A More Refined View

Obviously, defining the most hazardous part of Australia just on a state-by-state basis doesn’t tell us very much – though we can see from Figures 5-7 that the answer might not be as simple as we first thought. We could resolve the available data at a larger scale but with only around 5,000 events in the database spread over 100-200 years and more than 10,000 locations, it is not surprising that data are a bit thin in many areas. We used these “actual” natural hazards consequences in our map construction, but we decided that they should only count 30% towards our final map. For the other 70% we tried a different tack.

We began by choosing the best single-hazard potential maps we could find for Australia. Many of these came from the Natural Hazards Potential Map of the Circum-Pacific Region – Southwest Quadrant (Johnson et al., 1994), but we created our own maps for tornadoes, landslides and floods, and for each of these we also introduced buffer zones of varying dimensions and intensities. The maps were digitized and converted to a common co-ordinate system.

Maps of natural hazards risk usually consider only one hazard and often use rating systems such as Severe, Moderate, Low, and Don’t Worry. Trying to compare an earthquake risk map at a scale of 1:1,000,000 using such a scale with a tropical cyclone map at 1:2,000,000 that rates 10% probability of exceedance gust wind speeds as >30m/s, 31-40m/s, 41-50m/s and so on, is like comparing oranges and wheelbarrows.

We then converted the hazard potential terms such as Low, High etc to fuzzy numbers and crisp numbers using the methodologies of Chen and Hwang (1992). Effectively, this methodology allows the oranges and wheelbarrows to be added (or multiplied) together. As our maps are in a GIS, we now have a potential risk rating for each hazard for each 2 km by 2 km cell – 1,907,377 cells for Australia.

Combining maps of individual hazards potential still requires decisions about the relative importance of each hazard. As our interest was, primarily, in potential building damage we have combined the maps using relative weightings similar to those suggested by Figure 3. We used a Weighted Linear Combination (WLC) method as it is the most-widely used and the best known of the Multi-Criteria Evaluation methods. These maps counted 70% towards our final product.

Figure 8 illustrates one of the integrated natural hazards maps. This is a risk map in the sense that it combines a 30% weighting on past vulnerability to natural hazard impacts with a 70% weighting on hazard potential. Here, the scale has been divided into six equal divisions. Given the methodology, we can produce integrated natural hazard risk maps at postcode, local government area, insurance CRESTA zones, state or any other divisions, using any number of categories.

Figure 8 shows a number of small red dots in the south east of the country (resulting mainly from a smearing of past tornado impacts) and a larger pink area on the northwest coast where the strongest tropical cyclone winds are expected. Vast areas on the SA-WA border, on the SA-NSW-QLD border and in northern interior Queensland have the lowest natural hazards risk.

Readers who appreciated Mark Twain’s view of science quoted earlier will probably have read Darrell Huff’s valuable little book “How to lie with statistics”. Equally fascinating is Mark Monmier’s delightful “How to lie with maps”. Compare Figures 8 and 9. Both maps use the same data and the same 2 km x 2 km resolution; Figure 8 uses six equal divisions on the risk scale, whereas Figure 9 divides the country into six categories so that the total area in each category is the same.

Figure 8: Integrated natural hazards risk map for Australia, using six equal divisions.

Figure 9: Integrated natural hazards risk map of Australia using six categories with equal areas.

«back to top«