Age characteristics of walleye pollock school echoes

ICES Journal of Marine Science, 63: 1465e1476 (2006)
doi:10.1016/j.icesjms.2006.06.007
Age characteristics of walleye pollock school echoes
Myounghee Kang, Satoshi Honda, and Tatsuki Oshima
Kang, M., Honda, S., and Oshima, T. 2006. Age characteristics of walleye pollock school
echoes. e ICES Journal of Marine Science, 63: 1465e1476.
The purpose of this study was to investigate the possibility of identifying the age of walleye
pollock (Theragra chalcogramma) using acoustic information. Acoustic data targeting
walleye pollock were collected at 38 and 120 kHz from 16 June to 12 July 2000 in the
Pacific, off Hokkaido, Japan. To complement these data 33 trawl hauls were made and
the species and age of the sample fish were accurately examined. The echoes of walleye
pollock schools according to age were used to determine the morphological and bathymetric characteristics such as mean height, maximum length, centre depth, seabed depth, and
distance from the seabed, as well as the frequency characteristics, this latter being the
difference of mean volume backscattering strengths at 38 and 120 kHz, respectively
(DMVBS). The DMVBS method is elaborated using MVBS (mean volume backscattering
strength) from an integration cell of optimal size, the cell being examined by means of
various integration periods to highlight the characteristics of the walleye pollock schools
resulting in 20 pings (120 m), and by applying this method only in a common observation
range for two frequencies. The ages of the schools are identified by a combination of morphological and bathymetric characteristics, and DMVBS characteristics. Age-0 groups are
easy to distinguish from other age groups because they exist in distinct, small schools, are
close to the coast, and have a narrow range of DMVBS regardless of time of day. Age-1
schools are low in height and very long, are distributed close to the sea floor, and have
an DMVBS range of 1 to 8 dB, with most between 3 and 5 dB. These characteristics
of age-1 schools are distinct from other age groups. As age-2 and age-5 schools have similar
maximum length and distribution depth, it is almost impossible to identify these two by just
morphological and bathymetric characteristics. However, the DMVBS of age-2 and age-5
schools show characteristic patterns that can be used as a means of identification. The pattern of DMVBS, which reflects an internal structure (swimming angles) of a school, is different for each age class, and is essential in the identification of the age of a walleye pollock
school.
Ó 2006 International Council for the Exploration of the Sea. Published by Elsevier Ltd. All rights reserved.
Keywords: age identification, common observation range, morphology and bathymetry,
optimal size of an integration cell, pattern of DMVBS.
Received 18 November 2004; accepted 8 June 2006.
M. Kang: Tokyo University of Marine Science and Technology, Tokyo, Japan. Now at
SonarData Pty Ltd, GPO Box 1387, Hobart 7001, Australia. S. Honda: Hokkaido National
Fisheries Research Institute, 116, Katsurakoi, Kushiro, Hokkaido 085-0802, Japan.
T. Oshima: Marine Fisheries Research and Development Department of Fisheries Research
Agency, 2-3-3 Minato-mirai, Nishi, Yokohama, Kanagawa 220-6115, Japan. Correspondence to M. Kang: tel: þ61 3 6231 5588; fax: þ61 3 6234 1822; e-mail: size100@hotmail.
com, [email protected].
Introduction
Four acoustic methods are used in identifying fish species.
The first is the frequency characteristics method, which is
based on the difference in the frequency characteristics of
fish school scattering by means of a wideband echosounder
(Simmonds et al., 1996; Zakharia et al., 1996) or multiple
frequencies (Conti and Demer, 2002; Kang et al., 2002;
Korneliussen and Ona, 2002). The second is the distribution characteristics method in which morphological,
1054-3139/$32.00
bathymetric, and energetic characteristics are extracted
from school echoes on the echogram. ICES (2000) gives
a comprehensive overview of this method by defining
acoustic school echoes, raising issues of acoustic school images, and suggesting correction algorithms that use simulated school echoes. Many software packages have
applied this method, including Echoview (Higginbottom
et al., 2000), EIPS (Lu and Lee, 1995), FASIT (LeFeuvre
et al., 2000), MOVIES (Weill et al., 1993; Masse et al.,
1996; Scalabrin et al., 1996), SHAPES (Coetzee, 2000),
Ó 2006 International Council for the Exploration of the Sea. Published by Elsevier Ltd. All rights reserved.
1466
M. Kang et al.
TSAN (Sawada, 2000), SCHOOLS, and SCHOOLBASE
(ICES, 2000). It seems that it is a very common method
and easy to apply for species classification. However, if
several species are present together it becomes less effective. The third method is that of signal characteristics.
While the distribution characteristics method relies on the
geometric features of school echoes on the echogram, this
approach deals with the echo signal itself, and in particular
the characteristics of the echo envelope. Rose and Leggett
(1988) found that the distance between the peak and the
trough of the envelope, and the distance between two peaks
were important for identification, whereas Scalabrin et al.
(1996) pointed out that a short echo length restricted the
use of a complicated spectrum analysis. This finding suggested that further investigation was required. The fourth
is the acoustic result method, which is based on data produced from an echosounder, including the volume backscattering strength (SV), target strength (TS), and
swimming speed by echo trace analysis (Richards et al.,
1991; Sawada, 2000). Moreover, environmental information by non-acoustic methods has been employed. The relationship between species and environmental characteristics
such as water temperature, current, and seabed geography
has been studied widely (Cooke et al., 2002).
Although several species have been accurately identified
using the above methods, the methods have not yet been
adopted at a practical level over a wide range of times
and areas (Scalabrin et al., 1996). Moreover, even if the
methods are utilized, acoustic species identification is not
easy because characteristics of echoes within the same
species may be different because of school formations, distribution patterns, and age composition (Misund, 1993;
Coetzee, 2000). Factors causing a variety of characteristics
of echoes within the same species must be examined for
species identification. However, if the characteristics of
the echo per se vary with fish age, it is feasible to identify
an age of the species by acoustic methods. This can be useful for estimating individual stock sizes by age.
In this study we examine the age characteristics of echoes of walleye pollock (Theragra chalcogramma) schools
to identify the representative ages of the schools. When
more information of this type has been collected and the
results verified, it will be easier to identify the species of
marine organisms acoustically (MacLennan and Holliday,
1996; Horne, 2000); this is just a first step. For age identification, a wealth of information is created by combining
the frequency characteristics approach (DMVBS characteristics) and the distribution characteristics method (morphological and bathymetric characteristics). Species were
confirmed by trawl sampling, and ages were measured via
a precise experiment. In this study, the advanced DMVBS
method discussed in Kang et al. (2002) is further improved.
The advanced DMVBS method was applied over a common
observation range to compare the frequency characteristics
of only the target marine organisms at two frequencies. In
this case, the method of calculating observation range has
been enhanced by using measured noise. By applying
a small integration cell, it was possible to reduce the
contamination of multiple species, but hitherto the most
appropriate size of the integration cell has not been investigated sufficiently. One purpose of this study is to determine an optimal size of integration cell that can best
highlight the characteristics of schools. Finally, age identification using two characteristics is attempted for echoes
from walleye pollock schools collected from the Pacific
Ocean off Hokkaido, Japan.
Methods
Acoustic and trawl survey
An acoustic survey targeting walleye pollock schools was
conducted from 16 June to 12 July 2000 in the Pacific
Ocean off Hokkaido, Japan, using the RV ‘‘Kaiyo-maru
#3’’ (460 grt), a vessel originally built for fishing but now
used for research. In all, 19 parallel transects were covered
on the east side of Cape Erimo and ten transects on the
west, as shown in Figure 1. Survey data were collected
from the calibrated 38 and 120 kHz channels of a SIMRAD
EK500 echosounder. Table 1 shows the specifications of
the echosounder and the parameters of the calibration.
Two transducers were mounted at the same depth, on the
bottom of a protruding instrument keel, with the 38-kHz
transducer in front of the 120-kHz transducer. The beam
widths of both transducers were close to 7 . Simultaneous
transmission of pulse was used. The calibration was carried
out on 18 June 2000, and two calibration spheres made
from tungsten were used at 38 and 120 kHz. The water temperature was 4.6 C, salinity was 32.3, and sound speed was
1465 m s1.
Figure 1 also shows the trawl survey stations where the
age and species of walleye pollock were confirmed. To examine differences between day and night echoes, the vessel
covered the same survey line by both day and night of the
same day, when possible. Trawl positions were determined
by daytime echograms, but taking into consideration the
night-time ones. Samples were collected by trawling during
daylight. In all, 33 trawls were used to ground-truth the
acoustic data. A dual purpose, midwater/bottom trawl net
(JAMARC-98, Nichimo Co. Ltd, Tokyo, Japan), made up
of four panels, was designed to sample rather diversely distributed fish schools. The main purpose of using a new
trawl net was to catch walleye pollock schools regardless
of various distribution patterns by time and season. The
net was 63.2 m long, with 11-mm stretched mesh in the codend. When a midwater trawl was used, the size of the
mouth opening was approximately 25 m vertically and
18 m horizontally, and for bottom trawls, the measurements
were 20 m vertically and 23 m horizontally. Bottom trawling was used in most cases.
The samples caught by the trawls were classified by
species onboard. A double-checking method for samples
Age characteristics of walleye pollock school echoes
1467
Figure 1. Transects and trawl stations in the Pacific Ocean off Hokkaido, Japan, surveyed by the RV ‘‘Kaiyo-maru #3’’ from 16 June to 12
July 2000. The interval between two continuous transects was approximately 8 nautical miles. Each line was designed to be nearly perpendicular to the isobaths at depths of 30e500 m (isobaths at 50, 100, 150, 200, 300, and 500 m are shown). Vessel speed during the
survey was approximately 8 knots. The numbers of trawl stations used to confirm the age of fish in the schools acoustically recorded
are shown in coloured circles, red for age-0, orange for age-1, yellow for age-2, and blue for age-5.
was applied in order to obtain accurate age and size. At
every trawl station, fork lengths of 300 fish were
measured, and 100 fish were randomly selected, for age,
length, and weight to be determined. The age of a fish
was accurately determined by counting the yearly translucent zones on a vertical section of an otolith embedded in
black resin.
Observation range
The observation range of an echosounder is derived by finding the border of an acceptable signal-to-noise ratio; this is
Table 1. Echosounder specifications and calibration parameters.
Frequency (kHz)
Transducer
Beam type
Transducer depth (m)
Calibration sphere range (m)
SV transducer gain (dB)
TS transducer gain (dB)
Angle sensitivity (dB)
Alongship
Athwartship
Angle offset (degrees)
Athwartship
Alongship
Beam width (degrees)
Athwartship
Alongship
Equivalent beam angle (dB)
Pulse duration (ms)
Maximum power (W)
Receiver bandwidth (kHz)
Absorption coefficient (dB km1)
38
ES38B
Split
4.1
16.1
26.4
28.6
120
ES120e7
Split
4.1
12.0
23.0
26.0
21.9
21.9
21.0
21.0
0.01
0.1
0.32
0.25
6.7
6.8
20.7
1
2 000
3.8
10
7.7
8.1
20.6
0.3
1 000
12
38
related to the insonified marine organisms, instrument
parameters, vessel noise, and acoustic propagation. The
DMVBS method makes use of the frequency characteristics
of the scattering of marine organisms. It is essential for this
method to be used within a common observation range
across multiple frequencies, in which the effects of the
frequency difference by noise and directivity of the
echosounder are minimal, in order to obtain pure frequency
characteristics of scattering of the target marine organisms.
The concept of a common observation range was introduced by Furusawa et al. (1999) and applied to the DMVBS
method by Kang et al. (2002).
In Kang et al.’s (2002) paper, the noise spectrum level,
an essential parameter for obtaining the common observation range, was calculated using an experimental equation
by Nishimura (1969). However, during the current study,
noise was measured from the echo-integrator output (Takao
and Furusawa, 1995). The noise spectrum level was
accurately calculated for each frequency using noise
measurements from the EK500 echosounder and the RV
‘‘Kaiyo-maru #3’’. Takao and Furusawa (1995) showed
that when the SV corresponding to noise was measured at
echo-integrator output, it was related directly to the mean
power of the noise. This is a practical method, because
the noise component of SV, the so-called noise SV, is measured in the same way as normal SV. The noise spectrum
level can be directly calculated from
ð1Þ
N2P ¼ P20 DI ctJCSV DN = 2Df r2 expð4arÞg ;
where P0 is source pressure, DI the directivity index of a
receiving transducer, c the sound speed in seawater, t
the pulse width, J the equivalent beam angle, Df the bandwidth of the receiver, r the range, a the absorption
attenuation coefficient, g the coefficient caused by the
1468
M. Kang et al.
integration process, and CSVDN is the mean noise SV. The
coefficient g is calculated as follows:
g ¼1=32a3 rw r2 P8a2 r22 h r2 4aðr2 h rÞ þ h 1R;
h ¼expð4arw Þ;
ð2Þ
where the integration is performed for the range between r
and r2 ¼ r þ rw. Table 2 lists the parameters used to calculate the noise spectrum level. Some parameters in Table 1
are used to calculate it. According to Equation (2), g will
be w1 when r is large and rw is small. When there was
no echo of marine organisms, the noise SV was measured
in an integration cell 10 min long in the horizontal and
150e170 m in the vertical. The CSVDN was 83.3 dB at
38 kHz and 69.0 dB at 120 kHz. Calculations using Equations (1), (3), and (4) of Kang et al. (2002) were utilized to
find values of P0 and DI.
There was regrettably an error in Equation (6) of Kang
et al. (2002). The parameter r (the density of seawater)
in the numerator was erroneously displayed as P. The
following corrected equation should be used to calculate
observation ranges:
2 pafq
a4 f 5:8 100:2ar B
4p3 rhW exp 2
c
SN ¼
:
ð3Þ
c3 r4 N2P0 Df
Here, B (the backscattering strength of scatterer(s)) was
shown as TS in Equation (6) of Kang et al. (2002). TS
referred not only to the backscattering strength from a single
fish, but also that from a school. The TS of a school was
defined as the mean TS of a single fish multiplied by the
number of fish in the school. Here, B is used instead of
TS to avoid confusion in terminology.
It is difficult to assume B, because there can be diverse
types and sizes of fish or schools or, indeed, both parameters
together, present when carrying out an acoustic survey.
Table 2. Parameters used for calculating the noise spectrum level.
Frequency (kHz)
Integration coefficient
Range (m)
Layer width of echo
integration (m)
Backscattering strength
of fish (or school) (dB)
Mean volume
backscattering
strength of noise (dB)
Noise spectrum level
(dB re mPa Hz1/2)
Noise spectrum level
extrapolated to 1 Hz
(dB re mPa Hz1/2)
g
r
rw
B
CSVDN
38
1.2
150
20
120
1.4
150
20
20, 30,
40, 50
83.3
20, 30,
40, 50
69.0
NP
68.8
61.0
NP0
151.2
152.4
Moreover, some part of a large school insonified at the
edge of a beam can appear to be a small weak school. Therefore, if a relatively small value is used for B, there will not
be an issue when calculating an observation range. The TS
of the smallest single fish is calculated as 49.2 dB using
the reduced TS (TScm ¼ 66 dB), and the average body
length (6.9 cm) confirmed at trawl station 2. Therefore, we
examine a common observation range for a range of B
from 50 to 20 dB.
Optimal size of an integration cell
Acoustic data from multiple frequencies should physically
and spatially be as similar as possible in order for them
to be compared (Korneliussen and Ona, 2002). In this study,
the acoustic parameters at 38 and 120 kHz are almost the
same except for pulse length. Although the pulse lengths
were different (1 ms at 38 kHz, 0.3 ms at 120 kHz), SV data
have been processed by correcting pulse lengths at different
frequencies and can be used for comparison. The size of
the integration (averaging) cell should be considered as an
essential parameter in comparing multi-frequency data, especially for the purpose of highlighting the characteristics
of a fish school. There has not been any study on cell size
with this in mind. As it is very likely that there would be
a mixture of different body lengths, ages, and species in
a large integration cell (Kang et al., 2002), a small cell would
seem to be preferable. However, if a cell is too small, the
averaging process within the cell does not serve to represent
adequately the mean backscatter on respective frequencies.
Further details of this issue will be explained later, in the
Discussion section.
Consequently, for highlighting school characteristics, an
optimal size of an integration cell is necessary, so that the
dominant age of a school can be ascertained. To select
the optimal size of the integration period, i.e. the horizontal
interval of a cell, it should be, first, small enough to retain
the shape of a school, second, small enough to distinguish
the boundary between the target school and others, and
third, large enough to be able to read the DMVBS pattern,
i.e. a change of DMVBS. The first two points are related to
the exterior characteristics of a school, and the last bears on
its interior characteristics. To help decide the minimum
threshold of horizontal cell size, rough parts of the perimeter of a school can be calculated by distance or by the
number of pings. If integration was conducted on a cell
of larger size than the parts, the boundary of the school
becomes too smooth to retain the shape of a school. Simple
investigation of the horizontal lengths of schools of different species, and the distance between a targeted school
and a school of another species, provides an idea to assist
with understanding the second point. The relationship
between a target school and other species should be clear
after processing integration. Regarding the third point, the
DMVBS characteristics of a school should be readily apparent. If a cell is too small, the range of DMVBS becomes too
Age characteristics of walleye pollock school echoes
large, and if a cell is too large, the range of DMVBS becomes
too small for a specific DMVBS range for a school to be
obtained.
Several integration periods, such as 1, 6, 10, 20, 30, and
50 pings, i.e. 5.2, 31.2, 52, 104, 156, and 260 m, were used
to investigate the optimal size of a cell for school of age0 walleye pollock sampled at trawl station 1.
DMVBS characteristics
The improved DMVBS method attempts to discriminate
walleye pollock schools by age. First, MVBS values from
an optimal size of integration cell are obtained at 38 and
120 kHz. The DMVBS characteristics of walleye pollock
schools are examined up to the common observation range
on the DMVBS echograms. As schools were distributed
very close to the sea floor by day, data suitable for the
DMVBS method were rarely collected. Therefore, only
data collected at night were used.
To understand DMVBS characteristics in a school, we
take the example as follows. Consider the TS directivities
(change of TS according to fish tilt angle) and DMVBS
of two fish which have different body lengths (short and
long) and are swimming at different angles (horizontally
and some other direction) at different frequencies (low
and high). When a fish is swimming horizontally, the maximum TS at two frequencies is the same regardless of body
length, so DMVBS does not appear. However, when a fish
is swimming in another direction, the TS directivity of fish
with the longer body length will change relative to the
small fish. TS directivity is sharp and fluctuating with
body length especially at high frequency (Sawada, 2000).
Therefore, if fish with longer body length are swimming
in many directions, various DMVBS will be observed
easily. If a school comprises fish of the same species and
similar body lengths, DMVBS in a cell can represent
information on the swimming angles of fish. Therefore,
characteristics of DMVBS (or patterns of DMVBS) in
schools provide characteristics of swimming angle within
schools. Even if a part of the DMVBS range overlaps between schools of different ages, the DMVBS patterns within
schools may be different and can provide unique information to identify the internal distribution of the schools at
each age. Hence, it is necessary for the DMVBS echogram
to have the same scales in the horizontal and vertical
dimensions for the purpose of viewing and comparing the
DMVBS ranges and patterns of schools.
Morphological and bathymetric characteristics
Mean height, maximum length, centre depth, seabed depth,
and distance from the seabed of the echoes of walleye
pollock schools were examined to obtain the morphological
and bathymetric characteristics of the schools by age.
Figure 2 shows these school descriptors. Echoview software was utilized to extract the characteristics. First, echoes
1469
Centre depth
Maximum length
X
Mean height
Distance from
the seabed
Seabed depth
Figure 2. Definition of fish school predictors illustrating the
morphological and bathymetric characteristics of walleye pollock
school echoes. The cross shows the centre of the fish school and
is the middle of the mean height and the maximum length.
of a targeted school on the echogram were selected manually. Mean height was calculated for each ping in the enclosed region. The depths from the surface to the centre
of the school at each ping in the region were averaged,
and the average was referred to as the centre depth. Seabed
depth is the average depth from the water surface to the
seabed in the region where walleye pollock schools were
distributed. Distance from the seabed is the distance from
the centre depth to the seabed. Maximum length is calculated by multiplying the average distance between two
continuous pings by the number of pings within the
enclosed region. As the ping distance depended on vessel
speed, several ping distances were selected and averaged
at 6 m.
Results
Composition of age and body length
Figure 3 shows the age composition of the walleye pollock
schools and the ratio of the number at each age to the total
numbers of samples at each trawl station. The data from
trawl stations when fewer than 100 samples were taken
were not included. The data in which a specific age group
is the most prominent were used as target age school. The
coloured rectangles under the trawl number show the
prominent ages that correspond to the colour of the trawl
stations in Figure 1. Age-0 schools comprised almost completely fish of that age, and age-1 schools were 53e99%
1 year olds. From age-2, fish tend to mix with schools
of older age. Age-2 and age-5 schools have been labelled
as such because they comprise 50% or more of that age
group.
Figure 4 shows the composition of the body length of walleye pollock at the trawl stations. The body length of walleye
pollock is divided into five groups according to average body
1470
M. Kang et al.
100%
Age
9
8
Age composition
80%
7
6
60%
5
4
40%
3
2
20%
1
0
0%
1 2
4
5
8 9 12 14 16 17 18 19 21 22 24 25 27 28 29 33
Trawl station
Figure 3. Age composition of walleye pollock schools at each trawl station. Trawl data were used only when the number of walleye pollock sampled was >100. The age of the walleye pollock school is defined by its colour. The numbers and the coloured rectangles at the
abscissa correspond to those of trawl stations in Figure 1.
length, based on grouped trawl stations by age. Figures 3 and
4 show that age composition is strongly related to body
length. In general, walleye pollock schools seem to contain
fish of similar size, but the longer the body length, the more
diversity of body length there tends to be within a school.
Observation range
The noise spectrum levels using the measured noise in
Equation (1) were 68.8 dB re mPa Hz1/2 at 38 kHz, and
61.0 dB re mPa Hz1/2 at 120 kHz, for the calculation of
the observation range.
Figure 5 shows that the observation ranges at 38 and
120 kHz are dependent on the backscattering strength of
the insonified marine organisms, such as 20, 30, 40,
and 50 dB. As the backscattering strength of scatterer(s)
B becomes smaller, the observation ranges at both frequencies and the difference between the observation ranges
become smaller. The observation breadths at two frequencies are similar due to the similar beam widths. When B
is 50 dB, the maximum observable depth is about
169 m at 38 kHz and 120 m at 120 kHz, and the maximum
observable breadth is approximately 17 and 14 m. Clearly,
the observation range is identical up to a depth of 100 m for
38 and 120 kHz, even when B is small. In other words, even
though only a single fish with a body length of approximately 10 cm is insonified at both frequencies, observation
ranges at the two frequencies are the same up to a depth of
100 m. Therefore, the minimum common observation range
is approximately 100 m at both frequencies for the various
sizes of fish schools.
The common observation range at two frequencies
becomes larger as school size increases. For example, B
of the school confirmed age-1 at trawl station 16 results
in 26.6 dB using SV, TS, and reverberation volume. In
Composition by body length
100%
Body
length (cm)
45-61
80%
34-45
60%
23-34
40%
10-23
20%
3-10
0%
1
2
4
5
8
9 12 14 16 17 18 19 21 22 24 25 27 28 29 33
Trawl station
Figure 4. Length composition of walleye pollock schools at the trawl stations.
Age characteristics of walleye pollock school echoes
1471
Finally, an observation range is dependent on the backscattering strength of scatterer(s) at a similar condition of
the acoustic system. However, a common observation range
at multiple frequencies should be decided by considering
the approximate backscattering strength of a target school.
A depth of 100 m is the safest common observation range
regardless of size of fish school and will be used in this
study, though occasionally modified to suit the backscattering strength of a school.
Optimal size of an integration cell
Figure 5. Observation ranges of the echosounder (EK500) operating at 38 and 120 kHz installed onboard RV ‘‘Kaiyo-maru #3’’.
Backscattering strengths (B) are shown, and the minimum common
observation range, shown by a dashed line, extends approximately
to a depth of 100 m regardless of frequency and size of fish.
Figure 5, when B is 20 dB, the observable depth shown as
an overlapped maximum depth between two frequencies is
about 300 m, and when B is 30 dB, the depth is about
220 m. As the B of the school is about 25 dB, a common
observation range can be used at 200 m. Therefore, for the
age-1 school at trawl station 16, the DMVBS method can be
used up to a depth of 200 m.
To find the optimal size of an integration cell, a change in
DMVBS (Figure 6) compares the characteristics of the
school with various integration periods in ping number or
distance, or both parameters. In the echogram with a oneping integration period, the DMVBS range is approximately 3 to 14 dB, with noticeable patches from 10 to
14 dB (Figure 6a). When the number of pings in one integration period increases from 6 to 20 (31.2e104 m) the
range of DMVBS decreases to 3 to 6 dB (Figure 6bed).
The blue (13 to 10 dB of DMVBS) and orange
(10e12 dB of DMVBS) colours in the upper part of the
school may indicate different species, perhaps planktonic
organisms, because plankton has a variety of frequency
characteristics according to the physiology of its constituent
species, and it often comprises different species, even if
their taxon cannot be confirmed. The state of a mixture of
Figure 6. Changes in DMVBS by integration period, used to find the optimum size of an integration cell for the purpose of highlighting
target school characteristics. The age-0 school was used at trawl station 1 in daylight. The vertical extent of the integration cell is uniformly 1 m. The vertical axis reaches 130 m deep, and the horizontal axis represents the integration period. Only the one-ping echogram
(a) indicates the distance in kilometres on the horizontal axis. One ping corresponds to approximately 5.2 m horizontally. The school
boundary, with different species and school shape, and the distribution pattern of DMVBS must be considered when determining the
optimum size of an integration cell.
1472
M. Kang et al.
multiple species in the case of >30 pings is difficult to observe (Figure 6e, f). When the integration period is as large
as 50 pings (260.4 m), the edge of the school is not clear,
and its inner portion has a very narrow range of DMVBS,
approximately 1 to 1 dB. When a cell is too large, all
the DMVBS of a school become nearly one value. There
is no value in averaging echoes in such a large integration
cell to describe the frequency characteristics of the school
(Figure 6f). Therefore, the optimal integration period for
the walleye pollock school should be set at 6e20 pings
(31e104 m).
We found the optimal integration period to be about 20
pings (124 m calculated at 6.2 m of ping distance between
two pings) for the school of age-5 fish at trawl station 33
during daylight. In this study, therefore, an integration
period of 20 pings (120 m) is used as the optimum size
of an integration cell for walleye pollock schools.
Morphological and bathymetric characteristics
To illustrate the distribution characteristics of walleye
pollock schools based on age, a typical example of
a night-time echogram for each school by age is shown in
Figure 7aed, and for daylight in Figure 7eeh. Daylight
echograms were obtained at the same locations as those
recorded at night.
In the night-time echogram the school of age-0 is widely
distributed in depth from the surface down to approximately 100 m, mainly at a distance from the seabed, but
parts do reach the seabed. The age-1 school is distributed
through a height of some 40 m at a depth of 180e190 m
over the continental shelf. The age-2 school has a height
of 40e60 m, is distributed approximately 25 m above the
seabed, and is located at the edge of the continental shelf.
The age-5 school starts at a depth of approximately
140 m on the upper continental slope, and has a height of
40e120 m.
In the daylight echogram, the age-0 school is distributed
more densely in the centre of the school than is observed at
night. In general although the sizes of age-0 schools vary
greatly by day, they have a tendency to cluster, which
makes them easily distinguishable from other age groups.
The age-1 schools, indicated by a brown arrow in Figure 7f,
are very small, and are located near the sea floor. The age-2
school observed by day has different distribution characteristics from those observed at night: it is largely scattered
and forms only a few small clusters. The age-5 school is
distributed close to the seabed, so it is hard to discriminate
between seabed and fish.
The number of schools analysed during this study is
shown in Table 3. In all, 23 small schools of age-0 were
observed by day, but just two schools of the same age at
night. Most age-1 schools could barely be distinguished
from seabed echoes by day (e.g. Figure 7f), so these
daytime schools have not been included for further analysis
(Table 3, Figure 8).
In order to understand the general distribution characteristics of walleye pollock schools based on this survey,
school descriptors by age are illustrated in Figure 8. This
shows that age-0 fish are small and live in shallow water,
so it is easy to distinguish their schools from schools of
different age. The age-1 schools are distributed in relatively
deep water at about 180 m, and have little height but great
length, e.g. 7770 m. When comparing the age-2 schools
with those of age-5, their distribution characteristics are
very similar, especially their maximum length, the centre
depth, and the seabed depth, even though the schools of
age-5 fish show a greater change of mean height between
day and night and are distributed near the seabed by day.
DMVBS characteristics
Figure 9 presents the DMVBS echograms of walleye pollock schools according to age. Each echogram has a vertical
axis of 300-m depth and a horizontal axis of 6000 m. The
optimal cell size of 20 pings is used as the integration
period. The safest common observation depth of 100 m
(shown as a white dotted line) is necessary when observing
the DMVBS echograms, regardless of the size of the fish
school. The common observation depth expands to some
200 m in Figure 9b because of the strong backscatter
(26.6 dB). The age-0 school has a narrow range (3 to
2 dB) of DMVBS (Figure 9a), which means similar scattering strengths at 38 and 120 kHz. The age-1 school has
DMVBS distributed from 1 to 8 dB (Figure 9b). Most
DMVBS is between 3 and 5 dB (yellow) and 0 dB (light
green spots). The school of age-2 fish has DMVBS between
3 and 8 dB (Figure 9c). A mixture of DMVBS is shown
mainly between 3 and 4 dB (yellow) throughout the school,
and some DMVBS of 3 dB (green) at the upper edge and
at the middle of the school. The school of age-5 fish has
a wide range of DMVBS, from 0 to 12 dB (Figure 9d). A
section of DMVBS between 6 and 12 dB in orange is
shown at the centre of the school.
Not all school groups can be easily identified using the
range of DMVBS, because parts overlap. However, the pattern of DMVBS, shown as a range of colours, is distinctive
between age groups. When looking at all echograms, it is
very easy to see that the school of each age has its own particular characteristic in terms of a pattern of DMVBS. The
pattern of DMVBS is crucial information in identifying
the age of schools, and can be assumed to be a reflection
of the character of the movements of the fish.
A great variation in DMVBS values in schools of older
age might be due to distributed tilt angles. When fish in
an integration cell are distributed randomly, fish tilt angles
(or TS directivity) are averaged in the cell. Hence, there is
little effect of fish tilt angle on the MVBS of the cell. If fish
in a school uniformly face a certain direction in an integration cell, as can be supposed for older walleye pollock
schools, TS will not be appropriately averaged across
a large and random distribution of tilt angles. A specific
Age characteristics of walleye pollock school echoes
0.5 nautical miles
1473
Night
20 m
100 m
200 m
(a) age-0 (#2)
(c) age-2 (#19)
100 m
200 m
(b) age-1 (#16)
(d) age-5 (#28)
Day
20 m
100 m
200 m
(g) age-2 (#19)
(e) age-0 (#2)
100 m
200 m
(f) age-1 (#16)
(h) age-5 (#28)
Figure 7. A set of examples of echograms by age at night (aed) and by day (eeh), to display the distribution characteristics of walleye
pollock schools. Trawl station numbers are in parenthesis under the echograms. Note that horizontal scales on the echograms differ. The
brown arrow indicates the age-1 school.
tilt angle affects MVBS differently at different frequencies,
and causes fluctuating DMVBS. Therefore, if there are
several subgroups of fish with various orientations in
a school, a diversity of DMVBS can be seen. If a school
of a certain age has a particular swimming direction, then
that characteristic appears in the pattern of DMVBS. It
can be concluded that DMVBS indicates the internal
distribution structure of a school.
Discussion
Observation range
When the acoustic signals of a school or single fish are not
more than a given threshold, the weak echoes do not
contribute to the received signals, and so the sampled
volume effectively becomes smaller. To calculate this volume an equivalent beam angle is often used, and the term
0
1
2
5
Day
Night
23
0
5
5
2
3
5
5
given to the end result is often called the acoustic sample
volume (Aglen, 1982; Foote, 1991; Reynisson, 1996).
The tilt angle of fish is known to be a more important
factor than transducer directivity and threshold in this situation (Foote, 1991; Reynisson, 1996). However, this
method did not consider the acoustic system and the noise
of the vessel. The ‘‘observation range’’, which gives the
detectable range of an echosounder, does consider both,
as well as the backscattering strength of a school, and so
provides a better measure of the various underwater
conditions.
Diverse DMVBS
A diverse range of DMVBS was apparent when the integration cell was very small, for example with an integration period of one ping. If a cell is too small, no averaging in the
cell can be carried out effectively. Two possible reasons
can be assumed. The first may be frequency difference in
TS directivity. It is well known that TS directivity is narrower and more complicated with a long body length (i.e.
body length is large compared with wavelength) (Sawada,
2000). If a cell is too small to have a large number of fish,
the tilt angles of the fish cannot be considered to be
effectively random. Hence, the effect of TS directivities is
different at different frequencies, which results in a diverse
DMVBS. The second reason may be the frequency difference caused by the interference of multiple echoes.
Multiple echoes are generated when single echoes with
a carrier signal overlap. The amplitude of multiple echoes
comprises the basic component and the interference.
When the number of echoes is small, variations of both components are large. Moreover, in some cases the interference
becomes complicated, and may occur at two frequencies
(Furusawa, 1991). If the data set is large, a variation of
the basic component becomes even, and interference can
be considered to be uniformly distributed and ignored. If
many echoes in an integration cell are averaged, the frequency difference attributable to interferences will become
smaller. Therefore, the size of the cell should be large
enough for it not to be affected by interference and TS
directivity.
20
0
Max. length (m)
Age
40
10000
Centre depth (m)
Number of schools
200
Seabed depth (m)
Table 3. Number of schools used for examining school
characteristics.
Mean height (m)
M. Kang et al.
Distance from
the seabed (m)
1474
8000
6000
4000
2000
0
150
100
50
0
200
150
100
50
0
60
40
20
0
0
1
2
5
Age group
Figure 8. Morphological and bathymetric characteristics of walleye
pollock schools by age. White bars (s.d.) refer to schools identified by day and grey bars (s.d.) to schools identified at night.
TS characteristics by age
TS characteristics by age were investigated using single targets detected at the boundaries of schools recorded at trawl
stations. Only the average TS, such as approximately
46 dB, the assumed age-0, is clearly distinguishable
from that of other groups, and the body length deduced
from this average TS and that from trawls are quite similar.
However, in groups older than age-1, average TS can be
similar while the body lengths from each average TS do
not correspond with that of specimens found in the trawls.
Therefore, it is difficult to discriminate the age of walleye
pollock among older age groups using TS. The study of
Sadayasu (2005) on the TS of walleye pollock used
measurements of suspension and free swimming, and
a Kirchhoff ray-mode model. The relationship between
TS and body length differed beyond a body length of
Age characteristics of walleye pollock school echoes
1475
Figure 9. DMVBS echograms of walleye pollock schools by age. The age of fish in a school, the number of the trawl station, and the depth
of the trawl are shown. The colour scale corresponds to DMVBS in decibels. Pink arrows indicate the target age school. One integration
period corresponds to approximately 120 m (optimal size of a cell) on the horizontal axis. Each echogram has the same scale of axes,
a transit distance of 6000 m (50 integration periods) horizontally, and a depth of 300 m vertically. The white dotted line shows the depth
(100 m) of the common observation range, regardless of the size of a school and its frequency. However, for an age-1 school (b), the
common observation range extends to a depth of 200 m owing to the backscattering strength of the school.
10 cm. This can be explained by the fact that the relative
growth ratio of the swimbladder (strongly affecting TS)
changed greatly at that body length. Therefore, to demonstrate satisfactorily whether age can be discriminated by
TS characteristics, data from in situ TS measurement of
targets close to a transducer may be required, and further
study of fish behaviour will be needed.
It is imperative when using the DMVBS method that
DMVBS be measured with negligible noise. The ICES
report on the underwater noise of research vessels (ICES,
1995) suggested that noise spectrum levels of a research
vessel should be 47.6 and 42.1 dB at 38 and 120 kHz,
respectively. As the vessel used in this study was not
designed for acoustic research, its noise spectrum was
higher than that of the ICES recommendation. However,
the noise of the vessel was measured accurately and used
to calculate the observation ranges.
Many morphological, bathymetric, and energetic characteristics are used in fish species identification (Weill et al.,
1993; Lu and Lee, 1995; Scalabrin et al., 1996; Coetzee,
2000; Lawson et al., 2001). Of the three, morphological
characteristics tend to provide the best predictor of species.
Even if descriptors of height, area, length, and perimeter of
school have proven more effective, the contributions of
those parameters varied slightly between studies. It is
important to use many school descriptors, however, in order
to ascertain those that are most effective. One study used
distribution characteristics along with DMVBS between
three frequencies to identify Antarctic krill from other
plankton (Woodd-Walker et al., 2002). If a pattern of
DMVBS among frequencies in that study had been examined, the results may have been interesting.
It is well known that age-0 and age-1 schools contain
pre-recruitment walleye pollock, and these age groups
should be managed for sustainability (Honda, 2004). The
results presented here are therefore relevant to fisheries
resource management, because they can be used to discriminate age-0 and age-1 schools from other older age groups
in order to estimate the biomass of younger age groups
individually. Also, for commercial fisheries, the method
can be applied to prevent younger fish from being inadvertently caught.
Acknowledgements
The study was supported by grants from the Fisheries
Agency of Japan. We thank cruise members of RV
‘‘Kaiyo-maru #3’’ at Nippon Kaiyo for assistance, and
two anonymous reviewers for pertinent comments on an
1476
M. Kang et al.
earlier draft. We also thank Matthew Wilson of SonarData
for smoothing the English in this paper.
References
Aglen, A. 1982. Impact of fish distribution and species composition
on the relationship between acoustic and swept-area estimates of
fish density. ICES Journal of Marine Science, 53: 501e505.
Coetzee, J. 2000. Use of a shoal analysis and patch estimation system (SHAPES) to characterize sardine schools. Aquatic Living
Resources, 13: 1e10.
Conti, S. G., and Demer, D. A. 2002. Wide-bandwidth acoustical
characterization of anchovy and sardine from reverberation measurements in an echoic tank. ICES Journal of Marine Science,
60: 617e624.
Cooke, K., Kieser, R., and Stanley, R. D. 2002. Acoustic observation and assessment of fish in high relief habitats. ICES Journal
of Marine Science, 60: 658e661.
Foote, K. G. 1991. Acoustic sampling volume. Journal of the
Acoustical Society of America, 90: 959e963.
Furusawa, M. 1991. Designing quantitative echosounders. Journal
of the Acoustical Society of America, 90: 26e36.
Furusawa, M., Asami, T., and Hamada, E. 1999. Detection range of
echosounders. In Proceedings of the Third JSPS International
Seminar ‘‘Sustainable Fishing Technology in Asia towards the
21st Century’’, pp. 207e213. Ed. by T. Arimoto, and J. Haluan.
Tokyo University of Fisheries International JSPS project volume 8.
Higginbottom, I. R., Pauly, T. J., and Heatly, D. C. 2000. Virtual
echograms for visualization and post-processing of multiplefrequency echosounder data. In Proceedings of the Fifth European Conference on Underwater Acoustics, ECUA 2000, pp.
1497e1502. Ed. by M. E. Zhakharia, P. Chevret, and P. Dubail.
Office for Official Publications of the European Communities.
Honda, S. 2004. Abundance estimation of the young cohorts of the
Japanese Pacific population of walleye pollock (Theragra chalcogramma) by acoustic surveys. Bulletin of the Fisheries Research Agency (Japan), 12: 25e126 (in Japanese).
Horne, J. K. 2000. Acoustic approaches to remote species identification: a review. Fisheries Oceanography, 9: 356e371.
ICES. 1995. Underwater noise of research vessels: review and recommendations. ICES Cooperative Research Report, 209.
ICES. 2000. Report on echo trace classification. ICES Cooperative
Research Report, 238.
Kang, M., Furusawa, M., and Miyashita, K. 2002. Effective and
accurate use of difference in mean volume backscattering
strength to identify fish and plankton. ICES Journal of Marine
Science, 59: 794e804.
Korneliussen, R. J., and Ona, E. 2002. Synthetic echograms generated from the relative frequency response. ICES Journal of
Marine Science, 60: 636e640.
Lawson, G. L., Barange, M., and Fréon, P. 2001. Species identification of pelagic fish schools on the South African continental
shelf using acoustic descriptors and ancillary information.
ICES Journal of Marine Science, 58: 275e287.
LeFeuvre, P., Rose, G. A., Gosine, R., Hale, R., Pearson, W., and
Khan, R. 2000. Acoustic species identification in the Northwest
Atlantic using digital image processing. Fisheries Research, 47:
137e147.
Lu, H. J., and Lee, K. T. 1995. Species identification of fish shoals
from echograms by an echo-signal image processing system.
Fisheries Research, 24: 99e111.
MacLennan, D. N., and Holliday, D. V. 1996. Fisheries and plankton acoustics: past, present, and future. ICES Journal of Marine
Science, 53: 513e516.
Masse, J., Koutsikopoulos, C., and Patty, W. 1996. The structure
and spatial distribution of pelagic fish schools in multispecies
clusters: an acoustic study. ICES Journal of Marine Science,
53: 155e160.
Misund, O. A. 1993. Dynamics of moving masses: variability in
packing density, shape, and size among herring, sprat, and saithe
schools. ICES Journal of Marine Science, 50: 145e160.
Nishimura M. 1969. Study on the optimum frequency of
echosounders. DSc thesis, Tohoku University, Japan (in
Japanese).
Reynisson, P. 1996. Evaluation of threshold-induced bias in the
integration of single-fish echoes. ICES Journal of Marine Science, 53: 345e350.
Richards, L. J., Kieser, R., Mulligan, T. J., and Candy, J. R. 1991.
Classification of fish assemblages based on echo integration
surveys. Canadian Journal of Fisheries and Aquatic Sciences,
48: 1264e1272.
Rose, G. A., and Leggett, W. C. 1988. Hydroacoustic signal
classification of fish schools by species. Canadian Journal of
Fisheries and Aquatic Sciences, 45: 597e604.
Sadayasu, K. 2005. Study on accurate estimation of target strength
of fish. DSc thesis, Hokkaido University, Japan (in Japanese).
Sawada, K. 2000. Study on the precise estimation of the target
strength of fish. DSc thesis, Tokyo University of Fisheries, Japan
(in Japanese).
Scalabrin, C., Diner, N., Weill, A., Hillion, A., and Mouchot, M-C.
1996. Narrowband acoustic identification of monospecific fish
shoals. ICES Journal of Marine Science, 53: 181e188.
Simmonds, E. J., Armstrong, F., and Copland, P. J. 1996. Species
identification using wideband backscatter with neural network
and discriminant analysis. ICES Journal of Marine Science,
53: 189e195.
Takao, Y., and Furusawa, M. 1995. Noise measurement by echo
integrator. Fisheries Science, 61: 637e640.
Weill, A., Scalabrin, C., and Diner, N. 1993. MOVIES-B:
an acoustic detection description software. Application to
shoal species’ classification. Aquatic Living Resources, 6:
255e267.
Woodd-Walker, R. S., Watkins, J. L., and Brierley, A. S. 2002.
Identification of Southern Ocean acoustic targets using aggregation backscatter and shape characteristics. ICES Journal of
Marine Science, 60: 641e649.
Zakharia, M. E., Magand, F., Hetroit, F., and Diner, N. 1996.
Wideband sounder for fish species identification at sea. ICES
Journal of Marine Science, 53: 203e208.