QUALITY CONTROL OF GROUND WATERS BY CLUSTERING AND PLS-DA CLASSIFICATION

The quality of the ground waters is an important field to be valued and its level could destine the water to the human, irrigation or industrial use. Water quality depends by several factors such as clime, soil, rock type in which the water is born but by the use of soil too; for these reasons it is necessary to have systems of continuous water monitoring. Multivariate analysis appraises simultaneously a lot of data referred to a lot of sample. In this work a multivariate classification model has been developed for the monitoring of eighteen spring waters in the land of Serra St. Bruno, Calabria, Italy,. Thirty analytical parameters for each water source were investigated and reduced to eight by means of Principal Component Analysis (PCA). The samples were grouped in five distinct clusters by clustering techniques (CA). Cluster analysis methods provide means for classifying a given population into groups (clusters), based on similarity or closeness measures, afterwards a model for their classification is built by a Partial Least Squares–Discriminant Analysis (PLS–DA) procedure. The graphic shows the variation of the analytical parameter values, within the defined single clusters. It can be seen that some parameters, as conductivity, NO3−, hardness, Ca2+ and Mg2+, were subject to the highest fluctuation, furnishing so information more influencing the clustering distribution. PLS-DA model proved to be able to notice deviations of the global analytical characteristics, by pointing out in the course of time a different distribution of the samples within the classes. The final classification model was applied to two data sets during the years 2005 and 2006. This operation allowed to verify eventual variations of the water characteristics along the time. The same eight analytical parameters used in the calibration set were used as independent variables. Both data sets resulted distributed in five classes, analogously to the grouping obtained for the data of the year 2004. Really, the most of the water sources showed to have kept the same analytical characteristics during the three-year period investigated, without any variation of the classes they belonged. The variation of nitrate concentration was demonstrated to be the major responsible for the observed class shifts. The shifting sources were localized in areas used as sowable lands. High variability of nitrate content was ascribed to the practice of crop rotation, involving a varying use of the nitrogenous chemical fertilizers.