In order to minimize the catastrophic effects of floods, it is essential to have good forecasts of the flood dynamics. To carry out these forecasts, commercial computing tools use hydraulic models based on the Saint-Venant equations. Instead of these hydraulic models, this paper proposes the use of input-output adaptive predictive expert (APE) models with properly adjusted parameters. For the initial parameter setting of the APE models used in this paper, four flood episodes occurred in the Ebro river in 2001, 2003, 2007 and 2008 were analysed. In a second stage, the flood episode occurred in February 2009 was forecasted with these adjusted models, and the results were compared to the ones made with the commercial forecasting model MIKE11.