Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Influence of Fuels, Weather and the Built Environment on the Exposure of Property to Wildfire

  • Trent D. Penman ,

    trent.penman@unimelb.edu.au

    Current address: Department of Forest and Ecosystem Science, University of Melbourne, Victoria, Australia.

    Affiliation Centre for Environmental Risk Management of Bushfires, Institute of Conservation Biology and Environmental Management, University of Wollongong, NSW, Australia

  • Luke Collins,

    Affiliation Hawkesbury Institute for the Environment, University of Western Sydney, Penrith, NSW, Australia

  • Alexandra D. Syphard,

    Affiliation Conservation Biology Institute, La Mesa, California, United States of America

  • Jon E. Keeley,

    Affiliations U.S. Geological Survey, Western Ecological Research Center, Sequoia-Kings Canyon Field Station, Three Rivers, California, United States of America, Department of Ecology and Evolutionary Biology, University of California Los Angeles, Los Angeles, California, United States of America

  • Ross A. Bradstock

    Affiliation Centre for Environmental Risk Management of Bushfires, Institute of Conservation Biology and Environmental Management, University of Wollongong, NSW, Australia

Influence of Fuels, Weather and the Built Environment on the Exposure of Property to Wildfire

  • Trent D. Penman, 
  • Luke Collins, 
  • Alexandra D. Syphard, 
  • Jon E. Keeley, 
  • Ross A. Bradstock
PLOS
x

Abstract

Wildfires can pose a significant risk to people and property. Billions of dollars are spent investing in fire management actions in an attempt to reduce the risk of loss. One of the key areas where money is spent is through fuel treatment – either fuel reduction (prescribed fire) or fuel removal (fuel breaks). Individual treatments can influence fire size and the maximum distance travelled from the ignition and presumably risk, but few studies have examined the landscape level effectiveness of these treatments. Here we use a Bayesian Network model to examine the relative influence of the built and natural environment, weather, fuel and fuel treatments in determining the risk posed from wildfire to the wildland-urban interface. Fire size and distance travelled was influenced most strongly by weather, with exposure to fires most sensitive to changes in the built environment and fire parameters. Natural environment variables and fuel load all had minor influences on fire size, distance travelled and exposure of assets. These results suggest that management of fuels provided minimal reductions in risk to assets and adequate planning of the changes in the built environment to cope with the expansion of human populations is going to be vital for managing risk from fire under future climates.

Introduction

Wildfires can pose a significant risk to people and property [1][4]. Large losses of property and life have been recorded from individual fires or fire complexes in fire prone regions throughout the globe [5][7]. Such events can impact individuals and communities for many years [8][11]. As a result, fire management agencies invest significant budgets in reducing the risk of loss from wildfires, primarily through investment in fuel management and active fire suppression [12], [13].

Fuel management is a commonly used risk management tool in fire prone landscapes [14], [15]. The justification for the approach derives from the fundamentals of fire behaviour with a reduction in fuel loads expected to result in a subsequent lowering of the fire intensity and rate of spread [16][20]. In turn, these changes to fire behaviour are expected to increase the probability of successfully containing the fire with active fire suppression [21], [22].

Fuel breaks are a commonly applied type of fuel management treatment in a variety of ecosystems. These are mechanical reductions or removal of fuel, typically as linear features along ridgetops, to enable safe access for fire suppression crews to manage fires [23]. Empirical analysis has found fuel breaks are more effective when readily accessible and well-maintained and when used for backfire operations [23], [24]. Under these conditions the intensity and rate of spread are lower and containment of the fires through active suppression is more likely to be successful [21], [22], [24]. Simulation studies have found individual fuel breaks have the potential to reduce the size and intensity of wildfires [25], [26]. While studies examining the impact of fuel breaks on the behaviour of individual fires are valuable, to quantify the extent to which fuel breaks reduce risk to lives and property, we need to examine the role of fuel breaks at the landscape scale [27]. Although simulation studies have shown management of fuels can alter fire regimes in forested ecosystems, there is a need to quantify how fuel breaks, in particular, can reduce the risk of exposure of assets. Here we examine the performance of fuel breaks in mitigating risk in Mediterranean shrubland (chaparral) landscapes.

Fuel breaks are the main fuel treatment carried out in the chaparral shrublands of southern California, with a long history of extensive deployment [23], [24]. Thus case studies of their effectiveness in mitigating risk can provide valuable insight into mitigation of risk in a region with some of the highest exposure of fire-prone urban and peri-urban developments in the world. Case studies in this context may also be valuable for assessing fuel treatment options in other fire-prone, temperate environments which share similar elements of the problem [28].

Bayesian Networks (BN) provide a suitable methodology for the analysis of risk management problems [29][31]. They are depicted as directed acyclic graphs with nodes representing the variables and arrows representing the directional relationships between nodes [32]. There is a conditional probability table for each node that contains the joint probability distributions for the variable [33]. Root nodes occur at the top of the model and are not influenced by other variables in the model. These nodes have a conditional probability table containing a single probability for each state in that node. Child nodes are variables that are influenced by one or more variables (parent nodes). These nodes have a conditional probability table that represents the probability of a given state in the child node given the state(s) in the parent node(s). Uncertainty is propagated throughout the model, providing probability distribution for all output nodes. Results are likelihoods that form the basis for risk management calculations [29]. Bayesian Networks have been found to be a valuable method for examining fire risk management problems at the landscape scale [31], [34].

Here we develop a Bayesian Network model to examine the role of fuel breaks in reducing the risk of assets being exposed to wildfire using San Diego county as a case study area. San Diego county has a history of major fire losses (circa 5000 houses destroyed between 2000 and the present), which reflects extensive and rapidly growing developments, exposure to regular episodes of severe fire weather and terrain and vegetation conducive to the rapid spread of intense fires [35]. Thus the county comprises encapsulates key elements and a significant portion of the wildfire risk problem in southern California. As a case study it therefore provides the potential for key insights into fire management that are regionally, nationally and globally significant. The BN model combines data from a fire simulation model (FARSITE) and environmental data. We specifically seek to determine the extent to which risk posed by wildfires to properties at the wildland urban interface is influenced by the environment (weather, fuel moisture, natural environment), developmental patterns (built environment) and fuel management (fuel load, fuel breaks).

Methods

The study area was San Diego County, California, USA, which supports a population of approximately 3.2 million people in an area of approximately 11 000 km2 (US Census, http://quickfacts.census.gov/qfd/states/06/06073.html, Accessed December 2013). In the county, there is a long and complex wildland urban interface [36], along which thousands of homes have been destroyed in major fire events in the last decade [37]. Native vegetation of the area is dominated by chaparral, coastal sage scrub, and oak woodland. The county experiences a Mediterranean climate with hot dry summers and winter rainfall with moderate temperatures. Fires in the county occur most frequently in summer months, but most area burned occurs in the autumn when annual fuel moisture is lowest and Santa Ana winds are most frequent [38]. San Diego County was selected as it is dominated by highly flammable shrubland vegetation and falls within known Santa Ana wind corridors, making the region prone to recurrent large fire events [39], [40]. The simulation landscape was defined as a 60 km×60 km area east of San Diego (Figure 1).

A Bayesian Network model was used to examine the relative influence of weather, the built and natural environment and fuel breaks on the risk of exposure to wildfire. Here we broadly follow the methods for developing Bayesian Networks recommended by Marcot et al. [41] and Chen and Pollino [42]. The primary steps used were to construct a conceptual model of the problem, develop influence diagrams to depict the relationships of the conceptual model and finally populate all the conditional probability tables within the model.

A basic conceptual model for the study was derived from previous fire management research [31], [34]. This model assumed an ignition had occurred and predicted the subsequent spread and potential impact of the wildfire upon property, which was dependent upon environmental conditions and management decisions.

Influence diagrams encapsulated the conceptual framework and the relevant environmental factors (Figure 2). These were developed for the study area by the authors through a series of workshops held by the United States Geological Survey (December 2010, September 2011, May 2012) involving twelve researchers with expertise in fire management, Bayesian Network analysis and landscape ecology. Iterations of the influence diagrams were presented to the group, discussed and then further refined until a consensus was reached. In the final set of influence diagrams, fire size and distance travelled were assumed to be influenced by the key variables considered in the simulation modelling – weather, fuel moisture, landscape fuel load and the occurrence of fuel breaks within the National Forest. Elements of the natural environment at the ignition location, specifically fuel type, fuel load and elevation, were considered to have an influence on fire size and distance travelled. The built environment also influences fire spread and exposure of property. Exposure to fire was taken as simple function considering the distance the fire could potentially travel and the distance from the ignition point to property (Figure 2).

thumbnail
Figure 2. Influence diagrams for the Bayesian Network Model.

See Table 1 for node definitions and states.

https://doi.org/10.1371/journal.pone.0111414.g002

Data for the conditional probability tables in the analysis (Table 1) were derived from either a simulation study or empirical data for the study region. We undertook a comprehensive simulation of fires in the area using the Fire Area Simulator (FARSITE) using random ignition locations. FARSITE is a two dimensional spatially explicit model that models fire spread using Huygens' principle [43]. The simulations examined all combinations of fire weather (low, high and Santa Ana), live fuel moisture (LMF 60% and 90%), fuel loading (low and high) and the presence or absence of maintained fuel breaks. A total of 11,944 fires were simulated a FARSITE framework. Complete details of the approach are presented in Table S1 in Material S1. From the simulation data, we derived values for the nodes fire size and distance travelled. Weather, fuel moisture, landscape fuel load and the occurrence of fuel breaks were implemented as decision nodes to explore the relative influence of each factor. Environmental variables (i.e. elevation, ignition fuel load, ignition fuel type and distance to the interface) were derived from GIS data from the region (www.landfire.gov Accessed October 2012). All nodes and methods of discretisation are described in more detail in Table 1. The Bayesian Network model is available from the ABNMS data repository (www.abnms.org).

thumbnail
Table 1. Nodes, definitions and states used in the Bayesian Network model.

https://doi.org/10.1371/journal.pone.0111414.t001

Two methods were used to examine the relative influence of variables. Firstly, the relative influence of each of the modelled factors was assessed using values of the terminal node – “exposure to fire”. This node reflects the risk of property being exposed to a wildfire under the given conditions. We considered all 24 combinations of the key predictor variables – weather (3 levels), fuel moisture (2 levels), landscape fuel load (2 levels) and fuel breaks (2 levels) (Table 1). Secondly, the sensitivity of nodes was assessed using the sensitivity to findings function in Netica (http://www.norsys.com/netica.html, Accessed December 2013). This function examines the extent to which changes in one variable affects the variable of interest. We examined the sensitivity of findings for “exposure to fire” as the terminal node and “distance travelled”.

Results

The size of fires after a 12 hour simulation ranged from 0.1 ha to 28,480 ha, with a mean of 2896 ha (±40 S.E.) and a median 882 ha. The distance travelled ranged from 18 m to 47,300 m with a mean of 6171 m (±57 S.E.). Responses of fire size and distance travelled to the predictor variables were consistent. Fire size and distance travelled increased with the severity of fire weather and the landscape fuel load, and decreased with increasing fuel moisture. The presence of fuel breaks had little influence on either individual fire size or distance travelled (Figure 3).

thumbnail
Figure 3. Relationships from the FARSITE simulation data between weather and fire size with fuel moisture of a) 60% and b) 90%, and distance travelled with fuel moisture of c) 60% and d) 90%.

Open symbols are for simulations with no fuel breaks, closed symbols for simulations with fuel breaks. Circles represent a high landscape fuel load scenario and triangles represent a low landscape fuel load scenario. NB 95% confidence intervals were too small to depict in the graphics.

https://doi.org/10.1371/journal.pone.0111414.g003

Risk of exposure to fires was influenced most strongly by weather (Figure 4). The risk of exposure was >99% for all fuel scenarios considered under Santa Ana conditions. Under low and high weather conditions, fuel load and fuel moisture had a strong influence of risk of exposure, whereby risk was greater when landscape fuels were high compared to low and when fuel moisture was 60% compared to 90% (Figure 4). These effects were lost under Santa Ana conditions, where fuel load and fuel moisture had very little effect on the risk of exposure. Fuel breaks had very little influence on the risk of exposure. Risk varied significantly across elevation with fires starting in low elevation sites having significantly higher risk than fires starting at higher elevation sites (Figure 4).

thumbnail
Figure 4. Risk of exposure for the 24 scenarios modelled for a) fires igniting at elevations of 300 to 600 m; b) fires igniting at elevations of 1000 to 4000 m; c) all locations across the landscape.

Open symbols = fuel break scenarios; closed symbols = no fuel breaks; Grey symbols = fuel moisture of 60%; Black symbols = fuel moisture of 90%; Circles = high landscape fuel loads; Triangle = low landscape fuel loads.

https://doi.org/10.1371/journal.pone.0111414.g004

Exposure to fires was most sensitive to changes in the built environment, as well as fire parameters, i.e. fire size and distance travelled (Figure 5a). Distance to structures from the ignition had the strongest influence, followed by the distance travelled by a fire and the fire size. This is expected as parent nodes are likely to have the strongest influence on a node. Of the variables a greater distance from the terminal node, variables depicting the built environment (housing density and distance to road) were the next most influential variables, followed by weather. Variables describing the natural environment had only a modest influence on exposure to fires, with the fuel variables having no influence (Figure 5a). Distance travelled by a fire was primarily influenced by the weather on the day of the fire (Figure 5b). Natural environment variables and fuel load all had minor influence (<2.1%), with all built environment variables having low influence (<1%) (Figure 5b).

thumbnail
Figure 5. Sensitivity to findings for nodes a) Exposure to fire and b) Distance travelled.

White bars = fire variables, dark grey bars = built environment variables; light grey bars = natural environment variables; Black bars = simulation model variables. D2S = distance to structure; DistTrav = distance travelled by the fire; HouseDens = housing density; D2Rd = Distance to road; D2C = distance to the coast; IgFuelType = fuel type at the point of ignition; IgFuels = fuel load at the point of ignition; FuelBreaks = presence of fuel breaks.

https://doi.org/10.1371/journal.pone.0111414.g005

Discussion

Consistent with previous research, fire size and distance travelled is most sensitive to changes in weather [18], [44][48] and the risk of exposure is most strongly influenced by attributes of the fire (size and distance travelled) and the nature of the built environment [34], [49]. Measured attributes of fuels had only a minor influence on fire parameters and risk of exposure. Fuel breaks in the National Forest did not affect fire size, distance travelled or the risk of exposure at the interface.

Risk of exposure

Weather had the strongest influence on fire size and distance travelled (Figure 5b) and indirectly, the risk of exposure. It has been well documented that fire weather strongly influences fire size, the rate of spread, spotting distance, fire intensity and severity [16], [18], [40], [50][52]. As a result, wildfires burning under extreme conditions account for the majority of area burnt for many regions [44], [53][55]. Furthermore, it is under extreme fire weather conditions wildfires pose the greatest threat to people and property [50], [56][59].

Exposure was also influenced by the built environment, namely distance to road and housing density (Figure 5). Higher densities of properties occur at lower elevations (less than 300 m) and these have a higher risk of exposure compared with higher elevation sites independent of weather (Figure 4). Fires starting close to the interface are more likely to impact upon assets under any weather conditions [60], [61], whereas fires starting considerable distances from property are only likely to impact on property under extreme weather conditions conducive to fire spread [34], [57]. These results suggest that adequate planning of the changes in the built environment to cope with the expansion of human populations is going to be important for managing risk from fire [49], [62], [63]. We do note that the model only considers exposure to fire and we have not considered the size of the fire or the extent of the interface exposed to the fire. Fires that start away from populations will be significantly larger when they do impact on the interface compared with those that start nearby [60]. Larger fires would be expected to impact on a greater number of assets than smaller fires and these relationships require further investigation.

Of the factors relating to fuels, landscape fuel load had the strongest influence on fire size, distance travelled and risk of exposure (Figure 3; Figure 4). However, our model is more sensitive to the effects of weather and the built environment (Figure 5). Price et al. [64] found no effect of antecedent area burnt on the annual area burnt by wildfire in southern California coastal systems. The authors argue that the low effect of past fire is related to the low level of wildfire in the system (<2% per annum) and the rapid development of fuels (1–2 years). A large proportion of the study area (∼22%) burnt in wildfires during 2007 and was consequently in the early stage of fuel development (i.e. 1 year old) in the 2008 fuel layer, which may explain the strong influence of fuel load under low and high fire weather. However, in San Diego County, the majority of annual area burnt occurs under extreme Santa Ana fire weather [55] where our model found no effect of fuel load. These results support the finding of Price et al. [64] that landscape fuel treatments in these systems are unlikely to reduce the risk of fire to assets. (Figure 3; Figure 4). Scenarios with 90% fuel moisture had significantly lower risk than those with 60% under low and high conditions, but not under Santa Ana (Figure 4). Live fuel moisture is related to fire activity in southern California, with large fires generally being associated with low levels (∼60–80%) [65], [66]. Early in the fire season live fuel moisture is generally greater than 90% (Keeley et al. 2009), the resulting area burnt in southern California typically is relatively small [65], [66]. Our selection of 60% and 90% may not have truly captured the variable effect of live fuel moisture, particularly when these values exceed 100% and fire activity is expected to be low. However, the greatest risk to assets comes during Santa Ana weather conditions where there is no distinguishable effect of live fuel moisture, providing further support for our existing results. Regardless, it is typically lowest at the end of summer drought when Santa Ana winds and hence large fires are most likely [67].

Fuel breaks were ineffective at altering risk of exposure of property under any weather scenario in our study. Here we modelled fires assuming that all mapped fuel breaks in San Diego County were maintained (see Material S1; Figure 1), which exceeds current practice. Fuel breaks have been found to affect individual fire size and distance travelled [25], [26]. The network of fuel breaks in San Diego County is highly clustered (Figure 1) presumably to protect particular assets from future wildfires, although fuel breaks continue to be constructed. Clustering of the fuel breaks will result in low encounter rates with wildfires that will result in a low efficacy of this management technique [24] when considering the landscape level risk. Here we assumed a random ignition model, however ignitions do not occur randomly across landscapes [60], [68][70] and tend to occur close to roads and development. This is important with regards to the result that fires that start closer to homes are most likely to reach those homes. Simulations have revealed that fire size and burn probability are sensitive to the use of random against non-random ignition locations, though these biases are minimised under extreme weather conditions [71] when the greatest risk of exposure occurs (Figure 4).

The effectiveness of fuel breaks is contingent on suppression resources and access [24]. In our study, the fuel breaks were constructed in FARSITE in a manner that simulated suppression along the fuel breaks. Fuel breaks tend to be constructed to allow for control of the fire flanks and not the head fire, i.e. the point of the greatest forward rate of spread. As a result, fuel breaks are unlikely to affect the distance travelled by a fire and have negligible impacts on total fire size. Our model did not model the interaction of suppression through direct attack of the fire front or indirect attack from other breaks in the landscape, e.g. roads and rivers. Inclusion of the suppression at these points may have altered the efficacy of fuel breaks when estimating risk. Similarly, we did not consider the impact on fires of igniting backburns from fuel breaks. However, as the severity of the fire weather increases the effectiveness of suppression actions are severely diminished [22], [48], [72]. Therefore, we would only consider fuel breaks in conjunction with suppression as having potential to further reduce risk under low fire weather and not under moderate fire weather or Santa Ana conditions [24], [39].

Fire management

Management agencies seek to reduce risk to assets acknowledging that there are no practical means to remove the risk. Weather is the primary determinate of risk to assets from fire [34], [55], [57]. While management actions can be effective under relatively benign fire weather, understanding the effectiveness of management under extreme fire weather is fundamental to determining the extent to which management can reduce risk to people and property [48]. In our model, we considered the role of fuel treatments both fuel breaks for suppression and fuel treatments through examining the role of fuel loads. Neither of these was effective under extreme fire weather despite our model considering extreme levels of fuel treatment (>20% in 1 year old fuel) and fuel breaks (all mapped breaks in the county).

A range of other fire management approaches are available that were not tested here. Three broad management areas have the potential to significantly reduce risk to assets. Firstly, ignition management to reduce the occurrence of ignitions and subsequent fires will reduce the risk to assets [44]. Included in ignition management would be rapid response or initial attack [47], [73], whereby resources are used to aggressively attempt to suppress fires before they become established fires. Secondly, improved urban planning policies to better develop the built environment to reduce the extent of the exposure [49], [62], [74]. This would include building in low risk areas outside Santa Ana wind corridors [35] and incorporating adequate offsets between vegetation and structures [74], [75]. In southern California, the best urban planning practices would be to focus on infill-type development, as low to intermediate housing density, and isolated clusters of development are the strongest risk factors for a house being destroyed in a fire [63]. Finally, reduce the vulnerability of residents and properties at the urban interface. Residents can be educated to reduce the vulnerability of their property through adequate preparation [76], [77]. Furthermore, properties can be built or retrofitted to appropriate construction standards to be more resilient to the impact of fire [78], [79]. While each of these is likely to reduce risk, only through an expanded analysis of these approaches across all weather scenarios will it be possible to identify an optimal management strategy.

Conclusion

Weather determines the risk of exposure for assets in the landscape. Under extreme weather, where the risk of fire is greatest, landscape fuel treatments are unlikely to have a significant influence on risk. These results suggest that managing the occurrence of fire and the spatial distribution of the built environment across the landscape is likely to be the best way to alter the risk profile. Further research is needed to examine the cost trade-offs of each of these approaches.

Supporting Information

Material S1.

Supplementary text outlines the modelling process in farsite. Table S1, Fuel moisture conditions used in the simulations. Dead fuel moisture values are from Scott and Burgin (2005). See text for description of LFM categories.

https://doi.org/10.1371/journal.pone.0111414.s001

(DOCX)

Acknowledgments

Data for the study was provided by U.S. Geological Survey. The paper was made possible due to the workshop participants - Teresa Brennan, Will Forney, CJ Fotheringham, Rick Halsey, Bill Labiosa, Ben Landis, Owen Price, Robert Taylor and Marti Winter. Stuart Brittain and Mark Finney kindly provided a command line version of FARSITE and assistance with its use. Robert Taylor provided advice on manipulating input layers for FARSITE to simulate spotting in shrubland fuels.

Author Contributions

Conceived and designed the experiments: TP LC AS JK RB. Performed the experiments: LC TP. Analyzed the data: TP LC AS. Contributed to the writing of the manuscript: TP LC AS JK RB. Designed the specifics of the computer simulation study: TP LC AS.

References

  1. 1. Cohen JD (2000) Preventing Disaster: Home Ignitability in the Wildland-Urban Interface. Journal of Forestry 98: 15–21.
  2. 2. Lampin-Maillet C, Jappiot M, Long M, Bouillon C, Morge D, et al. (2009) Mapping wildland-urban interfaces at large scales integrating housing density and vegetation aggregation for fire prevention in the South of France. Journal of Environmental Management 91: 732–741.
  3. 3. Gill AM, Stephens SL (2009) Scientific and social challenges for the management of fire-prone wildland–urban interfaces. Environmental Research Letters 4: 034014.
  4. 4. Gill AM, Stephens SL, Cary GJ (2013) The worldwide “wildfire” problem. Ecological Applications 23: 438–454.
  5. 5. McCaffrey S, 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? Journal of Forestry 107: 9–15.
  6. 6. Leonard J (2009) Report to the 2009 Victorian Bushfires Royal Commission Building performacnce in Bushfires. Highett, Victoria, Australia: CSIRO Sustainable Ecosystems. 74 p.
  7. 7. Boustras G, Boukas N, Katsaros E, Ziliaskopoulos A (2012) Wildland fire preparedness is Greece and Cyprus: Lessons learned from the catastrophic fires of 2007 and beyond. In: Paton D, Tedim F, editors. Wildfire and community: Facilitating preparedness and resilience. Springfield, Illinois, USA: Charles C Thomas Publisher Ltd. pp. 151–168.
  8. 8. Ganewatta G (2008) The economics of bushfire management. In: Handmer J, Haynes K, editors. Community Bushfire Safety. Collingwood: CSIRO Publishing. pp. 151–159.
  9. 9. McFarlane AC, Clayer JR, Bookless CL (1997) Psychiatric morbidity following a natural disaster: An Australian bushfire. Social Psychiatry and Psychiatric Epidemiology 32: 261–268.
  10. 10. Langley A, Jones R (2005) Coping Efforts and Efficacy, Acculturation, and Post-Traumatic Symptomatology in Adolescents Following Wildfire. Fire Technology 41: 125–143.
  11. 11. Papadatou D, Giannopoulou I, Bitsakou P, Bellali T, Talias MA, et al. (2012) Adolescents' reactions after a wildfire disaster in Greece. Journal of Traumatic Stress 25: 57–63.
  12. 12. Schoennagel T, Nelson CR, Theobald DM, Carnwath GC, Chapman TB (2009) Implementation of National Fire Plan treatments near the wildland-urban interface in the western United States. Proceedings of the National Academy of Sciences of the United States of America 106: 10706–10711.
  13. 13. Stephens SL, Ruth LW (2005) Federal forest-fire policy in the United States. Ecological Applications 15: 532–542.
  14. 14. Fernandes PM, Botelho HS (2003) A review of prescribed burning effectiveness in fire hazard reduction. International Journal of Wildland Fire 12: 117–128.
  15. 15. Penman TD, Christie FJ, Andersen AN, Bradstock RA, Cary GJ, et al. (2011) Prescribed burning: How can it work to conserve the things we value? International Journal of Wildland Fire 20: 721–733.
  16. 16. Noble IR, Bary GAV, Gill AM (1980) Mcarthurs Fire Danger Meters Expressed as Equations. Australian Journal of Ecology 5: 201–204.
  17. 17. Price OF, Bradstock RA (2010) The effect of fuel age on the spread of fire in sclerophyll forest in the Sydney region of Australia. International Journal of Wildland Fire 19: 35–45.
  18. 18. Bradstock R, Hammill K, Collins L, Price O (2010) Effects of weather, fuel and terrain on fire severity in topographically diverse landscapes of south-eastern Australia. Landscape Ecology 25: 607–619.
  19. 19. Fulé PZ, Crouse JE, Roccaforte JP, Kalies EL (2012) Do thinning and/or burning treatments in western USA ponderosa or Jeffrey pine-dominated forests help restore natural fire behavior? Forest Ecology and Management 269: 68–81.
  20. 20. Collins L, Bradstock RA, Penman TD (2013) Can precipitation influence landscape controls on wildfire severity? A case study within temperate eucalypt forests of south-eastern Australia. International Journal of Wildland Fire Accepted April 3rd 2013.
  21. 21. Hirsch KG, Podur JJ, Janser RF, McAlpine RS, Martell DL (2004) Productivity of Ontario initial-attack fire crews: results of an expert-judgement elicitation study. Canadian Journal of Forest Research 34: 705–715.
  22. 22. Plucinski MP (2012) Factors Affecting Containment Area and Time of Australian Forest Fires Featuring Aerial Suppression. Forest Science 58: 390–398.
  23. 23. Syphard AD, Keeley JE, Brennan TJ (2011) Comparing the role of fuel breaks across southern California national forests. Forest Ecology and Management 261: 2038–2048.
  24. 24. Syphard AD, Keeley JE, Brennan TJ (2011) Factors affecting fuel break effectiveness in the control of large fires on the Los Padres National Forest, California. International Journal of Wildland Fire 20: 764–775.
  25. 25. Agee JK, Bahro B, Finney MA, Omi PN, Sapsis DB, et al. (2000) The use of shaded fuelbreaks in landscape fire management. Forest Ecology and Management 127: 55–66.
  26. 26. Finney MA (2001) Design of Regular Landscape Fuel Treatment Patterns for Modifying Fire Growth and Behavior. Forest Science 47: 219–228.
  27. 27. Syphard AD, Scheller RM, Ward BC, Spencer WD, Strittholt JR (2011) Simulating landscape-scale effects of fuels treatments in the Sierra Nevada, California, USA. International Journal of Wildland Fire 20: 364–383.
  28. 28. Keeley JE, Bond WJ, Bradstock RA, Pausas, editors (2012) Fire in Mediterranean Ecosystems: Ecology, Evolution and Management. Cambridge: Cambridge University Press.
  29. 29. Marcot BG, Holthausen RS, Raphael MG, Rowland MM, Wisdom MJ (2001) Using Bayesian belief networks to evaluate fish and wildlife population viability under land management alternatives from an environmental impact statement. Forest Ecology and Management 153: 29–42.
  30. 30. Johnson S, Mengersen K, de Waal A, Marnewick K, Cilliers D, et al. (2010) Modelling cheetah relocation success in southern Africa using an Iterative Bayesian Network Development Cycle. Ecological Modelling 221: 641–651.
  31. 31. Penman TD, Price O, Bradstock RA (2011) Bayes Nets as a method for analysing the influence of management actions in fire planning. International Journal of Wildland Fire 20: 909–920.
  32. 32. Pearl J (1986) Fusion, propagation, and structuring in belief networks. Artificial Intelligence 29: 241–288.
  33. 33. Korb KB, Nicholson AE (2011) Bayesian Artificial Intelligence. Computer Science and Data Analysis. Second Edition. Boca Raton: CRC/Chapman Hall.
  34. 34. Penman TD, Bradstock RA, Price OF (2014) Reducing wildfire risk to urban developments: Simulation of cost-effective fuel treatment solutions in south eastern Australia. Environmental Modelling & Software 52: 166–175.
  35. 35. Syphard AD, Keeley JE, Massada AB, Brennan TJ, Radeloff VC (2012) Housing Arrangement and Location Determine the Likelihood of Housing Loss Due to Wildfire. PLoS ONE 7: e33954.
  36. 36. Hammer RB, Radeloff VC, Fried JS, Stewart SI (2007) Wildland-urban interface housing growth during the 1990 s in California, Oregon, and Washington. International Journal of Wildland Fire 16: 255–265.
  37. 37. Keeley JE, Syphard AD, Fotheringham CJ (2013) The 2003 and 2007 Wildfires in Southern California. In: Boulter S, Palutikof J, Karoly D, Guitart D, editors. Natural Disasters and Adaptation to Climate Change. Oxford: Cambridge University Press. pp. 42–52.
  38. 38. Keeley JE, Fotheringham CJ (2003) Impact of past, present, and future fire regimes on North American mediterranean shrublands. In: Veblen TT, Baker WL, Montenegro G, Swetnam TW, editors. Fire and climatic change in temperate ecosystems of the Western Americas. New York: Springer. pp. 218–262.
  39. 39. Keeley JE, Safford H, Fotheringham CJ, Franklin J, Moritz M (2009) The 2007 Southern California Wildfires: Lessons in Complexity. Journal of Forestry 107: 287–296.
  40. 40. Moritz MA, Moody TJ, Krawchuk MA, Hughes M, Hall A (2010) Spatial variation in extreme winds predicts large wildfire locations in chaparral ecosystems. Geophysical Research Letters 37: L04801.
  41. 41. Marcot BG, Steventon JD, Sutherland GD, McCann RK (2006) Guidelines for developing and updating Bayesian belief networs applied to ecological modeling and conservation. Canadian Journal of Forest Research 36: 3063–3074.
  42. 42. Chen SH, Pollino CA (2012) Good practice in Bayesian network modelling. Environmental Modelling & Software 37: 134–145.
  43. 43. Finney MA (1998) FARSITE: Fire Area Simulator-Model Development and Evaluation. United States Department of Agriculture Forest Service.
  44. 44. Cary GJ, Flannigan MD, Keane RE, Bradstock RA, Davies ID, et al. (2009) Relative importance of fuel management, ignition management and weather for area burned: Evidence from five landscape-fire-succession models. International Journal of Wildland Fire 18: 147–156.
  45. 45. Archibald S, Roy DP, Van Wilgen BW, Scholes RJ (2009) What limits fire? An examination of drivers of burnt area in Southern Africa. Global Change Biology 15: 613–630.
  46. 46. Bessie WC, Johnson EA (1995) The Relative Importance of Fuels and Weather on Fire Behavior in Subalpine Forests. Ecology 76: 747–762.
  47. 47. Arienti MC, Cumming SG, Boutin S (2006) Empirical models of forest fire initial attack success probabilities: the effects of fuels, anthropogenic linear features, fire weather, and management. Canadian Journal of Forest Research 36: 3155–3166.
  48. 48. Penman TD, Collins L, Price OF, Bradstock RA, Metcalf S, et al. (2013) Examining the relative effects of fire weather, suppression and fuel treatment on fire behaviour – A simulation study. Journal of Environmental Management 131: 325–333.
  49. 49. Syphard AD, Keeley JE, Bar Massada A, Brennan TJ, Radeloff VC (2012) Housing Arrangement and Location Determine the Likelihood of Housing Loss Due to Wildfire. PLoS ONE 7.
  50. 50. Price OF, Bradstock RA (2012) 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. Journal of Environmental Management 113: 146–157.
  51. 51. Collins BM, Kelly M, van Wagtendonk JW, Stephens SL (2007) Spatial patterns of large natural fires in Sierra Nevada wilderness areas. Landscape Ecology 22: 545–557.
  52. 52. Thompson JR, Spies TA (2009) Vegetation and weather explain variation in crown damage within a large mixed-severity wildfire. Forest Ecology and Management 258: 1684–1694.
  53. 53. Bradstock RA, Cohn JS, Gill AM, Bedward M, Lucas C (2009) Prediction of the probability of large fires in the Sydney region of south-eastern Australia using fire weather. International Journal of Wildland Fire 18: 932–943.
  54. 54. Minnich R, Chou Y (1997) Wildland Fire Patch Dynamics in the Chaparral of Southern California and Northern Baja California. International Journal of Wildland Fire 7: 221–248.
  55. 55. Keeley JE, Fotheringham CJ (2001) Historic Fire Regime in Southern California Shrublands. Conservation Biology 15: 1536–1548.
  56. 56. Blanchi R, Lucas C, Leonard J, Finkele K (2010) Meteorological conditions and wildfire-related houseloss in Australia. International Journal of Wildland Fire 19: 914–926.
  57. 57. Bradstock RA, Gill AM, Kenny BJ, Scott J (1998) Bushfire risk at the urban interface estimated from historical weather records: Consequences for the use of prescribed fire in the Sydney region of south-eastern Australia. Journal of Environmental Management 52: 259–271.
  58. 58. Keeley JE, Fotheringham CJ, Moritz MA (2004) Lessons from the October 2003 Wildfires in Southern California. Journal of Forestry Research (Harbin) 102: 26–31.
  59. 59. Bar Massada A, Radeloff VC, Stewart SI, Hawbaker TJ (2009) Wildfire risk in the wildland-urban interface: A simulation study in northwestern Wisconsin. Forest Ecology and Management 258: 1990–1999.
  60. 60. Syphard AD, Radeloff VC, Keuler NS, Taylor RS, Hawbaker TJ, et al. (2008) Predicting spatial patterns of fire on a southern California landscape. International Journal of Wildland Fire 17: 602–613.
  61. 61. Price OF, Bradstock RA (2013) The spatial domain of wildfire risk and response in the Wildland Urban Interface in Sydney, Australia. Natural Hazards and Environmental Management 1: 4539–4564.
  62. 62. Syphard AD, Radeloff VC, Keeley JE, Hawbaker TJ, Clayton MK, et al. (2007) Human influence on California fire regimes. Ecological Applications 17: 1388–1402.
  63. 63. Syphard AD, Bar Massada A, Butsic V, Keeley JE (2013) Land Use Planning and Wildfire: Development Policies Influence Future Probability of Housing Loss. PLoS ONE 8: e71708.
  64. 64. Price OF, Bradstock RA, Keeley JE, Syphard AD (2012) The impact of antecedent fire area on burned area in southern California coastal ecosystems. Journal of Environmental Management 113: 301–307.
  65. 65. Dennison PE, Moritz MA (2009) Critical live fuel moisture in chaparral ecosystems: a threshold for fire activity and its relationship to antecedent precipitation. International Journal of Wildland Fire 18: 1021–1027.
  66. 66. Dennison PE, Moritz MA, Taylor RS (2008) Evaluating predictive models of critical live fuel mois-ture in the Santa Monica Mountains, California. International Journal of Wildland Fire 17: 18–27.
  67. 67. Keeley JE, Zedler PH (2009) Large, high intensity fire events in southern California shrublands: debunking the fine-grained age-patch model. Ecological Applications 19: 69–94.
  68. 68. Penman TD, Bradstock RA, Price O (2013) Modelling the determinants of ignition in the Sydney Basin, Australia: implications for future management. International Journal of Wildland Fire 22: 469–478.
  69. 69. Hering A, Bell C, Genton M (2009) Modeling spatio-temporal wildfire ignition point patterns. Environmental and Ecological Statistics 16: 225–250.
  70. 70. Diaz-Avalos C, Peterson DL, Alvarado E, Ferguson SA, Besag JE (2001) Space-time modelling of lightning-caused ignitions in the Blue Mountains, Oregon. Canadian Journal of Forest Research-Revue Canadienne De Recherche Forestiere 31: 1579–1593.
  71. 71. Bar Massada A, Syphard AD, Hawbaker TJ, Stewart SI, Radeloff VC (2011) Effects of ignition location models on the burn patterns of simulated wildfires. Environmental Modelling & Software 26: 583–592.
  72. 72. Hirsch KG, Corey PN, Martell DL (1998) Using expert judgement to model initial attack fire crew effectiveness. Forest Science 44: 539–549.
  73. 73. Plucinski MP, McCarthy GJ, Hollis JJ, Gould JS (2012) The effect of aerial suppression on the containment time of Australian wildfires estimated by fire management personnel. International Journal of Wildland Fire 21: 219–229.
  74. 74. Gibbons P, van Bommel L, Gill AM, Cary GJ, Driscoll DA, et al. (2012) Land Management Practices Associated with House Loss in Wildfires. PLoS ONE 7: e29212.
  75. 75. Cohen JD (2004) Relating flame radiation to home ignition using modeling and experimental crown fires. Canadian Journal of Forest Research 34: 1616–1626.
  76. 76. Penman TD, Eriksen C, Blanchi R, Chladil M, Gill AM, et al. (2013) Defining adequate means of residents to prepare property for protection from wildfire. International Journal of Disaster Risk Reduction 6: 67–77.
  77. 77. 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 Safety Journal 61: 175–184.
  78. 78. Standards Australia (2009) A.S.3959–2009: Construction of buildings in bushfire prone areas. Sydney, Australia: SAI Global Limited. pp. 129.
  79. 79. CBC (2007) California Build Code Materials and construction methods for exterior wildfire exposure Title 24, part 2, volume 1 of 2. Sacramento, CA.