Which Configurations Cause Entrapment Risk?

Full text

Proceedings for the 5th International Fire Behaviour and Fuels Conference
April 11-15, 2016, Melbourne, Australia
Published by the International Association of Wildland Fire, Missoula, Montana, USA
Which Configurations Cause Entrapment Risk?
Sébastien Lahaye*a, Thomas Curtb, Christelle Hély c
aBouches-du-Rhône Fire Service, Marseille, France, slahaye@sdis13.fr
bIRSTEA, Aix-en-Provence, France, thomas.curt@irstea.fr
cISEM, Montpellier, France, christelle.hely-alleaume@univ-montp2.fr
Introduction
Forest fires in southern Europe are characterized by their high intensity and rapid spread. The
most devastating fires occur in the hot and dry summer, especially during windy episodes
(Curt et al. 2013). Furthermore, threatened issues are increasingly important. The growth of
population in recent decades has created an extension of the wildland-urban interfaces.
This population, often rural background-free, is poorly aware of the fire risk and prevention
methods. It entirely relies on Fire Services to protect. Firefighters should thus engage to stop
fire progression before it threatens populated areas, regardless of the violence of fire. This is
why many stakeholders are trapped by the fire line and smoke. The consequences can be
severe and even fatal (Viegas and Simeoni 2011).
Some studies in Europe have described in detail accidents in order to reconstruct the
concatenation of events and human behavior that lead to tragic consequences (Viegas et al.
2009). Two scientific hypotheses are now trying to explain how fire suddenly turned to
accident causing intensity. The first considers that a number of entrapments could be the
result of a sudden and massive inflammation of fuel distilled volatile organic compound gases
(Carbonell et al. 2004). The second hypothesis suggests a runaway kinetics fire front made
possible by specific terrain conditions "trench effect" (Chatelon et al. 2014).
In the present study, we focused on the behavior of trap causing fires in recent decades, no
matter what the consequences were. From a sample of 64 events (Fig.1), we sought if the
accidents were related to weather conditions or particular fuel. We also investigated the
conditions in which accidents could take place in areas seemingly less risky.
Materials and methods
In Europe, there is no database of entrapments occurred during wildfires. However, the most
serious accidents in recent years have been the subject of investigation reports or
communications. Moreover, firefighters have left oral or sometimes written evidences of
entrapments or potentially dangerous fire behaviors. For this study we analyzed 36 reports
and undertook 18 face-to-face interviews with firefighters or foresters. The investigated
period was 1979-2015.
For 11 of the 64 recorded events, there were no casualties or destruction. Considering the
violent behavior described, it is however evident that there would have been victims in case of
staff presence.
For every reported trap we first looked for: date and time of the fire start and final surface of
the fire. Then we wondered what type of fire it was regarding (Lahaye et al. 2014)
classification.
The entrapment time (H trap) was compared with fire start time (H ignit) (Promethee 2015)
using a homogeneity test of mean values. We also compared the start-to-trap time (D trap)
with the total duration of fire (D tot). All analysis were performed in the R environment (R
Core Team 2013).

Proceedings for the 5th International Fire Behaviour and Fuels Conference
April 11-15, 2016, Melbourne, Australia
Published by the International Association of Wildland Fire, Missoula, Montana, USA
Environmental drivers
We collected air temperature (T), relative humidity (H) and windspeed (Wind) from the
closest weather station from the fire (Meteo France 2016). Depending on the year and the
station, the measuring time step could be from one to four hours but this data can have a
significant diurnal variability. This is why we also collected Air temperature at the 850 hPa
pressure level (T850) that is a representative indicator of severe fire weather (Cardil et al.
2014). We used reanalyzed data from the National Center for Environmental Prediction
(Kalnay et al. 1996) to obtain daily T850 at 12:00 UTC.
Figure 1: Entrapments location
We used the Fire Weather Index (FWI) and its sub-indices to assess the fuel sensibility (Van
Wagner 1987). We compared the entrapment data with the average values collected by Météo
France since 2001(Manach et al. 2007).
Vegetation composition within the area burned was characterized using the 1986 and 2006
French Forest Institute maps (Institut Forestier National, BDForêtV1; http://inventaire-
forestier.ign.fr). We found three main fuel classes around the entrapment locations: pine
forests (Pin=1), shrublands (Brush=1) or broadleaved forests (Pin=0, Brush=0).
Trap location
We first considered if the entrapment occurred in the main axis of the fire (Front=1) or due to
a flanking fire. Then we looked at the topography between the fire line and the trap location
(Institut Geographique National, BDAlti25; http://www.ign.fr). We noted Slope=1 events in an
upward slope or atop a ridge. We noted ParaVal=1 traps in a parallel to the main axis of fire
valley and PerpVal=1 when the valley was perpendicular. The other cases were assumed to
be Flat. By combining these positions according to a flowchart (Tab.1), we assigned each
event a risk level (Risk).

Proceedings for the 5th International Fire Behaviour and Fuels Conference
April 11-15, 2016, Melbourne, Australia
Published by the International Association of Wildland Fire, Missoula, Montana, USA
Given the number of weather data collected, we firstly tested their autocorrelation in order to
select only the most relevant using a FactomineR package Principal Component Analysis
(Husson et al. 2013). We then used an Ade4 co-inertia analysis (Dray et al. 2003) to find the
environmental drivers associated to each Risk level.
Table1: Risk level regarding entrapment position
Front
Slope or Vall
Above
Risk
1
Slope=1 or ParaVal=1
1
10
0
9
Flat
-
8
PerpVal=1
1
7
0
6
0
Slope=1
1
5
0
4
Flat
-
2
ParaVal=1 or PerpVal=1
1
2
0
1
Results
About Spain, Portugal, Greece and Croatia the entrapments that have been entered here were
the most serious whereas in France we obtained information on a broader sample, including
minor events (Tab. 2).
Table 2: Entrapment results
Trap number
fatalities
Seriously
injured
Lightly
injured
Destroyed
vehicles
Sub total in France
53
27
14
34
45
Sub tot in Po Sp Cr Gr
11
75
5
0
8
Total
64
102
19
34
53
Min
0
0
0
0
Max
24
5
10
5
Mean value
1.6
0.3
0.5
0.9
Standard deviation
3.7
0.8
1.7
1
The average size of fires with entrapment was 989Ha and the first quartile was 221Ha, greater
than the mean size of fires in France, 6.5Ha. The distribution of fires with entrapment did not
conform to the distribution of large fire types. Trap probability was higher during Oneway
fires and lower during Multiway fires.
Entrapments occurred around 16.00, ie 3 hours after the mean time of fire breakout (CI 95%,
Fig. 2). Most entrapments occurred during the first two hours of fire. This period between fire
start and trap followed a decreasing exponential that is not correlated with the total duration of
fires (CI 95%, Fig. 3).
Figure 2: Comparison between entrapment hour (H trap) and fire ignition time (H ignit)

Proceedings for the 5th International Fire Behaviour and Fuels Conference
April 11-15, 2016, Melbourne, Australia
Published by the International Association of Wildland Fire, Missoula, Montana, USA
Environmental drivers
The FWI of entrapment fires was higher than the average FWI of fires (CI95%, Fig. 4/a). The
DC and DMC were not significantly different (CI95%, Fig. 4/b and 4/c). The FFMC was
smaller for entrapments (CI95%, Fig. 4/d). T and H were not different (CI95%, Fig. 4/e and
4/f). Wind was stronger for entrapments (CI95%, Fig. 4/g).
Figure 3: Comparison between time before trap (D trap) and fire total duration (D tot)
Boîtes à moustaches
0
40
80
120
160
200
FWI IFM
Boîtes à moustaches
0
200
400
600
800
1000
1200
DC IS
Boîtes à moustaches
0
200
400
600
800
1000
DMC IH
Boîtes à moustaches
62
72
82
92
102
FFMC ICL
Boîtes à moustaches
0
10
20
30
40
50
T TEMP
Boîtes à moustaches
0
20
40
60
80
100
HHU
Boîtes à moustaches
0
20
40
60
80
100
Vent FFraf
Figure 4: Comparison between traps (left boxplot) and general population of fires (right). a/ FWI; b/DC; c/DMC;
d/FFMC; e/Temperature; f/Humidity; g/Windspeed
Trap location
Because of variables auto-correlation we rejected from co-inertia analysis T and H in favor of
T850. We kept DMC and not DC; we also kept Wind, Pin and Brush.
The entrapments scaled on 8 Risk levels. R1 to R8 were well discriminated by environmental
drivers whereas R9 and R10 were not (Fig.5). The lowest levels R1 and R2 were driven by the
presence of pines and strong wind. R5, R6 and R7 were driven by DMC and Brush and R8
was driven by T850.

Proceedings for the 5th International Fire Behaviour and Fuels Conference
April 11-15, 2016, Melbourne, Australia
Published by the International Association of Wildland Fire, Missoula, Montana, USA
Discussion
We will not make any connection between environmental drivers and entrapment injuries or
deaths because we didn’t investigate here human factors such as training, personal experience,
protective equipment, lack of communication…
Most entrapments occur during the first hours of fire. It is therefore necessary to determine in
advance the days and places where the conditions are right for this type of large fire.
Figure 5: Co-inertia analysis between environmental data and risk level
In a previous study, we showed that the probability of an entrapment increases with the
amount of vehicles involved in the fight (Lahaye et al. 2014). This is consistent with the fire
types. Multiways are less urbanized mountainous inland fires. The human issues are less
vulnerable than on the coast. So there are less resources assigned for fight and less
entrapments.
Environmental drivers
Beyond the risk of outbreak, a high FWI also gives rise to a greater danger for stakeholders.
The entrapments we studied are neither driven by fuel dryness (DC and DMC) nor by high air
temperature or low relative humidity. So it is unlikely that these entrapments are flashover
explosions caused by volatile organic gases accumulation. There is no intention to question
the conclusions of (Carbonell et al. 2004); their study focused on a limited number of cases.
However, it helps demystify the flashover explosions subjectively blamed for many accidents
by firefighters.
Strong wind seems to be a driver of entrapments but we need to be cautious as the measure is
uncertain. Local wind is often very different to the weather station registered data. More
frequent use of mobile weather stations during fires, with recorded statements, could improve
the results.
Trap location
Facing fire front in an upslope position or in a flow direction valley are the most risky
situations. Then entrapment can occur whatever the weather and fuel environment.
In pronounced drought conditions, linked to a high temperature and low humidity, there could
be entrapments for crew facing the fire on flat ground or downhill, in shrublands. This means
that in these drought conditions, any frontal attack remains dangerous.
Finally, entrapments are more likely to happen on flanking fire when the wind is very strong
and in dense pine forests. Crews that were supposed to be secured by low intensity fire lines
and by down slopes or valleys, have nevertheless been trapped.

Proceedings for the 5th International Fire Behaviour and Fuels Conference
April 11-15, 2016, Melbourne, Australia
Published by the International Association of Wildland Fire, Missoula, Montana, USA
Acknowledgements
The valuable testimonies of French firefighters and foresters and the data provided by Météo-
France made this study possible. We also thank Laure Paradis for the realisation of the map.
References
 Carbonell G, Monet JP, Dusserre G, Sauvagnargue-Lesage S (2004) 'Embrasement généralisé
éclair en feu de forêt.' Ecole des Mines d'Alès - Service Départemental d'Incendie et de
Secours des Bouches du Rhône, Alès, France.
 Cardil A, Molina D, Kobziar L (2014) Extreme temperature days and potential impacts in
Southern Europe. Natural Hazards and Earth System Sciences 2, 3863-3886.
 Chatelon F-J, Sauvagnargue S, Dusserre G, Balbi J-H (2014) Generalized Blaze Flash, a
"Flashover" Behavior for Forest Fires - Analysis from the Firefighter's Point of View. Open
Journal of Forestry 4, 547-557.
 Curt T, Borgniet L, Bouillon C (2013) Wildfire frequency varies with the size and shape of fuel
types in southeastern France: Implications for environmental management. Journal of
Environmental Management 117, 150-161.
 Dray S, Chessel D, Thioulouse J (2003) Co-inertia analysis and the linking of the ecological
data tables. Ecology 84, 3078-3089.
 Husson F, Josse J, Le S, Mazet J (2013) FactoMineR: multivariate exploratory data analysis and
data mining with R. R package version 1, 102-123.
 Kalnay E, Kanamitsu M, Kistler R, Collins W, Deaven D, Gandin L, Iredell M, Saha S, White G,
Woolen J, Zhu Y, Leetma A, Reynolds R (1996) The NCEP/NCAR 40-year reanalysis project.
Bulletin of American Meteorologic Society 77, 437-471.
 Lahaye S, Curt T, Paradis L, Hely C (2014) Classification of large wildfires in South-Eastern
France to adapt suppression startegies. In '7th International Conference on Forest Fire
Research'. Coimbra, Portugal. (Ed. DX Viegas) pp. 696-708. (ADAI)
 Manach J, Bertrand E, Bidet J (2007) 'Assistance météorologique à la prévention et à la lutte
contre les feux de forêt en zone méditerranéenne.' Société météorologique de France, Paris,
France.
 Meteo France (2016) http://www.meteofrance.com. In.
 Promethee (2015) Base de données sur les incendies de forêt en région méditerranéenne en
France depuis 1973. https://www.promethee.com/. In.
 R Core Team (2013) R: A Language and Environment for Statistical Computing. In. (R
Foundation for Statistical Computing: Vienna, Austria)
 Van Wagner C (1987) 'Development and structure of the Canadian Forest Fire Weather Index
System.' Canadian Forestry Service, Ottawa.
 Viegas DX, Simeoni A (2011) Eruptive Behaviour of Forest Fires. Fire Technology 47, 303-320.
 Viegas DX, Simeoni A, Xanthopoulos G, Rossa C, Ribeiro L, Pita L, Stipanicev D, Zinoviev A,
Weber R, Dold J, Caballero D (2009) 'Recent Forest Fire Related Accidents in Europe.'
European Commission, Luxembourg.