Use este identificador para citar ou linkar para este item: http://repositorio.uem.br:8080/jspui/handle/1/5250
Autor(es): PENCZAK, T.; LEK, S.; GODINHO, F.; AGOSTINHO, A.A.
Orientador: 
Título: Patterns of fish assemblages in tropical streamlets using SOM algorithm and conventional statistical methods
Palavras-chave: Paraná, Rio;Análise correspondente canônica;SOM;Brasil.;Paraná River;Tropical fish population;Assemblage structure;Self-Organizing Map;Canonical correspondence analysis;Romaniszyn diagram;Brazil.
Resumo: 
Abstract: Six sites located on the Caracu River and five sites located on the Água do Rancho River (small tributaries of the Paraná River, Brazil) were chosen for quantitative electrofishing sampling to examine patterns in fish assemblage structure using the dualism ordination technique developed by Romaniszyn. Canonical correspondence analysis was also used for verification of the data. A similar separation of sites from both streams by both methods confirmed the Romaniszyn method (RD) was useful for the study of ordination. The RD was also tested using Self-Organizing Maps (SOM), which are a variant of the Artificial Neural Network (ANN) methods. Similarly, the SOM confirmed that sites belonging to the two investigated streams were distinct, but also unveiled a minor weakness of the RD resulting from its linear character. However, in spite of this weakness, we recommend the RD for assemblage analysis by scientists unfamiliar with canonical correspondence and ANN analyses.
Descrição: PENCZAK, Tadeusz; LEK, Sovan; GODINHO, Francisco; AGOSTINHO, Angelo Antonio. Patterns of fish assemblages in tropical streamlets using SOM algorithm and conventional statistical methods. International Journal of Ecohydrology & Hydrobiology, Warsaw, v.4, no.2, p.139-146, 2004.
URI: http://repositorio.uem.br:8080/jspui/handle/1/5250
Aparece nas coleções:1.2 Artigo - Ciências Biológicas (CCB)

Arquivos associados a este item:
Arquivo Descrição TamanhoFormato 
228.pdf1,61 MBAdobe PDFVisualizar/Abrir


Os itens no repositório estão protegidos por copyright, com todos os direitos reservados, salvo quando é indicado o contrário.