Elsevier

Environmental Modelling & Software

Volume 52, February 2014, Pages 166-175
Environmental Modelling & Software

Reducing wildfire risk to urban developments: Simulation of cost-effective fuel treatment solutions in south eastern Australia

https://doi.org/10.1016/j.envsoft.2013.09.030Get rights and content

Highlights

  • Fuel treatment in the interface zone gives the greatest reduction in risk of property loss.

  • Costs of managing the interface zone are comparable to landscape treatments.

  • Treating the interface zone is cost-effective and significantly reduces the risk of property loss.

Abstract

Wildfires can result in significant economic and social losses. Prescribed fire is commonly applied to reduce fuel loads and thereby decrease future fire risk to life and property. Fuel treatments can occur in the landscape or adjacent to houses. Location of the prescribed burns can significantly alter the risk of house loss. Furthermore the cost of treating fuels in the landscape is far cheaper than treating fuels adjacent to the houses. Here we develop a Bayesian Network to examine the relative reduction in risk that can be achieved by prescribed burning in the landscape compared with a 500 m interface zone adjacent to houses. We then compare costs of management treatments to determine the most cost-effective method of reducing risk to houses. Burning in the interface zone resulted in the greatest reduction in risk of fires reaching the houses and the intensity of these fires. Fuel treatment in the interface zone allows for a direct transfer of benefits from the fuel treatment. Costs of treating fuels in the interface were significantly higher on a per hectare basis, but the extent of area requiring treatment was considerably lower. Results of this study demonstrate that treatment of fuels at the interface is not only the best means of reducing risk, it is also the most cost-effective.

Introduction

Wildfires can result in significant social and economic losses when they encounter the interface between human and natural areas. Wildfires in California, USA, in 2007 resulted in the evacuation of 300,000 people and the loss of 2223 houses (McCaffrey and Rhodes, 2009). The Black Saturday fires (February 2009) in Victoria, Australia resulted in the loss of over 2000 houses and 173 lives, as well as significant costs to agricultural and timber production (Leonard et al., 2009). Losses are expected to increase as urban populations grow and the extent of the interface between native vegetation and urban areas increases (Keeley et al., 1999, McCaffrey and Rhodes, 2009, Syphard et al., 2007).

Prescribed fire is used as a pre-emptive fuel reduction treatment with the aim of reducing the rate of spread and intensity of subsequent wildfires and consequent risk of loss of assets (e.g. Hodgson, 1968, Reinhardt et al., 2008). Areas treated by prescribed burning fall into two main locations – landscape and interface burns. Landscape burning is applied in strategic areas in order to decrease the size of fires by slowing the rate of spread and reducing intensity. Such effects are intended to increase the chance of suppression success in truncating the spread of wildfires. Interface burns occur immediately adjacent to residential areas and aim to lower the intensity of the fire in these areas to increase the probability of safe and effective suppression and consequent protection of these houses. There has been considerable debate about the degree to which prescribed fire can alter the risk of house loss (see reviews by Fernandes and Botelho, 2003, Penman et al., 2011a).

Prescribed burning has been demonstrated to reduce the area burnt by wildfires and fire intensity in simulation and empirical studies in some regions (Boer et al., 2009, King et al., 2006, Loehle, 2004, Moghaddas et al., 2010, Penman et al., 2011a, Price and Bradstock, 2011), but not others (Price et al., 2012). Treatment of fuels at the interface between vegetation and houses are more likely to reduce risk to property than treatments distributed in the landscape (Bradstock and Gill, 2001, Cary et al., 2009, Ager et al., 2010, Bradstock et al., 2012). Gibbons et al. (2012) found that the probability of house loss was most strongly influenced by vegetation cover close to property (i.e., 40 m) and distance of the property from flammable forests, whereas landscape prescribed burning had small effects on the probability of loss. Gibbons et al. (2012) also found that fire weather conditions were the predominant effect on losses, with implied fuel effects being strongly conditional on the fire weather. This may indicate that both intra- and inter-regional variations in weather conditions during fires may have potential to affect the efficacy of any particular fuel treatment strategy, i.e., landscape or adjacent to property. As a result, insight is required to understand how differing treatment strategies may perform under the gamut of fire weather conditions.

Determining ideal treatments for maximum reduction in probability of property loss will involve trade-offs in cost and other management objectives, such as ecological and cultural heritage (Ager et al., 2010, Bradstock et al., 2012, Stockmann et al., 2010). If interface treatments are more effective in reducing risk, compared with landscape treatments, then any potential advantage gained from pursuing an interface treatment strategy will depend on relative cost. If such interface treatments are prohibitively expensive, then the overall advantage in risk mitigation may be diminished or negated. Quantification of the reduction in risk provided by combinations of treatment effort with comparative costs is required to develop an understanding of how to optimise investment in fuel treatment.

Bayesian Networks (BNs) are an ideal tool for understanding complex environmental problems (Johnson et al., 2010, Kelly et al., 2013, Pollino et al., 2007). Developed from graph theory, BNs are directed acyclic graphs with variables represented by nodes and arrows representing the directional relationships between them (Pearl, 1986). Two main types of nodes are used – decision nodes and stochastic nodes. Decision nodes represent actions which an agent must select between competing approaches. Stochastic nodes are random variables represented by a conditional probability table which contain the joint probability distributions for the variables (Korb and Nicholson, 2011). Variables represented by parentless nodes (i.e., those at the top of the model and are not influenced by other variables in the model) have a conditional probability table containing a single probability for each state in that node. Variables represented by child nodes (i.e., those variables that are influenced by one or more variables), have a conditional probability table which represent the probability of a given state in the child node given the state(s) in the parent node(s). These probability distributions and the associated uncertainty are propagated throughout the network. Outputs are then presented as likelihoods making the model ideal for inclusion in a risk management framework (Marcot et al., 2001). These models, therefore, allow for the risk of wildfires to houses to be estimated over the full range of climatic and environmental conditions in a single framework.

We develop a Bayesian Network (BN) to investigate the differing trade-offs involved in designing an optimal prescribed burning strategy for the reduction of risk to property in the Sydney Basin bioregion in south eastern Australia, where destructive fires are common (Bradstock et al., 1998; Blanchi et al., 2010). Within this region, loss of property is often, but not always, associated with large fires burning under severe weather conditions. The aim of the study was to determine the relative efficacy of an offensive (i.e., treatments broadly dispersed in the landscape), as opposed to a defensive approach (i.e., treatments concentrated close to property) to prescribed burning. In particular we aimed to determine which of these contrasting strategies most strongly diminish the risk of property loss across the full range of weather conditions and offers the most cost-effective reduction in risk to life and property. That is, which treatment combination provides the highest reduction in risk for a given level of expenditure?

Section snippets

Study area

The Sydney Basin Bioregion is the most populated area of Australia containing three large urban centres (Sydney, Newcastle and Wollongong) and a population of 5.5 million people (www.abs.gov.au, Accessed March 2011). All three urban centres are surrounded by large tracts of fire prone native vegetation which is primarily managed for conservation and water yield. These areas are dominated by Eucalyptus forests with dry sclerophyll forests on the ridge and upper slopes and wet sclerophyll forests

Risk to houses

Annual risk of a wildfire reaching houses was negatively related to level of prescribed burning in both the BUIZ and the landscape (Fig. 4). For 0% of BUIZ treatment, increasing prescribed burning in the landscape from 0% to 10% per annum resulted in an approximately 10% reduction in annual risk from 0.154 to 0.138. In contrast, if the landscape treatment was 0% per annum and treatment of the BUIZ increased from 0 to 10% per annum the annual risk was reduced by approximately 45% from 0.154 to

Treatment, wildfire spread and risk

Landscape treatments had a relatively small influence compared with BUIZ treatments, consistent with other simulation and empirical studies (Bradstock et al., 2012, Cary et al., 2009, Gibbons et al., 2012). One of the key reasons for this is that approximately 40% of ignitions occur within BUIZ where by definition the landscape treatments are incapable of influencing fire behaviour before the wildfire impacts on houses (Fig. 2). Ignitions that do occur in the landscape (60%) are only likely to

Acknowledgements

The study was funded by the NSW Rural Fire Service. Kevin Tolhurst and Simon Heemstra, provided useful comments in the design phase of the study. Comments of Kevin Tolhurst, Derek Chong and Simon Metcalf aided the development of earlier drafts of this paper.

References (75)

  • T.D. Penman et al.

    Climate and recent fire history affect fuel loads in Eucalyptus forests: implications for fire management in a changing climate

    For. Ecol. Manag.

    (2010)
  • C.A. Pollino et al.

    Parameterisation and evaluation of a Bayesian network for use in an ecological risk assessment

    Environ. Model. Softw.

    (2007)
  • O.F. Price et al.

    The efficacy of fuel treatment in mitigating property loss during wildfires: Insights from analysis of the severity of the catastrophic fires in 2009 in Victoria, Australia

    J. Environ. Manag.

    (2012)
  • O.F. Price et al.

    The impact of antecedent fire area on burned area in southern California coastal ecosystems

    J. Environ. Manag.

    (2012)
  • E.D. Reinhardt et al.

    Objectives and considerations for wildland fuel treatment in forested ecosystems of the interior western United States

    For. Ecol. Manag.

    (2008)
  • H.D. Safford et al.

    Effects of fuel treatments on fire severity in an area of wildland-urban interface, Angora Fire, Lake Tahoe Basin, California

    For. Ecol. Manag.

    (2009)
  • K. Stockmann et al.

    Guiding preventative wildland fire mitigation policy and decisions with an economic modeling system

    For. Policy Econ.

    (2010)
  • J.R. Thompson et al.

    Vegetation and weather explain variation in crown damage within a large mixed-severity wildfire

    For. Ecol. Manag.

    (2009)
  • G. Winter et al.

    The role of community policies in defensible space compliance

    For. Policy Econ.

    (2009)
  • M.A. Acuna et al.

    Integrated spatial fire and forest management planning

    Can. J. Forest. Res.

    (2011)
  • A.H. Berry et al.

    The effect of the wildland-urban interface on prescribed burning costs in the Pacific Northwestern United States

    J. For.

    (2004)
  • R. Blanchi et al.

    Property safety: judging structural safety

  • R. Blanchi et al.

    Meteorological conditions and wildfire-related houseloss in Australia

    Int. J. Wildland Fire

    (2010)
  • R. Bradstock et al.

    Effects of weather, fuel and terrain on fire severity in topographically diverse landscapes of south-eastern Australia

    Landsc. Ecol.

    (2010)
  • R.A. Bradstock et al.

    Prediction of the probability of large fires in the Sydney region of south-eastern Australia using fire weather

    Int. J. Wildland Fire

    (2009)
  • R.A. Bradstock et al.

    Living with fire and biodiversity at the urban edge: in search of a sustainable solution to the human protection problem in southern Australia

    J. Mediterr. Ecol.

    (2001)
  • G.J. Cary et al.

    Relative importance of fuel management, ignition management and weather for area burned: evidence from five landscape-fire-succession models

    Int. J. Wildland Fire

    (2009)
  • G.J. Cary et al.

    Comparison of the sensitivity of landscape-fire-succession models to variation in terrain, fuel pattern, climate and weather

    Landsc. Ecol.

    (2006)
  • N.P. Cheney et al.

    Grassfires: Fuel, Weather and Fire Behaviour, second ed.

    (1997)
  • D.A. Cleaves et al.

    Prescribed burning costs: trends and influences in the National forest system

  • B.M. Collins et al.

    Spatial patterns of large natural fires in Sierra Nevada wilderness areas

    Landsc. Ecol.

    (2007)
  • R.J. Conroy

    Fuel management strategies for the Sydney Region

  • S.G. Cumming

    Effective fire suppression in boreal forests

    Can. J. For. Res.

    (2005)
  • P.M. Fernandes et al.

    A review of prescribed burning effectiveness in fire hazard reduction

    Int. J. Wildland Fire

    (2003)
  • M. Flannigan et al.

    Future area burned in Canada

    Climatic Change

    (2005)
  • J. Fried et al.

    The impact of climate change on wildfire severity: a regional forecast for Northern California

    Climatic Chang.

    (2004)
  • P. Gibbons et al.

    Land management practices associated with house loss in wildfires

    PLoS ONE

    (2012)
  • Cited by (50)

    View all citing articles on Scopus
    View full text