Bayesian Network Model for the Assessment of the Effect of Antecedent Rainfall on Debris Flow Forecasting In Alborz Zone of Iran

Comprehensive assessment of debris flow hazards is a challenging issue due to the complexity and uncertainty of its factors. In this paper, the effect of antecedent rainfall (AR) on the debris flow occurrence was assessed by using of Bayesian networks (BN) in Alborz Zone, Iran. In this model, the effect of factors such as average basin height, average basin slope, watershed area, the current rainfall, AR (three days ago) and discharge one-day ago have been used as the model’s input. Six scenarios including the amounts of AR three days ago separately, AR two days ago separately, AR one day ago, cumulative rainfall of AR three days ago, cumulative rainfall of AR two days ago and the effect of excluding AR were considered. The results indicated the performance of BN model in the first case, 13% more than the second, and highest accuracy of the model was computed by the scenario of AR 3 days ago separately with a forecasting accuracy of 91%. Furthermore, excluding the effect of any of the events AR from the model causes reduction of its performance. The proposed model is able to provide reliable results in warning systems of debris flow hazards in watersheds.