Water Quality Classification Based on Minimum Qualitative Parameter (Case Study: Karun River)

Rivers are considered one of the most important resources of providing fresh water. Restrictions faced by such resources underline the necessity of preserving their quality. Water quality indices are usually resorted to qualitatively monitor water resources. Any of such indices is calculated with regard to a series of specific qualitative parameters. The process of sampling and quantification of the aforementioned parameters are, however, time-consuming and costly. Finding a reliable method comprised of minimum qualitative parameters could be, therefore, of great help in classifying water quality. As an alternative to the common NSFQWI, the advantages of the Probabilistic Neural Network (PNN) as a classifier are used in the present study to classify water quality of Karun River. In order to fulfill this objective, the qualitative statistics of 172 samples were used in a way that qualitative parameters and water quality classes derived from NSFWQI are used respectively as the input and output of the model. The assessment criteria of error rate, error value, accuracy and Spearman’s correlation coefficient were used to evaluate the performance of PNN model. The results showed that through making use of merely three parameters of turbidity, fecal coliform and total dissolved solids, PNN model is capable of classifying water quality with the accuracy of 94.365% and 90.7769% at two stages of training and test respectively which in turn indicates considerable accuracy of PNN in determining water quality classification.