Extreme events over time - Extreme Events: A Physical Reconstruction and Risk Assessment - page 267

critical question is knowing when this has occurred or is occurring and know-
ing how to determine this. Assumptions of stationarity and randomness will lead
many to think it is reasonable to look for changing trends in short historical
records and if an increase is observed over the past few decades then this is likely
evidence for the impact of climate change. However, such an approach could be
very misleading because the natural variability of the particular hazard has not
been decoupled from the observed trend. In other words it is important to ask
whether the trend is simply part of a natural cycle in hazard behaviour. The
critical test for increasing air temperature as a function of climate change has
been to compare trends occurring over the last century to that recorded, both
from proxy and other records, of temperature changes over multi-century to
millennial scales. In particular, much focus has been upon the extent of temper-
ature changes during the medieval warm period compared to the last century.
Hence comparisons with the long-term record have formed the critical test for
assumptions about recent global changes in air temperature.
The same is true when attempting to determine whether human-induced cli-
mate change is playing, or has played, a role in changing the frequency and
magnitude of natural hazards. It is difficult to decouple human-induced change
from the natural variability of the hazard without knowledge of the long-term
record. Therefore, comparisons with the long-term record are necessary to deter-
mine if climate change is playing a role. Only when this is established is it pos-
sible to ascertain how climate change is altering the behaviour of the natural
hazard. Unfortunately it is common to attempt to answer the latter ( how )with-
out knowing the former ( if )orassume that how is if when relying solely upon
the short historical record.
Long-term records of extreme events offer insights into the behaviour of nat-
ural hazards beyond that possible from short historical records alone. Already
numerous techniques for extracting and deciphering these natural records exist
and the data derived highlight that non-randomness or serial correlation is a key
characteristic of such events over longer time periods. Recognition of this cyclic-
ity and the lack of its incorporation into hazard risk assessments to date suggests
that many communities may be unnecessarily exposed to hazard impacts at least
in terms of perceived levels of community safety. Hazard risk assessments need to
incorporate variable temporal probabilities, determine the uncertainty margins
of such estimates and err on the side of caution. Variable temporal probabilities
also need to be considered when investigating the role of human-induced cli-
mate change on the behaviour of natural hazards. Hopefully, the investigation of
long-term records will be a corner stone in the near future in all natural hazard
risk assessments. Only then can we hope to understand the true behaviour of
the hazard in question.