Bayesian networks in levee reliability

We applied a Bayesian network to a system of levees for which the results of traditional reliability analysis showed high failure probabilities, which conflicted with the intuition and experience of those managing the levees. We made use of forty proven strength observations - high water levels with no evidence of failure - to refine the probability distributions of the random variables relevant for failure, and to improve the failure probability estimate of the system. We found that the use of these observations in the Bayesian network resulted in a decrease in the estimated faliure probability of over two orders of magnitude.