University of Tabriz Water and Soil Science 2008-5133 19 2 2010 01 21 Spatial Prediction of Fluoride Concentration Using Artificial Neural Networks and Geostatic Models Spatial Prediction of Fluoride Concentration Using Artificial Neural Networks and Geostatic Models 129 145 1483 FA A Asghari Moghaddam AA Nadiri E Fijani Journal Article 2014 06 15 In the last decades, researchers had high consideration on the presence of chemical anomalies in water, soil and air which threat human health. Anomalies in fluoride concentration values exceeding standard limit (>1.5 mg/l) in drinking water have high importance, because of direct influence on physiology of human body. Fluoride concentration values of water resources in Bazargan and poldasht plains exceed standard limit (WHO). The aim of this research is spatial prediction of fluoride concentration in these plains. For this purpose Artificial Neural Networks (ANNs) model was utilized as a nonlinear model. For spatial prediction of fluoride concentration in the study area, different structures of these models were tested and the best structure (FNN-BFG) was determined. Spatial modeling was carried out by this structure and using fluoride ion concentration, correlated ions values and position of each sample, for which the determination coefficients of training and test steps were equal to 0.9625 and 0.9019 respectively. Then, results of the model were compared to those of the geostatistical methods of kriging and cokriging and the determination coefficients for test steps were 0.7285 and 0.8556, respectively. The best results of the three developed models were related to ANNs models.  In the last decades, researchers had high consideration on the presence of chemical anomalies in water, soil and air which threat human health. Anomalies in fluoride concentration values exceeding standard limit (>1.5 mg/l) in drinking water have high importance, because of direct influence on physiology of human body. Fluoride concentration values of water resources in Bazargan and poldasht plains exceed standard limit (WHO). The aim of this research is spatial prediction of fluoride concentration in these plains. For this purpose Artificial Neural Networks (ANNs) model was utilized as a nonlinear model. For spatial prediction of fluoride concentration in the study area, different structures of these models were tested and the best structure (FNN-BFG) was determined. Spatial modeling was carried out by this structure and using fluoride ion concentration, correlated ions values and position of each sample, for which the determination coefficients of training and test steps were equal to 0.9625 and 0.9019 respectively. Then, results of the model were compared to those of the geostatistical methods of kriging and cokriging and the determination coefficients for test steps were 0.7285 and 0.8556, respectively. The best results of the three developed models were related to ANNs models.  https://water-soil.tabrizu.ac.ir/article_1483_8e2926a675cd9836c93e1459156dbd04.pdf