© 2003 by ICES/CIEM International Council for the Exploration of the Sea/Conseil International pour l'Exploration de la Mer
Acoustic Doppler current profiler observations of herring movement
a Department of Physics and Physical Oceanography, Memorial University of Newfoundland St. John's, Newfoundland, Canada A1B 3X7
b Institute for Marine Research Bergen, Norway
*Correspondence to L. Zedel. e-mail: zedel{at}physics.mun.ca; tor{at}imr.no; ranjan{at}physics.mun.ca.
Observations were made of over-wintering (December 1997) and migrating (January 1998) Norwegian, spring-spawning herring (Clupea harengus) using a moored 307 kHz acoustic Doppler current profiler (ADCP). The location of herring in ADCP data is identified by regions of volume-backscatter strength greater than –60 dB re 1 m–1. The presence of herring was verified using net trawls and 38 kHz, EK500 data. While the ADCP cannot make speed measurements of individual fish, the system does provide a measure of the swimming speed and direction of large herring schools. Herring were observed to move both horizontally and vertically: horizontal speeds were from 0 to 50 cm s–1. Higher speeds were observed during daylight hours for both deployments with somewhat increased activity at both dawn and dusk. At night-time, over-wintering herring demonstrated no well-defined velocity.
Keywords: herring, ADCP, migration, swimming speed, swimming direction
Received 18 October 2001; accepted 4 June 2002.
 |
Introduction |
|---|
 Top Introduction Doppler-sonar operationHerring in OfotenExperimental procedureData processingBackscatter dataSpeedDirectionFish localizationResultsDiscussionSummary and conclusionsReferences |
|---|
Doppler sonar provides some attractive possibilities for application to the measurements of fish movement and activity. These systems should be able to detect both the location of targets and determine their speed. An additional important consideration is that they also have the potential to provide information on fish movement in the context of current velocity. If a practical system could be realized it would have great potential for research into fish behavior as well as for contributing to fisheries management.
There have been sporadic experiments at applying Doppler-sonar technology to the problem of measuring fish movement that go back to the work of Bainbridge (1958) as referenced in Holliday (1974). A fundamental problem that surfaces in all of the earlier work is the trade-off between spatial resolution and Doppler-speed resolution. In practice it is impossible to achieve high accuracies in both the speed and spatial domain at the same time. As a result, Doppler sonar has only been applied in special circumstances where the measurement of one of these two components can be relaxed in some way.
Holliday (1974) reports on a 30 kHz sonar used to measure the collective motions of a fish school at distances of up to 1200 m. Using a 0.5 s duration pulse, speed measurements with a resolution of 0.01 m s–1 could be achieved. The long pulse, however, provided no spatial-sampling abilities (a 0.5 s pulse averages over about 375 m in range). The poor spatial-sampling abilities were accepted in the specific application where the distribution of velocities associated with fish movements within a large school were measured. The fish school was localized independently with an 11 kHz sonar. Analysis of this data allowed the extraction of tail-beat frequencies as well as fish-school speed relative to the water.
A domain of application in which the spatial-sampling constraints are somewhat relaxed occurs in monitoring the movement of fish in rivers. Pincock and Easton (1978) identify the Doppler signature in the frequency domain as a means of discriminating fish from various false targets. In addition, they discuss what operating frequencies would be suitable for either a continuous-wave Doppler system or pulsed system: they conclude that for a pulsed sonar system an operating frequency of the order 2.5 MHz would be required to achieve suitable speed and range resolution.
Some improvement in performance can be achieved by adjusting the signal-analysis approach; Hendershot and Acker (1984) describe a system where correlation between a transmit-pulse template and the received-backscatter signal are used to localize fish. Another technique is to avoid undertaking complete spectral analysis of the backscattered signal and only estimate the mean Doppler shift using the autocorrelation technique described in Miller and Rochwarger (1972) as discussed by Waite and Belcher (1985). This algorithm is used extensively in incoherent or narrow-band, Doppler current profilers and provides for enhanced range resolution.
Substantially greater speed and spatial resolution can be achieved by using coherent or coded-signal processing used in modern current-profiling systems as discussed in detail in a later section. Doppler current-profiling systems are not optimized to detect fish but there have been reports of current data being biased by fish presence (Freitag et al., 1992; Wilson and Firing, 1992): the occurrence of such bias clearly demonstrates that these instruments are capable of measuring fish velocities. Demer et al. (2000) report on an application of a ship-mounted, current-profiling system to monitor the movements of a fish school. Using an approach very similar to that used by Holliday (1974) they determined details of the fish-school movement relative to the water. They note many of the limitations of the current-profiling system for this work but their results demonstrate the potential of the technique when the limitations are taken into account.
In this paper we provide some background on the signal processing required in Doppler, current-profiling systems. We present observations of Norwegian herring made with a self-contained instrument (an RDI Workhorse acoustic Doppler current profiler (ADCP)). This instrument can only sample at a fixed point and so relies on fish to transit its location as in riverine applications of the technique and, because it is self-contained, the data rate is somewhat limited. The advantage of the moored instrument is that it eliminates in large part the need to correct for platform motion and there would be no technical reason why the instrument could not be deployed for several months. As a result, such an instrument could be left to monitor fish movements as a sentinel for an extended period of time in situations where fish are known to aggregate in a restricted area. The data presented provide an example of the capabilities of such instruments under circumstances when fish densities are adequate to meet the sampling requirements of the current-profiling ADCP geometry.
|
Doppler-sonar operation |
|---|
|
TopIntroductionDoppler-sonar operationHerring in OfotenExperimental procedureData processingBackscatter dataSpeedDirectionFish localizationResultsDiscussionSummary and conclusionsReferences |
|---|
The Doppler-sonar system tested in this trial is the RD Instruments, 307 kHz Workhorse system. This unit operates using the broad-band processing technique and is packaged as a self-contained, internally recording instrument.
The broad-band system is a variation of the fully coherent sonar, as described by Lhermitte and Serafin (1984), where backscatter from successive, independent acoustic pulses is analyzed. In broad-band sonars, a single pair of acoustic pulses are transmitted and the combined backscatter is recorded. This signal is analyzed to determine the phase change that has occurred between backscatter of the two pulses from a given range. The phase change results from the movement of the scattering particles and by knowing the time between the two pulses the speed can be inferred. These systems provide improved velocity accuracy over so-called narrow-band Doppler sonars where the actual frequency content of backscattered sound is used to estimate speed (Brumley et al., 1991).
An important consideration in coherent and broad-band systems is the occurrence of an ambiguity velocity. If scatterers move more than a quarter wavelength between the passage of the two acoustic pulses, then the speed cannot be resolved accurately; the ambiguity velocity is given by:
|
| (1) |
|
| (2) |
is the signal correlation coefficient and Ma is the number of independent samples realized per averaging interval. An important component of Equation (2) is that accuracy is independent of the length of the transmitted pulses. This characteristic is the important distinguishing factor between narrow-band Doppler sonar where short-pulse lengths lead to poor velocity accuracy (Pinkel, 1979).
In Equation (2), aside from the set operating parameters, the accuracy is limited by the correlation coefficient . For discrete targets the correlation can approach 1, but for volume scattering it is typically 0.5 because of interference between the two transmitted pulses (Brumley et al., 1991; Zedel and Knutsen, 2000). The correlation in returns between the successive pulses decreases with time and so places a limit on the maximum-usable lag. For 307 kHz systems this limit is about 20 ms as discussed by Brumley et al. (1991). The correlation coefficient can be used to discriminate the quality of data and a typical threshold value is 0.25: 64 counts as specified by RD Instruments documentation, RD Instruments (1999). No matter what data-processing mode is used, the Doppler-profiling systems can only measure speeds radial to the acoustic transducer. In a practical sense what this means is that a particular transducer will measure the component of velocity resolved along its beam axis: well-defined beams are needed to get well-defined velocity components (see discussion by Demer et al., 2000). However, if three-dimensional velocity is of interest, it is necessary to use multiple beams with different orientations to measure multiple components of velocity. The ADCP overcomes this limitation by employing (typically) four beams oriented at 0, 90, 180 and 270° in the horizontal plane, and directed 70° below the horizontal. The four beams do not make measurements at a single point but, if it can be assumed that the flow field is uniform over the sample area, then the four component measurements can be inverted to estimate the flow velocity. Theriault (1986b) provides a detailed analysis of this sampling problem.
The RD Instruments 307 kHz WorkHorse ADCP used in this study can profile velocities over a 100–150-m depth range. Each of the beams has a 3 dB beam width of 2.2° (average side lobe levels are –41 dB relative to the main lobe). This system is not a general-purpose sonar system but rather a purpose-built, current-profiling unit and as such it gives the user limited control over system operation. The parameters that can be adjusted include speed sensitivity (and consequently the ambiguity velocity), profile-bin size and sample rate. The system has an operating bandwidth of 75 kHz which would suggest a range resolution of about C/(75 000)/2=1 cm (where C = 1475 ms–1 is the speed of sound in water). Unfortunately, this bandwidth is only realized for the code transmissions: the range resolution of sample bins is limited to 20 cm in depth. In addition there is a limit of 128 sample bins so that larger bins are required to achieve greater maximum sample ranges. Speeds are measured along the four beams and these values are combined by firmware inside the instrument to produce velocity profiles (corrected to an earth-coordinate system) and these are provided as output to the user.
The ADCP also acquires backscatter-amplitude data from the four beams and accumulates these into depth bins as is done for the processed velocity data (bin sizes 20 cm and larger). The backscatter data generated by the ADCP is identical to the data that would be recorded using an echosounder and represents the range evolution of backscatter along the acoustic beam. An important distinction is that the ADCP generates a separate record for each of the four diverging beams.
|
Herring in Ofoten |
|---|
|
TopIntroductionDoppler-sonar operationHerring in OfotenExperimental procedureData processingBackscatter dataSpeedDirectionFish localizationResultsDiscussionSummary and conclusionsReferences |
|---|
The ability of the ADCP to measure fish speeds required a location where the fish behavior would best match its operating constraints. In particular, in order for component-velocity estimates from the diverging beams to be combined to produce a fish-velocity estimate, it was essential that the fish be organized into schools that spanned the distance between the diverging beams. In addition a good understanding of the particular fish stock being observed was desirable. Given these considerations, the herring that over-winter in the Ofoten area of northern Norway provided an ideal study opportunity. As an example of the concentrations of herring that occur in this area, Figure 1 shows 30 min of output from an EK500 echosounder observed on January 28, 1998: data from this area are discussed in greater detail later in the paper. A great deal of information on the behavior of over-wintering Norwegian spring-spawning herring is available (Vabø, 1999) offering a framework within which the observations of herring motions could be gauged. The over-wintering herring stock is also restricted to a limited geographical area that is shielded from adverse-weather conditions and the admixture of other species is insignificant (Huse and Korneliussen, 1995).
|
Large migrations between feeding grounds, over-wintering and spawning areas are common for herring. The Norwegian spring-spawning herring appear north of Iceland in the summer months and they usually feed in the Norwegian Sea until the end of July. It is likely that herring begin their winter migration when the abundance of their primary-prey organisms, the copepod Calanus finmarchicus, reduces significantly some time in autumn (Vabø, 1999). They approach the Norwegian coast in September and enter the over-wintering areas in both the Ofotfjord and Tysfjord around the end of October. The wintering season ends between the end of December and the middle of January when the southward-spawning migration is initiated.
Throughout the over-wintering and migration periods, the herring undertake a diurnal vertical migration, being close to the surface at night and at greater depth during the day. The vertical movement serves to conserve energy (Huse and Ona, 1996), provides the fish with an opportunity to refill their swim bladders (Blaxter and Batty, 1984), minimizes predation risk (Huse and Ona, 1996), and also maintains contact with possible navigational cues at the surface (Huse and Korneliussen, 1995).
While over-wintering, if daytime illumination is insufficient for schooling, herring either swim slowly with a significant upward tilt or alternate between swimming upward and gliding down (Huse and Ona, 1996). While schooling, the herring maintain relatively constant speed, heading and individual distances (Blaxter and Hunter, 1982).
|
Experimental procedure |
|---|
|
TopIntroductionDoppler-sonar operationHerring in OfotenExperimental procedureData processingBackscatter dataSpeedDirectionFish localizationResultsDiscussionSummary and conclusionsReferences |
|---|
The 307 kHz Workhorse ADCP can provide profiles over a depth range of up to 150 m. However, in Vestfjord herring can be located over a substantially greater depth range. Thus a method of positioning the instrument in order to allow herring schools to fall within the ADCP's sampling range was required. For this purpose, the instrument was deployed in a configuration suspended from the surface as shown schematically in Figure 2. The length of the mooring cable to the surface was adjusted so that the instrument would be located at an appropriate depth.
|
Deployments were made during December 1997 and January 1998 from the vessel R/V "Johan Hjort" during cruises 1997216 and 1998201. Figure 3 identifies the times and locations of the deployments. In each case, accumulations of herring were identified using the echosounder systems on the "Johan Hjort" and the mooring geometry was adjusted to position the ADCP above the herring. At the time of the January cruise, the herring had started to migrate southward and abundance had significantly decreased in Ofotfjord and Tysfjord compared to December 1997. As a result, observations during January 1998 were made in the southeastern part of the Vestfjord.
|
Between ADCP deployments, trawl samples were undertaken to measure fish size distribution and weight. Abundance and vertical distribution of herring were monitored continuously using the Simrad EK500 38 kHz echosounder system. Profiles of water-density structure were also acquired using the Seabird model 911 CTD.
The ADCP configuration was essentially the same during each deployment as indicated in Table 1. In particular, with 75 bins of 2-m length a total recorded-profile range of 150 m was possible. It is important to note that by choosing a bin length of 2 m, the depth resolution of the instrument is restricted to 2 m consistent with the transmit-pulse length of 2.36 m (Table 1). One important distinction between the 1997 deployments and those of 1998 is that in the latter year data were recorded for individual acoustic beams. This was intended to allow the investigation of discrete target movements but, unfortunately, depth averaging associated with the comparatively large (2 m) depth bins made target identification all but impossible.
|
|
Data processing |
|---|
|
TopIntroductionDoppler-sonar operationHerring in OfotenExperimental procedureData processingBackscatter dataSpeedDirectionFish localizationResultsDiscussionSummary and conclusionsReferences |
|---|
The Workhorse ADCP provides profiles of velocity, backscatter intensity and additional data-quality indicators arranged in discrete, depth-averaged bins. For precise positioning of data with respect to the instrument it is necessary to take into account the exact instrument orientation. We do not make any such correction in the present analysis so that our observations are made with respect to the vertical axis of the instrument in fact. Our data records show that this axis is as much as 5° from the vertical and so the profiles may be slightly off in depth. Slight mis-orientations might appear at first to cause significant errors in velocity estimation but the four-beam, "Janus" configuration provides velocity estimates that are effectively corrected to first order for this rotation (Theriault, 1986a). In this study we focus on backscatter data and the velocity data as represented by speed and direction profiles. In this section we describe the data processing that we have used for each of these data components and also comment on how the data types are used in combination.
|
Backscatter data |
|---|
|
TopIntroductionDoppler-sonar operationHerring in OfotenExperimental procedureData processingBackscatter dataSpeedDirectionFish localizationResultsDiscussionSummary and conclusionsReferences |
|---|
Backscatter data have been used to localize the presence of fish and it is therefore an important component of the present study. Backscatter data recorded by the Workhorse must be processed to convert observations into calibrated results. We use an approach described by Deines (1999) to recover volume-backscattering strength (Sv) from the recorded data.
Backscatter data in the Workhorse ADCP is extracted from the receiver circuit's "Receive Signal Strength Indicator (RSSI)" output. This value is proportional to the logarithm of the signal strength with a sensitivity (Kc) of about 0.45 dB/LSB; the exact slope varies between different receiver circuits. A calibration of this sensitivity can be done with basic laboratory equipment: we no longer have access to the instrument used for the present study and so we must use the approximate value of 0.45 dB/LSB. Using the recorded-backscatter signal, backscatter strength in dB re 1 m–1 are estimated using the relation given by Deines (1999),
|
| (3) |
is the sound-absorption coefficient = 0.0873 dB m–1 (at 4°C, 307 kHz), E is the recorded-backscatter signal, and Er is the minimum (background) level recorded by the instrument. Calibration of the sonar is achieved by comparing received levels to the background thermal-noise level (represented by Er) and the temperature scaling (Tx+273.16) adjusts for changes in the thermal noise associated with temperature changes. Use of the thermal noise as a calibration point is possible for higher-frequency sonar systems where there is little or no naturally occurring sound. Values of Sv are computed for each of the four ADCP beams separately and the maximum of these four values was chosen as the representative value for a given sample. |
Speed |
|---|
|
TopIntroductionDoppler-sonar operationHerring in OfotenExperimental procedureData processingBackscatter dataSpeedDirectionFish localizationResultsDiscussionSummary and conclusionsReferences |
|---|
Speed data were collected in this study using a beam radial, ambiguity velocity of 170 cm s–1, 72.3 ms time lag between transmit codes and 2-m depth bins with no initial averaging of profiles. This configuration can be converted into velocity uncertainty using Equation (2) by choosing the appropriate parameter values: the correlation coefficient for a broad-band ADCP is ideally 0.5, and the effective number of independent samples is determined by the bandwidth limited range resolution,
|
| (4) |
|
| (5) |
v is the single-beam, radial-velocity uncertainty (1 cm s–1) and the sin(20) accounts for the component of horizontal velocity resolved. We find for the present configuration vr=4.1 cm s–1. Brumley et al. (1991) note that the accuracy is degraded by hardware limitations by a factor of about 1.5 which, for the present case, gives an uncertainty of 6.2 cm s–1: this result is close to the uncertainty predicted by RD Instruments deployment-preparation, software-package "Plan" which predicts an uncertainty of 5.9 cm s–1. This value represents an upper limit on the single-ping accuracies that can be anticipated for the present application.
An additional factor that may be important in the present context is the fact that these instruments do not make a point measurement of velocity but rather include information from the four diverging beams. At any depth d below the instrument information from points separated by 2d cos(20°) in the horizontal are included in the measurement. It is necessary to assume that the flow is homogeneous over this interval in order to extract three-component, velocity estimates. While temporal resolution is clearly lost by extensive averaging of the data, the assumption of flow homogeneity is somewhat more reasonable if a long averaging time is considered. For the present analysis, a Butterworth filter is used to smooth out variations that occur over intervals of 20 profiles: this corresponds to a cut-off frequency of 0.005 Hz for the 1997 data and 0.01 Hz for the 1998 data. Assuming that each profile is independent for the purposes of averaging, the resulting profile uncertainty is reduced by 
to give 1.3 cm s–1 in the present case.
|
Direction |
|---|
|
TopIntroductionDoppler-sonar operationHerring in OfotenExperimental procedureData processingBackscatter dataSpeedDirectionFish localizationResultsDiscussionSummary and conclusionsReferences |
|---|
The ADCP converts direct observations of speed along the four acoustic beams into a single velocity profile referenced to the earth through use of an internal compass. Direction data suffers from the same spatial-sampling problems that the speed estimates have and that data has been smoothed using the same filters as used on the speed data. Direction cannot be filtered directly, rather the data were converted into northward- and eastward-velocity components, these component values were filtered and then re-computed into direction estimates.
|
Fish localization |
|---|
|
TopIntroductionDoppler-sonar operationHerring in OfotenExperimental procedureData processingBackscatter dataSpeedDirectionFish localizationResultsDiscussionSummary and conclusionsReferences |
|---|
Normally Doppler-velocity profiles are based on background levels of volume-backscattering strength and no distinction is made as to the presence or absence of targets. The ideal signal is realized when there is little variation in backscatter as a function of range. The quality of the velocity data does not provide an indication of fish presence. As a result, the backscatter strength must be used to discriminate fish presence and that information can then be used to interpret the velocity data. In the present study, we have used a threshold-backscatter strength of –60 dB re 1 m–1 to identify the presence of fish based on inspection of the data. At large ranges where the attenuation and spherical-spreading terms in Equation (3) begin to dominate, background (noise) levels are eventually amplified to the point where they can exceed the threshold level. We have restricted our analysis to a maximum range of 100 m to avoid this difficulty.
|
Results |
|---|
|
TopIntroductionDoppler-sonar operationHerring in OfotenExperimental procedureData processingBackscatter dataSpeedDirectionFish localizationResultsDiscussionSummary and conclusionsReferences |
|---|
A total of 10 deployments were completed during the two cruises being considered. Some of these deployments were made to test the instrumentation or duplicate results more clearly shown in other deployments while others were not successful in locating fish. We will focus our discussion on two of the deployments that demonstrate the abilities of the ADCP system: deployment 4 during December 1997 (referred to here as Dep9704) demonstrates a transition between the night-time and daytime behavior of over-wintering fish, deployment 4 during January 1998 (referred to here as Dep9804) demonstrates the transition between night and day behavior for migrating fish.
Deployment Dep9704: December 7, 1997, 03:00–12:30 Central European Time
The deployment location (68°28.9'N 17°24.8'E) is indicated by • in Figure 3 and with the ADCP positioned at a depth of 50 m profile coverage over the depth interval from 50 to 150 m was achieved. Based on the time of year we expect that the fish are over-wintering and not migrating during these observations. The density and temperature profiles collected for this deployment are shown in Figure 4. The profile reveals a complicated structure with three distinct regions; a mixed layer of cold (5.2°C) fresh water is found above 40 m, between 40 and 100 m there is evidence of warmer (7.5°C), somewhat saltier water, and below 100 m, there is relatively uniform water properties at an intermediate temperature of 
6°C. Despite the substantial temperature structure the density profile (Figure 4b) does not show any abrupt pycnoclines.
|
The ADCP and corresponding 38 kHz EK500 data for deployment Dep9704 are shown in Figure 5. The region of ADCP-data coverage is indicated in the EK500 data by lines at 50 and 150 m depth in Figure 5b. For the first four hours of this deployment, the backscatter strength (Figure 5a) shows patches of backscatter at all depths with perhaps increased bands of scattering at 70 and 90 m; the same depths at which large temperature gradients are observed). Beginning at 07:00, however, well-defined regions of scatterers become apparent at between 75 and 90 m depth. With time, the scattering layer evolves into a single well-defined layer which varies in depth reaching a minimum of 70 m and a maximum of 130 m.
|
The EK500 data for Dep9704 (Figure 5b) show two well-defined bands of herring centered at 80- and 150-m depth with evidence of fish at all depths between 50 and 150 m. At around 09:00, the lower band appears to end but the bulk of concentration that was initially at 80 m descends to about 150 m. The linking of these scattering layers with herring was verified by trawl surveys undertaken at the same time as the acoustic observations. A multi-sampler with three nets mounted on the cod-end was used to take stratified samples at 22:44 December 6 and 01:01 December 7, just prior to the Doppler system deployment. The trawl catches identify herring at depths of 45–58, 87–107 and 130–168 m. Size distributions from these deployments are shown in Figure 6 for each of the three net depths.
|
Profiles of direction are shown in Figure 5c. There are three bands of flow indicated: above 80 m, motion is toward 50° true, at depths between 80 and 100 m motion appears toward 200 to 250° and below 110 m, the motion is toward
0°. The velocity structure is broken up after 10:00 by motions associated with the well-defined scattering band. Profiles of observed speed are shown in Figure 5e. Most of the record shows a background low speed of less than 10 cm s–1 with somewhat distributed observations of higher speed events. There are two regions where "higher than background" speeds are seen consistently; between 06:00 and 08:00 at 90–100-m depth, and also after 10:00 at depths that follow the region of enhanced scattering.
The occurrence of fish throughout the region sampled by the ADCP—indicated by the black horizontal lines in Figure 5b—makes it difficult to distinguish fish speeds from water speeds. The observed speeds can be considered as an upper limit on water speeds, assuming that the motion of herring would only lead to higher speeds. Using this assumption, current speeds are generally very low <10 cm s–1 and this is certainly consistent with the enclosed nature of the observation area with tidal-model predictions for this location predicting current speeds less than 1 cm s–1 (Moe et al., 2002). There is some structure indicated in the current direction and it would appear to be associated with the temperature intrusion that occurs between depths of 40 and 100 m.
The EK500 data (Figure 5b) show that herring are present over the entire range of the ADCP measurements until about 09:00 (Central European Time (CET)). Over this period of time, strong backscatter is in fact seen at all depths in the ADCP data (Figure 5a). After 09:00 CET, the EK500 data indicates that the herring move into a thinner layer at about 100-m depth and the ADCP data mirror this showing a more densely concentrated scattering layer that has significant vertical movements especially after 10:00 CET. In this case, the motion of the herring can be isolated based on backscatter strength relative to background levels. Identifying all targets for which Sv>–60 dB re 1 m–1 as being caused by herring, Figure 5d, f shows the swimming direction and speed of the herring while excluding all other data. In Figure 5d, f most of the targets identified as fish drift with the prevailing current until about 10:00 when their direction of motion deviates substantially from that of the local currents and changes constantly. Also at this time, the average speed of detected targets increases to about 20 cm s–1 (Figure 5f).
Deployment Dep9804: January 28, 1998,12:00–16:07 CET
Deployment Dep9804 occurred at a position in the southeastern part of Vestfjord (67°57.06'N 14°40.7'E) as indicated by 
in Figure 3. The water-property profiles (shown in Figure 7) are similar to those observed during deployment Dep9704 in December: there is a mixed layer above 30 m and a strong thermocline at 75 m separating 4°C water from that of 7.5°C. At 150 m there is a temperature inversion going from 7.5 to 6.9°C. The density structure shows the well-defined mixed layer and there is a pycnocline at 75 m corresponding to the thermocline.
|
ADCP and EK500 data for this deployment are shown in Figure 8: the ADCP was positioned at 150-m depth and provided profile information over the depth interval 150–250 m. Through most of the time the ADCP-backscatter strength identifies a well-defined band (Figure 8a) that corresponds with a similar structure in the synoptic 38 kHz EK500 data (Figure 8b). This band of scatterers was verified as being caused by herring through trawl catches made at this time: the herring's size distribution from trawl sets are shown in Figure 9. Exact agreement between the ADCP and EK500 data cannot be expected because the ADCP remained at a fixed location while the EK500 data were collected while undertaking a survey in a region within 500 m east and west and 8 km north and south of the ADCP. The herring rise from a depth of 240 m up to 170 m through the deployment and then disperse or move above the instrument at the end. The scatterer band becomes somewhat less well defined between 13:00 and 14:30.
|
|
ADCP-direction data for entire profiles are shown in Figure 8c. The motion of the herring alone can be isolated by identifying all targets for which Sv>–60 dB re 1 m–1; based on this criterion, Figure 8d isolates the swimming direction of the herring. An interpretation of the background current structure can now be made by considering Figure 8c, d with reference to Figure 7. Prior to 14:30, the background current is northwestward (
300° true) above 175 m and northeastward (75° true) below 75 m. After 14:30, the entire water profile shifts to a direction of 275° true. The occurrence of velocity shear at 175 m is consistent with the temperature inversion at that depth. After 14:30, the uniform water motion may suggest that the water structure had changed but we have no CTD data to support that speculation. Current speeds throughout appear to be 20 cm s–1 or less while tidal-model results for this area and time indicate currents that are less than 10 cm s–1 (Moe et al., 2002). Speed profiles for this deployment (Figure 8e) show very low speeds except in the area marked by the presence of the herring. It is, in fact, possible to identify the herring school by the anomaly in the speed profile. Using the same approach to isolate the speed of the herring as that used to isolate their swimming direction, Figure 8f displays the herring-swimming speed alone.
The motion of the herring school as a whole can be extracted by averaging the direction and speed profiles (Figure 8d, f), the results of this extraction are shown in Figure 10 as both time series of speed and direction (Figure 10a, c), and also as histograms (Figure 10b, d). The fish were moving with a speed of 20–30 cm s–1 in a direction of 150 to 275° during the time interval of 12:00–13:00 (Figure 10a). From 13:00 to 14:30, the school is somewhat more dispersed (Figure 8a) and the speeds vary considerably between 15 and 45 cm s–1 while the direction becomes more uniform at around 200°. At 14:30, the speed abruptly drops as the school becomes more consolidated and this drop in speed coincides with a sudden change in swimming direction (Figure 10c). The histogram of swimming direction for the entire Dep9804 deployment shows that the distribution has a narrow peak at a direction of 200° (Figure 10d). The speeds range from a low of about 10 cm s–1 to a high of 45 cm s–1.
|
|
Discussion |
|---|
|
TopIntroductionDoppler-sonar operationHerring in OfotenExperimental procedureData processingBackscatter dataSpeedDirectionFish localizationResultsDiscussionSummary and conclusionsReferences |
|---|
Our observations demonstrate a clear signal from herring schools in ADCP-backscatter data and velocity estimates from these regions are distinct from those of the greater water column (see Figs. 5 and 8). Doppler-sonar systems provide a backscatter-intensity-weighted estimate of velocity (see the discussions of Ahn and Park, 1991). This means that when scattering from fish is dominant the system measures fish speed. When there are no fish present adequate backscatter is normally available from zooplankton (see, for example, Flagg and Smith, 1989) and even from the temperature microstructure (Seim et al., 1995) so that the water speed can be measured. Backscatter strength, then, provides a means to discriminate velocity signals that might be coming via fish targets from those coming from the water per se. Here we have used a backscatter-strength threshold of Sv>–60 dB re 1 m–1 to identify herring. In the event the backscatter from other sources approaches the level we expect from herring, this will result in biased velocities. However, background-volume backscatter has Sv<–70 dB re 1 m–1 (see Figure 8) in the present data. There is, therefore, a substantial margin in scattering strength between that of herring as compared to that of the water alone. In the present situation it is quite likely that water speeds are at times biased by the presence of fish but it is unlikely that fish speeds are biased.
With four separate acoustic beams the ADCP provides four independent measurements of volume backscatter. For the present analysis we have reduced the data by choosing the highest of the four values. This approach was taken because averaging in the logarithmic domain had the effect of smearing the boundaries of the fish schools so that the selection of a suitable backscatter threshold then became less obvious. This smearing of boundaries does, however, suggest differences in backscatter observed between the four beams that we have not explored in the present paper but intend to pursue in future studies.
If we accept the backscattering cross section as a means of distinguishing the fish signal in the present data we can generate statistics of the fish motion and direction. One of the basic parameters we have considered is the distribution of speeds (Figure 10). When evaluating these distributions it is important to recall that the uncertainty in individual velocity estimates is 1.3 cm s–1. The variance seen in Figure 10b is therefore an indication of the range of velocities measured and not uncertainty in the measurement itself.
One of the fundamental characteristics of Doppler-sonar systems is the use of three or four separate acoustic beams required to resolve various components of velocity. In order to recover three-dimensional velocities it is necessary to assume that the velocity measurements are homogeneous over the combined footprint of the acoustic beam. For the Workhorse ADCP being investigated here, at a range of 100 m this footprint is approximately 70 m across. As a result, it is impossible to measure the three-dimensional velocity of a single fish and only the velocity of large fish schools can be determined. If individual fish could be identified in the acoustic beams some statistical description of their motions could be extracted. In the present data we explored this approach but found that with the 2-m depth bins that were employed it was not possible to localize individual (fish) targets.
A severe limitation with the present data set is that we have no direct reference with which to compare the ADCP-velocity estimates. In fact, the direct-velocity measurements of the ADCP are hard to reproduce with other measurement techniques. Estimates of swimming speed can be inferred by tracking the movement of entire fish schools if the outline of the school can be delineated using repeated passes of a survey vessel. Acoustic tags and tag-recovery schemes can provide good estimates of mean swimming speeds but again these are not instantaneous measurements like those of the ADCP. Split-beam sonar systems can track individual swimming fish but this is not generally possible from a ship at sea. In the present case we take advantage of the large body of knowledge on over-wintering-herring behavior to evaluate the consistency of the ADCP observations.
There have been many reports on the vertical-swimming behavior of over-wintering herring (Huse et al., 1994; Huse and Ona, 1996) because this behavior affects the yearly abundance assessment (Røttingen et al., 1994; Vabø, 1999). Several factors may affect the fish behavior, particularly predator avoidance and the need to conserve energy during over-wintering (Huse and Ona, 1996; Huse and Korneliussen, 2000). In this setting the diurnal change in illumination is an important factor with regard to herring behavior as many of its predators like cod, saithe and killer whales are predominantly visual feeders. During this study, herring were observed moving to deep water at sunrise (Figure 5) and ascending at dusk (Figure 8). It is believed that this migration allows the herring to conserve energy through an increase in swim-bladder volume at shallow depths while affording some protection from predators in the dark (Blaxter and Batty, 1990). It is, however, important to note that a proportion of the herring remain in deeper water during the night as has been documented by Huse and Korneliussen (2000).
Our observations from the inner part of Ofotfjord made on 7 December 1997 represent a pure over-wintering situation. During the night the herring are seen near the surface at less than 150 m and the absence of any well-defined swimming speed at this time suggests that the fish are more or less drifting with the water current or at least not moving in any systematic direction (Figure 5c). There is some indication of a second, deeper, herring layer, note the EK500 data shown in Figure 5b, but these two layers appear to merge after 09:00. No prominent speed changes are seen until about 10:00 when, with the arrival of dawn, the swimming speed increases from 10 cm s–1 to a maximum of 30 cm s–1 (Figure 5f). On average, the herring migrate to deeper water from about 08:00 onwards but rapid changes in the vertical distribution suggest that there may be some predator-avoidance activity with increasing daylight.
The low (average) speeds of movement of the herring during December suggest that the herring might be in a state of hibernation. While in this state, a certain type of swimming behavior consisting of upward swimming and gliding has been documented and is possibly an adaptation to decreased buoyancy at depth (Huse and Ona, 1996). In our velocity data, this behavior would lead most likely to an increase in velocity variance and perhaps this behavior accounts for the frequent speed anomalies seen in the first 4 h of Figure 5f. Huse and Ona (1996) report that at intermediate depths herring will take on this swim-and-glide behavior during low light levels. At higher light levels, they report herring at 62–107-m depth having a near-horizontal tilt angle and a mean swimming speed of 33–27 cm s–1. Because of the range limitations of the ADCP, acoustic recordings deeper than 150 m could not be obtained for Dep9704 where the swim–glide behavior might have been expected during the day. The EK500 data (Figure 5b) do confirm that most of the herring population in this region is located at a band around 60-m depth and a second, deeper layer at around 150 m during the night. Huse and Korneliussen (2000) also report occurrences of shallow and deep layers of over-wintering herring in this area.
Based on the time of year and location, our January deployments represent a situation where most likely the herring have initiated their southward, spawning migration. For 28 January (Dep9804) the sun is now above the horizon for about 5 h, from approximately 09:40–14:50. Hence, the herring are active for a longer period than during the December deployments. Prior to about 13:00 the herring school has speeds of around 15–30 cm s–1 with a well-defined direction of movement between 200 and 300°, suggesting that they are heading south or south-west and leaving the Vestfjord region. In daytime observations the recorded speed was as high as 45 cm s–1 (Figure 10a). Given the observed fish length of 30–35 cm (Figure 9) this speed is consistent with a swimming speed of 1 BL s–1 as reported by Huse and Ona (1996). Around sunset and thereafter the herring migrate toward the surface but now have a swimming direction that varies with time and a somewhat lower absolute-swimming speed. The fish disappear from the ADCP recordings at around 15:45 as, in all probability, they migrate above the instrument and disperse. The EK500 data from this time confirm that the herring that were observed in deep waters below 200-m depth during the daytime, migrate to the surface toward the end of the deployment Dep9804 (Figure 8a).
|
Summary and conclusions |
|---|
|
TopIntroductionDoppler-sonar operationHerring in OfotenExperimental procedureData processingBackscatter dataSpeedDirectionFish localizationResultsDiscussionSummary and conclusionsReferences |
|---|
In this paper we have presented observations of herring-school motions made using a moored, 307 kHz, RDI Workhorse ADCP. This deployment approach avoids interference from ship noise and can provide a time series when fish are constrained to move in a well-defined geographic area. Profiles were limited to a 100-m interval because of the 307-kHz operating frequency of the system: a lower frequency system would be able to achieve a greater range. Velocity accuracy is calculated to be ±1.3 cm s–1 in data averaged over 100–200 s. Greater variation in observed velocities appears to be representative of real variations in fish-school movements. In finding these velocities it is necessary to assume that school movement is homogeneous over the footprint of the four acoustic beams used by the ADCP and this is as much as 70 m in the present application. The requirement for this spatial homogeneity restricts this approach only to large aggregations of fish.
Backscatter data recorded by the ADCP has been calibrated using the procedure described by Deines (1999). In this approach, measured receiver sensitivities are combined with factory-specified, source-level settings and observed background-noise levels to arrive at volume-backscatter strength. This procedure made the backscatter values from the four independent beams comparable. We do not have an independent check on the resulting calibration but the observed-backscatter levels from the schools of herring are consistent with the expected levels. A volume-backscattering strength of greater than –60 dB re 1 m–1 was used to discriminate areas of herring in the ADCP data.
The behavior of schools of Norwegian, spring-spawning herring were observed while over-wintering (November 1997) and during the spring migration (January 1998) and have been compared to, and are found to be consistent with, prior observations of this herring stock. Many of the fish were seen to undertake a diurnal, vertical migration rising close to the surface at night (i.e. to depths below 100 m) and to prefer greater depths (200–300 m) during the day. We must, however, stress that not all of the fish undertook this migration. Well-defined schools were formed during hours of daylight and dispersed at dusk. Swimming speeds of no more than 20 cm s–1 during the over-wintering period were less than the maximum of 45 cm s–1 observed during the migration period. They were greater during the day than at night. Some daytime observations were marked by rapid changes in swimming speed and direction suggesting, possibly, that these were attempts to evade predators.
|
Acknowledgements |
|---|
The following are thanked for their support of this work: the EU through RTD-contract no. MAS3-CT95-0031 (BASS), the Norwegian Research Council through grant no. 113809/122 and the Bergen Large-Scale Facility for Marine Pelagic Food Chain Research. The support of Lucio Calise through Convenzione IMC (IMC-Centro Marino Internazionale, Loc. Sa Mardini, I-09072 Torregrande (OR), Italy) Regione Autonoma Sardegna L.R. 2/94, art. 32 titolo 11.3.10/I, is gratefully acknowledged too. This is a contribution to the Mare Cognitum Program at IMR.
Thanks are due to Kenneth G. Foote for providing the opportunity to participate in herring-survey cruises and to Lee Gordon, who made available the RD Instruments WorkHorse Sentinel and FishMass 307 kHz systems for these experiments, and Mark Vogt of RD Instruments, who gave technical assistance. The help of Bjørn Gjevik who generated the tidal-model results for our field sites is much appreciated also.
|
References |
|---|
|
TopIntroductionDoppler-sonar operationHerring in OfotenExperimental procedureData processingBackscatter dataSpeedDirectionFish localizationResultsDiscussionSummary and conclusionsReferences |
|---|
-
Ahn B.Y. and Park S.B. (1991) Estimation of mean frequency and variance of ultrasonic Doppler signal by using second-order autoregressive model. IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control 38:172–182.[CrossRef][Web of Science]
Blaxter J.H.S. and Batty R.S. (1984) The herring swimbladder: loss and gain of gas. Journal of the Marine Biological Association of the United Kingdom 64:441–459.[Web of Science]
Blaxter J.H.S. and Batty R.S. (1990) Swimbladder "behaviour" and target strength. Rapports et Proces-Verbaux Reunions du Conseil International pour l'Exploration de la Mer 189:233–244.
Blaxter J.H.S. and Hunter J.R. (1982) The biology of clupeids. Advanced Marine Biology 20:1–223.
Brumley B., Cabrera R., Deines K., Terray E. (1991) Performance of a broad-band acoustic Doppler current profiler. IEEE Journal of Oceanic Engineering 16:402–407.[CrossRef][Web of Science]
Deines K. L. (1999) Backscatter estimation using broadband acoustic Doppler current profilers. Proceeding of the IEEE Sixth Working Conference on Current MeasurementSan Diego.
Demer D.A., Barange M., Boyd A.J. (2000) Measurements of three-dimensional fish school velocities with an acoustic Doppler current profiler. Fisheries Research 47:201–214.[CrossRef][Web of Science]
Flagg C.N. and Smith S.L. (1989) On the use of the acoustic Doppler current profiler to measure zooplankton abundance. Deep-Sea Research 36:455–474.
Freitag H. P., McPhaden M. J., Pullen P. E. (1992) Fish-induced bias in acoustic Doppler current profiler data. IEEE Oceans 1992 Proceedings pp. 112–117.
Hendershot R. G. and Acker W. C. (1984) Doppler techniques applied to fisheries hydroacoustics. IEEE Oceans 1984 Proceedings pp. 15–20.
Holliday D.V. (1974) Doppler structure in echoes from schools of pelagic fish. The Journal of the Acoustical Society of America 55:1313–1322.[CrossRef][Web of Science]
Huse I. and Korneliussen R. (1995) Diurnal variations in acoustic density measurements of wintering Norwegian spring spawning herring. ICES CM 1995/B: 12, Ref. H. 17 pp.
Huse I. and Korneliussen R. (2000) Diel variation in acoustic density measurements of over-wintering herring (Clupea harengus L.). ICES Journal of Marine Science 57:903–910.
Huse I. and Ona E. (1996) Tilt angle distribution and swimming speed of over-wintering Norwegian spring spawning herring. ICES Journal of Marine Science 53:863–873.
Huse I., Foote K., Ona E., Rottingen I. (1994) Angular distribution of over-wintering spring spawning Norwegian herring. ICES CM 1994/B: 19, Ref. H. 14 pp. (Mimeo).
Lhermitte R. and Serafin R. (1991) Pulse-to-pulse coherent Doppler sonar signal processing techniques. Journal of Atmospheric and Oceanic Technology 1:4293–308.[CrossRef]
Miller K.S. and Rochwarger M.M. (1972) A covariance approach to spectral moment estimation. IEEE Transactions on Information Theory 18:588–596.[CrossRef]
Moe H., Ommundsen A., Gjevik B. (2002) A high resolution tidal model for the area around The Lofoten Islands, northern Norway. Continental Shelf Research 22:485–504.[CrossRef][Web of Science]
Pincock D.G. and Easton N.W. (1978) The feasibility of Doppler sonar fish counting. IEEE Journal of Oceanic Engineering OE-3:37–40.[CrossRef][Web of Science]
Pinkel R. (1979) Acoustic Doppler techniques. In Dobson F., Hasse L., Davis R. (Eds.). Instruments and Methods in Air–Sea Interaction(Plenum Press, New York) pp. 171–199 Chapter 10.
RD Instruments. (1999) Workhorse Technical Manual. (RD Instruments, San Diego).
Røttingen I, Foote K., Huse I., Ona E. (1994) Acoustic abundance estimation of wintering Norwegian spring spawning herring, with emphasis on methodological aspects. ICES Conference: Joint Session on Estimating Abundance from Fishing Surveys and Acoustic Measurements ICES-CM-1994.
Seim H.E., Gregg M.C., Miyamoto R.T. (1995) Acoustic backscatter from turbulent microstructure. Journal of Atmospheric and Oceanic Technology 12:367–380.[CrossRef][Web of Science]
Theriault K.B. (1986) Incoherent multibeam Doppler current profiler performance. Part I—estimate variance. IEEE Journal of Oceanic Engineering OE-11:7–10.[CrossRef][Web of Science]
Theriault K.B. (1986) Incoherent multibeam Doppler current profiler performance. Part II—spatial response. IEEE Journal of Oceanic Engineering OE-11:16–25.[CrossRef][Web of Science]
Vabø R. (1999) Measurement and correction models of behaviourally induced biases in acoustic estimates of wintering herring (Clupea harengus L.). PhD thesis, University of Bergen, Norway.
Waite J. W. and Belcher E. O. (1985) Sonar detection of riverine fish using the pulse pair covariance Doppler frequency estimator. IEEE Oceans 1985 Proceedings pp. 700–706.
Wilson C.D. and Firing E. (1992) Sunrise swimmers bias acoustic Doppler current profiles. Deep-Sea Research 39:885–892.
Zedel L. and Knutsen T. (2000) Measurement of Fish Velocity Using Doppler Sonar. Proceedings Oceans 2000September 2000.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  L. Zedel and F.-Y. Cyr-Racine Extracting fish and water velocity from Doppler profiler data ICES J. Mar. Sci., October 1, 2009; 66(9): 1846 - 1852. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. Zedel, R. Patro, and T. Knutsen Fish behaviour and orientation-dependent backscatter in acoustic Doppler profiler data ICES J. Mar. Sci., January 1, 2005; 62(6): 1191 - 1201. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||