Contaminant Evaluation in Fish from the Mining-Impacted Morphou Bay, Cyprus, Using Statistical and Artificial Neural Network Analysis

The aim of this work was to statistically evaluate the presence of various contaminants in fish in the copper mining-impacted Morphou Bay. Multivariate data techniques were used to analyse the contamination in seasonally encountered fish. Principal component analysis was used to classify the contaminants with respect to factor loadings. Red mullet, a local deep feeder, was found to be affected at 66% of factor loadings in variance, with a linear correlation between arsenic (As) and copper (Cu). Agglomerative hierarchical clustering linked three similar clusters: (As, Cu, Se, Zn), (Cd, Cr, Hg), and (Ni, Pb) and we merged them in a dendrogram. The self-organising maps were developed by combining the artificial neural network analysis classified data with the sub-patterns, while considering the results of principal component and clustering analysis. The resultant maps effectively distinguished the correlations between the contaminants in the sampled fish. Based on the results, we suggest that human consumption of fish from the polluted area should only take place with caution and that the study be repeated using a more elaborate sampling strategy.