Development of Hybrid Bayesian Network Model for Multi-Hazards Risk Assessment of Irrigation Network

Since the majority of water resources is used for agricultural purposes, irrigation and drainage networks become important. Moreover, these networks are threatened by various natural and unnatural hazards that each one can affect the performance of the network. This research seeks to develop a risk analysis model of agricultural water distribution systems with the help of hybrid Bayesian networks. The proposed hybrid Bayesian model evaluates the risk of agricultural water distribution system and its components with the following inputs: inflow of water distribution system and its fluctuation, and the demand of water. The hazards which threat the system are identified and the model nodes are determined according to these risks and the system components. This model was investigated on the distribution of Roodasht irrigation network located in Isfahan, which is under the threat of improper performance of the ditch-riders and operational losses. The average risk value of the distribution system was 14.8% and the risk of components was calculated in the range of 0.01-49.2%. The hybrid Bayesian network model shows a good accuracy and performance in training and test sets with root mean square error of 0.07% and 0.08%, and coefficient of determination of 0.65 and 0.63, respectively. The proposed model helps operators and decision-makers to be aware of the causes and potential failures of the system’s components. This can lead to better planning for the allocation of irrigation water based on the anticipated risks in the occurrence of various hazards.