|
1L - Exposure Modelling (WEP/93) Applying Self-Organizing Map to assess the heavy metals contamination of a chemical / petrochemical complex in Tarragona (Catalonia, Spain). Nadal, Marti1, Espinosa, Gabriela2, Schuhmacher, Marta 1, 2, Domingo, Jose Luis1, 1 Laboratory of Toxicology and Environmental Health, 'Rovira i Virgili' University, Reus, Spain2 Environmental Engineering Laboratory, ETSEQ, 'Rovira i Virgili' University, Tarragona, Spain ABSTRACT- Advances in the theory and technology of Artificial Neural Networks (ANN) provide the potential for new approaches to the problem of classification and diagnosis of areas polluted by a large number and variety of contaminants. In this paper, neural networks trained with unsupervised learning algorithms have been used. This network can detect regularities in its space in order to develop different kinds of behaviour of the input data. In this case, data about concentration of heavy metals in soils has been taken to assess the environmental pollution in a strongly industrialised zone where several chemical factories and an oil refinery are present. Kohonen's Self-Organizing Map (SOM) has been executed to achieve a double purpose: firstly, to sort the heavy metals into several groups according to their environmental behaviour, and secondly, to establish the most impacted points of the area of study. Results appear in a friendly visualisation system where differences can be observed easily. Key words: heavy metals , Artificial Neural Networks, SOM, petrochemical complex |
