Evaluation of Aquifer Hydrograph Prediction with Approaches of Single and Integrated Bayesian Networks

Utilization of groundwater resources in arid areas is an important factor in the development and landscape of the region. This has led to the use of different approaches to assess and determine the volume of groundwater resources. In this study, Bayesian network, which is a probabilistic network based on recorded data, was used to reduce uncertainties. The use of two single states of piezometers and a combination of piezometers in the estimation of aquifer hydrographs using Bayesian network was evaluated using HUGIN v8.3 software. In order to implement two simulation approaches with Bayesian network, single state simulation for each observation well and integrated mode simulation for five observation wells at the aquifer level were performed. The results of two simulation models for two years predicting the future trend of the aquifer indicate a high level of statistical indicators between observational data and simulation. The final results in the single method indicate an average explanation coefficient of 0.85 with an average error square of 0.42 and in the integrated method with an average explanation coefficient of 0.8 with an average error square of 0.25. The results also showed that the use of Bayesian method to predict groundwater level and aquifer volume is highly accurate and the use of a single approach to predict groundwater level in each observation well and integrated method in predicting aquifer hydrograph has good accuracy.