767 - TB

B2
Overhead lines
Vegetation fire characteristics

and the potential impacts on
overhead line performance
Reference: 767
June 2019
TB 767 - Vegetation fire characteristics and the potential impacts on overhead line performance
Vegetation fire
characteristics and the
potential impacts on
overhead line performance
WG B2.45
Members
H.F. VOSLOO, Convenor ZA H. HAWES AU

A. BRITTEN ZA P. FROST ZA
F. LIRIOS AU M. LEE AU
R. NEL ZA J. CALITZ ZA
J. FERNANDES BR H. VALENTE PT
Reviewers
V. NAIDOO NO P. DULHUNTY AU
Copyright © 2019
“All rights to this Technical Brochure are retained by CIGRE. It is strictly prohibited to reproduce or provide this publication in any
form or by any means to any third party. Only CIGRE Collective Members companies are allowed to store their copy on their
internal intranet or other company network provided access is restricted to their own employees. No part of this publication may
be reproduced or utilized without permission from CIGRE”.
Disclaimer notice
“CIGRE gives no warranty or assurance about the contents of this publication, nor does it accept any responsibility, as to the
accuracy or exhaustiveness of the information. All implied warranties and conditions are excluded to the maximum extent permitted
by law”.
WG XX.XXpany network provided access is restricted to their own employees. No part of this publication may be
reproduced or utilized without permission from CIGRE”.
ISBN : 978-2-85873-469-6
Disclaimer notice
“CIGRE gives no warranty or assurance about the contents of this publication, nor does it accept any
ISBN : 978-2-85873-469-6
Executive summary
The subject of fires under power lines proved to be a wide one. It covers many related areas in an attempt to reach
a wide readership with an interest in this subject. Chapter 1 introduces the subject of fires.
The question of why electric utilities should be concerned about fires under their lines is addressed in chapter 2
where the costs associated with fires are discussed. This not only deals with the cost of fire to society, but also
focusses on those costs to the electric utility stemming from fires.
Because by its very nature, fires are dangerous, and that danger increases when they burn close to live
conductors, chapter 3 is dedicated to understanding the risks and indicate ways in which these risks may be
reduced.
The subject of the insulation breakdown of gasses during a fire has been the subject of many studies, spanning a
number of decades. In chapter 4 the history of work that was done is reviewed, and further expanded by looking at
work that was done outside of the classic electrical engineering fraternity by adding the biochemical processes
which occur during the combustion of vegetation material under power lines. The theory of insulation breakdown
and conductivity of fires is discussed in annexure A.
In order to find a model to describe the breakdown process in engineering terminology, one of the correspondents,
Mr Tony Britten, developed a model for this purpose and also tests the assumptions with some real data. This is
dealt with in chapter 5.
The terminology used with fire is an important first step in understanding how fires burn. In 6, the “anatomy” of fire
is explained to the reader as a precursor to 7.
In any study of fire, fire behaviour must be considered. In 7 this concept is defined for the reader and the aspects
that influence this behaviour are discussed. This work has largely been done by institutions and persons involved in
the prevention and suppression of fire as well as students of pastoral science, where the effects of fire on
vegetation are of concern. This knowledge will not only prove important in the planning of vegetation management
strategies but may also be used in the design and placement of towers for new lines in fire-prone areas.
As will be seen in 8, climate and weather have a profound effect on the occurrence and behaviour of fires. Because
of these influences, it is of course also possible to predict the onset of fires by prediction of the weather. This has
been proven as a valuable aid to system operators.
It was not possible to deal with every subject in detail; however, the Technical Brochure will raise the awareness of
the reader to those aspects that needs to be considered when dealing with a fire under power lines.
Some of the mathematical treatments appear in the text. More detailed work is taken up in the annexure.
5
Table of Content
Executive summary ....................................................................................................5
1. Introduction and Background .............................................................................9

1.1 Background................................................................................................................................... 9
2. Costs associated with fires under power lines ................................................11

2.1. Societal costs of fires................................................................................................................. 11
2.2 Costs associated with loss of supply ....................................................................................... 12
2.3 The effect of fires on overhead conductors ............................................................................. 13
2.4 The effect of fires on Tower structures (steel and wood) ....................................................... 15
2.5 Fire damage to other equipment on lines ................................................................................. 18
2.6 Fire effects on substations,transformers and switchgear ...................................................... 21
2.7 Conclusion .................................................................................................................................. 22
3. Danger of fires under OHL to firemen and the public ......................................23

3.1 Introduction ................................................................................................................................. 23
3.2 Conditions under which fire induced flashover are likely to occur ....................................... 23
3.3 Step and touch potential ............................................................................................................ 25
3.4 Minimum safe working distance from overhead line............................................................... 28
3.5 Calculating the horizontal distance to a point of discharge under windy conditions .......... 28
3.6 Resistance to ground of the Arcing Point ................................................................................ 30
3.7 Magnitude of the fault current ................................................................................................... 31
3.8 The generated step voltage ....................................................................................................... 33
3.9 Safe distance between a fireman and the point of discharge ................................................. 33
3.10 The extent of the hazardous zone ............................................................................................. 36
3.11 An approximate solution ............................................................................................................ 36
3.12 Conclusion .................................................................................................................................. 38
4. Insulation breakdown during fires under overhead lines................................41

4.1 Introduction ................................................................................................................................. 41
4.2 The electrical breakdown of a gas ............................................................................................ 41
4.3 Overview and conclusions of fire experiments ....................................................................... 42
4.4 The reduced air density theory ................................................................................................. 50
4.5 Particle initiated flashover theory ............................................................................................. 51
4.6 Chemistry in flames and combustion ....................................................................................... 53
4.7 Temperatures of flames ............................................................................................................. 56
4.8 Flame conductivity ..................................................................................................................... 57
5. Proposed model for the breakdown mechanism in a fire................................59

5.1 Introduction ................................................................................................................................. 59
5.2 Basis and rationale for the model used to predict the likelihood of flashover ..................... 59
5.3 The DC Case. .............................................................................................................................. 61
5.4 Application of the DC (direct current) model to fire flashovers ............................................. 68
5.5 The AC (alternating current) case ............................................................................................. 69
5.6 Measurement of AC voltages and currents in fire conditions ................................................ 73
5.7 Measurements of fire-induced corona in transmission lines ................................................. 76
5.8 Predicted performance of transmission lines under cane fire conditions ............................ 78
5.9 Comparison of Eskom results with Lanoie & Mercure (1997) ................................................ 81
5.10 Discussion .................................................................................................................................. 83
5.11 Concluding remarks ................................................................................................................... 84
6
6. Fire ......................................................................................................................85
6.1 Introduction ................................................................................................................................. 85
6.2 Fire initiation ............................................................................................................................... 87
6.3 The anatomy of a fire.................................................................................................................. 90
7. Fire behaviour ....................................................................................................93

7.1 Factors influencing fire behaviour ............................................................................................ 94
7.2 Available heat energy. ................................................................................................................ 96
7.3 Rate of energy release and fire intensity. ................................................................................. 97
7.4 Vertical distribution of heat energy........................................................................................... 98
7.5 Fuel and its characteristics ..................................................................................................... 100
7.6 Air Temperature and relative humidity ................................................................................... 111
7.7 Topography ............................................................................................................................... 112
7.8 Fire barriers............................................................................................................................... 113
7.9 Conclusion ................................................................................................................................ 114
8. Fire climate and fire weather ...........................................................................115

8.1 Introduction ............................................................................................................................... 115
8.2 Fire and wind ............................................................................................................................ 115
8.3 South Africa .............................................................................................................................. 117
8.4 Australia .................................................................................................................................... 121
8.5 The United States of America .................................................................................................. 124
8.6 Mediterranean countries .......................................................................................................... 126
8.7 Conclusion ................................................................................................................................ 126
9. Fire danger rating systems..............................................................................127

9.1 Overview.................................................................................................................................... 127
9.2 US National fire danger rating system .................................................................................... 127
9.3 Australian fire danger systems ............................................................................................... 128
9.4 Southern African fire danger index systems ......................................................................... 131
9.5 Conclusion ................................................................................................................................ 133
10. Predicting fire induced flashovers ...............................................................135

10.1 Introduction ............................................................................................................................... 135
10.2 Fire Induced Flashover Probability Index (FIFPI) .................................................................. 135
10.3 Conclusion ................................................................................................................................ 141
11. Fire tracking systems ....................................................................................143

11.1 Introduction ............................................................................................................................... 143
11.2 Remote sensing: monitoring fires from Space ...................................................................... 143
11.3 Background to the Advanced Fire Information System (AFIS) ............................................ 144
11.4 Extending AFIS Functionality to other users ......................................................................... 147
11.5 New developments ................................................................................................................... 150
11.6 Conclusions .............................................................................................................................. 151
12. Mitigation of fires under power lines ............................................................153

12.1 Background............................................................................................................................... 153
12.2 Routing of power lines ............................................................................................................. 153
12.3 Vegetation management .......................................................................................................... 153
12.4 Fire Free Servitudes in sugar cane fields ............................................................................... 153
12.5 The legal approach ................................................................................................................... 153
12.6 Design and technical considerations ..................................................................................... 157
12.7 The approach followed in Mexico ........................................................................................... 157
12.8 The approach in Brazil ............................................................................................................. 157
12.9 The approach followed in Australia ........................................................................................ 158
12.10 The approach of South Africa ................................................................................................. 159
7
12.11 Fire risk mapping ...................................................................................................................... 159

12.12 Noise detection ......................................................................................................................... 160
12.13 Conclusion ................................................................................................................................ 160
13. Conclusion .....................................................................................................161
14. References .....................................................................................................163
15. Abbreviations and definitions .......................................................................171
16. Annex..............................................................................................................175
16.1 Comments received about the effect of fires on conductors ............................................... 175
16.2 The Breakdown of the insulating properties of a gas ........................................................... 177
16.3 Electrical conductivity in wildfires .......................................................................................... 195
16.4 Calculation example: The safe distance to a discharge of fault current into the soil ........ 196
16.5 Example of a cost analysis by Eskom to acquire cane free servitudes. .............................. 200
8
1. Introduction and Background

1.1 Background
Vegetation fires have been occurring on Earth for millennia and will continue to do so. Steven Pyne (1997) in his
book “World Fire” says that Earth is a uniquely fire planet. Other planetary bodies in the universe have elements of
fire: Jupiter has an ignition source in lightning, Mars has traces of free oxygen; Titan (the largest moon of Saturn)
has methane based fuel. Only Earth has all three and only Earth has the means to combine them - Earth has life.
Marine life has pumped the atmosphere with oxygen and terrestrial life has stocked the continents with carbon
fuels. (Pyne, 1997)
Earth is also the only planet with a species that can both start and stop fires. Payne states that by interfering with
the natural cycle of fires, fuel loads are not contained but rather expand and when fires then occur, they are
devastating and in many cases impossible to control (Pyne ,1997).
Except for the arctic continents, all the other continents experience fires and these fires are such a part of the
landscape that plant species subjected to them have over time adapted and are in many cases dependent on fire
for reproduction and vigorous growth. The presence of fires was recorded by many early European explorers. The
southern tip of the South American continent was named “Tierra del Fuego” (Spanish) by the Portuguese explorer
Ferdinand Magellan in 1520, who witnessed fires and smoke visible from the sea. Magellan undertook this voyage
on behalf of king Charles 1 of Spain. (Bergreen, L., 2003).
At the southern tip of Africa a similar case exists. The first recording of a veld fire was probably done by
Bartolommeo Dias when he became the first European to round the Cape in 1488. He called the current Cape St.
Francis “Ponta das Queimadas” (Axelson, 1973) (Queimada = forest fire Anon (no date). Ponta de Queimada, São
Jorge Island in the Azores, is another example.
Where power lines traverse areas where fires occur, a number of effects are experienced by these lines and, major
system disruptions have resulted from wild fires under lines. Although huge wild fires occur in Australia, South
Africa, the United States, Russia, and Mediterranean Europe, the effects of sugar cane fires and smaller grass fires
also have a considerable impact on both power lines and the rest of the electrical supply system.
Because of the problems associated with fire under power lines in many countries around the world which have
considerable impact on the total electrical system as well as on the supply of electricity to customers a work group
(WG B2-45) under Gigré Study Committee B2 was established to study this problem and produce a Technical
Brochure on the subject. This paper contains some extracts from the work done to date.
This document will discuss certain aspects of fire such as the costs resulting from fire, the weather and its influence
on the onset and behaviour of fires. Attention will also be devoted to ways to predict the inception of dangerous fire
weather as well as a novel approach which refines the conventional systems for power lines.
Finally the steps that can be taken by the designer of overhead lines (OHL) as well as the vegetation manager in
order to manage and minimize fire effects on OHL within their managed easements.
Where OHL traverse areas where fires occur, a number of effects are experienced by these lines. Major system
disruptions have resulted from wild fires under lines. Although huge wild fires in Australia, California, Colorado in
the Rockies, Russia, Greece, the effects of cane fires and smaller grass fires also have considerable impact on
OHL.
Because the problems associated with fire are experienced in many countries around the world and have
considerable impact on the total electrical system, as well as on the supply of electricity to customers, it has
become necessary to study this subject and produce a Technical Brocure based on contributions from around the
World.
This document will be the first to consider in one document, factors such as fire and the various parameters that
describe its behaviour, the various fuels and its effect on fires and the weather and it influence on the onset and
behaviour of fires, in addition to those subjects that have been studied to date by engineers and physicists.
9
Attention will also be devoted to the various impacts of fires under power lines on the whole electrical system and
its customers.
Finally, the document is of interest for the utilities, namely OHL designers, operation and maintenance (O&M)
personnel dealing with vegetation management, the system operators, but also for firefighting authorities to be
aware of the risk when working near an OHL.
10
2. Costs associated with fires under power lines

2.1. Societal costs of fires
The costs associated with fires, specifically wildfires can be enormous notably the destruction of property and even
human lives. These costs are often difficult, if not impossible to quantify. Fires affect wide areas and can
encompass all areas and components of the electrical system within a fire zone, inflicting heavy costs on the utility
and communities.
Fires are present all over the world. Uncontrolled slash and burning in Amazonia, Indonesia and India all threaten
biodiversity. Smoke plumes cover the Amazon Basin and pastoral burns pull the Sahara southwards. Multimillion-
acre wildfires regularly burn throughout the boreal forest and endless annual savanna fires burn throughout the
tropics. The Great Black Dragon fire incinerated the Hinggan forest in China, the East Kalimantan fires that turned
9 million acres of Borneo rain forest into a smoking hole. In Australia the Ash Wednesday almost brought an
industrialized nation to its knees. The conflagration of 1988 through the Yellowstone National Park swallowed
$130 million for the provision of fire fighters, without any effect on the fire. (Pyne, 1997) ). The debates and
discussions within communities after such fires have wide ranging effects. An early example is the proclamation of
forest reserves in 1891 as well as huge fires in the USA since had a significant effect on the debate and policies
concerning fire, its prevention and suppression.
America developed the most expensive firefighting operation in the world, but with a capacity to mobilize against
wildfire, being a marvel of the modern world. Only Canada could demonstrate anything like it. Within days the fire
establishment could assemble and deploy around the country tens of thousands of fire fighter, hundreds of fire
engines and tractors and the mobile kitchens, power tools and gasoline tanks, even pay fire fighters and support
staff. The suppression apparatus could flood the sky with scores of air tankers, helicopters and observation planes,
but at a cost of $1 million per day. The rehabilitation of the landscape after the Foothills fire near Boise cost
$24 million. Over the 1994 season $925 million was spent by the fire establishment (Pyne, 1997).
Fires in Greece, Italy, Portugal and Spain make the news headlines from time to time. Driven by hot, dry winds,
they leave destroyed property, vegetation and livestock in their wake. People that are not able to flee, perish.
On 12 August 2014 the press service of the Russian forestry administration reported heavy forest fires which broke
out in the Russian Far East since the forest fire season began. These fires have already destroyed huge forest
territories. According to the administration the fires have already destroyed more than 1.6 million hectares of
forests, this year alone (2014). This is double the area of forest territory destroyed in the same period the previous
year.
11
Figure 1 - Example of a fire whirl in a commercial forest fire in South Africa (photo Mauritz Bam)
Given this huge society cost of fire effects, fire research experienced an intensive era since the early 1970’s, with
the focus changing to predicting fire behaviour, and not only establishing fire danger as given by the national fire
danger rating system (NFDRS). A model for this purpose was developed by Richard Rothermel, of the Northern
Fire Lab in 1972, with the prospect of mathematical forecast of fire spread. Due to the popularity of the model, fire
officers embraced it but extrapolated it far beyond its originating conditions, necessitating redefining the model and
making it more universal. The computer based fire behaviour system that emerged was the now well-known
BEHAVE system (Pyne, 1997).
2.2 Costs associated with loss of supply

The problem with fire-induced flashover is experienced worldwide in countries such as the U.S.A., Australia, South
Africa, Brazil, Mexico, and others. The experience of each of these countries is discussed in the annexure.
Fires burning under overhead lines normally cause short-circuits resulting in the operation of the protection system.
In some extreme cases, however, the line may suffer damage to either the conductor or the structure, leading to a
permanent and sometimes prolonged outage.
In cases where the outage occurs and the protection system operates, successfully re-closing the breakers, this is
referred to as a momentary outage. Depending on the voltage of the line and the fault current, this may last
between 0.1 s to 2 s and could affect customers with continuous process plant.
The case with line faults caused by fires is regarded as more serious than other transient faults, as it has been
observed that after a successful auto-reclose the breakdown conditions still exists under the line and a second
breaker operation then results in a lock-out. (Sadurski, 1977).
A study conducted in 2001 in the USA commissioned by the Electric Power Research Institute (EPRI), the
Consortium for Electric Infrastructure for a Digital Society (CEIDS) indicated that poor power quality cost the
American economy between US$15,000 million to US$25,000 million per year (Primen, 2001).
12
In South Africa, a single fire-induced flashover causes an average of three voltage dips on the transmission
system, which can cause damage to a customer’s production ranging between R5000 and R150 000 per dip
(Taylor 1999, Vajeth et al. 2003).
Outages caused by two fires 295 km distant on two 765 kV lines in South Africa, lead to the separation of the Cape
load network on the 29th of July 2002. After the trip and lockout of the second line, the remaining 400kV lines were
not able to transmit sufficient power, causing the Cape network to island. The deficit of active power in this islanded
network resulted in a frequency drop to 47.72Hz. This in turn caused the activation of the under-frequency load
shedding scheme and a collapse of the Cape network. The outage lasted for 10 minutes and resulted in a load loss
of 1,593 MW.
Bushfires in the Sydney region, during the 4th to 6th December 2002, caused an unprecedented number of faults
on the main NSW 330 kV and 500 kV networks.
The main 330 kV / 500 kV network was subjected to approximately 70 faults due to fires between 2.30pm and
12.00am on the 4th December 2002. On the following two days more than 45 faults were recorded. These faults
caused considerable number of network outages, resulting in major reductions in supply security and the service
quality.
It was estimated that on the 4th December alone, 4,000 MWh of energy were not consumed by customers due to
loss of load caused by voltage dips. The economic impact on electricity users, on that afternoon, was in the order
of A$40 million using the value of lost load (cost of unserved energy) as A$10,000/MWh. (Ref. Investigation on the
impact of Bushfires on TransGrid Extract supplied by Francis Lirios,AU)
In other cases in Australia with a long interconnected transmission grid extending over 4,500 km there is a danger
during high fire danger season, for sustained trips to cause power swings on the network that can cause voltage
instability and major and extensive regional outages. For example, in a transmission line with 1,000 MW of
continuous load over a 500 km interregional transfer, a suddend loss of the line would cause voltage instability and
major black outs; but if for a short time the electrical load is reduced to 400MW, to allow a fire front to pass under
the line, it significantly reduces the potential risk.
2.3 The effect of fires on overhead conductors

Overhead lines are subject to different environmental conditions such as fires, lightning and even faults caused by
birds. The resulting outages not only affect the line but also impact the equipment in substations.
Although equipment is normally designed to withstand the effects of short circuits caused by these events,
repeated faulting stresses the equipment, sometimes outside their design parameters and cause premature
ageing.
A variety of responses has been obtained from utilities about damage caused to conductors of OHL during fires. As
the duration and intensity of the fire plays an important role in the effect on conductors, it is important to distinguish
between low intensity grass fires and high intensity forest fires.
13
Figure 2 - Examples showing a phase to phase fault during a bushfire and arc damage to conductors after a fire at
different location (Powerlink, 2014)
As a rule grass and shrub fires are of low intensity depending on the prevailing weather and fuel conditions. After
being exposed to a high intensity forest fire, utilities may want to determine if conductors have been damaged or
not and if the tensile strength of the conductors have been affected. Responses can be seen in Annexure B -
Damage to power lines due to fires – See Summary of Correspondence.
The response from Australian utilities with regard to the effect of fires on conductors varies and this is probably due
to the variation in fire intensities which prevail. The correspondence received about the effect of fires on the
mechanical properties of overhead conductors indicated considerable damage to overhead conductors.
TransGrid reported substantial damage done to ACSR conductors (Mango) during the Canberra fires of 2003. In
one case about 80 m of the outer layer was melted by the fire. Elsewhere there was bird-caging and it appeared
that the conductor expanded during the fire and did not return to its original diameter upon cooling. These
conditions, however, were regarded as highly unusual and evidence indicated that this fire was extreme and that a
fire tornado may have been responsible. (pers. comm. Henry Hawes, AU).
In a recent case the Electricity Department of the Ethekwini Municipality (Durban, South Africa) a fire that started in
some rubble in the servitude, spread to an adjacent timber yard where stacked timber started burning. A very
intense fire resulted and eventually the AAAC ELM conductor failed as a result of the fire. The conductors were
about 25 m above the fire. It can be concluded that because the fire remained in the area, the conductor was
heated sufficiently to cause its failing. (Duma Mhaule, pers. comm.)
PowerLink reported the impact of cane fires on older lines in north Queensland. These lines were designed before
the cane fields made an appearance and so had no extra clearance to allow for cane fires.
Flames would not normally reach conductors with their modern lines, but in these old lines the flames could reach
conductors. Being quite small, these conductors had a high heat transfer and some strand melting did occur. It was
not known how many strands were melted but repairs were required and took the form of aluminium repair rods.
These were not always very successful as they were not always at sufficient mechanical tension to lock on and
provide excellent electrical connection. As plenty of current passed through the rods, they were sometimes found to
be arcing and would later fail.
A number of respondents also raised the matter of the annealing effect of fires on aluminium conductors. It will
become clear that different opinions exist in this regard. These opinions could be as a result of the great variation in
fire intensities experienced in different parts of the world and the duration and heat yield of those fires.
According to the response received from Eskom (ZA) on this subject, indicated that annealing of the conductor is
not regarded as a problem of any significance. The reason, from a material point of view, is that most of their lines
are built with ACSR conductors in order to get the cost saving associated with building longer spans. Eskom stated
that what needs to be kept in mind for this problem is:
14
 type of conductor being exposed,

 number of strands (layers)
 time the conductor is exposed to fire
 fire intensity.
According to the response received from a materials specialist at Eskom (South Africa) various types of conductor
will respond differently to fires. Certain grades of aluminium must be heated to 300°C to 400°C and then be kept at
those temperatures for long periods ranging from one to as many as eight hours to anneal.
As the aluminium used for ACSR conductors are in the annealed condition (T0), hence any exposure to veld fires
will most possibly result in an increase in strength (heating and fast cooling due to wind). This will result in an
increase in resistance and result in lower electrical performance. Not all ACSR use fully annealed conductor;
exposure to high heat can make the grease drop out. Heat treated alloy 6201 is more susceptible to annealing.
1120 and 1350 are work hardened and annealing is well documented in the Australian Standard AS7000.
The steel core will not be affected as it is shielded by the aluminium and you need a temperature in the range of
700°C, which will result in the aluminium melting.
Figure 3 - Annealed and broken copper conductor (provided by Transend, 2014)
Transend reported a case of a broken and annealed copper conductor after a fire during December 2012 in the
Arthurs Lake area of Tasmania.
2.4 The effect of fires on Tower structures (steel and wood)

Apart from one respondent, limited damage to steel structures was reported due to fires. SPAusNet in Victoria
Australia, have reported damage to tower leg ground-level corrosion protection system as well as a case which
looks like the galvanizing that may have been damaged by a fire. ( Figure 4).
An extreme case, however, needs mentioning. During an extreme fire that occurred on the 9th of February 2009 in
Australia, a tower collapsed during the event. After analysing the steel for any material changes as a result of the
fire, it was concluded that the tower didn’t collapse as a direct result of the heat generated by the fire but by the
combination of strong winds that day with the convective effect of the fire accelerating up the head of a narrow
15
valley. Indications are that the speed up effects of fire in such situations could develop velocities in the range up to
180 km/h. Also see the extreme wind generated by the fire whirl in Figure 1and its effect on the large trees.
A structural consultant found that the failure was initiated by buckling on the two outside legs of the tower due to
the wind loads acting on the conductors and the tower.
Figure 4 - Example of fire damage ground level corrosion protection (left) and damage to galvanising (right)(provided
by SPAusNet, 2014)
Figure 5 - A tower collapsed during a fire on the 9th February 2009 at Strathewen Australia
16
Figure 6 - A failed wood pole structure caused by a grass fire in South Africa.
During a fire event, pole mounted transformers are also at risk, as is shown in Figure 7
Figure 7 - Extensive damage caused to the pole structure and pole-top transformer in South Africa
Wood pole structures can be damaged extensively by intense fires and whole structures may fail as is shown in
Figure 6
17
2.5 Fire damage to other equipment on lines
General observations from Respondents
In addition to the damage to structures discussed above, equipment attached to these structures is similarly
destroyed by fires.
TransGrid from New South Wales, Australia, reported damage from bushfires to spiral dampers and fittings with
neoprene inserts (AGS units, spacers and vibration dampers) which have required replacement. No information on
fire impacts on spacer dampers was available, but it was assumed that the heat could affect the neoprene on these
as well.
Powerlink in Queensland Australia, investigated cathodic protection systems on old tower lines recently. Where
plastic joint boxes were in use and the towers subjected to grass fires, damage occurred as shown in Figure 8. It
was noted that the plastic cover was damaged but the CP wiring was still functional. Any future replacement of
these boxes will be in fire resistant materials. They were not aware of any reported damage to overhead fittings
from fire.
SP AusNet in Victoria Australia, has experienced outages due to bushfires. The most prevalent impact of bushfires
on their lines is the flashover events involving disc insulators caused by the presence of soot and dust (Figure 9).
These required washing after the fire events (Figure 10).
They also reported cases of some polymeric strings covered with soot due to the bushfire but none thus far have
suffered any flashovers. (Figure 11)
ElectraNet in South Australia reported a similar occurrence of soot on polymeric insulators but whether this or the
flames caused the outage could not be determined. Some damage to vibration dampers by fires was reported.
(Figure 12)
Reports of fire damage to non-metallic items in fittings (e.g. neoprene inserts in helical support units, vibration and
spacer dampers) caused by the elevated temperatures were reported by some of the Australian respondents.
Figure 8 - Fire damage to plastic joint boxes (Powerlink 2013)
18
Figure 9 - Flashover marks left on glass discs due to a fire fault (SP AusNet)
Figure 10 - Glass discs pre- (left) and post-washing (right) (SP AusNet)
19
Figure 11 - An example of soot on a polymeric insulator after being exposed to bushfire in Australia (Francis Lirios)
Figure 12 - An example of damage to a spiral vibration damper after being exposed to bushfire in Australia (John Mc
Cormack)
Fire impacts on optic fibre installations

Ausgrid in New South Wales, reported that during recent bushfires over summer in the Port Stephens / Raymond
Terrace area both ADSS and OPGW on their 33 kV poles were affected. The OPGW and Joint Boxes suffered no
damage but the ADSS was damaged in areas over about 2 km, with sections of it down on the ground and other
sections having bubbling of the outer jacket. Following this, Ausgrid replaced the damaged ADSS with OPGW. In
their opinion it made a good argument to install OPGW in bushfire prone areas rather than going the cheaper
option of ADSS.
Powerlink in Queensland reported that their joint boxes will be positioned high on the tower, clear of any bushfire
danger. The fibre near ground level is no longer protected by metal and is standard fibre cable, sometimes
contained within a poly duct. Their practice now is to use a steel conduit and to provide a concrete apron around
the undergrounding point, for about 500 mm in all directions. Where this has not been done, grassfires can burn
the poly duct and cause some charring. No damage to the fibre cable inside, however, has been experienced.”
Transend in Tasmania stated that no bushfire damage/impacts to OPGW or joint boxes have been reported. For
reasons of OPGW security their joint boxes out on the line (away from a station) are located at reasonable height
up the structure and well above the climbing barrier. This helps to keep them away from bushfires. There is always
an aspiration from fibre jointers to have the joint boxes lower for easier access but Transend resists this as much
20
as possible. As a rule, the only place where they bring the fibre down the leg to ground level is within the security of
a substation site, with no vegetation fuel loads nearby and normally with a concrete pad installed around as well.
TransGrid in New South Wales has had no reports of bushfire damage to OPGW or joint boxes. The joint boxes
are normally located 5 m to 6 m above ground and above the anti-climbing device, which significantly reduces the
heat and fire impact.
2.6 Fire effects on substations,transformers and switchgear

During any short circuit condition on power lines such as may be caused by fire flashovers for instance, the effect is
evident in the substation connected to the line. The effects of these short circuits on substation equipment has
been the subject of a number of Cigre documents which will be referred to briefly below. Although often difficult to
quantify in monetary terms, these faults all have an economic inpact as it affects both the design- as well as the life
cycle costs of substations and its equipment.
TB227 (Life Management Techniques for Power Transformers) states that for large capital items such as
transformers, the direct capital cost of a replacement is usually by far the largest cost element, and for this reason it
is often difficult to justify replacement before end of life failure. However, there are sometimes situations, usually
when the indirect outage costs are very high, when a replacement can be justified before an end of life failure if the
costs of keeping suspect equipment in service are sufficiently high.
Operational stresses are usually dominated by intermittent events such as lightning strikes or short circuits. As an
example of the changing stresses over the life of a transformer, it is helpful to consider the mechanical stresses
imposed on a winding. When the transformer is new, the windings will be well clamped and therefore have a high
strength, while the electromagnetic centres of the windings will be aligned to minimise the stresses of
electromagnetic forces during short circuits. As the transformer insulation ages, the paper insulation will shrink and
may result in a reduction of clamping pressure, thereby reducing mechanical strength. If a short-circuit occurs and
the windings move slightly, the electromagnetic centres of the windings may move slightly, which will lead to much
higher stresses during subsequent faults.
The forces that act on the transformer during a short circuit event is discussed in TB209 (The Short-Circuit
Performance of Power Transformers) by WG12.19. The buckling of transformer windings during short-circuits is
also well demonstrated in this work. It is expected that a transformer will experience and survive a number of short
circuits during its service life, but sooner or later one such event will cause some slight winding movement, and the
ability of the transformer to survive further short circuits will then be severely reduced.
Periodic checking of the mechanical condition is therefore recommended. It is probably through such a process of
falling strength and increasing stresses that the mechanical condition of a transformer will degrade rapidly over a
few short-circuits immediately preceding the final failure.
In TB06 (The Mechanical Effects of short circuit currents in Open Air Substations) the authors discuss the effects
that occur during short circuit events and its implication for the design of bus systems to negate these effects. The
study concludes that based on international experience and although faillures caused by these mechanical effects
are rare, substation designers must nevertheless ensure that their designs meet the extreme requirements
associated with short circuit conditions.
TB156 (replaced by TB 528, April 2013) (Guide for Customers Specifications for Transformers -100MVA and
100kV and above) states that the extent of supply, system operating conditions and environmental considerations
need to be considered in the specification of transformers. In South Africa for instance, fire induced line faults
caused by medium intensity grass fires, account for a total of 20% of annual line faults. Fault types and fault levels
has implications for the design in order to protect the transformer. Environmental considerations such as fire
hazards and its risk of producing contaminants must be considered and the effects mitigated by means of the
transformer design.
21
2.7 Conclusion
From the above examples, it is clear that the impact of fires is wide-ranging and the financial impacts vast. It is
recommended to the reader to use the examples above in determining the economic impact of fires to the Utility
and use these to prepare the financial justification for any mitigation. Cognisance of the impact on fires also needs
to be taken during the design process of new equipment.
The reader is also referred to Vajeth (2008). This guideline shows the methodology of calculating the cost of a dip
for a stated case study and this may be used to aid the justification for new projects.
During 1992, a study was carried out to justify rising costs for the purchase of cane-free servitudes in the then
Natal Province of South Africa (see Annexure A 16.4). In the example the reader can see which costs were
considered. Please note that the currency in the example is 1993 Rand and not easily converted to current values
due to inflation and exchange rates. The example, however, shows which costs were considered and the reader
must use own current costs in any calculations.
22
3. Danger of fires under OHL to firemen and the public

3.1 Introduction
In an incident in British Columbia, Canada, a man was injured when dousing a fire under two 500kV lines
connecting BC Hydro and Bonneville Power Administration (BPA). A person burning refuse under the line, caused
a flashover which resulted in him being knocked down, as well as sustain burns due to the “step potential”
(explained below). Following the incident, the responsible Fire Department issued a “study bulletin” about the
event. This document highlighted the dangers to any person that is in the vicinity of a power line when grass or
other material burns below or close to the line. This document was then distributed to municipal fire halls and BC
Forest Service.
Due to a concern of BC Hydro officials that this document may overstate or may contain inexact information, it was
decided to conduct a study to try to quantify the problem and determine the safe and hazardous zones in the case
of a flashover as well as conditions under which flashovers are likely (or not likely) to occur. This resulted in a
bulletin on the subject by Mousa (Mousa, 1990).
During 2005, a similar incident took place in South Africa (Anon, 2005). During an extremely hot day in the
KwaZulu-Natal province, a sagging conductor made contact with a tree, which started a fire. This fire spread to a
nearby building with a thatch roof. The farm manager and assistant went to douse the fire with water from a hose.
According to witnesses, the farm manager was holding a metal rod in his hand with which he was poking at the
burning material and was wearing very thin-soled sandals. At that moment, the line was re-closed according to
standard operating procedure and a further flashover occurred in the fire, killing the farm manager.
These two incidents, although severe and not very usual, highlight the dangers to people in the vicinity of fires
burning under power lines. The utilities’ eagerness to ensure safe working practices close to power lines during fire
incidents, lead to three documents being produced to determine the safe and hazardous zones around fire induced
flashover sites (Mousa, 1990), (Bologna & Sadurski, 2001) (Chrzan, 2004). The information contained in these
documents is presented below.
The mathematical treatment of the subject is shown in this chapter, not only for the convenience of the reader, but
also to show how the results were obtained. As the lives of people may be affected by fires under power lines, the
users of this information are urged to ensure that the conditions assumed in this chapter closely matches their local
conditions and, in that way, yield correct values of the safe zones, rather than simply use the values derived here.
But first the reader will be introduced to the factors and conditions which may warn of an impending flashover
during a fire under a power line.
3.2 Conditions under which fire induced flashover are likely to occur
In the bulletin mentioned above (Mousa, 1990), the erroneous assumption was made that dense smoke in a fire
would yield a conductive path for a flashover to occur. It has been shown that dense smoke alone will not result in
a flashover. (Deno, &. Zaffanella, 1982). Because BC Hydro feared that firemen would stand back and not fight the
fire when these conditions prevailed, it was decided to look at conditions under which flashovers were likely to
occur or not.
The time at which the fire burns under the OHL is also an indicator of the likelihood of flashover. According to data
collected over 15 years in South Africa, the occurrence of grass fire induced flashovers at night is extremely rare.
This is due to low fire intensity resulting from low ambient temperatures, higher humidity and lower wind speeds
which prevail at night, compared to the daytime. This makes fire at night time less dangerous than those burning
during the day. In high intensity wild fires, however, this may not be the case and extreme caution needs to be
exercised by personnel close to power lines.
As will be explained more fully in chapter 10.2, atmospheric conditions play an important role in the occurrence or
otherwise of fire flashovers. The most important weather parameters are:
 Ambient air temperature  23 ºC
 Relative humidity  22 %
 Wind speed  4 m/s
23
When conditions approach or exceed these values, the occurrence of fire induced flashovers increases
dramatically. In the interior of South Africa these conditions are normally associated with pre-frontal weather
systems and the winds will be Westerly to North Westerly. Similar weather systems in other continents are
discussed in 8 and this should be included as part of the risk assessment.
Reduced clearance under power lines should also be considered as part of the risk assessment for possible fire
induced flashover. When lines are built, the minimum mid-span clearance is checked but subsequent actions of
land owners may change this situation. Contours in ploughed fields or dam walls are examples. This situation can
lead to a higher probability of flashover and should be checked before any burns are carried out.
The fuel which is burning is another indicator of fire intensity and possible risk of flashover. Certain plant species
contain a high percentage of volatile oils, which leads to increased intensity of the fire. South African examples are
khaki bush (tagetes minuta) or cosmos (Cosmos bipinnatus)). These plants should also be noted and taken into
account during burns. In certain cases, reduced clearances (e.g. contours) coincide with good stands of Khaki bush
or cosmos due to the leaching of agricultural nutrients to the contours. This combined situation calls for extra
caution during fires as these will increase fire intensity.
The likely discharge paths during a fire flashover is a further factor that should be considered in determining the
possible threat to persons present at an OHL during a fire. In Table 1Bologna & Sadursky shows line voltages and
likely flashover paths during a fire.
Table 1 - Likely discharge paths (Bologna & Sadurski, 2001)
Line Voltage Flashover Path *

(kV) Phase - Ground Phase - Phase Phase – Shield
22 No Yes Yes
33 No Yes Yes
44 No Yes Yes
66 No Yes Yes
88 No Yes Yes
132 Yes ** Yes Yes
275 Yes Yes Yes
400 Yes Yes *** Yes
765 Yes Yes *** Yes ***
*Flashover paths are shown in Figure 13.

**The probability of flashover of 132kV line to ground at mid-span is low. It can occur during very hot fires (flames
reaching phase conductors) which creates whirls of hot air and generates large amounts of burning coal, dust, soot
and debris. (Bologna & Sadurski, 2001)
***Low probability of flashover.
In their report, Bologna & Sadurski, (2001) found as follows:
 At least 50 % of the gap (clearance above vegetation) must be bridged by flame to cause flashover of
400 kV and 765 kV transmission lines.
 There is high probability of flashover to ground of 132 kV, 275 kV, 400 kV and 765 kV lines when the flame
is near the phase conductor. This probability increases when the flames are above phase conductors.
 For lower voltage lines (below 220 kV) the probability of flashover of phase-to-phase and of phase-to-shield
wire are much higher than phase-to-ground.
24
1 : Phase to Phase fault

Shield Wire
2 2 : Phase to Shield fault
1
3 : Phase to Ground fault
Phase Conductors 3
Figure 13 - Flashover Paths on Overhead lines (Bologna & Sadurski, 2001)
3.3 Step and touch potential

During the flash-over two types of rising voltage (or potential) need to be considered. As the current tries to find its
way back to the source (in the substation) a rise in potential will be the highest at the point where the flashover
strikes the ground. As this will normally be in or in close proximity to the flame, fire fighters will normally not be
there, but the returning current will cause a rise in potential radially away from the flash point.
The reader is referred to the Technical Brochure produced by WG B2.56 (2017) titled “Ground potential rise at
overhead AC transmission line structures during power frequency faults”. This work gives a comprehensive
treatment of step and touch potential in section 6.2.
Humans (especially when wearing well insulated boots) standing with the feet together, will typically not experience
any significant effect but in the case where a person’s feet are poorly insulated and are about 1m apart, he/she will
be subjected to an electric shock. This effect is referred to as step potential. (see a more comprehensive definition
in the TB referred to above)
In the case of animals facing the flash, for instance, the front legs will experience a higher voltage than the rear
legs. This is due to the resistivity of the soil. The body of the animal will present a path with lower resistivity and the
animal will experience a shock.
Under certain circumstances, the fault current may also find its way through the tower, raising the voltage. This rise
in voltage is referred to as touch potential. In addition to the tower, other metallic objects such as fences may also
experience a rise in voltage and could become a source of danger to the fire fighter. Depending on the materials
used in the fence, a fence could experience a rise in voltage many kilometres distant from the actual flash point. It
is especially true of poorly earthed fences. It is important that fire fighters remain clear of fences or metallic pipe
lines during firefighting operations.
25
Figure 14 - The figure shows a fence line running between Eskom’s Pluto – Minerva 400kV line and the Lulamisa –
Pluto 400kV lines where a fire induced flashover occurred.
Figure 15 - Pieces of burnt-off barbed wire is evidence of the current that was flowing through the fence during the
flashover
26
Figure 16 - This fence post was over one kilometre from the flashover site. Note how the wires were welded to the post
during the flashover
Figure 17 - A close-up of a fence wire that was welded to the post during the flashover
27
3.4 Minimum safe working distance from overhead line

The calculation of the safe working distance from a fire under a power lines, is shown by both Mousa (1990) and
Bologna & Sadurski (2001). The mathematical treatment of the problem by these authors will be shown below for
the reader who wishes to carry out calculations for their own specific set of circumstances and for easy reference.
Some tables generated by these authors will also be shown and can be used as a convenient reference without
having to do any calculations.
When there is likelihood of flashover of a line, the minimum safe working distance between fireman and the outer
conductor of the line must be maintained. This distance consists of two components, namely:
 Potential distance (horizontal) between outer phase conductor and a point of discharge of fault current into
the soil under windy conditions
 Minimum distance between a fireman and a point of discharge of fault current into the soil.
3.5 Calculating the horizontal distance to a point of discharge under windy

conditions
When the flashover occurs, the most dangerous conditions (for firemen or field staff) are created by phase-to-
ground flashovers. When there is an intensive fire on the one side of the line, a strong wind can force the flame
towards the phase conductor the flashover to ground can then occur well before the fire reaches the line (see
Figure 18). In this section hypothetical examples will be given to show how far ahead of the fire a flashover can
occur under very windy conditions and to warn personnel of this danger.
These flashovers occur on 132 kV and higher voltage lines. The phase-to-phase flashovers predominate on lower
voltage lines. Phase to shield wire flashovers can occur on all distribution and transmission lines.
Windspeed sufficient
to force flame 76°
towards line conductor
22m
8.5m
Phase Conductors
21.4m
Absence of high Area of fire with

flames long flames
Figure 18 Example of the distance between phase conductor and point of discharge under windy conditions (400 kV
line) (Bologna & Sadurski, 2001)
The distance between the outer phase of the high voltage lines (132 kV and above) and the point of strike to
ground during windy conditions are given in Table 2. The distance for a 400 kV line operating at UMAX = 420 kV
can be determined as follows:
(1)
𝑈𝑚𝑎𝑥 420
𝑑= = = 22𝑚
𝐸 11√3
28
for 400 kV where:
UMAX - Maximum line voltage (line-ground)

E- minimum value of the dielectric strength of sugar cane flame (11 kVrms/m) (Bologna & Sadurski, 2001)
This is the distance between the phase conductor of a 400kV line and a point of discharge of fault current into the
soil ahead of the fire. This distance depends on the position of the fire in relation to a line and on wind speed.
Wind speed, in this example, must be sufficient to force the flame into contact with the phase conductor – as shown
in Figure 18. These are however, exceptional conditions. During actual firefighting operations, the leader of the
team must assess both the likelihood of flashover and the possible distance between the line and the striking point
due to wind. This should be done, taking the following into account: -
 Fire intensity and height of the flames
 Weather conditions as outlined above
 Position of the highest flames in relation to conductors (the flashover will always take place along the
column of the flame)
 Concentration of particles – dust, soot, burning debris in the column of the flame. (Bologna & Sadurski,
2001)
The magnitude of this distance was determined using Eskom’s findings, namely:
 Dielectric strength of flames – 15kVrms/m
 Withstand strength – 11kVrms/m
 The dielectric strength of burning vegetation is found to be negligible.
 The heights of burning vegetation – 3m.
Adjustments will be necessary for different fuel and weather conditions.
Possible strike distances for very windy conditions are given in Table 2.
Table 2 The possible Strike Distances for very windy conditions. (Bologna & Sadurski, 2001)
Maximum Line Maximum Distance between outer

Voltage conductor and discharge point (m)**
(k𝐕rms ) Actual Horizontal
145* 7.6 6.8

300 15 14.3
420 22 21.4
800 41 40.3
*The probability of a mid-span flashover of lower voltage lines is negligible. **See Figure 21.
29
Conductor
Dangerous
Flashover step potential
Safe
zone (d)
𝜶
r
R Ionized zone
1m dr
Figure 19 The generation of step potentials
The high current density and associated high voltage gradient in the immediate vicinity of the of the arc termination
point cause breakdown of the soil.
3.6 Resistance to ground of the Arcing Point

The resistance-to-ground is derived by Mousa (1990) as follows (Figure 19):
The resistance of a thin hemi-spherical layer is given by
(2)
𝜌𝑑𝑟
𝑑𝑅𝑔 =
2𝜋𝑟 2
Where 𝑅𝑔 = resistance to ground and ρ the resistivity of the soil
 resistivity of the soil
 radius of the Ionized zone.
And the corresponding voltage drop is given by
(3)
−𝐼𝜌
𝑑𝑉 = 𝑑𝑟
2𝜋𝑟 2
Thus, the voltage gradient is given
(4)
𝑑𝑉 −𝐼𝜌
=
𝑑𝑟 2𝜋𝑟 2
The radius of the formed conducting hemisphere is determined by:
(5)
|𝑑𝑉/𝑑𝑟| 𝑎 = 𝐺
From (4) and (5) the radius a is given by
30
(6)
𝐼𝜌
𝑎=√
2𝜋𝐺
The resistance-to-ground at the arcing point is determined by integrating (2) between the limits α and ∞. Thus
(7)
𝜌
𝑅𝑔 =
2𝜋𝛼
Substituting from (6) into (7) the resistance-to-ground is given as a function of the magnitude of the fault current.
(8)
𝜌𝐺 3𝜌
𝑅𝑔 = √2𝜋𝐼 =√2𝜋𝐼
Where
I = magnitude of fault current.
Mousa (1990) cites numerous authors which used different values for the soil breakdown value. He accepts the
value according to Eriksson (1981) as:
(9)
𝐺 = 3.0 × 105 V𝑚−1
3.7 Magnitude of the fault current

The magnitude of the phase-to-ground fault current is determined by the circuit shown below.
From Figure 20 it follows:
3EN (10)
𝐼 =
(Z1 + Z2 + Z0 + 3RG )
Where
Z1 , Z2 , Z0 = the positive, negative and zero sequence impedances of the system respectively for the fault
at the subject site.
EN Phase-to-ground voltage of the power line.
RG Resistance to ground given by (9) when impedance of arc is neglected.
Note that R G is resistive while the other impedances in (10) are highly inductive. For the purpose of this
analysis, those impedances will be assumed to be pure reactances.
31
𝐸𝑛
𝑍1
𝐼
𝑍2 3
𝑍0
3𝑅𝑔
Figure 20 - Calculating the Fault Current (Mousa, 1990)
The fault level I0 available from system studies is based on zero grounding resistance and is given by
Mousa (1990) as
(11)
3𝐸𝑛
𝐼0 =
(𝑍1 + 𝑍2 + 𝑍0 )
This gives:
(12)
3𝐸𝑛
|𝑍1 + 𝑍2 + 𝑍0 | =
𝐼0
From (10) and (12)
(13)
𝐸𝑛
𝐼=
𝐸
[𝑗 ( 𝐼 𝑛 ) + 𝑅𝑔 ]
0
Substitution from(7) in (13) gives
(14)
𝐸𝑛
𝐼=
𝐸 𝜌𝐺
[𝑗 ( 𝐼 𝑛 ) + √2𝜋𝐼 ]
0
Which reduces to:
32
(15)
2
𝐼0 𝜌𝐺
𝐼2 + ( ) 𝐼 − 𝐼02 = 0
𝐸𝑛 2𝜋
The solution of this second-degree equation is given by:
(16)
2𝐼0 2
𝐼 = 0.5𝛽 [−1 + √1 + ( ) ]
𝛽
Where
(17)
2
𝐼0 3𝜌𝐺
𝛽=( )
𝐸𝑝 2𝜋
And EP is the phase-to-phase voltage. (Mousa,1990)
3.8 The generated step voltage

From (3) the voltage rise with respect to remote earth of a point of the soil at a distance 𝑅 from the arcing point
(also see Figure 19) is given by:
𝑅 (18)
𝐼𝜌
𝑉 = ∫ 𝑑𝑉 =
∞ 2𝜋𝑅
In calculating the step potential, the distance between the person’s feet are taken to be 1m. Thus:
(19)
𝑉𝑠 = 𝑉(𝑅) − 𝑉(𝑅 + 1)
Substitution from (18) in (19) gives:
(20)
𝐼𝜌
𝑉𝑠 =
2𝜋𝑅(𝑅 + 1)
(Mousa, 1990)
3.9 Safe distance between a fireman and the point of discharge

When phase-to-ground flashovers occur, the fault current of the line is discharged into the mass of earth. The fault
current generates a potential which is highest at the striking point and then drops as we move away. The
hazardous zone ends where the step potential reaches a level below the tolerable step voltage. A person standing
near the point of discharge is subjected to a step potential (a potential difference between the feet that are spaced
apart), which is generated by the fault current flowing radially away from the termination point of the arc. (See
Figure 19 and Figure 21)
The magnitude of this distance was determined using Eskom’s findings, namely:
 Dielectric strength of flames – 15 kVrms/m
 Withstand strength – 11 kVrms/m
33
 The dielectric strength of burning vegetation is found to be negligible.

 The heights of burning vegetation – 3 m (sugar cane) (Bologna & Sadurski, 2001).W
(Withstand” refers to the 90% of the flashover voltage during the fire.)
This step voltage will cause the current to flow through the legs and lower part of the body. This depends on the
magnitude of fault current and also on resistivity of the soil.
The magnitude of safe value of step voltage is determined from the allowable limits of the body current, and is
given by IEEE Standard 80-1976, (IEEE Guide for safety in AC substation Grounding) as:
(21)
(116 + 0.7𝜌)
𝑉=
√𝑡
Where:
t -duration of fault current in seconds.

ρ -soil resistivity in ohm meters.
V -tolerable step voltage (distance between feet-1 meter)-in volts.

In the case of a 500kV line fault cleared by the primary protection, the duration of the faults is only about 0.1
second. If the fault needs to be cleared by backup protection, the duration of the fault will be longer. At lower
voltages clearing times will be longer due to the less sophisticated protection. For the purpose of safety
calculations, the fault duration is generally assumed to be 0.5 seconds.
The exposure increases inversely with the resistivity of the surface layer of the soil. To be on the safe side, 𝜌𝑠 is
taken as 100 𝛺𝑚.
Using the above assumptions:
𝑉 = 236 𝑉 (22)
(Mousa, 1990)
Shield
Wire
Phase conductor Ustep

Potential rise above earth
During a short circuit
Flashover
path
Figure 21 - Step Voltage Potential (Bologna & Sadurski, 2001)
34
In order to calculate the minimum safe zone, the following assumptions are used: -
 The fault duration is equal to 0.5 s (as explained above)

 The contact resistance between soil and the person’s feet is negligible, assuming the individual is not using
special protective boots.
The safe zones will be determined for: -
 Soil resistivity of - 200 .m (agricultural areas).

 Soil resistivity of - 500 .m (semi desert areas)
 Soil resistivity of - 1,000 .m (rocks)
 Fault levels of - 20 kA, 30 kA, 40 kA and 50 kA.
(Bologna & Sadurski,2002)
 Probability of flashover to the surface of the ground is negligible for 22 kV, 33 kV, 44 kV and 66 kV lines.
 Agriculture areas are characterized by the average soil resistivity of 200 m.
 Average soil resistivity in South Africa – 300 m – according to CSIR*.
 Semi-desert areas (bush-grass) – soil resistivity 500 m.
 Rocky ground – soil resistivity 1,000 m.
* CSIR = Council for Scientific and Industrial Research in South Africa
Table 3 - The minimum safe distance (m) between the arcing point (point of fault current entering the ground) and
firemen for step potential. (Bologna & Sadurski,2002)
Umax
Ground Resistivity Ground Resistivity Ground Resistivity
Line
Voltage 200m 500m 1000m
(kV)
Fault Level (kA) Fault Level (kA) Fault Level (kA)
20 30 40 50 20 30 40 50 20 30 40 50
100 6.0 6.0 6.0 6.0 6.0 6.0 6.0 6.0 6.0 6.0 6.0 6.0
145 8.9 8.9 8.9 8.9 8.9 8.9 8.9 8.9 8.9 8.9 8.9 8.9
300 19.0 19.0 19.0 19.0 19.0 19.0 19.5 19.5 19.0 19.0 19.0 19.0
420 26.2 26.3 26.5 26.6 26.7 26.7 26.8 26.8 26.8 26.8 26.8 26.8
800 39.1 43.5 45.0 47.6 48.0 50.0 50.3 50.7 50.3 50.9 51.1 51.3
Refer to 16.4 for an example of the calculation of the safe distance between a person and the point on the ground
where the fault current enters.
35
3.10 The extent of the hazardous zone

From (20) the distance D from the arcing point beyond which the step potential drops to a safe value is given by:
𝑖𝜌 (23)
𝐸𝑠 =
[2𝜋𝑑(𝑑 = 1)]
This is a second-degree equation in D and the solution is
(24)
2𝐼𝜌
𝑑 = 0.5 [−1 + √1 + ( )]
𝜋𝐸𝑠
Equations (16),(17) and (24) give the complete solution to the problem. By substituting the values given in (9) and
(22) in the above equations, the following is obtained:
2 (25)
𝐼05
𝛽 = 1.4324 ∗ 10 ( ) 𝜌
𝐸𝑝
𝑑 = 0.5 [−1 + √1 + 2.4206 ∗ 10−3 𝐼𝜌] (26)
Note that (25) and (26) are based on:
 Fault duration = 0.5 s.

 Resistivity of the top layer of the soil = 100 Ωm
 The breakdown gradient of the soil is 3.0 kV/cm
3.11 An approximate solution

Mousa (1990) also proposes the following simplified solution.
Equation (16) can be written as
𝐼 = 0.5 𝛽 [−1 + √1 + 𝐾 2 ] (27)
Where
4𝜋 𝐸𝑝2 (28)
𝐾=
3𝐺 𝜌𝐼0
If the short circuit level I0 is high, the ground resistivity Ρ is high, or the product ΡI0 is high, so that K is small
compared to I0 then (27) reduces to
𝐼 ≅ 0.25 𝐾 2 𝛽 (29)
Substituting from (17) and (28) in (29) gives:
2𝜋 𝐸𝑝 2 (30)
𝐼≅
3𝐺 𝜌
36
Substituting from (30) in (24) gives:
(31)
4𝐸𝑝 2
𝑑 = 0.5 [−1 + √1 + ( )]
3𝐺𝐸𝑠
Substituting the numerical values of (9)and (28) in (31) and expressing EP in kV gives:
(32)
𝑑 ≅ 0.5 (−1 + √1 + 0.0169𝐸𝑝 2 )
Where d is in meters and EP is in kV (Note Except for (32), in all other equations, EP is in Volts and I0 in
Amperes)
(Mousa, 1990)
In Table 4 below the hazardous zone has been evaluated for different voltages and for two different values of
ground resistivity namely 100Ωm (low) and 1000Ωm (high) using (16), (25) and (26). In the last column, the value
using the approximate approach is shown for comparison.
Table 4 Effect of soil resistivity and fault current level on extent of the Hazardous Zone (Mousa,1990)
ρ=100Ωm ρ=1000Ωm
Line
Voltage Approximate
kV 10kA 20kA 50kA 10kA 20kA 50kA solution
25 1.2 1.2 1.2 1.2 1.2 1.2 1.2
69 4.0 4.0 4.0 4.0 4.0 4.0 4.0
138 8.4 8.5 8.5 8.5 8.5 8.5 8.5
230 13.6 14.2 14.4 14.4 14.4 14.5 14.5
287 16.1 17.5 18.0 18.1 18.2 18.2 18.2
345 18.0 20.4 21.6 21.9 21.9 21.9 22.0
500 20.9 26.0 30.4 31.5 31.9 32.0 32.0
765 22.6 30.3 40.7 46.0 48.3 49.1 49.0
From the table is becomes clear that:
 Where the soil resistivity is high, the error using the approximate approach is practically insignificant.
 Where the soil resistivity is low, the approximate approach will still give adequate accuracy for voltages
below 345 kV.
 The approximate approach, however, yields significant errors for the case of EHV lines traversing low
resistivity soils. (Mousa, 1990)
37
3.12 Conclusion
The conclusions reached by (Bologna FF, KJ` Sadurski, 2001) are as follows:
It is safe to conduct fire-fighting operations without de-energising on:
 Line voltages of 66kV and below (however the zone within 10m from the tower should be avoided).
 Where the fire is burning under the power line and the flames are very low – less than 50% of the top of the
vegetation – conductor gap is spanned by flame (This applies to lines of 88kV, 132kV and 275kV).
 When there is likelihood of flashover (flames in the vicinity of conductors) the firefighting operation can only
be carried out in the areas of tolerable step potential (minimum safe distance from the point of expected
discharge of fault current into the soil should be maintained – see Table 4).
 Firemen should avoid contacting any metallic object (such as fences, pipes, etc.) which might exist within
the work area on or near the servitude of overhead lines. Metallic fences (especially with the wooden
droppers or posts) can conduct electricity over large areas if the fence runs through a fire under a power
line during a flashover.
 As it often not possible for fire fighters to know the voltage of an overhead line, it is recommended that in all
cases boots be worn that have a high resistivity.
The conclusions further drawn by Mousa (1990) are:
Fires under power lines can cause mid-span flashovers with consequent hazardous step potentials.
For the sake of both public safety and power line reliability, the public education programs of power utilities should
include warning against starting fires in the proximity of power lines. When fires accidently occur, the guidelines for
safety of firemen are as follows:
1. It is safe to conduct firefighting operations on/near the Right-of-way (ROW) without de-energizing the power
line in the following cases:
(a) Where the fire has not reached the ground area beneath the wires.
(b) Where the fire is burning under the wires but the gap between the wires and the top of the flame is not
expected to increase to 2 m below the conductor. (This applies to voltages up to 500 kV.)
2. If a major fire is detected on the ROW, the utility should be requested to de-energize the affected power
line(s). For the period that the line remains energized, the hazardous step potential zone should be identified
based on Table 2 and Table 4, and firefighting should proceed outside that zone.
3. If a fire is spreading toward a wood pole line, it is advisable to drench the lower part of the structures with
water to decrease their susceptibility to ignition. Solid stream nozzles not exceeding 518" or spray/fog nozzles
are permitted to be used for this purpose.
4. The area within 10 m from individual towers should be avoided at all times to protect against the risk of
flashover of insulators under the wet polluted conditions which may be generated by the fire and firefighting
activities.
5. Firemen should avoid contacting any metallic fences or pipeline valves which might exist within their work
area on or near the ROW. The above summary can be used by a power utility to produce a Guide* to be
issued by the Fire Departments.
In doing so, the following is to be noted:
1. Fire fighters are not expected to know the voltage of the line. The Guide should include a table showing the
number of disks per string (case of porcelain and glass insulators) and the approximate length of the insulator
(case of polymer insulators) for the different voltage levels. That table is to be used by firemen in cases where
help is not immediately available from a qualified utility worker. Upon arriving at the site, firemen can quickly
38
count/estimate the insulation level, then consult the table to determine the voltage. If in doubt, firemen should
use the separation corresponding to the next higher voltage level.
2. The Guide should include a single table for the extent of the hazardous zone. This should be produced from
Table 2 and Table 4 as follows:
(a) For EHV levels, only a single distance should be included for each voltage class. This is to be selected
from Table 4 based on the applicable system parameters.
(b) To simplify the table, the voltages not used by the subject utility should be deleted from the listing.
* Such a guide was produced and distributed to BC Forest Services and forms the basis of handling fire incidents in
BC Hydro’s service area. Care should however be exercised at all times. It will be shown that the hot gas above the
flame can also permit a flashover to develop. Therefore, people should stay clear of the area when fires burn close
to the line.
39
40
4. Insulation breakdown during fires under overhead lines

4.1 Introduction
The breakdown process of air during a fire is a subject that is not fully understood. In many cases experiments
have been conducted with fires under power lines and observations made and conclusions drawn. These
conclusions are not always the same and very difficult, if at all to repeat (Sukhnandan, 2004).
The common issue of whether it is the smoke or the flame that causes the flashover is often debated. The
contribution of the chemistry in the flame was not considered in all but a few studies. Furthermore, the atmospheric
conditions that existed during fire experiments have also not been taken into account during most of these studies.
Finally, the properties of the fuel which was used was also not considered.
The simulation of fires during experiments remains difficult. Relative humidity, the dryness of the fuel and wind
conditions are elements that are difficult, if not impossible to control or replicate. As a result, the results of the
experiment will differ from that of the case of real fires.
The classic process of the breakdown of a gas is discussed comprehensively in 15. It will be seen that the ionized
molecules play a fundamental role in the breakdown process when a gas is subject to an electric field. This
approach considers a gas at normal temperature and pressure and under the influence of an electric field. In this
chapter it will be discussed the conditions that exist during a fire and will be shown how these contribute to the
supply of ionized molecules that contribute to the breakdown process.
4.2 The electrical breakdown of a gas

At normal temperature and pressure, a gas may be almost a perfect insulator. Under certain conditions, however,
when subjected to an electric field for instance, the same gas may become an almost perfect conductor. The
transition from insulator to conductor manifests itself as an electrical discharge and is known as the electrical
breakdown of the gas. This transition from insulating to conducting states occurs at a critical potential difference
referred to as the breakdown potential and is the lowest potential difference at which breakdown can occur
providing initiating electrons are present.
The mechanism that results in the breakdown of electrical insulation in a gas has been studied for many years.
Before 1900 and up until 1902 no theories explaining the phenomenon were published. It was with the discovery of
x-ray in 1895 and the electron in 1896 that J.S. Townsend developed a theory for the conduction of electricity
through a gas involving two collision processes. (Martinussen & Bozzoli, 1964). Later Raether, Loeb and Meek
advanced the streamer mechanism for spark formation. (Loeb & Meek, 1940).
At normal temperature and pressure, the gasses are excellent insulators allowing only a current density of the
order of 10−10 A/cm² (Wadhwa, 2001). This small current conduction results from some naturally occurring ionizing
phenomena, such as gamma rays, emanating from radioactive processes in the soil or from cosmic radiation.
Atmospheric air at sea level contains approximately 1000 positive ions and almost a similar number of negative
ions per cubic centimetre due to these natural ionisation processes (Maruvada, 2005).
At raised electric fields, charged particles may gain sufficient energy between collisions to cause ionisation on
impact with neutral molecules. During these elastic collisions, an electron loses little energy and rapidly builds up
its kinetic energy which is supplied by an external electric field. During these elastic collisions, a large part of the
kinetic energy is transformed into potential energy by ionizing the molecule struck by the electron. The ionization by
electron impact under strong electric field is one of the most important processes leading to the breakdown of
gasses. (Wadhwa, 2001). Further processes leading to ionization and breakdown will be discussed below.
Clean, dry air contains about 78% Nitrogen as 𝑁₂, 21% Oxygen as O₂ and 1% of noble gases such as Helium
(He), Neon (Ne), Argon (Ar), Krypton (Kr), Xenon (Xe) and Radon (Rn) with argon amounting to about 0.9%.
Normal atmospheric air also contains carbon dioxide ( CO₂) and water vapour (H₂O) in varying amounts and small
quantities of other extraneous gases as nitrogen and Sulphur compounds.
41
In his study of the breakdown mechanism, Cobine (1941) states that an idealized gas is considered as consisting
of small elastic spheres in continual random motion striking each other. For a given gas in its normal state, these
particles, or molecules all have the same size, weight elasticity, etc. and it is assumed that their dimensions
compared to the distance they travel are so small that their volume may be neglected. Gravitational forces between
individual particles may also be neglected. In their motion and collisions, they obey Newton’s laws of motion and
general gas laws such as Avogadro’s hypothesis, the Gay-Lussac law, Boltzmann’s relation, the Maxwell-
Boltzmann distribution function, and others.
The mathematical treatment of the classic breakdown mechanism has been comprehensively shown in several
publications. (Vosloo, 1992), (Wadhwa, 2001). Sukhnandan, (2004) approaches the subject in a concise and lucid
way and his work is presented in detail in 6, annexure A.
4.3 Overview and conclusions of fire experiments

The research using fire experiments stretch over several decades. In order to see how this theme was developed
over time, a short overview will be given of the most relevant work in this regard.
One of the earlier fire experiments conducted was that of Sadurski (Sadurski & Robertson, 1977). In a paper by
Sadurski & Reynders (1989) they show that with small air gaps (𝑑 ≤ 1𝑚) the effect of thermal ionisation is
negligible for the temperature range they considered (20°𝐶 < 𝑇 < 500℃) and that the flashover voltage may be
considered from a practical point of view, as a function of the gap size and air density alone. They also showed that
by introducing various materials into the flame such as aluminium, sawdust and fine ash, the breakdown strength of
the gas reduced to as low as 7 kV r.m.s. No explanation of any chemical processes involved was given in that
work. A discussion to elaborate on this topic will be given in 4.6.
West & McMullen (1997) carried out tests during the late 1970’s and did two different fire experiments. The first
consisted of a horizontal sphere gap where heated air from a furnace was forced through. This experiment clearly
showed that at temperatures normally associated with fires (350 °C and higher) the dielectric strength had been
reduced to less than half of its standard value. When the temperature approaches 700 °C, this standard value was
reduced to one third. (Figure 22)
A further experiment that was carried out was to examine the combined effect of particulates and gasses using a
set of 40.6 cm aluminium spheres separated at 12.7 cm. Several different materials were burned including saw
dust and motor oil, rubber and some plastics. They found that the relationship between particulate intensity on
flashover voltage to be similar to those of Martinussen & Bozzoli (1964), which stated that although dust particles
reduce the impulse breakdown voltages significantly, their influence on AC gaps is minimal. Of this minimal effect,
the greatest influence is due to carbon deposits on the electrodes.
Finally, line tests were conducted using a 76 m two phase test setup with the lines energised at 495 kV phase to
phase. Five identical structures (3 m x 3 m x 4.9 m) were built to be used for fuel, which were placed successively
at the mid-span, halfway between the phases. The bases of these structures were earthed.
42
After the first two fires failed to produce flashovers in spite of flames reaching the conductors, nine automobile tyres
were added. As the fire intensity increased, an equal number of flashovers between the phases and down to
ground ensued.
140%
120%
Ratio of Vs to Vt (%)
100%
80%
60%
40%
20%
0%
Temperature of the air gap (°C)
Figure 22 - The relationship between the rise in temperature and reduction of dielectric strength in an air gap (Vosloo
W.L, 2002. West & McMullan, 1979)
Three fire components were observed: the smoke column, the flame column, and the heat column. When these
three came together, the flashover of the line was imminent.
The study concluded by stating that flashovers become a certainty when the following criteria are met: Sufficient
heat, large pressure drop due to plume configuration, high gas and particulate emissions and large quantities of fire
brands and debris carried in the plume activity.
The first fire line test experiment cited by Deno & Zafanella (1982) is taken from West & McMullen (1997) and was
discussed above.
Two laboratory tests were carried out at a Transmission research facility named Project UHV.
In the first, a pail of flaming gasoline rags produced a small flame and a large amount of smoke. This was done to
determine if the flashover was caused by smoke or the flame. After several tests, only one flashover resulted and
this occurred when the flame became unusually high for an instant. In many cases the smoke rose straight up to
the conductor without causing any flashovers and it was concluded that large flames would bring about a flashover
where smoke by itself would not.
In the second test a propane flame apparatus was used to generate a large, controlled flame, without much smoke.
Numerous flashovers occurred during this test.
 Transmission lines may flash as a result of fires near or under the lines
 These flashovers are a function of the voltage of the line and the distance between the line and flame.
 Smoke composed of particles of burnt fuel did not seem to cause flashovers even if it came into contact
with the conductors.
 A flame has a high ion and electron concentration making it conductive, similar to corona and streamer-
type partial breakdown. When a flame is large enough to reach the conductor, a flashover will immediately
occur
 Although smoke of burned particles does not have the same effect as the flame near the conductor, the
smoke column must be considered partially conductive and will contribute to the flashover.
Fonseca et al. (1987), (1990) conducted fire experiments in an attempt to determine the dimensions of overhead
lines to prevent flashovers where those are subject to fires. In this work, they also stated that the highest reduction
43
in breakdown strength is caused by the presence of floating particles. In the work, discussions were recorded
where A.M. Mousa suggested that” the existence of floating particles is a symptom of the intensity of sugar cane
fires and that the associated larger reduction in strength of air gaps is caused by the higher temperature of the fire
rather than by the floating particles themselves.”
Galluci and Pigini also noted a big discrepancy between the withstand voltages with floating particles between this
work and that of Sadurski. This was attributed to the fact that the Sadurski work was done in a laboratory versus
actual field results of this work.
Islas (1988) reported that due to numerous faults on the 400kV interconnector between the Malpaso hydroelectric
plant and Mexico City, investigations were launched to find solutions to this problem. The faults were caused by the
burning of sugar cane. Research conducted in 1978 and 1979 resulted in the determination of the following in
dielectric strength.
Withstand (kV/cm)
3
2,5
2,5
Withstand (kV/cm)
2 1,9
1,7
1,5
1
1 0,8
0,5 0,3
0
Without fire Without fire Without fire With fire With fire (cane With fire
(15°C) (100°C) (120°C) (gasoline) alcohol) (sugar cane
leaves)
Figure 23 - Withstand values determined for six different cases (redrawn from Islas (1988)
The value of 0.3 kV/cm was the point of departure that was used in the re-design of existing lines as well as for
new transmission lines in Mexico. The success of this intervention is evident from the reduction in faults shown in
Figure 24
Several actions were implemented on existing lines prior to 1983. These included the raising of towers as well as
re-tensioning conductors. In order to reduce phase to phase faults, the phase separation was also increased. The
success of this strategy is clear from the graph below.
44
Number of outages '71-'86

140
120
100
80
60
40
20
0
1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986
Figure 24 - The line performance pre- and post-1983 is proof of the success achieved in Mexico
(Redrawn from Islas 1988)
Robledo-Martinez (1991) carried out experimental burns using a 70 kV three phase setup where the conductor
height could be changed from 0.85 m to 2 m in 10 cm steps above the combustion tray. Five different fuels were
used, namely:
 Assorted gardening refuse

 Sugar cane bagasse
 Sugarcane leaves
 Wood and wood derivatives
 Butane gas.
All experiments were conducted in open air under dry conditions with typical values of T=20℃, RH=50-70% and
p=79kPa. The 60 Hz ac voltage was applied to the line with no load.
Robledo-Martinez found evidence that the chemical nature of the fuel played an important role. He also found that
the temperature/ionization seemed to be a dominant factor and that the presence of solid particles can play an
important role only at small conductor spacings.
Baldo et al (1992) in the Cigré Technical Brochure 72 prepared by Study Committee 33 WG07 states: “As is well
known, the basic process in gas discharge development is the formation of free charges of opposite sign mainly
due to ionization of molecules by collisions with free electrons accelerated by the electric field. Due to both the
different mobility of electrons and of positive ions and to the high value of air density at atmospheric pressure,
charge concentrations can be created, which change the electric field distribution and allow the propagation of
ionization phenomena also in parts of the gap where the field due to the applied voltage alone would be too low.
This causes the propagation of ionization processes along channels which are partially ionized and can cross the
whole gap causing breakdown.”
Lanoie & Mercure (1997) carried out some fire experiments, using a 70 m single phase line with a 13.7 m mid-span
clearance. The conductor was a hollow aluminium tube, 4.4 cm in diameter with a minimum of 100 spruce trees
45
below. These trees were placed vertically, leaving a gap of 10.7 m along the entire line. The AC power was
supplied by a 1 MVA voltage transformer protected by a 15 kΩ resistor with the maximum operating voltage of
425 k𝑉𝑟𝑚𝑠 /m during the tests. The line was connected to a voltage divider and a current transformer as well as
several recorders and a transient digitizer. A video camera and other recorders were used to record the main
electrical and physical test parameters.
Simultaneous analysis of voltage, current and video recordings allowed the flame resistivity (𝑅𝑓 ) to be estimated,
namely:
5𝑘Ω/𝑚 < 𝑅𝑓 < 35𝑘Ω/𝑚 (33)
Lanoie & Mercure (1997) found that under certain conditions, the most important was the value of the electric field
E in the flames, during the time when the current flows in the flames and concentrated into an arc within a few
cycles. They found that two areas were defined: one with high impedance conduction and one with low impedance,
corresponding to the occurrence of an arc in the flames. Based on these results they proposed a minimum critical
field 𝐸𝑐 which establishes a threshold for the transition to the arc regime, and which is given by:
𝑘𝑉𝑟𝑚𝑠 (34)
𝐸𝑐 ( ) ≈ 4.16√𝑅𝑓
𝑚
Where 𝐸𝑐 is the average electric field in the flame when the current starts to flow in the flames.
𝑅𝑓 is the average resistivity (kΩ/m) of the flames before transition to the arc regime.
The lowest equivalent flame resistivity value that was obtained was about 4 kΩ/m, yielding a critical field value
9 k𝑉𝑟𝑚𝑠 /m during the burning of very dry spruce needles (water content of about 30 %). Where the needle water
content reached values of 120 % (sic) the equivalent flame resistivity increased to 12 kΩ and 𝐸𝑐 to 15 k𝑉𝑟𝑚𝑠 /m
(Lanoie & Mercure (1997))
Lanoie & Mercure (1997) reported that the water content on live trees are at a minimum of 80 % during the dry
summer months (July) and if a simple linear relationship between water content and flame resistivity is assumed,
the critical 𝐸𝑐 can be estimated at:
𝐸𝑐 ≈ 12𝑘𝑉𝑟𝑚𝑠 /𝑚 (35)
This value may be found by direct interpolation of figure 4 in Lanoie & Mercure (1997) (see Figure 25 below) and is
considered as representative for Hydro- Québec’s power lines running through evergreen forests in central Québec
during dry summer months. They state that this value may be used for planning and maintenance purposes.
46
Figure 25 - Critical electric field concept establishing high and low impedance flame conduction regimes
(from Lanoie & Mercure, 1997)
Lanoie & Mercure (1997) offers some practical information that may be used in cases similar to those of the spruce
tree forests of central Québec. The basic arcing criterion for spruce tree fires is estimated as
𝐸𝑐 = 12𝑘𝑉𝑟𝑚𝑠 /𝑚 (36)
From which the critical breakdown voltage 𝑉𝑐 (𝑘𝑉𝑟𝑚𝑠) can be calculated:
𝑉𝑐 = 𝐸𝑐 ∗ 𝑙𝑓 (37)
Where 𝐸𝑐 is the minimum critical electric field (𝑘𝑉𝑟𝑚𝑠 /𝑚)
𝑙𝑓 is the flame length (m)
Equation (37) gives the breakdown voltage for a conductor-flame configuration for various gap lengths (d)
assuming that the flames fill the entire space between the line and the ground (Figure 26 curve a). Curve b is
obtained in the absence of any flames in the gap. By modifying equation (37) slightly, a minimum clearance 𝐷 is
obtained, corresponding to extreme conditions.
𝑉 𝑉 (38)
𝐷=
=
𝐸𝑐 12
Where 𝐷 is the clearance (m) between a phase conductor and ground, 𝑉 is the AC line voltage (k𝑉𝑟𝑚𝑠 ) and 𝐸𝑐 is the
critical electric field (12 k𝑉𝑟𝑚𝑠 /m).
47
Figure 26 Breakdown voltage for a conductor-ground gap (d) covered by flames

(From Lanoie & Mercure, 1997)
Lanoie & Mercure (1997) present two tables as a guide or maintaining safe line to tree top clearances, applicable to
the conditions that pertain to the lines of central Québec. The reader is cautioned to consider local conditions
pertaining to vegetation and weather when assessing these values
In Table 5 the minimum clearance distance is given for the case where the fire burns directly under the conductor
for five nominal phase-phase voltages.
Table 5 - Minimum clearance distance suggested in the case where the flames are immediately under the line
conductors (from Lanoie & Mercure, 1997)
Nominal phase-to-phase Minimum clearance distance

Configuration
voltage levels (𝑉𝑟𝑚𝑠 ) h suggested. (m)
735 35
315 15
230 11
161 8
120 6
In Table 6 the minimum cleared ROW distances are given as lateral distances from the conductor as well as a
practical minimum height clearing for each of the nominal phase to phase voltages. This deals with the case where
the flames are blown towards the conductor by a cross wind. It is to be noted that these distances only pertain to
phase to ground flashovers. For lower voltages, somewhat longer distances are suggested in order to account for
flash tree-top fires that may occur where tall trees are present next to the ROW.
48
Table 6 Minimum cleared right-of-way distance, given as lateral distances from each line conductor (from Lanoie &
Mercure, 1997)
Nominal phase-to- Minimum line clearance

Lateral distances 𝒅𝒍
phase voltage levels Configuration height 𝒉𝒄 .
suggested (m)
(𝑽𝒓𝒎𝒔 ) (m)
735 13.6 25
315 6.5 11
230 6.1 8
161 5.8 8*
120 5.5 6*
*taking in account the height of tree tops
Lanoie & Mercure (1997) states that Figure 25 can be used to illustrate the concept of critical electric field which is
required to switch from a high impedance fault (HIF) regime to a low impedance fault (LIF). It is shown that the 𝐸𝑐
level increases with increasing equivalent flame resistivity (𝑅𝑓 ), or that 𝐸𝑐 is a function of 𝑅𝑓 . This critical field 𝐸𝑐 in
fact corresponds to the minimum heat input that must be injected into the flames to generate thermal instability
conditions. Such unstable conditions can indeed be achieved once the heat input (by Joule effect) into the
convection column (flames) exceeds heat losses through turbulence.
A minimum critical field value could thus be determined for each test by recording the lowest equivalent flame
resistance before arcing occurred. It must be noted that different types of burning vegetation or different fire
conditions imply different flame resistivity values. Therefore, a new minimum critical field (𝐸𝑐 ) -levels will have to be
established for each type of fire.
Lanoie & Mercure (1997) explains that the concept of a critical field 𝐸𝑐 applies to an average field through the
flames because it was observed that the fault current preferably flowed through the flames rather than along or
through the tree limbs and trunks. This was exemplified with the recording of fires inducing arcs that were parallel
and away from the tree trunks. When analysing test results, they accounted for the presence of the trees, - whether
they were burnt or not, when calculating the equivalent flame resistivity. For tree-top fires and flashovers arcs
attaching to the tip of the trees, they estimated that the tree itself adds about 40 kΩ to the flame equivalent
resistance. Generally speaking, since it is reasonable to assume that the flames might extend from the ground up
to the line conductor, it is wise to consider the critical electric field value 𝐸𝑐 for line maintenance purposes.
Although large volumes of smoke were generated at the beginning of each test, it had no significant effect on the
breakdown voltage and change in air insulation was only observed when the tips of the flames began to approach
the conductor.
The flame temperature was approximately 1,000 °C, which yields a local relative air density of≈ 0.2. According to
IEEE standards (Anon,1978) one would expect a 50 % reduction in the efficiency of air insulation based on
temperature alone. However, their test results yielded a reduction figure of ≈ 90% as is shown in Figure 26. A
simple reduced-air-density effect therefore cannot explain the observed dramatic modification of the air insulation
properties, both for these experimental tests as well as for actual power system conditions. They conclude that
“one very significant factor other than their thermal properties is probably the chemical characteristics of the flames.
The presence of minerals is also likely to play a major role. We observed that when the flames were produced by
burning spruce trees without their needles, which are known to contain far more dissolved mineral salts than the
bark, it was more difficult to initiate fire induced flashovers.” Lanoie & Mercure (1997).
The role which chemistry plays in the conductivity of flames has largely been omitted during previous fire research,
with the focus on the ionization mechanisms mostly that are caused by electric field, heat and other effects. In
recent work published by some authors, (Sukhnandan & Hoch. 2002), (Sukhnandan, 2004). (Mphale & Heron, 2007),
49
(Mphale & Heron, 2008) showed the important role that the chemical ionization in the flame plays during electrical
breakdown.
These works will be discussed in more detail below
From the work discussed above, it is shown that three major areas were identified, which are major contributors to
the breakdown process during fires. These are:
 Reduced Air Density theory

 Particle initiated flashover theory
 Flame conductivity theory.
In the next sections an overview of each mechanism will be given.
4.4 The reduced air density theory

West & McMullan cites pressure drops in fire whirlwind could be assumed to be the most significant, albeit very
difficult to measure. He cites Mr. Countryman, who has measured reduction in thermal columns of greater than
25mm Hg (3.331kPa) and others have recorded pressures in tornadoes as low as one tenth of an atmosphere. He
concludes that pressures of one half of an atmosphere can easily exist within a fire whirl. (West & McMullan. (1979)
The insulation strength of air decreases as the air density is reduced and the temperature increases. Under
standard temperature and air pressure conditions, meteorological factors affect the breakdown voltage by less than
10%, but under fire conditions assume a much greater importance. The flashover voltage depends on temperature
and humidity:
(39)
𝐻
𝑉𝑠 = 𝑉𝑡
𝐷
where H is a humidity correction factor, D is the relative air density VS is the flashover voltage under standard
conditions and VT is the flashover voltage under actual conditions (e.g. in a fire). D affects the flashover value the
most and it varies according to:
𝑝𝑇0 (40)
𝐷=
𝑝0 𝑇
Where 𝑝 is the barometric pressure in kPa, 𝑝0 is equal to 101.3kPa, 𝑇is the temperature in °K and 𝑇0 is equal to
293°K (Sukhnandan & Hoch, 2002).
During fire conditions, the humidity correction factor is negligible thus,
2.892 𝑝 (41)
𝑉𝑡 = 𝑉𝑠
𝑇
This shows that the flashover voltage varies inversely with temperature and directly with barometric pressure. The
inverse relationship of temperature and breakdown voltage is demonstrated in
Figure 22. (Deno & Zafanello, 1987)
In the late 80’s, a high temperature furnace was used by Sadurski and Reynders (Sadurski & Reynders, 1989) to
detect the effects of temperature on air insulation. A temperature range of 20°C to 500°C was investigated. The
results showed that the effect of thermal ionisation over this range of temperatures is negligible and that the
flashover voltage may be considered as a function of the gap size and air density only.
They stated further that at 1,100 °C the breakdown voltage tends (due to extensive thermal ionization) to be only
about 15 % lower than compared to what was obtained from the density effect alone. They concluded that since
the clearance distances involved are too large to permit flashovers of the high voltage transmission lines (even at
50
1,100 °C), factors other than flame temperature must therefore govern the mechanism of such breakdowns. This
led to the investigation of particle-initiated flashover.
4.5 Particle initiated flashover theory

The subject of air gap breakdown at high temperatures with floating particles was studied by scientists such as
Akazaki, Hara, Ryan and Powell during the early 1970’s. (as cited by Sadurski & Reynders, 1989). Work done by
Martinussen and Bozzoli (1964), (Naidoo & Swift, 1993). Robledo-Martinez, 1991) also investigated the effect of
particles on the breakdown strength of air which will be discussed below.
All natural or artificially generated aerosols are to some extent charged electrically. Agents such as ionizing
radiation, high voltage corona discharge, static electrification, high temperature or chemical reactions, produce
gaseous ions that affect the electrical charge of the aerosol particles they interact with. (Vosloo, 1992; Sukhnandan
& Hoch 2002).
An aerosol can acquire an electrical charge when gaseous ions produced by ionizing radiation, high voltage corona
discharge or high temperature are captured by the particle.
Aerosol particles are defined according to
 Particle size
 Number of unit charges and sign (+ - or bi-polar)
 Particle mobility
There is a maximum electrical charge which can be carried by a particle which is limited by the physical properties
of that particle. At reduced pressure particles can be charged to substantially higher levels than possible at
atmospheric pressure. (Low pressures exist in fires hence the charge of particles will be increased).
The role in the electrical breakdown of a gas by particles in the flames of vegetation has been suggested by some
authors although not all agree about this mechanism in large gaps which exist on Transmission lines (Sukhnandan
& Hoch, 2002), (Mousa in Fonseca, et al. 1987).
During 1977, Sadurski recorded breakdown voltages for air gaps of up to 1.2 m spanned by a Bunsen flame. He
found that, the introduction of fine grass ash, fine aluminium particles and saw dust at room temperature, reduced
the breakdown voltage by approximately 20 % - 30 %. A clean butane flame without any particles reduced the
breakdown strength by 75 %. Introducing the particles into the flame, however, reduced the breakdown voltage in
the order of 15 times. It was deduced that the mechanism leading to the lowering of the breakdown voltage was
the same for both cases. This test found no appreciable differences between the various particles which were used
in the test. (Sadurski, 1977)
In further tests carried out for gaps up to 1m the breakdown voltage was found to be between 6 kV/m and 7 kV/m
for both a rod to rod as well as a rod to plane configurations with particles introduced to the Bunsen flame. Sadurski
K.J., J.P. Reynders, 1989 Sadurski (1989) found that introducing floating particles present in a butane flame
reduced the breakdown voltage to below 10 kV/m
Sadurski presents the following theory on the local field enhancements by air gap particles (Sadurski & Reynders,
1989)
The field induced on the surface of a conductive body placed in a uniform electrostatic field (𝐸0 ) is given by:
𝐸 = 2𝐸0 𝑐𝑜𝑠 𝑥 − 𝑐𝑦𝑙𝑖𝑛𝑑𝑒𝑟 (42)
𝐸 = 3𝐸0 𝑐𝑜𝑠 𝑥 − 𝑠𝑝ℎ𝑒𝑟𝑒 (43)
𝑎2 𝑒0 (44)
𝐸= 𝑎 − 𝑒𝑙𝑙𝑖𝑝𝑠𝑜𝑖𝑑
4𝑟 2 (𝑙𝑛 − 1)
𝑟
51
The following direct conclusions can be obtained from the above equations:
1) The particle raises the field strength near itself and lowers the corona threshold voltage.
2) Corona discharge from the particle brings a space charge into the gap, even when the external field is very
low.
3) The space charge produced by the corona will change the local field distribution in the vicinity of the
particle.
4) Discharges are attracted to the particle due to the concentration of short-range fields created by the particle
and due to the extension of the leaders from the particle.
5) The uniform field between the electrodes will change to a non-uniform one when the spark will bridge the
gap between the particle and the electrode. (Sadurski & Reynders, 1989)
In order to investigate the effects large particles related to sugar cane fires have on the AC breakdown strength of
air insulation, Naidoo and Swift used an experimental method with parallel tubular conductors 500 mm apart
(Naidoo & Swift, 1993).
The following three parameters were varied and the effects were recorded (Sukhnandan, 2004):
(a) Type of material: No set pattern can be seen when the type of material was changed. Thus, no solid conclusion
can be drawn.
(b) Length of material: There is an exponential decrease in breakdown strength as particle length increases.
(c) Spacing between particles: There is a greater reduction in breakdown strength if one end of the particle is in
contact with the electrode surface. When the total particle length is fixed at 300 mm and the number of particles is
varied the following is observed – Smaller individual spacing does not necessarily mean a lower breakdown
voltage. The effect of particles in contact with the electrode is greater but not so noticeable when particles are
many. (Noted case =10)
In the above experiments, it was noted that the breakdown path was through these particles, whose configuration
substantially reduced the breakdown voltage, and along the surface of those particles, whose configuration did not
substantially reduce the breakdown voltage. It was found that in an electric stress of 10kV/m the resistance of the
sugar cane particles was about 10MΩ/m for those particles that substantially did reduce the breakdown voltage
(Sukhnandan, 2004).
From the obtained results in (Sadurski & Reynders, 1989), it was apparent that the flashover voltage of the gap
spanned by the flame is independent of the shape of electrodes but is influenced by a parameter other than the
temperature.
To investigate this parameter, the leakage current was monitored during the test and it was noticed that the high
frequency peaks of discharge currents are caused by the discharge to and from the carbon particles released from
a soot covered (from the fire) electrode. The effect was less pronounced after cleaning the electrode. It is said that
the discharge between particles and the electrodes changes the field distribution in the gap from uniform to non-
uniform. For long air gaps, a series of AC tests were carried out by Sadurski, to determine the influence of the
flame and the presence of floating particles. The following statements describe their test setup and procedure:
(a) Rod-plane and conductor-plane electrode configurations were used.
(b) A particle injector was used to inject a fixed quantity of particles (between 2 mg and 5 mg) into the gap.
(c) First a clean butane flame spanned the air gap and tests were conducted.
(d) With the air gap at room temperature the following floating particles were injected:
(i). Aluminium -size 0.005 mm to 0.2 mm

(ii). Sawdust –size 0.01 mm to 0.2 mm
(iii). Fine ash obtained from burned grass.
(e) Lastly the gap was spanned by flame with floating particles obtained from burned grass.
52
The following are the conclusions that can be drawn:
(a) There was 75 % reduction in the breakdown voltage of the air spanned by a clean butane flame.
(b) The presence of injected floating particles in the air gap at room temperature reduces the flashover voltage by
20 % to 30 %.
(c) The following was observed for a particle filled flame-spanned gap:
(i). Non-self-sustained breakdown
(ii). Flashover was triggered by floating particles
(iii). Corona was extinguished when particles move to non-crucial areas.
(iv). No conditions leading to a stable arc, even when large amounts of particles are introduced into the gap.
(v). The resistance of the gap is high and depends on the quantity of particles in the gap.
It must however, be kept in mind that this applies to small gaps of up to 1m (Sukhnandan, 2004).
Sukhnandan & Hoch (2002) states that although the presence of particles floating in the air gap does reduce the
breakdown voltage due to their influence on the electric field, the probability of a large number of particles forming
a chain-like structure and shorting out a significant part of a large (10m) gap appears to be remote.
“In conclusion, for the existing conditions (particle size and flow, especially) the temperature/ionization seems to be
the predominant factor in breakdown; solid particles can play an important role but only at small conductor
spacing.” (Robledo-Martinez et al, 1991).
The effect of chemical reactions of combustion of the particles during their injection into the flame may need to be
investigated further. In his M.Sc. Thesis titled “Aluminium dust concentration effect on combustion in hydrocarbon
Bunsen flames” Michael Soo (2012) shows that when a premixed methane–air Bunsen-type flame is seeded with
micron-sized atomized aluminium powder it was found that at a concentration of 180 g/m3 the aluminium
combustion regime changes from slow, low-temperature oxidation to full-fledged aluminium flame front
propagation. The transition is manifested by a sharp increase in temperature up to 2,600 K. The actual combustion
of floating particles in the flame and its effect has not been considered in some of the above fire experiments and
may need further investigation.
4.6 Chemistry in flames and combustion

The subject of the chemical processes in flames is a wide and complex subject, especially to those readers who
are not well versed in the biochemistry and the physics of this subject. In this section, we will give an overview and
examples of some of these processes in order to demonstrate the contribution of the chemical processes and its
role in the electrical breakdown process.
“Combustion commences in chemistry, with the occurrence of self-supporting exothermic reactions” (Bradley,
(1965) in Sukhnandan, (2004). The conduction of thermal energy, the diffusion of chemical species, and the bulk
flow of gasses all follow from the release of chemical energy in the exothermic reaction (Sukhnandan, 2004).
Wildfires are also described as exothermic oxidation reaction in nature with temperatures that can range up to
2000°C in its luminous reaction zone. A flame is caused by this self-propagating exothermic reaction and may also
be described as a weakly ionized medium (Mphale & Heron, 2007).
The effect of an electric field on flames

During the 18th century many experiments were conducted in Europe that showed that flames were conductors of
electricity and could be deflected by electric fields. This is shown on the Internet
(https://www.youtube.com/watch?v=a7_8Gc_Llr80)
When an electric field is applied to a hydrocarbon-air flame, the flame is attracted towards the cathode. This is
independent of the nature of flame being premixed or diffusion. The following are the effects that electric fields can
have on flames:
(a) Increases the flammability limits of premixed flames

(b) Shortens the flame lengths
53
(c) Increases stability

(d) Flame temperature variation
(e) Impacts on heat release rates
(f) Changes noise and soot forming characteristics.
(g) Electrons and ions are produced in small concentrations as by-product of complex chemical kinetics of
combustion.
(h) Charged particles are also produced in flames seeded with alkali metals and salts. (Sukhnandan, 2004)
“The electric field affects the motion of the charged particles, which then interact strongly enough with the
surrounding neutral gas molecules to affect the concentration and flow of the neutral species” Lewis & von Elbe,
1987)
Thermal ionization
This high temperature environment in the reaction zone thermally excites incumbent flame particles and the
energised particles become electronically unstable to the extent that they lose their outer shell electrons during
collisions with other flame particles. This process occurs on a selective basis which is determined by temperature
and ionization potential and is referred to as the thermal ionization process (Mphale & Heron, 2008).
Potassium and graphitic carbon ( 𝐶𝑛 ) are vegetation fire particles which are likely to produce appreciable ionization
due to their low ionization energy and work functions of 4.34 eV and 4.35 eV (Sorokin et al (2002) in Mphale &
Heron 2008). Thermal ionization of the exited flame species (𝐹𝐿∗ (g)) occurs by the following reaction equation:
𝐹𝐿∗ (𝑔) ↔ 𝐹𝐿+ (g) +𝑒 − (45)

(Mphale & Heron, 2008)
Chemi-ionization
Another process by which ionization may occur in the flame is chemi-ionization. In chemi-ionization, dissociation
reactions provide part of the energy required for ionization since they are exothermic. The rest of the energy comes
from the flame.
A number of complicated disturbances take place in a flame which is subjected to strong electric field. Firstly, the
concentration of ions rises rapidly due to ionization of neutral molecules by collisions with the accelerated ions and
electrons in the flame. This gives rise to atoms and radicals and since the electrons are much more mobile than the
gas, they are drawn to the positive electrode, leaving the body of the gas positively charged. This rapid migration of
the charged particles causes a Chattock (or combustion) wave or an electric wind and affects the flame in both
macroscopic and microscopic ways (Lewis & von Elbe, 1987).
Absolute values of ions and electrons measured in propane-air flames are in the order of 108 to 109 ions/cm³
(Lewis & von Elbe, 1987) and concentration of electrons in vegetation can be as high as 1018 /m³. These ions and
electrons, which are in motion, both contribute to the electrical conductivity of the fire (Mphale & Heron, 2008). Also
see 4.8 below.
There are two main types of flames, viz. diffusion flames and premixed flames (Sukhnandan, 2004). Bush, grass
and cane fires do not specifically fall under either premixed or diffusion flames. They can be categorized as having
some characteristics specific to both divisions (Mphale & Heron, 2007).
In essence, the wildfire is viewed by Mphale & Heron (2007) as an impure hydrocarbon diffusion flame seeded with
alkalis, alkaline earth metals and graphitic carbon. Since potassium and graphic carbon exists in significant
amounts in the flame and have low ionization potential and work function of 4.34 eV and 4.35eV respectively
Mphale & Heron, (2007) assumes that thermal ionization in the fires is predominantly due to these species.
54
A mechanism of ion formation suggested by the eminent Professor O. Stern is based on the consideration that the
thermionic work function of a comparatively small aggregate of carbon atoms may be expected to sensibly be that
of a large carbon particle which has a thermionic of only 3.93 eV. Thus, a source of ionization may arise from
graphitic carbon nuclei that are formed from dehydrogenated hydrocarbon molecules (Lewis & von Elbe, 1987)
(Pedersen & Brown, 1993).
The organic structure of plants is mainly composed of three interconnected biopolymers; cellulose, hemicellulose
and lignin and these compounds account for over 90 % of the plant’s dry weight. The remainder consist of a
mixture of low molecular weight organic compounds, water and inorganic elements such as alkali and alkaline
earth metal species (e.g. 𝐾, 𝑁𝑎 𝑎𝑛𝑑 𝐶𝑎 and graphitic carbon 𝐶𝑛 ) (Mphale & Heron, 2007). Atoms of alkali metals
have fairly low ionization potentials of the order of 4-5 eV and thus generate ions in energetic collisions (Lewis &
von Elbe, 1987).
Potassium is an omnipresent nutrient in soils and is incorporated into the tissues of plants during growth. Plant
tissue contains typically 0.4 % K – 3.4 % K by dry weight (Bowen 1979). During combustion with high enough
temperatures, thermal dissociation of potassium species take place because of their low dissociation energies (e.g.
𝐾2 𝐶𝑂3 𝑤𝑖𝑡ℎ 3.7𝑒𝑉) (Mphale & Heron, 2007). Because of its low excitation energy (4.34 eV) (NIST 2001), a
significant portion (10 % – 20 % or greater) of the K in burning biomass may be ionized, (Vodacek et al, 2002)
All simple molecules or atoms present in the reaction zone, including the various free radicals such as 𝐶2 and CH
have ionization potentials in excess of 10 eV and as a result cannot ionize in the flame to a detectable degree
because the most energetic elementary reactions imaginable do not liberate energy much in excess of 4 eV. No
ions are therefore detectable in flames of 𝐻2, 𝐶𝑂, 𝐶𝑆2 and 𝐻2 S. if the gases are free of contamination by dust
particles containing sodium and potassium. However, ions are found in hydrocarbon flames and other flames that
are contaminated with hydrocarbons and the source of these ions is the chemi-ionization reaction
𝐶𝐻 + 𝑂 → 𝐶𝐻𝑂+ + 𝑒 (46)
Followed immediately by the charge exchange reaction
𝐶𝐻𝑂+ 𝐻2 𝑂 ↔ 𝐻3 𝑂 + + 𝐶𝑂 (47)
The 𝐻3 𝑂+ is the dominant ion in both fuel lean and slightly rich hydrocarbon flames. In very rich and near-sooting
flames the dominant ion is 𝐶3 𝐻3+ formed from electronically excited 𝐶𝐻 ∗ and acetylene according to
𝐶𝐻∗ + 𝐶2 𝐻2 → 𝐶3 𝐻3∗ (48)

Ions decay by reactions such as
𝐻3 𝑂+ + 𝑒 → 𝐻2 𝑂 + 𝐻 (49)
(Lewis & von Elbe, 1987)
Chemi-ionization occurs in hydrocarbon flames but not in flames that do not form free carbon, such as flames of
hydrogen, carbon monoxide, hydrogen sulphide and carbon disulphide. Therefore, peak ion concentrations is much
higher in rich than in lean flames as the concentration of transient graphite nuclei is much lower in lean flames than
in rich flames. Likewise, the ion concentration of methane flames is found to be much lower than that in flames of
higher hydrocarbons (Lewis & von Elbe, 1987).
Investigations done on flames containing metals which were introduced by means of salt sprays or in some other
methods. Atoms of alkali metals (Li, Na, K, Rb, Cs) have a low ionization potential of the order of 4 eV to 5 eV and
thus generate ions in energetic collisions (Lewis & von Elbe, 1987). The reaction of particulate suspensions of light
metals, aluminium particles, in particular, with the products of hydrocarbon flames is one of the critical stages in the
combustion of solid and metalized gelled propellants, pyrotechnics, and metalized explosives (Mphale & Heron,
2007).
55
The ion concentration attains a high peak value in the reaction zone and drops off sharply as the chemical process
goes to completion. Evidently, the ions owe their existence to chemical processes occurring in the reaction zone,
rather than to thermal collision processes (Lewis & von Elbe, 1987).
4.7 Temperatures of flames

An important consideration is temperature of hydrocarbon flames. These severe temperatures shown below will
greatly affect the degree of ionisation as described by Saha in equation (44). Thus, the conductivity of the flame will
increase with this increased ionisation (Sukhnandan, 2004).
Thermal Ionisation occurs when atoms or molecules gain sufficiently high velocity to cause ionisation on collision
with other molecules or atoms. This is the main source of ionisation in flames and high-pressure arcs. Work done
by Saha resulted in an equation to describe the degree of ionisation 𝜃 in terms of gas pressure and absolute
temperature:
3 (50)
𝜃2 1(2𝜋𝑚𝑒 )2 5 𝑊𝑖
= (𝑘𝑇)2 𝑒 𝑘𝑇
1−𝜃 𝑝ℎ
Where:
p = pressure in torr
Wi = ionisation energy of the gas
k = Boltzman constant
T = absolute temperature
𝑚𝑒 = mass of electron
h = plank constant (Sukhnandan, 2004).
Table 7 Flame Temperatures for some common fuel gases (Sukhnandan, 2004)
Reactants Flame Temperature(C)
𝐻2 + 𝑂2 2500
CO + 𝑂2 (+ 𝐻2 O) 2925
C𝐻4 +𝑂2 2780
𝐻2 + AIR 2045
CO + AIR(+𝐻2 O) 2000
C𝐻4 + AIR 1960
𝐶2 𝐻3 + AIR 2250
𝐶2 𝐻4 + AIR 1975
𝐶2 𝐻6 +AIR 1895
𝐶2 𝐻8 + AIR 1925
𝐶2 𝐻10 + AIR 1895
56
Lewis & von Elbe (1987) (Table 1, p717) shows no less than 79 combustibles, with flame temperatures between 1422°C (Amyl
acetate) and 2930°C (natural gas 1025 B.T.U.).
Several authors cited temperatures of fires. Wildland fires, with temperatures ranging from about 875°C to 2000°C
noted by Chandler et al. (1983) as cited in (Vodacek et al, 2002) Lanoie & Mercure (1997) estimated flame
temperatures of approximately 1,000 °C.
Flame temperatures usually lie in the range 1,000 K to 3,500 K (727 ºC – 3,227 ºC) (Bradley, 1965) (Bernard & von
Elbe, 1987), Palmer & Beer, !987), (Karlowitz, 1962) as cited in Sukhnandan, 2004).
4.8 Flame conductivity

It is explained in section 4.7 that in high temperature combustion of hydrocarbons that the major part of the
enthalpy of reaction is released rapidly in a narrow reaction zone leading to the production of very high
temperatures and because of the high temperature, all reactions will become correspondingly faster.
Sugar cane fires are noted to be fast burning fires. It therefore leads one to suspect that the fire itself must have
very high temperature flames Sukhnandan (2004). From section 4.6 and Table 8 it is shown that there are a large
number of ions present in the hydrocarbon flame.
Two of the most important conclusions that can be drawn from section 4.6.1 is that the presence of the electric field
in a flame changes the heat release rate and causes an ionic wind. This allows us to believe more easily that the
flame could possibly get very hot at times and thus becoming very conductive.
Table 8 - Ion Densities in various flames. Data taken from Lewis & von Elbe (1987) as cited in Sukhnandan (2004)
Typical Ion Density

Flame (ions/m³)
Observed Predicted
Hydrocarbon-air 1018 1012
CO or 𝐻2 -air 1012 1012
Gas seeded with K of 𝐾𝐶𝑙 1018 1018
All the above points to the fact that there is a high possibility of there being a very conductive flame present, even
for a very short time. This is the basis for the flame conductivity theory. It is believed that the flame channel itself
becomes very conductive due to the presence of ions and allows most of the voltage to be dropped across the air-
gap between the flame tip and conductor. (Sukhnandan, 2004) This causes an enhanced electric field in the air-
gap itself. Often the enhanced field is sufficient to engage corona inception and sustenance throughout the gap.
Thus, flashover is the result.
The conductivity of a fire was observed to be temperature dependent and found to be in the range of
0.0058 mho/m - 0.0079 mho/m in a fire with a maximum temperature of 1,240 K (967 ºC) (Mphale & Heron, 2008).
Flame resistivity 𝑅𝑓 found to be estimated by Lanoie & Mercure in equation (33) page 46 as
(5𝑘Ω/𝑚 < 𝑅𝑓 < 35𝑘Ω/𝑚)
According to Sukhnandan (2004) the flame resistance Rf and flame resistivity pf were estimated from the analysis
of voltage and video recordings in a 6.7m gap as
(50𝑘Ω ≤ 𝑅𝑓 ≤ 1000𝑘Ω)
57
(70𝑘Ω𝑚 ≤ 𝑝𝑓 ≤ 500𝑘Ω𝑚)
The current flowing in the flame is concentrated into an arc in less than 5ms (Lanoie & Mercure (1997) sic as cited
by Sukhnandan (2004)) and with the mean electric field gradient
In Figure 27 below it is shown that the ion concentration continues to increase outside of the luminous zone of the
flame. This may account for the observations that the flashover ensued as the flame (visible portion) approached
the live conductor.
Figure 27 - The distribution of ion-electron concentration in the combustion wave of a stoichiometric natural gas-air
flame (Karlowitz et al as cited in Lewis & von Elbe (1987)
It can be assumed that there are three sources of ions:
(a) Ions present due to the chemi-ionisation process involved in hydrocarbon combustion
(b) Ions that are created due to thermal-ionisation due to the temperature of the flame.
(c) Ions present due to photo-ionisation. Sukhnandan (2004)
It can be concluded that the flame itself is conductive due to the presence of ions produced from two sources,
namely from the fuel oxidation reaction described above and from the thermal ionization of the gas (Sukhnandan &
Hoch, 2004). When a flame is subjected to an electric field, the heat release rate changes and an ionic wind is
caused. The flame channel itself becomes very conductive due to the presence of ions and allows most of the
voltage to be dropped across the air-gap between the flame tip and the conductor. This causes an enhanced field
in the gap, which is then sufficient to engage corona inception and sustenance throughout the gap. The flashover is
the result (Sukhnandan, 2004).
58
5. Proposed model for the breakdown mechanism in a fire

5.1 Introduction
It is widely accepted that at the normal line clearances and rated operating voltage, a given line will inherently
experience flashovers under fire conditions (Cowan et al., 1991) (Chrzan & Wroblewski, 2004) (Claverie, 1971)
(Nel et al, 2012) (Vosloo et al, 2009). In the South African context, this applies mainly to 275 kV and 400 kV lines,
and on occasion to the 765 kV network and those clearances used with transmission line voltages. In South Africa
for example, no provision was made for the effect of grass fires in the determination of minimum midspan clearance
to prevent fire flashover until about 2010, when the minimum midspan clearance of 8m was increased to 10m
(Vosloo, Britten & Burger, (2009)).
The prevention and reduction of fire-induced flashovers have been extensively researched in several countries.
Increasing the minimum height of the phase conductors has been identified as one possible means of mitigation
(Islas 1988).). This chapter describes a method for predicting the relative probability of flashover as function of the
conductor height. The theory on which the method is based is given in the body of the chapter, with some
augmentation in the annexure.
The method being presented is novel in that it considers the criteria that must be satisfied to allow the development
of the power arc and hence establish a fault, which needs to be cleared by the line protection. This is distinctly
different from consideration of just the leakage currents which may or may not cause breakdown (Kunhardt &
Luessen, 1983, appendix). It is based on the theory developed by (Kunhardt & Luessen, 1983), (Claverie, 1971),
(Abdel-Salam et al., 2000). This method is not the only one that may be used for flashover predictions, but it is
backed up by some empirical data and considered appropriate.
The fire flashover is partially analogous to the pollution flashover process; however, the scale of currents and
voltages are much larger in magnitude than those present in the pollution flashover mechanism.
The study summarised here was to a considerable extent prompted by the finding made in Mexico that raising the
height of the conductors of various 400 kV lines by some 8 m to 10 m, over sugar cane fields, reduces the
occurrence of flashovers “substantially” (Islas, 1988).
This version deals with the fundamentals of fire-induced breakdown under AC, in particular, the question of the
conditions that lead to the transition from low leakage current to the formation of the power arc.
5.2 Basis and rationale for the model used to predict the likelihood of
flashover
If we consider phase-to-ground flashovers first, the following physical mechanisms are postulated:
 The engineering parameters of the mechanisms which cause flashover are similar to those (but not
identical to) leading to development of the pollution flashover on ceramic insulators.
 Some experimental evidence will be used to back up this postulate. Whether flashover occurs or not, is in
essence, determined by the resistance of the flame and attainment of a critical current and voltage.
 The key process is the transition from a low resistive current flowing through the flame to the creation of a
stable or unstable arc in which the current is high enough to cause complete flashover of the line-to-ground
insulation.
It is important to realise that it is usually only when a power arc occurs that the line’s protective relaying will
operate, and so make the utility aware that a fire fault has occurred. The fire fault caused in this manner will also
produce a voltage dip, whereas a burst of sub-critical current does not cause a fault to develop, and nor are the
relays set to detect such events- this can thus be said to represent a fire-withstand event. This will be covered in
more detail later.
It is further necessary to understand that breakdown of a high voltage gap (conductor-to-ground) must include a
non-linear or negative resistance element in series with the line-to-ground insulation. Hence, it is postulated that
59
the following simple circuits (Figure 28), expressed in terms of electric circuit parameters, may be used to study the
criteria for flashover:
Figure 28 - Conceptual diagram showing two main flashover paths (Shield wires not shown)
The essential elements of the phase-to-ground flashover path are now summarised in qualitative terms. This is
being done based on the simple electric equivalent circuit shown in Figure 29 (Lambeth, 1984).
Assume that the flame is a plasma having a resistance distributed uniformly along the length of the flame. As the
tip approaches, say, the outer phase, it can be considered to be the upward projection of an electrode which at
some point causes the gap to spark over. Depending on the magnitude of the resulting current, a stable arc may or
may not develop. The conditions for the development of a stable arc, which can grow in length, is complex, and will
not be explained for the time being.
The breakdown of the gap by means of a spark over is well explained in terms of the ionisation and gas chemistry
of the combustion processes in 4. It is now important to note that research done by Evert (Evert, 2003) into the
onset of corona on 275 kV and 400 kV lines showed that the positive corona inception voltage under real sugar
cane fire conditions is much lower than the corresponding values in normal dry and wet conditions, as is illustrated
in Figure 43 to Figure 49. The result of this difference is that the much higher numbers and amplitudes of corona
pulses produced in fire conditions, greatly increase the incidence of gap spark overs and the statistical likelihood of
any spark over event developing into a stable arc.
What basic factors determine whether a spark over event can lead to an arc or not?
The key factors appear to be combined effects of the instantaneous resistance of the fire, as a whole, and whether
the resulting current reaches a high enough value (“critical current”) to cause the development of a stable arc. The
current in this case must reach a calculable, critical magnitude. If the arc is in fact stable, the high currents which
can then flow, are determined only by the source resistance of the ac system, the system voltage, and the
resistance of the arc itself.
The transition from spark over to stable arc happens in microseconds, whereas in the case of a polluted ceramic
insulator, the whole flashover process takes much longer.
It is now timely and helpful to explain an important difference between the mechanisms of pollution and fire
flashovers; it is this:
 in a pollution event in which dry bands are formed, it is the current itself which, by means of heating of the
conductive pollution layer, causes shortening of the conductive path and lengthening of the arc
60
 this in turn causes the arc current to rise and the arc voltage to drop. This can be thought of as the transfer
of the system voltage to a decreasing length of the insulation path.
Flashover is inherently a very complex process and will thus only occur if the right combination of critical factors,
namely, electric field, leakage current and flame resistance, exist. (See 4.8).
5.3 The DC Case

In the study of arc phenomena in AC power systems, it is helpful to begin such a study with a review of arc
behaviour in HVDC circuits. (The analysis is simpler than in the case of HVAC).
The theory being presented here is based on extrapolations of the work on arc behaviour of polluted insulators
(Claverie, 1971), (Kunhardt & Luessen, 1983), (Islas, 1988), (Le Roy, et al. 1984, p268-275), (Lambeth, 1984),
(Cobine, -1958).
The starting point is to consider the following elementary circuit, as shown in Figure 29; the aim here is to derive
and understand the relationships between the variables. It is necessary to point out that the resistance of the flame
reduces with a rise in temperature and can be viewed as a negative resistance. This is one of the critical conditions
for flashover.
Figure 29 An equivalent circuit which has been adapted to the fire case (Lambeth,1984)
Unlike the case of the pollution flashover, the dry bands are not formed by the action of the leakage current but
occur randomly according to the movement of the flame. Hence, the only controllable factor to prevent flashover is
to increase the size of the gap between the flame and the conductor and is not a function of the magnitude of the
leakage current. It can therefore be deduced that flashover will develop with a flame approaching the conductor but
in theory, should the flame approach the conductor at or near zero voltage and then very quickly envelop the
conductor, a flashover is unlikely to occur. This example, however, is to illustrate the essential nature of the gap
between the flame and conductor for the formation of the arc. In practice, the vast difference in the time scale of the
frequency of the voltage (50 or 60Hz) and the time of the flame development, makes this an unlikely scenario in
real life.
The basic equation which describes the circuit is given by:
𝑉 = 𝑉𝑎𝑟𝑐 + 𝑉𝑅 + 𝑉𝐸 (51)
𝑉 = 𝐴𝑥𝑖 −𝑛 + 𝑖(𝐿 − 𝑥)𝑟 + 𝑉𝐸 (52)
where:
𝐴 and n are the arc constants

I is the peak current in the circuit in A
x is the length of the arc in m
L is the gap length in m
61
r is the resistance in ohms per unit length (cm or m) of the electrolyte or flame height
R is the total resistance in ohms
𝑅𝑥 is the total resistance over a path length of x cm or m.
𝑉𝐸 is the volt drop across the anode and cathode in volt.
See Figure 30 for an example of the calculated minimum voltage for flashover. (𝐴 = 150 and n=0.8)
The form of the arc equation being used is given by (EDF, 1984), (Lambeth,1984):
𝐴𝑥 (53)
𝑉𝑎𝑟𝑐 =
𝑖𝑛
where the symbols are as listed above and 𝑉𝑎𝑟𝑐 .is directly proportional to the length of the arc.
The values of A and n for the pollution case are respectively 150 and 0.8 (Rizk,1989). Values for fire are not known
at this stage. Although there is likely to be spread in the values of the constants, these will not change the
principles involved, nor the basic functioning of the equivalent circuit.
It is very important at this point to understand that V is not necessarily the supply voltage; it merely gives the
minimum voltage which must appear across the input to the circuit containing the arc and resistance elements. The
question of whether the arc is stable or not, will be considered later.
40000
Component voltages in basic DC circuit
35000
30000
TOTAL VOLTAGE IN VOLTS
25000
Volt drop across residual

20000 resi9stance (L-x)r
Arc voltage for the length x
15000
Total voltage
10000
5000
CURRENT IN A
Figure 30 - V-I characteristic for a fixed value of the arc length x=30cm and r=2000 Ω/cm
62
40
ARC characteristic for the arc length= 20 cm
35
30
VOLTAGE IN KV
25
VR resistive volt drop
20 arc voltage
total voltage
15
"maximum" voltage
10 system voltage
CURRENT IN A
Figure 31 - V-I characteristic for x=10 cm and r= 600 Ω /cm (Note that the x-axis is in amperes)
The graphs shown in Figure 30 and Figure 31can be combined, for a number of different values of the arc length x,
into the family of curves shown in Figure 32. Note that the curves all refer to the same electrode geometry and
value of series resistance.
63
Figure 32 - Family of curves for different arc lengths in a simple dc circuit (Le Roy, et al. 1984, p275)
To get a deeper insight into the behaviour of the dc arc, the minimum total voltage seen in Figure 30, Figure 31 can
be taken to be the minimum voltage above which an arc, whether stable or unstable, can exist. Thus, if the supply
voltage is above this value, which also depends profoundly on the value of x and the series resistance R, an
indication of the nature of the arc can be gained. A stable discharge is not observable to the right of point P where
the zone is weak (Le Roy, et al. 1984, p275)
The criteria for complete flashover of the gap may then be deduced from this analysis.
Referring to equation (52) the minimum voltage condition is given by
𝜕𝑉 𝜕 (54)
= (𝑉𝑎𝑟𝑐 + 𝑉𝑅 + 𝑉𝐸 )
𝜕𝑖 𝜕𝑖
𝜕 (55)
= (𝐴𝑥𝑖 −𝑛 + 𝑖(𝐿 − 𝑥)𝑟 + 𝑉𝐸 ) = 0
𝜕𝑖
After some manipulation, this gives:
64
1 (56)
𝑛𝑁𝑥 𝑛+1
𝑖0 = ( )
(𝐿 − 𝑥)𝑟
1 (57)
𝑛1 (𝐿 − 𝑥)𝑟 𝑛+1
𝑉0 = 𝑉𝐸 + (𝑛 + 1)(𝑁𝑥)𝑛+1 ( )
𝑛
Where
𝑁 = the arc constant (for guidance on finding 𝐴 and 𝑁 see Cobine, (1958))
𝑉0 = 𝑡he minimum voltage needed to allow the arc to exist is given by the voltage
Equation (52) has been plotted in Figure 33 for R = 3,000 Ω and L = 30 cm. (Note that 𝑉𝐸 has been neglected.)
For a given value of L, the maximum value, 𝑉𝑚 on the voltage minima or stability curve Figure 33 is found from the
relationship
𝜕𝑉0
=0
𝜕𝑥
Thus, 𝑉𝑚 is given by:
1 𝑛 (58)
𝑉𝑚 = 𝑁 𝑛+1 𝐿𝑟 𝑛+1
The value of 𝒙 which corresponds to 𝑉𝑚 is given by:
𝐿 (59)
𝑥=
𝑛+1
We now come to a very important point concerning the behaviour of the dc arc: it is this, namely, that if the arc is to
be able to extend from zero length to completely bridge the gap, the supply voltage must exceed 𝑉𝑚 . “This is a
necessary, but not sufficient requirement for flashover of a polluted insulator” (Lambeth, 1984).
It is accordingly postulated that this criterion can be adapted to the fire situation. This is based on the observation
from equation (58) that the maximum voltage depends on R and the length of the gap L. If the supply voltage of the
system exceeds the value of 𝑽𝒎 , then one of the key conditions for flashover, and not just spark over, has been
met.
The above observation provides a rational basis for the co-ordination of electric field stresses with the degraded
insulation strengths and low flame resistances caused by fire.
The second criterion for the growth of the arc and flashover is that a critical level of current must also be attained.
This current is derived in two ways, and both ultimately give the same result:
𝜕𝑉
By differentiation of (52) such that = 0, for a fixed value of x. Thereafter, the maximum value 𝑉𝑚 of 𝑉0 is found
𝜕𝑖
𝜕𝑉0
by putting = 0 . This yields equations (58) and (59) for 𝑉𝑚 and the critical value of 𝑥 respectively.
𝜕𝑥
1
The critical current is then derived by the substitution of 𝑥 = 1+𝑛 into equation (57), namely,
65
1 (60)
𝑛𝑁𝑥 𝑛+1
𝑖0 = ( )
(𝐿 − 𝑥)𝑟
Therefore
1 (61)
𝑁 𝑛+1
𝑖𝑐𝑟𝑖𝑡𝑖𝑐𝑎𝑙 =( )
𝑟
The second criterion, which is related to an energy increase and compensating balance principle and the resulting
onset of circuit instability, is that the power or the current drawn from the supply shall rise as the critical current is
𝜕𝑖
approached. This is embodied by the requirement that > 0.
𝜕𝑥
Put another way, if the supply voltage is constant, and if the arc root moves so that x increases, the total resistance
in the circuit is reduced. If it is assumed that when the arc grows in length by an amount δx, the arc voltage
increases by 𝑉𝑎𝑟𝑐 δx and the voltage across the wet pollution layer drops by an amount equal to the increment in
the arc voltage, namely, irδx. In the limiting case, thus, the following relationship will hold:
𝑁𝑖𝑐−𝑛 𝛿𝑥 = 𝑖𝑐 𝑅𝛿𝑥 (62)
Therefore
1 (63)
𝑁 𝑛+1
𝑖𝑐𝑟𝑖𝑡𝑖𝑐𝑎𝑙 =( )
𝑟
18000 minimum voltage profile

16000
minimum voltage profile
Stable
14000
region
12000
VOLTAGE IN VOLTS
10000
No arc possible
8000 in this region
The curve is the locus of
6000 the minimum voltage at
which a stable arc can exist
4000
2000
0
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31
ARC LENGTH IN CM
Figure 33 - Arc existence and stability limits for the elementary dc equivalent circuit shown in Figure 29
The graph in Figure 33 thus shows the minimum voltage V0 which must be attained to allow the formation of an arc
for arc lengths ranging from 0 cm to 30 cm.
66
1,2
Current which corresponds to the maximum
voltage
1
0,8
0,6
CURRENT IN A
0,4
current which corresponds to

the maximum voltage
0,2
0
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31
THE VARAIBLE X IN CM
Figure 34 - Current which corresponds to the voltage minima 𝑽𝟎 in Figure 33
It is interesting to observe in Figure 333 and Figure 34 that as the length of the arc extends beyond the critical
length, in this case 16.67 cm, the current increases rapidly and the voltage also drops rapidly; this appears to be
caused by the predominance of the arc characteristic, at least for this set of circuit parameters.
The fact that the voltage across the arc and the resistance of the pollution layer, (or the resistance of the flame),
drops as the current rises is the key phenomenon which allows flashover to occur.
Minimum
voltage
Resistance
Found from
V/I
Current
which
Responds to
Max voltage
Figure 35 - minimum voltage (orange), current (blue) and resistance (green) relationships for the 30 cm case
67
Note that the resistance values have not been shown in Figure 355 but they can be easily found from the data
contained in Figure 333 and Figure 344.
5.4 Application of the DC (direct current) model to fire flashovers

The following three questions are now to be studied:
 How exactly should the dc theory developed above be applied in practice to fires and the prediction of
flashovers?
 What values of flame resistance should be used in the predictions?
 How does the fire flashover methodology differ from that of the pollution flashover?
First question
Consider Figure 35 which is derived from Figure 33 and the two curves showing the minimum voltage profiles.
Assume as a starting point that a fixed height of flame and conductor are being considered. The variable in this
case is the resistance R/unit length of the flame. Thus, for some value of R, the equation for 𝑽𝒎 (see equation
(58)), the maximum voltage will be equal to the applied or system voltage. The creation of the unstable arc will be
preceded by the spark over of the heated air gap; this will in turn cause the development of the unstable arc and its
subsequent extension in length, and thereafter lead to complete flashover from the conductor to ground.
1 𝑛 (64)
𝑉𝑚 = 𝑁 𝑛+1 𝐿𝑟 𝑛+1
𝑛 (65)
= 𝐾𝑟 1+𝑛
It is emphasised that L in the above analysis is regarded as a constant. This is not unrealistic, since ac tests have
shown that the flame heights change relatively slowly, in hundreds of milliseconds, typically. (See 5.7, p76). A
theoretical analysis of the sensitivity of 𝑉𝑚 to variations in N, n and R has shown that the spread in the values of
𝑉𝑚 is within acceptable limits; this thus allows 𝑉𝑚 to be taken as a constant.
The arc constants in fire conditions are, for the time being, assumed to be invariant with the properties of the fire. It
can be argued at length that this is a reasonable assumption, but this will not be done now.
The basis of the insulation co-ordination methodology is thus to compare the deterministic value of 𝑽𝒎 with the
applied voltage 𝑽𝒔 ; if 𝑽𝒔 exceeds 𝑽𝒎 , then there is a high probability that flashover will occur. The probability is to
be taken as the ratio
𝑉 (66)
P1 = 𝑉 𝑠 for P1 ≤1
𝑚
and
P1= 1 for P1˃1 (67)
A fuller analysis must include the influence of variations in the value of L, but this is still to be done. In practice, this
would probably be a significant variable. (See Figure 36 below for more detail.)
Second question
The question of what values of the flame resistance to assume is very complex.
68
It has been found that the flame presents a resistive load at dc and power frequencies. Values in the range of
10 kΩ/m to about 30 kΩ/m are used for grass and cane fires.
Figure 36 - Example of dc stability characteristic for resistance values of 100 kΩ/m and 300 kΩ /m
(derived from Lambeth, 1984)
As regards the influence of ambient temperature, the likely mechanism is associated with enhanced combustion in
the fire which causes the resistance of the flame to drop. This must be investigated further by the working group.
Third question
The key differences between the ac pollution and fire flashover mechanisms are the following:
 The gaps under fire conditions are much longer than for typical pollution gaps.
 The pollution leakage current influences the length of the flashover path by causing dry bands to develop;
this does not happen in fire flashovers
 The flashover process is much quicker in fires than on polluted insulators.
Similarities include:
 The parameters which control the growth of the arc appear to be basically the same in both cases
 The resistances of the conductive path appear to be strikingly similar.
5.5 The AC (alternating current) case

The essential difference between the characteristics of ac and dc arcs is that, under ac, the arc current is
extinguished as it passes through zero in every half cycle. For the arc to be re-ignited, the voltage in the next half
cycle must be high enough for breakdown to occur.
69
Figure 37 - Typical arc voltage and current waveforms in a highly inductive a circuit (Abdel-Salam et al., 2000)
Figure 38 - Time domain responses for two different cases; inductively and resistively limited current waveforms
shown in the left and right-hand diagrams respectively (Abdel-Salam et al., 2000)
When the gap becomes too large for re-ignition to take place the arc is said to have been quenched. The voltage
near the current zero at which the current is interrupted is the extinction voltage. Typical waveforms of the voltage
and current as function of time are shown in Figure 39and Figure 40. An example of a related volt-ampere
characteristic is shown in Figure 41.
70
Figure 39 - Superposition of the static dc and dynamic ac volt-ampere characteristics (Abdel-Salam et al., 2000)
The overall breakdown process under ac is much more complex than for dc and will not be explored further. As a
result, it is considered acceptable for the key engineering results merely to be stated.
However, the simplified picture that the practitioner can use is that the flashover still depends on the attainment of a
critical voltage and current that will result in the complete flashover of the gap. This requires, as in the dc case, the
arc to extend to about two thirds of the length of the gap.
The ac re-strike voltage as a function of the length of the arc x is given by:
𝑉𝑐𝑥 = 𝐴𝑥𝑖 −𝛼 𝑣𝑜𝑙𝑡𝑠 (68)
Typical values of A and α are 1000 and 0.5 respectively.
71
Figure 40 - Dc and ac stability diagram for the pollution flashover model. (Lambeth, 1984)
Note that the solid curves refer to dc and stippled curves to ac. The dotted curve refers to the critical ac case for
the parameters chosen, namely, L=30 cm, r= 3000 Ω/cm, N=150, n= 0.8, A= 1000 and α=0.5.
It has already been mentioned that for flashover to occur under dc, the critical maximum voltage and the critical
magnitude of the current must be met before flashover can occur.
(See equations (59) and (65)).
The same requirements apply to ac as well, but the form of the equations differs.
It should be clear from the two points of intersection, which correspond to Ic (dc) and Iv (ac) that the ac arc
requires more power than the dc arc to sustain it. The critical ac current is also much higher than that needed for
dc.
The critical equations for the ac case are given as (EDF, 1984), (Lambeth, 1984):
Critical Voltage
2 (69)
𝑉𝑐 = 90(𝑥𝑐2 𝑟(𝑥))3
Critical current
1 (70)
700𝑥𝑐 3
𝐼𝑐 = ( )
𝑟(𝑥𝑐 )
72
Critical arc length
2 (71)
𝑥𝑐𝑟𝑖𝑡 = 𝐿
3
The above equations have been used in 5.8 p78 to calculate the resistance of the flames as a function of their
height.
5.6 Measurement of AC voltages and currents in fire conditions

The central hypothesis contained in this submission is that the arc equations which govern pollution flashovers can
be adapted to describing how fires cause line-to-ground faults on both dc and ac lines. Whilst this has not been
rigorously demonstrated by experiment, there is some practical evidence to support this contention.
A selection of relevant waveforms is shown in Figure 41 to Figure 45. The data is taken from single phase tests
done with sugar cane on gap sizes of 2, 3 and 4 m. The gap size refers to the separation between the conductor
and the top of the cane. The short circuit current of the test source was about 3 amps. The maximum output
voltage was 50 kV.
Figure 41 - Voltage and current waveforms over an intermediate period
The arc is preceded by a period of leakage current flowing through the flame.
73
Figure 42 - Voltage and current waveforms over an intermediate period
The arc is preceded by a period of leakage current flowing through the flame.
Figure 43 - Voltage and current waveforms which show the development of the arc
The detail shows the onset of the critical current and the distorted arc voltage waveform which suggests that
extinction and re-ignition are occurring.
74
Figure 44 - Voltage and current waveforms recorded over a long period
Significant current activity which was below the critical current is evident.
Figure 45 - Voltage and current waveforms recorded over a long period
Significant current activity which is below the critical current level is evident.
75
5.7 Measurements of fire-induced corona in transmission lines

The role of conductor corona is significant in the breakdown of the line insulation under fire conditions. In the late
1990s, Eskom, the South African power utility, did (and is still doing) research into the increase in radio noise
caused by fire-enhanced conductor corona (Evert, 2003)
Although the aim of this research was not primarily to study the breakdown mechanisms, the results did, however,
throw some light on the underlying physical processes which contribute to flashover.
It was established experimentally in Eskom’s Corona Cage that the quasi-peak radio frequency noise increased by
typically 20 dB above rain noise levels over a wide range of conductor surface gradients. The corresponding
increase with respect to dry conditions was nearly 40 dB. Such noise would appear to be easily detectable, but
exploiting this finding is another story.
Examples of the time and frequency responses are given in Figure 46 to Figure 50.
Figure 46 - Radio noise measurement at 0.5 MHz done in a corona cage in FIRE, RAIN and DRY conditions
Figure 47 - Radio noise measurement at 0.5 MHz done in a corona cage in RAIN conditions
76
Figure 48 - Radio noise measurement at 0.5 MHz done in a corona cage in FIRE conditions
Figure 49 - Broad-band time-domain measurement done in operational cane fire conditions
Note that in the above figure, the noise is generated by three phases; it appears that the coupling in this case is
dominated by two phases, probably the centre and an outer phase.
The waveforms shown in the four diagrams allow the following somewhat generic observations to be made:
 The fire noise at typical operational conductor surface gradients is substantially higher than in worst-
case rain conditions
 The corona inception gradients are much lower in fire conditions than in ambient dry conditions
 Fire noise occurs predominantly in the positive half-cycle
 The repetition rates of fire noise pulses are recognisably lower than for those in rain and dry conditions
Possibly the most useful practical finding is that each corona pulse or discharge is a potential cause of flashover;
thus, during a fire event, hundreds of such discharges act on the fire itself, but without necessarily causing
flashover. Only when a critical combination of flame height, flame resistance, ambient temperature and electric field
occur, will an arc form and flashover take place! This pulse pattern is further illustrated in the pattern of neutral
current pulses recorded during a cane fire under a 275kV line (Lambeth, 1984). See Figure 50
77
Figure 50 - Neutral current measured during a real cane fire under a 275kV line in DRY conditions (Britten & Hill, 1997)
The above pattern of pulses tends to confirm that a typical fire fault may cause numerous sub-critical current pulses
to be generated by a fire under a line. The critical pulse is certainly not a one-off event.
5.8 Predicted performance of transmission lines under cane fire conditions

The physical processes leading to the derivation of the critical electric field along the arc critical arc current are very
briefly introduced in the appendix. This refers mainly to corona, flame conductivity and the arc voltage gradient
which must be reached in order to make it likely that a stable or unstable arc will form, and so trigger independently
sustained, i.e., by the system voltage, earth or phase-to-phase faults.
The methodology used in the calculations is as follows:
 Assume the height of the vegetation, in this case sugar cane to be 5 m. (Figure 51)
 Take the body of the flame be 3 m high on average. (Other combinations of the above two heights can
of course be used.)
 The resistance of the cane and the body of the flame is taken as 20 kΩ/m
 Superimposed on the body of the flame is shaft of flame the average resistance of the shaft is taken to
be 1,300 kΩ/m If a conductor height of 17 m is assumed, the average resistance of the flame will be (9
x 1300 +8x20)/17 = 697.6 kΩ/m = R
1
The critical arc gradient is given by 𝐸𝑐 =10.43𝑅 3 = 92.511 kV/m. This represents the strength of the insulation.
2
 The peak value of the applied AC stress E0 = 420√ = 343 kV over the distance of 17 m is 20.2 kV/m
3
E0
 The ratio = 20.2/92.5 = 0.218. This is a component of the probability of flashover.
Ec
78
 The probability of the flames reaching a height of 17 m is deduced from Figure 53. In this case, Ph =
1
0.45. The temperature variation with height is proportional to where n is an exponent.
(shaft length)n
For 17 m, this gives 1/91.7 = 0.024.
 The overall probability is given by 0.218x0.45x0.024 = 0.00235 or 2.35x10 -3. The above procedure is
then repeated for each conductor height, flame combination and temperature exponent.
The results for phase-to-earth flashovers are shown for 275 kV, 400 kV and 765 kV lines in Figure 29 and Figure
30 respectively. Phase-to-phase flashovers are still being evaluated.
Figure 51 - Simplified structure of the flame. (Claverie, 1971), (Kunhardt & Luessen, 1983)
79
Probability vs height for 275 kV
TOTAL HEIGHT OF CONDUCTOR (M)

1
8 9 10111213141516171819202122232425262728293031323334
0,1
PROBABILITY OF FIRE FLASHOVER
H=7, exp=1
0,01 H=7,
exp=1.7
H=5, exp=1
0,001
H=5,
exp=1.7
0,0001
0,00001
Figure 52 - Predictions for 275 kV
10
1
PROBABILITY OF FIRE FLASHOVER
0,1
Mexican data lies here; in H=7, exp=1
range 9-16 m, ratio of outages
H=7, exp=1.7
0,01
H=5, exp=1
H=5, exp=1.7
0,001
0,0001
Matimba data, 11-14 m; ratio of outages about
0.2/0.07 ≈ 3
0,00001
8 9 10111213141516171819202122232425262728293031323334
Figure 53 - Predictions for 400 kV (Measured data shown)
80

1
PROBABILITY OF FIRE FLASHOVER 0,1
H=7, exp=1
0,01
H=7, exp=1.7
H=5, exp=1
0,001 H=5, exp=1.7
0,0001
1516171819202122232425262728293031323334
Figure 54 - Predictions for 765 kV (Note that the minimum conductor height is 15 m)
Figure 55 - Measured distribution of flame heights under “hot” ambient conditions
5.9 Comparison of Eskom results with Lanoie & Mercure (1997)

The purpose of this section is to review the interesting approaches taken by Eskom and IREQ (Lanoie & Mercure,
1997) (refer to section 4.3, Figure 25, p47) in the determination of the power flashover in the presence of
vegetation fires.
Figure 56 shows results of a brief review of the mechanisms of ac breakdown as a function of the flame resistance.
It suggests that the results are mutually consistent.
The IREQ data (Lanoie & Mercure, 1997): on forest fire flashover gradients and flame resistance differs from that
which has evolved in the so-called Eskom method, in the following respects
81
 The Eskom method focusses on the critical electric field gradients and the non-uniform flame resistance of
the fire which it is argued is a necessary contributor to the creation of the power arc.
 The Eskom method thus allows the probability of flashover to be calculated as a function of the height of a
line, but with many simplifying assumptions being made.
 The IREQ method is based on the determination of a critical flashover electric field gradient which just
precedes the transition to the arc state. It considers only the resistance of the flame without a gap in series
and implies the fundamental concept that as the flame resistance changes, so will the critical gradient. The
Eskom method, on the other hand, does consider a gap to be in series with the flame resistance. One of
the arguments used here is that flashover cannot occur unless an arc with its negative resistance is able to
elongate to a critical length which will allow the complete flashover to occur; without the negative resistance
element, flashover will not be possible.
 In essence, what IREQ have shown is that if the flame resistance just before flashover is known, or can be
estimated, the corresponding critical gradient can be used to obtain a deterministic value of the flashover
voltage. This is done by means of the empirical relationship
𝑘𝑉𝑟𝑚𝑠 (72)
𝐸𝑐 ( ) ≈ 4.2√𝑅𝑓𝑙𝑎𝑚𝑒
𝑚
 It appears that the IREQ method cannot easily be used to estimate the influence of the conductor height on
the risk of flashover; the main reason for making this statement is that flashover gradient has not been
shown to vary with the flame reach; this would inherently bring in height as a variable.
 The Eskom data appears to fit the IREQ transition curve more closely than the curve for the
𝑘𝑉𝑟𝑚𝑠 3
𝐸𝑐 ( ) ≈ 10.4√𝑅 assumed by Eskom.
𝑚
 The scatter in the 275kV data is probably attributable to variations in the flame reach; conversely, the data
obtained from the 66 kV laboratory tests (group of points on the left of the graph of Figure 56 is probably
enveloped by the flames to a greater degree than for higher voltages and larger gaps. One can thus expect
a smaller scatter at the lower voltages.
It is clear that both models are basically realistic, and in the opinion of the writer, show some promise in terms of
their applicability and validity in being able to predict fire-related clearance criteria in engineering terms.
Overall, it is felt that if more research were to be done, this would help to harmonise the two methodologies, and
thus to exploit their practicability. However, the information provides some guidance on possible determination of
the potential for power flashover in the presence of vegetation fires.’
82
Figure 56 - Comparison between flashover data obtained by IREQ (Lanoie & Mercure) and Eskom
5.10 Discussion
The calculation model contains many simplifying assumptions, but it does represent a rational starting point, being
based as it is on the development of a stable power arc fed from a strong source.
The experience gained in Mexico showed that by increasing the conductor height of the affected 400 kV lines from
8 to 16 m, the number of cane fire induced flashovers was reduced by a factor of about 10-20. Another way of
quantifying the improvement in performance was that the raising of the conductor by 8 m to a total of 16 m above
ground, reduced the operational stress from about 30 𝑘𝑉𝑟𝑚𝑠 /m to about 15.1 𝑘𝑉𝑟𝑚𝑠 /m.
The latter value appears to be a practical withstand gradient for the assumed hot fire conditions.
The question may be asked of how this value relates to the above predictions, for a height of 16 m, in Figure 30
(the 400-kV case).
The Mexican and Canadian data agree tolerably with the uppermost curve, (Figure 56) both as regards the
absolute values and the relative changes. The curve in question refers to high flames in hot conditions.
The Matimba data is taken from the fire performance statistics of two adjacent 400 kV lines (Vosloo et al., 2011).
The minimum conductor height of one line is 2-3 m greater than that of the other. The ratio of the total numbers of
fire faults was about 34:15 = 2.25 for the period under review. The agreement with the “hot” curve in Figure 30 is
tolerable at best.
83
5.11 Concluding remarks

The approach presented in this chapter was based on material presented in 2012. (Britten, 2012)
The paradox here is that the justification for the ac methodology is yet to be satisfactorily completed, but it is
contended that this submission is at least “a step in the right direction”.
The simple model presented in this very brief review serves as an introduction to the prediction of probability of
flashover to ground under fire conditions. It is emphasised that it is based on the critical arc gradient, and in this
sense, may be regarded as novel.
The phase-to-phase flashover model is being developed.
However, the information provides some guidance on possible determination of the potential for power flashover in
the presence of vegetation fires. The results presented here compare very well with the Matimba data (Figure 53)
84
6. Fire
6.1 Introduction
An understanding of the complex nature of fires in Nature and those factors that influence its behaviour, is crucial in
the in study of the subject of this work.
Figure 57 - A devastating fire with flames dwarfing huge eucalypt trees

(from 2009 Victorian Bushfires Royal Commission Report)
Many different terms exist, describing wildfires. In Australia, they may be called bushfires because they occur in
rural areas or “the bush”. It is more accurate to describe a wildfire by using the dominant fuel type, e.g. forest fire,
scrub fire or grassfire (Cheney & Sullivan, 1997)
Numerous experiments have been done in the study of the electrical breakdown of a gas during fires and results
were sometimes varied and interpretation divergent. One explanation could possibly be the variation in fires due to
different fuels which were used or atmospheric conditions that existed during those experiments. In very few cases
85
were these two factors brought into consideration, (Robledo-Martinez et al, 1991; Deno & Zafanella, 1982; Cowan
et al, 1991; Sadurski, 1997; Fonseca et al, 1987, 1990; Sadurski & Reynders, 1989).
It is therefore necessary to give an overview of those aspects and factors that influence the intensity of fires, the
rate at which they propagate as well as the effect of atmospheric conditions during a fire. This knowledge will not
only help in improving the interpretation of the results of experimental work but will aid in the design of overhead
lines that will be better at withstanding the effect of fires as well as aid utilities in the execution of vegetation
management strategies in fire prone areas. During fire suppression in particular, understanding these factors are
crucial to life and property.
Fire is a rapid chemical reaction which produces heat and light and the process is illustrated by the well-known
combustion triangle involving fuel, oxygen, and heat. When heat is applied, the fuel will first be heated and then
proceed to break down, releasing combustible gasses and water vapour.
Fuel
Figure 58 - The Combustion Triangle
At temperatures below 250°C the gaseous products are mostly non-combustible and the fuel becomes charred,
(e.g. charcoal). This process is an endothermic reaction that is expressed as follows:
𝐶6 𝐻10 𝑂5 + ℎ𝑒𝑎𝑡 → 6𝐶 + 5𝐻2 𝑂 (73)
This reaction, however, does not usually proceed directly to carbon as written in equation (74). The partial
breakdown of the cellulose molecules at temperatures above 250°C produces flammable gasses hydrocarbon
gasses and at high enough temperatures, these will ignite combining it with oxygen to produce the characteristic
diffusion flames. These gasses will burn on the outside and the overall reaction of the combustion of cellulose will
produce carbon dioxide, water and heat as follows:
𝐶6 𝐻10 𝑂5 + 6𝑂2 → 6𝐶𝑂2 + 5𝐻2 𝑂 + ℎ𝑒𝑎𝑡 (74)
(also see equations (73) and (76))
As discussed earlier, this process is exothermic and consequently heats up the adjacent fuel to the level for
combustion to take place. When the heating is very rapid, the breakdown becomes chaotic and large volumes of
gas, other hydrocarbons such as tars and molecular carbon (soot) are produced. Under these conditions the mixing
with oxygen and hence the combustion is incomplete and is evident by the billows of smoke (Cheney & Sullivan,
1997).
86
6.2 Fire initiation

Vegetation Management Plans to prevent or curb fires under overhead line network provide the means to manage
wildfire impacts. During the planning stage, it is important to understand how fires are ignited and how wild fires
then develop. The prevention of ignition of fires should be an important first step in the prevention and the
management of fires.
Lightning strike initiation

Electrical storms commonly occur in subtropical countries during periods of heat wave and ahead of frontal weather
systems. Lightning strikes to ground, trees and overhead lines have the potential to initiate a fire.
Figure 59 - A remarkable photo showing a lightning strike on a tree (photo H Hawes)
Certain areas like the Southwest of the United States are well known for lightning setting off major fires and some
statistics are astonishing. During 1960 to 1974, there were 12 days in Arizona and New Mexico where well over
100 lightning fires started. On the 28th of June 1960 lighting kindled 143 fires. In 1970 lighting initiated 100 fires on
the 18th of July with a further 100 the next day. On the 24th of June 1971, 103 fires started consuming the
vegetation on 75 713 acres. The Southwest’s national forest average more fires per year than any other region and
have the second-highest rate of burned area from both wild and controlled fires. Critical fire weather occurs
herewith greater frequency and persistence than anywhere in the rest of the USA (Pyne, 1997).
In South Africa, lightning storms occurring at the start of spring are known to cause fires in the dry grass following a
rainless winter. Overall, lightning is not a major cause of grass fires in that country because of its normal
occurrence during rainy summer when the grass is green and does not burn readily.
Pyro cumulonimbus cloud lightning

Pyro cumulonimbus are thunderstorms clouds that form above smoke plumes of intense fires with when very hot,
dry air masses rise above the fire column. These in turn caused lightning strikes to occur ahead of the advancing
fire storm and initiate further fire starts.
87
Experience in Victoria, Australia in the 2009 bushfires indicates that these fire starts can occur some 12 km ahead
of the main fire front. This phenomenon occurred during the Victorian Kilmore – Kinglake bushfire of 2009 and as
illustrated in Figure 60 (2009 Victorian Bushfires Royal Commission – Interim Report)
Figure 60 - Pyro cumulonimbus cloud lightning (Extract from 2009 Victorian Bushfires Royal Commission –
Interim Report Chapter 1, photo H Hawes)
(Source: Exhibit 126 – Bushfire CRC Interim Report (CRC.300.001.0001R) at 006744 Source: Exhibit 22 – Statement of M
Williams (WIT.013.002.0001) at 006845)
Anthropogenic actions
Man is the only species with the ability to make fires and have used this skill not only for preparing its food and to
warm its shelter, but also used fire in hunting, war and the manipulation of vegetation in its environment. Evidence
of aboriginal fires was found on virtually all continents colonized by Europeans and many plant species have not
only adapted to anthropogenic fire but some need fire for their germination. The farming model which the
Europeans introduced in the New World, however, excluded fire and fire controls and an almost total ban on fires
were introduced (Pyne 1997).
Many fires which get out of control, causing widespread property damage, are deliberately or accidently started by
irresponsible people, young and old. In most parts of the world this behaviour is considered criminal negligence in
terms of local legislation. In some rural areas of the world, for instance, fire is often used to smoke out game or
bees and these fires then spread if not extinguished properly.
Where strict fire bans were enforced during the past, fuel loads eventually accumulate over time. When the
inevitable fires start, they rage out of control and suppression becomes extremely difficult if not virtually impossible
(Cheney & Sullivan, 1997) (Pyne,1997).
Burning off of pastures by farmers

Prior to the regular severe bush fire season, it is common practice in rural areas of most countries to burn off
excessive grass fuel in the cooler months to promote new growth of grasses for cattle. Burning is also used in
places to reduce fuel loads. Unfortunately, weather changes can occur at short notice. In addition, fires that
continue to smoulder in larger tree remnants can suddenly rekindle and become wildfires.
88
It is becoming a modern practice of farmers in South Africa to burn off crop remnants and in this way enriching the
soil in preparation for the next crop. This a practice that is predominantly used where the same field is used to plant
both summer and winter crops and plant remains must be reduced in preparation of the soil. Neglecting to consult
the weather forecast prior to burning can lead to fires burning out of control if the weather changes.
Inadvertent ignition events

Fire starts during severe fire weather from the causes listed below, have previously occurred where the public and
landowners have not exercised due care.
 Cigarette embers
 Vehicle exhaust embers in grasslands
 Poorly extinguished bush camp fires
 Sparks from grinders and field welding
Arcing from conductor clashes in strong winds

Experience has shown that some older overhead lines with horizontally spaced conductors may clash under high
wind conditions resulting in flashover of phase conductors. Molten metal particles falling to ground have been
shown by research to have an elevated temperature when they land on the ground sufficient to cause spontaneous
combustion of cured dry grass trash. (Rallis & Mangaya, 2002).
Arcing from conductor contact with fast growing trees and shrubs
When periods of heavy rainfall in fertile soil areas are followed by periods of summer temperatures rapid growth of
certain species of small trees can occur. Experience with trial plots of different tree species has confirmed that
some species particularly those of the acacia family can grow from a seedling 1.0m tall before rain to 7.5m within 1
year and cause flash over. Fire initiation in vegetation under the line is thus a concern.
89
Figure 61 - A rouge tree under a 275kV line showing arc tracking to tree top and down tree trunk to ground where the
arc-initiated fire in grasses
6.3 The anatomy of a fire

In the study of fires, reference is made to various types of fires. In this section, the anatomy and terminology used
with fires will be explained. The intensity of forest or bush fires differ from that of grass and bush fires but the
terminology remains the same for all.
A fire front that moves with the wind is referred to as a head fire (or heading fire) and conversely the fire front that
moves against the prevailing wind is called a back fire (or backing fire). This is illustrated Figure 62.
These fires also spread at different rates and the ash residue for each can be distinguished from each other. The
head fire moves with the wind and the flames are blown towards the fuel, drying, and heating it, ready for ignition.
This fire has tall flames and the combustion can be quite inefficient in severe weather conditions, leaving very black
ash.
90
Figure 62 - the anatomy of a fire (Cheney & Sullivan, 1997)
Back fires move against the wind with flames that lean over the burnt fuel and ignite the fuel bed near the ground.
These fires move slowly and burn very efficiently leaving very little partially burned carbon and mostly a fine white
ash residue.
Flanking fires, on the edge of the fire, burn with the wind. Gusts of wind can cause high flames to travel in waves at
very high speed along the flanks equivalent to the gust speed. A flanking fire can change into a head or back fire in
response to wind changes (Cheney & Sullivan, 1997). This could turn a moderate fire in a catastrophic fire as
demonstrated in Figure 81.
The main characteristics of a moving fire are the flame height, flame depth, flame angle and flame length. Each of
these is difficult to estimate due to the dynamics of the combustion flame and the influence of wind flow into the
turbulence of the burning zone of the fire as shown in Figure 63
91
Figure 63 - further aspects of a typical fire Cheney & Sullivan (1997)
The resident time taken for a typical fire to remain burning over one spot depends on the thickness of the fuel bed.
Heavy pastures have a residence time from 10-15 seconds while light and well cured pastures have a residence
time of 5 seconds or less. All fires increase their rate of spread after ignition until a quasi-static equilibrium is
reached for the prevailing wind and fuel height (Cheney & Sullivan, 1997).
Surface fire (Thomas & McAlpine, 2010, p42), as the name indicates, are those fires burning the surface fuels.
When the intensity of the fire, is sufficient, it ignites the top of trees and is referred to as a crown fire (Thomas &
McAlpine, 2010, p42). After crowning, fires have been observed to increase their rate of spread. A model for the
predicting the ignition of crown fuels above a spreading surface fire was developed by Crux et al. (2006). The
model uses the characteristics of the surface fire energy, buoyant plume dynamics, heat sinks as described by the
crown fuel particle characteristics and energy transfer to the crown fuels.
When two fire fronts approach each other, the rising air accelerates and forms convection columns. (see Figure 1).
These columns can be responsible for extreme winds and even when the columns form in relatively benign grass
fires, they have been observed to cause flashovers due to the increased fanning of the fuel, raising the fire
intensity.
92
7. Fire behaviour
The way in which fire responds to various influences, will be discussed in this chapter. The manner in which, and
the factors that influence the release of heat energy, involves the study of fire behaviour which is defined as the
release of heat energy during combustion as described by fire intensity, rate of spread of the fire front, flame
characteristics and other related phenomena (Trollope, 1981).
This subject discussed in this chapter emanates from the research of fire ecologists and pastoral scientists. In
particular, the research done by scientists in Australia, Southern Africa and North American on grass fuels will be
discussed in this chapter (Trollope, 2002). The work of these scientists encompasses studying the effect of fire on
natural ecosystems and the response of living organisms to the release of heat energy through the combustion of
plant material. (Trollope, 2002) This chapter therefore will give the reader a good insight in understanding how fires
will behave.
Research on the effects of fire has been conducted throughout the grassland and savanna areas of Africa,
particularly in southern Africa, since the early period of the 20th century. Fire behaviour refers to the release of heat
energy during combustion as described by fire intensity, rate of spread of the fire front, flame characteristics and
other related phenomena (Trollope et. al., 1990) (Trollope, 2002).
The primary factors affecting fire behaviour are fuel load, fuel moisture, air temperature, relative humidity and wind
speed (Luke & McArthur, 1978).
In their study of fire behaviour, fire ecologists have developed parameters to quantitatively describe the behaviour
of fires as it pertains to the effect that the fire has on flora. Head fires for instance, release its heat above the
ground and has little effect on the growth points of grasses, whilst it has a marked effect on bushes and trees
higher up where the released heat directly affects the growth points. The basic parameters of fire that these
ecologists describe, however, also apply to the work done in this study on air gap breakdown and will be discussed
below.
The release of heat from the combustion of fuel has been shown to be an important factor in setting the conditions
for flashover during fires under overhead lines. The way, as well as the factors that influence the behaviour of fires
and in particular the release of heat, have been studied in countries such as Australia, the USA and South Africa by
fire ecologists. (Trollope et al, 2002).
The study of fire behaviour requires an understanding of the phenomenon of combustion. Brown & Davis (1973)
states that combustion is an oxidation process comprising a chain reaction. The heat released during a fire
originates from solar energy via the photosynthesis process and combustion is similar to photosynthesis in reverse.
It can be illustrated as follows:
For Photosynthesis:
𝐶𝑂2 + 𝐻2 𝑂 + 𝑠𝑜𝑙𝑎𝑟 𝑒𝑛𝑒𝑟𝑔𝑦 → (𝐶6 𝐻10 𝑂5 )𝑛 + 𝑂2 (75)
For Combustion:
(𝐶6 𝐻10 𝑂5 )𝑛 + 𝑂2 + 𝑘𝑖𝑛𝑑𝑙𝑖𝑛𝑔 𝑡𝑒𝑚𝑝𝑒𝑟𝑎𝑡𝑢𝑟𝑒 → 𝐶𝑂2 + 𝐻2 𝑂 + ℎ𝑒𝑎𝑡 (76)
where the kindling temperature merely has a catalytic role of initiating and maintaining the combustion process.
(Brown & Davis, 1973)
93
Three phases of combustion are described by Brown & Davis (1973) namely:
 The pre-heating phase, where plant material ahead of the fire is raised to its ignition point, driving off
moisture and generating flammable hydrocarbon gasses.
 The gaseous phase- The pre-heated fuel breaks down into gases and charcoal and flaming combustion
occurs with visible flames
 Combustion phase – The gasses burn off and residual charcoal is consumed by glowing combustion, the
stage of combustion where oxygen is limited, the ignition of these gasses characterized by flaming
combustion
 The final phase where the remaining charcoal is consumed by glowing combustion, leaving a small
amount of residual ash.
The fuel type will determine the amount of heat energy released during the combustion and glowing phase.
According to Brown and Davis (1973) heavy fuels (dense wood) with low flames will release a large portion of their
heat energy during the glowing combustion phase. Light fuel, such as grass will release most of its heat energy
during the flaming combustion phase. The three phases overlap and will occur simultaneously during a fire.
The intense combustion of the flammable hydrocarbon gasses during a fire is clearly visible in Figure 74. Also note
the marked difference in fire intensity with the rest of the surrounding fuels. This will be discussed further in 7.5.3.
The chain reaction of combustion is maintained and heat energy is transferred to fuel through a process of
conduction, convection and radiation as well as the movement of burning material through spotting (Steward,
1974). Besides spotting, convection and radiation are the prime factors in transferring heat to the fire front. The
effect of convection is clearly observed pre-heating higher shrubs layers and tree crowns. Radiation accounts for
most of the preheating of the fuel ahead of the fire front (Luke & McArthur, 1978).
7.1 Factors influencing fire behaviour

Trollope (2003) states that the study of fire behaviour is well advanced in countries such as the USA and Australia
and very sophisticated mathematical models have been developed in the USA to predict the behaviour of fires by
Rothermel (1972) and others.
Luke & McArthur, however, believed that for practical field use simpler models can be used based on general fuel
characteristics (particle size, distribution moisture content), slope, relative humidity, air temperature and wind
speed. Similarly, the rate at which a fire propagates through vegetation, can also be complex and more practical
and simplified approaches will be presented.
A review of the literature done by Trollope (2002) shows that the factors influencing fire behaviour are
acknowledged by numerous authors. (Brown & Davis, 1973; Luke & McArthur, 1978; Cheney, 1981; Leigh & Noble,
1981; Shea, Peet &Cheney, 1981; Wright & Bailey, 1982)
Rate of spread (ROS)

The rate at which a fire will spread is an important factor in fire behaviour. This parameter will permit fire fighters or
control centres to estimate the time that a fire will take to reach a power line or fire barrier and how much time is
available to implement a back burn, for instance.
The ROS can be complex to estimate as it is influenced by many factors, which will be discussed below. The
complexity of the ROS is well described in the work of Cheney & Gould, (1995) and shown graphically in Figure 64.
94
Figure 64 - The complex growth fire subject to frequent changes in wind direction (from Chaney & Gould, 1995)
The isochrones in Figure 64 represent the position of the fire perimeter at 2- minute intervals and the effective
head fire width was defined as the width of the fire front which influences the head fire spread for the next period.
The rate of spread is the maximum rate of spread normal to the previous isochrone (Cheney & Gould, 1996).
Although this approach appears to be complicated, it is well suited to computer applications where computing
power is not a problem.
For field use, a more simplified and practical approach will be shown to demonstrate the ROS of fires in grass fuels.
95
According to Trollope the rate of spread is
𝐴 (77)
𝑅𝑂𝑆 =
𝑇𝐿
Where ROS is the mean rate of spread (m/s), A is the area burnt, (m²), T is the period of flaming combustion (s), L
is the mean length of the fire front (m).
Predicting rate of spread

Predicting the rate of spread can be difficult, as seen from the above example as it is influenced by the atmosphere
and fuel with all its variations. Grass fires can spread at a rate of 1.7km/h (Frost, 2010 p64). Computer simulations
for predicting ROS are currently being developed using accurate, real-time weather parameters and fuel models.
The reader is also referred to Cheney and Sullivan, (2008) for further reading on the subject.
7.2 Available heat energy

The total amount of heat energy available from a fire is determined in part by the fuel load (Luke & McArthur,
1978). The heat combustion determined by using a Gallencamp automatic adiabatic bomb calorimeter in South
African grasslands indicated a mean heat combustion of 18,024±149 kJ/kg (Trollope ,1983).
This value is similar to those determined in the USA and Australia. (Trollope et al, 2002). (Luke & McArthur, 1978;
Vines, 1981). Not all of the potential heat energy of the fuel is released during combustion as some of the energy is
used to drive off moisture present in the fuel.
The heat release from head- and back fires also differ. Results obtained from burns carried out in fine grass fuels in
South Africa yielded heat releases of 16,890kJ/kg for head fires and 17,781kJ/kg for back fires (Trollope, 1983).
These values agree well with those quoted in the U.S.A and Australia (Albini, 1976; Luke & McArthur, 1978). See
Table 9 below.
Table 9 - Heat yield of typical grass species in Australia. (Cheney & Sullivan, 1997)
Grass species Moisture content Heat yield

(% Oven-dry (kJ/kg)
weight)
Phalaris tuberosa 6 16000-16300
Phalaris tuberosa 10 13700-13900
Phalaris tuberosa (partially cured) 47 11500
Themeda australis 10 14500-14900
Themeda australis 8 16900-17800
Eriachne spp. 11 15100-16700
Eriachne spp. 10 15200-18500
Sorghum intrans 10 16900-17600
Sorghum intrans (stems only) 13 17400-18600
Sorghum intrans (leaves only) 13 18600
In Table 9 heat yields are shown for grass species in Australia. As can be seen, the moisture content of the fuel
plays an important role in the amount of heat released during combustion.
96
Figure 65 - An illustration of the fire intensity around the perimeter of a fire (Cheney & Sullivan,1997)
Tests done on a number of grass species in the Eastern Cape region of South Africa resulted in an average heat
yield of 18,024 kJ/kg for a composite grass sample. This value is similar to the mean heat of combustion (18,558 +
550 kJ/kg) of herbaceous fine fuels determined by Smith (1982). These values compare well with those obtained in
the U.S.A (Trollope, 1983).
7.3 Rate of energy release and fire intensity

The rate at which a fire releases heat energy (fire intensity) was defined by Byram (1959) as the release of heat
energy per unit time per unit length of fire front and is numerically the product of the available heat energy and the
forward rate of spread of the fire front. It can be expressed by the equation:
𝐼 = 𝐻𝑤𝑟 (78)
Where I=fire intensity (kJ/s/m), H=heat yield (kJ/kg), w=mass of available fuel (kg/m²), r=rate of spread of the fire
front (m/s)
The formula proved to be useful in studies of fire behaviour and Albini (1976) noted that it to correlate well with the
height of lethal scorching of coniferous tree crowns. Similarly, Trollope et al (1985; 1990) found significant effects of
fire intensity on the top kill of stems, branches of trees and shrubs in the Kruger National Park in South Africa.
(Trollope et al, 2002). This will also influence the likelihood of the fire to initiate a flashover.
This formula can be converted to a relationship between fire intensity and fuel load, which is easier to estimate in
the field. It is based on using an average value for heat yield of fuel.
1 (79)
𝐼 = 𝑚𝑟
2
where, I = intensity (kW/m) w= fuel load (tonnes/hectare) r = spread (m/h)
It has been reported by forest fire experts that the range of intensities for forest fires is enormous – varying from
20 kilowatts per metre to greater than 100,000 kilowatts per metre (kW/m).
The flame height is related to the intensity of the fire. A typical relationship between flame height and fire intensity
is given in Table 10. Fire intensity is given as power per linear metre of flame front. It should be recognised that this
is an approximate relationship and will vary with topography, weather and fuel conditions at the site and the type of
fire fuel.
97
Table 10 - Flame Height versus Intensity
Flame height (m) Fire Intensity (kW/m)

0-0.2 0-50
0.2-0.5 100-200
0.5-1.5 200-400
1.5-3.0 500-1000
3.0-5.0 1000-2000
7-8 3000
10 4000
14 8000
16 10000
The reader is also referred to Cheney & Sullivan for further reading and Ryan (1981) for the estimate of forest fire
intensity.
7.4 Vertical distribution of heat energy

The perpendicular height of flames from ground level has proven to be a reliable indicator of the vertical distribution
of heat energy released during a fire. The vertical energy distribution was also measured at different heights above
the ground and showed considerable differences at increased heights. A significant relationship between these
measurements and the response of grass sward and woody plants were found in experiments conducted in South
Africa (Trollope, 2002).
Two examples of the overall and vertical release of heat energy are presented according to Mphale & Heron
(2007).
In the example below, Mphale & Heron (2007) demonstrates an example typical of high intensity forest fires typical
of those in Canadian boreal forests Mediterranean regions of Spain and Australian deciduous forests, where a
peak intensity of 100MW/m can be reached (Williams et al, 2003) in jack pine crown fires (Pinus banksiana). This is
demonstrated in Table 11 and Figure 66.
Table 11 - Fire behaviour in jack pine crown fire (Mphale & Heron, 2007)
Max fire intensity Rate of spread Flame height Av. canopy height Max. Temp
MW/m m/min m m °C
89.7 69.8 20-30 12 1330
Figure 66 - Vertical temperature variation with plume height for a 90MW/m crown fire (Mphale & Heron, 2007)
98
Temperature above the jack pine canopy is considered to decrease exponentially to about a 1,000 K at about 21 m
aboveground, that is, at B (Figure 66). This is at the intermittent flame zone of the plume. It then decreases
gradually to the ambient temperature higher up in the plume. The maximum temperature of 1330°C for the
simulated crown fire is at A in Figure 66 (Mphale & Heron, 2007).
The height of the flame (Lf) above the crown canopy is calculated from the relation given by Butler et al. (2004) as:
𝐿𝑓 = 0.0175𝐼 0.67 (80)
Also see the estimate by Nel et al. (2013). They cite Kenhardt (http://www.forestencyclopedia.net/p/p492:
𝐿 = 0.0175𝐼 0.46 (81)
The next simulation is done for very high intensity forest fires in olive tree (Olea europaes) shrub.
Table 12 and Figure 67 show the behaviour of a shrub fire in olive tree forest (Santoni et al., 2006) in Mphale &
Heron (2007).
Table 12 - Fire behaviour in olive shrub fire
Max fire intensity Rate of spread Flame height Av. canopy height Max. Temp
MW/m m/min m m °C
20.500 0.4 7.5 3 1,061
Temperature above the shrub canopy also decreases exponentially to an intermittent flaming zone which is about
7.5mabove the ground level. The maximum temperature of 1060°C occurs at about a meter below the shrub
canopy (A) in Figure 67.
Flame length from the ground level has been determined from the equation for fire line intensity which is given by
Santoni et al. (2006) in Mphale & Heron (2007) as:
𝐼 = 259.83𝐿2.174 (82)
Figure 67 - Vertical temperature variation with plume height for a 21 MW/m shrub fire (Mphale & Heron, 2007)
99
Further data on fire intensities can be found in Nel et al. (2013).
7.5 Fuel and its characteristics

The characteristics of the fuel burning under a power line have a marked influence on its ability to cause line faults.
The chemistry associated with the combustion of various fuels has been discussed in 4 (4.6). The fuel type will
determine the amount of heat energy released during the combustion and glowing phase. According to Brown and
Davis (1973) heavy fuels (dense wood) with low flames will release a large portion of their heat energy during the
glowing combustion phase. Light fuel, such as grass will release most of its heat energy during the flaming
combustion phase. The three phases overlap and will occur simultaneously during a fire.
The assessment of the hazard remains a difficult task. As far as measurable parameters are concerned it is
possible to measure fuel load and fuel moisture. The hazard that vegetation poses is highly influenced by weather
conditions at the time. See the example in 8.3.1. This does unfortunately not exonerate the utility from carrying out
the task of vegetation management. Examples of the fire hazard posed by different types of vegetation are shown
in the Overall Fuel Hazard Guide (2009) Published by the Victorian Government in Australia. It states that:
Overall Fuel Hazard = (the sum of the influences of) Bark Hazard + Elevated Fuel Hazard + Surface Fine
Fuel Hazard
This approach represents a significant change in the philosophy of assessing the fuel hazard affecting fire
behaviour. Rather than simply considering surface fine fuel loads (in tonnes/hectare) only, as is customary in grass
fuels, it shifts the emphasis to considering the whole fuel complex, and particularly the bark and elevated fuels in
woodland and shrublands. The bark and elevated fuels are the fuel elements principally responsible for both first
attack failure and also for general suppression difficulty in Victorian forests, woodlands, deserts, heathlands and
shrublands (Anon, 2002)
Fuel load
In grass fires the amount of fuel available in a fire, or fuel load, is regarded as one of the most important factors
influencing fire behaviour. This is because the total amount of heat energy available for release during a fire is
related to the quantity of fuel. (Luke & McArthur, 1978). Brown & Davis (1973) stated that if a constant heat yield for
the fuel is assumed, then the intensity of a fire will be directly proportional to the amount of fuel available at any
given rate of spread of the fire front.
Trollope (1983) found a near linear relationship between fire intensity and fuel load during controlled burns carried
out in parts of South Africa. He showed that a fuel load of 1,500 kg/ha yields a 1,000 kJ/s/m fire intensity and 7,500
kg/ha a fire intensity of 6,500 kJ/s/m. He found the average grassland fuel loads to be in the region of 4,500 kg/ha
resulting in fire intensities of 4,000 kJ/s/m.
7000
FIRE INTENSITY - kJ/s/m
6000
5000
4000
3000
2000
1000
0
1500 2500 3500 4500 5500 6500 7500
FUEL LOAD - kg/ha
Figure 68 - The effect of fuel load on fire intensity as recorded in South Africa (Trollope, 2002)
100
Experience has shown that while the effect of fuel load on fire intensity is valid, some extreme weather conditions
can override this and line faults can occur in very light fuel loads. An example is discussed in 8.3.1.
The surface fine fuels, bark and elevated fuels in Woodland requires the estimating of each of these fuels. A
method which is described the Fuel hazard guide for estimating the fine fuels, uses a method akin to that of the
disc pasture meter developed by Bransby & Tainton (1977) (in Trollope, 2002).
This technique involves relating the settling height of an aluminium disc dropped onto the grass sward to the
standing crop of grass holding up the disc, expressed in kilograms per hectare. There is a simple relationship
between the settling height and the standing crop of grass based on the disc pasture meter. This device has been
successfully calibrated for much of the grasslands and savannas in southern and east Africa. Research and field
experience indicate that the calibration developed in the Kruger National Park (Trollope & Potgieter, 1986), in
South Africa, can be used as a general calibration for estimating grass fuel loads for management purposes in
these regions of Africa (Trollope, 2002). The calibration equation is: y =
𝑦 = −3019 + 2260√𝑥 (83)
where y = mean fuel load - kg/ha, x = mean disc height of 100 readings - cm.
Figure 69 - An example of the disc pasture meter used to estimate fuel loads in grass (Trollope, 2002)
The method for sampling the surface fine fuel hazards in the Overall Fuel Hazard guide (Anon, 2002) uses a simple
depth gauge constructed from a ruler and a 150mm circular piece of Masonite or plywood with an opening that can
accept the ruler. The disc is pushed down with a very light pressure, and the ruler is read off level with the top of
the disc. Measurements should be taken at convenient intervals (suggested 200 m to 1000 m) around a burning
block so that the prevailing range of fuel types is sampled
Surface Fine Fuel Hazard can be assessed by measuring litter-bed height. Litter-bed height should be measured
using a simple depth gauge constructed from a ruler and a circular piece of aluminium as shown in Figure 69. A
small gap is made in the litter-bed down to mineral soil and the end of the ruler is placed resting on the mineral soil
surface. The disc is pushed down with a very light pressure, and the ruler is read off level with the top of the disc.
(Note that the end of the ruler beyond the scale has been adjusted to match the thickness of the disc). At least 5
101
measurements of litter-bed should be made at each sampling site. Measurements should be taken at convenient
intervals (suggested 200m to 1000m) around a burning block so that the prevailing range of fuel types is sampled.
The Table 13 can be used to convert litter-bed height to Surface Fine Fuel Hazard.
Near-surface fuels – i.e. grass tussocks, dead bracken, low shrubs or low wiregrass up to 0.5 m high – interact
with the surface litter to increase fire behaviour, and therefore need to be compensated for when assessing
Surface Fine Fuel Hazard (see McCarthy in prep). If the site has high cover values of near-surface fuels, then you
should assign the next higher Surface Fine Fuel Hazard rating to it than would normally have been the case for that
litter-bed height. For example, a site with 30 mm of litter-bed height, but with the site covered by grass tussocks
with cover values of 40% or greater, would go into the Very High class rather than the High class.
Table 13 - Surface Fine Fuel Hazard rating based on litter bed height (Anon,2009)
Surface fine
fuel hazard low medium high Very high extreme
rating
Litter bed < 15 15-25 25-35 35-50 50 >

Height (mm)
Equivalent <4 4-8 8-12 12-20 20+

Litter Load
(t/ha)
Elevated fuel comprises shrub, heath, and suspended material. The level of hazard depends on the fuel continuity
(horizontal and vertical), height, amount (weight), proportion of dead material, thickness of the foliage and twigs,
and flammability of the live foliage. The flammability of the elevated fuel is highest when: the foliage, twigs and
other fuel particles are very fine (e.g. maximum thickness 1-2 mm); the proportion of dead material is high; the fuels
are arranged with a high level of density and horizontal and vertical continuity that promotes the spread of flames;
and the live foliage has low live fuel moisture contents. The vegetation type and the time since the last fire,
substantially determine the level of the elevated fuel hazard
The elevated fuel level hazard is explained by means of characterising descriptions such as “Easy to walk though
in any direction. Elevated fuel virtually absent” for low elevated fuel hazard and coupled with photographs showing
typical examples. Similar examples exist for the other classes.
Finally, the hazard of bark is done by means of descriptions of each hazard class.
Moderate bark hazard, for instance is described as:
“Very little bark is available to allow spotting to occur. Fires with a flame height of 0.5m to 1.0 m will not ‘climb’
these trees, and so spotting generally does not cause a problem. Moderate Bark Hazard is characterised by:
 Stringybarks – the bark is black for a substantial distance up the bole of most trees
 other bark types – where the fibrous/sub-fibrous bark is fine textured and held very tightly to the bole of the
tree, e.g. Box, Ironbark”
The Overall Fuel Hazard rating for a site done by using a table for each bark hazard category with associated
elevated fuel hazard on the one axis and surface fine fuel hazard on the other. The starting point for these tables is
Bark Hazard in each case.
102
Table 14 Example of Bark Hazard: Low to Moderate
Surface Fine Fuel Hazard

L M M H VH E
Fuel hazard
L L M M M H H
Elevated
L L M M M H H
L L M M H VH VH
VH VH VH VH VH VH VH
E E E E E E E
The equivalent fuel loads for the various hazard levels for each fuel component is given by the following table;
Table 15 - Equivalent fuel loads (t/ha) for given ratings. (Anon 2009)
FUEL Low Mod High V High Extreme

Bark 0 0 2 5 7
Surface fine 2 5 10 16 20
Elevated 0 0 3 6 10
To calculate the overall fuel load for the whole fuel complex, add the values for each component, e.g.
High Bark High Surface V H elevated

2 + 10 + 6
= 18 t/ha Overall
This total fuel load can then be applied to the McArthur Forest Fire Danger Meter MK V (1973) for predictions of
forward rate of spread and flame height (Anon, 2009) http://www.csiro.au/en/Research/Environment/Extreme-
Events/Bushfire/Fire-danger-meters/Mk5-forest-fire-danger-meter
Figure 70 - McArthur Forest Fire Danger Meter MK V (1973)
103
The FFDM gives a fire danger index that is related to the chances of a fire starting, its rate of spread, its intensity,
and its difficulty of suppression, according to various combinations of air temperature, relative humidity, wind speed
and both the long and short-term drought effects.
An index of 1 means that a fire will not burn, or, will burn so slowly that control presents little difficulty. An index of
100 means that fires will burn so fast and hot that control is virtually impossible.
The index is divided into five fire danger ratings (Low, Moderate, High, Very High, and Extreme) that represent the
degree of difficulty of suppression.
In 1973, a metrified version of the FFDM was produced. Since that time, the FFDM has been widely accepted and
is used by all rural fire authorities in Australia (except WA) and the Bureau of Meteorology.
Figure 71 - The CSIRO Fire spread calculator for Northern Australia (Cheney & Sullivan, 1997)
Fuel moisture
Fuel moisture is an important factor that influences fire intensity as it affects ease of ignition, combustion rates of
different types of fuel as well as the quantity of fuel that is consumed. The fuel moisture has a smothering effect on
the fire as the water vapour is released. It also reduces the amount of oxygen in the vicinity of the fuel, thereby
decreasing the rate of combustion. (Brown & Davis, 1973, as cited by Trollope, 2002).
The inverse relationship between fire intensity and fuel moisture was recorded in South Africa by Trollope and is
shown in Figure 72 - The effect of fuel moisture on fire intensity as recorded in South Africa (Trollope, 2002)
104
5000
FIRE INTENSITY - kJ/s/m

4500
4000
3500
3000
2500
2000
1500
1000
500
0
5 10 15 20 25 30 35 40 45 50 55 60 65 70
FUEL MOISTURE - %
Figure 72 - The effect of fuel moisture on fire intensity as recorded in South Africa (Trollope, 2002)
Trollope (2002) cites Luke and Mc Arthur (1978) that distinguish between the moisture content of living plants,
which vary in response to seasonal and climatic changes, and cured plant material. Due to the hygroscopic nature
of the latter material, its moisture content is affected on an hourly basis by absorption and dispersion in response to
atmospheric changes in temperature and relative humidity. The dominance of fire faults during the warm (and dry)
part of the day possibly relates to the higher fire intensity of dry material as compared to the more moist conditions
of the early morning and nights (Figure 78).
According to Cheney & Sullivan (1997) dead fuel moisture content and wind speed are the two most important
variables affecting how a fire will spread. It is unfortunately not easy to estimate in the field. Under laboratory
conditions, the fuel is weighed before and then dried out in an oven for 24 hours at 103°C (Cheney & Sullivan,
1997).
Apart from visual signs of wet limping grass or brittle leaves curling when dry, there is no reliable way of visually
estimating moisture content in the relatively narrow range of 2%-10%oven-dry weight where a change of 2% to 3%
can substantially alter a fire’s rate of spread.
The moisture content of dead fuel is determined by the relative humidity of the air, changing from around 100% at
night when dead fuel can absorb around 35% of its weight in moisture. This is referred to as the saturation content.
As the humidity falls during the day, so does the moisture content albeit more slowly. The opposite happens at
night time again. A generalized relationship between the relative humidity and ambient temperature have been
developed and is shown in Figure 73.
105
Figure 73 - The relationship between ambient air temperature, relative humidity and dead fuel moisture content under
field conditions (Cheney & Sullivan, 1997)
Type of fuel, veld condition fuel factor and fuel score

During uncontrolled wildfires, certain plant species were observed on Eskom transmission lines capable of
generating particular flame heights and fire intensities sufficient to bridge the 8m gap on 400kV lines between the
earth and the conductor, resulting in a line fault. The awareness of these plant species directly under the conductor
proved to be important in vegetation management and the reduction of fire risk. Other aspects of fuel types were
discussed above in 7.5.1.
The effect of different plant species was observed by Lanoie & Mercure (1997) in 4.3.
Research in the Kruger National Park in South Africa has shown that it is both feasible and practical to describe
quantitatively the condition of the grass sward in terms of its potential to produce grass fuel that significantly
influences the intensity of a fire (Trollope, 1990).
In the context of the vegetation growing underneath the power lines the veld condition relates to its potential to
produce plant fuel that will generate fire intensities and flame heights sufficient to cause a line fault.
A fuel factor is a number allocated to the different herbaceous species occurring in the grass sward on a scale of 1-
10. A fuel factor of 10 indicates a high genetic potential for a plant species to produce high fuel loads of flammable
plant fuel and vice versa.
The Fuel Score is calculated by summing the products of the percentage frequency and the fuel factor for each
grass species recorded in the grass survey. The range in the fuel scores from very high (>500) to very low (<200)
reflect the potential of the grass sward to produce to support a high intensity grass fire. These categories have
proven to be ecologically meaningful with highly applicable practical management implications (Trollope, 2004).
Eskom conducted a fire behaviour trial in to determine the potential of specific plant species to generate different
fire intensities and flame heights under different atmospheric conditions, particularly air temperature and relative
humidity. This study executed by Trollope (2004) is presented below.
106
The study focused on plant communities containing significant stands of Cosmos bipinnatus (Cosmos), Tagetes
minuta (Khakibos), Protoasparagus laricinus (Katbos) and Acacia mearnsii (Black Wattle).
The project was conducted in the Bankenveld veld type (Acocks, 1975) in the Gauteng Province of South Africa
where these plant species form abundant components of the vegetation in certain areas.
Results from the fire trials are shown below in Table 16 below.
The results in Table 16 indicate that there were marked differences in the fire intensities and flame heights
recorded during the controlled burns.
It must be noted that the atmospheric fire danger conditions that prevailed during the test were rather benign
because it is not possible to safely conduct experimental burns under more severe conditions. Consequently, the
fire intensities were much lower than those that would occur under wild fire conditions.
In spite of the very moderate conditions, evident from wind speeds and rate of spread (ROS), flame heights of 8 m
to 10 m were observed. High fuel scores, resulting from certain plant species and significant fuel loads, were
responsible for these flame heights. These flame heights would be sufficient to cause a line fault.
Table 16 - The environmental conditions and flame heights, fire intensities and rates of spread recorded during the
application of the controlled burns in the ESKOM fire behaviour trial
SITE Date Time Air Wind FDI Max FIRE FUEL

Fuel RH ROS
Temp Speed Flame INTENSITY SCORE
load
% m/s Height
kg/ha °C km/h kJ/s/m
m
1b 29/8/03 12:35 2071 25.6 18.2 9.4 62 0.8 8 2798 321
4b 29/8/03 10:55 3390 23.4 21.0 7.2 57 1.1 8 6127 468
5 28/8/03 15:45 2853 25.5 24.2 9.7 62 0.8 5 3903 358
6 29/8/03 13:56 751 26.5 17.7 3.5 62 0.2 4 190 151
7 29/8/03 16:15 4535 26.3 18.2 7.5 62 0.9 10 6511 516
8 29/8/03 17:35 1871 20.8 27.3 2.7 48 0.8 8 2370 412
Legend: Air Temp = Air Temperature; RH = Relative Humidity; FDI = Fire Danger Index; ROS = Rate of Spread.
The vegetation of each site is shown in Figure 75 and the flame height, fire intensity and fuel score is shown here.
The effect of the environmental factors on these two fire behaviour parameters were determined with simple
regression analyses where fuel load, air temperature, relative humidity, FDI and fuel score were the independent
variables and fire intensity and flame height were the dependent variables. The resultant correlation coefficients are
presented in Table 17.
An important outcome of a fire trial by Eskom was the objective assessment of the fire intensity potential of certain
herbaceous plant fuels. The identification of Cosmos and Khakibos as types of fuel that have a significantly higher
potential than any of the dominant grass species in the Bankenveld of Gauteng Province i.e. Eragrostis curvula,
Hyparrhenia fillipendula and H. hirta is of particular significance
It provides quantitative confirmation of previous field observations that these two plant species are prime culprits in
exacerbating fire intensities and flame heights of wildfires burning under extreme atmospheric conditions resulting
in line faults in Eskom transmission lines.
107
Figure 74 - The contrasting low fire intensities generated by the short grass sward and the high intensities generated
by the green Katbos shrubs during an ESKOM fire behaviour trial
This difference in fire intensities of plant species is often observed by field staff during fires, as can be seen in the
photo (Figure 76). This is attributed to the oils and terpenes that the plant releases during a fire. The terpenes are
aromatic hydrocarbons and are thought be a protective function of plants. Laboratory experiments done by Eskom
with oil from Tagetes minuta (Khakibos) showed that flashovers readily occur in flames when the oil was
introduced.
Table 17 Result of regression analysis
DEGREES CORRELATION
COMPARISON SIGNIFICANCE
FREEDOM COEFFICIENT
FIRE INTENSITY vs FUEL LOAD 4 0.9726 P<0.01

FIRE INTENSITY vs AIR TEMPERATURE 4 0.0168 NS
FIRE INTENSITY vs RELATIVE HUMIDITY 4 -0.0312 NS
FIRE INTENSITY vs WIND SPEED 4 0.5148 NS
FIRE INTENSITY vs FDI 4 0.0814 NS
FIRE INTENSITY vs FUEL SCORE 4 0.9078 P<0.05
FLAME HEIGHT vs FUEL LOAD 4 0.7092 P<0.11
FLAME HEIGHT vs AIR TEMPERATURE 4 -0.2648 NS
FLAME HEIGHT vs RELATIVE HUMIDITY 4 -0.0344 NS
FLAME HEIGHT vs WIND SPEED 4 0.1594 NS
FLAME HEIIGHT vs FDI 4 0.2505 NS
FLAME HEIGHT vs FUEL SCORE 4 0.8412 P<0.05
108
The result of the fire trial permitted a comparison of fire intensities of various herbaceous plant fuels and is shown
in Figure 74.
3000
FIRE INTENSITY - kJ/s/m 2500
2000
1500
1000
500
TYPE OF PLANT FUEL
Figure 75 - Fire intensities associated with different types of fuel as recorded two meters above the canopy of the
vegetation in the ESKOM fire behaviour trial
Figure 76 - Contrasting flame heights of fires generated by Cosmos/Khakibos (left) and Eragrostis curvula dominated
plant communities (right)
The difference in fire intensity between grass, dry and wet Katbos was also observed and is shown below. This is
attributed to the green leaf material of the GREEN KATBOS containing greater amounts of the
flammablesubstances (oils and terpenes) resulting in higher fire intensities than that of the DRY KATBOS shrubs
(Trollope, 2003).
109
450
FIRE INTENSITY - kJ/s/m 400
350
300
250
DRY KATBOS
200
SHORT GRASS
150 GREEN KATBOS
100
50
0
GROUND CANOPY CANOPY CANOPY CANOPY
LEVEL HEIGHT HEIGHT+1m HEIGHT+2m HEIGHT+3m
MEASURING HEIGHT
Figure 77 - A comparison of fire intensities of dry Katbos, Short grass and green Katbos shrubs
It is clear that the type of fuel is an important factor in determining the risk that it poses for flashover during a fire
and the work presented above will enable utilities to identify and manage the vegetation under power lines
Fuel size
In their study of a simple and practical approach, Luke and Mc Arthur (1978), as cited by Trollope (2002), classify
plant fuels into two sizes. Plant material with a diameter up to 6mm including standing grasses and other
herbaceous plants, fallen leaves, bark and twigs are classified as fine fuels. These burn readily and the combustion
is almost complete in grass fires. Heavy fuels constitute plant material larger than 6 mm and these are normally
found in forests. The ignition is slower and combustion is not as complete as with fine fuels. (Also see the
discussion in 7.5.1).
Fuel distribution
Authors recognize three broad groups in the vertical distribution of fuels that have a marked influence on fire.
These are ground, surface and aerial fuels (Brown and Davis, 1973) as cited by Trollope (2002) (Anon, 2002). The
structure of surface fine fuel has been found to be better related to fire behaviour than the surface fine fuel load
(McCarthy in prep in Anon, 2002). Also read in conjunction with 7.5.1 and Anon (2001)
 Ground fuels include all combustible material below loose surface litter and are difficult to ignite but
persistent once ignited. It supports glowing combustion (Trollope, 2002).
 Surface fuels include standing grass swards, shrublet communities, seedlings, and forbs (Trollope, 2002).
These fuels are all fine fuels and support intense fires in direct proportion to the quantity of fuel per unit
area (Brown and Davis, 1973) as cited by Trollope (2002).
 Aerial fuels include all combustible material found in the canopies of trees and shrubs and these fuels
support crown fires (Trollope, 2002).
110
Discontinuities in the grass fuels, such as the tussock grasses in Australia, will hamper the ROS (Cheney &
Sullivan, 1997)
Fuel compaction and fuel structure

For successful combustion, enough oxygen must be available. The amount of compaction of plant material
influences the ability of oxygen to reach the plant material during a fire as well as the efficiency of heat transfer. It is
more critical in the case of heavy fuels (Luke and McArthur, 1978) as cited by Trollope (2002).
In the case of grassy fuels, the amount compaction will inhibit flaming combustion and favour smouldering
combustion and result in producing smoke. This compaction of the fuel will retard the movement of the flame and
increase the residence time of the flame. It was observed in South Africa that older grass cuttings which started to
decompose, remained moist and consequently difficult to ignite during controlled burns (V. Fourie, pers. Comm.)
The smoulder time is the length of time the fuel, will smoulder after the flames of the head fire have passed over a
spot and is dependent on the structure and compaction of the fuel. The smoulder time of perennial Themeda
pastures in Northern Australia is around 60 seconds. Annual pastures and crop lands with little compacted material
on the surface have very short smoulder times and are relatively easy to put out. Where very large tussocks burn,
the oxygen supply to the core is inhibited, these may smoulder for several hours or even days. Such fires will be
difficult to suppress and may later start flaring up again if the wind picks up. (Cheney & Sullivan, 1997)
When controlled burns are carried out, the personnel must be mindful of these types of fire and keep an extended
lookout for flare-ups long after the mopping up is completed to avoid claims from land owners.
It can be concluded that the fuel in a fire is a complex subject but it should be understood as it not only affects the
possibility of a flashover, but knowledge of the plan species will aid in the vegetation management of the
servitudes.
7.6 Air Temperature and relative humidity

Air temperature and relative humidity are two factors that directly influence fire behaviour. Relative humidity can be
described as the relation between the amount of water vapour a unit of air contains and the maximum amount of
water vapour that the unit of air could contain at similar temperature and pressure.
Air temperature can be described as a measure of the heat content of the air (Tyson, 1964). Air temperature and
relative humidity are strongly inversely correlated through the partial pressure of water that is increased with
temperature (Tyson, 1964). Higher temperatures will thus result in lower relative humidity values and vice versa.
Raised air temperatures will dry out fuels (especially fine fuels). A direct effect of high temperatures is to influence
the temperature of the fuel and therefore the quantity of the heat energy necessary to raise it to its ignition point
(Brown & Davis, 1973). Relative humidity influences the moisture content of the fuel and influences the flammability
of the fine fuels (Luke and McArthur, 1987).
According to Heikkila et al. (1993) the following rules of thumb can be applied:
 For every 20º C decrease in air temperature, the relative humidity is doubled.
 A relative humidity of less than 30% will contribute to an increased fire danger.
As expected, fire intensities are reduced by an increase in relative humidity values. The importance of relative
humidity and air temperature as indicators of fire intensity is evident in the fact that humidity values of 5% -30% all
relate to constant high fire intensities of around 3500 kJ/s/m. During the fire season on the South African Highveld
average relative humidity values are typically around 25% creating the conditions for intense fires daily (Preston-
Whyte and Tyson, 1988). The dramatic increase in fire induced flashovers with the rise in ambient temperature
coupled with reduction in relative humidity, is also vividly demonstrated later in 10.2.
111
Veld Fire Time of day of faults
300
250
Faults 200
150
100
50
0
0 2 4 6 8 10 12 14 16 18 20 22
Hour
Figure 78 - Graph showing the time of day during which fire faults occur on the Eskom transmission network
Eskom Line faults caused by fires for the period 1993 to 2003 was analysed and shows that fire faults occur during
the hot time of the day, with a peak at 14:00. This is in spite of the fact that grass fires burn through the night during
the winter season. This phenomenon is attributed to the lower fire intensities associated with low temperatures and
higher humidity’s during the night and the converse during the day (Vosloo, 2005).
7.7 Topography
The term “topography” refers to the earth’s surface, in particular to the character or physical features of the earth’s
surface.
 The shape of the terrain will impact on the speed and direction of wind.
 Ridges and mountains are barriers to the horizontal movement of air. The wind is deflected over them
adding to the local up-slope convective winds from the surface heated by the sun.
 Ravines and gullies will form paths for the flow of air and may change the direction of the fire. In narrow
ravines, heat will dry out fuels increasing their flammability (Cheney, 1981).
 Certain plant species are therefore associated with specific habitats which are influenced by conditions
such as soil, shade, heat, moisture as well as aspect (Grant & Thomas, 2000). Consequently, the type of
fuel, fuel load and fuel moisture will all be influenced by aspect. (Teie, 2003) and should be considered with
the routing of new power lines in fire prone areas as well as during the management of the vegetation on
existing lines.
Slope and aspect

The aspect of the ground refers to the direction in which the slope faces. Northerly aspects in the southern
hemisphere are more exposed to sunlight which results in drier vegetation. Southerly mountain slopes receive less
sunlight, enabling vegetation to retain moisture for longer periods. The opposite is true for slopes in the northern
hemisphere (Pyne et al. 1996).
112
Figure 79 - The effect of slope on fire in metres per hour (Cheney & Sullivan, 1997)
As is the case with wind, slope significantly increases the forward rate of spread of a fire by increasing the
preheating of fuel immediately ahead of the fire as it moves up the slope. This occurs due to the flames assuming
an acute angle with the sloped surface and in slopes exceeding 15-20°, the flames are almost in continuous
contact with the fuel ahead of the fire. This increases the rate of spread by four times for a slope of 20° Conversely,
a fire that moves down the slope, will move at a decreased rate and at low wind velocity behave like a back fire
(Luke & McArthur, 1978).
Wind combined with rising ground slope can also dramatically affect the wind speed associated with the fire zone
as well as the rate of forward ignition of fire, particularly when preceded by heat wave conditions. As illustrated in
Figure 79 the velocity up slope can be amplified by a factor of the order of 2 for each 10 degrees of slope.
Cheney (1981) showed that in Australia a general exponential relationship is used to estimate the effect of slope on
the rate of spread of surface head fires by
𝑅 = 𝑅0 𝑒 𝑏𝑥 (84)
where R=rate of spread (m/s); R 0= rate of spread on level ground (m/s); e = exponential function; b=0.0693 an
x=angle of the slope (degrees).
This relationship should, however, not be used on slopes steeper than 30° as the surface fuel usually becomes
discontinuous on steep slopes (Cheney, 1981).
7.8 Fire barriers

Barriers are any man made or natural structure that might stop or slow down the rate of spread of a fire. It could be
a road or a river, a back burn or an old burn scar. In many cases barriers only slow down fires, which enable fire
fighters to regroup. Under severe weather conditions very few barriers will be able to stop a fire.
A popular misconception exist that a fire break is made to stop a fire. Depending on the severity of the fire, some
would jump major highways of several hundred meters wide. If coupled with spotting, firebreaks are worthless to
113
stop fires. At best, a fire break can serve as a place where a back burn can be started, and then under relatively
moderate conditions.
Experience gained in South Africa shows that the controlled burning of grass in the servitude proved to be the most
reliable barrier against grass fire induced flashovers (pers. comm. Veon Fourie, snr Supervisor Lines, Eskom).
Figure 80 - Examples of controlled burns of the grass in servitudes
These burns are carried out in in the mornings between 10:00 and 12:00 and in the late afternoon after 16;00 and
evenings when the relative humidity increases, wetting the grass and reducing the fire intensity and flame
temperatures substantially. No effect on the tower members have been reported and this practice can be accepted
as safe.
The fire was lit downwind of the line and the backing fire with its low flames allowed to burn under the lines. No
flashovers were ever experienced during these burns. It was of course most important to consider the weather both
at the time of burning as well as following the burn as smouldering fires could of course start a fire the next day is
fresh winds blow over the area. Cow dung proved to be one of the substances which proved to be a culprit in
restarted fires from smouldering material.
These burned servitudes were also used by land owners as part of their preparation for the fire season.
7.9 Conclusion
In this chapter, the behaviour of fires which is influenced by several factors was discussed. Because fire behaviour
is complex and local conditions must be considered, this chapter was merely an overview of the subject. The
reader is referred to the literature for further study.
This subject will prove to be important for planning mitigation measures as well as for the reader which is involved
in observing the progress of a fire.
114
8. Fire climate and fire weather

8.1 Introduction
The climate and weather differ from one country to the next. There are, however, very distinct common features
around the world when it comes to the climates and weather patterns that precede as well as subsist during fire. In
this chapter, these will be discussed for several countries and the common features will be highlighted.
Fire seasons occur in countries contingent on the location on the continent in question and the type of climate that
the area experiences.
Mediterranean type climate for instance, is found around the world, notably in the Mediterranean countries, South
East Australia and on the southern Cape coast in South Africa. It normally consists of wet winters and dry summers
during which hot dry weather lead to the occurrence of fires. In South Africa, the opposite is true for the interior of
this subcontinent which experiences wet summers and dry winters, during which grass fires abound.
During each of these climes, the weather varies and often during hot and windy conditions, wild fires occur. These
winds will be shown to be flowing over the warm ground, high pressure with adiabatic increase in temperature. In
this section, the climate and the resultant fire weather will be described for some countries.
8.2 Fire and wind

Wind serves to raise the intensity of a fire by providing extra oxygen to the flames. Even under moderate heat wave
conditions, the wind also dries out fuels and thus preparing it for easy ignition.
Wind is the most dynamic variable and biggest controlling factor during a fire. Wind supplies the fire with oxygen
and dries out fuel ahead of the flames which causes preheating, preparing the fuel for quick ignition. Wind speed
will increase the rate of spread while wind direction will determine the direction in which the fire front and back fire
will move. (Also see section 6.3 p 90 for fire terminology).
The acute angle of the flames caused by the wind leads to more efficient preheating of fuel ahead of the fire and
consequently to a greater rate of spread of surface head fires. Flame height, however, does not necessarily
increase with increased wind speed because of the more acute angle the flames assume, which may prevent the
ignition of aerial fuels and crown fires. (Luke & McArthur, (1978); Cheney, 1981).
Combustion rate is positively influenced by rate of oxygen supply to the fire. Doubling the wind speed will
quadruple the rate of spread of the fire. However above about 50 km/h this relationship begins to break down, and
above 80 km/h the rate of spread in grasslands reduces. This occurs because the head fire breaks up into narrow
tongues, many of which become self-extinguishing. Winds also carry burning embers downwind, which can start
new fires. This is known as spotting (Anon, 2011)
Increased wind speed result in more intense fires (Luke & McArthur, (1978); Brown & Davis, (1973); Cheney,
(1981)) Wind speeds of 0 to 3. 6 msˉ¹ increased the rate of spread of surface head fires exponentially but with no
effect on backfires (Beaufait, 1965). Brown and Davies (1973) and Luke and McArthur (1987) both found a strong
relationship between wind speed and fire intensity for fires between 0 m s⁻¹ to 13 m s⁻¹ in grasslands.
When the wind velocity exceeds 50km/h (13,9 m/s) the rate of spread in grassland tends to decrease. (Luke &
McArthur, 1978) According to Brown & Davis (1973) and Cheney (1981) this phenomenon only appears in low fuel
loads of grass.
Trollope and Potgieter (1985) and Trollope and Tainton (1986) found that under atmospheric conditions suitable for
controlled burning wind plays a significant role in fire intensity, but not a dominant one.
115
Figure 81 - Simulated time line of Victorian Kilmore-Kinglake /Murrindindi bushfires of 2009. These maps are a
modelled reconstruction of the fires using Phoneix RapidFire and photography records
(This illustration is an extract from Fire Storm 2009 First Edition –published by Committee of Parents & Friends of
Glenvale School Lilydale Victoria)
Wind direction variability is a factor with profound effects on fire. A graphic illustration of this is shown in the satellite
image time simulation of the Victorian Kilmore-Kinglake /Murrindindi bushfires of 2009. Fires were initiated during
strong gusting winds ranging up to 70km/hr. Uncontrollable wild fires progressed on narrow fronts for several
hours.
There was then a sudden frontal system change with a resultant wind shift of 90 degrees. The consequence was
that a severe narrow fire front quickly became a very broad uncontrollable fire front of immense proportions. It
resulted in loss of 173 lives, 2200 houses and 352,000 hectares of land burn out, and most utility infrastructure
destroyed.
116
The significant association between wind and the occurrence of fire induced flashovers is discussed in 10.2, p135
and shown in Figure 99.
One important aspect of wind during fire conditions, is that of offshore wind conditions. A recurring theme that can
be seen on synoptic charts is that of wind blowing from the mountainous interior to the coast. The air is heated up
as it blows over land. As the air descends, it further heats up at the dry adiabatic lapse rate of 9.8°C/km (or
3°C/1000ft).
These offshore winds have different names and in South Africa it is referred to as Berg wind (from Afrikaans berg
"mountain" + wind "wind", i.e. mountain wind). They occur in winter in South Africa. Berg winds blow from the
Kalahari High to a coastal low-pressure area.
Many other names exist for these warm winds (föhn or foehn, Chinook or katabatic wind). A katabatic wind, from
the Greek word κατάβᾰσις katabasis meaning "descending", is the technical name for a drainage wind, a wind that
carries high density air from a higher elevation down a slope under the force of gravity. Such winds are sometimes
also called fall winds.
Katabatic are mainly winds in the order of 10 knots (18 km/h) or less, but sometimes they can rush down elevated
slopes at hurricane speeds. Examples of true katabatic winds include the bora (or bura) in the Adriatic, the
Bohemian Wind or Böhmwind in the Ore Mountains, the Santa Ana in southern California, and the Oroshi in Japan.
As this subject can become very technical, it is sufficient to conclude that fires are fanned by this dry, descending
air, which is adiabatically heated and which dries the fuel, ready for ignition. For the purpose of this document, we
will refer to these as a berg wind.
8.3 South Africa

The African continent suffers extensively from grass and brush fires. Fire is generally used for burning of vegetation
for agricultural reasons. In many cases negligence with fire lead to runaway fires and some folklore has it that a fire
season with many fires lead to good rains.
Most of the fires in Southern Africa are grass fires which are not comparable with the intense fires that are
experienced in Australia and California, USA. These fires in South Africa, however, are intense enough, given the
correct atmospheric conditions, to cause flashovers from the live conductor to ground. Very intense forest fires
occur at times and those can be devastating, as is shown in Figure 1 on p12. Brush fires comparable to the
Californian fires occur in the Western Cape areas. The burning of sugar cane prior to harvesting is a common
practice which invariably leads to intense fires and power outages.
In the case of Eskom, approximately 20% of line faults on the Transmission system are caused by fires (Vosloo,
2005). Fire faults also affect equipment such as breakers and transformers, which can lead to premature ageing of
these components. The consequence of these fire faults is difficult to quantify in monetary terms.
Distinct fire seasons are experienced. The interior of the Southern African continent has summer rainfall with the
fire season occurring during the winter months (April to October). The Western Cape region has a Mediterranean
climate with winter rains and a dry summer when fires occur (November to March).
Vegetation types that support fires vary from one biome to the next. The interior of South Africa consists of open
grass lands. In the north Savannah occurs which consist of grass, shrubs and trees. The Western Cape has a floral
biome known as “Fynbos” and which is highly regarded by botanists because of the wide range of species that
occur here. This biome is also prone to fires and leads to a number of fire faults during the fire season.
Tyson and Preston-Whyte (2000) refer to ten southern African weather types. Of these coastal low and Berg winds
(hot, off shore winds) as well as cold fronts are of particular concern during the fire season. Cold fronts are
important determinants of the weather in South Africa. They occur most frequently during winter when the
amplitude of westerly disturbances is greatest and are associated with the invasion of cold air from the south
(Tyson and Preston-Whyte, 2000).
117
Figure 82 - The Berg wind conditions ahead of the cold front lead a rapid increase in the occurrence and spread of fires
[Nigrini 2004]
During these events the demand for electricity rises in response to the cold weather. As the cold front approaches,
the continent, a coastal low develops on the south-east coast. At the same time the Atlantic high ridges in over the
sub-continent from the West. The result is that Berg wind conditions develop over the interior as shown below. The
descending air heats adiabatically and creates hot, dry, and windy conditions, shown as red arrows. Under these
circumstances, fuels dry out and any fire is fanned by the high winds. It becomes extremely difficult to control fire.
During the winter the demand for electricity is naturally higher than in summer, resulting in a stressed transmission
system that can ill afford line faults.
Figure 83 - Dangerous fire weather conditions in the Cape associated with a high-pressure system over the Indian
Ocean
118
The situation in the Western Cape is different to that of the interior. This area receives most of its rain during winter
with hot dry summers. During summer Berg winds, these conditions occur when the Indian High causes offshore
north easterly winds over the Western Cape. These winds are also hot and dry and cause extreme fire danger
conditions. (shown by red arrows)
Fire event on 765kV lines in South Africa

A significant fire outage took place in South Africa in 2002 (Vosloo, (2002) (1) & (2). It was important because of the
following:
 It was a fire induced outage on two 765kV lines with ground clearance of 15m,
 It was caused by a fire burning in very sparse grass,
 Two fire faults occurred at almost the same time, although they were 295km apart,
 It was the first fire outage in Eskom where the importance of weather was highlighted
On Monday 29 July 2002, two separate veld fires caused fire flashovers on the Alpha-Beta 1 and 2 lines. These
two lines carry the bulk of the electricity supply to the Western Cape load centre. The first fault occurred at 11:48
on the Alpha-Beta No.2 765kV line in central South Africa.
The second fault, at 12:29, was reported on the Alpha-Beta No. 1 765kV, 295km from the first fault. During the field
investigation, a flash mark was found on the conductor between towers 34 and 35. The Fire Chief was at the scene
of the fire when it passed under the line and heard a loud noise as the line flashed over. He reported a lot of smoke
and estimated flame lengths to be about ten meters.
These two faults interrupted the main supply to the Cape Town area. A total load of 1593MW was lost for 10
minutes during the incident.
Three aspects of these incidents were of significance. In the first instance the fuel loads at both fires were
extremely low. This is evident from Figure 84 and Figure 85.
Figure 84 - Picture of the low fuel load at the site of the Virginia flashover on a 765kV line in South Africa
119
Figure 85 - Picture of the low fuel load at the site of the Standerton flashover on a 765kV line in South Africa
The second was that the fires in these two cases were of sufficient intensity to cause a flashover on the mid-span
gaps of 15m used on these 765kV lines.
The third aspect was that these two faults occurred under similar, extreme Fire Danger conditions.
This emphasizes the fact that weather conditions play a major role, not only with the onset and spread of fires, but
also in the ability of these fires to cause flashovers on power lines. In Figure 86 the graph depicts the FDI (Fire
Danger Index) for July 2002. The passage of a cold front on the 19th of July 2002 is associated with a very low
FDI.
During the ensuing week, pre-frontal conditions begin to prevail with an associated rise in FDI as well as fire faults
(see Figure 86). When this combination of high temperature and wind and the low RH reached a peak on the 29th
of July, no fewer than 36 fire faults were recorded on the transmission system, including the 765kV lines.
120
Fire Fault correlation with F.D.I. July 2002
80 40
70 35
No. of fire faults

60 30
50 25
F.D.I.
40 20
30 15
20 10
10 5
0 0
11
13
15
17
19
21
23
25
27
29
31
1
3
5
7
9
Day of the month
Fire Faults F.D.I
Figure 86 - As the graph indicates, the FDI and line faults reached a maximum value on the 29 th of July 2002
Vosloo, (2002)
This incident brought the realization to Eskom of the importance effect that weather conditions have on fires and
subsequently on the reliability of the network. As a result, Eskom introduced regular fire weather forecasts.
8.4 Australia
Bushfires are an inevitable occurrence in Australia, fuelled by over 800 endemic species dominated by the fire-
adapted eucalypts and grasslands. The latter covers nearly 75% of the country.
Fire is most common over the tropical savannahs of the north, where certain parts burn on an annual basis. It is
estimated that more than 2.5 Mha is burnt annually by wildfire with much more in extreme years (Cheney &
Sullivan, 1997).
The biggest threat to people and property by devastating wild fires, occur in the southeast, where the majority of
the population resides. This is also where the majority of overhead power lines are situated.
A unique factor in these fires of the southeast is the climate of the region. The south-eastern region experiences
what is known as a Mediterranean climate with its hot, dry summers and mild, wet winters. Whilst the winter and
spring rains allow fuel to grow, the dry summers dries the fuel and augments the fire risk. This risk is further
aggravated by periodic droughts that occur as a part of natural inter-annual climate variability.
Figure 87 shows a map of Australia showing the peak fire season for different areas. was compiled by Luke and
McArthur (1978) as cited in (Lucas et al., (2007), Anon (2009)).
121
Figure 87 - Map of peak fire seasons across Australia - From Lucas et al., (2007), Anon (2009)
Figure 88 - Synoptic chart for 16 February 1983 - From Lucas et al., (2007), Anon (2009)
122
Heavy arrows depict general wind direction ahead of a cold front. Fire symbols denote approximate locations of
bushfires observed on the day fanned by a berg wind.
Figure 89 - Synoptic chart for 7 January 1994. Fires on the East coast are fanned by strong berg winds, denoted by the
thick arrow - From Lucas et al., (2007), Anon (2009)
Figure 90 - The off-shore wind in this synoptic chart generates fire conditions in West Australia -
From Lucas et al., (2007), Anon (2009)
For most of Australia's east coast, the fire season extends from spring to midsummer. The greatest danger occurs
after the dry winter/spring period, before the onset of the rainy weather common in summer, and when deep low-
123
pressure systems near Tasmania bring strong and dry westerly winds to the coast, as occurred in the major New
South Wales fires in January 1994 (Figure 89) as well as during the 2001-02 summer.
Climate change projections indicate that south eastern Australia is likely to become hotter and drier in future with
the number of ‘very high’ and ‘extreme’ fire danger days increasing between 4 and 25% by 2020 and 15 to 70% by
2050. (Lucas et al, 2007)
Comparison of these times of drought with the Forest Fire Danger Index (FFDI) time series shows that there is a
strong correlation, with the ‘high’ periods of FFDI (or longer fire season length or larger values of the 90th-
percentile FFDI) corresponding to drought episodes. These droughts, in turn, are often associated with El Niño
events (Fig. 31). However, the relationships between ENSO and rainfall (and hence droughts) are not necessarily
related in some simple, linear way (e.g. Power et al 1998). Droughts can often last longer than a single El Niño
event. Even a weak El Niño event (e.g. 2002-03) can have a large impact on rainfall and vice versa (1997- 98).
(Lucas, et al., 2007).
The Third Dimension - the vertical profile of a severe fire day

A layer of warm air present over a layer of cold air is called a temperature inversion. Inversions are common during
the night and early morning when cool night air collects close to the ground (Figure 91).
This arrangement of air is stable because cold air near the ground is heavy and tends to stay near the ground. The
warm air above the inversion is lighter and tends to stay above the inversion.
There are often strong winds in the warm air above inversions, but while the inversion lasts these winds cannot
come down to ground level (Figure 91)
As the sun heats the ground during the day, the inversion weakens and strong winds may begin to blow near the
ground. This is one reason why many bushfires burn more fiercely in the afternoon, as happened with the fires in
Hobart in 1967 (Anon, 2009)
Figure 91 - Vertical profile during severe fire weather (Anon, 2009)
8.5 The United States of America

(This information was taken from Wikipedia. https://en.wikipedia.org/wiki/Santa_Ana_winds accessed on 26 August
2015)
The theme of warm, dry winds blowing from the interior is also seen in the United States of America. The Santa
Ana winds are strong, extremely dry down-slope winds that originate inland and affect coastal Southern California
and northern Baja California. Santa Ana winds blow mostly in autumn and winter but can arise at other times of the
year also. These winds, which can easily exceed 40 miles per hour (18 m/s), are warm and dry and can severely
exacerbate brush or forest fires, especially under drought conditions
They can range from hot to cold, depending on the prevailing temperatures in the source regions - the Great Basin
and upper Mojave Desert. The winds are known especially for the hot dry weather (often the hottest of the year)
that they bring in the fall and are infamous for fanning regional wildfires. For these reasons, they are sometimes
known as the "devil winds" across Southern California. (Masters, 2012).
124
The air heats up from adiabatic heating during its descent. While the air has already been dried by orographic lift
before reaching the Great Basin as well as by subsidence from the upper atmosphere, the relative humidity of the
air is further decreased as it descends from the high desert toward the coast, often falling below 10 percent.
The images above illustrate a characteristic high-pressure area centred over the Great Basin, and upper Mojave
Desert with the clockwise anticyclone wind flow out of the high-pressure center. This gives rise to a Santa Ana wind
event as the air mass flows through the passes and canyons of southern California, manifesting as a dry north
easterly wind (Fovell, 2012).
Figure 92 - The Santa Ana wind in the South West United States (source
http://www.csun.edu/science/geoscience/meteorology/santa-ana-winds2/index.html
125
Figure 93 - Satellite image showing fires fanned by the SantaAna winds (Source:
http://www.csun.edu/science/geoscience/meteorology/santa-ana-winds2/index.html)
It is often said that the air is heated and dried as it passes through the Mojave and Sonoran deserts, but according
to meteorologists this is a popular misconception. The Santa Ana winds usually form during autumn and early
spring when the surface air in the elevated regions of the Great Basin and Mojave Desert (the "high desert")
becomes cool or even cold (although they may form at virtually any time of year).
As the cool, dense air from the desert blows out toward the coast, it tends to channel down the valleys and
canyons and through the major mountain passes. Gusts can attain hurricane force at times. As it descends, the air
not only becomes drier, but also warms adiabatically by compression.
The southern California coastal region gets some of its hottest weather of the year during autumn while Santa Ana
winds are blowing. During Santa Ana conditions, it is typically hotter along the coast than in the deserts (Anon,
2011).
8.6 Mediterranean countries

Several countries in the Mediterranean experience dangerous fire weather at timed. These include Portugal, Spain
Italy, Greece and others. These countries all experience wet cool winters with hot, dry and windy conditions during
summer.
The southern parts of Croatia (Dalmatia, Primorje and Istria and Archipelago) are situated exactly in the middle of
northern Mediterranean coast. The climate is similar to other Mediterranean coasts including North-African and
may be considered as close to sub-tropical. In summer air temperatures often exceed 40 ° C, variations of air
humidity are extreme and sea salinity is among the highest in the world (up to 40 ‰).
Strong winds are experienced at times, with northerly winds named “bura” and southerly winds known as “jugo” or
“sirrocco”). As a result, Dalmatia, (the long and narrow southern strip of Croatia) experiences wildfires during the
dry summer months and which turn into disastrous wildfires (Sekso-Telento et al 2001).
8.7 Conclusion
In this chapter, the fire climates and fire weather of several fire-prone countries around the world was briefly
discussed. There is a remarkable similarity in countries with Mediterranean climates; cool wet winters with vigorous
plant growth followed by dry hot summers. Even in cases where wet summers are followed by dry winters, fires
abound when the weather patterns result in hot descending air resulting in winds that both dry the fuels and fans
the fires. Signs of these weather patterns, should warn the utility of impending fires.
126
9. Fire danger rating systems

9.1 Overview
The potential for the occurrence and development of fires is dependent upon the interaction of fuels with a number
of weather elements that vary over long and short time-scales. Several methods have been developed over the last
40 years to combine information on weather and fuels in a fire danger index.
Fire danger ratings describe conditions which reflect the potential for a fire to ignite, spread as well as indicate the
difficulty of suppression. Fire danger Index is the result of both relatively constant and variable factors affecting the
chance of a fire igniting in a particular fuel, its rate of spread, intensity and difficulty of suppression and control, as
well as its potential to do damage (Chandler et al. 1983). This is achieved through various combinations of
temperature, relative humidity, wind speed and drought effects (McRae, 1999)
Fire behaviour on the other hand deals with an existing fire in a given time and space (See 7 p93). Fire behaviour
describes the movement (rate of area increase), intensity (flame length), and indicators of rapid combustion
(spotting, crowning, and fire whirls) of that fire. It is expressed as real time or predicted condition for on-going fires
(Everson et al. 1985).
Fire danger indices are also used to declare fire bans, allocate resources, and inform public awareness of fires in
addition to assessing fire behaviour potential in an operational setting.
The concepts of fire danger and fire behaviour could be confused due to the similarity of the factors affecting these
concepts. Factors such as fire fuels, weather, topography, and intensity are part of fire danger rating and are very
similar to those that affect fire behaviour predictions. The principle difference is that fire danger is a broad scale
assessment while fire behaviour is site specific (Everson et al. 1988).
The three main fire danger rating systems of the world originated in the United States, Canada, and Australia. The
US and Canada developed complex models that required input data that is not always available. These models
also have a strong focus on forest fires. The Australian McArthur indices were based on empirical models
developed for forest and grassland conditions.
9.2 US National fire danger rating system

In 1968, the United States Forest Service (USFS) initiated the development of a National Fire Danger Rating
System (NFDRS) based on Rothermel’s fire behaviour model (Rothermel, 1972).
The US NFDRS (Figure 94) is a complex model that requires very specific inputs in order to perform on a national
level. The complexity of the model originates from the diversity of the vegetation types and climatic conditions
across the United States.
The fuel models cover a wide range of fuel types and include live fuel moisture content. The outputs from the
model include a spread component (SC), an energy release component (ERC) and burning index (BI).
More recent development of the system has incorporated ignition components for both human-caused and
lightning-caused ignitions, so that the fire load or fire severity index can be estimated.
127
Figure 94 - US National Fire Danger Rating System (Willis et al. 2001)
Six indices in the form a spread, energy release, burning, human ignition, lightning ignition and fire load index
(Willis et al. 2001).
The system consists of three main components:
 Observations which relate to fuel models and fuel types, topography and weather parameters in the form of
relative humidity, air temperature, wind speed and rainfall duration.
 The Fuel Moisture component is derived from combinations of the observations discussed above to
produce moisture contents estimations for different fuels (grass swards to tree stumps)
 The Fire behaviour component produce a fire severity index which consists of a Spread Component (SC),
a Burning Index (BI) produced from the Energy Release Component (ERC) and a Lightning and human
Component (Figure 94).
9.3 Australian fire danger systems
McArthur forest fire danger index

McArthur, (1958), started the development of the McArthur Forest Fire Danger Index (FFDI) in Australia. The
McArthur FFDI is an empirical model of fire behaviour in open forests, but the fire danger index is applied broadly
to all forest types and scrublands.
The index is classified into five classes ranging from Low, Moderate, High, and Very High to Extreme. The structure
of the McArthur FFDI is shown in Figure 95. The inputs required to run the model are air temperature, relative
humidity, wind speed, rainfall, time since last rainfall, and the Keetch-Byram Drought Index (KBDI) (Keetch and
Byram, 1968). The KBDI combined with the Days since Last Rainfall produce the Drought Factor sub-model which
feeds again in to the Rate of Spread sub-model when combined with wind speed.
128
Figure 95 - McArthur Forest Fire Danger Index (Willis et al. 2001)
The temperature and relative humidity inputs are used to produce the Fuel Moisture. The Rate of Spread and Fuel
Moisture is combined to produce the Suppression Difficulty sub-model which determines the McArthur Fire Danger
Index value.
McArthur Grassland fire danger index (MK 4)

McArthur’s Grassland Fire Danger Index (McArthur, 1958) as illustrated in Figure 96, is similar to the FFDI, except
that the fine fuel availability calculated using the Drought Index and Drought Factor components is calculated using
the degree of curing (Griffiths, 1999).
Curing is defined as the proportional weight of dead grass to live grass. A curing factor of 80% thus means that, of
the total grass biomass, 80% of the grass is dead. The moisture content of green grass is affected by physiological
processes and is generally above the moisture of extinction, while the dead grass moisture content, is affected by
local atmospheric conditions such as air temperature and relative humidity. The fuel moisture content, is therefore
determined by the combination of the degree of curing and environmental conditions. The fine fuel moisture content
combined with the wind speed, are the main determinants of fire spread rate (Figure 96).
The MK 4 model was derived for the predicted spread rate of a fire in lightly grazed, temperate perennial grassland
on flat ground. Fire spread rates would increase where the topography is generally hilly and where the grass is tall,
and, decrease where grasslands are heavily grazed or recently burnt.
129
Figure 96 - McArthur Grassland Fire Danger Index (McArthur, 1958)
The CSIRO fire danger rating system

The conditions that affect relative fire danger do not always affect the rate of spread in the same way. Fire danger
and difficulty of suppression are related exponentially to wind speed. As wind speed increases, the difficulty of
putting out a fire rises at an ever-increasing rate. The rate of forward spread, on the other hand, has a near linear
relationship to wind speed. Thus, while wind speed is an important factor in predicting both fire spread and fire
danger, fire spread cannot be directly linked to a fire danger index.
The CSIRO Fire Danger Rating System is based on the McArthur Mk 4, but with two important differences:
 The rate of spread removed. This must be calculated using the CSIRO Fire Spread Prediction System.
 The maximum value of 100 of the McArthur system can also be exceeded in the CSIRO system.
In McArthur’s original system an index of 100 was representative of the worst possible fire weather conditions likely
to be expected in Australia. This has been exceeded on several occasions since 1966 (Cheney & Sullivan, 1997)
Australian weather office system

Following devastating fires in Australia in 2003, 2005, and February 2009, the Australian Bureau of Meteorology
incorporated changes to the Fire Danger Ratings used by fire agencies in its weather forecasts and warnings from
October 2009.
These ratings also provide an indicator of ambient conditions that, should fires occur, are favourable to the
occurrence of fire trips to overhead lines. Fires that occur in less than Very High would seem, from operational
experience, to be generally slow burning and controlled, which combined with low vegetation growth decreases fire
trip to low risk.
Fire danger ratings of Very high and above, however, would seem to pose a high-risk condition for electrical
flashover, should these fires progress across or under overhead line easements.
130
Table 18 - Forest Fire Danger Index (FFDI) and the Grass Fire Danger Index (GFDI)
The critical elements that drive the fire danger ratings and risk of wildfires under the higher rated conditions are low
rainfall or droughts preceding the fire season and prolonged heat wave conditions.
The previous Extreme category has been divided into three levels - Severe, Extreme, and Catastrophic (Code
Red). These new levels are based on the Forest Fire Danger Index (FFDI) and the Grass Fire Danger Index (GFDI)
as indicated in Table 18.
Fires that occur in less than Very High would seem from operational experience to be generally slow burning and
controlled and provided vegetation is low growth then risk of fire trip is low. Fire danger ratings of Very High and
above, however, would seem to pose a high-risk condition for electrical flash over should these fires progress
across or under overhead line easements.
The critical elements that drive the fire danger ratings and risk of wildfires under the higher rated conditions are low
rainfall or droughts preceding the fire season and prolonged heat wave conditions.
These ratings also provide an indicator of ambient conditions that should fires occur may provide a high potential
risk of fire trips to overhead lines.
Prolonged heat wave conditions can also be associated with strong winds. These combined conditions provide the
critical situation that promotes the more extreme fire events.
Droughts are a significant feature of the climate of some countries and extended dry spells significantly lower the
moisture content of fire fuels and create a very high fire risk, particularly if the dry spell follows a period of good rain
that has encouraged lush growth and increased fuel load.
9.4 Southern African fire danger index systems

As southern Africa seldom experience the devastating fires in natural vegetation as in other countries with
extensive forest, the introduction of commercial forests here highlighted the need for the prediction of dangerous
fire weather. As residential development in areas such as the Cape Fynbos areas with its intense fires, increased,
this need further increased. During the 1998 legislation was introduced which enacted inter alia the introduction of
a FDI system (NATIONAL VELD AND FOREST FIRE ACT, Act 101 of 1998).
131
Lowveld Fire Danger Index

The current fire danger rating index used by most in South Africa is an adaptation of a Fire Hazard Index
developed by Michael Laing in Zimbabwe (then Rhodesia) in 1968 (Laing 1978).
The basic model uses the same inputs as the McArthur models, which are scaled to produce a simple model that
can calculate numbers easily without needing any complex calculations (McArthur, 1958). It is often referred to as
the ‘Lowveld Fire Danger System’ since this is where it has been most widely used. However, the process by which
the system was developed, and the assumptions that were made, were never documented (M. Laing, pers. comm.
As cited In Frost, 2010).
The Lowveld system consists of a Burning Index BI which is derived from air temperature and relative humidity.
The equation for the BI calculation is:
𝐵𝐼 = 32.8937 + 1.07188𝑇 − 0.394077𝑅𝐻 (85)
where air temperature 𝑇 is measured in degrees Celsius and relative humidity 𝑅𝐻 in %. (Frost, 2010)
The calculation of the Burning Index (BI) can also be performed using a nomogram (a two-dimensional diagram
designed to allow the approximate graphical computation of a function) as demonstrated in Figure 97.
The scale to the left represents air temperatures and the scale to the right relative humidity percentages. In order to
calculate the BI, a relevant air temperature (e.g. 15° C) and relative humidity (e.g. 30%) and linked on the
nonogram with a straight line. The line will cross the centre scale corresponding to a BI value (35%) (Figure 97)
Trollope (2005) introduced a modified version of the Lowveld model by adding a wind and curing factor to the
model.
The new equation is as follows:
𝐹𝐷𝐼 = (32.8937 + (1.07188 ∗ 𝑇) − (0.394077 ∗ 𝑅𝐻)) + (1.1782 + 0.7212 ∗ 𝑊𝑆) ∗ 𝐶 (86)
With air temperature T in degrees Celsius, relative humidity RH in %, wind speed WS in km/h and grass curing C as
a percentage. The Lowveld model is currently being used by the Fire Fighting Associations (FFA) in the
Mpumalanga and KwaZulu-Natal provinces of South Africa and can be described as the unofficial fire danger rating
system of that country (Frost, 2010).
132
Figure 97 - Nomogram of the Lowveld model Burning Index (Willis, et al. 2001)
9.5 Conclusion
The various methods of predicting dangerous fire weather was discussed in this chapter. These approaches were
aimed primarily at fire fighters. Although this may prove to be useful to the electrical utility, they all have some
shortcomings in this application. In 10 a refinement to the conventional fire danger ratings will be discussed. This
method enables the user to predict conditions which would lead to fire induced flashovers instead of just dangerous
fire conditions.
133
134
10. Predicting fire induced flashovers

10.1 Introduction
The awareness of the danger of fire has been improved around the world to a large extent by news reports of these
conflagrations. In many countries, weather forecasts will warn of impending dangerous fire weather. The weather
systems which point to dangerous fire weather were discussed above and those can readily be identified by the
utility using standard weather forecasts.
In South Africa, forecasts of fire weather by the national weather service are legislated and are broadcast widely on
radio and television. One shortcoming at present is that these forecasts are based on weather parameters alone
and do not consider the fuels which will support the fire. That leads to fire forecasts in areas with extremely sparse
vegetation or where green fuels are dominant and the real danger of fires are reduced considerably.
During the development of South Africa’s Active Fire Information System (AFIS), the idea was born to develop a
Fire Induced Flashover Prediction Index (FIFPI) which will warn Eskom of weather conditions with a high likelihood
of fire flashovers, as a refinement to the normal fire danger index discussed in 9. This was the work of one of the
corresponding members of WG B2.45 and will be discussed below.
10.2 Fire Induced Flashover Probability Index (FIFPI)

The link between atmospheric conditions and fire intensities and fire induced line flashovers was made by
numerous authors and is discussed in 8. Whilst Fire danger indices were mainly developed for the fire suppression
community to describe or predict fire behaviour parameters, no system existed to link actual weather parameters
with the probability of a fire under a power line to cause a flashover.
A study was undertaken by Frost (2011) to investigate the relationship between atmospheric conditions that existed
at the position where actual fire flashovers occurred. The aim of this study was to develop a flashover probability
index based on the conditions prevailing during a fire.
The data sets that were used included:
 MODIS active fire satellite data

 Eskom Transmission line location data
 Historic fire induced flashover point data
 Numerical weather forecast data
 Vegetation Condition Index data
 Digital Elevation Model (DEM) data.
The predictor variables used in this study were air temperature, relative humidity, wind speed, wind direction
altitude, and the vegetation condition index (amount of curing of the grass as derived from satellite data). Fire
Danger values for three variations of the South African “Lowveld” fire danger index as well as the Mc Arthur
Grassland Index was also used as predictor variables.
The main inputs for the development of the target variables were the MODIS active fire data, a shape file of the
Eskom Transmission grid and the historical fire flashover point data.
If MODIS detected a fire within a radius of 3km of the position of a fire fault that occurred on the same day, it would
be classified as a “true” and it was accepted as the cause of that flashover. MODIS fires falling outside of the 3km
radius from the flashover point or that occurred on a different day were classified as “false”.
Frost then performed both a decision tree and a logistic regression analysis on the target and predictor variables to
determine which model would predict the likelihood of a fire induced flashover best and the model with the lowest
error. The FIFPI was able to correctly predict 98.9% of the flashovers in the validation data set using the LR model,
while correctly predicting 96% of the flashovers using the decision tree model. Both approaches showed that the
combination of air temperature, relative humidity, wind speed and wind direction provide the most accurate
predictors for fire flashovers.
135
Based on these results, Frost then used the logistic regression to derive a Fire Induced Flashover Potential Index.
The following expression gives it:
𝟏
𝑭𝑰𝑭𝑷𝑰 = ( 87 )
𝟏 + 𝒆−(𝟏𝟗𝟔+𝟔.𝟐𝟒𝑹𝑯+𝟎.𝟎𝟓𝟎𝟓𝑻+𝟎.𝟓𝟓𝟕𝑾𝑫−𝟎.𝟗𝟎𝟗𝑾𝑺)
Where 𝑅𝐻 is the relative humidity in %, T the air temperature in °C, WD the wind direction in degrees from North
and WS the wind speed in m/s.
Figure 98 - Development of the target variable (Frost, 2010)
The standard error and Chi squared probability indicate a strong relationship (p˂0.00001) between the predictor
variables and the “true” flashovers, indicating that all variables contribute to the outcome, while a low standard
error was calculated for each of the predictor parameters.
The rather surprising importance of wind direction and temperature increase on the flashover probability was
further investigated. Flashover probabilities were calculated for north-easterly (300°) and south-easterly (150°)
winds, while keeping a constant RH of 16% and wind speed at 5m/s which is shown in Figure 99
136
Figure 99 - Temperature simulations in relation to different wind speeds
Figure 82 shows that a north-westerly flow is normally associated with warm, dry and windy conditions associated
with pre-frontal bergwinds. The relationship between these conditions ahead of a cold front with high FDI and the
increased occurrence of fire induced faults is also shown in Figure 86.
Strong winds and low humidity have a direct bearing not only on the intensity of the fire but also on the onset of
runaway fires, fuelled by dry vegetation. This is further discussed in 8.
In a study conducted by Xu & Schoenberg (2007) where they correlated wind direction with burned area, it was
found that wind direction provided key prediction information that other variables such as relative humidity and
temperature could not.
The prediction capability of FIFPI against that of the fire danger indices mentioned above, was carried out next.
Using the decision tree as well as the logistic regression models, a comparison of the relative importance of each
model’s ability to predict flashovers was done. The results show that FIFPI had a 100% relative importance,
followed by the Australian Mc Arthur model at 85%. The South African Lowveld systems scored lowest with values
of below 56%.
Finally, a multiple regression analysis was performed to determine the correlation between the different models as
well as the relationship with the flashover (target) data.
FIFPI showed a 0.98 correlation with the flashover data followed by a 0.80 correlation for the Mc Arthur model
while the Lowveld systems had a correlation of 0.64 for two of them and 0.38 for the other. The correlation between
FIFPI and the Mc Arthur system was good at 0.84.
The study found that wind speed and wind direction proved to be the most important variables in predicting fire
induced flashovers. It showed a sharp increase in flashover probability as soon as the wind speed reaches 4m/s.
137
Figure 100 - Wind speed simulations in relation to different wind directions
Figure 101 - Relative humidity simulation for different wind directions
In summary, this study has shown that the following atmospheric conditions lead to a rapid increase in the ability of
a fire to cause a flashover:
 North-westerly, pre-frontal berg winds

 Ambient temperature of more than 24° C (with a rapid increase above32° C)
 Relative humidity of below 23% (50% probability at18% and 100% probability below 15%)
 Wind speed of more than 4m/s (100% probability above 5m/s)
138
Figure 102 - A graph showing the Flashover Probability Index against air temperature
This work shows that the prediction of atmospheric conditions that will cause flashover during a fire is feasible with
the use of satellite technology and global weather data. When used in conjunction with a fire tracking system, the
utility can continually be aware of the risk that a fire may pose to an overhead line.
Wind direction
Figure 103 illustrates the change in flashover probability when wind speed increases for winds from a 300-degree
(north-westerly winds), and 150-degree (south-easterly winds) direction with a constant air temperature value of
299 K (26º C) and a relative humidity of 16%. A 150-degree wind had little effect on the flashover probability with a
maximum probability of 60% at wind speeds of 6 m/s, while a 300-degree wind caused a sharp increase in
flashover probabilities from 4 m/s with a 100% probability at 5 m/s. The strong relationship between wind speed
and wind direction lie in the occurrence of the prefrontal conditions during the winter. Wind speeds will very seldom
go above 5 m/s on the Highveld during winter in the absence of pre-frontal conditions.
Figure 104 illustrates the change in flashover probability when wind direction increases for wind speeds of 4.5 m/s
and 5 m/s with a constant air temperature value of 299 K (26º C) and a relative humidity of 16%. Wind speeds of
4.5 m/s resulted in an increase of flashover probability with a wind direction of 250 degrees, and a 50% probability
increase with wind directions near 250 degrees.
The 5 m/s winds increased the flashover probability from 250 degrees with a 50% probability close to 350 degrees.
The relative difference (50º) between the wind directions for 4.5 m/s and 5 m/s wind speeds is not fully understood.
The main aim of the wind direction/wind speed test was to illustrate the wind direction zone where most flashovers
occur. The results clearly show (Figure 104) that winds blowing from a 200-degree to 350-degree angle have a
much higher potential to increase the risk of fire-induced flashovers than those winds blowing from a 0-degree
to150 degree direction emphasising the role of pre-frontal conditions in flashover events. (Vosloo, 2005).
139
Figure 103 - Wind speed simulations in relation to different wind directions (Frost, 2011)
Figure 104 - Wind direction simulations in relation to different wind speeds
140
10.3 Conclusion
The work done on predicting weather parameters which closely correlates with fire flashovers rather than relying on
fire danger index is a development that will prove very helpful to staff at National Control Centres. This will enable
them to judge when conditions approach the critical values on page 138.
This work was done using grassy fuels prevalent in the interior of South Africa, however, this exercise may be
carried out for any fuel which exist under the lines of the utility. The real-time weather data that is provided with the
AFIS system (see Figure 110, p148) will enhance the controller’s ability to manage fire threats to the transmission
system.
141
142
11. Fire tracking systems

11.1 Introduction
The science of remote sensing, or observing the Earth from space, started just over 60 years ago with the
launching of Sputnik 1 by the former Soviet Union on 4 October 1957 (Anon, 2009a). This was the World’s first
earth orbiting artificial satellite.
The first satellite images of the earth came from the American weather satellite TIROS1 on 1 April 1960 (Anon,
2009b). Seventeen years later, the first real-time satellite images were received on Earth from the American KH-11
satellite system (Anon, 2009c)
Today, satellites are used for a variety of purposes from navigation, weather forecasting, real-time tracking, radio
internet communications and even TV broadcasts. Satellites images have numerous uses namely: agriculture,
geology, forestry, regional planning, education, conservation and national intelligence.
11.2 Remote sensing: monitoring fires from Space

Numerous studies have demonstrated the ability to use remote sensing techniques for monitoring the Earth’s
surface and for generating information on vegetation fire (Davies et al. 2008). Data from sensors, such as the
Moderate Resolution Imaging Spectroradiometer (MODIS) on-board NASA’s Terra and Aqua polar-orbiting
satellites, are used to generate products of the land surface for global monitoring, modelling and assessments. The
approach is designed to transform raw data into information while at the same time minimizing the data processing
burden on users (Davies et al. 2008).
The MODIS products are relatively easy to access, but some expertise is required to order and use them. File sizes
can be prohibitively large (typically 50 MB) especially when products are accessed through low internet bandwidth.,
MODIS fire products need to be created as soon as possible after satellite data is acquired to be of value for
operational fire management. Direct acquisition satellite data via a dedicated reception dish during an overpass
overcomes this problem.
This direct broadcast capability enables near-real-time processing of MODIS data. For South Africa, direct
broadcast MODIS data was initially processed in at the Satellite Applications Centre (SAC, west of Pretoria)) but
more recently it was received via the dedicated receiving station mounted on the roof of a building on the Council
for Scientific and Industrial Research (CSIR) campus in Pretoria.
In 1999, the Terra polar-orbiting satellite was launched, with the Aqua satellite following in 2002. Both have a
MODIS sensor on board that can detect fires (Justice et al. 2006) with high precision (fires as small as 50 m x 50 m)
four times a day. The MODIS active fire detection algorithm is applied to day and night-time data providing
approximately four active fire products per day for most land surfaces. The time resolution of MODIS is suitable for
monitoring global trends but could not provide continuous monitoring for operational fire management. As Eskom
was also interested in more regular feeds, it asked CSIR to consider other possible sources of satellite data.
Observations from Meteosat Second Generation (MSG), a geostationary weather satellite from Eumetsat was
proposed. It provides coarser (fires as small as 350 m x 350 m) and less sensitive fire detection information but at
15-minute intervals. Continuous fire monitoring was achieved by using data from the Spinning Enhanced Visible
and Infrared Imager (SEVIRI) sensor, on-board MSG. This approach had been demonstrated in the Mediterranean
but had at the time not been applied in South Africa (Davies et al. 2008).
143
Figure 105 - One of the authors at a dedicated MODIS receiving station
The high temporal frequency of MSG is offset by a relatively poor spatial resolution (4 km compared to 1 km spatial
resolution of MODIS). The success of this approach will be discussed below.
11.3 Background to the Advanced Fire Information System (AFIS)

During the 2004 fire season (June 2004), Eskom, implemented this novel satellite-based fire alert and information
system to help combat flashovers caused by fires underneath transmission lines. The Advanced Fire Information
System (AFIS) was the first near real-time satellite-based fire monitoring system in Africa. It was originally
developed for, and funded by, Eskom as a research project to test the technology.
By combining transmission line data with the fire data in a web-based Geographic Information System (GIS), AFIS
was able to do the continuous monitoring of active fires close (3 km either side of a line) to any of the transmission
lines in South Africa.
With the detection of a fire, e-mail and/or text messages are immediately sent to the Eskom transmission line
manager responsible for the affected transmission line as well as to the National Control Centre.
AFIS was first implemented using propriety GIS technology, but has been re-engineered as an Open Geospatial
Consortium-compliant Sensor Web application.
Hotspot detection by satellites

AFIS currently relies on contextual algorithms for thermal anomaly or “hotspot” detection using data from the two
MODIS sensors on-board the Terra and Aqua satellites and the MSG satellite. The contextual hotspot detection
algorithm (Flasse and Ceccato, 1996 as cited in Frost, 2011) was originally developed for the Advanced Very High-
Resolution Radiometer (AVHRR) sensors flown aboard the TIROS satellites but was significantly improved for
MODIS (Giglio et al. 2003). The algorithm uses the short-wave infrared (3.9μm) and thermal bands (10μm) to
discriminate fire pixels from background pixels. The algorithm classifies a pixel according to a fixed threshold, e.g.
T(3.9μm) > 305 K and _T < 8 K, (where T is surface temperature) to identify potential fire pixels, and the remaining
pixels are called background pixels. The neighbourhood of this pixel is then searched for background pixels,
growing the area if necessary, to ensure that at least 25% of the neighbourhood pixels are background pixels.
From this set of background pixels, the mean and standard deviation statistics are calculated from the difference
144
between the mid-infrared and thermal band. The pixel under consideration is then classified as a hotspot if its mid-
infrared value exceeds the background mean by some multiple of the standard deviation. A similar test is
performed on the mid-infrared and thermal band difference (Davies et al., 2008).
Collaboration between the CSIR SAC and NASA enabled the installation of the MODIS Rapid Response software
within the SAC MODIS direct broadcast processing system. The main component of the MODIS Rapid Response
code from the University of Maryland and Goddard Space Flight Centre (GSFC) is the enhanced contextual fire
detection algorithm for MODIS (Giglio 2003). The new algorithm can detect fires roughly half the minimum size that
could be detected with the original algorithm while having an overall false alarm rate 10–100 times smaller.
The system generates an ASCII text file containing information regarding the Latitude, Longitude, Brightness
Temp, Along scan pixel size, Along track pixel size, Date, Time of acquisition, Satellite (A=Aqua and T=Terra) and
Confidence. This text file is automatically sent via to the ArcIMS and ArcSDE server where it will be displayed in
real time.
The CSIR SAC in collaboration with David Taylor developed the MSG Fire Tracker software package. The software
code is based on a contextual fire detection algorithm, originally developed for AVHRR data (Flasse and Ceccato
1996). The MSG Fire Tracker software is activated every 15 minutes using the Windows XP scheduler.
An ASCII text file is generated every 15 minutes containing the following information: Latitude, Longitude,
Brightness Temp, Date, Time of acquisition, and Satellite (M = MSG). This text file is then automatically sent to the
ArcIMS and ArcSDE server where it will be archived and displayed.
Figure 106 - The original AFIS Web viewer
Hotspot detection success rate

The success of AFIS as a management tool within Eskom is measured by its ability to detect fires close to
transmission lines before flashovers occur.
A validation study in 2007 determined that MODIS was able to detect an average of 44% of all fires that caused
flashover fires for the years 2003 – 2005, while MSG detected 46% of all such fires during the same period. By
combining the detection capability of MODIS and MSG within one system (AFIS), the detection rate increased to
60% due to each satellite sensors unique design characteristics (Frost et al. 2007).
145
Figure 107 - Satellite detection rates of fires associated with flashovers for MODIS and MSG separate and combined
(Frost et al. 2007)
Figure 107 illustrates the fire detection rate of the separate satellites as well as the rate when they are combined.
The statistics of the MODIS and MSG detections clearly demonstrate the limitations of each of these sensors as a
detection tool on its own. The MODIS sensor could detect many of the smaller fires, but due to its infrequent revisit
time, was unable to detect short-duration fires between satellite overpasses between 10:00 am and 15:00 pm. The
MSG sensor struggles to detect smaller fires but can provide 15-minute updates of bigger fires (Frost, 2011).
Surprisingly, despite the MSG’s lower resolution and less advanced detection algorithm, it had a 2% higher
detection accuracy than MODIS, which highlights the advantage of frequent observations (every 15 minutes) Both
the MODIS and MSG sensors were able to detect a high percentage of fires before actual flashovers occurred.
MODIS could detect 70% of fires while MSG achieved a higher percentage of 81% of fires detected before the fire-
induced flashover had occurred (Frost et al. 2007).
In order to further improve the detection rate of MSG, a more sensitive non-contextual hotspot algorithm was
developed. The basic approach for the new algorithm was to build a general model of the diurnal cycle of the
thermal and infrared bands, and then fit this model to the observed data for the preceding 24 hours. The model is
able to generate accurate estimates of the expected background temperatures. If a statistically significant
difference between the current observed temperature and the predicted background temperature is observed, then
the pixel in question is classified as a hotspot (Frost, 2011).
146
Figure 108 - The percentage of fires detected by AFIS before the fire flashover had occurred (Frost et al. 2007)
The first implementation of this algorithm relied on a Kalman filter to provide the estimates of the background
temperature. Initial results indicate that this method is significantly more sensitive, particularly in cases where the
background temperature is below 300 K e.g. early morning (Van den Bergh and Frost, 2005).
11.4 Extending AFIS Functionality to other users

AFIS is not only providing SMS and email fire alerts to Eskom transmission line managers, but in 2007 also started
to provide alert messages to Fire Protection Associations (FPA’s) all around South Africa. FPA’s are registered
associations managed by farming communities around South Africa for predicting, preventing, managing and
extinguishing veld fires. More than 40 FPA’s are registered in AFIS and receive daily alerts whenever fires are
detected within the boundaries of the FPA.
As part of on-going development in response to user needs, several enhancements have been added to the
system. As older satellites reach the end-of –life, new satellites are being launched, each with improved
capabilities. The latest VIIRS satellite for instance, has a spatial resolution of 375mx375m.
In Figure 110 the satellite input sources in AFIS is shown and these are now complimented by weather and other
data sets such as vegetation types. Inputs to the system can also now come from smartphones and static fire
cameras. The various ways of viewing the information shown on the right of the figure includes smartphones and
tablets in addition to the original computer screens. Mobile apps can be downloaded free of charge from app
stores.
147
Figure 109 - A diagram show successive satellites and their improvement is spatial resolution
Data processing and

Input Sources storage Products Output
Platforms
Figure 110 - A Schematic representation of the total AFIS system
AFIS products supplement the near-real-time active fire monitoring with seven-day fire danger forecasts, burned
area maps and vegetation dryness reports. Real-time weather data is also obtainable from weather stations on the
ground, which is particularly helpful during actual fire events. Historic fire data kept by the system proved a
valuable resource for fire planning in many forms.
148
Figure 111 - The current global coverage of AFIS also showing the occurrence of fires
Figure 112 - The extensive global access of AFIS by mobile users during 2015
149
Figure 113 - The extensive global access of AFIS by mobile and online users during 2015
The mobile app has the following functionality:
 Available on iOS and Android platforms

 5-day Fire Danger forecasts
 Fire reports (date of last burn, fire frequency, vegetation curing)
 Geo-tagged picture uploads to the AFIS server Map of active fires within a radius of my location
 Track my location on the AFIS viewer (using the phone as a tracking device)
 View all my assets equipped with the AFIS GPS tracking device.
11.5 New developments
Figure 114 - Comparison with Figure 110 show number enhancements of AFIS
New enhancement planned for AFIS shown in Figure 114 are:
 Customer data uploads included in AFIS e.g. overhead lines or Fire Protection Zones.
 Precision 2 hourly Fire weather forecasting presented in the FDI system chosen by the user.
 Multi satellite active fire monitoring, updating the satellite constellation as new satellites are launched.
150
 30m Burn area. The MODIS data used previously used (16 daytime lag) will be done using Landsat and
Sentinel with 8-day return period
 Dynamic Fire Spread model using local fuel models, combined with high precision weather data.
 Mobile alerting and crowd sourcing application, where geo-tagged photos can be uploaded to AFIS from
smartphones.
 FDI and high-resolution burn area combined with precision weather to produce a fire risk report
 New AFIS viewer dashboard combining several the functions and reports mentioned above for a quick
overview of the fire situation.
11.6 Conclusions
The synergy of remote sensing, GIS, internet and mobile phone technologies described in this case study is
revolutionizing the way in which satellite-derived fire information is being used. The improved ability for control
centres to be aware of the situation on the ground, coupled with sound history and dependable forecasting of fire
spread by means of the dynamic fire spread facility is evidence of the degree of sophistication which the first fire
tracking systems have attained.
151
152
12. Mitigation of fires under power lines

12.1 Background
The effects of fire induced faults on power lines as well as numerous aspects pertaining to this, have been
discussed in the previous chapters. The remaining question of how to prevent or deal with these fires remains. In
this chapter the various approaches followed by utilities around the World will be discussed.
12.2 Routing of power lines

In areas with high fire frequency, new line routes should be chosen to avoid high fire risk areas. This approach
would require good historical information on fires, preferably over a period of more than 10 years to ensure that
peak fire seasons are included in the data set. This type of information is readily available these days from satellite
data. It must be stressed, however, that the choice of the route will inevitably also be influenced by factors such as
cost, environmental requirements as well as the specification of the required quality of supply of the new line.
A common practice in Eskom in South Africa is to attempt as far as possible, not to run lines adjacent to each other
in order to curb common cause faults. In the case of fires as one of these common causes, appears on the face of
it to be a good practice. In the case of the 765 fire faults (8.3.1, page119) it can be seen that the common cause for
those faults were in fact the weather on the day, responsible for fires over a vast area. These fires resulted in two
fire faults, 295km distant. It is therefore important to consider the bigger picture of the topology of the system during
the planning phase as well as the routing of new lines through areas of reduced fire threats. During the life of the
power line, the management of the vegetation will also be important to ensure reliability of the line
12.3 Vegetation management

Where high intensity fire areas cannot be avoided, the utility will have to turn to vegetation management in order to
minimize the risk of fires. This is a vast subject with many alternatives, each with its pros. Some examples of the
approach by some utilities will be shared below.
12.4 Fire Free Servitudes in sugar cane fields

One approach that has been followed in some countries is to negotiate a ROW that is devoid of sugar cane.
Alternative methods may include:
 the planting of alternative crops where fire is not used as a means of harvesting
 Green harvesting of sugar cane
 Removal of sugar cane
 Use of the ROW as roads or as fire breaks
 Controlled burns under power lines
In 16.4 the cost associated with the above strategies were discussed and shown to be very high. Some success
has been achieved by some of these strategies, although a certain number of trips were still experienced especially
in the case of runaway fires.
Some attempts by utilities to exclude burning of sugar cane under power lines failed and eventually compelled the
utilities to request legislation that would prohibit this form of harvesting using fire. (E-mail communication Alexandre
E. M. Meloni (meloni@marteng.com.br) SNC-Lavalin and Alcedo Quintiliano (alcedo@marteng.com.br) Marte
Engenharia).
12.5 The legal approach

A case of legislation preventing fires under power lines, has been reported by the Brazilian public utility CESP. In
1998, Federal Decree Nº. 2661 was published which prohibits sugar cane burning within servitudes and up to 15 m
away from transmission lines. The law also forbids any burning within 15 m from distribution lines.
153
At the same time, utilities have pursued with greater intensity agreements with land owners in transmission line
servitudes to plant corn, soybean or orange under the lines. Five main actions are followed to prevent fire induced
flashovers:
 Sugarcane plantations prohibition within transmission line servitudes (since 1999)

 Search for agreements with farmers to replace existing sugar cane in servitudes by corn, soybean, orange
and Brachiaria grass pasture in the existing transmission lines.
 The legal ban of sugar cane burning in the servitudes and up to 15m from their boundaries (Decree No.
2661/1998)
 Constant clearing or cutting down of Guinea grass, with special attention to dry seasons
 Permanent educational campaigns
These actions have resulted in the reduction of fire induced line faults in Brazil.
In Portugal, the utility REN reported that in 2010 about 65% of their lines cross forest land.
Figure 115 - Typical landscape in Portugal with transmission line crossing forest areas (photo REN)
Article 28, paragraph 3 of Regulatory Decree n. º 1/92 of 18 February 1992 provides buffer zone up to 45 metres
wide, 22.5 metres from the centre line of the power line, where trees can be felled or cut back as necessary to
ensure the minimum safety distance.
Forest Fire Prevention Networks were created with the promulgation of Decree‐Law no. 124/2006 of 28 June
2006 (superseded by Decree‐Law no. 17/2009 of 14 January 2009) and the Law no. 76/2017 of 17 August 2017.
This legislation provides structures for the territorial coordination of rural areas, following from the planning strategy
for preventing forest fires. Fire breaks are established by managing combustible material in primary, secondary and
tertiary areas.
To ensure the safety of the transmission lines under the terms of current legislation, the utility has established a
45 m wide protection zone – the buffer corridor (ROW) –, which extends for 22.5 m on either side of the lines.
Within this area, the utility must either prune or remove trees to maintain minimum safety distances, while the
landowners must not allow the planting or preservation of trees that could jeopardize the safety of the lines. (Figure
116)
154
Combustible-material management
strip
ROW (45 m)
Figure 116 - A schematic representation of the areas managed to prevent fires (REN)
The National Defence System against Forest Fires has outlined a series of measures and institutional coordination
actions for planning and intervention to protect forests from fires. One of these measures is that the electricity
transmission line buffer corridors are required for forest fires prevention strategies, becoming a secondary network
for fire prevention and combat.
For this it is necessary to proceed to vegetation management including bush and trees removing, along the
electricity transmission line buffer corridors. (Figure 116).
In complement, the utility also has a program for ROW conversion in what concerns:
 Land use – for instance conversion into for agriculture lands (vines, pasture, orchard);
 Soil occupation – replacing existing trees with others that are capable of remaining within the minimum
safety distances between them and the power lines, such as Portuguese oak, holm oak, walnut tree,
olive tree, stone pine, cork oak, etc.
This program allows the utility to ensure:
 Compatibility of vegetation with the corridors;

 Reduction of fire risk;
 Enhancement of the landscape and promotion of native species (such as oak, cork oak, holm oak, and
others);
 Stop the proliferation of areas of land with the same species of vegetation;
 Increase the intervention cycles;
 Ensure a shared responsibility with landowners;
 Cut maintenance costs.
155
Figure 117 - Example of a managed servitude in Portugal (photo REN)
Figure 118 - Example of a converted servitude, in Portugal, changing the soil occupation (photo REN)
Considerable progress has been made to date allowing the utility to convert approximately 1,450 ha of land, which
represents almost 4.5 % of the total servitude area (Figure 119).
156
Figure 119 - Progress made with the conversion of land use in servitudes between 2010 and 2018. (REN)
(in green confirmed values; in blue estimates)
12.6 Design and technical considerations

The effect of fires under power lines can also be mitigated by way of design in the case of new lines or the
modification of existing lines. The major thrust of such action will be to increase the height of the conductor, placing
it outside of the reach of the fire. The design engineer will have to create a balance between the extra cost of such
an approach and the quality of supply that is required for the line.
A further problem that should be addressed is that of the height of flames that has to be catered for in the design. It
has been shown that flame height and fire intensity vary markedly with changes in atmospheric conditions. (See
7.4 p98) As the determination of flame heights is carried out during controlled burns, which normally take place
during moderate fire weather, results will always be lower than that which occurs during extreme fire weather with
runaway fires. The experience of some utilities on design criteria of lines will be discussed below.
12.7 The approach followed in Mexico

Due to excessive line faults experienced on lines crossing sugar cane fields, the Federal Electricity Commission of
Mexico (FEC), the Testing Laboratory Equipment and Materials (LAPEM) conducted a series of experiments in
search of a solution to this problem. During these experiments the dielectric strength of the gap under fire
conditions was found to be 30kV/m and this value was used as a benchmark for raising the conductors where the
lines crossed sugar cane fields (Islas (1988).
A variety of methods were employed to lift the conductors some of which use intermediate towers, higher towers,
re-tensioning of conductors as well as inserting extensions in the tower body. The results show a marked reduction
in the number of line faults experienced, although not completely eliminating all faults. (See Figure 24 p45)
12.8 The approach in Brazil

In the 1980-90's, the Brazilian public utility CESP system, (34,000 structures and 12,000 km of transmission lines
ranging from 69 kV to 500 kV), experienced between 10 to 40 outages due to fires. 40 to 50% of them were due to
sugar cane fires with the remainder by other causes such as fire in Guinea grass or tree branches breaching safe
clearances.
During one particular year (1994) 110 outages occurred due to fires. It was a very dry year further compounded by
delays in contracting for vegetation management.
157
With the sugar cane 4 or 5 m high and the flames reaching 12 m or more, it would be necessary to raise the tower
an additional 15 m or more over the normal clearance. This would seriously affect the financial feasibility of 138 kV
transmission lines and as well as higher voltages.
After unsuccessful negotiations in the 1990’s with the sugar cane industry to stop the use of fire during harvesting,
90's Brazilian utilities (CESP, CTEEP, CPFL, CEMIG, FURNAS, CHESF, etc.) reverted to legislation for help. As a
result of Decree Nº 2661/1998, the practice of burning the sugar cane in right-of-ways was no longer permitted and
farmers were forced to perform green harvesting in the right-of-ways as well as 15 meters beyond their boundaries.
At the same time, utilities actively pursued agreements with land owners in right-of- ways to plant alternative crops
such as corn, soybean or orange or plant pastures with Brachiaria grass (Brachiaria decumbens), which is not as
tall as Guinea grass and therefore less susceptible to fire.
Five actions listed below were implemented and outages have reduced to only about five per year in the CTEEP
system in Sao Paulo, a company originally from CESP, now with 14,000 km of 88 to 500 kV TLs (this overall length
accounts for other lines of the ISA group in Brazil).
1) Sugarcane free right-of-ways (since 1999);

2) Agreements with farmers to replace existing sugar cane in right-of-ways by corn, soybean, orange and
Brachiaria grass;
3) The legal ban of sugar cane burning in the right-of- ways and up to 15 m from their boundaries (Decree
No. 2661/1998);
4) Constant clearing or cutting down of Guinea grass, especially during dry seasons;
5) On-going educational campaigns.
Figure 120 - The planting of alternative crops in Brazil to mitigate against fire trips (Ref)
12.9 The approach followed in Australia

Because of problems resulting from sugarcane fires in Queensland Australia conductor height was increased by
6.0 m mainly on affected 275kV lines. Although the results were positive some fire trips still occurred occasionally
during more intense fires.
In areas where a high frequency of faults still occurred farmers were paid to 'green harvest' under and near the
lines before burning off to minimise line trips. However, over the last 15 years, while some small isolated areas are
still burnt, current practice in the majority of areas is now to use mechanical harvesting without the use of fire.
Sugar Cane trash are left on the ground for nutriment or reclaimed for use in associated small power generation
plants and other agricultural purposes such as a mulch blanket to conserve moisture and provide nutriment to the
soil.
158
12.10 The approach of South Africa

The transmission lines of Eskom in South Africa are not only affected by sugar cane fires but experience most fire
faults from grass and shrub fires. The fire intensity and flame heights differ with the type of fuel as well as with the
weather that exists at the time of the fire. This fact makes the fire withstand gradient, and therefore, the design
parameters highly variable. Some examples will be discussed below to illustrate this.
The case of two 400kV lines subjected to similar fire conditions illustrated the effect of increased mid-span
clearance on the reduction of fire induced faults in a Bushveld savannah type environment. It was found that the
estimated dielectric strength under fire conditions for this biome was 15kV/m. The line with only 2m additional mid-
span clearance experienced less than half of the fire faults of that of the adjacent line. Vosloo et al. (2009)
The tests performed at CSIR and NETFA HV laboratory shows that the dielectric strength of the gap spanned by
sugar cane flames is as follows:
1) Test performed at CSIR 15kV/m

2) Test performed at NETFA 12.5kV/m (cane stored for a few days)
3) Test performed at NETFA 9kV/m (cane stored for three weeks)
Like other parts of the world, the problems associated with fire induced flashovers of power lines sugarcane fields
is receiving rigorous attention and is the focus of on-going research.
The approximately 35 300 registered sugarcane growers annually produce on average 20 million tons of sugarcane
mainly on the eastern seaboard of South Africa. Approximately 14 445 small cane growers produced 8.4% of the
total crop delivered cane in 2013. The removal of the cane within the line servitudes therefore is not always an
option as some of these areas are crucial to the sustainability of small cane growers, with some of these farmers
owning land as small as half a hectare. The participation of emerging farmers in sugarcane production is increasing
and they cannot easily afford to lose part of their land due to new power lines.
In South-Africa green harvesting is not a favoured option at this time, as it requires costly procedures and
modifications to the farmers harvesting process and would incur additional costs and may even at times cost the
farmers more than what it would cost Eskom to build taller towers. Taller towers, although costly, will reduce the
probability of line faults, and may be the only option for Eskom when the options discussed above are not viable.
The issue of the dielectric strength under fire conditions and the flame height is still being researched by way of
field measurements as well as experimental burns in laboratories. Eskom researchers recorded a maximum flame
height of 32 meters in 1995 but the current study in KZN has indicated that fires exceeding 20 meters are not very
common. Further research regarding flame heights will be necessary.
The economic implications of the various options discussed above and using existing as well as new towers, have
also received special attention. For both 400kV and 765kV the raising of tower height and mid-span clearance is
considerably cheaper than cane free servitudes.
Of the three alternative solutions compared in this research (green harvesting, cane free servitudes and taller
towers), the most viable option for improved line performance is the increase of the mid-span clearances of lines
traversing cane fields. It is suggested that a minimum mid-span clearance of 24 meters is maintained for all 400kV
and 28 meters for 765kV lines spanning over cane areas. The clearances of course apply to South African cane
fields and weather conditions and may have to be adopted to the conditions existing in other countries.
12.11 Fire risk mapping

The use of historic data obtained from fire induced line faults as well as the fires which occurred in close proximity
to power lines, has been shown to be valuable in deducing the fire induced flashover index in 10.
That data can also be used to generate a graphic rendition of power lines with high fire risk by plotting the power
lines, the fire faults for a given period and the history of fires observed from satellite sources in a geographic
information system (GIS).
159
In the example shown in Figure 121 shows the low risk lines as green, with the higher risk lines in darker colours. In
addition to risk to the line, the transmission system should further be analysed to determine how a permanent line
outage caused by fire, will affect the remaining lines. The operator should also be mindful of the fact that fires will
occur over a widespread are during poor fire weather and may threaten other lines as well.
Figure 121 - A geographic depiction of the risk of fire induced faults

for the Eskom Transmission system for the period 1994-2014
12.12 Noise detection

The detection of fires by using AC corona noise has been the subject of numerous research work (Vosloo,1992,
Evert & de Klerk, 1994, Evert, 2003).
A number of studies have been carried to show that fire induced faults under 275 kV and 400 kV transmission
network can be detected with the aid appropriate HF background noise identification. Fires under these lines cause
increase levels of conductor corona which generate, inter alia, increased levels of background radio noise
detectable in the carrier system communication channels.
Although the approach shows promise, it is not clear to what extent it will allow the operator to respond in time to
prevent a line fault by, say, switching the line out or taking it off ARC. It will, however, certainly help the operator to
be certain that fires exist under the line.
12.13 Conclusion
In this chapter, several approaches followed by utilities to mitigate against fires under power lines are presented.
These measures are of course not complete and comprehensive but, as research and developments proceed, new
and better methods will be found.
160
13. Conclusion
When Workgroup B2.45 was formed with the task to investigate Bushfire Characteristics and the potential impacts
on line performance, the task ahead appeared to be straight forward.
As the contributions came in and the subject was researched in the literature, it became clear that the subject was
a wide one, which involved many (and new) disciplines in all aspects of fires and combustion.
This TB is an attempt to put together a fairly wide document on what turned out to be a fascinating subject. It is
believed that no such document exists today.
It is hoped that it will serve a wide group of readers, from the design engineer in a utility, to the Operations and
Mantenace staff. Further it may also serve as a reference for persons who wish to do further study on this subject.
161
162
14. References
Abdel-Salam,M., H Anis, A El-Morshedy and R Radwan, (2000). High voltage engineering: theory and practice. Marcel Dekker,
2000.
Albini, F.A., (1976). Estimating wildfire behaviour and effects. U.S.D.A. For. Serv. Gen. Tech. Rep. I.N.T.30.
Allen N L and A. Gaffar, (1993). Temperature and density effects on streamer propagation in air. in 8th International Symposium
on High Voltage Engineering. Yokohama, Japan.
Allen N L and A. Ghaffar, (1995). The variation with temperature of positive streamer properties in air. Journal of Physics D:
Applied Physics, 1995. 28: p. 338-343.
Aleksandrov, N.L., E.M. Bazelyan, I.V. Kochetov & N.A. Dyatko. (1997). The ionization kinetics and electric field in the leader
channel in long air gaps. Journal of Physics D: Applied Physics, 1997. 30:p 1616-1624.
Anon (1976). IEEE Standard 80-1976, IEEE Guide for safety in AC substation Grounding
Anon, (1978). American National Standard, “IEEE Standard Technique for High Voltage Testing”, IEEESTD 4-1978.
Anon, (1983). “Protective grounding for Lineman”. Transmission and Distribution Vol 35, No 6, 1983.
Anon, (1998). NATIONAL VELD AND FOREST FIRE ACT, 101 OF 1998 AS AMENDED BY THE NATIONAL FOREST AND
FIRE LAWS AMENDMENT ACT, 2001. Government Printer Pretoria.
Anon, (2005). Report of an Investigation into a Non-Eskom Fatal Contact Incident which occurred at Farm Misty in the
Camperdown area. Unpublished Eskom Investigation report.
Anon, (2009). Bushfire Weather. Australian Government Bureau of Meteorology, Weather services http://www.bom.gov.au Last
updated 01 October 2009
Anon. (2009). Overall Fuel Hazard Guide Third Edition May 1999 – Reprint 2009 Department of Sustainability and
Environment. Victorian Government Melbourne Australia.
Anon, (2009). Victorian Bushfires Royal Commission Report)
Anon, (2009a). SPUTNIK 1.http://en.wikipedia.org/wiki/Sputnik_1. Accessed 6 April 2009
Anon, (2009b) Anon.2009.TIROS-1.http://en.wikipedia.org/wiki/TIROS-1.Accessed 6 April 2009
Anon (2009c). Fire Information for Resource Management System. from, http://maps.geog.umd.edu/firms
Anon (2011). "Santa Ana Winds - Wildfires". NOAA Watch All Hazards Monitor. NOAA National Weather Service. Retrieved 10
February 2011
Axelson, E. (1973). Portuguese in South –East Africa 1488 – 1600. Struik, Johannesburg.
Baldo G, I. Gallimbeti, H. N. Gracia, B. Hutzler, J. Jouaire and M. F. Simon. (1975). Breakdown Phenomena of Long Gaps
Under Switching Impulse Conditions Influence of Distance and Voltage Level. IEEE Transactions on Power
Apparatus and Systems, 1975. PAS-94(4): p. 1131-1139.
Baldo , G., R.R. Diaz & G. Pesavento. (1993) Discharge Mechanisms in presence of flames. In 8th International Symposium on
High Voltage Engineering. 1993 Yokohama, Japan.
Beaufait, W.R., (1965). Characteristics of back fires and head fires in a pine needle fuel bed. USDA For. Serv. Res. Note
I.N.T.30.
Benard L., and G. von Elbe, (1987). Combustion, flames and explosion of gases. Orlando: Academic press.
Bergreen, L., (2003). Over the Edge of the World: Magellan's Terrifying Circumnavigation of the Globe. HarperCollins. p.
179. ISBN 978-0-06-186588-6.
Bologna, FF, KJ` Sadurski, (2001). Assessment of Safety Risks Associated with Fire Induced Flashovers on High Voltage
Transmission Lines. Eskom Research report RES/RR/00/12054
163
Bondiou, A., I. Gallimberti, (1994). Theoretical modelling of the development of the positive spark in long air gaps. Journal of
Physics D: Applied Physics, 1994. 27: p. 1252-1266.
Boutlendji J, Allen. (1994). Dielectric strength of air insulation for coordination of minimum clearances of overhead lines: a
review. IEE Proceedings Science, Measurement and Technology, 1994. 141(6): p. 449-461.
Bradley J. N., (1965). Flame and Combustion Phenomena. London, Methuen.
Britten, A.C., P. Hill. 1997. Controlled burning of sugar cane fires under the No.1 Impala-Avon 275kV line on 3 August 1997:
Assessment of 50Hz Fire-induced leakage currents. Eskom Research report TRR/CON/3/1997.
Britten, A.C. 1997 The detection of sugar cane fires under 275 and 400kV Transmission lines: the feasibility of using power
frequency (50Hz) leakage current through the fire as a detection parameter. Report no TRR-QS-96_Q503 Nov
1996.
Britten A.C. 2012. SUGAR CANE FIRES: minimum conductor heights; Comparisons between Eskom Predictions and Mexican
Data. Unpublished Eskom presentation by A.C. Britten.
Brown, A.A. & K.P. Davis, (1973). Forest fire: control and use. McGraw Hill Book Co., New York.
Bowen, H. J. M., (1979). Environmental Chemistry of the Elements (London: Academic Press).
Butler, B.W., J. Cohen, D.J. Latham, R.D. Schuette, P. Sopko and co-author. (2004). Measurement of radiant emission power
and temperature in crown fires. Can. J. For. Res. 34 1577-1587
Cameron, C. (1992). Planning Proposal –Cost justification of Purchase of servitude right over cane fields in Natal. Unpublished
Eskom report.
Carrara G, L. Thione, (1976). Switching Surge Strength of Large Air Gaps: A Physical Approach. IEEE Transactions on Power
Apparatus and Systems, 1976. PAS-95(2): p. 512-521.
Chandler, C., P. Cheney, P Thomas, L. Trabaud, D. Williams, (1983). Fire in Forestry (New York: John Wiley and Sons).
Cheney, N.P. (1981). Fire behaviour. In: Fire and the Australian biota. Eds. A.M. Gill, R.H. Groves, I.R. Noble. Australian Acad.
Sci., Canberra. 151-175.
Cheney, P., J.S. Gould. (1995). Fire growth in Grassland Fuels. Int. J. Wildland Fire 5(4): 237-247, 1995
Cheney, P., Sullivan, A., (1997). Grassfires: Fuel, Weather and Fire behaviour. CSIRO Publishing, Collingwood, Australia.
Cheney, P. and A. Sullivan. (2008). Grassfires- Fuel, weather and fire behaviour – Second Edition- CSIRO Publishing
Cheney, N. P. (1981). Fire Behaviour. In Gill, A.M., Groves, R.H. & Noble, I.R. (Eds). Fire and the Australian Biota. Canberra:
Australian Academy of Science.
Chrzan, K.L., Z. Wroblewski, (2004) The threat caused by fires under high voltage lines. International Conference on Advances
in Processing, Testing and Application of Dielectric Materials. Scientific papers of the Institute of Electrical
Engineering Fundamentals of the Wroclaw University of Technology, No. 40, Conference 15.
Cobine, J.S., (1958). Gaseous conductors. New Dover edition. Dover Publications, New York
Cowen, P.M., J.A. Dunn, P. Naidoo, J. Masters (1991) Sugar cane fires induced transmission line flashovers. SAIEE electron
Jan 1991
Claverie, P., (1971) Predetermination of the behaviour of polluted insulators. IEEE Transactions PAS, Vol 90, no 4, July-August 1971.
Cruz, M.G., B.W. Butler, M.E. Alexander, J.M. Forthofer, and R.H. Wakimoto.(2006). Predicting the ignition of crown fires above
a spreading surface fire. Part 1: model idealization. Int J of Wildland Fire, 2006, 15 ,47-60
Danielson, Levin, and Abrams (2003). Meteorology, McGraw Hill
Davies, D.K., H.F. Vosloo, P.F. Frost, S. S. Vannan. (2008). Near real-time fire alert system in South Africa: from desktop to
mobile service. DIS 2008, February 25-27, 2008, Cape Town, South Africa
Deno, D.W., L.E. Zaffanella, (1982). Flashovers caused by Fires. In La Forest, J.J., (ed.): Transmission Line Reference Book
345kV and above, (2nd ed) California: EPRI. p384-8.
164
Domens, P., A. Gilbert, J Dupuy & B. Hutzler, (1991). Propagation of the positive streamer-leader system in a 16.7m rod-plane
gap. Journal of Physics D: Applied Physics1991, 24: p. 1748-1757.
Eriksson A.J., (1981). “Transient Impedance of Earthing Systems”, Part II of Cigre Working Group Paper No 33.81 (SC). Rio de
Janeiro 1981
Everson, T. M., Smith, F. R. & Everson, C. S. (1985). Characteristics of fire behaviour in the montane grasslands of Natal.
Journal of the Grassland Society of Southern Africa, 2, 13-21.
Everson, T. M., Van Wilgen, B. W. & Everson, C. S. (1988). Adaptation of a model for fire danger rating in the Natal
Drakensberg. South African Journal of Science, 84, 44
Evert, C.R. (2003). The Detection of Fires under High Voltage Transmission Lines. M.Sc. Thesis University of KwaZulu Natal.
Evert, C. R., P.J. de Klerk. (1994). CORONA NOISE AS A DIAGNOSTIC TOOL. Proceedings of the 4th International
Conference on Properties and Applications of Dielectric Materials July 3-8,1994. Brisbane Australia
Flasse, S. & Ceccato, P. (1996). A contextual algorithm for AVHRR fire detection. International Journal of Remote Sensing, 17,
419-424.
Fonseca, J.R., A.L. Tan, V. Monassi, W.S. Junquera, R.P. Silva, L.A.R. Assunção, M.O.C. Melo. (1987) Insulation,
dimensioning in Transmission Lines in regions subject to grass fires. (Portuguese) Tx National Seminar on
Electric Energy Generation and Distribution. Belo Horizonte – Minas Gerais Brazil.
Fonseca, J.R., A.L. Tan, V. Monassi, W.S. Junquera, R.P. Silva, L.A.R. Assunção, M.O.C. Melo. (1990) Effects of Agricultural
Fires on the Performance of Overhead Transmission Lines. IEEE Transactions on Power Delivery, Vol 5, No. 2
April 1990.
Foster, E.O. (1982) The search for Universal Features of Electrical breakdown in Solids, Liquids and Gasses. IEEE
Transactions on Electrical Insulation. EI-17(6): p517-521.
Fovell, R. (2012). "UCLA explains the naming of the Santa Ana winds" . http://sciencedude.blog.ocregister.com/. Orange
County Register. Retrieved May 3, 2012.
Frost, L. S. (1961). Conductivity of seeded atmospheric pressure plasmas.J. Appl. Phys. 32, 2029–2037.
Frost P., H. Vosloo, (2006). Providing satellite-based early warning of fires to reduce fire flashovers on South Africa’s
transmission. Bushfire conference 2006 – Brisbane 6-9 June 2006
Frost, P. E., Vosloo, H. & Annegarn, H. (2007). Providing satellite-based early warnings of fires to reduce fire flashovers on
south Africa’s transmission lines. Geoscience and Remote Sensing Symposium, 2007. IGARSS 2007. IEEE,
Barcelona.
Frost, P.E. 2011, Development of a Fire-induced Flashover Probability Index (FIFPI) for Eskom Transmission Lines. M.Sc.
Thesis, University of Johannesburg.
Giglio, L., J. Descloitres, C. O. Justice, & Y. Kaufman (2003). An enhanced contextual fire detection algorithm for MODIS.
Remote Sensing of Environment, 87, 273-282.
Gilbert A and F Bastien, (1989). Fine structure of streamers. Journal of Physics D: Applied Physics, 1989. 22.
Goelian N, P. Lalande, A Bondiou-Clergerie, G.L.Bacchiega, A. Gazzani and I. Gallimberti, (1997). A simplified model for the
simulation of positive-spark development in long air gaps. Journal of Physics D: Applied Physics, . 30: p. 2441-
2452.
Goldman, M. (1981). Corona Discharges and their Application. IEEE Proceedings, 1981. 128 Part A(4): p298-302.
Gracia H N and B. Hutzler, Electrical Breakdown in long air gaps - The Final Jump. : p.206-210.
Grant, R., and V. Thomas. (2000). Sappi Tree Spotting Bushveld. Johannesburg: Jacana.
Griffiths, D. (1999). Improved formula for the drought factor in McArthur’s Forest Fire Danger meter. Australia Forestry, 62, 202-
206.
165
Heikkila, T. V., R. Gronovist, & M Jurvelius. (1993). Handbook on forest fire control. A guide for trainers, Helsinki, National board
of Education of the Government of Finland.
Hileman, A.R., (1980). Transmission line insulation coordination – Twenty-eight Bernard Price Memorial Lecture. The
Transactions of the SA Institute of electrical Engineers, 1980. June; p145-157.
Hutzler B., D. Hutzler-Barre. (1978). Leader Propagation Model for Predetermination of switching Surge Flashover Voltage of
Large Air Gaps. IEEE Transactions on Power Apparatus and Systems. PAS-97(4):p1087-1096.
Islas R.R., (1988). Problems and solutions to transmission line outages due to the burning of cane fields under them. (Spanish)
Proceedings of the IEEE Summer Power Meeting, Acapulco, Mexico, 1988.
Justice, C. O., L. Giglio, L. Boschetti, D. Roy, I. Csiszar, J. Morisette, & Y Kaufman. (2006). MODIS Fire Products. Algorithm
Technical Background Document. Retrieved on February 7, 2009 from
http://modis.gsfc.nasa.gov/data/atbd/atbd_mod14.pdf
Karady G.G, Shah M, and Dumora D,. (1996) “Probabilistic Method to Assess Insulating Link Performance for Protection of
Crane Workers”, IEEE Trans. Power Delivery Vol 11, No 1, January 1996. Karlovitz B, (1962). Flames
Augmented by Electrical Power. in International Union of Pure and Applied Chemistry: Radio-activation Analysis
Symposium. 1962. Vienna, Austria: Pergamon Press.
Keetch, J.J. & Byram, G.M. (1968). A Drought Index for Forest Fire Control. US Department of Agriculture Forest Service
Research Paper SE-38.
Kuffel & W.S. Zaengl. (1984). High Voltage Engineering. ISBN 0-08-024213-8 ed. A. Wheatman & Co.
Kunhardt, E.E. and L.H. Luessen, (1983) Electrical Breakdown and Discharges in Gases. Part A Fundamental Processes
and Breakdown. Plenum Press New York.
Laing, M. V. (1978). Forecasting bush and forest fire weather in Rhodesia. Rhodesia, Dept. Meteorological Services.
Lambeth (1984). Notes on Insulator pollution. Unpublished notes issued at in-house Eskom Insulator course, 1984 Available
from antony.britten@gmail.com
Lanoie, R., H.P. Mercure, (1997). Influence of Forest fires on Power Line Insulation: AC Line Case. 10 th International
Symposium on High Voltage Engineering, Montréal, Quebec, Canada, August 25-29, 1997.
Leigh, J.H. & Noble, J.C., (1981). The role of fire in the management of rangelands in Australia. In: Fire and the Australian biota.
Eds. A.M. Gill, R.H. Groves, I.R. Noble. Australian Acad. Sci., Canberra. 471-495.
Le Roy, G., C. Gary, B. Hutzler, J. Lalot, C. Dubanton, 1984. Les propriétés diélectriques de l’air et les très hautes tensions.
Ouvrage réalisé sous la direction de Claude Gary. Collection de la Direction des études et recherches d’
Électricité de France. Editions Eyrolles, Paris.
Lewis, B., G. von Elbe. (1987). Combustion, Flames and Explosion of Gases. Third edition, Academic Press, Inc. Harcourt
Brace Jovanovich, Publishers. New York
Lucas, C., K. Hennessy, G. Mills and J. Bathols. (2007). Bushfire Weather in Southeast Australia: Recent Trends and Projected
Climate Change Impacts. Consultancy Report prepared for The Climate Institute of Australia (Bushfire CRC and
Australian Bureau of Meteorology, CSIRO Marine and Atmospheric Research)
Luke, R.H. & A.G. McArthur, (1978). Bush fires in Australia. Australian Govt. Pub. Serv.., Canberra.
Loeb, L.B., J.M. Meek, (1940). The Mechanism of Spark Discharge in Air at Atmospheric Pressure. I. Journal of Applied
Physics, June 1940, Volume 11, p438-446.
Masters, N. (2012). "SoCal's Devil Winds: The Santa Anas in Historical Photos and Literature". www.ket.org. Retrieved 3 May
2012.
McRae, R.H.D. (1999). Operational calculation of drought index and drought factor. ACT Emergency Services Agency,
Technical Note TN018. Retrieved on February 7, 2009 from www.esa.act.gov.au/Documents/TN018_-_df_algs.pdf
Meek, (1969). Inaugural Address-The Vital Spark. IEE Proceedings, 116(1): p. 1-11.
166
Martinussen, K.H.H, G.R. Bozzoli, (1964). Study of the effect of dust particles on breakdown voltages in air. Transactions of the
South African Institute of Electrical Engineers, Vol 55, April 1964.
Maruvada, P.S.2005. Corona and Gap Discharge Phenomena. EPRI Transmission Line Reference Book 200kV and above,
Third edition. EPRI, Palo Alto, CA:2005.1011974.
Mhaule, Duma. Senior Engineer: HV Lines, eThekwini Electricity, ZA. Email: mhauled@elec.durban.gov.za
Mousa, A.M., (1990). Protecting Firemen against fire-Induced Flashovers. IEEE Transactions on Power Delivery, Vol. 5 ,No.1,
January 1990.
Mphale, K., M. Heron, (2007). Wildfire plume electrical conductivity. Tellus 59 (2007), 766-772.
Mphale, K., M. Heron, (2008). Measurement of electrical conductivity for a biomass Fire. Int. J Mol. Sci 2008,9,1416-1423; DOI:
10.3390/ijms9081416. www.mdpi.org/ijms.
Naidoo, P., D.A. Swift (1993). Large Particle Initiated Breakdown of an Atmospheric Air gap: Relating to AC Power Line Faults
Caused by Sugar Cane Fires. in 8th International Symposium on High Voltage Engineering. 1993. Yokohama,
Japan.
Nel, R., S Narain, K.Sadurski, P. H. Pretorius and H. Vosloo. (2013). Fire induced flashovers - recent findings and
recommendations for improved line design. 7th Southern Africa Regional Conference. Somerset West, October
2013
NIST Atomic Spectral Database, 2001, URL; http://physics.nist.gov.
Palmer H. B., and J. M. Beer. (1987) Combustion, flames and explosion of gases. 3 ed. 1987, Orlando: Academic press.
Pedersen T., R.C. Brown. (1993). Simulation of electric Fields in Premixed Methane Flames. Combustion and Flame. 1993. 94:
p.433-448.
Preston-Whyte, R. A. & Tyson, P. D. (1988). The atmosphere and weather of Southern Africa. Cape Town, Oxford University
Press: NECC.
PRIMEN. (2001): The cost of Power Disturbances to Industrial and Digital Economy Companies. Report to EPRI’s Consortium
for Electric Infrastructure for a Digital Society (CEIDS). Madison. WI.
http://ceids.epri.com/ceids/Docs/outage_study.pdf
Pyne, S. J., Andrews, P. L. & Laven, R. D. (1996). Introduction to Wildland Fire. Wiley: New York.
Pyne, S.J., (1997). World Fire: The culture of Fire on Earth. University of Washington Press, 1997
Radtke, R., K. Gunther. (1976). Electrical conductivity of highly ionized dense hydrogen plasma: 1 Electrical measurements and
diagnostics. Journal of Physics D: Applied Physics 1976.9.
Rallis, C.J. B., M. Mangaya. (2002). Ignition of veld grasses by hot aluminium particles ejected from clashing overhead
transmission lines. Fire Technology 38, 81-92. University of Witwatersrand South Africa
Rizk, F.A. M., (1989) Switching impulse Strength of Air insulation: Leader inception criterion. IEEE Transactions on Power
Delivery, 1989.4(4): p. 2187-2194.
Rizk F A M (2), (1989). A Model for switching impulse leader inception and breakdown of long air gaps. IEEE Transactions on
Power Delivery, 4(1): p. 596-606.
Rizk F A M, (1996). Critical switching impulse breakdown of long bundle conductor gaps. IEEE Transactions on Power Delivery,
1996. 11(1): p. 373-381.
Rizk F A M., (1997) An Overview of Recent High Voltage Engineering Research in Canada. In 10th International Symposium on
High Voltage Engineering. 1997. Montreal, Quebec, Canada: ISH.
Robledo-Martinez, A., E. Guzman, J.L. Hernandez, (1991). Dielectric Characteristics of a Model Transmission Line in the
Presence of Fire. IEEE Transactions on Electrical insulation, Vol 26 no.4, August 1991.
Rothermel, R.C. (1972). A mathematical model for predicting fire spread in wildland fuels. Research Paper INT-RP-115. Ogden,
UT: USDA Forest Service, Intermountain Forest and Range Experiment Station.
167
Ryan, K. C. (1981) Evaluation of a Passive Flame-Height Sensor to Estimate Forest Fire Intensity. United States Department of
Agriculture. Research Note PNW-390 June 1981.
Sadurski K.J., E.W.J Robertson., (1977). Effects of bush, grass and cane fires below transmission lines. Eskom internal report
1997, Report no. 123/2/97. Project no. 72E9, May1977
Sadurski, K. J., J. P. Reynders (1989), High Voltage AC breakdown in Presence of Fires. In Sixth International Symposium on
High Voltage Engineering. 1989. New Orleans, LA, USA.
Sakao, F. and Sato, H. (1969). Nonequilibrium electrical conductivity of potassium – seed argon plasma. Phys. Fluids 12, 2063–
2071.
Santoni, P. A., A. Simeoni, J. L. Rossi, F. Bosseur, F. Morandini and co-authors. (2006). Instrumentation of wildland fire:
characterization of a fire spreading through a Mediterranean shrub. Fire Safety J. 41, 171–184.
Schneider, J., and F.W. Hofmann. (1959). Absorbsion and dispersion of microwaves in Flames. Phys. Rev. 116, 244-249
Shea, S.R., Peet, G.B. & Cheney, N.P., (1981). The role of fire in forest management. In: Fire and the Australian biota. Eds.
A.M. Gill, R.H. Groves. I.R. Noble. Australian Acad. Sci. Canberra. 443-470.
Sekso-Telento, A., S. Boji, Z. Bertalanić. (2001). Fires in Mediterranean area and high voltage lines. Paper 100-05 Cigre
Symposium, CAIRNS 2001.
Sorokin, A., X. Vancassel, P. Mirabel. (2002). Emission of ions and charged soot particles by aircraft engines. Atmos. Chem.
Phys. Discuss. 2002, 2, 2045-2074.
Sukhnandan A. (2004). A Theoretical and Experimental Investigation into Fire Induced Flashover of High Voltage Transmission
Lines. MSc Thesis. University of KwaZulu Natal.
Sukhnandan A, D.A. Hoch., (2002). Fire Induced Flashover of Transmission Lines: Theoretical Models, Proceedings of the
IEEE Africon Conference, George South Africa, 2002
Soo, M. (2012). Aluminum dust concentration effect on combustion in hydrocarbon Bunsen flames. M.Sc thesis, McGill
University Montreal Canada.
Taylor, P.V. (1999): Investigation into bird streamer caused transient earth fault on a 275kV Transmission Grid. M.Sc. Thesis,
University of Natal, Pietermaritzburg, Department of Electrical Engineering.
Tobazéon R., (DATE). Liquid Breakdown and its Relation to Gas Breakdown, Laboratoire d'Electrstatique et de Materiaux
Dielectriques: Grenoble Cedex France.
Teie, W.C. (2003). Fire manager’s Handbook on Veld and Forest fires Strategy, Tactics and Safety. C.F.Pool (ed). Menlo Park:
S.A. Institute of Forestry.
Thomas P. A. and R.S. McAlpine. (2010) Fire in the Forest. Cambridge University Press, New York, USA.
Trollope, W.S.W. (1983). Control of bush encroachment with fire in the arid savannas of southeastern Africa. PhD thesis,
University Natal, Pietermaritzburg.
Trollope, W. S. W. & A. L. F. Potgieter. (1985). Fire Behaviour in the Kruger National Park. Journal of the Grassland Society of
Southern Africa, 2.2, 17-22.
Trollope, W. S. W. & N. M. Tainton. (1986). Effect of fire intensity on the grass and bush components of the Eastern Cape
Thornveld. Journal of the Grassland Society ofSouthern Africa, 2, 27-42.
Trollope, W. S. W., Trollope, L.A. and Bosch, O. J. H., (1990). Veld and pasture management terminology in southern Africa. J.
Grassld. Soc. South Afr. 7,1:52-61.
Trollope W.S.W., L.A. Trollope, D.C. Hartnett, (2002). Fire behaviour a key factor in the fire ecology of African grasslands and
savannas. In Forest Fire Research & Wildland Fire Safety, Viegas (ed). 2002 Milpress, Rotterdam, ISBN 90-
77017-72-0
Trollope, W.S.W. & L.A. Trollope, (2004). Final report on the fire behaviour trial conducted to estimate flame heights and fire
intensities generated by different plant fuels in the Bankenveld veld type in Gauteng province. Eskom
Unpublished research report.
168
Tyson, P. D. (1964). Berg winds of South Africa. Weather, 19, 7-11.
Tyson, P.D. and R.A. Preston-Whyte. 2000: The Weather and Climate of South Africa. Cape Town: Oxford University Press
Vajeth, R. (2008). Methodology for calculating the cost of a network fault and the cost of unserved energy. Eskom Guideline
TRMAGABK5.
Vajeth, R., D. Mtolo and D. Dama. (2003): Cost of a Network Fault Affecting a Transmission Supply Point. Eskom Enterprises
Report no. RES/RR/03/20237
Van Den Bergh, F. & P.E. Frost, (2005). A Multi-Temporal Approach to Fire Detection using MSG Data. The third international
workshop on the analysis of multi-temporal remote sensing images (Multitemp’05), Biloxi, Mississippi, USA.
Vines, R.G., (1981). Physics and rural chemistry of fires. In: Fire and the Australian biota. Eds. A.M. Gill, R.H. Groves, I. R.
Noble. Australian Acad. Sci., Canberra.
Vodacek, A .,R. L. Kremens, A. J. Fordham,s. C. Vangorden, D. Luisi and, J. R. Schott, (2002). Remote optical detection of
biomass burning using a potassium emission signature. INT. J. REMOTE SENSING, 2002, vol. 23, no. 13,
2721–2726
Vosloo, H.F., (2002). Report on Fire incidents on 765kV lines on 29 July 2002. Eskom report TT/07/02.
Vosloo, H.F., (2002) (1). Report on Fire fault on the 765kV line in the Standerton area on 29 July 2002. Eskom report TT/07/02.
Vosloo, H.F., (2002) (2). Report on Fire fault on the 765kV line in the Virginia area on 29 July 2002. Eskom report TT/08/02
Vosloo, H. F., W.S.W. Trollope and P. E. Frost, (2003). Right-of-Way Management by Eskom, South Africa. Unpublished
Assignment for M.Sc. in Geograpy, Rand Afrikaans University.
Vosloo, H.F., (2003). Vegetation scanning Research project: Description of background Information to and structure of an
experiment to demonstrate hyper spectral scanning technology in the identification of plant communities with
high fire risk. Unpublished Assignment for M.Sc. in geography, Rand Afrikaans University.
Vosloo, H.F., (2005). The need for and contents of a life cycle Management plan for Eskom Transmission line Servitudes. M.Sc.
Thesis., University of Johannesburg.
Vosloo, H.F., A.C. Britten and A.A. Burger, (2009). The Susceptibility of 400 kV Transmission Lines to Bird Streamers and Bush
Fires: A Definitive Case Study. Proceedings of the CIGRE 6th Southern African Regional Conference, Somerset
West Western Cape 17-21 August 2009
Vosloo, W.L., (1992). A Study into the effect of sugar cane fires on AC corona noise. M.Sc. Thesis, University of Stellenbosch
Wadhwa, C. L., (2001). High Voltage Engineering, New Age International Publishers, New Dehli , India.
Waters R T, (1981). Breakdown in non-uniform fields. IEE Proceedings, 1981. 128(Part A No 4): p. 319-325.
West, H.J., D.W. Mc Mullan. (1979). Fire induced Flashovers of EHV Transmission Lines. IEE PES Winter meeting New York,
NY, 4-9 February1979. Paper A 79 047-2
Willis, C., Van Wilgen, B. W., Tolhurst, K., Everson, C., D'abreton, P., Pero, L. & Flemming, G. (2001). The development of a
national fire danger rating system for South Africa, Report No. ENV-P5C 2000-073, Division of Water,
Environment of Forestry Technology, CSIR, Pretoria.
Williams, R.P., R.A. Congdon, A.C. Grice and J.P.Clark. (2003). Effect of fire regime on plant abundance in tropical eucalypt
savanna of North Eastern Australia. Austral Ecol. 28. 237-338
Wright, H.A. & Bailey, A.W., 1982. Fire Ecology: United States and Southern Canada. John Wiley & Sons, New York.
WG A2.18, (2003). Life Management Techniques for Power Transformers.Cigre TB227
WG12.15 (2000). (Guide for Customers Specifications for Transformers -100MVA and 100kV and above) (replaced by TB 528,
April 2013). Cigre TB156
WG12.19, (2002). The Short-Circuit Performance of Power Transformers. Cigre TB209

WG23 SC23, (1987). The Mechanical Effects of short circuit currents in Open Air Substations. Cigre TB06
169
170
15. Abbreviations and definitions

As this work covers a number of different disciplines, some readers may find abbreviations and concepts with
which they are not familiar. The list below gives the abbreviations and their meanings.
4 eV. 4 electron Volt
60 Hz = 60 Hertz (frequency)
7 kV r.m.s 7 kilo Volt root mean square
AAAC All Aluminium Alloy Conductor
AAC All Aluminium Conductor
AC Alternating Current
ACSR Aluminium Conductor Steel Reinforced is a type of high-capacity, high-strength stranded conductor
typically used in overhead power lines
ADSS (All-Dielectric Self-Supporting) fibre optic cable is a non-metallic cable which supports its own weight
without the use of lashing wires or messenger cables.
AFIS The Advanced Fire Information System, a satellite-based system tracking fires on the Earth’s surface
AGS units Armour grip suspension units, which use rubber inserts and are used for the attachment of
conductors involving fibre optic cables
ARC Auto re-close of a breaker
ArcIMS (standing for Arc Internet Map Server) is a Web Map Server produced by Esri accessible through a web
browser. It is a GIS that is designed to serve maps across the Internet.
ArcSDE (Spatial Database Engine) is a server-software sub-system (produced and marketed by Esri) that
aims to enable the usage of Relational Database Management Systems for spatial data. The spatial data
may then be used as part of a geodatabase
ASCII American Standard Code for Information Interchange, a widely used standard for encoding text documents
on computers
AVHRR Advanced Very High-Resolution Radiometer sensors flown aboard the TIROS weather satellites
BEHAVE system Fire Behavior Prediction and Fuel Modeling system
CEIDS Consortium for Electric Infrastructure for a Digital Society ()
CESP The Companhia Energética de São Paulo (CESP) is the largest producer of electricity in the state of São
Paulo, with total installed power of 7,455 MW, and the third largest in Brazil. It owns and operates six
hydroelectric plants integrated into the National Interconnected System
CSIR Council for Scientific and Industrial Research in South Africa
CSIRO Commonwealth Scientific and Industrial Research Organisation
DEM Digital Elevation Model
EHV lines Extra High Voltage lines
ENSO The El Niño-Southern Oscillation (ENSO) is a recurring climate pattern involving changes in the
temperature of waters in the central and eastern tropical Pacific Ocean. El Niño and La Niña are the
extreme phases of the ENSO cycle; between these two phases is a third phase called ENSO-neutral
171
EPRI Electric Power Research Institute
Esri Environmental Systems Research Institute) is an international supplier of geographic information system
(GIS) software, web GIS and geodatabase management applications. The company is headquartered in
Redlands, California.
Eumetsat European Organisation for the Exploitation of Meteorological Satellites
FDI Fire Danger Index
FEC Federal Electricity Commission (Comisión Federal de Electricidad or CFE) in Mexico
FFDI Forest Fire Danger Index
FIFPI Fire Induced Flashover Probability Index
FPA Fire Protection Associations
GIS Geographic Information System
GSFC Goddard Space Flight Centre
HIF high impedance fault
HVAC High Voltage Alternating Current
IEEE The Institute of Electrical and Electronic Engineers is the world's largest technical professional organization
dedicated to advancing technology for the benefit of humanity
IREQ Hydro-Québec's research institute.
kWh losses kilo Watt hour losses
KZN KwaZulu Natal, one of the provinces in South Africa
LAPEM Testing Laboratory for Equipment and Materials, Mexico
LIF low impedance fault
McArthur FFDM McArthur Forest Fire Danger Meter () gives a fire danger index that is related to the
chances of a fire starting, its rate of spread, its intensity, and its difficulty of suppression, according to
various combinations of air temperature, relative humidity, wind speed and both the long and short-term
drought effects
MODIS Moderate Resolution Imaging Spectroradiometer
MSG Meteosat Second Generation, a geostationary weather satellite from Eumetsat
MVA MVA stands for Mega Volt Amp or Volts X Amp /1000,000. If your total load requirement is 1000 volts and
5000 amps (1000 x 5000 = 5000,000 VA) it can be expressed as 5MVA. This is called "apparent power"
because it takes into consideration both the resistive load and the reactive load
MVAr reactive power measured in mega voltamperes reactive
NASA North American Space administration
NETFA South Africa's National Electrical Test Facility
NFDRS national fire danger rating system.
O&M operation and maintenance (personnel)
172
OHL overhead lines
OPGW An optical ground wire (also known as an OPGW or, in the IEEE standard, an optical fibre composite
overhead ground wire is a type of that is used in overhead power lines. Such cable combines the functions
of grounding and communications The OPGW cable is run between the tops of high-voltage electricity
pylons.
p=79kPa Pressure =79 kilo Pascal
REN Redes Energéticas Nacionais, SGPS, S.A. (formerly Rede Eléctrica Nacional S.A.) is a Portuguese energy
sector company
RH Relative humidity
ROS Rate of spread (of a fire)
ROW Right-of-way, Easement or Servitude on which a power line is built,
S.O.G. Standard Operating Guideline for an electrical control centre
SAC Satellite Application Centre at Hartebeeshoek South Africa
SEVIRI Spinning Enhanced Visible and Infrared Imager sensor, on-board MSG
SMS Short Message Service or text message on cellular phone
T=20℃, Temperature = 20 degrees Centigrade
TB Technical Brochure
the CP wiring Cathodic protection wiring
VIIRS satellite The Visible Infrared Imaging Radiometer Suite (VIIRS) is a sensor designed and manufactured by
the Raytheon Company on board the Suomi National Polar-orbiting Partnership (Suomi NPP) and NOAA-
20 weather satellites.
173
174
16. Annex
16.1 Comments received about the effect of fires on conductors
Eskom South Africa
Jacques Calitz. Senior Consultant, Research, Testing and Development
Regarding the question, what the effect of fire and especially veld fires will be on the mechanical properties of
overhead conductor, my comments are as follows:
The behaviour of the conductor will be depending on the conductor type, namely
• ACSR – Consist of annealed EC grade aluminium wire (1000 series) and steel
• AAC – Hard drawn EC grade aluminium (1000 series)
• AAAC – “Mechanical” grade aluminium alloy (6000 series)
Although, different grades within a specific series are used by different conductor suppliers to manufacture the
above conductor types, the most comment grades used are:
• EC grade 1100 (1050, 1100 and 1350)
This grade cannot be hardened by heat treatment. It can only be hardened by cold working (strain hardening).To
produce the annealed condition (TO temper), it must be heated to 345°C long enough (about 1 hour) to allow for
thorough heating and then air cooled (cooling rate is not important). , however, the material is sensitive to high
temperatures ranging between 200°C and 250°C and might lose some of its strength depending on the time of this
exposure. Melting range is 645°C to 665°C
• Alloy 6063 (Alloy 6061 similar Temper temp)
To produce the annealed condition (TO Temper) it must be heated to 415°C and held at this temperature for three
hours. Thereafter it must, cool at a rate of 10°C per hour down to 260°C, and then further air cooled.
To produce the annealed condition (T4 Temper Solution Heat Treating) it must be heated to 520°C and then held
at this temperature for one hour. Thereafter it must be quenched in water.
To produce the annealed condition (T6 Temper, Solution Heat Treating plus Aging) age T4 material by reheating to
175°C, hold temperature for six to eight hours, and then cool at any desired rate.
Based on the above, it can be argued that the heat intensity of a veld fire can affect the behaviour of the conductor,
but what needs to be kept in mind is the type of conductor being exposed, the number of strands (layers) and the
time the conductor is exposed.
The reason veld fires are not considered to be a major problem for Eskom from a material point of view is that most
of our lines are built with ACSR conductors in order to get the cost saving associated with building longer spans.
As the aluminium used for ACSR conductors are in the annealed condition (T0), hence any exposure to veld fires
will most possibly result in an increase in strength (heating and fast cooling due to wind). This will result in an
increase in resistance and result in lower electrical performance.
The steel core will not be affected as it is shielded by the aluminium and you need a temperature in the range of
700°C, which will result in the aluminium to melt.
175
In the case of AAAC and AAC, annealing of the aluminium is a possibility and it will depend on the temperature,
time of exposure and the number of layers and diameter of the strands. In short, I don’t think a fast-moving
grass/sugar cane fire will have a significant affect, but fires such as forest fires where the flame is maintained for a
longer period could result in annealing (Over-ageing of the material). This could result in reduced conductor ground
clearance.
Notes:
The temper designation appears as a hyphenated suffix to the basic alloy number. Four basic temper designations
are used for aluminium alloys. They are
• – F: as fabricated;
• – 0: annealed; − Material has undergone an annealing treatment to reduce strength and hardness levels.
This is also known as “dead soft”.
• – H; strain hardened and
• –T: thermally treated. Material has been thermally treated for added strength. The “T” is always followed
by one or more digits. The first digit after the “T” indicates the following:
o –T4: Solution heat treated and naturally aged.
o –T6: Solution heat treated and artificially aged.
• –W: is a fifth designation which is used to describe an as quenched condition between solution heat
treatment and artificial or room temperature aging.
176
16.2 The Breakdown of the insulating properties of a gas

In order to understand the breakdown of the insulating properties of air during a fire, it may be helpful to some
readers to revisit the current knowledge on the processes that lead to a gas losing its insulating properties. In this
chapter the reader will be taken from the basic concepts of the kinetic theory of gases through to the concepts of
Townsend’s coefficients and Meek’s streamer theory, which has resulted in a better understanding of the relatively
more complicated leader type breakdown of large air gaps. The different mechanisms of ionisation and electron
recombination will be presented as well as explaining the phenomena of corona. The latter part of the chapter will
give attention to the leader mechanism and its observed properties.
A number of studies exist which discuss this subject [Loeb & Meek, (1940), Meek, (1969), Vosloo (1992), Wadwa
(2001), Evert (2003).] The work of Sukhnandan (2004) is a relatively modern work with a clear and concise
approach and for the convenience of the reader his work will be presented below.
“Ever since Edison developed the use of electric energy as a means to do work, man has looked for ways to isolate
the two wires required to convey electricity to the place of usage.” (Forster,1982).
In a power distribution network air insulation is widely used making the stability and smooth operation of the power
system dependent upon the performance of the air insulation. A large amount of work has been performed to
investigate, from a physical point of view the phenomenon of dielectric discharges in air gaps subject to high
voltages. The purpose of this chapter is to present an understanding of the basics of electric discharges and air
breakdown. High voltage discharge in air gaps can be categorized into two broad sections: Small air gap
breakdown (streamer mechanism) and large air gap breakdown (leader mechanism). This chapter deals with small
air-gaps first and later progresses into large air-gaps.
Meek’s streamer theory and corona discharges are also described. Scientists till now have been exploring the two
mechanisms in order to understand their operation and to predict their behaviour with the aid of mathematical
models. An important motive of most air insulation research is to determine the insulation strength of air so that the
design of electrical transmission lines and apparatus can be aided to estimate minimum clearances for
transmission line applications (Hutzler et al.,1978).
The limitations encountered in the prediction of air insulation clearances are directly related to the physical
mechanisms of corona effects. Thus, the understanding of the behaviour of air pre-discharges plays a fundamental
role in the refinement of models and in the improvement of electrical system design.
The Kinetic Theory

The fundamental principles of the kinetic theory of gases are pertinent to the study of electrical breakdown and
gaseous ionisation. The results of the kinetic theory need to be examined.
Boyle’s Law
For a given amount of enclosed gas at a constant temperature, the product of pressure and volume is constant.
𝑝𝑉 = 𝐶 = 𝐶𝑜𝑛𝑠𝑡𝑎𝑛𝑡 (88)
also,
𝑣 𝑇 (89)
=
𝑣0 𝑇0
(Gay Lussac’s Law) (2.2)
From these formulae, the equation that describes the state of an ideal gas is formed, namely:
𝑝𝑉 = 𝑛𝑅𝑇 (90)
where:
R = universal gas constant equal to 8.314 joules/K mol
n = number of kilo moles of gas
177
and T = absolute temperature in Kelvin.
An alternate form of this equation is:
𝑝𝑉 = 𝑁𝑘𝑇 (91)
where:
k = universal Boltzman’s constant equal to 1.3804 x10-23 joules/K

Also, the mean energy per molecule of gas is determined by:
3 (92)
𝑤 = 𝑘𝑇
2
The free path l of molecules and electrons

The free path is defined as the distance molecules or particles travel between collisions. The free path is a random
quantity and its mean value depends upon the concentration of particles or the density of the gas.
Collisions and energy transfer

There are two types of collisions between gas particles:
(1) Elastic or simple mechanical collisions in which the energy exchange is always kinetic
(2) Inelastic – in which some of the kinetic energy of the colliding particles is transferred into potential
energy of the struck particle or vice versa.
Townsend’s First Ionisation Coefficient

In air under normal conditions the recombination process counterbalances the creation of free electrons and
positive ions and equilibrium exists (Radtke, et al.,1976). If however the external electric field is sufficiently high the
state of equilibrium will be upset. Townsend studied the gas current variation with respect to changes in the applied
field. The current at first increased proportionally with the applied voltage and then remained nearly constant at a
value 𝑖0 which corresponded to the background current (saturation current). If the cathode was irradiated with
ultraviolet light then the emitted photocurrent was also equal to 𝑖0 When the external voltage was increased further
the current increased above the value of 𝑖0 at an exponential rate. Townsend ascribed this increase in the current
beyond a threshold value to the ionisation of gas due to electron collision. As the field increases, electrons leaving
the cathode are accelerated more and more between collisions until they gain enough energy to cause ionisation
on collision with gas molecules or atoms.
Townsend introduced the first ionisation constant called a defined as the number of electrons produced by an
electron per unit length of path in the direction of the field. If n is equal to the number of electrons at a distance x
from the cathode, the increase in electrons dn in additional distance dx is given by:
𝑑𝑛 = 𝛼𝑛 𝑑𝑥 (93)
Integration over the distance from anode to cathode (d) gives
𝑛 = 𝑛0 𝑒 𝛼𝑑 (94)
Where:
𝑛0 = number of primary electrons generated at the cathode
The above equation becomes:
𝐼 = 𝐼0𝑒 𝛼𝑑 (95)
Where:
I0 = current leaving the cathode.
178
Photoionization
If on collision an electron has energy lower than the ionisation energy of the other colliding body then the collision
may excite the gas atoms to higher states. i.e.:
𝐴 + 𝑒 + 𝐾𝑖𝑛𝑒𝑡𝑖𝑐 𝐸𝑛𝑒𝑟𝑔𝑦 → 𝐴∗ + 𝑒 (96)
𝐴∗ → 𝐴 + ℎ𝑣 (97)
where:
A represents the neutral atom or molecule
𝐴+ represents the positive ion
hv represents the photon energy
The excited atom on recovering from its excited state radiates a quantum of energy or photon (hv) that in turn may
ionise another atom whose ionisation energy is equal to or lower than the photon energy. This process is known as
Photoionization and is represented as:
𝐴 + ℎ𝑣 → 𝐴+ + 𝑒 (98)
Photoionization is a secondary ionisation process and may be acting in the Townsend breakdown mechanism and
is essential in the streamer breakdown mechanism and in some corona discharges. Photo excitation is a process
whereby an atom or molecule just absorbs a photon and is raised to a higher energy level due to the energy in the
photon being not sufficient in order to ionise the atom or molecule.
Ionisation by Interaction of Metastables with Atoms

A metastable state is an excited state that lasts for several seconds. The atoms in this state are referred to as
metastables and have a relatively high potential energy and are able to ionise neutral particles. Consider the
following case:
𝐴𝑚 + B → 𝐴+ + B + e (99)
Where:
𝐴𝑚 represents the metastable
B represents the neutral atom

Other examples of ionisation by interaction of metastables are as follows:
𝐴𝑚 + 𝐴𝑚 → 𝐴+ + 𝐴 + 𝑒 + 𝐾𝐸 (100)
(usually when the density of metastables are high)
𝐴𝑚 + 2𝐴 → 2𝐴∗ + 𝐴; 2𝐴∗ → 𝐴 + 𝐴 + ℎ𝑣 (101)
The photon released in the above reaction may cause ionisation in the cathode rather than the gas molecules due
to its low energy value.
Thermal Ionisation
Thermal Ionisation occurs when atoms or molecules gain sufficiently high velocity to cause ionisation on collision
with other molecules or atoms. This is the main source of ionisation in flames and high-pressure arcs. Work done
by Saha resulted in an equation in order to describe the degree of ionisation 𝜃 in terms of gas pressure and
absolute temperature:
179
3 (102)
𝜃2 1(2𝜋𝑚𝑒 )2 5 −𝑊𝑖
= (𝑘𝑇)2 𝑒 𝑘𝑇
1−𝜃 𝑝ℎ
Where:
p = pressure in torr
𝑊𝑖 = ionisation energy of the gas
k = Boltzman constant
T = absolute temperature
me = mass of electron
h = Plank constant
Deionization by Recombination
Recombination is a process that occurs wherever there are negative and positive ions present. Upon
recombination of these ions a quantum of energy is released. In some cases, a third body may be involved and
may absorb the excess energy released in the recombination. The reactions are represented as follows:
𝐴+ + 𝑒 → 𝐴 + ℎ𝑣 (Radiation) (103)
𝐴+ + 𝑒 → 𝐴𝑚 + ℎ𝑣 (recombination) (104)
𝐴+ + 𝐶 + 𝑒 → 𝐴∗ + 𝐶 → 𝐴 + 𝐶 + ℎ𝑣 (105)
𝐴+ + 𝑒 + 𝑒 → 𝐴∗ + 𝑒 → 𝐴 + 𝑒 + ℎ𝑣 (106)
Deionization by Attachment – Negative Ion Formation

Electronegative gases are gases that are lacking one or two electrons in its outer shell and thus tend to readily
acquire a free electron to form a stable negative ion. For these negative ions to remain stable for some time, the
total energy must be lower than that of an atom in the ground state. The electron affinity is the change in energy
that occurs when one electron is added to an atom in the gaseous state. The following are the processes for
negative ion formation:
 Radiative attachment: Excess energy upon attachment is released as a quantum.
𝐴 + 𝑒 ⟺ 𝐴 + ℎ𝑣 (107)
The captured electron can be released by absorption of a photon (photo detachment). Thus, the process is
reversible.
 Third body collision attachment: Excess energy upon attachment is absorbed by a third body as kinetic
energy.
· 𝑒 + 𝐴 + 𝐵 → 𝐴 + (𝐵 + 𝑊𝑘 ) (108)
where:
𝑊𝑘 = change in kinetic energy of body B
180
 Dissociative attachment: Excess energy is used to separate the molecule into a neutral particle and an
atomic negative ion.
𝑒 + 𝐴𝐵 ⟺ (𝐴𝐵)∗ + ℎ𝑣 ⟺ 𝐴 + 𝐵 (109)
This process is predominant in molecular gases.
 In the above process the molecular ion is at a higher potential level and upon collision with a different
particle this excitation energy may be lost to the colliding particle as potential and/or kinetic energy.
𝑒 + 𝐴𝐵 ⟺ (𝐴𝐵)∗ (110)
(𝐴𝐵 − )∗ + 𝐴 ⟺ (𝐴𝐵)− + 𝐴 + 𝑊𝑘 + 𝑊𝑝 (111)
 There are other processes of negative ion formation; one such process involves the splitting of a molecule
into negative and positive ions upon impact of an electron without attaching the electron.
The process of electron attachment cumulatively describing the removal of electrons by attachment from ionised
gas by any of the above processes may be expressed by a relation analogous to the expression that describes
electron multiplication in a gas. The loss of electron current in a distance dx due to attachment is:
𝑑𝐼 = −𝜂𝐼𝑑𝑥 (112)
Where 𝜂 is the attachment coefficient indicating the number of attachments produced in a path of a single electron
travelling a distance 1cm in the direction of the field.
Thus, for a gap length d and with electron current 𝐼0 starting at the cathode,
𝐼 = 𝐼0 𝑒 −𝜂𝑑 (113)
The number of electrons produced by collision in distance dx is,
𝑑𝑛𝑖 = −𝑛𝛼𝑑𝑥 (114)
At the same time, the number of electrons lost in dx by attachment is
𝑑𝑛𝑎 = −𝑛𝜂𝑑𝑥 (115)
So then the number of free electrons is
𝑑𝑛 = 𝑑𝑛𝑖 + 𝑑𝑛𝑎 = 𝑛(𝛼 − 𝜂)𝑑𝑥 (116)
Integration from zero to x with no electrons starting from the cathode gives the number of electrons at any point x
in the gap as:
𝑛 = 𝑛0 𝑒 (𝛼−𝜂)𝑥 (117)
The steady state current will have two components.
 Electron flow
181
 Negative ion flow
The increase in negative ions in distance dx is,
𝑑𝑛 = 𝑛𝜂𝑑𝑥 = 𝑛0 𝜂𝑒 (𝛼−𝜂)𝑥 𝑑𝑥 (118)
Integration from zero to x gives,

𝑛0 𝜂 (119)
𝑛− = [𝑒 (𝛼−𝜂)𝑥 − 1]
𝛼−𝜂
The total current equals the sum of the two components or
𝑛 + 𝑛− 𝛼 𝜂 (120)
= 𝑒 (𝛼−𝜂)𝑑 −
𝑛0 𝛼−𝜂 𝛼−𝜂
and the current becomes,
𝛼 𝜂 (121)
𝐼 = 𝐼0 [ 𝑒 (𝛼−𝜂)𝑑 − ]
𝛼−𝜂 𝛼−𝜂
𝛼 − 𝜂 is defined as the effective ionisation coefficient.
Mobility and De-ionisation by diffusion

In the presence of an electric field, charged particles in a gas will experience a force causing them to drift with a
velocity that varies directly with the field and inversely with the density of the gas through which it moves.
𝑢 m2 (122)
𝑘= ( sec)
𝐸 v
u is the average drift velocity in field direction and E is the electric field strength. In a region of low values of E/p it
is possible to derive an expression for mobility of ions in a gas since the drift velocity will be small compared to
thermal velocity. An expression for the reduced mobility at 0°C and 760mm Hg is as follows:
𝑝 273 (123)
𝑘0 = 𝑘
760 𝑇
By letting 𝜏 equal the time between two successive collisions,
𝜆̅𝑖 (124)
𝜏=
𝑐̅
where 𝜆𝑖 is the ionic mean free path and c is the mean thermal velocity of the ion. The acceleration a is,
𝑒𝐸 (125)
𝑎=
𝑚
and the distance travelled in time 𝜏 is,
which makes drift velocity,

𝑒𝐸 2 (126)
𝑠=
𝜏
2𝑚
𝑒𝐸 𝑒𝜏 𝑒𝜆𝑖 (127)
𝑢= 𝜏 = ( )𝐸 = ( )𝐸
2𝑚 2𝑚 2𝑚𝑐̅
182
and
𝑢 𝑒𝜆𝑖 (128)
𝑘= =
𝐸 2𝑚𝑐
Taking the statistical distributions of the mean free path into account, the average value of s becomes,
𝑒𝐸 2 (129)
𝑠̅ = 𝜆
𝑚𝑐 −2 𝑖
If mean free time,
𝜆𝑖 (130)
𝜏̅ =
𝑐
Then
𝑒𝐸 −2 (131)
𝑠̅ = 𝜏
𝑚
and the drift velocity,
𝑠̅ 𝑒𝐸 (132)
𝑢= = 𝜆
𝜏 𝑚𝑐̅ 𝑖
and finally,
𝑢 𝑒𝜆̅𝑖 (133)
𝑘= =
𝐸 𝑚𝑐
This shows that when the distribution of free paths is taken into account, k doubles. If the initial velocities of the
particles are taken into account i.e. the velocities of the particles after collision, then k can be defined as follows:
(134)
0.815𝑒𝜆̅ 𝑚 + 𝑀
𝑘= √
𝑚𝑐 𝑚
For thermal equilibrium,
𝑚𝑐𝑖2 𝑚𝑐 2 3 (135)
= = 𝑘𝑇
2 2 2
For the special case of an electron:
̅̅̅𝑒
𝑒𝜆 (136)
𝑘 + 0.815
𝑚𝑐1
The presence of impurities is found to have a significant effect on the measured mobility. This ion and electron
mobility can be used for the determination of conductivity of an ionised gas.
The following is a simple case:
Let the concentration of electrons and positive ions be equal, i.e.:
[𝑛+ ] = [𝑛𝑒 ] = [𝑛] (137)
183
Therefore, the total current density,
𝐽 = 𝐽𝑖 + 𝐽𝑒 = 𝑛(𝑢𝑖 + 𝑢𝑒 )𝑒 (138)
therefore,
𝐽 = 𝑛𝑒𝐸(𝐾𝑒 + 𝐾𝑖 ) (139)
and,
𝐽 (140)
𝜎= = 𝑛(𝐾𝑒 + 𝐾𝑖 )
𝐸
Since 𝐾3 >>𝐾𝑖 , the conductivity is given by,
𝜎 = 𝑛𝐾𝑒 (141)
However, in the presence of appreciable space charge, 𝑛𝑒 ≠ 𝑛𝑖 , the conductivity components must be considered
separately.
Cathode processes – Secondary effects

In a gas, discharge electrons are supplied by the cathode for initiation, sustaining and for completion of the
discharge. The energy required to remove an electron from the surface of a cathode is called the work function and
is specific to that material. The required energy may be supplied in several ways:
Photoelectric emission
If the energy of photons that hit the surface of the material exceeds the work function of the material then electrons
may be ejected from the material. Any excess energy may be transferred into kinetic energy of the electron. The
following relation applies:
𝑚𝑢𝑒2 (142)
= ℎ𝑣 − ℎ𝑣0
2
where:
m = electron mass
𝑢𝑒 = electron velocity
hv = energy of the photon = (Plank’s constant * photon frequency)
h𝑣0 = 𝑊𝑎 , the work function of the material
Electron emission by Positive ion and excited atom impact

Due to bombardment by positive ions or metastables, electrons are emitted from the metal surface. To cause a
secondary emission of an electron the impinging ion must release two electrons, one of which is utilized to
neutralize the ion charge. The minimum energy required for positive ion electron emission is twice the work
function, since the ion is neutralized by one electron and the other electron is ejected.
184
Thermionic Emission
If a metal temperature is increased to some 1500-2500K, the electrons will receive energy from the violent thermal
vibrations. Their energy will sometimes be large enough to cross the surface barrier and leave the metal. The
mission current is related to the temperature.
Field Emission
In this case, the electrons are drawn out of the field by the high electric field at the surface of the metal. An
important effect to mention at this point is the tunnel effect where the strong electric field at the surface of the metal
modifies the potential barrier at the metals surface to such an extent that electrons in the upper level close to the
Fermi level will have a definite probability of passing through the barrier (Kuffel & Zaengl, 1984).
Townsend Second Ionisation Coefficient 𝜸

Townsend, during early measurements of current in parallel plate gaps observed that at higher voltages the current
increased at a more rapid rate than explained by the first mechanism postulated. Townsend thus postulated a
second mechanism. Meek sums it up as follows: “These secondary mechanisms cause the production of additional
electrons from the cathode as a result of the impact on the cathode of positive ions, photons and metastable
molecules. The various secondary processes may be described quantitatively by a secondary ionisation coefficient
𝛾 which is defined as the number of secondary electrons produced at the cathode per ionising collision in the gap.”
(Meek, 1969).
Consider the following:
Let 𝑛0 = the number of electrons liberated from cathode per second. (e.g.: by uv radiation from an external source)
Let 𝑛𝑒 = the number of electrons reaching the anode each second
Let 𝑛𝑔 = the number of electrons released from the cathode per second by secondary mechanisms related to the
production of positive ions in gas.
𝑛𝑒 = (𝑛0 + 𝑛𝛾 )𝑒 𝛼𝑑 (143)
The number of positive ions 𝑛+ produced within the gas per length of path:
𝑛+ = 𝑛𝑒 − (𝑛0 + 𝑛𝛾 ) (144)
and so, for 𝑛𝛾 = 𝛾𝑛+
𝑛𝛾 = 𝛾(𝑛𝑒 − 𝑛0 − 𝑛𝛾 ) (145)
therefore:
𝛾(𝑛𝑒 − 𝑛0 ) (146)
𝑛𝛾 =
(1 + 𝛾)
and:
𝑛0 𝑒 𝛼𝑑 (147)
𝑛𝑒 =
[1 − 𝛾(𝑒 𝛼𝑑𝑑 − 1)]
Thus, the steady state current may be described as:
𝐼0 𝑒 𝛼𝑑 (148)
𝐼=
[1 − 𝛾(𝑒 𝛼𝑑 − 1)]
185
As the voltage applied to the gap is increased the quantity 𝛾𝑒 𝛼𝑑 increases until, when it reaches the value 1, the
denominator becomes zero and the current reaches infinity theoretically. In reality the current is limited by the
limitations of the circuit.
In the 1960’s many studies were conducted by different scientists around the world to examine the variation of the
current in uniform field gaps in different gases. Values for a and g have been determined as functions of electric
field and gas pressure. The influences of the electron attachment and detachment processes have been carefully
studied. Their results confirm the correctness of the basic ideas of the Townsend theory. Although the Townsend
theory gives a consistent account for spark growth and streamer onset in a uniform field it does not give a clear
picture of other aspects of spark development. Problem areas:
 Explanation of short formative times of gaps subject to high over voltages.

 Breakdown of a long gap between a positive point and an earthed plane appears to be independent of
cathode material.
 Cathode effects can hardly be invoked to explain the growth of lightning discharge between negative cloud
and ground.
 The branched and irregular nature of sparks is also difficult to reconcile with the Townsend theory.
Meek’s Streamer Theory

In 1939 Meek put forward a theory known as the streamer theory of spark. This theory has the following ideas:
 Depends primarily on the ionisation processes occurring in the gas.

 Explanation is independent of secondary emissions of electrons from the cathode.
The theory is well presented in (Meek, 1969) where Meek states: “Consider the application of a voltage gradient of
E volts per centimetre across a gap of length d centimetres between parallel plane electrodes in a gas of pressure
p millimetres Hg. If the ratio E/p is sufficiently high, an electron leaving the cathode will ionise the gas molecules
and the additional electron so created will be accelerated in the applied field and so cause further ionisation. When
the original electron has moved a distance of x centimetres in the direction of the applied field, the number of
additional electrons created is 𝑒 𝛼𝑥 . The process is rapidly cumulative and is appropriately named an ‘electron
avalanche’. In a field of the magnitude required to cause breakdown, the electrons travel at a speed of the order of
2 x105 m/s, while the positive ion from which the electron has been detached has a speed of 2 x103 m/s. The
positive ions may therefore be considered as virtually stationary in comparison with the faster moving electrons,
and the avalanche develops across the gap as a cloud of electrons behind which is left a positive ion space charge
in roughly conical channel…”
Figure 122 - Diagram illustrating the distribution of electrons and positive ions in an electron avalanche, redrawn from
(Meek, 1969)
186
Figure 122 shows a pictorial view of the streamer. The space charge contained in the head of the avalanche
produces a distortion in the electric field. The distorted electric field is the greatest at the head of the avalanche
where the ion density is the greatest. A field radial to the direction of the avalanche axis is also produced. When the
tip of the avalanche has reached the anode, the electrons are absorbed into it and the slower moving positive ions
are left in the cone shaped volume extending from the anode to the cathode. The ion presence does not cause
breakdown of the gap. In the gas surrounding the avalanche, photoelectrons are produced by the photons
produced from the densely ionised gas forming the avalanche track.
These electrons initiate auxiliary avalanches which, if the space charge field is 𝐸𝑟 is of the order of the external field
E, will be directed towards the axis of the main avalanche as shown in Figure 123.a.
The greatest occurrence of these avalanches will be where the space charge field supplements the external field.
The positive ions left behind by these avalanches effectively lengthen and intensify the space charge of the main
avalanche in the direction of the cathode. The process continues as a self-propagating streamer as shown in
Figure 123 b. The streamer proceeds across the gap to form a filament of highly ionised gas between the
electrodes as shown inFigure 123 c. This filament formation constitutes breakdown of the gap. The voltage
collapses and the external circuit discharges through the gap.
If a voltage is applied to the gap in excess of the minimum breakdown value, the space charge field Er developed
by the avalanche attains a value of the order of the external field E before the avalanche reaches the anode. In
this case mid gap streamer formation occurs with both cathode directed and anode directed-streamers developing
from the head of the avalanche. The value of the radial field 𝐸𝑟 caused by space charge in the electron avalanche
can be calculated and is given by:
5.3∗10−7 𝛼𝑒 𝛼𝑥 (149)
𝐸𝑟 = 1 volt/cm
𝑥 2
(𝑝)
Where x centimetres is the distance travelled by the avalanche, p millimetres Hg is the gas pressure and a is the
primary coefficient of ionisation by electrons corresponding to the externally applied field E. The transition from an
avalanche to the more rapidly growing positive and negative growing streamers is considered to occur when Er ≅
E” (Meek, 1969).
Figure 123 - Diagram illustrating the transition from an electron avalanche to a streamer and the subsequent growth of
the streamer across the gap (left to right a, b, c) - (redrawn from (Meek, 1969))
187
Rapid breakdown occurs for higher applied voltages due to the streamer transition taking place in the mid gap
region, thus a shorter formative time to produce a spark. The ion multiplication in an avalanche required to cause
the avalanche streamer transition is approximately 108 (Goldman, 1981).
Through investigations scientists have concluded that the minimum breakdown value of gaps up to 10 cm is set by
the Townsend mechanism while the streamer mechanism initiates the breakdown if an over voltage is applied.
Meek states that it is possible to reconcile the Townsend theory and the streamer theory of spark if one considers
that the streamer theory is an elaboration of the Townsend process of ionisation by electrons and secondary
ionisation in the gas.
The Corona Discharge

The word “corona” literally means the disk of light that appears around the sun. However, in the field of high voltage
it refers to the partial discharges that develop in zones of highly concentrated electric fields. The corona discharge
is quite distinct in nature and appearance from the complete breakdown of air gaps between electrodes. The
following is a list of observable effects:
 Visible light
 Audible noise
 Electric current
 Energy loss
 Radio interference
 Mechanical vibrations and
 Chemical reactions producing small amounts of 𝑂3 and nitrogen oxides.
The mechanism of corona discharge depends on the polarity of the voltage applied.
Positive Corona
Positive corona starts off due to ionisation by collision at the anode, which then leads to an electron avalanche.
Positive ions formed due to the ionisation process form an extension to the anode. Secondary generation of
avalanches called onset streamers leads to a glow discharge where numerous onset streamers that are short in
length overlap in space and time. The current through the HV electrode is said to be quasi steady.
Negative Corona
Negative corona is sometimes referred to as Trichel pulse corona that has a rapidly and steadily pulsating mode
(Meek, 1969). The photons from initial avalanches get radiated in all directions. Photoelectrons thus start
subsidiary avalanches. With an increase in applied voltage, the Trichel pulses increase at a repetitive rate up to a
critical level at which the negative corona gets into the steady “negative glow” mode.
AC Corona
The basic difference lies in the periodic change in direction of the applied field and its influences on the residual
space charges left over from the discharge during preceding half cycles (Hileman, 1980). If the applied voltage has
a suitable magnitude, both positive and negative glows and streamer coronas can be observed in each cycle.
The Corona Onset Level

Corona starts at the HV electrode and conductor surface when voltage gradients reach a critical value 𝐸0 .
𝐸0 . depends on:
 Voltage polarity
 Pressure
 Temperature
 Conductor Diameter
 Conductor surface irregularities
188
Large Air-Gap Breakdown

Investigations carried out by Allibone and Meek (1937-38) using a rotating film camera enabled finer details of the
corona and leader stroke growth to be recorded (Meek, 1969). By adjusting circuit conditions, it was found that the
speed of development of the spark could be controlled. An increase in the series resistance causes a decrease in
the speed of the leader stroke. Leader propagation is an important phase in the discharge breakdown of long gaps.
The voltage drops along the leader channel and therefore the flashover voltage is largely governed by the electric
properties of the leader channel (Aleksandrov et.al., 1997)
A typical discharge between a positive high voltage point and an earthed plane (using a 130cm gap as an example)
follows the following observable pattern:
 The discharge is seen to be initiated at the positive point by a leader stroke that grows across the gap to
the negative plane.
 When it reaches the plane a return stroke develops along the filamentary ionised track traced out by the
leader stroke.
 A branch of the leader stroke from the point electrode is retraced by the main stroke of the discharge. The
time interval between the initiation of the leader stroke and the main stroke was observed to be 95𝜇s.
 The initial velocity of the leader stroke was 8 x103 m/s.
 The final velocity of the leader stroke was 2.5 x 104 m/s.
 Total time from the initiation of the leader stroke to the occurrence of the main stroke depends on the gap
length and circuit conditions and varies appreciably between successive discharges. Typical values for a
100cm gap are: 10𝜇s for 1kΩ series resistor and 16𝜇s for a 1000kΩ resistor.
If the set-up was changed and a point electrode was placed on the negative plane then the mechanism of spark is
altered. After the downward growing positive leader travels a part of the way through the gap a negative leader
starts growing upwards from the earthed point. The two leaders then grow towards each other and when they meet
the main stroke occurs. This main stroke follows the track traced out by the two leader strokes.
When a negative impulse voltage is applied to the point electrode distinct differences are noticed. The discharge is
initiated by a negative leader stroke that develops in a series of sharply defined steps from the negative point. Each
step extends the path traced out by the previous step. When the process has developed across more than one fifth
of the gap length, a continuous leader stroke proceeds. While all this is happening a positive leader stroke is
initiated at the other electrode and these two leaders grow to meet in the mid gap region. When the two meet, the
return stroke occurs. The upward growing positive leader stroke and the downward growing negative leader stroke
are branched in the direction of their propagation. A voluminous shower of discharge occurs between them. A
frequent occurrence of mid gap streamers which may be several centimetres in length is an interesting feature of
negative discharges. Meek states that the height above the plane of the junction point of the two leader strokes,
when expressed as a percentage of the gap length, decreases as the gap length increases. If a point electrode is
placed on the earthed positive plane the upward growing positive leader stroke is initiated at the point and its
progress is enhanced, with the result that the junction point with the downward growing negative leader occurs
relatively closer to the high voltage negative point.
Meek has stated the following in (Meek, 1969): “There is, as yet, no complete quantitative theory of corona and
spark breakdown in uniform fields, although various theoretical explanations have been given of some of the
observed features and leader-stroke growth.”
Quantitative criteria based on the Townsend theory and the streamer theory are of limited value to the high voltage
engineer since the results are not sufficiently accurate and can often not be applied at all, because the necessary
physical data are not available.
Pedersen in a paper on calculation of corona starting voltages in non-uniform fields points out that Photoionization
is important to the formation of the streamer (Pedersen et.al., 1993). Pedersen uses an empirical method to
calculate a function from which the breakdown voltage is inferred. This was just another effort by another scientist
to give some sort of calculation to determine the spark breakdown or corona starting voltages.
189
Parameters Affecting the Breakdown Process
16.2.25.1 Distance and Voltage

With long gaps under divergent conditions, the breakdown develops in three main phases (Domens et.al. ,1991),
(Bondiou et. al., 1994):
 First corona
 Leader propagation
 Final jump
Each of the above stages has statistical fluctuations in their parameters. These fluctuating values are combined
together to form the variance in the 50% breakdown voltage and the spread in the time to breakdown for each
voltage level. There is also observable discontinuity in the spread of the 50 % breakdown voltage. Baldo and
Gallimberti (Baldo et. al., 1994), studied the influence of the different phases upon 𝜎𝑈50 (the variance in the 50%
breakdown voltage), 𝑇𝐵 (time to breakdown) and 𝜎𝑇𝐵 (the variance in the time to breakdown), by varying in a wide
range both the gap spacing and the peak value of the applied voltage (Rizk, 1989).
It was found that if the crest voltage and hence the field rate of rise 𝑑𝐸 ⁄𝑑𝑡 is increased, the minimum and the mean
inception time lag together with the dispersion in the distribution decreases. The opposite behaviour was observed
for measured inception fields. It was observed that both the minimum and the mean values of the measured
inception fields decrease with 𝑑𝐸 ⁄𝑑𝑡. Baldo and Gallimberti found that both quantities had the minimum value of
31.2 kV/cm at 𝑑𝐸 ⁄𝑑𝑡 equal to zero (Baldo et al., 1975). The amount of charge injected into the gap at a specific
inception field depends upon the gap length because the streamer path extends into the region where the field
distribution is no more the same.
It was found that the mean value of 𝑇𝐿 (time from leader inception to final jump) together with its standard deviation
𝜎𝑇𝐿 , have a tendency to decrease as the over voltage ratio is increased. The real leader length is proportional to
the time of leader propagation. This means that for a fixed applied voltage, the average value of the velocity and
the standard deviation of the average velocity are independent of gap length.
Baldo and Gallimberti state that the average current associated with the leader propagation is strongly dependent
on both the gap length and the over voltage ratio.
The final jump occurs within 1 or 2 𝜇𝑆 after the arrival of the streamers at the plane. Baldo and Gallimberti have
also proven, just as Gracia and Hutzler did, that the axial length of the leader channel at the transition time to the
final jump and hence the height of the final jump depends strongly on the applied voltage (Cracia & Hutzler as cited
in Sukhnandan, 2004).
“The corresponding time duration of the final jump 𝑇𝑗 is of the order of 20𝜇s and this value together with its
dispersion 𝜎𝑇𝑗 (about 2𝜇s) is practically independent of both the voltage and the gap length. This depends on the
fact that the velocity of the leader grows almost exponentially in time: therefore, the total duration of the final jump
does not vary appreciably with its length.” (Cracia & Hutzler as cited in Sukhnandan, 2004).
16.2.25.2 Effect of Wave Shape Variation

One of the most important parameters affecting the breakdown process is the effect of wave shape variation due to
switching surges of different shapes.
Under this topic one of the main considerations is that the change in the amount of the charge injected into the gap
by pre-discharges generated at the high voltage electrode depends upon the time taken by an impulse to reach its
crest value. Thus, a larger amount of charge is injected for longer times. Since the electric field along the gap is
modified by space charges, this process affects the magnitude of spark over voltage.
The effect of impulse time-to-crest on spark over voltage is noticeable when the 50% spark over voltage is plotted
as a function of time to crest. “The characteristic, shaped like a ‘U’ is called the ‘U-curve’ and is also observed in
most other configurations such as rod-plane, rod-rod or conductor-plane for gaps >1m” (Boutlendji & Allen,1994).
The minimum of this curve represents the most extreme case, since it determines the lowest electrical strength for
190
a given air gap arrangement. (See figure 1 in (Boutlendji & Allen,1994)). The coordinates of this minimum are
called ‘the critical time-to-crest’ and ‘the critical flashover’.
Thione expressed the critical time-to-crest as:
𝑇𝑐𝑟 (𝑐𝑟𝑖𝑡) = (50 − 35[𝑘 − 1]𝑑) 𝜇𝑠 (150)
where:
d = gap spacing in meter

𝑇𝑐𝑟 (𝑐𝑟𝑖𝑡) = critical time-to-crest in 𝜇s, and
k = the gap factor defined by Paris.

The definition and a description of gap factor is given in Appendix 1 of Sukhnandan, 2004).
When the critical time-to-crest is plotted as a function of gap spacing the general trend is that the critical time to
crest increases linearly with gap spacing. It has also been found that for gap spacing ≥ 2m the critical time to crest
for the rod-rod gap is lower than that of the rod plane arrangement. Since the dielectric characteristic of the rod-rod
gap is comparable to that of a conductor-conductor arrangement, the result will indicate that for the conductor-
conductor geometry, the critical time to crest can be lower than for that of a reference rod-plane gap.
By experimentation it has been seen that an increase in air humidity shifts the critical time-to-crest towards lower
values (Boutlendji & Allen,1994). The influence on the breakdown voltage of wave tail time can be summarized as
follows:
“The 10% and the 50% spark over voltages were found to increase with wave tail reduction when the time to crest
is kept more or less constant” (Boutlendji & Allen,1994).
The most crucial geometry used to assess the electrical strength of external insulation is the rod-plane gap
stressed with positive switching impulses. Gallet expressed the minimum breakdown strength of any air-gap
configuration under critical conditions as:
𝑘 ∗ 3400 (151)
𝑉50 (𝑚𝑖𝑛) = 𝑉50 (𝑐𝑟𝑖𝑡) = 𝑘𝑉
8
1+
𝑑
Paris expressed the spark over voltage for the standard switching impulse of 250/2500 𝜇s as (Boutlendji &
Allen,1994).
𝑉50 = 𝑘 ∗ 500𝑑 0.6 (152)
It has been stated by Boutlendji and Allen in (Boutlendji & Allen,1994). that the value of k depends upon the
atmospheric conditions and the minimum of the U-curve and thus the critical value of the spark over voltage
becomes a function of both air density and humidity variations.
For small times to crest, close to the critical value there is no extinguishing of the leader propagation. There are
also no primary dark periods for electrodes with large curvature radii, however at least one primary dark period
exists for electrodes with small curvature radii.
Tests performed on rod plane gaps with impulses having critical time to crest have shown that spark over occur
practically on the impulse crest (Carrara & Thione, 1976). Waters in (Waters, 1981). has stated: “The minimum
spark over voltage arises when the time to breakdown 𝑇𝐵 is equal to the time to crest 𝑇𝑐𝑟 . Under these critical
conditions 𝑇𝐵 is mainly controlled by the leader velocity. Neither gap length nor geometry have any influence upon
this minimum stable velocity.”
191
The whole phenomena of the discharge process present dispersions in each of its stages. The effect of these
dispersions is shown by the existence of a probability function. The lowest impulse that could lead to spark over is
that for which the most favourable conditions occur, i.e. leader progression along shortest path and minimum
continuous leader inception time.
16.2.25.3 Influence of Air Humidity

The influence of air humidity is important when positive streamers are the determinant pre-discharges during the
pre-breakdown process. A humidity coefficient e is defined using the linear characteristic 𝑉50% versus absolute
humidity h as follows:
𝜀 (153)
𝑉ℎ = 𝑉11 [1 + (ℎ − 11)]
100
valid for 0.5 ≤ d ≤ 8m and
5𝑔/𝑚3 ≤ ℎ ≤ 20𝑔/𝑚3
Where:
𝑉ℎ = 50% spark over voltage at humidity h
𝑉11 = 50% spark over voltage at humidity 11g/𝑚3 (corrected to standard air density 𝛿, numbered to unity, when the
pressure and temperature are equal to 1013 mbar and 293K respectively.
The humidity correction factor is then defined as:
𝑉ℎ 𝜀 (154)
𝐾ℎ = =1+ (ℎ − 11)
𝑉11 100
16.2.25.4 Temperature and density effects

The progress of a streamer in air depends upon the availability of negative ions in the atmosphere to provide
successive initiating electrons and also upon the dynamics of negative ion reformation behind the electron head. A
review of data on the dependence of negative ion type and relative concentrations show that there are sharp
dependencies upon temperature.
Allen and Ghaffar have shown that the principal effect of a variation in temperature is caused by the consequent
change of air density (Allen & Ghaffar, 1993), (Allen & Ghaffar, 1995). The specific effects of temperature upon
ionic species, thermal energy of neutrals, etc. are insignificant. They conducted some experiments on the effect of
temperature on the propagation field. When the results were adjusted by a factor(𝛿)1.5, the temperature
dependence of the propagation field is insignificant. Relative air density 𝛿 is defined as:
𝑝 293 (155)
𝛿=( )∗( )
1013 𝑇
Where p is the pressure and T is the temperature in Kelvin.
This phenomenon was constant over the temperature range -14°C to 148°C (259K to 421K)
Properties of the Leader
16.2.26.1 Leader Gradient

The electrical gradient within the leader cloud is one of the most important single parameters governing the spark
over characteristics in highly non-uniform fields. Field probes have played an important role obtaining
measurements of this gradient.
192
A typical example of the leader gradient as given in (Waters, 1981). is in tableTable 19.
Table 19 - Typical values of leader gradient and length (10m rod/plane gap at V50% (critical sparkover))
Leader Length 2 3 4 5
Gradient MV/m 0.5 0.35 0.25 0.18
This shows that the leader tip potential varies slowly during its development.
16.2.26.2 Leader Diameter

Measurement of the leader radius by direct photography is difficult due to the limited resolving power of this
method. Strioscopy and high-speed image recording have overcome this problem.
Ross found that the thermal boundary of a leader channel could be measured with a precision of ± 0.1 mm. By
using this method, it was noticed that the diameter of a given section of the channel increased with time, giving a
tapered appearance to the leader at any instant. The rate of radial expansion was less than 100m/s in all cases
measured (Waters, 1981). This indicated that the gas pressure within the leader channel remained constant at the
ambient value.
With the aid of image converter recording and strioscopy, Gilbert was able to detect the density change within the
leader core and also the generation of a sonic wave associated with the leader (Gilbert & Bastien, 1989).
16.2.26.3 Gas Density Reduction

The expansion of the leader channel and the decreasing axial gradient characteristic are clearly related
parameters. Waters states that the records of the channel show that the neutral particle density within the channel
is a decreasing function of time (Waters, 1981). This reduced density is the basic explanation of how the leader
channel conductivity can be studied.
At atmospheric density, a field of about 3 MV/m is necessary to maintain direct impact ionisation. An expansion of
the leader radius by a factor of 3 will reduce the required field to little over 0.3 MV/m. This value approaches the
average fields available in long gap flashover.
The Final Jump

Gracia and Hutzler used an image converter camera on different gap length rod plane configurations in order to
observe the effect impulse wave shape has on the final jump (Cracia & Hutzler as cited in Sukhnandan, 2004).
From the observed images, it can be seen that as the leader develops into the gap, it is preceded by a corona
cloud effect. The streamers forming the corona have a length depending on the external field strength. The final
jump is said to begin when the streamers reach the negative plane due to the electric field in the unbridged gap
being sufficiently high.
A luminous phenomenon starts from the negative electrode following probably the same path of the streamers and
climbing up to the leader’s head. As soon as the final jump-starts, there is an exponential growth in the leader
speed. The return stroke is said to travel at approximately one fifth the velocity of light (Cracia & Hutzler as cited in
Sukhnandan, 2004).
It was found that the height of the final jump increased with the crest voltage applied to the gap (Cracia & Hutzler
as cited in Sukhnandan, 2004). When the height of the final jump was plotted against the value of the breakdown
voltage, the data fitted well about a straight line having the equation as follows:
𝑈𝐵 = 0.33𝑆 + 0.15 (156)

Where:
S = the height of the final jump in meters, and
𝑈𝐵 = the breakdown voltage in Mega volts.
193
The same dependence was seen for different electrode configurations, different distances and different wave
shapes. This would mean that the final jump height is independent of gap geometries and time to crest. Circuit
parameters also have no influence on the start of the final jump. (Cracia & Hutzler as cited in Sukhnandan, 2004).
found that for an electrode spacing less than 2.4m direct breakdown occurs, i.e. no leader and no final jump. For
gap distances, greater than 2.4m, the curve follows a linear law. This means that the condition D = 10R, where R
is the curvature radius of the hv electrode, usually used to define the transition between direct breakdown and
leader breakdown is not observed.
The duration of the final jump was seen to be constant and had a value of between 15 to 25𝜇s. The mean axial
velocity during the final jump is a function of the applied voltage. The values of the mean electric fields in the
leader, in a first approach, can be estimated to be 1 kV/cm and the mean electric field in the streamers during the
final jump can be approximated to be around 4_kV/cm.
Conclusion
This chapter has taken us from the basic concepts of the kinetic theory of gases through to the concepts of
Townsend’s coefficients and Meek’s streamer theory. This has resulted in a better understanding of the relatively
more complicated leader type breakdown of large air gaps. The different mechanisms of ionisation and electron
recombination have been presented. The phenomena of corona have also been explained. The latter part of the
chapter gave attention to the leader mechanism and its observed properties.
Many scientists and engineers have made great effort to model the leader breakdown mechanism. One such
attempt was made by Rizk (Rizk F A M (2), (1989) which tried to address the following issues:
(a) There exists no formula for continuous leader inception voltage, related to gap length and electrode geometry.
(b) Several useful empirical formulae are available for calculation of spark over voltage of a rod plane gap;
however, each is naturally valid within a certain range of gap spacing and bears little or no physical correlation
to our present-day knowledge of the discharge mechanism.
(c) Critical radius can only be determined from experimental results or empirical formulae.
(d) There appears to be basic contradiction between a widely-used formula for the height of the final jump and the
latest spark over data of very large air gaps at that time
By modelling the space charge in the air gap as a cylinder having its axis on the rod-plane axis, was able to derive
a formula for the electric field strength at any position along the air gap. Using this it was possible to derive an
expression for the leader inception voltage of the rod-plane configuration. Meek also provided an analytical tool to
predict the critical radius of the HV electrode.9Rizk F A M ,1989).
There has been great effort to model the development of the positive spark in long air gaps (Bondiou et. al., 1994),
(Goelian et al.,1997), (Hutzler et al.,1978), (Rizk F A M,1997), (Tobazéon R, (Rizk F A M, 1996). This type of
research is still on-going especially with the advances that are made in computing and imaging resources. Our
study is not concentrated on the mathematical modelling of the mechanisms but rather on the physical attributes of
the fire induced flashover. (Sukhnandan, 2004).
194
16.3 Electrical conductivity in wildfires

The mathematical treatment of this subject is covered in Mphale & Heron (2007) and (2008). The work in (2007) is
presented here for the convenience of the reader.
Conductivity of an un-magnetized ionized gas subjected to alternating electric field is given by Schneider and
Hofmann (1959) as:
∞ (157)
4𝜋𝑒 2 1 𝑓0
𝜎=− {∫ [ 𝑢3 ] 𝑑𝑢 }
3𝑚𝑒 (𝜑𝑒𝑓𝑓 + 𝑖𝜔) 𝜕𝑢
0
where 𝑚𝑒 , 𝑢 and 𝑒, are electron mass, velocity and charge respectively and 𝑓0 , 𝜑𝑒𝑓𝑓 and ω are electron distribution
function, effective collision frequency and propagation cyclic frequency, respectively.
Effective collision frequency is a function of energy of the electrons (ε) and is expressed by Sakao and Sato (1969)
as:
(158)
𝜑𝑒𝑓𝑓 (𝜀) = 𝑣̅ ∑ 𝑛𝑖 𝑄𝑖 (𝜀)
where 𝑛𝑖 and 𝑄𝑖 (𝜀) are number density and electron-neutral momentum transfer collision cross section of
component flame gases. Flame components used in the calculation of effective collision frequency of the alkali
1⁄
8𝑘𝑇 2
seeded fire are those given by Frost (1961). 𝑣̅ = ( ) is the mean electron velocity K is the Boltzmann
𝑚𝑒 𝜋
constant. The electron – neutral collision cross-section in flames is inversely proportional to mean electron velocity,
consequently collision frequency is constant (Schneider and Hofmann, 1959). This simplifies the expression for
electronic conductivity to;
∞ (159)
4𝜋𝑒 2 𝜕𝑓0
𝜎=− {∫ 𝑢3 𝑑𝑢 }
3𝑚𝑒 (𝜑𝑒𝑓𝑓 + 𝑖𝜔) 𝜕𝑢
0
Temperatures of forest fires seldom exceed 1500◦C, therefore free electrons in the flames can be assumed to be
in thermal equilibrium with neutral gases therefore the electron velocity distribution is Maxwellian (e.g. in Schneider
and Hofmann, 1959). Maxwellian velocity distribution for electrons is given as:
𝑚𝑒 −𝑚𝑒 𝑢2 (160)
𝑓0 = 𝑁√ 𝑒𝑥𝑝 ( )
2𝜋𝑘𝑇 2𝑘𝑇
𝑚𝑒
and if one lets α be √( ) 𝑢 then equation (160) becomes;
2𝑘𝑇
8𝑁𝑒 2 ∞
2 (161)
𝜎= {∫ 𝛼 4 𝑒 −𝛼 𝑑𝛼 }
3√𝜋 𝑚𝑒 (𝜑𝑒𝑓𝑓 + 𝑖𝜔) 0
∞ 2 3
But {∫0 𝛼 4 𝑒 −𝛼 𝑑𝛼 } = √𝜋 , therefore (161) becomes
8
𝑁𝑒 2 (162)
𝜎=( )
𝑚𝑒 (𝜑𝑒𝑓𝑓 + 𝑖𝜔)
195
which could be rewritten as:
𝑁𝑒 2 (163)
𝜎=[ 2
] (𝜑𝑒𝑓𝑓 − 𝑖𝜔)
𝑚𝑒 (𝜔 2 + 𝜑𝑒𝑓𝑓 )
An example for typical wildfire conductivity would be when one considers x-band microwaves (frequency = 5 GHz,
ω = 3.142 × 1010 rads−1) illuminating a wildfire with an average electron density (N)=1×1016 mˉ³
Momentum transfer electron-neutral particle collision frequency (φeff ) of the flame gases is around 1×1011 s −1 .
Electrical conductivity as calculated from equation (163) is complex and its value is 0.0026 + 0.0008i.
At low electromagnetic wave propagation frequencies (ω), eq. (163) reduces to:
𝑁𝑒 2 (164)
𝜎=( )
𝑚𝑒 𝜑𝑒𝑓𝑓
If frequencies as low as 50 Hz are considered, equation (163) gives 0.0028 + 0.000i as electric conductivity of the
fire, which is the same value given by equation (164). Thus, conductivity has no imaginary part.
16.4 Calculation example: The safe distance to a discharge of fault current into
the soil
a) RESISTANCE TO GROUND OF THE ARCING POINT

The resistance is given by: Mousa (1990)
 (165)
𝑅𝑔 =
2𝜋𝛼
 = resistivity of the soil
𝛼 = radius of the formed conducting hemisphere.

This radius is a function of magnitude of fault current and is given by: Mousa (1990)
(166)
I
𝛼=√
2𝜋𝐺
Where
𝐺= critical ionisation gradient of the soil. (According to Eriksson (1981) 𝐺=3,0.x105 V/m)
Resistance to ground – as a function of fault current – is given by: -
(167)
G 3𝜌
𝑅𝑔 = √ =√
2𝜋𝐼 2𝜋𝐼
where
𝐼 = magnitude of fault current.
196
b) MAGNITUDE OF THE FAULT CURRENT
The magnitude of the phase-to-ground fault current is given by
3𝑈 (168)
𝐼=
(𝑍0 + 𝑍1 + 𝑍2 + 3𝑅𝑔
Where
Z1, Z2, Z0 – The positive, negative and zero sequence impedances of the system respectively for the fault
at the subject site.
𝑈 phase-to-ground voltage of the power line.
𝑅𝑔 resistance to ground given by equation 3 (impedance of arc is neglected).
The fault level of the system is given by: (Eriksson (1981))
3𝑈 (169)
𝐼0 =
(𝑍1 + 𝑍2 + 𝑍0
The solution of the set of equations i.e., (167), (168) and (169) is given in Eriksson (1981)
The magnitude of fault current in Eriksson (1981) is
2𝐼0 2 (170)
𝐼 = 0.5𝐵[−1 + √1 + ( ) ]
𝐵
Where
𝐼0 3𝛿𝐺 (171)
𝐵=( )
𝑈𝑝 2𝜋
and
𝑈𝑝 = the phase-to-phase voltage.
c) STEP VOLTAGES
National Electrical Safety Codes of many countries permit a step voltage of 263V.
This figure is obtained as follows:
The tolerable step voltage is given by: IEEE Standard 80-1976 (1976).
197
(116 + 0.7𝛿𝑠 ) (172)

𝑉𝑠𝑡𝑒𝑝 =
√𝑡
Where
s – resistivity of the top layer of the soil – in ohm meters.

𝑉𝑠𝑡𝑒𝑝 – tolerable step voltage – in volts.
𝑡 -. duration of fault current – in seconds
In safety calculations the fault duration is usually taken equal to 0,5 seconds [(IEEE Standard 80-1976, Anon,
(1983) and Karady et al. (1996)].
The resistivity of the top layer of the soil is typically taken equal to 100m.
𝑉𝑠𝑡𝑒𝑝 = 263𝑉 (173)
The step voltage generated by fault current (between the man’s feet 1m apart) is given by:
𝐼𝛿 (174)
𝑉𝑠 =
2𝜋(𝑟 + 1)
Where
𝑟 distance from the arcing point
Figure 124 shows the relationship between distance, ionisation current and the “safe” zone.
I
Safe Zone
Conducting radius given by (2)
d
r
1
m
Figure 124 Relationship between distance, ionisation current and the “safe” zone
The relationship is described
𝐼𝛿 (175)
𝑉𝑠 = 𝑉𝑠𝑡𝑒𝑝 =
2𝜋𝑑(𝑑 + 1)
From this equation distance (𝑑) beyond which the step voltage drops to a safe value is determined to be:
198
𝑑 = 0.5[−1 + √1 + (2𝐼𝜌/(𝜋𝑉_𝑠𝑡𝑒𝑝 )) (176)
Where
𝐼 – magnitude of fault current given by (170) and (171)
A calculated example of the above derivation is given below: To calculate the minimum safe distance from
transmission line
Given: 𝑈 = 765kV, 𝑈𝑚𝑎𝑥 = 800kV, Fault level 20kA and Ground resistivity 200m
Magnitude of fault current is (equations(170) and (171)
2 (177)
𝐼 = 0.5𝐵 [−1 + √1 + (2𝐼0/ 𝐵) ]
When
(178)
2 2
10 3𝛿𝐺 10 3.3∗105 𝑣/𝑚𝛿 5 10
𝐵=( ) = =1.4324∗ 10 (𝑈 ) 𝛿
𝑈𝑝 2𝜋 𝑈𝑝 2𝜋 𝑝
(units are V, A, , m)
Answer
B = 17905,
I = 12959,8A
D = 39,1m
199
16.5 Example of a cost analysis by Eskom to acquire cane free servitudes.

This example serves to show how Eskom evaluated all costs incurred by 50 - 100 faults per annum: Costs are
shown in South African Rand (ZAR) 1992
 Equivalent of 2 to 3 extra 3 yearly breaker maintenance per annum R60000

 12-yearly overhead brought forward by 4 years (2 breakers) R10000
 Increased tap-change operations because of line MVAr changes. R12000
Assume 10% increase in tap-change maintenance.
 Conductor damage due to burning:
Total conductor length involved= 205km x 3 (phases) x 2 (twin) 1 230 km
Cost/km of zebra/dinosaur conductor = R10 OOO/km

Life reduction estimated at 30%, i.e. 50-year life reduced to 33 years.
Conductor depreciation increased from R24 600 p.a. to R37 300 p.a., R127 000
i.e. annual cost of R127 000
Annual repair cost as a result of bird-caging and arcing damage R60 000
estimated at R60 000 p.a.
 Additional tower corrosion due to fires.
Cost per tower to paint == R5 000
Number of towers in cane fields; 620.
Cost involved = R3,1 m
Towers painted every 7 years because of corrosion. Rate of

corrosion increased due to heat damage of galvanising.
Annual anti-corrosion cost = R3,1m/7 yr. = R443 000 p.a.
Portion due to tires R10000
 Increased losses due to line switching: Adjacent lines are switched and
using present S.O.G. (Standard Operating Guideline)
Assumptions:
a) Approximately 500 operations p.a.
b) Line lengths average 100km
c) Line loading increases from 30% to 60% of full load
d) Outages average of 30 minutes.
Assume common resistance equivalent to twin zebra is 0.04Ω/km
Increase in kWh losses:
(480² - 240²) *4,04 * 100 *0,5 = 345 600kwh @ 3c/kWh≈ R10000p.a
200
R10000
.
Loss of revenue due to potential loss of supply once per annum (statistically) of 400 MW for 5
minutes R20000
 Cost of operating:
a) Control:
Annual charge from Georgedale = R444 000. During 6-month cane

burning season, 10% of time is dedicated to cane- fire management, R 23 000
i.e.
.
b) operating staff:
Staff travel to site of burning and also to sites of unreported/runaway

fires.
For 60 faults (average)
LABOUR @ R37/hour x 60 x 2hr/trip = R4 440

TRANSPORT: 60 x 100km x 0,6 = R3 600 = R8000
 Communication/administration
(Cane-Fire Management):
Regular visits to cane-growers, correspondence, meetings/ socials, R20000

purchase of gifts, etc.
 Reduction of transformer life:
Transformers have a typical life-span of 50 years (or more). Though faults on

transformers cause a chafing action on the insulation and with the associated
switching surges will cause eventual breakdown. The American IEEE Standard
C57.1 09-1 985 suggests that major through faults on transformers have a
progressive damaging effect on the windings and insulation. Discussions with
transformer specialists on the effects of multiple lower magnitude faults support the
premise that premature ageing and faulting can be ascribed to the adverse faulting
situation as experienced in Natal.
On the assumption of an equivalent 10% average reduction of life from 50 years to

45 years, the annual increased depreciation is approximately
R700000
 The acquisition of servitudes in cane fields will significantly impact the time and
costs associated with patrolling and carrying out other essential maintenance such
as anticorrosion treatment and fault locating.
The constraints involved with access to towers often increases the line restoration
time by an hour or more. Likewise, line-patrol and anti-corrosion treatment costs
could be reduced by at least 10% (labour and transport).
e.g. Of R443 000, ± R20 000 = labour and transport
∴ SAVINGS R20000
201
 Line patrol
Present cost of ± R130 per km reduced to R100 per km

∴ SAVINGS OF R6000
TOTAL R1176000
202

767 - TB

Uploaded by

Copyright:

Available Formats

767 - TB

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

767 - TB

Uploaded by

Copyright:

Available Formats

B2

Vegetation fire characteristics

H.F. VOSLOO, Convenor ZA H. HAWES AU

Executive summary ....................................................................................................5

1. Introduction and Background .............................................................................9

2. Costs associated with fires under power lines ................................................11

3. Danger of fires under OHL to firemen and the public ......................................23

4. Insulation breakdown during fires under overhead lines................................41

5. Proposed model for the breakdown mechanism in a fire................................59

7. Fire behaviour ....................................................................................................93

8. Fire climate and fire weather ...........................................................................115

9. Fire danger rating systems..............................................................................127

10. Predicting fire induced flashovers ...............................................................135

11. Fire tracking systems ....................................................................................143

12. Mitigation of fires under power lines ............................................................153

12.11 Fire risk mapping ...................................................................................................................... 159

13. Conclusion .....................................................................................................161

14. References .....................................................................................................163

15. Abbreviations and definitions .......................................................................171

1. Introduction and Background

2. Costs associated with fires under power lines

2.2 Costs associated with loss of supply

2.3 The effect of fires on overhead conductors

 type of conductor being exposed,

Figure 3 - Annealed and broken copper conductor (provided by Transend, 2014)

2.4 The effect of fires on Tower structures (steel and wood)

2.5 Fire damage to other equipment on lines

General observations from Respondents

Figure 8 - Fire damage to plastic joint boxes (Powerlink 2013)

Fire impacts on optic fibre installations

2.6 Fire effects on substations,transformers and switchgear

3. Danger of fires under OHL to firemen and the public

Table 1 - Likely discharge paths (Bologna & Sadurski, 2001)

Line Voltage Flashover Path *

*Flashover paths are shown in Figure 13.

In their report, Bologna & Sadurski, (2001) found as follows:

1 : Phase to Phase fault

Figure 13 - Flashover Paths on Overhead lines (Bologna & Sadurski, 2001)

3.3 Step and touch potential

3.4 Minimum safe working distance from overhead line

3.5 Calculating the horizontal distance to a point of discharge under windy

Absence of high Area of fire with

for 400 kV where:

UMAX - Maximum line voltage (line-ground)

Maximum Line Maximum Distance between outer

145* 7.6 6.8

Figure 19 The generation of step potentials

3.6 Resistance to ground of the Arcing Point

The resistance of a thin hemi-spherical layer is given by

 resistivity of the soil

 radius of the Ionized zone.

And the corresponding voltage drop is given by

Thus, the voltage gradient is given

The radius of the formed conducting hemisphere is determined by:

I = magnitude of fault current.

3.7 Magnitude of the fault current

From Figure 20 it follows:

EN Phase-to-ground voltage of the power line.

RG Resistance to ground given by (9) when impedance of arc is neglected.

Figure 20 - Calculating the Fault Current (Mousa, 1990)