1. Introduction
Widely heralded in the media as Australia’s worst natural disaster (e.g., Rule 2009), the 7 February 2009 Black Saturday bushfires (wildfires) in Victoria were but the most recent reminder of the potential for natural hazards to impact Australian communities (Crompton and McAneney 2008). Fueled by record high temperatures and high winds in the midst of a protracted drought, the Black Saturday fires claimed 173 lives and 2298 houses (Victorian Bushfires Royal Commission 2009) as well as numerous other structures, including schools and police stations. This paper attempts to place these most recent bushfire impacts into a historical context.
Following a method analogous to Crompton and McAneney (2008) and other recent work (Bouwer 2010), this paper asks: What would have been the impact of past bushfires if they were to recur under current societal conditions? Without accounting for the known influence societal factors have on disaster records, it is impossible to know whether the devastation inflicted by the Black Saturday fires was truly anomalous, whether this provides a glimpse of the future under expected changes in climate, and what policy changes might prove effective in reducing the impact of future disasters. In examining such questions, we shall also reevaluate work undertaken before the Black Saturday fires (e.g., McAneney et al. 2009) and present some patterns of building destruction in these particular bushfires.
Despite claims that the Black Saturday fires were Australia’s greatest natural disaster (e.g., Rule 2009), several previous events have been more destructive in terms of loss of life and property damage, even before the societal influence has been accounted for: in 1974, Cyclone Tracy laid waste the city of Darwin, demolishing about 3700 dwellings and damaging another 3300 to the point that only 6% of the building stock was left habitable (Walker 1975); in 1899, Cyclone Mahina, a category-5 tropical cyclone, claimed about 410 lives; and the heat wave that preceded the 1939 Black Friday bushfires in Victoria is blamed for 438 deaths (from the research organization Risk Frontiers’s “PerilAUS” natural disaster database, described below in section 2a; Blong 2004; McAneney et al. 2009; Haynes et al. 2010). Of the more extreme bushfires, 1694 houses were lost in the 1983 Ash Wednesday fires in Victoria and South Australia (PerilAUS; Blong 2004; McAneney et al. 2009; Haynes et al. 2010), and, although we have been unable to verify this independently, 2000 buildings were reportedly destroyed in the 1898 Red Tuesday fires in Victoria (State Government of Victoria 2003, p. 10). Regardless of its ranking in terms of numbers of fatalities and property damage, the extreme impacts in the Black Saturday fires warrant critical examination. This same sentiment led the Victorian state government to form a royal commission with wide executive powers to scrutinize all aspects of bushfire management leading up to and during the bushfires (Victorian Bushfires Royal Commission 2009).
The process of adjusting time series of disaster losses for changes in population, wealth and inflation and, in some cases, improved construction standards is known as normalization and has been applied in a wide range of locales for a range of phenomena (e.g., Pielke and Landsea 1998; Pielke et al. 2008; Crompton and McAneney 2008; Zhang et al. 2009; Barredo 2009, 2010; Vranes and Pielke 2009). Accounting for inflation/deflation is necessary because the value of a currency changes over time while increases in population and wealth mean more people and property are located in exposed areas.
In respect to Australian bushfire, McAneney et al. (2009) argued that the stability over the last century of exceedance loss statistics for building damage suggested that it was premature to conclude that a signal of greenhouse-gas emissions was present. The authors contend that, given that these loss statistics had proved so stable in the face of the vast societal changes that took place over the twentieth century, any greenhouse-gas signal cannot be large or significant. This study revisits this question using a different approach by explicitly accounting for these societal changes.
Whereas a greenhouse gas–driven climatic-change signal has thus far not been detected in normalized disaster loss records for a wide range of perils in locations around the world [see review by Bouwer (2010) and references therein], and is unlikely to be detected in at least storm and flood losses in the near future (Höppe and Pielke 2006), patterns of behavior characteristic of meteorological cycles such as El Niño–Southern Oscillation (ENSO) have been identified in normalized Atlantic Ocean hurricane damages (Pielke and Landsea 1999). ENSO and another coupled ocean–atmosphere oscillation, the Indian Ocean dipole (IOD), are also known to influence the weather and climate of eastern Australia (McBride and Nicholls 1983; Power et al. 2006; Ashok et al. 2003; Cai et al. 2009b); the former oscillation is in the equatorial Pacific Ocean and the latter in the Indian Ocean.
An El Niño (La Niña) phase of the ENSO cycle refers to the situation in which sea surface temperatures in the central to eastern Pacific Ocean are significantly warmer (cooler) than the long-term average, whereas a positive IOD (pIOD) event is when the eastern Indian Ocean is cooler than normal and the western Indian Ocean is anomalously warmer (Saji et al. 1999). El Niño events increase the chance of drought along eastern Australia (Kiem and Franks 2004) and bushfire (Williams and Karoly 1999), whereas La Niña events often presage widespread increases in rainfall (Power et al. 2006) and chance of flooding (Kiem et al. 2003). Ummenhofer et al. (2009) showed that a lack of negative IOD (nIOD) events was strongly related to drought in southeastern Australia, and Cai et al. (2009a) report a link between pIOD events and enhanced bushfire risk over Victoria. Moreover, Cai et al. (2009a) found that pIOD events were more effective than El Niño events in preconditioning Victorian bushfires, a robust result that was not conditional on the definitions adopted for each. This paper will examine the relationships between ENSO and the IOD and normalized bushfire building damage in Australia.
The remainder of this paper is structured as follows: we begin with a description of Risk Frontiers’ PerilAUS inventory of Australian bushfire building damage and the bushfire fatality database of Haynes et al. (2010). The normalization methods, ENSO and IOD definitions, and the method used to examine patterns of building damage in the Black Saturday fires for two of the most severely impacted towns, Marysville and Kinglake, are then detailed. We then present key results, including those from two historic case studies (the 1967 Hobart fires and the 1983 Ash Wednesday fires) used to “ground truth” the normalization method. The paper concludes with a discussion of results and some implications for public policy with regard to bushfire in Australia.
2. Data and methods
a. Bushfire building damage and fatality data
The current study draws upon Risk Frontiers’ databases of natural disasters in Australia (hereinafter referred to as PerilAUS). Data entries were derived mainly from archival searches of the Sydney Gazette and Sydney Morning Herald (dating from 1803 and 1831, respectively) and were cross referenced against other local newspapers and official documents in other states or territories where necessary and other reference material where available (Coates 1996; Blong 2004; Haynes et al. 2010). In the case of bushfire and except for some years prior to 1926 for which data are incomplete, it provides a comprehensive national record of Australia’s loss events. Although it is expected that any bushfire that resulted in significant building damage and numbers of fatalities has already been cataloged, the database is constantly being improved and revalidated. In the course of this study, further events, including a bushfire that destroyed 454 houses in Victoria in 1962 (State Government of Victoria 2003, p. 10), were identified and added to the record. These additional events and those from more recent bushfire seasons were not included in the McAneney et al. (2009) analysis.
For almost 1200 events listed in PerilAUS, it is possible to estimate the number of buildings destroyed, with damaged buildings (residential, public, commercial and industrial, etc.) converted to house equivalents HE using relative building costs and floor areas for different types of buildings (Blong 2003). One HE can correspond to the complete destruction of one median-sized house, two such houses each 50% destroyed or, for example, a suburban police station experiencing damage amounting to 47% of its replacement value. McAneney et al. (2009) note that most outcomes from bushfires tend to be binary in nature, with buildings either being completely destroyed or surviving relatively unscathed. Damage to building contents, cars, machinery, aircraft, crops, and so on is not included in the HE estimates.
In addition to building damage information, PerilAUS also contains details of bushfire-related fatalities, including names of the deceased. This information was used by Haynes et al. (2010) as the entry point to forensic, witness, and police statements contained in coroners’ inquest reports for each known death from 1901 to 2007/08. An outcome of that study was a database of civilian bushfire fatalities; our study will analyze those entries over a common time horizon with building damage for bushfire years 1925–2008. The definition of bushfire years, in which 1925 represents the 12-month period beginning 1 July 1925, reflects the Southern Hemisphere bushfire season.
b. Normalizing house equivalents
Equation (1) ignores any explicit correction for inflation and wealth as measured in economic terms. The HE representation avoids the need for an inflation adjustment; whether an adjustment for increasing economic wealth is required is less obvious. An argument for its inclusion stems from the manifest increase in the average size of Australian dwellings over time: for example, the average increase in the average number of bedrooms per dwelling between 1976 and 2006 was 0.3% per year (from the ABS data). If this rate of increase had held constant over the entire analysis period, then the average dwelling size would have increased by 28% between 1925 and 2008. On the other hand, we expect most of that increase has been implicitly accounted for in the manner by which the HE data were derived: if, by way of example, we imagine a hypothetical bushfire event in which 100 houses were destroyed, then we assume that this equates to 100 HE whether the event occurred in 1930 or 1990. Although Blong (2003) differentiates between small, median, and large houses based on floor area, this level of detail is not often included in the source documents and so, for most building types, numbers of HE were based on a single (median) size of each building type. This being the case, we have chosen not to further adjust the HE data for changes in wealth; however, any adjustment of economic losses would also require both an inflation and economic wealth adjustment.
c. Normalizing fatalities
d. Validation of normalization methods
Equations (1) and (2) assume that growth in the exposure—number of bushfire-prone dwellings and population in the areas impacted—occurred at the same rate as the growth in total number of dwellings and population for each state or territory. Except for a few particular bushfires, data are not available to allow a more precise estimate of growth in exposed areas over time.
We can get some sense of the relative accuracy of this assumption by comparing state/territory-based dwelling number and population event factors with those derived by weighting equivalent local-level factors by each local area’s proportional contribution to event building damage and fatalities. Urban center/locality (UCL)-based factors were calculated for two of the most damaging historical bushfires: the 1967 Hobart fires and 1983 Ash Wednesday fires. Although UCL-level growth may not necessarily mirror bushfire-prone dwelling and population growth, we expect this to be a more accurate representation than the state/territory-level figures.
The UCL structure is one of the seven interrelated classification structures of the Australian standard geographical classification that groups census collection districts together to form areas defined according to population size (from the ABS). In broad terms, an urban center is a population cluster of 1000 or more people, and a locality comprises a cluster of between 200 and 999 people. The number of dwellings and population in each UCL is reported in census years in the census of population and housing (available from the ABS).
e. ENSO and IOD
There exist multiple definitions for the El Niño and La Niña phases of ENSO, based upon either the Southern Oscillation index (SOI) or various sea surface temperature (SST)-based metrics, but these generally concur for the major El Niño and La Niña events. Here we adopt the Japan Meteorological Agency (JMA) index of 5-month running mean of spatially averaged SST anomalies over a region of the tropical Pacific (4°S–4°N, 150°–90°W). An ENSO year from October through to the following September is then categorized as El Niño (La Niña) if JMA index values are 0.5°C (−0.5°C) or greater (less) for at least 6 consecutive months (including October, November, and December). All other years are classified as neutral. The JMA index for the post-1949 period is based on observed data and, for the years 1925–48, upon reconstructed monthly mean SST fields (Meyers et al. 1999).
In a similar way, IOD events are definition dependent, and we adopt that of Cai et al. (2009a). They define an event using an index of the IOD called the dipole mode index (Saji et al. 1999) in spring (September, October, and November), referenced to the climatological mean over the period 1880–2008. A pIOD (nIOD) event occurs when the index is greater (less) than 0.75 of its long-term standard deviation. Cai et al. (2009a) focused on the spring season as this is when pIOD events peak, and they relate the classification to the following summer season (December, January, and February). All other years are classified as neutral.
The above ENSO and IOD definitions correspond for the worst months for Australian bushfire impacts—December, January, and February—with bushfire years defined earlier as 12-month periods starting 1 July. Classifications according to the above definitions are given in Table 1 (see http://coaps.fsu.edu/jma.shtml). Cai et al. (2009c) noted the recent high frequency of pIOD events, with five occurring during 2002–08.
f. Post–Black Saturday observations
After the Black Saturday fires, Risk Frontiers undertook an aerial reconnaissance for the Kinglake area and Melbourne’s northeastern suburbs, which are interfaced with extensive bushland. On-the-ground surveys were not possible at the time (11 February 2009), with access to many of the impacted areas prohibited while police conducted crime-scene investigations.
Quantitative damage analysis focused on Marysville and Kinglake, the two towns most severely damaged. The main aim was to reveal the spatial pattern of destroyed properties in relation to distance from surrounding bushland boundaries. A Melbourne-based company, Airtech (http://www.airtechaust.com/), provided 15-cm-resolution, georeferenced postfire imagery captured on 22 and 24 March 2009. These images were manually interpreted, and locations of a total of 1156 destroyed buildings and other surviving structures were digitized. For the distribution and extent of prefire bushland, we performed various supervised image classifications with the 2.5-m-resolution, orthorectified imagery in the 2009 SPOTMaps series (http://access.spot.com/). It was possible to reliably evaluate the best classification results given the fine resolution of imagery employed and the relative small size of the study area. Once the locations of buildings and bushland boundaries were known, we then calculated distance-based statistics relevant to land use planning and insurance pricing.
3. Results
a. Case studies
We first test the legitimacy of our assumptions that HE and fatalities have increased in proportion to the state/territory-level increase in the total numbers of dwellings and population. Table 2 shows state-based dwelling and population factors (as defined previously) for the 1967 Hobart and 1983 Ash Wednesday bushfires as well as UCL-weighted event factors, calculated by weighting UCL dwelling and population factors by their relative contribution to the total event HE and fatality numbers.
For the Hobart fires, state dwelling and population factors closely mimic their UCL-weighted equivalents (Table 2). Seven UCLs were used to calculate the weighted dwelling factor, with a 60% weight given to the Hobart UCL factor. Only the Hobart UCL was used for the weighted population factor as all 64 fatalities occurred there. The closeness of the state and UCL-weighted factors is not surprising given the size of Hobart as compared with the rest of Tasmania—in 2008 the population in the Hobart UCL stood at approximately one-quarter of the total for the entire state (from the ABS). In other words, the state-based figures are also highly weighted toward Hobart.
In contrast to the Hobart fires, 12 UCLs were used to calculate the weighted dwelling factor for the Ash Wednesday fires, with no one contributing more than 22% of the total building damage. Similar treatment was used for the weighted population factor, where nine UCLs were used with a weighting of not more than 19% applied to each of the contributing UCL factors. All of the UCLs impacted by this event were small relative to the state in which they are located, and the differences are greater than was the case for Hobart, with both state-level factors underestimating growth in dwellings and population at the local level (Table 2).
Table 2 gives some confidence that, while it is possible for state- and local-level normalization factors to diverge, the variation does not appear to be systematic; if anything, it provides some indication that our assumption may be conservative as state-level data may underestimate dwelling growth in exposed areas. It was not possible to derive UCL-weighted factors for the entire analysis period as the UCL structure did not exist prior to the 1966 census (from the ABS).
Table 3 compares the Hobart and Ash Wednesday UCL-weighted normalized HE and fatalities with the data recorded for these events together with those experienced on Black Saturday. The key observation is that the Black Saturday death toll appears to be an aberration: after normalizing the data, the ratio of fatalities to building damage in the Black Saturday fires is more than double that for Hobart and Ash Wednesday. We note that the normalization factors are different for HE and fatalities and so the values shown in the final column in Table 3 are not just a simple arithmetic ratio of the recorded data.
b. Time series of building damage and fatalities
Figures 1a and 2a show time series of the annual aggregated bushfire HE and fatalities recorded in PerilAUS and the Haynes et al. (2010) database for bushfire years 1925–2008; Figs. 1b and 2b present the corresponding normalized values. Regression analysis on the recorded data reveals increasing trends (Figs. 1a and 2a), more pronounced in the case of Fig. 1a—trends that are reversed in the normalized data (Figs. 1b and 2b). None of these trends are statistically significant at the 10% level, and the overriding impression is of a time series that is dominated by occasional extreme excursions from the mean.
The average annual normalized HE over all years is 301 (Table 4), and the equivalent figure for fatalities is 14. The former is some 3.6 times that determined by McAneney et al. (2009), a difference that arises primarily from normalizing the data. Other factors that influence this difference are 1) the inclusion of other events that had not been previously identified, 2) extending the analysis period to include Black Saturday, and 3) beginning the analysis in 1925 rather than in 1900—the years between 1900 and 1926 for which data exist being characterized by low levels of building damage.
The seemingly anomalous loss of life in the Black Saturday fires and 2008 bushfire year is subject to further scrutiny in Fig. 3, which shows the ratio of annual aggregate normalized fatalities [Eq. (2)] to normalized HE [Eq. (1)] for those bushfire years for which the normalized HE is greater than 600. Adoption of a 600-HE threshold, which conveniently reduces the data to the 10 most damaging years, was done simply to eliminate those years with little or no building damage and/or few or no fatalities. The generally decreasing pattern in Fig. 3 over time is broadly insensitive to the threshold of building damage adopted: a very similar pattern is revealed if a threshold of 100 HE is applied. Of the 10 most damaging years, not since close to the beginning of the analysis period, the 1938 bushfire year, has there been a higher ratio of normalized fatalities to building damage (Fig. 3) than in the 2008 bushfire year. The ratio of total normalized fatalities to HE over the entire analysis period is 4.7%.
c. ENSO and IOD relationship with normalized bushfire building damage
Table 4 shows the median, average, and standard deviation (of normalized HE) over the 84-yr study period for ENSO and IOD classified years. There are distinct differences in the median annual and average annual normalized HE for El Niño and La Niña years as there are for pIOD and nIOD years. As expected, the average building damage is highest in El Niño and pIOD years and the median damage in La Niña and nIOD years is zero. The distribution of damage over all years is highly skewed.
The relationship between ENSO and bushfire building damage is reasonably robust, although the strength of the relationship is weakened if an alternative SOI-based definition is applied (Table 4). Under the SOI definition, an El Niño (La Niña) year occurs when the average of June–December monthly SOI values is less (greater) than −5 (5) (S. Power 2009, personal communication). The difference between the average building damage in El Niño and La Niña years is substantially reduced under the SOI definition but the median in La Niña years is still zero.
It is important to note the statistics in Tables 4 and 5 are sensitive to building damage in the most destructive of bushfire years. The 10 most damaging years in terms of normalized HE account for almost 80% of total normalized damage, and the ENSO classification of only one of these bushfire years (2008—the fourth largest) differs between the two ENSO definitions: using the SST definition, the 2008 bushfire year is classified as neutral (Table 1) whereas under the SOI definition, it is categorized as La Niña.
Tables 4 and 5 suggest that the IOD is more discriminating than ENSO in relation to normalized bushfire building damage in Australia (the SST definition of ENSO was used in Table 5). The two most damaging combined phases are pIOD/neutral and pIOD/El Niño, which together make up 17 of the 84 bushfire years in the study period (Table 5). The 1938 bushfire year (La Niña year, neutral IOD year) is the only example in which extreme building damage (normalized HE > 1000) occurred in either a La Niña or nIOD year.
As pointed out earlier, bushfire years (beginning 1 July), ENSO years (beginning 1 October), and IOD years (relating to the summer season) do not completely overlap. The effect of this difference is negligible as less than 0.2% of the total normalized HE occurred in the months from 1 July to 30 September inclusive and almost 95% of the total normalized building damage occurred during summer (December, January, February).
d. Post–Black Saturday analysis: Kinglake and Marysville
Destroyed buildings in Kinglake and Marysville were categorized as a function of distance from bushland boundaries, and these data are presented in Fig. 4. A key feature is that about 25% of destroyed buildings were located physically within the bushland boundary, and 60% and 90% were within 10 and 100 m of bushland (Fig. 4). Most buildings in Marysville lay within 200 m of the bushland boundary and, given the wind change that occurred early in the evening on 7 February 2009, would have been subject to ember attack from multiple directions (Victorian Bushfires Royal Commission 2009).
4. Discussion
In assuming bushfire-related building damage and fatalities change in proportion to dwelling numbers and population, Eqs. (1) and (2) estimate the number of HE and fatalities in a given event had it occurred under 2008 bushfire year societal conditions. There are other factors not accounted for in the normalization methods, although we expect their influence, particularly on the building damage record, to be minimal relative to societal change. For example, it is likely that some historical bushfires occurred in what were formerly unpopulated areas and thus would have registered no building damage, whereas in these same areas large losses may now be possible. The opposite is also true where original bushlands have been converted to suburbs so that some historical bushfire impacts may now be physically impossible.
Haynes et al. (2010) suggest that a reduction, over time, in the number of people living and working in isolated rural locations would have influenced the fatality data. The effect of this shift was evident in the decreased number of fatalities due to late evacuation, the most common activity at time of death (Haynes et al. 2010). More specifically, Haynes et al. (2010) found a marked decline in those who died while evacuating from working outside and they concluded that this in part explained the absence of a trend in the fatality data (prior to the Black Saturday fires) despite considerable population growth. Notwithstanding these and other qualifications, Figs. 1b and 2b show our best estimates of normalized bushfire building damage and fatalities.
Is the normalized building damage realistic? The average nominal value of a new house (excluding land) in Australia in the 2008 bushfire year was approximately AU$260,000 (from the ABS) so that, in dollar loss terms, the average annual building damage of 301 HE (Table 4) equates to AU$78 million. As noted earlier, this amount excludes building contents, cars, and so on and so will underestimate the property loss, but it does include both insured and uninsured building damage. From an independent dataset, but using a conceptually similar normalization method, Crompton and McAneney (2008) found the average annual insured property loss from weather-related natural disasters between 1967 and 2006 to be around AU$820 million (in 2006 dollars), of which about 12%, or AU$98 million, can be attributed to bushfire. Despite the stated differences, the closeness of these two independent estimates provides some confidence in the method and results. The relationship between normalized building damage and ENSO and the IOD provides additional confidence.
Similar to the result of Cai et al. (2009a) we found normalized Australian bushfire building damage to be more strongly related to the IOD than to ENSO. This is unsurprising as Cai et al. (2009a) follow the Ellis et al. (2004) definition of a significant bushfire, and this incorporates historical impacts (fatalities, property, and livestock losses) rather than meteorological variables or indices. The significant Victorian summer bushfire seasons that the Cai et al. (2009a) study is based upon are therefore correlated with years of high normalized Australian building damage (61% of the total normalized HE occurred in Victoria), at least over the common time period since 1950.
The Black Saturday fires rank fourth in terms of normalized building damage. There were 173 fatalities, more than double the recorded number in any other bushfire event over the analysis period. After normalization, the Black Saturday death toll ranks second to the 1939 Black Friday fires with 214 normalized fatalities. In other words, history suggests that even larger impacts are possible under the climate of past decades. However, this ranking should not detract from the extreme impacts and high ratio of normalized fatalities to building damage in the Black Saturday fires and the need for policy changes to reduce the likelihood of this happening again.
One unequivocal result from our analyses is the absence of any significant trend in normalized HE over time (Fig. 1b). This being the case, a reasonable conclusion at this time, consistent with similar studies summarized by Bouwer (2010), is that it is not possible to detect a greenhouse-gas climatic-change signal in the time series of Australian bushfire building damage once it has been normalized. Such an influence is not ruled out by our analysis, but, if it does exist, it is clearly dwarfed by the magnitude of the societal change and the large year-to-year variation in impacts. Moreover it seems highly implausible that the net effect of other factors such as changes in bushfire risk management is being exactly balanced by a greenhouse gas–driven climatic-change influence.
5. Policy implications
This study has shown that increasing building damage due to bushfire in Australia is largely being driven by increasing dwelling numbers and that the impact of greenhouse gas–driven climatic change is not detectable at this time. With this in mind, to reduce the impact of future bushfire events, investments to reduce societal vulnerability need to be made and are likely to bring immediate benefit. Adaptation should be undertaken concurrently with mitigation so that success in addressing bushfire risk in Australia in the short term at least is not misunderstood in terms of obtaining global agreement on reduction of greenhouse-gas emissions.
The Black Saturday tragedy occurred in the face of significant investments (Ashe et al. 2009) and improvements in bushfire risk management and suppression. We take the view that the extreme property losses were in part related to the close proximity of many dwellings to bushland. Chen and McAneney (2004) showed that, although distance to bushland is not the only variable determining bushfire vulnerability, it is demonstrably the most important, with the probability of home destruction decreasing strongly as a function of this distance, a result interpreted as being indicative of ember density and flammability. In the towns of Kinglake and Marysville, where the majority of building damage occurred in the Black Saturday fires, we have shown that 25% of destroyed buildings were literally located within bushland and that 60% were within 10 m of the bushland boundary. Under the extreme conditions prevailing on that day, it is difficult to imagine that homes in the flame zone could have been successfully defended against the combined threats of flames, radiant and convective heating, and embers.
The “prepare, stay, and defend or leave early” bushfire policy, adopted by all Australian fire authorities at the time of the Black Saturday fires, arose on the basis of concerns about the likelihood of large losses of life occasioned by late evacuation (Handmer and Tibbits 2005; Haynes et al. 2010) and the perceived impracticability of evacuating large numbers of people every time severe bushfire conditions exist, circumstances that might arise in some years and some parts of the country for much of the summer. The policy put the actions of residents as central in the protection of lives and property and has the commendable attribute of discouraging an unwarranted dependence upon emergency services. On the other hand, an incorrect interpretation of this policy during Black Saturday may also have contributed to a mistaken belief that homes constructed within or in close proximity to the bushland could be successfully defended against bushfires in extreme conditions.
Our results raise serious questions about land use planning in Australia in relation to bushfire risk. The comparison of state- and UCL-level normalization factors (Table 2) is cause for further concern as it suggests dwelling growth in areas of high bushfire risk may be occurring faster than state averages. We echo the sentiments of McAneney et al. (2009) that, without changes in policy, particularly in land use planning, further bushfire catastrophes are inevitable.
REFERENCES
Ashe, B., McAneney K. J. , and Pitman A. J. , 2009: Total cost of fire in Australia. J. Risk Res., 12 , 121–136.
Ashok, K., Guan Z. , and Yamagata T. , 2003: Influence of the Indian Ocean dipole on the Australian winter rainfall. Geophys. Res. Lett., 30 , 1821. doi:10.1029/2003GL017926.
Barredo, J. I., 2009: Normalised flood losses in Europe: 1970–2006. Nat. Hazards Earth Syst. Sci., 9 , 97–104.
Barredo, J. I., 2010: No upward trend in normalised windstorm losses in Europe: 1970–2008. Nat. Hazards Earth Syst. Sci., 10 , 97–104.
Blong, R., 2003: A new damage index. Nat. Hazards, 30 , 1–23.
Blong, R., 2004: Residential building damage and natural perils: Australian examples and issues. Build. Res. Inf., 32 , 379–390.
Bouwer, L. M., 2010: Have disaster losses increased because of anthropogenic climate change? Bull. Amer. Meteor. Soc., in press.
Cai, W., Cowan T. , and Raupach M. , 2009a: Positive Indian Ocean dipole events precondition southeast Australia bushfires. Geophys. Res. Lett., 36 , L19710. doi:10.1029/2009GL039902.
Cai, W., Cowan T. , and Sullivan A. , 2009b: Recent unprecedented skewness towards positive Indian Ocean dipole occurrences and its impact on Australian rainfall. Geophys. Res. Lett., 36 , L11705. doi:10.1029/2009GL037604.
Cai, W., Pan A. , Roemmich D. , Cowan T. , and Guo X. , 2009c: Argo profiles a rare occurrence of three consecutive positive Indian Ocean dipole events, 2006–2008. Geophys. Res. Lett., 36 , L08701. doi:10.1029/2008GL037038.
Chen, K., and McAneney K. J. , 2004: Quantifying bushfire penetration into urban areas in Australia. Geophys. Res. Lett., 31 , L12212. doi:10.1029/2004GL020244.
Coates, L., 1996: An overview of fatalities from some natural hazards in Australia. Proc. Conf. on Natural Disaster Reduction, Queensland, Australia, Institution of Engineers, Australia, 49–54.
Crompton, R. P., and McAneney K. J. , 2008: Normalised Australian insured losses from meteorological hazards: 1967–2006. Environ. Sci. Policy, 11 , 371–378.
Ellis, S., Kanowski P. , and Whelan R. , 2004: National inquiry on bushfire mitigation and management. Commonwealth of Australia, 451 pp. [Available online at http://www.coagbushfireenquiry.gov.au/findings.htm#downloading].
Handmer, J., and Tibbits A. , 2005: Is staying at home the safest option during bushfires? Historical evidence for an Australian approach. Global Environ. Change, 6 , 81–91.
Haynes, K., Handmer J. , McAneney J. , Tibbits A. , and 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 (3) 185–194.
Höppe, P., and Pielke R. A. Jr., Eds. 2006: Workshop on climate change and disaster losses: Understanding and attributing trends and projections. Final Workshop Rep., Hohenkammer, Germany, 266 pp. [Available online at http://sciencepolicy.colorado.edu/sparc/research/projects/extreme_events/munich_workshop/workshop_report.html].
Kiem, A. S., and Franks S. W. , 2004: Multi-decadal variability of drought risk, eastern Australia. Hydrol. Processes, 18 , 2039–2050.
Kiem, A. S., Franks S. W. , and Kuczera G. , 2003: Multi-decadal variability of flood risk. Geophys. Res. Lett., 30 , 1035. doi:10.1029/2002GL015992.
McAneney, J., Chen K. , and Pitman A. , 2009: 100 years of Australian bushfire property losses: Is the risk significant and is it increasing? J. Environ. Manage., 90 , 2819–2822.
McBride, J. L., and Nicholls N. , 1983: Seasonal relationships between Australian rainfall and the Southern Oscillation. Mon. Wea. Rev., 111 , 1998–2004.
Meyers, S. D., O’Brien J. J. , and Thelin E. , 1999: Reconstruction of monthly SST in the tropical Pacific Ocean during 1868–1993 using adaptive climate basis functions. Mon. Wea. Rev., 127 , 1599–1612.
Pielke, R. A. Jr., and Landsea C. W. , 1998: Normalized hurricane damages in the United States: 1925–95. Wea. Forecasting, 13 , 621–631.
Pielke, R. A. Jr., and Landsea C. W. , 1999: La Niña, El Niño, and Atlantic hurricane damages in the United States. Bull. Amer. Meteor. Soc., 80 , 2027–2033.
Pielke, R. A. Jr., Rubiera J. , Landsea C. , Fernandez M. L. , and Klein R. , 2003: Hurricane vulnerability in Latin America and the Caribbean: Normalized damage and loss potentials. Nat. Hazards Rev., 4 , 101–114.
Pielke, R. A. Jr., Gratz J. , Landsea C. W. , Collins D. , Saunders M. , and Musulin R. , 2008: Normalized hurricane damage in the United States: 1900–2005. Nat. Hazards Rev., 9 , 29–42.
Power, S. B., Haylock M. , Colman R. , and Wang X. , 2006: The predictability of interdecadal changes in ENSO activity and teleconnections. J. Climate, 19 , 4755–4771.
Rule, A., 2009: Our darkest day. The Age, 9 February 2009. [Available online at http://www.theage.com.au/national/our-darkest-day-20090208-810q.html].
Saji, N. H., Goswami B. N. , Vinayachandran P. N. , and Yamagata T. , 1999: A dipole mode in the tropical Indian Ocean. Nature, 401 , 360–363.
State Government of Victoria 2003: Report of the inquiry into the 2002–2003 Victorian bushfires. State Government of Victoria, Victoria, Australia, 370 pp.
Ummenhofer, C. C., England M. H. , McIntosh P. C. , Meyers G. A. , Pook M. J. , Risbey J. S. , Gupta A. S. , and Taschetto A. S. , 2009: What causes southeast Australia’s worst droughts? Geophys. Res. Lett., 36 , L04706. doi:10.1029/2008GL036801.
Victorian Bushfires Royal Commission cited. 2009: Interim report. Parliament of Victoria, Victoria, Australia. [Available online at http://www.royalcommission.vic.gov.au/getdoc/208e8bcb-3927-41ec-8c23-2cdc86a7cec7/Interim-Report].
Vranes, K., and Pielke R. Jr., 2009: Normalized earthquake damage and fatalities in the United States: 1900–2005. Nat. Hazards Rev., 10 , 84–101.
Walker, G. R., 1975: Report on Cyclone Tracy: Effects on buildings, December 1974. Australian Department of Housing and Construction, Melbourne, Australia, 100 pp.
Williams, A. A. J., and Karoly D. J. , 1999: Extreme fire weather in Australia and the impact of the El Niño–Southern Oscillation. Aust. Meteor. Mag., 48 , 15–22.
Zhang, Q., Liu Q. , and Wu L. , 2009: Tropical cyclone damages in China 1983–2006. Bull. Amer. Meteor. Soc., 90 , 489–495.
(a) Annual aggregate HE for bushfire events in PerilAUS for years beginning 1 Jul; (b) as in (a) but with HE normalized to 2008 bushfire year values.
Citation: Weather, Climate, and Society 2, 4; 10.1175/2010WCAS1063.1
(a) Annual aggregate HE for bushfire events in PerilAUS for years beginning 1 Jul; (b) as in (a) but with HE normalized to 2008 bushfire year values.
Citation: Weather, Climate, and Society 2, 4; 10.1175/2010WCAS1063.1
(a) Annual aggregate HE for bushfire events in PerilAUS for years beginning 1 Jul; (b) as in (a) but with HE normalized to 2008 bushfire year values.
Citation: Weather, Climate, and Society 2, 4; 10.1175/2010WCAS1063.1
(a) Annual aggregate fatalities for bushfire events in the Haynes et al. (2010) database for years beginning 1 Jul; (b) as in (a) but with fatalities normalized to 2008 bushfire year values.
Citation: Weather, Climate, and Society 2, 4; 10.1175/2010WCAS1063.1
(a) Annual aggregate fatalities for bushfire events in the Haynes et al. (2010) database for years beginning 1 Jul; (b) as in (a) but with fatalities normalized to 2008 bushfire year values.
Citation: Weather, Climate, and Society 2, 4; 10.1175/2010WCAS1063.1
(a) Annual aggregate fatalities for bushfire events in the Haynes et al. (2010) database for years beginning 1 Jul; (b) as in (a) but with fatalities normalized to 2008 bushfire year values.
Citation: Weather, Climate, and Society 2, 4; 10.1175/2010WCAS1063.1
Ratio of annual aggregate normalized fatalities [Eq. (2)] to HE [Eq. (1)] for years (beginning 1 Jul) in which normalized building damage exceeded 600 HE.
Citation: Weather, Climate, and Society 2, 4; 10.1175/2010WCAS1063.1
Ratio of annual aggregate normalized fatalities [Eq. (2)] to HE [Eq. (1)] for years (beginning 1 Jul) in which normalized building damage exceeded 600 HE.
Citation: Weather, Climate, and Society 2, 4; 10.1175/2010WCAS1063.1
Ratio of annual aggregate normalized fatalities [Eq. (2)] to HE [Eq. (1)] for years (beginning 1 Jul) in which normalized building damage exceeded 600 HE.
Citation: Weather, Climate, and Society 2, 4; 10.1175/2010WCAS1063.1
Cumulative distribution of buildings destroyed in relation to distance from nearby bushland. The number of samples for Marysville and Kinglake is 540 and 616, respectively.
Citation: Weather, Climate, and Society 2, 4; 10.1175/2010WCAS1063.1
Cumulative distribution of buildings destroyed in relation to distance from nearby bushland. The number of samples for Marysville and Kinglake is 540 and 616, respectively.
Citation: Weather, Climate, and Society 2, 4; 10.1175/2010WCAS1063.1
Cumulative distribution of buildings destroyed in relation to distance from nearby bushland. The number of samples for Marysville and Kinglake is 540 and 616, respectively.
Citation: Weather, Climate, and Society 2, 4; 10.1175/2010WCAS1063.1
Years 1925–2008 (each beginning 1 Jul) identified as either El Niño or La Niña (ENSO), and pIOD or nIOD (IOD). Other years are classified as neutral for each oscillation.
State and UCL-weighted dwelling and population normalization factors for the 2008 bushfire year.
Recorded and UCL-weighted normalized HE and fatalities in the 1967 Hobart, 1983 Ash Wednesday, and 2009 Black Saturday fires.
Summary statistics by ENSO and IOD phase for annual aggregate normalized building damage for years 1925–2008 (each beginning 1 Jul). Numbers in parentheses were derived using the SOI definition of ENSO.
Average annual normalized HE by ENSO (SST definition) and IOD phase for years 1925–2008 (each beginning 1 Jul). Numbers in parentheses are the number of years on which the average is based.
