Application of a cluster-bootstrapping method for identifying the dietary patterns of fish populations

Dietary data usually exhibit considerable statistical awkwardness and the analytical techniques commonly used in the study of fish diet suffer from various problems (especially at the intraspecific and intrapopulation levels) resulting from difficulties in group definition and pooling of data. Considering that, we advocate the use of a combination of cluster analysis and bootstrapping to identify statistically significant groups of individual predators. Both basic and extended versions of the procedure (yielding slightly different results) are outlined. We also suggest the use of detrended correspondence analysis (DCA) for further interpretation of significant clusters. When applied to dietary data from a brown trout population, the method successfully finds different feeding patterns, even though possible sources (geographical and temporal) of dietary variation were very limited. Dietary patterns appear to be related to the seasonal dynamics of the prey community coupled with seasonal shifts in microhabitat use by the predator; and, in a finer time-scale, to diel macroinvertebrate drift periodicity or interindividual, size-related differences in preyhandling capabilities.