Principal component analysis: an appropriate tool for water quality evaluation and management-application to a tropical lake system

An eutrophic lake system characteristic of Ivory Coast provided us with the opportunity to check that the values of all analytical variables are linked to both causes and effects of eutrophication (feedback effect). Therefore, none of these values can accurately describe a trophic state alone. To solve this difficulty we suggest here, that relationships between analytical variables are able to generate better descriptors than variables themselves. We show that principal component analysis (PCA) using coefficients of linear regression is, by construction, an appropriate tool for this purpose. The graphic representations obtained underline that: (i) the first principal component is linked to the trophic potential and the second one to the trophic level; (ii) the graphical locations of the different lakes studied are consistent with their apparent features; (iii) allochthonous inputs have a spreading effect on the graphic representation. Extension of this model to other lakes, located in the same geographical area, was successfully carried out. Furthermore, it has been shown that it is possible to reduce the number of analytical parameters to four (pH, conductivity, UV absorbance at 254 nm and permanganate index for raw water) without notably impairing the quality of the PCA representation. Moreover, these very simple parameters are easier to quantify than classical one (nutrients, chlorophyll-a, etc.) and make their use easier for the water resources management. (C) 2004 Elsevier B.V. All rights reserved.