NOTE: This is a direct translation from the French report, as close as possible in content and meaning to the original document. (It should be noted that this translation retains the use of mixed verb tenses found in the original.) If you have any questions, please call me at 617-946-3050 or e-mail me at email@example.com
Sergiu F. Luchian, P.E.
CA/T, Boston, MA
August 26, 1999
1. MONT BLANC TUNNEL - DESCRIPTION
1.1. Tunnel geometry
1.3. Smoke extraction capacities
1.3.1. Smoke removal requirements in French bulletin no. 81.109 of 29 December 1981 on vehicular tunnel safety
1.3.2. Recommendations of the Committee of Vehicular Tunnels of the World Highways Association
1.3.3. Regulations in proposed new French bulletin currently in preparation
1.4. Tunnel operating procedures
1.5. Traffic rules for moving and halted vehicles
1.5.1. Distance between vehicles
1.5.2. Speeds and passing
1.5.3. Traffic signals
1.6. Evolution of traffic since the opening
1.7. Comparison with similar vehicular tunnels
1.7.2. Means of evacuation and protection of motorists
1.7.3. Safety equipment
1.7.4. Hazardous cargo
1.7.5. Response team
2. ORGANIZATION OF TUNNEL SAFETY
2.1. Original institutional framework
2.2. Common administrative organization
2.3. Intergovernmental control commission (CIG)
2.4. Organization of rescue efforts by the operators
2.4.1. Original documents contain few specific instructions on this subject
2.4.2. Organization of the operators' rescue assets
2.4.3. Explanations of the choice of a rescue service on the French side for the entire tunnel
2.4.4. Firefighting drills
2.5. Public safety plans
2.6. French-Italian cooperation
3. DEVELOPMENT OF EVENTS
3.1. Traffic conditions and tunnel functioning, before and after the alarm
3.1.1. Traffic status before the alarm
3.1.2. Alarm conditions
3.1.3. Tunnel closure
3.2. Rescue operations
3.2.1. The alarm
3.2.2. The response
3.2.3. Information given to public authorities and rescue services
3.2.4. First response by French rescue assets
3.2.5. Organization of Italian rescue assets
3.2.6. Comments on organization and implementation of the rescue services
3.2.7. Smoke reversal attempts
3.2.8. Contacts between French and Italian firefighters
3.2.9. Firefighters' response
3.3. Results of the catastrophe
3.3.1. Human loss
3.3.2. Physical damage to the tunnel
4. ANALYSIS OF CAUSES OF THE CATASTROPHE
4.1. Operational characteristics of the Mont Blanc Tunnel
4.2. Historical review of accidents and fires in vehicular tunnels
4.3. Review of truck fires in the Mont Blanc Tunnel
4.3.1. Seventeen truck fires in the tunnel since 1965
4.3.2. Circumstances of the five other major fires
4.4. Sizes of the fires and smoke volumes of the truck fires and their cargos
4.4.1. Caloric potential
4.4.2. Thermal power
4.4.3. Smoke emission
4.4.4. Smoke characteristics
4.4.5. Truck flammability
4.5. Circumstances of development of the first fire
4.5.1. Description of the development of the first fire
4.5.2. Specifics of the truck and its cargo
4.6. Ventilation conditions
4.6.1. Ventilation levels before and after the alarm
4.6.2. Consequences of ventilation levels on smoke movement and development of the fire
4.6.3. Spreading of the fire among the vehicles
4.6.4. Comments on the ventilation procedures put into effect during the fire
4.7. The role of the other systems
4.7.1 Electrical equipment
4.7.2 Traffic signals
4.7.3 Refuge areas
4.7.4 Safety corridors
4.7.5 Other equipment
After the 24 March 1999 Mont Blanc tunnel fire catastrophe, the Minister of the Interior and the Minister of Equipment, Transportation and Housing assigned a technical investigation to Mr. Pierre Duffé, Inspector General of the Administration, and to Mr. Michel Marec, State Engineer of Bridges and Highways.
Their written orders (attached) were to investigate and determine the reasons why this tragedy reached such a magnitude, to examine the tunnel safety systems and their functionality, and to present proposals to enhance the operational safety of vehicular tunnels of this kind.
The technical investigation task force includes:
- Mr. Pierre Duffé, Inspector General, and Mr. Mathias Pestre-Mazieres, Inspector of the Administration, assisted by Colonel Gilardo and Lieutenant Colonel Vergnault, for the Ministry of the Interior (DDSC);
- Mr. Michel Marec, State Engineer of Bridges and Highways, assisted by Mr. François Barthélemy, Chief Engineer of Mines, Mr. Jacques Demoulin, Board of Bridges and Highways, and Mr. Didier Lacroix, Chief Engineer of Bridges and Highways, for the Ministry of Equipment, Transportation and Housing.
The task force was supported by additional experts as necessary.
A status report!was submitted on 13 April 1999. Its goal was to present as rapidly as possible information on the circumstances and causes of the catastrophe. The very brief period in which it was prepared (two and a half weeks) did not always permit performing all the necessary verifications and in-depth research. Nevertheless, this report revealed the basic circumstances of the catastrophe. We repeat them here:
- the speed and magnitude of the fire development in the first truck, which caused the fire, and its spreading to other vehicles;
- the smoke extraction limitations of the tunnel ventilation;
- the manner in which the ventilation was activated.
In addition, other circumstances were exposed:
- the insufficient coordination between the two companies;
- the inadequacy or poor functioning of certain equipment.
This report!clarifies all of the above, while providing in-depth information on the circumstances of the accident.
The authors have attempted to rank the various causes leading to the magnitude of the catastrophe. In-depth studies will be necessary to recreate as much as possible the development of the fire and define the role of each of the factors leading to it, particularly the ventilation.
Proposals for restoring service in the Mont Blanc Tunnel, in accordance with the most modern tunnel safety requirements, and the lessons learned for other tunnels in terms of equipment, operation and organization of rescue forces, are the subjects of the joint report!developed with the Italian administrative task force.
This report!is attached to the joint report.
Chapter 1 describes the tunnel. Chapter 2 describes the legal background and its evolution, as well as the role of the common administrative organization and the intergovernmental control commission. The development of events and particularly the rescue operations are the subject of Chapter 3. Chapter 4 analyzes the causes of the catastrophe. Chapter 5 consists of a conclusion. The recommendations of the task force, together with those of the Italian task force, are included in the joint report.
|1. MONT BLANC TUNNEL - DESCRIPTION|
The tunnel was built jointly by France and Italy and opened in 1965. It is 11,600 m long. The longest part is on French territory: 7,640 m, with 3,960 m in Italy. Two operating entities were created, the STMB (Société du Tunnel du Mont Blanc) in France, which became ATMB (Autoroute et Tunnel du Mont Blanc), and SITMB (Societa Italiana del Traforo di Monte Bianco) in Italy, each operating half of the tunnel.
1.1 Tunnel Geometry
Connecting, in the direction France-Italy, the Chamonix Valley with the Val D'Aoste, located at elevations 1,274 m (French portal) and 1,381 m (Italian portal), at the top of a severe slope (up to 7%) about 4 km long on the French side, the tunnel has a 7 m wide highway with two 0.8 m walkways on the sides; every 300 m there are vehicle rest areas, 3.15 m wide by 30 m long, situated on alternating sides of the highway and numbered from 1 to 36 in the France–Italy direction; opposite each area there is a designated U-turn for trucks. Every other rest area has a safe refuge area (also called shelter) for people, supplied with fresh air and protected by an enclosure with a two-hour fire rating. Safety niches every 100 m contain a fire pullbox and two fire extinguishers. Fire niches every 150 m provide water supply for firefighters; they are across from the rest areas and are equipped with telephones and pullboxes.
The ventilation had been designed, at the outset, to be identical for the two halves of the tunnel, starting at each of the portals, with ducts underneath the roadway. Four supply air ducts, numbered from 1 to 4, start from each portal and each services a quarter (1,450 m) of the half length of the tunnel by ventilation slots located every 10 m at the bottom of the lateral wall in the Italy-France direction. Each duct can supply 75 m³/s of fresh air, which corresponds to a maximum flow of 300 m³/s at each portal, and 600 m³/s total. A fifth duct was originally built to exhaust air polluted by traffic. In principle, it is capable of removing 150 m³/s from each half of the tunnel by exhaust openings located in the ceiling at the level of the rest areas, i.e., every 300 meters, in the France to Italy direction. It should be noted that, originally, the extraction of fire smoke was not a design criterion for the ventilation system.
In 1979, modifications were made to the exhaust at each portal in order to use the exhaust ducts for additional supply when air pollution by traffic (especially trucks) justifies it. This way a total of 300+150= 450 m³/s can be supplied starting from each portal, or 900 m³/s of fresh air in the entire tunnel. The exhaust is then completely removed through the portals.
In 1980, in the half operated by the French, motorization allowed the exhaust of any of the one-, two- or three-thirds of the half length. On the Italian side, individual motors were installed in 1997 at each exhaust opening. These motors allow exhaust from two, three or four openings.
In 1981-1982, an old service tunnel used during construction and situated in the main tunnel axis, with the connection to the main tunnel located 110 m from the French portal, was modified to exhaust with a maximum flow of 450 m³/s, which was the total exhaust previously coming out of the portal. This was to mitigate air pollution on the French side and permit better control of the air flow in the tunnel.
1.3. Smoke Extraction Capacities
In each of the two tunnel halves, smoke caused by a fire is extracted by reversible duct no. 5 under the roadway. The direction of the air flow in these ducts can be reversed to permit the extraction of smoke in case of fire (approximately 150 m3s for each tunnel half). It has been seen above that the operator can remotely control the activation of all or some of the exhaust openings:
In the French half of the tunnel, there are three exhaust sections and it is possible to concentrate the extraction on any one, two or three of them; in the Italian half of the tunnel, it is possible to concentrate the extraction on two to four openings in the smoke area.
However, the utilization of these possibilities greatly varies the extraction flow actually available.
According to various flow measurements in the past, the smoke extraction capacities are the following:
these values are significantly lower than those measured at the head of duct no. 5 during the startup testing of the original installation, with extraction over the entire length: 179 m3/s on the French side and 224 m3/s on the Italian side.
In France, this is due partially to the fact that the exhaust fans are not being used at full capacity 4/4, but at level 3/4 only; it was indicated that this is due to the risk of fan vibrations.
In France and Italy, the dampers installed for smoke control are without a doubt leaking air. It is also possible that there are other leaks within the ducts and the tunnel.
Given the margin of tolerance due to test conditions and the accuracy of the instrumentation used for the different tests, it can be stated that, according to where the extraction is concentrated, the flow actually removed by duct no. 5 from half of the tunnel may vary between these maximum limits:
- from about 15 m3/s with only one active opening on the Italian side,
- to 100-135 m3/s when all duct openings are active on the French or the Italian side.
Narrowing down to a one-kilometer of extraction, the ventilation capacity varies within these limits:
- in France, from 17 m3/s/km (conservative flow assumption for removal over three thirds) to 44 m3/s/km (optimistic flow assumption for removal over one third),
- in Italy, from 22 m3/s/km (measured in 1998 over the entire length of duct no. 5) to 57 m3/s per 900 m (activation of the first three openings).
It is interesting to compare the smoke extraction capacity of the Mont Blanc Tunnel with those of regulations or recommendations published after the initial opening of the tunnel. This is the subject of the following three paragraphs.
1.3.1. Smoke Removal Requirements in French Bulletin No. 81.109 of 29 December 1981 on Vehicular Tunnel Safety
The 29 December 1981 bulletin, still current at the time of the fire, doesn't exactly apply to the Mont Blanc Tunnel. Signed 16 years after the commissioning of the tunnel, it applies only to new vehicular tunnel construction projects in the French national network, especially those in non-urban areas and of more than 1,000 m in length. However, it is recommended to apply the requirements of this bulletin to older tunnels, if at all possible, especially if this becomes necessary due to changing traffic conditions. Incidentally, it does not apply directly to binational tunnels.
The bulletin sets forth the following requirements (article 3.1 - Ventilation):
"The case of a tunnel fire must be studied and the following practical measures must be taken:
Dimensioning of the mechanical ventilation:
- semi-transverse or transverse ventilation (Mont Blanc Tunnel): provide an exhaust of 80 m3/s/km for a two-way tunnel; the possibility of concentrating this removal in the immediate fire area must be studied.
The levels to be used during fires will be studied case by case by a special commission made up of representatives of the contractor, the Tunnel Design Center and the local directorate of Civil Defense. All measures shall be taken as regards both structure and equipment to avoid the total destruction of ventilation during a fire."
In the Mont Blanc case, the initial actual extraction flow, of 179 m3/s (French side) or of 224 m3/s (Italian side), gave only an average range of 31 to 39 m3/s/km, this being at the time weaker at the portals (about 8 m3/s/km) than in the center of the tunnel (around 45 m3/s/km) due to an increasing exhaust flow toward the middle of the tunnel. It has been shown that the average extraction flow over an entire half tunnel length is from 17 to 22 m3/s. On the French side, if the flow is at its highest, it will result at best in 44 m3/s/km, which is about half of the value required by the 1981 bulletin. On the Italian side, removing through three openings achieves from 47 to 57 m3/s over 900 m, which is still significantly lower than the 80 m3/s/km of the bulletin.
The reasons cited for not increasing these ventilation capacities after the tunnel commissioning include the thickness of the tunnel's rock cover (over 1,600 m at the center) which would have made drilling intermediate ventilation galleries impossible.
It is appropriate to point out that this bulletin, applicable to new projects, is currently being revised (see 1.3.3. below).
1.3.2. Recommendations of the Committee of Vehicular Tunnels of the World Highways Association (PIARC)
On the international level, one single technical association works on the operation and safety of vehicular tunnels and makes recommendations on this subject. This is the World Highways Association (formerly called Permanent International Association of Roadways Congress, which explains the initials PIARC), and more precisely, its Committee of Vehicular Tunnels.
The recommendations of PIARC are generally established during the Roadway World Congresses held every four years. They don't actually provide an extraction flow but give the necessary guidelines to define it.
The report!of the Committee of Vehicular Tunnels at the 1987 Brussels Congress indicates the thermal power and the smoke volume from personal vehicle, bus, truck, or gasoline spill fires and presents ventilation configurations allowing control of smoke propagation within certain limits. For example, a smoke flow of 50 m3/s would require an exhaust of 80m3/s/km of tunnel so that the length of propagation at stable levels would be limited to 600 m.
The report!for the 1995 Marrakech Congress gives operational procedures during fires but does not address exhaust dimensioning.
The report!for the last Congress which took place in Montreal in 1995 complements the Brussels Congress information, by using the results of the Eureka 499 "Firetun" project, but only shows the methods of studying smoke removal, the principles of ventilation, or examines the importance of fire detection and extinguishing systems. Fire probabilities and their analysis are given, but no flow dimensioning is provided.
A report!titled Control of Fires and Smoke in Vehicular Tunnels, issued before the 24 March 1999 fire, has just been published. It contains general recommendations on the use of transverse and semi-transverse ventilation systems (as in the case of the Mont Blanc Tunnel), but it does not give flow dimensioning values. It indicates that the extraction value of 80 m3/s is too low for a fire in a truck not carrying hazardous cargo, but there is no consensus on recommending another value.
1.3.3. Regulations in Proposed New French Bulletin Currently in Preparation
The previously referenced bulletin of 29 December 1981 is currently under revision and a draft has been proposed (before the 24 March 1999 fire). This gives more accurate and complete recommendations than the existing text. For a tunnel like the Mont Blanc, the following requirements must be satisfied, based on a design fire for a truck developing a thermal power of 30 MW and generating 80 m³/s of smoke:
"- The goals are to maintain to the best ability the natural smoke stratification at the higher part of the tunnel to preserve a layer of clean air near the pavement, and to extract the fire smoke by a system of removal at the ceiling level. The exhaust of design fire smoke must be obtained over a length of 600 m for non-urban tunnels.
- These goals are made more difficult by the longitudinal air flow that has to be controlled in such a way that the smoke stratification in the high area of the tunnel be maintained to the greatest extent possible. To this effect, in the fire area the longitudinal air flow speed must be reduced as much as possible. In bidirectional tunnels of more than 3,000 m, the ventilation system has to be able to limit the average speed of longitudinal air flow to 1.5 m/s across the section of the tunnel, even during reasonably adverse meteorological conditions.
- In areas where there may be motorists, it is required to provide a certain amount of fresh air (longitudinally from tunnel areas with smoke or transversely through fresh air registers). However, maintaining the natural smoke stratification, where present, implies not moving fresh air to the ceiling and providing fresh air at the bottom of lateral walls at low levels.
- When the tunnel has a permanent surveillance system, it is recommended to extract the smoke through remotely controlled ceiling openings. This system is the most efficient, if correctly used, as it permits concentration of the exhaust near the fire site. The openings are spaced at a maximum of 100 m in the non-urban tunnels. In all tunnels longer than 3,000 m, this requirement is mandatory.
- The exhaust flow has to be greater than the smoke flow to account for the additional volume of fresh air at the level of removal. An increase of one third will be used for ceiling openings. For the design fire (which develops 80 m³/s of smoke), this results in a minimal exhaust capacity of 110 m³/s. This will be distributed over a maximum of 60 m of non-urban tunnel. Openings that are not located in the ceiling, but laterally at the top of side walls, are less efficient and are not allowed unless the removal capacity is increased significantly with specific justification based on modeling."
This draft document is therefore much more precise and more stringent on several subjects than the 1981 bulletin; although it doesn't yet have regulatory status, it is well known by the experts. The main differences between the recommendations in this document and the systems in the Mont Blanc Tunnel are:
- the smoke extraction capacity, already discussed:
- on the French side, it doesn't go up to 110 m³/s but to 85 m³/s, and it covers 2,000 m as opposed to 600 m,
- on the Italian side, it exhausts either 85 m³/s over 1,200 m, or 35 m³/s over 600 m.
- control of the longitudinal air flow in the fire area, which was never studied (although it seems it must have been lower than 1.5 m/s at the outset of the 24 March fire);
- the supply of fresh air at the bottom of the lateral walls, required by the Mont Blanc Tunnel operating procedures to be put at maximum levels in case of fire, whereas the proposed draft document and widespread practice recommend a very low level.
1.4. Tunnel Operating Procedures
The requirements pertaining to fire ventilation levels are defined by the safety procedures included in the operating regulations common to both companies.
For fresh air, they evolved over the years.
The regulation of 29 March 1967 required that fresh air supply be completely cut throughout the tunnel during a fire.
The regulation of 14 September 1969 had the same requirements.
Then smoke removal tests were performed on 5 October 1972 with a fresh air supply at level 1/2, or half of its maximum capacity.
The regulation of 1 February 1974 directed that the fresh air supply be at the maximum during a fire.
The last regulation, of 23 May 1985, repeats the same requirement.
Over the years, the trend has been to increase the fresh air flow. This is to allow the motorists closer to the fire area to breathe fresh air supplied by openings located every 10 m at roadway level.
But this fresh air feeds the fire and heats the air in the tunnel if it comes at high speed and in great quantity.
The operating procedure of 23 May 1985, applied to the Mont Blanc Tunnel, doesn't appear to have been appropriate.
As for the smoke extraction, the safety procedures of the Mont Blanc Tunnel never changed. All operating requirements issued directed that exhaust be set at the highest level.
1.5. Traffic Rules for Moving and Halted Vehicles
1.5.1 Distance Between Vehicles
There are tunnel traffic rules included in the prefectoral ordinance of 31 January 1997, which require a distance of 100 m between moving vehicles (and 200 m for hazardous cargo vehicles). This rule aims to prevent chain-reaction accidents but does not protect stopped vehicles from the risk of fire spreading nor does it protect motorists from smoke. At the beginning of the tunnel operation there was a detection system for vehicles not keeping to the 100 m distance with a signal alarm, but that system has been removed. From statements by the operating companies, it was not obeyed and some drivers had fun triggering it; it caused maintenance problems.
1.5.2. Speeds and Passing
The regulation sets both the maximum (80 km/h) and minimum (50km/h) speeds and prohibits passing. Joint teams of French and Italian police were tasked to enforce these rules. The operators repeatedly notified police of nonobservance of these rules.
1.5.3. Traffic Signals
There are traffic signals every 1,200 m in the tunnel (at rest areas 3, 7, 11, 15, etc.). They are not located above the lanes, but on the sides; they are three-color signals.
1.6. Evolution of Traffic Since the Opening
As shown on the following graphic, the increase of traffic has been constant since commissioning of the tunnel on 19 July 1965. For tourist vehicles (passenger cars, motorcycles, and tour buses), the numbers went from 548,000 in 1966 (the first full year of tunnel operation) to 1,221,000 in 1998, or more than double (over 33 years, not a significant increase). For trucks, the increase in traffic was much greater, since it went from 45,000 in 1966 to 777,000 in 1998: the truck traffic was multiplied by a factor of 17 in 33 years. In all, this accounts for more than 45 million vehicles, of which a third were trucks, that had passed through the tunnel by 31 December 1998. Moreover, the size of trucks has significantly increased.
The following table shows the distribution of 1998 traffic by category and direction. It should be noted that trucks represent 39% of the total, and tour buses 1.4%.
However, the significance of the preceding numbers must not be misinterpreted. If the Mont Blanc traffic, especially the trucks, increased greatly, it still remained relatively low as an absolute value. A traffic flow of 5,473 vehicles per day is much lower than a roadway of the same capacity can take.
Actually, the capacity of a tunnel is set not only by its geometry but also by its equipment and safety setup. The toll revenue generated by this traffic was 518 million francs in 1998.
1.7 Comparison with Similar Vehicular Tunnels
There are 14 vehicular tunnels of longer than 8,000 m currently in service in the world. Five of these tunnels have two bores (three in Japan, one in Italy, one in Switzerland). Three others have only one bore, but their traffic is very low (Norway and Japan). The remaining six tunnels (including the Mont Blanc) have only one bore with two lanes and significant two-way traffic, and their features are relatively similar.
A longer tunnel, with a single bore 24,500 m long, is currently being built in Norway between Aurland and Laerdal.
In the following table, only the Mont Blanc and Fréjus tunnels cross national borders.
The other major border tunnels, especially the Karavanken (7,864 m long) between Austria and Slovenia and the Grand-Saint-Bernard (5,828 m long) between Switzerland and Italy, are shorter.
The Mont Blanc Tunnel is the oldest of all these preceding tunnels. Only Switzerland has tunnels commissioned at around the same era but they are much shorter (San Bernadino with a bore 6,600 m long opened in 1967, Grand-Saint-Bernard with a length of 5,828 m opened in 1964).
The table includes only toll mountain tunnels, except for the Saint-Gothard tunnel where the regular highway tax sticker is required. Only at the Mont Blanc and Fréjus tunnels are there barriers at the entrance toll plazas.
The pavement of these tunnels varies in width from 7 to 7.8 m, except in the Fréjus and Gleinalm tunnels, where it is 9 m wide. The authorized vehicular width is 4.5 m, except in the Saint-Gothard tunnel where it is 4 m.
The horizontal alignments include wide curves or sharper curves toward the ends.
The vertical profiles have a high point, except for the Fréjus tunnel, which is on a continuous slope of 0.54%. The Mont Blanc Tunnel has a steeper slope (2.4%). Following it is the Arlberg tunnel, with a slope of 1.67%, then the Saint-Gothard tunnel, with 1.4%.
1.7.2. Means of Evacuation and Protection of Motorists
Only the Saint-Gothard tunnel has a parallel safety tunnel, pressurized and with access from the main tunnel every 250 m. The Mont Blanc and Fréjus tunnels are the only ones with pressurized refuges. In the Saint-Gothard tunnel, the access galleries between the main and the safety tunnels are built as refuges. The Austrian tunnels have no safety tunnel or pressurized refuges.
Only in the Fréjus tunnel can the fresh air duct be used for access by rescue teams.
Only the tunnel of Gleinalm has no rest areas. In the other tunnels, a passage across from each rest area allows truck U-turns (with difficulty).
1.7.3 Safety Equipment
From the equipment standpoint, these six tunnels are relatively similar. They all have safety niches in their side walls. Because of the location of the fire standpipe, the fire niches are only on one side, whether or not they are in the safety niches.
They all have interconnected electrical power at each end.
The tunnel lighting is carried on two distinct circuits in the Mont Blanc and Fréjus tunnels, while it has only one circuit in the Swiss and Austrian tunnels. Only in the Arlberg tunnel is there a continuous drainage system for roadway spills.
The ventilation system is transverse in all cases. The six tunnels, with the exception of the Mont Blanc, have several intermediate ventilation rooms connected on one side with vertical ducts and on the other with two ducts (supply and exhaust) in the tunnel ceiling. Because of the thickness of its cover, the Mont Blanc Tunnel has only one ventilation room at each portal. The numerous ventilation ducts are located under the roadway.
The ventilation system in the Mont Blanc Tunnel has been modified to account for the increase in traffic and to improve smoke removal. The French part of the Fréjus tunnel has recently been equipped with remote-controlled exhaust openings. The other tunnels do not have remote-controlled exhaust openings.
All the tunnels have traffic signals, an alarm network, CCTV, two-way radio, and a fire detection system (except the Fréjus tunnel), managed from a unique command and control center (PCC). The exception is the Mont Blanc Tunnel, which has two PCCs, one at each portal, independent of each other. The Saint-Gothard tunnel also has a PCC at each end, but they alternate every two weeks.
1.7.4 Hazardous Cargo
Hazardous cargo transit is regulated in the six tunnels. In the Mont Blanc Tunnel, it is allowed only in limited quantities. In the Saint-Gothard tunnel, it is allowed in larger quantities. It is the same in the Arlberg tunnel, which has either hour restrictions or mandatory escort. Just recently, all authorized hazardous cargo in the Fréjus tunnel is required to have a double escort. There do not seem to be any restrictions in the Plabutsch and Gleinalm tunnels.
1.7.5. Response Team
The Saint-Gothard tunnel has an around-the-clock firefighting team at each end of the tunnel for the quickest possible response. After the severe 1997 truck fire, the team changed from three to four firefighters. The additional person is especially responsible for the evacuation of motorists from their cars to a safe area.
In the Fréjus tunnel, there is a first response team with a fire engine at each entrance; the technical staff is made up of voluntary firefighters. They are trained to respond to any incident inside the tunnel with the equipment provided by the safety chief.
In the Mont Blanc Tunnel, a team made up of a firefighter and volunteers is stationed at the French entrance and has specialized vehicles; it can be mobilized for response throughout the tunnel (see chapter 2).
|2. ORGANIZATION OF TUNNEL SAFETY|
2.1. Original Institutional Framework
The original framework sought to closely tie safety and operations through unified management of the tunnel. The 14 March 1953 convention between France and Italy relating to the construction and operation of a tunnel under Mont Blanc established the role of the toll companies and of the governments in the operation and safety of the tunnel. Two organizational levels were defined. On the one hand, the French and Italian toll companies (article 7) were supposed to relinquish the common management of the tunnel to an operational branch. This French-Italian company would have had responsibility for ensuring global consistency of the operation and therefore safety of the tunnel. On the other hand, a French-Italian control commission (article 8), an intergovernmental body, received the mission to control the operation, maintenance and preservation of the tunnel.
This framework was repeated and defined by the lease documents and contracts of 7 July 1959 that specified the safety obligations of both companies. Article 7 of the French lease document returns to the principle of one French-Italian branch "for the operation on both the French and Italian sides". A framework "to limit as much as possible the danger of a vehicular fire in the tunnel" was requested of the companies (article 16 of the French contract). Additional works to ensure a better operation of the tunnel or better safety could be proposed by the government and the lessee (article 18). It was also requested of the lessee to establish rules for the operation and "the functioning of the safety system both for appropriate ventilation and firefighting measures".
2.2. Common Administrative Organization
The institutional organization based on a common operating company was never accomplished. At the opening of the tunnel in 1965, a so-called temporary and flexible coordination framework was established between the French and Italian companies. It endured to this day, following several major modifications.
The pact of 23 March 1965 established (article 1) that the tunnel would temporarily be operated by both toll companies working together, as the common branch was not yet set up.
To limit the effects of this institutional change departing from the principle of a unified management, article 2 of the pact required that each company commit to taking all measures necessary to operate the entire tunnel. A common administration committee was created to replace the common branch. This committee was responsible for organizing the liaison between the two companies and taking all measures necessary to operate the tunnel, "as a branch company would have done". The collaboration agreement of 27 April 1966 between the two companies specified, "pending the development of the statutes of the common operating company", the operational framework of the common administration committee. A single operations director, employed by either one of the two companies, was to be named by both companies (article 6), although his authority was not clearly defined. Functionally, the operations director would appear to act on the decisions of the common committee, although actually appointed by the two companies.
The shortcomings of this compromise, acceptable for a temporary framework but crippling for the long term, have had important consequences on the efficiency of tunnel management since 1966. Little by little, the operation moved toward two overlapping managements for the two halves of the tunnel, without sufficient coordination. Thus, since 1968, a minor disagreement on the operation of the tunnel has been reflected in the works of the common committee. The companies were debating over the hierarchical authority -- common committee or employing company -- of the operations director. In 1972-1973, the committee decided, without modifying the 1966 agreement, to substantially modify the principle and the organization chart of the common management. Two operations directors, French and Italian, were placed at the head of the common management. This two-headed organization repeated and reinforced the uncertainties of the 1966 agreement, the two directors being appointed by the common committee but "having decision powers given by the two companies".
In 1979, the separation into two operating entities was finalized. By then, each of the two companies had a full-time operations director specifically responsible for half of the tunnel. This new reform did not, however, eliminate the need for a common operation. In February 1999, a new debate shook the common committee, with two opposing positions. For some, the two directors should continue to report!to the common committee while, for others, they should both represent their companies and act according to their companies' instructions.
In this context, the capital improvements for safety, especially for ventilation, were planned and implemented without coordination between the two companies. The common administration committee saw its funding decision-making powers limited, or even opposed, by the two companies, on their way to "take control". This was the case with the updating of the ventilation openings, implemented in 1980 by the French company and in 1997 by the Italian company, and for the tunnel safety service organization. Similarly, to date, the tunnel operation and control appear only partially coordinated. For example, at the
time of the accident, the ATMB operator was not fully aware of the status of the ventilation on the Italian half, and vice versa. Aware of this shortcoming, in 1990 the companies started to implement an upgrade program that they hoped would be coordinated. The common committee supervised this program. This flexible approach to the coordination of capital improvements initially started with lighting, wall finishes, and tunnel signals. The two companies mutually informed each other of their progress and tried to coordinate it. An additional step toward a strengthened coordination of safety capital improvements was made starting in 1996 with the centralized technical management program, then in 1998 with the automatic incident detection program. In these two projects, one company was designated as leading the study and contract formation for the entire tunnel.
Finally, it should be noted that the coordination and consistency of capital improvements has been set back seriously by different procedures of the two companies. These procedures, particularly cumbersome for the Italian company, explain the major delays between the deciding on and the implementation of the capital improvements.
2.3. The Intergovernmental Control Commission
The French-Italian control commission, made up of representatives of the two governments, is responsible for the control and supervision of tunnel safety, according to article 8 of the 1953 convention. The 1 March 1966 exchange of correspondence between the French and Italian foreign ministers directed that the commission control the execution of additional works needed to improve the tunnel operation and to increase its safety.
However, the control commission appears to have met several obstacles in this mission.
First, the inexact definition of responsibilities and the absence of formal ties between the commission, the common committee, the common management and the two companies appear to have presented an obstacle early on. In 1967, the common committee pointed out to the government that the control commission could not give instructions to the companies and, even less, to the common management. But it was above all on the issue of safety that the relationship between the control commission and the common committee failed. On its side, the common committee had protested since 1967 the terms of the 1 March 1966 correspondence confirming the control power of the commission. The latter regretted not being informed until after the fact of the work inside the tunnel and recalled, in support!of their case, the need for the tunnel to appear to the motorists as a unique and homogenous entity, both in its operation and its looks.
In 1988, a legal study was finally conducted to clarify the authority of the control commission to approve contracts. A classification into three categories was adopted, in strict adherence to the letters of 1 March 1966. While the commission would simply be informed after the fact about regular maintenance work and before the fact for major repair work, it has to review any supplementary construction. This classification has been correctly applied since 1990 with regard to the updating of the tunnel.
Secondly, starting in 1973, the rapid growth of vehicular traffic, especially trucks, led the control commission to focus its efforts on two recurring and sensitive subjects, the adjustment of tolls and access to the tunnel, while disregarding safety issues, which took a back seat in the discussions. The rules of hazardous cargo traffic in the tunnel, which the commission has the authority to submit, were regularly addressed. Likewise, in 1982, the emergence of a mutual aid project between the two countries in the case of a disaster was initiated by the commission. A draft convention was put together by the Foreign Ministry. This convention was only completed in 1992, and it was not published until 1995.
Finally, the ever-changing composition of the commission, difficult to avoid since the majority of its members were appointed, plus its formality and the frequency of its meetings, appear to have raised a third obstacle to a quality follow-up of the safety issue.
2.4. Organization of Rescue Efforts by the Operators
2.4.1. Original Documents Contain Few Specific Instructions on this Subject
The lease contract of 7 July 1959 states that the tunnel operation is at the risk of the lessee (article 7), which has to ensure its safety and operational continuity (article 8).
The operating procedures do not further specify the operator's organizational duties and the assets to be used in case of fire. An operating procedure must nevertheless provide for a fire safety system.
2.4.2. Organization of the Operators' Rescue Assets
The operating procedures of 23 May 1985, common to both operators, describe the alarm procedures, the rescue vehicles at each portal, the roles of the operators and motorcycle patrols and, only for the French side, firefighters.
A major organizational difference can be seen between the available assets on the French and Italian sides:
On the French Side
- 6 professional firefighters (all volunteers at the rescue centers of Chamonix or d'Aiguebelle) work 3 shifts of 8 hours;
- 6 motorcycle patrols work on the same schedule;
- the toll takers are all would-be volunteers during a potential rescue operation.
Every day, a firefighter and a motorcycle patrol are on duty.
Complemented by the toll takers, the team can go up to 10 people a day, with a minimum of 4 people at night. The team is available 24 hours a day.
As for the fire and rescue vehicles:
- an FPTL (2,000 liter light pumper engine) with extinguishers and individual breathing devices (ARI);
- a PS (600 liter first rescue vehicle) with extinguishers and ARI;
- an ambulance (VSAB).
On the Italian Side
- a team of eight motorcycle patrols;
- a multi-use fire vehicle with three extinguishers staffed by a driver "accompanied by voluntary Italian employees for rescue operations".
The operating and safety procedures state that "inside the tunnel, the firefighter is responsible for the use of vehicles and equipment, the response techniques, as well as the safety of motorists and employees". The first response service available on the French side is therefore aimed at the entire tunnel.
This response method actually excludes the scenario on 24 March when the first response could come only from the Italian side. In fact, aside from the motorcycle patrol, this entrance had neither firefighters nor volunteers, since for reasons of principle none of the employees had signed a response agreement (minutes of the common administration committee of 25 October 1972).
2.4.3. Explanations of the Choice of a Rescue Service on the French Side for the Entire Tunnel
In 1971-1972, the two companies acknowledged in the common administration committee that they had to improve their rescue assets. A decision was made to set up a permanent rescue team made up of a firefighter, a coordinator and a driver.
However, the local circumstances eventually led to a situation depriving the Italian side of both firefighters and volunteer operators for combating fires.
A review of the minutes of the common administration committee shows that the two companies initiated contact with Chamonix and Courmayeur. Finally, ATMB decided to hire five firefighters to be located at the French portal and placed under the common operator while, on the Italian side, the SITMB was being asked by the administration of Aoste for a financial contribution to implement a fire station at Courmayeur.
The investigating commission did not find, in the 1971 and 1972 minutes of the intergovernmental control commission, any trace of these decisions, nor its own formal agreement to this asymmetrical rescue organization by the operators.
2.4.4. Firefighting Drills
Since the implementation of this service, the common administration committee sought to organize common drills between the tunnel and the rescue centers of Chamonix and Courmayeur.
Two such drills were organized and were the subject of reports by the fire and rescue services of Haute-Savoie. The first (15 October 1972) addressed smoke tests in the tunnel. The second (25 March 1973) was a real rescue drill including the firefighters and the tunnel rescue teams, but involved only the French teams.
The conclusions of the report!by
Lieutenant Colonel Leverge were formal:
"in all circumstances and at any time during the year, four specially trained men must start simultaneously from each side immediately after the alarm is triggered, without regard to the size of the accident or the support!of the rescue centers of Chamonix and Courmayeur".
These drills led the committee to increase the personnel presence during the night on the French side. No other follow-up was mentioned in the following meeting minutes.
The commission did not know of any other fire drills inside the tunnel. The 1989 drill, according to witnesses, took place on the Italian side and was aimed at training the Italian firefighters in rescuing trapped motorists from their vehicles.
The common operating procedure does not actually require the organization of drills for the tunnel teams alone or with adjacent public service firefighters.
The major fire of January 1990 did not lead the operators to plan further, although the common administration committee meeting minutes of 26 January 1990 state that all lessons learned from this fire will be applied.
The question would be frequently raised by the French firefighters within the framework of cooperation started in 1996, but without actually resulting in a French-Italian drill. The prefect of Haute-Savoie had planned for 1999 a rescue drill inside the tunnel with the goal of establishing a first response plan common to Haute-Savoie and Val d'Aoste.
2.5. Public Safety Plans
There is no common French-Italian public safety plan.
On the French side, the specialized safety plan developed by the Haute-Savoie prefecture is dated 1 July 1994. Currently, a specialized plan is not mandatory for tunnels.
On the Italian side, the equivalent safety plan (piano di emergenza) was developed in 1995 by the civil defense services of the autonomous region of Val d'Aoste, the president of the region acting as the prefect in matters of civil defense.
The two plans are very different in both form and content.
The French plan seems more complete. It recalls the role and the methods of first response of each operator. But it also indicates, wrongly, in an appendix titled "Participation of Rescue from Val d'Aoste" that, on the Italian side, the first response is provided by a team of 2 to 6 operators. Actually, as noted earlier, besides the motorcycle patrols, there are no firefighting team and no volunteer operators for first response on the Italian side. The Italian plan does not mention this team and it only refers to the three-extinguisher fire vehicle that the operators were not trained to use, as the SITMB stated to the investigating commission.
Regarding the possible support!to be provided by a government on the side of the tunnel under the jurisdiction of another government, the two rescue plans refer to the establishment of an ad hoc agreement. However, the Italian plan proposes that the Italian and French rescue teams work in collaboration throughout the length of the tunnel under the coordination of the jurisdictional government authority.
2.6. French-Italian Cooperation
Since 1996, the organization of the tunnel safety has been the focus of genuine French-Italian efforts, but with no real results so far.
The starting point was the 10 September 1995 French-Italian convention on mutual assistance during disasters. Article 16 of the convention states that agreements and special arrangements may be reached to establish the first response conditions for the neighboring country in the Mont Blanc Tunnel. On this basis, between 1996 and 1998 the civil defense services of Haute-Savoie and Val d'Aoste and the firefighters from each country developed an administrative agreement setting the practical framework of requests for assistance, and planning common communications and training activities, as well as drills. This proposed agreement did not, however, get signed, pending the Rome agreement received by the president of the Val d'Aoste region... on 24 March 1999.
Besides working out of this agreement, the development of regional French-Italian cooperation has been intense, especially in 1997 and 1998, with the goal of acquiring the INTERREG European funding. It is thus that two agreements to improve safety in the tunnel were signed on 2 October 1997 and on 19 February 1999 between the prefect of Haute-Savoie, the president of the Val d'Aoste region, and the president of the administrative board of the fire and rescue services of Haute-Savoie. Activities to inform and train firefighters from Haute-Savoie and Val d'Aoste, common drills, and the construction of a firefighting training facility for response in confined spaces have been planned and introduced for EU financing. Only the activities to inform and train the French and Italian firefighters, by making tunnel visits, took place before the 24 March 1999 fire.
|3. DEVELOPMENT OF EVENTS|
3.1. Traffic Conditions and Tunnel Functioning, Before and After the Alarm
3.1.1. Traffic Status Before the Alarm
The traffic was of average intensity before the alarm, but increasing:
- from 08:00 to 09:00, in the direction France-Italy, 131 vehicles, of which 82 were trucks,
- from 09:00 to 10:00, 163 vehicles, of which 85 were trucks.
In the minutes preceding the alarm, the number of vehicles coming from France passing through the tolls averaged 4 per minute. If a tunnel speed of 60 km/h is assumed, the average distance between vehicles would be 250 m. If a tunnel speed of 90 km/h is assumed, the average spacing would be 375 m. In actuality, two vehicles or several could have followed each other more closely, but it cannot be said that there was congestion. In fact, traffic was flowing.
The fact remains that, during the 9 minutes separating the entry of the truck which caught fire first (10:46) and the tunnel closure at the time of the alarm (10:55), one motorcycle, 10 passenger vehicles, including one pickup truck, and 18 trucks entered the tunnel. Four trucks passed the truck on fire after it had stopped. There were 26 vehicles caught in the fire, including the motorcycle.
At 10:55, five vehicles had passed the tolls in that minute, but 2 out of the 5 continued into the tunnel. The 3 others were stopped by the ATMB siren.
3.1.2. Alarm Conditions
A detailed chronology of the facts is found in the Appendix. Only included here is the part that appearing to the task force to have particular importance.
The access closure at the two tunnel portals occurred at 10:55, i.e., within the first one or two minutes following the alarm. It should be noted that the clocks are not coordinated between the two tunnel entrances, nor between the various teams, and that the times indicated can differ, despite the recalculations effected, by about 1 minute.
The truck that ignited and started the fire had entered at 10:46.
9 minutes passed between these two events. What happened in the meantime?
The truck stopped at the toll plaza at 10:46. It started to enter the tunnel; accelerated, seemingly reached its cruising speed and then, realizing it emitted smoke, slowed down and stopped at rest area 21; it was about 10:53. We know from prior studies that the average speed of trucks going from France to Italy is about 56 km/h. This is consistent with the fact that the truck covered 6,700 m from the toll plaza between 10:46 and 10:53, in about 7 minutes. It can be concluded that, when the smoke was observed on the screens at 10:53, the truck had just stopped (or was close to stopping).
The functioning of the other alarm equipment:
At 10:52 (French time) or 10:51 (Italian time) in rest area 18, the obscuration monitor gave a "coefficient Westinghouse" value of more than 30%. That set off an audio and visual alarm at the French operator's station. That station is equipped with such alarms, which do not exist on the Italian side. In the same minute, the obscuration meter in rest area 14 had also indicated surpassing 30% (maximum saturation of the obscuration monitor). But that was not reported on the main board in the control room.
On the French Side
The 30% saturation level of the monitors in rest areas 14 and 18 at 10:52 had automatically activated the display on the control screens of the camera images for this zone, without automatically triggering the register.
At 10:53 the French operator cleared the alarm, i.e., indicated to the system that it had been acknowledged. He observed the camera images at rest area 18 and those at rest areas 16, 17 and 19 and saw the smoke in the tunnel.
The French fire detection system measures temperatures and detects those which go beyond 50B at sensors located every 8 m. The sensors did not trigger during the truck's travel, which is not unusual. They did detect the high temperatures later on.
The temperature measurements are recorded and kept in memory for four days, then are automatically replaced by the measurements of the day. Unfortunately, in the four days following the outbreak of the fire, the system was not stopped and these numbers were not collected, and therefore are no longer available.
On the Italian Side
The Italian fire detection system works on a different principle (based on heating of a gas in tubes 70 to 80 m in length). According to the Italians, the system gave frequent false alarms and, because of one of these coming from the sensor in rest area 21 (where the Belgian truck on fire stopped), the relevant section had been placed out of service the night before.
Several other elements confirmed the alarm:
- a phone call around 10:54 from a person at rest area 22, received by the Italian control center,
- an alarm from rest area 21 around 10:57 (usage of a fire pullbox), followed around 10:58 by an alarm showing the lifting of a fire extinguisher in the same rest area.
3.1.3. Tunnel Closure
On March 24, when the alarm had been given by the French and Italian control centers after telephone communications and following the appearance of thick smoke on the control screens of both countries between rest areas 16 and 21, the siren was set off at the French portal at 10:54. At 10:55, all the traffic signals in the France-Italy direction turned red. A truck that was entering was quickly backed up to make way for emergency access, and vehicles which had already cleared the tolls were removed from the entrance. The Italian entrance was simultaneously closed at 10:55 and 10:56, by setting off the siren, turning the signals to red, and closing the gate.
3.2. Rescue Operations
3.2.1. The Alarm
On the French Side
The Central Alarm Center (CTA) in Annecy, where all communications are clock-dated and recorded, was alerted at 10:58:30. It immediately forwarded the alarm to the Main Rescue Center (CSP) in Chamonix, allowing the first rescue vehicle to leave its base at 11:02 and arrive at the tunnel at 11:10.
With regard to for the alarm and dispatch of the first rescue vehicle, there was no unusual delay.
On the Italian Side
The Courmayeur firefighters were alerted at 11:02, according to the telecommunications operator's records.
The first firefighting vehicle from Val d'Aoste left the Courmayeur rescue center at about 11:04 and arrived at the tunnel entrance at 11:11.
3.2.2. The Response
On the French Side
The initiation of the private and public rescue tunnel operations was carried out in accordance with the "Mont Blanc Tunnel" Specialized Rescue Plan, adopted by the Haute-Savoie Prefect on 1 July 1994.
Two rescue plans were activated by the Haute-Savoie Prefect:
- at 13:04 the Specialized Rescue Plan for the tunnel,
- at 13:35 the Red Plan, given the foreseeable number of injured after the entrapment of the firefighters and ATMB personnel in the tunnel .
Two Command Posts were installed:
- the base Command Post (CP) at the Prefecture at 14:00, led by the chief of staff who, as of the following day, together with the mayor of Chamonix attended to the sensitive issue of calling the affected families and determining the number of victims;
- the site CP at the ATMB facility at the French entrance. The Bonneville deputy prefect, present at the site since 15:00, represented the Prefect in leading the rescue operations.
Lt. Colonel Laurent, director of the regional rescue and fire fighting services (SDIS), took command of the rescue operations, starting at 13:32.
The SDIS had anticipated the decision to activate the Red Plan and deployed the major assets requested by the plan (rescue and paramedic vehicle [VSAB] - helicopters, etc.).
This brought out the assets called for in the first and second stages of the Specialized Rescue Plan.
On March 24 at 15:00 the following assets were dispatched on the French side: 26 firefighting and rescue vehicles and 98 firefighters, including ten equipped with airpacks (ARICF).
A medical outpost was installed at the site and staffed by SAMU and SDIS personnel, but would be used only for the benefit of the rescue workers, since nobody could reach the motorists trapped inside the tunnel.
On the Italian Side
The activation of the Italian rescue plan (piano di emergenza) was not initiated by the president of the Val d'Aoste region, responsible for civil defense, who did not consider the plan justified, based on the information given to him by his staff around 12:00: no need for coordination presented itself, the situation relating only to the firefighters already dispatched. The regional services took the necessary actions concerning traffic.
The rescue operation was directed by the fire chief of the province of Aoste, Engineer Badino, and by Captain Marlier, the Courmayeur fire chief, without the formal installation of an operational command post in the SITMB center, or setting up a means of communication for contact with the French firefighters.
3.2.3. Information Given to Public Authorities and Rescue Services
It is important to emphasize the fact that the French and Italian public authorities and the rescue services did not have, during the course of the first day and even until Thursday, information from the operators on the possibility of there being motorists trapped in the tunnel.
The density of the smoke from the first minutes on the French side had rendered the surveillance cameras incapable of reporting the number of vehicles and people trapped in the tunnel. However, a pullbox alarm was set off in rest area 22 on the Italian side, and an alarm followed by removal of a fire extinguisher was registered at 10:57 on the French side in refuge area 21.
These uncertainties would last until Thursday at 22:30, even after the discovery of three victims on Wednesday at 19:04 in the vicinity of rest area 18. According to the ATMB, this can be explained by uncertainties in interpreting, without risk of error, the information lists from the French toll plaza, given the lack of a system for counting vehicles present in the tunnel or control at the exit.
3.2.4. First Response by French Rescue Assets
Only the main events will be recorded here, with the detailed chronology included in the Appendix.
From 10:55 to 11:36, all rescue vehicles entering the tunnel were successively blocked, far from the Belgian truck on fire.
The chronology of the stopped vehicles indicates the speed of the smoke propagation:
10:53 - stoppage of the Belgian vehicle in the tunnel at the level of refuge area 21;
10:55 - an ATMB agent enters the tunnel and is stopped shortly after rest area 18, i.e., about 750 m from the Belgian truck;
10:57 - the ATMB fire engines enter and are blocked at refuge area 17 (about 1200 m from the truck);
11:10 - the first Chamonix vehicle enters (FPTGP with six men) and is blocked at rest area 12 (about 2700 m from the truck);
11:36 - the second Chamonix vehicle enters (FPTL with five men) and is blocked at refuge area 5 (about 4800 m from the truck).
Priority was then given to rescuing the 17 people who were trapped. This operation began at 12:55 with the entry into the duct no. 5 of a team directed by Captain Comte, head of the Chamonix Rescue Center, through the ventilation ducts. It ended at 18:35 with the evacuation of the six ATMB personnel.
This effort!to rescue the trapped firefighters and operating personnel was efficiently carried out. It was very dangerous.
The courage of the rescuers, like that of the trapped personnel, certainly led to the avoidance of a greater human loss. It must, however, be recalled that a junior officer of the Chamonix center died shortly after his evacuation and a total of 14 firefighters were hospitalized.
On the French side, it was therefore impossible to attempt any rescue operation of the motorists, about whom it must be remembered that the rescuers did not know anything and whose survival time, for those who had not found shelter, must have been short because of the enormous quantities of carbon monoxide and other toxic gases generated by the fire.
The emergency equipment was adequate; in all, the assets of four departments were mobilized, to which must be added the firefighters and the marine firefighting equipment of Marseille and Geneva. It was impossible to use them on Wednesday to combat the fire, despite the very short reaction time of the ATMB safety personnel and the Chamonix firefighters.
3.2.5. Organization of Italian Rescue Assets
Between 10:57 and 11:01, the tunnel was successively entered by an SITMB agent, the French patrolman, who was at the Italian portal at the time of the alarm, and the three-extinguisher fire engine, i.e., three men and three vehicles. The second, in his vehicle, crossed through the smoke between rest areas 21 and 22 and came within some 10 meters of the Belgian truck on fire, stopped at rest area 21. Exposed to danger, he had to back up and quickly return to rest area 22, where he found his Italian colleagues. Together they decided to evacuate the drivers from the trucks (including the Belgian driver), whose vehicles were stopped in the Italy-France direction about 300 m from the truck on fire. The passenger vehicles had succeeded in making U-turns and exiting toward Italy.
The first vehicle from the Courmayeur Center (FPTGP with three men) arrived at the portal about 11:11. It entered the tunnel and arrived at rest area 22 at about 11:16. It could not continue its path toward the truck on fire.
Blocked, they tried to proceed toward the truck on fire, first with their vehicle, then on foot, carrying their breathing devices. But, unable to see through the smoke, they had to turn around and return to rest area 24. They were joined there at 11:45 by two firefighters in a vehicle from the d'Aoste center carrying airpacks (ARICF).
Sheltered since 12:02 in rest area 24, the five firefighters were evacuated at about 15:00 by Captain Marlier, head of the Courmayeur center who reached them by ventilation duct no. 5.
For March 24, the count on the Italian side: 9 vehicles, 10 men with 2 airpacks and 19 breathing devices from the Courmayeur and Aoste centers. The Turin assets were not mobilized on either Wednesday or the days following, the rescue chief preferring to rely on the firefighting teams familiar with the tunnel.
3.2.6. Comments on Organization and Implementation of the Rescue Services
The PCO (Operation Command Post) on the French side was set up in the offices of the ATMB, since there was no planned control room. The available office was small and lacked an adequate number of communication equipment. The head of the rescue operations had to share the ATMB safety director's office, where his only means of communication was the director's telephone.
The setup of the PCO was not without consequence on its functionality. According to witnesses, the operator played an important role in the rescue operations, going beyond his normal role of technical adviser in carrying out a rescue plan.
The advanced medical post, thanks to its own setup, functioned well.
3.2.7. Smoke Reversal Attempts
On Wednesday the 24th and Thursday the 25th, the operator made attempts to reverse the direction of the smoke and facilitate the access of the French firefighters who were unable to proceed due to the smoke and heat.
All these attempts did not succeed in facilitating the operations of the French firefighters, and probably delayed the decision to transfer the assets to the Italian portal in order to concentrate them toward the entrance where it was possible to proceed, as testified by the Italian firefighters who, as of Thursday morning, had extinguished the fire in the trucks stopped in the direction Italy-France and approached the truck which started the fire.
From the beginning of the fire, the tunnel situation was notably different on the French and Italian sides, due to the direction of the smoke:
- at about 11:05, the French patrolman was about ten meters from the Belgian truck (on the Italian side), while the first firefighters' vehicle from Chamonix was stopped 2,700 m from the truck on the French side at 11:15;
- between 11:20 and 11:30, the Italian firefighters came within about 300 m of the truck, then were forced to go back to refuge area 24, which was 900 m from the truck, while the second Chamonix firefighters' vehicle was stopped at 4,800 m.
3.2.8. Contacts Between French and Italian Firefighters
Until Wednesday evening, communications were few: the fire chiefs informed each other briefly of their response activities, particularly concerning the evacuation of people trapped in the tunnel.
Thereafter, lacking liaison officers and also due to the communications difficulties previously mentioned, they were not in a position to examine the overall situation together and arrive at a common strategy.
When, during the night of Wednesday into Thursday, an offer of a "fresh" reinforcement team equipped with breathing devices was proposed to the leaders of the Italian rescue effort, it was first accepted and then declined, due to the expected arrival of Italian reinforcements.
A clear but after-the-fact review of the assets mobilized at each portal shows that the proposed transfer of assets toward Italy was wise and timely, coming immediately at the end of the rescue operations on the French side.
The lack of experience and operational coordination, lack of practice and the difficulties with equipment and communications explain in part the delay of the proposed concentration of the assets on the Italian side.
It must be recalled that the proposed transfer of assets was considered on the French side and decided upon Thursday morning, but could not take place until the night of Thursday into Friday because of bad weather conditions on Thursday afternoon.
3.2.9. Firefighters' Response
This fire, exceptionally serious due to its strong caloric power, presented particular extinguishing difficulties due to its location and the impossibility of approaching it from the French side because of the density of the smoke. It should be noted that it took 53 hours to extinguish the fire.
The only person who saw the vehicle that started the fire was the French patrolman, at about 11:05. He has repeated his testimony that, at that moment, the fire could have been fought. Although this patrolman is not a professional firefighter, he did participate in extinguishing the truck fire in January 1990.
It must be stated, however, that even if the fire had been addressed at that moment, it cannot be confirmed that it could have been extinguished and the motorists saved, given the density of the toxic smoke which had already permeated the tunnel where the motorists were trapped at 10:55 (the time when the first ATMB agent was stopped on the French side at 750 m from the truck on fire).
All the French and Italian firefighters experienced great hardship with their response: almost no visibility, extreme heat, and great difficulty with airpacks in a very hot environment. The status report!mentioned the need for training before using this type of equipment, and the fact that the ATMB breathing devices were not compatible with those of the firefighters. In general, all the team leaders have insisted that this kind of response requires a high level of physical and psychological strength, in addition to special training.
The response also revealed technical problems: insufficient water pressure on the French side of the tunnel, short-circuited pump, standpipe connection incompatibility for the different teams, and communication problems inside the tunnel, a portion of the communication equipment having been quickly damaged by the fire, complicating and increasing the psychological pressure on the firefighters during their rescue efforts.
Finally, the response proved the important role of foggers, relatively rare equipment brought in from Marseille and Geneva on the Italian side.
3.3. Results of the Catastrophe
3.3.1. Human Loss
Besides the two ATMB emergency vehicles, 23 trucks, 10 passenger vehicles and 1 motorcycle were destroyed.
All these vehicles were between rest areas 19 and 23, in the tunnel half operated by the Italian company, close to the part operated by the French company (see attached map).
38 victims were identified. Among these, 27 were found in their vehicles, 2 in other vehicles, and 9 outside of vehicles. To these must be added the death of Chief Tosello of the Chamonix fire department after his evacuation.
Two drivers of trucks up front, and thus close to the fire, left their vehicles and fled toward France. They probably died of asphyxiation, after having gone about 200 to 240 meters. We do not know what their speed was: whether they walked quickly or ran and covered this distance in 1 to 2 minutes, or whether they were delayed by the beginning of asphyxiation or by the lack of visibility and needed more time to cover it.
Most of the other drivers, both in trucks and in passenger vehicles, stayed inside or near their vehicle. At the beginning of the fire and doubtless for a certain period before the fire spread to neighboring vehicles, they couldn't see the fire. The black smoke quickly filled the entire tunnel section, as shown by the television screens and attested by the truck drivers coming from Italy who had passed the first truck on fire.
We know that the Italian motorcycle patrol who was headed toward Italy sought shelter in refuge area 20, followed by the driver of a passenger car. Both of them died in this refuge, which was located close to the fire that lasted for more than 50 hours.
Among the drivers or occupants of the cars stuck in line, four also left their vehicles. They died of asphyxiation after going about 100 to 500 m.
Of the 10 passenger vehicles, 4 had started to make U-turns, but were stopped practically at their point of departure.
All these facts attest to the speed with which the entire tunnel section was filled with dense smoke, which must have quickly limited or suppressed all visibility. This smoke was probably very toxic.
The circumstances surrounding the outbreak of the fire and the conclusions that can be drawn will be discussed further in this report.
3.3.2 Physical Damage to the Tunnel
In addition to this terrible human loss, there were significant material losses, currently difficult to evaluate; damage for a length of over 900 m to the tunnel crown, and more localized but on significant lengths, of the roadway pavement and slab. In addition, tunnel equipment over a considerable distance was destroyed or severely damaged by the high temperatures and fire byproducts.
The vehicles traveling from France to Italy, which burned, were found between rest areas 19 and 21, over a distance of 600 m. The 8 trucks traveling from Italy to France, which stopped before arriving at the level of the first truck on fire, were evacuated by their occupants and burned between rest areas 22 and 23, over a distance of about 200 m.
In addition, the ATMB firefighting vehicle (FPTL), as well as another light rescue vehicle (PS), deserted by their occupants shortly beyond refuge area 17, were also destroyed.
|4. ANALYSIS OF CAUSES OF THE CATASTROPHE|
4.1. Operational Characteristics of the Mont Blanc Tunnel
The Mont Blanc Tunnel was built as a result of the French-Italian convention of 4 March 1953 and opened in 1965. The lease and contract documents of 7 July 1959 established the role of the lessee. Capital investments follow, as has been seen before, a complex process involving the common administration committee and the companies, a process that does not necessarily result in a coherent technical solution or synchronized schedule. In recent years, the capital improvement coordination had, however, been implemented for computerized central management, automatic incident detection and variable message signs, but these systems were only at an experimental stage at the time of the fire.
This dual operation for the same facility is at the base of various difficulties of great importance with regard to its safety. The two companies have agreements to share the regular revenue and expenses for maintenance and operation, but capital improvement funds are completely independent, so that safety modifications of the facility, especially the ventilation, were made by the two companies in a non-coordinated way: for example, the work on the exhaust openings on the ATMB-operated side was performed in 1980, while different work on the exhaust was performed on the SITMB side starting in 1997.
What appears more serious is that the operation of the two tunnel halves is only partially coordinated between the two companies, neither of the operators having a complete knowledge of the status of ventilation on the tunnel side operated by the other company.
The procedure requiring that, during an alarm, the first operator alerted takes operational command is not really applicable in the absence of central command and control.
4.2. Historical Review of Accidents and Fires in Vehicular Tunnels
Vehicular tunnel accidents are infrequent and data is available for only a few of them. The accident rate (number of incidents for 108 vehicles.km) has a limited significance and must be interpreted very carefully.
The results of a survey of several French tunnels are shown on the following table. It should be noted that the category of "large tunnels with two-way traffic" includes only the Mont Blanc and Fréjus tunnels.
It can be seen that roadway safety in large two-way tunnels, such as Mont Blanc, is better than that of open roadways in the same category (national highways) and is, rather, comparable to non-toll divided highways.
There are some tunnels with high accident levels. This is due mainly to a peculiarity which is difficult to correct. Narrow lanes, unusual alignments (sharp curves) and profiles (severe slopes), or a mixture thereof, cause accidents. Poor pavement or traffic congestion are also causes of accidents.
The most serious tunnel accident in the records caused 11 dead and 35 injured (the Melarancio tunnel on the Autostrada del Sole in Italy in 1983). This occurred between a heavy truck and a school bus in a tunnel bore operating with two-way traffic due to work in the other bore.
In vehicular tunnels, fires are rare events.
Statistical data has been established for fires between 1965 to 1992 in the Mont Blanc Tunnel and from 1980 to 1991 in the Fréjus tunnel. The following table summarizes the results.
Whatever the roadway type, the ratio of passenger vehicle fires is the same. On the other hand, it can be seen that the ratio of truck fires is much higher in large, two-way tunnels (Mont Blanc and Fréjus) than in tunnels on highways or in urban or semi-urban areas.
The average annual number of truck fires is similar and the fire ratios are identical for the Mont Blanc and Fréjus tunnels. It should be noted that both these tunnels are located at about 1,200 m of altitude at the top of long climbs. This could explain, at least partially, the higher likelihood of fire in the trucks using them.
All the fires recorded in the survey started spontaneously and none resulted from an accident. No deaths occurred. One major fire was recorded in the Fréjus tunnel: a fire in a truck carrying plastic drums damaged a roof slab that had to be reinforced; there were no victims.
The world's major vehicular tunnel fires causing death or a large number of injuries are summarized in the following tables. In the first table, only fires not caused by hazardous materials are shown. The most serious caused 8 deaths. The second table shows fires involving hazardous cargo. The most serious caused 7 deaths. The majority of these fires followed a collision.
4.3. Review of Truck Fires in the Mont Blanc Tunnel
4.3.1. Seventeen Truck Fires in the Tunnel Since 1965
Most of these were put out with extinguishers on board or available in the tunnel. At least 5 fires were followed up with response by firefighters, on an average of one every 5 or 6 years. These fires were rapidly controlled, although the 11 January 1990 fire presented certain response difficulties.
For these five fires, the fire durations and nature of the damages were:
In four cases out of five, trucks caught fire probably due to overheated engines from the steep climb. None of the fires described above was detected by the tunnel equipment, other than the television cameras.
They were all able to be approached and extinguished by the fire services.
None of these fires spread to a second vehicle.
Attempts have been made to reconstruct the ventilation conditions after the alarm for each of these fires. Unfortunately, until now the only information has come from the reports of these fires. These include the following information:
In conclusion, it appears that: the fresh air supply was set at 1/2 or 4/4 levels, according to the case. Whenever functioning, the exhaust duct was used to extract at maximum levels in the fire area. This is in accordance with the safety procedures.
4.3.2. Circumstances of the Five Other Major Fires
28 January 1974: Vehicle on fire at 0.45 km (French portal). At 0.15 km, close to the fire, the air flow was 4 m/s in the Italy-France direction. The response team coming from the French portal passed through very dense smoke at the tunnel entrance to the "penetration limit". They were able to fight and extinguish the fire very quickly because the truck was only 450 m from the entrance. The cargo was 21.5 tons of steel billet.
15 April 1978: The vehicle was on fire at 400 m from the Italian entrance it had just passed. The air speed was more than 5 m/s in the France-Italy direction. A French motorcycle patrol coming from France was the first to see the truck on fire. He stopped the vehicles (escorted convoy) following him, went through dense smoke, evacuated the four passenger vehicles stopped behind the truck, and alerted the Italian operator. The alarm was triggered and rescue assets coming from France arrived at the scene, followed by those of Courmayeur. The cargo, consisting of shoes, wood and marble, did not burn.
17 September 1981: The truck was going from France to Italy. The fire took place at 4.5 km. The air flow speed is not known. Abundant smoke was coming out of the engine, but the vehicle did not catch on fire. The report!does not mention how the fire was extinguished. It does not mention the nature of the cargo.
2 September 1988: The truck on fire was at 5.05 km. No information on the air speed. Toward 07:10, when the driver stepped down from the cab, he saw flames on the right front of the truck, gave the alarm with the pullbox at rest area 16. He fought the fire with an extinguisher and succeeded in controlling it. At 07:17 the Italian motorcycle patrol arrived and, since the fire had reignited, the patrolman put it out with two extinguishers, one from rest area 16 and the other from his vehicle. At 07:22 the French rescue team arrived after crossing 500 m through thick smoke. The fire started to ignite the cargo. The fire was then completely extinguished. The cargo consisted of 800 bags of 25 kg Hostalit chemical powder, not classified as hazardous, since these powders apparently had low combustibility levels.
11 January 1990: The truck was going from Italy to France. It stopped, on fire, at 5.81 km between rest areas 19 and 18. The driver had observed smoke since about 1.6 km, but tried to continue on until flames appeared in his cab. He gave the alarm by telephone from rest area 18 at 10:42, then returned and turned back a passenger car coming from France. At 10:50, the safety personnel coming from France and Italy, about 8 minutes after the alarm, were blocked by the smoke, the former at rest area 18 and the latter at rest area 19. However, at 10:52 the two French firefighting patrols succeeded in coming within 3 m of the truck on fire with their tanker engine (CCI) and fighting the fire. At 10:55, the fire in the truck cab was extinguished; the entire back of the truck was on fire, but the truck nevertheless accessible. At 10:58, when the Courmayeur firefighters arrived, they were able to control the fire. The cargo consisted of 20 tons of industrial cotton spools, wrapped in plastic bags, 10 to 12 tons of which burned.
What conclusions can be drawn from these five fires?
In all cases, the alarm was given only when the truck on fire was stopped.
No fires were ever observed at the tunnel entrances.
After the alarm, the longitudinal air flows in the fire areas were not recorded.
No testimonial mentioned stratified hot smoke. This is not surprising, as a relatively weak (1.5 to 2 m/s) longitudinal air flow is enough to homogenize smoke in a tunnel section.
After the alarm, the fresh air levels were set at 1/2 of their capacity or at maximum capacity, according to the case.
In 4 out of 5 cases, the smoke extraction was set at maximum. In the fifth case, it did not function at all.
The cargo was either inert (steel), or undetermined (when it was not ignited), or combustible (shoes, wood, cotton or thread), or of unknown combustibility (Hostalit chemical powder).
4.4. Sizes of the Fires and Smoke Volumes of the Truck Fires and Their Cargos
In the following, information is given on heat quantities produced by truck combustion (caloric potential), released thermal power, smoke volumes, their toxicity and the inflammability of a truck.
4.4.1. Caloric Potential
It is relatively easy to estimate the caloric potential of a vehicle and its cargo, in other words, the quantity of energy released if they burn completely. This estimation can be made by adding the caloric potential of all the parts of the vehicle, its fuel and its cargo. The result is generally expressed in GJ (giga-joules, or billions of joules, the joule being the legal energy measuring unit).
The currently agreed-upon values are from 5 to 12 GJ for a passenger vehicle, according to its size, about 50 GJ for a bus, from 150 GJ for a tractor trailer carrying a cargo of medium combustibility, and 1,000 GJ for a tanker carrying 30,000 liters of gasoline.
Actually, the caloric potential of trucks can vary widely, according to their cargos. Therefore, some cargos, not classified as hazardous in the strict sense of the rules, generate when burning caloric potentials close to those of inflammable liquids (classified as hazardous cargo). This is especially the case with margarine (present in the first two trucks coming from France, that burned) and of polyethylene (carried by two other trucks coming from France). Thus for burning cargo, the caloric potential varies from several dozen GJ (fruits, vegetables) to about 900 GJ (all margarine cargo). For the vehicle that started the fire (PL0), the cargo caloric potential was between about 500 and 600 GJ.
Compared with these numbers, the caloric potential of tractor trailers is relatively low: from about 5 to 10 GJ. Semi-trailers and trailers (without cargo) add 20 to 30 GJ each, as well as the fuel carried (17 GJ for 500 liters of diesel).
A rough estimation results in a total caloric potential of 5,000 to 7,000 GJ for the total number of vehicles entering from France, that burned (or the caloric potential of from 5 to 7 gasoline tanker trucks), and from 1,100 to 1,800 GJ for those which entered from Italy and burned.
4.4.2. Thermal Power
While caloric potentials may be easily estimated, this is not the case for the thermal power emitted at each moment, or the speed of energy released. This would require knowing the fire dynamics and there is, unfortunately, little information on this subject. The thermal power is usually measured in MWs (megawatts, or millions of watts, the watt being the legal unit of power).
Numerous tests have been run over the years to study fires in tunnels. Unfortunately, most of these tests used pans of hydrocarbon liquids for fuel, and it is difficult to establish the relationship with vehicular fires, in regard to fire dynamics. However, some tests used actual vehicles, notably in the EUREKA project involving nine European countries from 1992 to 1995, and in which French public and private entities widely participated.
Currently agreed-upon values representing power levels over periods of time are from 2 to 5 MW for a passenger vehicle (fire duration of about 45 min.), of about 20 MW for a bus (fire duration of about 1 hour), approximately 30 MW for a truck carrying mildly combustible merchandise (fire duration of about 1-1/2 hours), and from 100 to 200 MW for a tanker truck carrying liquid hydrocarbons (fire duration from 2 to 3 hours). However, values clearly higher than the 30 MW indicated above are possible for a truck not carrying hazardous cargo: a tractor trailer carrying furniture released more than 100 MW during several minutes in conditions of very strong ventilation during one of the EUREKA tests. In general, the powers previously mentioned are not reached immediately, but after at least ten minutes in the case of trucks.
It is not possible at this time to reconstruct the exact thermal history of the 24 March 1999 fire, and it would be difficult to proceed on this subject. At best, some of the elements which allow an estimation of the thermal power can be cited: dividing the total caloric potential in the vehicles (from 6,000 to 9,000 GJ) by the total duration of the fire (about 50 hours) results in an average power for the duration of the event of between 30 and 50 MW. It is certain, however, that higher power was released during some time periods and lower during others; moreover, the available oxygen at the fire level limited the release of power at any moment. There is actually a known relationship between the quantity of energy released by the combustion of an organic matter and the mass of oxygen being used. This matter will be discussed later.
4.4.3. Smoke Emission
In order to understand the development of events in the tunnel, it is very important to estimate the amount of smoke generated by the fire. It would be convenient to first define the terms used. In fact, to be able to pinpoint generated volumes, it is necessary that the smoke must not mix with the surrounding air: the smoke is then made up of the combustion gases and gases that have crossed the fire core without burning (unburned nitrogen and oxygen from ambient air). This is what can be found in a tunnel when smoke stratifies close to the ceiling. When the combustion products are fed with ambient air, they form a mixture that looks like smoke, with a volume that is the sum of the volumes of smoke and mixed-in fresh air. It is essentially this latter condition which occurred during the 24 March fire.
The smoke emission values generated by vehicular fires are agreed upon when the smoke is stratified in the tunnel: about 20 m3/s for a passenger vehicle, 50 m3/s for a van, 80 m3/s for a truck carrying mildly combustible cargo, 200 to 300 m3/s for a gasoline tanker truck. Actually, it is the nature and quantity of the cargo that determine smoke emission for trucks. The tractor trailer that burns alone would generate from 20 to 40 m3/s of smoke.
When the smoke emitted by the fire is mixed with air brought in by ventilation, as was the case of the 24 March fire, at least between the core and the French entrance, one could probably estimate that the volume of smoke (i.e., the mixture of combustion products and the unburned portion of the ventilation air) is slightly higher than the volume of the air brought in by ventilation, giving consideration of course to gas dilation due to high temperatures. In fact, oxygen represents only 21% of the air present in the fire core, the other four fifths not being really part of the combustion. The combustion does not generate a gas volume much higher than that of the oxygen used (aside from the effect of rising temperature). In total, the gas volume coming out of the fire core does not exceed by much (at the maximum about 20%) that of the drawn gases, given the same temperature.
4.4.4. Smoke Characteristics
The main dangers presented by smoke are:
- obscurity, which prevents people from fleeing due to lack of visibility;
- toxicity, which incapacitates, a condition preventing flight and finally causing death, its effects depending on the nature of the poisonous elements present in the smoke, their concentration, and the duration of exposure;
- temperature, which also incapacitates and may cause death, depending on exposure.
These hazards usually occur in this order: loss of visibility, then exposure to toxic gases, then excessive temperature. It is important to note that the alarms triggered the 24 March followed this succession of events as the smoke progressed toward the French side. Triggering the obscuration sensors of rest areas 14 and then 9 (alarm at 20% very quickly followed by a reading higher than 30%) preceded by minutes that of CO detectors (alarm at 150 ppm quickly risen beyond 250 ppm). As for the fire alarm, it triggered at a temperature of 50º C at the level of rest area 15, only 15 minutes after the over-the-limit CO reading at rest area 14, although this was located 300 m farther from the center of the fire.
While the temperature is dependent upon the thermal power released by the fire, the obscurity and smoke toxicity largely depend upon the burning materials and the combustion conditions. Some data is available on obscurity and carbon monoxide (CO) produced by tunnel vehicular fires. Among the other very toxic gases that could have been released during the fire can be cited hydrochloric acid, released especially by PVC combustion, cyanhydric acid, generated, as well as nitrogen oxides, by the combustion of polyurethane, widely used for refrigerated trailer insulation, and acrolein, produced by the combustion of margarine. Toxic gas production is usually increased when the combustion lacks oxygen, which was probably the case during most of the fire, with the exception of the first minutes.
4.4.5. Truck Flammability
Truck fires similar to that of the Volvo truck on 24 March 1999, but also like truck fires in other circumstances, can develop very quickly, and especially increase in the few seconds immediately upon the truck's slowing down and then stopping. In fact, the fire is no longer limited by the travel air speed. Thinking about the flame of a candle: if one blows too hard, one extinguishes it; if one blows gently, one feeds it.
The dramatic consequences that may be generated underground by the development of such quick fires lead to the question of what the safety rules have to say on the subject.
It is known that no fire resistance standard exists for truck gas tanks, which are sometimes made of plastic (this was not the case of the Volvo truck that caught on fire, which was a light alloy). It is known that the volume authorized for these tanks has been increased. Since 1 January 1997, the capacity can reach 1500 liters, of which 500 liters are for the trailer tanks.
Nothing is mentioned concerning the flammability risks of trailers or semi-trailers. The shells of refrigerated trailers may be combustible. In today's industrial and urban civilization, the number and length of underground roadways will likely continue to increase. It appears very desirable that the safety requirements for trucks and their trailers be improved with regard to fire risk.
4.5. Circumstances of Development of the First Fire
4.5.1. Description of the Development of the First Fire
The vehicle which caught fire was a tractor trailer consisting of a Volvo FH12 cab and a refrigerated trailer containing margarine (9 tons) and flour (12 tons).
The task force is quite familiar with the conditions in which the fire developed in this first truck, thanks to the testimonies of its driver and the drivers of vehicles which passed the truck in both directions after it had stopped.
The first signs of smoke were reported by trucks coming from Italy, passing the Volvo truck in the opposite direction. This happened toward km 2 or 3, thus 2 or 3 minutes after the truck entered the tunnel. White smoke was coming out of the cab, passing under the trailer and coming out behind and swirling up toward the ceiling.
When the driver, alerted by flashing headlights, looked into his rear view mirrors, he saw the white smoke behind his truck, on the right.
He slowed down and stopped. He allowed a truck passing him in the opposite direction to go by and then got out. White smoke was coming out of the cab and rising up between the cab and the trailer. He tried to reach his extinguisher located under the seat of the cab on the left side. At that moment, for the first time, flames burst out on both sides of the cab. He stepped back and could not do any more.
From this time on, all testimonies described the smoke as black.
The fire had quickly entered the cab.
The fire only spread to the trailer afterwards. Nevertheless, one testimony noticed the beginning of a fire on top of the trailer; this was offered by the driver of a truck passing the truck on fire in the opposite direction.
At the present time, the source of the original white smoke is not known. The continuing investigation of the truck will reveal this.
4.5.2. Specifics of the Truck and its Cargo
The tractor trailer's fuel tank was built of a light aluminum alloy. It had a capacity of 920 liters and contained about 550 liters of diesel at the time of the fire. The tractor pulled a refrigerated trailer.
The refrigerated trailer was constructed of an isothermal foam shell which was easily flammable. The cargo of margarine, as it melted, was transformed into very combustible liquid oils capable of spreading onto the roadway while setting off a fire of considerable power. The investigations taking place will try to determine how long it took for the margarine to melt and the strength of the fire caused by the combustion of this oil. The type of wrapping of each packet of margarine should be considered. The role of the flour in the combustion of this fire will also be estimated as much as possible.
4.6. Ventilation Conditions
4.6.1. Ventilation Levels Before and After the Alarm
Since the tunnel's control stations did not have computerized command control systems (aside from a partial system on the French side called "mini GTC"), it was not always easy to know what actions had been taken. Nevertheless, the gathering of testimonies, and data provided by certain recordings and by visual observation, allow the following reconstruction of events:
Before the alarm:
The fresh air ducts:
On the French side: the supply was at level 2/4, or about half of maximum capacity. This is normal for average traffic levels.
On the Italian side, they were working at level 2/3.
The reversible supply/exhaust ducts:
On the French side, the duct had worked in exhaust at level 1/2 since the morning. It had been set in exhaust position, as was sometimes the case when staff were performing work inside the tunnel, per their request.
On the Italian side, the duct was used as supply of fresh air and was working at level 1/3. This is completely normal, being the most economical way to improve air quality inside the tunnel during normal traffic conditions.
The exhaust at the French entrance:
The fan installed in the old corridor to extract tunnel air coming to the French side was working at 2/3 of its maximum capacity.
After the alarm:
The supply ducts:
On the French side, three out of four ducts were set at full level and the one servicing the first part of the tunnel was only set at level 3/4.
On the Italian side, the four ducts were raised progressively to level 3/3. This level was attained at about 11:01.
The reversible supply/exhaust ducts:
On the French side, the reversible duct, which was in exhaust even before the alarm at level 1/2, was raised to 3/4 with concentration of the exhaust in area 3, apparently toward the center of the tunnel in the zone close to the fire.
On the Italian side, the reversible duct, which was in a supply position at level 1/3 before the alarm, was left in this position and the level went gradually to 3/3, attained at about 11:02.
At 11:13 on the Italian side, fire ventilation procedure was ordered, by reversing the duct to exhaust and configuring the openings for a concentrated exhaust around rest area 20. At 11:14 this fire configuration was deactivated, returning to supply of fresh air. The fan levels were lowered.
At 11:15, the Italian duct returned to supply.
From 11:43 to 12:04, the levels of supply ventilation were increased by this reversible duct.
Starting at 12:29: still by the same duct, an attempt was made to run a fire procedure by exhaust and configuration of exhaust openings to concentrate around rest area 24, then at 12:40, a return to supply. These operations are discussed in more detail later.
The exhaust at the French entrance:
After the alarm, the fan in the old corridor was set at its maximum exhaust capacity for the tunnel air coming to the French entrance.
4.6.2. Consequences of Ventilation Levels on Smoke Movement and Development of the Fire
At this preliminary stage of the studies to determine what happened during the catastrophe, it is too early to reconstruct the actual development of the fire, and further studies on this subject will be difficult, due to the lack of available information. At best, an attempt can be made to give some yet uncertain and relatively general indications about the beginning of the fire.
The contents of the following paragraphs show that the air flow that carried the smoke mostly toward France at the beginning of the fire was essentially due to the unbalanced ventilation levels between the two halves of the tunnel (higher supply of fresh air and absence of exhaust on the Italian side, exhaust at the French portal).
This imbalance created a relatively weak air flow (1 to 1.5m/s) in the Italy-France direction at the level of the Belgian truck that started the fire. It explains the speed with which the smoke reached the vehicles stopped behind the truck, surrounding all the people there, then moving faster and faster toward the French portal exhaust. On the other side, the smoke advanced only slowly in the direction of the Italian portal, against the air flow, over a distance that did not go beyond 300 meters in over an hour. This is what allowed all the people who entered by the Italian portal to escape.
The reading of paragraphs 18.104.22.168 to 22.214.171.124 and 4.6.3, all very technical, is not necessary in order to understand the rest of this report.
126.96.36.199. Air Movements in the Tunnel
As usually utilized and notably before and during the 24 March fire, the tunnel ventilation system supplies much more air into the tunnel than it exhausts. The additional air exhausts by the two portals where the air is always flowing out. On the French side, this takes place slightly more than 110 m from the portal where the exhaust fan usually collects the entire air flow coming out of the tunnel.
With the air exhausting this way through the two openings, there is always a point in the tunnel called "neutral point", where there is no longitudinal air flow. On each side, the air flow is directed to each portal, with a speed that increases as it gets further from the neutral point. The air movements inside the tunnel, and therefore the position of the neutral point, are determined by several complex factors:
- the influence of the supply and exhaust volumes in each of the tunnel sections
- the role played by the French portal exhaust fan
- the difference in barometric pressure between the two portals (due to winds and different meteorological conditions on either side of the mountain)
- the "chimney" effect resulting from the temperature difference between the tunnel interior and the exterior atmosphere, since the Italian entrance is located 107 m higher than the French portal (even without a fire, the air is always warmer inside the tunnel because of the geothermal heat due to the high cover over the tunnel)
- the piston effect of vehicles, especially trucks, that push the air inside the tunnel in the direction of their travel.
Knowing, at least approximately, the ventilation flows before and after the start of the fire, two elements may be used to estimate the position of the neutral point: the measurements of the only working anemometer, close to the Italian portal, and the observations on smoke movements. The way the smoke propagated at the beginning of the fire strongly toward France and covered only 100 to 130 m toward Italy shows that, at the level of the truck that started the fire, there was a relatively limited air flow (probably about 1 to 1-1/2 m/s) and it was directed toward France. This is consistent with a neutral point from rest areas 22 to 24 and therefore to an Italy-France air flow direction in the middle of the tunnel.
Simple calculations, not taking into account thermal effects produced by the fire, show that passing from the ventilation levels prior to the fire to those after the fire started had a tendency to move the neutral point slightly from Italy toward France (about 500 m). The air exhaust speeds were increased at the two portals but the air flow in the area where the vehicles entering from France were was not modified and remained in the Italy-France direction. Since the smoke coming from rest area 21 toward Italy closed in progressively on rest area 22, without, however, passing it during almost an hour, it can be guessed that the neutral point was closer to rest area 24 when the Belgian truck stopped, and then moved toward rest area 22. This scenario is compatible overall with the Italian portal anemometer measurements that indicated widely varying values (due to vehicular traffic) between 4 and 6 m/s before the fire, and more stable, around 6 m/s, afterwards.
These simple calculations show that at the beginning of the fire, the chimney effect going in the France-Italy direction and due at that moment to the natural heat (geothermal) of the rock and not of the fire, must have been more or less compensated by opposing atmospheric effects. The 24 March there was a current of warm air (fœhn) that tended to push the tunnel air from Italy toward France. Such conditions occur about twenty times a year, the air moving from France to Italy the rest of the time. In these conditions where the chimney effect and exterior winds almost compensated each other, the unbalanced air flow that went toward France at the level of the Belgian truck was due mostly to the imbalance of ventilation levels between the two tunnel halves (higher fresh air supply and lack of exhaust on the Italian side, exhaust working at the French portal).
188.8.131.52. Consequences of Smoke Movement
The chronological graph (Appendix 3) shows two distinct phases of smoke emission by the Belgian truck.
The first phase is related to the truck's travel in the tunnel until it stopped at rest area 21: While the rest area 4 and 9 obscuration sensors sent an almost normal obscurity value during the truck's passage, those of rest areas 14 and 18 reached their maximum limits very quickly. This seems to show that the truck's smoke emission developed suddenly between rest areas 9 and 14, remained the same to rest area 18, then decreased while the truck continued beyond. Under the effect of the air flow from Italy to France, and also due to dilution by fresh air supplied at the bottom of the side walls, this first smoke emission, which was clear in color according to observers, almost disappeared after 6 minutes at rest area 18 and 14 minutes at rest area 14. The exit of the diluted smoke slug in the French direction saturated the obscurity reading at rest area 9 for 1 or 2 minutes, while the obscurity sensor of rest area 4 later sent only a brief increase of the obscurity at a slightly lower level.
The second phase is related to the stopping of the Belgian truck, with the sudden start of the fire, as has been described by witnesses:
The smoke become darker and thicker than in the first phase and was pushed with increasing speed toward the French portal, successively saturating the four obscuration sensors. It took only about 6 minutes between the exhaust of the first phase smoke and the arrival of the second phase smoke, at the level of the obscurity sensors of rest areas 14 and 18. During this period, the visibility was mediocre but acceptable for normal operation at rush hour.
The second phase smoke behavior will be examined in more detail. It has been seen that at the beginning of the fire this smoke advanced progressively toward Italy from rest area 21 to rest area 22, but didn't pass the latter for almost an hour. This is what allowed the people who had entered at the Italian side to escape.
The phenomenon of smoke advancing against the longitudinal air flow inside the tunnel is well known (it is called "backlayering" in English). It happens when the longitudinal air speed is lower than the speed called "critical speed", depending mainly on the fire's magnitude. The smoke rises to the ceiling over a distance related to the thermal power of the fire, the air flow speed, and the tunnel slope. It may stay up for quite some time, forming a rather thick smoke layer that regenerates at the ceiling level or, if longer, it may cool down and descend to the pavement where it is again drawn in the direction of the fire.
This latter phenomenon was observed by the ATMB patrolman, who drove his vehicle within meters of the fire core shortly after 11:05: coming from rest area 22, he first passed through a "wall" of smoke filling the cross section of the tunnel for 100 to 200 m, then, when he saw the truck in flames, found himself in an area which at his level was clear and where the smoke, while very hot, was locally stratified at the ceiling.
On the other hand, toward France the smoke seemed to have been completely mixed with fresh air. The air flow drew most of the air mix – smoke over the entire length separated the fire zone from the exhaust corridor close to the French portal. This happened despite the exhaust, which was apparently concentrated on the third third of the French-operated side. In this section, the flows before the fire were about 25 m³/s per km for fresh air and 15 m³/s per km for exhaust. After the start of the fire, they increased to about 50 m³/s per km for supply and about 44 m³/s per km for exhaust. The exhaust did not match the supply, in either of these two cases, and could exhaust only a limited amount of smoke.
It has been previously estimated that the longitudinal air flow was about 1 to.5 m/s toward France at the level of the truck that started the fire. Due to the continuous fresh air supply over the tunnel length, which was not balanced by the exhaust, this air flow was accelerating with its progression toward the French portal. Without accounting for the thermal effects that later increased these values, the simple calculations mentioned above show that the air speed, and therefore the smoke, must have been as follows, with the ventilation levels used after the beginning of the fire:
- from 2 to 2.5 m/s (slightly less before the fire) at the level of rest area 19 (located closed to the middle of the tunnel, 100 m after the last burning truck)
- from 3.5 to 4 m/s (2.5 to 3 m/s before the fire) at the level of rest area 9 (in the middle of the part operated by the French company)
- from 6 to 6.5 m/s (3 to 3.5 m/s before the fire) at the French portal.
First, it should be pointed out that these speeds are higher than those at which most people, even in good health, can flee over a certain distance. Then, it is worth noting that these are the speeds with which the smoke traveled while exceeding the range of the obscurity sensors:
- obscurity of 20% measured at rest area 14 at 11:11, at rest area 9 at 11:19 (or an advancing speed of a little more than 3 m/s between the two sensors), at rest area 4 at 11:25 (a smoke speed of a little more than 4 m/s from the preceding sensor).
184.108.40.206. Effects of the Fire Development
As previously indicated, it has been extremely difficult to reconstruct the development of the fire and especially the dynamics of heat and smoke release.
For the smoke, it was first estimated that, when it was not stratified, the release volume only slightly changed the volume of gas traveling through the tunnel (with the condition, of course, of allowing for heat dilation). Actually, combustion gases do not take up a much larger volume than the oxygen being used, and this does not represent more than a fifth of the ambient air volume, at a maximum. Therefore, whatever the volume of the released smoke, it is the longitudinal air flow that governs its travel, and not the characteristics of the fire.
The longitudinal air flow speed helps explain that, in the French direction, no more than 10 minutes were needed to fill the 900 m length between the truck that started the fire and rest area 18 with smoke. This speed depends very little on the size of the smoke release. On the other hand, the composition and thus the obscurity, the toxicity and the temperature of the traveling mixture of air and smoke are directly related to the generation and the type of released smoke.
Since the smoke characteristics in various areas at the beginning of the fire and their evolution are not well known, it is hard to learn from available observations the released quantity of smoke and thermal power.
At this point, it is only possible to give the maximum limits of the thermal power, based on the available oxygen volume. In fact, in an organic material combustion, very common with vehicular fires, the burning of a kilogram of oxygen corresponds to a thermal energy release of 13.1 MJ (mega-joules, or millions of joules).
At the level of the vehicle starting the fire, the estimated longitudinal air flow between 1 and 1.5 m/s allowed a maximum thermal power between 75 and 110 MW. These are values that assume that no more than half of the oxygen from the air that crossed the fire area was burned, which is likely. At these combustion speeds, the fire duration for this tractor trailer would be between an hour and a half and more than two hours, accounting for a caloric potential of about 600 GJ. It is worth noting that, in regard to the thermal power released by the first tractor trailer, the supply through the opening of duct no. 5 at the ceiling level of rest area 21 did not have a significant impact, because it was upstream of the truck and had flows of up to 10 m³/s, compared to the 50 to 70 m³/s brought in by the longitudinal air flow.
The same type of reasoning could be used for the group of vehicles between the truck that started the fire, located at rest area 21, and the last vehicle, located 100 m before rest area 19. It must recognized, however, that the ventilation conditions had definitely changed: this will be an issue to be discussed later. With the disputable assumption that the ventilation conditions were the same, there were available in this area 46 to 70 m³/s brought in by the air flow coming from rest area 21 in the direction Italy-France (assuming that the oxygen was not already burned by the fire of the trucks which had entered from Italy and stopped between rest areas 22 and 23), as well as 25 m³/s supplied at the base of the side walls by duct no. 4 and 10 m³/s provided in each of the rest areas 21 and 20. In these conditions, the maximum possible power for the entire area where the vehicles which had entered from the French side were, would be between 150 and 190 MW, still assuming that, at maximum, half of the available oxygen was burned. At similar combustion speeds, all the vehicles in this area could have burned in 7 to 13 hours, which clearly shows the conservative nature of these estimations.
It should be noted that gases from overheating the vehicles and their cargoes, without burning due to the lack of oxygen, could have progressed beyond rest area 19 and burned farther away, when the oxygen became available due to the supply of fresh air at the base of the side walls.
4.6.3. Spreading of the Fire Among the Vehicles
The question remains as to how the fire was able to spread to such a large number of vehicles, some of which were distant from each other.
220.127.116.11. Vehicles That Burned
Two groups of vehicles can be distinguished, on the one hand based on whether they entered at the French portal or the Italian portal and, on the other hand, the ATMB vehicles:
- Vehicles entering at the French portal behind the Belgian tractor trailer (PL0):
26 vehicles which burned were found behind the Belgian truck, including the truck: 15 tractor trailers with semi-trailer and/or full trailer, a pickup truck, 9 passenger vehicles, and the SITMB agent's motorcycle. The distances between the remains of the tractor trailers varied from 3 meters to almost 45 meters. The passenger vehicles and the motorcycle were interspersed among them. All of these remains were located between about 100 meters beyond rest area 19 (in the direction of the traffic) and rest area 21, i.e., over a stretch of about 500 meters.
- Vehicles entering at the Italian portal:
The eight tractor trailers that burned had stopped between about 100 meters beyond rest area 23 (in the direction of their route) and rest area 22, i.e., over about 200 meters. The distances between the vehicles varied from 2 to almost 20 meters. The closest to the Belgian truck was 290 meters from it. It should be recalled that the passenger vehicles driving behind the trucks had made U-turns and that the tunnel operators had backed up the first tractor trailers to free up rest area 22, which brought the vehicles closer together but distanced them slightly from the Belgian truck.
- ATMB rescue vehicles:
The ATMB fire engine (FPTL) stopped shortly before rest area 18, and the rescue vehicle (PS) shortly after rest area 17. The former, located 450 meters behind the last tractor trailer entering from the French side, also burned. The latter, stopped 230 meters behind, was greatly heated up and damaged, but did not burn.
18.104.22.168. Ways by Which Fire Spreads
A certain number of causes can lead to the spread of fire from one vehicle to another, or from one group of vehicles to another. This report!will mention those that could have occurred during the fire of 24 to 26 March, although the numbers should only be taken as orders of magnitude, giving a general idea:
- Flame and radiation
The fastest way fire can spread at short distances is by flame and thermal radiation greatly heating the surface of another vehicle and leading to its catching on fire.
Heat may be carried by convection or by movement of very hot gases. This convection usually occurs in the direction of the tunnel air flow, but it may happen counterflow, under certain conditions, as has been testified by the ATMB patrolman when he came close to the fire core shortly after 11:05 on 24 March. Materials catch on fire most often by self-ignition generated by the thermal fluxes they receive. Numerous polymers, present in various ways in vehicles, start to decompose from 200 to 300º C. Some of them, when heated, produce gases that are easily flammable by a spark or an open flame starting from 300 to 400º C. Even without the spark, the same gases can self-ignite, starting from 400 to 500º C. Truck fuel may itself be a spreading agent, due to its relatively low self-ignition temperature (260ºC).
Moreover, an electrical fire can occur if the heat creates short circuits in the vehicle's electrical equipment.
One particular way of spreading fire by convection is related to the "backdraft" phenomenon: combustion lacking oxygen, as certainly was the case during some periods of time, can produce unburned gases that are highly flammable. The cloud thus formed is drawn by the ventilation air flow and can violently ignite whenever air mixes in. There can also be a sudden ignition of the cloud, i.e., an explosion.
- Burning liquids
Fire can also be spread by a burning liquid that flows on the pavement or in the drainage system. Some of the likely liquids are diesel fuel, margarine, and polyethylene.
- Pavement combustion
Another possible way of spreading fire could be the combustion of the bituminous wearing layer covering the roadway slab. If it is clear that the bitumen can burn, and it burned in certain areas of the roadway, it is not known at this time in what conditions it can spread fire when it is used as roadway pavement.
22.214.171.124. Likely Ways the Fire Spread From Vehicle to Vehicle
The progress of studies to date is still not adequate to ascertain what were the actual mechanisms that led to the propagation of the fire.
The flame and the radiation may certainly explain the fire spreading from a truck cab to its trailer, as well as to a nearby vehicle. The convection-related phenomena may have caused the ignition of all other vehicles behind the Belgian truck and backed up over 500 m behind it. But propagation by a burning liquid spill is equally possible, at least nearby. The case of the ATMB FPTL is perhaps more complex, but appears not to relate to anything other than convection, i.e., backdraft.
The most difficult explanation is the one regarding the ignition of the eight trucks that entered from Italy across a 290 m long area free of vehicles. The 100 m long area ahead of the first of these vehicles is remarkable for the minor damage to the ceiling (cables and light fixtures are still there), while the areas ahead and behind are very damaged.
If propagation was by convection, it could have been a layer of smoke traveling at a temperature high enough to ignite one of the trucks but insufficient to cause much damage to the ceiling or even to equipment. It is also possible that a spill (of margarine, for example) developed from the first two trucks over perhaps 50, 100 or 150 m (the slope is gentle: 0.5% toward Italy). The fire from this spill (with relatively high flames) would have destroyed the equipment and the pavement in the "very damaged" area, causing a spreading by the pavement fire (short flames) to the level of the first tractor trailer of the group (PL-1). Another hypothesis could be a burning liquid flow or hot gases traveling through a ventilation duct.
4.6.4. Comments on the Ventilation Procedures Put Into Effect During the Fire
From all this, the following statements can be made.
The fire procedure for supply of fresh air requires full levels in the entire Mont Blanc Tunnel, including the fire area. This was done. It has been noted that this procedure had changed since the commissioning of the tunnel (initially, supply air was not provided). In fact, it is normal to supply fresh air to help people who could be in the fire area, but too much supply churns up the air and mixes it with the smoke, pushing it far inside the tunnel and feeding the fire.
It is also generally recommended to have the supply air partially, rather than fully, open.
The exhaust procedure in the case of fire calls for a maximum exhaust in the fire area.
This was not done on the Italian side where, on the contrary, the operator configured for maximum supply. He had seen on the CCTV at rest area 22, 300 m east of the fire, vehicles making U-turns to return to Italy. He declared that, after observing the situation, he sincerely thought this is what he needed to do to save lives. Later, at 11:13, a short exhaust command was attempted in the presence of his supervisor, by simulating a fire at rest area 20, with the participation of a technician from the company which had installed the equipment allowing the extraction of smoke by concentrating exhaust. It lasted only a minute. Another attempt was made between 12:29 and 12:40 by simulating a fire at rest area 24. The managers of the Italian company confirmed that the smoke extraction equipment was in good working order and that it was, in fact, a voluntary decision on the part of the operator and then eventually his supervisors to supply fresh air to save lives. It must be noted that, according to Italian personnel statements, the exhaust attempts, even that of 12:29, did not lead to smoke arrival at the exhaust stack of duct no. 5, located at the end of the Italian approach. Perhaps they did not last long enough for that.
Although it is not known to what effect the short exhaust attempts were made, the task force does not doubt the good will of the staff who thought this was the best approach.
However, this was against the procedures designed for such situations. Moreover, it is indisputable that this contributed instead to feed the fire, and helped destratify the hot smoke and push it toward France and toward Italy.
At about 11:45, five Italian firefighters, who had entered the tunnel, took shelter in refuge area 24. They would be eventually rescued through the reversible duct which supplies fresh air. This would not have been possible if, at that moment, between three and four hours after the alarm, it had carried exhaust air.
4.7. The Role of the Other Systems
4.7.1. Electrical Equipment
Very quickly, that is from the first minutes, the lighting equipment was destroyed in the fire area; besides the difficulty caused by the smoke, the lack of lighting probably increased the hardship to motorists, while the destruction of this equipment resulted in short circuits leading to the progressive shutdown of electrical circuits affecting power supply serving other equipment.
This is what happened to the fire sprinkler on the French side and the exhaust dampers on the Italian side. It is clear that the tunnel electrical equipment does not have automatic redundant systems necessary to function in a failsafe mode or in difficult situations and that, generally, the safety equipment (such as sprinklers or exhaust dampers) must be able to function and should therefore have a viable electrical system.
These issues must be carefully studied again during preparations to recommission the tunnel, then verified on paper during safety studies, and finally tested in real life during site drills.
4.7.2. Traffic Signals
In the tunnel, there are traffic signals every 1,200 m (at the entrance, then at the levels of rest areas 3, 7, 11, 15, etc.). According to the testimonies received, these signals turned red in the France-Italy direction at 10:55 and in the Italy-France direction at 10:56. There is, however, some doubt as to the effective functioning of the red lights, even if the command was actually given, because there is no reliable recording system for the working conditions of the major safety equipment. The traffic lights at the French entrance to the tunnel were indeed working and at least one motorist saw them, but there is doubt as to the functioning of the traffic lights in the tunnel. If, in fact, the lights had turned to red, some motorists could have run through them either unintentionally (the lights are not very visible) or in thinking that, in the absence of visible danger, they could do so without risk.
On the other hand, if even one motorist had stopped for a red light, that would have required the vehicles following him to do the same and, several minutes later, the stopped vehicle or vehicles would have been passed by Mr. Roiget, the Italian operator, who had departed to inspect the tunnel, and the number of victims would have been reduced accordingly. As that did not happen, there are only two hypotheses: either the motorists ran through the red lights, or else the red lights in the tunnel were not all functioning.
4.7.3. Refuge Areas
Since the opening of the tunnel, there have been 6 closed areas which could qualify as refuge areas. But neither their fire resistance, nor their ventilation, nor their interior equipment corresponded to what can really be called refuge areas.
After the serious fire of 11 January 1990, the two companies installed 18 real refuges, located in every other rest area, every 600 m. It must be noted, however, that, on the French side, these refuges are at the level of the odd-numbered rest areas and, on the Italian side, at the level of the even-numbered rest areas; thus there are two consecutive rest areas in the center of the tunnel without refuge areas, leaving a length of 900 m without a refuge.
The refuges are supplied with fresh air by the ventilation ducts, are represented as fire resistant during two hours (2-hour fire rating), and are connected by telephone to the control rooms. These shelters played an essential role in saving the lives of the firefighters and operating company personnel working inside the smoke-filled tunnel to rescue passengers. On the French side, refuge 17, where 6 ATMB personnel found shelter, protected them from toxic smoke and heat for nearly 7 hours. On the Italian side, the firefighters similarly took shelter in refuge 24. They were protected there until their evacuation. Unfortunately, it wasn't the same for refuge 20, where two people, including an SITMB representative, died. It is very likely that, even if these refuge areas had been fire-rated at 4 hours, their deaths would not have been avoided. In fact, the fire burned for more than 50 hours.
The problem of locating the refuge areas, given the lack of visibility, must also be noted. They are indicated by a lighted panel above the door (the refuges on the Italian side have two doors), carrying the sign "refuge" (in French on the French side, in four languages on the Italian side), and by graphics for "extinguisher" and "telephone". In the conditions of poor visibility due to the smoke, the panels above the doors were probably hardly visible and the exterior appearance of the refuge with the light visible through glazed windows differs little from other technical rooms of the tunnel, which are neither protected from fire nor ventilated.
Fresh air is supplied into the refuges by a direct link with one of the ducts on the French side, and by a local ventilation system on the Italian side. This results in a real constraint in the operation of tunnel ventilation in case of a fire, since it is necessary to assure a minimum flow in the fresh air ducts to assure a sufficient supply in the refuges when people are present or likely to be there (flow greater than half of the maximum, on the French side).
4.7.4. Safety Corridors
Generally, a safety corridor parallel to a main tunnel can serve to evacuate people in danger and facilitate access for emergency personnel.
The Mont Blanc Tunnel does not have any.
With regard to the evacuation of people in the case of this fire, it has been shown that the majority of the victims did not leave their vehicle. They apparently did not attempt to reach an emergency exit. Therefore, they would not have been saved by a safety corridor. The victims who did get out of their vehicle and died in the tunnel did not enter a refuge area. They were surrounded by smoke and in darkness (the lighting was out from 11:01 on). They must have simply tried to flee via the tunnel. It is likely, given past experience noted in other vehicular tunnel fires around the world, that they would not have reached an emergency exit without having been led there by qualified personnel. On the other hand, the 2 victims who died in refuge 20 could have without a doubt been saved, if it had been linked to a safety corridor.
The 1981 internal policy circular requires that, "if a pedestrian safety corridor cannot reasonably be made available, then:
- either provide refuge areas at U-turns,
- or provide fresh air ducts for the purpose of using them for the evacuation or shelter of passengers".
Refuge areas were therefore chosen for the Mont Blanc Tunnel (every other rest area). They were not connected to fresh air ducts for the evacuation of users, probably because that would have required significant rock excavation with explosives and an extended tunnel closure.
On the other hand, the reversible supply/exhaust duct may be used, but only in difficult and uncertain conditions, as a safety room, as long as no smoke is exhausted at that time. On the Italian side, this allowed the evacuation of the firefighters trapped in the refuge area of rest area 24. On the French side, it allowed the partial progress of Captain Comte, who came to rescue the other Chamonix firefighters trapped in the tunnel. It did not, however, allow their evacuation, nor that of the ATMB agents trapped in rest area 17, for it was too filled with soot, brought in by the fresh air ventilation, as a result of its previous function as exhaust duct.
4.7.5. Other Equipment
126.96.36.199. Emergency Telephone Network
The tunnel is equipped with 72 telephones installed on opposite sides of the tunnel in the rest areas or the U-turn areas across from them. Calls made from these telephones go to control rooms and the caller is then in touch with an operator. It should be noted that the niches between the rest areas are not equipped with telephones.
During the fire, a telephone call from rest area 22 was received at about 10:54 by the Italian control room. Then several telephone communications were established with rest areas 5, 9, 12 and 28. The French operator attested to several communications with the ATMB agents trapped in rest area 17, but no record appears in the mini-GTC listing (see further in report).
It should be pointed out that the Chamonix firefighters attested to difficulties in reaching the operator by using the telephone in rest area 12. This statement is contested by the ATMB agents, who say they spoke with the occupants of this rest area. The listing shows that "telephone hang-ups" did indeed occur in rest area 12. It would seem that the communication could not always be established between the tunnel and the control room, but it could be in the other direction.
The emergency telephone network played an important role for the operators during the alarm phase. In particular, the equipment allowed them to maintain contact with the rescue team trapped in the tunnel. A reservation must nevertheless be expressed concerning a seemingly unequal functioning on the French side. It was cut completely from 15:49, however, eliminating all direct communications between the two portals.
Six-kilogram ABC powder extinguishers are located in groups of two every 300 m in the rest areas and in the U-turn areas opposite them. Between these rest areas, they are also located every 100 m in niches situated alternately on one or the other side wall. Thus four pairs of extinguishers were available every 300 m.
In each control room, an alarm is triggered when an extinguisher is lifted and the zone of the rest area where it is located is indicated.
It is thus that at 10:58 an unhooked extinguisher in rest area 21 was signaled. However, after a visual examination of the spot, it does not appear that an extinguisher was actually lifted: it is possible that the alarm was created by a faulty cable which passed on the information.
With the exception of rest area 21, for which a slight doubt remains, it must thus be concluded that the extinguishers were not used during the fire of 24 March. A great majority of fires in vehicles (including trucks) which break out in tunnels are, however, put out with such extinguishers: in the Mont Blanc Tunnel, 14 out of 21 fires which occurred previously were extinguished in this manner. The driver of the Belgian truck tried to take the extinguisher from his vehicle, but was unable to do so because of the fire in the cab. He did not, it seems, attempt to use a tunnel extinguisher. Without a doubt, the strength of the fire core had already gone beyond the level which such an apparatus could successfully combat.
188.8.131.52. Fire Protection System
Fire hydrants are available to rescue services in the tunnel every 150 m along the southern side wall (on the right in the France-Italy direction). There are in fact two different water systems in each half of the tunnel. They are not interconnected and thus do not allow a mutual rescue operation.
On the French side, the installation took place in 1991-1992. Mountain water seeping into the tunnel is collected in a 120 m³ reservoir at rest area 16. Because of the limited water infiltration, the fill time is about eight hours. A pump under the reservoir feeds the standpipe on the French side and provides a flow of 60 of m³/h for two hours under a minimum pressure of 6 bars at each hydrant, which conforms to the requirements of circular no. 81-109 of 29 December 1981 relative to safety in vehicular tunnels. Each fire niche contains two 65 mm diameter connections, compatible with the French firefighting equipment.
The Italian system, installed later in 1997, has generally similar characteristics. The tank is located at rest area 29 and has a capacity of 117 m³. Its fill time is shorter than on the French side, since the water coming from the rocky massif is more abundant. Pressure pumps assure a flow of 96 m³/h under 7 or 8 bars at each hydrant. The connections, built to Italian standards, all have adapters for French equipment.
During the 24 to 26 March fire, only the Italian standpipe was used to combat the fire. Water was available at 60 m³/s over a period far longer than the two hours set par the French circular mentioned earlier.
A power outage occurred on the French side at about 21:42 on 24 March, due to a short circuit caused by the destruction of cables crossing in the rest area 18 ceiling. The pump was returned to service by the firefighters only in the evening of 26 March. The standpipe then fed a fogger and a hose set up in rest area 17, but when the tank was emptied, new repairs were required by the pump's power supply.
184.108.40.206. Radio Rebroadcasting
Radio transmissions cannot be received underground. Since 1995, over the total length of the tunnel, radiax cables installed in the ceiling enabled the transmission of certain two-way radio services (French and Italian operators, police and fire services of the two countries), as well as four FM stations (two French and two Italian) for the general public which can be listened to by motorists on their car radios. The control room operators have the ability to break into these latter stations, to broadcast alarm messages (the French operator can only insert messages on the French FM frequencies and the Italian operator only on the Italian frequencies).
Such messages were broadcast on the French frequencies at about 11:15. In addition, the two-way radio system was extremely valuable in allowing staff from both companies who were in the tunnel to communicate with the control rooms. A radio connection on the SITMB frequency was maintained for about an hour with the motorcycle patrol who later died in the refuge at rest area 20.
The firefighters of the two countries were able, in the same way, to stay in contact with their control room during their first entries into the tunnel. However, it appears that, at least on the French side, there were only mobile radios installed in the vehicles: when they left them, the firefighters lost the possibility of communicating with the control room by radio and became dependent on the telephone stations in the tunnel, located 300 meters apart.
220.127.116.11. Fire Detection
Two different fire detection systems were installed in the two halves of the tunnel. The Italian system, installed at the beginning of the 1990s, detected temperature increases by measuring pressure variation in 70 to 80 meter long tubes located at the top of the ceiling.
On the French side, a cable installed on the ceiling linked temperature sensors located every 8 meters. An alarm is triggered by this cable when it measures a local increase in temperature in relation to the average of the adjacent readers, or a temperature above 50º C. The data on exceeding the set thresholds is transmitted to the operators.
The Italian system provides for the automatic start up of smoke removal ventilation, after confirmation from the operator. However, this detection system had registered numerous false alarms: it was for this reason, for example, that the section located at rest area 21 had been placed out of service the evening before the fire. During the fire, the cable which collects data was lost very quickly: after input on a fault at 10:57, communication was lost at 11:02 between the data acquisition cabinet at rest area 23 and the Italian control room.
On the French side, the first alarm occurred at 11:13 from the rest area 19 zone. The next alarms occurred 6 minutes later at rest area 18 and then at intervals of approximately 4 minutes from rest areas 17, 16 and 15. At 16:51 there was a loss of data coming from the zone from rest area 11 to rest area 19.
The data acquisition system connected to the cable records the temperatures measured by the readers and saves them for four days. In effect, current day readings erase and take the place of those recorded four days earlier. Because of this record, it should have been possible to reconstruct the development of the rising temperature in the French half of the tunnel until the loss of the data acquisition system. In reality this data was lost, since the system was evidently allowed to continue functioning after the end of the fire and thus erased the readings recorded during the fire.
The Italian fire detection system did not notify the controller of the presence of a fire and also did not allow the temperature's evolution to be followed. All in all, the fire detection systems on the two sides did not give the alarm in time and did not record data that could actually be used afterwards. They were both based on temperature measurements, as is most often the case in vehicular tunnels. In reality, in the majority of fires, obscurity increases much faster than temperature, and the detection of the former often occurs too late to give an alarm.
18.104.22.168. Video Surveillance
The first video surveillance system was installed simultaneously by the two companies in the entire tunnel in 1974 and was renovated in 1994. Today it includes 40 color cameras. Each control room has access to all screens and can control them as it wishes.
In the tunnel, each camera surveys a rest area zone and is directed toward the nearest portal (toward France between the French entrance and the middle of the tunnel, toward Italy beyond). Therefore, the cameras are located about every 300 meters. Some additional cameras were installed to allow complete video coverage of the tunnel.
In the control rooms, five monitors are normally used for a cyclical visualization of the tunnel. Supplementary monitors are assigned to alarms (lifting of the extinguisher, pullbox, unhooking the telephone, etc.) and to certain cameras chosen by the operator. In case of alarm, the five cyclical monitors are automatically assigned to the camera aimed at the zone in alarm and to those located on both sides. In addition, a recording is automatically started on the alarm zone camera.
The television system allowed the two operators to see the smoke in the tunnel, when they had already been alerted (the alarm from the obscuration sensor on the French side, the telephone call from rest area 22 on the Italian side). In these two cases, the screen images of the affected zones allowed the operators to very quickly realize the gravity of the situation. The video surveillance equipment thus played a very important role in the very first moments at the beginning of the fire; then the images were no longer useful because of the dense smoke advancement.
While the video recording in the Italian control room allowed afterwards for important observations on certain events during the fire, no French recording is available. ATMB states that the operator put an unrecordable cassette into the video recorder and did not realize this. Whatever the reason, the absence of such a recording, essential to understanding the fire, is to be strongly regretted.
It can also be regretted that the systematic recording of vehicles entering the tunnel was not carried out, which would have more quickly provided the number and type of vehicles present at the time of the catastrophe.
22.214.171.124. Automatic Incident Detection
In 1994, during the installation of the new video surveillance system, arrangements were made to allow the later installation of automatic incident detection equipment based on image analysis.
Since that date, studies have progressed: a bid was offered and a company was hired. The work was to cover the entire tunnel. A first full-scale test was to take place on Monday, the 29 March 1999: eight cameras located throughout the tunnel were to be linked up to a video processor. Completion of the installation of all 40 cameras in the tunnel was planned for the summer vacation period of 1999. Each control room, in both France and Italy, would have then initiated the same procedures to generate an alarm when a vehicle stopped in a lane or in a rest area, when traffic slowed down, or when a single vehicle traveled slowly. This would have allowed a small time gain for an alarm.
126.96.36.199. Computerized Central Management (GTC)
From the beginning, information regarding the traffic management equipment (traffic signals), the air quality sensors (obscurity and CO monitors), as well as status information relating to telephones, extinguishers and fire pullboxes, has been sent simultaneously to the two ends of the tunnel by cables (called "DMR" and "telemetry") routed under the walkway.
On the French side, during 1992 and 1993, an interim computerized central management system (called "mini-GTC" by ATMB) was implemented as a test. Although not perfect, it allowed the reduction of the data from the tunnel, and the processing and recording on a printer of the status changes of certain alarms, as they come in.
The information traveling on the DMR cable was transmitted to the control centers (and processed by the mini-GTC) until 13:19 and the telemetry (obscurity and CO levels) until 13:35. After this time, no other information was received. Therefore, the mini-GTC worked well for as long as the information came to the French side.
The mini-GTC's role was, however, limited for the following reasons:
- the number of available twisted pair cables did not allow the collection of all the data coming from the tunnel (some of them were connected in parallel);
- the transmission architecture was not redundant and the loss of the DMR and telemetry cables halted the system;
- the wealth of information sometimes caused the loss of some of the data.
The available records from the mini-GTC seem reliable, but are not very detailed: there could have been data that was not recorded. It is regrettable that a tunnel of the importance of the Mont Blanc did not have a real GTC, currently available in all new tunnels. The installation of a GTC was, however, under way in a coordinated effort!by the two companies which, on the 24 March, were in the process of selecting a consultant.
188.8.131.52. Control of Tunnel Environment
The obscurity and the air concentration of carbon monoxide are monitored by nine obscurity sensors and nine carbon monoxide (CO) monitors installed close to the portals and in the tunnel approximately every 1,500 m. All measurements are recorded as a graph in each control room. This equipment, checked periodically, was all working on 24 March.
In the case of high obscurity increase, two threshold levels are applicable to the equipment readings:
- a very rapid increase threshold if the difference is higher than 5% (Westinghouse unit) between two readings,
- an alarm threshold at 20% Westinghouse obscurity, which corresponds to the motorist to the appearance of smoke that may impede driving.
The obscurity sensors proved to be very valuable, and gave the first alarm in the French control room: an obscurity alarm coming from rest area 18 was triggered at 10:52 and automatically concentrated the video coverage in that area, where the smoke could be visually verified. As the response time of the equipment was very fast, the exceeding of the thresholds recorded by the mini-GTC and the reading of the printouts give an accurate image of the evolution of the ambient air transparency and of the smoke development; these latter led to a rapid saturation of the equipment, which has a scale limit of 30% Westinghouse.
The CO monitors also have two concentration thresholds, with an alarm at 150 ppm and a scale limit of 300 ppm. These sensors are infrared and require some response time, due to the piping bringing tunnel air to the equipment. They thus give a slightly different image of the pollution characteristics.
The control and recording of the air speed is accomplished by three vortex anemometers, which replaced the original propeller anemometers. The equipment is placed close to each portal, as well as in the middle of the tunnel at rest area 18. Two units are operated by the French company and the third by the Italian company.
On the French side, it is very regrettable that the units have not been operational since 1996-1997. The operators, rescue services and investigating teams therefore did not have the measured values at the French portal and in the middle of the tunnel that could have been a great help in understanding the air and smoke movement inside the tunnel. However, on the Italian side, the anemometer was functioning; it is checked every three months, it was serviced in 1997, and its calibration was verified in the smoke tests of October 1998. In general, the readings from tunnel anemometers must be considered more as an order of magnitude, due to the unknowns such as the position of the unit with relation to the main air flow and the turbulence generated by moving vehicles. However, the information delivered by the Italian anemometer is very important in order to understand what happened. It appeared to match the various initial data (obscurity) and the calculations attempting to reconstruct the air flows.
The catastrophe which occurred on 24 March 1999 in the Mont Blanc Tunnel is the result of several concurrent causes.
- The truck that caused the fire was particularly combustible, for a vehicle which did not carry hazardous cargo in the legal sense of the term. The smoke released by its combustion was very toxic.
- The strong supply at the roadway level and the air flow at the ceiling contributed, together with the longitudinal air flow, to feed the fire and destratify the smoke, which instead of staying at the ceiling and clearing up at the motorists' level, filled the entire cross section of the tunnel.
- The hot and toxic smoke was not extracted in sufficient quantities; due to, on one hand, the exhaust capacity limitations of this particular tunnel and, on the other hand, the use of exhaust ducts in supply mode.
The role played by each one of these factors in the magnitude of the catastrophe will be estimated only when the fire development will be reconstructed by calculations. This work will take several months.
The vehicles stopped a very short distances from each other. This contributed to the rapid propagation of the fire and to trapping of the motorists in a toxic smoke cloud.
In the tunnel, there are traffic signals every 1,200 m. They were turned to red several minutes after the alarm, but they did not limit the losses, either because some of them were not working or because they were not obeyed (these signals are hardly visible).
With regard to the rescue assets of the operators, the fire configuration made the first response, which could not come close to the fire, extremely difficult.
The analysis of the fire's circumstances brought to light numerous other factors that had or could have had a negative effect:
- The tunnel does not have a safety corridor (allowing the approach of the rescue teams or the evacuation of people in refuges).
- The tunnel has two distinct control centers, one at the Italian portal and one at the French portal. Their coordination is not good.
- The operators did not know, even approximately, the number of motorists present inside the tunnel.
- The tunnel is operated by two different companies, and their actions have not always been well coordinated.
- Some of the tunnel equipment, although an upgrading program started in 1990, was not at the level of those of new tunnels. This slow catch-up is due partly to disagreement between the two leasing companies on capital investments.
- The existence of the intergovernmental control commission did not change this situation.
- The safety requirements dating from 1985 did not apply to fires.
- The number of fire drills performed was completely inadequate.
- The operators' first response assets were inadequate: on the Italian side, there was a fire engine that could not be used immediately for lack of qualified personnel.
Besides the analysis of circumstances surrounding the fire and the activities to implement its control, the task force in close collaboration with the Italian team has proposed recommendations for the Mont Blanc Tunnel that are contained in the joint report.
These will not be discussed in detail here. The following overview includes only those that are applicable to similar tunnels.
The general goal of these recommendations is to avoid the tragic chain of events leading from a technical incident to a catastrophe. Their implementation will have to be adapted to the specific facilities: length, one-way or two-way, traffic, locale, and distance from a public rescue center... It will be recalled that interministerial task forces established after the Mont Blanc Tunnel fire are already working on these recommendations.
1. The truck that started the Mont Blanc fire leads to the examination of means to diminish the potential risks of truck fires. This especially raises the issue of fuel tanks (type and capacity), and the materials used to build a tractor cab and refrigerated trailer. It also points to the possibility of a quick truck inspection before entering large mountain tunnels.
2. The truck cargo – margarine and flour – raises the issue of reexamining the hazardous cargo definition. It must take into consideration the caloric power and the possible smoke quantity that could be produced by flammable food products. This must necessarily be part of the international framework of hazardous cargo transit regulations.
3. An automatic incident detection system was being installed on the French side of the Mont Blanc Tunnel. The use of this type of installation appears necessary in all large tunnels.
4. The analysis carried out by the investigating task force demonstrates the need for a unified operating and investment policy for each of the binational leased tunnels. This must be accomplished by one operating company, a branch of the lessees.
5. For binational tunnels, it appears mandatory to implement a unique control center as well as an integrated equipment management.
6. The importance of a speedy operator response during an alarm requires a command and control system providing continuous awareness of the number of vehicles inside the tunnel and ensuring that quick and reliable first response be taken by the operator.
7. In general, the tunnel equipment and especially the electrical system must be protected from failing during a fire. In particular, the securing of networks must allow communications inside the tunnel, indispensable during a crisis.
8. In order to prevent motorists from parking behind the vehicle starting the fire, risking being trapped by the smoke and having their vehicle quickly engulfed in fire ("domino effect" as seen in the Mont Blanc Tunnel), it is recommended, on the one hand, to install systems, at least in longer tunnels with no recurrent congestion, which allow spacing of moving and stopped vehicles as well as an effective stop signal system and, on the other hand, to better inform motorists in case of emergency, perhaps by radio.
9. Faced with a fire, motorists should be able to easily find a shelter close by which is protected, equipped, and clearly marked. Information on these rescue installations must be made available to the motorists in advance. If possible, these installations must be linked to an evacuation corridor. In general, it is worth looking at the interior architecture and finishes of tunnels.
10. The operator must organize a first rescue service which is identical at each portal, allowing around-the-clock response by a team of three to five, led by a professional firefighter and capable of responding in the first five minutes of an alarm.
11. An internal safety plan must be developed by the sole operator, based on a safety study. For binational tunnels, the internal safety plan will consist of bilingual flash cards. It will provide for the alarm conditions for public agencies, the formation of rescue teams, and the activities to be conducted.
12. A unified public rescue plan must also be developed for binational tunnels. These plans must require at least one annual drill and clarify the principle of unified command and control: the public authority having jurisdiction. In general, the stringent requirements of firefighter activities in a confined environment, such as a tunnel, demand specialized training and recruitment efforts.
13. For binational tunnels, the intergovernmental control commissions must have real legal powers, i.e., their composition must be stable and especially be supported by a technical safety committee consisting of technical local and headquarters staff, competent in subjects of tunnels, civil defense and safety. This committee must be able to query the operators, to review the need for safety investments, to control the organization and the equipment of rescue teams, and to report!to control commissions.
14. For tunnels that are not binational, it will be worthwhile to implement the legal basis allowing control before construction and operation. For important tunnels and following safety studies and traffic level evolution, the documents must require the organization of a first response service at each end. In all cases, it will be mandatory to develop an internal safety plan.
© Ministère de l'Equipement, des Transports et du Logement.