Flood events occur rarely and are therefore characterized by annuality in extreme value statistics. In order to provide protection from such incidents, flood protection structures are designed with a recurring interval of a 100 years. According to the state of the art, measurements of impact and resistance are being compared. In this process, sta-bility can be proven with a relation between stress and resistance below one. Such an evaluation indicates binarily either a failure or a non-failure. Regarding the probability of failure, a proof of stability means the probability of failure equals zero. A failure is thus ruled out in practice. Failures at river levees combined with flood events as for example in the Saale-estuary clearly demonstrate that levees have a probability of failure even if they comply to safety standards. However, the reasons for such an example of failure can be of different nature. For instance, hydraulic loads might have exceeded the design values, the basic geotechnical parameters gained from small sets of samples might have been assessed too optimistically or physical processes have not been considered due to a lack of knowledge.
In order to sensitize for the uncertainties of geotechnical parameters, uncertainties at a dike section at the mouth of the Saale that failed in 2013 are shown. In addition, a method for the structured investigation of the parameter effect and for the description of the uncertainties is presented and illustrated by short examples.