The relationships between fish assemblage structure (species occurrences) and environmental variables were elucidated using a canonical correspondence analysis (CCA) (ter Braak, 1986). It is a fact that rare species are often positioned as outliers in correspondence analysis ordinations (Greenacre, 2013). Furthermore, statistical estimation problems are posed by multicollinearity (ter Braak, 1986). In this sense, species occurring at less than 10 stations and significantly correlated environmental
variables (correlation test — Pearson, 1957) were omitted from analysis prior to conducting CCA. A Monte-Carlo randomization test (1000 permutations) (McCune and Grace, 2002)was used to assess the probability of the observed pattern being due to chance.