Archetypes and change in wildfire risk perceptions, behaviours and intentions among adults in Tasmania, Australia

Introduction

Globally elevated wildfire risk to human communities is driven by numerous interacting factors. These include an increasing number of days of extreme heat, extended drought, and longer fire seasons due to a warming climate (Bowman et al. 2020; Intergovernmental Panel on Climate Change 2022), and an increasing expansion of the wildland–urban interface (WUI) regions exacerbating the risk to communities located in these zones (McLennan et al. 2019; Koksal et al. 2020). This risk is driven in part by the past success of suppression of smaller fires in these regions, leading to excessive fuel loads which feed larger fires that defy suppression – a concept described as the ‘wildfire paradox’ (Ingalsbee 2017). In this context, understanding community perceptions of wildfire risk, and how these may influence response options and actions during a wildfire event, are vital to policymakers when planning community education and engagement programs aiming to build resilience and reduce loss of life in a wildfire emergency.

Unlike other fire-prone regions globally, Australia does not have mandatory, government-organised evacuation and relocation policies related to wildfire threat (McLennan et al. 2019), making self-evacuation a widely understood and accepted response. Nationwide community advice is provided through the Australian Fire Danger Rating System (AFDRS), which outlines four levels of wildfire risk rating and an associated set of recommended actions (National Emergency Management Agency 2024). Furthermore, Australian government agencies consider natural disaster mitigation and adaptation actions a ‘shared responsibility’ between individual landowners and authorities (Commonwealth of Australia 2011, 2018), however recent studies highlight the significant challenges associated with this narrative in relation to wildfires (Cooper et al. 2020; McDonald and McCormack 2022). Communities in fire-prone regions of Australia therefore ideally need to have an informed understanding of their risks and options, in order to facilitate safe decision-making under the threat of a wildfire event, especially given the heightened risk of more extreme fire behaviour associated with climate change. In parallel, policymakers need to understand how local community members understand, evaluate and respond to their risks, and how these perceptions have changed over time, in order to provide targeted, timely and relevant information and community messaging.

The current understanding of community perceptions, behaviours, preparedness and response actions associated with wildfire risk is mixed, complex and incomplete, with much still to learn (Ahmed and Ledger 2023). For example, a review of 10 studies by Koksal et al. (2019) found only a small overall correlation between risk perception and risk mitigation actions related to wildfire, with a wide variation between individuals. Howe et al. (2024) recently identified that an individual’s world view more accurately predicts their support for wildfire risk mitigation actions than wildfire experience, while Haghani et al. (2022) identified a major knowledge gap in understanding human behaviour related to wildfires, especially in risk perception.

One method of achieving a more nuanced and comprehensive understanding of community characteristics and response actions related to a wildfire threat is by defining archetypes from the studied population. Originally conceived by Carl Jung, archetypes embody a ‘typical character’ which is universally understood in a social and cultural context (Jung 2014). Examples of commonly understood literary archetypes include the ‘villain’, the ‘victim’ and the ‘scapegoat’. Archetypes are routinely used in market research to summarise demographic categories (Dominici et al. 2016), and this approach has previously been used in studies of human behaviour associated with wildfire both in Australia (Strahan et al. 2018) and the United States (Paveglio et al. 2014) to help understand preparedness levels, decision-making and leave intentions for a community. An understanding of local archetypes may assist in developing policy, communication, and messaging for more targeted intervention approaches.

Also required is understanding how community-wide attitudes to wildfire threat are changing. We would expect that there is heightened awareness of wildfire threats given the combined effect of targeted education campaigns and wide media reporting of global wildfire disasters. To our knowledge, however, there are no published longitudinal studies of changes to community-wide attitudes to wildfire threat and response behaviours.

The aim of this study therefore was to gain an increased understanding of wildfire self-evacuation intentions over time, alongside current perceptions, understanding, and behaviours associated with wildfire risk and threat for residents in the highly wildfire-prone jurisdiction of Tasmania, Australia.

Setting

Tasmania is an island state situated off the south coast of Australia, with an estimated population of 571,500 in 2022 (Australian Bureau of Statistics 2022). There are 29 local government areas (LGAs), with the major population centres of Hobart in the south, Launceston in the north and Burnie-Devonport in the northwest comprising the urban regions (see Fig. 1). Tasmania is largely classified as regional and remote (Australian Bureau of Statistics 2011).

Fig. 1.

Map of Tasmania, Australia, showing urban/regional and rural/remote local government areas and major population centres.

With substantial areas of bushland extending across the state, approximately 98% of Tasmania’s land area is classed as fire-prone (Tasmania Fire Service 2022a). Fire risk is distributed across both urban and regional/remote areas (Campbell et al. 2024). Wildfire is considered to have the highest risk profile of all natural hazards in Tasmania (White et al. 2016), with recent major uncontrolled fires occurring in 2013 (Dunalley-Forcett, Lake Repulse, Giblin River and Bicheno fires), 2016 (North West fires) and 2019 (Gell River, Riveaux Road and Miena fires) (Bowman et al. 2022). Furthermore, and in line with other increasing fire risks globally, Tasmania’s wildfire risk is projected to increase in a warming climate, with likely increases in air temperature and soil dryness, longer warm seasons and earlier season occurrences of dry lightning ignitions, all contributing to greater frequency and intensity of wildfire events (Fox-Hughes et al. 2014; Love et al. 2017; Remenyi et al. 2022).

The 2016 Tasmanian Population Health Survey (TPHS) found that for those living in a location at risk of wildfire, fewer than a third (31.5%) planned to leave immediately when a wildfire was burning near their home (Department of Health and Human Services Tasmania 2016; Campbell et al. 2024), despite this being the primary protective advice from government authorities in a wildfire event (Tasmania Fire Service 2022b). Furthermore, current understanding of wildfire risk perception in Tasmania is relatively limited, with Lucas et al. (2022) finding that residents in the highly fire-prone city of Hobart had concerns appropriate to biophysical risks, but these did not necessarily translate into preparedness actions.

Methods

Results

Cluster analysis

When asked if their home was in an area at risk of a wildfire in the 2023 survey (see Box 1, Question A), 41.5% (n = 626) of participants responded ‘Yes’ and 57% (n = 860) of participants responded ‘No’. Twenty-two respondents (1.5%) responded ‘Unsure’. A cluster analysis was performed for those that identified as being at risk (i.e. responded ‘Yes’ to Question A). Cluster analysis revealed four distinct clusters or groups (see Fig. 2). Based on survey responses, each group demonstrated clear patterns and reasons for self-evacuation intention, enabling the development and characterisation of four archetypes.

Fig. 2.

Dendrogram showing node height, calculated by agglomerative clustering with Ward’s distance. Leaf nodes coloured by leave intention, where red = stay, green = leave.

Based on the top five variables by group (see Supplementary Material S2, Supplementary Table S1), and using Fig. 3 below, each group can be characterised to an archetype as follows:

• Group 1 (red; n = 167) – the ‘immediate leavers’ strongly tended towards (i) leaving in response to a threat and (ii) being uninfluenced by official warnings or visual cues. Group 1 tended to be female, reside in urban areas and own their own home.

• Group 2 (yellow; n = 116) – the ‘prepared stayers’ were (i) ready to stay in response to a threat (ii) prepared for wildfire threat, and (iii) not influenced by official warnings or visual cues. Group 2 tended to be male, rural, older, have less income, own their home, and have greater bushfire experience than other groups.

• Group 3 (green; n = 83) – the ‘unprepared leavers’ tended to (i) leave at the threat of a wildfire, (ii) be unprepared, and (iii) have less comprehension of warnings. Group 3 tended to be more urban, younger, have greater income, rent their property, and have less bushfire experience than other groups.

• Group 4 (purple; n = 260) – the ‘informed leavers’ tended to (i) leave at the threat of a wildfire, and (ii) be influenced by warnings and advice. Group 4 tended to be female and reside in urban areas.

Fig. 3.

Proportions of various sociodemographic characteristics for each cluster: (a) gender; (b) location; (c) age category; (d) income category; (e) home ownership status; (f) previous wildfire experience; (g) leave intention; and (h) preparedness actions.

Fig. 3 shows proportions of various sociodemographic characteristics for each group, while Fig. 4 shows how each group compares to other groups across various categories (above the line) and sociodemographic variables (below the line). Significant differences in gender occur between Group 2 and Groups 1, 3 and 4; in home ownership between Group 3 and Groups 1, 2 and 4; and in resident location between Group 2 and Groups 3 and 4 (see Table 2).

Fig. 4.

Spectrum of responses by archetype across the categories of informed/uninformed, prepared/unprepared and leave/stay above the line, and sociodemographic variables of female/male; urban/rural; younger/older; lower/high income; own/rent home, and no/previous wildfire experience below the line, where 1 = immediate leavers; 2 = prepared stayers; 3 = unprepared leavers; and 4 = informed leavers. Colour codes follow classification of archetypes in Fig. 2.

Table 2.Sociodemographic variables by group, showing analysis type and significance level.

Variable	Group 1	Group 2	Group 3	Group 4	Analysis type and significance
Gender A	(a) 40% Male	(b) 65% Male	(a) 39% Male	(a) 42% Male	ANOVA P < 0.01, df = 3
Income	75% > $50,000	74% > $50,000	90% > $50,000	80% > $50,000	ANOVA P < 0.05, df = 3
Home ownership A	(a) 87% Home owner	(a) 92% Home owner	(b) 64% Home owner	(a) 84% Home owner	ANOVA P < 0.01, df = 3
Urban/Rural B	(ab) 41% Rural	(a) 48% Rural	(b) 25% Rural	(b) 33% Rural	ANOVA P < 0.01, df = 3
Previous bushfire experience A	(b) 57% bushfire experience	(a) 73% bushfire experience	(c) 37% bushfire experience	(b) 55% bushfire experience	ANOVA P < 0.01, df = 3
Modal age category	65+	65+	35–44	65+	Significantly different age distribution (chi squared P < 0.01, df = 15)

^A Significant difference between groups (a) and (b), or (a), (b) and (c).

^B Significant difference between groups (a) and (b). No significant difference between groups (ab) and (a), or (ab) and (b).

Regression analysis

Self-evacuation intention

Regression analysis for self-evacuation intention in 2023 demonstrated that women were significantly more likely to leave (RR = 1.25, 95% CI (1.17–1.33)), as were those living in an urban area (RR = 1.09, 95% CI (1.03–1.15)). Those living in a household with an income above $50,000 were also more likely to leave, although this fell marginally outside statistical significance (RR = 1.06, 95% CI (1.00–1.13)) (see Fig. 5 and Supplementary Material S10, Supplementary Table S2). Age, wildfire risk at home, previous wildfire experience and home ownership status were not clearly associated with a leave intention.

Fig. 5.

Risk ratios and 95% confidence intervals for the association between leave intention and sociodemographic outcomes for all participants from a representative sample of 1270 people in Tasmania, Australia (2023).

Preparedness actions

Regression analysis demonstrated that undertaking preparedness actions was significantly associated with previous direct experience with a wildfire (RR = 1.18, 95% CI (1.08–1.28)) as was owning a home as opposed to renting (RR = 1.49, 95% CI (1.22–1.84)). Living in an urban area was significantly and negatively associated with preparedness actions (RR = 0.92, 95% CI (0.86–0.98)). Gender, age and income were not associated with undertaking wildfire preparedness actions (see Fig. 6 and Supplementary Material S11, Supplementary Table S3).

Fig. 6.

Risk ratios and 95% confidence intervals for the association between undertaking preparedness actions and sociodemographic outcomes for all participants from a representative sample of 557 people in Tasmania, Australia (2023).

Changes in wildfire risk and self-evacuation intention (2016 and 2023)

In the 2023 survey, a greater proportion of participants (42%) identified as living in an area at risk of a wildfire than in the 2016 TPHS (29%), with a statistically significant association between year and wildfire risk at home (χ²(1) = 92.4, P < 0.001) (see Supplementary Material S12, Supplementary Table S4 for full results). Fig. 7 shows a comparison between responses from the 2016 TPHS and this 2023 survey.

Fig. 7.

Self-reported home wildfire risk,^A Tasmania, Australia (2016 and 2023).

^AHome wildfire risk refers to living within 100 m of bushland larger than about 1 ha (100 m × 100 m).

In both 2016 and 2023 surveys, if a participant responded ‘Yes’ to being in an area at risk from wildfire, they were subsequently asked about their leave intentions during a wildfire (see Box 1, Question B). A greater proportion of respondents intended to leave in 2023 than in 2016, with a statistically significant association between year and leave intention (χ²(1) = 41.3, P < 0.001). Fig. 8 shows a comparison between responses for leave intention from the 2016 TPHS and the 2023 survey by gender and location (see Supplementary Material S12, Supplementary Table S5 for full results).

Fig. 8.

Wildfire leave intention proportion by gender and location, Tasmania, Australia (2016 and 2023).

In both the 2016 and 2023 surveys, if a participant responded to Question B ‘I would not leave’ as their intention, they were subsequently asked about the main reason they would stay (see Box 1, Question C). In both 2016 and 2023, the majority of participants who would not leave responded ‘I think I can protect my home and/or myself’ (72.5%, n = 58 in 2016; 55.2%, n = 16 in 2023) (see Supplementary Material S12, Supplementary Fig. S13 for full results).

Discussion

Our 2023 survey highlights substantial social and geographic variation among adults’ perceptions and their likely responses to wildfire threats across the fire-prone island of Tasmania. Gender and location were significantly associated with leave intention, with women more likely to leave than men, and those residing in urban areas more likely to leave than those in rural areas. Additionally, previous wildfire experience and home ownership were significantly associated with increased wildfire preparedness actions, while residing in an urban location was significantly and negatively associated with preparedness actions. Considering the 2023 survey results collectively, we can identify four distinct archetypes – the ‘immediate leaver’, the ‘prepared stayer’, the ‘unprepared leaver’ and the ‘informed leaver’, each with diverse sociodemographic characteristics with respect to gender, age, income, home ownership, residential location and previous wildfire experience.

Our classification of archetypes relevant to wildfire self-evacuation intention in Tasmania has some similarities with previous studies initiated by the Fires Services Commissioner in Victoria, Australia (Fire Services Commissioner Victoria 2013). The Victorian study identified seven distinct archetypes related to the attitudes and behaviours of people responding to a wildfire threat, five of which loosely match the identified Tasmanian archetypes (see Table 3).

Archetypes identified in our study also reflect the changing demographics of Tasmania, with more recent population growth driven by the migration of people in their mid-20s to mid-30s from other Australian states and overseas, who tend to live in predominantly urban locations (State Growth Tasmania 2021). For instance, Group 3, the ‘unprepared leavers’, who tend to be younger, live in urban areas, and are less familiar with the risks and messaging associated with wildfires, may be representative of these newer arrivals to Tasmania.

With reference to Fig. 4 (showing where each group sits on the spectrum of informed/uninformed, prepared/unprepared and leave/stay), ideally residents living in areas at risk of wildfires tend towards the left side of this spectrum, identifying as informed, prepared and choosing to leave their property. An arguably ‘riskier’ approach is to remain on the right side of the spectrum as uninformed, unprepared and choosing to stay. None of the identified archetypes in Tasmania are fully to the right side of the spectrum, while Group 4 are placed fully on the left side, suggesting this group may identify as having the most ideal approach to wildfire safety compared to other groups. This diagram also shows the spread of groups across the spectrum of various sociodemographic variables, highlighting clear opposing factors from Groups 2 and 3, and a less clear differentiation between Groups 1 and 4.

Our results from analyses related to sociodemographic drivers of self-evacuation intention and preparedness actions are congruent with previous studies in Australia, including gender, location and home ownership. With regards to gender, McLennan et al. (2011) found that women were more likely than men to have left their homes during the Victorian ‘Black Saturday’ fires in 2009, noting these findings related to decisions made during the threat of a wildfire rather than self-reported intentions, as were recorded in our survey. In a survey of households in southeast Australia regarding leave intentions and preparedness actions related to wildfires, McLennan et al. (2014) found those residing in an urban setting were most likely to report an intention to leave, compared to those residing in a rural or remote setting, a result aligned with our findings about resident location and intention to leave. Conversely, McLennan et al. (2011) found that location did not influence resident actions with regard to leaving or staying under the threat of a wildfire. These opposing findings may be explained by differing definitions of urban and rural, or the differing contexts of intention to leave versus actual decision to leave. Furthermore, in a survey of residents in Hobart, Tasmania, Paton et al. (2008) found that those having a strong attachment to their home and property were more likely to have undertaken wildfire preparedness actions around their home, as opposed to those that reported a low attachment to their place of residence, a result congruent with our finding that home ownership was associated with preparedness actions (noting preparedness actions included in our survey covered both actions around the home and planning actions such as creating a bushfire survival plan and packing an emergency kit).

We have used the concept of archetypes to discern broad contours of public perception and response to wildfire threat. We accept that further research can refine and provide greater depth to our archetypes, particularly focusing on more intangible sociocultural and value elements and dimensions that affect wildfire and climate change adaptation, such as world views (Ruane 2018) and political orientation (Hazlett and Mildenberger 2020). While the approach of using archetypes has been used in prior wildlife research, we acknowledge that the findings of such studies cannot necessarily be generalised across all fire-prone regions. This is because of the diversity of national and regional-specific legislation, policy and advice messaging for wildfires that shape residents’ preparation and evacuations intentions (McLennan et al. 2019). A fertile area of wildfire disaster risk reduction research is the exploration of behavioural archetypes in contrasting cultural, biophysical and political settings. Such targeted research will help resolve how archetypal behaviours are shaped, and whether there are consistent patterns despite markedly different sociocultural milieu. This knowledge is important for further tailoring of community education and adaptation to the escalating risk of extreme wildfire globally (Cunningham et al. 2024).

Compared to the 2016 TPHS, we found a significant increase in the proportion of Tasmanian residents intending to leave their property at the threat of a wildfire in 2023. To the best of our knowledge, our study is the first to capture and report changes to wildfire self-evacuation intention over time for an Australian jurisdiction. This evidence goes part way towards evaluating the national change in advice messaging in response to a wildfire threat, from ‘prepare, stay and defend or leave early’ (commonly known as ‘stay or go’) which was prevalent during the 2009 Black Saturday bushfires (Teague et al. 2010) to the more recent national advice of ‘leave early’ (Australasian Fire and Emergency Service Authorities Council 2019).

Conclusion

By using an archetype framework, our study helps fire managers understand variability in wildfire preparation and response among residents of a highly fire-prone environment in terms of social and demographic patterns. Such knowledge aids fire managers in tailoring messaging and educational campaigns for those most at risk. Through using repeated survey questions over a 5-year period, we demonstrate clear changes in intention to self-evacuate among broad demographic groups (men/women, urban/rural), suggestive of the differential effectiveness of current generic educational programs and bushfire advice. Further work is required to determine if self-reported proximity to wildfire hazards is accurate, and whether intentions closely match actual behaviour.

Supplementary material

Supplementary material is available online.