Throughout human history, water has undergone changes in quality. This problem is more serious in dry areas, where there is a natural water deficit due to climatic factors. The aims of this study, therefore, were (i) to verify correlations between physical attributes, chemical attributes and biological metrics and (ii) from the biological attributes, to verify the similarity between different points of a body of water in a tropical semi-arid region. Samples were collected every 2 months, from July 2009 to July 2011, at seven points. Four physical attributes, five chemical attributes and four biological metrics were investigated. To identify the correlations between the physicochemical properties and the biological metrics, hierarchical cluster analysis (HCA) and canonical correlation analysis (CCA) were applied. Nine classes of phytoplankton were identified, with the predominance of species of cyanobacteria, and ten families of macroinvertebrates. The use of HCA resulted in the formation of three similar groups, showing that it was possible to reduce the number of sampling points when monitoring water quality with a consequent reduction in cost. Group I was formed from the waters at the high end of the reservoir (points P1, P2 and P3), group II by the waters from the middle third (points P4 and P5), and group III by the waters from the lower part of the reservoir (points P6 and P7). Richness of the phytoplanktons Cyanophyceae, Chorophyceae and Bacillariophyceae was the attribute which determined dissimilarity in water quality. Using CCA, it was possible to identify the spatial variability of the physicochemical attributes (TSS, TKN, nitrate and total phosphorus) that most influence the metrics of the macroinvertebrates and phytoplankton present in the water. Low macroinvertebrate diversity, with a predominance of indicator families for deterioration in water quality, and the composition of phytoplankton showing a predominance of cyanobacteria, suggests greater attention to the management of water resources.
Keywords: Canonical correlation analysis; Hierarchical cluster analysis; Water quality.