Save Article Instructions

Assessing Fish Bypass Flow: Combining Numerical and Physical Models

To determine the flow performance of a fish bypass structure proposed for 1,038-MW Wanapum Dam, researchers combined results from a physical model with numerical modeling. Using these two tools allowed researchers to more accurately replicate and study the complex geometry of the structure and to take advantage of each model’s strengths.

By Cagri K. Turan, Pablo M. Carrica, Troy C. Lyons, Duncan Hay, and Larry J. Weber

The 1,038-MW Wanapum Dam project is one of many hydro projects on the Columbia River in Washington that has affected anadromous fish.1 To counter these effects, regulatory agencies have focused on promoting development of fish passage facilities that minimize juvenile mortality.2

One such facility being tested by IIHR – Hydroscience and Engineering is an ogee-crested spillway that will act as a passage route from the forebay to the tailrace of Wanapum Dam. This testing combines use of a physical model and computational fluid dynamics (CFD) modeling. Together, these two methods of analysis give more information. However, this combination of methods is harder to conduct because the data should be validated for both the CFD and physical models.

First, in 2003, researchers developed a 1:50 scale model of the Wanapum Dam forebay to analyze large-scale forebay hydraulics, flow patterns, and entrance conditions. Then, in 2004, a 1:24 scale model was built to obtain detailed information for the spillway entrance, chute, and exit hydraulics and to design the control gates and air vents.3 The data in this last model was used to calibrate the bypass design and is presented in this article.

Next, CFD simulations were performed using Fluent 6.1 software.4 The VOF method, an interface capturing technique, was used to model the free surface flow through the spillway. The complex geometry of the spillway was modeled using hybrid structured/unstructured grids. The model was validated against data collected from the physical model for flow rates at different headwater elevations and gate settings, free surface elevations, pressure distribution at the ogee surface, and velocities at selected regions.

Once proven accurate, the CFD model was used to calculate information, including pressure and velocity distributions not available with the physical model. This information guided the design of the spillway in the assessment of cavitation potential, the analysis of inlet conditions in order to predict fish capture efficiency, and the inference of fish stress at different locations.

Design goals for the spillway

Project owner Grant County Public Utility District (PUD) wants to achieve 95 percent survival of anadromous salmonid smolts passing Wanapum Dam via a combination of turbine and non-turbine passage routes. The ogee crested spillway is intended to provide a non-turbine passage route that will maximize both the usage and survival rate of fish. Design discharge for the spillway was set at a nominal value of 566.3 cubic meters per second (cms). However, the ability to regulate flow was incorporated in the design, given the possibility of achieving the 95 percent survival at lower flows. Design goals for this spillway were set in consultation with the National Marine Fisheries Service (NMFS).

Based on the premise that powerhouse flows would draw smolts to the vicinity of the bypass, the spillway was located adjacent to the powerhouse on an unused intake bay.

Traditional hydraulic design considerations for the spillway include flow capacity, pressure, and cavitation potential. Hydraulic design considerations unique to this spillway include oversized flow fairing (to reduce drag), a 1:1 sloped floor upstream of the ogee crest, an emergency stoplog location, angled control gates, nappe air vent openings, elliptical expansion of the downstream sidewalls, a crowned apron to spread bypass flows, and cantilevered side wall extensions to minimize jet submergence in the tailrace. (See Figure 1 on page 42.)

The entrance was designed to provide access over a considerable range of depths so that smolts entering the spillway would not need to change elevation in the forebay. (See streamlines on Figure 2 on page 46.) At the same time, it was desirable to have unimpeded flow from the surface of the forebay that led to a typical ogee crest for the spillway.

Figure 1: The prototype fish bypass for 1,038-MW Wanapum Dam will be located adjacent to the powerhouse on an unused intake bay, to take advantage of attraction flows from the powerhouse intakes. (Courtesy Jacobs Civil Inc.)
Click here to enlarge image

To meet these two considerations, the entrance invert was ramped upward toward the spillway crest, forcing smolts that entered at depth to change elevation.

To prevent fish from escaping the spillway, the entrance was designed to develop a high flow velocity before the start of the ramp, thus trapping smolts. The corners formed by the walls and entrance invert were rounded leading up to the spillway crest to reduce the likelihood of smolts holding in the corners or leaving the spillway from an area of lower velocity. Considerable care was applied in the entrance design to avoid developing flow conditions that might lead smolts to reject the spillway, such as upwelling, flow separations, and vorticity.

The spillway exit also merited several hydraulic design considerations, including limiting the increase in percent saturation of dissolved gas in the tailrace, minimizing riverbed and bank erosion, and improving fish survival.3 In May 2004, exit flow conditions for the spillway design were field tested at Wanapum Dam by releasing balloon-tagged smolts in the existing ice-trash sluiceway, which was modified to give a unit discharge of 20.7 cms per meter from the forebay to match the fish bypass design flow density. Data showed 100 percent smolt survival and indicated injury due to minor abrasions was less than 1 percent.5,6

The regulating gates for the spillway were another unique element of the design. The spillway will be equipped with two sets of gates. The first will be cable-supported bulkhead gates near the dam face, to be deployed under a balanced head but also designed for emergency closure. The second is regulating gates near the spillway crest, designed to achieve three hydraulic objectives:

  1. ) Permit flows of less than 566.3 cms, namely 424.7, 283.2, 141.6, and 56.6 cms, respectively. The latter flow is similar to the magnitude of releases from the existing sluiceway;
  2. ) Draw flow from the top of the forebay so that surface-oriented smolts do not have to sound and pass under a gate; and
  3. ) Have the trajectory of the top-spill flow intercept the spillway chute at a flat angle, tangential to the chute.

A 2.54-centimeter-wide air slot downstream of the regulating gate slots provides air to the underside of the discharge nappe during regulated flows. The air slots run continuously over the length of the gate slots to provide air access during all operating conditions, to avoid large pressure gradients and maintain the nappe smooth.

Developing and testing the physical model

Froude number scaling relationships were used to calculate expressions relating model and prototype values. Geometrical similarity with a length scale of 1:24 was selected for the physical model. Using spillway width and chute velocity, typical Reynolds (Re) numbers on the model scale were well over 106. For the relationship between Reynolds numbers, the Re for the prototype is 117.6 times the Re for the physical model. Because the Re is very high, scaling effects will not be important, except very near the walls.

As in the case of the Reynolds number, the Weber (We) number similitude was not achieved. For the relationship between Weber numbers, the We for the prototype is 576 times the We for the physical model. This mismatch on the Weber number affects the scalability of air entrainment processes.

The physical model consisted of a head box, the spillway structure, and a tailrace box.3 The head box was a rectangular tank 4.9 meters long by 3 meters wide by 2.4 meters tall. A 75-horsepower variable frequency drive pump conveyed water to the head box through a 25.4-centimeter-diameter pipe, terminating in a 36-centimeter-diameter submerged diffuser pipe. The diffuser pipe was sealed at each end and had uniformly spaced outlet holes, creating a uniform flow distribution over the width of the head box. A 2.4-meter-high porous wall was installed downstream of the diffuser, creating a settling chamber to reduce turbulence at the upstream end of the tank. An additional porous wall was installed horizontally in the settling chamber to dissipate energy vertically. A 1.2-meter by 2.4-meter floating timber mat was positioned just downstream of the porous wall to further dampen surface waves within the head box.

A 30.5-centimeter-high false floor was constructed in the head box to raise the floor to prototype elevation 140.3 meters, approximately corresponding to the Wanapum Dam forebay bed elevation at the spillway location. Powerhouse pier noses were installed on the downstream wall of the head box to replicate the surface geometry of the powerhouse and future units.

Figure 2: The free surface topology (top) measured using the computational grid (bottom) shows that the various streamlines follow a smooth path and do not cross each other. Blue is water coming from the bottom of the forebay, red comes from close to the free surface, and yellow and green streamlines originate at intermediate depths in the forebay.
Click here to enlarge image

The bypass structure was fabricated and supported with dimensionally stable materials. The ogee profile was manufactured from polyvinylchloride (PVC) sheeting using computer numerically controlled milling machines. The ogee surface was replicated by forming stainless steel sheeting over the PVC templates. Seven 1.159-millimeter-diameter holes were drilled along the centerline of the ogee and fitted with pressure taps to measure gage pressures. Two coats of white paint were applied to the stainless steel ogee surface to provide a smooth, uniform finish. The chute sidewalls were fabricated from Plexiglas and were reinforced with Plexiglas members to prevent deformation under hydraulic loading. A removable Plexiglas panel was integrated into each side- wall to allow for gate slot modifications. Gate slots were milled into the Plexiglas sidewalls, and the regulating gate segments were milled from Plexiglas sheeting. Air vents and air supply chimneys also were milled from Plexiglas.

Downstream, the apron sidewalls and floor were fabricated from PVC sheeting. The crowned flow spreader was formed from concrete and given a smooth finish with waterproof auto body filler and paint. To minimize flow disturbances in the chute, all sidewall construction joints were smoothed with filler and then sanded.

The tailrace box was a rectangular tank 7.3 meters long by 3.7 meters wide by 1 meter deep. Part of the right wall was constructed of Plexiglas to provide visibility beneath the water surface. The basin was filled with a 0.3-meter-thick layer of 2-centimeter-nominal-diameter washed gravel to approximately replicate the tailrace floor elevation of 137.2 meters. An overflow tailgate provided tailwater elevation control.

Experimental equipment measured model inflow, headwater elevation, tailwater elevation, ogee pressure, velocities at selected regions, and volumetric airflow. Model flow rate was measured with a calibrated annular orifice meter installed in the supply pipe. Headwater and tailwater elevations were measured with hook-type point gages installed inside acrylic pipes to dampen water surface fluctuations. Tailwater elevations were measured immediately to the left of the bypass, corresponding approximately to the field measurement location. Ogee crest pressures were measured on a piezometer panel equipped with a vernier scale, referenced directly to the headwater point gage. Both were referenced to the ogee crest using a GTS-226 total station, an optical positioning instrument from Topcon Positioning Systems Inc. Velocities were measured with an acoustic doppler velocimeter (ADV) supplied by SonTek Inc. or a Pitot tube (for locations where the ADV saturated as a result of velocities above the ADV operational limit).

Experimental runs were performed for four different headwater elevations, including the normal pool elevation (173.74 meters), the highest expected elevation (175.26 meters), and two lower elevations (170.69 meters and 172.21 meters), chosen from typical possible operational conditions. Design conditions were investigated for free flow (ungated) and with the vertical gate raised, then with one, two, or three of the inclined gates inserted within the inclined gate slots, all for headwater elevation 173.74 meters.

Developing and testing the CFD model

The numerical model was constructed at prototype scale in order to obtain valid predicted stresses. The grid used was hybrid (structured/unstructured), with about 1 million nodes. Typical run times were about 24 to 72 hours in eight processors of a Linux Cluster parallel machine.

Figure 2 and Figure 3 (on page 48) show the numerically computed free surface geometry and a close-up of the flow near the gate slots. The colored ribbons are streamlines on the core of the flow, showing little perturbation by the gate and air slots. Note the free surface deflection on the gate slots caused by strong vortices pushing the fluid down. These vortices are shown in Figure 3 with black ribbons.

Figure 3: The bypass contains five gate slots, two vertical (at left) and three inclined (at right). The black streamlines indicate recirculations inside the gate slots. The gates in the vertical slots are introduced first to regulate the flow rate. The inclined slots contain gates that can be used to further reduce the flow rate.
Click here to enlarge image

The air slots also entrained air into the flow and had a lower free surface elevation. CFD predictions of air entrainment and vertical structures were in good qualitative agreement with visual observations from the physical model. However, the bubbles predicted by the CFD model were much larger than those entrained in the laboratory. In addition, some air entrainment was observed on the physical model gate slots that was not predicted by the CFD model. This was a natural consequence of the grid spacing on the CFD model being much greater than the physical model scale bubble size, thus preventing the CFD model from resolving the bubble evolution.

To validate the CFD model, the data from the physical model was scaled to prototype equivalents and compared with the CFD predictions. Figure 4 shows a comparison of measured and predicted rating curves. The agreement, obtained with no adjustable parameters, was excellent. The difference never exceeded 2 percent, which is on the order of the uncertainty in the data from the physical model.

For all conditions, free surface profiles predicted by the CFD model were compared to results from the physical model. For all cases, the agreement between CFD and the physical model was very close.

Numerical gage pressure values at the ogee symmetry line were compared with the physical model results for headwater elevations 173.74 meters and 175.26 meters in Figure 5. The minimum measured pressure is accurately predicted with the CFD model. The CFD model detected a localized low pressure peak at the start of the ogee curve upstream of the crest, due to the discontinuity of the slope at that location.

Figure 4: Rating curves developed for performance of the physical and CFD models show excellent agreement. After scaling up data from the physical model to prototype equivalents, the difference between the two never exceeded 2 percent.
Click here to enlarge image

Velocity predictions also compared well against experimental measurements. At the inlet, reasonable agreement against ADV measurements was found for all three components of velocity. At the ogee crest, the x-velocity in a vertical plane was compared against experimental data measured with a Pitot tube. The agreement between CFD and the physical model was very good, with some discrepancy near the free surface, where CFD tends to slightly under-predict velocity. The reasons for this discrepancy have not been quantified.


Researchers used the CFD model to determine the cavitation number and velocities for the bypass.

Cavitation number

The cavitation number, Ca, was calculated for all the operational conditions according to:

Equation 1:


Smaller Ca values represent higher cavitation probability. Cavitation potential was evaluated for unregulated flow with headwater elevation 175.26 meters, which is the worst case. Because the velocity is exactly zero at the bottom wall, the velocity in the cavitation number equation was evaluated at the end of the boundary layer. The cavitation number was lowest at the ogee crest and at the end of the apron, where velocities are high and absolute pressure is small due to the small water depth. The minimum value of Ca at the crest was 0.58, while at the end of the apron where the velocity was highest Ca was 0.37. These values were within Ca .0.30, the acceptable limit taken for this project.


The CFD model was interrogated to check that flow conditions were appropriate to avoid unnecessary stress on the fish, including large pressure and velocity gradient, excessive accelerations, and streamline curvatures. These requirements condition the design of the fish bypass and are very difficult to verify experimentally.

Concluding remarks

This integrated study of the flow on a fish bypass and the design criteria leading to the bypass included free surface elevations, pressure, velocity, and discharges measured on a 1:24 scale model. A fully three-dimensional CFD model was set up and validated against measured discharges and free surface elevations at two headwater elevations for unregulated flow and at nominal pool elevation for regulated flows. Pressure distributions and velocities at selected locations were also used for validation. The comparison of the CFD model results with data from the physical model showed excellent agreement for all variables. Cavitation indexes and velocity fields were also reported, showing that the design did not lead to extreme conditions that might cause cavitation or large stress on fish.

Though only the final design is discussed in this paper, several designs were evaluated during the course of the study.

Mr. Turan may be reached at Bechtel Power, 5275 Westview Drive, Frederick, MD 21703; (1) 301-228-6000; E-mail: Messrs. Carrica, Lyons, and Weber may be reached at IIHR – Hydroscience and Engineering, The University of Iowa, Iowa City, IA 52242; (1) 319-335-5237; E-mail:, troy-lyons@uiowa. edu, or Mr. Hay may be reached at Oakwood Consulting, 237 Turtlehead Road, Belcarra, British Columbia V3H 4P3 Canada; (1) 604-936-5161; E-mail: duncanhay@


Dotson, Curtis L., and Kathleen R. Kiefer, “Future Unit Fish Bypass Planned for Wanapum Dam,” Hydro Review, Volume 25, No. 6, October 2006, pages 22-25.


  1. National Research Council, Upstream: Salmon and Society in the Pacific Northwest, National Academic Press, Washington, D.C., 1996.
  2. Weitkamp, D.E., “Hydraulic Models as a Guide to Fish Passage Design,” Proceedings of Issues and Directions in Hydraulics, The University of Iowa, Iowa City, Ia., pages 287-293, 1995.
  3. Lyons, Troy C., P.T. Haug, Pablo M. Carrica, and Larry J. Weber, “Hydraulic Model Studies for Fish Diversion at Wanapum/Priest Rapids Devel- opment Part XXII: Model Studies of a Future Unit Fish Bypass for Wanapum Dam,” IIHR-Hydroscience & Engineering, Report No. 325, 2005.
  4. Fluent 6.1 User’s Guide, Fluent Inc., Lebanon., N.H., 2003.
  5. “Characterization of Sluiceway Chute Conditions at Wanapum Dam, Columbia River, Washington, 2004,” Limited Distribution Report PNWD-3490, Prepared by Batelle for Normandeau Associates, 2004.
  6. “Estimation of Direct Survival and Condition of Juvenile Salmonids Passed by a Prototype Bypass Sluice Chute at Wanapum Dam,” Prepared by Normandeau Associates for Grant County PUD No. 2, 2005.


The authors greatly appreciate the support by Grant County Public Utility District No. 2 of Grant County, Wash.

Cagri Turan, formerly with IIHR-Hydroscience and Engineering at the University of Iowa, is an engineer with Bechtel Corp. Pablo Carrica is associate professor and research engineer, Troy Lyons, P.E., is research engineer, and Larry Weber, PhD, P.E., is director with IIHR-Hydroscience and Engineering. Duncan Hay, P.Eng., is president of Oakwood Consulting Inc. Turan and Carrica worked on the numerical model, Lyons and Weber worked on the physical model, and Hay worked on the bypass design.

µ Peer Reviewed

This article has been evaluated and edited in accordance with reviews conducted by two or more professionals who have relevant expertise. These peer reviewers judge manuscripts for technical accuracy, usefulness, and overall importance within the hydroelectric industry.

To access this Article, go to: