| A storm loss model
for Germany is presented. Input data to the model are the daily maximum
gust wind speeds measured at stations distributed over the country. The
individual daily peak gust values are scaled with the local climatological
upper 2% quantile at each station. This scaling serves to take local
conditions at the stations into account, and thus permits a simple spatial
interpolation of the storm field. The next step is the computation of a
loss index for each storm. It is based on the excess of (scaled) wind
speed over the upper 2% quantile, and on population numbers in the
individual districts within Germany, with the latter serving as a proxy
for the spatial distribution of values that could be affected by a storm.
Using wind speeds in excess of the percentile value also serves to take
spatial heterogeneity of vulnerability against storms into account. The
aggregated storm index gives an estimate of the severity of an individual
storm.
Finally, the relation between actual loss produced by a
storm and the index is estimated using published annual insurance loss due
to windstorm in Germany. Index values are accumulated for each year, and
the relation to actual loss is computed. The average ratio for the whole
reference period is eventually used. It is shown that the interannual
variability of storm-related losses can be reproduced with a correlation
coefficient of r = 0.96, and even individual storm damages can be
estimated. Based on these evaluations we found that only 50 storms account
for about 80% of insured storm losses between 1970 and 1997. |