Surveys of 262 pools in 3 small streams in eastern Tennessee demonstrated a strong positive relationship between pool depth and the size of the largest fish within a pool (P<0.001). Similarly, the largest colonizers of newly-created deep pools were larger than the colonizers of shallow pools. We explored the role of predation risk in contributing to the "bigger fish - deeper habitat" pattern, which has been noted by others, by conducting five manipulative field experiments in two streams. Three experiments used stoneroller minnows (Campostoma anomalum); one used creek chubs (Semotilus atromaculatus); and one used striped shiners (Notropis chrysocephalus). The stoneroller experiments showed that survival of fish approximately 100 mm in total length (TL) was much lower in shallow pools (10 cm deep) than in deep (40 cm maximum) pools (19% versus 80% survival over 12 d in one experiment) and added cover markedly increased stoneroller survival in shallow pools (from 49% to 96% in an 11-d experiment). The creek chub experiment showed that, as for stonerollers, pool depth markedly influenced survival: the chubs survived an average of 4.9 d in shallow pools and >10.8 d in deep pools. In the striped shiner experiment in shallow artificial streamside troughs, no individuals 75-100 mm TL survived as long as 13 d, where-as smaller (20-25 mm) fish had 100% survival over 13 d. The results of the experiments show that predation risk from wading/diving animals (e.g., herons and raccoons) is much higher for larger fishes in shallow water than for these fishes in deeper water or for smaller fish in shallow water. We discuss the role of predation risk from two sources (piscivorous fish, which are more effective in deeper habitats, and diving/wading predators, which are more effective in shallow habitats) in contributing to the bigger fish - deeper habitat pattern in streams.

I examined downstream displacement of young‐of‐the‐year fishes (primarily centrarchids and cyprinids less than 25 mm total length, TL) by field sampling of drift during floods and by an experiment conducted in a 4.9‐m‐long artificial stream channel. The pattern of drift during a June 1985 flood (stage increase of 3.2 m) in Brier Creek, Oklahoma, indicated that both centrarchids and cyprinids smaller than 10 mm TL were extremely susceptible to downstream displacement. Drift rates of larger fish (10–25 mm) during both the June flood and a smaller flood (stage increase of 0.4 m) in July 1985 suggested that these fishes were much less susceptible to displacement. The rapid decline in susceptibility to displacement with increase in size from the field study coincided with results obtained in the artificial channel. These results suggest that the effects of floods on stream fish communities can depend on small differences in the timing of reproduction and of flooding.

A wide variety of resource management activities can affect surface discharge in small streams. Often, the effects of variation in streamflow on fish survival and growth can be difficult to estimate because of possible confounding with the effects of other variables, such as water temperature and fish density. We measured the effect of streamflow on survival and growth of rainbow trout Oncorhynchus mykiss in a small stream in northwestern California by manipulating the flow entering four of eight enclosed stream sections (9.0-15.3 m long) containing one pool and 2.5-4.0 m of upstream riffle habitat. In the four manipulated experimental units, we reduced inflow by 75-80% over a 6-week period in summer 2003. Flow diversion substantially decreased water velocity in riffle-pool transition areas but did not strongly affect habitat volume or water temperature. Fish in control units grew about 8.5 times as much as those in units with reduced streamflow; however, discharge manipulation did not affect survival, which exceeded 90% in both treatments. The input of invertebrate drift to pools within control units greatly exceeded the input into units with reduced streamflow; the concentrations of drifting invertebrates differed to a lesser extent between control and manipulated units. Managers concerned about salmonid populations in small streams should consider dryseason streamflow to be a key variable affecting fish growth.

Despite their promise for simulating natural complexity, individual-based models (IBMs) are rarely used for ecological research or resource management. Few IBMs have been shown to reproduce realistic patterns of behavior by individual organisms. To test our IBM of stream salmonids and draw conclusions about foraging theory, we analyzed the IBM's ability to reproduce six patterns of habitat selection by real trout in simulations contrasting three alternative habitat-selection objectives: maximizing current growth rate, current survival probability, or ''expected maturity'' (EM). EM is the product of (1) predicted survival of starvation and other mortality risks over a future time horizon, and (2) the fraction of reproductive size attained over the time horizon. Minimizing the ratio of mortality risk to growth rate was not tested as a habitat-selection rule because it produces nonsensical results when any habitat yields negative growth rates. The IBM simulates habitat selection in response to spatial and temporal variation in mortality risks and food availability as fish compete for food. The model fish move each daily time step to maximize their habitat-selection objective with no other restrictions (e.g., territoriality) imposed.Simulations with habitat selected to maximize growth reproduced three of the six habitatselection patterns; maximizing survival reproduced two patterns; and maximizing EM reproduced all six patterns. Two patterns (shifts in habitat with changes in temperature and food availability) were not reproduced by the objectives that consider only current growth and risk but were explained by the EM objective that considers how future starvation risk depends on current energy reserves and energy intake. In 75-d simulations, populationlevel survival and biomass accumulation were highest for fish moving to maximize EM. These results support the basic assumptions of state-based dynamic-modeling approaches to habitat selection. Our IBM appears successful because it avoids restrictive assumptions, incorporates competition for food, assumes salmonids make good habitat-selection decisions at a daily time step, and uses a habitat objective (EM) that provides reasonable trade-offs between growth and mortality risks.