In this issue:
NHRC is kindly sponsored by:
For the last 30 years the Insurance Council of Australia has maintained a record of catastrophe payments for insured events in Australia. While this record provides a valuable insight into the incidence and costs of insured natural hazards in Australia it is short compared with the return periods of many natural events. For example, a record of 50-100 years may be necessary to establish an adequate idea of the magnitude, frequency and consequences of tropical cyclones or bushfires. Similarly, we know from statistical studies of hailstorms that detailed records for 40-50 years are required to define adequately the risks associated with these events.
Insured property is damaged by a variety of perils in Australia - tropical cyclones, bushfires, earthquakes, hail and windstorms, tsunami, floods and landslides. Many of these perils are complex in the sense that damage (and injury or death) can result from more than one agent; for example, damage in tropical cyclones can result from strong winds, storm tides, flooding or heavy rainfalls. Insurance records rarely differentiate the proportions of damage produced by each agent (whether insured agents or not) but such attempts are required for improved risk identification and management.
The Natural Hazards Research Centre has received a grant from the Insurance Foundation (a division of the Insurance Council of Australia) for a project entitled "Perils, Postcodes and Risk Accumulation Zones" (PPRAZ). This project will compile and present up-to-date, factual information on the magnitude, frequency, and consequences of natural perils in Australia. The aim is to educate underwriters and others in the industry about natural perils, the primary purpose being to assist in the formulation of underwriting strategies, the control of risk accumulation and the management of portfolios.
The project will result in two CD-ROM products:
The first CD-ROM is due for release mid 1999 and the second expected by October 2000.
For further information please contact:
Researchers at the NHRC have created some valuable databases on the occurrence and consequences of natural hazards in Australia. They are distinguished from other such databases by the wealth of descriptive detail contained therein, concerning the hazard impact, and the inclusion of information about any fatalities caused by that hazard.
In the present study, we have utilised these databases in a preliminary investigation of some of Australia's highest-ranked natural hazards (in terms of fatalities) with respect to monthly SOI ranges (a measure of ENSO (El Niño-Southern Oscillation) strength) for eastern Australia. Do any correlations exist? To what degree are climatic hazard fatalities a result of sudden change (natural variability), chance (randomness or coincidence) or immutable fate (destiny or inevitability)?
As mentioned at a previous NHRC Seminar, heatwaves are the most significant natural hazard in Australia in terms of loss of life (with the possible exclusion of disease hazards). Floods and tropical cyclones tie for second place and bushfires rank third. The hazards investigated in this study were floods, heatwaves and bushfires (the fatalities data within the tropical cyclone database is as yet incomplete). Climatic extremes in the Australian region are, to a certain degree, a function of ENSO, the signal being particularly strong in eastern Australia. For this reason, the eastern states - Queensland, NSW, ACT and Victoria - were chosen for study.
Generally, the best correlations between climatic variables, such as precipitation, and the SOI are those done on a monthly or seasonal basis. It does not necessarily follow that the same is true on a hazard event basis - for example, a local high-intensity 1-or-2-day event that has resulted in fatalities may not necessarily correlate well with a high positive or negative SOI value. In fact, the databases show that, with the possible exception of bushfires, there was no direct correlation between the total fatalities per event or per month and the monthly SOI value. There are just too many other factors (socio-economic and circumstantial) involved in the fatalities for a significant result to be obtained.
To overcome this, the data were examined from another angle. The fatalities data were examined on an "event day" basis and were categorised into eight SOI classes (according to the corresponding monthly SOI value). One "event day" indicates a day where one or more fatalities occurred (whether 1, 2 or 60).
These event days were then summed for each month. The resulting values were standardised, to take into account the fact that very few months experience ± 30 SOI values. Using this methodology, clear trends emerged for floods and bushfires, but not for heatwaves (C/F fig 1). A possible explanation for this is the complexity of heatwaves.
However, to find out which groups of society are most at risk for more direct, specific targeting of the public (for warnings) and to enable the most feasible mitigation strategies to be employed, the NHRC databases still hold the key. As well as containing information on the physical characteristics of the hazard and the damage caused, these databases have been interrogated for trends in fatalities. Lines of investigation have concentrated on population vulnerability with respect to location and seasonality of the hazard event, and the gender, age and activity of the victim at the time of death.
The NHRC Flood Fatalities database, completed near the end of 1997, illustrates some of these trends.
For more information please contact:
In severe urban thunderstorms, automobile damage can be substantial and extensive. Often the cost of replacing and repairing the many damaged vehicles represents a large proportion of the total insured property losses for the storm. It is reported that close to half of the total insured losses following the last two major hailstorms in Sydney (March 1990 and October 1986) were for motor vehicle damage. Severe urban hailstorms in North America and Europe have yielded similar damage proportions. The Natural Hazards Research Centre is currently developing a numerical modelling approach to estimating motor vehicle hail losses. The model will simulate relevant physical characteristics of hailstorms and incorporate vehicle exposure and vulnerability information.
Relevant physical characteristics of hailstorms include:
The terminal velocity of a hailstone (the vertical velocity due to gravity) is related in quite a complex manner to its mass, size, shape, and surface characteristics. Because of its mass and velocity, the hailstone embodies an amount of kinetic energy. The impact kinetic energy of the hailstone is critical with respect to damage potential and is strongly related to hailstone diameter (Fig 1). When a hailstone hits a car panel its kinetic energy is converted to heat, noise, elastic and plastic deformation of the impacted panel, fracturing energy of the hailstone and rebound kinetic energy of the hailstone (or pieces thereof). The most significant effect from a damage perspective is of course plastic deformation of the panels (i.e. denting) or, in the case of windscreens and plastic components, failure by cracking. As hailstones get larger, even by relatively small amounts, the damage potential increases dramatically.
Figure 1: Theoretical relationships between terminal velocity, kinetic energy and hailstone size.
The number of vehicles exposed to the impact of a particular storm simply depends on the number of vehicles located in the affected region at the time of the storm and how many of those vehicles are protected from the hailstones (i.e. under shelter). The number of vehicles in a storm-affected area will vary with the time of day, the day of the week and the time of year. The proportion of vehicles under shelter also varies temporally and spatially. Because the time of hailstorm occurrence is variable and the duration quite short, these descriptions of vehicle exposure must be represented as accurately as possible.
For the same storm, the cost of the damage in dollar terms and as a proportion of the insured value will be different for different cars - some cars dent more easily than others and some are more costly to repair. Thus, estimating the cost of damage sustained by a group of cars exposed to hail also requires some understanding of impact physics, and the relative dentability and repairability of vehicle types and components. Loss curves exist for other hazards and other units at risk (such as houses and commercial buildings) but have not yet been developed for motor vehicles exposed to hail.
For further information please contact:
We are living in a world that is increasingly subject to uncertainty. Flood risk and costs are no exception. The flooding that occurred in Nyngan in April 1990 caused damage estimated at A$47.27 million. Flood losses were most likely enhanced by the highly uncertain flood forecasting in the Bogan catchment and a misplaced confidence in the levee banks. Damage estimates did not take into account costs related to reconstruction and improvements of roads elsewhere in the catch-ment, private sector expenses in establishing protection from future floods, costs to individual farming properties in the catchment, or ongoing damage caused by lesser subsequent stream flows that were exacerbated by geomorphic changes in the river channel network.
Total costs to the various communities may there-fore be substantially higher than originally estimated and are likely to be ongoing. In addition to these short-comings, there are two other factors which are only partly accounted for and which will impinge on future flood risk: hydrologic changes caused by land management and develop-ment practices over the past 100 years; and the possible hydrological effects of climate variability and climate change.
Since the early 1900s until 1947, rainfall amounts have decreased
relative to the long-term average (Fig. 1). Thereafter, rainfall time-series
show a positive trend. Runoff anomalies mirror the negative trend in rainfall
until 1947 and the trend is reversed after 1947 (Fig. 2). However, runoff
values experience a negative trend from the early 1960s which do not
correspond to rainfall trends. These results imply a change in the catchment's
rainfall-runoff relationship. Flood probabilities based on pre- and
post- 1947 flood records will be considerably different, influencing mean
annual flood damage estimates. Changes in rainfall and runoff trends
and relationships increase the uncertainty in estimating flood frequency and
magnitude, and risk and costs. The importance of the latter problem
to the insurance industry is apparent.
We cannot afford to disregard the lessons learned at Nyngan if a similar disaster is to be avoided in the future. Future flood damage may be even greater than the damage incurred in 1990, unless appropriate management action is taken at every opportunity. At present we lack the coherent knowlege to do this effectively.
For further information please contact: