Size-based responses of prey to piscivore exclusion in a species-rich neotropical river

Characteristics used to group species, and thus generalize ecological interactions, can aid in constructing predictive models in species-rich food webs. We tested whether size could be used to predict a behavioral response of multiple prey species (n > 50) to exclusion of large-bodied fishes, including seven abundant piscivore species, in a lowland Neotropical river. A randomized block design (n = 6) included three experimental treatments constructed on sandbank habitats: large fish exclusion, cage control (barrier with gaps), and natural reference plot (no barrier). Exclosures prevented passage of all large-bodied fishes, but the mesh size allowed passage of prey fishes. After two weeks, experimental areas were sampled once during the day and once at night. Total abundance of prey fishes was not significantly different among treatments, and effects on species density were variable. Analyses based on fish size class, however, demonstrated significant size-based effects of large-fish exclusion. Abundance of medium fishes (40-110 mm) in exclusion treatments increased significantly relative to controls in day (248%) and night (91%) samples, and this trend was apparent for many species (n > 13). Species density of medium fishes increased significantly in exclusion treatments. There was evidence of an intraspecific, size-dependent response for the three most common species. Ninety percent of prey in stomach contents of seven common large-bodied piscivores (collected from river sandbanks) were 40-110 mm, corresponding to size of fishes that responded to large-fish exclusion.. Behavioral responses of medium prey likely resulted from exclusion of these large-bodied piscivores. These results extend size-based predator avoidance behavior demonstrated in temperate systems to a highly connected tropical river food web. In this species-rich web, size is a strong predictor of behavioral responses of multiple prey species to large piscivores, and it can be used to generalize outcomes of multiple species interacting simultaneously.