SUMMARY. The main physical aspects of forest fire propagation are described with the intention of illustrating its complexity and the risks associated with fire-fighting activities. The various modes of fire propagation and corresponding heat transfer mechanisms are described in relation to their effects on personnel working on the fireline. Some study cases of fatal accidents caused by forest fires are presented in order to emphasize the importance of having an adequate assessment of fire behaviour, together with training and adequate protective equipment, in order to prevent this type of accident.

Forest fires are among the most dangerous natural accidents that occur in practically all Mediterranean countries. These fires cause great damage to the environment, including wildlife, but also socially and economically. One cause of major concern is the threat that these incidents pose to human life. Many accidents in the past have proved that forest fires constitute a hazardous environment and the effort to extinguish them is a dangerous activity. One of the main causes of this is certainly the great variability of fire propagation properties, owing to a vast and complex set of influencing factors. Although in principle the behaviour of a flame front can be predicted, sufficient inforiflation may not be available to make a very accurate prediction. Scientific research is working on the development of models that are both robust and simple to provide support in the management of this type of accident. The problem of fireline safety has been addressed by many authors, including Chandler et al. (1983) and Palmer (1988). It is an issue that is present in all fire-fighter training programmes and in basic texts that support these courses, of which those published by NWCG (1986) and by NARTC (1983) are good examples.
The analysis of past accidents is helpful as it provides information about capabilities and decisions that did not work for one reason or another, which may be helpful to prevent future disasters. Descriptions and reports on accidents can be found in the literature, namely in Chandler et a]. (1983), Pyne (1984), Simard et al. (1983) and Washbum et a]. (1986). With this purpose four case studies are presented that show how a misjudgement of fire behaviour led to disaster, with loss of human life.

A forest fire is generally a dynamic phenomenon that changes its properties and behaviour from one place to another and with the passage of time. Owing to the fact that the forest fuel available in a given location is limited, for a fire to continue it must spread to neighbouring fuel. This is performed through the complex heat-transfer and thermochemical processes that determine fire behaviour.
Without going into these process in detail, it is important to recognize that there are various stages of fire growth and decay, and some typical fire propagation regimes. In order to understand the safety problems created by forest fires to human life it is necessary to be familiar with the above-mentioned concepts.

Fire growth and decay

A forest fire starts when a heat source - usually a small flame - ignites available fuels. If conditions are favourable the fire will propagate and grow both in size and intensity.
Combustion may be either non-flaming or with a flame. In the first case ground fuels usually bum very slowly, and this type of fire does not create a great threat to human activity, unless it changes to flaming combustion which grows and propagates much faster.
If there is sufficient fuel on the ground and near the surface the fire will attain a steady state of propagation called a surface fire (cf. Fig. 1). Secondary growth., spots consist of burning embers that are projected great distances, creating secondary fires that may cause great danger to people near the fireline, as can be easily understood.
When there is no fuel available for combustion, or when topographical or meteorological conditions change, the fire decays and we have a similar phenomeno-

Fig. 1 Phases of development and propagation regimes of a forest fire.

In forested areas a surface fire may grow vertically and reach the tree crowns, with the support of an ascending slope or strong wind. The fire will then propagate as a crown fire, which is extremely dangerous and difficult to extinguish. This is the so-called large fire regime. Large fires may be propagated either by wind and slope or by their own convection. In the first case we have a conflagration fire and in the second a mass fire. In large fires we have not only crown fire propagation but also spot fire propagation. These logy but in a reverse sense. A large fire may decay and become a surface fire, which may degenerate into a ground fire until it is extinct and ceases.

Fire spread regimes

As said before there are three main regimes of fire propagation: (i) ground fire; (ii) surface fire; and (iii) crown fire. The rate of spread and in particular the heat released by the fire varies by several orders of magnitude between these regimes and we must therefore analyse each one separately to show its respective threat to human life. Ground fire is not important in this context and we shall thus concentrate our attention on surface and crown fire.
Surface fire is by far the most common in the large majority of forest fires in the Mediterranean environment. Given the large variability of fuel and meteorological and topograplical conditions, a surface fire may propagate in a wide range of values of rate of spread and of fireline intensity (defined as the heat released per unit length of fire front per unit of time). A very important and visible effect of fire propagation is the flame front. Flames are the loci of very intense exothennic reactions and constitute the main heat source in a forest fire. Flames consist of gases at temperatures varying between 1000 and 2000 'C and are characterized by a certain length or height, inclination and depth. All these properties are relevant to personal safety in a fire front. Surface fires may have flames ranging from 0. 1 to 5 m and a rate of spread ranging from a few metres per hour to 10 km per hour.

Crown fires are characterized by much larger flames, of the order of 20 to 50 m and consequently higher rates of spread.

Although large fires and crown fires are potentially the most dangerous, N~ (1986) found that most fatal accidents in the past occurred in relatively small fires or quiet sectors of large fires. This is understandable as normally no one would risk his life or that of others in clearly dangerous situations, as is obviously the case in large fire propagation. There are of course also exceptions, due to an erroneous perception of the situation or to changes in the conditions affecting fire spread that were not noticed by the victims.

Factors affecting fire propagation

We have already mentioned the three groups of factors that affect fire behaviour. Let us recall them: (i) forest fuels; (ii) topography; and (iii) meteorological conditions.
Forest fuels have different properties depending on the plant species present in the area, on whether they are alive or dead, on the size of their particles, on the amount of fuel per unit area and on its spatial distribution. The moisture content of fuel particles is of great importance for fire propagation, as high values of moisture content damp or even prevent fire spread. As a rule fine dry particles propagate fire faster and more easily, as we see in grass or light brush fields.
Topography is of utmost importance in fire behaviour. A fire spreads much faster uphill than on a horizontal surface or downhill. Even in the absence of wind this fact may be overwhelming and a fire front may run uphill, catching people in its path, as too many past accidents have unfortunately demonstrated.
Meteorology affects fire propagation with both longterm and short-term action. Long-term action is determinant for the moisture content of forest fuels, including that of fire fuels. Short-term action concerns air temperature, relative humidity and the wind conditions that directly affect fire behaviour. Of these by far the most important is wind, defined by its velocity and direction.
Wind blowing in the same direction as that of flame propagation will increase the rate of spread from five to one hundred times the value of windless conditions. This action is stronger with light fuels, such as grass or light brush. These are usually considered not very dangerous fuels, although they may become so owing to the potentially high rates of spread that make entrapment relatively easy.
Wind may vary because of general changes in the atmosphere (large scale or weather changes) that usually occur slowly. The interaction of wind with topography creates a very complex flow pattern near the ground, with large variations of wind velocity and direction, sometimes with reversed flow, i.e. with a direction contrary to the main wind, in certain zones. As the flame front moves from one zone to another, even under constant wind conditions, it may experience considerable changes in its behaviour. It has been reported in the past (NWCG, 1986) that some fatal accidents occurred following unexpected wind shifts or increase in windspeed.
This brief description of the various regimes of fire spread and of the factors that affect them shows that accidents during forest fires can happen very easily. A person near a forest fire, even one with experience in the field, does not normally have all the information and knowledge necessary to evaluate how the fire will develop in the near future. There are so many factors and, especially for a person on the ground, it is not easy to form an overall view of the scene and predict if the fire will maintain its properties or accelerate, change direction, decay or do something else. With the exception of persons caught quite inadvertently inside a forest fire, the main cause of fatal or near fatal accidents, in the writer's opinion, is misjudgement of fire behaviour.

Heat transfer mechanisms

In a forest fire the main heat transfer mechanisms are radiation and convection. Given the low thermal conductivity of the materials in a forest, thermal conduction is a relatively minor heat transfer mechanism.
Working in the proximity of a flame front exposes persons to important heat fluxes. If these are greater than the body cooling capacity a heat stress situation may arise, with all its medical consequences.
Chance contact with flames, very hot particles or gases may cause bums whose severity depends on the temperature of the heat source and on the exposure time. In a military context it was been established (Chandler et al., 1983) that a person exposed to a radiation source of about 1000 'C cannot survive if the time of exposure is greater than 18 sec. At 15 see about 50% incapacity is attained and practically no incapacity occurs at times of less than about 9 see. If we apply these considerations to the case of a person crossing a flame front we have the following situation. Given a flame front 3 m in height and an average temperature of 1000 'C, and a person running through the flame at a velocity of 5 m/sec, we may calculate that in the approach phase the person is exposed to that temperature for about 2 sec. This leaves some 7 sec to cross the flame front. Supposing that the person is able to maintain the same velocity, we may conclude that the person could cross a flame front 35 m of depth. Although some of these considerations may be unrealistic, they give some encouragement to the decision to cross a flame front in the event of entrapment. It also shows that if a person tries to run in front of the flame front he is more likely to be exposed to severe bums and heat stress.
The convection of hot gases is another factor of heat stress in individuals. The inhalation of hot gases in particular may cause severe damage to the respiratory system which can be fatal, even if no major bums occur.
This must also be kept in mind when considering the question of crossing a flame front. In the previous example it was assumed that there was non-inhalation of hot gases from the flames, which is not easy to achieve, as can be easily understood.

Smoke

The combustion reaction in a forest fire is far from being complete and efficient, and large amounts of smoke composed of gaseous, condensed and solid products, with important effects on people working near the flame front, are released. Given the diffusion of smoke near the ground a situation of long exposure to smoke particles and to toxic gases can arise, with all its consequences.
In terms of fireline safety the loss of visibility created by smoke particles is very dangerous because it creates a sense of disorientation and aggravates the difficulty for a person to judge and evaluate the situation correctly. This particularly applies to drivers caught in roads full of smoke even if only for a short time-span. Even more dangerous is the case of pilots of low-flying planes or helicopters who may lose sight completely of the ground, with easily imaginable consequences.
In the composition of smoke we find carbon monoxide (C0) and occasionally some other toxic gases. The concentration of CO varies from around 10 to 20 ppm away from the fire to around 100 to 200 ppm at the edge of the fire. Protracted exposure to CO, even at low concentrations, can affect the nervous and cardiac systems and in extreme cases even cause death.

The foregoing considerations clearly show that it is very important to have the capacity to predict the behaviour of a fire front, in order to prevent situations dangerous to human life.
Past accidents have shown that even experienced fire-fighters are sometimes caught by surprise by unexpected danger, either because they did not correctly evaluate the situation, or because they did not have complete information, as for example the perception of some overall change.
In order to assist the persons in change of firefighting operations, especially in the event of a large fire, it is important to have a team working on fire behaviour prediction. With the support of mathematical models of fire behaviour and modem techniques of data collection and processing the process of decision making can be facilitated. Fire suppression can thus be carried out more efficiently and more safely. Decisions to evacuate fire crews from a flame front or to evacuate the population of a house or a village can be supported objectively and taken well in advance.

Actions to promote the safety of personnel dealing with forest fires must be directed first above all at those involved in fire suppression. We cannot however forget the general public, because in many instances persons who have nothing to do with fire-fighting activities are caught by inadvertence, curiosity or the desire to protect nature.
It is beyond the scope of this work to give a detailed treatment of the problem of fireline safety. We shall mention only briefly some aspects of two basic conditions for safety: training and equipment.
Personnel training is very important in order to teach crews to work as a team, to improve their technical skills and to provide the essential concepts of fire behaviour and the necessity of putting safety first.
Besides the essential tools and equipment, it is necessary to provide each fire-fighter with personal gear that affords him comfort and protection in the hard conditions of their mission. Jukkala and Putnam (1986) present a short description of the properties of each item of protective personal equipment, and similar descriptions can be found in NWCG and Chandler et al. The equipment should include flame-resistant overall clothing, leather boots and woollen socks, gloves, a hard hat and goggles. There have been interesting developments in Europe in the quality of the various pieces of equipment, especially in overall clothing. The work by Rouget and Drouet (1987), for example, describes tests with protective equipment for fire-fighters and contains some suggestions for protective equipment to be adopted in the forest fire environment. It is very important to ensure good protection while working near the flame front, without loss of comfort while travelling or walking. Adequate clothing in this environinent can prevent the risk of bums, especially in delicate parts of the body.
One very important piece of safety equipment is the so-called fire shelter, which is a sort of aluminiumcoated tent that can be carried in a small bag and put up rapidly in case of danger. Each fire shelter can provide excellent protection for one person, even in the event of entrapment by a large fire. In the writer's opinion this item should be mandatory for all those involved in firefighting. Its effectiveness has been well proved in many past cases and some fatal accidents could probably have been avoided if such shelters had been available. The report by Mutch and Rothermel (1986) is quite impressive in this respect.
A final remark to conclude this comment on safety: it is sometimes argued that protective equipment may induce overconfidence and thus provide ground for accidents. This will happen only if adequate training is not given in order to emphasize the capacities and the limitations of the equipment and of the human body.

In order to illustrate some of the ideas developed in this article we shall now briefly present four case studies of fatal accidents caused by forest fires. Unfortunately there are many examples that could have been quoted here in order to give some insight into the problem and to learn how to avoid danger.
In Portugal 55 persons were killed by fire in three major accidents: in 1966, 25 soldiers were trapped in Sintra; in 1985, 14 fire-fighters were killed in Annamar; and in 1986, 16 fire-fighters were caught in Agueda. In 1992 seven persons were killed in five different accidents, including two pilots of fire-fighting aircraft.
From these and other accidents we choose four cases that will be briefly described below.

This accident occurred on 8 September 1985 near Armamar, in the north of Portugal. Fig. 2 gives a schematic view of the area. On the third day of a large fire, a crew of 15 men, most of them well-experienced fire-fighters, had just extinguished a fire in a pine-stand (point A in Fig. 2) and was moving across a valley to help a small village that was experiencing some fire starts around it during a thunderstorm.
They were descending down a slope covered by heavy bush in a sort of small valley (B). In the meantime, under a heavy wind, the main fire made a long run along both sides of the valley. This was not noticed by the men until it was too late. They tried to run to their right, cross the small valley and climb to the other side, because they could see only the flames on their left.
In spite of the problems of progressing in such a terrain and in difficult psychological conditions five men reached the top of the valley (C), but there they were caught by the flame front approaching along the right side of the valley. In the meantime the others were caught in the middle of the two flame fronts. Only one man escaped: he was the last in the line and decided to run down the valley along the water line and survived (D). Everything was over in less than twenty minutes after the team first saw the danger.

Fig. 2 Schematic view of the accident area at Armamar.

This accident occurred near the village of Nagoselo do Douro, also in the north of Portugal, not far from the site of the previous accident, on 31 July 1992.
A small fire was caused by a thunderstorm on the top of a hill, near a small chapel (see location A in Fig. 3). The owner of the land, which was covered in fine oak trees, went there and tried to extinguish the fire in its initial stages. A soldier who had just arrived at home on leave also went immediately to the area, taking with him four boys he met on the way.
The six people were at the bottom of a canyon approximately two-thirds up the side of the hill (B), with very poor visibility to their back and sides and with very bad escape routes. Unbeknown to them the fire spread down the opposite slope, around the hill (C) and appeared at their back at the base of the hill on the bank of the River Douro (D). With the help of wind the flames roared uphill giving very little chance of escape. The four boys were more fortunate and reached a safe point (E), but the soldier and the land-owner, who was a 56-year-old man, perished in the fire (F).

Fig. 3 Perspective of the accident site at Nagoselo do Douro.

This accident occurred near the Fronhas dam on the River Alva, 40 km from Coimbra, in central Portugal, on 6 August 1992.
At about 2 p.m. a fire started about 5 km away from the Fronhas dam, where a crew of workers were installing a steel structure in the exterior of the central building of the dam (location A in Fig. 4). Although they saw the fire they went on working until at about 4 p.m. they noticed that the flame front was coming in their direction and was very close to them. Although the span of the top of the dam is about 100 m and there is a road along it they decided to run to the bottom of the dam and wait for the fire there (B). There was a jet of water that sprinkled them and kept them cool and comfortable.
For some reason two of the men - one of them was actually an off-duty fire-fighter - decided to walk along the river (C) and cross to the other side. When they were climbing the steep slope on the other side, the flames appeared on the other bank (D) and propagated to the slope where they were (E). The first man managed to reach the top of the slope but the firefighter did not and was caught by the flames rapidly ascending through grass and light bush (F).

Fig. 4 Perspective view of the site of the accident at the Fronhas dam.

Celavisa

This accident occured on 7 August 1992, near the village of Celavisa, about 20 kin from the place of the previous accident and actually in the same large forest fire.
A young man who had just arrived to spend his holidays in his native village accepted the suggestion of his 12-year-old daughter to go and get a closer view of the fire. They went by motorcycle along the road and suddenly were surprised by a column of smoke from a fire downhill (location A in Fig. 5). Instead of turning back, perhaps because the fire had cut their escape, they left the motorcycle at the side of the road (B) and walked on down the road, probably looking for shelter in a house some 300 m from the place where they left the motorcycle. Unfortunately they did not reach the house and were both caught by the ascending flames (C), ironically in the middle of the only stretch (about 400 m long) where the fire crossed the road (D).

Fig.5 Perspective view of the Celavisa accident.

In this paper the author has tried to give a general idea of the complex phenomenology of forest fire propagation, in order to emphasize the danger that this type of natural accident represents for human life.
As a general rule one can say that the basic cause of every forest fire accident is some misjudgement of the behaviour of the propagating fire front. Given the vast amount of variables that govern this behaviour, it is not difficult to appreciate this difficulty. Scientific research on fire behaviour and the development of fire prediction systems and protective equipment are valuable tools to improve fireline safety and can be very helpful in the prevention of such accidents.
The case studies presented support the idea that even experienced and well-trained fire-fighters can sometimes be trapped in apparently nondangerous situations. As untrained persons can also be caught in a fire, the problem of public awareness is very important.
Joint research between medical doctors, scientists and operational institutions is needed in order to provide better ways of preventing injuries and fatalities due to forest fires. The author hopes that this work may be of help in this direction.

RESUME. L'auteur décrit les principaux aspects physiques de la propagation des incendies des forêts dans le but d'illustrer la complexité du problème et les risques associés aux activités anti-incendie. Les différentes modalités de la propagation du feu et les méchanismes correspondants du transfert de la chaleur sont décrits pour ce qui concerne les effets sur les personnes engagées dans l'extinction de l'incendie. L'auteur présente des cas particuliers d'accidents mortels causés par les incendies de forêt avec l'intention de souligner l'importance de posséder une compréhension précise du comportement des incendies, ainsi qu'une formation complète et un équipement protectif approprié pour éviter ce type d'accident.

Acknowledgements

The author wishes to thank the institutions that support the research programme on the physical aspects of forest fires, in the framework of which this paper was written, namely the Junta Nacional.

InvestigaQAo Cientffica e TecnoWgica, the Comissdo Nacional Especializada de Fogos Florestais, and the EEC. A particular acknowledgement is made to the Servigo Nacional de Bombeiros for the information provided regarding the accidents described.

1. Chandler C.P., Cheney P., Thomas L., Trabaud, Williams D.: "Fire in forestry", vol. 2, 231-48. J. Whiley & Sons, New York, 1983.
2. Gaylor H.P.: "Wildfires", 167-87. R.J. Broady Co., 1974.
3. Jukkala A.H., Putnam S.E.: Protective equipment for the wildland. Seminar on Firefighters, Methods and Equipment for the Prevention of Forest Fires, Valencia, 30 Sep. - 4 Oct.,1986.
4. Mutch R.W., Rothermel R.C.: 73 Fire-fighters survive in shelters. Fire Command, 30-2: March, 1986.
5. National Advanced Resource Technology Center: El combate de incendios forestales. Curso Internacional Avanqado, Arizona, vol. 2, 1983.
6. National Wildfire Coordinating Group: Fire Behavior S-390 (Student Guide and Text), Boise Interagency Fire Center, 1981.
7. National Wildfire Coordinating Group: Fire-fighters Guide, USDA, NFES 1571, Boise Interagency Fire Center, 1986.
8. Palmer S.L.: Emerging issues and concerns about wildland fire-fighter safety. In: "Proceedings of the art and science of fire management", Alexander M.E., Bisgrave G.F. (eds), 1988.
9. Pyne S.: "Introduction to wildland fire - fire management in the United States", 390-400, J. Whiley & Sons, New York 1984.
10. Rouget Y., Drouet J.C.: - Essais de vêtements et dispositifs de protection individuelle des personnels luttant contre les feux de forêts. R.G.S., 62: 50-7, 1987.
11. Simard A.J., Haines D.A., Blank R.W., Frost J.S.: The Mack Lake Fire. USDA, Forest Service, North Central Forest Experiment Station, Gen. Techn. Rep. NC-83, 1983.
12. Washburn A.E., Harlow D.W., Fahy R.F.: U.S. fire-fighter deaths in 1985. Fire Command, 21-38: June, 1985.