The Brazilian coastal zone presents a large extension and a variety of environments. Nevertheless, little is known about biological diversity and ecosystem dynamics. Environmental changes always occur; however, it is important to distinguish natural from anthropic variability. Under these scenarios, the aim of this work is to present a Data Mining methodology able to access the quality and health levels of the environmental conditions through the biological integrity concept. A ten-year time series of physical, chemical and biological parameters from an influenced upwelling area of Arraial do Cabo-RJ were used to generate a classification model based on association rules. The model recognizes seven different classes of water based on biological diversity and a new trophic index (PLIX). Artificial neural networks were evolved and optimized by genetic algorithms to forecast these indices, enabling environmental diagnostic to be made taking into account control mechanisms of topology, stability and complex behavioral properties of food web.
Data Mining; Intelligent Systems; Environmental Diagnostic; Ecosystem Management; Biological Integrity