The aim of this study is to implement a reliability levee model in order to assess the probability of failure, conditioned to the hydraulic load, for a given river reach. A physically–based performance function with stochastic variables is implemented into the reliability model to represent the triggering condition of piping failure, the main collapse mechanism occurred on the 98km–reach of Po River test site. The probability of failure is provided through different reliability methods, as first order reliability method and Monte Carlo, for integrating the joint probability distribution of the performance function over the failure region for a deterministic range of values of load condition. A comparison between the proposed reliability methods, for three different discrete river reaches, is provided. Flood risk mitigation and management actions can be evaluated using the proposed method for generating different levee breach scenarios taking into account sources of uncertainty which are seldom considered in the current practice of flooding hazard mapping.