Un modèle de prévision de débit et d'analyse d'erreur

Incorporating the variance in forecast flow at various lead times was attempted in this study taking into account the uncertainty in models. The probability of exceeding a specific discharge was therefore estimated to provide further information for flood warning. The model for forecasting flow was composed of an autoregressive model for upstream flow forecasting and a Muskingum-type model for channel routing. The uncertainties in both models were represented by the variance of input parameters and previous forecast values. Both Mean Value First-Order Second Moment and Monte Carlo methods were adopted to estimate a reliability index and the probability of exceedence of a specific discharge at various lead times. Both methods yielded similar results. From the stochastic viewpoint, the model for flow forecasting had reasonable forecasting ability around the peak flow based on the verification from five historical storm events.