In 1999 NHRC completed production of PerilAUS I. These spatial databases focus on nine natural perils – tropical cyclones, floods, bushfires, wind gusts, hail falls, earthquakes, tornadoes, landslides and tsunamis. PerilAUS includes nearly 5,000 events in the last 100 years. A Damage Index has been estimated for almost 1,200 events, with total damage to buildings equivalent to the complete destruction of 38,300 houses.
With the enormous task of assembling these databases complete, at least to the stage where the information can be utilised, PerilAUS II uses these data to develop relative risk ratings for each of the nine perils for each of the 2,573 postcodes in Australia. The map presented with this issue of NHQ provides one view of the spatial distribution of natural perils and relative risk on a postcode basis.
Given the data available, there are innumerable ways of assembling maps of natural perils. We have chosen to combine the historical data in PerilAUS I with various maps of natural hazards potential. Our focus is on damage to buildings – both past and future damage. Here we represent the risk from the nine natural perils on a postcode-by-postcode basis – an appropriate format for insurance industry purposes.
Figure 1 illustrates the framework for analysing and combining PerilAUS I and Potential Hazards data.
|click here to view Figure 1|
From the PerilAUS databases we have extracted information on locations affected, maximum and average magnitudes or intensities at each location, and frequency of occurrence for each of the nine perils. One natural hazard event may produce several records (e.g., for different suburbs) in the one postcode. The point location information has then been converted to an Australia Post postcode basis. This conversion creates a particular perspective on natural perils in Australia as postcodes range in area from 0.01 km2 to 638,100 km2.
The nine natural perils are not equal. Tropical cyclone winds can affect thousands of square kilometres, with winds decreasing in intensity from the cyclone eye, while a single landslide or tornado rarely impacts more than a few square kilometres. Similarly, the built environment is more vulnerable to the impact of some perils than it is to others. A wide variety of weighting schemes is available. We have examined several alternatives but focussed on the consequences for buildings of natural perils in the last 100 years.
Damage Index information in PerilAUS I indicates the relative importance of each peril as illustrated in Figure 2. This figure is based on data from 853 events and 830 postcodes, all the events and postcodes for which we have nformation about damage to buildings. For the maps in PerilAUS II we have used a global weight-ing scheme similar to that suggested in Figure 2. Had our primary concern been the vulnerability of humans or of crop production, different weighting schemes would have been appropriate.
The postcode data and the global weightings have then been combined using the Weighted Linear Combination (WLC) method. Other combination methods, such as TOPSIS – Technique for Order Preference by Similarity to Ideal Solution – were also trialled, but we have used the WLC method as it is the best known and most widely-used of the traditional Multiple Criteria Evaluation methods (see Yoon and Hwang, 1995). Data for each peril for each postcode has been multiplied by the corresponding weight for that peril. The products have been summed to provide a single value for the historical (PerilAUS I) data for each postcode.
A number of natural hazards potential maps for Australia have been published. The best of these are in Johnson et al. (1995), though we have also used a variety of other sources. As national maps of flood, landslide and tornado hazard potential have not been produced, we have derived a potential map for each of these perils from PerilAUS I data. For the point data available for each of these three perils, buffer zones of varying dimensions and intensities have been introduced.
The potential hazard maps have been digitised and converted to a common coordinate system. For maps which rate hazard potential as Low, Moderate, or High, for example, the linguistic terms have been converted to fuzzy numbers and crisp numbers using the methodologies of Chen and Hwang (1992). Processing of these values produces a potential risk rating for each hazard for each 2 km x 2 km cell in Australia (1,909,377 cells in all).
The nine values for each cell can then be combined using global weightings similar to those in Figure 2 and the WLC method. A slightly generalised map of natural hazards potential in Australia for the nine perils is presented in Figure 3.
Alternatively, as indicated in Figure 1, potential hazards can be extracted at the postcode level. In the case of bushfire potential, a mask has been used to reduce the apparent potential in highly-urbanised inner-city suburbs to more realistic levels.
The attached map of Australia illustrates one of the final products of more than a decade’s research effort. Here the historical (PerilAUS I) and potential maps have been combined with the historical data contributing 30% of the total weighting for each postcode. Fifteen levels of relative risk from nine perils are shown [1 = lowest]. Equal numbers of postcodes occur in each level.
The attached map raises issues about the relative significance of natural perils in Australia. Why do some postcodes have higher/lower values than those nearby? To what extent does the map reflect population concentrations? How would the map differ if ICA/CRESTA Zone boundaries been used instead of postcodes? Can a simple scheme be devised so that insurance premiums reflect the spatial distribution of natural perils risk? How do the risk ratings change if uninsured perils are omitted?
PerilAUS II provides a range of maps produced using the broad framework outlined in Figure 1. These maps include those developed for individual perils, those based on CRESTA/ICA Zones and those formed from values for 2 km x 2km cells.
PerilAUS II provides peril-by-peril and postcode-by-postcode risk ratings in a spreadsheet format that allows easy comparison. All the issues raised above can be addressed readily.
PerilAUS I and PerilAUS II are available to Insurance Council of Australia members at a combined price of AUD$8,800. For non-members the price is AUD$9,900. Special prices can be negotiated for data providers. A PerilAUS demonstration is available on the web at http://www.riskfrontiers.com/perilaus/index.htm.
Chen S J and Hwang C L, 1992, Fuzzy Multiple Attribute Decision Making, Springer-Verlag, Berlin.
Johnson, R.W., Blong, R.J., Ryan, C. and seven others. 1994. Natural hazards potential map of the Circum-Pacific Region - Southwest Quadrant, US Geological Survey, 1:10 million map
Yoon K P and Hwang C L, 1995, Multiple Attribute Decision Making: An Introduction, Sage Publications, Thousand Oaks.
|Natural Hazards Quarterly Homepage|
|Natural Hazards Research Centre Homepage|