Takao Uno

Logit model for river levee stability evaluation considering the flood return period

Abstract Probabilistic models of the binary logit model for both middle/small and large-scale levees are presented for the evaluation of the stability of river levees. Stability can be expressed as the probability of collapse of a levee as estimated by this logit model. The expectation cost (sum of the damage risk potential plus cost of restoring the levee) is adopted for the comprehensive evaluation of river levees where the application is limited to damage done to the levee itself and excludes damage to human life and property. This method, therefore, can be used to determine appropriate types of improvements. The evaluation is statistical and requires a large amount of data including data about non-damaged levees. To circumvent this, we consider the probability of a flood and its return period and show that a practical probability of levee damage actually occuring must be calculated using a combination of the logit model and the return period.