Applications of the Sensor Fish Technology

The sensor fish – a fish-sized sensor package invented by Pacific Northwest National Laboratory – is used to better understand the physical experience of fish during passage through hydroelectric turbines and dam bypasses. The sensor fish provides data for developing more “fish-friendly” facilities.

By Zhiqun Deng, Thomas J. Carlson, Joanne P. Duncan, and Marshall C. Richmond

The sensor fish is an autonomous device Pacific Northwest National Laboratory (PNNL) developed for the U.S. Department of Energy (DOE) and the U.S. Army Corps of Engineers. The device is intended to help researchers better understand the physical conditions fish experience during passage through hydroelectric turbines and dam bypasses, such as spillways.

Since its development in 1997, the sensor fish has undergone several changes to improve its function and extend its range of use. The most recent design – a “six-degree-of-freedom” device deployed in February 2005 – is being used to study passage of fish at hydro projects along the Columbia and Snake rivers in the Pacific Northwest.

Understanding the sensor fish

To develop more “fish-friendly” hydro facilities, it is necessary to characterize the conditions fish experience while passing through complex hydraulic environments. Researchers also must identify the locations and operations where conditions are severe enough to injure fish. Four general approaches are being applied to develop safer turbines and spillways in the Pacific Northwest:

  1. ) Evaluation of turbine passage injury and mortality rates and other aspects of injury risk in field studies using balloon tags, passive integrated transponder (PIT) tags, and radio tracking;1,2
  2. ) Using hydro-acoustics to monitor passage and infer passage efficiency (the percent of fish that pass a dam via a bypass or other passage facility);3
  3. ) Laboratory studies to establish the biological criteria for fish injury and mortality (this is done by quantifying the hydraulic forces required to produce the biological responses observed in field studies);4,5,6 and
  4. ) Applying reduced-scale physical models and computational fluid dynamics (CFD) modeling techniques to simulate pressure and velocity and evaluate design and operation alternatives.7

While valuable, these approaches do not provide information on the specific hydraulic conditions or physical stresses fish experience or the specific causes of biological responses. To measure conditions in situ, PNNL created the sensor fish as an internal development initiative in 1997. This functional prototype was field tested in spring 1999 and used extensively during the winter of 1999-2000 to evaluate the first minimum gap runner installed at Bonneville Dam’s 532-MW First Powerhouse on the Columbia River. The DOE Wind/Hydropower Program has sponsored ongoing sensor fish development activities, and the Corps has supported many field trials using this technology.

The first sensor fish prototype was molded from di-electric polymer into the size and shape of a yearling Chinook salmon smolt. (See left side of Figure 1 on page 36.) This prototype successfully demonstrated proof-of-concept for acquiring complex hydraulic data during turbine passage using an autonomous device.8 For the second generation device, PNNL sought to improve data downloading and accessibility of electronics inside the device. To this end, PNNL chose PVC (polyvinyl chloride) tube housing with screw-in end caps for this device, and designed and built a cylindrical sensor fish using sturdy clear polycarbonate plastic in May 2000.

Developing the “six-degree-of-freedom” sensor fish

The current model (see right side of Figure 1 on page 36) of the sensor fish was deployed in February 2005. During the design process, PNNL developed governing equations of motion for the sensor fish and simulated them. The intent was to understand the design implications of instrument selection and placement within the device.9 PNNL also conducted experimental and computational studies to develop a set of force and moment relationships that are specific to the sensor fish body over the range of flow it experiences. These relationships are used to interpret measurements of sensor fish motion.

Figure 1: The first sensor fish prototype (left) was molded from di-electric polymer into the size and shape of a yearling chinook salmon smolt. The new six-degree-of-freedom sensor fish (right) is a cylindrical device made of sturdy clear polycarbonate plastic.
Click here to enlarge image

The six-degree-of-freedom sensor fish measures six components: three components of linear acceleration (up-down, forward-back, and side-to-side) and three components of angular velocity (pitch, roll, and yaw). The sensor fish also measures pressure and temperature. The new sensor fish uses a sampling frequency of 2,000 Hertz (Hz) and a recording time of about 4 minutes (233 seconds). This compares to a sampling frequency of 200 Hz and a recording time of about 2 minutes with the previous model. This increased sampling frequency allows more accurate capture of the rapid change of motion and pressure that occurs during passage through hydroelectric turbines.

The six-degree-of-freedom sensor fish is 0.96 inch in diameter and 3.54 inches long and weighs about 43 grams. These measurements closely match the dimensions of a yearling salmon smolt. Like a fish, the new sensor fish design is nearly neutrally buoyant in fresh water.

In actual use, the sensor fish is only one part of a system necessary to acquire data on hydraulic conditions. The system consists of modules that charge the sensor fish’s internal battery, program the sensor settings, acquire data, convert the analog signal to digital form, download and analyze the data, and interpret the results. Data acquired during use of the sensor fish are stored in an internal memory card and transferred to computers via a wireless infrared link with an external modem.

Use of the sensor fish requires calibration of several internal devices. To calibrate the pressure transducer, PNNL uses a hyperbaric chamber programmed to simulate the rapid pressure changes fish encounter when passing through a large turbine or spillway. Pressure data measured by sensor fish during this exposure are compared to the programmed pressure-time histories of the hyperbaric chamber to obtain calibration coefficients.

PNNL calibrates acceleration measurements of the device using an acceleration test mechanism. Each axis of the tri-axial accelerometer set is individually checked, and sensor fish movement is recorded using high-speed videography. Equipment used is the PCI FastCAM 1280, supplied by Photron USA Inc. in San Diego, Calif. Researchers conduct advanced motion analysis to obtain trajectories, from which velocity and acceleration are computed.

Finally, calibration of the tri-axial rotation sensors is conducted in a rotation test mechanism in a manner similar to that for acceleration. The relative errors of the linear acceleration and angular velocity measurements are less than 5 percent.

For retrieval after passage during field studies, the sensor fish is equipped with a radio transmitter supplied by Advanced Telemetry Systems in Isanti, Minn., and HI-Z balloon tags from Normandeau Associates Inc. in Bedford, N.H. Contact with water activates a chemical-filled capsule that produces gas, inflating the balloons sufficiently to bring the device to the surface minutes after its release. Researchers then use a directional radio receiver antenna to home in on the radio transmitter attached to the sensor fish.

Sensor fish and live fish are deployed concurrently through an injection pipe mechanism. In this way, physical conditions measured by the sensor fish are linked to the actual injury and mortality results of live fish. However, sensor fish are not fish. The device cannot swim or adjust its position. Unless the sensor fish is embedded in a fish, it does not quantify the exact conditions a specific fish has sustained. To bridge the gap, tests on concurrently released sensor fish and live fish also were conducted in a well-defined laboratory environment. From these laboratory tests, a response relationship between live fish injury and sensor fish measurement was developed for exposure to turbulent shear flow.

Applications of the current model

The sensor fish has evolved to be a major tool for improving understanding of injury mechanisms of fish passing through dams with hydro facilities in the Pacific Northwest. It has been deployed at 1,090-MW Bonneville, 1,812.8-MW The Dalles, 6,809-MW Grand Coulee, 603-MW Ice Harbor, 2,160-MW John Day, 810-MW Lower Monumental, 980-MW McNary, and 1,038-MW Wanapum.

Results using early versions of the sensor fish already have been reported.10 Here, we discuss three representative applications of the six-degree-of-freedom sensor fish, including a turbine environment at Wanapum, a spillway deployment at Lower Monumental, and a laboratory study.

Turbine passage at Wanapum Dam

After years of economic, engineering, and biological study and assessment, Grant County Public Utility District (PUD) began installing a novel Kaplan turbine at Wanapum Dam on the mid-Columbia River in April 2004. The new turbine, Unit 8, features many elements of the advanced-design concept. It was designed to increase power production and to provide better passage conditions for fish. Installation was completed in early 2005, and Grant County PUD sponsored an investigation to evaluate biological performance of the new turbine relative to an existing turbine unit.

Each passage condition sampled for this study was one of 24 unique treatments consisting of: 1) one of four target turbine discharges [9,000, 11,000, 15,000, and 17,000 cubic feet per second (cfs)]; 2) one of three turbine intake bays; and 3) one of two intake injection elevations. Researchers released one sensor fish after each test lot of ten live fish. The study involved the use of about 10,000 live test fish and 1,000 sensor fish releases.

Sensor fish pressure and acceleration measurements have characteristic signatures for particular events that are helpful in identifying the occurrence and location of severe events, such as strikes. Readily identifiable information related to these signatures includes time of passage from the injection pipe exit to the turbine intake, through the turbine stay vane-wicket gate cascade, upstream of the runner, through the runner, through the runner wake, and through the draft tube. (See Figure 2 on page 38.) Researchers use these signature events to estimate the probable location and time of collision or shear exposure events. For example, a collision with a stay vane can be identified from the corresponding spike of pressure or acceleration. This collision also may be confirmed by the immediate increase of the sensor fish rotational velocity. With such a large impact (in the example, an acceleration of 160 times gravity’s acceleration), there is a high probability of fish being injured.

Detailed analysis of the events permits detailed quantitative and statistical assessment that provides additional information for biological evaluation of advanced-design and existing turbines.

Spillway passage at Lower Monumental Dam

Lower Monumental Dam, on the Snake River in south-central Washington, is 3,791 feet long, with an effective height of 100 feet. The dam features an eight-bay spillway. The powerhouse contains six turbine-generating units.

Figure 2: Measurements taken during passage of the new six-degree-of-freedom sensor fish through the advanced-design turbine at Wanapum Dam show a spike in pressure (black line), linear acceleration (red line), and rotational velocity (blue line). These spikes indicate a collision with a stay vane.
Click here to enlarge image

Fish management agencies list spillways as a preferred alternative for downstream passage of migrating juvenile fish because of their generally higher survival rate. However, spillway operations can increase total dissolved gas (TDG) concentrations downstream, potentially exposing fish to supersaturated gas conditions and the associated risk of gas bubble trauma. As a result, many hydro project owners have constructed spillway deflectors and divider walls to reduce production of TDG.

The objective of the study at Lower Monumental Dam was to characterize the physical conditions fish would experience during spill passage from specific release locations at specific spill discharge levels, as well as the effects of the addition of new structures, such as deflectors.

Researchers conducted 123 sensor fish releases from two spillways (Bay 7 and Bay 8) at two elevations (deep and shallow water). Results indicated that rapid pressure changes occurred during passage through a spillway tainter gate. The rate of pressure change during gate passage was dependent on the release elevation of the sensor fish and, consequently, the location of the fish relative to the bottom edge of the tainter gate. Specifically, the deeper sensor fish releases experienced a lower rate of pressure change during gate passage compared to the shallow releases. Rapid pressure changes can cause direct mortality and sub-lethal physical injuries to juvenile fish. Temporary disability or stunning, as well as other sub-lethal injuries, have been documented in laboratory studies of pressure cycling and observed in balloon tagged fish after turbine passage.11 Sub-lethal injuries of sufficient severity could be factors in indirect fish mortality from predation or delayed mortality.

Figure 3: Researchers compared sensor fish measurements with computational fluid dynamic (CFD) simulation results during a spillway passage. Results show that the pressure-time histories and general features of the acceleration magnitude time histories are similar to those from CFD simulation.
Click here to enlarge image

Acceleration measurements at Lower Monumental Dam supported the observation that the lowest rates of pressure change occurred for fish released in deeper water. The maximum acceleration values during gate passage were much smaller for sensor fish released in deep water than for those released in shallow water. CFD simulation and tracking of numerical particles also were performed to provide the timing information (i.e., where sensor fish are at a certain time) and further pinpoint the location of sensor fish during such severe events as collisions and strong shear exposures. The signatures of sensor fish spillway passage from upstream of the gate to the stilling basin and the relationship between rate of pressure change and acceleration also were evident in the results from the numerical simulations. (See Figure 3.)

Laboratory studies

To establish correlation metrics between sensor fish measurements and live fish injuries, researchers exposed sensor fish and hatchery-raised fall chinook salmon to turbulent shear flows in the laboratory. Two synchronized high-speed, high-resolution cameras captured digital video images of the exposures. Researchers performed advanced motion analysis to obtain three-dimensional trajectories of sensor fish and juvenile salmon. Time series of the velocity, acceleration, fish body bending, and total force magnitudes were calculated from the trajectories. A complete exposure consisted of a before-exposure zone, a high turbulence/shear zone, and an ambient flow zone. (See Figure 4 on page 40.) In the before-exposure zone, the sensor fish was rotating slowly and had minimum acceleration. Entering the high turbulence/shear zone caused both acceleration and rotational velocity to increase rapidly. The live fish had similar response to exposure.

Figure 4: Researchers exposed sensor fish and hatchery-raised fall chinook to turbulent shear flows in the laboratory to establish correlation metrics between sensor fish measurements and live fish injuries. This figure shows the results from sensor fish measurement. The black line is acceleration and the red line is rotational velocity.
Click here to enlarge image

For sensor fish passing hydro turbines (See Figure 2), passage occurs in the same fashion: the before-exposure zone (in the intake bay approaching the turbine runner), the high turbulence/shear zone (wicket gate and runner vicinity), and the draft tube zone (corresponding to ambient flow in the lab studies). Therefore, the signatures of the sensor fish releases in laboratory-generated turbulent shear flows were similar to those of sensor fish passing turbines.

Researchers used binomial logistic regression analysis to statistically establish a relationship between the expected biological injury rate and a subset of the computed kinematic and dynamic parameters. Researchers discovered that the bulk acceleration of the fish was the best predictor for all fish injury types and overall injury level. Researchers then conducted quantile regression analysis to correlate live fish and sensor fish accelerations. Finally, researchers repeated the binomial logistic regression to relate the probability of specific biological responses to sensor fish measurements.

Ongoing studies focus on relating the findings of field investigations and CFD modeling to the sensor fish measurements in the turbine environment, using the correlation metrics developed in the lab studies.

Future work on the sensor fish

With this latest generation of sensor fish, researchers and fishery managers have the opportunity to gather scientific data on the physical environment juvenile fish experience when traveling past a dam and associated hydroelectric plant. Correlation metrics developed in the lab environment provide a link between laboratory and field, as well as the correlation of sensor fish measurements and live fish injury/mortality observations in field studies. This information is leading to the design, testing, and operation of more fish-friendly turbines that will balance the use of the Pacific Northwest’s hydropower resource with the protection of its migrating fish.

Future development will include:

– Closer coupling of sensor fish with CFD simulations to determine the flow fields sensor fish go through; and

– CFD tracking of numerical six-degree-of-freedom particles, which are of the same size and mass as the sensor fish.

The authors may be reached at Pacific Northwest National Laboratory, P.O. Box 999, Richland, WA 99352; (1)-509-372-6120 (Deng), (1) 503-417-7562 (Carlson), (1) 509-376-7899 (Duncan), or (1) 509-372-6241(Richmond); E-mail:, thomas.carlson@,, or


  1. eisey, P.G., Dilip Mathur, and T. Rineer, “A Reliable Tag-Recapture Technique for Estimating Turbine Passage Survival: Application to Young-of-the-Year American Shad (Alosa sapidissima),” Canadian Journal of Fisheries and Aquatic Sciences, Volume 49, No. 9, September 1992, pages 1,826-1,834.
  2. athur, Dilip, P.G. Heisey, E.T. Euston, J.R. Skalski, and S. Hays, “Turbine Passage Survival Estimation for Chinook Salmon Smelts (Oncorhynchus tshawytscha) at a Large Dam on the Columbia River,” Canadian Journal of Fisheries and Aquatic Sciences, Volume 53, No. 3, March 1996, pages 542-549.
  3. kalski, J.R., G.E. Johnson, C.M. Sullivan, E. Kudera, and M.W. Erho, “Statistical Evaluation of Turbine Bypass Efficiency at Wells Dam on the Columbia River, Washington,” Canadian Journal of Fisheries and Aquatic Sciences, Volume 53, No 10, October 1996, pages 2,188-2,198.
  4. urnpenny, A.W.H., “Mechanisms of Fish Damage in Low-Head Turbines: An Experimental Appraisal,” Fish Migration and Fish Bypasses, M. Jungwirth, S. Schmutz, and S. Weiss, eds., Blackwell Publishing, Oxford, United Kingdom, 1998, pages 300-314.
  5. eitzel, D.A., et al, “Survival Estimates for Juvenile Fish Subjected to a Laboratory-Generated Shear Environment,” Transactions of the American Fisheries Society, Volume 133, No. 2, March 2004, pages 447-454.
  6. eng, Zhiqun, et al, “Evaluation of Fish-Injury Mechanisms during Exposure to Turbulent Shear Flow,” Canadian Journal of Fisheries and Aquatic Sciences, Volume 62, No. 7, July 2005, pages 1,513-1,522.
  7. Rakowski, Cindy L., Laurie L. Ebner, and Marshall C. Richmond, “Using CFD to Study Juvenile Salmon Bypass at Bonneville,” Hydro Review, Volume 24, No. 1, March 2005, pages 56-65.
  8. Carlson, Thomas J., and Joanne P. Duncan, “Evolution of the Sensor Fish Device for Measuring Physical Conditions in Severe Hydraulic Environments,” DOE/ID-11079, Pacific Northwest National Laboratory, Richland, Wash., 2003.
  9. Deng, Zhiqun, Marshall C. Richmond, C.S. Simmons, and Thomas J. Carlson, “Six Degrees of Freedom Sensor Fish Design: Governing Equations and Motion Modeling,” PNNL-14779, Pacific Northwest National Laboratory, Richland, Wash., 2004.
  10. Carlson, Thomas J., Joanne P. Duncan, and Theresa L. Gilbride, “The Senor Fish: Measuring Fish Passage in Severe Hydraulic Conditions,” Hydro Review, Volume 22, No. 3, June 2003, pages 62-69.<
  11. Abernethy, C.S., B.G. Amidan, and Glenn F. Cada, “Laboratory Studies of the Effects of Pressure and Dissolved Gas Supersaturation on Turbine-Passed Fish,” DOE/ID-10853, U.S. Department of Energy, Idaho Operations Office, Idaho Falls, Idaho, 2001.


The work described in this article was conducted at Pacific Northwest National Laboratory (PNNL) in Richland, Wash., which is operated by Battelle Memorial Institute for the U.S. Department of Energy (DOE). We thank Jim Ahlgrimm from DOE’s Hydropower program and Dennis Dauble at PNNL for their leadership and support of this research.

Zhiqun (Daniel) Deng, PhD, is research scientist with Pacific Northwest National Laboratory (PNNL). Tom Carlson, PhD, is chief scientist with the ecology group of PNNL. Joanne Duncan is research scientist with PNNL. Marshall Richmond, PhD, is chief engineer with the hydrology group of PNNL.

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: