Abstract. A Bayesian approach is described for dealing with the problem of infilling and generating stochastic flow sequences using rainfall data to guide the flow generation process, and including bounded (censored) observed flow and rainfall data to provide additional information. Solutions are obtained using a Gibbs sampling procedure. Particular problems discussed include developing new procedures for fitting transformations when bounded values are available, coping with additional information in the form of values, or bounds, for totals of flows across several sites, and developing relationships between annual flow and rainfall data. Examples are shown of both infilled values of unknown past river flows, with assessment of uncertainty, and realisations of flows representative of what might occur in the future. Several procedures for validating the model output are described and the central estimates of flows, taken as a surrogate for historical observed flows, are compared with long term regional flow and rainfall data.
A Bayesian approach for condition assessment and damage alarm of bridge expansion joints using long-term structural health monitoring data
