Prediction and modeling of fluoride concentrations in groundwater resources using an artificial neural network: a case study in Khaf

 Background: One issue of concern in water supply is the quality of water. Measuring the qualitative parameters of water is time-consuming and costly. Predicting these parameters using various models leads to a reduction in related expenses and the presentation of overall and comprehensive statistics for water resource management. Methods: The present study used an artificial neural network (ANN) to simulate fluoride concentrations in groundwater resources in Khaf and surrounding villages based on the physical and chemical properties of the water. ANN modeling was applied with regard to diverse inputs. Results: The MLP1 model with eight inputs of parameters such as root mean square error (RMSE) and correlation coefficient of actual and predicted outputs exhibited the best results. The lowest fluoride concentration (0.15 mg L-1) was found in Sad village, and the highest concentration (3.59 mg L-1) was found in Mahabad village. Based on World Health Organization (WHO) standards, 56.6% of the villages are in the desirable range, 33.3% of them had fluoride concentrations below standard levels, and 10% had higher than standard concentrations of fluoride. Conclusion: The simulation results from the testing stage for MLP1 as well as the high conformity between experimental and predicted data indicated that this model with its high confidence coefficient can be used to predict fluoride concentrations in groundwater resources.