Using multivariate statistics and artificial neural networks to determine the colonization behavior of aquatic macrophyte populations in Santana reservoir.
Cluster analysis, principal components analysis and Kohonen artificial neural networks were tested to determine the behavior patterns of aquatic macrophyte colonization on Santana reservoir, Piraí-RJ, 2004. The colonization behavior of the populations was divided into two groups, representing the two year semesters. The principal components analysis showed that the behavior of the first group (first semester) was mainly influenced by S. montevidensis, Heteranthera reniformis, Ludwigia sp., Rhynchospora aurea, C. iria, C. ferax e Aeschynomene denticulate and the second group by Echinochloa polystachya, Polygonum lapathifolium, Alternanthera phyloxeroides, Pistia stratiotes, Eichhornia azurea, Brachiaria arrecta e Oxyscarium cubense. The artificial neural networks (ANN) grouped the aquatic macrophyte populations into nine groups according to their colonization behavior during the year. The application of principal components analysis (PCA) on frequency values of the population determined by the artificial neural networks for the first three groups allowed to discriminate three groups of populations with different colonization behaviors. The application of artificial neural networks allowed a better discrimination of communities (months) and species colonization behavior than the application of PCA on raw data.