Systemic Estimation of Dam Overtopping Probability: Bayesian Networks Approach

Abstract

Dam overtopping is one of the major causes of dam breach accidents. Many studies have focused on the estimation of dam overtopping probability. However, the occurrence of a dam overtopping incident is subject to many factors including hydrologic and management factors. In this paper, a model for estimating dam overtopping probability considering spillway gate maintenance activity is established based on Bayesian networks (BN). The interactions between factors that can form a loop in causality are decomposed, and the probability representing the feedback effect is calculated for BN construction. The model is applied to an arch dam with three spillway gates. Statistical data and expert domain knowledge are used to quantify the relationships between variables. The results of a case study demonstrate that gate maintenance activity can significantly influence the occurrence probability of dam overtopping. Inspection frequency and individuals’ situational awareness are important factors that need to be enhanced during dam operation. Moreover, the method’s potential as a management tool is illustrated by examining the effect of embedding a reward system.