Published works

This study examines building damage as recorded in PerilAUS to determine the national profile of natural peril impacts and frequencies. The analysis employs Risk Frontiers’ Damage Index based on a House Equivalent (HE) loss metric introduced by Blong (2003); a normalisation correction based on Crompton and McAneney (2008) and Crompton et al. (2010); and a lower bound event threshold of 25 normalised HE. The latter is equivalent to a monetary loss of around $10m in 2015-16. Normalisation puts historical events on a common footing with losses that would be incurred given 2015 societal and demographic conditions. It answers the question: what would be the building losses if historic events were to recur today?

While more validation and analysis remains to be done, we are confident that the relative importance of the various perils and their spatial distribution across states and territories has been faithfully captured. Broadly we find that there have been on average 5.85 events per year causing losses in excess of 25 normalised HE. This frequency exhibits no statistically significant change since 1900. The mean loss per event is $118m with a standard deviation of $430m.

The most costly event in terms of building damage is the 1999 Sydney hailstorm, which was also the most expensive insured loss. The losses broadly follow a Pareto distribution in which 20% of events account for 80% of the aggregated normalised building losses and the top 20 are responsible for 50% of those losses. We can expect natural disaster events as costly as the 1999 Sydney hailstorm to occur about once per century, events like the Brisbane floods once every 30 to 40 years and that of the Hobart Bushfires about once a decade.

Just why most of the extreme losses – the 1974 Brisbane floods and Cyclone Tracy in the same calendar year, the 1999 Sydney hailstorm and the 2011 Queensland and Victorian floods – are clustered post-1970 requires further investigation. We do not believe this to be a reporting bias and know in the case of floods, for example, that Brisbane experienced much higher floods in the early and late 19th century than either the 1974 and 2011 floods. Regardless of the reason, the pattern of losses demonstrates clearly the ‘heavy-tailed’ character of the distribution of natural peril losses: in other words, there is always the possibility of event losses far in excess of the historical mean. This may be occur because of an event of higher intensity or larger footprint, that footprint impacting an area of higher-valued exposure, or all of these together.

Of all the perils, tropical cyclones have been most destructive and responsible for 30% of the national building damage since 1900. Bushfires, floods and hail have all been similarly costly each accounting for another 18% of building losses, although when hailstorms are combined with other storm events (excluding cyclones), thunderstorms similarly contribute 30% of the losses. Compared with meteorological hazards, geophysical perils have had a minor influence on building damage over the last 116 years with earthquake losses dominated by a single event -- the 1989 Newcastle earthquake. However this time period is too short to predict the frequency of damaging seismic events and in the case of this peril, as with some others, the spatial pattern of losses shown here could be overturned by another extreme loss.