ASSESSMENT OF WATER QUALITY AND ENVIRONMENTAL SUSTAINABILITY USING MULTIVARIATE STATISTICAL

Abstract

A publicly available dataset of 3,276 samples of drinking water was analysed using multivariate statistical methods to evaluate the water quality and its implications for environmental sustainability. Nine physicochemical parameters (pH, hardness, TDS, chloramines, sulfate, EC, OC, THMs, turbidity) and a binary classification for potability were analyzed. The missing values were imputed using the median and all of the variables were standardized before analysis. Descriptive statistics, Spearman correlation, principal component analysis, factor analysis, hierarchical cluster analysis and comparisons based on the degree of potability were performed. The results showed high physicochemical variability with the highest coefficient of variation being in the case of total dissolved solids. The overall correlations between the parameters were relatively low, suggesting that their water quality was influenced by several independent processes in relative independence. 49.074% of the total variance was explained by 4 principal components, and the 4 factors extracted in the mineralization process were dissolved-mineral composition, mineralization and acid–base conditions, ionic-organic characteristics, and treatment-related influences. The class of potable samples was not clearly distinguished from the non-potable observations; the samples were grouped into three similar classes using cluster analysis based on their physicochemically similar observations. The overall result indicated that 61.0% of the samples were non-potable and 39.0% potable with statistically non-significant differences between the two for individual parameters. The results indicate that none of the indicators alone can adequately define potable water and support multi-dimensional water-management, source protection, treatment optimization, and integrated monitoring.