Lismore, located on the far north coast of New South Wales, has an urban population of approximately 30,000 (ABS 1996). The town is located at the junction of two tributaries of the Richmond River: Leycester Creek and Wilson’s River. The streams that run into these creeks have a total catchment area in excess of 1400 square kilometres in one of the wettest areas of New South Wales. Rainfall run-off tends to be concentrated at Lismore by the unusual fan shape of the catchment and the steep stream gradients. These factors combine to make Lismore one of the most flood-prone towns of its size in Australia.

Flood severity

A low pressure system over the Queensland – New South Wales border produced sustained and heavy rainfall in late January 2001 which lead to flooding in Lismore and a number of other north coast towns. During the week starting 29th January, Lismore received 336mm of rain. The Wilson River at Lismore started rising early on the 1st of February and peaked at 5pm on the 2nd February. The peak height recorded was 10.42m AHD at the Lismore Gauge.

Flood heights at Lismore have been collected since 1857. The peak height recorded on the 2nd February 2001 flood is the 13th highest on record (see Figure 1). The Average Recurrence Interval (ARI) for the flood is estimated to be about 6 years. Flood heights for less frequent events are shown in Table 1. The NHRC’s flood model (FloodAUS) was used to model the expected patterns of inundation shown on Figure 2. Investigations carried out on a field trip in the days following the flood verified that the FloodAUS map was a fair representation of the actual extent of inundation.

Commercial Impacts Many streets in the CBD were inundated (see Figure 2). The depth of inundation ranged from almost three metres to a few centimetres. A preliminary survey of 39 flooded businesses revealed that seven had suffered no significant damage while some others estimated their losses in tens of thousands of dollars. The length of interruption to business ranged from one to eleven days, with most businesses only losing a day or two of trading. Damage was predominantly to carpets, shop fittings and fixtures. Most stock losses were due to damage incurred while lifting or relocating the items; direct water damage was unusual. Most business owners considered that the major cost of the flood was in the form of business interruption and wages paid during mitigation and clean-up activities.

Some businesses had well prepared and rehearsed loss mitigation strategies that were implemented before the flood, while others appeared to be caught by surprise. Mitigation strategies included:

  • ·removable carpet tiles
  • mezzanine floors for stock and equipment storage
  • easily disassembled counters and fittings
  • elaborate pulley systems for raising fittings in-situ.

The last major flood occurred twelve years ago, and this may partially explain the variation in flood awareness and preparedness.

It is worth noting that the proposed CBD levee has a design crest height of 10.9 metres AHD, so would theoretically have protected the CBD area from this flood.

Residential Impacts

Dozens of residences in North and central Lismore were affected, as were two high schools and three caravan parks. Although most houses on the floodplain are elevated, the living area of some were lower than the height of this flood and many others use the lower level for a multitude of other purposes. Our inspections produced the following short list:

  • locations for hot water heaters, air-conditioners and spas
  • storage for everything from gardening equipment to antique furniture
  • laundries
  • self-contained flats
  • (in one case) studio and small cinema

Some houses experienced more than half a metre of water through the living areas. The most commonly damaged items were undoubtedly carpets, whitegoods and particleboard fittings. Only limited structural damage was apparent from our inspections. This included subsided footings, dislodged foundation piers, damaged stairs and warped doors.

Although Lismore City Council had distributed property-specific flood information to all addresses on the floodplain just three weeks before this flood, there was still considerable variation in awareness and preparedness. In general the more experienced residents were better prepared than those who had not experienced a flood before. Other factors that may have affected preparedness were that this was the first Lismore flood for which flood height forecasts were issued relative to the Australian Height Datum rather than Ballina LWOST (low water ordinary spring tides) datum. Some residents were unsure which datum was being referred to in the warnings, so were expecting the peak flood height to be 0.8 metres lower than it was. In addition, the forecast peak flood height was revised upwards as the time of the peak approached.

Conclusion

Although Lismore has a long history of flooding and many houses on the floodplain are elevated, floods with relatively low ARIs still have the potential to cause disruption and damage. In particular it is worth remembering that not all houses on the floodplain are elevated, and that the area under raised houses is often used for a range of purposes including storage of vulnerable goods. Also the Lismore CBD is very low lying and adjacent to the river – and raising or relocating existing businesses in an established commercial area is not always practicable.

References

ABS 1996, Census of Population and Housing, Australian Bureau of Statistics
WBM Oceanics 1999, Lismore Levee Scheme Environmental Impact Statement

Figure 1: Peak heights for major floods in Lismore from February 1880 to February 2001

Figure 2: Approximate area of Lismore inundated during February 2001 flood (as modelled by FloodAUS)

For more information contact Roy Leigh or Andrew Gissong on
Tel: +61-2-9850 9683 or email: NHRC@laurel.ocs.mq.edu.au

For several years NHRC staff have suspected that earthquake Probable Maximum Loss estimations are influenced significantly by the level of detail recorded (or the scale) used in the ground zonation scheme used to reflect variations in the intensity of earthquake ground shaking. This report summarises the results of a simple test of this hypothesis.

Three CRESTA Zones (also called ICA Zones 41-43) with a total area of 3,534 km2 cover the area of much of Sydney. The areas of the three CRESTA Zones vary from 862 to 1,434 km2. A total 208 postcodes are confined within the boundaries of CRESTA Zones 41-43. The average postcode area is about 17 km2. The 208 postcodes contain a total 5,204 Census Collection Districts (CDs). The average area of a CD is about 0.7km2.

A simple five-category ground zonation scheme has been devised, with the most intense ground shaking (5) on unconsolidated and poorly-drained alluvial and estuarine deposits and landfill. The areas judged to be least affected by ground shaking (1) are found in areas with shallow soils on competent bedrock on gentle slopes with limited relief. The details of the scheme are not particularly important to this test, but the scheme is the same as that used in the Greig Fester (now Benfield Greig), 1997 report Earthquake PML – Household Buildings – Sydney II.

Table 1 indicates the numbers of CDs in each ground-shaking zone and the proportion of the Sydney area in each ground-shaking zone. The proportion of total dwellings in each zone is based on the distribution of the more than one million dwellings in CRESTA Zones 41-43.

Portfolio data on 423,672 houses of brick, timber, fibre-cement (fibro) and “other” construction, about 40% of the current total housing stock in Sydney, has been used in the test runs. The portfolio has a total value of about AUS$87 billion. In general, double brick houses are located in inner suburbs, brick veneer in post-1970 suburbs, and fibro homes in the suburbs that developed in the 15-20 years after World War II.

For the test, the ground zonation class at the centroid of each CD is assumed to extend across the whole Collection District. Ground shaking zones determined for each CD have then been combined as a weighted, area-dependent average for each postcode. The postcode averages have then been combined to provide a weighted, area-dependent average for each CRESTA Zone.

Loss curves for each of the five residential construction types were developed from the loss experience in the 1989 Newcastle earthquake. Again, these Modified Mercalli intensity versus % damage curves are the same as those used in the Greig Fester Sydney report. Similarly, the attenuation curves, which express the decrease in ground shaking with distance from the earthquake epicentre, are from that study.

A simple deterministic PML model was developed for a ML 5.6 earthquake at a depth of 7 km using epicentres spaced on a 1-km grid. A deterministic model with regular epicentral spacing was used rather than a stochastic model with random epicentral locations so that we could be sure that the variations in estimated PMLs reflected variations in ground zonation scale. A total of 6,750 earthquakes were simulated.

Table 2 summarises the results of the simulations for each of the five construction types as well as for the entire housing portfolio. Results are presented as the average PML% for the 6,750 earthquakes used in the test.

The results summarised in Table 2 confirm that PMLs based on ground zonation averaged over large areas (CRESTA Zones) are lower than those derived from more detailed ground zonation. PMLs based on CDs (average area about 0.7 km2) are about twice those based on entire CRESTA Zones. Values based on postcodes (average area about 17 km2) are 10-18% higher than those based on CRESTA Zones.

It is unlikely that the increase in estimated PML with finer resolution of ground conditions is peculiar to Sydney. Further testing is required but it seems reasonable to assume that similar results would be found in most areas where a range of construction types are spread across variable ground conditions.

Probable Maximum Loss estimates have always been fraught with difficulty – black or grey box models, a range of assumptions varying from model to model, results expressed as average or 90th percentile PMLs, and only rarely explicit statements about the actual time frame for which the PML has been estimated. The results reported here indicate that the scale at which ground zonation has been averaged is another variable that needs to be considered carefully. The simple test reported here demonstrates that, other factors being equal, more refined models will give higher PML estimates – perhaps double those from models with simpler ground characterisation.

This report is based on Blong R J and Hunter L J, 2000, Ground zonation scale and loss estimation: Sydney, Australia, in Proceedings of the Sixth International Conference on Seismic Zonation (6ICSZ), November 12-15, 2000, Palm Springs, California, Earthquake Engineering Research Institute CD-ROM.

For further information contact Russell Blong or Laraine Hunter on Tel: +61-2-9850 9683 or email: NHRC@laurel.ocs.mq.edu.au

Return to NHQ homepage
Return to NHRC homepage