Flood insurance and emergency management activities require detailed risk assessment on a property-by-property basis. Risk assessment at the postcode level is unsatisfactory for flooding. The distribution of flooded properties is rarely uniform across a postcode, the proportion of flood-prone properties is generally quite small, and the proportion of flooded properties for different flood intensities is different for each postcode. Similarly risk assessments that only consider floods with Average Recurrence Intervals (ARIs) less than 100 years are also unsatisfactory - in many catchments the bulk of properties lie between the 100 year ARI flood level and the Probable Maximum Flood (PMF) level.
FloodAUS (formerly FIRM) the NHRC’s flood model allows risk assessment on a property basis for a range of floods up to the PMF. For a given catchment it combines a digital elevation model with available flood surfaces and geo-coded street addresses to estimate the (ARI) of inundation of the land surface and points 1m and 2m above the surface for each parcel of land. The detailed output from FloodAUS highlights the importance of considering floods less frequent than the 100-year ARI event and reveals the different inundation patterns for individual postcodes.
A FloodAUS analysis of the Georges River catchment indicates that, of the approximately 14,000 street addresses with a ground level lower than the PMF level (ARI approximately 5,000 years), only 28% are below the 100 year ARI flood level. Over half of the remaining 72% of addresses are between the 100-year ARI and the 250-year ARI flood levels.
Figure 1 represents the relationships between the proportion of inundated properties (land parcels) per postcode and flood intensity in the Georges River catchment. The green lines represent the cumulative proportions of affected properties in individual postcodes for floods of increasing severity (increasing ARIs), while the black line represents the characteristic of the whole catchment. It is evident that different postcodes have very different inundation characteristics. In some postcodes the proportion of flood-prone properties inundated by the 100-year ARI flood exceeds the catchment average of 28%. Other postcodes are only significantly affected by the more severe (rarer) floods with less than 20% of properties affected by the 100-year ARI flood. The different characteristics are related to the spatial distribution of properties, the orientation of streets and the topography within each postcode.
FloodAUS is currently being applied to the Hawkesbury-Nepean catchment and other catchments in New South Wales. The results of these analyses will be available progressively from the end of 2000. For further information contact Roy Leigh or Laraine Hunter Telephone: +61-2-9850 9683. Facsimile: +61-2-9850 9394. Email: NHRC@ocs1.ocs.mq.edu.au
Determining insured flood damage relies on an understanding of at least four components – contents insurance, building insurance, clean-up costs and accommodation costs. Here attention is focused on just the first two of these components. Flood water can damage residential property in at least four ways:
- Building materials and contents are damaged by immersion – plaster board may disintegrate, wood may swell or warp, perishable contents rot, electrical parts short out
- Mud, sediments and other contaminants in the flood water can cause corrosion or other decay
- Dampness promotes the growth of mildew – mould or fungus that can grow on anything
- The physical force of the water and objects swept along in the flow may damage the building structure – usually only where the velocity of the flood water is more than a few metres per second.
While the depth of overfloor inundation is usually seen as the most important control on residential damage, other factors may also be important – for example, duration of inundation, sediment content, water velocity, building materials, interior construction, building age, content location, and warning time.
Contents
Figure 1 shows an NHRC integrated Contents loss curve together with data collected after the Georges River 1986 floods by Water Studies Pty Ltd. The integrated loss curve has been constructed from lists of room components and % damage-depth functions developed by FLAIR (1990). Only bedrooms, dining/lounge rooms and kitchens have been considered, with the final curve using room data in proportions derived from Water Studies data from Georges River (1986) and Nyngan (1990). The most significant feature of the loss curve is the very steep rise in the proportion of Contents losses at shallow overfloor depths losses – about 50% of contents value has been lost at a water depth of 0.5 m, and more than 80% at a depth of 1 m.
The Georges River points are based on estimates of Actual and Potential damage for bedrooms, kitchens and dining/lounge rooms expressed as % Damage. The figure suggests reasonable agreement with the NHRC loss curve up to a depth of about 1.2 m. Scatter could relate to efforts by occupants to reduce actual damage, the range of asset values or multiple floor levels.
The NHRC loss curve needs to be tested against a greater range of data. Critical issues include:
- Does the very steep rise in losses for shallow overfloor inundation depths indicate that accurate estimates of losses will be difficult?
- Does duration of inundation affect estimates of Contents losses?
- How significant is moral hazard? and
- Does widespread Contents’ underinsurance increase or decrease insurers’ losses?
Buildings
Most of the flood damage data collected in Australia show that building losses are relatively minor except for the costs of replacing built-in furniture. However, recent data and overseas loss curves paint a different view. Figure 2 indicates contributions to damage of structure components at each stage of inundation based on FLAIR data modified for Australian conditions. The dramatic changes occur in the first 10-30 cm of overfloor inundation.
Critical issues include the relevance of both available Australian data and the FLAIR data to a flood-insured Australia, the drying of buildings, and the quantum of damage external to the structure.
Conclusion
The NHRC has developed new contents and building damage curves for flood loss estimation, but better data spanning a range of inundation depths and building styles are required for testing. As the two curves are different shapes, combining them into a single curve requires information about the ratio of contents to buildings sums insured.
Our preliminary estimates of potential (and actual) insured flood losses in Australia are much larger than earlier estimates suggest. If our estimates are correct, insurers will need to get involved not only in flood risk assessment but also in the fundamental issues of land use planning and inundation-friendly building codes.
FLAIR, 1990, Flood Loss Assessment Information Report, Flood Hazard Research Centre, Middlesex Polytechnic, 378p (by A N’Jai, S M Tapsell, D Taylor, P M Thompson, and R C Witts).
For further information please contact Russell Blong Telephone: +61-2-9850 9683; Facsimile: +61-2-9850 9394 Email: Russell.Blong@mq.edu.au
Hailpaths are amongst the most “directional” damage patterns caused by natural perils. Over a medium-sized area, such as Greater Sydney (5000 km2), these paths tend to have a preferred alignment (although a larger spectrum of storm paths is possible). Due to the geographic location of the city and the synoptic settings favoring hailstorm development in these latitudes, many severe hailstorms in the Sydney area tend to arrive from the southwestern quadrant and progress towards the northeast. Keeping this fact in mind and considering that the metropolitan area, covered by ICA Zones 41 to 43, has a (roughly) quadratic shape (Figure 1), the spatial pattern and structure of company portfolios located in these three ICA zones can significantly influence the associated loss profiles.
Traditional risk evaluation methods (based on loss extrapolation) are not very appropriate for the assessment of changes in risk potential arising from different spatial structures of the portfolio, or for its optimisation. This task can be better accomplished by probabilistic loss models equipped with comprehensive and interactive hazard occurrence, exposure and vulnerability modules (provided these modules are based on good quality climatological and exposure data).
As an example of a possible model-based portfolio evaluation we use the HailAUS model to compare the risk potential of two hypothetical Sydney house portfolios with different spatial structure. As shown in Figure 1, the first portfolio is arranged from the southwest to the northeast, while the other displays a northwest to southeast alignment. Both portfolios cover the same area (ca. 250 km2), contain the same number of units and proportions of house types (ca. 50,000) and have the same total sum insured (ca. $10,000 million). In this model run the latter figures are used as substitutes for the real exposure and vulnerability data, while the parameters of the hazard occurrence module, (which simulates the physical characteristics of storms, i.e. hailstone size, storm area, alignment, time etc.), are left unaltered.
In the process of the above described simulation, the model produces ten thousand synthetic storms with random characteristics, combines this information with the supplied exposure and vulnerability data and calculates the resulting losses. The HailAUS simulated loss profiles are shown in Figure 2, while the numerical losses for selected return periods are listed in Table 1.
The results of this simulation indicate that the losses of the two (artificial) portfolios are only comparable in the 8-10 year return period range. The NW-SE oriented portfolio shows substantially higher losses for shorter return periods, whereas the NE-SW aligned portfolio has a larger risk of suffering catastrophic losses. The same principle and differences in risk profiles apply to real portfolios which are not sufficiently spread and which contain an elongated core area. Wherever such spatial portfolio structure arises, attention might have to be paid to a better estimation of the appropriate retention and reinsurance levels.
Probabilistic models, such as HailAUS, are the most appropriate tools to solve this type of optimisation task. Furthermore, they can also be of assistance in other areas, such as the evaluation of the risk underlying reinstatement of treaties. i.e. the probability of two losses going through a particular reinsurance layer in the same year.
In HailAUS, the latter risk analysis can be extended into another dimension through a choice of one of several (inbuilt) El Niño-Southern Oscillation cycles.
For further information contact Ivan Kuhnel or Roy Leigh Telephone: +61-2-9850 9683, Facsimile:+61-2-9850 9394 Email: nhrc@ocs1.ocs.mq.edu.au
PerilAUS I contains the most comprehensive collection of Australian natural peril data for the period from 1900 to 1999. Detailed magnitude/intensity information for nine perils for a total of 10,067 affected locations nationwide is provided. The primary objective of PerilAUS II is to develop indicative Relative Risk Ratings (RRR) for each of nine perils for each of the 2,573 postcodes in Australia, using PerilAUS I historical data and additional information on natural hazards potential.
The development of the RRR is conducted using a well-established multi-criteria evaluation (MCE)-GIS methodology in an open and rational manner. A detailed framework was presented in NHQ 6(2). To accommodate various peril magnitude/intensity scales, a normalisation approach for converting linguistic risk terms to fuzzy numbers to crisp values is employed. A weighting scheme, specifically concerned with damage to buildings, is also developed using historical building damage data and other damage-oriented factors under pairwise comparisons between individual perils. Finally, the weighted linear combination (WLC), as an averaging aggregation method that is neither risk taking nor risk averse, is used to produce composite risk ratings.
Final products in PerilAUS II include a series of spreadsheets and maps (more than 20) with different peril perspectives. Table 1 shows an excerpt from one spreadsheet that expresses the RRR at a postcode level (RRR are also provided for ICA Risk Accumulation Zones). The rating values in the table are explicitly put on a comparable basis, allowing arithmetic calculations – addition, subtraction, multiplication and division. RRR serve as “risk meters” and can be compared from the following two ways:
- Postcode-by-postcode: For a total of 2573 postcodes, risk ratings can be viewed individually or collectively. The ratings for each peril and the accumulated ratings for nine perils (shown in the right-hand column) vary from one postcode to another, from one region to another. At the bottom of the table, minimum and maximum rating values for each of nine perils and the total value are given for simple comparisons. It is also possible to calculate the relative standing of a rating for a postcode within all postcodes.
- Peril-by-peril: By comparing ratings horizontally, it is simple to obtain the relative importance of different perils in a postcode. Which peril is the most significant and/or its contribution to the total accumulated risk rating of the postcode can be easily determined. Also, we can calculate the sub-total risk ratings for a selected set of perils.
This RRR project is the first of its kind in the world. We use the most comprehensive data possible and focus on the damage to buildings – perhaps the primary insurance industry concern with natural perils. Procedures are fully documented in the report accompanying PerilAUS II. The spreadsheets and maps of PerilAUS II should provide valuable assistance to the insurance industry in the formulation of underwriting strategies.
For further information please contact Keping Chen or Russell Blong Telephone: +61-2-9850 9683. Facsimile: +61-2-9850 9394 Email: nhrc@ocs1.ocs.mq.edu.au
