TECHNIQUES IN CHEMOMETRICAL MODELING FOR MICRO-POLLUTANTS ASSESSMENT

Chemometrics have been described as effective tools for exploring chemical data, and many software packages are now available on micro-computers. This work evaluate their suitability for environmental analytical chemistry. Guidelines for multivariate method selection are proposed. They are based on the type of variables and the goal of the study. Two examples are proposed to illustrate these methods and their efficiency. Firstly it is shown that a global assessment of Rhine basin mercuric pollution in the Alsace region is possible with Multiple Correspondence Factor Analysis. Several goals are simultaneously reached: a mapping of pollution, a detection of pollution changes during the study period and an evaluation of bio-accumulation as a function of fish species. Secondly the modeling of industrial soil pollution by heavy metals is studied by Multiple Linear Regression. It demonstrates that chemometrics provide us with necessary tools for environmental analytical chemistry but also for toxicological studies or ecosystem modeling.