The complex 2-dimensional movements of fish during an annual migration circuit were generated and simulated by a chaotic model of fish movement, which was expanded from a small-scale movement model. Fish migration was modeled as a neural network including stimuli, central decision-making, and output responses as variables. The input stimuli included physical stimuli (temperature, salinity, turbidity, flow), biotic factors (prey, predators, life cycle) and landmarks or navigational aids (sun, moon, weather), values of which were all normalized as ratios. By varying the amplitude and period coefficients of the klinokinesis index using chaotic equations, model results (i.e., spatial orientation patterns of migration through time) were represented as fish feeding, spawning, overwintering, and sheltering. Simulations using this model generated 2-dimesional annual movements of sea bream migration in the southern and western seas of the Korean Peninsula. This model of object-oriented and large-scale fish migration produced complicated and sensitive migratory movements by varying both the klinokinesis coefficients (e.g., the amplitude and period of the physiological month) and the angular variables within chaotic equations.