Introduction - Extreme Events: A Physical Reconstruction and Risk Assessment - page 11

capabilities. But as Baker (1994)states, prediction can be powerfully wrong as
well as powerfully right. For many natural hazard risk assessments, determina-
tion of the nature of the hazard, being the first part of the risk assessment
process, is dominated by an engineering and physics approach. This is especially
the case with floods, droughts and tropical cyclones. The reasoning process of
physics is analytical, meaning that first principles are assumed and consequences
deduced according to structured logic, which is often mathematical. This deduc-
tion is often thought of as a prediction, but these are not prophecies of future
events. Predictions are logical deductions whose accuracy, or contact with reality,
is in their match with a measured property of nature (Baker, 1994). The match
can be in the form of an experiment, which in some situations can be based
upon the results of numerical models. Many models, however, are based upon
first principle assumptions concerning the nature of the hazard. Testing these
models cannot realistically be undertaken by comparisons with observations of
future events because the suggested answers to the problem have been used as
thebasis of constructing the model, which is in turn being used to derive a
solution to the same problem. This, therefore, is a circular argument.
Assessing the probability of occurrence of tropical cyclones is a case in point.
The return period of extreme cyclone-generated storm tides is often calculated
by simulating a large number of storm tides from cyclones randomly selected
from frequency distributions of observed (historical) cyclone characteristics. Ran-
domly selected cyclones are then run through a numerical storm tide model.
The total number of modelled events represents a number of years that depends
on the observed cyclone frequency for a region. For example, if cyclones occur
on average once every 2 years at a particular location each storm tide simulation
represents 2 years' data (1000 simulations 2000 years' data). The storm tide

simulations are ranked in decreasing order (highest

=

= 2000 year event, 2nd =

1000 year event, 3rd 667 year event). In Australia, the cyclone frequency anal-

ysis is confined to the last 40 years (since 1960) because detailed instrumented
data only exist for this period. In the Cairns region of northeast Queensland,
forexample, cyclones have occurred on average once every 10 years over the
last 40 years. Based on this recurrence interval, 1000 storm tide simulations
can be deemed to be equivalent, at the determined frequency of 1 cyclone per
10 years, to 10 000 years of storm tide simulations. In this instance the 100
year storm tide = 2.15 m above Australian height datum (AHD) + wave set-up
( ∼ 0.5 m) and the 1000 year storm tide = 3.75 m (AHD) + wave set-up. Based
on this analysis Cairns can expect a large storm tide from a Category 5 cyclone
( ∼ 900 hPa) approximately every 1000 years.
The approach adopted here is mathematically and statistically rigorous and
based upon sound physics. However, an assumption is made that the 10 000 years

=