Development Bayesian Model for Forecasting Groundwater Quality Index (Case Study: Zanjan Plain)

Determining and forecasting groundwater quality can be a primary step for managing aquifer sustainability. This study investigates and forecasts groundwater quality in Zanjan Aquifer. In the previous studies, the GWQI index is a simple weighting based on expert opinions. Thus, in the developing new index (C-GWQI), for weighting, the Shannon entropy method and the COPRAS multi-criteria decision making technique were used. In this research, COPRAS Multi Criteria Decision Making Technique was used to develop the new index (C-GWQI). By defining two permissible and desirable points of drinking water according to WHO standard, aquifer quality was classified into three ranges including, desirable, permissible and non-permissible for drinking water. The results showed that in all periods of time, groundwater quality is lower in urban areas than in other areas. However, in most of the wells surveyed, the water quality was evaluated in desirable range for drinking. The developed index was forecasted using the Bayesian network model under eight structural strategies and the best-case strategy was selected according to mean absolute relative error (MARE) and correlation coefficient (R). The best strategy was forecasted next month's groundwater quality with MARE of training and test respectively of 1.932% and 0.992%. This strategy was able to forecast the following month with good accuracy with predictor parameters such as return water, discharge, precipitation, temperature, and quality of this month. The results of this study can assist managers to better conserve and manage aquifer.