Assessment of Effective Factors on the Forecasting of Debris Floods Using Bayesian Network Model

It is important to predict debris flood for reducing its damages. The aim of this study is the prediction of sediment concentration of debris floods and ordinary floods using bayesian network (BN) and artificial neural network (ANN) models in Ammameh, Navrood and Casilian basins which were located in Tehran, Gilan and Mazandaran provinces, respectively. Accordingly, average basin elevation (EL), average basin slope (S), watershed area (A), current day rainfall (R), antecedent rainfall (AR) of three-days ago and discharge of one-day ago were selected as input variables. Then, 32 scenarios were tested to determine the most effective factors on the sediment concentration of flood. For the scenario derived from all selected factors, indices R2 and MAPE in the test stage were obtained 0.97 and 8.55%, respectively. Assessment of the effect of different factors shows that the most effective factors on the BN model’s prediction accuracy are EL, R, PQ, A and AR one-day ago. Indices R2 and MAPE for this scenario were obtained 0.916 and 11.01%, respectively. It was selected as the best scenario because the least number of predictors and the highest accuracy. The most effective factors identified in this study can be used to predict debris flood in similar basins.