Geoscience Reference
In-Depth Information
show that natural events, including natural hazards, display serial correlation
or cyclicity over time. Serial correlation is a condition where the occurrence of
an event in a time series is dependent upon or related to the previous event and
it suggests cyclicity in the record. The apparent 'white noise' or randomness of
hazards evident in short historical records is in reality part of a longer term trend
towardscyclicity. If hazards display serial correlation over time, then statistical
techniques based upon the random occurrence of events will not accurately
estimate the probability of hazard occurrence. In other words, if a hazard does
follow a pattern of serial correlation then the probability of its occurrence must
change with time. Most assessors of risk assume that the probability remains
constant over time so the probability of an event occurring during the next 50
or 100 years is seen to be exactly the same as during any 50--100 year period. This
assumption derives from the hypothesis that the behaviour of a hazard over the
short historical record is a true reflection of its longer term behaviour. However,
if serial correlation is recognised in a longer term record then the probability
of an event occurring over the past 50 years is likely to be different from its
probability of occurrence over the next 50 years, especially where the cycle of
hazard occurrence is greater than 100 years.
Serial correlation is evident in many of the long-term records discussed in
earlier chapters. This chapter briefly reviews the characteristics of these long-
term records and then presents the statistical evidence to demonstrate that
these time series display serial correlation and hence cyclicity over time.
Atmospherically generated extreme events
Despite the assumptions made by many who undertake risk assessments
of the tropical cyclone hazard, there is little evidence, if any, to support the
notion that these events occur randomly over the long term. Indeed, several stud-
ies have found that there have been discrete periods of time when these events
tend to cluster. Liu and Fearn's (2000, 2002)analysis of washover deposits along
the Louisiana and Alabama coasts show that the frequency of severe hurricanes
has not been constant over time. A 2400 year period of increased cyclogenesis
occurred here between 3200 and 1000 14 CyearsBP.Eleven of the twelve high-
intensity hurricanes that struck this region over the last 4800 years occurred
during this period and only 1 occurred over the past 1000 years. Furthermore,
no hurricanes of the intensity registered in the prehistoric record have struck
theregion during the past 130 years. The 1600 year period prior to 3200 years
BP was also relatively quiescent. A similar story is evident for the long-term
Chinese record of tropical cyclones. This record is an historical one, but one
that is obviously much longer than occurs in most other countries. Its length,
Search WWH ::




Custom Search