1. Pham Quang, Harold J. Geiger, Alaska Dept. of Fish & Game, USA

A statistical approach to estimating fish passage using a sonar for dynamic salmon management.

Following the Moose and Ehrenberg approach, the total voltage output of echoes from a single ping, returned by a fish is thought of as a filtered inhomogeneous Poisson process. This information can be used in echo integration to estimate the number of fish passing through the sonar beam. Requirements are a good knowledge of the geometry of the beam, the pulse and echo envelopes must be rectangular, and the sonar circuitry must be able to distinguish the different echoes. We derive alternate estimators and examine hybrid counting algorithms that use scaled echo integrators, echo counting, and echo tracking.

2. G.A. Rose, S. Gauthier, G. Lawson, Memorial University of Newfoundland, Canada

Acoustic biomass estimates are often made with incomplete or no assessment of error. Many estimates are made in which sampling error is the only source of error considered. It is difficult to judge the certainty of estimates without consideration of error. There are several sources of error in acoustic estimates: 1) system and calibration; 2) target strength; 3) detectability; and 4) sampling. These error sources act simultaneously and interact to produce the observed variation in acoustic measures. A monte-carlo simulation of error from known sources in acoustic estimates was conducted using data from shallow water acoustic measures of Atlantic cod (Gadus morhua) in Placentia Bay, Newfoundland, and from redfish (Sebastes spp.) in deeper waters. Results indicate that errors counter each other when considered simultaneously, and are less than when single sources are considered independently. The major sources of error varied between inshore and offshore surveys, but were associated with biological variations in fish behavior, especially detectability and TS. Methods to identify, reduce and quantify error are discussed.

3. G.L. Lawson, Rose, G.A., Fisheries Conservation , Canada

A major source of bias and imprecision in inshore shallow-water acoustic abundance estimates of semi-demersal fish is detectability, defined here as the detected proportion of the true number of fish within the ensonified volume, from surface to bottom. In autumn 1996 and 1997, acoustic surveys for Atlantic cod (Gadus morhua L.) in inshore Placentia Bay, Newfoundland, Canada, indicated mean daytime densities (e.g., 0.016 fish m-2 in 1997) an order of magnitude higher than at night (0.001fish m-2). A corresponding downward shift in vertical distribution at night was also observed. At the same site in 1996, in situ video censuses of cod from a submersible showed equivalent cod abundance by day (n = 3) and night (n = 1).  Submersible observations also indicated that at night cod were located nearer to bottom, and preferred rocky and boulder-strewn substrates and not open sandy bottoms (p < 0.001). Submersible acoustic estimates and video census indices were positively associated during daytime. We conclude that diel change in acoustic density resulted from variations in detectability caused by cod vertical movements and habitat preferences.

We advocate the inclusion of a detectability coefficient in the scaling of acoustic backscatter to abundance for semi-demersal fish.

4. Vikas S. Sonwalkar, Barbara. L. Adams, University of Alaska Fairbanks, USA

We present analytical and numerical models to gain insights into various uncertainties associated with estimating fish abundance present in rivers from acoustic data. Specifically we model the physics and signal analysis involved in estimating size and number of salmon moving upstream and downstream of a river using sonar equipment. Both direct problem, i.e., given a distribution of fish, determine the output of the sonar instrument, as well as the inverse problem, i.e., given the sonar data, estimate fish distribution present in the river, are analyzed using the model. The model developed can be used to analyze sonar data obtained from dual-beam, split beam, and wideband sonars used in fisheries acoustics. The numerical model is used to test the direct model under realistic conditions. The model is applied to analyze the data obtained during the ADFG-July 1998 split beam sonar experiment performed at Wood River near Dillingham, Alaska. We determine quantitative limits on the uncertainties involved in fish population estimation and suggest methods to reduce them.

5. T. J. Mulligan, R. Kieser and John Ehrenberg, Department of Fisheries and Oceans, Canada

Experiments conducted at the Qualark Creek acoustic site on the Fraser River have shown a systematic error in the split-beam angle measurement. A 10cm diameter plastic sphere at a range of 3.7m range was used for the target. The on-axis signal-to-noise ratio for this target was approximately 12dB. Although this signal to noise ratio is quite low for most environments in which measurements of individual fish echoes are attempted, it represents the typical target strength and signal-to-noise ratio obtained when observing migrating adult salmon at this Fraser River site. The target's location was measured both from the frame used to position the target in the beam and from the acoustic data. Comparison of these two sets of paired measurements demonstrated an angle bias that increased with the target's distance off the beam axis (note that the signal to noise ratio decreased as the target is moved off axis). The acoustically measured target locations were closer to the beam axis than the actual locations. Each acoustic estimate was obtained by averaging individual angle measurements for the target for 3000 acoustic pings of data.  A Monte Carlo simulation was developed that modeled the effects of random noise on the split-beam angle measurements. The simulation results also show a systematic underestimate of the acoustic angle estimates. This bias increases rapidly with increasing off-axis target location and decreasing signal-to-noise ratio. The simulation accurately predicted the magnitude and direction of this bias for circular cross section beams. The simulation results for elliptical cross section beams did not provide as good a match. The simulation provided insight in to the reason why noise causes angle bias. This mechanism will be discussed in the paper.

The presence of the angle bias can affect the estimates of fish density and migration speed obtained with a split beam system. It should be noted that even with the bias, the split beam system will provide more information than can be obtained with either a dual beam or single beam system. The effect of noise on dual beam systems is even greater than it is on split beam system and neither dual beam nor single beam systems provide measurements of the three-dimensional position of the fish in the beam. The fish location estimation obtained from a split-beam system can be used to account for variations in fish density as a function of distance from the bottom. The angular estimates can also be used to correct for variable fish detection probability as a function of location in the beam.

6. Stratis Georgakarakos, Gerlotto, Francois, Orstom, France

The use of single beam echo sounder is limited in shallow waters (less than 5 meters) by several factors: small sampling volume and fish avoidance in vertical echo sounding, and multiple reverberation and TS variability in horizontal acoustics. The use of multibeam sonar may overcome these problems : the sonar has sixty 1.5 beams which receive echoes from a 90 field. The instrument is deployed on a vertical plan, the central axis being horizontal, and perpendicular to the line of travel of the research ship. For each ping an image of the water mass on the side of the boat is recorded on a video tape. The paper describes a image analysis method allowing to extract the pixels representing the fish inside each ping. The ratio between the number of pixels representing fishes and the total sampled surface is used as occurrence index for measuring fish abundance. Some methods to extract quantitative indexes from these data are presented. Some applications are presented and the meaning of the results is discussed.

7. François Gerlotto, Peter K. Eriksen, Orstom, Reson A/S, France

The paper describes the main sources of drawbacks in the application of conventional acoustics in shallow water areas, and reviews the advantages and limitation that the existing multibeam sonar presents in this kind of ecosystem. Then a new technology for adapting multibeam sonars to shallow waters is proposed and discussed. The « ideal » acoustic device is described taking as an example a new RESON system: the sonar operates at 455kHz with a coverage of 120 degree in one direction and, depending on the needs of the user, 19 degree or 1 degree (can be modified easily) in the perpendicular one. The beam opening angle is 0.5 degree in the center beam growing up to 1.0 degree at the 60 degree steer angle, giving a total number of 240 beams. The data collected by this tool can be used for several purposes, mainly fish density estimates, individual fish biomass, and spatial and temporal behaviour of fish. The theoretical/practical ranges and capabilities of this system are analysed, and a description of the way to deploy it is described.

8. Robert L. Johnson, Russell A. Moursund, Battelle Pacific Northwest National Laboratory, USA

Bonneville Dam, on the Columbia River, has been the site of a concerted effort by the U.S. Army Corps of Engineers to develop and evaluate a prototype surface collector (PSC) device to divert juvenile salmonids from turbines at Powerhouse I. Turbine Unit 3 was evaluated based on two slot widths, 1.5 m and 6.0 m. An important component of the evaluation was monitoring juvenile salmon behavior in the forebay as they approached the structure. The overall objective of the behavioral evaluation was to determine the travel direction and velocity of juvenile salmon in the region 15 m upstream of the structure, and relate that information to the local environmental conditions, principally flow. This was accomplished using a newly developed, non-intrusive, Dual-Head Multibeam (DHM) sonar technique. The multibeam sonar provided a coherent view of a large volume of water and protracted 3-D fish track data. Data processing, analysis and visualization tools were developed based on data collected at Lower Granite Dam during the juvenile salmon migration in 1997. Range related metrics provided a quantitative basis for comparing the two slot configurations under evaluation. An interactive 3-D visualization tool was developed to provide biologists, managers, and engineers with a view of fish behavior unobstructed by turbidity.

9. Juha Jurvelius, Timo J.Marjomäki, Juha Lilja & Raimo Riikonen, Finnish Game and Fisheries Research Institute, Finland

Winter fishing under ice is in many areas of Finland and northern Europa the most profitable fishing for local people. Winter seining and fish behaviour under ice has been studied eg. by videocameras. However, this method is restricted by lightness and water transparency. Limitations for under-ice hydroacoustics are principally the same as during ice-free season. Thus the spatial distribution and the stock density of wintering fish can be studied by a mobile under-ice echosurvey.  Diurnal fish density and distribution was studied from January to April in winter seining areas and unfished areas in shallow Finnish lakes. The catchability of the seine and the effects of seining to the fish density was described, and fish migrations under-ice were compared with those of ice-free season.

10. B.H. Ransom, S.V. Johnston and T.W. Steig, HTI, USA

Many anadromous salmonid populations (Oncorhynchus and Salmo spp.) are declining as pressure from over harvesting, habitat degradation, and other sources increases. To aid the management of these stocks, hydroacoustic techniques have been used since the 1960's to estimate adult salmonid escapement in nearly 50 rivers in North America and Europe. Initial evaluations used single-beam hydroacoustic techniques, with dual-beam techniques introduced in the mid-1980's. Since 1992, digital split-beam hydroacoustic techniques have been used in over 50 studies in 17 rivers. Due in large part to it's improved spatial resolution and three-dimensional tracking capabilities, the split-beam technique has proven more useful than single-beam or dual-beam techniques. Uneven bottom bathymetry and nonlaminar hydraulics are typical. The major issues to properly address include hydroacoustic equipment and techniques, site selection, transducer deployment, and fish behavior. Typically, narrow-beam transducers are mounted near shore and aimed horizontally into the river, perpendicular to flow, monitoring migrating fish in side-aspect. A bottom substrate of low acoustic reflectivity (e.g., sand, small rocks) enables the acoustic beam to be aimed close to the bottom. Sample sites are sought where fish are actively migrating, not holding or milling. In addition to escapement counts, results include estimated fish sizes, spatial distributions, diel distributions, and velocities.

11. T. Brock Stables, George Kautsky, BioSonics Inc, USA

Estimating abundance of out-migrant juvenile salmonids, primarily chinook salmon (Oncorhynchus tschawytsha), is a cornerstone of salmon management in Northern California’s Trinity River. In a three year feasibility study (1996-1998) to evaluate use of sonar for this purpose, discrimination of fish from drifting debris (mainly leaves with TS similar to out-migrants) was deemed essential. On the premise that fish may respond to an external stimulus by moving in a way that would distinguish them on echograms from passively drifting debris, in-situ tests of electricity and light as stimulants were conducted in the Trinity River. Fish responded noticeably to both stimulae, but an apparent attraction to visible light caused the strongest and most potentially useful reaction. Underwater video showed that nearly all out-migrants dove toward a submersed light, while nearly all leaves followed a linear trajectory past the camera. Although side-looking split-beam sonar did not detect this response in concurrent tests, refinement of sonar methods and better transducer siting would likely improve future results. Video data about type and magnitude of fish response provide an initial estimate of sonar sensitivity necessary for this application.

12. John Pham, Vikas Sonwalkar, University of Alaska Fairbanks, USA

In fisheries acoustics, the choice of filter type and properties are important for characterizing return echoes.  Typically the sonar data is first filtered through a bandpass filter to improve the signal-to-noise ratio and then analyzed for various characteristics such as pulse arrival time, width, amplitude and phase. Often the echoes arriving from two different fish overlap. The problem now is to extract individual echo parameters from this superposition. We review various techniques used to measure the characteristics of the pulse. Then we present an analysis that addresses the following questions: (1) what filters are optimal to measure arrival time, pulse width, phase, and amplitude? (2) What filter or combination of filters is optimal to isolate an individual echo from a superposition of echoes? We present results from MATLAB simulations that use typical pulse widths (0.1 to 1.0 ms) and frequencies (100 to 400 kHz) found in sonar applications to fisheries. Various types of digital filters, including matched filters, as well as various designs (IIR, FIR, impulse-invariant, bilinear, etc.) are considered.

The Advantages of Broadband Sonar for Riverine Fish Stock Assessment

The broadband sonar techniques possess unique processing advantages when compared with the classically utilized, single frequency split beam riverine sonar systems. The techniques, mature signal processing methods developed in the radar and seismic communities for over thirty years, have not-been used to the same degree in riverine sonar applications until relatively recently. The authors will give a short presentation to introduce the basic advantages of the broadband sonar techniques when compared to single frequency split beam processes.  The talk will provide a brief overview in the following two topic areas:

Broadband Sonar Definitions - The basic meaning of broadband sonar as it applies to the familiar subject of acoustic echo ranging. The generation of the sonar pulse using digitally encoded techniques, the choice of operational and engineering parameters for practical systems, and the constrains applied by existing sonar transducer technology will be discussed.

Broadband Sonar Capabilities - The detection and classification application of the broadband sonar in the riverine environment. The authors will present a brief discussion of fish target detection and target size estimation, tracking a swimming target using broadband Doppler techniques and estimating fish target spectral response. Examples of successful application of a broadband sonar will illustrate the techniques.

Split-beam hydroacoustic techniques have been used to enumerate adult Atlantic salmon (Selma solar) passage on the River Wye since 1994. Aquatic weed aggregations are seasonally present at the monitoring site. Echo amplitude returns from high densities of weeds can be similar to those of adult salmon. Direction-of -movement information readily resolves upstream-migrant salmonids from static or downstream-travelling weeds. However, Atlantic salmon are iteroparous, and downstream-migrant kelts must be factored in the acoustic counts to avoid bias. In 1995, a data subset was selected to determine if non-directional acoustic parameters could be used to separate fish from macrophytes (n=7l). Available parameters included echo position, pulse width (at -6, -12, and -18 dB power points), amplitude, beam pattern factor, and target strength. Videotape records were used for ground truth. Individual targets were visually identified as fish cr weeds. Pulse width standard deviation (PWSD) measurements were determined to be the most effective individual parameters for discriminating weeds from fish. Based on -6 dB PWSD selection, 77% of all weed targets were removed. Multiple selection criteria increased total target discrimination. Applying a combination of -6 and -12 dB PWSD, and Y-axis target slope criteria rejected 94% of all weeds, without excluding valid fish targets.

An overview of Acoustic Fish Counter projects in the UK The Environment Agency for England and Wales is leading an R&D project in collaboration with the agencies of two Government departments, the Ministry of Agriculture, Fisheries and Food (MAFF) and the Scottish Office, and the River.Spey Research Trust. The primary objective of the project is to maximize the operational value of hydroacoustic for the enumerating adult salmonid migration in the UKand to encourage the controlled uptake of the technique. It sets out to fully evaluate and test this application of hydroacoustic technology, from initial deployment through to validation methods, in a variety of differing river types.

Both MAFF and the Scottish Office have a common interest with the Environment Agency in developing effective methods of monitoring salmonid stocks. Collaboration between the Agency, MAFF and SOAEFD together with the Spey Research Trust makes political and economic sense in workng towards realizing the potential of hydroacoustics in the UK.

There are currently five sites in the UK at which acoustic counters are deployed. The Tavy in Devon, the Wye on the English/Welsh border and the rivers Teifi and Towy in West Wales have counters operated by the Environment Agency. In Scotland, the River Spey has a site operated by the Spay Research Trust. These represent a cross section of applications and river types and form the basic network of sites upon which a notional collaborative project, dedicated to the development of hydroacoustic counter within the UK, is based.

Validation is carried out at each site by an array of underwater video cameras and image analysts. A major component of the R&D Project is to develop a methodology for validating acoustic counters which can be applied to future acoustic applications and form a thoroughly cost part of any fish counter proposal.  This presentation gives an overview of the UK's acoustic fish counter program. It briefly describes the management responsibilities of the Environment Agency regarding salmon and sea trout and outlines the monitoring equipment and deployment methods employed. Results are presented from a range of sites. The validation methods are discussed together with an overview of validation results and the problems of species apportionment.

Hydro-acoustic Fish Counter Development on the River Spey, Scotland Migratory salmonid stocks in the River Spey are managed by the Spay District Fishery Board and in order to improve their understanding of returning adults a reliable fish counter is considered essential. In Great Britain resistivity fish counters have been developed to enumerate adult numbers however, these rely on the construction if a weir to attach the counting electrodes. The volatile nature of the river and the considerable expense of such a construction secluded development of this type of counter in the lower Spey. Thus and alternative was required. After successful trials in 1994 and 1995 the Spey Board purchased a hydro-acoustic counter (HTI Model 240, Later upgraded to Model 243) and have embarked on a six yiar development program. This paper reviews the development of the system and considers some of the practical and data analysis problems encountered to date.   Four potential counting sites were examined in 1996 with the most promising at Collie Pot, some 16km from the river mouth, selected for longer term development. This site offers a good triangular profile, laminar flow during a range of levels, fish tend to pass through the area for most of the year and is secure from public disturbance.

Initial results are promising although problems relating to the behaviour of Atlantic salmon and sea trout are evident. Both species remain close to the river bed during their migrations thus positioning the beam as close as possible to the river bed is critical. This is difficult given the cobble and boulder substrate present at the site. Fish ascend the Spey during all months of the year so a permanent installation is required resulting in considerable quantities of data processing. Sea trout and smaller salmon often migrate during similar periods of the year and currently it is difficult to separate these species acoustically. Downstream targets are also problematic particularly during spates and when there is high quantity of woody debris in the water column during Autumn.

Ongoing research to address the critical problem of fish passage close-to the riverbed and under the acoustic beam was initiated in 1997. An underwater video camera system was developed and' installed directly under the acoustic beam. In this way an assessment of fish passage in and around the beam is underway. Early results indicate that counter efficiency varies through season and across the river width.  Video data has also provided and insight into the- criteria used to determine fish using TRACKER/TRAKMAN and problem areas and improvements will be explored.  The Spey counter is still in the early stages of development, progress to date has been very good with video data beginning to provide counting efficiency estimates which should allow more accurate estimates of fish passage. Some plans for the future development of the counter will be described including smoothing the river bed with sand bags and installation of a portable resistivity counter.

5. Atso Romakkaniemi, Juha Lilja, Mari Nykanen, Timo J. Marjomaki & Jurvelius

The River Tornionjoki supports the largest wild Atlantic salmon stock in the Baltic Sea. The river is 500 km in length and the mean discharge is 381 m3/s. Finnish Game and Fisheries Institute initiated hydroacoustic studies of salmon escapement in 1995.  A site located a few kilometers upstream from the river mouth has been chosen for extended sampling. The site is 250-280 m wide, with maximum depth of 8-10 m. Since 1997, two hydroacoustic systems have been sampling horizontally into the channel from the east and west banks of the river. Each system consists of an HTI Model 243 split-beam echo sounder sampling two elliptical-beamwidth transducers (2.80x140 and 40x 100). Each transducer was sequentially sampled for 10-15 min each hour. In 1998, a fifth transducer was added and moved to various locations to investigate passage in unmonitored areas. Net weirs directed migrants offshore into the acoustically-monitored area. Sampling has begun in early June, following river ice breakup, and has terminated in late August. Peak MSW (multi-sea-winter) salmon migration occurs in June, but grilse (1-sea-winter salmon) run occurs in July and August.  In 1997-1998, hydroacoustic monitoring detected a few thousand upstream-migrant targets within the acoustic size range expected for salmon. Extrapolating total passage through the sampled volume resulted in annual escapement estimates of 10,000-15,000 fish. This estimate is slightly greater than the recorded 1997-1998 river catches. Hydroacoustic estimates are therefore considered to underestimate the true escapement. Some of the probable reasons for this apparent underestimation include:

1) The slope of the river bottom is irregular, partly preventing good fit of transducer beams along the bottom Salmon may pass in the unsampled areas below the acoustic volume.

2) Targets of acoustic size consistent with salmon were observed also in upper water columns. To date, these targets have not been included in the expanded escapement estimates.

3) 411 salmon were not moving upstream perpendicular to the transducer beam axes. Therefore, salmon passing at greater aspect angles may be excluded because of their small detected target size.

4) Significant escapement may occur in the unmonitored mid-channel area. Many salmon-sized targets were observed near the extreme sampling range of the offshore transducers

In 1998, daily mean TS distributions followed similar trend over the season for each transducer. This trend was in agreement with expected changes in migratory fish length. However, mean fish TS varied by sampling location. TS measurements appeared to be relatively sensitive to transducer horizontal aiming angle, and presumably fish aspect. This phenomenon caused difficulties in selecting comparable fish length groups at all monitoring locations and combining data sets. Several migratory and resident non-target species (e.g. Selma trutta m. trutta, Coregomus lavaretus,.Esox lucius) are present at the site. Peak migrations of non-target species are distinct from those of salmon, but overlap occurs. Commercial and recreational catch data were used as indices of the size variation of salmon and run intensity of different species. These indices were in agreement with observations from the echo sounders. Bimodality in fish TS distributions was observed during the MSW salmon migration. Laboratory measurements showed that a specimen of certain length of Coregonus lavaretus or Esox lucius gave about 5 dB larger TS than salmon of the same length.

The results of the River Tomiojoki hydroacoustic project to date reveal difficulties of sampling in large river with multi-species environment. The results indicate that implementation of this technique in Baltic salmon rivers is net trivial, and requires additional site-specific research. However, other Baltic saloon rivers are smaller and many of the problems encountered on the liver Tomionjoki will probably be less evident on other rivers.

6. Yong-Jun Yang, John Hedgepeth, Jie Cao, and David Fuhriman,BioSonics, Inc. 4027 Leary Way N.W. Seattle WA, USA

Improving fish target strength and location estimates from split-beam measurements

The split-beam echosounder is considered to be theoretically superior to a dual-beam or single-beam echosounder for fish target strength estimates (MacLennan and Simmonds 1992, Ehrenberg 1983). The objective of this research was to use simultaneous split-beam and dual-beam measurements to determine- if additional information provided by the dual-beam improves fish target strength and location estimates. The method used dualbeam angle estimates to determine if split-beam measurements were phase-wrapped. While this method was successful, additional improvement could result from a vernier approach for determining angles from the transducer to fish targets.  Collections of simultaneously acquired split-beam, dual-beam and single-beam echoes were used to determine target strength and fish density. After methods were tested in simulation, modifications were made to a BioSonics model DT6000 split-beam echosounder to provide simultaneous single, dual and split-beam measures on fish. Comparative results are presented for the three methods as well as for the improved splitbeam method.

7. Debby Burwen, ADF&G, Anchorage Alaska

Evaluation of Multifrequency Parameters as Potential Hydroacoustic Discriminatory of Fish Species in Rivers

Side-looking sonar has been used to assess chinook salmon Oncorhynchus tschwatscha returns to the Kenai River since 1987. Hydroacoustic assessment of chinook salmon in the Kenai River is complicated by the presence of more abundant sockeye salmon O. nerka which migrate concurrently with chinook salmon. Most counts from sockeye salmon are censored using a combination of range and acoustic size thresholds (sockeye Salmon generally migrate closer to the bank and are smaller than chinook salmon). However, some inflation of chinook estimates by sockeye salmon is known to occur. Efforts to improve our ability to discriminate between chinook and sockeye salmon continue through research using fathered and flee-swimming fish of known size and species. These studies explore the potential for discriminating size groups of fish using a combination of acoustic parameters. In 1998, we investigated the use of multifrequency data to assist in discriminating between these two species. Target strength and other acoustic parameters were measured on fathered chinook and sockeye salmon at 120kHz, 200kHz, and 420kHZ. Preliminary analyses suggest a differential response in target strength between the frequencies. Results on continued research into the use of the returned echo pulse width as a potential species discriminator will also be presented.

8. Deborah A. Hart and Harold J. Geiger, Alaska Department of Fish and Game, Commercial Fisheries Management and Development, Juneau, Alaska

Alaska Department of Fish and Game's Bendix Replacement Project

The salmon management strategy of the Alaska Department of Fish and Game is to achieve specified ranges of annual escapements for each stock or run. Measures of escapement and in-season abundance estimates are key components to this strategy. Since many streams in Alaska are laden with Glacial silt and too turbid for visual counting techniques an acoustical means for acquiring abundance estimates was developed for these systems. During the 1960's, the Department along with Bendix Electrodynamics Corporation developed sonar techniques and equipment for counting upmigrating adult salmon. These techniques and equipment were improved upon and even now, decades later, these matured Bendix systems are still in use on many of Alaska's highly valued commercial salmon streams. Within recent years, improvements in digital and computer technology has enhanced the field of riverine hydroacoustics. New techniques and equipment for detecting and analyzing fish echoes has become available. Currently, the State has integrated both dual and split beam sonar in river systems such as the Yukon and Kenai. Since growth and expansion will continue to occur, the Department is developing protocols for integrating newer hydroacoustic developments into the State's riverine sonar program. In March of 1997, the Department hosted the Riverine Sonar Workshop and invited experts in the fields of hydroacoustics and fish behavior to join in discussions about the development of a new riverine sonar system and procedures for the State of Alaska. As a result of this meeting and subsequent discussions, the Department began the Bendix Replacement Project and issued a Request for Proposal for development of a new riverine sonar system. The Department awarded the system development contract to BioSonics, Inc. and established various review contracts with the University of Washington's Applied Physics Laboratory and the University of Alaska Fairbanks. The first field season of the project was completed during the summer of 1998. On going reviews of the new system and the development project are currently in progress.

9. Peter Dahl, Harold J. Geiger, Jim Dawson, Donald Degan and Deborah A. Hart

Properties of background reverberation during the Wood River salmon-counting sonar experiment

Sonars used to count migrating salmon in Alaskan rivers must operate in a complicated, and often highly time-varying, acoustic environments. Very little is known, however, about the background reverberation structure expected in such environments, which affects the performance of salmon-counting sonars. A notable exception is the work by , Trevorrow on the Fraser river. Here we present measurements of background reverberation taken during the 1998 Wood River Experiment, sponsored by the Alaska Department of Fish & Game, and designed to compare sonar counts with both visual, video-derived counts in the Wood river (an Alaskan river with exceptional clarity due to the absence of glacial silt which render the waters opaque). Reverberation measurements in the form of ensemble-averaged backscattered intensity are presented and discussed in terms of geometry, and sonar parameters such as frequency, beam width and pulse length. The reverberation measurements are then interpreted using scattering models for acoustic backscatter originating from the boundaries and volume. Ultimately we seek a reasonably accurate model for background reverberation which will be important for more advanced fish counting algorithms.

10. Harold J. Geiger and Debbie Hart, Alaska Department of Fish and Game, Division of Commercial Fisheries, PO Box 25526 Juneau AK 99801

The Alaska Wood River Sonar Project

In 1998 the Alaska Department of Fish and Game established a sonar research site on the Wood River, in Bristol Bay Alaska. The research site is located just below the outlet of Lake Aleknagik, a short distance from the village of Aleknagik. This clear-water river is home to a large run of sockeye salmon, and a long-established counting-tower project that fishery managers have used to assess the size of sockeye runs into Lake Aleknagik. The tower-project estimates are assumed to be very accurate at low and medium density, and provide conformation of estimates derived from hydroacoustics. Additionally, a video system was used to try and capture images of the sound field at each ping, to provide an interpretation of the raw sonar data Water levels were the highest in our experience, and the water clarity was poor. throughout the season, 196 hours of sonar data was collected. which amounted to 1.4 gigabytes of raw digitized data and 0.98 gigabytes of tracked data. Prior to the season, we hoped to demonstrate echo tracking at passage rates from a few hundred fish per hour to passage rates in excess of three of four thousand fish per hour. Remarkably, passage rates exceeded 40,000 fish per hour. Using a pulse width of 0.4 ms, and 15 pangs per second, sonar fish-tracking estimates and tower-count estimates showed a very strong linear relationship up to about 3,000 fish per hour. At passage rates much in excess of 3,000 fish per hour, the sonar estimates failed to increase at the same rate as tower estimates. The video images collected during the 1998 field season were disappointing. Water cleric was not adequate to collect underwater images. Images collected from above were affected by wavy surface structure caused by wind, boat wake, and specular lighting conditions.

At present, our plans for the 1999 season call for improvements to the video system, designed experiments to look at the effects of changing pulse width, ping rate, and transducer configuration as density varies.

11. James J. Dawson, Daniel Wiggins, Donald Degan, Harold J. Geiger, Deborah A. Hart, and Barbara L. Adams

The introduction of split-beam scientific echo sounders to the riverine environment allows echo-tracing algorithms to evolve from two-dimensional time/range methods to three-dimensional true spatial correlation algorithms. These spatial algorithms were used to count sockeye salmon in the 1998 Alaska Statewide Sonar project at the Wood and Kenai Rivers. During analysis of fish tracks passing through the sound field, we observed that split-beam positional estimates in the horizontal plane exhibited greater variability that those in the vertical plane. Given the relatively close range of the fish, we propose that this variability is due to degradation of the split-beam phase measurement by violation of the point source assumption. This paper presents an evaluation of the effects of point source violation based on observations of positional viability as a function of range, and on empirical results from measurements of standard spheres and models of fish bodies.

12. David Daum, U.S. Fish and Wildlife Service, Fairbanks, Alaska

The Intricate Art of Riverine Fish Tracking

In recent years, the art of counting upstream migrating fish in rivers has advanced tremendously. Four-dimensional data (time and 3-D space) from split-beam technology gave the riverine sonar worker additional target behavior. information required to make fish counting more accurate and scientifically defendable than earlier attempts. Prior. fish enumeration techniques relied on interpretation of target traces from echograms using only two-dimensional information (time and range). Angle of target trajectory (change in range), gaps between contiguous echoes, and target duration in the acoustic beam were criteria used to differentiate individual targets into upstream fish, downstream fish, downstream debris, and stationary river bottom reflections. Many of the behavioral assumptions used for echogram interpretation were untested in the field, based on preconceived notions of how specific target types should acoustically act.

Knowledge gained from individually examining over 675,000 acoustic targets using fixed-location, split-beam sonar data has lead to a better understanding of specific acoustic target identification. These data were collected during fall chum salmon migrations in the Chandalar River, Alaska ( 1994-1998) and from a chinook salmon run in the Yukon River (1993). Most upstream fish were easily separated from downstream fish, debris, and stationary targets using upstream/downstream directional information. But under certain conditions, fish tracking became a difficult task. Acoustic noise and high fish density can confound individual fish identification. Close-range targets may exhibit high variability in upstream/downstream positional data due to the "barndoor effect", makng directional determinations challenging. Forward scattering of the acoustic signal by fish targets may generate upstream-traveling "shadows" under certain river bottom conditions which are hard to distinguish from real fish targets. Acoustic multi-paths from bottom reflections off sand/silt substrate may produce non-distinct fish targets at close range and "ghost" images it far range. All of these conditions may affect the ability to accurately track fish, both manually and automatically. The complications in tracing fish targets presented here should be considered when developing future automatic trackng systems. Otherwise, the utility of such a system will be limited.

13. John E. Ehrenberg, Hydroacoustic Technology Incorporated,715 NE Northlake WaySeattle, WA 98105

A Comparison of Single Echo Isolation Techniques

The first processing step in many in situ acoustic assessment techniques, such as target strength measurement, echo counting and target tracing, isolates echoes from individual fish. If overlapping returns from multiple fish are improperly selected as single fish echoes, the performance of the acoustic technique is adversely affected. In particular, accepting overlapping echoes as single echoes can result in a significant increases in the bias-and variance in the estimates provided by echo counting and target strength measurements and can introduce so much variability in a target's angular location estimates that tacking the target is nearly impossible. The traditional method used to isolate single echoes is based on the measured width of the pulse at one or more positions in the pulse. Other techniques that have been proposed use variability of the phase within the pulse and the normalized correlation coefficient that is obtained by correlating the received signal with an undisturbed version of the transmitted pulse. This talk investigates the relative performance of these various techniques. In particular results that show how the performance of these techniques is dependent on the signal to noise ratio, receiver bandwidth, echo separation etc. are presented. Earlier claims that state that the phase measurement and the normalized correlation coefficient technique are superior to the methods based on pulse width measurement are shown to be false. Some other approaches for minimizing the likelihood of accepting multiple echoes as single echoes are discussed.

Side-looking deployments of short pulse hydroacoustic systems in rivers produce data consisting of groups of echoes from fish as they pass through an acoustic beam. These groups of echoes or tracks can be counted to provide estimates of migrating fish populations. or accurate estimates to be obtained, the echoes must be grouped in such a way that: a) each exulting track is composed of echoes from one and only one fish, b) each group of echoes from a fish target is identified and tracked, and c) non-fish groups of echoes are not tracked as fish targets. With the introduction of split-beam techniques to riverine monitoring in 1992, the three-dimensional spatial location of each echo in a group of echoes became available for use in target racking systems. Since that time rapid advances in software operating systems and computer capabilities have allowed data collection systems to track targets in three-dimensions as they are collected, as well as provide user friendly environments for post-processing manual target tracking, aromatic target tracking, and track editing.

Using Split-Beam Hydroacoustic Techniques for Monitoring Adult Migrating Chum Salmon in the Chandalar River, Alaska

Split-beam hydroacoustic techniques have been used to monitor migrating adult chum salmon (Oncorhynchus keta) in the Chandalar River throughout their primary migration period each summer since 1995. Narrow beam elliptical transducers were deployed on opposite banks of the river aimed horizontally out into the river, at right angles to the direction of river fiow. Spatial coverage of the acoustic beam was measured using the spit-beam position information from reference targets. Individual fish targets were visually identified from digital split-beam echo data using software that displayed individual echoes and allowed selecting fish tracks. The number of individual fish tracks recorded ranged from- a low of 59,259 individuals in 1998 to a high of 256,090 individuals in 1995. The digital data contained measurements of target strength and three dimensional position of each echo from each fish. From this data the average target strength, direction of travel, and position in the beam of each fish target were calculated. All years showed most fish traveled upstream, migrated near the bottom, and were primarily shore oriented. Escapement estimates generated from these data have high precision due to high temporal and spatial coverage of the migrating population.

A video system was deployed together with an HTI split-beam sonar system near the town of Spencer Bridge to observe migrating adult Pacific salmon in the Fraser River of British Columbia, Canada. The video system consisted of two video cameras: one placed underwater and the other above the water surface, with a field-of-view covering part of the acoustic beam area, allowing determination of 3D positions of fish when they pass through the acoustic beam. This talk will describe the video photogrammetry and data analysts, and compare acoustic tracks with the corresponding video tracks. We will use the video data as a reference to evaluate the overalls performance of the acoustic system in detecting and tracking positions of single fish. Limitations of the acoustic system in detecting and tracking multiple fish will also be discussed.

(This work was performed in collaboration with members of the Riverine Acoustics Program, Pacific Biological Station, Department of Fisheries and Oceans of Canada.)

17. Tim Mulligan and Peter Withler, Department of Fisheries and Oceans, Pacific Biological Station, Nanaimo, BC Canada

Target Tracking and Track Editing

How do you know that your estimates of the number of migrating salmon are correct when you have no direct comparison with estimates using another method? How do you know that your tracking algorithm is providing reliable, unbiased estimates? We will use an example from our work at Qualark Creek on the Fraser River and from a comparison with visual counting on the Thompson River to demonstrate some of the problems we encountered and how we are attempting to overcome them. Data from the visual-acoustic comparison and from extensive use of a track editor have been combined to generate an empirical approach.

18. H. Enzenhofer, R. Kisser, J. Ehrenberg, Division of Fisheries and Ocean Canada, Pacific Biological Station, Nanaimo, B.C

Split-beam Angle Measurement Bias at Low Signal-to-Noise

The measurements that stimulated this work were carried out at the DFO fisheries acoustic facility at Qualark Creek on the Fraser River. A split beam system is used at this site to provide daily salmon escapement estimates. In order to estimate detection probability as a function of target location in the beam, a series of measurements were made using a target frame that allowed us to move the target-to any desired position in the beam cross section. Given a known target location we found that the detected location was closer to the beam axis than its true position. This split-beam angle measurement bias was substantial at the relatively low signal to noise that is typical for riverine acoustic measurements. This paper will present a stochastic model that provides a good explanation for these observations.

19. Peter N. Johnson, AScI, Inc., Fisheries Engineering Team, Columbia River Basin, P.O. Box 40, North Bonneville, WA 98648

Hydroacoustic Evaluation of Strobe Light Effectiveness for Reducing Entrainment of Juvenile Salmon at Hiram M. Chittenden Locks, Seattle

In FY 98, we evaluated the use of strobe lights for vertically redistributing salmon smelts in front of a culvert used to fill a navigation lock chamber in Seattle, WA. Fish densities near the culvert entrance were monitored hydroacoustically using a pair of 420 kHz 6degree single-beam transducers mounted on a pole and aimed down across the center of the intake. Density estimates of fish in the acoustic beams were based on echo integration intervals of 15 seconds in each of 11 1-m depth strata. We used a sigma value equivalent to 14 cm total length (average fish size based on purse seine catches) to calculate the echo scaling constant.

Flows into the culvert during low tides were up to 2200 cfs, with maximum instantaneous water velocities of 5.S fps. We installed ten strobe light flash heads around the perimeter of and in front of the north filling culvert. We ran strobe lights at 600 watts with a pulse rate of 300 flashes per minute. Strobe lights were evaluated using paired on and off treatments during day (n=32) and night (n=31). Nighttime strobe light evaluations were inconclusive due to low numbers of fish near the culvert entrance during both test and control treatments. We evaluated effects before and during fill events.

10-minutes before fill: In control treatments, juvenile salmon were mostly distributed within the bottom 5 m of the acoustic beams, with the greatest mean density observed in the deepest strata (12-13 m). With strobe lights fumed on, the fish distribution shifted upward, with the greatest mean density observed in the 5-6 m depth strata. Mean fish densities at the depth of the culvert (8-13 m) decreased by over 96% during strobe light-on treatments. Based on paired t-tests, mean fish densities in 1-meter strata below a depth of 8 m were all significantly lower (alpha = 0.05) during on treatments than during off treatments.

During fill events: In control treatments, the distribution of juvenile salmon was observed to be relatively uniform, with greatest mean densities 8-10 m's from the transducers. During strobe light-on treatments, juvenile salmon were distributed higher in the water column, with greatest mean densities occurring 5-7 m from the transducers, and very low densities at the depth of the culvert. Mean fish densities at the depth of the culvert decreased by 87% during strobe light-on treatments. Based on paired t-tests, mean fish densities in 1-meter strata below a depth of 9 m were all significantly lower (alpha = 0.05) during on treatments than during off treatments.

20. Edward O. Belcher, Applied Physics Laboratory, University of Washington, Seattle, WA

The fishing industry has harvested an average of 25 million Sockeye Salmon each year in Bristol Bay, Alaska between 1977 and 1996. To remain sustainable, this fishery needs accurate forecasts of abundance. Most Sockeye Salmon smelts migrate to Bristol Bay via the Kvichak River. The last three years the Alaska Department of Fish and Game (ADF&G) estimated over 300 million smelts migrating down that river.

The Kvichak River is a challenge for hydroacoustics. At the measurement site, it is 3 m deep and 130 m wide. Its water velocity ranges between 3 Ws and 6 Ws. Counts begin when the ice breaks until the migration tails off-usually between mid-May to mid-June. The environment is noisy with entrained bubbles from wins) rain, and boat traffic. The river has a rock bottom and contains debris and ice.  The smelts present a challenge to hydroacoustics as well. They are between 80-110 mm long and travel in schools. Some schools are so large that their silvery mass takes minutes to drift by.  The current acoustic assessment is accomplished with a Bendix acoustic system that has three arrays of transducers. The 3.3-m arrays are placed on the river bottom and look up to the surface. The arrays are calibrated to count once each time 41.5 g of biomass pass in a unit time. The three arrays together cover 7.6% of the river width. The ADF&G interpolates between array counts to estimate total count.

In 1990 and 1991 Daniel Huttunen and Paul Skvorc, II (ADF&G) used a side-looking sonar and found that the distribution of the smelt significantly varied over the river width. They were not successful in obtaining TS measurements using a dual-beam approach because the transducer beamwidth was too large and thus the effective range was too short. They also estimated counts using echo integration techniques and found the method very susceptible to noise. They used one beam to interrogate the river width and suggested that a transducer be placed on either side with each measuring one-half of the width (65m).

Our plan is to use the existing upward-looking sonar as a base line and install a modem side-looking sonar with a transducer on each bank. We will also install underwater video that images a volume of water over one of the Bendix transducer arrays and part of one of the side-looking beams. We plan to use ambient light for the video. According to past surveys 32% of the smelts travel down the Kvichak River from 0600 to 1800 each day. The side-looking system will have split-beam capability for TS measurements and fish tracking when possible. The system will use chirp processing for echo integration and echograms with left and right transducers differing by up-chirp and down-chirp codes to minimize crosstalk. The transducers will be custom made with either 20 conical or 20 by 40 elliptical patterns. Raw data will be recorded for post analysis and comparison with video and the upward-looking sonar counts. The goal of the study is to develop methods and algorithms to minimize false counts and missed counts for these challenging river and smelt-school conditions. The study will provide a second independent abundance estimate to compare with the Bendix units and provide tighter confidence intervals than currently obtained.

Correspondence: Email: ed@apl.washington.edu Phone: (206)685-2149 FAX: (206)543 6785