Abstract
The identification of pollution levels by numerical classification-ordination and the statistical confirmation of the detected trends-were attempted in a eutrophication assessment study. Special emphasis was placed on the importance of data scaling and the selection of a distance coefficient that would accentuate discrete states within the system. Among metric, binary and ordinal variable scaling, ordinal numbers showed the maximum sensitivity in discriminating pollution levels; the observed trends were further enhanced by using the absolute distance coefficient as a resemblance measure. The eutrophic patterns identified were statistically confirmed by a non-parametric permutation test. Finally a step-by-step multivariate procedure is proposed for assessing environmental quality in aquatic ecosystems.