Dynamic real-time prediction of flood inundation probabilities

The Bayesian Generalised Likelihood Uncertainty Estimation (GLUE) methodology, previously used in rainfall-runoff modelling, is applied to the distributed problem of predicting the space and time varying probabilities of inundation of all points on a flood plain. Probability estimates are based on conditioning predictions of Monte Carlo realizations of a distributed quasi-two-dimensional flood routing model using known levels at sites along the reach. The methodology can be applied in the flood forecasting context for which the N-step ahead inundation probability estimates can be updated in real time using telemetered information on water levels. It is also shown that it is possible to condition the N-step ahead forecasts in real time using the (uncertain) on-line predictions of the downstream water levels at the end of the reach obtained from an adaptive transfer function model calibrated on reach scale inflow and outflow data.