The study
and investigation of fires is an essential component in the development
of fire safety engineering as a discipline. But, argues Dougal Drysdale,
misleading conclusions can be drawn from fire scenes unless the investigator
understands the wide range of factors that can influence the fire process.
ACCIDENT
INVESTIGATION is an essential part in the development of any engineering
discipline. It provides a means of improving our knowledge and understanding
of the subject. For example, we accept that civil engineers are able to
design and construct bridges that will be safe. But structural design
and the relevant theory of structures is underpinned by the analysis of
past failures. The same argument is true in the context of fire. Fire
investigation is an essential component in the development of fire safety
engineering as a discipline.
Certain fires
have provided information that has been invaluable in improving our understanding
of the fire process, which in turn should allow a more rational approach
to the identification and mitigation of fire hazard. However, the response
to significant failures (defined by a large life loss) has been dominated
by societal reaction, followed by the imposition of new regulations to
prevent recurrence. This is well illustrated by the aftermath of the Bradford
City Stadium fire (1985)1 and the King's Cross Underground Station fire
(1987)2. While such tragedies continue to provide valuable information
for the fire safety scientist, that information is seldom used to its
maximum advantage. It is gradually being absorbed into the body of knowledge
required by the fire safety engineer but is seldom fed back into the fire
investigation community. Although some fires provide the opportunity to
expand our knowledge of fire behaviour, most fire investigations are routine
and require only a rudimentary knowledge of fire dynamics. Occasionally
contemporary knowledge is insufficient to resolve a particular problem,
or the characteristics of the fire were such that the damage pattern is
ambiguous, leading to possible misinterpretation. In both cases, a thorough
investigation will have the dual purpose of identifying fire behaviour
that may not previously have been recognised, as well as adding to the
knowledge base.
Understanding
the past
The study
of fire dynamics has developed rapidly since the 1960s, but its adoption
by the forensic community has been slow. Fire behaves according to the
basic laws of physics, particularly those relating to fluid dynamics and
heat transfer. Whether one is trying to predict how a fire might behave
following ignition or to locate from the physical evidence at a fire scene
where a fire started and how it developed, the same laws of physics must
be applied.
From time
to time a fire occurs leaving physical evidence and eyewitness accounts
that are very difficult to understand. For example, there may be insufficient
evidence to identify the locus of origin, or there may be features of
the damage pattern which lead to conclusions inconsistent with eyewitness
accounts, or even with other physical evidence. Some of the earlier texts
on fire investigation were misleading in that they preserved some of the
myths of fire behaviour, ignoring research results that clearly showed
otherwise. One of these is the adage that all wood chars at the same,
constant rate of one fortieth of an inch per minute. Not only does this
seem unlikely, but research published in 19713 showed it to
be false: the rate of charring depends strongly on the level of radiant
(or convective) heating to which the wood has been exposed. Nevertheless,
depth of char can be used as a valuable guide when attempting to interpret
a fire scene. With care, understanding and the application of common sense,
useful corroborative conclusions may be drawn.
Most fire
investigators rely on their experience and have a general appreciation
of fire behaviour, often allowing some details to be overlooked to no
significant disadvantage. However, fires do occur where such details are
of vital importance and can be of great assistance in the interpretation
of the evidence. This will be illustrated with reference to three fires:
the Stardust Club fire (Dublin, 1981), the King's Cross Underground Station
fire (London, 1987) and the Garley Building fire (Hong Kong, 1996).
The investigation
of the Stardust Club fire4 revealed significant shortcomings
in regulatory control that allowed an extremely hazardous configuration
of combustible materials to be used in a place of public assembly. A full-scale
test was carried out to demonstrate how and why the fire developed so
rapidly but, in fact, enough was known about the particular configuration
of combustible materials to have been able to make at least a qualitative
assessment of the hazard. On the other hand, the configuration of fuel
(the wooden escalator) which led to the extremely rapid growth of the
King's Cross fire had not previously been identified as a specific hazard
(other than the undesirability of having extended combustible surfaces
in a public space). The Court of Inquiry uncovered the so-called 'trench
effect'. In the case of the Garley Building5, the fire risk
could be said to have been obvious, but the characteristics of the fire
were such that the patterns of damage in the area where the fire is alleged
to have started were ambiguous and open to misinterpretation.
The Stardust
Club
The fire
at the Stardust Club in Dublin in February 1981 was responsible for the
deaths of 48 people. The investigation and subsequent Inquiry revealed
a large number of factors which had contributed to the high toll (including
locked escape doors), but the root cause was the rapidity with which the
fire developed.
In circumstances
where there has been rapid fire development, there is always a suspicion
that the fire has been started maliciously with an accelerant. This was
found not to be the case in the Stardust fire, but it was appreciated
that the combination and configuration of combustible materials - namely
carpet tiles which were used to line the walls - in the alcove where the
fire is known to have started was hazardous. The Fire Research Station
carried out a full-scale demonstration of the fire's development, which
confirmed the speed with which the fire grew in its early stages, as the
carpet tiles on the rear wall became involved6, 7.
Indeed, the ferocity of development demonstrated not only the hazard of
combustible wall-lining materials but also how a low ceiling can make
matters even worse. The fire scientist knew that this combination was
very dangerous but at that time (and perhaps even now) was unable to quantify
the problem. This is a major disadvantage. Quantification would have allowed
the fire safety engineer to make a substantive case to the authorities
that such a combination was totally unacceptable. Similarly, the fire
investigator would benefit if after the event he was able to demonstrate
by rational argument, backed up by calculation, that this had been the
cause of such rapid fire development. In a sense, the Stardust Club fire
was analogous to the failure of a bridge that alerted the design engineer
to a fault to which his attention had not previously been drawn.
The fire
investigator who is aware of the basic principles of fire science and
fire dynamics will be able to recognise configurations of fuel that could
be hazardous. As our knowledge of fire behaviour increases, reliable modelling
techniques will become available which will allow the validity of conclusions
to be demonstrated8, 9. For example, ongoing research
into flame spread on vertical surfaces is showing how data from small
scale tests (specifically, the cone calorimeter) can be used to model
the rate of upward fire spread. Further developments will soon enable
the rate of fire growth, even under complex circumstances, such as those
in the Stardust Club, to be predicted. With such models available, there
would be no further need to carry out expensive full-scale tests simply
for demonstration purposes.
King's
Cross
The King's
Cross Underground Station fire in November 1987 revealed a significant
gap in our appreciation of fire behaviour. The location of origin was
in no doubt. The fire was first seen underneath the steps of escalator
No. 4 (one of three escalators which served the Piccadilly Line from the
King's Cross Station booking hall), approximately one-third of the way
up from the lower level. This was confirmed in the subsequent investigation
as the lowest point of damage10. The fire spread above the
escalator steps and the damage to the ceiling of the escalator shaft suggested
that flames had 'corkscrewed' diagonally upwards from the lowest point
to the other side of the escalator shaft, meeting the opposite side about
8m from the booking hall. Eyewitness accounts indicated that the fire
had spread extremely rapidly in the two or three minutes before the booking
hall became totally engulfed in flames. One witness referred to a 'jet
of flame' emerging into the booking hall from the Piccadilly Line escalator
shaft11. It was some time before an explanation was found to
account for this rapid development.
During the
early stages of the investigation, considerable attention was paid to
the flame spread properties of the multiple layers of paint on the ceiling
of the escalator shaft as it was known that multiple paint layers could
be a serious fire hazard. Other suggestions included unusual cross radiation
effects within the escalator trench, but ultimately it was shown that
a very different mechanism was responsible. This came to be known as the
'trench effect'. The first indication of this phenomenon was revealed
by some early computational fluid dynamic (CFD) modelling12
of the flows of hot gases in the escalator trench (inclined at an angle
of 30°). Instead of the hot buoyant gases rising vertically, out of the
confines of the trench, they became 'attached' to the trench floor and
flowed upwards within the trench, hinting for the first time that the
flames from the burning wooden escalator would flow in a similar fashion
and preheat unburnt fuel (wooden balustrades, steps and risers), thus
causing very rapid flame spread up the trench towards the booking hall.
This mechanism was subsequently confirmed in laboratory tests on a small
scale model13, and ultimately in a one-third scale mock up
which included the escalators, the escalator shaft and the booking hall14.
Subsequent CFD modelling of the behaviour of flames on inclined surfaces
(with and without 'sidewalls') has confirmed the experimental findings
and has provided valuable insight into the general problem15,16.
The technical aspects of the report of the Inquiry into the King's Cross
fire should be required reading for any fire investigator. Some of the
evidence was confusing and capable of being misinterpreted if only the
basic rules of thumb were applied. A combination of an understanding of
fire dynamics and the liberal application of common sense will help avoid
misinterpretation of ambiguous clues that may be found at the fire scene.
The Garley
Building
An interesting
example of an ambiguous clue was encountered during the investigation
of the fire in the Garley Building in which 40 people lost their lives.
This occurred on 20 November 1996 in Tsim Sha Tsui, Hong Kong.
The
building, which was 16 storeys high, was operating as a multiple occupancy
commercial building. The first three floors were occupied by a department
store (CAC) and the remaining floors housed a variety of businesses. One
stairway (A) had connected the lift lobbies from the ground to floor 3
(3/F) but had been sealed off at ground and 3/F to form a series of storerooms.
These were separated from the lift lobbies on 1/F and 2/F by plywood partitions.
Furthermore the fire-resisting walls between the lift lobbies and CAC
on 1/F and 2/F had been replaced by roller shutter doors. These closed
automatically on activation of smoke detectors but were not fire rated.
Two of the four lifts serving the building were under refurbishment at
the time of the fire. Lift shafts 1 and 3 had been completely stripped
and were open at every level, only protected by wooden hoardings. Bamboo
scaffolding had been erected within the shafts to allow access for the
installation of fixed hardware.
Fire was
first observed in the 2/F lift lobby by a workman who climbed up the bamboo
scaffolding in lift shaft 3. He described flames behind some wooden planks,
which were leaning against the hoarding, about 2m from the edge of lift
shaft. Within a short period of time flashover appears to have occurred
within this hoarding and shafts 1 and 3 were soon involved. Unusually,
the fire spread downwards to involve the 1/F lift lobby within a matter
of minutes. While unexpected, this may be a consequence of the burning
characteristics of bamboo, which when involved in fire explodes as the
air trapped between the internal nodes expands, scattering burning fragments
over a wide area17.
From
this stage onwards, serious fires were burning in the lift lobbies on
the first and second floors, drawing air from the two open lift shafts.
There was serious fire damage on the ground, first, second, 13th, 14th
and 15th floors. All but one of the fatalities were trapped on the upper
floors. Levels 4 to 12 were scarcely touched by the fire.
It was concluded
by the Commissioner of the Public Inquiry that the fire had started in
the second floor lift lobby as a result of welding debris falling from
the 13th floor, where work was being carried out at the rear wall of the
lift shaft. However, significant aspects of the evidence of the witness
who first observed the fire conflicts with some of the conclusions that
were accepted by the Inquiry.
Figure
1: The lift lobby on the second floor of the Garley Building as
it was prior to the fire. Lift shafts 1 and 3 we re completely open,
protected by wooden hoardings (WH). The fire wall separating the lift
lobby from the department store (CAC) had been re m oved and replaced
by ordinary roller shutters (RS). Stair A closed off at G/F and 3/F
had been conve rted into store rooms, with no fire separation between
the stairwell and the lift lobbies at 1/F and 2/F. |
|
|
Figure
2: Plan of the 2/F lift lobby, showing the location of the low
leve l damage outside Lift Shaft 3 at A. The wooden hoardings (WH)
were almost totally consumed. The fire damage to wooden boards and
planks at B was consistent with an intense fire burning in the vicinity
of A, as we re the remains of a pile of corrugated card board at C.
The roller shutter doors (RS) were in place at the end of the fire.
A wooden Chinese Altar at D was totally consumed. |
Figure
3: The early and late stages of the fire. (a) The flows associated
with fires in both lift lobbies on 1/F and 2/F before the involvement
of the store rooms in Stair A. (b) The flow that would have developed
after the wooden partitions in Stair A had burned through, allowing
the fire in the 1/F lift lobby to vent through the 2/F lift lobby
into the lift shafts. |
|
Post-fire
evidence
The post-fire
evidence showed that there was serious low level damage in front of the
opening to lift shaft 3, which was interpreted as identifying the locus
of ignition. There was also evidence on the metal architrave around the
opening to the shaft that suggested that flames had emerged into the shaft
at a very low level. There were further directional markers which were
taken to confirm that the fire had started on the floor close to the lift
opening and spread from there outwards. However, when the damage pattern
is considered in the context of the way in which the fire subsequently
developed, it must be concluded that this evidence is at least ambiguous.
The fire
burned for many hours. The storerooms in the enclosed stairwell between
the ground and 3/F eventually became involved. At some stage during the
fire, a flow path would have been created from the first floor lift landing
to the second floor lift landing, through the storerooms, which contained
large quantities of combustible material. As the contents became increasingly
involved, the partitions would have burnt away and allowed a vigorous
fire flow to develop between the two lift lobbies. Prior to this event,
the fires in the first and second floor lift lobbies would have entrained
air from the lift shaft (Figure 3(a)). Each level would have behaved as
a conventional compartment fire, in which fresh air is entrained through
the lower part of the opening and flames and hot combustion products emerge
from the upper part. In the unusual conditions of the Garley fire, the
2/F lift lobby fire would have been entraining the products arising from
the fire in the 1/F lift lobby. Once the partitions between the lift lobbies
and the storerooms had burned through, the flames and gases from the fire
on the 1/F lift lobby would have vented through the stairwell into the
2/F lift lobby.
The two lift
lobbies and the stairwell would then become one compartment, fresh air
being entrained at the lower level (1/F) and the hot combustion products
flowing into the lift shaft from the 2/F lift lobby (Figure 3(b)). The
damage to the floor outside lift shaft 3 on the second floor lift lobby
is entirely consistent with a flow of this nature, as flames would have
completely filled the opening as they emerged into the lift shaft. (It
has been observed that when a fully developed fire vents through a single
opening, the emerging flames fill the upper half, to two-thirds of the
opening18, 19). This flow would have produced intense
heat transfer to the floor of the lift lobby outside lift 3 and provided
high levels of radiant heating, which would have enhanced the burning
rate of any combustible materials in the vicinity. The low-level damage
to the floor outside lift shaft 1 was much less severe, but it may have
had some degree of protection from the fire when it was at its most intense.
The fuel load in that corner of the lift lobby was relatively low and
the single hoarding which surrounded the opening may not have been consumed
as quickly as the hoarding around lift 3.
The interpretation
of the evidence in the 2/F lift lobby is consistent with this mechanism.
The directional markers that were used to help pinpoint the seat of the
fire as being immediately in front of lift shaft 3 would have been caused
by intense low level burning in this location at a later stage in the
fire. In fact, there is a strong argument that the damage observed in
the 2/F lift lobby would have been caused by any long-burning fire which
came to involve the contents of the storerooms in Stair A - regardless
of where ignition first occurred. Consequently, the low level damage to
the floor and surrounding materials outside lift shaft 3 cannot be assumed
to be evidence of the locus of origin.
Using
all available evidence
The three
fires discussed were all major incidents involving multiple loss of life.
Such tragedies are fortunately rare, but they do provide unusual opportunities
for in-depth studies of fire behaviour. Many, if not most, fires that
are investigated are quite clear-cut in much of the detail. But it is
a duty of the fire investigator to draw on as much available evidence
as possible to confirm any conclusion about the way in which the fire
started and spread.
In many cases, it is entirely sufficient to identify the most likely cause,
but this should be combined with identification of the evidence that confirms
where the fire started. Certain clues that may be found at the fire scene
can be misleading. For example, deep charring at specific locations may
be interpreted as evidence for the use of flammable liquids, particularly
if this damage is at low level in an enclosure. However, serious burning
can be a result of the way in which the fire has behaved within the compartment,
either as a consequence of ventilation being provided close to that region,
or from some unusual fire flow condition that gave rise to high rates
of heat transfer in that locality, such as in the Garley Building fire.
In principle,
the fire investigator should be able to identify any unusual features
of the building geometry that could account for the evidence at the fire
scene. Evidence for high rate of heat transfer or high temperatures in
unusual locations need to be explained. He or she should also be able
to identify circumstances that might lead to high rates of heat transfer
and, if this is likely to compromise interpretation of evidence, it must
be taken into account when coming to final conclusions q
Professor
Dougal Drysdale is based at the School of Civil and Environmental Engineering,
University of Edinburgh
References
1 Popplewell,
O., Committee of Inquiry into Crowd Safety at Sports Grounds: Final Report,
HMSO, London, 1986.
2 Fennell,
D., Investigation into the King's Cross Underground Station Fire, HMSO,
London, 1988.
3 Butler,
C. P., 'Notes on the charring rates in wood', Fire Research Note 896,
1971.
4 Report
of the Tribunal of Inquiry on the Fire at the Stardust, Artane, Dublin,
Stationary Office, Dublin, 1982.
5 Woo, K.
H., Final Report of the Inquiry into the Garley Building Fire on 20th
November 1996, Hong Kong: Government Printer, 1997
6 Morris,
W. A., 'Stardust Disco Investigation - some observations on the full-scale
fire tests', Fire Safety Journal 7, 255-65, 1984.
7 Video:
'Anatomy of a Fire', Building Research Establishment, Garston, Watford,
1982.
8 Karlsson,
B, 'A mathematical model for calculating heat release rate in the room
corner test', Fire Safety Journal 20, 93-113, 1993
9 Grant,
G. B. and Drysdale D. D., 'Numerical modelling of early fire spread in
warehouse fires', Fire Safety Journal 24, 247-278, 1995.
10 Moodie,
K., 'The King's Cross Fire: Damage assessment and overview of the technical
investigation', Fire Safety Journal 18, 13-33, 1992.
11 Roberts,
A. F., 'The King's Cross Fire: a correlation of eyewitness accounts and
results of the scientific investigation', Fire Safety Journal 18, 105-121,
1992.
12 Simcox,
S., Wilkes, N. S., and Jones, I. P., 'Computer simulation of the flows
of hot gases from the fire at the King's Cross Underground Station', Fire
Safety Journal 18, 49-73, 1992.
13 Drysdale,
D. D., Macmillan, A. J. R., and Shilitto, D., 'The King's Cross Fire:
experimental verification of the 'trench effect', Fire Safety Journal
18, 75-82, 1992.
14 Moodie,
K., and Jagger, S. F., 'The King's Cross Fire: results and analysis from
the scale model tests', Fire Safety Journal 18, 83-103, 1992.
15 Woodburn,
P., and Drysdale, D. D., 'Fire in Inclined Trenches: the dependence of
the critical angle on the trench and burner geometry', Fire Safety Journal
31, 143-164, 1998.
16 Woodburn,
P., and Drysdale, D. D., 'Fire in Inclined Trenches: time-varying features
of the attached plume', Fire Safety Journal 31, 165-1172, 1998.
17 Janssen,
J. J. A., Building with Bamboo: a Handbook, pp.21 Intermediate Technology
Publications, London, 1988.
18 Babrauskas,
V., and Williamson, R. B., 'Post flashover compartment fires: basis of
a theoretical model', Fire and Materials 2, 39-53, 1978.
19 Drysdale,
D. D., Introduction to Fire Dynamics, Second Edition, JohnWiley and Sons,
Chichester, 1998.
The author
wishes to thank Dr Sheilah Hamilton, Forensic Focus, Hong Kong, for valuable
discussions.
This paper
was presented at Interflam '99, the 8th Fire Science and Engineering Conference
organised by Interscience Communications Ltd at Heriot Watt University,
Edinburgh, in July 1999.
|