A model of habitat selection by drift-feeding stream salmonids at different scales

Whole-stream size gradients of drift-feeding stream salmonids have received practically no attention, perhaps because the smaller-fish-upstream pattern that is thought to prevail is consistent with knowledge of habitat selection by other stream fish at the local scale. However, a rather counterintuitive larger-fish-upstream pattern has recently been documented for Arctic grayling (Thymallus arcticus) in Alaska and brown trout (Salmo trutta) in New Zealand, and habitat selection theory cannot explain these observations. My goal in this paper is to improve this situation by developing a model that predicts the distribution of a population of fish both within single pools and over the length of the entire river, as well as the behavioral mechanism responsible. The model uses information on invertebrate drift density and water temperature to predict the growth rate of different sizes of fish at the positions available in the stream. Fish then distribute themselves so that each individual occupies the most profitable position it call defend, with the largest fish winning any disputes. The model suggests that streams can be classified into categories based on the way temperature and drift density vary with the passage downstream. Each of these categories favors a different combination of size gradient and behavioral mechanism. A larger-fish-upstream pattern due to size-dependent habitat preference is predicted for streams with cool temperatures and low drift densities, conditions found in Arctic grayling streams. A larger-fish-upstream pattern due to competition between fish of different size is predicted in warm streams, irrespective of drift density, conditions found in many New Zealand trout streams. A smaller-fish-upstream pattern due to competition between fish of different size is expected in cool streams with high drift abundance; currently no data are available to test this prediction.