Bayesian statistical methods for modeling and predication of major landslides in coastal cliffs

The ability to predict coastline evolution in the long term is of great importance to coastal managers for social, economic and environmental risk assessment purposes. A Bayesian statistical approach to the problem of modeling and predicting the occurrence of major failure events in soft coastal cliffs is presented. This approach combines available data with expert judgement about the behaviour of the cliff under study. Expert judgement is described in the form of expected values for cliff failure occurrences, confidence or degree of belief and subjective probability of occurrence of extremes. WinBUGS software is used to derive predictive probabilities, estimates of the rate of failure and standard deviations, which combine all information. The statistical method is illustrated in the cases of cliffs situated in Herne Bay and Scarborough, UK. The results can be used in a risk-based assessment of coastal cliff recession.