Development and application of a mechanistic model to predict juvenile salmon swim paths

Utility companies and regulatory agencies are interested in understanding juvenile salmon swimming patterns as they approach hydropower dams because it can allow them to assess fish bypass efficiency and conduct fish survival studies. A model capable of predicting juvenile salmon swim paths can assist in the design of fish bypasses and diversion structures. This thesis presents a mechanistic model tailored to simulate swimming patterns of juvenile salmon swimming in forebays, tailraces, and free-flowing rivers. The model integrates information on juvenile salmon behavior at both field and laboratory scale and literature on juvenile salmon swimming capabilities.

Simulated fish swim paths are determined by solving Newton's Second Law. Most of the model parameters are represented by probability distributions. Behavioral responses are triggered for the most part by the flow acceleration and pressure. The model uses conditional probability distributions of thrust magnitude and direction, given flow acceleration. Simulated fish select a swimming direction referenced to the flow velocity vector. To consider juvenile salmon's tendency to coast with the flow, the model intersperses periods of active swimming and gliding. Chinook salmon measured swim paths were analyzed. The flow variables at the fish locations were obtained from CFD simulations. Juvenile salmon mean thrust was determined from solving Newton's Second Law at every measured location. Results show that as flow acceleration increases, the juvenile salmon average thrust increases and the probability of gliding decreases. Chinook salmon tend to migrate tail-first as flow acceleration increases. For the flow accelerations of 5x10-4 m/s2 and 1x10-2 m/s2, approximately 85% and 95% of the analyzed fish migrated tail-first, respectively. The model capacity to predict fish migration route selection, fish-like trajectories, and residence times was tested at two hydropower dams. On average, migration routes were predicted with 17 percent of relative error. Model predictions for fish average residence times were within 10 percent of measured values.