| Downstream passage of outmigrating juvenile salmon (migrants) at hydropower dams on the Columbia and Snake Rivers has, generally, not been entirely successful. To date, research in the region has generated considerable, but often inconclusive, results as to the factors influencing observed migrant guidance and passage. This dissertation synthesizes existing fisheries and sensory biology literature together with fundamentals of fluid mechanics and river hydrogeomorphology to develop a theoretically rigorous hypothesis of observed migrant movement behavior at Lower Granite Dam on the Snake River, Washington USA (chapter 1). The hypothesis, the Strain-Velocity-Pressure (SVP) Hypothesis, is supported by existing 3-D acoustic-tag telemetry data, multi-beam hydroacoustic passage data, 3-D computational fluid dynamics (CFD) modeling, and 3-D spatially-explicit, time-varying virtual fish simulation that outputs model results in similar form to data collected by field instrumentation.;The virtual fish model, a coupled Eulerian-Lagrangian agent- and individual-based model (CEL Agent IBM), is a simple, yet theoretically- and computationally-robust, means for simulating individual fish movement behavior. The CEL Agent IBM framework (chapter 2) is based on well-established principles in computer science, foraging theory, and CFD modeling. Virtual fish behavioral rules are biological tractable and mechanistic (i.e., rule-based) in nature. Therefore, one can “plug-and-play” a rule-based hypothesis of fish movement behavior for objective evaluation. In addition, coefficients can be further calibrated or validated in laboratory experiments and rule structures emerging from back-casting simulation analyses are amendable to forecast simulation.;The CEL Agent IBM for migrants at Lower Granite Dam, i.e., the Numerical Fish Surrogate (NFS), is developed from four independent CFD model and biological data sets from year 2000 studies. The NFS was then ‘blindly’ validated using seven independent CFD modeled flow conditions from year 2002 studies, i.e., with no prior knowledge of 2002 biological results. NFS predictions from the ‘blind’ validation runs were first compared to 2002 biological data by an independent entity. The NFS and embodied SVP Hypothesis capture 80% (r2 = 0.80) of the observed variation in 2000 and 2002 migrant passage and 85% (r2 = 0.85) of the observed variation in dawn, dusk, and nighttime passage, i.e., when migrant visual acuity is limited. |
