FATALITIES IN WILDLAND FIRES FROM 1945 TO 2015 IN SARDINIA

: The worst outcome of wildland (cid:192)res is the loss of human lives, a recurrent phenomenon over the last few decades in Sardinia, Europe and worldwide. This work analyzes all recorded fatalities in wildland (cid:192)res in Sardinia from 1945 to 2015 and trends in terms of annual number of fatalities. This time period was split due to legal and socioeconomic changes periods (1945-1975; 1976-2000; 2001-2015). We classi(cid:192)ed accident types during wildland (cid:192)res to study the most frequent causes of fatalities and how they were related 1) to involved human groups (professional (cid:192)re(cid:192)ghters, auxiliary (cid:192)re(cid:192)ghters, and civilians, 2) to (cid:192)re size and 3) to extreme weather conditions. We observed that the annual number of victims was higher in the 1981-1999 period than in other periods with 2.6 fatalities per year. Entrapment is the most frequent cause of death within the (cid:192)re professional (cid:192)re(cid:192)ghters (75.6 %). The rate of fatalities seemed to be higher in the 1981-1999 period for “civilians” and lower for “professional (cid:192)re(cid:192)ghters”. We detected that the annual number of “civilian” fatalities is higher in the 1981-1999 period with 1.6 fatalities per year. The calculated rate for “army forces & volunteer (cid:192)re(cid:192)ghters” group was in the middle ground. Entrapment is the most frequent cause of death with a percentage of 75.6 %. Fire size is a key factor in the fatality occurrence because over 80 % of deaths in wildland (cid:192)res (without considering aerial accidents) happened in (cid:192)res


INTRODUCTION
Wildland res are a growing threat worldwide, mainly in re-prone areas (CARDIL et al., 2014;SALIS et al., 2012;PEREIRA et al., 2011).Europe has experienced a large number of wild res that have caused enormous losses in terms of human lives, environmental damage and economic disruptions.The average annual burnt area in the European Union from 2000 to now is approximately 500,000 ha, causing estimated annual losses of 2 billion € (JRC-European Commission, 2013).The loss of human lives is the worst outcome of wildland res and populated areas have a higher exposure to large wildland res (LWF) than ever (VIEGAS, 2009).Wildland urban interface areas have increased in Sardinia, Italy and many other countries in last years.Both rural and coastal areas have been populated by homes which have contributed to the increase of wildland urban interface areas.Additionally, climate change and rural land abandonment are in uencing re occurrence and extreme re behavior.Climate change is providing extreme weather conditions in terms of drought and temperature (CARDIL et al., 2013;FLANNIGAN et al., 2000) and rural land abandonment provides high fuel loads and fuel complex continuities across the landscape, increasing the probability of developing large wildland res (VEGA-GARCÍA AND CHUVIECO, 2006).
The number of fatal victims in forest res was over 500 individuals from 1980from to 2007from (VIEGAS et al., 2009) ) and wildland res are one of the main threats for forests degradation (FAO, 2013).Many catastrophic wildland res with losses happened in Sardinia including 13 fatalities in Portisco re (1989) and 9 fatalities in Curraggia wild re (1983).Portisco re involved 13 civilian people in an entrapment when escaping from their houses by car during a mistral wind-driven wild re of 700 ha and maquis vegetation.The second one (Curraggia wild re; 40,000 ha) involved 9 people during a heat wave that lasted 10 days.The incident occurred in a steep slope near the village of Tempio (North Sardinia) with a high amount of dry herbaceous fuels.Fatalities were recorded in other regions of the world such as Spain (Horta de San Joan, Tarragona, 2009, 5 fatalities; Riba de Saelices, GUADALAJARA, 2005, 11 fatalities, CARDIL andMOLINA, 2015), Greece (78 people died in the summer of 2007), Portugal (22 victims in 2007(22 victims in , Viegas, 2009)), United States of America (133 re ghters died on 94 separate events in the 1990-1998period (MANGAN, 2007)), Mexico (19 volunteer re ghters lost their lives in Puebla in 1998 in a single re event, Rodríguez-Trejo et al., 2000).The last large fatality in the USA involved the Granite Mountain Hotshots; 19 crew members lost their lives on June 30, 2013, on the Yarnell Hill Fire in Arizona), Australia or Chile (CARDIL AND MOLINA, 2015;MANGAN, 2007;VIEGAS, 2009;CABBIDU et al., 2011;HALTENHOFF, 2003).Lastly, these two are the most impressive historic accounts on re events with fatalities: 1) In Southwestern France, near Bordeaux, a forest re kills 230 people and burns more than 100,000ha (19-August 1949) (DEVILLE, 2009), and 2) In USA (Idaho-Montana), The Great Fires of 1910 (also commonly referred to as the Big Blowup) killed 87 people and burned about 1,200,000ha (EGAN, 2011).
Most re managers consulted (MOLINA et al., 2010) agreed that a re with victims escalate to a completely different emergency level.First, a new emergency (i.e., taken care of the victim) occurs during the propagation of the wildland re, involving lots of resources to address the victim.Second, when there is a dead one in a wildland re, social pressure increases quickly and re suppression resources are less ef cient in these situations (i.e., thinking about affected teammates (victims), suffering uncertainty).High quality training on safety protocols such as LACES or Dead Man Zone (GLEASON, 1991) should be critical to pursue a lower accident rate (MOLINA et al., 2010).The LACES protocol includes Lookout, Anchor lines, Communications, Escape routes, and Safety zones.Elsewhere are speci c details on LACES safety protocol (MOLINA et al., 2010) or LCES protocol (USDA Forest Service, 2004).
Both re ghters and citizens are at risk in wildland res and many people were involved in several wildland re accidents in Sardinia.Large amounts of resources have been invested in re suppression resources to reduce the negative consequences of res and to provide safety to the population.However, re ghters sometimes do work in a high-risk environment and fatalities may occur.Individuals involved in different tasks of re management are subject to diverse dangers including burnover, vehicle and aircraft accident, and medical emergencies (MANGAN, 2007)

Study area
Sardinia, Italy, is a large island in the Mediterranean Sea (24,235 km 2 ) located between 38º51´N and 41º15´N latitude and 8º8´E and 9º50´E longitude (Figure 1) and has a population of about 1.7 million inhabitants.Overall, the island has a complex topography with hills and low mountains (RICOTTA et al. 2012).The average elevation of the island is 338 m a.s.l. and the highest point is Punta la Marmora with1834 m a.s.l. in the center of the island.The climate is classi ed as Mediterranean, with dry hot summers and an important water de cit from May to September (CHESSA AND DELITALA, 1997).The mean annual temperature ranges from 17.8 ºC in the Southern coast to 12.8 ºC in the mountainous areas.Maximum temperature peaks are higher than 30.8 ºC during the summer season.Average annual rainfall is 1300 mm in the mountains, but slightly less than 500 mm in the coast, and most of the annual rainfall occurs in fall and winter.The ora includes 2407 taxonomic species, with 10% of endemic species (DE ANGELIS et al., 2012).Large areas of the island are covered by scrub and/or herbaceous vegetation associations, comprised primarily of Pistacia lentiscus L., Arbutus unedo L., Erica arborea L., Myrtus communis L., Olea europea L., Phyllirea spp., Juniperus spp.and Cistus spp.(SALIS et al., 2012).Woodlands and forest area is approximately 55 % of Sardinia (National Forest Inventory, C.F.S., 2005) with the main tree species being Quercus ilex L., Q. suber L., Q. pubescens Willd., and Q. congesta Presl.Pine plantations with Pinus pinea L., P. halepensis Mill.,other pines and Eucalyptus sp.only spread over 3% of the island and are mainly concentrated along the coast (SALIS et al., 2012).Pastures and agricultural lands represent about 36% while the urban areas cover 3% of the island.

Fire and fatalities data
Historical re data records with fatalities from the Sardinian Regional Forest Agency (CFVA -Corpo Forestale e di Vigilanza Ambientale) were used from 1945 to 2015 in Sardinia.The CFVA database has an entry for each wild re ignition and provides information on date, municipality and location of the ignition, and area burned.Although information about re size was not available before 1971 for all res, re size for those res with fatalities was accessible.All fatalities recorded in the databases were veri ed through interviews with senior managers and by analyzing documents of the accidents and news in newspapers.Therefore, data about fatalities in wildland res in re suppression systems is accurate.
On average, from 1971 to 2015 (from re data base collected burned areas systematically), Sardinia experienced 2,931 res per year and about 18,800 ha burned per year.Wild res are typically concentrated from June to September, with the maximum of both ignitions and area burned in July (SALIS et al., 2012).Similar to elsewhere in Euro-Mediterranean ecosystems, a few large wildland res account for most of the burned area (CARDIL and MOLINA, 2013).However, the aim of this work is to analyze if re size could be an in uencing factor in the victim occurrence.For this purpose, we assessed the relation between fatalities and re size, segregating all wildland res with victims in ve size classes (Table 1).We assessed if the re size was associated with the number of victims and if the largest res might have the highest percentage of victims.
In the case of fatalities, we evaluated the death cause in all wildland res using re of cial CVFA data as well as the available information in several newspapers.We classi ed all victims according to the following death causes: a) Physical cause: this category includes different medical causes (including heart attack and dizziness); b) Entrapment: either when a wildland re reaches the victim or death by hot gases;  in the re suppression system from the Italian National Firemen Agency, CFVA and FORESTAS agency (including forestry seasonal workers); 2. Auxiliary re ghters (AF).Armed forces and volunteer re ghter groups; 3. Civilians: Many of the victims here fall in the sub-category of shepherds.
We have worked to assign a normalized presence of these groups in suppressing active res to improve the study.We strongly recognize that amount of hours-worker (or days of work) that is committed to suppress active res is higher for the group "professional re ghters" than for "army forces & volunteer re ghters".Our best educated guess is that it is 4.0 times higher for "professional re ghters".This is to say that, in average, there are four "professional re ghters" for each "army & volunteer" worker at any given re event.This will be important latter to calculate a normalized fatality ratio.

Statistical analysis
We determined if there were signi cant changes in the number of dead people in Sardinia from 1945 to 2015 using three different time intervals considering socioeconomic (Law 47/1975 that assigned in the whole Italy the wild re coordination to State Forest agency and law 353/2000 that de ned a wildland re as an emergency of civil protection), land abandonment and re suppression changes.We used three time intervals , [1981][1982][1983][1984][1985][1986][1987][1988][1989][1990][1991][1992][1993][1994][1995][1996][1997][1998][1999] and [2000][2001][2002][2003][2004][2005][2006][2007][2008][2009][2010][2011][2012][2013][2014][2015].The rationale behind this partition is 1) from 1981 more human resources are committed to res suppression, and 2) from 2000, it is fully applied the legislation piece entitled "occupational safety consolidated text" and wild res became a civil protection issue, not a forest problem as in previous time.In the rst time interval, fuel loads and continuity were very low due the intensive agroforestry use of the territory and res were smaller in number and mean c) Aerial accident: either landing, or taking off, or in cruising ight; d) Terrestrial accident: including a person stepping on a power line, accidents while driving vehicles, falling down with lethal injures; Sardinia developed an original re-ghting command system that had different phases since 1945.Originally, res were extinguished by volunteers groups (slightly trained) and coordinated by the Italian National Forest Corps.In 1975 a new framework law (L.47/75) assigned the wild re coordination responsibility to Italian National Forest Corps.In 1985 the new CFVA (Sardinian Regional Forest and Environmental Corps) was structured (from personnel from the Italian National Forestry Corps) and renewed its components with a great recruitment in the '90s".CFVA took care of re suppression tasks using forestry workers and rural police.In the '70's the use of helicopters and terrestrial machinery in suppression increased and the military aerial eet appeared as a new tool.In 2000, a new Italian national law (L.353/00) de ned wildland res as emergencies of civil protection although re suppression efforts were still the responsibility of the CFVA.In 2015, many personnel were involved in re suppression in 377 municipalities: some 1,400 within CFVA, approximately 2,000 forest workers and re ghers (now in FORESTAS Agency), 3,500 volunteer re ghters (from different Civil Protection), 4,000 rangers from private rangeland states ("barraccellos"), and some 700 Structural re ghters (Italian National Firemen Agency).All fatalities in Sardinia from 1945 to 2015 are shown in Table 2.
All fatalities were analyzed and classi ed in relation to the cause of death and the job duties of the victims in the re suppression system.The categories used to describe the involvement of the victim in the wildland re were the following (Table 2): 1. Professional re ghters (PF).In this category, we included forest rangers, fuel and forest managers or supervisors, re crew members and people who worked size than in the second period (DELOGU et al., 2013).
In the second period, land abandonment started to be widespread and both the fuel load and re occurrence increased.In the third period, re agencies improved remarkably their training, re behavior knowledge and suppression resources.Land abandonment was even more evident (in the third period) and large wildland re occurrence with extreme behavior larger than ever.In the third period, the culture of working hard to avoid job hazard is widely spread in the re ghing operations (DELOGU et al., 2013).
All fatalities in entrapments, terrestrial accidents or physical causes were classi ed in relation to either high temperature days (HTD) or to strong wind patterns with gust from 80 to 100 km .h -1 (no available information about wind speed in all res) in order to identify relationships between victims and extreme re weather (wind and temperature).HTDs provide extreme weather conditions and in uence re occurrence, re size and behavior (CARDIL et al., 2013;CARDIL, et al., 2014).Additionally, HTDs could be a key factor affecting fatalities because they provide extreme conditions for re ghters (less effectiveness, tiredness, exhaustion, dehydration) and, additionally, re behavior is more extreme (CARDIL et al., 2013) (KALNAY et al., 1996) were used to characterize HTDs on a synoptic scale.NCEP outputs are available from 1948 to the present.Daily air temperature data at the 850 hPa pressure level at midnight 00:00 UTC were analyzed to assess the in uence of high lower troposphere temperatures.A high temperature day was de ned as a day with an air temperature at 850 hPa higher or equal than the 95 th percentile of that temperature in Sardinia (CARDIL et al., 2014).

RESULTS
In the period 1945-2015, 74 people died in wildland res in Sardinia (Table 2), of which 17 worked for the professional re ghters (PF).12 auxiliary re ghters (AF) died in the study period.Therefore, the rest, 45 people without relation to the re suppression systems (civilians, many shepherds) died in wildland res.All fatalities occurred in the re season (May to September, both included).Most of fatalities were in August (49.3%) and July (31.5 %).

Groups
Fatality rates were the higher for "civilians" and the lower for PF.Differences among AF and PF fatality rates were not signi cant at p-value<0.05(Table 4).During the 1981-1999 period (when most fatalities occur), if we consider that re exposure in PFs is much more than for AF, PF normalized losses are lower than AF normalized losses (p-value=0.02;Table 4).

Causes
In relation to the total number of fatalities in Sardinia, re entrapment was the most frequent cause with 56 victims (75.7 % of the total number of deaths).Aerial accidents comprised 12.2 % (9 victims), terrestrial accidents 9.5 % (7 victims) and physical causes 2.7 % (2 victims; Table 2).Table 2 shows the number of fatalities in relation to the cause of death and role of the victims in Sardinia from 1945 to 2015.

Fire size
Signi cant differences in the leading cause of mortality across different re sizes were found.For res larger than 100 ha, 82.1% of deaths fall in the re entrapments cause.For res over 500 ha, this percentage was 80.4% and for res over 1,000 ha, 37.5% (Table 1).Considering terrestrial accidents, aerial accidents and fatalities due to physical causes, all victims occurred in res smaller than 100 ha (none in res larger than 100 ha).

Extreme weather conditions
We studied both high temperature days (HTDs) and strong wind days (SWDs thereafter) with gust from 80 to 100 km .h -1 to check their share of total fatalities in these class days.In re entrapments fatalities, 26 re ghters (46.4 % of the total) died under HTDs.28 % of total deaths in terrestrial accidents occurred under HTD.However, the percentage of deaths in aerial accidents under HTD was 0.
Regarding extreme wind events recorded, we worked with data from 16,790 days in the register in Olbia Costa Smeralda weather station (40° 53' 56.49" N, 9° 30' 56.33" E).These days are from 1969 to 2015.Only 7 days (out of 16,790) had maximum wind speed 80km/he and higher.This is 0.0041 % of the days.Under SWDs, 18 people died in Portisco re (13 deaths) and Porto S. Paolo re (5 deaths).This happened near touristic settlements, generally trying to escape in narrow roads, into a car or waking, in days of strong mistral (NW) wind speed 80-123 km .h -1 .This is 24.3 % of total deaths in SWDs.
The increase in the number of fatalities could be related to more extreme re behavior in last decades (CARDIL et al., 2014) and more numerous human resources in re suppression.The more people involved in re suppression, the more likely to have fatalities.
Fire suppression is a high-risk profession (MANGAN, 2007;VIEGAS, 2009).Society and all involved agencies have to be informed about the job hazard situations and all available protocols to try to reduce victims in wildland res.We recommend: 1) establishing new approaches to reduce the number of victims (i.e., do not involve too many people, do not involve many volunteers and do involve mostly professional re ghters), 2) increasing re suppression safety of both personnel and civil population, 3) training with re in prescribed burns to gain experience for both professional and volunteer re ghters (simulating stress in decision making and suffering communication dif culties to simulate an eventual large re scenario).
The larger the prescribed burn sizes the better to feel closer to potential wildland re situations.As reported in Molina et al. (2010) in several regions of Spain and abroad, prescribed burning is seen as the key training tool to enhance re ghters safety.The reason is that specially in large prescribed res, we are forcing the crews to relay in ef cient communications as not all re ghters are able to see the whole area burning at that time.And in all prescribed burning actions we do enforce the safety protocols.This could ensure that re crews do follow safety protocols and ef cient communication protocols in an eventual re suppression action.
These training actions (with prescribed burning) could be useful for citizens to understand a re situation, how suppression agencies work and how they should behave in an eventual wildland re emergencies.It is remarkable to note that most entrapments are close to roads.Today, not all roads in Sardinia are safe enough to locate re engines to attack a wildland re.Eastaugh and Molina (2012) have shown how to assess which roads are safer using extensive simulations with FlamMap (FINNEY, 2002).With that Eastaugh and Molina (2012) criteria some roads could be classify as unsafe for most re ghters in suppression action.
The normalized dead accident rate (deaths per 100,000 burned ha) was 4.8 in Sardinia, a value similar to Spain with 4.5 fatalities per 100,000 burned ha.By contrast, in the United States, this value was 0.91 (MANGAN, 2007(MANGAN, , 2009)).It is true that in the US large burning areas are dealt with minimum intervention (not so many re ghters as in Europe).
The re entrapment is the most frequent cause of death in wildland res accounting for more than 75 % of fatalities.This percentage was lower in Spain, around 50 % (CARDIL and MOLINA, 2015) and United States, 14 % (MANGAN, 2007).Additionally, the normalized dead accident rate in entrapments in Sardinia was 4.3, a value higher than other countries like Spain (2.15, CARDIL and MOLINA, 2015) and United States (0.19;MANGAN (2007MANGAN ( , 2009))).This may indicate that our LACES and Dead Man Zone protocols are not working good enough.Therefore, we could improve our gures trying to reduce the entrapments by working harder in safety protocols during re suppression.
Considering the normalized dead rate in aerial accidents, we have got similar results in Sardinia than in Spain (0.78 vs 0.82, respectively).This value was lower in the United States (0.22).This could be related to the higher use of aerial resources in Spain and Sardinia versus the USA.Both Spanish and Italian re suppression systems use aerial resources even in small res; therefore, increasing the probability of aerial accidents.Indeed, most aerial fatalities occurred in res smaller than 100ha (100 % in Sardinia and 56.5 % in Spain).Aerial accidents require an aeronautical review of the accident causes, and this is not in the database, and therefore, we cannot go further in this review regarding cause.
Fire size is a key factor in victim occurrence in re entrapments in Sardinia.This fact was also found in Spain (CARDIL and MOLINA, 2015).This could be related to two factors.First, in large wildland res, re behavior is normally more adverse and the probability of entrapment could be higher.Second, when a fatality (or serious injury) occurs in a wildland re, an additional emergency starts.
Therefore, many resources become diverted to try to mitigate the consequences of the accident, allowing the re to propagate more freely and thus increasing the re size.Implementing suitable forest management could moderate the potential re behavior and diminish the re potential, eventually decreasing the burned areas and the probability of accidents (CARDIL and MOLINA, 2013).
It would be interesting to investigate the relationship among fatalities and vegetation types (i.e., grassland, shrub-land and forest) although we did not have enough data to perform the analysis.
Extreme weather conditions (HTD and SWD) could also in uence the victim occurrence due to a high percentage of fatalities in entrapments under this type of condition that represent a mean of 6 days per year.This is similar to what was reported in Spain with a 60% of the fatalities in entrapments in wildland res occurred under high temperature (CARDIL and MOLINA, 2015).In this HTD conditions, re behavior is more adverse to re ghters, although as Beaver (2002) states that it (an extreme re behavior) is not necessary the cause accident but the lack of matching human behavior to re behavior.Misperception of re risk by re ghters matters.However, this high clumpy distribution is not clear in aerial and terrestrial accidents (only 14 % under HTDs).Therefore, we have to focus the safety protocols to avoid entrapments under high temperature days (HTDs).Additionally, the annual average number of HTDs increased in South Europe and Sardinia from 1980 to 2010 (CARDIL et al., 2014).Climate change projections (IPCC, 2014) indicate that we will experience more extreme temperatures that could in uence new accidents in years to come.Extreme weather conditions in uence re behavior and, therefore, increase the probability of entrapment.In addition, these conditions could affect re ghters and people due to dehydration, stress and worse working conditions.Many accidents were related to strong NW winds (Mistral winds) with a gust wind speed close to 100 km .h -1 .These are the most important situations to lead to fatalities: extreme temperature conditions from dry air masses from Africa and strong NW wind.

CONCLUSIONS
We have more deaths (fatalities/year) since 1981 than in the 1945-1980 period.After 1981, the number of fatalities has been reduced up to 0.6 fatalities per year in 2000-2015 period but this change was not signi cant.The entrapment is the most frequent death cause (75.6 %) and it is here where agencies have room for doing things better.Fire size is a key factor in fatality occurrence because over 80 % of deaths in wildland res (without considering aerial accidents) happened in res larger than 100 ha.
Extreme weather conditions (high temperature or strong winds) in uenced victim occurrence because almost half of entrapments occurred under extreme temperature days that only represent approximately a 5 % of the total days in the re season (June-September).Under extreme re behavior, re-ghters should act in very a different way and the accidents should not be more likely.Unfortunately, we have more accidents and this is because there is a misperception of re risk by re ghters more often under extreme re behavior circumstances.High quality safety protocols and extensive training (i.e., prescribed burning operations with emphasis on safety protocols are critical to pursue a lower accident rate.Lessons learned, and shared among different agencies (national and internationally) should encourage agencies to opt for high quality training on safety protocols.Fatality rates are higher for "civilians" and lower for PF.In the 1981-1999 period (when most fatalities are present), PF normalized losses are lower than AF normalized losses.Therefore, it is important to train PF thoroughly to maintain the recent (2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015) low fatality rates in these workers that are exposed to high structural job hazard in suppression actions.

FIGURE 1
FIGURE 1 Geographic location of Sardinia (Italy) in the Mediterranean.

FIGURE 2 FIGURE 3
FIGURE 2 Total burned area and re number in Sardinia from 1971 to 2015 .

TABLE 2
Number of fatalities according to the cause of death and role for victims inSardinia (1945Sardinia ( -2015)).

TABLE 4
Number and normalized number of fatalities considering the exposure factor for professional re ghters, Armed forces and volunteer re ghters and civilians in three study periods inSardinia (1945Sardinia ( -2015)   )