Una evaluación crítica del análisis cluster para identificar sedimentos contaminados en la Ría de Vigo (NW España)

Abstract

The indiscriminate use of cluster analysis to distinguish contaminated and non-contaminated sediments has led us to make a comparative evaluation of different cluster analysis procedures as applied to heavy metal concentrations in subtidal sediments from the Ría de Vigo, NW Spain. The use of different cluster algorithms and other transformations from the same departing set of data lead to the formation of different clusters with a clear inconclusive result about the contamination status of the sediments. The results show that this approach is better suited to identifying groups of samples differing in sedimentological characteristics, such as grain size, rather than in the degree of contamination. Our main aim is to call attention to these aspects in cluster analysis and to suggest that researchers should be rigorous with this kind of analysis. Finally, the use of discriminant analysis allows us to find a discriminant function that separates the samples into two clearly differentiated groups, which should not be treated jointly.