Probabilistic prediction of peak flood discharges

A probabilistic model for the prediction of flood discharges is presented. The objective is to make one-step-ahead predictions of river peak flood discharges. The model is based on a filtered Poisson process to represent the process of instantaneous river flows. An analytical expression is found for the conditionnal probability density of next flood discharge given the last flood discharge. A case study is also presented. The results obtained using two predictors based on the model are compared to those obtained using an order one autoregressive model.