DYNAMIC REAL-TIME PREDICTION OF FLOOD INUNDATION PROBABILITIES

The Bayesian Generalised Likelihood Uncertainty Estimation (GLUE) meth odology, previously used in rainfall-runoff modelling, is applied to t he distributed problem of predicting the space and time varying probab ilities of inundation of all points on a flood plain. Probability esti mates are based on conditioning predictions of Monte Carlo realization s of a distributed quasi-two-dimensional flood routing model using kno wn levels at sites along the reach. The methodology can be applied in the flood forecasting context for which the N-step ahead inundation pr obability estimates can be updated in real time using telemetered info rmation on water levels. It is also shown that it is possible to condi tion the N-step ahead forecasts in real time using the (uncertain) on- line predictions of the downstream water levels at the end of the reac h obtained from an adaptive transfer function model calibrated on reac h scale inflow and outflow data.