The trench effect and eruptive wildfires: lessons from the Kings Cross

Underground disaster.
Jason J. Sharples1,2, A. Malcolm Gill3 and John W. Dold4
1.
2.
3.
4.

ACT Rural Fire Service.

Applied and Industrial Mathematics Research Group, School of Physical,
Environmental and Physical Sciences, University of New South Wales at the Australian
Defence Force Academy, Canberra, ACT 2600, Australia.
Fenner School of Environment and Society, Australian National University, Canberra
ACT 0200, Australia.
Fire Research Centre, School of Mathematics, University of Manchester
Manchester, M13 9PL, United Kingdom.
Abstract

Fire eruption (also known as blow-up or flare-up) is a very dangerous aspect of wildland
fire behaviour. It poses a significant threat to the safety of wildland fire-fighters and
complicates the broader problem of wildfire risk management in complex terrain. Despite
the seriousness of eruptive fire events, operational wildland fire spread models generally do
not account for the possibility of these events occurring and there is little in the literature
devoted to obtaining a model that consistently describes the phenomenon. To address this
situation international researchers have recently proposed models that account for eruptive
fire behaviour as a feedback process, with possible mediation by other factors. One such
factor is flame attachment to steep slopes.
The concept of flame attachment is also of interest in the structural fire-fighting context.
For example, it is a key component of the so-called trench effect which is a well
established mechanism for fire propagation on enclosed slopes such as escalators or
stairwells in building structures. Studies into structural fires have revealed that the trench
effect can produce extreme fire behaviour and rapid fire spread, as was apparent in the 1987
Kings Cross (London) Underground fire disaster, for example. Despite the well known
implications of the trench effect in the structural fire context it has received little attention
in the wildfire context.
We therefore consider the trench effect in a wildfire context as a possible trigger for
eruptive fire behaviour. In particular we consider the findings of a number of inquiries into
fatal and near-fatal incidents apparently involving eruptive fire behaviour and discuss the
possible involvement of a landscape-scale trench effect in these incidents.

1. Introduction
It is well known that a wildfire will spread more rapidly up a slope than it will on flat terrain
(McArthur, 1967; Van Wagner, 1977; Noble et al, 1980). Indeed, it is common lore amongst
Australian bushfire-fighters that the rate of spread of a fire will double for every 10 increase in
topographic slope. This rule-of-thumb is well entrenched in wildland fire-fighting training materials
where, for example, it is included on the back of the circular slide rules used to calculate expected
fire behaviour characteristics based on the McArthur Forest Fire Danger Rating System (McArthur,
1967). Implicit in such rules is the notion that constant environmental and topographic conditions
(e.g. fuel, weather and slope) will result in a roughly constant rate of fire spread: significant changes
in rate of spread arise only through changes in the underlying conditions. In particular, on a slope of
constant gradient a fire is expected to spread at a well-defined, constant rate of spread. However,
there are now many documented cases from around the globe where wildfires have violated this
expectation (Viegas, 2009).

Analyses of fire behaviour in both laboratory and wildfire settings have indicated that in some
circumstances a fire can accelerate significantly up a slope of constant gradient (Viegas, 2006). This
type of fire behaviour has recently been termed eruptive, though blow-up or flare-up may have
been traditionally used to describe the same phenomenon. In the wildland fire setting, instances of
eruptive fire behaviour pose significant dangers and are commonly associated with serious injuries
to fire-fighters. In the worst cases they have resulted in multiple fatalities.
Understanding the physical mechanisms that result in eruptive fire behaviour is thus an important
aspect of wildfire risk management, but until recently this problem had received little concerted
attention amongst wildfire researchers. While a number of models have been devised to
successfully account for eruptive fire behaviour (Viegas, 2005; Dold & Zinoviev, 2009), they are
not widely known amongst fire management practitioners, nor are they definitive in explaining the
phenomenon. For example, a majority of eruptive fire incidents have been observed to occur in
connection with steep slopes or canyons (Viegas, 2006), yet topographic attributes have not been
explicitly included in eruptive fire behaviour models. Instead topographic effects are considered a
priori to modulate the initial conditions required as input to these models (Viegas, 2009; personal
communication).
The intent of this article is to discuss the possible role that topography has in the incidence of
eruptive fire behaviour. In particular, the trench effect is introduced and discussed as a potential
trigger for eruptive behaviour of wildfires. The trench effect refers to a specific interaction
between the buoyant plume of a fire and an inclined trench, and became known through
investigation into the tragic Kings Cross Underground fire in London, 1987 (Crossland, 1992).
This fire, which started as an innocuously small fire in an escalator trench, took fire-fighters by
surprise as it rapidly escalated in intensity and ultimately claimed the lives of 31 people and caused
extensive damage to the Underground railway station.
In what follows the characteristics of fire behaviour driven by the trench effect are described along
with some of the most important findings of the Kings Cross fire investigation. Implications of the
trench effect for wildland fire behaviour are then discussed. In particular, two serious burn-over
incidents, one in Australia and one in the United States, are considered with the trench effect in
mind. It is hoped that this will stimulate interest, discussion and better collaboration amongst
Australian fire researchers and practitioners, and promote better communication between the
structural/urban and wildland fire-fighting fraternities.
2. Wildland fires and topographic slope
Topographic slope has a strong effect on the rate of spread of a wildfire. The slope of the terrain
essentially brings the ground, and hence the fuel upon it, into closer proximity with the flames
ahead of the fire. This in turn extends the preheating range and allows for faster rates of spread. In
practice there are a number of methods for incorporating the effects of topographic slope into
wildfire rate of spread calculations (Sharples, 2008). Of particular importance in the Australian
context is the method introduced by McArthur (1967). In his fire behaviour guide McArthur noted
that the rate of spread of a fire doubles for each 10 increase in slope. This can be expressed
mathematically as (Noble et al., 1980):

RS = R0 exp(0.069 S )

(1)

where RS is the slope-induced rate of spread, R0 is the corresponding rate of spread on level ground
and S is the topographic slope in degrees.
While equation (1) is accurate in expressing McArthurs ideas, it may not be the best formulation
for describing the effect of slope on rate of spread. In fact, it indicates that the rate of spread will
increase to an implausible extent as topographic slope increases to 90. This is inconsequential for
practical use of the model within an appropriate data domain but suggests that the model should be

considered with some caution when applied to steep slopes. Indeed, McArthur (1967) only
considers slopes below 20. Despite its limitations, equation (1) is the most common method
currently used in fire behaviour calculations in Australia and, in the absence of more appropriate
methods, is often used to account for fire behaviour on slopes exceeding 20 (e.g. Newnham et al.,
2007).
3. Eruptive fire behaviour
In a series of articles (Viegas, 2002, 2004; 2006; Viegas and Pita, 2004; Viegas et al., 2005; Viegas
et al., 2006) Viegas and his co-workers demonstrate and discuss the variable nature of fire spread,
even when wind and topographic conditions are unchanging. Such behaviour contradicts traditional
slope correction methods such as equation (1), which imply that a fire on a uniform slope will
spread at a constant rate, assuming homogeneity in all other factors (Sharples, 2008). Of particular
concern in the context of fire-fighter safety is the phenomenon of eruptive fire behaviour, whereby a
fire can exhibit a continually increasing rate of spread, with associated increases in intensity, flame
height and flame depth. Eruptive fire behaviour has been linked with many fatal incidents in the
past, the majority of which occur in connection with steep slopes and canyons (Viegas et al., 2005;
Viegas et al., 2006; Viegas, 2009).
Viegas and Pita (2004) provided confirmation of the existence of eruptive fire behaviour in the
laboratory by considering a number of symmetrical canyon-like configurations. The experimental
results indicated an apparent dichotomy in the observed rates of spread. For canyons inclined at 0
and 10 the rates of spread remained relatively constant, with only small variations over time. On
the other hand, for canyons inclined at 30 and 40 the fire fronts were observed to accelerate
significantly. For inclinations of 20 different dynamic behaviour was observed depending on how
closed the canyon was; for more closed canyons a distinct acceleration in spread was observed,
while for fairly open canyons the acceleration was not as pronounced.
Viegas (2005; 2006) introduced and discussed a mathematical model designed to account for
eruptive fire behaviour. In this model heat produced by the flames induces air movement that acts to
intensify the combustion process and the rate of spread. The increase in intensity further enhances
the induced air movement leading to a feedback effect, which results in accelerated fire growth.
Dold & Zinoviev (2009) also derived a mathematical model that predicts eruptive fire spread under
certain circumstances. In this model, the predicted rate of spread is critically dependent on a
parameter . For < 1 the model predicts a stable, constant rate of spread, while for 1 the model
can exhibit accelerated fire growth. Dold & Zinoviev (2009) argue that the case 1 corresponds
to a situation where the buoyant fire plume attaches to the surface over which the fire is spreading.
The hypothesised role of plume attachment was supported in a series of laboratory experiments and
also in a field experiment.
While the models of Viegas and Dold & Zinoviev do both predict eruptive fire behaviour under
certain circumstances, they do not have an explicit dependence on topographic attributes such as
slope. As such they do not provide any direct explanation as to why incidences of eruptive fire
behaviour are so commonly observed to occur on steep slopes or in canyons. Similarly they do not
offer any direct insight into why the geometric properties of the terrain, such as how closed a
canyon is, should have any effect on the incidence of eruptive fire spread.
4. The Kings Cross Underground fire and the trench effect
The Kings Cross Underground (railway) fire occurred on the 18th of November 1987. It exhibited
behaviour that was both unexpected and disastrous, resulting in 31 deaths and the grievous injury of
many people including commuters, public transport workers and fire-fighters. The fire had ignited
within a wooden escalator trench and burnt for approximately 15 minutes, during which time firefighters had assessed its size and behaviour and concluded that although it was a significant fire,

there was nothing to suggest that it would rapidly develop (Crossland, 1992). However, within an
unexpectedly short amount of time the fire spread with extreme ferocity up the escalator trench and
into the ticket hall and surrounding areas with tragic consequences. Photographs taken immediately
after the incident attest to the extreme intensity of the fire. For example, see the photograph at
http://en.wikipedia.org/wiki/File:Kings_Cross_Fire1.jpg
Investigations into the Kings Cross fire revealed that the extreme fire growth was primarily caused
by a distinct phenomenon, which caused the flames and combustion products to be confined and
concentrated within the escalator trench below the balustrades. This phenomenon was called the
trench effect. The scientific aspect of the investigation involved both experimental and theoretical
research into fire spread in trench-like configurations. The experimental work focused on scale
models of the escalator trench (Moodie & Jagger, 1992; Drysdale et al., 1992), while the theoretical
work mainly concerned computer simulations of the flow of hot gases within the escalator shaft
(Simcox et al., 1992).
Experiments using scale models of the Kings Cross escalator shaft and ticket hall found that the
flame front from a fully developed fire across the escalator trench remained virtually within the
channel and spread upwards at an exponentially increasing rate. As in the Kings Cross fire a
sustained jet of flame was observed to emerge from the top of the channel. As such the experiments
confirmed the existence of a mechanism for accelerated fire spread that occurs primarily within the
confines of a trench. Subsequent experiments found similar effects in various trench configurations,
even in trenches with only one side-wall (Drysdale et al., 1992). Similarly, the main feature of the
theoretical simulations was the way that the hot gases in the buoyant plume lay along the trench
floor.
Subsequent experiments and simulations with trenches of various inclinations and geometries found
that flames would attach to a surface inclined above a critical gradient of around 24-27 (Drysdale
& Macmillan, 1992; Smith, 1992; Wu and Drysdale, 1996; Wu et al., 2000). The effect was found
to be independent of the amount of heat released and surface conditions, and was observed even
when the trench side-walls were removed. Thus, the term trench effect is actually a misnomer;
while a trench-like configuration will cause the particular behaviour more readily, the essential
requirements are a slope of sufficient gradient and the inhibition of lateral entrainment of air into
the plume. This can occur with long line fires on flat, inclined surfaces or by restricting lateral
entrainment with a trench-like configuration (Smith, 1992). This suggests that the terms plume
attachment or flame attachment are perhaps more appropriate to describe the phenomenon.
Simcox et al. (1992) concluded that flame attachment is the result of two separate effects: a
significant chimney (Venturi) effect caused by the heat source, and the Coand effect, which causes
the buoyant plume to attach to the floor of the trench. Once flame attachment is established the
enhanced preheating and pyrolysis of the fuel above the fire leads to accelerated fire spread. The
Coand effect is a response to the pressure differential induced by differences in the capacity for
entrainment of air upslope and down-slope of the fire plume. It is named after aerodynamics pioneer
Henri Coand (1886-1972), who patented a number of devices based on the effect, but appears to
have been known of much earlier (Young, 1800).
5. Flame attachment and wildfires: two case-studies
The trench effect, or more properly flame attachment, arises through the interaction between a fires
buoyant plume and the terrain geometry. The physics underlying the phenomenon applies over a
wide range of geometries and scales. As such it is natural to consider flame attachment as a
potential trigger for eruptive fire behaviour in landscape-scale wildfires. Indeed, such a
consideration is entirely consistent with the observations and analyses of Dold & Zinoviev (2009).
Moreover, the fact that confined slopes of over about 25 are the most prone to flame attachment
provides a physical rationale as to why eruptive wildfire behaviour is more prevalent on steep

(b)

(a)

(c)
Figure 1. Wildfires burning on steep slopes and in montane forest fuels (approx. 30m tall eucalypts) to the
west of Canberra on 18 January 2003. The slopes in (a) and (b) are approximately 30o and the slope in
panel (c) is approximately 40o. Photographs taken by Stephen Wilkes (NSW Rural Fire Service Air
Observer).

slopes and in canyons. In this section this rationale will be applied to two case-studies of burn-over
incidents. In particular, the similarities between the extreme fire behaviour resulting in the burnovers and that observed in the Kings Cross fire will be discussed.
Before discussing the case-studies, it is interesting to note that Rothermel (1985) mentions the
possibility of flame attachment in a wildland setting. In particular he refers to steep slopes over
50%. It is remarkable that a slope of 50% equates to 26.5, which is about the same figure derived
from the investigations into the Kings Cross fire. It is also remarkable that Rothermel reported this
figure two years before the Kings Cross disaster and suggests that the wildfire and structural/urban
fire-fighting fraternities might benefit from closer communication and collaboration on research
matters. Moreover, Van Wagner (1977) notes that on slopes of 60-70% (30-35) flames tend to
bathe the slope directly, leading to intense and unstable fire behaviour, while Cheney and Sullivan
(1997) note that under very strong winds the convection column may not lift away from the surface
of steep slopes and result in very high rates of spread. Cheney and Sullivan (1997) make no mention
of the possibility of such an occurrence under conditions of light or no wind, however. Alexander et
al. (2009) also mention the possible role that flame attachment played in the development of the
escape fire used by fire-fighters to survive the Mann Gulch fire.
There is also strong evidence of flame attachment playing a role in the extreme fire behaviour
during the 2003 Canberra fires and the recent Black Saturday fires of 7 February 2009. Figure 1

shows photographs taken by Stephen Wilkes, an air observer with NSW Rural Fire Service. Figures
1a and 1b show entire slopes involved with flame with very little smoke directly above them.
Similarly, figure 1c shows flames on a steep slope that appear to be attached to the terrainvegetation surface. The slopes depicted in the photographs are approximately 30-40. Photographs
taken on 7 February 2009 also indicate flame attachment on slopes of around 20-30. For example,
see the photograph at:
http://www.worldweatherpost.com/2009/02/13/greenies-blamed-for-victoria-bushfires-scale/
5.1 Tuolomne Fire, California 2004.
The Tuolumne Fire occurred on 12 September 2004 in California. A Helitack crew from the
California Department of Forestry and Fire Protection were tasked with constructing a control line
with hand tools near the back of the fire. The terrain was very steep, with slopes of 80-120%
(approx. 38-50). During the initial stages of control line construction fire behaviour was reported
as relatively benign: the crew was working near a portion of the fireline that was backing slowly
across-slope against a steady, light up-canyon wind with surface flame lengths of 6 to 12 inches.
However, with an abrupt wind shift of approximately 90 the rate of spread and flame lengths
increased dramatically and the fire rapidly burnt upslope, emitting a loud roaring sound and overrunning the crew within 8-30 seconds (Cole and Edgar, 2005). Three fire-fighters sustained minor
burns and other injuries, while one fire-fighter was entrapped and killed.
The report into the burn-over incident indicates that, as in other fatal wildfire accidents, an
unexpected wind shift was a primary causal factor. The wind shift was supposed to be part of a
localised weather pattern created by the interaction of the ambient up-canyon winds with the
complex topography, but no similar wind shift was observed for over thirty minutes before the
burn-over, nor was one observed in the remainder of the day or in the next several days. Cole and
Edgar (2005) concluded that the cause of the wind shift was not apparent.
As part of the investigation into this incident an analysis of the fire behaviour using the BEHAVE
system (Andrews, 1986) was conducted. The analysis indicated that with the assumption of a
moderate increase in wind strength associated with the change in direction, the fire would have
exhibited a rate of spread of approximately 2-3 km h-1. At this rate, one would expect to be able to
outrun the fire, even after considering the effects of a 40 slope on maximum running speed
(Minetti et al., 2002). It therefore seems likely that the fire was propagating much faster than the
modelling suggests.
Testimony of fire-fighters that survived the incident appears to favour fire spread driven by
attachment of the buoyant plume to the terrain-vegetation surface. Indeed, one fire-fighter describes
seeing the fire sheeting in association with the wind change. Specifically, the sheeting behaviour
was described as fire about 8 to 12 inches tall and 10 to 15 feet deep, spreading 10 to 15 feet into
the green with intense heat. Another fire-fighter describes feeling heat at his back as he scrambled
up the hill and rolled over the lip of the road within the flame front. The incident occurred at a
location with a slope of approximately 42, which is sufficient to enable flame attachment via the
Coand effect, even without the shift in wind. In fact in this light it seems plausible that the upslope
wind, which unexpectedly occurred and is cited as the cause of the escalation in fire behaviour, was
actually the flow within the buoyant plume itself, attached to the terrain surface. Based on the
evidence presented in Cole and Edgar (2005) it is difficult to rule out this possibility.
5.2 Mansfield Fire, Victoria 2006
A more recent example, in which fire eruption driven by flame attachment may have been a
contributory factor, occurred during the Great Divide Fires of 2006/07 in Australia, southeast of
Mansfield, Victoria. A team of New Zealand fire-fighters assisting in the suppression efforts were
constructing control lines just above a side-cut road near the top of a hill, with a 25 slope above
them and a 30-36 slope below. Both slopes contained unburnt fuel but weather conditions were

benign and observed fire behaviour was relatively mild (Newnham et al., 2007). The majority of the
crew felt comfortable with their position. However, within seconds the fire-fighters became aware
that there was an escalation in fire behaviour in the gully below them and that they were in danger.
Witnesses reported hearing the fire before seeing it, with the sound described as the roar of a jet
engine or freight train. Those members of the crew that didnt escape in time were burnt-over, with
eleven fire-fighters suffering burns and other injuries.
An investigation into the burn-over incident (Newnham et al., 2007) concluded that the escalation in
fire behaviour was mainly due to the effects of slope on the rate of spread of the fire, as described
by equation (1). The predicted rate of spread on level ground was estimated as 0.27 km h-1, and so
equation (1) suggests that on a 35 slope the fire would have been travelling at a little over 3 km h-1.
This implies that the fire would have traversed the 180 m slope in approximately 3.5 minutes.
However, given that the length of the road impacted by the fire was around 250-300m with a
maximum incline of less than 10, and that the fire-fighters were physically fit, they should have
been able to reach safety within 3.0 minutes1, or even less given that they would have been running.
This combined with the details of the burn-over incident contained in Newnham et al. (2007)
indicates that the fire spread much more quickly than the steady rate obtained from equation (1).
Indeed, it seems unlikely that experienced remote area fire-fighters would be surprised by a fire
spreading up a slope at 3 km h-1, which corresponds to a moderate walking pace. Eruptive fire
spread induced by flame attachment again offers a plausible explanation for the rapid spread
upslope. Even without any wind the slope of 30-36 would enable the effect.
6. Discussion and Conclusions
The characteristics of fire behaviour driven by the trench effect, such as in the Kings Cross
escalator fire, bear a strong resemblance to the fire behaviour described in many blow-up fires and
burn-over incidents. In particular, the rapid rates of spread and blasts of superheated air described in
burn-overs such as those discussed above and in other fatal incidents (for example, see Butler et al.,
1998; Viegas, 2009) are consistent with the results of experimental and numerical work that lead to
the discovery of the trench effect or flame attachment phenomenon. Flame attachment driven by
the Coand effect on the buoyant plume applies at a multitude of scales and so it is entirely
plausible that such an effect could play a role in these eruptive occurrences. Indeed, the fact that
eruptive fire incidents tend to occur almost exclusively in landforms that satisfy the geometric
prerequisites for flame attachment, supports the hypothesis that it is a key aspect of eruptive fire
behaviour in these cases. Furthermore, the rapidity of the transition from a relatively low-intensity,
slow spreading fire to a high-intensity, fast moving fire that is associated with the flame attachment
phenomenon is consistent with the fact that in many burn-over incidents experienced fire-fighters
are caught as if by surprise, with little time to reach safety. The sudden upslope wind change that
is often reported as a main causal factor in such occurrences is also consistent with what would be
experienced in the initial stages of plume attachment.
The presence of various types of vegetation, the intricacies of the terrain and the impacts of highly
variable winds in a wildland setting will undoubtedly affect whether or not flame attachment, and
hence eruptive fire behaviour, can occur. For example, in cases where the topography doesnt quite
possess the required geometric properties, the conditions necessary for flame attachment might be
met with the assistance of relatively mild upslope winds. Fuel structure can also be expected to have

A standard fitness test for fire-fighters in Australia requires them to carry a 20 kg weight over 4.8km within
45 minutes (without running). This implies an average walking pace of 6.4 km h-1. Moreover, remote area
fire-fighters are required to complete the 4.8 km in 40 minutes or less, implying an average walking pace of
7.2 km h-1.

an influence on any resultant fire behaviour (Gould et al., 2008). Further research is clearly
indicated and will be pursued by the authors.
Acknowledgements
The authors would like to thank Prof. Domingos Viegas, Mr. Rick McRae and Assoc. Prof. Rodney Weber
for illuminating discussion on the topic at hand. J.J. Sharples is grateful for the financial support of the ACT
Rural Fire Service. This work was supported by the Australian Academy of Sciences Scientific Visits to
Europe Program 2010-2011.

References
Alexander, M.E., Ackerman, M.Y., Baxter, G.J. (2009) An analysis of Dodges escape fire on the 1949 Mann
Gulch fire in terms of a survival zone for wildland fire fighters. Paper presented at the 10th Wildland
Fire Safety Summit, International Association of Wildland Fire, Phoenix, Arizona, 27-30 April 2009.
Available from:
http://fire.feric.ca/36702008/AnalysisDodgeEscapeFireon1949MannGulchFireSurvivalZone.pdf
Andrews, P.L. (1986) BEHAVE: Fire behavior prediction and fuel modeling system-BURN subsystem, Part 1,
General Technical Report INT-194. Ogden, UT: U.S. Department of Agriculture, Forest Service,
Intermountain Research Station. 130p.
Atkinson, G.T., Drysdale, D.D., Wu, Y. (1995) Fire driven flow in an inclined trench. Fire Safety Journal 25,
141-158.
Butler, B.W., Bartlette, R.A., Bradshaw, L.S., Cohen, J.D., Andrews, P.L., Putnam, T., Mangan, R.J. (1998)
Fire behavior associated with the 1994 South Canyon Fire on Storm King Mountain, Colorado. USDA
Forest Service, Rocky Mountain Research Station, Ogden, Utah. Research Paper RMRS-RP-9. 82pp.
Cheney, P., Sullivan, A. (1997) Grassfires: fuel, weather and fire behaviour. CSIRO Publishing.
Cole, M., Edgar, B. (2005) CDF Helitack Crew 404 Burnover Accident Investigation Report. Tuolumne Fire
CA-STF-2191 Stanislaus National Forest, September 12 2004. United States Department of
Agriculture Forest Service and California Department of Forestry and Fire Protection.
Crossland, B. (1992) The Kings Cross Underground Fire and the setting up of the investigation. Fire Safety
Journal 18, 3-11.
Dold , J.W., Zinoviev, A. (2009) Fire eruption through intensity and spread-rate interaction mediated by flow
attachment. Combustion Theory and Modelling, 13(5), 763793.
Drysdale, D.D., Macmillan, A.J.R. (1992) Flame spread on inclined surfaces. Fire Safety Journal 18, 245254.
Drysdale, D.D., Macmillan, A.J.R., Shilitto, D. (1992) The Kings Cross Fire: Experimental verification of
the trench effect. Fire Safety Journal 18, 75-82.
Gould, J.S., McCaw, W.L., Cheney, N.P., Ellis, P.F., Knight, I.K., Sullivan, A.L. (2008) Project Vesta Fire in Dry Eucalypt Forest: fuel structure, dynamics and fire behaviour. Department of
Environment and Conservation, Western Australia.
McArthur, A.G. (1967) Fire behaviour in eucalypt forests. Commonwealth Forestry and Timber Bureau
Leaflet 107.
Minetti, A.E., Moia, C, Roi, G.S., Susta, D., Ferretti, G. (2002) Energy cost of walking and running at
extreme uphill and downhill slopes. Journal of Applied Physiology, 93, 1039-1046.
Moodie, K, Jagger, S.F. (1992) The Kings Cross Fire: Results and analysis from the scale model tests. Fire
Safety Journal 18, 83-103.
Newnham, R., McKay, R., Rasmussen, J., May, R. (2007) Report of the investigation of the Mansfield
burnover incident on Saturday 16th December 2006. A joint report by the Department of Sustainability
and Environment, Country Fire Authority and New Zealand Rural Fire Authority Investigation Team.
(Can be obtained through http://nrfa.fire.org.nz).
Noble, I.R., Bary, G.A.V. and Gill, A.M. (1980). McArthur's fire-danger meters expressed as equations.
Australian Journal of Ecology 5, 20l-203.
Rothermel, R.C. (1985) Fire behaviour considerations for aerial ignition. In Mutch, R.W. (technical
coordinator) Prescribed fire by aerial ignition, Proccedings of a workshop. Intermountain Fire Council,
Missoula, Montana.
Sharples, J.J. (2008) Review of formal methodologies of wind-slope correction of wildfire rate of spread.
International Journal of Wildland Fire 17, 179-193.
Simcox, S., Wilkes, N.S., Jones, I.P. (1992) Computer simulation of the flows of hot gases from the fire at
Kings Cross Underground Station. Fire Safety Journal 18, 49-73.

Smith, D.A. (1992) Measurements of flame length and flame angle in an inclined trench. Fire Safety Journal,
18, 231-244.
Van Wagner, C.E.. (1977) Effect of slope on fire spread rate. Canadian Forestry Service Bi-monthly Research
Notes, 33, 7-8.
Viegas, D.X. (2002) Fire line rotation as a mechanism for fire spread on a uniform slope. International
Journal of Wildland Fire 11, 11-23.
Viegas, D.X. (2004) On the existence of a steady state regime for slope and wind driven fires. International
Journal of Wildland Fire 13, 101-117.
Viegas, D.X. (2005) A mathematical model for forest fires blow-up. Combustion Science and Technology
177, 27-51.
Viegas, D.X. (2006) Parametric study of an eruptive fire behaviour model. International Journal of Wildland
Fire 15, 169-177.
Viegas, D.X. ed. (2009) Recent forest fire related accidents in Europe. European Commission Joint Research
Centre Scientific and Technical Reports, European Communities.
Viegas, D.X., Pita, L.P. (2004) Fire spread in canyons. International Journal of Wildland Fire 13, 253-274.
Viegas, D.X., Pita, L.P., Ribeiro, L., Palheiro, P. (2005) Eruptive fire behaviour in past fatal accidents.
Proceedings, Eighth International Wildland Fire Safety Summit, April 26-28. Missoula, MT.
Viegas, D.X., Pita, L.P., Caballero, D., Rossa, C., Palheiro, P. (2006) Analysis of accidents in 2005 fires in
Portugal and Spain. Proceedings, Ninth International Wildland Fire Safety Summit, April 25-27.
Pasadena, CA.
Wu, Y., Drysdale, D.D. (1996) Study of upward flame spread on inclined surfaces. HSE Contract Research
Report No. 122/1996
Wu, Y., Xing, H.J., Atkinson, G. (2000) Interaction of fire plume with inclined surface. Fire Safety Journal
35, 391-403.
Young, T. (1800) Outlines of Experiments and Inquiries Respecting Sound and Light. Philosophical
Transactions of the Royal Society of London, 90, 106-150.