Pattern and process in the distribution of cisco, Coregonus artedii, in Trout Lake, Wisconsin
In Trout Lake, small cisco, Coregonus artedii, were vertically segregated from large cisco and exhibited a different temporal-spatial pattern of dispersion. The small fish, which were relatively high in the water column (10-20m depth), aggregated into dense schools after sunrise, and remained aggregated throughout the day where their depth distribution intersected the bottom. At dusk they dispersed horizontally, and remained dispersed all night. Large cisco tended to be relatively deep in the lake (\textgreater20m), and were dispersed throughout the diel period. Day-to-day variability in the horizontal distribution of cisco occurred at a large scale during both day and night. Schools did not exhibit site fidelity. A modeling approach combining a bioenergetics model with a composite consumption rate model identified the dispersed phase as a feasible foraging window for each behavioral group. An individual-based model of schooling behavior and an animal movement simulator to execute the model were developed to examine the transient process of school development in Trout Lake. The model produced realistic schooling behavior. The rate of school development depended on both fish density and spatial extent of the simulation volume. Mean school size and mean number of schools from the simulations were used in a simple aquatic predation model to estimate potential consumption rate of the schooling fish by a predator. At high fish densities similar to those in Trout Lake, schooling behavior provided substantial protection from predation soon after schooling behavior was initiated. Consumption rate by the predator reached a maximum at intermediate prey densities. The schooling behavior model reproduced major characteristics of cisco behavior in Trout Lake. In both the lake and the simulations, small schools formed across the horizontal extent of the fish distribution. In both systems, school size progressively increased and fish aggregated near the lake boundary. Schools in the lake were 5-lOm tall. Simulated schools were \textlesslm tall. A sonar simulator produced homologous data from the simulations for comparison with hydroacoustic data. Spectral coherence between simulated sonar data and data from the lake was poor. Recurring spatial structure in the lake at a scale of 17m was detected in transects spanning more than one hour.