Author(s): T. Heyer; H. -B. Horlacher; J. Stamm
Linked Author(s): Jürgen Stamm, Torsten Heyer
Keywords: No Keywords
Abstract: In August 2002 heavy precipitation caused extreme floods on several rivers in Saxony, Germany. Subsequently about100 embankment failures were reported. In a post-event analysis, intensive research was undertaken in order to identify the hydraulic, geotechnical, structural and biological conditions of both failed and non-failed embankment sections during the flood. Having established a comprehensive database about numerous embankment sections, multivariate data analysis was conducted in order to determine the main drivers that influence or cause failure. During the in-depth analysis the necessity of combining parameters of different scales became apparent. In particular biological factors, such as tree growth on embankments, infestation by burrowing animals or the quality of the grass cover are not measurable on metric scales and thus difficult to include in conventional models. In addition"embankment failure"can be regarded as a nominal parameter with binary realisations. Under these conditions logistic regression offers best opportunities for modelling dependencies between a range of input variables X i (the local characteristic of an individual embankment section) and a dichotomous response variable Y, e. g. embankment failure. A logistic regression model is calibrated on observed data by weighting the input parameters with individual regression coefficientsβ i using maximum likelihood techniques. Such models can be used to compute failure probabilities for river embankments section by section depending on the intensity of certain flood events. Thus the locations most likely to fail can be identified on a regional scale. Possible applications for the presented approach emerge from the"Directive on the assessment and management of flood risk"published by the European Commission. Within this directive the evaluation of flood risk along large rivers is a major topic. The product formula for the quantification of risk requires the determination of event-specific failure probabilities of protection facilities. This can be achieved by means of the proposed method.
Year: 2010