Fish communities in New Zealand rivers and their relationship to environmental variables

• Published in: New Zealand journal of marine and freshwater research Vol. 37; no. 2; pp. 347 - 366
• Main Authors: Jowett, Ian G., Richardson, Jody
• Format: Journal Article
• Language: English
• Published: Taylor & Francis Group 01.06.2003
• Subjects: abundance; community; distribution; Freshwater; freshwater fish; habitat; land use; New Zealand; Pisces; rivers; New Zealand
• ISSN: 0028-8330; 1175-8805
• DOI: 10.1080/00288330.2003.9517172

Fish abundance data from selected electric-fishing records in the New Zealand Freshwater Fish Database (NZFFD) were used to identify riverine fish communities and to examine their relationship to environmental variables included in the NZFFD. Only 21 fish species, 18 native species, and three introduced salmonids, were present at 1% or more of the sites. We defined 12 fish communities characterised by a dominant species: two salmonid communities, two non-diadromous native communities, and eight communities characterised by diadromous native species. Altitude and distance inland were the two most significant variables, reflecting differences between communities dominated by diadromous and non-diadromous species. The next most important variables were related to the geographic location of the site. Other environmental variables that were highly correlated with the fish community assignments were stream width and percentage of native forest or farming land use in the catchment upstream of the site. Of the local habitat variables, percentage of cascade habitat and percentage of sand substrate were the most important discriminators between communities. Despite having only a limited set of broad-scale environmental variables, we achieved 47% success in the prediction of community membership using multiple discriminant analysis, with another 21% of sites being near misses. A greater knowledge of barriers to migration of diadromous species and fine-scale variables describing in-stream habitat would probably increase the predictive ability of the model, although collection of such data is time consuming and therefore impractical on a national scale.