A General Model of Selectivity for Fish Feeding: A Rank Proportion Algorithm

Abstract

Given that various prey are available to fish in a particular ecosystem, by default fish feed selectively. Studies of fish feeding ecology have provided key insights into the dynamics of aquatic ecosystems, yet prey selectivity is the least addressed component of these studies. This may be due to the higher level of effort associated with examining both the stomach contents and ambient prey abundance, the assumption that a determined diet composition is static, or the lack of a predictive protocol for a priori estimates of prey selectivity and diet composition. Here I present a rank proportion algorithm (RPA) model that predicts prey preference from first principles of predation that, when coupled with ambient prey concentrations, can predict prey utilization (i.e., diet composition). I applied the model to benthivore, planktivore, and piscivore examples from lentic, lotic, estuarine, and marine ecosystems. Compared with observed stomach contents, the RPA model's predictions of diet composition exhibited more than 83% accuracy, and in most of the cases the model predicted the predominant prey item accurately; the entire prey rank order was predicted correctly on the order of 70–80% of the time. Additionally, more than 85% of the prey items were predicted to be within 10% of observed values, and over 70% were within 5%. The results of the RPA model were notably different from those of the null model of no selectivity. The results suggest that the RPA model is a useful tool when prey preference or stomach composition data are limited but required for other applications and that a general knowledge of the predation process is useful in obtaining quantitative information about fish diet.