An index of effluent aquatic toxicity designed by partial least squares regression, using acute and chronic tests and expert judgment

The evaluation of the ecotoxicity of effluents requires a battery of biological tests on several species. In order to derive a summary parameter from such a battery, a single endpoint was calculated for all the tests : the EC10 obtained by nonlinear regression, with bootstrap evaluation of the confidence intervals. Principal component analysis was used to characterize and visualize the correlation between the tests. The table of the toxicity of the effluents was then submitted to a panel of experts, who classified the effluents according to the tests results. Partial least squares (PLS) regression was used to fit the average value of the experts judgments to the toxicity data. Furthermore PLS regression resulted in an optimum battery, with two chronic tests and one acute test. The index is intended to be used for the classification of effluents based on their toxicity to aquatic species. / L'évaluation écotoxique des effluents nécessite une batterie de tests biologiques. Nous avons utilisé la CE10 calculée par régression non linéaire avec estimation bootstrap des intervalles de confiance. Un panel d'experts nous a servi à classer les effluents et d'optimiser par régression PLS une formule de calcul corrélée au jugement des experts sur les 30 effluents. Seuls 2 tests chroniques et 1 test aigu suffisent.