Assessment of Self-Organizing Map artificial neural networks for the classification of sediment quality

The application of mathematical tools in initial steps of sediment quality assessment frameworks can be useful to provide an integrated interpretation of multiple measured variables. This study reveals that the Self-Organizing Map (SOM) artificial neural network can be an effective tool for the integration of multiple physical, chemical and ecotoxicological variables in order to classify different sites under study according to their similar sediment quality. Sediment samples from 40 sites of 3 estuaries of Cantabria (Spain) were classified with respect to 13 physical, chemical and toxicological variables using the SOM. Results obtained with the SOM, when compared to those of traditional multivariate statistical techniques commonly used in the field of sediment quality (principal component analysis (PCA) and hierarchical cluster analysis (HCA)), provided a more useful classification for further assessment steps. Especially, the powerful visualization tools of the SOM, which offer more information and in an easier way than HCA and PCA, facilitate the task of establishing an order of priority between the distinguished groups of sites depending on their need for further investigations or remediation actions in subsequent management steps.