Mathematically, natural disasters of all types are characterized by
heavy tailed distributions. The analysis of such distributions with
common methods, such as averages and dispersions, can therefore lead to
erroneous conclusions. The statistical methods described in this book
avoid such pitfalls. Seismic disasters are studied, primarily thanks to
the availability of an ample statistical database. New approaches are
presented to seismic risk estimation and forecasting the damage caused
by earthquakes, ranging from typical, moderate events to very rare,
extreme disasters. Analysis of these latter events is based on the limit
theorems of probability and the duality of the generalized Pareto
distribution and generalized extreme value distribution. It is shown
that the parameter most widely used to estimate seismic risk - Mmax, the
maximum possible earthquake value - is potentially non-robust. Robust
analogues of this parameter are suggested and calculated for some
seismic catalogues. Trends in the costs inferred by damage from natural
disasters as related to changing social and economic situations are
examined for different regions.
The results obtained argue for sustainable development, whereas entirely
different, incorrect conclusions can be drawn if the specific properties
of the heavy-tailed distribution and change in completeness of data on
natural hazards are neglected.
This pioneering work is directed at risk assessment specialists in
general, seismologists, administrators and all those interested in
natural disasters and their impact on society.