Example: flood risk analysis.
Occurrence rates of heavy floods (magnitude classes 23) in central Europe.
a, b, Elbe, winter; c, d, Elbe, summer; e, f, Oder, winter; g, h, Oder, summer.
Flood data from Weikinn’s (19582002) documentary sources (b, d, f, h),
covering the interval up to 1850, were augmented with measured records.
The analysis tool employed a nonstationary, nonparametric risk estimation technique
with a Gaussian kernel, a bandwidth of 35 years and 2000 bootstrap simulations.
This yielded (a, c, e, g) occurrence rates (solid lines) and 90% confidence bands (grey);
occurrence rates using data from CLIMDAT for 15001799 are shown as dashed lines.
Records before 1500 are likely not homogenous (no confidence bands drawn).
Arrows indicate the results (downward/no trend) from the statistical test (90% level)
for trend in the flood occurrence rate (Elbe, 18522002; Oder, 18501920 and 19202002);
results for uncorrected and reservoir-size corrected data are identical.
Source: Mudelsee M, Börngen M, Tetzlaff G, Grünewald U (2003)
No upward trends in the occurrence of extreme floods in central Europe. Nature 425:166.
External link to paper >>
Software: Caliza™. Climate risk analysis software >>