Fuzzy c-means and K-means clustering with genetic algorithm for identification of homogeneous regions of groundwater quality

In this study, two different clustering algorithms, fuzzy c-means (FCM) and K-means with genetic algorithm, were used to identify the homogeneous regions in terms of groundwater water quality. For this purpose, data of 14 hydrochemical parameters from 108 wells were sampled in 2016, Golestan province, northeast of Iran. The results showed that the optimal clusters of the K-means and FCM were 5 and 6, respectively. The evaluation of water quality by FCM for drinking uses showed that in terms of total dissolved solid (TDS) and chlorine (Cl) parameters, cluster 3 was in an unfavorable condition. Moreover, according to the K-means algorithm, cluster 1 was in inappropriate condition in terms of the TDS and Cl. Water quality assessment by FCM for agricultural use showed that in general, cluster 3 was not in a good condition, especially for the electrical conductivity (EC) parameter. Also, according to the K-means, in general, cluster 1 had an inappropriate state for the EC and sodium adsorption ratio parameters. Investigating the hydrochemical facies of clusters using the FCM and K-means showed that in the northern half of the Golestan province, most samples are Cl–Na and in the southern half, most of the samples are HCO3–Ca. In general, by comparing the results of clustering algorithms, it was found that the FCM algorithm has better results than the K-means clustering algorithm, mainly due to consideration of uncertainty conditions in determining the class boundary.