Urban water quality evaluation using multivariate analysis

Abstract

A data set, obtained for the sake of drinking water quality monitoring, was analysed by multivariate methods. Principal component analysis (PCA) reduced the data dimensionality from 18 original physico-chemical and microbiological parameters determined in drinking water samples to 6 principal components explaining about 83 % of the data variability. These 6 components represented inorganic salts, nitrate/pH, iron, chlorine, nitrite/ammonium traces, and heterotrophic bacteria. Using the PCA scatter plot and the Ward's clustering of the samples characterized by the first and second principal components, three clusters were revealed. These clusters sorted drinking water samples according to their origin - ground and surface water. The PCA results were confirmed by the factor analysis and hierarchical clustering of the original data.

Keywords

• Water quality
• drinking water
• principal component analysis
• multivariate methods
• data mining