions in the form of radiation. The radiative properties of air
are complex functions of frequency and temperature. Radiation at a given frequency can
be absorbed and re-radiated at different frequencies while traversing the channel in the
outward direction. Paxton et al. (1986, 1990) and Dubovoy et al. (1991, 1995) used
tables of radiative properties of hot air to determine absorption coefficients as a function
of temperature for a number of selected frequency intervals to solve the equation of
radiative energy transfer in the diffusion approximation.
The pinch effect due to the interaction of electric current with its own magnetic
field was included in the gas dynamic model of Dubovoy et al. (1991, 1995). This
phenomenon counteracts the channels gas dynamic expansion, resulting in 10-20%
increase in input energy for the same input current because of reduced channel size.
Table 2-1,which is found in Rakov and Uman (1998), summarizes predictions of
the various gas dynamic models for the input energy and percentages of this energy
converted to kinetic energy and radiated from the channel. Additionally included in Table
2-1 are energy estimates based on experimental data (Krider et al; see Section 2.5.1 and
on electrostatic considerations (Uman, 1987; Borovsky, 1998; see Section 2.5.2). Brief
comments on each of these estimates follow the table.
Table 2-1: Lightning Energy Estimates [Rakov and Uman, 1998].
% Converted
Source
Current Input
to Kinetic
Energy,
Peak,
3
Energy
kA
10 J/m
Hill (1971,1977)
21
15
9+ (at 25 s)
(~3)
Plooster (1971)
20
2.4
4 (at 35 s)
Paxton et al.
20
4
2 (at 64 s)
(1986,1990)
% of Energy
Radiated
2*+ (at 25 s)
50 (at 35 s)
69 (at 64 s)
27
Continued Table 2-1.
Source
Dubovoy et al.
(1991,1995)
Borovsky (1998)
Krider et al. (1968)
Current
Peak,
kA
20
Input
Energy,
103 J/m
3
% Converted
to Kinetic
Energy
-
Singlestroke
flash
0.2-10
230
(200-2000)
Uman (1987)
% of Energy
Radiated
25 (at 55
s)
0.38#
Hill (1971, 1977) overestimates the input energy by a factor of 5 or so due to the
underestimation of electrical conductivity. The corrected value is given in the
parentheses. Ploosters (1971) model gives a crude radiative transport mechanism
adjusted to the expected temperature profile. Paxton et a/. (1986, 1990) gives individual
temperature dependent opacities for several wavelength intervals. Dubovoy et al.s
(1991, 1995) model is in principle the same as the previous one, except for the fact that
the pinch effect was taken into account. Uman (1987) estimates the input energy by
assuming that tens of coulombs are lowered from a height of 5 km to ground. An
assumption made is that the potential difference between the ground and charge center
inside the cloud is 108-109 V.
Krider et al. (1968) estimated the average energy per unit length and peak power to
be 2.3105 J/m and 1.2109 W/m (see Section 2.5.1). In this experiment the radiative
efficiencies of the long spark and the lightning channel are assumed to be constant in the
wavelength range of 0.4 to 1.1 m. Since the input energy for long spark energy is
known, the radiative efficiency can be determined (0.38%) and applied to the lightning
28
return stroke. This estimate appears to be consistent with the thunder theory of Few
(1965, 1995).
Borovsky (1995, 1998) describes the dart leaders and return strokes as
electromagnetic waves that are guided along the conducting channels (see Section 2.5.2).
In this electrodynamic representation of lightning, the stored electrostatic energy Wstored
around a charged channel is the source of power for a return stroke. Borovsky, based on
electrostatic considerations, estimates the energy per unit channel length to be around
11031.5104 J/m, which is consistent with that predicted by the gas dynamic models.
Hence, there are one to two orders of magnitude differences in the estimates of
energy per unit length. The higher end of the energy range is likely to have included a
significant fraction of the energy dissipated by processes other than the return stroke.
These include the in-cloud discharge processes like the one in which charges are
collected from isolated hydrometeors in volumes measured in cubic kilometers and
transported into the developing leader channel. Additional experiments are required to
resolve the up to two orders of magnitude uncertainty in the estimate of lightning energy
input. In chapter 3, we will attempt to estimate lightning energy using recently acquired
experimental data for rocket-triggered lightning.
CHAPTER 3
ESTIMATING POWER AND ENERGY
3.1 Methodology
Power per unit length and energy per unit length, each as a function of time are
estimated from the vertical (longitudinal) component of the electric field in the immediate
vicinity of the triggered-lightning channel and associated lightning return stroke current.
Additionally, channel resistance per unit length and channel radius are estimated. The
vertical electric field was measured by Miki et al. (2002) using a Pockels sensor placed at
a radial distance of 0.1 m from, and at a height of 0.1 m above the tip of the 2-m vertical
rod. The measured field was assumed to be equal to the longitudinal electric field inside
the channel. Indeed, according to Borovsky (1995), the longitudinal electric field at radial
distances of 10 cm and 1.6 m from the channel axis differs from the field at the channel
axis only by 2.1 10-4 % and 18 10-4 %, respectively (see Ez in Figure 2-12). The
average values of leader electric field changes (approximately equal to return stroke field
changes) at 0.1 to 1.6 m, 15 m, and 30 m from the lightning channel are 577 kV/m,
105kV/m, and 60 kV/m, respectively (Miki et al., 2002; Schoene et al., 2003, JGR).
Lightning current was measured at the base of the 2-m strike rod. We will assume that
this current is representative of the current flowing in the lightning channel at a height of
the Pockels sensor. Under these assumptions, the power and energy per unit length can be
t
estimated as P(t) = I(t) E(t) and W(t) = P()d , respectively (Figure. 3-1). This energy is
0
associated with joule heating of the lightning channel and can be viewed as the input
energy for the return-stroke process that is spent for ionization, channel expansion, and
29
30
production of electromagnetic (including optical) and acoustical radiation from the
channel.
Figure. 3-1: Illustration (not to scale) of the method used to estimate power, P(t), and
energy, W(t), from measured lightning channel current, I(t), and vertical
electric field, E(t), near the channel.
Very close (0.1 to 1.6 m) vertical electric fields and associated channel-base
currents were obtained for 36 strokes in nine triggered lightning flashes (see section 3.2
for the experimental setup). Out of 36 strokes, only 31 strokes in 12 flashes were suitable
for the analysis presented here. For the remaining strokes, though the current records
were available, the corresponding electric field records were saturated. All the acquired
electric field signatures can be divided in three types: 1) classical V-shaped signature
with return-stroke electric field change ERS being approximately equal to the leader
electric field change EL (ERS = EL); 2) V-shaped signature with ERS being
appreciably smaller EL; 3) same as 2, but with the return stroke portion exhibiting no
flattening that is expected to occur within 20 s or so of the beginning of the return
stroke. These three types of waveforms are illustrated in Figure 3-7. The reason for the
31
residual electric field some tens of microseconds after the return stroke for Types 2 and 3
is apparently due to the fact that the return stroke fails to neutralize all the leader charge
in the corona sheath surrounding the channel core (Kodali et al., 2003). The statistics for
peak power and energy are produced separately for the three types of electric field
signatures (no energy estimates for Type 3). Since Type 1 represents the classical
leader/return stroke sequence, while Types 2 and 3 indicate the presence of an additional,
slower process involved in the removal of charge from the channel (not all the
electrostatic energy deposited along the channel by the leader is tapped by the return
stroke), all the analysis concerning the channel resistance per unit length and channel
radius is presented only for Type 1. Further, the power and energy estimates for Type 2
were performed after adjusting the E- field waveforms to account for the residual field.
Events of Type 3 were used for estimating peak power only.
3.2 Experiment
Experimental data used in Section 3 have been acquired at the International Center for
Lightning Research and Testing (ICLRT) at Camp Blanding, Florida, in 2000. The
experiment was a joint University of Florida / CRIEPI, Japan, project, which is described
by Miki et al. (2002).
3.2.1 Pockels Sensors
The Pockels sensors used in this experiment had a stated dynamic range of 20 kV/m
to 1 MV/m (Miki et al., 2002). The lower measurement limit was determined by noise. In
this study we applied filtering (moving time averaging; see section 3.4.1) that allowed us
to significantly reduce this noise and, hence, to lower the measurement limit. The residual
noise translated to noise in power waveforms but did not materially influence energy
estimates.
32
During the calibration in laboratory in Japan, the 2/50 s voltage waveform from a
1-MV impulse generator was applied across a plane-plane gap formed by two electrodes
separated by 2-3 m for creating fields less than 1 MV/m and 0.1-0.2 m for creating fields
between 1 and 2 MV/m (Miki et al., 2002). 1.2-kV and 18 kV generators were also used.
The calibration setup is shown in Figure 3-2. The electric field was obtained by dividing
the voltage by the gap length, h (Figure. 3-2). The Pockels sensor was placed in this gap
and its output voltage was measured. The variation of the sensor output voltage as a
function of the external electric field is shown in Figure 3-3. The sensor output voltage
varies linearly with the E-field, and this linear relationship was applied to all
measurements analyzed here, even when the field values were less than the lowest field
used in the calibration process.
Field calibration of the Pockels sensors was performed at the ICLRT and
accomplished by comparing the outputs of Pockels sensors with that of a flat-plate
antenna, both installed 5 m from the triggered-lightning channel. Figure 3-4 shows
1.2-kV
18-kV, or
1-MV
Impulse
Voltage
Generator
Figure 3-2: Calibration of the Pockels sensor. Courtesy Megumu Miki of CRIEPI, Tokyo,
Japan. h=2 or 3 m for creating fields less than 1 MV/m and h=0.1 or 0.2 m for
creating fields between 1 and 2 MV/m.
examples of the two types of observed electric field waveforms, termed slow and fast,
measured simultaneously with a Pockels sensor and a flat-plate antenna.
33
The flat-plate antenna was calibrated theoretically [e.g., Uman, 1987], and the
Pockels sensors were calibrated (up to about 2 MV/m) in plane-plane gaps by CRIEPI
Figure 3-3: Variation of the Pockels sensor output voltage as a function of the applied
electric field: (a) sensor No.6 (used to measure vertical electric field
component) (b) sensor No.7 (used to measure horizontal electric field
component). Courtesy Megumu Miki of CRIEPI, Tokyo, Japan.
34
personnel (see above). Figure 3-5 shows a scatter plot of the magnitude of the vertical
electric field due to lightning measured with the Pockels sensor versus that measured
with the flat-plate antenna. Figures 3-4 and 3-5 show that the magnitudes of slow
waveforms are essentially the same for the flat-plate antenna and the Pockels sensor
records. However, the magnitudes of the relatively fast waveforms measured with the
Pockels sensor are on average about 60% of those measured using the flat-plate antenna.
This implies that electric field peaks measured using Pockels sensors may be
underestimates by 40% or so, provided that the frequency content of the electric field in
the immediate vicinity of the channel is not much different from that of relatively fast
waveforms at 5 m. The difference in the response of the Pockels sensors to slow and fast
waveforms is presumably caused by the insufficient upper frequency response of 1 MHz
of the Pockels sensor measuring system. If the frequency content is higher very close to
Figure 3-4: Comparison of the electric field waveforms simultaneously measured with a
Pockels sensor and a flat-plate antenna, both at 5 m.
35
the channel than at 5 m, the field peaks measured by the Pockels sensors may be
underestimated by more than 40%.
Figure 3-5: Comparison of magnitudes of the vertical electric field peaks measured with
Pockels sensors and a flat-plate antenna, both at 5 m. Pockels sensors No.6
and No.7 were subsequently used for measuring the vertical and horizontal
electric field components, respectively, in the immediate vicinity of the
lightning channel. [M. Miki, V.A. Rakov, K.J. Rambo, G.H. Schnetzer, and
M.A. Uman; "Electric Fields Near Triggered Lightning Channels Measured
with Pockels Sensors," J. Geophys. Res., vol.107 (D16), Figure. 4, pp. 4,
2002]
3.2.2 Experimental Setup
Pockels sensors were installed on the underground rocket launching facility at the
ICLRT [Rakov et al., 2000, 2001; Crawford et al., 2001], as shown in Figure 3-6.
36
Figure 3-6: Experimental setup. [M. Miki, V.A. Rakov, K.J. Rambo, G.H. Schnetzer, and
M.A. Uman; "Electric Fields Near Triggered Lightning Channels Measured
with Pockels Sensors," J. Geophys. Res., vol.107 (D16), Figure. 3, pp. 3,
2002]
The vertical field sensor was placed at a radial distance of 0.1 m from, and at a
height of 0.1 m above the tip of the 2-m vertical strike rod, and the horizontal field sensor
was placed directly below it. A metal ring having a radius of 1.5 m was installed around
the strike rod. The ring was connected to the base of the strike rod, which was grounded.
Since the lightning channel could attach itself either to the strike rod or the ring, the
horizontal distance between the channel and the Pockels sensor varied between 0.1 m to
1.6 m. The corresponding lightning currents are measured using a current viewing
resistor (shunt), placed at the base of the strike rod. Currents were also measured using a
different method as discussed in Section 3.4.1. There are two types of current records: 1)
low-current records, whose duration is about 250 ms and the measurement range is from
2 kA to 2 kA. The amplitude resolution of low-current records is about 1.8 A. The
sampling interval was 100 ns; 2) high-current records, whose duration is 50 s and the
measurement range is from 31 kA to 18 kA. The resolution of high-current records was
37
about 450 A. The current sampling interval was 20 ns. The sampling interval for electric
field records was 0.5 s.
3.3 Electric Field Waveforms
3.3.1 V-Shaped Signatures with ERS = EL
In this type, the return stroke apparently neutralizes all the charge deposited by the
leader and thus the entire waveform exhibits a V-shaped signature in which the leader
and return stroke field changes are nearly equal to each other. The rise time of the return
stroke electric field is of the order of 1 s.
EL
ERS = EL
EL
ERS < EL
EL
Time, s
Figure 3-7: V-shaped electric field signatures with a) the return stroke field change, ERS,
being equal to the leader field change, EL. b) ERS < EL, field flattening
within 20 s or so of the beginning of the return stroke (of the bottom of the
V), c) ERS (t) < EL, no flattening within 20 s.
38
Table 3-1: Summary of peak current and EL statistics for 8 strokes exhibiting V-shaped
electric field signatures with ERS = EL.
Min
9.85
Peak Current, kA
Max
Mean GM
21.5
16.6
15.5
Min
52.5
EL, kV/m
Max
Mean
305
122
GM
109
3.3.2 V-Shaped Signatures with ERS < EL and Field Flattening within 20 s
These electric field waveforms are characterized by residual electric fields (Kodali
et al., 2003) and, hence, residual charge (and associated electrostatic field energy) that is
apparently dissipated via a slower process lasting in excess of some hundreds of
microseconds, other than the return stroke. Therefore, such waveforms cannot be used
with confidence for estimating the input energy of a lightning return-stroke using the
method illustrated in Figure. 3-1, which is based on the assumption that all the
electrostatic field energy of the leader is converted to the Joule heating of the channel by
the return stroke. However, these waveforms can be used to estimate the peak power,
which is expected to occur within the first few hundred nanoseconds, long before the
flattening takes place. We also used these waveforms for computing the input energy
after adjusting them to eliminate the residual field. The statistics for the peak current and
EL for this category of strokes are given in Table 3-2.
Table 3-2: Summary of peak current and EL statistics for 5 strokes with ERS < EL
and flattening within 20 s or so.
Peak Current, kA
EL, kV/m
Min Max Mean GM Min Max Mean GM
15.4 26.3 20.6 19.6 105 227 160
155
39
3.3.3 Signatures with ERS (t) < EL and no Flattening within 20 s
In these E-field signatures, the electric field after the beginning of the return stroke
continues to increase during a time interval of the order of a few milliseconds. Such
behavior is indicative of a residual charge (and associated electrostatic field energy)
located near the attachment point and a process other than the return stokes being at work
to neutralize this residual charge. E-field waveforms with ERS (t) < EL and no
flattening with 20 s were used for computing only the peak power, which occur before
the abnormal behavior of the return-stroke E-field begins. The statistics of the peak
current and EL associated with this type of waveforms are given in Table 3-3.
Table 3-3: Summary of peak current and EL statistics for 18 strokes with ERS (t) < EL
and no flattening within 20 s
Min
5.1
Peak Current, kA
Max
Mean GM
26.4
11.4
10.5
Min
EL, kV/m
Max
Mean
GM
175
1150
474
554
40
portion of the electric field record following the initiation of the return stroke was
filtered, since only this portion was needed for estimating power and energy input. A
moving-averaging window of 100 data points, which acts as a low-pass filter, was used
for this purpose. Averaging was done after a suitable time interval after the beginning of
the return stroke, so that the initial (fast-varying) portion of the return stroke is not
modified, as illustrated in Figure. 3.8 b. In this example the E-field waveform is averaged
2.5 s after the start of the return stroke. One can see that the main features of the
waveform are preserved, while the noise is significantly reduced. Such filtering was
performed for all the strokes analyzed here. The resultant electric fields are shown in
Figures 3-10 to 3-17.
Leader
Return Stroke
Original
Time-Averaged
2.5s
B
Time, s
Figure 3-8: Stroke S0013-1. A) Original E-field record. B) Filtered (100-s movingwindow time averaged) version of the E-field waveform shown in A).
In 2000, lightning currents were measured using two methods. In the first method,
the total lightning current was measured using a current viewing resistor, CVR (shunt),
41
placed at the base of the strike rod. In the second method, currents entering the launcher
grounding system (ground screen and ground rod) were measured and summed to obtain
the total lightning current. In this latter case, the current into the 70 70m2 buried
metallic grid (ground screen) was measured using two CVRs and the ground rod current
was measured using two P 110As current transformers (CT). The current range of P
110As is from a few amperes to 20 kA, when terminated with a 50-ohm resistor. A
passive combiner was used to sum the two signals from the ground rod CTs to a total
ground rod current. The ground screen current was measured by two separate
instrumentation systems IIS-S (south ground screen current) and IIS-N (north ground
screen current). The sum of the ground rod current and north and south screen currents
gives the total screen
IR = 0.85+1.02 IS
R2 = 0.8
n = 36
Figure.3-9: Scatter plot of screen current, IS vs. strike rod current, IR, for 2000. [V.
Kodali, Characterization and analysis of close lightning electromagnetic
fields, Masters thesis, University of Florida; 2003].
current. A scatter plot of ground screen current (the sum of the ground screen and ground
rod currents to be exact) vs. strike rod current is shown in Figure 3-9. With a few
42
exceptions, the two current values are very close to each other. In the following sections
(also in Tables 3-1 to 3-3), we used the strike-rod current, although in Table 3-4 the
power and energy were also computed using the ground-screen current, when available.
3.4.2 Power and Input Energy
Power as a function of time, obtained as the product of longitudinal E-field and
strike-rod current, and energy, the integral of the power curve, are shown in Figures 3-10
to 3-17, for the eight strokes having the V-shaped E-field signatures with ERS = EL.
The estimated peak power and energy values are given in Table 3-4. Histograms of the
various quantities and associated scatter plots are found in Section 3.4.3. Error analysis is
presented in Section 3.4.4.
Figure 3-10: Time variation of electric field, current, power, and energy for stroke
S006-4.
43
Figure 3-11: Same as Figure. 3-10, but for Stroke S008-4. Negative values in the
variation of power with time are due to residual noise in the electric field
waveform.
44
Time, s
Figure 3-13: Same as Figure. 3-10, but for Stroke S0013-4. Negative values in the
variation of power with time are due to residual noise in the electric field
waveform.
Time, s
Figure 3-14: Same as Figure. 3-10, but for Stroke S0015-2. Negative values in the
variation of power with time are due to residual noise in the electric field
waveform.
45
46
Leader
Return Stroke
Time, s
Figure 3-16: Same as Figure. 3-10, but for Stroke S0015-6. The time scale here is
different from that in Figures. 3-10 to 3-15 and 3-17 (50 s vs. 55 s). The
energy is obtained at 10 s after the return stroke, because at later times the
electric field becomes positive causing the power waveform to change polarity
(become negative), which is physically unreasonable.
Time, s
Table 3-4: Power and energy estimates for strokes having V- shaped E-field signatures with EL= ERS.
EL,
Peak power,
Date
Flash
Stroke
Termination Peak current,
Energy,
8
ID
order
point
kA
kV/m
10 W/m
103 J/m
Rod
Screen
Rod
Screen
6/13
S0006
Rod
14.3
10.2
52.5
2.35
2. 9
6/17
S0008
Ring
20.9
19.3
60.0
2.2
4.8
Rod
11.6
125.0
5.2
6/18
S0013
4
Ring
11.6
122.5
8.6
6/23
Rod
19.3
Noisy
105.0
9.9
Rod
21.5
Noisy
112.5
8.7
Rod
20.0
Noisy
92.5
14.5
Ring
9.85
16.3
305.0
25.1
22.0
6/23
S0015
6/23
7/11
S0023
Rod
1.8 (at
45.7
s)
Screen
1.7(at
41.3
s)
0.9 (at
45.6
s)
2.57 (at
45.7
s)
1.34 (at
45.5
s)
6.35 (at
45.5
s)
0.8 (at
45.5
s)
-
5.0 (at
45.5
s)
1.3 (at
10.0
s)*
6.2 (at
45.7
s)
Classical
trigger, 5
RSs
Classical
trigger, >8
RSs
Classical
trigger, 6
RSs
Classical
trigger, 6
RSs
5.4 (at
47.3
s)
* For this stroke, after 10 s ERS > EL causing the power waveform to change polarity (become negative) which is physically unreasonable.
Classical
trigger, 3
RSs
47
6/18
Remarks
48
3.4.3 Statistical Analysis
Histograms displaying the distributions of peak current, EL, peak power, energy,
and action integral for the V-shaped E-field waveforms with ERS = EL are shown in
Figures 3-18 through 3-22. The means and standard deviations are given separately for
different lightning channel termination points (rod or ring) and for all data combined.
Additionally presented in Figure 3-23 and Figure 3-24 are histograms of risetimes for
Number
ring
n=3
Peak Current, kA
Mean = 14.1 kA
rod
n=5
Mean = 18.1 kA
n=8
Mean = 16.1 kA
Min = 9.9 kA
Figure 3-18: Histogram of peak current for strokes characterized by V- shaped electric
field signatures with ERS = EL.
Number
49
EL, kV/m
ring
rod
n=3
n=5
n=8
Number
50
n=3
rod
n=5
n =8
Figure 3-20: Histogram of peak power for strokes characterized by V- shaped electric
field signatures with ERS = EL.
Number
51
rod
n=5
rod
n=8
Figure 3-21: Histogram for input energy for strokes characterized by V- shaped electric
field signatures with ERS = EL.
Number
52
rod
n=5
n=8
Figure 3-22: Histogram for action integral for strokes characterized by V- shaped electric
field signatures with ERS = EL.
53
Number
Mean = 0.85 s
Std. dev = 0.37 s
Min = 0.40 s
Max = 1.6 s
Risetime, s
Figure 3-23: Histogram of the risetime of current for strokes characterized by V-shaped
electric field signatures with ERS = EL. The risetime is defined as the time
taken by current to rise from 0% to 100% of the peak value.
Number
Mean = 0.43 s
Std.dev = 0.12 s
Min = 0.28 s
Max = 0.60 s
Risetime, s
Figure 3-24: Histogram of the 0-100 % risetime of power per unit length for strokes
characterized by V-shaped electric field signatures with ERS = EL.
54
Scatter plots showing correlation between the various parameters are presented in
Figures 3-25 through 3-29.
P = -0.49 I + 17.6
R2 = 0.1
Peak Current, kA
Figure 3-25: Peak power vs. peak current for strokes characterized by V- shaped electric
field signatures with ERS = EL. The filled circles and the hollow circles
represent the strokes for which the lightning channel terminated on the rod
and ring, respectively.
It can be observed from Figures 3-25 and 3-26 that the determination coefficient,
R2, which is the square of the correlation coefficient, R, in both cases is close to zero.
Two possible reasons for the low correlation coefficients are the following: 1) the
influence of the electric field is more significant than that of current and 2) the electric
field decays to a negligible value before the current attains its peak magnitude. Indeed,
the mean risetime to peak current is 0.85 s (see Figure. 3-23), and the average risetime
of power per unit length to its peak 0.43 s (see Figures. 3-23 and 3-24). Hence, electric
55
field has a more pronounced effect on the peak power value, since the current attains its
peak value only after the power does, resulting in the lack of dependence of the peak
power on the peak current value. It is likely that both reasons listed above can contribute
to the observed lack of correlation between the peak power and peak current. There
appears to weak positive correlation between the energy per unit length and action
integral. The latter was computed as the integral of the square of current over the same
time interval (typically between 40 and 50 s) as the corresponding power per unit
length. The unit for action integral is A2s, which is the same as J/.
W = -0.02 I + 3.6
R2 = 2.3 10-3
Peak Current, kA
Figure 3-26: Energy vs. peak current for strokes characterized by V- shaped electric field
signatures with ERS = EL. The filled circles and the hollow circles represent
the strokes for which the lightning channel terminated on the rod and ring,
respectively.
56
P = 0.08 EL 0.66
R2 = 0.78
EL, kV/m
Figure 3-27: Peak power vs. EL for strokes characterized by V- shaped electric field
signature with ERS = EL.
W = 0.02 EL + 1
R2 = 0.37
EL, kV/m
Figure 3-28: Energy vs. EL for strokes characterized by V- shaped electric field
signature with ERS = EL.
57
AI = 348 W + 936.21
R2 = 0.31
3
2
1
0
0
1000
2000
3000
4000
5000
6000
7000
Energy, J/m
Figure 3-29: Energy vs. Action Integral for strokes characterized by V- shaped electric
field signature with ERS = EL.
3.4.4 Error Analysis
In this section, three sources of uncertainty involved in power and energy estimates
are examined: 1) as noted in Section 3.2.1, electric fields measured using Pockels sensors
may be underestimated by 40% or so due to the insufficient upper frequency response of
1 MHz of the measuring system (Miki et al., 2002), (2) the sampling interval for electric
field records was 0.5 s vs. 20 ns for current records (see section 3.2.2), and (3) electric
field and current records were aligned manually using the bottom of the V-shaped electric
filed signature and the beginning of the current waveform. There is little ambiguity in
selecting the return-stroke current starting point, while the bottom of the V is somewhat
uncertain within 0.5 s due to insufficient sampling rate for electric field. Sensitivity of
power and energy estimates to changes in the electric field peak magnitude and its
58
position on the time scale are examined using the following procedure. The peak power
and energy are calculated after applying a correction factor of 1.6 at the instant of peak
electric field (to take into account the 40% potential error due to the insufficient upper
frequency response of the measuring system). Then the electric field waveform is
modified in three different ways: (1) keeping the position of the negative field maximum
intact (2) moving the negative maximum (the bottom of the V) 0.24 s to the left, and (3)
moving the maximum, 0.24 s to the right from its original position, in order to partially
account for the 0.5 s uncertainty noted above. These three steps are illustrated in
Figures 3-30 to 3-32 for stroke S0013-1. Similar procedure was applied to all the strokes
analyzed here, and results are summarized in Table 3-5, along with the peak power and
energy values estimated from the original records without correction.
E-Field, kV/m
59
Figure 3-30: Flash 0013, stroke 1; Correction factor of 1.6 is applied at the instant of
negative E-field peak. The filled circles and the arrows indicate the original
data points (the sampling interval of the electric field record was 0.5 s). The
hollow symbols represent fictitious data points based on linear interpolation so
as to match the sampling interval of 0.02 s of the current record. The circles
represent the original E-field waveform and the triangles represent the E-field
waveform after the correction factor of 1.6 is applied. Calculated values of
peak power and energy before and after correction are given.
E-Field, kV/m
60
Figure 3-31: Flash 0013, stroke 1; Correction factor of 1.6 is applied at the instant of
negative E-field peak, which is shifted by 0.24 s to the left in order to
partially account for the 0.5 s uncertainty in the position of the peak. The
filled circles and the arrows indicate the original data points. The hollow
symbols represent fictitious data points based on linear interpolation so as to
match the sampling interval of 0.02 s of the current record. Calculated values
of peak power and energy before and after correction are given.
61
Time, s
Figure
igure 3-32: Flash 0013, stroke 1; Correction factor of 1.6 is applied at the instant of
negative E-field peak, which is shifted by 0.24 s to the right in order to
partially account for the 0.5 s uncertainty in the position of the peak. The
filled circles and the arrows indicate the original data points. The hollow
symbols represent fictitious data points based on linear interpolation so as to
match the sampling interval of 0.02 s of the current record. Calculated values
of peak power and energy before and after correction are given.
Table 3-5: Dependence of peak power and energy on errors in the value of E-field peak and its position on the time scale.
Flash
Stroke Original
E-Field peak multiplied by
E-Field peak multiplied by 1.6
E-Field peak multiplied by 1.6
ID
order record
1.6
and shifted by 0.24 s to the left and shifted by 0.24 s to the
right
P.P,
E,
P.P,
E,
P.P,
P.P,
E,
E,
8
3
8
3
8
3
8
%
%
%
%
10 , 10 ,
10 , %
10 ,
10 ,
10 ,
10 ,
103, %
W/m
J/m
W/m
J/m
W/m
J/m
W/m
J/m
S006
4
2.3
6.4
2.6
13
1.8
0.0
4.5
96
1.92
6.7
1.74
-24
1.77 -1.7
4
1.97
0.8
2.83
44
0.82
2.5
6.58
234
0.96
20
1.96
-0.5
0.76
-5.0
S0013
5.2
2.6
6.1
17
2.6
0.0
8.0
54
2.8
7.7
4.1
-21
2.4
-7.7
S0013
8.6
1.34
8.6
2.3
-3.0
13
51
1.7
27
5.9
-31
1.2
-10
S0015
S0015
2
4
9.9
8.7
6.35
5.0
12.7
11.9
28
37
6.43
5.1
1.0
2.0
14
13.9
41
60
6.2
5.38
-2.0
8.0
8.1
5.9
-18
-32
6.1
4.86
-4.0
-3.0
S0015
14.5
1.3
14.5
0.0
1.33
2.3
17.5
21
1.7
31
10.7
26
0.94
-28
S0023
25.1
6.2
31.5
25
6.4
3.0
26.9
7.0
6.9
11
23.8
-5.0
5.6
-10
Mean
9.53
3.75
11.3
21
3.35
1.0
13
71
3.5
14
7.8
-20
2.9
-9.0
St. deviation
7.53
2.5
9.3
16.1 2.3
1.96
7.1
71.2
2.3
11.3
7.1
19.4
2.2
8.4
62
S008
63
Table 3-5 suggests that the peak power estimates are sensitive to the considered
uncertainties, the maximum mean error being 71%. Intuitively it makes sense, because
the peak power occurs in the first few microseconds or less of the beginning of the return
stroke. On the other hand, energy estimates are relatively insensitive to the uncertainties
examined here, the variation not exceeding 31% (14% on average for the 8 strokes
analyzed).
3.4.5 Channel Resistance and Radius
The expression, R (t) = E (t) / I (t), gives the evolution of resistance per unit channel
length with time. Since we cannot measure leader currents, R (t) can be evaluated only
for the return stroke. The evolution of channel radius can be estimated from the channel
resistance using the expression, r (t) = [ R (t)] 0.5, where is the electrical
conductivity of the channel, assuming = 104 S/m. In reality, increases with time (as
the channel temperature increases), but this variation is rather weak for the expected
temperature range ( 20,000 K or so) (e g., Rakov, 1998). The assumption of =
constant implies that R (t) decreases only due to expansion of channel (increase in r (t)).
In principle, the channel radius and resistance can be evaluated for the entire length
of the field record. But the results become dominated by noise once the electric field
magnitude decreases below 20 kV/m because of the limitations on the dynamic range of
the Pockels sensor. The evolution of resistance and channel radius along with
corresponding E-field, current, and power profiles, for eight strokes exhibiting V-shaped
E-field signatures with ERS = EL for the time interval when the E-field magnitude is
greater than 20 kV/m is shown in Figures 3-33 to 3-40. Table 3-6 shows the resistance
and channel radius at the instant of peak power. The time scale over which the evolution
64
is shown differs from that of other Figures in Section 3.4.2 because as the magnitude of
E(t) falls below about 20 kV/m, it attains small values during zero crossings forcing R(t)
to very small values, which in turn causes r (t) to go to unreasonably high values
(Figure 3-41).
a)
b)
c)
d)
S006-4
e)
Figure 3-33: Evolution of the various quantities for the first 0.58 s for Flash S006,
stroke 4. a) E-field; b) current; c) power per unit length; d) channel resistance
per unit length; e) channel radius.
65
a)
b)
c)
d)
e)
Figure 3-34: Evolution of the various quantities for the first 0.4 s for Flash S008, stroke
4. a) E-field; b) current; c) power per unit length; d) channel resistance per
unit length; e) channel radius.
66
Figure 3-35: Evolution of the various quantities for the first 1.4 s for Flash S0013,
stroke 1. a) E-field; b) current; c) power per unit length; d) channel resistance
per unit length; e) channel radius
67
a)
b)
c)
d)
S0013-4
e)
Figure 3-36: Evolution of the various quantities for the first 1.2 s for Flash S0013,
stroke 4. a) E-field; b) current; c) power per unit length; d) channel resistance
per unit length; e) channel radius
68
a)
b)
c)
d)
e)
Figure 3-37: Evolution of the various quantities for the first 2.1 s for Flash S0015,
stroke 2. a) E-field; b) current; c) power per unit length; d) channel resistance
per unit length; e) channel radius
69
a)
b)
c)
d)
e)
Figure 3-38: Evolution of the various quantities for the first 1.5 s for Flash S0014,
stroke 4. a) E-field; b) current; c) power per unit length; d) channel resistance
per unit length; e) channel radius
70
a)
b)
c)
d)
e)
Figure 3-39: Evolution of the various quantities for the first 1.3 s for Flash S0015,
stroke 6. a) E-field; b) current; c) power per unit length; d) channel resistance
per unit length; e) channel radius
71
a)
b)
c)
d)
S0023-3
e)
Figure 3-40: Evolution of the various quantities for the first 5 s for Flash S0023, stroke
3. a) E-field; b) current; c) power per unit length; d) channel resistance per
unit length; e) channel radius
Channel Radius, cm
72
Time, s
Figure 3-41: Evolution of channel radius for S0008-4. The peaks are formed due to small
values attained by the noisy electric field waveforms.
Table 3-6: Resistance and channel radius for strokes having V- shaped E-field signatures
with ERS = EL
Flash Stroke Resistance*,
Channel
Termination
ID
order
radius*, cm
point
/m
S006
4
0.67
0.69
Rod
S008
4
1.3
0.49
Ring
S0013
1
4
2
4
6
3
5.1
8.0
4.5
5.1
4.3
30.8
0.25
0.22
0.31
0.25
0.27
0.10
Mean value
7.5
0.32
S0015
S0023
Rod
Ring
Rod
Ring
* Resistance and channel radius are evaluated at the instant of peak power
73
As expected, the channel resistance decreases and channel radius increases with time as
the return-stroke current heats the channel.
In Rakov (1998), the propagation mechanisms of dart leaders and return strokes are
analyzed by comparing the behavior of traveling waves on a lossy transmission line and
the observed characteristics of these two lightning processes. The R in the transmission
line model is assumed to be a constant but different ahead of and behind either the dartleader or the return-stroke front with any nonlinear effects occurring at the front. The
channel radius and resistance ahead of return-stroke front are estimated to be around 0.3
cm and 3.5 /m. The channel radius and resistance in Table 3-6 are obtained at the
instant of peak power which occurs at around 0.4 s. The mean values for channel radius
and resistance in Table 3-6 are 0.32 cm and 7.5 /m. In Rakov (1998), the channel radius
and resistance behind the return-stroke front are estimated to be around 3 cm and 0.035
/m. For the pre-dart-leader channel, these two quantities are estimated to be around 3
cm and 18 k/m, respectively.
3.5 Analysis of V-shaped E-field Signatures with ERS < EL and Field Flattening
within 20 s
An example of such a waveform exhibiting the residual electric field, EL - ERS, is
shown in Figure. 3-5 b. The product of channel-base current and close vertical
(longitudinal) electric field, each as a function of time, yields the power per unit channel
length vs. time waveform. The energy per unit length is obtained by integration over time
of the power waveform, as discussed in section 3.1. The electric field waveform due to
__ __
the return stroke was adjusted by subtracting the residual electric field, EL- ERS. ERS
is defined as the average value of the electric field from 44 s to 50 s after the return
stroke. This time frame is selected since most of the energy estimates for most of the
74
events analyzed in Section 3.4 was obtained at around 45 s after the beginning of the
return stroke. The adjustment was needed (see Section 3.3.2) to eliminate the electrostatic
energy not involved in the return stroke process. Figure 3-42 shows this procedure for the
stroke S0008-3. This methodology enables us to reasonably compare the energy estimates
for the two classes of electric field waveforms. Table 3-7 gives the peak power and
energy values for the 5 strokes exhibiting V-shaped E-field signatures with ERS < EL
and field flattening within 20 s. Histograms giving the distribution of peak current, EL,
peak power, energy, and action integral are shown in Figures 3-43 to 3-47.
Table 3-7: Power and energy estimates for strokes having V- shaped E-Field Signatures with ERS < EL and field flattening within 20
s or so.
Peak power,
Date Flash
Stroke
Termination Peak current,
EL,
Energy,
(EL- ERS)*
kA
ID
order
point
kV/m 108 W/m
kV/m
103 J/m
Screen Rod
Screen
Rod
Screen Rod
Screen Rod
6/13
S0008
Ring
13.9
17.8
180.5
8.69
12.7
1 (at
20 s)
6/18
S0012
Ring
21.0
189.9
14.9
8.1 (at
50 s)
18.1
36.7
17.9
62.0
47.5
75
2.9 (at
50 s)
2
12.3
124.9
9.40
1.9 (at
47.5
s)
Ring
S0013
6/18
3
26.3
219.9
16.8
3.66 (at
50 s)
22.7
277.3
20.2
3.3 (at
5
45.5
s)
* ERS in (EL- ERS) is the average value of electric field from 44 s to 50 s after the return stroke.
20.0
76
__
EL
ERS
Time, s
__
__
Figure.3-42: V-shaped signature with EL> ERS. ERS represents the average electric
field between 44 s to 50 s after the beginning of the return stroke (at 50 s).
The electric field waveform due to the return stroke was adjusted by
subtracting the residual electric field, EL- ERS.
Number
Min = 12.3 kA
Max = 26.3 kA
Mean = 20 kA
Std. Dev. = 5.3 kA
n =5
Peak Current, kA
Figure 3-43: Histogram of peak current for strokes characterized by V-shaped electric
field signatures with ERS < EL and flattening within 20 s.
77
Number
Min = 107.0 kV
Max = 229.8 kV
Mean = 162.1 kV
Std. Dev. = 43.90 kV
n=5
EL, kV
Figure 3-44: Histogram of EL for strokes characterized by V-shaped electric field
signatures with ERS < EL and flattening within 20 s.
Number
78
Number
Number
Min = 2.05 10 A s
3 2
Max = 6.18 10 A s
3 2
Mean = 3.92 10 A s
3 2
Std. Dev. = 1.79 10 A s
n=5
79
The mean values of peak power and energy for this category of strokes are similar
to their counterparts for the first category (see Section 3.4). Scatter plots showing
correlation between the different quantities are presented in Figures 3-48 to 3-52.
P = 0.56 I + 3.17
R2 = 0.72
Peak Current, kA
Figure 3-48: Peak power vs. peak current for strokes characterized by V- shaped electric
field signatures with ERS < EL and flattening within 20 s.
80
W = 0.02 I + 3.6
R2 = 1.6 10-3
Peak Current, kA
Figure 3-49: Energy vs. peak current for strokes characterized by V- shaped electric field
signatures with ERS < EL and flattening within 20 s.
P = 0.08 EL + 0.98
R2 = 0.83
EL, kV/m
Figure 3-50: Energy vs. EL for strokes characterized by V- shaped electric field
signatures with ERS < EL and flattening within 20 s.
81
9
W = 0.01 EL + 2.28
R2 = 0.04
8
7
6
5
4
3
2
1
100
120
140
160
180
200
220
240
EL, kV/m
Figure 3-51: Energy vs. EL for strokes characterized by V- shaped electric field
signatures with EL< ERS and flattening within 20 s.
AI = -0.05 W + 4.13
R2 = 5.2 10-3
1000
2000
3000
4000
5000
6000
7000
8000
9000
Energy, J/m
Figure 3-52: Energy vs. Action integral for strokes characterized by V- shaped electric
field signatures with EL< ERS and flattening within 20 s.
82
In contrast with the strokes for which ERS = EL, there appears to be moderate
positive correlation between the peak power and peak current (or EL) and essentially no
correlation between the energy and action integral, although the sample size is small.
3.6 Analysis of V-Shaped E-field Signatures with ERS (t) < EL and no Field
Flattening within 20 s
An example of such a signature is shown in Figure 3-5 (c). Table 3-8 gives the
values of the strike-rod peak current (no usable ground-screen currents are available),
EL, and peak power for the 18 strokes of this type. Since there is no field flattening
within 20 s (in fact, the field varied on the millisecond time scale) of the start of the
return stroke, energy computation was not performed for the events of this type.
Table 3-8: Power estimates for strokes having V- shaped E-Field Signatures with ERS (t)
< EL (t) and no flattening within 20 s
Peak
Peak
power,
EL,
Flash Stroke Termination
current,
Date
Remarks
kV/m
ID
order
point
109
kA
W/m
1
11.8
492
4.4
7/11
Classical
S0022
Rod
trigger,
3 RSs
3
8.9
486.6
3.7
7/11
7/16
7/16
S0023
S0025
S0027
11.5
451.5
3.4
15.2
15.0
9.4
308.2
743.4
763
3.9
11.0
5.9
7.1
432
2.7
26.4
1149
25.1
11.4
865.1
17.0
1102
17.1
15.3
1108
12.8
2
1
2
3
2
3
Ring
Ring
Ring
Classical
trigger, 3 RSs
Classical
trigger, 4 RSs
Classical
trigger, 9 RSs
83
Continued Table3-8.
Date
S0029
Peak
power,
Remarks
109 W/m
Peak
current,
kA
EL,
kV/m
6.7
256.8
1.6
5.1
358.8
1.8
5.9
175.7
0.64
11.3
446.1
4.8
8.2
209.5
1.3
12.0
364.9
4.1
6.9
263.6
1.8
Stroke Termination
order point
Rod
Classical
trigger, 9 RSs
Number
7/20
Flash
ID
Ring
Rod
n=9
n=9
n = 18
Peak Current, kA
Mean = 14.3 kA
Mean = 8.5 kA
Mean = 11.4 kA
Min = 5.1 kA
Number
84
EL, kV
n=9
Mean = 769.1 kV
n=9
Mean = 339.3 kV
n = 18 Mean = 554.2 kV
Min = 175.7 kV
Number
85
Rod n = 9
n = 18
Figure 3-55: Histogram of peak power for strokes characterized by V- shaped electric
field signatures with ERS (t) < EL and no field flattening within 20 s.
Number
86
n=9
Rod
n=9
As seen in Figure. 3-53, the mean peak power for this stroke category is several
times larger than for the first two categories considered in Sections 3.4 and 3.5. Scatter
plots showing correlation between the different quantities are presented in Figures. 3-55
and 3-56.
87
P = 1.1 I 6.23
R2 = 0.83
Peak Current, kA
Figure 3-57: Peak power vs. peak current for strokes characterized by electric field
signatures with ERS (t) < EL and no field flattening within 20 s. The filled
circles and the hollow circles represent strokes for which the lightning channel
was attached to the rod and the ring, respectively.
88
P = 0.02 EL 3.5
R2 = 0.83
EL, kV
Figure 3-58: Peak power vs. EL for strokes characterized by electric field signatures
with ERS (t) < EL and no field flattening within 20 s. The filled circles and
the hollow circles represent the strokes for which the lightning channel was
attached to the rod and the ring, respectively.
Similar to the strokes considered in Section 3.5, there appears to be moderate
positive correlation between the peak power and peak current (or EL).
CHAPTER 4
CHARACTERIZATION OF PULSES SUPERIMPOSED ON STEADY CURRENTS
4.1 Initial Stage in Rocket-Triggered Lightning
4.1.1 Introduction
Rocket-triggered lightning is initiated by an upward leader propagating from the
upper end of a vertical grounded wire extended below the charged cloud by a small
rocket. The upward-leader stage, including explosion of the triggering wire and its
replacement by an upward-leader plasma channel, is followed by an initial continuous
current (ICC). ICC has duration of some hundreds of milliseconds and amplitude of some
tens to some thousands of amperes. The upward leader and the ICC constitute the initial
stage (IS) of rocket-triggered lightning. ICC pulses are the current pulses superimposed
on the slowly varying continuous current of the initial stage. After the cessations of the
ICC, one or more downward leader/upward return stroke sequences may occur. The
magnitudes of ICC pulses are smaller than those of return-stroke pulses. Overall, the
initial stage in rocket-triggered lightning is apparently similar to that of lightning initiated
from tall structures (e.g., Rakov, 1999). This chapter analyzes the action integral (energy
per unit resistance at the channel termination point) and other characteristics of the ICC
pulses, including duration, rise-time, pulse peak, half-peak width, and charge in rockettriggered lightning The following comparisons are made: (a) ICC pulses in triggered
lightning recorded at the ICLRT in 2002 and 2003 (relatively high sampling rate) vs.
their counterparts recorded earlier (relatively low sampling rate), (b) ICC pulses in
triggered lightning vs. those in object-initiate lightning, and (c) ICC pulses in triggered
89
90
lightning vs. M-component (pulses superimposed on the continuing current that follows
the return strokes) pulses in triggered lightning.
RS2
Wire explosion
Initial stage
ICC pulses
RS1
Figure 4-1: Flash 03-31, bipolar flash. The negative initial stage is followed by 2 return
strokes (RS1 and RS2) of opposite polarity. In the inset the instant at which
wire explosion takes place is shown.
In Figure 4-1, the initial stage contains a number of ICC pulses superimposed on a
slowly varying current waveform. Figure 4-2 illustrates ICC pulses with relatively short
(ICC1 and ICC2) and relatively long (ICC3 and ICC4) risetimes, as well as a typical
return-stroke pulse (RS1). Definitions of the characteristics of ICC pulses analyzed here
are illustrated in Figure 4-3.
91
ICC3
ICC4
ICC2
ICC1
RS1
Figure 4-2: Flash 03-31. Comparison of ICC pulses with relatively short (ICC1 and
ICC2) and relatively long (ICC3 and ICC4) risetimes and a typical returnstroke pulse (RS1).
While computing the charge and action integral for the ICC pulses, the background,
slowly-varying current is approximated by an imaginary line (dashed line in Figure 4-4),
the background is subtracted, and charge and action integral are computed using the
modified current pulse shown in the inset of Figure 4-4.
92
4.1.2 Statistical Characteristics of ICC Pulses
The ICC pulses analyzed here occurred in 11 flashes triggered at the ICLRT at
Camp Blanding, Florida, in 2002 and 2003. Histograms of the distributions of the
parameters are shown in Figures 4-5 to 4-22. These are compared with statistics, found in
Miki et al. (2004), of the ICC pulses in rocket-triggered lightning analyzed previously
and ICC pulses in object-initiated lightning derived from current measurements on 1) the
Gaisberg tower (100 m, Austria), 2) the Peissenberg tower (160 m, Germany), and 3) the
Fukui chimney (200 m, Japan). Miki et al. (2004) found that the characteristics of ICC
pulses in object-initiated lightning are similar within a factor of two, but differ more
Charge = i(t) dt
Action Integral = i(t)2dt
Flash 03-31
Figure 4-3: Definitions of parameters (peak, duration, rise time, half-peak width;
additionally shown is the preceding continuous current level) of ICC pulses.
Integration in evaluating charge and action integral is over the duration of the
pulse. Peak= 0.31 kA, duration=1.9 ms, risetime=137 s, half-peak
width=0.43 ms, charge = 86 mC, Action integral = 6.1 A2s.
Flash F029
Current, A
Current, A
93
Time, s
Time, s
Figure 4-4: Illustration of the removal of the background continuous current in computing
charge and action integral. The dashed line represents the imaginary line,
which is used as reference. The charge and action integral are computed by
integrating the modified current (or the square of current) as shown in the
inset. In this example, charge = 80 mC, action integral = 7.8 A2s.
significantly from their counterparts in triggered lightning. The triggered-lightning data
analyzed by Miki et al. (2004) were acquired at the ICLRT in 1996, 1997, 1999, and
2000. The ICC pulses in object-initiated lightning exhibit larger peaks, shorter rise times,
and shorter half-peak widths than do the ICC pulses in 1996, 1997, 1999, and 2000
rocket-triggered lightning. The rocket-triggered lightning currents were recorded on a
magnetic tape and later digitized. The sampling interval was 40 s for the years 1996 and
1997, 80 s for the year 1999, and few microseconds for 2000. In contrast, sampling
intervals for the years 2002 and 2003 considered here were 1 s and 0.5 s, respectively.
The original objective of the analysis whose results are presented in Figure 4-5 through
94
Figure 4-22 was to examine the dependence of the statistics of the parameters of ICC
pulses in rocket-triggered lightning on the sampling interval. This is particularly
important for the rise time that can be smaller than the sampling intervals used in 1996
and 1997 (40 s) and 1999 (80 s).
18
GM=232 A
Max=1076 A
Min=6.3 A
Sample size=66
16
14
Number
12
10
8
6
4
2
0
0
32
64
128
256
Pulse Peak, A
512
Figure 4-5: Histograms of the peak of the ICC pulses for 2002.
1024
2048
95
18
GM=76.8 A
Max=490.6 A
Min=17.7 A
Sample size=50
16
14
Number
12
10
8
6
4
2
0
32
64
128
256
512
1024
2048
Pulse Peak, A
Figure 4-6: Histograms of the peak of ICC pulses for 2003.
25
GM=144.1 A
Max=2082 A
Min=6.3 A
Sample size=116
20
Number
15
2002
2003
10
32
64
128
256
512
1024
Pulse Peak, A
Figure 4-7: Histogram of the peak of ICC pulses for 2002 and 2003.
2048
96
25
20
GM=3.7 ms
Max=16.3 ms
Min=0.8 ms
Sample size=66
Number
15
10
0.5
16
32
Duration, ms
Figure 4-8: Histogram of the duration of ICC pulses for 2002.
30
25
GM=4.6 ms
Max=15.4 ms
Min=0.76 ms
Sample size=50
Number
20
15
10
5
0
0.5
Duration,Fms
16
32
97
50
GM=4.1 ms
Max=16.3 ms
Min=0.76 ms
Sample size=116
45
40
Number
35
30
2002
2003
25
20
15
10
5
0
0.5
16
32
Duration, ms
Figure 4-10: Histogram of the duration of ICC pulses for 2002 and 2003.
20
GM=0.36 ms
Max=2.1 ms
Min=0.04 ms
Sample size=66
18
16
Number
14
12
10
8
6
4
2
0
16
32
64
128
256
512
1024
Risetime, s
2048
4096
98
20
GM=0.46 ms
Max=2.9 ms
Min=0.06 ms
Sample size=50
18
16
Number
14
12
10
8
6
4
2
0
4
16
32
64
128
256
Risetime,s
512
1024
2048
35
30
Number
25
2002
2003
20
15
10
5
0
4
16
32
64
128
256
512
1024
2048
Risetime, s
Figure 4-13: Histogram of the risetime of ICC pulses for 2002 and 2003.
4096
4096
99
25
GM=0.83 ms
Max=4 ms
Min=0.18 ms
Sample size=66
Number
20
15
10
0
16
32
64
128
256
512
1024
2048
4096
8192
Half-peak width, s
Figure 4-14: Histogram of the half-peak width of ICC pulses for 2002.
25
GM=1.4 ms
Max=5.4 ms
Min=0.26 ms
Sample size=50
20
Number
15
10
16
32
64
128
256
512
1024
2048
4096
Half-peak width, s
Figure 4-15: Histogram of the half-peak width of ICC pulses for 2003.
8192
100
40
GM=0.98 ms
Max=5.4 ms
Min=0.18 ms
Sample size=116
35
Number
30
2002
2003
25
20
15
10
5
0
16
32
64
128
256
Half-peak
width,
512
1024 s 2048
Half-peak width, s
4096
8192
Figure 4-16: Histogram of the half-peak width of ICC pulses for years 2002 and 2003.
20
GM=100 mC
Max=550 mC
Min=4 mC
Sample size=66
18
16
14
Number
12
10
8
6
4
2
0
0
50
100
150
200
250
300
350
400
450
Charge, mC
Figure 4-17: Histogram of the charge of ICC pulses for 2002.
500
550
101
Number
GM=90 mC
Max=450 mC
Min=10 mC
Sample size=50
50
100
150
200
250
300
350
400
450
500
550
Charge, mC
Figure 4-18: Histogram of the charge of ICC pulses for 2003.
GM=96 mC
Max=550 mC
Min=4 mC
Sample size=116
Number
2002
2003
50
100
150
200
250
300
350
400
450
500
Charge, mC
Figure 4-19: Histogram of the charge of ICC pulses for years 2002 and 2003.
550
102
Number
GM=15.2 A2s
Max=378.8 A2s
Min= 0.08 A2s
Sample size=66
50
100
150
200
250
300
350
400
Figure 4-20: Histogram of the action integral of ICC pulses for years 2002.
Number
GM=4.5 A2s
Max=63.1 A2s
Min= 0.11 A2s
Sample size=50
50
100
150
200
250
300
350
400
103
GM=9.0 A2s
Max=378.8 A2s
Min= 0.08 A2s
Sample size=116
Number
2002
2003
50
100
150
200
250
300
350
400
Action Integral, A s
Figure 4-22: Histogram of the action integral of ICC pulses for years 2002 and 2003.
Histograms of the peak, duration, rise time and, half-peak width of ICC pulses in
lightning triggered from tall objects and in rocket-triggered lightning reported by Miki et
al., (2004) are shown in Figures 4-23 to 4-26. They did not present any charge and action
integral statistics for ICC pulses. As stated earlier, the statistics for the rocket-triggered
lightning were obtained from experiments conducted at the ICLRT in 1996, 1997, 1999,
and 2000.
104
Figure 4-23: Histograms of the peak of ICC pulses. The geometric mean (GM),
maximum (MAX), and minimum (min) values are indicated on each
histogram. The rocket-triggered lightning data presented in this figure were
obtained at the ICLRT at Camp Blanding, Florida, in 1996, 1997, 1999, and
2000.
105
Figure 4-24: Histograms of the duration of ICC pulses. The geometric mean (GM),
maximum (MAX), and minimum (min) values are indicated on each
histogram. The rocket-triggered lightning data presented in this figure were
obtained at the ICLRT at Camp Blanding, Florida, in 1996, 1997, 1999, and
2000.
106
Figure 4-25: Histograms of the risetime of ICC pulses. The geometric mean (GM),
maximum (MAX), and minimum (min) values are indicated on each
histogram. The rocket-triggered lightning data presented in this figure were
obtained at the ICLRT at Camp Blanding, Florida, in 1996, 1997, 1999, and
2000.
107
Figure 4-26: Histograms of the half-peak width of ICC pulses. The geometric mean
(GM), maximum (MAX), and minimum (min) values are indicated on each
histogram.
Table 4-1 summarizes the ICC pulse parameters (geometric mean values) for 2002 and
2003 rocket-triggered lightning experiments and compares those with their counterparts
for 1996, 1997, 1999, and 2000, analyzed by Miki et al. (2004).
ICLRT
Miki
et al.
(2004)
This
Study
Gaisberg
Tower
Peissenberg
Tower
Fukui
Chimney
Miki
et al.
(2004)
Action
Integral,
A2s
110
113
N=296
76.6
2.59
N=254
3.18
464
N=267
517
2002
66
232
3.7
360
800
100
15.2
2003
50
76.8
4.6
460
1400
90
4.5
2002+2003
116
144
4.1
400
1000
96
8.9
1996-2003
473-522
111.4
3.0
461
987
2000
344
377
1.2
10
276
1996-1999
124
512
0.83
61
153
1996-1999
231
781
0.51
44
141
247-296
108
ObjectInitiated
Lightning
Charge,
mC
Halfpeak
width,
us
943
N=247
1079
1996,1997,
1999
2000
Duration, Risetime,
ms
s
109
As seen in Table 4-1, the geometric means of the parameters of the ICC pulses in
rocket-triggered lightning from 2002 and 2003 are consistent with those for 1996, 1997,
1999, and 2000, suggesting that they are not influenced by different sampling intervals.
ICC pulses in object-initiated lightning exhibit larger peaks, shorter risetimes, and shorter
half-peak widths than do the initial-stage pulses in rocket-triggered lightning.
Two possible reasons are proposed in Miki et al. (2004). First, multiple upward
branches could have facilitated the simultaneous occurrence of a continuous current in
one branch and a downward leader in another branch in object-initiated flashes, as
observed for Monte San Salvatore and Ostankino tower flashes (Berger, 1967; Gorin et
al., 1975). Second, the charge sources for initial-stage current in thunderclouds over the
tall objects in Austria, Germany, and Japan might be located closer to the lightning
attachment point than the sources of initial-stage current pulses in Florida. Hence,
because of the shorter propagation path between the in-cloud source and the lightning
attachment point, the fronts of the downward-propagating current waves in objectinitiated flashes might have suffered less degradation due to dispersion and attenuation
than their counterparts in Florida rocket-triggered flashes. Table 4-2 which compares the
parameters of ICC pulses in Gaisberg tower flashes in winter and summer supports the
latter hypothesis. It is known that the cloud charge sources in winter are lower than those
in summer.
Hence, it is expected that the ICC pulses in winter flashes should exhibit larger
peaks, shorter rise times, and shorter half-peak widths (HPW) than the ICC pulses in
summer flashes. As seen in Table 4-2, all parameters except for the risetime are similar.
More data are needed to arrive at a more decisive conclusion. Charge and action integral
110
for ICC pulses are considerably smaller than their counterparts for return strokes (e.g.,
Rakov, 1999).
Table 4-2: Parameters of ICC pulses in Gaisberg tower flashes as a function of season.
Adapted from Miki et al. (2004)
Season
Sample size Peak, A
Duration, ms Risetime, s HPW, s
Winter
36
319
1.03
74.9
298
Summer
38
368
1.25
134
269
4.2 M-Components
4.2.1 Introduction
M-components are impulsive processes that occur during the continuing current
following the return strokes. In this chapter, statistics are compiled for the following
parameters of the M-component pulse: magnitude, rise time, duration, half-peak width,
charge, and action integral. The purpose of this is to compare these statistics to their
counterparts for the ICC pulses occurring during the initial stage. The same triggeredlightning current records as in Section 4.1 are used here. More information on the return
stroke and M-component pulses can be found in Fisher et al. (1993) and Rakov and
Uman (2003).
4.2.2 Statistical Characteristics of M-Components
An example of a triggered-lightning flash, which contains the initial stage, return
strokes, and M-components, is shown in Figure 4-27. Definitions of the characteristics of
M-components analyzed here are illustrated in Figure 4-28. A total of 72 M-components
in 14 flashes triggered at the ICLRT at Camp Blanding, Florida, in 2002 and 2003 are
used for analysis here.
111
Current, A
Current, A
Initial stage
M
RS1
Time, s
RS1 RS2 RS3
Time, s
Figure 4-27: Flash F0213. The initial stage is followed by 3 return strokes (RS1, RS2 and
RS3). In the inset, 3 M-components (labeled M) following return stroke RS1
are shown.
The histograms of duration, half peak width, peak and risetime of the Mcomponents are shown in Figures 4-29 to 4-34. Table 4-3 compares the characteristics of
M-components obtained at ICLRT, NASA Kennedy Space Center, Florida, and in
Alabama with those of ICC pulses obtained at ICLRT during 1996 to 2003.
112
Pulse Peak
Half-peak width
Risetime (10-90%)
Charge = i(t) dt
Action Integral = i(t)2dt
Time, ms
Figure 4-28: Flash F0213. Definitions of parameters (peak, duration, rise time, half-peak
width) of M-components. Peak= 70.3 A, duration=15.2 ms, rise time=0.9 ms,
half-peak width=1.8 ms.
30
2002
GM = 2.2ms
N =32
25
20
Number
Current, A
Duration
2003
GM = 2.4 ms
N = 40
15
GM = 2.3 ms
Min = 0.06 ms
Max = 17.2 ms
N = 72
10
5
0
10
12
14
Duration, ms
Figure 4-29: Histogram of duration of M-component pulse.
16
18
113
14
2002
GM = 198 A
N = 32
12
2003
GM = 125 A
N = 40
GM = 154 A
Min = 10.8 A
Max = 3578 A
N = 72
Number
10
8
6
4
2
0
16
32
64
128
256
512
1024
2048
4096
Peak, kA
Figure 4-30: Histogram of peak of M-component pulse.
16
2002
GM = 220.3 s
N =32
14
12
2003
GM = 252 s
N = 40
Number
10
GM = 237 s
Min = 23 s
Max = 2640 s
N = 72
6
4
2
0
0
64
128
256
512
Risetime, s
1024
2048
4096
114
35
2002
GM = 426 s
N =32
30
2003
GM = 452 s
N = 40
Number
25
20
GM = 440 s
Min = 44 s
Max = 3600 s
N = 72
15
10
5
0
128
512
1024
2048
4096
Half-peak width, s
Figure 4-32: Histogram of the half-peak width of M-component pulse.
30
2002
GM =70 mC
N = 32
35
2003
GM = 117 mC
N = 40
30
Number
25
GM =91.5 mC
Min = 1 mC
Max = 909 mC
N =72
20
15
10
5
0
100
200
300
400
500
600
700
800
Charge, mC
Figure 4-33: Histogram of the charge of M-component pulse.
900
1000
115
70
2002
GM = 11.0 A2s
N = 32
60
Number
50
2003
GM =13.0 A2s
N = 40
40
GM = 12.1 A2s
Min = 0.03 A2s
Max = 1940 A2s
N = 72
30
20
10
0
200
400
600
800
1000
1200
1400
1600
1800
2000
Action Integral, A s
Figure 4-34: Action integral of M-component pulse.
The geometric means of M-component current peak and duration obtained here are
similar to those previously reported by Thottappillil et al. (1995). The geometric means
of risetime and half-peak width are smaller by a factor of about two compared to those
obtained by Thottappillil et al. (1995). The probable reason for this discrepancy might be
the fact that in Thottappillil et al. (1995), some overlapping M components which do not
allow unambiguous measurement of such parameters as risetime, duration, and half-peak
width were not used while compiling the statistics. These overlapping M components
usually occur during the first 5 ms following the beginning of the return stroke and have a
faster rise time. Added to this, sample sizes are very small compared to those used in
116
Thottappillil et al. (1995). Table 4-3 gives the statistics of the characteristics of M
components along with those of ICC pulses.
Table 4-3: Geometric means of the various parameters of M-components and ICC pulses.
M- components
Experimental
Site
Lightning
triggering
sites in
Florida and
Alabama
Thottappillil
1990,1991
et al. (1995)
This
Study
2002
2003
20022003
HalfPeak
width,
us
Action
Charge,
Integral,
mC
A2s
422
N=124
800
N=113
129
N=104
2.2
2.4
220
252
426
452
70
117
11
13
2.3
237
440
92
12
Sample
Size
Peak,
A
Duration, Risetime,
ms
s
113124
117
N=24
2.10
N=114
32
40
198
125
72
154
ICC pulses
Miki et al.
(2004)
ICLRT
This study
(Table 4.1)
1996,
97,99
2000
2002
2003
20022003
19962003
247296
110
66
50
113
N=296
76.6
232
76.8
2.6
N=254
3.2
3.7
4.6
464
N=267
517
360
460
943
N=247
1079
800
1400
100
90
15.2
4.5
116
144
4.1
400
1000
96
9.0
473522
111
3.0
461
987
117
ICLRT
Source
Year
118
As seen Table 4-3, the duration, risetime, and half-peak width of the ICC pulses are
approximately twice those of M-components, suggesting a frequent occurrence of slower
pulses in the initial stage compared to the pulses superimposed on the continuing currents
following return strokes. The charge and action integral of M-components and ICC
pulses are similar.
CHAPTER 5
RECOMMENDATIONS FOR FUTURE RESEARCH
1.
The frequency range of the Pockels sensor measuring system used in 2000 was
relatively narrow, from 50 Hz to 1 MHz. This led to underestimation of the peaks
of the fast electric field waveforms by about 40%. Pockels sensors with higher
upper frequency response should be used to overcome this limitation.
2.
The sampling interval of the vertical electric field was 0.5 s compared to 20 ns of
the high current records. As seen in Section 3.4.4, peak power estimates are
sensitive to the uncertainties related to the relatively large electric field sampling
interval. Hence, smaller sampling intervals (higher sampling rates) should be used
in future experiments to get more accurate estimates of the peak power.
3.
The high-current record length was 50 s. Longer current records are needed to
obtain the power and energy curves for later times.
4.
The estimation of power and energy in this study assumes that the lightning
channel is vertical, but in reality the lightning channel is tortuous and drifts because
of wind. Hence, optical sensors should be placed around the strike rod to identify
those flashes for which the channel is relatively straight and vertical, so that more
accurate peak power and energy estimates could be obtained.
5.
Along with the vertical electric fields, horizontal electric fields were also measured
for eight strokes. Unfortunately, the unavailability of high-current records for those
flashes prevented the estimation of the Poynting vector associated with the upwardmoving wave. It would be interesting to obtain such estimates in the future.
6.
7.
Sample sizes in this study were rather small. Additional measurements are needed
to obtain a larger sample that would allow one to draw more statistically significant
conclusions.
119
LIST OF REFERENCES
A. S. Bizjaev, V. P. Larionov, and E. H. Prokhorov, Energetic characteristics of
lightning channel, in Proc. 20th Int. Conf. Lightning Protection, Interlaken,
Switzerland, pp.1.1/11.1/3, Sept. 1990.
J. E. Borovsky, An electrodynamic description of lightning return strokes and dart
leaders: Guided wave propagation along conducting cylindrical channels,
J.Geophys. Res., pp.1002697-1002726, Feb. 1995.
J. E. Borovsky, Lightning Energetics: Estimates of energy dissipation in channels,
channel radii, and channel-heating risetimes, J.Geophys. Res., vol.103, pp.1153711553, May. 1998.
J. D. Cobine, Gaseous Conductors, sec. 7.11, McGraw-Hill, New York, 1941.
D. E. Crawford, V. A. Rakov, M. A. Uman, G. H. Schnetzer, K. J. Rambo, M. V.
Stapleton, and R. J. Fisher, The close lightning electromagnetic environment:
Dart-leader electric field change versus distance, J. Geophys Res., vol.106,
pp.1490914917, 2001.
E. I. Dubovoy, V. I. Pryazhinsky, and V. E. Bondarenko, Numerical modeling of the
gasodynamical parameters of a lightning channel and radio-sounding reflection,
Izvestiya AN SSSR-Fizika Atmosfery i Okeana., vol. 27, pp. 194203, 1991.
E. I. Dubovoy, V. I. Pryazhinsky, and G. I. Chitanava, Calculation of energy dissipation
in lightning channel, Meteorologiya i Gidrologiya, no. 2, pp. 4045, 1991.
E. I. Dubovoy, M. S. Mikhailov, A. L. Ogonkov, and V. I. Pryazhinsky, Measurement
and numerical modeling of radio sounding reflection from a lightning channel, J.
Geophys. Res., vol.100, pp. 14971502, 1995.
A. A. Few, J. Dessler, D. J. Latham, and M. Brook; Power spectrum of thunder,
J.Geophys. Res., pp.746926-746934, 1969.
R. J. Fisher, G. H. Schnetzer, R. Thottappillil, V. A. Rakov, M. A. Uman, and J. D..
Goldberg, Parameters of triggered-lightning flashes in Florida and
Alabama, J. Geophys Res., vol.98, pp.22887-22902, 1993.
P. Henry, G. Portier, and H. Coin,Proceedings in Atmospheric Electricity, Deepak,
Hampton, Virginia, 1994.
120
121
R. D. Hill, Channel heating in return stroke lightning, J. Geophys. Res., vol. 76, pp.
637645, 1971.
R.D. Hill, Energy dissipation in lightning, J. Geophys. Res., vol. 82, pp. 49674968,
1977.
V. Kodali, Characterization and analysis of close lightning electromagnetic fields,
Masters thesis, University of Florida; 2003.
E.P. Krider, Some photoelectric observations of Lightning, J.Geophys. Res., pp.30953098, 1966.
E.P. Krider, G.A Dawson, and M.A.Uman, The peak power and energy dissipation in a
single-stroke lightning flash, J.Geophys. Res., pp.733335-3339, 1968.
M. Miki, V.A. Rakov, K.J. Rambo, G.H. Schnetzer, and M.A. Uman; "Electric Fields
Near Triggered Lightning Channels Measured with Pockels Sensors ," J. Geophys.
Res., vol.107 (D16), 2002.
A. H. Paxton, R. L. Gardner, and L. Baker, Lightning return stroke: A numerical
calculation of the optical radiation, Phys. Fluids, vol. 29, pp. 27362741, 1986.
A. H. Paxton, R. L. Gardner, and L. Baker, Lightning return stroke: A numerical
calculation of the optical radiation, in Lightning Electromagnetics, New York:
Hemisphere, pp. 47-61, 1990.
J.M. Picone, J.P. Boris, J.R. Grieg, M. Rayleigh, and R.F. Fernsler, Convective cooling
of lightning channel, J. Atmos. Sci, pp.382056-62, 1981.
M.N. Plooster, Numerical model of the return stroke of the lightning discharge,
Phys.Fluids, pp.142124-2133, 1971.
V. A. Rakov, Lightning Discharges Triggered using Rocket-and-Wire Techniques," J.
Geophys. Res., vol.100, pp.25711-25720, 1999.
V.A. Rakov, Some inferences on the propagation mechanisms of dart leaders and return
strokes , J. Geophys. Res., vol.103, pp.1879-1887, 1998.
V. A. Rakov, D. E. Crawford, K. J. Rambo, G. H. Schnetzer, M. A. Uman, and R.
Thottappillil, M-component mode of charge transfer to ground in lightning
discharges, J. Geophys. Res., vol.106, pp.22817 22831, 2001.
V.A. Rakov, and M. A. Uman, Lightning: Physics and Effects; Cambridge University
Press, New York; 2003.
122
V. A. Rakov, M. A. Uman, D. Wang, K. J. Rambo, D. E. Crawford, and G. H. Schnetzer,
Lightning properties from triggered-lightning experiments at Camp Blanding,
Florida, (1997 1999), paper presented at the 25th International Conference on
Lightning Protection, Univ. of Patras, Rhodes, Greece, Sept. 1822, 2000.
J. Schoene, M.A. Uman, V.A. Rakov, V. Kodali, K.J. Rambo, and G.H. Schnetzer,
Statistical characteristics of the electric and magnetic fields and their time
derivatives 15 m and 30 m from triggered lightning, J. Geophys. Res., vol.108,
2003.
D. F. Strawe, Non-linear modeling of lightning return strokes, in Proc. Fed. Aviat.
Administra./Florida Inst. Technol.Workshop Grounding Lightning Technol.,
Melbourne, FL, Rep. FAARD- 79-6, pp. 915, Mar. 1979.
R. Thottappillil, J. D.. Goldberg, V. A. Rakov, and M. A. Uman, Properties of M
components from currents measured at triggered lightning channel base, J.
Geophys Res., vol.100, pp.25711-25720, 1995.
M. A. Uman, The Lightning; sec. 7.3.3, Dover Publications, Minneola, New York; 1984.
M. A. Uman, The Lightning Discharge; Dover Publications, Minneola, New York; 1987.
M.A. Uman, and R.E. Voshall, Time interval between lightning strokes and the
initiation of dart leaders, J.Geophys. Res., vol.73, pp.497-506, 1968.
BIOGRAPHICAL SKETCH
Vinod Jayakumar was born in Vellore, India, in 1980. He graduated with a
bachelors degree in electronics and communications from P.S.G college of technology at
Coimbatore, Tamil Nadu, India in 2002. In 2002, he went to the USA to pursue graduate
studies at the University of Florida.
123