Comprehensive at-site flood frequency analysis using Monte Carlo Bayesian inference

In flood frequency applications where the design flood is required to have a specified exceedance probability, expected probability should be used. Its computation, however, presents formidable difficulties. This study presents a Monte Carlo Bayesian method for computing the expected probability distribution as well as quantile confidence limits for any flood frequency distribution using data on gauged flows, possibly corrupted by rating curve error, and on censored flows. This is achieved by a three-step process: (1) Formulate the likelihood function for the given data; (2) approximate the likelihood function using a multinormal distribution; and (3) integrate the expected probability integral using importance sampling. The FLIKE software for performing this is described, and an example is given.