A large dataset of detection and submeter-accurate 3-D trajectories of juvenile Chinook salmon



JSATS acoustic receiver deployments

Two types of acoustic receivers were deployed for this study: cabled and autonomous acoustic receivers (Table 1). These systems were deployed in the Snake River in eastern Washington State and in the lower Columbia River along the Oregon/Washington border. Two hydropower dams were outfitted with cabled acoustic receiver arrays: LGS and LMN. This allowed passage routes to be defined at each of these dams for each of the tagged fish. Transects of autonomous acoustic receivers (i.e., receiver arrays) were deployed immediately upstream and downstream of each of these dams for computing dam passage survival, and at several other locations to estimate cumulative mortality. From the release site to the final autonomous acoustic receiver array, fish detected at the final array will have passed through LGS, LMN, Ice Harbor, McNary, John Day, The Dalles, and Bonneville Dams. Each of these dams feature powerhouses with large, vertical Kaplan runners19 and spillways utilizing radial or vertical lift spillway gates. Most also include juvenile bypass systems (JBS; Table 1) and surface spill weirs20 to provide safer passage routes for fish.

Before deployment, all hydrophones and receivers were evaluated in an acoustic tank lined with anechoic materials at the PNNL Bio-Acoustics & Flow Laboratory (BFL21). The BFL is accredited by the American Association for Laboratory Accreditation to ISO/IEC 17025:2005, which is the international standard for calibration and testing laboratories. The accreditation scope (Certificate Number 3267.01) includes hydrophone sensitivity measurements and power-level measurements of sound sources for frequencies from 50 to 500 kHz for both military equipment and commercial components. The evaluation involved simulating transmissions from tags located at increasing distances. This allowed the performance of each receiver to be validated prior to deployment to ensure the expected detection range will likely be achieved.

Cabled receivers – deployment, hardware, and data processing

A JSATS cabled acoustic receiver system2 consists of up to four narrowband hydrophones; various types of hydrophone cables (e.g., four-channel “deck” cables, “y-blocks” that split the “deck” cables to individual connectors, and “wet” cables that run from the surface down into the water); a signal-conditioning, variable-gain amplifier; a data acquisition card that features a high-speed, analog-to-digital converter, a digital signal processor and field-programmable gate array; a GPS receiver for synchronizing time among multiple systems; a data acquisition computer; and software for detecting22 and decoding23,24 the acoustic waveforms. Deploying these systems within the forebay of a hydropower dam typically involves rigidly mounting slotted pipes to the upstream edge of the pier nose between the powerhouse and spillway bays. The cabled hydrophones are mounted on “trolleys” that have an L-shaped arm that protrudes and rides in the slot in the pipe and allows the hydrophone to be mounted pointing upstream. Conical baffles containing anechoic material are installed around the hydrophones to block noise coming from behind the hydrophone—either noise from the dam or reflections off the concrete. To obtain the locations of hydrophones that have been lowered below the water surface, survey equipment is used to sight the tip of the hydrophone as it is lowered down the pier nose. This provides the true direction and slope of the pier nose, which is used, along with the length of braided steel cable attached to the trolley to lower it into place, to calculate the 3-D location in space for each hydrophone. The individual “wet” cables for each hydrophone are routed to the forebay deck, where they are combined into “deck” cables carrying four signals using the “y-block” cables. The deck cables are routed to mobile trailers that house the acoustic receiver equipment. Acoustic beacons that send out JSATS tag-code signals every 15 to 60 s are deployed alongside several of the hydrophones across the array. These beacons are used primarily for quality control, to monitor (typically through the internet from an off-site location) the performance of each hydrophone to determine whether there is a reduction in performance so that any malfunctioning hydrophones can be repaired as soon as possible.

Two main programs run on the JSATS cabled receiver data acquisition computer. The first program is an energy-based detector software22 that collects the raw acoustic waveforms whenever the hydrophone signal meets a prescribed set of criteria. The second is decoder software24 that processes the waveforms saved by the detector to determine whether there is a valid detection (i.e., a decoded signal that has a valid CRC). If there is a valid detection, the decoded tag-code is saved to a text file, along with the detection time and other metadata. The detector software writes the binary waveform files to the hard drive using the *.bwm file type. The decoder is configured to wait for *.bwm files to be generated. Once a new *.bwm file is detected, the decoder will open the file, decode the data contained in the file, and then change the file extension to *.com to indicate that this file was decoded. If the detector saves data faster than the decoder can process it, for example at a hydraulic structure that is generating large amounts of acoustic noise (e.g., spillways with vertical lift gates), the decoder is configured to skip waveform files to avoid falling behind. Although these files may be skipped, the data contained within these files will still be used because the two different file extensions allow for readily identifying which files were not decoded in real time so that they can be decoded offline in a separate processing step after retrieving the data. The data is physically collected every 1–2 weeks, by swapping out the data collection hard drives. To make sure that all detection waveforms are processed by the decoder, the collected hard drives are put into data processing machines to decode any files that still have the *.bwm file extension. After confirming that all files have been decoded, either in real time or through post-processing after data retrieval, the decoded data from every hydrophone is checked for gaps in data. If a hydrophone is functional, there should be no large gaps in decoded data, since there are multiple stationary acoustic beacons deployed with the cabled hydrophone receiver array.

Data is filtered to remove potential false positive decodes. Data filtering for the JSATS cabled acoustic receivers (Fig. 2a) begins with a multipath filter, which removes decodes from multipath signal propagations (e.g., acoustic reflections off the surface/bottom). The multipath filter is used on the data from each individual hydrophone. Any decodes of the same tag-code that occur a very short time (e.g., typically <≈0.3 s) after an initial decode are removed, where the initial decode is assumed to be the original direct-path signal propagation. After the suspected multipath signals are removed, the remaining data from each hydrophone is combined into a single dataset and the decodes from each tag-code are grouped into “messages” suspected to be from the same tag transmission based on the detection time. A single detection filter removes any messages that were only detected by a single hydrophone, since the hydrophone spacing should enable any real signal from a transmitter within the detection range to be detected by multiple hydrophones. A message ping-rate interval (PRI) filter is then used on the data to remove messages that do not follow the expected transmission pattern inherent to all JSATS transmitters. The message PRI filter also groups messages into “events” of nearly continuous decodes of a tag-code. A predictable corruption of the tag-code bit pattern (e.g., a bit inversion at a specific bit position) can cause the incorrect decoding of a real tag-code signal and the generation of a different valid tag-code (i.e., a valid CRC). The final filter used on the dataset is a mimic filter, which removes these “mimic” decodes which result from the method to encode the data to be transmitted and is thus inherent to all JSATS acoustic receivers. To obtain the event history of each tag-code, the dataset is checked for events from known mimic tag-codes that occur during the same period. If an event is suspected to be from a mimic tag-code, the 3-D trajectories for the two tag-codes are compared to determine whether they appear to originate from the same spatial location. The result of this data filtering is a clean list of event histories for each tag-code, which is used in subsequent analysis of detection histories. A more detailed description of the data filtering process, including assessments of field detection probability and false-positive detection probability, are provided in Deng et al.11.

Computing the 3-D trajectories starts with the data from the multipath filter. The accurately time-stamped, decoded data is used to calculate the time difference of arrival (TDOA) between the detection times of a transmission on each of the different hydrophones. The TDOAs are used with the 3-D location of each hydrophone and water temperature data to compute the sound speed, to calculate the source location using an approximate maximum likelihood solver12. At each dam, the geographic coordinates of the hydrophones are converted to a local Cartesian coordinate system, in which the X-axis extends out into the forebay, the Y-axis runs parallel to the dam, and the Z-axis is normal to the water surface. Once the raw 3-D trajectories have been computed for each tag-code, a series of filters removes 3-D tracked points that are outliers. These filters include ones to remove points that are too far away spatially from adjacent points (>45 m), too far away temporally from adjacent points (>10 min), and that result in unrealistic velocities (~2 m/s for the size of fish we studied). Further quality assurance filters remove points that occur before a transmitter was released, after the PIT tag in the fish is detected downstream, and after the acoustic transmitter is detected by downstream autonomous receiver arrays.

Once the 3-D trajectories have been computed, they are used to assign passage routes through the dam for each tagged fish (Fig. 3a). The route assignment for each tagged fish is divided into three parts: main route, subroute, and hole (Table 2). The main route describes the part of the general dam structure through which the tagged fish passed. This includes the powerhouse, the spillway, and a generic category, “dam,” which is used for rare scenarios where there is confidence that the fish was physically present but a lack of confidence in specifically where the fish passed the dam. The subroute further divides the main passage route into different subcategories. For a main route of “spillway,” the two subroutes are the traditional (deep) spillbays and special surface weirs (surface spillbays20). The surface weirs reside within one of the spillbays and assignment to either of these two subroutes is made directly using the acoustic telemetry results. For a main route of “powerhouse,” the two subroutes are turbine and JBS. Assignment of the JBS subroute requires that the PIT tag of a fish assigned to the powerhouse was detected by the PIT tag readers within the JBS system; otherwise the fish is assigned the turbine subroute. PIT tag detections at dams where cabled hydrophone arrays have been deployed can be used to assign the JBS subroute, and PIT tag detections at the other dams along the migration route can serve as additional detection events. The hole assignment defines the specific powerhouse intake or spillbay where the passage occurred.

Table 2 Dam passage routes (main route, subroute, and hole) at LGS and LMN Dams.

Passage routes are assigned using the last 3D tracked location and the last detection. Two methods are used because the ability to consistently track the transmitter can diminish as the tagged fish approaches or passes through the plane containing the hydrophones, and the last decoded transmission could be later than the last 3-D tracked location.

When the last 3-D tracked location is used, a route is assigned according to whether the last tracked point is within a specific area. This area spans the entire dam plus 25 m on each side and extends from the dam face to 30 m upstream into the forebay. If the last tracked point is within this boundary, the 3-D track passage route is assigned to the bay corresponding to the Y coordinate in the local dam coordinate system. If the last tracked point is outside the piers on either side of the dam, the passage route is assigned to the nearest bay.

Route assignment based on the last detection uses the last transmission that was detected by multiple hydrophones. The detections associated with this transmission are sorted by time, and the pier numbers for the two hydrophones on different piers that first detected this transmission are averaged; the passage bay corresponding to this average pier is assigned as the last detection passage route. The default final route assignment is the 3-D tracked route assignment. However, when the two methods indicate different main routes, subroutes, or a different hole that is more than two bays away, the 3-D tracks are manually reviewed, and a decision is made regarding which method should be relied on for the final route assignment.

After the final route assignment, a final quality assurance step is to compare the final route assignment to the dam operations. In case a tag-code has been assigned to a closed passage route, the 3-D tracks are reviewed to consider the trajectory of the tagged fish and the location of the nearest open passage route. As previously mentioned, the ability to consistently track a transmitter is diminished as it approaches or passes through the plane containing the hydrophones. An example of when a tagged fish could initially be assigned to a closed passage route would be when a passage route with a strong attractive flow (e.g., surface spill weir) is adjacent to a closed passage bay.

Autonomous receivers – deployment, hardware, and data processing

A JSATS autonomous receiver (SR5000, Advanced Telemetry Systems [ATS], USA), along with the necessary deployment accessories, consists of a hydrophone that is connected to a cylindrical, positively buoyant, self-contained, battery-powered, autonomous acoustic receiver; a submerged buoy line; an acoustic release; a braided stainless-steel cable; and a steel anchor. These receivers are typically deployed according to the methods described by Titzler et al.10, which involves using a 34 kg or 57 kg (depending on flow) steel anchor to deploy the autonomous receiver system to the river bottom. The anchor is attached to the release side of an acoustic release using a braided stainless-steel cable. The fixed end of the acoustic release is attached to the autonomous receiver using a submerged buoy line. When the acoustic release is remotely triggered, it detaches from the anchor line, and the combined buoyancy of the acoustic receiver and the submerged buoy line bring the system up to the surface. To maintain the receiver orientation in the water column during deployment, a thin plastic sheet is folded around the cylindrical body of the receiver, creating an airfoil-like shape that keeps the receiver oriented in the flow direction. Attached to each autonomous receiver is a JSATS beacon that is similar to the JSATS beacons deployed with the cabled hydrophone receiver arrays and used in the same way. The length of time that the JSATS autonomous receivers can be deployed is largely dependent on the battery life, with data retrieval and battery changes typically done every 2–3 weeks.

Although it is possible to use autonomous receivers to conduct 3-D tracking, the process is much more challenging than using the cabled hydrophone receivers because the receivers are not fixed at well-defined locations (e.g., changes in river currents could change the receiver’s depth and horizontal location relative to the anchor) and are not time synchronized with each other. The autonomous receivers are typically used to detect presence of tagged fish, although recent research has investigated methods to improve the ability to conduct 3-D tracking25. Deploying an autonomous receiver array entails deploying several receivers in a line spanning the width of a river with the detection ranges of the receivers overlapping slightly. This creates a virtual detection gate that can be used to determine when a tagged fish passes through this location in the river. In addition to simply determining the migration timing of tagged fish, the autonomous receiver arrays are typically used for analyzing dam passage survival and near-dam behavior (e.g., forebay residence time, tailrace egress time).

The data collected by the autonomous receivers is filtered similarly to data from the cabled receivers (Fig. 2b). The primary difference is that the data from each individual autonomous receiver is processed entirely by itself. As a result, the single-hydrophone filter is not used, and a single-node PRI filter is used instead of the message PRI filter.

After the event histories for both the cabled and the autonomous acoustic receivers have been determined, individual routes were cross-checked by tracing the chronology of detections of every tagged fish as it was detected along the river in the sequence of acoustic receiver arrays. Upstream movement past a dam or out-of-sequence detections were deemed anomalous detection events. These anomalous detection events could be a few receptions resulting from noise or repeated detections of a transmitter that had been dropped near a receiver array after fish or bird predation. If the apparent behavior was impossible for a live fish, the anomalous detection was excluded from the detection history used for subsequent analysis.

JSATS transmitters

The injectable transmitters used in this study (Fig. 1b) were manufactured by PNNL. Each transmitter (model microV26, which is licensed to, and currently commercially available from, Advanced Telemetry Systems as Model SS400) was 15 mm long, had an outside diameter of 3.35 mm, a volume of 0.111 mL, and a mass of 0.216 g in air and 0.105 g in water. The transmitters are generally cylindrical; excess epoxy was eliminated to reduce the weight, and epoxy surrounding the transducer element was minimized. The transmitters had a nominal transmission rate of 1 pulse every 4.2 s. Nominal transmitter life was expected to be about 28 d at a 4.2 s pulse rate. The acoustic signal is transmitted using a carrier frequency of 416.7 kHz, a source level of approximately 156 dB (ref. to 1 µPa at 1 m), and a total signal duration of 477 µs. The transmitter emits a uniquely coded 31-bit signal2, resulting in more than 65,000 individual tag-codes, using binary phase-shift keyed (BPSK) signal encoding.

Each fish also bore a PIT tag (HPT12, Biomark, USA; 12.5 mm x Ø2.03 mm; 0.106 g in air). PIT tag detections were used to assign fish to passage through the JBS at LGS and LMN to distinguish between fish that were assigned to a main route of powerhouse and the turbine or JBS subroutes.

Tagged fish

For this study, 682 subyearling Chinook salmon (Oncorhynchus tshawytscha) were tagged with the injectable acoustic transmitter and released upstream of LGS Dam on the Snake River in Washington State, USA (Fig. 1a). The fish were obtained from the JBS at LMN Dam and selected using existing fish screening criteria utilized in previous juvenile salmon survival studies26. The fish selected for the study were held in holding tanks for 18 to 30 hours prior to tagging, and for 10 to 25 hours after tagging prior to release. The size criteria for tagged fish was also identical to other recent juvenile salmon survival studies26. For this study the fork-lengths ranged from 95 to 143 mm, and the weights ranged from 7.5 to 29.3 g (see Tagged Fish Data for information on each tagged fish).

Tagging procedure

While each anesthetized fish was at the data station for recording physical parameters, a second person inserted both a disinfected PIT tag and an injectable acoustic transmitter, assigned to a specific fish, into a sterilized 8-gauge stainless-steel hypodermic needle17. First, the injectable transmitter was placed into the needle, battery-end first. The PIT tag was then also inserted in the same needle. A sanitized plastic cap was then placed over each end of the needle to retain the tags. Once both tags had been placed in the needle, the tag loaded needle was handed to the surgeon working at the tagging station. Additional details for the tagging procedure are documented by Deng et al.11.

Release procedure

The fish implanted with the injectable acoustic transmitters were released using the same methods as fish tagged with commercially available acoustic transmitters for a separate large-scale survival study26. All fish were tagged at LMN and transported in insulated totes by truck to the single release site (Fig. 1a). There were five release locations across the river at the release site, and equal numbers of fish were released at each of the five locations. Releases occurred for 11 consecutive days (between 22 June and 2 July, 2013) and were staggered between day and night.

Data management

Use of JSATS can generate a large volume of data. An integrated suite of science-based tools known as the Hydropower Biological Evaluation Toolset (HBET; https://hydropassage.org/hbet)27 was developed to assist the characterization of hydraulic conditions at hydropower structures and to understand the potential impacts on aquatic life. HBET was initially developed to be utilized to facilitate use of the autonomous sensor technology known as Sensor Fish28. HBET allows researchers to use previously collected Sensor Fish data to design studies to evaluate hypotheses, archive field-collected data, process raw sensor data, compare different hydraulic structures or operating conditions, and to estimate the biological response for species with known dose-response relationships. More recently, HBET was adapted to also provide the functionality of archiving new or previously collected acoustic telemetry data and to produce visualizations from that data. Although it is not necessary to visualize the data set associated with this manuscript, PNNL offers free government and academic use of the HBET software package in the U.S. and a free 90-day trial version of the package to interested parties.



Leave A Reply