Surviving bushfire: the role of shelters and sheltering practices during the Black Saturday bushfires

Environmental Science and Policy 81 (2018) 86–94

Contents lists available at ScienceDirect

Environmental Science and Policy journal homepage: www.elsevier.com/locate/envsci

Surviving bushfire: the role of shelters and sheltering practices during the Black Saturday bushfires

T

⁎

Raphaele Blanchia, , Joshua Whittakerb,d, Katharine Haynesc,d,e, Justin Leonarda, Kimberley Opiea a

CSIRO Land & Water, Australia Centre for Environmental Risk Management of Bushfires, Institute for Conservation Biology and Environmental Management, University of Wollongong, Australia c Department of Geography and Planning, Macquarie University, Australia d Bushfire & Natural Hazards Cooperative Research Centre, Australia e Risk Frontiers, Macquarie University, Australia b

A R T I C L E I N F O

A B S T R A C T

Keywords: Sheltering WUI policy bushfire wildfire

The decision of whether to leave or stay and defend is a well communicated public safety policy for those at risk from bushfire in Australia. Advice relating to sheltering practices during bushfire is less developed. This paper presents findings from a study of sheltering practices during the 2009 Black Saturday bushfires. The study examined the circumstances and challenges experienced by residents when sheltering and/or exiting houses, sheds, and personal bunkers. The analysis considered a number of factors including human behaviour and decision making, house design and construction, the surrounding landscape and fire behaviour. The results show the need for contingency planning and the need for active sheltering, involving regular monitoring of conditions inside and outside the shelter and actions to protect the shelter and its occupants. Also discussed is the tenability and location of the shelters and key questions around how bushfire-related building controls can improve the predictability of shelter failure, reduce the rate of shelter tenability loss and facilitate egress. This research highlights the need for enhanced community engagement and education to encourage residents to plan and prepare for active sheltering.

1. Introduction Urban development and population growth in the Wildland Urban Interface (WUI) is increasing the exposure of communities to bushfire (or wildfire) risk in many parts of the world (e.g. Gill, 2005; LampinMaillet et al., 2010; Theobald and Romme, 2007). In response to bushfire threat, people have the option to evacuate, shelter-in-place or shelter as part of their home defence strategy (e.g. Cova et al., 2009; McCaffrey and Rhodes, 2009; Tibbits and Whittaker, 2007). However, as many studies have highlighted, the decision is not a simple one (e.g. Cova et al., 2009; Handmer et al., 2005; Haynes et al., 2010, 2009; McCaffrey and Rhodes, 2009; Paveglio et al., 2010; Stephens et al., 2009; Tibbits et al., 2008; Tibbits and Whittaker, 2007; Whittaker et al., 2013). Early evacuation has been the preferred option in the United States (US) and in some countries in Europe, but this action is becoming increasingly challenging, and alternatives to evacuation are now being widely considered (Cova et al., 2009; McCaffrey et al., 2015; McCaffrey and Rhodes, 2009; Paveglio et al., 2008, 2010). Early evacuation is not always possible due to an inability to provide early warnings, land use ⁎

planning and development that is not conducive to swift egress, and dangers associated with mass evacuations such as accidents and traffic jams (McCaffrey et al., 2015). Research on a range of hazards suggests that sheltering during flash floods, cyclones, and radioactive and chemical emergencies may be safer than late evacuations (Haynes et al., 2009; Scanlon, 1992; Yard, 2000). In Australia, fire agencies have historically encouraged a ‘shared responsibility’ approach with the objective of developing a range of risk mitigation measures (including self-protection measures) to protect life and assets during bushfires. The fire services have endorsed the ‘Prepare, stay and defend or leave early’ policy (Australasian Fire Authorities Council 2005, AFAC 2005). Under the policy, residents are advised to prepare, stay and defend their homes against bushfire, or leave before a fire threatens them or blocks their evacuation route (Handmer et al., 2005; Whittaker et al., 2013). The policy, colloquially known as the ‘stay or go’ policy, drew heavy criticism following the 2009 ‘Black Saturday’ bushfires, which resulted in the deaths of 173 people, 118 of these people perished while sheltering in structures, including 104 in residential buildings (Blanchi et al., 2015). A

Corresponding author at: CSIRO Land & Water, Private Bag 10, Clayton South VIC 3169, Australia. E-mail address: [email protected] (R. Blanchi).

https://doi.org/10.1016/j.envsci.2017.12.013 Received 3 September 2017; Received in revised form 21 December 2017; Accepted 21 December 2017 1462-9011/ Crown Copyright © 2017 Published by Elsevier Ltd. All rights reserved.

Environmental Science and Policy 81 (2018) 86–94

R. Blanchi et al.

subsequent Royal Commission into the fires (Teague et al., 2010) concluded that the basic premise of the policy remained sound and specified that “leaving early, before there are fires, is the safest option; staying to defend a well-prepared, defendable home is a sound choice in less severe fires for those who are mentally and physically able”. The Royal Commission also recognised that early evacuation is not always possible and recommended that people have contingency plans in the event that they are unable to leave. The Commission identified sheltering as one of the contingencies that people should be prepared for (Teague et al., 2010). The practice of sheltering during bushfires is not a recent phenomenon in Australia and has been used in rural areas since European settlement (Handmer et al., 2005). Early accounts of sheltering come from people working or living in the bush at the time of the 1939 Black Friday bushfires in Victoria, when people retreated inside structures such as houses, timber mills or dugouts (Stretton, 1939). However, to date, there has been no specific research on sheltering practices in Australia apart from McLennan’s report on informal places of shelter used in Black Saturday bushfires (McLennan, 2009). Research on resident planning and preparation identifies a number of factors and circumstances that may increase the risk to residents taking shelter during a bushfire (Handmer et al., 2005; McCaffrey and Rhodes, 2009; McLennan et al., 2012; Penman et al., 2013; Tibbits et al., 2008; Whittaker et al., 2013). Residents who wait until the last moment before taking action are more likely to be faced with multiple high-risk options ranging from unsafe late evacuation to sheltering in a poorly-prepared structure (McLennan et al., 2012; Whittaker et al., 2009). Buildings and surroundings could be designed to sustain a certain fire exposure and mitigate some of these risks identified. However, variation and unpredictability in fire behaviour make it difficult to characterise the fire exposure and the vulnerability of structures (Blanchi et al., 2014; Cova et al., 2009; McCaffrey and Rhodes, 2009; Paveglio et al., 2008). In addition, there is little research on the compatibility between peoples’ behaviour and these measures or their efforts to prepare and maintain structure survivability. A better understanding of the factors influencing safe sheltering in terms of preparedness, behaviour, types of shelters, response to fire, smoke and vulnerability of the shelter, is needed. Fundamental questions remain regarding residents’ knowledge and understanding of safe sheltering practices, the adequacy of peoples’ preparation and responses, the adequacy of shelters to withstand the bushfire and the challenges people face when trying to shelter. In this paper, we examine the following questions:

bushfires on 7 February 2009 in Victoria, south-eastern Australia. The broader study used both quantitative and qualitative analyses of sheltering practices. This paper focuses mainly on the quantitative analyses; results of the qualitative analyses are presented in Whittaker et al. (2017). While the Black Saturday fires may be considered an extreme fire event, the data was chosen because it covers a variety of demographics (urban, peri-urban, rural areas), has a significant amount of consistent data available and removes variation in fire weather, operational procedures, policy and guidelines. A spatial database was specifically developed to facilitate the quantitative analysis. This dataset included tabular data stored in a Microsoft Access database and associated spatial data stored in the Geographical Information System (GIS) software ESRI ArcGIS. A large amount of information on the 2009 bushfires was available to determine accurately the spatial location of the people and shelter(s). A high accuracy level was obtained for most of the cases (precision of ± 10 m). The database contains 325 incidents involving 169 fatalities and 861 survivors. Some of the spatial locations could not be identified and the data associated with those locations are not included in the spatial analysis (n = 169 fatalities, n = 838 surveyed survivors location(s)). 2.2. Data sources The sources of data used included witness statements presented at the Victorian Bushfire Royal Commission (VBRC, Teague et al., 2010), semi-structured interviews conducted by the Bushfire CRC Research Taskforce with residents in affected areas (Whittaker et al., 2009), a dataset containing bushfire related life and house loss (Blanchi et al., 2012) and other available information from books, reports, journals and personal accounts that are publicly available. The location of people was recorded utilising available geo-registered high-resolution aerial photography and existing spatial datasets (Blanchi et al., 2012). Information on house locations was obtained from the life loss database (Blanchi et al., 2012), the Bushfire CRC Research Taskforce 2009 bushfire survey (Leonard et al., 2009) and the National Exposure Information System (NEXIS) database that was developed by Geoscience Australia (Nadimpalli et al., 2007). In addition, other data were also included such as distance and density of surrounding forest using the National Carbon Accounting Forest dataset (Furby et al., 2009). Surrounding vegetation was characterised by the Dynamic Land Cover Map (Lymburner et al., 2011) and the Ecological Vegetation Class (EVC) dataset. A fire severity layer was produced for the Kilmore region (Cruz et al., 2012) and further vegetation characterisation was derived from the vegetation layer developed for this specific study.

1. Where did people shelter during the 2009 Black Saturday bushfires? 2. What actions did people take? What risks did they face while sheltering? 3. Is there a relationship between proximity to fuel, or fire severity, and the residents sheltering experiences? 4. What other factors influence shelter survival during a bushfire?

2.3. Data collection and analysis Initial analysis through word searches of the 611 (total) interviews revealed that 315 contained references to sheltering. These interviews were used for this research. All transcripts of evidence from lay witnesses (100 witness statements) and hearings presented to the VBRC were considered in this study together with any associated material such as photography or documentation provided by the witnesses (Teague et al., 2010). Half of the witness statements (50 cases) described sheltering practices and experiences during the fire and were included in this study. Data were collected for a number of variables (Table 1). The intentions and actions of people involved in sheltering and descriptions of the circumstances of sheltering were recorded as free text. Where available, data were collected on the number of people, and whether they survived or perished while sheltering. The geographic location of the originating residential address and place of shelter(s) was recorded for each individual (where possible) using Google Earth and Geographical Information System (GIS)

Much can be gained from a better understanding of resident sheltering and egress experiences in combination with a better understanding of how house design and fire severity influence the modes and rates of loss of house tenability. In this study, particular attention has been given to understanding how global parameters such as fire severity, exposure to heat, fire and smoke; and the type and vulnerability of the shelter relate to the perceptions, behaviour and conditions experienced by residents. 2. Methods 2.1. Objective The work presented here was part of a broader study focused on the sheltering experiences of people affected by the Black Saturday 87

Environmental Science and Policy 81 (2018) 86–94

R. Blanchi et al.

Table 1 Data summary. Data

Type of data and coding

Data sources

Intention, decision making and circumstances of sheltering (including challenges, shelter failure, tenability, location and movement) House construction and defence Number of people sheltering Active or inactive behaviour Type of shelter

Free text

Combination of sourcesa Combination Combination Combination Combination

Details of type and location of shelter Number of shelters used Shelter status after bushfire Shelter failure while sheltering Shelter location

Free text Number Active/inactive/unknown - House (residential building), - Commercial and public building (e.g. school, hotel/ pub, shop), - Other structure (shed, bunker, cellar, water tank), - Open space (e.g. paddock free from elevated vegetation), - Outside behind a physical barrier (car, wall,…), - Water body (dam, swimming pool, lake), - Car/vehicle, - Unknown, - Non Applicable, - Defending outside. Free text Number Destroyed/survived/unknown/Non Applicable Yes/no/unknown/Non Applicable Spatial (GIS)

Person’s residential location

Spatial (GIS)

People movement Distance to forest

Spatial (GIS) Meters

Vegetation effect

1. 75–100% Crown consumption, 2. 25–75% Crown consumption, 3. < 25% Crown consumption, > 60% Crown scorch, 4. 30–60% Crown scorch, 5. 0–30% Crown scorch, 6. Burnt grassland 7. Unburnt grassland

of of of of

sourcesa sourcesa sourcesa sourcesa

Combination of sourcesa Combination of sourcesa Combination of sourcesa Combination of sourcesa - Combination of sourcesa - National Exposure Information System (NEXIS) - Combination of sourcesa - National Exposure Information System (NEXIS) Combination of sourcesa Forest extent (FNF) layer, Furby et al., 2009) Fire severity (Cruz et al., 2012) (Kilmore fire only)

a Semi-structured interview (Bushfire CRC Research Taskforce), witness statements (VBRC, Teague et al., 2010), other sources (books, reports, etc.), Attorney General Database Life and house loss dataset (Blanchi et al., 2012), Victorian bushfires (N = 173), Taskforce 2009 bushfire survey (Leonard et al., 2009).

personalised narratives that offer unique insights into people’s experiences and understandings of events and phenomena. Caution should be taken when utilising these findings as the sample might not be representative of the entire population of people who survived while sheltering. There is no intention to use these results as a statistically confident predictive model. A range of descriptive statistics were derived using the quantifiable and categorical data extracted from the interviews, witness statements, other sources and spatial analysis to examine: the type of shelter used by residents when sheltering, the number of shelters used, the shelter vulnerability, the relationship between shelter location, surrounding forest, fire severity and the residents’ sheltering experiences. We also analysed the effects of individual variables and combinations of variables on the probability of survival using general linear models to determine those variables that had the largest impact on the chance of an individual’s survival.

software (ESRI ArcGIS). The type of shelter used was coded into eight categories (Table 1). Where people used a residential house for shelter, the location and known movements of these people within the house was recorded to understand better the behaviour adopted by fatalities and survivors and the modes in which the house may have lost its tenability. If multiple places of shelter were used, then the movement, each of the locations and the type of shelter were recorded along with the status of the shelter after the fire, where it was known (destroyed, minor or no damage). The distance from a location of shelter to residence and between shelter(s) was recorded where available. The straight line distance to forest and the fraction of forest within 100 m of all locations (people, place of residence and shelter(s)) was collected. Information on house construction and defence was recorded as free text to provide additional context regarding house vulnerability to fire exposure and the influence of occupant behaviour if available (Leonard et al., 2009). Several limitations were encountered when using witness statements and interviews as data sources. Members of the public were invited to submit witness statements for the VBRC or submit evidence to the VBRC (N = 100). The interviewed sample (N = 611) covered the major fire complexes of the Victorian 2009 bushfires and can be considered broadly representative of what happened in different communities and locations. However, these interviews were semi-structured, meaning that researchers asked a series of open-ended questions to explore key issues, but allowed interviewees to frame and structure their responses (Rubin and Rubin, 2005). This process can provide rich,

3. Results 3.1. Type of shelter 3.1.1. Sheltering Location Results from the data show that many people used residential buildings as their main place of shelter (n = 575 for their first shelter, n = 14 for their second shelter and n = 11 for their third location of shelter), representing the largest proportion (60%) of all shelter types (Table 2). In most cases, residents took shelter inside the house during 88

Environmental Science and Policy 81 (2018) 86–94

R. Blanchi et al.

Many people mentioned sheltering behind a physical barrier to protect themselves against the effects of the fire. Several residents mentioned the need to shelter around the house, behind water tanks, sheds, cars or tractors while defending the house and property. In a few cases, people evacuating a burning shelter had to temporarily find protection behind a physical barrier such as a car, machinery, carport, trees or fences. Residents reported that they needed to be sufficiently close to the house to be able to go inside if necessary, but also to be far enough away to not be affected by burning objects. Sixty-six people fled their property by car (Table 2). Most of these people found shelter in a stationary position in an area cleared from vegetation (driveways, burnt paddock) or in a community place of refuge such as a sports oval or golf course. A few residents sheltered in their cars while driving around to protect themselves from burning elements. Overall, residents experienced difficulties due to poor visibility, obstacles such as fallen trees and other vehicles and burning objects. Different types of water bodies (e.g. swimming pools, dams, lakes; see Table 2) were used by residents, often as a secondary place of shelter. Residents described them as places of last resort and many used fabric (woollen blankets, towels) to protect themselves against the effect of the fire front and burning objects. Some shelters failed, and some people sought alternative places of shelter, having up to three shelter locations. A summary of these data showed that 82% of people survived in their first location of sheltering. Subsequently, of the 14% of people who moved to a second location, 86% survived. Finally, 2% of people moved to a third location, and all survived (Table 2). Most residents’ first place of shelter was inside a residential house (including 100 fatalities and 475 survivors). The most prevalent place of shelter for the second location was in open space (n = 65). Fourteen people died in open space after exiting a house, and other shelter location (Table 2).

Table 2 Number of location(s) of people sheltering categorised by shelter type. Shelter location/type

Number of fatalitiesa

Number of survivorsb

Total

First Shelter location Residential building Commercial building School Other structure (cool room, garage, shed, spa room, water tank) Bunker Fire service shedc Open airc Behind physical barrier Open air/space Vehicle Driving Stationary Water body Dam Lake Pool

150 100 4 0 7

700 475 119 9 14

850 575 123 9 21

2 0 25 18 7 12 0 12 0 0 0 0

3 7 29 19 10 26 7 19 18 7 1 10

5 7 54 37 17 38 7 31 18 7 1 10

Second Shelter location Residential building Commercial building Other structure Bunker Fire service shedc Open air/spacec Behind physical barrier Open air/space Vehicle (stationary) Water body Dam Pool

19 4 0 1 0 0 14 0 14 0 0 0 0

121 10 1 9 6 4 51 31 20 25 15 6 9

140 14 1 10 6 4 65 31 34 25 15 6 9

Third Shelter location Residential building Open air/spacec Vehicle (stationary) Total (shelter location)

0 0 0 0 169

17 11 3 3 838

17 11 3 3 1007

3.1.2. Shelter locations within a residential building Where people used a residential house for shelter, the movements between locations of these people within the house varied substantially. 37% of people sheltered in a room with poor visibility to the outside, mainly the bathroom and also rooms under the house (which includes enclosures under the house and cellars). Of the 57 people known to have sheltered in bathrooms, there were 37 fatalities (65% of total fatalities). Others moved around the house and monitored the conditions of the fire outside and potential ignitions inside the house (Table 3).

a

Including all known fatalities. Surveyed survivor’s location (s) (Note: some people sheltered at 1 or more location). c Indicative numbers from the sample. A large number of residents and fire personnel (in the order of tens or hundreds) sheltered at different fire station sheds or inside vehicles on the sports oval (McLennan, 2010). b

the passage of the fire front and then returned outside to defend the house and property. Occasionally, some members of the household (often younger or elderly persons) stayed in the house before, during and after the passage of the fire front. A number of structures around the main residence were used for shelter including sheds, fire bunkers, water tanks, cool rooms, and kennels. These structures were often used because they were the closest option for those who could no longer withstand exposure to the fire while defending. They were also used as a secondary shelter when the first place of shelter, usually the house, had failed. Interestingly, more than 120 people (12%) sheltered inside commercial buildings such as hotels, pubs, wineries and ski huts (Table 2). Some residents took refuge in open spaces close to their house such as in burnt paddocks, non-combustible driveways or olive groves. Others sheltered in locations further away such as sports ovals or school grounds, or sheltered in their car. Many people reported experiencing difficulties getting to a place of shelter due to poor visibility and burning obstacles. While precise information on the number of people who sheltered in fire service sheds was not available, Table 2 present figures based on the interview and witness statement data set. Data from elsewhere indicates that a large number of residents and fire personnel (in the order of tens or hundreds) sheltered at the fire service sheds in Kinglake, Kinglake West and Dixons Creek (McLennan, 2010).

3.2. Active versus inactive sheltering A broad distinction was drawn between active and inactive sheltering (see Whittaker et al., 2017). Active sheltering is characterised by regular monitoring of the fire and conditions inside and outside the place of shelter, as well as actions taken to protect occupants. Such actions include extinguishing fires; preventing the entry of smoke; and caring for children, the elderly and people who may have been injured. Inactive sheltering is characterised by a lack of regular monitoring of the bushfire and lack of actions to protect occupants or shelters. Evidence suggests that the majority of residents who sheltered engaged in monitoring and activity to protect occupants. However, there were households where one or more people sheltered inactively, often children and the elderly, while others defended or actively sheltered. Cases of inactive sheltering were more common among those who had bunkers, and those who found themselves sheltering unexpectedly, such as those who were forced to shelter outside while defending. There were only 22 cases out of 325 incidents where all members of the household sheltered inactively. Of those who sheltered, 31% of fatalities actively sheltered, and 58% of survivors were people who actively sheltered (Table 4). If we look at the roles of people, 90% of 89

Environmental Science and Policy 81 (2018) 86–94

R. Blanchi et al.

Table 3 Details of shelter locations inside houses. Shelter location within house

Number of fatalities

Table 5 Number of shelter failures while people were sheltering in them (forcing residents to exit or become trapped in the house). Number of survivors

Houses destroyed

Houses survived

Total number of people

First Shelter Bathroom Pantry/Storeroom under stairs Enclosure under house (Cellar, workshop) Cellar Bedroom Entrance Hallway Kitchen Kitchen and moving throughout Laundry Bathroom with external door Lounge Study Sun room Verandah Upstairs and downstairs workshop Back of house Move through Room away from fire direction/Lee side of house No information available

100 37 0

127 NIAb 7

348 20a NIA

575 57 7

6

1

3

10

0 9 9 0 20 0

NIA NIA NIA 3 NIA 1

5 NIA NIA 12 5 NIA

5 9 9 15 25 1

3 1

NIA NIA

NIA NIA

3 1

3 6 1 0 0

NIA 5 NIA 2 NIA

NIA NIA NIA 1 10

3 11 1 3 10

1 0 2

NIA 43 NIA

NIA NIA 2

1 43 4

2

65

310

357

Second Shelter Study No information available

4 4 0

5 NIA 5

5 NIA 5

14 4 10

Third Shelter No information available Total

0 0

1 1

10 10

11 11

104

133

363

600

Number of inactive people

Number of people with unknown behaviour

Total number of people sheltering

Fatality Survive Total

52 (31%) 482 (58%) 534 (53%)

71 (42%) 202 (24%) 273 (27%)

46 (27%) 154 (18%) 200 (20%)

169 (100%) 838 (100%) 1007 (100%)

Number of survivors

Total number of people

Bunker Commercial building Residential building Other structure Total

2 4 104a 8 118

0 11 133 16 160

2 15 242 18 276

One case of shelter under the house largely untouched but the house was destroyed.

more than 30 people sheltering in a commercial building. 3.3. Shelter vulnerability and structure failure Those who sheltered actively were able to detect ignitions on, in and around the house, which many then attempted to extinguish. In some cases, early detection meant that fires were extinguished relatively easily. However, there were cases where undetected fires took hold or where residents were unable to safely go outside to extinguish fires. While some people exited immediately, others waited until the fire outside subsided. A number of interviewees described moving progressively through the house and exiting at the last moment. Many were able to recognise the threat presented by the burning house and were able to exit safely. A few residents gave vivid accounts of their experiences and depicted the considerable risks taken while sheltering and exiting a burning house (see Whittaker et al., 2017). A large number of people experienced structural failure of their shelter. They faced major challenges and issues, including 118 persons dying inside structures that failed and 160 (out of 838) exiting a burning house and having to find a second, and sometimes a third, place of shelter (Table 5). Although the numbers of examples are low in some categories, the data infer that the survival rate involving burning shelters is poorest for residential buildings and best for other structures, including commercial buildings. The higher survivability of occupants in commercial buildings is not a surprise given that they are subject to building controls that facilitate egress in the event of a fire. Some residents identified factors that made structures unsafe to shelter in. For example, combustible elements around the house, such as cars, created additional exposure to the house and increased the potential number of ignitions to address. If ignitions are not suppressed quickly, they can develop to a point where they cannot be suppressed by the occupant. The duration that residents sheltered varied substantially. Some mentioned that they sheltered from a few minutes to up to an hour during the passage of the fire front. In other cases, consequential fire (fires in adjacent structures or heavy fuels) following passage of the fire front occurred over many hours and extended the time that shelter was needed. This could be attributed to the severity of the bushfires and weather conditions on the day, as well as building and landscaping design. Investigation of the relationships between shelter location and fire severity (Fig. 1) revealed that the highest number of fatalities in destroyed houses (n = 38) were in areas that experienced the highest fire severity (75% crown consumption, equivalent to an active crown fire). However, in this category, some people managed to survive the fire even when the shelter was destroyed (n = 30). In observing the distance to forest with respect to the cumulative percent of people, over 90% of locations were within 100 m of forest (Fig. 2). The data show that 59% of fatalities were in sheltering locations within 5 m of forest and 90% within 35 m. Conversely, 50% of those who survived were in sheltering locations within 15 m of forest and 90% at 90 m. From this, we can state that a greater proportion of fatalities occurred closer to forest.

Table 4 The number of people with active, inactive and unknown behaviour categorised by their survival (frequency and percentage of total for each row). Number of active people

Number of fatalities

a

a Most cases involve children sheltering in the bathroom, sometimes unsupervised or partially supervised (see Whittaker et al., 2017). b No Information Available.

Survival

Shelter type

those who actively sheltered survived while 74% of those who were inactive while sheltering survived. However, many inactive people had someone in their household or immediate vicinity defending or actively sheltering and were, therefore, able to protect them or alert them to hazards and the need to take action. There were examples of group sheltering, including many people and pets, in residential and commercial buildings. Most of the incidents involved one person (n = 90) or two persons (n = 107). However, of the limited data, 28 incidents involved six persons or more, with a maximum of 26 people sheltering inside a residential building, and 90

Environmental Science and Policy 81 (2018) 86–94

R. Blanchi et al.

Fig. 1. Comparison of survival by shelter damage and burn severity for the 2009 Kilmore East bushfire.

3.4. Statistical analyses of sheltering variables influencing survival

4. Discussion

A summary of the sheltering variables found to have significant (Pr ( > |z|) < 0.1) effects on the probability of survival are presented in Table 6. The number of people actively sheltering and the total number of people sheltering both had highly-significant positive effects on survival (Table 6). Survival decreased when the number of shelter locations increased. Further, survival reduced with increasing fire severity (only available for the Kilmore East fire), distance to the forest and use of a non-residential building (Table 6). The most reliable model for positive survival included the number of active people sheltering, the number of people sheltering, sheltering outside of a forest in a shelter location with little to no damage. These variables combined have an area under receiver operating (AUROC) curve of 92%. Due to the limited sample of survivors and complete sample of fatalities these analyses should be used with caution.

4.1. Types of shelter and actions when sheltering People sought shelter in a range of locations and shelter types, mainly inside houses and commercial buildings such as hotels, pubs, wineries and ski huts (Table 2). A large number of residents and fire personnel sheltered in and around fire service sheds. Occasionally people sheltered in open spaces such as sports ovals or paddocks. Alternatively, they found protection in water bodies such as pools, dams or lakes. When people are exposed to a bushfire, they seek refuge and protection from the radiant heat and smoke, which are the main killers. In some cases, successive shelters were used to find protection. These refuges may take different forms according to specific geographic constraints, needs and protection available. While the domestic house is the most common option, other choices can be made depending upon the situation at the time. For those sheltering in a house the location of people is critical for their safety (Table 3). Some people sheltered in a room with poor

Fig. 2. Cumulative percent of people with respect to their location to forest, in metres, and survival.

91

Environmental Science and Policy 81 (2018) 86–94

R. Blanchi et al.

Table 6 Summary of significant variables (Pr(> |z|) < 0.1) from the analyses of individual variables on the probability of success of survival from sheltering. SE standard error of the mean; DF degrees of freedom. Variable

Coefficient estimate

SE

z value

Pr(> |z|)

AIC

Fisher

Residual deviance

Null DF

Residual DF

Number of people actively sheltering Number of people sheltering Number of shelters used Vegetation effect (Fire severity for Kilmore East only) Within forest (0 m from forest) Distance to forest Shelter at residential Building

0.80 0.47 -0.64 0.18 -0.68 0.01 0.75

0.141 0.111 0.182 0.053 0.225 0.004 0.445

5.66 4.25 -3.53 3.38 -3.02 2.42 1.68

< 0.0001 < 0.0001 0.0004 0.0007 0.0026 0.0155 0.0937

461 478 492 492 495 495 494

5 6 4 4 4 5 4

457 474 488 488 491 491 488

493 493 493 493 493 493 493

492 492 492 492 492 492 491

beyond the capability of the occupants to suppress them (section 3.3). In these cases, the occupant needs to recognise the inevitability of leaving a house that has ignited past the point of suppression. The rate at which the house loses tenability from this point is a function of house design, whether doors and windows in the house are open or closed and the nature of the developed fire. The fire could develop at a rate that is faster than what is anticipated by the occupants. Occupants need to locate themselves at the most viable exit locations in anticipation of this occurring. The most viable exit will depend on the location of the developing house fire(s), the presence of combustible elements attached to or near the exit and the location(s) outside to provide them with the highest likelihood of survival. Several studies have indicated that egress is made difficult or impossible due to combustible elements such as vegetation, cars, garden furniture and woodpiles, or elements that lost structural integrity such as decks and stairs (Blanchi and Leonard, 2008; Cohen, 2000; Davis, 1990; Gill and Stephens, 2009). These difficulties support the findings of Blanchi et al. (2012), where 40% of fatalities were found to have occurred in close proximity to the home (< 20 m). Distance to forest Fig. 1 Comparison of survival by shelter damage and burn severity for the 2009 Kilmore East bushfire(Fig. 1) has been used in other studies to predict house loss, and while being a simple metric, it provides an indication of the potential impact of the fire on the structure (Ahern and Chladil, 1999; Chen and McAneney, 2010). In their studies of fire behaviour measures and community loss, Harris et al. (2012) and Kilinc et al. (2013) found that an estimate of the energy released from the fire front using Byram’s fire line intensity (Byram, 1959) was the best predictor of house loss. As fire line intensity was not available for this study, burn severity was used (where available) as a surrogate for fire severity (Fig. 2).

visibility to outside conditions, such as a bathroom. This is consistent with past research, which highlights that many people believe bathrooms to be the safest room in which to shelter (Blanchi et al., 2012; Krusel and Petris, 1999; McArthur and Cheney, 1967). The perceived safety of the bathroom appears to stem from the ready availability of water and the contained nature of the room. It is noteworthy that children and pets were often confined to bathrooms, with and without the presence of adults (Whittaker et al., 2017). This is consistent with previous studies on fatalities in Australia (Blanchi et al., 2012) where 41% of the fatalities with known locations occurred in a room with reduced visibility to outside conditions (bathroom, enclosure, laundry, study, toilet block, and bunker). Results of this study suggest that sheltering should occur with active monitoring and extinction of fire ignitions (Table 4). This differs from the practice of ‘shelter in place’ (SIP) (Cova et al., 2009; Paveglio et al., 2010) where people passively shelter while the fire passes. The statistical analyses confirmed the importance of people actively sheltering (Whittaker et al., 2017) and the number of people sheltering on survival. A large number of people experienced failure of their shelter (Table 5). This raised the question of the adequacy of the shelter to withstand specific weather conditions and severe bushfire exposure. High winds during a bushfire event have the potential to cause either superficial or major damage to buildings. Leonard et al. (2009) reported that 13% of houses surveyed after Black Saturday bushfires were affected by wind (such as missing roof tiles, lifted roof sheeting or missing sheeting from eaves).This can occur before, during or after the passage of the fire. In each case, even superficial wind damage can render a building more vulnerable to the other effects of the bushfire (e.g. ember attack). The key processes are damage to the building façades, breakage of windows and lifting of roof materials. Winds can also provide an air pressure difference that can force flames through small gaps in the building envelope, increase the rate and extent to which building elements lose moisture and increase the intensity and distribution of embers and combustible debris that reaches or enters the building (Blanchi et al., 2011). In addition, winds make survival outside of the protection of a shelter difficult. An able-bodied person may have some difficulty walking in winds greater than 54 km/h (15 m/s) (Blanchi et al., 2011). High-velocity debris also creates an added hazard to a person. Given that 60% of our sample used a residential house as shelter, house safety presents as an important factor to consider in bushfire safety policy and planning. House vulnerability is one of the main challenges to consider because houses are complex. They have many occupiable and non-occupiable compartments. Unless specifically designed, these compartments are made of or contain combustible elements. An occupant is faced with the task of monitoring all of these compartments and having the resources to address any fire starting within them (some house compartments, for example in roof or underfloor, are more difficult or impossible to monitor). The building design and material would govern the number of multiple ignitions possible (Cohen, 2000; Leonard et al., 2009; Ramsay et al., 2003). Another key challenge is the ability to recognise ignition in and around the house and the circumstances where these ignitions develop

4.2. Policy implications For those actively sheltering in a house, fatalities occurred either because the occupants were overwhelmed by the rate at which the house lost tenability and/or they failed to recognise the developing house fire because it was in a building cavity or un-monitored room. Building regulations could play a prominent role in specifying provisions that reduce the rate of tenability loss and reduce the prevalence of combustible cavities where fires can develop without the occupant’s knowledge. These points have been previously highlighted as potential deficiency is some bushfire related building standards (Leonard, 2009) and currently, AS3959 (Standards Australia, 2009) does not identify either of these factors in its considerations while the National Association of Steel-framed Housing (NASH) standard1 for bushfire construction does. Policies that encourage occupants to leave are likely to reduce the number of people faced with these circumstances; however, it is unlikely that such policies will result in complete evacuation due to the broad range of factors that influence peoples’ decisions to stay or go 1 http://www.nash.asn.au/nash/publications/nash-standards 2015).

92

(accessed

October

Environmental Science and Policy 81 (2018) 86–94

R. Blanchi et al.

Acknowledgements

(Cova et al., 2009; McCaffrey et al., 2015; McLennan et al., 2013). Indeed, Australian bushfire research highlights the tendency for people to wait until they are directly threatened before taking action to evacuate (e.g. McLennan et al., 2015; McNeill et al., 2015; Whittaker et al., 2013). The inevitability that some people will delay evacuating and find themselves unable to safely leave was recognised by the VBRC and prompted its recommendation to expand the range of options available to people, including provision of community refuges and bushfire shelters (Teague et al., 2010). Given that over 70% of deaths and 60% of houses lost in bushfires occurred on days exceeding Forest Fire Danger Index 100 (Blanchi et al., 2014), the question now remains whether building and planning regulations should more specifically target the circumstances leading to life and house loss in these events. This can be specifically addressed in several ways. First, by reducing the fragility of houses in extreme fire weather events with particular attention to wind effects. Second, by improving the robustness principles on building design codes. Third, by incorporating building design principles that support egress from the building. Fourth, by managing heavy fuels in landscaping design to improve egress provisions in house and urban design. Last, by informing people on the limitations of building regulations and design to provide confidence in the reliability of the building as a temporary shelter. Greater uptake of information about how houses and other structures are impacted by bushfires is needed to help residents make more informed decisions about sheltering. Hence it may be beneficial for building regulations to provide prescriptions that address building performance and perceptions of building performance. Some of the occupants’ accounts identified rapid loss of house tenability at a rate that could challenge a well-informed and vigilant occupant. This suggests that improved building prescriptions that consider egress could reduce future losses of life during a bushfire. Overall, greater consideration of alternatives to evacuation is needed to increase community safety during bushfires, as noted by Paveglio et al. (2015) and McCaffrey et al. (2015). Australian fire services now emphasise the need for people to have a backup plan in the event that they are unable to safely stay and defend or evacuate. In Victoria the Country Fire Authority (CFA) developed guidelines for the installation and use of private bushfire shelters. The use of community shelters is also promoted as an option of last resort in many States in Australia.

This project is funded by the Australian Government under the Natural Disaster Resilience Grants Scheme (NDRGS). In the State of Victoria, the NDRGS is facilitated by Emergency Management Victoria (EMV). We gratefully acknowledge all the people who contributed to this research, especially the residents impacted on Black Saturday who gave their time to share their knowledge and experience. We also thank the Country Fire Authority, and in particular Mark Holland, for their valuable assistance and support. We thank Stephanie Dreyfuss for her assistance in compiling literature related to sheltering in Australia. We acknowledge the support of the Australasian Fire and Emergency Service Authorities Council and the Bushfire and Natural Hazards CRC, which provided access to the data collected by the Bushfire CRC’s 2009 Victorian Fires Research Taskforce. References AFAC, 2005. Position Paper on Bushfires and Community Safety. AFAC Limited, Victoria, Australia (http://38c69b050a3d5d1eb1e3-aa923a4231e15c57e2802c896554e8a6.r6.cf4.rackcdn.com/D/D565388D-165C-412B-819CF367B218C2D7.pdf (Accessed October 2015). Ahern, A., Chladil, M., 1999. How far do bushfires penetrate urban areas? Proc. Aust. Disaster Conf. 1999, Disaster Prev. 21st Century. Blanchi, R., Leonard, J., 2008. Property safety: judging structural safety. In: Handmer, J., Haynes, K. (Eds.), Community Bushfire Safety, pp. 77–85. Blanchi, R., Leonard, J., Haynes, K., Opie, K., James, M., Kilinc, M., Dimer de Oliveira, F., Van den Hornet, R., 2012. Life and house loss database description and analysis – Final report. CSIRO, Bushfire CRC report to the Attorney-General’s Department, CSIRO EP129645, Melbourne. Blanchi, R., Leonard, J., Haynes, K., Opie, K., James, M., Oliveira, F.D., 2014. Environmental circumstances surrounding bushfire fatalities in Australia 1901-2011. Environ. Sci. Policy 37, 192–203. Blanchi, R., Whittaker, J., Haynes, K., Leonard, J., Opie, K., Holland, M., Dreyfuss, S., 2015. Sheltering practices during bushfire. Melbourne, CSIRO report EP 158583. Natural Disaster Resilience Grants Scheme. Byram, G.M., 1959. Combustion of forest fuels. In: Davis, K.P., Byrma, G.M., Krumm, W.R. (Eds.), Forest fire – Control and use. McGraw-Hill Book Co., New York. Chen, K.P., McAneney, J., 2010. Bushfire penetration into urban areas in Australia: a spatial analysis [WWW Document]. URL http://www.royalcommission.vic.gov.au/ Documents/Document-files/Exhibits/CRC-304-001-0001.pdf. Cohen, J.D., 2000. Preventing disaster. J. For. 98, 15. Cova, T.J., Drews, F.A., Siebeneck, L.K., Musters, A., 2009. Protective Actions in Wildfires: Evacuate or Shelter-in-Place? Nat. Hazards Rev. 10, 151–162. http://dx. doi.org/10.1061/(ASCE)1527-6988(2009)10:4(151). Cruz, M.G., Sullivan, A.L., Gould, J.S., Sims, N.C., Bannister, A.J., Hollis, J., Hurley, J.J., 2012. Anatomy of a catastrophic wildfire: The Black Saturday Kilmore East fire in Victoria, Australia. For. Ecol. Manage. 284, 269–285. http://dx.doi.org/10.1016/j. foreco.2012.02.035. Davis, J.B., 1990. The Wildland-Urban Interface – Paradise or Battleground. J. For. 88, 26–31. Furby, S., Caccetta, P., Wallace, J., Lehmann, A., Zdunic, K., 2009. Recent Development in Vegetation Monitoring Products From Australia’s National Carbon Accounting System. Internaitonal Geosci. Remote Sens. Symp. 276–279. Gill, A.M., Stephens, S.L., 2009. Scientific and social challenges for the management of fire-prone wildland–urban interfaces. Environ. Res. Lett. 4, 34014. Gill, M.A., 2005. Landscape fires as social disasters: an overview of ‘the bushfire problem’. Glob. Environ. Change B Environ. Hazards 6, 65–80. Handmer, J., Tibbets, A., Tibbits, A., 2005. Is staying at home the safest option during bushfires? Historical evidence for an Australian approach. Environ. Hazards 6, 81. http://dx.doi.org/10.1016/j.hazards.2005.10.006. Harris, S., Anderson, W., Kilinc, M., Fogarty, L., 2012. The relationship between fire behaviour measures and community loss: an exploratory analysis for developing a bushfire severity scale. Nat. Hazards 63, 391–415. http://dx.doi.org/10.1007/ s11069-012-0156-y. Haynes, K., Coates, L., Leigh, R., Handmer, J., Whittaker, J., Gissing, A., Mcaneney, J., Opper, S., 2009. Shelter-in-place vs. evacuation in flash floods. Environ. Hazards 8. http://dx.doi.org/10.3763/ehaz.2009.0022. Haynes, K., Handmer, J., McAneney, J., Tibbits, A., Coates, L., 2010. Australian bushfire fatalities 1900–2008: exploring trends in relation to the Prepare, stay and defend or leave early policy. Environ. Sci. Policy 13, 185–194. Kilinc, M., Anderson, W., Anderson, D., Price, B., 2013. On the need for a bushfire scale that represents the bushfire hazard. Monash University bushfire CRC report to DSE, Melbourne. Krusel, N., Petris, S.N., 1999. A Study Of Civilian Deaths In the 1983 ash wednesday bushfire victoria, CFA Occasional Paper No 1. Contry Fire Authority, Mt Waverley. Lampin-Maillet, C., Jappiot, M., Long, M., Bouillon, C., Morge, D., Ferrier, J.P., 2010. Mapping wildland-urban interfaces at large scales integrating housing density and vegetation aggregation for fire prevention in the South of France. J. Environ. Manage. 91, 732–741. http://dx.doi.org/10.1016/j.jenvman.2009.10.001.

5. Conclusion This research highlights that safe sheltering requires considerable planning and preparation by residents. Regardless of whether they intend to stay and defend or leave, residents should identify multiple places where they can go if they need to take shelter during a bushfire. It is important that residents shelter actively by continually monitoring conditions inside and outside the shelter, and by taking action to protect the shelter occupants. Critically, sheltering should always be planned in conjunction with property defence and as a contingency in the event that evacuation or defence is not possible. There is a risk that greater provision of sheltering information and shelters may give people a false sense of security, encouraging them to wait until the last moment before taking action. This could increase the risk that people are caught sheltering with minimal or no preparation, or that they undertake late and dangerous evacuations. Nevertheless, there is a need for public information to encourage planning and preparation for sheltering and safe sheltering behaviours. The findings of this research indicate that safe sheltering requires appropriately designed and located shelters that are occupied by residents who are prepared and shelter actively. Information should emphasise that sheltering is not an alternative to leaving early or defending, but may be necessary as part of defence or when safe evacuation is not possible. 93

Environmental Science and Policy 81 (2018) 86–94

R. Blanchi et al.

677–688. http://dx.doi.org/10.1071/WF09034. Penman, T., Eriksen, C., Blanchi, R., Chladil, M., Gill, A., Haynes, K., Leonard, J., McLennan, J., Bradstock, R.A., 2013. Defining adequate means of residents to prepare property for protection from wildfire. Int. J. Disaster Risk Reduct. 6, 67–77. Ramsay, C., Rudolph, S., CSIRO, 2003. Landscape and building design for bushfire areas. CSIRO Publishing, Melbourne. Rubin, H.J., Rubin, I.S., 2005. Qualitative interviewing: the art of hearing data, Second edition. Sage Publications, Thousand Oaks. Scanlon, J., 1992. Evacuations and Shelter. Disaster Preparedness: Some Myths and Misconceptions. Easingwold, UK. Standards Australia, 2009. AS 3959-2009 (Amend. 1) Construction of buildings in bushfire prone areas. Standards Australia. Stephens, S.L., Adams, M.A., Handmer, J., Kearns, F.R., Leicester, B., Leonard, J., Moritz, M.A., 2009. Urban-wildland fires: how California and other regions of the US can learn from Australia. Environ. Res. Lett. 4. Stretton, L., 1939. Report of the Royal Commission to inquire into the causes and measures taken to prevent the bush fires of January, 1939 and to protect life and property and the measures to be taken to prevent bush fires in Victoria and to protect life and property in the event of future bush fires. Teague, B., McLeod, R., Pascoe, S., 2010. 2009 Victorian Bushfires Royal Commission final report: summary. State Government of Victoria, Melbourne. Theobald, D.M., Romme, W.H., 2007. Expansion of the US wildland-urban interface. Landsc. Urban Plan. 83, 340–354. http://dx.doi.org/10.1016/j.landurbplan.2007.06. 002. Tibbits, A., Handmer, J., Haynes, K., Lowe, T., Whittaker, J., 2008. Prepare, stay and defend or leave early: evidence for the Australian approach. In: Handmer, J., Haynes, K. (Eds.), Community Bushfire Safety. CSIRO publishing, Melbourne, pp. 59–71. Tibbits, A., Whittaker, J., 2007. Stay and defend or leave early: Policy problems and experiences during the 2003 Victorian bushfires. Environ. Hazards 7, 283–290. Whittaker, J., Blanchi, R., Haynes, K., Leonard, J., Opie, K., 2017. Experiences of sheltering during the Black Saturday bushfires: Implications for policy and research. Int. J. Disaster Risk Reduct. 23, 119–127. http://dx.doi.org/10.1016/j.ijdrr.2017.05.002. Whittaker, J., Haynes, K., Handmer, J., McLennan, J., 2013. Community safety during the 2009 Australian Black Saturday bushfires: an analysis of household preparedness and response. Int. J. Wildl. Fire 22, 841–849. http://dx.doi.org/10.1071/WF12010. Whittaker, J., McLennan, J., Elliot, G., Gilbert, J., Handmer, J., Haynes, K., Cowlishaw, S., 2009. Victorian 2009 Bushfire Research Response: final report. Bushfire CRC, Melbourne. Yard, C.R., 2000. Evacuation vs. shelter-in-place: a situational comparison. Disaster Recover. J. 13, 28–31.

Leonard, J., 2009. Report to the 2009 Victorian Bushfires Royal Commission. Building performance in bushfires. Leonard, J., Blanchi, R., Leicester, R., Lipkin, F., Newnham, G., Siggins, A., Opie, K., Culvenor, B., Cechet, B., Corby, N., Thomas, C., Habili, N., Jakab, M., Coghlan, R., Lorenzin, G., Campbell, D., Barwick, M., 2009. Building and Land use planning research after the 7th February 2009 Victorian bushfires. Preliminary findings. Interim report USP2008/018 – CAF122-2-12, Melbourne. Lymburner, L., Tan, P., Mueller, N., Thackway, R., Lewis, A., Thankappan, M., Randall, L., Islam, A., Senarath, U., 2011. The National Dynamic Land Cover Dataset. McArthur, A.G., Cheney, N.P., 1967. Report on the southern Tasmanian bushfires of 7 February 1967. Forestry Commission Tasmania, Hobart. McCaffrey, S., Rhodes, A., Stidham, M., 2015. Wildfire evacuation and its alternatives: perspectives from four United States’ communities. Int. J. Wildl. Fire 24. http://dx. doi.org/10.1071/WF13050. McCaffrey, S.M., Rhodes, A., 2009. Public response to wildfire: Is the Australian stay and defend or leave early approach an option for wildfire management in the United States? J. For. 107, 9–15. McLennan, B.J., 2009. Use of informal places of shelter and last resort on 7 february 2009. Bushfire CRC. McLennan, J., 2010. Use of informal places of shelter and last resort on 7 February 2009. Melbourne. McLennan, J., Elliott, G., Omodei, M., 2012. Householder decision-making under imminent wildfire threat: stay and defend or leave? Int. J. Wildl. Fire 21, 915–925. http://dx.doi.org/10.1071/WF11061. McLennan, J., Elliott, G., Omodei, M., Whittaker, J., 2013. Householders’ safety-related decisions, plans, actions and outcomes during the 7 February 2009 Victorian (Australia) wildfires. Fire Saf. J. 61, 175–184. McLennan, J., Paton, D., Wright, L., 2015. At-risk householders’ responses to potential and actual bushfire threat: an analysis of findings from seven Australian post-bushfire interview studies 2009–2014. Int. J. Disaster Risk Reduct. 12. http://dx.doi.org/10. 1016/j.ijdrr.2015.02.007. McNeill, I.M., Dunlop, P.D., Skinner, T.C., Morrison, D.L., 2015. Predicting delay in residents’ decisions on defending v. evacuating through antecedents of decision avoidance. Int. J. Wildl. Fire 24, 153–161. Nadimpalli, K., Edwards, M., Mullaly, D., 2007. National Exposure Information System (NEXIS) For Australia: Risk assessment opportunities. pp. 1674–1680 MODSIM07. Paveglio, T., Carroll, M.S., Jakes, P.J., 2008. Alternatives to evacuation – protecting public safety during wildland fire. J. For. Paveglio, T.B., Carroll, M.S., Jakes, P.J., 2010. Adoption and perceptions of shelter-inplace in Californias Rancho Santa Fe Fire Protection District. Int. J. Wildl. Fire 19,

94