pro-O R I G I N A L A R T I C L E FisheriesComparisons of monthly and geographical variations in abundance and size composition of Pacific saury between the high-seas and coastal fishing
Trang 2O R I G I N A L A R T I C L E Fisheries
Classification of fish schools based on evaluation
of acoustic descriptor characteristics
Aymen Charef•Seiji Ohshimo •Ichiro Aoki•
Natheer Al Absi
Received: 27 May 2009 / Accepted: 15 October 2009 / Published online: 8 December 2009
Ó The Japanese Society of Fisheries Science 2009
Abstract Acoustic surveys were conducted from 2002 to
2006 in the East China Sea off the Japanese coast in order
to develop a quantitative classification typology of a
pelagic fish community and other co-occurring fishes based
on acoustic descriptors Acoustic data were postprocessed
to detect and extract fish aggregations from echograms
Based on the expert visual examination of the echograms,
detected schools were divided into three broad fish groups
according to their schooling characteristics and ethological
properties Each fish school was described by a set of
associated descriptors in order to objectively allocate each
echo trace to its fish group Two methods of supervised
classification were employed, the discriminant function
analysis (DFA) and the artificial neural network technique
(ANN) We evaluated and compared the performance of
both methods, which showed encouraging and about
equally highly correct classification rates (ANN 87.6%;
DFA 85.1%) In both techniques, positional and then
morphological parameters were most important in
dis-criminating among fish schools Fish catch composition
from midwater trawling validated the fish group
classifi-cation through one representative example of each
group-ing Both methods provided the essential information
required for assessing fish stocks Similar techniques of fishclassification might be applicable to marine ecosystemswith high pelagic fish diversity
Keywords Acoustic descriptor Artificial neuralnetwork Discriminant function analysis Fishclassification Species identification
IntroductionThe northern part of the East China Sea represents one ofthe main spawning and nursery areas of small pelagicfishes in the waters off of the Japanese coast It also con-stitutes an important fisheries ground for commerciallyvaluable pelagic fishes During the last half decade, theaverage landing was estimated to be roughly 250,000 tonsper year and was composed of Japanese anchovy Engraulisjaponicus, round herring Etrumeus teres, jack mackerelTrachurus japonicus, chub mackerel Scomber japonicusand spotted chub mackerel Scomber australasicus(according to statistics from the Ministry of Agriculture,Forestry and Fisheries, Government of Japan) The fishstock size assessment is crucial for fisheries management inthese waters Broadly, the main assessment techniques arebased on the virtual population analysis (VPA) method.This method makes use of commercial catches, whichmight bias the assessments and then generate very seriousoverfishing problems [1,2] To eliminate such complica-tions, reliable and fishery-independent data are needed.Hydroacoustic methods are one of the few techniquesused in order to provide fisheries independent quantitativeestimates of fish stocks Fisheries acoustics have experi-enced dramatic development in technologies and datamanagement Acoustic surveys using quantitative scientific
A Charef (&) I Aoki
Graduate School of Agriculture and Life Science,
University of Tokyo, Bunkyo, Tokyo 113-8657, Japan
e-mail: aymen_charef@yahoo.com
S Ohshimo
Seikai National Fisheries Research Institute,
Fisheries Research Agency, Nagasaki 851-2213, Japan
N Al Absi
Ocean Research Institute, University of Tokyo,
Nakano, Tokyo 164-8639, Japan
DOI 10.1007/s12562-009-0186-x
Trang 3echo sounders commonly employed to determine the
abundance and biomass of pelagic fish are becoming
increasingly important for the management of pelagic
fisheries [3] Owing to the common aggregative behavior,
small pelagic species appear in echograms as a mixture of
diverse fish assemblages [4] Echo integration is used to
estimate fish quantity since the sampled volume contains
overlapping target fish echoes [3] The obtained target
strengths and the backscattering strength can be translated
into biomass units if the proportions of different species
and their length distribution and target strength on fish size
are known In such a context, distinguishing among fish
targets is greatly needed to deal with each target fish echo
separately Therefore, identification of echo traces of fish
schools is crucial in conjunction with accurate acoustic
surveys to give reliable estimates of target strength and
consequently improve the fish stock assessment
The classification and subsequent identification of
acoustic targets to taxa or species are still the great
chal-lenge of fisheries acoustics [5,6] Species identification has
been limited by the difficulty in objectively classifying
backscattered energy of echo traces to species [6,7]
Echo-trace classification defined as the detection and description
of aggregations in acoustic data can be used to study
behavioral and ecological processes in aquatic
environ-ments [8] It is generally agreed that besides integration of
target species’ biomass, useful information, such as
fea-tures from digitized echograms, can be extracted from the
acoustic data Many studies have attempted to develop
echo-trace classification in order to study shoaling behavior
and predator-prey interactions, to characterize fish
aggre-gations, their spatial distribution and their relationship to
environmental variables; see Horne [9] for a review
First attempts at fish identification introduced basically
subjective and time-consuming methods These methods
involved expert scrutiny of echograms combined with
concurrent trawling data Visual scrutiny of acoustic data
depends on human experience and is therefore subject to
biases and difficult to be quantified This makes objective
methods more efficient, timely, less or not dependent on
subjective interpretation, and controlled by evaluating their
accuracy [10] These automated methods require data
processing and detection of acoustic features from
echo-grams as a first step, and secondly, description of selected
schools characteristics with a set of descriptors [11] They
aim to train an algorithm on a set of identified, single
species schools Then the algorithm is adopted to identify
other schools [12,13] Success of objective methods relies
primarily on a suitable choice of acoustic descriptors
concerning number and efficiency In the case of high
diversity ecosystems, such as the East China Sea, where
small schools are numerous, species classification highly
depends on verification via trawl data
In the East China Sea, some attempts to estimate thepelagic fish populations’ biomass were made with acousticsurveys These studies were restricted to subjective clas-sification of fish species [14, 15], while limited to singlespecies such as anchovy [14, 16] and sardine [17, 18] inother works In this work, we applied two objective tools ofsupervised echo-trace classification, discriminant functionanalysis (DFA) and artificial neural network (ANN) Theaim of this paper is to describe and to evaluate the efficacy
of the two methods, based on a set of acoustic descriptors,
in objectively classifying fish schools of pelagic fishcommunity and other co-occurring fishes such as pearlsideand lantern fish
Materials and methodsData collection
Acoustic surveys were conducted annually in the latesummer from 2002 to 2006 by the Japanese FisheriesResearch Agency on board the RV Yoko Maru Surveyswere carried out along 27 parallel transects spaced by 10nautical miles (Fig.1) During surveys, vessel speed wasapproximately 10 knots and total length of transects rangedfrom 593 to 828 nautical miles (Table1)
Fig 1 Study area and acoustic survey scheme
Trang 4Acoustic data were collected using a calibrated
hull-mounted SIMRAD EK505 scientific echo-sounder system
operating at 38 kHz with a time-varied gain function set
at 20 log R The echo-sounder pulse length was 1 ms, its
ping rate was 0.33 ping s-1, and its estimated sound
speed 1500 ms-1, giving a target resolution of 0.001 s 9
1500 m s-1/2 = 0.75 m Acoustic measurements were
logged continuously during all surveys and recorded only
during daytime
Small pelagic fish species may reduce the risk of daytime
predation by schooling [19] The schooling behavior
typi-cally characterizes each fish school in daylight, which is
essential for the fish identification However, during twilight
and nighttime, fish schools scatter and overlap, which biases
the fish identification in acoustic processing [4,20]
Acoustic data processing
Acoustic data were postprocessed using Echoview
Soft-ware version 4.50 [21] The seafloor was automatically
detected using the ‘‘maximum Sv backstep’’ algorithm,
where the backstep was set at 1 m Data deeper than 1 m
above the selected bottom line were removed due to the
false bottom detection Data shallower than 10 m were also
removed from analyses to eliminate the transmit pulse and
reduce backscatter by surface bubbles
A background threshold of -67 dB was applied
equiv-alently to all echograms The threshold was determined by
analyzing a subset of data collected from each year and
allowed accurate detection of all possible aggregations of
target fishes Fish aggregations were detected and
charac-terized using the ‘‘Schools detection’’ module implemented
in Echoview Input parameters were set according to
schools’ features observed in acoustic records The
algo-rithm pattern required schools to be at least 8 m long and
4 m high Adjacent aggregations were linked to shape one
school if the maximum horizontal linking distance was
15 m and maximum vertical connection distance 5 m
Then echograms were visually inspected, and doubtful and
‘false’ detections (scattering layer, acoustic interference)
were removed Connected aggregations with dimensions
smaller than the minimum school length and height
parameters were discarded
For each detected acoustic target, a set of five schooldescriptors was calculated and extracted, and they fell intothree categories (Table2): (1) morphological: schoollength, height and height mean; (2) energetic: mean vol-ume backscattering strength (Sv); (3) positional: meanschool altitude (Depth)
Midwater trawl catch dataMidwater trawling was used to identify acoustic targets and
to establish their weight composition Midwater trawlingwas only performed at nighttime because of the high net-avoidance rate of fish targets in the daytime, which makes
it difficult to sample the observed fish schools in acousticrecordings [22] Visual inspection of echograms for severalhours permitted the characterization of schooling behaviorand swimming depth of target species The position of the
Table 2 Definitions and units of school descriptors used in both analysis methods
Morphological
along the transect from the first to last ping crossing the school
separating the maximum and minimum depths of the rectangle bounding the school
upper to lower limit along each ping crossing the fish school
Energetic Mean volume backscattering strength (Sv)
dB The mean energy produced
by pixels shaping a fish school, which indicates its mean density
Positional Mean school depth (Depth)
m The distance from the sea
surface to the geometric center of the fish school
Table 1 Year, beginning and
end dates, total length of
transects, number of detected
schools and number of stations
of CTD casts and midwater
trawls during each acoustic
survey
Year Begin date End date Total length of transects
(nautical miles)
Number of detected schools
Trang 5trawl stations was decided beforehand according to the
location of peculiar fish concentrations detected during
acoustic surveys in the daytime
A total of 88 midwater trawls were conducted (Table1)
Towing speed was approximately 3 knots for a towing time
of 30 min Towing depth was targeted to fish schools by
adjusting the towing speed and warp length The mouth of
the trawl net was approximately 20 m by 20 m, and the
mesh sizes of the cod end and the inner bag were 60 and
20 mm, respectively The trawl catch was separated by
species, and the total weight of each species was
determined
Other data
Conductivity-temperature-depth (CTD) profiles were taken
along the survey tracks at the beginning of each trawling
operation The on-board data recording and entry system
was deployed to record series of time (GMT), geographic
position and the EK 500 vessel log
Fish-group classification
School images were selected and allocated to a species
through visual expert examination of the echogram
dis-plays based on prior experience knowledge, in conjunction
with the interpretation of echograms The identified target
fishes were classified into three types of fish groups
according to their schooling characteristics and ethological
properties The verification of this typology also involved
the results of the midwater trawl catch amount and
composition
The classification was partially based on the previous
findings of Ohshimo [15] from acoustics surveys conducted
following a similar survey scheme on the same study area
The first type (G1) consisted of compactly aggregated
schools, assumed to be Japanese anchovy and round
her-ring, within the upper layer of the water column The
second group (G2) appeared in the midwater layers, mostlyabove the bottom rise structure, and it was thought to becomposed of jack mackerel and chub mackerel The lastgroup (G3), assumed to consist of lantern fish and pearl-side, occurred in demersal layers mainly along slopes andformed horizontally elongated schools in contact with theseabed (Fig 2) Some detected fish schools that did not fallwithin this typology were neglected
Statistical analysisDiscriminant function analysis (DFA) is a well-knownstatistical procedure used to predict group membershipbased on a combination of the interval variable [23] Thefive school descriptors constituted the predictor variablesfor this discrimination analysis, whereas the dependentvariable was fish group (G1, G2, G3) defined a priori on thebasis of visual expert scrutiny and direct sampling results.DFA was performed using SPSS (version 6.0) based onMahalanobis distances (D) Mahalanobis distance is thedistance between a case and the centroid for each fishgroup (of the dependent variable) in attribute space By thisprocedure, each school is allocated to the fish group forwhich D has the smallest value [24] Classification accu-racy was estimated with leave-one-out cross-validation, inwhich the discriminant function is first derived from only
n - 1 schools and then used to classify the other schoolleft out The procedure is repeated n times, each timeomitting a different observation [25] DFA was applied foroverall years data pooled together
Artificial neural networksArtificial neural networks (ANN) were also used as themethod of species classification and identification of fishschools from acoustic data They imitate human neuronfunctioning and solve problems by applying knowledgegained from past experience to new situations [26]
Fig 2 Acoustic recordings showing typical schools of three different fish groups
Trang 6A multiple layer perceptrons (MLPs) neural network
was constructed and computed using Matlab 6.0 MLPs are
the most commonly and the simplest network type used,
primarily due to their speed and versatility [27] They
consist of three feed-forward layers: input, hidden and
output (Fig.3) The input layer was composed of five
variables The number of nodes in the hidden layer was
determined by testing the performance of the model using a
range of node numbers The dependent variable fish groups
represented the output layer The data set was split into a
training set and validation set consisting of 70 and 30% of
the identified schools, respectively, with the same
propor-tion of each fish group Based on supervised learning, the
neural network was trained by means of a backpropagation
learning algorithm (BP) in order to develop the ability to
correctly classify new fish schools from further acoustic
data [28] The school fish’s classifications based on their
relative descriptors occurred in two major phases First,
during the learning phase, internal parameters within the
network were adjusted iteratively The performance of the
network, equivalent to classifying schools into fish groups
accurately, was maximized; this stage continued until there
was no further increase in network performance or
classi-fication success Although the aim of the training is to
reduce the error as much as possible, reducing the error too
much leads to the network learning the noise rather than
underlying relationships Precautions were taken to avoid
over-fitting (over-training) of the network’s model Finally,
during the validation phase, which is the second phase, the
optimal network was applied to test sets, along with
cross-validation
ResultsClassification using discriminant function analysisDiscriminant function analysis was computed using 830detected schools and five acoustic descriptors (Tables1,3).Since the dependent variable, fish school, has three groups,two canonical discriminant functions were determined.Both functions were significant, but nearly all of the vari-ance in the model is captured by the first discriminantfunction The small Wilk’s lambda coefficients indicatedalso that only the first function is useful The eigenvaluesconfirmed the significant difference between both dis-criminant functions The standardized discriminant func-tion coefficients were used to compare descriptorsmeasured on different scales Coefficients with largeabsolute value correspond to variables with greater dis-criminating ability This implies that within the first func-tion, for instance, depth contributed the most Thus,descriptors in rank order of efficacy in discriminating fishschools are depth, height, height mean and length, whilemean volume backscattering strength Sv comes last.The confusion matrix showed the results of the DFAusing five acoustic descriptors for discriminating fishschools from survey data of 5 years (Table 4) Emboldenedvalues on the main diagonal of each confusion matrixrepresent the number of schools that were correctly iden-tified within every fish group The overall correct classifi-cation was evaluated at 85.1%
The correct recognition rates per group showed highscores for G1 schools Almost 95% were well assigned anddistinguished from other groups G2 schools represent 57%
of the total number of schools and were the least correctlyclassified with a relatively low rate of 80.3% The pro-portion of G3 schools is small, with only 13.25%, and had acorrect classification score of 81.8%
Classification using an artificial neural networkApplication of the trained network to 5 years of pooledacoustic data resulted in predicted species compositionsthat corresponded well to those observed with an overallcorrect classification evaluated at 87.6% for the validationdata set (Table5) The model performed well for
Fig 3 Network architecture for the model used in this study
Table 3 Results of discriminant analysis using five descriptors for overall 5 years data
Trang 7classifying G1 schools with a correct classification rate of
95.94%, but less for G3 and G2 schools, with 84.84 and
83.91%, respectively
The contribution factor of a variable is the sum of the
absolute values of the weights generated from this
partic-ular variable It reveals the importance of input variables,
descriptors, to classify fish schools The analysis showed
similar ordering of descriptor categories to DFA results and
indicated that the heaviest impact in classifying was
assigned to positional, morphological and then energetic
properties of a school However, the ascending order within
the morphological descriptors category differs slightly,
though depth was the most efficient descriptor (Fig.4)
Validation with catch data
Midwater trawling catch assisted in fish identification
simultaneously with visual scrutiny of echograms The
recorded acoustic data in daytime permitted to observe
typical shapes of fish schools and then facilitated their
identification Examination of catch data over all 88 towsshowed that the dominant target species was jack mackerel,which contributed 22% by weight of the total catch, fol-lowed by Japanese anchovy (18.4%) and lantern fishes(16%) (Table 6) Round herring was an exception in 2002and was the most abundant species, reaching 20% of thetotal catch by weight in the same year The non-targetspecies that did not fall in the three identified categories ofmajor species were clustered into one group as ‘‘others’’and represented around 28% of the total catch amount(Table6) Catch composition was also valuable to verifythe classification of target species into three groups of fishschools Table7 shows catch composition data fromselected trawls hauled near the locations where schools ofG1, G2 or G3 were observed in daytime Each group ofspecies was assigned according to the most dominantspecies comprised in each trawl catch
A summary of trawl hauls with fish schools matchingwith acoustically detected schools is shown in Table8.Looking at both tables simultaneously (Tables7, 8) per-mitted examining the catch composition according to theamount and number of trawl hauls In the overall data for
5 years, the number of detected schools evenly matchedwith the catch amount of target species The correspon-dence between detected and caught G1 schools was esti-mated to be 44% of the catch from 11 hauls, mainly made
up of Japanese anchovy as it is the most abundant species
in G1 The mismatch is primarily due to the high amount ofcatch of the G2 and G3 species Around 34% of thedetected G2 schools were validated by catch data from 11hauls Other co-occurring species, mainly represented bypuffer fishes and squid, made up 43% of the total catchamount and were fairly abundant in 15 hauls; some of themwere small catches (less than 2 kg) In the case of G3schools, nearly 41% of identified schools were validated bycatch results Bycatch species that were caught during thesame trawl hauls represented 33% of the total catch butbelonged to one trawl haul
Table 4 Confusion matrix of DFA analysis
Number of schools from each group (true classification) distributed
over predicted groups Values in bold denote correctly classified
schools
Table 5 Results of ANN classification from the two data sets
Values in bold represent correct assignment
Fig 4 Proportion of the contribution factor of each descriptor used
as input into the artificial neural network
Trang 8Comparison of classification techniques
In this study, ANN and DFA models were optimized in
order to classify fish schools Both techniques showed
nearly similar recognition performance The overall
clas-sification rate was higher for ANN than DFA, but
nevertheless was only slightly higher As for the three fishgroups’ relative classifications, there were minor differ-ences in classification success based on the two specifiedmethods In particular, differences were trivial for G1schools, whereas the successes of discrimination of G2 andG3 schools were significantly more important with ANNthan DFA (Tables4,5) The particularly effective power ofANN to classify fish schools is attributed to its ability to
Table 6 Catch amount (kg) by midwater trawling of abundant species assumed to compose acoustically detected fish schools
Scientific name Common name
G1 Engraulis japonicus Japanese anchovy 44.2 (13.4) 7.2 (1.8) 55.4 (42.0) 36.9 (9.3) 139.7 (49.1) 283.4 (18.4)
Etrumeus teres Round herring 67.7 (20.5) 10.8 (2.7) 0.9 (0.7) 5.5 (1.4) 23.0 (1.8) 107.8 (7.0) Sardinops
Trachurus
japonicus
Japanese jack mackerel
Others Arothron spp Puffer fishes 113.8 (34.4) 60.6 (15.3) 8.6 (6.5) 0.8 (0.2) 0.6 (0.2) 184.3 (12.0)
Loglio edulis Swordtip squid 5.9 (1.8) 8.7 (2.2) 8.3 (6.3) 16.6 (4.2) 14.4 (5.1) 53.8 (3.5)
Todarodes pacificus Japanese common
Values between brackets represent percentage (%)
Table 7 Comparison of acoustically detected schools with trawl catch composition
Japanese sardine
Scomber spp.
Trachurus spp.
Decapterus spp.
Lantern fishes
Upper line of each row represents catch amount per kg Lower line indicates percentage of catch amount
Values in bold represent fish species included in each group
Trang 9handle non-linear relationships between descriptors and
dependent variables, through the presence of many
inter-vening information-processing units, which each uses the
binary logistic activation function [27] A further
advan-tage of ANN is the small impact of extreme values on
discrimination success and the absence of any specific
assumptions on the distribution of the data In fact, ANN
established functional relationships of the data by learning
from the input training data set [29,30] On the other hand,
despite these advantages, a liability of its application is that
it needed much more computing time than discriminant
analysis, especially during optimization procedures such as
weight analysis
Similar performances of ANN and DFA in identifying
fish schools that have been found in several studies
cor-roborate our finding that ANN is more effective than DFA
[12,13,31] Moreover, they reported better, sometimes far
better, overall classification rates These mentioned case
studies inferred that an increasing number of descriptors
should lead to an improvement in discrimination
effec-tiveness However, Scalabrin et al [32] found a lower rate
when classifying only three species using nine school
parameters Theoretically, the greater the number of
parameters that can be included in the model, the more
likely the analysis will assign a school image to the correct
group [13] However, in our practical analysis, for both
classification methods we were limited to five acoustic
descriptors as input variables
Parameters controlling fish-group classification
The fish schools’ classification is defined as the
discrimi-nation of acoustic backscatters to the species, genus or
group level, depending on the richness of fish diversity
[10] In this work, the classification of fish echo traces into
three fish groups was reliable due to the high fish diversity
in the East China Sea The acoustic aggregations of the
numerous target species were categorized based solely on
their schooling characteristics The feasibility of this
approach is justified by the existence of acoustic
popula-tions; groups of echo traces show a consistent pattern in
space and time at a regional scale [33] In tropical waters,Gerlotto succeeded in dividing highly multispecific fishcommunities into four fish acoustic populations [34].However, for some marine systems at high latitudes, such
as the North Atlantic Ocean, species richness is relativelypoor The low number of target fishes and the occurrence ofmonospecific schools permitted a lower level of discrimi-nation and yielded a higher successful classification rate[32]
The vertical distribution of fish schools in the watercolumn gave evidence of the typology applied in thisstudy G1 schools existed predominantly in the upperlayer of the water column above the thermocline detected
at approximately 50 m depth (Fig.5a) The G2 schoolswere observed within a deeper layer below the thermo-cline G3 schools were distributed in the bottom half of thewater column below 150 m depth until the closest layer tothe sea bottom The vertical distribution of G3 species is
in agreement with the vertical range (160–200 m) ted by Fujino et al [35] in the case of pearlsides and below
repor-200 m depth in the case of mesopelagic lantern fishes[36]
The vertical distribution of fish schools exhibited anoticeable pattern that corresponds to the overlap of G1–G2 schools and G2–G3 schools in water layers at 60–80and 160–180 m, respectively Fish schools co-occurringwithin these depths could not be discriminated properly onthe basis of the positional descriptor Both methods (DFAand ANN) resulted in relatively weaker performancewithin these overlap layers; the correct classification ratedid not exceed 88%
Notwithstanding the fair limitation of fish-group fication within ‘overlap’ layers, the results of DFA andANN revealed that the school’s altitude in the water col-umn was the most effective acoustic descriptor in suc-cessfully discriminating schools into the three groups Onthe other hand, morphological acoustic descriptors andbackscattered volume Sv contributed to distinguishingamong species In fact, the G3 species pearlsides and lan-tern fishes formed generally large elongated aggregationsthat were fairly dense and characterized by relatively low
classi-Sv values The G2 species jack mackerel, spotted mackereland chub mackerel aggregated in relatively smaller schoolsmarked by higher Sv values (Fig.5b, c, d)
Although the vertical distribution of the target fishescannot be addressed in detail within the scope of this paper,
it provided valuable information about the environmentaland physiological properties of identified target species.The occurrence of Japanese anchovy, Japanese sardine andround herring above the thermocline was most likelyrelated to temperature gradient patterns Temperature at thesea surface varied between surveys from 26.5 to 28.8°C(Fig.6), and below 60 m depth, temperature profiles were
Table 8 Summary of acoustically detected schools with the most
abundant caught species in number of trawl hauls
Trang 10fairly homogenous The thermocline might have played the
role of a barrier that restricted the migration of these
spe-cies to deeper layers Thus, the thermal barrier implicitly
facilitated the identification of species confined to theupper layers [14]
The availability of food as well as avoidance of tion could also be plausible key factors concerning thevertical distribution patterns [37] Small fish may havemigrated to a depth level with a lower concentration oflarger fish to avoid predation [36] Myctophids and pearl-sides fishes feed on zooplankton They ascend from the seabottom at night following food and prey patterns and arethought to compete for food with pelagic fish within theupper layer [14,15]
preda-Fish identification improvementSeveral works have been using multiple frequency echosounding to allocate fish echoes to species by using thefrequency difference in mean volume backscatteringstrength (MVBS) and target strength differencing [38–40].These methods have shown considerable promise andprovided high rates of correct classification in restrictedecological situations, that is, none have provided a classi-fier that can be applied over broad ranges of time and space[41] In the East China Sea particularly, owing to the highfish diversity, the use of an extended number of narrow-band acoustic frequencies may facilitate the identification
of fish species More precisely, low frequencies might bethe best aid to increase species discrimination, for instance,midwater layers of mesopelagic fish appear much stronger
on 12 kHz than on 38 kHz [42–44] Simultaneously, withmore accurate acoustic surveys, additional trawl datashould facilitate the identification of fish species withineach group In parallel, the increase in the amount of col-lected data enhances ANN training and thus its efficacy.Taking the advantage of its fast performance and the speed
of processing using modern computers, the application ofANNs in real-time classification would be advantageous infisheries stock assessments
In the same order of magnitude, further statisticalanalysis should be performed to evaluate the consistency ofacoustic data and trawl data Ideally, the fish schoolsdetected during daylight acoustic surveys will be caughtusing the midwater trawling conducted only at nighttime.The horizontal migration of fish may bias the verification
of identified fish schools using trawl data However, in thiswork, the time lag was neglected since midwater stationswere meticulously chosen to correspond to locations oftarget fish schools observed previously in echograms.Quantification of uncertainty of the match between bothdata sets (acoustic and trawl data) may lead to improvingthe objectivity of fish identification and classification
In conclusion, this study demonstrated that the neuralnetwork can perform reasonably well in classifying fishschools and that it performs slightly better than DFA This
Fig 5 Distribution of detected schools in relation to depth (a),
height (b), length (c) and mean backscatter volume Sv (d)
Trang 11achievement was guaranteed by an integration of prior
knowledge, direct sampling in conjunction with the two
patterns of recognition and classification More
specifi-cally, the use of a set of five descriptors that combines
positional, energetic and morphologic criteria provided the
best fish-group discrimination The choice was made to
cover many aspects of the school while avoiding
parame-ters likely to generate redundant information In our
prac-tical analysis, for both methods, we concluded that
compiling a different set of descriptors and adding other
acoustic parameters (such as skewness and school
elon-gation) during model optimization led to a decrease in the
overall classification rate In some studies, more complex
criteria were implemented to parameterize the shape and
intrinsic structure complexity of the school [13,41] These
authors recognized that using such descriptors is
satisfac-tory for classification purposes for large schools, but likely
to become not valuable to some extent for smaller schools
This study succeeded in potentially improving the
objectivity of the identification and discrimination of fish
species, illustrated by high correct classification rates
accounting for both tools of analysis Further work on these
approaches should continue with an expanded acoustic data
set for all species of the three groups Subsequently, these
two methods will represent powerful means able to
increase the accuracy of the stock size assessment in the
East China Sea using hydroacoustic techniques For the
foreseeable future, acoustic surveys must be viewed as a
substitute for rather than a necessary complement to ventional survey methods
con-Acknowledgments We are grateful to Dr Hiroshige Tanaka (Fisheries Research Agency) for data collection, and Dr Tadanori Fujino and Dr Kazushi Miyashita (University of Hokkaido) for data analysis initiation Thanks are due to Dr Vidar Wespestad (University
of Alaska Fairbanks), Dr Hideaki Tanoue and Dr Teruhisa Komatsu (University of Tokyo) for providing advice at various stages of the work We thank Dr Takaomi Kaneko for his thorough editorial assistance.
References
1 Hansson S (1999) Human effects on the Baltic Sea ecosystem— fishing and eutrophication In: Anonymous (eds) Ecosystem approaches for fisheries management Fairbanks University of Alaska, pp 405–406
2 Kuikka S, Hilden M, Gislason H, Hansson S, Sparholt H, Varis O (1999) Modelling environmentally driven uncertainties in Baltic Cod management by Bayesian influence diagrams Can J Fish Aquat Sci 56:629–641
3 Simmonds J, MacLennan D (2005) Fisheries acoustics: theory and practice, 2nd edn Blackwell, Oxford
4 Brehmer P, Gerlotto F, Laurent C, Cotel P, Achury A, Samb B (2007) Schooling behaviour of small pelagic fish: phenotypic expression of independent stimuli Mar Ecol Prog Ser 334:263–272
5 MacLennan DN, Holliday DV (1996) Fisheries and plankton acoustics: past, present, and future ICES J Mar Sci 53:513–516
6 Mackinson S, Freeman S, Flatt R, Meadows B (2004) Improved acoustic surveys that save time and money: integrating fisheries and ground-discrimination acoustic technologies J Exp Mar Biol Ecol 305:129–140
7 Lundgren B, Nielsen JR (2008) A method for the possible species discrimination of juvenile gadoids by broad-bandwidth backscattering spectra vs angle of incidence ICES J Mar Sci 65:581–593
8 ICES (2000) Report on echo trace classification ICES ative Research Report, Denmark
Cooper-9 Horne J (2000) Acoustic approaches to remote species cation: a review Fish Oceanogr 9:356–371
identifi-10 Jech JM, Michaels WL (2006) A multifrequency method to classify and evaluate fisheries acoustics data Can J Fish Aquat Sci 63:2225–2235
11 Reid D, Scalabrin C, Petitgas P, Masse J, Aukland R, Carrera P, Georgakarakos S (2000) Standard protocols for the analysis
of school based data from echo sounder surveys Fish Res 47: 125–136
12 Simmonds EJ, Armstrong F, Copland PJ (1996) Species fication using wideband backscatter with neural network and discriminant analysis ICES J Mar Sci 53:189–195
identi-13 Haralabous J, Georgakarakos S (1996) Artificial neural networks
as a tool for species identification of fish schools ICES J Mar Sci 53:173–180
14 Ohshimo S (1996) Acoustic estimation of biomass and school character of anchovy Engraulis japonicus in the East China Sea and the Yellow Sea Fish Sci 62:344–349
15 Ohshimo S (2004) Spatial distribution and biomass of pelagic fish
in the East China Sea in summer, based on acoustic surveys from
1997 to 2001 Fish Sci 70:389–400
16 Iversen SA, Zhu D, Johannessen A, Toresen R (1993) Stock size, distribution and biology of anchovy in the Yellow Sea and East China Sea Fish Res 16:147–163
Fig 6 Vertical profiles of temperature during surveys
Trang 1217 Takeshita K, Ogawa N, Mitani T, Hamada R, Inui E, Kubota K
(1988) Acoustic surveys of spawning sardine, Sardinops
mel-anosticta, in the coastal waters of west Japan Bull Seikai Reg
Fish Res Lab 66:101–117
18 Ohshimo S, Mitani T, Honda S (1998) Acoustic surveys of
spawning sardine Sardinops melanostictus in the waters off
western and southern Kyushu, Japan Fish Sci 64:665–672
19 Connell SD (2000) Is there safety in numbers for prey? Oikos
88:527–532
20 Sassa C, Moser HG, Kawaguchi K (2002) Horizontal and vertical
distribution patterns of larval myctophid fishes in the Kuroshio
Current region Fish Oceanogr 11:1–10
21 Myriax (2007) Echoview Version 4.50 Myriax software Pty Ltd
1995–2008
22 Suuronen P, Lehtonen E, Wallace J (1997) Avoidance and escape
behaviour by herring encountering midwater trawls Fish Res
29:13–24
23 Duda RO, Hart PE, Stork DG (2001) Pattern classification, 2nd
edn Wiley, London
24 McLachlan GJ (2004) Discriminant analysis and statistical
pat-tern recognition Wiley, NY
25 Landau S, Everitt BS (2004) A handbook of statistical analyses
using SPSS Chapman & Hall/CRC, London
26 Arbib M (2003) The handbook of brain theory and neural
net-works, 2nd edn MIT Press, Cambridge
27 Basheer IA, Hajmeer M (2000) Artificial neural networks:
fun-damentals, computing, design, and application J Microbiol
Methods 43:3–31
28 Rumelhart DE, Hinton GE, Williams RJ (1986) Learning
repre-sentations by back-propagating errors Nature 323:533–536
29 Chen DG, Ware DM (1999) A neural network model for forecasting
fish stock recruitment Can J Fish Aquat Sci 56:2385–2396
30 Cabreira AG, Tripode M, Madirolas A (2009) Artificial neural
net-works for fish-species identification ICES J Mar Sci 66:1119–1129
31 Woodd-Walker RS, Watkins JL, Brierley AS (2003)
Identifica-tion of Southern Ocean acoustic targets using aggregaIdentifica-tion
back-scatter and shape characteristics ICES J Mar Sci 60:641–649
32 Scalabrin C, Diner N, Weill A, Hillion A, Mouchot MC (1996)
Narrowband acoustic identification of monospecific fish shoals.
ICES J Mar Sci 53:181–188
33 Petitgas P, Masse J, Beillois P, Lebarbier E, Le Cann A (2003) Sampling variance of species identification in fisheries-acoustic surveys based on automated procedures associating acoustic images and trawl hauls ICES J Mar Sci 60:437–445
34 Gerlotto F (1993) Identification and spatial stratification of tropical fish concentrations using acoustic populations Aquat Living Resour 6:243–254
35 Fujino T, Miyashita K, Aoki I, Masuda S, Uji R, Shimura T (2005) Acoustic identification of scattering layer by Maurolicus japonicus around the Oki Islands, Sea of Japan Nippon Suisan Gakkaishi 71:947–956 (in Japanese with English abstract)
36 Watanabe H, Moku M, Kawaguchi K, Ishimaru K, Ohno A (1999) Diel vertical migration of myctophid fishes (Family My- ctophidae) in the transitional waters of the western North Pacific Fish Oceanogr 8:115–127
37 Zwolinski J, Morais A, Marques V, Stratoudakis Y, Fernandes
PG (2007) Diel variation in the vertical distribution and schooling behaviour of sardine (Sardina pilchardus) off Portugal ICES J Mar Sci 64:963–972
38 Anderson CIH, Horne JK, Boyle J (2007) Applying a robust probabilistic classification technique to multi-frequency fisheries acoustics data J Acoust Soc Am 121:EL230–EL237
39 Kang M, Furusawa M, Miyashita K (2002) Effective and accurate use of difference in mean volume backscattering strength to identify fish and plankton ICES J Mar Sci 59:794–804
40 Gauthier S, Horne JK (2004) Potential acoustic discrimination within boreal fish assemblages ICES J Mar Sci 61:836–845
41 LeFeuvre P, Rose GA, Gosine R, Hale R, Pearson W, Khan R (2000) Acoustic species identification in the Northwest Atlantic using digital image processing Fish Res 47:137–147
42 Barr R (2000) A design study of an acoustic system for entiating orange roughy and other New Zealand deepwater spe- cies J Acoust Soc Am 109:164–178
differ-43 O’Driscoll RL (2003) Determining species composition in mixed-species marks: an example from the New Zealand hoki (Macruronus novaezelandiae) fishery ICES J Mar Sci 60:609– 616
44 Nero RW, Magnuson JJ (1989) Characterization of patches along transects using high resolution 70-kHz integrated echo data Can
J Fish Aquat Sci 46:2056–2064
Trang 13O R I G I N A L A R T I C L E Fisheries
Acoustic pressure sensitivities and effects of particle
motion in red sea bream Pagrus major
Takahito Kojima•Tomohiro Suga •Akitsu Kusano•
Saeko Shimizu•Haruna Matsumoto• Shinichi Aoki•
Noriyuki Takai• Toru Taniuchi
Received: 30 April 2009 / Accepted: 2 November 2009 / Published online: 15 December 2009
Ó The Japanese Society of Fisheries Science 2009
Abstract The auditory pressure thresholds of red sea
bream were examined using cardiac response in the field by
placing fish subjects far from the sound source to prevent
particle motion Pressure and particle motion thresholds
were also obtained using the auditory brainstem response
(ABR) technique The thresholds at 100 and 200 Hz were
significantly higher when measured using the cardiac
response in the far field than those obtained in previously
conducted experiments in experimental tub However,
thresholds obtained using ABR from 200 to 500 Hz were
not remarkably lower, although significantly different
(0.01 \ P \ 0.05), compared with those obtained using
cardiac response in the far field Furthermore, calculated
particle velocity thresholds indicated that fish probably
detected particle motion within the frequency range of
50–200 Hz, even in fish with a deactivated lateral line
Although the ABR method is widely applied in fish
audi-tory study, hearing thresholds are apparently affected by
particle motion
Keywords ABR Audiogram Dipole Far field
Inner ear Lateral line Near field
IntroductionAuditory stimuli to control the behavior, activity, and phys-iological condition of fish in coastal waters or culturingfacilities have persistently been targeted for investigation [1].Fish have been conditioned to be attracted by sound uponemission of a signal for marine ranching [2,3] When a soundsignal is emitted in water, water particle motion attenuatesfaster than pressure waves because pressure decreases line-arly while displacement decreases with the square of thedistance from the source according to the acoustic law [4,5].Consequently, in the field, it is difficult to propagate particlemotion to fish from a distant sound source Therefore, it might
be more important to evaluate inner ear sensitivity to pressurewaves instead of particle motion when auditory stimuli areused to control fish behavior
Two pathways are known for the detection of soundstimuli One is by transformation of sound pressure to particledisplacement by the swimbladder, and the other is by directdetection of particle motion in the inner ear [6] The auditorybrainstem response (ABR) technique is a well-known, non-invasive, far-field recording method in which the neuralactivity of the eighth nerve and brainstem auditory nucleielicited by acoustical stimuli are detected through the skulland skin at the head region [7] Furthermore, fish with eitherintact or deactivated lateral line have identical thresholds [7].However, it remains unclear whether the audiogram obtained
by ABR is affected by particle motion, because particle placement is detected not only by the lateral line but also bythe inner ear directly, as described above
dis-Red sea bream (Pagrus major) is a major industrial fishspecies in Japan for which audiograms were previouslyexamined using cardiac suppression [8 10] To date, theclassical method for assessing fish auditory sensitivity is bydetermining hearing thresholds by cardiac suppression due
T Kojima (&) A Kusano S Shimizu H Matsumoto
S Aoki N Takai T Taniuchi
College of Bioresource Sciences, Nihon University,
Fujisawa, Kanagawa 252-8510, Japan
e-mail: kojima.takahito@nihon-u.ac.jp
T Suga
Graduate School of Fisheries Science, Hokkaido University,
Hakodate, Hokkaido 041-8611, Japan
Present Address:
T Suga
National Research Institute of Fisheries Engineering,
Kamisu, Ibaraki 314-0408, Japan
Fish Sci (2010) 76:13–20
DOI 10.1007/s12562-009-0194-x
Trang 14to sound signals after conditioning [11–14] Although two
speakers are placed face to face to eliminate water particle
displacement at the center of the tub, there is no evidence
that fish only perceive pressure-dominated sound stimuli by
the inner ear The setup for sound emission in the ABR
technique usually includes a speaker suspended beyond the
tub or a submerged underwater speaker, which might
generate water particle motion around the fish In
mea-suring the hearing abilities of Acipenseridae and
Cyprinidae, Lovell et al [15,16] used two sound
projec-tors, either driven out of phase to create an area associated
with high particle motion or driven in phase to create a
field dominated by sound pressure; the recorded thresholds
were lower in the sound field dominated by particle
motion Additionally, hearing experiments using auditory
evoked potentials (AEP) in response to dipole or monopole
sound stimuli consisting of particle motion reveal that
elasmobranches are effective in receiving stimuli from
dipole sources and are more sensitive to dipole sound than
to monopole sound [17, 18] In goldfish, which has a
swimbladder that acts as a pressure transducer, there is no
difference in detecting dipole and monopole stimuli in a
small tub using respiration response [19] To compare
results from ABR without the influence of the lateral line,
measurements of pressure sensitivity by the inner ear
sys-tem are conducted in air [20] to avoid the effect of water
particle motion generated by sound However, it remains
unclear whether the threshold levels determined by ABR
are affected by the sensitivity to particle motion, or to what
extent the sensitivity is influenced by other sound
components
In the present study, we analyzed the thresholds of red
sea bream, a hearing generalist because it lacks mechanical
connections between the swimbladder and the inner ear
We first used a classical conditioning method by cardiac
suppression to determine its hearing thresholds in the field,
with the sound source set apart from the subject fish to
prevent particle motion We also employed the ABR
method to measure both fish hearing sensitivity and the
threshold for particle motion generated by a vibrating
sphere Sensitivities to particle motion were measured in
both fish with intact and fish with pharmacologically
deactivated lateral line We used these results to examine
the hearing sensitivity of this fish species We also assessed
whether the ABR method, which is usually conducted in
the near field, represented the sensitivity of the inner ear for
sound pressure or was affected by particle motion
Materials and methods
In all, 63 red sea bream obtained from a local fish
dis-tributor were used for the measurements Fish were kept in
a 2000-l tank at the Marine Laboratory of Nihon University
in Shimoda, Shizuoka, Japan A total of 35 fish (16 cm FL;
90 g BW) were used to measure the cardiac response tosound signals in the loch The response of 18 individuals(16 cm FL; 95 g BW) to an air speaker and of another 10fish (18 cm FL; 89 g BW) to a vibration generator wereassessed by ABR method
Cardiac methods with a distant sound source
An iron frame (5.5 9 5.5 m2) with floats was constructedand moored in a loch off the shore of the marine laboratory
at a depth of 3 m The water temperature was 21–25°C.The fish were anesthetized by phenoxyethanol (Wako,Tokyo, Japan) solution (1 ml/l) A silver line (ca 10 mmlong; 1 mm diameter), insulated with a thin polyethylenetube and open at the tip for conduction of electricity, wasattached to a small silver disk (ca 5 mm diameter) Theline was inserted into the chest of the fish and bonded to thedisk using instant glue (Kony Bond, Osaka, Japan) Afterthe operation, fish cardiograms were examined using abiomedical amplifier (MEG-1200; Nihon Koden, Tokyo,Japan) to confirm whether a distinct cardiogram wasdetected The fish were placed inside a plastic net cagesuspended from the metal frame at 1 m depth (Fig.1) Apair of small copper plates (10 9 50 mm2), one at each end
of the cage, was attached to apply electric shock
Conditioning was conducted by emitting sound ing an electric shock A pure-tone sound signal was gen-erated using a function synthesizer (1915; NF, Kanagawa,Japan), attenuated (AL-255; Ando Electric-YokogawaElectric, Tokyo, Japan), amplified (AD-1; Pioneer, Tokyo,Japan), and emitted from an underwater loudspeaker(US300; Fostex, Tokyo, Japan), which was suspended fromthe frame at a distance of 7.7 m from the test fish, hereindefined as far field At each frequency examined, theconditioning was performed using 1-s sound signals at 105,
follow-205, 305, 505, 1010, 1510, and 2010 Hz, followed by anelectric shock (5–7 V AC) Sound frequencies were shifted
r e z i s h t n y S
m 7 7
p m a o i d A
p m a o i B
Fig 1 Schematic drawing of the experimental setup for ment using cardiac response in the far field
Trang 15measure-slightly to avoid the influence of electrical noise at
fre-quencies that are multiples of the AC power supply
(50 Hz) The boundary between the predominant sound
pressure and water particle motion for the sound source
was calculated at 4.8 m for 100 Hz; therefore, at the set
distance between the test fish and underwater loudspeaker
(7.7 m) used in this study, it was assumed that sound
pressure was more dominant than particle motion [4,21]
Sound pressure was calibrated using a hydrophone and an
underwater sound level meter (SW1020, ST1020; Oki
Electric Industry, Tokyo, Japan) in the absence of the fish
The ambient noise level was also measured 10 times using
the hydrophone and averaged at the tested frequencies
Sound pressure of 135 dB (0 dB re 1 lPa) for 1 s was used
during conditioning, which is detectable to the fish at
100–1000 Hz according to Ishioka et al [8] The
stimula-tion for each individual fish was repeated 10 times at 5-min
intervals After a recovery period of around 10 h, the
audi-tory response to sound was determined as positive when
heartbeat intervals immediately after the sound emission
were extended significantly before 25 interbeats (Fig.2)
The auditory detection thresholds were determined as the
level between the positive and negative responses
In addition to measurements using normal (intact)
sub-jects, the swimbladder of several fish that were lightly
anesthetized with 0.1% phenoxyethanol was punctured
from the side with a needle (22G; Terumo, Tokyo, Japan)
The gas inside the bladder was removed using a syringe
(30 ml; Terumo, Tokyo, Japan) according to the
method-ology described by Sand and Enger [12], Yan and
Curtsinger [22], and Yan et al [23] After removal of the
gas, the subjects were conditioned, and the auditory
thresholds were measured using the procedures described
above To confirm the removal of gas from the
swim-bladder, the subjects were again anesthetized with the same
phenoxyethanol solution after the measurements, and a
radiograph image of the swimbladder’s shape for each
subject was taken by a soft X-ray machine (Softex
SFX-130, Softex, Kanagawa, Japan) (Fig 3) The old levels (n = 5: 18.2 cm FL, 132 g BW) were used afterconfirming the removal of gas from the swimbladder.Auditory brainstem response to sound from an airspeaker
thresh-The schematic design of the ABR experiment is shown inFig.4 Test fish were injected with 0.2–0.4 ml gallaminetriethiodide solution (0.02% Flaxedil; Sigma, St Louis,USA) to inhibit skeletal muscle movements Everyimmobilized subject was immersed into a seawater-filledplastic tub (25 9 37 9 13 cm3) and secured by a holder.The body position and water level in the tub were adjusted
so that the nape was just above the water surface Aeratedwater was gravity-fed via a plastic tube through the mouth
to irrigate the gills during the experiments A small piece oftissue paper was placed on the head region Plastic insu-lation was peeled from the tip of the thin silver wire(0.5 mm diameter), and this was placed on the mid-line ofthe skull over the medulla region A reference electrodewas placed about 5 mm anterior to the recording electrode.The two electrodes were clamped to a manipulator (SM-15;Narishige, Tokyo, Japan), which led to a biomedicalamplifier (MEG-1200; Nihon Koden, Tokyo, Japan), andthe ground terminal of the amplifier was connected to thewater in the tub The air speaker (WS-A10-K; Panasonic-Matsushita Electric Industrial, Osaka, Japan) was mounted
60 cm above the subject Pure-tone sound signal emittedfrom the speaker was generated by the same functionsynthesizer, attenuator, and audio amplifier that were used
to analyze cardiac response in the loch The subject fish inthe tub and air speaker were placed in a fine metal meshcage (170 9 130 9 70 cm3) to shield the apparatus elec-trically, and the cage was surrounded with polystyrene
Fig 2 Example of heartbeat extension of fish conditioned to the
signal sound (300 Hz, 120 dB re 1 lPa) emission Arrow indicates
the sound emission
Fig 3 Photograph of representative subject goldfish with fully deflated swimbladder
Trang 16foam boards (ca 40 mm thickness) to prevent ambient
noise Sound pressure was calibrated in the absence of fish,
with the hydrophone positioned at the designated position
of the fish during the experiments Background noise was
also recorded by the hydrophone 10 times and averaged
The sound signal and ABR waveform recording were sent
to a personal computer via an analog-to-digital (A/D)
converting data recording system (NR-500; Keyence,
Osaka, Japan)
The sound stimuli consisted of a 10-cycle sinusoidal
waveform repeated 300 times Half of the 300 waveforms
were shifted in phase by 180° to reduce contamination
between the sound signal and ABR waveform The evoked
bioelectrical responses were recorded with 100-ls interval
and averaged (Excel; Microsoft) Positive responses with
the ABR waveforms were usually determined by visual
inspection to distinguish the response from noise (Fig.5)
Because it has been noted that the ABR waveform showed
a doubling of the stimulus frequency [24], we used power
spectral analysis (fast Fourier transform, FFT) for
wave-forms using 2048 data sets when it was difficult to
determine a positive response visually and analyzed for thepresence of significant peaks at twice the frequency of thestimuli that were obviously above background levels[24, 25] (Fig.6) At each frequency, the highest soundpressure level (ca 120–130 dB) was first projected andthen attenuated in 4-dB steps until a positive response wasnot obtained in the power spectra The threshold level ofauditory sensitivity at each frequency was defined as the
Vibrator
Fig 4 Schematic drawing of the experimental setup for
measure-ment of ABR in the experimeasure-ment under the condition in which sound
was emitted by an air speaker (upper) and particle motion was
generated by a vibrating sphere (lower)
Fig 5 Examples of auditory brainstem responses from subject fish.
In the upper two traces, a positive response was visually identified at
84 dB re 1 lPa, although no response was observed at 76 dB In the lower two traces, it was difficult to distinguish a positive response visually between 96 and 100 dB
Fig 6 Fast Fourier transform (FFT) of the ABR response to the sound indicated in the lower two traces (105 Hz sound) in Fig 5 The frequency for sound projection at 100 Hz was shifted slightly to
105 Hz to avoid the effects of electric noise (50 Hz) Therefore, the positive response was observed at the doubled frequency of 210 Hz
Trang 17intermediate value between the lowest sound pressure level
that elicited an obvious response and the level with no
detected response
Auditory brainstem responses generated by detecting
particle motion
We also measured particle motion (dipole stimuli)
sensi-tivity using ABR The treatment of test fish and the setup to
detect ABR potentials resembled those described above,
except for the generation of particle motion A plastic
sphere (0.5 cm diameter) supported by a thin metal rod
(2 mm diameter) was set in the tub at 1 cm from the fish
The sphere was vibrated in the horizontal direction (head to
tail) using a self-modified air pump connected to the
function synthesizer, attenuator, and audio amplifier
described previously (Fig.4) The function synthesizer
generated a 20-cycle sinusoidal waveform repeated 300
times, and half of the waveforms were shifted in phase by
180° The frequencies were set at 50, 150, and 200 Hz
Particle motion intensity was measured using a hydrophone
and an underwater sound level meter, as previously
described For this reason, particle motion is expressed as
the sound pressure level (dB re 1 lPa) in this study When
it is necessary to convert the sound pressure level to
par-ticle velocity, we applied the definition that sound pressure
is equal to the acoustic impedance multiplied by the
par-ticle motion, p = qcv, where p signifies pressure, q denotes
the density of water, c represents the speed of sound in
water (1500 m/s), and v represents the particle velocity
generated by a propagating wave in the far field, even
though the measurements were conducted in the near field
[26] Therefore, only relative changes in the value of
par-ticle velocity can be compared A positive response was
determined when obvious peaks at twice the frequency of
the stimulus in the power spectral analysis (FFT) were
detected for the waveform Streptomycin sulfate (Wako,
Tokyo, Japan) solution was used to deactivate the lateralline function to determine the effect of lateral line sensi-tivity on ABR measurements [27] Five fish were placedfor 3 h in seawater containing 0.5 g/l streptomycin sulfate.Measurements were conducted identically, using the sameapparatus and procedures used for intact fish
All audiograms obtained in the experiment were pared using one-way analysis of variance (ANOVA) to testfor difference among the thresholds obtained using thedifferent methods at a significance level of a = 0.05
com-ResultsThe thresholds obtained using cardiac response in the farfield are plotted as an audiogram in Fig.7, together withthe results of Ishioka et al [8] and Iwashita et al [10], whoconducted similar experiments in tanks where soundsources were set nearer to their subjects, herein defined asnear field Thresholds obtained using cardiac response afterconfirming deflation of the fish swimbladder are alsoshown, although only one threshold was obtained at eachfrequency Thresholds with the deflated bladder tended to
be higher than those of intact subjects; the differences weregreater at 200, 300, and 500 Hz than at 100 and 2000 Hz.Cardiac responses in the far field and those previouslytaken in the near field were significantly different(P \ 0.01) at 100 and 200 Hz [8, 10] The hearingthresholds determined by cardiac response in the far fieldand ABR, superimposed with the thresholds for particlemotion calculated from records of sound pressure levels at
150 and 200 Hz, are shown with ambient noise in Fig.8.Although the audiograms obtained by cardiac response inthe far field and ABR are similar and have their lowestlevels at 300 Hz, the thresholds obtained using ABR weresignificantly lower than those from cardiac response in thefar field at 200, 300, and 500 Hz (P \ 0.05) Meanwhile,
+
40 60 80 100 120 140
10000 1000
100
ECG in far-field removed gas in the bladder ECG in near-field by Ishioka et al.
ECG in near-field by Iwashita et al.
noise far-field noise near-field
Hz) ( y n e u e r F
*
Fig 7 Comparison of audiograms obtained using cardiac response in
the field with the sound source distant from the fish (far field, present
study), and in the near field (Ishioka et al [ 8 ] and Iwashita et al.
[ 10 ]), and the cardiac response in the far field after deflating the
swimbladder (present study) Signs indicate significant differences (P \ 0.01) between thresholds obtained in the near field by Ishioka
et al (asterisk), by Iwashita et al (plus), and far field
Trang 18the particle velocity thresholds calculated from sound
pressure level for both intact and lateral line deactivated
fish are depicted in Fig.9 The particle velocity thresholds
of around 10-6 to 10-5 cm/s at 100 and 200 Hz, which
were taken by indirect measurements in this study, are
similar to those of sciaenid species where the particle
velocity was measured directly [28] The particle velocity
thresholds of intact and lateral line deactivated fish were
not different (P [ 0.05)
Discussion
Particle motion generated by an underwater speaker
sepa-rated by 7.7 m from fish might be damped according to
hypothetical near-field and far-field boundaries for 100 and
200 Hz [4] Thresholds in the range of 200–500 Hz in fish
with deflated swimbladders were higher than in fish with
intact bladders, which is consistent with results obtained
for goldfish [29], gourami fish [23,30], and cod [12] This
result suggests that the swimbladder plays an important
role in the sensitivity of fish [12, 30–32] Even though
cardiac response experiments in the near field are
per-formed using two speakers facing one another to reduce
particle motion [9, 10], the lower thresholds at 100 and
200 Hz relative to similar experiments in the far fieldsuggest that the thresholds might be affected by sensitivity
to particle motion Meanwhile, differences in hearingthresholds recorded by cardiac response in the far field and
by ABR using air speaker at 200, 300, and 500 Hz werelikely caused by higher ambient background noise between
200 and 300 Hz in cardiac response in the far field because
it was pointed out that the critical ratios were at around 10–
20 dB [9,33] Nevertheless, the threshold levels obtainedusing these methodologies were very close at 100 Hz
It has been suggested that teleost fishes can detect waterparticle motion using the lateral line at frequencies ofaround 100 Hz or lower [19,21,34] The ABR techniquesupports several approaches to project signal sounds to thesubject fish, as with an air speaker suspended above the testsubject (e.g., [7]), or a submerged projector and twounderwater projectors facing each other, driven in phase tocreate a sound-pressure-dominated field [15, 16] Irre-spective of the position of the sound projector in ABRexperiments, the auditory thresholds for some teleost fishestended to be higher at 100 Hz than at 200 Hz [16, 24],except in the case of elasmobranches [17, 35] Kenyon
et al [7] referred to unchanged ABR thresholds at 100 Hzfollowing pharmacological treatment by a lateral linefunction blocker in goldfish; it is likely that lateral linesensitivity is not recorded as the ABR waveform A com-ponent at twice the frequency of the signal sound in thepower spectrum of the ABR wave, which was used todetect whether fish responded, is the characteristic response
of otolithic organs, where the otolith is supported byunderlying hair cells that are oriented oppositely in theinner ear [6] Moreover, the threshold levels of intact andlateral line deactivated fish obtained by particle motionwere not significantly different in the present study(Fig.9) However, sensitivity to water particle motion hasusually been detected for ABR waveforms in elasmo-branches, which have an area on the top of the head wherethe cranium is depressed ventrally with a jelly-like tissue—
40 60 80 100 120 140
10000 1000
100
ABR ECG in far-field Vibration noise ABR noise far-field
Frequency (Hz)
×
Fig 8 Comparison of audiograms obtained by ABR technique and
cardiac response in the far field Two thresholds for particle motion
detected as sound pressure level are superimposed Ambient noise
levels in the ABR and in the far field are also presented as dotted lines Crosses indicate significant differences (P \ 0.05) between thresholds obtained by ABR and cardiac response in the far field
150 100
50 0
Intact Deactivated
Frequency (Hz)
Fig 9 Comparison of audiograms of particle motion (in cm/s) for
intact fish and fish with pharmacologically deactivated lateral line
Trang 19parietal fossa [17,18,36] The inner ear of fish can detect a
dipole source directly in the near field or indirectly by the
swimbladder in the far field [19,37] The findings that the
thresholds obtained by ABR in a
particle-motion-domi-nated field were lower than in a sound-pressure-domiparticle-motion-domi-nated
field [15], and that the dipole stimulus, with calculated
thresholds resembling those of sound pressure in the
present study, was really detected (Fig.8), suggest the
existence of particle motion sensitivity in ABR The
par-ticle motion generated by the air speaker might be
mono-pole and different from that caused by the vibrating sphere
[38], and the intensity of particle velocity might be
underestimated as the sound pressure level in the present
study Thus, some concerns persist that audiograms
mea-sured using ABR technique, which is usually considered as
pressure wave sensitivity, includes sensitivity to particle
motion detected directly by the inner ear In addition, it
was considered that the otolithic ear can sense the inertia of
denser otolith to create a shearing force on hair cells by
moving succulus sensory epithelia with respect to lagging
otoliths [6] Therefore, the dual sensitivity to pressure and
particle motion makes the study of hearing in fish difficult
and confusing [6]
As described previously, the pressure component is
considered more important than particle motion when
controlling red sea bream behavior by sound stimuli
Although lateral line sensitivity might not affect the
thresholds obtained by ABR, it remains unclear whether
the frequency providing the best sensitivity of red sea
bream at 300 Hz was the actual pressure sensitivity or the
sensitivity to particle motion of the inner ear Therefore,
further studies should be conducted to clarify the
sensi-tivities to particle motion and sound pressure in order to
assess and select effective modes for transmitting
under-water sound stimuli in fish culture systems
Acknowledgments We would like to express sincere thanks to Dr.
Ricardo Babaran, University of the Philippines in the Visayas, who
kindly read and gave us useful suggestions on this manuscript We are
grateful to the staff of the Shimoda Marine Laboratory of Nihon
University for their helpful support Thanks are also extended to the
students of our laboratory at Nihon University who helped with the
experiments This work was partially supported by a Nihon
Univer-sity Grant-in-Aid in 2006.
References
1 Popper AN, Carlson TJ (1998) Application of sound and other
stimuli to control fish behavior Trans Am Fish Soc 127:673–707
2 Okamoto M (1982) Studies on the community ecology of fishes
near the experimental fish nursery of Pagrus major using acoustic
conditioning Nippon Suisan Gakkaishi 48:1113–1119
3 Anraku K, Matsuda M, Nakahara M, Shigesato N, Kawamura G
(1997) Sound learned by red sea bream conditioned to
intermit-tent 300 Hz sound Nippon Suisan Gakkaishi 63:934–938
4 Coombs S, Janssen J, Montgomery J (1992) Functional and evolutionary implications of peripheral diversity in lateral line systems In: Webster DB, Fay RR, Popper AN (eds) The evolutionary biology of hearing Springer, New York
5 Schellart NAM, Popper AN (1992) Functional aspects of the evolution of the auditory system of actinopterygian fish In: Webster DB, Fay RR, Popper AN (eds) The evolutionary biology
of hearing Springer, New York
6 Popper AN, Fay RR (1999) The auditory periphery in fishes In: Fay RR, Popper AN (eds) Comparative hearing: fish and amphibians Springer, New York
7 Kenyon TN, Ladich F, Yan HY (1998) A comparative study of hearing ability in fishes: the auditory brainstem response approach J Comp Physiol A 182:307–318
8 Ishioka H, Hatakeyama Y, Sakaguchi S (1988) The hearing ability of the red sea bream Pagrus major Nippon Suisan Gakkaishi 54:947–951
9 Hatakeyama Y (1989) Masking effect on the hearing of red sea bream, Pagrus major, by ambient noise Int J Aquat Fish Technol 1:271–277
10 Iwashita A, Sakamoto M, Kojima T, Watanabe Y, Soeda H (1999) Growth effects on the auditory threshold of red sea bream Nippon Suisan Gakkaishi 65:833–838
11 Buerkle U (1967) An audiogram of the atlantic cod, Gadus morhua L J Fish Res Board Can 24:2309–2319
12 Sand O, Enger PS (1973) Evidence for an auditory function of the swimbladder in the cod J Exp Biol 59:405–414
13 Hawkins AD, Chapman CJ (1975) Masked auditory thresholds in the cod, Gadus morhua L J Comp Physiol 103:209–226
14 Sawa M (1976) The audiogram of the goldfish determined by a heart rate conditioned method Bull Fac Fish Hokkaido Univ 27:129–136
15 Lovell JM, Findlay MM, Moate RM, Nedwell JR, Pegg MA (2005) The inner ear morphology and hearing abilities of the paddlefish (Polyodon spathula) and the lake sturgeon (Acipenser fulvescens) Comp Biochem Physiol A 142:286–296
16 Lovell JM, Findlay MM, Nedwell JR, Pegg MA (2006) The hearing abilities of the silver carp (Hypophthalmichthys molitrix) and bighead carp (Aristichthys nobilis) Comp Biochem Physiol
A 143:286–291
17 Casper BM, Mann DA (2006) Evoked potential audiograms of the nurse shark (Ginglymostoma cirratum) and the yellow stingray (Urobatis jamaicensis) Environ Biol Fish 76:101–108
18 Casper BM, Mann DA (2007) Dipole hearing measurements in elasmobranch fishes J Exp Biol 210:75–81
19 Coombs S (1994) Nearfield detection of dipole source by the goldfish (Carassius auratus) and the mottled sculpin (Cottus bairdi) J Exp Biol 190:109–129
20 Suga T, Kawabe R, Hiraishi T, Nashimoto K (2004) Sound pressure values in the audiogram of marbled sole Pleuronectes yokohamae measured in air Fish Sci 70:33–40
21 Popper AN, Fay RR (1973) Sound detection and processing by teleost fishes: a critical review J Acoust Soc Am 53:1515–1529
22 Yan HY, Curtsinger WS (2000) The otic gasbladder as an ancillary auditory structure in a mormyrid fish J Comp Physiol A 186:595–602
23 Yan HY, Fine ML, Horn NS, Colo´n WE (2000) Variability in the role of the gasbladder in fish audition J Comp Physiol A 186:435–445
24 Mann DA, Cott PA, Hanna BW, Popper AN (2007) Hearing in eight species of northern Canadian freshwater fishes J Fish Biol 70:109–120
25 Egner SA, Mann DA (2005) Auditory sensitivity of sergeant major damselfish Abudefduf saxatilis from post-settlement juve- nile to adult Mar Ecol Prog Ser 285:213–222
Trang 2026 Siler W (1969) Near- and far fields in a marine environment J
Acoust Soc Am 46:483–484
27 Baker CF, Montgomery JC (1999) Lateral line mediated
rheo-taxis in the Atlantic fish Pagothenia borchgrevinki Polar Biol
21:305–309
28 Horodysky AZ, Brill RW, Fine ML, Musick JA, Latour RJ (2008)
Acoustic pressure and particle motion thresholds in six sciaenid
fishes J Exp Biol 211:1504–1511
29 Fay RR, Popper AN (1974) Acoustic stimulation of the ear of the
goldfish (Carassius auratus) J Exp Biol 61:243–260
30 Yan HY (1998) Auditory role of the suprabranchial chamber in
gourami fish J Comp Physiol A 183:325–333
31 Akamatsu T, Nanami A, Yan HY (2003) Spotlined sardine
Sardinops melanostictus listens to 1-kHz sound using its gas
bladder Fish Sci 69:348–354
32 Chapman CJ, Sand O (1974) Field studies of hearing in two
species of flatfish Pleuronectes platessa (L.) and Limanda
limanda (L.) (family pleuronectidae) Comp Biochem Physiol A
47:371–385
33 Scholz K, Ladich F (2006) Sound production, hearing and sible interaction under ambient noise conditions in the topmouth minnow J Fish Biol 69:892–906
pos-34 Coombs S, Janssen J (1990) Behavioral and neurophysiological assessment of lateral line sensitivity in the mottled sculpin, Cottus bairdi J Comp Physiol A 167:557–567
35 Casper BM, Lobel PS, Yan HY (2003) The hearing sensitivity of the little skate, Raja erinacea: a comparison of two methods Environ Biol Fish 68:371–379
36 Myrberg AA Jr (2001) The acoustical biology of elasmobranches Environ Biol Fish 60:31–45
37 Fritzsch B (1999) Hearing in two worlds: theoretical and actual adaptive changes of the aquatic and terrestrial ear for sound reception In: Fay RR, Popper AN (eds) Comparative hearing: fish and amphibians Springer, New York
38 Harris GG (1964) Considerations on the physics of sound duction by fishes In: Tavolga WN (ed) Marine bio-acoustics Pergamon, New York
Trang 21pro-O R I G I N A L A R T I C L E Fisheries
Comparisons of monthly and geographical variations
in abundance and size composition of Pacific saury between the
high-seas and coastal fishing grounds in the northwestern Pacific
Wen-Bin Huang
Received: 24 June 2009 / Accepted: 17 November 2009 / Published online: 17 December 2009
Ó The Japanese Society of Fisheries Science 2009
Abstract The monthly and geographical abundances
and size compositions of Pacific saury were compared
between the high-seas and coastal fishing grounds in
the northwestern Pacific during 2000–2005 based on
Taiwanese fishery data The large-sized saury was
domi-nant (44.3–71.4% of the catch) in the beginning of the
fishing season, while the medium-sized saury followed and
dominated from September to the end of the fishing season
(70.1–92.4% of the catch) In the high seas, the total catch
per unit effort (CPUE) (about 71.2% of the mean coastal
value) and both the large- (about 55.0%) and medium-sized
saury CPUEs (about 81.8%) were significantly smaller than
those in the coastal waters The mean proportions of the
large- and medium-sized saury in the high-seas catch were
about 86.6 and 107.0% of the coastal values, respectively
CPUEs for the total catch and the catch of medium-sized
saury varied in a highly consistent way The total and
medium-sized CPUEs were negatively correlated with the
sea-water temperature When the temperature was held the
same statistically, the total and medium-sized CPUEs were
larger in the shoreward, southward, and shallower waters of
the fishing grounds, while the large-sized CPUE was larger
in the shoreward waters
Keywords Cololabis saira Environmental correlations
High-seas fishing ground Size composition
Spatiotemporal distribution Taiwanese saury fishery
IntroductionPacific saury, Cololabis saira (Brevoort), is a major pelagiccommercial fish in the Far East, particularly for Japan andTaiwan Both abundance and size composition of Pacificsaury exhibit large variations among years [1 3] In Japan,the largest harvesting country, annual catches of Pacificsaury have fluctuated greatly from 575,000 mt in 1958 to63,000 mt in 1969, with an annual average of about258,000 mt over the last half century [4] The ratio of thelarge-sized Pacific saury (knob length[290 mm [5]) in theharvest of Japan fluctuated from 0.09 in 1977 to 0.93 in
2005 [6] The Japanese Pacific saury fishery is a coastalfishery, since its fishing grounds are generally locatedwithin the Exclusive Economic Zone (EEZ) of 200 nauticalmiles (about 370 km) Most of the averaged distances forthe Japanese fishing fleets from the Pacific saury fishinggrounds to shore were \150 km in 1971–1991, except in1981–1986 when distances were 170–330 km [7]
The migration of the Pacific saury to the coastal waters,including inshore and offshore areas, have been well studiedand documented since the 1950s [1 3,8 11] However, thePacific saury migrating to the high-seas fishing grounds areexclusively exploited by the Taiwanese fishing fleets, whichalso catch Pacific saury in the coastal waters as participants
in cooperative fishing with Russia and Japan (Fig.1) Theannual catch of Pacific saury in Taiwan increased dramat-ically from 27,900 mt in 2000 to 111,500 mt in 2005 [4](Fig.2) and 139,500 mt in 2008 (Overseas FisheriesDevelopment Council in Taiwan, 2009, personal commu-nication) The mean annual catch (63,800 mt) during2000–2005 made Taiwan the second largest Pacific sauryharvesting country after Japan Yet, in contrast to manystudies of the coastal waters, the state of the Pacific saury onthe high seas has largely been unexamined
W.-B Huang (&)
Graduate Institute of Biological Resources and Technology,
National Dong Hwa University, Meilum Campus,
Hualien 970, Taiwan
e-mail: bruce@mail.ndhu.edu.tw
Fish Sci (2010) 76:21–31
DOI 10.1007/s12562-009-0196-8
Trang 22Pacific saury is a scomberesocid fish, distributed widely
on both sides of the subarctic and subtropical North Pacific
[12] The physiological longevity of Pacific saury is
2 years [13] It grows quickly to around 28–30 cm in the
first (0?) year and reaches 30–33 cm in the second (1?)
year The smallest recorded mature fish was 25.3 cm, with
fish larger than 28.0 cm comprising the principal part of the
spawning stock [14]
Annual migratory patterns of Pacific saury in the
northwestern Pacific (NWP) have been proposed in several
studies [11,15] Pacific saury migrates extensively between
the summer feeding grounds in the Oyashio waters around
Hokkaido and the Kuril Islands and the winter spawning
areas in the Kuroshio waters off southern Japan [1,2] This
annual migratory cycle is believed to allow for the optimal
utilization of planktonic food resources, since the plankton
biomass in the Oyashio Current area is much higher than
that in the Kuroshio Current area during the summer
[16, 17] Most of the adults mature and are ready for
spawning during their southward migration [2, 18] The
Pacific saury fishing season, coinciding with this migration
period, begins generally in August and continues through
to mid-December [1,19] In addition to its importance to
fisheries, Pacific saury also plays an important role in theNWP ecosystem as a predator of zooplankton [16] and as aprey species for ichthyophagous fishes [20], sea birds, andmarine mammals [21,22]
Understanding the age (cohort) composition and lation structure of Pacific saury stocks can reduce theuncertainty of stock assessment and management [11].Along with production and demand, the size of fish is one
popu-of the dominant factors that affects the price popu-of fish infisheries and aquaculture [23] The knob lengths of thelanded Pacific saury in Japan are generally graded intothree size groups, namely large ([29.0 cm), medium(24.0–29.0 cm), and small (20.0–24.0 cm) [24] The large-sized group is considered to mostly consist of the 1? yearcohorts and most of the medium- and small-sized saury are0? year [13,25]
Due to an excessive supply in Japan, a landing limitationhas been enforced throughout the fishing season since thelate 1980s to avoid an abrupt breakdown of the Pacificsaury market price [26] Yamamura [19] suggested that,under this landing limitation, fishermen would discard thesmall-sized saury that had been found in the stomachs ofdemersal fishes to maximize their profits Fish sortingmachines began to be used by fishermen on boats toselectively land members of the larger-sized class ofPacific saury with a higher price starting in the mid-1990s[27, 28], making the issue of discarding worse [6] Inconsequence, there was a series of price collapses caused
by the excessive supply of large-sized fish in the 2000s, andthe fishermen decided to remove the sorting machines fromtheir vessels in 2006 [6]
Knowledge of environmental effects on the variability
of distributions and abundances of exploited fishery stocks
is key for future management strategies [29] Locations andsizes of the Pacific saury fishing grounds are largelydependent on oceanographic conditions [7] Temperaturehas been reported to be a dominant factor in determiningthe boundaries of migratory paths of Pacific saury [30] Thesea surface temperature (SST) in the Kuroshio region inwinter and in the Kuroshio–Oyashio Transition Zone andOyashio region has been found to affect abundances of thelarge-sized (winter-cohort) and medium-sized (spring-cohort) groups of Pacific saury, respectively [3] Recently,
a modeling of the spatial and temporal migration of thePacific saury stock in the NWP was proposed to incorpo-rate the effect of the SST [28] However, the high-seas datawere sparse in the northern region, located east and north
of 149°E and 40°N, respectively, and had to be assumed inthe modeling process In addition, understanding the spa-tiotemporal distributions of the size composition of Pacificsaury stocks in the fishing grounds could increase theeffectiveness of the stock assessment and management andreduce the unnecessary fishing mortality of the less
Fig 1 Spatial distribution of the mean catch per unit effort (CPUE)
for Pacific saury in the fishing season (July–November) during 2000–
2005 in the northwestern Pacific from the Taiwanese saury fishery
with relevant sea surface temperature (SST) The dotted line
represents a boundary of the Exclusive Economic Zone of Japan
1975 1980 1985 1990 1995 2000 2005
Year 0
4
8
12
Fig 2 Annual catches of Pacific saury in the northwestern Pacific
during 1977–2005 from the Taiwanese Pacific saury fishery
Trang 23profitable Pacific saury including the small-sized
individ-uals that are discarded
The objectives of this study are to compare the
varia-tions in the abundance and size composition of Pacific
saury in the monthly aggregation and geographical
distri-bution between the high-seas and coastal fishing grounds of
the NWP in 2000–2005 and to relate those variations to sea
water temperature (SWT) and spatiotemporal variables in
order ultimately to better understand and manage the
fishery
Materials and methods
Data sources
Fishery data of Pacific saury used in this study were
pro-vided by the Overseas Fisheries Development Council,
authorized by the Fisheries Agency, Council of
Agricul-ture, in Taiwan The data were compiled from daily
log-books submitted by skippers of Taiwanese stick-held
dipnet fishing vessels operating in the NWP from 2000 to
2005 There were 24,373 daily vessel records in the
2000–2005 data and their spatial distribution is shown in
Fig.1 The fishing season during 2000–2005 generally
extended from July to November, although there was no
fishing operation in July in 2000 and 2001 and November
2000 for Taiwanese saury fishery The biological
mea-surements of fish body, such as body lengths and weights,
of Pacific saury caught by the Taiwanese fishing vessels
were not collected until 2004 Therefore, only the 2004 and
2005 biological data were available for the Taiwanese data
SWT was measured by a thermometer under the vessel
when fishing was underway Considering that SWT was
recorded only at the place of fishing operations, SST was
additionally used to illustrate the geographical variations in
environmental temperature throughout the NWP SST data
were obtained from the website of the Physical
Oceanog-raphy Distributed Active Archive Center (PO.DAAC)
and derived from the Advanced Very High Resolution
Radiometers (AVHRR) on board the NOAA-series polar
orbiting satellites Digital data of the bathymetric depths in
the NWP were derived from the Centenary Edition of
the General Bathymetric Chart of the Oceans (GEBCO)
Digital Atlas [31]
Conversion of catch from weight categories
to length grades
Catches in the logbooks and fish-landed market of the
Taiwanese Pacific saury fishery are divided by weight into
five categories of commercial packs, including extra-large
(\6 ind./kg), no 1 (6–9 ind./kg), no 2 (9–12 ind./kg), no 3
(12–15 ind./kg), and no 4 ([15 ind./kg) Conversion of thecatch from the five weight categories to the three lengthgroups of large (LS[29 cm), medium (MS 29–24 cm), andsmall sizes (SS\24 cm) was made in this study to allow forcomparison with other studies of Pacific saury, graded bylength, and for speculating on the spawned cohorts.The standards of the contents in the five Taiwanesecommercial pack categories of Pacific saury are set by theTaiwan Squid Fishery Association in order to facilitatesales Therefore, the contents of each commercial packshould not change based on the time, place, and vessels.Nevertheless, two replications of the sampling vessels, one
in 2004 and another in 2005, were used in order to reduce thepossibility of differences in the time, place, or vessels in thisstudy Sixty specimens of Pacific saury were randomlysampled from commercial packs in each of the five weightcategories, in 2004 and 2005, and their knob lengths weremeasured There were, therefore, 600 fish specimens(60 individuals 9 5 size categories 9 2 years) used for theconversion; their sampled time and area were early October
2004 and 2005 and 150–153°E, 41–43°N We assumed thatthe length frequency distribution was a normal distribution
in each weight category, and the distribution was simulated
by the mean and standard deviation of the length Then, ineach weight category, three proportions were divided at thelengths of 24 and 29 cm into the three length grades LS, MS,and SS (Table1) The catch data of 2000–2005 in the log-books were converted to the three length size groups usingthe average division proportions of 2004–2005 for each ofthe five weight categories For example, the catch of the LSgroup (Fig 3b) was estimated by the sum of the proportions,0.976, 0.456, 0.236, 0.023, and 0.000, multiplied by thecatch of the five weight categories, extra-large, no 1, no 2,
no 3, and no 4 in the logbooks, respectively (Table1).Data analysis
Catch per unit effort (CPUE) is used as an index ofabundance in weight for stock assessment of Pacific saury[3,11,29,32], as well as in our study on the basis of themetric tonnes of catches per day per vessel (t/day/vessel).The CPUE and size-composition data were converted into1° mean and transferred onto a geographic grid comprisingcells of 1° latitude by 1° longitude Thus, a 375-grid celldatabase (15 and 25 grid cells in latitude and longitude,respectively) was constructed from 140 to 165°E and 35 to50°N for the CPUE and size composition MapInfo Pro-fessional 6.0 (MapInfo, Troy, NY) was used to create thegeographical distribution of the CPUE and size composi-tions on the SST contour maps, allowing the overlay ofbiological and oceanographic spatial data, in order tovisually analyze the geographical interaction of the seatemperature with the abundance and size-composition
Trang 24distributions of Pacific saury in the NWP Before being
incorporated into the MapInfo software, SST contours were
prepared using the Kriging grid method of the SUFER
software (Golden Software, Golden, CO)
A paired t test was used to determine the significance
level of differences in the monthly CPUEs and size
com-positions of Pacific saury catches between the high-seas
and coastal fishing grounds The 6 years of data
(2000–2005) were regarded as the replicated samples of the
monthly CPUEs and size compositions of Pacific saury
catches The high-seas fishing ground in our study was
defined as being located to the east of the Japanese EEZ
boundary close to 151°E in the NWP, and the coastal
fishing ground was defined as being located to the west of
the EEZ boundary (Fig.1) Coefficients of Spearman’s
correlation were carried out to examine the relationship of
the SWT to the abundance indices (total and three size
CPUEs) of Pacific saury and the spatiotemporal variables
(month, latitude, longitude, and bathymetric depth) Partial
correlation coefficients were then calculated to evaluate
the relationship of the spatiotemporal variables to the
abundance indices while taking away the effects of the
temperature (SWT) on this relationship In order to reduce
the effect of the unequal sample sizes of daily records in
the years of 2000–2005 on the analysis, 7-day-averagevalues in a geographical scale of 1° 9 1° square, loga-rithmically or rank transformed if necessary, were used.There were 154, 217, 289, 152, 249, and 225 averagevalues from 2000 to 2005, respectively, with a total of1,286 The ratio of the largest sample size to the smallestwas thereby reduced from 3.58 to 1.88 Also, the 7-day-average value in the 1° 9 1° square could reduce the bias
of a similar problem that a number of the vessel-dailyrecords were duplicated in a certain time and area, undergood conditions for fishing when the fishing vessels wereaggregating intensely Statistical analyses were performedusing SPSS 12.0 for Windows (SPSS, Chicago, IL)
ResultsVariations in the monthly CPUE, catch, and proportions ofthe three length-size groups in the catch of Pacific sauryfrom the Taiwanese saury fishery in the fishing seasons areshown in Fig.3 The CPUE generally increased from July
to November (Fig.3a), while the highest monthly catchoccurred in September–October (Fig.3b) The LS saurywas dominant (44.3–71.4% of catch) in the first fishingmonth but decreased over time to the end of the fishingseason (5.1–28.1%) (Fig.3c) In contrast, the MS saurywas subordinate (28.4–55.3% of catch) in the first monthand increased throughout the fishing season The MS saurywas most dominant (70.1–92.4% of catch) in November,the late fishing season The SS proportion of Pacific saurywas the least in the catch (0.1–5.6%), and its monthlyvariation pattern was generally consistent with the MSproportion (Fig.3c) Since most of the SS saury propor-tions were too small (\1.0% in the catch of 2000–2005,except for a range of 2.1–5.6% in 2004) to compare withthe large- and medium-sized saury, the SS group was notincluded in the statistical analysis
The monthly total CPUEs of Pacific saury in the seas fishing ground (13.13 ± 1.35, mean ± SE) were sig-nificantly smaller than those in the coastal waters(18.45 ± 2.28) in 2000–2005 (paired t test: t = -3.08,
high-df = 19, p = 0.006) (Fig.4), as were the monthly LSsaury CPUEs (t = -2.51, df = 19, p = 0.021) and themonthly MS saury CPUEs (t = -2.78, df = 19,
p = 0.012) (Table 2) The mean CPUEs of the total catch,
LS saury, and MS saury in the high seas were about 71.2,55.0, and 81.8% of those in the coastal waters, respectively
In terms of the monthly size compositions in the catch, the
LS saury proportions in the high seas (32.85 ± 4.12%)were significantly smaller than those in the coastal waters(37.93 ± 4.72%) (t = -3.06, df = 19, p = 0.006), whilethe MS proportions in the high seas (65.10 ± 3.67%) weresignificantly larger than those in the coastal waters
Table 1 Mean and standard deviations of the knob lengths (KLs) in
the five weight categories of commercial packs for the Taiwanese
Pacific saury fishery in 2004 and 2005, and the proportion in each
weight category converted to the large- (LS), medium- (MS), and
small-sized (SS) groups of length with an assumption of normal
distribution for the length
large
Extra-No 1 No 2 No 3 No 4
Trang 25(60.82 ± 4.57%) (t = 2.70, df = 19, p = 0.014) (Table2;
Fig.5) The mean proportions of the LS and MS saury in
the high-seas catch were about 86.6 and 107.0% of the
coastal values, respectively There was no Pacific saury
fishing activity by the Taiwanese fishing fleets in the
coastal waters before August in 2000–2005
Variations in the geographic distributions of the CPUE
and size composition of Pacific saury in the coastal waters
and high seas from July to November in 2000–2005 are
shown in Fig.6, as is the SST Most saury were found in
the waters where the SST ranged between 10 and 20°C InJuly, most saury were located in the high seas around154–160°E, 41–47°N The LS saury were almost com-pletely dominant throughout the high-seas fishing ground,except for the northern area (about [45°N) where the MSsaury were dominant Subsequently, the saury divided intotwo connected parts In August, some saury aggregated inthe high seas, and the others aggregated southwestwardaway from the cold intruding water (\10°C SST) and kept
to the coastal waters around 145°E and 41°N The LS saury
Monthly mean
Jul Au g Sep Oct Nov
00' - 05' Month
Small size Medium size Large size
0 5 10 15 20 25 30 35
0.0 0.2 0.4 0.6 0.8 1.0
Jul Au g Sep Oct Nov
00' - 05' Month
Small size Medium size Large size
(a)
0 5 10 15 20 25 30 35
Fig 3 Annual and fishing
seasonal changes of Pacific
saury in size structure, by catch
per unit effort (CPUE) (a), catch
(b), and proportion (c), from
2000 to 2005 in the
northwestern Pacific Asterisk
indicates no fishing operation of
Taiwanese saury fishing vessels
in this month
Trang 26generally made up more than 50% of the catch in both the
coastal waters and the high seas In September, the saury
continued to aggregate with greater abundance
southwest-ward to the coastal waters around 143°E and 41°N During
this time, the MS saury, rather than the LS, were dominant
in the fishing grounds, except in the offshore area of the
coastal waters, around 146–150°E and 41–44°N, and the
northeastern area of the high seas, around 155–163°E and
45–48°N, where the LS saury made up more than 50% of
the catch The geographical distribution pattern of Pacific
saury in October was similar to that in September, but with
more abundance and concentrated aggregation, and the
dominant MS saury continuously increased to around 75%
of the catch In November, saury dispersed in the coastal
waters around Hokkaido and the Kuril Islands which had
become occupied by cold water (\10°C SST) The fishing
area of Pacific saury then moved back to the high seas for
the Taiwanese saury fishing fleets The proportion of MS
saury was larger than 75 and 50% in the waters to the west
and east of 152°E, respectively
The SWT was negatively correlated with the total
CPUE and MS CPUE (p \ 0.01, third row in Table3)
Coefficients of the partial correlation indicated that, if thetemperature was adjusted to equal, the total CPUE wascorrelated positively to both the LS (r = 0.786) and MS(r = 0.915) CPUEs (p \ 0.001; Table3) The r value of
MS CPUE was higher than that of LS The fishing monthwas positively correlated to the total and MS CPUE(p \ 0.001), while it was negatively correlated to thefishing latitude, longitude, and bathymetric depth(p \ 0.01) (Table3) The fishing latitude, longitude, andbathymetric depth were negatively correlated with the totaland MS CPUE (p \ 0.01), while only the fishing longitudewas negatively correlated to the LS saury CPUE (p \ 0.05)(Table3)
DiscussionSpatiotemporal differences in saury abundance and sizecomposition between the high seas and coastal waters
In the high-seas fishing ground, the total CPUE (about71.2% of the mean coastal value) of Pacific saury and both
0 10 20 30 40
Fig 4 A comparison of
monthly catch per unit effort
(CPUE) for Pacific saury
between the coastal waters and
high seas in the northwestern
Pacific from July to November
of 2000–2005 based on
Taiwanese fishing data
Table 2 Paired comparisons of monthly CPUEs and size proportions of Pacific saury between the high-seas and coastal fishing grounds in the northwestern Pacific during the fishing season of 2000–2005 (n = 20)
a Ratio of the mean values of the high-seas to the coastal fishing grounds
Trang 27the large- (about 55.0%) and medium-sized saury CPUEs
(about 81.8%) were found to be significantly smaller than
those in the coastal waters in the NWP, particularly in
September and October (Figs.4,6) In July, the large-sizedsaury were almost completely dominant (44.3–71.4% ofcatch) throughout the high-seas fishing ground except for
0.0 0.2 0.4 0.6 0.8 1.0
Small size Medium size Large size (a)
0.0 0.2 0.4 0.6 0.8 1.0
Small size Medium size Large size
0.0 0.2 0.4 0.6 0.8 1.0
monthly proportions of large-,
medium-, and small-sized
groups of Pacific saury between
the coastal waters (a) and high
seas (b) in the northwestern
Pacific from July to November
of 2000–2005 based on
Taiwanese fishing data
Fig 6 Geographical variations in monthly mean catch per unit effort (CPUE) (t/day/vessel) and size compositions of Pacific saury in the northwestern Pacific with sea surface temperature (SST) in fishing seasons of 2000–2005 The size of the circles indicates the CPUE quantiles
Trang 28the northern area, while from September the saury schools
were apparently more abundant and the medium-sized
saury were dominant, instead of the large-sized (Figs.3,6)
Novikov [33] has indicated that the large-sized saury are
the first off the southern Kuril Islands in the spawning
migration This implies that a high proportion of
large-sized saury migrate from the high-seas to the coastal waters
at the beginning of the southwestward migration, while a
high proportion of medium-sized saury follow and
domi-nate from September, even though at that time the school
abundances of both the large- and medium-sized saury
increase in these two waters In the central North Pacific
from July to August, the large-sized saury group was
composed of spawning individuals as well as individuals
ready for re-spawning, while all medium-size saury were
immature [25] Therefore, the large-sized saury mature are
fast and lead the saury spawning schools, including the
immature, from the high seas to the coastal waters, making
the abundance of Pacific saury in the coastal waters higher
than in the high seas
In the coastal waters, large-sized saury were found to be
more abundant in the offshore areas than in the inshore
areas [33] In our study, a similar but more precise result
was found The proportion of the large-sized saury in the
catch was larger in the offshore area around 146–150°E
and 41–44°N in comparison with the inshore areas in
September–October (Fig.6) Novikov [33] indicated that
Pacific saury start their prespawning southward migration
in August–September mainly along inshore and offshore
branches of the Oyashio Current, which shift to the south
during fall, and the offshore branch is the main migration
pathway unless this branch is weak With the progressive
offshore shift of the Oyashio Current, the southwardprespawning migratory routes of the large-sized saury alsomoved more farther from the inshore area [10]
The effect of westward intensification could be one ofthe important factors that make saury more abundant in thecoastal waters than the high-seas fishing grounds Deepcurrents are intensified along western boundaries of theoceans and tend to flow in gyre-like motions within thenorthern and southern basins of each ocean in response tothe forces of the Earth’s rotation and the Coriolis effect[34] The effect of westward intensification yields an SWTgradient, resulting in a clear oceanic front Frontal areas arethe favored conditions for migratory routes of marine fishes[30] During the spawning migration season, saury could beattracted to aggregate intensively at around 15°C SST ofthe frontal areas and move in the direction of the thermalgradients to find spawning grounds with SSTs of around20°C, which are favorable for the survival and growth oftheir offspring [32] Strictly speaking, the high CPUE ofthe coastal waters off Japan found in our study meant thatPacific saury were abundant on the migratory routes in thecoastal areas However, before the prespawning migration,most saury fishing stock is found in the feeding grounds inthe Oyashio waters around Hokkaido and the Kuril Islands.Comparatively, the fishing stock could be low in thetransition zone between the feeding ground and the mainspawning grounds, the Kuroshio waters off southern Japan,but abundant in the migratory routes during the migratoryseason Pacific saury is an oceanic migratory species, andits CPUE in the frontal areas of the migratory routes could
be an abundance index of the fishing stock in the oceanduring the migratory season
Table 3 Coefficients of Spearman’s correlation of the sea water
temperature (SWT) and coefficients of partial correlation of the fixed
SWT to and between the abundance indices and spatiotemporal
variables, including total CPUE (CPUE), large-size CPUE (LS), and
medium-size CPUE (MS) of Pacific saury; month; latitude (LAT); longitude (LONG); and bathymetric depth (BD) in the northwestern Pacific during 2000–2005, based on the 7-day-average values at a geographical scale of 1°9 1° square (n = 1,286)
CPUE (t/day/vessel) LS (t/day/vessel) MS (t/day/vessel) Month LONG (°E) LAT (°N) BD (m) Spearman’s correlation
Trang 29Relationship of abundance and size composition
of saury
In the analysis of length composition, two distinct groups
of Pacific saury were generally found, namely large- and
medium-sized [2,13,35], although the saury caught during
the fishing season have previously been divided into either
three or four length groups [13,24] It is reasonable to find
that both the large- and medium-sized abundance indices
(CPUEs) of Pacific saury significantly contribute to the
total abundance index (Table3) However, in our study the
variations of the medium-sized abundance index were
found to be more highly consistent with the total
abun-dance index The hatch period of the medium-sized saury
in the fishing season ranged from the previous autumn to
winter with their otolith increments ranging from ca 220 to
270, estimated by the growth functions of this species [36,
37] Saury spawns with a peak in winter, and the winter
cohort plays an important role in recruitment [18]
Recently, most Pacific saury documents agree that the
medium- and large-sized groups are 0? and 1? year
classes, respectively [2,3] The latest document of the age
determination of Pacific saury also indicates that the
medium- and small-sized groups closely correspond to the
0? age group, while the large-sized group corresponds to
the 1? age group [13] The abundance of age 0? fish was
found to correlate positively with the abundance of age 1?
fish in the next year (Y Ueno, 2008, personal
communi-cation) In our study, on average, the medium-sized saury
comprised more than 60% of the fishing stock (Table2)
and had a high correlation to the total abundance of Pacific
saury fishing stock (Table3) However, the catches of the
small-sized saury were very small by the commercial
fishing vessels, possibly due to low price and/or the net
mesh size, thus their abundance is unclear but could be
large in the ocean
Environmental effects on the spatiotemporal variations
of saury abundance and size composition in the high
seas and coastal waters
Fish stocks are known to fluctuate extensively over a large
range of spatial and temporal scales, and several biotic and
abiotic processes, as well as their interactions, may induce
such fluctuations [38] Environmental changes, such as
variations in temperature, salinity, wind field, and currents,
can affect both the productivity and the distribution of fish
stocks [39, 40] Temperatures associated with
oceano-graphic conditions have been found to be strongly linked to
the migration route, distribution, and abundances of Pacific
saury fishing stocks in the NWP, which results in annual
variations in saury abundances and beginning times of the
fishing season [3, 7, 28, 30, 32, 41] In our study, the
monthly distribution pattern (Fig.6) was consistent with themigration pattern in previous studies from Japan [3,11,15].Some saury migrated southwestward to the coastal waters inAugust–September keeping away from the colder intrusivewater with SST \10°C, and in November the saury dis-persed in the coastal waters around Hokkaido and the KurilIslands, which had become occupied by cold water with SST
\10°C (Fig.6) The SSTs below 10°C and above 20°Cwere the threshold temperatures of the extremely cold andwarm SST waters, which correlate closely with the closure
of the fishing season and a hard barrier blocking thesouthward migration of the saury, respectively [28].Pacific saury prefer cold waters as migration routes, andthe first and second Oyashio intrusions are importantsoutherly migration routes for the species [30] The fishinggrounds of Pacific saury are formed in the Oyashio water,along the oceanic front between the Oyashio cold watersand warm waters originating from the Kuroshio [28,30] Incontrast, most SWTs [15°C in the fishing grounds were anindicator of low stock abundances for the Pacific saury[32] In our study, the SWT was also found to correlatenegatively with the abundance indices (CPUEs) of the totalcatch and medium-sized Pacific saury (Table3), indicatingthat the migrating Pacific saury, particularly the medium-sized saury, favor the cooler temperature areas of thefishing grounds However, the negative relationship of theCPUE and SWT was significant (p \ 0.05) but weak(r2\ 0.01), and more data are needed to further verify thisrelationship in future studies
In addition to temperature, our study particularly ined the effects of geographic factors, including the fishinglatitude, longitude, and bathymetric depth, on the variations
exam-of abundance distributions exam-of Pacific saury when the perature effect was statistically held equal Then, the fishinglatitude, longitude, and bathymetric depth were found tocorrelate negatively with the abundance indices of the totalcatch and medium-sized Pacific saury, while the abundanceindex of the large-sized Pacific saury only correlated neg-atively with the fishing longitude (Table 3) This indicatesthat if the temperature was the same, the migratory aggre-gation schools of Pacific saury, particularly the medium-sized saury, would be more abundant in the westward,southward, and shallow waters of the fishing grounds, whilethe large-sized schools would be abundant in the westwardwaters Planktonic food resources are one of the importantfactors guiding the annual migration route of Pacific saury[16,17] Fukushima et al [42] also indicated that the geo-graphical distribution of saury schools for the southwardmigration coincided with the appearance of dense zoo-plankton zones in the NWP Generally, an open ocean has alower net primary productivity than the coastal watersincluding the continental shelf, upwelling zones and estu-aries [43,44] The proximity to land and the shallowness of
Trang 30the waters ensure productivity in the continental shelf
region; in contrast, the open ocean is a more homogeneous
environment with low rates of primary production [45] The
westward and shallow-water migratory pattern of Pacific
saury schools in the NWP could result from the higher
primary productivity in the coastal areas than the high-seas
to ensure a sufficient supply of food for their offspring and/
or themselves In addition, the southward migratory pattern
of the medium-sized saury, which are immature at the
beginning of the spawning migratory season in the central
North Pacific [25], could imply that the medium-sized saury
need time to mature during the spawning migration and then
migrate distantly to the winter spawning areas in the
Ku-roshio waters off southern Japan
From the sustainable utilization point of view, the
Pacific saury fishery should be managed The main
spawning group is large fish of age 1? Based on the
findings in our study, a high proportion of the large-sized
saury migrates first in the early stage (July–August) of the
fishing season when there is a comparatively low CPUE
The fishing effort at this stage should not be large to allow
the main spawning individuals to migrate safely southward
to the main spawning ground in the Kuroshio Current
waters off southern Japan In the middle stage (September–
October) of the fishing season, the fishing effort could be
allowed to increase and the catch per effort (fishing
effi-ciency) is good at this stage
In fact, a similar fishery management practice has
effectively been carried out in Japan for many years
Small fishing vessels are allowed to catch Pacific saury in
the early stage of the fishing season, while the large
fishing vessels are only allowed to fish, generally, after
about mid-August However, the management of the
Taiwanese saury fishery could be different from Japan,
since the saury fishing vessels from Taiwan are very close
in size and large (700–900 gross tons) Limiting the
Taiwanese saury fishing vessels operating in the NWP to
a certain number is a way to reduce the fishing effort in
the early stages of the fishing season In the middle stage
of the fishing season, starting about late-August, the
fishing could be opened for all saury fishing vessels from
Taiwan if the stock is in a good condition In addition, at
this stage, it is economical for Taiwanese fishing vessels
to operate on the high-seas in abundant years due to the
large number of fish and lower costs In contrast, to
increase their catch in average and less abundant years,
the Taiwanese fishing vessels could choose, depending on
the profitability, to cooperate with Japan or Russia to fish
in the coastal waters where the saury abundance is larger
than the high-seas
Acknowledgments This research was funded by the Fisheries
Agency, Council of Agriculture, Executive Yuan, and the National
Science Council, Taiwan (95AS-14.1.2-FA-F1(3), NSC 026-004, 96AS-15.1.2-FA-F2) We are very grateful to the Overseas Fisheries Development Council in Taiwan for providing logbook data from the Taiwan saury fishery for this study We also wish to thank two anonymous reviews for their constructive comments on the manuscript.
95-2511-S-References
1 Fukushima S (1979) Synoptic analysis of migration and fishing conditions of saury in northwest Pacific Ocean Bull Tohoku Reg Fish Res Lab 41:1–70
2 Kosaka S (2000) Life history of Pacific saury Cololabis saira in the Northwest Pacific and consideration of resource fluctuation based on it Bull Tohoku Natl Fish Res Inst 63:1–96 (in Japanese with English abstract)
3 Tian Y, Akamine T, Suda M (2003) Variations in the abundance
of Pacific saury (Cololabis saira) from the northwestern Pacific in relation to oceanic-climate changes Fish Res 60:439–454
4 FAO (Food and Agriculture Organization of the United Nations) (2006) Fisheries productions statistics (databases and statistics).
http://www.fao.org/fi/statist/statist.asp
5 Kimura K (1956) The standard length of the Pacific saury, olabis saira (BREVOORT) Bull Tohoku Fish Lab 7:1–11 (in Japanese with English abstract)
Col-6 Oyamada S, Ueno Y, Makino M, Kotani K, Matsuda H (2009) Bioeconomic assessment of size separators in Pacific saury fish- ery Fish Sci 75:273–283
7 Yasuda I, Watanabe Y (1994) On the relationship between the Oyashio front and saury fishing grounds in the north-western Pacific: a forecasting method for fishing ground locations Fish Oceanogr 3:172–181
8 Fukusima S (1958) Relation between the fishing of the Pacific saury and the oceanographical conditions in the northeastern sea area along the Pacific coast of Japan Bull Tohoku Natl Fish Res Inst 12:1–27 (in Japanese with English abstract)
9 Hotta H (1962) Fluctuations in the abundance of saury on the north-eastern sea area in Japan (I) Bull Tohoku Natl Fish Res Inst 21:1–20 (in Japanese with English abstract)
10 Novikov Y (1982) Some notions of mechanism of long-term variations of stock composition and abundance of Pacific saury Bull Tohoku Reg Fish Res Lab 44:101–107
11 Tian Y, Akamine T, Suda M (2004) Modeling the influence of oceanic-climatic changes on the dynamics of Pacific saury in the northwestern Pacific using a life-cycle model Fish Oceanogr 13(suppl 1):125–137
12 Hubbs CL, Wisner RL (1980) Revision of the sauries (Pisces, Scomberesocidae) with descriptions of two new genera and one new species Fish Bull US 77:521–566
13 Suyama S, Kurita Y, Ueno Y (2006) Age structure of Pacific saury Cololabis saira based on observations of the hyaline zones in the otolith and length frequency distributions Fish Sci 72:742–749
14 Suyama S (2002) Study on the age, growth, and maturation process of Pacific saury Cololabis saira (Brevoort) in the North Pacific Bull Fish Res Agency 5:68–113
15 Ito S, Sugisaki H, Tsuda A, Yamamura O, Okuda K (2004) Contributions of the VENFISH program: meso-zooplankton, Pacific saury (Cololabis saira) and walleye pollock (Theragra chalcogramma) in the northwestern Pacific Fish Oceanogr 13(suppl 1):1–9
16 Odate K (1994) Zooplankton biomass and its long-term variation
in the western North Pacific Ocean, Tohoku sea area, Japan Bull Tohoku Natl Fish Res Inst 56:115–163 (in Japanese with English abstract)
Trang 3117 Sugisaki H, Kurita Y (2004) Daily rhythm and seasonal variation
of feeding habit of Pacific saury (Cololabis saira) Fish Oceanogr
13(suppl 1):63–73
18 Kurita Y (2001) Seasonal changes in spawning grounds and the
abundance of egg-laying of Pacific saury In: 49th Annual report
of the research meeting on saury resource Tohoku National
Fisheries Research Institute, Aomori, pp 203–205 (in Japanese)
19 Yamamura O (1997) Scavenging on discarded saury by demersal
fishes off Sendai Bay, northern Japan J Fish Biol 50:919–925
20 Chikuni S (1985) The fish resources of the northwest Pacific.
FAO Fish Tech Pap No 266 FAO, Rome
21 Tamura T, Fujise Y, Shimazaki K (1998) Diet of minke whales
Balaenoptera acutorostrata in the northwestern part of the North
Pacific in summer, 1994 and 1995 Fish Sci 64:71–76
22 Springer AM, Piattt JF, Shuntov VP, Van Vliet GB, Vladimirov
VL, Kuzin AE, Perlov AS (1999) Marine birds and mammals of
the Pacific Subarctic Gyres Prog Ocenogr 43:443–487
23 Huang WB, Chiu TS (1997) Effects of stocking density on
sur-vival, growth, size variation, and production of tilapia fry Aquac
Res 28:165–173
24 Fukushima S, Watanabe Y, Ogawa Y (1990) Correspondence of
spawned season to large, medium, and small size Pacific saury
exploited in the northwestern Pacific Ocean Bull Tohoku Natl
Fish Res Inst 52:17–27 (in Japanese with English abstract)
25 Suyama S, Sakurai Y, Shimazaki K (1996) Age and growth of
Pacific saury Cololabis saira (Brevoort) in the western North
Pacific Ocean estimated from daily otolith growth increments.
Fish Sci 62:1–7
26 Anonymous (1991) Progress and summary of saury fishing In:
39th Annual reports of the research meeting on saury resources.
Tohoku National Research Institute, Shiogama, pp 3–136
(in Japanese)
27 Ueno Y, Honma T (2002) Estimating the effect of fish sorting
machine for Saury In: 50th Annual report of the research meeting
on saury resource Tohoku National Fisheries Research Institute,
Aomori, pp 244–255 (in Japanese)
28 Watanabe K, Tanaka E, Yamada S, Kitakado T (2006) Spatial
and temporal migration modeling for stock of Pacific saury
Cololabis saira (Brevoort), incorporating effect of sea surface
temperature Fish Sci 72:1153–1165
29 Hofmann EE, Powell TM (1998) Environmental variability
effects on marine fisheries: four case histories Ecol Appl 8(suppl
1):S23–S32
30 Saitoh S, Kosaka S, Iisaka J (1986) Satellite infrared observations
of Kuroshio warm core rings and their application to study of
Pacific saury migration Deep Sea Res 33:1601–1615
31 IOC, IHO, BODC (2003) Centenary edition of the GEBCO digital atlas, published on CD-ROM on behalf of the intergov- ernmental oceanographic commission and the international hydrographic organization as part of the general bathymetric chart of the oceans British Oceanographic Data Centre, Liver- pool, UK
32 Huang WB, Lo NCH, Chiu TS, Chen CS (2007) Geographical distribution and abundance of Pacific saury fishing stock in the Northwestern Pacific in relation to sea temperature Zool Stud 46:705–716
33 Novikov GF (1967) The main biological features and stocks of Pacific saury Izv Tikhook NI Inst Rybn Khoz Okeanogr 56:3–50 (in Russian)
34 Segar DA (1998) Introduction to ocean sciences Wadsworth, Belmont
35 Matsumiya Y, Tanaka S (1974) Considerations on the so-called large- and intermediated-sized fish of saury on the basis of the analysis of the length composition Bull Tohoku Reg Fish Res Lab 33:1–18 (in Japanese with English abstract)
36 Watanabe Y, Butler JL, Mori T (1988) Growth of Pacific saury, Cololabis saira, in the northeastern and northwestern Pacific Ocean Fish Bull US 86:489–498
37 Nemoto Y, Kurita Y, Oozeki Y, Honma T, Hyashizaki K, Ida H (2001) Otolith microstructure of Pacific saury, Cololabis saira Bull Tohoku Natl Fish Res Inst 64:69–78
38 Cushing DH (1982) Climate and fisheries Academic Press, London
39 Southward AJ, Boalch GT, Maddock L (1988) Fluctuations in the herring and pilchard fisheries of Devon and Cornwall linked to change in climate since the 16th century J Mar Biol Assoc UK 68:423–445
40 Alheit J, Hagen E (1997) Long-term climate forcing of European herring and sardine populations Fish Oceanogr 6:130–139
41 Yasuda I, Kitagawa D (1996) Locations of early fishing grounds
of saury in the north-western Pacific Fish Oceanogr 5:63–69
42 Fukushima S, Odate K, Takahashi S (1976) Relationship in the distribution between zooplankton as prey organisms and Pacific saury in the feeding area of the northwestern North Pacific Bull Tohoku Reg Fish Res Lab 36:1–8
43 Whittaker RH, Likens GE (1975) The biosphere and man In: Lieth H, Whittaker RH (eds) Primary productivity of the bio- sphere Springer, New York, pp 305–328
44 Kennish MJ (1994) Practical handbook of marine science, 2nd edn CRC Press, Boca Raton
45 Wootton RJ (1992) Fish ecology Chapman and Hall, New York
Trang 32O R I G I N A L A R T I C L E Biology
Reproductive biology of the commercially and recreationally
important cobia Rachycentron canadum in northeastern Australia
Tonya D van der Velde•Shane P Griffiths•
Gary C Fry
Received: 8 September 2008 / Accepted: 8 September 2009 / Published online: 17 November 2009
Ó The Japanese Society of Fisheries Science 2009
Abstract The reproductive biology of 315 cobia,
Rachycentron canadum, from northeastern Australia was
studied for an 18-month period Cobia ranged from 181 to
1,470 mm FL (0.06–55 kg) Length–frequency
distribu-tions of males and females did not differ significantly The
sex ratio of females to males was 2.18:1 Histological data
showed that cobia developed hydrated oocytes during a
protracted spawning season between September and June
Gonadosomatic index peaked from October to December,
coinciding with the monsoon or ‘‘wet’’ season Estimated
length at first maturity for female cobia was 671 mm FL
Length at 50% maturity (L50) for females was estimated at
784 mm FL (1–2 years of age) Batch fecundity ranged
from 577,468 to 7,372,283 eggs with a mean of 2,877,669
(± SD 1,603,760) eggs Relative batch fecundity was 249
eggs per g, and no relationship between relative fecundity
and fork length was found There was a significant positive
relationship between the total number of eggs produced
and fork length Spawning frequency, estimated by the
post-ovulatory follicle method, was 7.6 days Based on the
detection of hydrated oocytes in fish caught at night, cobia
most likely spawn at night Cobia also feed throughout the
spawning period This is the first report on the reproductive
biology of cobia in Australian waters, and provides
valu-able data for future population assessments of cobia
throughout the Indo-Pacific
Keywords Black kingfish Cobia Fecundity
Maturity Reproduction
IntroductionCobia, or black kingfish, Rachycentron canadum is acommercially and recreationally important species that iswidely distributed throughout the neritic regimes of almostall tropical, subtropical and warm temperate waters of theworld, except for the eastern Pacific Cobia are the onlymember of the Rachycentridae family and grow to amaximum length of 2 m TL and a weight of 68 kg, andhave a reported maximum age of 15 years [1] (seehttp://www.fishbase.org/Summary/SpeciesSummary.php?id=3542,accessed 15 July 2007)
Cobia are generally solitary fish that occasionally occur insmall schools or ‘‘pods’’ that are often associated withfloating objects such as fish-attracting devices, buoys andlogs, fixed structures such as oil rigs, piers and jetties [2], andlarge oceanic animals such as sharks, rays, turtles andwhales They are often referred to as a pelagic species,although dietary studies performed both in Australia [3,4]and the United States [5 8] reveal that a large portion of theirdiet consists of benthic and demersal prey, including crabs,stingrays, flatfishes and stomatopods As a result, cobia may
be a key contributor to the transfer of energy between thepelagic and benthic communities in tropical ecosystems.Cobia are relatively common but are rarely encountered
in large numbers As a result, they are rarely a target ofcommercial fishers, but are a valued incidental catch, par-ticularly in the hook and line fisheries of the United States[1] The total global catch has steadily increased since 1981from 2,300 t to 11,000 t in 2000 In Australia, cobia are notcurrently a primary target species of any state or Com-monwealth fishery, with total reported landings \30 t.However, cobia are a valuable bycatch in a number of hookand line and gillnet fisheries, and juveniles are often a dis-carded bycatch in trawl fisheries in northern Australia [3,9]
T D van der Velde (&) S P Griffiths G C Fry
CSIRO Marine and Atmospheric Research,
PO Box 120, Cleveland, QLD 4163, Australia
e-mail: tonya.vandervelde@csiro.au
DOI 10.1007/s12562-009-0177-y
Trang 33The large size, fighting ability and superb eating
quali-ties of cobia have made the species an important sportfish
in Australia and the United States In the Gulf of Mexico,
catches of cobia by the recreational fishery (1,008 t) far
exceed the commercial sector (80 t) (National Marine
Fisheries Service, Fisheries Statistics Division, Silver
Spring MD, 2008, personal communication) Similarly, in
Australia, cobia is an important component of the
recrea-tional and charter boat catch in northern Australia Because
of its fast growth rates—reaching 6–10 kg in 12–
14 months [10,11]—and superb flesh quality [10], cobia is
also an excellent candidate for aquaculture
Despite the importance of cobia to many commercial
and recreational fisheries worldwide, few studies have
investigated the biology or ecology of the species in order
to provide useful data for stock assessment and
manage-ment Most of the biological studies on cobia have been
conducted on the southeastern coast of the United States,
and in particular the Gulf of Mexico This research has
focused on age and growth [1, 5, 11,12], descriptions of
eggs, larvae and juveniles [13, 14], and reproductive
biology [15–17] These studies have shown that cobia are
fast-growing and early-maturing batch spawners and can
produce up to 2 million eggs per spawning However, these
studies have not estimated length or age at maturity, which
has been highlighted as a key deficiency in cobia stock
assessment [18]
The present paper reports the first reproductive biologystudy of cobia in Australian waters aimed at providing datafor stock assessment The specific aims of the study were(1) to determine the sex ratio of the population, (2) toinvestigate the timing of spawning using histology and agonadosomatic index (GSI), (3) to estimate the length andage at sexual maturity from histological data, and (4) toestimate the batch fecundity of mature females
MethodsCollection of specimensMonthly collections of cobia were made opportunisticallybetween January 2003 and September 2005 from a broadregion in northeastern Australia ranging from the easternGulf of Carpentaria (GoC) to southeastern Queensland(SEQ) using a number of methods (Fig.1) Cobia weretreated as a single stock in this study, since anecdotal evi-dence and tagging data have shown that fish move exten-sively in coastal regions in both eastern Australia and theeastern United States [19] The fish were collected usingthree main methods: from commercial gillnet vessels inQueensland’s N9 fishery, from sportfishing tournaments,and from fishery-independent collections using hook andline Some fish collected in the commercial gillnets and
Fig 1 Map showing the
northeastern Australian region
where cobia samples were
collected during the period
January 2003 to September
2005
Trang 34sportfishing tournaments were collected as fish frames
where no whole weights were obtained The length–weight
relationship for the fish caught in the fishery-independent
collections was used to assign weights to these fish Since
gillnets are selective for larger-size fish and specimens from
the sportfishing sector are regulated by a minimum legal
length of 750 mm TL, juveniles were opportunistically
collected from scientific trawl surveys conducted, in the
study region, by CSIRO Marine and Atmospheric Research
After capture, specimens were kept on ice until they
could be frozen and freighted to the laboratory Frozen fish
were then thawed, weighed (0.01 g), measured for their fork
and total length (FL and TL in mm), and initially sexed by
macroscopic examination of the gonads Only fork length is
given in this paper unless otherwise stated The stomach
was removed, including all food contents The total wet
weight of the stomach contents of each individual was used
to relate feeding intensity to reproductive activity A full
description of the diet is not provided here (S Griffiths and
G Fry, 2008, unpublished data) Both thawed gonad lobes
were also removed, weighed (±0.001 g), trimmed of fat,
and placed in a labeled perforated plastic bag Gonads were
then fixed in a 10% formaldehyde solution for at least
14 days and stored in 70% ethanol until histological
anal-ysis Because the tissue was left for extensive periods before
processing, we found that transferring to 70% ethanol
enhanced the structural condition of the preserved cells and
made them less brittle during sectioning
Reproductive dynamics
Reproductive activity was assessed using a gonadosomatic
index (GSI) and histology GSI was determined for each
fish, for both males and females, using the equation
whole body wtðg) gonad wt ðg)
100
All immature fish were excluded from the analysis of
GSI data to describe peak spawning activity To investigate
the intensity of feeding with respect to reproductive
activity, a quantitative index of feeding intensity was
obtained by dividing the wet weight of the stomach
contents (0.01 g) by the wet weight of the eviscerated fish
Monthly changes in the stomach fullness index were then
compared with monthly GSI data
Length at first maturity (LMAT) for both males and
females was determined by plotting fish length against GSI,
and the length at which the greatest increase in GSI was
observed was deemed the length at first maturity Fish were
grouped into 100 mm length intervals, and the proportion of
mature fish in each size class was calculated Length at 50%
maturity (L50) was estimated using histology for females
only, which is the length at which 50% of females had
mature ovaries (i.e stages IV or V) The following logisticfunction was then fitted to the data to determine L50:Proportion of mature females
1þ expðK½ðfork lengthÞL50Þwhere K is the curvature of the function The value for L50was then substituted into the von Bertalanffy growthfunction of Fry and Griffiths [11] to provide an estimate ofage at 50% maturity (A50)
HistologyFor the histological examination of gonads, a subsample ofapproximately 1 g was taken from the middle of the ovary,placed in a histological cassette, infiltrated with paraffin, and
sectioned at 6 lm At least three sections were taken from one
lobe of the ovary and together they were mounted and stainedwith Harris’ hematoxylin and eosin counter-stain Onlyfemales were examined since (1) ovary development is moreindicative of spawning activity and (2) it is generally only thefemale reproductive parameters that are considered in stockassessment models, because female energy investment intoreproduction is usually higher than for males [20,21] Ova-ries were staged according to the most advanced group ofoocytes present using the methods of Cyrus and Blaber [22]and Blaber et al [23] as guidelines: unyolked (stage I), earlyyolked (stage II), advanced yolked (stage III), migratorynucleus (stage IV), fully mature/hydrated (stage V), and spent(stage VI) Mature fish are referred to as being stage IV to VI.Seasonal variation in reproductive activity was examinedusing GSI and histological data by plotting mean GSIs and theproportion of mature females across months, respectively.Batch fecundity and spawning frequency
Batch fecundity was estimated from ovary subsamplescontaining final oocyte maturation (FOM) and hydratedoocytes (stages IV and V) One subsample (0.4–0.6 g) perfemale was weighed (±0.001 g), oocytes were teased apartfrom connective tissue, and the number of hydrated ripeoocytes was counted under a stereomicroscope Fecundity(F) was calculated by the formula
F¼ gonad wt ðg) subsample egg count
gonad sumbsample wtðg)
Fecundity estimates from all samples were averaged toprovide a mean estimate of batch fecundity for the species.Relative fecundity estimates were calculated by dividingfecundity by body weight
Spawning frequency of females was estimated by thepost-ovulatory follicle (POF) method of Hunter andMacewicz [24] This method uses the incidence of mature
Trang 35females with POF \24 h old to define the fraction of the
population spawning
Sex ratio
The sex ratio was calculated by using the number of males
and females caught pooled across months for the entire
study A chi-square test was used to determine if the sex
ratio was significantly different to the expected ratio of 1:1
A Kolmogorov–Smirnov (K–S) test was used to determine
statistical differences in length–frequency distributions
between sexes [25]
Results
A total of 315 cobia were examined, ranging from 125 to
1,633 mm FL and 0.06 to 55 kg (Table1) No differences
in external morphology between sexes were identified
Males (n = 93) ranged from 125 to 1,280 mm FL and from
0.06 to 25 kg, while females (n = 203) ranged from 181 to1,470 mm FL and 0.03 to 38 kg (Fig.2) The length–weight relationship for 98 fish that were measured andweighed for both sexes combined was:
Total wet body weight
¼ 9:7 107FL3:3466 n¼ 98; r2¼ 0:9897Sex could not be determined for 19 fish (199–1,633 mmFL), as their gonads were either immature or removed prior
to sample collection Length–frequency distributions ofmales and females were not significantly different (K–Stest D = 0.167; P = 0.051) (Fig 2) The sex ratio offemales to males was 2.18:1 and was significantly different
to the expected 1:1 (chi-square df = 1; P \ 0.05).Gonad development
We examined histological sections from either the left orright lobes of the ovaries of 184 females The possibility of
Table 1 Summary of collection
methods and dates, numbers of
fish, fork length (mm) and
weight (g) ranges for cobia
increments) for male and female
cobia Rachycentron canadum
captured using gillnet and rod
and line in northeastern
Australia between January 2003
and September 2005
Trang 36a difference in oocyte development depending on the
location inside each lobe was examined by comparing
sections from three positions (anterior, middle, and
pos-terior) in both right and left lobes from 20 randomly
selected females The females used in this analysis
inclu-ded ovaries ranging from immature to fully mature As
observed in cobia from Central Mexico [15], no obvious
difference in oocyte development was detected between the
locations of each ovary Therefore, we concluded that any
section from the gonad provided a good indication of its
development stage All stages of development were found
amongst the samples, including hydrated oocytes and POF
in the ovaries of fish with fully mature vitellogenic oocytes,
which confirms that cobia are a multiple-spawning species
in Australia (Fig 3)
Length and age at maturityHistological data indicated that the smallest mature femalewas 671 mm FL This female had a GSI of 2.4% and wasprobably about 1 year of age Mean length at 50% maturity(L50) for females was estimated at 784 mm FL (Fig.4).The estimate of LMAT for females using GSIs was higherthan determined by histology: about 770 mm FL GSI datashowed that males reach LMATat about the same size andage as females; 770 mm FL and 1 year of age (Fig.5)
Fig 3 Histological sections of
ovaries of cobia Rachycentron
canadum showing oocytes at
different stages of development.
a Immature ovary containing
oogonia (O); b immature ovary
containing perinuclear (P) and
cortical alveolar (CA) oocytes;
c actively spawning ovary
containing vitellogenic oocytes
(V) and post-ovulatory follicles
(POF); d actively spawning
ovary containing hydrated
oocytes (HO)
Fig 4 Proportion of mature
female cobia Rachycentron
canadum in each 100 mm size
interval captured during the
spawning season The trendline
is a logistic curve fitted to the
data Dashed line indicates the
length at which 50% of the fish
are considered mature (L50), as
determined by histology
Trang 37Using the length-at-age growth curve from Fry and
Grif-fiths [11], this translates to an age of 1–2 years
Spawning season
GSIs were calculated for 217 females and 90 males, and
histology was undertaken on 184 females Mean monthly
GSIs reveal that cobia have a protracted spawning season
between September and June and display their lowest
reproductive activity in July and August (Fig.6) Mean
monthly GSIs were highest during October and November:
4.7 and 5%, respectively Among the females with highly
developed ovaries (stage V), GSIs were between 1.95 and
6.67% of body weight, with a maximum GSI of 11.66%
(Fig.5) Male and female GSIs showed similar trends
(Fig.6)
Results of histology closely mirror GSI data and showpeak spawning activity between September and May,although some ripe fish were present every month (Fig.7).Specimens of sexually mature fish could not be obtainedfor February A total of nine specimens (805–1,232 mmFL) showed evidence of spent tissue indicating recentspawning prior to capture These fish were caught inMarch, May, August, September and December
Fecundity and spawning frequency
A total of 64 female gonads were suitable for assessingbatch fecundity The estimated total number of eggs pro-duced by individual fish ranged between 577,468 and7,372,283, and the average batch fecundity was 2,877,669(± SD 1,603,760) eggs There was a strong positive rela-tionship between the estimated total number of eggs pro-duced versus fork length (Fig.8) Relative batch fecundity
0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500
Fork length (mm)
0 1 2 3 4 5 6 7 8 9 10 11 12
Females Males
Fig 5 Plot of gonadosomatic
index and fork length (mm) for
male and female cobia
Rachycentron canadum
collected in northeastern
Australia between January 2003
and September 2005 Solid and
dashed lines indicate estimated
length at first maturity (LMAT)
for males and females,
Fig 6 Mean (± SE) monthly gonadosomatic indices for male and
female cobia Rachycentron canadum collected in northeastern
Australia between January 2003 and September 2005 Immature fish
below LMAThave been excluded
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Month
0 10 20 30 40 50 60 70 80 90
5 19
17 19
12
26 17 28
7 18
6
Fig 7 Monthly percentages of female cobia Rachycentron canadum
in spawning condition (histological stages IV–VI) in northeastern Australia between January 2003 and September 2005 Numbers above bars show the sample size
Trang 38was 249 eggs per g (± SD 68.76) No relationship was
found between relative fecundity and fork length (Fig.9)
The ovaries of the nine females sampled showed
evi-dence of recent spawning activity The fraction of mature
females with POF was 0.134, giving a mean spawning
interval of 7.6 days
Discussion
Gonad development
Histological examination is the most accurate way of
determining the maturation status of fish [20] Although
histological analysis is more laborious and expensive than
using GSIs or macroscopic staging, it is the most
appro-priate method for estimating maturity because
post-spawning or resting females may be misclassified as
immature using macroscopic staging [26], and this can
result in a less accurate determination of mean length at
maturity Histology can also detect POF, which is used todetermine the number of spawning events throughout thespawning season [27] This study found that cobia produceoocytes asynchronously, since each ovary examined con-tained oocytes at various stages of development This wasalso recorded in reproductive studies of cobia in the Gulf ofMexico [15,16,28] However, ovaries could be assigned tospecific categories based on the dominant oocytes present.The maturation classification used here was divided intofive development categories and one post-developmentcategory (VI) This classification regime is similar to that
of Lotz el al [15], who determined four maturation gories with the exclusion of hydrated oocytes as a separatestage FOM and hydrated oocytes are significant develop-mental stages that indicate spawning is imminent
cate-Very little is known about the population dynamics andfishery for cobia in Australia This is the apparent firststudy, to our knowledge, of this species in Australia, where
it is becoming an increasingly popular commercial, ational and aquacultural species The information from thisstudy can be used for stock assessments In this context, wepurposely did not include the very early development stage
recre-of maturation (stage III) in the estimation recre-of L50, althoughthis stage is usually classed as a vitellogenic oocyte Thiswas done as a precautionary measure so as to obtain aconservative estimate
Samples for this study were collected every month,which enabled the detection of all maturation stagesincluding hydrated oocytes Given that other studies failed
to find hydrated oocytes [15, 16], it is likely that thisdevelopment stage is very short and is probably missed inthe opportunistic sampling of commercial and sportfishingcatches This could be substantiated by future work thatintensively sample cobia over a 24-h period during peakspawning periods Another explanation could be stressassociated with hooking a fish during fishing operationsand prompting the fish to prematurely release oocytes Animportant observation was that most of the hydrated eggsdetected in this study were derived from fish caught atnight, which suggests that this species spawns at night.Lotz et al [15] suggested that the lack of fish with hydratedeggs in their study was because spawning cobia do not feedand therefore were not able to be captured in large numbersusing their hook and line method Findings from thepresent study show conclusively that cobia do feedthroughout the spawning period (see Fig.10)
There was no obvious relationship between monthlymean GSI and mean stomach fullness index (SFI)(Fig.10) There was also no significant correlation betweenGSI and SFI pooled across months (r2= 0.575;
P = 0.848) Of the 64 fish that were in spawning condition(stage V), only 11 fish had empty stomachs, while 19 fishhad high stomach fullness index values above 2.0 These
Fig 8 Relationship between batch fecundity and fork length (mm)
for female cobia Rachycentron canadum from northeastern Australia.
Only fish with gonads of late stage IV and V were examined (FOM
and hydrated oocytes)
Fig 9 Relationship between relative fecundity and fork length (mm)
for female cobia Rachycentron canadum from northeastern Australia.
Only fish with gonads of late stage IV and V were examined
Trang 39results indicate that reproductive activity has no significant
effect on feeding intensity by cobia
Length at maturity
Cobia are fast-growing and reach sexual maturity early in
life [11] Histological data indicate that the smallest mature
female was 671 mm FL, which is around 1 year of age
[11] When using the more accepted determination of
maturity of L50, which is the length at which 50% of the
population is mature, cobia reach maturity at 784 mm FL
(or an age of about 2 years) This is apparently the first
study of cobia where L50 has been determined Therefore,
comparisons with other studies can only be made in
rela-tion to the smallest mature female detected However,
caution must be exercised here, as it is often not an
accu-rate representation of first maturity, as these fishes may
represent outliers in the population that—for various
rea-sons—may mature earlier
Comparing the results of the present study with those
from other cobia studies, fish in Australia appear to mature
at a similar length From this study we could determine that
females appear to reach earliest maturity in their third year,
at 696 mm FL and 3.27 kg The smallest mature female
from Chesapeake Bay, USA was 696 mm FL [12], and
from the North Carolina Atlantic Coast was 700 mm FL
[6], similar to the smallest female (671 mm FL) found in
Australia with mature oocytes In the Indian Ocean, Rajan
et al [29] collected a 426 mm TL female with ovaries in
the third stage of maturity In contrast, length at first
maturity estimated by Lotz et al [15] was significantly
higher than in the present study, with the smallest female
with developing oocytes being 834 mm FL and around
2 years old This difference in length at first maturity could
be partially explained by regional differences caused by
factors such as food availability and environmental
conditions
Sex ratios and length frequency distributionsMany cobia studies have found a higher percentage offemales than males in their samples [5,30] The sex ratio offemales to males in the present study was 2.18:1, whichwas similar to the ratio of 2.7:1 found by Franks et al [5] inthe Gulf of Mexico The study of Smith [6] appears to be
an exception, with a 1:1 ratio found in North Carolina,which may be due to small sample size in the larger sizeclass Thompson et al [30] reported an overall sex ratio of2.1:1 that was skewed towards males, which is difficult toexplain, as Franks et al [5] and Thompson et al [30]conducted their studies concurrently Franks et al [5]suggested that the discrepancy may be due to differentialsegregation or a higher mortality rate for males withinregions rather than any sampling bias
Because most specimens in the present study were lected in the commercial gillnet and recreational fisheries,there was an obvious bias towards larger individuals thatexceeded the Queensland minimum legal length of
col-750 mm TL Although equal effort was invested inobtaining specimens from both fisheries, the sex ratioprobably adequately represents the adult population but notthe population as a whole The length–frequency distribu-tions for male and female fish were not significantly dif-ferent, indicating no obvious sexual dimorphism inAustralian waters In contrast, Franks et al [5] found asignificant difference in the length frequency distributions
of male and female fish in the Gulf of Mexico, where maleswere smaller than females Most of their samples werefrom sportfishing and may be biased towards larger-sizedindividuals
Spawning seasonHistological examinations of ovarian tissue and monthlychanges in GSI values both suggest that cobia have a
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Month
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Female GSI Stomach fullness -2
-1 0 1 2 3 4 5 6
Fig 10 Relationship between
mean (± SE) monthly
gonadosomatic index values and
stomach fullness index for male
and female cobia Rachycentron
canadum collected in
northeastern Australia between
February 2003 and April 2005.
Immature fish have been
excluded
Trang 40protracted spawning season from September to June This
coincides with the monsoonal ‘‘wet’’ season in northern
Australia Several other reproductive studies of cobia
suggest that the reproductive season for cobia in the north
central Gulf of Mexico is also protracted and extends from
April through early October, with the greatest spawning
actively occurring in spring and early summer [15–17,28,
30]
A number of other pelagic species, including longtail
tuna Thunnus tonggol and talang queenfish Scomberoides
commersonnianus, also spawn at similar times in the GoC,
and the cue for spawning could be environmental factors
[4] Water temperature is often cited as an important cue
for spawning The temperature required for spawning in
many pelagic fishes is often in excess of 24°C [27, 31],
which is satisfied for most of the year in northern Australia
(CSIRO SST, unpublished data), so other environmental
factors may be more important For example, the monsoon
season in northern Australia delivers an enormous amount
of freshwater to the inshore region from numerous large
river systems The timing of spawning coincides with this
high volume of nutrient-rich water, which may influence
food resources and provide a dispersal mode for larvae
[32]
Brown-Peterson et al [16] demonstrated multiple
spawning events in cobia, as determined by the presence of
POF in the ovaries of fish with fully mature, vitellogenic
oocytes Lotz et al [15] and Brown-Peterson et al [16] also
provided additional evidence of multiple spawning of fish
with ovaries containing oocytes undergoing FOM, which
were concurrent with fully mature vitellogenic oocytes
The results of the present study support the suggestions of
Richards [12], Lotz et al [15] and Brown-Peterson et al
[16] that cobia spawn more than once during the spawning
season This is evident by the detection of POF amongst
maturing oocytes The average interval between spawnings
in female cobia, estimated from the evidence of POF, was
7.6 days This spawning frequency is longer than that
reported in other cobia studies, where Franks et al [5] and
Brown-Peterson et al [16] reported a spawning frequency
of 5 days in waters of the southeastern United States and
the north-central Gulf of Mexico However, the spawning
frequency was slightly shorter than that reported by
Brown-Peterson et al [16] in the western Gulf of Mexico (9–
12 days) The longer intervals between spawnings could be
due to the longer distances that cobia need to travel
between feeding and spawning grounds, as suggested by
Brown-Peterson et al [16], and regional differences in
temperature and productivity [27] The small monthly
sample size of females with POF in this study is not
suf-ficient to enable a precise estimation of the monthly
spawning frequency of Australian cobia, and future studies
should address this issue
Gonadosomatic index values were shown in the presentstudy to be a useful indicator of the duration of thereproductive season for cobia, since they agreed well withthe histological findings The GSIs were lowest in Junethrough to August, and then increased during spring(September through to November) to reach their maximumvalues in summer Brown-Peterson et al [16] also foundthat GSI values for cobia in the Gulf of Mexico agreed wellwith their histological data and used mean monthly GSIvalues to assess the duration of the reproductive season.However, Jons and Miranda [33] advised caution in the use
of GSI values for determining reproductive status because
of regional variations in values Brown-Peterson et al [16]therefore suggest that GSI values should not be used forcomparing or indexing maturity stages, particularly inmultiple-spawning fish GSI peaks for cobia from thenorthern Gulf of Mexico have been shown to vary amongstudies, but generally peak between April and June, whenwater temperatures are highest, as in the present study.Fecundity
Our estimates of cobia fecundity, as the total number ofeggs produced by individual fish (577,468–7,372,283oocytes), were similar to those made by Richards [12] (up
to 5,200,000 oocytes per spawn) However, they are siderably lower than the range estimated by Lotz et al [15](2,600,000 and 191,000,000) Their estimates were based
con-on the most advanced group of developing oocytesregardless of their maturation stage, and could overestimatefecundity if all eggs in the group are not released Incontrast, Brown-Peterson et al [16] found that their batchfecundities, 377,000–1,980,000, were lower than all pre-vious estimates made The mean relative fecundity of 249eggs per g (± SD 68.76) determined in this study is highwhen compared to the findings of Brown Peterson et al.[16] (29.1 ± 4.8 – 53.1 ± 9.4) These findings were alsoconsiderably higher when compared to other pelagic spe-cies where relatively fecundity values were available, such
as the southern bluefin tuna Thunnus maccoyii (57 eggs perg) [34] Differences in the methodology used to determinefecundity estimates may explain the vast range of estimatesbetween previous studies Fecundity estimates from thepresent study, which included only final maturation (FOM)and hydrated oocytes, may be more accurate, as thesestages of development are most likely to lead to spawningsuccess
Implications for fisheriesCobia is an increasingly popular commercial, recreationaland aquaculture species in Australia and other parts of theworld As a result, basic biological information on cobia in