Visualization of fish community distribution patterns using the self-organizing map: A case study of the Great Morava River system (Serbia)

Abstract

There is little information on how fish communities are in accordance with landscape classification. In this study, we have chosen the Great Morava River system to characterize fish assemblages and to assess how well they correspond to the landscape classification. Fish assemblage data was collected during a period between 2003 and 2011 at 99 sampling sites. The sites were patterned using the self-organizing map (SOM) based on fish biomass data. The SOM analysis distinguished two main clusters of samples, X and Y. The Y concerns the downstream areas, while the X was subdivided to the sub-clusters X1 (mid-stream areas) and X2 (upstream areas). Generally, the classification of the fish assemblages derived by SOM is in accordance with the a priori landscape classification to a greater extent. The significant anthropogenic influence is responsible for the misclassification of some particular sites. The classification strength, derived from the fish community classification obtained by SOM, was higher than for a priori defined groups. Environmental parameters, which the landscape classification is based on, such as altitude and dominant substrate, are found to be very important for community structure. The results of this study reveal the spatial organization of fish communities which could help their implementation in rapid bioassessment programs.