Kurdistan University of Medical Sciences Journal of Advances in Environmental Health Research 2676-3478 8 2 2020 04 01 The modeling and prediction of the quality of the groundwater resources in Tuyserkan plain using the optimized artificial neural network 100 110 108123 10.22102/jaehr.2020.210891.1150 EN Mohamad ParsiMehr Department of Environmental Sciences, Faculty of Natural Resources and Environment, Malayer University, Malayer, Hamedan, Iran 0000-0001-7559-9570 Kamran Shayesteh Department of Environmental Sciences, Faculty of Natural Resources and Environment, Malayer University, Malayer, Hamedan, Iran Kazem Godini Environmental Health Research Center, Research Institute for Health Development, Kurdistan University of Medical Sciences, Sanandaj, Iran 0000-0001-5954-3964 Journal Article 2019 12 07 Tuyserkan plain is an important agricultural plain located in Hamadan province, Iran. Despite the severe decline of the water levels in aquifers, the quality of the plain has not been evaluated in recent years. The present study aimed to analyze the data of 15 wells during 12 years to evaluate the quality of groundwater in this area using the Wilcox diagram for the aquifer. The electrical conductivity (EC) of the plain was interpolated using the Kriging method to evaluate its spatial distribution since this parameter has caused a decline in the quality of the groundwater in the plain. According to the findings, the EC value was higher in the eastern parts of the plain and Tuyserkan city, which was described as the spatial distribution of the parameter. The Pearson's correlation-coefficient was used to assess the correlations between EC and other parameters. To predict and model the EC value, multi-layer perceptron artificial neural network (MLP-ANN) were used. According to the results of the Pearson's correlation-coefficient, the reduced number of the data led to the decreased expenditures of the experiments in obtaining the input data. The third model was finally obtained with the lowest number of the input parameters, low error, and high correlations between the predicted and actual data. In this model, two input parameters and five neurons were obtained in a hidden layer (R: 0.997, mean: 8.634, NRMSE: 0.05) using the momentum and hyperbolic tangent functions, indicating the high potential of MLP-ANNs in the prediction and modeling of groundwater quality. https://jaehr.muk.ac.ir/article_108123_48e73c4cff542e0717c517daf5a35a20.pdf