Multivariate analysis for the evaluation of temporal variations in water quality of the South-Paris region

The recent trends on environmental protection indicate that, in the immediate future, the treatment rules and the controls about water are tending to increase. Furthermore, the ability to monitor parameters containing critical information on the analysis or treatment of water is very important for research and development. The modern analytical techniques provide an enormous amount of data, which, if processed using multivariate procedures, offers great possibilities for a decision-making process. By means of multivariate techniques, grouping of samples can be made by unsupervised methods which identify the natural clustering pattern and group objects on the basis of similarities between the samples. The most common methods of unsupervised pattern recognition are cluster analysis (CA) and Classification methods. In particular, classification methods are able to build mathematic models able to individualize the affiliation class of a new object, by using a limited number of independent variables. We have defined a rapid and accurate methodological procedure, based on CA and soft independent modelling class analogy (SIMCA) classification, applied to the analytical data obtained in the year 2004 from 68 water sources in Paris and its neighbouring south-region. A wide set of data of 73.440 values was collected and the variables carrying largest useful information were selected by Principal Component Analysis (PCA). These parameters included: conductivity, alkalinity, chloride, sulphate, nitrate, hardness, turbidity, Escherichia coli and Streptococcus faecalis. A SIMCA model was used to classify the same sources in the following year 2005. The most of the water sources had kept the same analytical profile during the two-year period investigated, but five sources resulted shifting to a different class because of a significant variation of some analytical parameters. This class change was above all caused by the considerable variation of the microbiological parameters. The proposed procedure proved to be a useful tool to rapidly highlight any change occurring in the time about the water quality. It resulted very effective to handle a complex data matrix carried out from the analysis of a high number of water sources and could be implemented as a fast and efficient method in routine analysis.