Extreme events over time - Extreme Events: A Physical Reconstruction and Risk Assessment - page 258

which may influence the ease with which ruptures initiate and also the nature
of fault strength. Whatever the cause, palaeoseismic evidence in the USA and
elsewhere such as the eastern Mediterranean (Pirazolli et al. , 1996;Biasi et al .,
2002)suggests that earthquakes may not occur entirely randomly over time
and that events may increase in frequency at times and decrease during other
periods.
Theoretically, if earthquakes can cluster over time then so too could seismi-
cally generated landslides. Crozier et al .(1995)noted several episodes of region-
ally synchronous landslide events in the Taranaki region of New Zealand's North
Island between 1200 and 1400 years BP and around 30 000 years BP. These events
were triggered by seismic events as opposed to episodes of heavy rainfall. Crozier
et al .did not discuss whether these events represent clusterings of landslides over
time but the interesting conclusion from their study was that none of these
multiple deep-seated landslides had occurred during the 150 years of European
settlement. Clearly, without the prehistoric record, the obvious conclusion to
draw is that this region rarely, if ever, experiences large landslides. However, the
long-term record demonstrates otherwise.
Tsunamis may also cluster, but there is little definitive independent evidence
to support this. However, like the Taranaki landslide events, tsunamis in some
locations can often be absent within short historical time frames. Australia is
a case in point as this country is rarely impacted by sizeable tsunamis and
certainly not within the time frame of European occupation (last 230 years).
Nott and Bryant's (2003)study of the palaeotsunami record in Western Australia,
being Australia's most tsunami-prone coast, showed that sizeable tsunamis had
overtopped cliffs 25--30 m above sea level only several hundred years ago. The
ages of shell deposits on headlands and within mixed sand and gravel deposits
along the coast of Western Australia suggests that the region has experienced
several large tsunami events over the past 5000 years. Most of these appear to
have occurred over the last 2000 years (Fig. 10.1)with a substantial gap between
events in the period around 3000--5000 years BP. Part of this chronology may
be biased by the preservation potential of shells in deposits along this coast,
such that more recent deposits are more likely to be preserved. However, the
palaeorecords highlight that tsunamis have occurred in the past and that they
were substantially larger than the small events that have occurred historically.
Quantitative evidence for non-randomness
Anumber of different statistical techniques can be used to test whether
atime series displays randomness or is serially correlated. Serial correlation in
atime series infers that the series displays cyclicity. The random occurrence