Fisheries science JSFS , tập 76, số 6, 2010 11

O R I G I N A L A R T I C L E BiologyCan research on the early marine life stage of juvenile chum salmon Oncorhynchus keta forecast returns of adult salmon?. A case study from eastern Ho

O R I G I N A L A R T I C L E Fisheries

Measurement of swimming speed of giant jellyfish Nemopilema

nomurai using acoustics and visualization analysis

Kyounghoon Lee•Bong-Seong Bae •

In-Ok Kim•Won-Deuk Yoon

Received: 22 April 2010 / Accepted: 10 September 2010 / Published online: 23 October 2010

Ó The Japanese Society of Fisheries Science 2010

Abstract A species of giant jellyfish, Nemopilema

nom-urai, which has appeared only recently in the East China

Sea, is an emerging nuisance in the northeastern region of

Asia because of its extensive damage to fisheries Until

now, the biomass estimates of these jellyfish have mainly

been obtained using trawl sampling and sighting survey

methods However, it is also necessary to determine the

origin and diurnal migration patterns of these jellyfish

Drawbacks of the trawl sampling method are that it is

effective only in estimating the density of jellyfish

popu-lation distributed throughout the entire water column and

requires considerable time Another common analysis

technique is the sighting method, which is effective only in

the estimation of he density of jellyfish distributed in

sur-face areas The sighting method can determine distributions

over wide areas in a short time This method has limitations

in investigating the vertical distribution and swimming

behavior of jellyfishes In our study, we utilized an echo

sounding method extensively and effectively to overcome

these limitations Our method involved the use of a entific echo sounder, acoustic camera, and conductivity-temperature-depth instrument during the drifting of aresearch vessel at various stations in the Yellow Sea Theacoustical method of particle tracking velocimetry (PTV)was used to analyze the swimming speed according to thevertical distribution of N nomurai jellyfish Results of thescientific echo sounder indicated that the jellyfish weremainly present in the water column from the surface up to adepth of 40 m The mean swimming speed of the jellyfishwas estimated as being 0.6 times the bell size (BS), with atendency to maintain a certain speed Further, results of aMonte Carlo simulation showed that the swimming speedwas in the range of 0.46–0.89 BS These results might beused as an index in a migration model, which may beuseful to forecast the behavior and origin of the giant jel-lyfish entering inshore areas on a massive scale in north-eastern Asia

sci-Keywords Giant jellyfish Nemopilema nomurai Swimming speed Vertical distribution 

PTV visualization analysis

IntroductionThe giant jellyfish Nemopilema nomurai, which hasappeared in the East China Sea only recently, is presumed

to be damaging the Korean and Japanese fishing industriesand causing a nuisance along the coasts of northeasternAsia Korea and Japan have recently undertaken research

on the distributed density of the jellyfish by employingtrawl sampling [1] and sighting methods They have alsodeveloped discharge devices for jellyfish to minimize thedamage to the fishing industry [2,3] It is very important to

K Lee ( &)

Fisheries System Engineering Division, Fundamental Research

Department, National Fisheries Research and Development

Institute, Busan 619-705, Korea

e-mail: khlee71@nfrdi.go.kr

B.-S Bae

Aquaculture Industry Division, East Sea Fisheries Research

Institute, Gangneung 210-861, Korea

I.-O Kim

Aquaculture Industry Division, West Sea Fisheries Research

Institute, Incheon 400-420, Korea

W.-D Yoon

Fishery and Ocean Information Division, Research and

Development Planning Department, National Fisheries Research

and Development Institute, Busan 602-092, Korea

DOI 10.1007/s12562-010-0294-7

understand the origin, migration, and behavior patterns of

the jellyfish as they move toward shore [4 6]

The spatial distribution of giant jellyfish has mainly

been determined by midwater trawling and sighting

sur-veys A drawback of the trawl sampling method is that it is

effective only in the estimation of the density of jellyfish

distributed in the main water column, and thus, it takes a

considerable amount of time Additionally, while the

sighting method provides a fast estimate of populations, it

provides only an effective estimate of jellyfish density on

the surface

Recently, underwater acoustics has been utilized

extensively and effectively to investigate the spatial and

vertical distribution of jellyfish, as well as their swimming

behavior patterns as biological information To estimate

population density in a water column using hydroacoustic

techniques, the sound-scattering characteristics of each

kind of jellyfish need to be elucidated, and these

charac-teristics should be differentiated from those of other

scat-terers in the same layer, such as zooplankton and nekton

[7 9]

Acoustic camera systems have recently been developed

and used to measure and monitor the swimming behavior

and speed of jellyfish and other types of fish [10,11] The

speed of fish with good swimming ability can be estimated

without considering the current field; however, the speed of

jellyfish needs to be extracted from the surrounding current

field because jellyfish are not strong swimmers

With this as background, this study observed the

dis-tributions of giant jellyfish moving with the current at sites

in the Yellow Sea during the summer, when the

thermo-cline layer is sufficiently developed; the observations were

carried out in water columns, and the vertical distribution

of giant jellyfish as a function of conditions in the column

was confirmed Additionally, the moving speed of giant

jellyfish was recorded using an acoustic camera system

along with suspended particles in the water column Then,

by assuming that the jellyfish move in the same direction as

the suspended particles, i.e., the current direction, the

swimming ability of the jellyfish was estimated by analysis

of the two-dimensional (2D) observed velocity of the

particles This analysis was done using the particle tracking

velocimetry (PTV) technique, where visualization analysis

was applied to a thin plane of the volume flow images that

were acquired by the acoustic cameras

Materials and methods

Survey area and data acquisitions

Measurements in this study were conducted in the Yellow

Sea using Tamgu 1 (R/V 2, 180G/T) of the National

Fisheries Research and Development Institute, Korea,during the summer, when giant jellyfish are found inmassive numbers mainly around northeastern Asia (Fig.1).The surrounding oceanographic data were collected at 10stations concurrently with the acoustic monitoring surveys.The water temperature and salinity were obtained using aconductivity-temperature-depth (CTD) instrument (SBE-

911, Sea-Bird, USA), and these data were compared withthe vertical distribution of the jellyfish, which was deter-mined using two frequencies of the scientific echosounder(EK60-38, 120 kHz, Simrad, Norway) installed on theresearch vessel at each station When the research vesselwas drifting at night, the jellyfish were monitored con-stantly using an acoustic camera (DIDSON, Sound Metrics,USA) and scientific echosounder (EK60-120 kHz, Simrad,Norway), each installed on the side of the research vessel(Fig.2) The acoustic camera systems were used to analyzethe swimming behavior of subject organisms The images

of the jellyfish moved by the current were analyzed todetermine their swimming ability in relation to their bellsize (BS) This was done by applying the flow visualizationtechnique

Flow visualization analysisThe PTV technique calculates particle displacement byanalyzing the quantitative velocity field This techniquewas originally used to calculate average velocity in order toFig 1 Map of Yellow Sea showing survey area for giant jellyfish

compute direct correlation coefficient or the Fourier

con-version for the scattered particles’ intensity distribution

within the observed section of flow visuals [12] Hence,

PTV calculates the cross-correlation function between

observation sections using two visuals obtained from

intervals of specific durations It assumes velocity by

considering the peak value achieved through the

calcula-tion to be the average displacement (Fig.3a) The PTV

software program used in this analysis was a Thinks 2D

PTV program (T&Tech, Korea), which employs a

two-frame cross-correlation algorithm to perform the visual

analysis The cross-correlation function is shown in Fig.3

and expressed in Eq.1 When this correlation condition is

satisfied, the particle-movement vector values can be

esti-mated The values of fi and gi are that of a pixel in the

image, and the resolution of our acoustic camera image is

The swimming speed of the jellyfish is extracted from

the current 2D velocity using PVT as shown in Fig.4

When jellyfish move as vector A!

with a swimming tiltangle h against suspended particles moving as vector B!

,the swimming speed can be estimated by the BS of jellyfish

acquired from the acoustic image data Then, the

swimming tilt angle in relation to the current direction is

represented by (?) and (-) in the case of upward and

downward directions, respectively

Additionally, it is certain that the moving behavior of

the jellyfish is affected by the current; the jellyfish are

considered to be moving faster or slower than the current

speed depending on the current direction relative to the

swimming direction of the jellyfish In this case, the causecan be considered in measured low values by resistanceagainst the jellyfish’s bell in a hovering state as a physio-logical conservation of its energy so that it will be con-sidered to any uncertainty of the mean swimming speed in

Fig 2 Experimental setup for

measurement of moving speed

of giant jellyfish; the setup

consists of an acoustic camera

and a 120 kHz split beam

second image of particle "A" at time t 2

(a)

(b)

Fig 3 Schematic diagram of particle tracking velocimetry (PTV)

relation to the jellyfish’s bell size A Monte Carlo

simu-lation was performed to analyze the minimum, maximum,

and mean swimming speeds of the jellyfish The simulation

was performed by assuming a normal swimming speed

distribution relative to the mean BS of the jellyfish The

variation in the swimming speed was established to be

within the range of ±5%

Results

Oceanographic environment and vertical distribution

Typically, during the summer season, the nighttime water

column in the Yellow Sea forms thermocline layers whose

temperature differs with depth However, strong

thermo-cline layers are generally formed at depths of 10–30 m

with a water temperature difference of more than 13°C

According to the salinity data, strong low-salinity troughs

were formed around the surface layer (Fig.5) Their

ver-tical distribution at each station was observed and analyzed

by two frequencies of the scientific echo sounder (Fig.6),

and the BS was obtained by bottom trawl sampling at each

station (Fig.7)

During the observation period in July 2007, the

popu-lation of giant jellyfish following the Kuroshio current

northward through the Yellow Sea was found to move day

and night from the surface layer to a depth of 40 m at 10

stations There was not nearly distributed in the water

columns deeper than 40 m depth with a stabilized watertemperature and salinity The average BS of giant jellyfishwas 0.30 m, and jellyfish with larger BSs had a higherdensity in the southern area than in the northern areas

Fig 4 Swimming speed of jellyfish as measured from 2D current

profiles

Fig 5 Variation of temperature

(a) and salinity (b) of water

column in survey area The box

marks the median and the 25th

and 75th percentiles of the

oceanographic data; the

whiskers show the 10th and 90th

percentile plots of the

Swimming speed

Table1 shows the relationship between the mean

swim-ming speed (based on the mean moving speed of suspended

particles) and the bell size of the jellyfish, obtained by

measuring the velocity fields using the PTV analysis of data

measured at each station and stored in the acoustic camera

The swimming behavior of jellyfish in the open sea

showed various tilt angles of hovering, which differed with

the current When the current direction was along the

horizontal axis in 2D vectors, the tilt angle was measured

in the up-downward direction, as shown in Table1 These

values were not considered in absolute terms because they

were measured while the test vessel rolled and pitched

owing to sea conditions In order to estimate the swimming

ability of the jellyfish with high precision, the acoustic

images were analyzed only under conditions where the

jellyfish moved and hovered along the current direction

Further, the detected jellyfish were only considered in the

recorded data when they were captured over a sufficiently

large temporal span by the acoustic camera

In our analysis of the mean swimming speed of

N nomurai jellyfish, the BS of jellyfish ranged from 0.24 to

0.68 m in two different current fields estimated by PTV

analysis using suspended particles (Table1) The

maxi-mum mean net swimming speed was 0.529 m/s (BS

0.680 m), whereas the minimum mean swimming speed

was 0.106 m/s (BS 0.247 m) The mean swimming speed

ranged from 0.43 to 0.78 BS, and the swimming speed was

estimated to be 0.60 times the BS

Discussion

When the thermocline layer was strongly formed within the

surface layer, the population of N nomurai jellyfish within

the first 10 m numbered 959, accounting for approximately42.7% of the total observed population The populationwas mainly distributed (98.6%) in the depth range betweenthe surface and 50 m This shows that the N nomuraijellyfish are distributed within a wide range of water tem-peratures, from 8 to 26°C, despite the presence of a strongthermocline layer From a general ecological perspective, it

is assumed that in water columns with rapidly changingtemperatures, the distribution of organisms tends to beseparated or limited to small zones; however, in our study,

N nomurai jellyfish were observed to be passing through astrong thermocline layer Such results confirm that, incontrast to the expected results [13], N nomurai jellyfishare distributed even in low-temperature zones althoughtheir population density around Korea drops rapidly inareas where the temperature is below 12°C Moreover, wefound that in high-temperature water columns and low-salinity waters in the upper thermocline layer, almost allgiant jellyfish were distributed at water depths shallowerthan approximately 40 m [1] In survey areas, N nomuraijellyfish are mainly distributed in water columns withsalinities in the wide range from 28.6 to 34.7 psu Low-salinity waters on the surface layer existed at stations 8–10

in the southern Yellow Sea These results differ from theexpectation that the influence of water temperature andsalinity variation on the vertical distribution of N nomurai

is not significant It seems that these jellyfish can adaptwell to widely ranging conditions in the oceanographicenvironment

Thus, considering the fact that zooplankton is the majorfood source of most jellyfish, it is necessary to elucidate thedistribution of zooplankton throughout the water column inrelation to their diurnal vertical migration Further, addi-tional experiments should be conducted to elucidate thedistribution of zooplankton in wide water columns close tothe surface, as well as midwater layers and bottom layers in

Table 1 Swimming speed in relation to bell size of N nomurai jellyfish

Bell size (BS, m) Speed (m/s) Direction (°) Speed (m/s) Direction (°)

consideration of the different environmental factors that

relate to jellyfish

The swimming behavior of jellyfish has been

deter-mined by the seawater circulation model in the field of

physical oceanography as a way of estimating movements

of N nomurai along currents This study was conducted in

order to estimate timings of jellyfish migrating to inshore

areas by using particle tracking under the assumption that

jellyfish have relatively slow swimming speed from a

temporary point of origin Moreover, jellyfish movement,

migration, and swimming speed are currently being

ana-lyzed using ultrasonic pingers along inshore areas of Japan

[11]

In general, it is difficult to estimate the swimming

speed of jellyfish because it is influenced by the current

field, which significantly affects the measured swimming

speed Recently, acoustic camera systems have been used

to analyze the swimming behavior of subject organisms

They are convenient tools for distinguishing the speed and

size of organism movement, but they require current

information in survey areas It is therefore difficult to

acquire actual information on swimming behavior Thus,

PTV analysis, applied to the flow visualization technique,

was used along with visual data acquired by acoustic

cameras to analyze a 2D velocity field at the specific depth

where jellyfish move This analysis was also used to

estimate the swimming ability of the jellyfish relative to

their BS

The mean swimming speed was estimated to be 0.6

times the BS, as shown in Table1 This mean speed was

estimated to be six times the swimming speed

(approxi-mately 0.1 BS), as measured by N nomurai jellyfish’s

vertical migration attached by pingers [11] These lower

measured values were possibly attributable to their vertical

migration speed, including horizontal current vectors at the

depth at which they were moving

Considering the relation between swimming speed and

BS size as shown in Table1, we suggest that there are

mainly two size domains centered around BSs of 0.25 and

0.65 m, and the swimming abilities of the jellyfish differ

The correlation between the BS and swimming speed of

N nomurai jellyfish can be estimated as shown in Fig.8

(R2= 0.92)

Further, it was clear that the movements of jellyfish are

affected by the current speed; the jellyfish were

consid-ered as moving faster or slower than the current speed In

the latter case, the slower speed is probably caused by

resistance to the current of the jellyfish’s bell in a

hov-ering state in an effort to conserve energy physiologically

In addition, the Monte Carlo simulation was performed to

analyze their swimming ability in relation to BS and

to subsequently consider the uncertainty of the mean

swimming speed in relation to the size of the jellyfish.The results of jellyfish’s swimming ability were estimatedover the entire range from 0.46 to 0.82 BS as shown inFig.9

The swimming performance of the jellyfish was lyzed by performing a comparison between the kinematicsacquired from video recordings and those simulated usingthe hydrodynamic model of jellyfish The comparisonshowed the experimental values of the swimming speedwere comparable to the body weight (BW, in kg) of moonjellyfish Aurelia aurita [14]: u¼ 1:4  ðBWÞ0:17where

ana-BW can be calculated from the length–weight ship: BW¼ 0:0003  ðBSÞ2:5364[13] Therefore, the swim-ming speed of N nomurai jellyfish ranged from 0.44 to0.78 BS The average swimming speed was estimated asbeing 0.67 BS While the morphological characteristics ofswimming speed of A aurita jellyfish are different fromthose of N nomurai jellyfish, the acceleration reactions oftheir swimming behaviors are similar; therefore, thisresult obtained by applying the PTV analysis is reliablefor estimating the swimming speed of jellyfish in thefield

relation-This study proves that visualization analysis using thePTV technique is more accurate than other analysis tech-niques (e.g., current meter, underwater acoustic camera) inestimating the swimming speed of N nomurai jellyfish asdistinct from the current speed The visualization analysisusing the PTV technique is useful even when the testvessel’s pitching and rolling are significant This studysuggests the usefulness of an index in forecasting themigration behavior and origin, with consideration of the

R2 = 0.92

0 0.1 0.2 0.3 0.4 0.5 0.6

Bell Size (m)

Fig 8 Relationship between swimming speed and bell size of Nemopilema nomurai jellyfish

current fields, of jellyfish that are found to be moving

inshore on a massive scale in northeastern Asia

Acknowledgments We thank the officers and crews of the R/V

Tamgu 1 for monitoring jellyfish in the survey periods We are

grateful to Dr T Arimoto, Dr Y Matsushita and two anonymous

reviewers for insightful comments that greatly helped to clarify and

refine the paper This study was partially supported by a grant

(RP-2009-FE-014) from the National Fisheries Research and

Devel-opment Institute of Korea and the YSLME Nemopilema nomurai

jellyfish monitoring project (2008–2009).

References

1 Honda N, Watanabe T (2007) Vertical distribution survey of the

giant jellyfish Nemopilema nomurai by an underwater video

camera attached to a midwater trawl net Nippon Suisan

Gakkaishi 73:1042–1048

2 Matsushita Y, Honda K (2005) Method of designing and

manu-facturing JET (jellyfish excluder for towed fishing gear) for

various towed fishing gears Nippon Suisan Gakkaishi 71:965–

967

3 Kim IO, An HC, Shin JK, Cha BJ (2008) The development of

basic structure of jellyfish separator system for a trawl net J Kor

Soc Fish Tech 44:99–111

4 Uye S, Ueno S, Hiromi J, Shiomi K (2005) Jellyfish blooms—

ecology, biochemistry and food science for utilization Nippon

Suisan Gakkaishi 71:968–994

5 Honda N, Watanabe T (2007) Observation of the giant jellyfish Nemopilema nomurai using an underwater acoustic camera Nippon Suisan Gakkaishi 73:919–921

6 Lee KH, Kim IO, Yoon WD, Shin JK, An HC (2007) A study on vertical distribution observation of jellyfish (Nemopilema nomu- rai) using acoustical and optical methods J Kor Soc Fish Tech 43:355–361

7 Colombo GA, Mianzan GH, Madirolas A (2003) Acoustic characterization of gelatinous-plankton aggregations: four case studies from the Argentine continental shelf ICES J Mar Sci 60:650–657

8 Brierley AS, Axelsen BE, Boyer DC, Lynam CP, Didcock CA, Boyer HJ, Sparks CAJ, Purcell JE, Gibbons MJ (2004) Single- target echo detections of jellyfish ICES J Mar Sci 61:383–393

9 Hirose M, Mukai T, Hwang DJ, Iida K (2005) Target strength measurements on tethered live jellyfish Nemopilema nomurai Nippon Suisan Gakkaishi 71:571–577

10 Rose CS, Stoner AW, Matteson K (2005) Use of high-frequency imaging sonar to observe fish behaviour near baited fishing gears Fish Res 76:292–304

11 Honda N, Matsushita Y (2009) In situ measurement of swimming speed of giant jellyfish Nemopilema nomurai Nippon Suisan Gakkaishi 75:701–703

12 Lee YH, Choi JW (1996) Principle and classification of PIV.

J Kor Soc Mech Eng 36:1146–1162

13 National Fisheries Research and Development Institute (NFRDI) (2005) Annual report of jellyfish research NFRDI, Busan, Korea

14 McHenry MJ, Jed J (2003) The ontogenetic scaling of namics and swimming performance in jellyfish (Aurelia aurita).

hydrody-J Exp Biol 206:4125–4137

Fig 9 Results of Monte Carlo

simulation according to the

uncertainty of swimming speed

of Nemopilema nomurai

jellyfish

O R I G I N A L A R T I C L E Biology

Are environmental conditions in Finnish streams limiting to early

life-history survival in the nonnative rainbow trout?

Kai Korsu•Ari Huusko

Received: 7 June 2010 / Accepted: 5 August 2010 / Published online: 17 September 2010

Ó The Japanese Society of Fisheries Science 2010

Abstract The nonnative rainbow trout Oncorhynchus

mykiss has been an unsuccessful invader in North European

streams, although it has been widely introduced Here we

studied whether early life history stages (egg incubation

and hatching, first overwintering) act as filters for the

establishment of hatchery rainbow trout Survival of

hat-ched alevins was approximately 80%, whereas only 47% of

the embryos survived However, the latter value was

impacted by the high number of unfertilized eggs

Corre-lation coefficients with embryo survival rate and

environ-mental variables (pH and temperature) were statistically

insignificant In the overwintering experiments, the

sur-vival of rainbow trout was 93% The growth was generally

slowed during the winter, but in the spring the growth of

rainbow trout exceeded that of the native brown trout Our

data demonstrated that the survival and growth of rainbow

trout during early life-history stages were relatively high

and comparable to those of the native brown trout Based

on the variables considered in our study, our results suggest

that environmental conditions during early life-history

stages are not detrimental for rainbow trout in the study

streams

Keywords Biological invasions  Egg incubation 

Hatching success Oncorhynchus mykiss  Overwintering

IntroductionInvasions by nonnative organisms are recognized to be amajor threat to global biodiversity, leading to speciesextinctions and worldwide homogenization of biota [1,2].The establishment success of nonnative species in recipientsystems has been relatively high, as up to 50% of speciesthat have entered recipient ecosystems have subsequentlybecome established [3, 4] An ability to predict whichspecies will establish and become nuisance invaders isconsidered highly important for prioritizing managementefforts such as prevention and eradication, which are likely

to be effective only in the early stages of invasion [5 7].Salmonids are widely introduced fishes that have suc-cessfully established themselves in many parts of the world[8] They also have wide-reaching impacts on native spe-cies and the trophic organizations of native stream com-munities [9, 10] Subsequently, the World ConservationUnion has listed two salmonids (the rainbow trout On-corhynchus mykiss and the brown trout Salmo trutta)among the eight fish species included in the list of 100 ofthe world’s worst invasive species [11] Among the sal-monids, rainbow trout is the most widely introduced spe-cies; nevertheless, its establishment success varies stronglyamong regions [12,13] For example, in Europe, rainbowtrout has become established primarily in southern regions,while the British Isles and northern regions have only afew self-reproducing populations [12–16] In Japan, NewZealand, and North America, established populations ofrainbow trout have frequently affected native speciesnegatively [10]

The reason for the low establishment success of rainbowtrout in North European streams has not been fullyexplained to date Fausch [15] suggested several influentialfactors, such as (1) high environmental resistance from

Finnish Game and Fisheries Research Institute,

Kainuu Fisheries Research, Manamansalontie 90,

88300 Paltamo, Finland

DOI 10.1007/s12562-010-0285-8

temperature and flow regimes, (2) high biotic resistance

from native fishes or diseases, (3) rainbow trout

domesti-cation, and (4) high angling pressure Water pH may also

affect establishment success in some regions, because

rainbow trout is vulnerable to acidic water [17–19]

Prob-ably the most powerful explanation thus far is associated

with environmental resistance: high floods during early

summer may wash away newly emerged rainbow trout fry

[12] However, climate change may alter the timing of

floods due to earlier snowmelt, thus facilitating fry survival

and subsequently the establishment of rainbow trout in

northern regions [13, 20, 21] In general, environmental

conditions during early life history stages have major

effects on salmonid populations For example, a lack of

swimming ability easily results in the displacement of

hatched alevins under high-flow conditions, and a lack of

adequate fat reserves may lead to fry starvation [15]

Moreover, the early life-history stages are more vulnerable

to acid water compared to adult fishes [18] Thus, any

change in environmental conditions in spring or early

summer may have a strong positive effect on the survival

of rainbow trout in North Europe [13, 20] After being

established in northern streams, rainbow trout might have

severe effects on the survival and habitat use of the native

brown trout [22] and various other stream organisms [9]

Moreover, being a spring spawner, rainbow trout might

superimpose their redds on those of the native

autumn-spawning salmonids, as has been observed in Japanese

streams [23]

Here, we explored the survival and growth of the

non-native rainbow trout during two critical early life-history

stages: (1) egg incubation and hatching in the spring, and

(2) the first overwintering We performed experiments in

Finland, where rainbow trout has been extensively stocked

for over 100 years [17], but reproduces only in a few

streams in southern regions (Saura et al.http://www.rktl.fi/

www/uploads/pdf/raportti289.pdf) The aim of the study

was to examine whether low survival in the early

life-history stages limits rainbow trout establishment in north

European streams, where rainbow trout has thus far been an

unsuccessful invader

Materials and methods

Incubation experiment

The experiment was conducted in five central Finland

streams where rainbow trout have not become established

The selected streams were: Arvaja, Kiertojoki,

Ko¨nkko¨j-oki, Ko¨hnio¨npuro, and Rutajoki (see [24–26]) The study

streams drain forested catchments and have natural flow

regimes with snowmelt-induced floods that typically occur

in April–May Each of these streams is known to support aself-sustaining population of the native brown trout [24], aspecies that spawns in autumn and for which the eggs hatch

in spring By contrast, rainbow trout spawn in spring andthe eggs hatch in late spring, approximately 1 month afterthe spawning

Rainbow trout eggs used in the experiment were ized in a local commercial hatchery (Savon Taimen

fertil-http://www.savontaimen.fi) on 24 April 2008 In theabsence of established populations of rainbow trout, eggand sperm were provided by hatchery fishes The fertilizedeggs were transferred to the incubation sites of the studystreams within 10 h after the fertilization Thirty eggs wereplaced in one plastic egg container (10 cm 9 10 cm 9

6 cm) with abundant gravel mixture (5–30 mm in ter) Regular water flow through containers was ensured byutilizing openings covered with plastic mesh (mesh size

diame-1 mm) Ten containers were placed in natural gravel beds

of each stream (water velocity, 12–30 cm s-1; depth,20–40 cm) Embryo survival, hatching, and the swimmingability of the hatched alevins were measured on twooccasions: on 30 May and 7 June Both times, five ran-domly selected containers per stream were examined.Water temperature and pH were measured four times ineach stream during the study At the end of the experiment,

we stored the hatched alevins in a solution of ethanol andformalin and determined their total lengths and dry masses(24 h at 60°C) later in the laboratory At the end of theexperiment, we also took a sample of hatchery alevins fromthe same strain that was used in our field experiment Thiswas done to compare the sizes of the alevins reared in thecontrolled hatchery environment and the field environment

We assume that a markedly smaller body size of the alevins

in the study streams compared to those reared in thehatchery would indicate the influence of some sublethalenvironmental factor

We used a paired t test to explore whether there was adifference between embryo survival on 30 May and 7 June

We also investigated the relationship between embryosurvival and three explanatory variables (degree days,minimum pH, and pH at the end of the experiment) byPearson’s correlation coefficient Because we had only fourtemperature measurements per stream, the degree daysreported here are rough estimates

Overwintering experiments

We studied the survival and growth of age-0 rainbowtrout compared to the native brown trout Sympatrictreatments were used because we assumed that the lowinvasion success of rainbow trout may result from its

lower ecological performance compared to the native

species, which is presumably well adapted to the

envi-ronmental conditions of North European streams From

late autumn 2007 to early summer 2008, we conducted

the experiments in six 26 m long and 1.5 m wide

out-door artificial stream channels located at the Finnish

Game and Fisheries Research Institute’s (FGFRI)

research station at Paltamo, Finland (64°300N, 27°100E)

The stream bed consisted of gravel and cobble

support-ing benthic invertebrate communities similar to those

present in a nearby river in terms of both species

com-position and densities Therefore, we did not feed the

fishes during the experiments Additional information on

the experimental system is given in our earlier papers

[27, 28]

In the six artificial streams, we performed two

experi-ments (three streams each) at two different spatial scales

(i.e., small scale and mesoscale, see below) This

separa-tion was made because ecological phenomena are typically

scale variant [29] In both experiments we used age-0

hatchery fishes reared at the FGFRI research station The

experiments started on 18 October 2007 and ended on 4

June 2008 Water temperature was about ?6°C at the start

of the experiments, decreased to below 2°C within a

month, and thereafter stayed relatively stable [average

1.9°C, standard deviation (SD) 0.3] until the end of April,

before rising to ?10°C by the end of the experiments At

the start of the experiments, the average (SDs in

paren-theses) masses and total lengths of the rainbow trout were

8.2 g (2.5) and 9.0 cm (0.9) (small-scale experiment) and

11.9 g (2.6) and 10.5 cm (0.7) (mesoscale experiment)

The corresponding values for brown trout were 6.6 g (1.4)

and 8.6 cm (0.6), and 9.9 g (0.7) and 10.0 cm (0.3),

respectively The size difference appeared because rainbow

trout tend to grow more rapidly than brown trout under

rearing conditions (see also [17,22]) In both experimental

settings, we measured water depth, water velocity, and

substrate size in cross-sectional transects Nine (three

transects) and 24 (eight transects) replicate measurements

were made in the small-scale and mesoscale experimental

streams, respectively

In the small-scale experiment, we used 15 plastic boxes

(width, 60 cm; length, 80 cm) that were placed at random

positions in three artificial streams, with five boxes in

each stream Here we consider only the results gained

from five boxes, because the other ten boxes held

inter-specific competition treatments that will be reported

separately Boxes were placed directly on the stream bed

Each box had gravel substrate (diameter 4 cm) and three

15 9 15 cm cobbles under which we placed shelter tubes

(32 mm in diameter and 10 cm in length) for use as

hiding places, thus mimicking typical overwintering

habitat for juvenile stream salmonids [30] Moderatewater flow [16.7 cm s-1 (SD 0.7)] was provided throughthe boxes by wire mesh panels (mesh size 5 mm)mounted at the heads and tails of the boxes Averagewater depth in the boxes was 14.8 cm (SD 0.4) No icecover formed in the boxes during the winter We releasedthree rainbow trout and three brown trout in each of thefive boxes (total fish density 12.5 ind m-2) used in thisstudy

In the mesoscale experiment, we divided each of thethree artificial streams into three 8.5-m long sections(upstream–middle–downstream) with wire mesh panels(mesh size 10 mm) [27, 28] Thus, we had nine similarsections, although we only consider three here (for anexplanation of why, see the previous paragraph) Eachsection comprised an upstream riffle (water velocity,20–60 cm s-1; depth, 15–25 cm; mean substrate diameter,

15 cm) and a downstream pool (0–20 cm s-1, 25–35,

4 cm, respectively) Partial ice cover formed to the edges

of the pool section temporarily during the midwinter Wereleased ten rainbow trout and ten brown trout (total den-sity 1.5 ind m-2) in each of the three sections of the threeartificial streams

Fish total length and mass were measured four timesduring the study: on 22 November, 1 April, 6 May, and 4June Fishes were collected from the streams by elec-troshocking the sections (mesoscale experiment) or liftingthe boxes from the stream (small-scale experiment).Because we observed a posteriori that fishes started togrow rapidly from 6 May to the end of the experiment (4June), the data were split into two successive periods:winter (17 October–6 May) and spring (7 May–4 June).Individual fish growth was analyzed as % mass (M) gain[(M2- M1)/M29 100, where M1 and M2 are the fishmasses at the beginning and end of the sampling period].Percent mass gain was used instead of increase in fishmass to avoid bias potentially caused by slightly differinginitial masses of the fishes Each fish was injected with apassive integrated transponder (PIT) to obtain growthrates of individual fishes (see [28, 29]) We studiedpotential differences in the % mass gains of the studyspecies by paired t tests separately for both periods(winter, spring) To avoid pseudoreplication, all analyseswere run with replicate averages (n = 5 and 3 in thesmall- and mesoscale experiments, respectively), resulting

in low statistical power Consequently, we set alpha at0.1 to increase the power of the test and thus reduce thepossibility of type II error The logarithmic specificgrowth rate could not be calculated because the fishestended to have negative growth rates during the winter.Dead fishes were calculated on each occasion whenmeasurements were performed One rainbow trout and

one brown trout disappeared from the artificial streams

during the winter

Results

Incubation experiment

During the incubation experiment, the water temperature

increased from between 3 and 5°C up to 18.9°C, while the

pH and water velocity decreased (Table1) At the end of

the experiment, the average survival rate of rainbow trout

embryos was 47% Although there seemed to be a slight

decrease in the survival rate between 30 May and 7 June,

the difference was statistically insignificant (paired t test:

t4= 1.93, P = 0.13) (Fig.1a) The fertilization success of

the eggs used in the experiment was only 67% (Savon

Taimen hatchery, unpublished), indicating that the true

survival in our field experiment, excluding unfertilized

eggs that we classified as dead, was potentially up to 80%

This higher estimate was further supported by the fact that

the survival of hatched alevins was approximately 80%

(Fig.1b) In the two streams with the highest water

tem-peratures, alevins hatched on 30 May (incubation time,

37 days; stream-specific degree day approximations, 352

and 385) In the three other streams, alevins hatched until 7

June (incubation time, 45 days; degree day

approxima-tions, 374–416) The hatched alevins were generally active

swimmers, and those reared in streams with higher water

temperatures had almost completely depleted their yolk sac

by 7 June The size of the hatched alevins in the field

experiment was also comparable to those individuals

reared in the hatchery: mean total lengths were 16.2 mm

(SD 0.8) and 15.7 mm (SD 1.0), and mean dry masses were

0.016 g (SD 0.002) and 0.015 g (SD 0.0001) in the

hatchery and natural streams, respectively Correlation

coefficients with stream-specific mean embryo survival and

three stream-specific explanatory variables—(1) degree

days, (2) minimum pH, and (3) pH—at the end of the

experiment were all negative (r = -0.54 to -0.65) but

Stream Geographic coordinates Temperature (°C) Water velocity (m s-2) pH

Latitude, N Longitude, E April 24 June 7 April 24 June 7 April 24 June 7

b

0 20 40 60 80 100

14%) In the mesoscale experiment, the mortality of both

rainbow trout and brown trout was 7% In the small-scale

experiment, the mortality was 7% for rainbow trout and

10% for brown trout The growth of both species was

generally slowed during the winter period, but fishes only

lost mass in the small-scale experiment (Fig.2) In the

spring period, however, both species started to grow: in the

small-scale experiment, the % mass gains of the study

species were almost identical (winter, t4= -0.33,

P = 0.758; spring, t4= -0.176, P = 0.869), but rainbow

trout tended grow slightly faster in the mesoscale

experi-ment (winter, t2= -3.25, P = 0.083; spring, t2= -2.18,

P = 0.161) (Fig.2)

Discussion

Our data demonstrated that the survival and growth of

rainbow trout during early life-history stages were

rela-tively high and comparable to those of the native brown

trout Syrja¨nen et al [25] found that the survival of brown

trout embryos in the same streams used in our incubation

experiment was 83–98%, which was only slightly higher

than that for the rainbow trout in our study In the wintering experiment, however, we observed slightlyhigher survival and growth of rainbow trout compared tobrown trout Based on the variables considered in ourstudy, our results suggest that environmental conditions atearly life-history stages are not detrimental to rainbow trout

over-in the study streams However, it is unclear how well theegg boxes mimic natural redds produced by the fishes andwhether the timescale of the study covers the mostimportant early life-history events Therefore, care should

be taken when drawing strong inferences from thisexperiment

Embryo survival in the incubation experiment was notrelated to either stream-specific water temperature (degreedays) or water pH, indicating that these two factors werewithin the tolerance limits of rainbow trout [15, 17, 18].However, the pH in the two most acid streams was \6,which is probably close to the levels that cause adverseeffects on rainbow trout embryos [18] Boreal streams such

as ours have snowmelt-induced floods in the spring, and the

pH may in some years concurrently drop to critical levelsfor various organisms, including rainbow trout [18] Thismay eliminate developing embryos, ultimately affectingthe reproduction success of the spring-spawning species.Fausch et al [12] showed that environmental resistancefrom flow regimes that differ from that to which rainbowtrout has adapted largely explains the inconsistency in theestablishment success of the species in different regions Inthe native range of rainbow trout, flooding occurs duringwinter, followed by low flows during summer after fryemergence [12] In our study streams, the flow regimeslightly differs from that of the native range of rainbowtrout, which probably hinders the reproduction successonly marginally The peak flow in our study streamsoccurred at approximately the start of the incubationexperiment (24 April) Discharges and water velocitiesthen started to decline, probably providing favorableemergence conditions for rainbow trout at the beginning ofJune The suitability of flow regimes for rainbow trout innorth European streams could even increase in the future ifclimate change alters the timing of floods due to earliersnowmelt [20,21]

Our results from the overwintering experiment strate that age-0 rainbow trout had a high survival andstarted to grow relatively quickly in the spring Theseresults suggest that the first overwintering of age-0 rainbowtrout is not a bottleneck period for the establishment of thespecies in boreal streams Rainbow trout even tended togrow faster than brown trout, a pattern also found else-where [17,22,31] The few age-0 rainbow trout originatingfrom natural reproduction in southern Finnish streams havealso been larger than sympatric age-0 brown trout (Saura

Fig 2 Mean (±SD) % mass gains of the sympatric rainbow trout

and brown trout in the small-scale and mesoscale experiments during

the two successive periods (winter = 17 October–6 May; spring = 7

May–4 June)

indicating good ecological performance of juveniles when

successful spawning and egg hatching have occurred

Given that embryo and overwintering survival were high in

our study streams, the reason for the low establishment

success of rainbow trout in Finnish streams remains

unsolved Besides the factors studied here, there are several

other possible explanations for the low establishment

suc-cess Firstly, acid water in the spring flood may affect

rainbow trout sperm negatively, thus dramatically reducing

reproduction success [17,32] Secondly, a European

my-xozoan parasite (Myxobolus cerebralis) that causes

whirl-ing disease in rainbow trout, but to which brown trout is

resistant, may hinder establishment [15] Thirdly, a lack of

rainbow trout males in natural waters may occur because

hatcheries commonly produce only female diploid fishes

[15] Nevertheless, it is common for only a few male fishes

to be present (P Latikka, pers comm., 2010), so these few

may be enough to fertilize a large number of eggs, thereby

increasing the risk of invasion [15]

We suggest that if the currently unknown factor(s) that

restrict rainbow trout establishment in North European

streams change, for example due to current climate change

[13, 20], this widely introduced species will rapidly

establish itself and spread across several stream systems

Blanchet et al [22] showed experimentally that rainbow

trout had a negative effect on the habitat use and survival

of brown trout in France, indicating that established

pop-ulations of rainbow trout could have serious

population-level effects on brown trout At present, another North

American salmonid species, brook trout Salvelinus

fonti-nalis, has become established and spreads effectively in

several European streams, subsequently posing a serious

threat to brown trout populations [33–35] Given that brook

trout is a headwater species, whereas rainbow trout is found

primarily in larger streams [10], these two invaders have

the potential to narrow the habitat niche of brown trout to

critical levels in several European stream systems [36]

Established populations of nonnative salmonids can also

have wider, ecosystem-level effects on native stream

communities that have not evolved with the exotic invasive

species [9] For example, in a Japanese stream, the effects

of rainbow trout invasion spread to the riparian community

because of the altered supply of adult aquatic insects

emerging from the stream and moving to the forest [37]

Fisheries managers should recognize the risk of rainbow

trout establishment and perform precautionary acts to

prevent this invader from gaining a foothold in Northern

Europe

Acknowledgments Kimmo Sivonen helped to perform the

incuba-tion experiment The staff of Kainuu Fisheries Research, FGFRI,

especially Teija Haataja, provided logistic help during the

experi-ments Juha Karjalainen, Yrjo¨ Lankinen, Ari Saura and Jukka

Syrja¨nen provided important information during the study Jani Heino

commented on an earlier version of the manuscript We also ciate the comments of two anonymous referees regarding a previous version of the manuscript.

4 Williamson M (2006) Explaining and predicting the success of invading species at different stages of invasion Biol Invasions 8:1561–1568

5 Copp GH, Garthwaite R, Gozlan RE (2005) Risk identification and assessment of non-native freshwater fishes: a summary of concepts and perspectives on protocols for the UK J Appl Ichthyol 21:371–373

6 Crooks JA (2005) Lag times and exotic species: the ecology and management of biological invasions in slow-motion Ecoscience 12:316–329

7 Simberloff D (2006) Invasional meltdown 6 years later: tant phenomenon, unfortunate metaphor, or both? Ecol Lett 9:912–919

impor-8 Rahel FJ (2007) Biogeographic barriers, connectivity and homogenization of freshwater faunas: it’s a small world after all Freshw Biol 52:696–710

9 Simon KS, Townsend CR (2003) Impacts of freshwater invaders

at different levels of ecological organisation, with emphasis on salmonids and ecosystem consequences Freshw Biol 48:982–994

10 Korsu K, Huusko A, Muotka T (2008) Ecology of alien species with special reference to stream salmonids Boreal Environ Res 13(Suppl A):43–52

11 Lowe S, Browne M, Boudjelas S, De Poorter M (2000) 100 of the world’s worst invasive alien species Invasive Species Specialist Group of the World Conservation Union, Auckland

12 Fausch KD, Taniguchi Y, Nakano S, Grossman GD, Townsend

CR (2001) Flood disturbance regimes influence rainbow trout invasion success among five holarctic regions Ecol Appl 11:1438–1455

13 Crawford SS, Muir AM (2008) Global introductions of salmon and trout in the genus Oncorhynchus: 1870–2007 Rev Fish Biol Fish 18:313–344

14 Povzˇ M, Sˇumer S (2005) A brief review of non-native freshwater fishes in Slovenia J Appl Ichthyol 21:316–318

15 Fausch KD (2007) Introduction, establishment and effects of native salmonids: considering the risk of rainbow trout invasion

non-in the United Knon-ingdom J Fish Biol 71(Suppl D):1–32

16 Ribeiro F, Elvira B, Collares-Pereira MJ, Moyle PB (2008) Life-history traits of non-native fishes in Iberian watersheds across several invasion stages: a first approach Biol Invasions 10:89–102

17 MacCrimmon HR (1971) World distribution of rainbow trout (Salmo gairdneri) J Fish Res Bd Can 28:663–704

18 Haines TA (1981) Acid precipitation and its consequences for aquatic ecosystems: a review Trans Am Fish Soc 110:669–707

19 Hulsman PF, Powles PM, Gunn JM (1983) Mortality of walleye eggs and rainbow trout yolk-sac larvae in low-pH waters of the LaCloche Mountain area, Ontario Trans Am Fish Soc 112:680–688

20 Rahel FJ, Olden JD (2008) Assessing the effects of climate change on aquatic invasive species Conserv Biol 22:521–533

21 Heino J, Virkkala R, Toivonen H (2009) Climate change and

freshwater biodiversity: detected patterns, future trends and

adaptations in northern regions Biol Rev 84:39–54

22 Blanchet S, Loot G, Grenouillet G, Brosse S (2007) Competitive

interactions between native and exotic salmonids: a combined

field and laboratory demonstration Ecol Freshw Fish 16:133–143

23 Taniguchi Y, Miyake Y, Saito T, Urabe H, Nakano S (2000)

Redd superimposition by introduced rainbow trout,

Oncorhyn-chus mykiss, on native charrs in a Japanese stream Ichthyol Res

47:149–156

24 Muotka T, Syrja¨nen J (2007) Changes in habitat structure, benthic

invertebrate biodiversity, trout populations and ecosystem

pro-cesses in restored forest streams: a boreal perspective Freshw

Biol 52:724–737

25 Syrja¨nen J, Kiljunen M, Karjalainen J, Eloranta A, Muotka T

(2008) Survival and growth of brown trout embryos and the

timing of hatching and emergence in two boreal lake outlet

streams J Fish Biol 72:1–16

26 Syrja¨nen J, Valkeaja¨rvi P (2010) Gillnet fishing drives

lake-migrating brown trout to near extinction in tje Lake Pa¨ija¨nne

region, Finland Fish Manage Ecol 17:199–208

27 Korsu K, Huusko A, Muotka T (2009) Does the introduced brook

trout (Salvelinus fontinalis) affect growth of the native brown

trout (Salmo trutta)? Naturwissenschaften 96:347–353

28 Korsu K, Huusko A, Muotka T (2010) Invasion of north

Euro-pean streams by brook trout: hostile takeover or pre-adapted

habitat niche segregation Biol Invasions 12:1363–1375

29 Werner EE (1998) Ecological experiments and a research

pro-gram in community ecology In: Resetarits WJ Jr, Bernardo J

(eds) Experimental ecology issues and perspectives Oxford University Press, New York, pp 3–26

30 Huusko A, Greenberg L, Stickler M, Linnansaari T, Nyka¨nen M, Vehanen T, Koljonen S, Louhi P, Alfredsen K (2007) Life in the ice lane: the winter ecology of stream salmonids Riv Res Appl 23:469–491

31 Kocik JF, Taylor WW (1995) Effect of juvenile steelhead (Oncorhynchus mykiss) on age-0 and age-1 brown trout (Salmo trutta) survival and growth in a sympatric nursery stream Can J Fish Aquat Sci 52:105–114

32 Ikuta K, Kitamura S (1995) Effects of low pH exposure of adult salmonids on gametogenesis and embryo development Water Air Soil Pollut 85:327–332

33 Korsu K, Huusko A, Muotka T (2007) Niche characteristics explain reciprocal invasion success of stream salmonids in dif- ferent continents Proc Natl Acad Sci USA 104:9725–9729

34 Spens J, Alana¨ra¨ A, Eriksson L-O (2007) Nonnative brook trout (Salvelinus fontinalis) and the demise of native brown trout (Salmo trutta) in northern boreal lakes: stealthy, long-term pat- terns? Can J Fish Aquat Sci 64:654–664

35 Cucherousset J, Aymes JC, Poulet N, Santoul F, Ce´re´ghino R (2008) Do native brown trout and non-native brook trout interact reproductively? Naturwissenschaften 95:647–654

36 Korsu K, Huusko A, Muotka T (2010) Impacts of invasive stream salmonids on native fish: using meta-analysis to summarize four decades of research Boreal Environ Res (in press)

37 Baxter CV, Fausch KD, Murakami M, Chapman L (2004) Fish invasion restructures stream and forest food webs by interrupting reciprocal prey subsides Ecology 85:2656–2663

O R I G I N A L A R T I C L E Biology

Can research on the early marine life stage of juvenile chum

salmon Oncorhynchus keta forecast returns of adult salmon?

A case study from eastern Hokkaido, Japan

Toshihiko Saito•Ikutaro Shimizu•

Jiro Seki•Toshiki Kaga• Eiichi Hasegawa•

Hiromi Saito•Kazuya Nagasawa

Received: 1 March 2010 / Accepted: 10 August 2010 / Published online: 15 September 2010

Ó The Japanese Society of Fisheries Science 2010

Abstract To examine the efficacy of juvenile salmon

research as a tool for forecasting adult returns, the results

from a study on the early marine life stage of juvenile

chum salmon, conducted in the Nemuro Strait during

1999–2002 (i.e., 1998–2001 brood years), were compared

with the return rates of adult salmon Among the four brood

years, the 2000 brood year (i.e., salmon migrating to the

sea in 2001) was previously reported as showing higher

abundance, higher growth rate and better somatic condition

during the coastal residency period Consequently, we

expected it to have the highest return rate, under a

hypothesis that juvenile survival in coastal residency

regulates brood-year strength Contrary to this expectation,the 2000 brood year had almost the lowest return rate.Alternatively, a statistical model in which sea surfacetemperature during the first year of marine life and size atrelease were utilized as explanatory variables reconstructedthe actual variability in return rates more accurately thanthat based on the early marine life stage Possible reasonsfor the discrepancy between the results of the juvenilesalmon research and adult returns are discussed, and wesuggest improvements for future research on juvenilesalmon

Keywords Brood-year strength Chum salmon Growth Juvenile salmon research  Return rate Size at release SST  Statistical model

IntroductionThe catch of Pacific salmon Oncorhynchus spp hasincreased substantially around the rim of the North Pacificsince the mid-1970s, and it is believed that in recent years

it has been near historic levels [1,2] This increased catch

is closely linked to climatic and oceanic conditions [1,3].Thus, if the favorable ocean conditions supporting highsalmon production over the last few decades change, it ispossible that production will decline Because of the rapidenvironmental changes caused by global climate change,such as increases in average air and ocean temperatures,widespread melting of snow and ice, and rising average sealevels [4], it seems unlikely that salmon production in thefuture will be as high as it is at present To evaluate cli-matic impacts on salmon production, an understanding isrequired of the oceanic processes regulating salmon sur-vival However, our ability to accurately assess the impact

T Saito ( &)  I Shimizu  J Seki  T Kaga  E Hasegawa

National Salmon Resources Center, Fisheries Research Agency,

Sapporo, Hokkaido 062-0922, Japan

e-mail: brochet@affrc.go.jp

Present Address:

I Shimizu

National Research Institute of Fisheries Science,

Fisheries Research Agency, Yokohama,

National Research Institute of Fisheries Engineering,

Fisheries Research Agency, Kamisu, Ibaraki 314-0408, Japan

H Saito

School of Biological Science and Engineering, Tokai University,

Sapporo, Hokkaido 005-8601, Japan

K Nagasawa

Graduate School of Biosphere Science, Hiroshima University,

Higashi-Hiroshima, Hiroshima 739-8528, Japan

DOI 10.1007/s12562-010-0286-7

of oceanic conditions on salmon survival through field

surveys is very limited; therefore, focused research is

required to obtain hypothesized ocean-driven processes,

such as that done in some recent studies which have used

statistical modeling approaches to analyze relationships

between salmon and oceanic factors [5 7]

The first few months after juvenile Pacific salmon enter

the sea results in mass mortality, so it is one of the most

influential life stages [8 12] Research on the early marine

life stage of Pacific salmon began more than 50 years ago

[13] In Japan, pioneering studies on the migration of

juvenile chum salmon Oncorhynchus keta began in the

1950s, and there is much information on the early marine

life stages of chum salmon [14] Mayama and Ishida [14]

reviewed the early marine life stage of juvenile chum

salmon based on more than 90 scientific articles published

in Japan However, there are only a few articles that link

results from the early marine life stage with adult returns

[15] Evaluations of the status of juvenile salmon

popula-tions during their coastal residency period (about

2–3 months [16]) in relation to factors affecting survival

(e.g., growth, food condition, and coastal environment

factors such as sea temperature) are required, and these

evaluations need to be linked to the number of adult

returns; this would not only identify the biological

mech-anisms that generate year-to-year variability in adult

returns of chum salmon, but would also advance our

understanding of the oceanic conditions that influence

salmon production

In the Nemuro Strait, eastern Hokkaido, Japan,

inter-annual variations in population parameters of juvenile

chum salmon, such as fish density, growth and length–

weight relationships, were investigated during 1999–2002

(i.e., the 1998–2001 brood years) [17] The authors

con-cluded that high values for the population parameters of

2000 brood-year fish (fish migrating to the sea in 2001)

were associated with high zooplankton abundance in the

strait during late May and June 2001 [17] If the early

marine life stage of juveniles is a highly influential phase

that regulates brood-year strength, then the 2000 brood

year would be expected to show relatively high adult

returns compared with other brood years As adults, most

of the 2000 brood year returned to the strait and the rivers

in which they were released, because the age at return for

Japanese chum salmon generally ranges from 2 (0.1) to 8

(0.7) years old (parentheses indicate age based on the

European designation system [18])

We compared the results obtained from the 1998–2001

brood years of juvenile chum salmon in the Nemuro Strait

to their return rates In addition, using a statistical modeling

approach proposed by Saito and Nagasawa [7], the return

rates for these brood years were calculated from the size at

release from hatcheries and coastal and offshore sea

surface temperatures (SSTs) during their first year oceanlife The main objective of this study is to examine whetherthe results of juvenile salmon research and those of sta-tistical models can accurately evaluate the actual brood-year strengths for the 1998–2001 brood years Throughthese comparisons, we point out the problems with previ-ous research on juvenile chum salmon and discussimprovements for future research aimed at assessing brood-year strength

Materials and methodsBiological variables of juvenile chum salmonduring the early marine life stage

Saito et al [17] studied the early marine life stage ofjuvenile chum salmon in the Nemuro Strait, easternHokkaido, Japan (Fig.1) during May and July in1999–2002 Juveniles were collected with purse seine nets

in coastal waters as far as 8 km from the shoreline Fishdensity per m2, specific growth rate (SGR) [19] and acondition index expressed by the fish length–weight rela-tionship [20] were compared among brood years The SGRduring coastal residency was estimated using the otolithback-calculated fork length (FL) at sea entry, FL at capture,and the number of daily otolith increments formed duringsea entry and the capture [17] Of these variables, in thepresent study, fish density was compared with return rates

of adult salmon, and SGR was judged against the earlymarine life stage growth rate estimated from scales takenfrom adult salmon (see below) The results were used toassess whether research on juvenile salmon can be used toforecast adult returns

Statistical modeling of the return rates of Okhotskand Nemuro chum salmon

The return rates of chum salmon for the 1976–1998 broodyears, as derived from indices of fry-to-adult survival ofhatchery-reared salmon, showed a similar year-to-yearvariation in the Okhotsk and Nemuro regions of Hokkaido,Japan [7], which are geographically close to each otheralong the Sea of Okhotsk The Nemuro Strait, where thejuvenile salmon research was carried out during 1999–2002[17], is located in the northern part of the Nemuro region.Saito and Nagasawa [7] investigated coastal SSTs duringNovember of the pre-release year and July in the releaseyear and size at release from the hatcheries in the regions

as potential factors that cause variations in return rate in theOkhotsk and Nemuro regions The underlying hypothesis

of their study was that brood-year strength is stronglyregulated during the early marine life stage in coastal

waters They revealed statistically positive relationships

between the aggregated return rates for the 1976–1998

brood years from these regions and the November–July

SSTs and the average size at release, and successfully

reconstructed actual return rates using a statistical model

that incorporated these variables

We constructed statistical models to reproduce

aggre-gated return rates for 1976–2001 brood-year chum salmon

from the Okhotsk and Nemuro regions The basic methods

used for statistical modeling and for calculating the actual

aggregated return rates of Okhotsk and Nemuro chum

salmon were similar to those described by Saito and

Na-gasawa [7] In the present statistical modeling, however,

we considered November–July SSTs, first wintering SSTs,

and size at release of each brood year as possible

explan-atory variables

November–July SSTs during 1976 and 2002 were first

grouped into principal components (PCs) using principal

component analysis The original SST data, analyzed as

ten-day mean SSTs for 1° latitudinal and longitudinal

square meshes, were obtained from the NEAR-GOOS

Regional Real Time Data Base (http://goos.kishou.go.jp/)

For the coastal area from 142°E to 147°E between 43°N

and 46°N, SST anomalies were calculated at each ten-day

interval from November to July during 1976 and 2002, and

then the anomalies were averaged for each month of year

i (i = 1976–2002) to generate monthly averaged SSTs

Accordingly, the November–July SSTs for each brood year

consist of nine SST variables In the principal component

analysis, these nine SST variables were grouped into a few

PCs under the criterion that each PC has an eigenvaluegreater than 1.0 Once the PCs had been obtained, the PCscores of each PC were calculated for each brood year, andthe PC scores were used as explanatory variables in mul-tiple regression models (see below)

Recently, the fall and winter period after the firstgrowing season during the early marine life stage of Pacificsalmon has been recognized to be a critical period in whichmass mortality takes place [21] During this period (i.e.,first wintering), young chum salmon originating fromJapan shift their distribution from the Sea of Okhotsk to thewestern North Pacific [22, 23] To examine a potentialeffect of the mass mortality during the first winteringperiod (December–May) on overall variability in the returnrates, we used the first wintering SSTs as a possibleexplanatory variable in statistical models The original SSTdata were also obtained from the NEAR-GOOS RegionalReal Time Data Base (http://goos.kishou.go.jp/) For broodyear i (i = 1976–2001), SSTs (anomalies) from December

in year i ? 1 to May in year i ? 2 were averaged for anarea from 156°E to 180°E and between 40°N and 50°N.The period and area were set based on the previous wintersalmon surveys [24, 25] and a hypothesized migrationroute of Japanese chum salmon [26,27] These averagedSSTs were used as an index of the first wintering condition.Size at release for each brood year released from theOkhotsk and Nemuro regions was calculated in the samemanner as described by Saito and Nagasawa [7] At eachbout of release, the number of released fish and the meanweight of individual fish obtained from subsampled fish are

Fig 1 a Location of the (I) Okhotsk and (II) Nemuro regions where

the return rates of 1976–2001 brood-year chum salmon were

calculated The dashed lines indicate the boundaries of the regions.

K, T, and S show the river mouths of the Kushiro, Tokachi, and

Shizunai rivers, respectively F represents Funka Bay b Detailed map

of the northern part of the Nemuro region Dashed lines indicate the

boundary of the region Numbered rivers show river systems where juvenile chum salmon were released during 1999–2002 (i.e., the 1998–2001 brood years) The river number 10 denotes the Ichani River, in which scale samples were collected from returning adults of 4-year-olds during 2002–2005 R and NP represent Rausu and Notsuke Peninsula, respectively

generally recorded in salmon hatcheries In Hokkaido, these

data for major river stocks have been compiled at ten-day

intervals over the release season [7] Based on these data,

each mean weight (weighted by its release number) was

averaged to generate size at release for each brood year

In the present study, five explanatory variables (i.e., the

PC1–3 scores on the basis of the November–July SSTs,

first wintering SSTs, and size at release) were available for

multiple regression modeling (see ‘‘Results’’) We first

constructed all possible models (25= 32 models) to

reproduce aggregated return rates for the 1976–2001

brood-year chum salmon from the Okhotsk and Nemuro

regions Interactions between explanatory variables were

not considered in the present modeling because sample size

(i.e., the number of brood years) was limited (n = 26) For

each model, a small sample AICc was calculated using the

following equation [28]:

AICc¼ AIC þ2PðP þ 1Þ

n P  1;

where AIC is Akaike’s information criterion, n is sample

size, and P is the number of estimable parameters in a

model (the number of explanatory variables plus 2) Since

we used least squares model fitting in this study, AIC can

where RSS is the residual sum of squares for the model,

and n and P are the same as in the above description

Based on the AICc values, we selected six models with

0–5 explanatory variables, where each model was the most

appropriate for a particular number of explanatory

vari-ables When a model had the lowest AICc value, that

model was judged to be the most appropriate model In the

same manner, we selected the best available model of the

six models mentioned above When the difference in

AICc values between the best and another models was

smaller than two units, a likelihood ratio test (LRT) was

carried out In general, the LRT statistic can be written as:

LRT¼ AICp  AICp þ i;

where AICp is the AIC value of the null model with

p parameters, and AICp ? i is that of the alternative model

with p ? i parameters [28] Since this statistic is expected

to be described using the v2distribution with i degrees of

freedom, LRT statistics were tested with the v2test

Both the aggregated return rates and size at release

showed increasing trends over 26 brood years Such

long-term trends may cause a spurious relationship between the

return rates and other explanatory variables To exclude the

effects of long-term trends, both time series were detrended

by calculating residuals from fitted linear regressions, and

these residuals (i.e., detrended aggregated return rate andsize at release) were also used for statistical modeling inthe same manner as mentioned above

Finally, the actual and calculated return rates for the1998–2001 brood years were compared with the results forfish density obtained from the juvenile salmon research, aswell as return rates of adult salmon that returned to thenorthern part of the Nemuro Strait and natal rivers

Adult returns of chum salmon in the Nemuro Straitand natal rivers

The 1998–2001 brood-year juvenile chum salmon werereleased into 10–11 river systems that enter the NemuroStrait during the spring of 1999–2002 (Fig.1) The num-bers of released juvenile chum salmon were 113.3 million

in 1999, 107.0 million in 2000, 105.3 million in 2001, and111.2 million in 2002 [29] These fish were expected toreturn as adults to the strait and the released rivers during2000–2009 because, as mentioned above, age at return ofJapanese chum salmon generally ranges from 2 to 8 yearsold; however, in a given a brood year, more than 95% ofthe fish that return to their natal rivers B6 years old [7].Thus, for the 1998–2001 brood years, we calculated thereturn rates based on total adult returns of 2- to 6-year-oldfish that returned to the strait and rivers during 2000–2007.Information on coastal (i.e., mixed-stock) and river (ter-minal) catches of adult salmon during 2000–2005 and age-composition data are available from the Salmon Database[30] Corresponding information for the 2006 and 2007 fishwas obtained from unpublished data collected by theNational Salmon Resources Center (NASREC) To assigncoastal and river catches of adult salmon to their brood years,age-composition data were required for each year Althoughthe number of river catches and age-composition data werenot available for all individual rivers where juveniles werereleased during 1999–2002, data for the former and latterwere collected from 5–6 and 2–4 rivers, respectively, during

2000 and 2007 ([30], NASREC, unpublished data,2006–2007) The age-composition data were generallycollected from rivers where adult returns were abundanteach year Accordingly, age-composition data from eachriver were annually weighted by the corresponding rivercatch to generate annual averages of age composition forriver catches, and these annual averages were then used toassign river catches to brood years in rivers where the actualage composition was unknown Coastal catches were alsoallocated to brood years using these averaged age compo-sitions, because coastal catches were thought to be mainlycomposed of stocks from the same rivers

To evaluate the 1998–2001 brood-year strengths ofsalmon that returned to the strait and their natal rivers as

adults, we calculated two types of return rates: the total

return rate, calculated from the sum of the coastal and river

catches of each brood year divided by the total release

number (as a percentage) from the 10–11 river systems;

and the river return rate, calculated from the sum of only

river catches of each brood year divided by their total

release number (as a percentage) Recently, the recovery of

otolith thermally marked adult salmon from coastal catches

in the Nemuro region demonstrated that 39.8% of the

marked fish landed in the Nemuro region were released

from rivers located in various other regions of Hokkaido

[31] This suggests that the total return rates may have been

overestimated due to landings of salmon originating from

other regions Taking this possibility into account, we used

the river return rates as well as the total return rates

Early marine life stage growth rates estimated

from adult scales

If early marine growth affects survival of juvenile salmon,

survivors would tend to have higher growth during early

ocean life In the juvenile salmon research conducted in the

Nemuro Strait, somatic growth rates, which were estimated

from daily otolith increments of juvenile fish, showed an

annual variability among the 1998–2001 brood years [17]

To examine whether the annual variability was observed in

adult returning salmon, scale circulus spacings were

examined on scales collected from 4-year-old adult salmon

that returned to the Ichani River (Fig.1) during

2002–2005 Scale circulus spacing is a good indicator of

the past growth of an individual fish [32] To evaluate the

early marine life stage growth of adult returning salmon,

about 100 individuals were measured for FL and body

weight, and sex was determined by examining external

features (i.e., sexual dimorphism) every ten days during the

fishery season (September–November) Scales were also

taken for age-composition analysis In the laboratory, the

scales were pressed on acetate slides to make impressions

of them From these impressions, we randomly selected

100 individual scales (i.e., 50 individuals of each sex) from

samples from each year and measured the scale circulus

spacings formed inside the first annulus along the longest

axis of scale radius Measurements of the scale circulus

spacings were made with a scale measurement system

(ARP/W v.5.20; Ratoc System Engineering Co., Ltd.,

Tokyo, Japan) at 259 magnification Juvenile chum

sal-mon at 100 mm FL have about 10–12 scale circuli on their

scales (Fig 8C in Kaeriyama [33]) Because most juvenile

salmon collected from the Nemuro Strait during

1999–2002 had FL \ 100 mm [17], we assumed that the

early marine life stage growth was mirrored by the scale

circulus spacings within the tenth circulus For each brood

year, we calculated two types of averaged scale circulus

distances: the distance from the scale focus to the fifthcirculus (5CW), and the distance from the fifth to the tenthcirculi (5–10CW) Both types of scale circulus distanceswere compared using a two-way analysis of variance(ANOVA) in which brood year and sex effects were themain factors, and with the interaction of both effectsincluded The sex effect was considered to regulate apotential source of variations that might affect somaticgrowth, because the difference in size at maturity is evidentbetween male and female Pacific salmon [34] The results

of these analyses were compared with the annual SGRmeans from the juvenile salmon research [17] Statisticalanalyses were based on the two-tailed condition with thesignificance level, a, at 0.05, and were performed usingSPSS 17.0 software (SPSS Japan Inc., Tokyo, Japan)

ResultsStatistical model for the 1976–2001 brood-year chumsalmon in the Okhotsk and Nemuro regions

The November–July monthly SSTs during 1976–2002were grouped into three PCs (Table1) These PCsaccounted for 76.6% of the variability of the original SSTs.PC1 factor loadings were higher for the SSTs in Januaryand April, suggesting that PC1 was mainly associated withthe SSTs during these months Likewise, PC2 was associ-ated with the SSTs in June and July, and the PC3 wascoupled with the SST in November Saito and Nagasawa[7] examined the November–July monthly SSTs during1976–1999 and extracted the same number of PCs.According to the best available model (model 5), thereturn rates for the 1976–2001 brood years that originatedfrom the Okhotsk and Nemuro regions were calculatedfrom the size at release, the first wintering SSTs, and thePC2–3 scores (Table2; Fig.2) However, differences in

Table 1 Factor loadings of principal components (PCs) obtained from a principal component analysis based on November–July sea surface temperatures in coastal waters off the Okhotsk and Nemuro regions (43–46°N, 142–147°E) during the period 1976–2002

AICc values were smaller than two units for model 3 (two

explanatory variables), model 4 (three explanatory

vari-ables) and model 6 (full model) (Table2) LRT revealed

statistically significant differences between model 3 and

model 5 (v2= 8.408, df = 2, P \ 0.05) and between

model 4 and model 5 (v2= 4.450, df = 1, P \ 0.05),

although there was no significant difference between model

5 and model 6 (v2= 3.786, df = 1, P [ 0.05)

Conse-quently, model 5 was regarded as the best in terms of the

principle of parsimony (i.e., the model with the smallest

possible number of parameters is best) Standardized

coefficients indicated that size at release was the most

influential variable for return rates among the explanatory

variables Because the regression coefficients of the size at

release and the PC2–3 scores were positive, a larger size at

release and higher PC scores were expected to be

associ-ated with a higher return rate in these regions In particular,

PC2 scores for the 1998–2001 brood years were higher

than those for the other brood years (Fig.3); the PC2 score

for the 2001 brood year was the highest among the 26

brood years examined Since the PC2 scores were ated with the SSTs in June and July, warmer SST condi-tions during the coastal residency were probablyadvantageous to the survival of juvenile salmon In contrast

associ-to these explanaassoci-tory variables, the regression coefficient ofthe first wintering SSTs had a negative value, even though

it was not statistically significant in model 5 (Table 2) Inthe other models (Table 2) the first wintering SSTs had allnegative coefficients

Statistically significant linear regressions could be fitted

to the time series of the aggregated return rates transformed return rate = 0.364 9 brood year - 709.682,

Table 2 The results for the most appropriate models with 0–5

explanatory variables (original data) that were used to reconstruct

interannual variability in the aggregated return rates (original data) for

1976–2001 brood-year chum salmon released from the Okhotsk and Nemuro regions, Hokkaido, Japan

-1.586*

(0.696) -0.202

-1.391*

(0.668) -0.177

0.560 (0.294) 0.173

-1.264 (0.631) -0.161

0.634*

(0.279) 0.196

0.524 (0.264) 0.162

-1.928*

(0.708) -0.246

0.503 (0.284) 0.156

0.612*

(0.266) 0.189

0.514 (0.252) 0.159

32.487*** (5, 20) 0.89 (0.863) 22.757 0.015

To construct the models, the return rates were arcsine-transformed

DAICc = AICc (i) - AICc (model 5), where i represents the i-th model

The model indicated by bold letters (model 5) was judged to be the best of these six models based on AICc values

Const, constant; SR, size at release; 1 W, first wintering SSTs; PC1–3, PC1–3 scores obtained from principal component analysis (Table 1 )

* P \ 0.05, ** P \ 0.01, *** P \ 0.001

a The upper, middle (in parentheses), and lower (in italics) values of the explanatory variables indicate model coefficients, standard errors, and standardized coefficients, respectively

consisted of four explanatory variables: the detrended size

at release, the first wintering SSTs, and PC1–2 scores

However, the differences in AICc values were trivial

between the model D5 and others: model D4 versus model

D5 (LRT: v2= 3.499, df = 1, P [ 0.05), model D5 versus

model D6 (LRT: v2= 3.761, df = 1, P [ 0.05), and

model D3 versus model D5 (LRT: v2= 7.047, df = 2,

P\ 0.05) Thus, model D4 and D6 were considered to fit

the data almost as well as model D5 In these tied models,

the first wintering SSTs was the most influential

explana-tory variable, and showed negative coefficients The other

explanatory variables (detrended size at release and PC1–3

scores) all had positive coefficients Although the effect of

size at release on variability in return rates was reduced bythe detrended manipulation, it still remained an influentialfactor in these models

Interannual variation in the return ratesfor the 1998–2001 brood-year chum salmon

in the Nemuro Strait and natal rivers

In the northern part of the Nemuro region, total return ratesfor the 1998–2001 brood years varied from 10.3 to 16.1%,and the river return rates for the same brood years fluctu-ated from 0.16 to 0.39% (Fig 4a) The former was 42–76times higher than the latter in these brood years Interan-nual variation in the total and river return rates showed asimilar trend with a few exceptions: the 2001 brood yearhad the highest total and river return rates among the fourbrood years, and both of the total and river return rates forthe 1999 and 2001 brood years were higher than those for

1998 and 2000 Of these four brood years, the 2000 broodyear had the lowest value for the total return rate and thesecond lowest value for the river return rate Average ages

at return (i.e., maturity) for the 1998–2001 brood yearscaught in coastal waters (their natal rivers) were 4.28(4.29), 4.35 (4.47), 4.25 (4.33), and 4.44 (4.40) years old,respectively There was a tendency for a higher return rate(Fig.4a) to be associated with older age at return.The multiple regression model (model 5) accuratelyreconstructed the aggregated return rates for the 1998–

2002 brood years from the Okhotsk and Nemuro regions,even though the actual return rate for the 2000 brood yearwas slightly below the lower limit of the 95% confidenceinterval (Fig.4b)

Annual average juvenile densities (fish per squaremeter) generated from Table1 of Saito et al [17] areshown in Fig.4c Interannual variations in juvenile densitywere very different from those of the total and river returnrates for the 1998–2001 brood years In particular, the 2000brood year had the highest fish density, but its return ratewas almost the lowest of the four brood years Further-more, among the 1976–2001 brood years, 1999 and 2001had the first and second highest return rates, respectively(Fig.2); however, there was no indication of high brood-year strengths for these years based on information fromtheir early marine life stage in the Nemuro Strait

Interannual variation in early marine life stage growthamong the 1998–2001 brood years

The scale distances 5CW and 5–10CW are shown in Fig.5.For the 5CW, there was no statistical significant differenceamong brood years; however, there was a significant dif-ference between males and females, with no interactionbetween brood year and sex effects (two-way ANOVA,

Fig 2 Actual and calculated return rates for 1976–2001 brood-year

chum salmon released from the Okhotsk and Nemuro regions The

calculated return rates were obtained from the results of a multiple

linear regression (model 5), as shown in Table 2 Dashed lines

indicate the upper and lower limits of the 95% confidence interval

Fig 3 Scores for principal component 2 (PC2) obtained from a

principal component analysis based on November–July sea surface

temperatures in the coastal waters of the Okhotsk and Nemuro regions

(Table 1 ) This was used as an explanatory variable in the multiple

linear regression analysis shown in Table 2 and Fig 2

brood-year effect: F(3, 392) = 2.616, P [ 0.05; sex effect:

F(1, 392)= 4.899, P \ 0.05; brood-year 9 sex effect:

neither a brood-year effect nor a sex effect; however, there

was significant interaction (two-way ANOVA, brood-year

effect: F(3, 392)= 1.216, P [ 0.05; sex effect: F(1, 392)=

3.086, P [ 0.05; brood-year 9 sex effect: F(3, 392)=

4.284, P \ 0.01) When these brood years inhabited the

Nemuro Strait as juvenile salmon during 1999 and 2002,

the 2000 brood year had higher SGRs than the other three

brood years (Fig 7 in Saito et al [17]) However, there was

no such trend in the 5CW and 5–10CW on the scales of

returning adult salmon

Discussion

The early marine life stage is critical for Pacific salmon,

and research on juveniles is expected to be able to identify

the mechanisms regulating mass mortality and to enable

the forecasting of returning adults Saito et al [17] basedtheir study on juvenile salmon research conducted in theNemuro Strait during 1999 and 2002; they reported that the

2000 brood year, which migrated to the sea in 2001,experienced better conditions during their early marine lifestage compared with those experienced by the other threebrood years However, the return rate for the 2000 broodyear was the lowest of the four brood years Furthermore,the 1999 and 2001 brood years had the first and secondhighest return rates, respectively, among the 1976–2001brood years in the Okhotsk and Nemuro regions (Fig.2);however, there was no indication from the juvenile salmonresearch of such high return rates Consequently, the resultsfrom the juvenile salmon research for the 1998–2001brood-year chum salmon in the Nemuro Strait were unable

to predict the number of adult returns

In our study, the return rates for the 1976–2001 broodyears in the Okhotsk and Nemuro regions were reproducedusing statistical models based on the average size at releasefor the brood years, coastal SSTs during November and

Table 3 The results for the most appropriate models with 0–5

explanatory variables that were used to reconstruct interannual

variabil-ity in the aggregated return rates (detrended data) for 1976–2001

brood-year chum salmon released from the Okhotsk and Nemuro regions, Hokkaido, Japan

1.068 (0.330) 0.237

6.020** (2, 23) 0.344 (0.287) 23.957 0.531

D4 -0.028

(0.261)

7.810 (4.355) 0.356

-2.256**

(0.765) -0.564

0.729 (0.368) 0.442

-2.156**

(0.734) -0.539

0.673 (0.353) 0.408

0.447 (0.257) 0.271

-2.117**

(0.700) -0.529

0.680 (0.337) 0.412

0.446 (0.245) 0.270

0.429 (0.243) 0.260

5.235** (5, 20) 0.567 (0.459) 23.466 0.040

To construct the models, the return rates were arcsine-transformed

DAICc = AICc (i) - AICc (model 5), where i represents the i-th model

The model indicated by bold letters (model 5) was judged to be the best of these six models, based on AICc values

Const, constant; SR, size at release; 1 W, first wintering SSTs; PC1–3, PC1–3 scores obtained from principal component analysis (Table 1 )

* P \ 0.05, ** P \ 0.01, *** P \ 0.001

a The upper, middle (in parentheses), and lower (in italics) values of the explanatory variables indicate model coefficients, standard errors, and standardized coefficients, respectively

July, and the first wintering SSTs (December–May) The

actual return rates for the 1998–2001 brood years were

accurately estimated using the best available model (model

5), even though the actual return rate of the 2000 brood

year was slightly below the lower limit of the 95%

confi-dence interval of the calculated return rates (Fig.4b)

Although a few actual return rates were beyond the 95%

confidence intervals of prediction, the statistical model

could reproduce the actual return rates over 26 brood years

(Fig.2) This demonstrates that the size at release and the

marine environmental conditions expressed by the SSTs

are influential factors in determining the brood-year

strengths of chum salmon in the Okhotsk and Nemuro

regions

Recently, the fall and winter period after the firstgrowing season during the early marine life stage of Pacificsalmon has been recognized as a critical period in whichmass mortality takes place [21] In this study, the effect ofthe first wintering SSTs was detected as a meaningfulexplanatory variable in several statistical models (Tables2,

3) Since all coefficients of this explanatory variable werenegative, this suggests that warmer SST conditions duringthe first wintering period are associated with reduced sur-vival of young salmon The winter distribution of salmon isconcentrated in northern cold waters of narrow latitudinalranges at SSTs ranging from 4 to 8°C [24] Nagasawa [24]hypothesized that young salmon prevent energy loss byreducing their metabolic rates at low temperatures underpoor food conditions during the wintering period Themetabolic rates of salmonids increase almost exponentiallyover the full range of tolerated temperature [35] Thus,warmer SSTs during the wintering period of young salmonmay put them at a disadvantage under limited food con-ditions However, research on young salmon during theirfirst wintering is limited [24, 25], so actual effects ofwintering conditions on survival of young salmon are stillunclear

Chum salmon that have migrated from rivers in Japanare distributed widely in the North Pacific and its adjacentseas before they return to their natal rivers as adults [26,

27]; hence, examining the relative importance of survival

in different ocean habitats is a major challenge because ofthe lack of survey information Greene et al [6] con-structed statistical models to evaluate return rates of wildChinook salmon in the Skagit River, Washington Theirmodels incorporated various environmental effects, such asthe flood recurrence interval in the river, SST, sea levelpressure and coastal upwelling at multiple life stages fromfreshwater to marine habitats, and explained up to 90% ofthe variation in the return rates; i.e., very high forecastingprecision Future research aimed at elucidating the relativeinfluence of factors regulating brood-year strength of chumsalmon in Japan at different life stages may benefit from asimilar statistical approach

Our statistical models were more effective for ing the effects of marine environmental conditions on thevariability of adult salmon returns than was research con-ducted on juvenile salmon during their coastal residency.Because the statistical models suggested that coastalenvironmental conditions during the coastal residencyperiod of juvenile salmon had a major influence on sur-vival, the juvenile salmon research might not have ade-quately assessed the status of the 1998–2001 brood yearsduring the coastal residency We believe that the reasonsfor the discrepancy between the results of juvenile salmonresearch and adult returns are important for improvingfuture juvenile salmon research

examin-Fig 4 a Total and river return rates for 1998–2001 brood-year chum

salmon in the northern part of the Nemuro Region The total return

rates were calculated from coastal and river catches of adult salmon,

and the river return rates were generated on the basis of river catch

only b Actual and calculated return rates for 1998–2001 brood-year

chum salmon in the Okhotsk and Nemuro regions Bars indicate 95%

confidence intervals This graph was redrawn from the data used in

Fig 2 to enable comparisons with a and c c Average juvenile salmon

densities observed during juvenile salmon research carried out in the

Nemuro Strait during 1999–2002 Bars denote standard deviations.

These data were obtained from Table 1 in Saito et al [ 17 ]

The total return rates for the 1998–2001 brood years

returning to the northern part of the Nemuro region ranged

from 10.3 to 16.1% (Fig.4a) These return rates are

extremely high compared with those previously obtained

from various regions of Japan: average return rates for the

1976–1998 brood years ranged from 0.3% in the Honshu

Sea of Japan to 5.9% in the Nemuro region [7]

Further-more, among the otolith thermally marked adult salmon

collected from landings in the Nemuro region (256

indi-viduals), 39.8% were fish released from rivers located in

various other regions of Hokkaido [31] Since this

pro-portion was calculated based on pooled samples collected

during 2001 and 2007, we should not discuss the degree of

interceptions in the Nemuro region with this information

However, these findings imply that coastal harvests in the

region probably contain a nonnegligible quantity of salmon

originating from other regions Consequently, interceptions

may have masked the real variability in the total return

rates for 1998–2001, which possibly affected the observed

discrepancy between the results of juvenile salmon

research and adult returns However, the total return rates

and the river ones varied annually with a similar trend over

the four brood years (Fig.4a) This indicates that relative

survival of these brood years could be fairly evaluated

regardless of the intercepted quantity in coastal fisheries

If age at maturity is delayed, for instance due to reduced

offshore growth [36], then a brood year would experience

higher mortality throughout their lifetimes Such a delayedage at maturity may cause a difference between juvenilesurvival and adult returns In this study, however, averageage at return tended to increase in brood years showinghigher return rates Accordingly, the discrepancy betweenhigh juvenile abundance and poor adult returns for the

2000 brood year cannot be explained with this scenario.The study area for the research conducted on juvenilesalmon in the Nemuro Strait during 1999–2002 extendedalong about 50 km of shoreline from near the NotsukePeninsula to Rausu [17] If juvenile salmon dispersedbeyond the study area during their coastal residency, thedata on salmon collected during the study were possiblybiased; the results from such biased data may not reflect thestatus of the brood years Recently, recaptures of otoliththermally marked juvenile chum salmon have revealed thatjuvenile chum salmon inhabit coastal waters at a greaterdistance from the point of sea entrance than previouslythought For instance, otolith thermally marked fishreleased from the Kushiro (42°5905000N, 144°2102500E) andTokachi (42°4104200N, 143°3905700E) rivers in easternHokkaido were recaptured in coastal waters near Shiraoi(42°290N, 141°160E) in western Hokkaido, and thosefish released from the Shizunai River (42°1904500N,142°2200000E) were found in Funka Bay (42°160N,140°190E) west of the Shizunai River [37] The distancebetween the sea entrance and recapture points for these fish

Fig 5 Boxplots of average

scale circulus spacings for

4-year-old adult chum salmon

returning to the Ichani River,

eastern Hokkaido, Japan.

Rectangles denote the ranges of

values within the 25th and 75th

percentiles of the distribution,

and the lines inside the

rectangles represent the median.

The horizontal lines show the

values of the 10th and 90th

percentiles of the distribution.

Circles within the plots indicate

the mean values of each

measurement The 5CW and

5–10CW mean scale distances

are those between the scale

focus and the fifth circulus and

those between the fifth and tenth

circuli, respectively

is estimated as 169–259 km in a straight line, which is

greater than the length of the 50 km study area

Further-more, the recapture points for these fish were in the

opposite direction from the migration route of juvenile

chum salmon reported by Irie [16] It is unknown whether

juvenile chum salmon occurring westward of their natal

rivers survive during their coastal residency and migrate

successfully to the Sea of Okhotsk However, these

find-ings strongly suggest that the coastal area in which juvenile

salmon reside extends far beyond the coastal area in the

vicinity of the river mouth of their natal river Thus, an

appropriate study area scale should be selected when

attempting to forecast brood-year strengths based on

juvenile salmon research

In our study, the early marine life stage growth rate

during the coastal residency, as estimated by scale circulus

spacings of 4-year-old adult salmon, did not differ among

the 1998–2001 brood years However, the SGRs, estimated

by otolith daily increments of juvenile salmon collected

from the Nemuro Strait, were statistically significant

dif-ferent among brood years In particular, the 2000 brood

year had higher SGRs than those for the other three brood

years [17] Consequently, the early marine life stage

growth rate showed a discrepancy between juvenile salmon

and returning adults In addition to the possibility of biased

sampling during the juvenile salmon research (as

men-tioned above), differences in the otolith and scale materials

may have caused the observed discrepancy, at least in part

Wells et al [38] demonstrated that increment patterns of

variation in the otoliths and scales of mature Atlantic

sal-mon Salmo salar are similar at a subseasonal resolution

However, the SGRs estimated by Saito et al [17] were

based on daily growth increments on otoliths of juvenile

salmon Therefore, in future studies aimed at comparing

the early marine life stage growth rates of juvenile salmon

and those of adult salmon, the question of whether

daily-based growth estimates on otoliths are comparable to

growth estimates based on scale circulus spacings should

be examined Alternatively, comparing the scale radius of

juvenile salmon with those of returning adults may give us

more comprehensive results for evaluating whether the

early marine life stage growth rate affects survival to

maturation (e.g., Moss et al [39])

Research on juvenile salmon in the Nemuro Strait [17]

failed to forecast the brood-year strengths of 1998–2001

brood-year chum salmon that returned as adults In more

than 50 years of juvenile salmon research in Japan, this

was a rare case study that attempted to link the results of

juvenile salmon research to adult returns According to a

review of research on the early marine period of Pacific

salmon [13] among countries where Pacific salmon exist

(Canada, Japan, Korea, Russia and the United States), only

Russia has, until recently, been using information from

juvenile salmon research to complement forecasts of thenumber of returning adults In Russia, growth indices ofjuvenile salmon, such as body size composition and therate of scale formation, and abundance estimates of youngsalmon in autumn surveys are used to adjust fishing fore-casts 8–9 months before the fishing is carried out [40] Inparticular, the autumn surveys, which are performed with aspecial pelagic trawl, cover vast offshore areas in the Sea

of Okhotsk and the Bering Sea [40], implying that anintensive survey over a spatially extended area may beneeded to quantitatively forecast returning salmon based oninformation from juvenile salmon research At present, insoutheast Alaska, peak average catch per unit effort (peakCPUE) of juvenile pink salmon O gorbuscha, which isobtained in a juvenile salmon research project named theSoutheast Alaska Coastal Monitoring (SECM) project, issuccessfully used to forecast pink salmon harvests of theregion for the subsequent year (Alaska Fisheries ScienceCenter: http://www.afsc.noaa.gov/abl/, accessed 18 Feb2010) The SECM research was started in 1997, and hascontinued annually since then Juvenile salmon are col-lected at monthly sampling intervals from May to Augustwith a surface trawl around Icy Strait, a major migrationcorridor for juvenile salmon moving to offshore waters.The sampled juvenile salmon are regarded as survivorsafter their initial marine residency where mass mortalitytakes place, and therefore the peak CPUE seems to be aneffective indicator for forecasting the future harvest.Japanese chum salmon are mainly produced in salmonhatcheries, and their early marine life in coastal waters is acritical period in which the brood-year strengths are largelydetermined; however, forecasting the number of returningadult salmon from the results of juvenile salmon research is

a challenging task More knowledge of juvenile chumsalmon in the coastal waters of Japan is required, such asspatial and temporal distributions, migration routes to theSea of Okhotsk, and growth and mortality variability, as isthe development of reliable methods for assessingabundance

Acknowledgments We are grateful to the staff of the Nemuro, Ichani and Nijibetsu field stations of NASREC for valuable advice and biological data on adult salmon used in the study Two anony- mous reviewers provided constructive comments that improved the early version of the manuscript.

3 Mantua NJ, Hare SR, Zhang Y, Wallace JM, Francis RC (1997)

A Pacific interdecadal climate oscillation with impacts on salmon

production Bull Am Meteorol Soc 78:1069–1079

4 Intergovernmental Panel on Climate Change (IPCC) (2007)

Cli-mate change 2007 synthesis report IPCC, Geneva

5 Briscore RJ, Adkison MD (2005) Biophysical factors associated

with the marine survival of Auke Creek, Alaska, coho salmon.

Trans Am Fish Soc 134:817–828

6 Greene CM, Jensen DW, Pess GR, Steel EA (2005) Effects of

environmental conditions during stream, estuary, and ocean

res-idency on Chinook salmon return rates in the Skagit River,

Washington Trans Am Fish Soc 134:1562–1581

7 Saito T, Nagasawa K (2009) Regional synchrony in return rates

of chum salmon (Oncorhynchus keta) in Japan in relation to

coastal temperature and size at release Fish Res 95:14–27

8 Parker RR (1968) Marine mortality schedules of pink salmon of

the Bella Coola River, central British Columbia J Fish Res Bd

Canada 25:757–794

9 Healey MC (1982) Timing and relative intensity of size-selective

mortality of juvenile chum salmon (Oncorhynchus keta) during

early sea life Can J Fish Aquat Sci 39:952–957

10 Bax NJ (1983) Early marine mortality of marked juvenile chum

salmon (Oncorhynchus keta) released into Hood Canal, Puget

Sound, Washington, in 1980 Can J Fish Aquat Sci 40:426–435

11 Pearcy WG (1992) Ocean ecology of North Pacific salmonids.

University of Washington Press, Seattle

12 Quinn TP (2005) The behavior and ecology of Pacific salmon and

trout University of Washington Press, Seattle

13 North Pacific Anadromous Fish Commission (2003) A review of

the research on the early marine period of Pacific salmon by

Canada, Japan, Russia, and the United States N Pac Anadr Fish

Comm Bull 3, Vancouver

14 Mayama H, Ishida Y (2003) Japanese studies on the early ocean

life of juvenile salmon N Pac Anadr Fish Comm Bull 3:41–67

15 Hayano H, Asami H, Hirano K (1997) Abundance of

micro-animals (especially, Harpacticoid copepods) during the release of

juvenile chum salmon, Oncorhynchus keta and adult return rate

into the coastal area off Mashike, Japan Sea, northern Hokkaido.

Sci Rep Hokkaido Fish Hatch 51:11–16 (in Japanese with English

abstract)

16 Irie T (1990) Ecological studies on the migration of juvenile

chum salmon, Oncorhynchus keta, during early ocean life Bull

Seikai Natl Fish Res Inst 68:1–142 (in Japanese with English

summary)

17 Saito T, Shimizu I, Seki J, Nagasawa K (2009) Relationship

between zooplankton abundance and marine life history of

juvenile chum salmon Oncorhynchus keta in eastern Hokkaido,

Japan Fish Sci 75:303–316

18 Koo TSY (1962) Age designation in salmon In: Koo TSY (ed)

Studies of Alaska red salmon University of Washington Press,

Seattle, pp 37–48

19 LeBrasseur RJ, Parker RR (1964) Growth rate of central British

Columbia pink salmon (Oncorhynchus gorbuscha) J Fish Res Bd

Canada 21:1101–1128

20 Vila-Gispert A, Moreno-Amich R (2001) Fish condition analysis

by a weighted least squares procedure: testing geographical

dif-ferences of an endangered Iberian cyprinodontid J Fish Biol

58:1658–1666

21 Beamish RJ, Mahnken C (2001) A critical size and period

hypothesis to explain natural regulation of salmon abundance and

the linkage to climate and climate change Prog Oceanogr

49:423–437

22 Ueno Y, Ishida Y (1996) Summer distribution and migration routes of juvenile chum salmon (Oncorhynchus keta) originating from rivers in Japan Bull Nat Res Inst Far Seas Fish 33:139–147

23 Ueno Y (1998) Distribution, migration, and abundance estimation

of Asian juvenile salmon Salmon Rep Ser 45:83–103

24 Nagasawa K (2000) Winter zooplankton biomass in the subarctic North Pacific, with a discussion on the overwintering survival strategy of Pacific salmon (Oncorhynchus spp.) N Pac Anadr Fish Comm Bull 2:21–32

25 Fukuwaka M, Sato S, Takahashi S, Onuma T, Sakai O, Tanimata

N, Makino K, Davis ND, Voklov AF, Seong KB, Moss JH (2007) Winter distribution of chum salmon related to environmental variables in the North Pacific N Pac Anadr Fish Comm Tech Rep 7:29–30

26 Urawa S (2000) Ocean migration route of Japanese chum salmon with a reference to future salmon research Natl Salmon Resour Center Newsl 5:3–9 (in Japanese)

27 Yatsu A, Kaeriyama M (2005) Linkages between coastal and open-ocean habitats and dynamics of Japanese stocks of chum salmon and Japanese sardine Deep Sea Res II 52:727–737

28 Burnham KP, Anderson DR (2002) Model selection and model inference: a practical information-theoretic approach, 2nd edn Springer, New York

multi-29 National Salmon Resources Center (2000–2003) Database on early life history of Pacific salmon in Japan; Salmon Database 8-11(3) National Salmon Resources Center, Sapporo (in Japanese)

30 National Salmon Resources Center (2002–2008) Database on biological assessment of Pacific salmon populations in Japan; Salmon Database 10-15(1) National Salmon Resources Center, Sapporo (in Japanese)

31 Takahashi H (2009) Findings obtained from otolith-thermally marked salmon FRA Salmon Res Rep 3:6–7 (in Japanese)

32 Fisher JP, Pearcy WG (2005) Seasonal changes in growth of coho salmon (Oncorhynchus kisutch) off Oregon and Washington and concurrent changes in the spacing of scale circuli Fish Bull 103:34–51

33 Kaeriyama M (1989) Comparative morphology and scale mation in four species of Oncorhynchus during early life Jpn J Ichthyol 35:445–452

for-34 Salo EO (1991) Life history of chum salmon (Oncorhynchus keta) In: Groot C, Margolis L (eds) Pacific salmon life histories USB, Vancouver, pp 231–309

35 Brett JR (1995) Energetics In: Groot C, Margolis L, Clarke WC (eds) Physiological ecology of Pacific salmon USB, Vancouver,

pp 1–68

36 Morita K, Morita SH, Fukuwaka M, Matsuda H (2005) Rule of age and size at maturity of chum salmon (Oncorhynchus keta): implications of recent trends among Oncorhynchus spp Can J Fish Aquat Sci 62:2752–2759

37 Nara K (2006) Results of the first five-year program of NASREC Natl Salmon Resour Center Newsl 16:1–3 (in Japanese)

38 Wells BK, Friedland KD, Clarke LM (2003) Increment patterns

in otoliths and scales from mature Atlantic salmon Salmo salar Mar Ecol Prog Ser 262:293–298

39 Moss JH, Beauchamp DA, Cross AD, Myers KW, Farley EV Jr, Murphy JM, Helle JH (2005) Evidence for size-selective mor- tality after the first summer of ocean growth by pink salmon Trans Am Fish Soc 134:1313–1322

40 Karpenko VI (2003) Review of Russian marine investigations of juvenile Pacific salmon N Pac Anadr Fish Comm Bull 3:69–88

O R I G I N A L A R T I C L E Biology

Development of sensory organs and changes of behavior in larvae

of the sutchi catfish, Pangasianodon hypophthalmus

Yukinori Mukai•Audrey Daning Tuzan •

Sitti Raehanah Muhamad Shaleh•

Bernardette Mabel Manjaji-Matsumoto

Received: 13 May 2010 / Accepted: 25 August 2010 / Published online: 20 October 2010

Ó The Japanese Society of Fisheries Science 2010

Abstract Larvae of the sutchi catfish Pangasianodon

hypophthalmus hatch with morphologically immature

fea-tures, but sensory organs develop rapidly as the fish grow

By 1 day old, yolk-sac larvae showed notochord flexion,

and by 2 days old larvae were observed to have consumed

a large part of the yolk sac At this stage, larvae had

well-developed eyes, olfactory organs with ciliated receptor

cells, inner ears with semicircular canals, and numerous

taste buds, and they commenced ingestion of rotifers,

Artemia nauplii, and artificial compound feed

Two-day-old larvae had many free neuromasts on the surface of the

head and flanks and clearly showed rheotaxis By 20 days

old, free neuromasts in postflexion larvae had sunk under

the skin At this later stage, larvae swam against a water

current and schooled along the side of the fish tank Rapid

development of sensory organs and notochord flexion

would be an adaptation for survival in conditions of

flowing water, as in the Mekong River In this study, we

show that development of the lateral line in the postflexion

stage seems to be closely related to larval behavior,

suggesting that these developments could be essential for

sutchi catfish larvae survival

Keywords Cannibalism
Larvae
Pangasianodon

hypophthalmus
Sensory organs
Sutchi catfish

IntroductionThe sutchi catfish Pangasianodon hypophthalmus (syno-nym, Pangasius hypophthalmus) originates from an areaextending from the Mekong River Basin in Vietnam to theChao Phraya River in Thailand [1,2] The sutchi catfish is

a popular fish for aquaculture It is widely cultured in largeamounts in Asian countries including Vietnam, Malaysia,Indonesia, Laos, Cambodia, and China [1 5] Vietnam, thelargest sutchi catfish producing country in the region [6],produced 1,200,000 tons of catfishes in 2007, of which95–97% were sutchi catfish [7] Although sutchi catfish is

an important species for aquaculture and the survival rate

of reared larvae is low due to high levels of cannibalism [2,3,

8,9], no detailed ecological studies have been done on theearly larval stage of this species In the wild, sutchi catfishlarvae are difficult to locate after they hatch from eggsattached to the roots of Gimenila asiatica trees due to strongriver flows and turbidity during the flood season [4,10–12].The behavior of fish larvae is closely related to thedevelopment of their sensory organs [13–23] When fishlarvae detect stimuli, they exhibit various types of behaviorconducive to survival in the given habitat However, thedetails of the development of sensory organs in sutchicatfish larvae remain largely unknown Mukai et al [24]tried to reconstruct the early life history of sutchi catfishusing limited morphological data, but the interpretation inthat study is limited Detailed analysis of sutchi catfishsensory organ morphogenesis will contribute to betterunderstanding of the early life cycle in the wild and to thedevelopment of more efficient larval rearing techniques inhatcheries [22] In this study, we examined the develop-ment of eyes, olfactory organs, taste buds, inner ears, andfree neuromasts (mechanosensory organs) in sutchi catfishlarvae Furthermore, we characterized larval behaviors to

Y Mukai ( &)
A D Tuzan
S R M Shaleh

B M Manjaji-Matsumoto

Borneo Marine Research Institute,

Universiti Malaysia Sabah,

88400 Kota Kinabalu, Sabah, Malaysia

e-mail: mukai9166@gmail.com

DOI 10.1007/s12562-010-0290-y

gain insight into how to improve larval rearing techniques.

Finally, we discuss the relationship between feeding

behavior and sensory organ development in sutchi catfish

Materials and methods

Morphological and behavioral observations

Fertilized eggs of sutchi catfish were obtained from brood

fish reared in the hatchery of the Borneo Marine Research

Institute, Universiti Malaysia Sabah Brood fish were

obtained from commercial fish farms in Kota Kinabalu

The eggs hatched 24 h after artificial fertilization Larvae

were reared in a fiber-reinforced plastic (FRP) tank (1 m3)

at temperature of 28–30°C and were fed with rotifers,

Brachinonus sp., Artemia nauplii, and artificial compound

feed (Gemma micro 150, Skretting Co.)

The larvae were observed under a light microscope to

identify morphological changes and to measure body

length Larval behavior was observed in the larval-rearing

FRP tank (1 m3) and in a 40-L transparent acrylic

aquar-ium Fish rheotaxis was examined in a circular glass

aquarium (20 cm diameter) A weak water flow was

cre-ated in the aquarium using a plastic laboratory spatula The

criterion for positive rheotaxis behavior was whether most

fish swam against the water flow Fish phototaxis was

examined using a small electric torch to observe whether

larvae gathered at or avoided the light beam

Histological experiments

Larvae were sampled every day for the first 10 days and

then again at 15 and 20 days after hatching Larvae were

anesthetized with 3-aminobenzoic acid ethyl ester (MS222,

200 ppm) and preserved in Bouin’s solution Specimens

were embedded in paraffin, cut into 6-lm-thick sections,

and stained with hematoxylin–eosin for histological

examination [newly hatched larvae, 6 individuals; 1 day

after hatching (1 day), 6; 2 days, 10; 3 days, 9; 4 days, 7;

5 days, 5; 6 days, 6; 7 days, 5; 10 days, 5; 15 days, 5;

20 days, 5] Other specimens from the same group of

larvae were anesthetized with MS222 (200 ppm) and

pre-served in modified Karnovsky’s fixative (1.6%

parafor-maldehyde, 1.5% glutaraldehyde in 0.08 M So¨rensen

phosphate buffer and 0.02 M s-collidin buffer at pH 7.4)

[25,26] These specimens were dehydrated in an ethanol

series, freeze-dried, and coated with platinum for

exami-nation of the larval body surface under a scanning electron

microscope (SEM, JSM 5610; JEOL, Tokyo) The number

of free neuromasts observed at each developmental stage

was expressed as a composite number derived from several

specimens (newly hatched larvae, 10 individuals; 1 day, 7;

2 days, 7; 3 days, 7; 4 days, 7; 5 days, 7; 7 days, 5;

10 days, 5; 15 days, 5; 20 days, 5)

ResultsMorphological and behavioral changesMorphological changes of sutchi catfish and free neuromastdistributions are shown in Fig.1 Table1 shows thechronology of morphological, sensory, and behavioraldevelopment of sutchi catfish larvae

Newly hatched larvae (BL ± SD, mm; 3.3 ± 0.1) hadlarge yolk sacs, and their mouths and anuses were not open

neuro-They were observed at the bottom of the rearing tank and

occasionally swam following the flow of water created by

aeration

One-day-old yolk-sac larvae (4.7 ± 0.2) had open

mouths and one pair of barbels each on the upper and lower

jaw The notochord ends bent slightly upwards Yolk-sac

larvae responded to water flow, actively swimming

verti-cally to the current but not against the flow Yolk-sac

lar-vae also showed positive phototaxis, swimming towards

the light of an electric torch

Two-day-old larvae (6.1 ± 0.2) had sharp teeth (Fig.2),

and the yolk sacs were already largely consumed by this

stage The ends of the notochord showed more flexion at

2 days old compared with 1 day old after hatching Small

primordia of pectoral fins appeared, and barbels

length-ened The larvae commenced to feed on rotifers, Artemia

nauplii, and artificial compound feed and began to exhibit

cannibalism Two-day-old larvae exhibited positive

pho-totaxis, swam actively in horizontal directions, and swam

against the water flow in a round glass basin, showing

positive rheotaxis

The notochords of larvae at 5 days old (7.9 ± 0.4) werefully flexed, and many fin rays were present on the caudaland anal fins Pigments appeared on the head and trunksurfaces At this stage, the larvae swam continuously in themiddle layer of the aquarium

Table 1 Relationship between larval morphology, behavior, and morphogenesis of the sense organs in sutchi catfish, Pangasianodon hypophthalmus

Large yolk sac

Mouth and anus not open

Stayed at the bottom Responded to water flow Swam in vertical directions at several hours old

Unpigmented eyes Cilia in nasal pits lacking depressions One pair of auditory vesicles One pair of free neuromasts on the head 1

Eyes pigmented Taste buds appeared Small free neuromasts appeared

Lagena, and three otoliths at 3 days old All layers of the retina at 4 days old 5

7.9 ± 0.4

Notochord fully flexed

Many fin rays on caudal and anal fins

Pigments on head and trunk surface

Swam continuously in the middle layer

of the aquarium Showed cannibalism

Sensory epithelium sinking in the nasal pits

Number of taste buds increased

10

9.3 ± 0.8

The pelvic fins with fin rays appeared

Fin rays on dorsal fin

Swam strongly against water currents Number of taste buds and free neuromasts

increased Some free neuromasts sunk or sinking on the head

20

11.4 ± 0.8

Deeply pigmented Swam in a school Four lamellae in nasal pits

Free neuromasts sunk on the head

a Each value shows the mean ± standard deviation (SD, n = 10)

Fig 2 Scanning electron micrograph showing a 2-day-old sutchi catfish Pangasianodon hypophthalmus larva Arrows indicate the barbel and sharp teeth Scale bar 200 lm

Ten-day-old notochord postflexion larvae (9.3 ± 0.8)

had pelvic fins with fin rays Dorsal fin rays were also

developed at this stage The larvae swam against water

currents

Twenty-day-old postflexion larvae (11.4 ± 0.8) were

deeply pigmented, though the number of fin rays was still

less than that of juvenile fish The larvae swam against the

water current and showed schooling at the side of the

Fig 3 Light micrographs of the

eyes of the sutchi catfish

Pangasianodon hypophthalmus:

a newly hatched larva, b

yolk-sac larvae at 1 day old, c

2-day-old larva, d 4-day-2-day-old larva,

e 15-day-old larva Scale bars

a–d 50 lm, e 10 lm

feature: the pigment layer protruded into the outside of the

retina (Fig.3c) This feature was not observed in notochord

flexion larvae at 4 days old All layers of the retina were

recognizable; with the exception of the rod cells, the eyes

were morphologically complete (Fig.3d) In 15-day-old

postflexion larvae, the number of visual cell nuclei in the

outer nuclear layer was greater than the number of cone

cells (Fig.3e), and thus rod cells were presumed to have

appeared, although they were not visible under light

microscopy

Olfactory organ

Small nasal pits without indentations were present in newly

hatched larvae (Fig.4a) and expanded with fish growth In

1-day-old yolk-sac larvae, the outline of the nasal pits was

clear (data not shown) In notochord flexion larvae at

2 days old, many cilia were observed in the sensory

epi-thelium Both ciliated nonsensory cells and ciliated

receptor cells, as described by Yamamoto [27], were

identified (Fig.4b) Ciliated receptor cells had only three

or four cilia, whereas ciliated nonsensory cells had many

cilia By 3 days old, the sensory epithelium of the pits

expanded and began to fold (data not shown) In 5-day-old

full-flexion larvae, the epithelium was observed to be

sinking under the skin (Fig.4c) In 10-day-old postflexion

larvae, olfactory pits were separated into anterior and

posterior portions (Fig.1e) By 20 days old, the sensory

epithelium of the postflexion larvae had four lamellae

(Fig.4d)

Taste buds

No taste buds were present in newly hatched larvae Tastebuds were first observed on barbels at 1 day old (data notshown), and 2-day-old notochord flexion larvae hadnumerous taste buds on the barbels (Fig 5a, b), around themouth, and in the buccal cavity (data not shown) Tastebuds were also observed on the gills in 3-day-old larvae(Fig.5c) The number of taste buds on the surface of thehead and in the buccal cavity increased gradually with fishgrowth (Fig.6d) In 20-day-old postflexion larvae, manytaste buds were distributed on the surface of the head(Fig.5e, f)

Inner earsRound auditory vesicles were observed in newly hatchedlarvae (Fig.6a) In 2-day-old notochord flexion larvae, theauditory vesicles had several cavities Three semicircularcanals (Fig.6b), a utricle, and a saccule, each with macula,were also observed at this stage Three pairs of otoliths and

a lagena were identified in the inner ears of 3-day-oldnotochord flexion larvae (Fig.6c, d)

Free neuromastsOne pair of free neuromasts was observed on the heads ofnewly hatched larvae (Fig.1a) Free neuromasts increased

on both the head and the flank (the trunk and tail region,but not on the caudal fin) during fish growth (Fig.1a–f)

Fig 4 Scanning electron

micrographs and light

micrograph of the olfactory

organs of the sutchi catfish

Pangasianodon hypophthalmus.

a Nasal pit of newly hatched

larva (arrow) b Sensory

epithelium of 2-day-old larva.

CR ciliated receptor cell, CN

ciliated nonsensory cell.

c Olfactory pit of 5-day-old

larva (arrow) d Olfactory pits

of 20-day-old larva Each arrow

indicates an olfactory lamella.

Scale bars a 20 lm, b 5 lm,

c 100 lm, d 200 lm

In 1-day-old yolk-sac larvae, eight free neuromasts were

observed on the head and three free neuromasts on the

flank (one side of the body) (Fig.1b) By 2 days old,

well-developed free neuromasts were observed on the head and

the flank, where they were distributed in two lines

(Figs.1c, 7b) Ten-day-old postflexion larvae had 22 free

neuromasts on the head and 61 free neuromasts on the flank

of the body By this stage, some free neuromasts had sunk

or were beginning to sink under the skin on the head

(Fig.7c) In postflexion larvae at 20 days old, the free

neuromasts on the heads had completely sunk under the

skin (Fig.7d), and 63 free neuromasts were distributed on

the flanks

Discussion

Sutchi catfish larvae had well-developed eyes at the

noto-chord flexion stage by 4 days old Two-day-old flexion

larvae showed positive phototaxis and eyeball movement;

thus, the sutchi catfish larval eye can be considered to be

functional by 2 days old after hatching The larvae hadwell-developed eyes at the notochord flexion stage by

4 days old However, the developing eyes had a peculiarshape in 2-day-old larvae (Fig.3c) Typically, the fishlarval eyeball is round in shape [15, 16, 28, 29], but weobserved a protrusion of the pigment layer to the outside ofthe retina Live larvae at the same stage also showed thesame eyeball shape, indicating that this unusual structure isnot simply an artifact of the fixation process Thesestructures have not been previously reported in other fishlarvae, and further study is necessary to understand thispeculiarity of eye development in the sutchi catfish.The olfactory organs of sutchi catfish developed simi-larly to those of the African catfish Clarias gariepinus [29]

We identified two types of ciliated cells, ciliated sory cells and ciliated receptor cells, as seen in the Africancatfish These two types of cilia have also been observed inthe early larval stages of other fish species, such as theJapanese flounder Paralichthys olivaceus and the Japanesestriped knifejaw Oplegnathus fasciatus [15,30] Morpho-logical descriptions of the sensory epithelia of many teleost

nonsen-Fig 5 Scanning electron

micrographs and light

micrographs of taste buds in

sutchi catfish Pangasianodon

hypophthalmus a Taste buds on

the barbel of a 2-day-old larva.

b Taste buds on the barbel of a

2-day-old larva c Taste buds in

the buccal cavity and gill region

of a 3-day-old larva d Taste

buds around the mouth of a

10-day-old larva e Taste buds on

the head surface of a 20-day-old

larva f Taste buds on the lower

jaw and upper jaw of a

20-day-old larva Each arrow indicates

a taste bud Scale bars a 5 lm,

b 10 lm, c–f 100 lm

fishes have been described in detail by Yamamoto [27],

who indicated that two adult Siluriformes fish, the Japanese

common catfish Sirulus asotus and Pelteobagrus nudiceps,

have a high density of nonsensory cells, whereas another

species, Plotosus lineatus, has no nonsensory cells Like

the Japanese common catfish, the sutchi catfish sensory

epithelium had a high density of nonsensory cells

According to Ueda [31], fish with a high density of

non-sensory cells have highly sensitive olfactory systems We

therefore predict that the sutchi catfish has a sensitive

olfactory system, though further study is needed to confirmthis hypothesis

According to Tanaka [32], fish that hatch from demersaleggs have taste buds from an early larval stage Like thesutchi catfish, the goldfish Carassius auratus, carp Cyprinuscarpio, and willow shiner Gnathopogon elongatuscaerulescens have many taste buds by the time feedingcommences and are all omnivorous feeders from an earlylarval stage [28, 32] In addition to optical selection,these fish have the ability to select food based on chemical

Fig 6 Light micrographs of the

inner ears of the sutchi catfish

Pangasianodon hypophthalmus.

a Arrows show auditory vesicles

of newly hatched larva.

b Semicircular canal of

2-day-old larva A anterior

semicircular canal, H horizontal

semicircular canal, P posterior

semicircular canal Each arrow

indicates the crista ampullaris;

bold arrow shows the dorsal

side c Three otoliths of the

inner ear in a 3-day-old larva.

d Inner ear of a 3-day-old larva.

U utricle, S saccule, L lagena.

Arrows show the maculae of the

utricle and lagena Bold arrows

show dorsal sides Scale bars

100 lm

Fig 7 Scanning electron

macrographs of free neuromasts

in the sutchi catfish

Pangasianodon hypophthalmus.

a Free neuromast on the head of

a 1-day-old larva b Free

neuromast on the head of a

2-day-old larva c Free

neuromast on the head of a

10-day-old larva (within the

white dotted line) d Head

region of a 20-day-old larva.

Arrows indicate the pits of the

canal organs Scale bars

a 2 lm, b 5 lm, c 20 lm,

d 200 lm

cues [32] In marine fish species that hatch from pelagic

eggs, the development of taste buds is delayed and their

distribution is limited [11,30,32] For example, taste buds

develop at 10 days old in larvae of the Japanese striped

knifejaw, and at 12 days old in larvae of the Japanese

flounder [15, 30] The taste buds of these fishes are

restricted to the gill arcs and the buccal cavity While larvae

from both demersal and pelagic eggs have taste buds

distributed in the buccal cavity and on the gill arcs,

omnivorous fish also have taste buds around the mouth and

on the barbels The taste buds in the buccal cavity and gill

arcs are innervated by the vagal nerve and are considered to

play a role in determining the suitability of food for

con-sumption Taste buds located on the barbels and around the

mouth are innervated by the facial lobe nerves and are

related to searching for and tasting food in adult fish

[33,34] The taste buds in the buccal cavity and gill arcs of

both types of larvae are considered to be important for

assessing food suitability, but further studies are required to

determine the nerve connections between taste buds and the

brain in the larval stage of these species

Swimming behavior with horizontal equilibrium is

linked to inner ear development in sutchi catfish Larvae in

the notochord flexion stage were able to swim horizontally

at 2 days old, by which time the semicircular canals of the

inner ear had formed Inner ear development and changes

in the swimming behavior of sutchi catfish larvae were

similar to those of the African catfish [29] African catfish

larvae also commence horizontal swimming just after the

yolk-sac stage, when the semicircular canals have

devel-oped Moreover, the sequence of inner ear development in

the sutchi catfish was similar to that observed in the

zeb-rafish Danio rerio [35,36] and the brown trout Salmo trutta

[37] However, the zebrafish lagena was observed at

15 days after fertilization, whereas the lagena of sutchi

catfish was observed at 3 days after fertilization It is likely

that this more rapid development is advantageous in the

conditions of the sutchi catfish’s natural habitats, such as

the Mekong River, during early larval life

Free neuromasts were distributed in two lines along the

flanks of sutchi catfish larvae, and further study is needed

to determine if there are differences between the free

neuromasts in the two lines These free neuromasts had an

appearance typical of mechanosensory organs and showed

configurations similar to those of African catfish Each free

neuromast was composed of 10–20 sensory cells, and each

sensory cell had one kinocilium and 30–40 stereocilia on

the apical surface However, the long microvilli detected

on the apical surface of the free neuromasts in African

catfish were not observed Sutchi catfish larvae also have

electrical receptors on the skin surface (Mukai,

unpub-lished data) When canal organs developed on the head, the

larvae swam against the water current along the side of the

rearing tank Such changes in behavior are likely essentialfor sutchi catfish to survive in river environments.The study of sutchi catfish larvae in their natural riverhabitats has been hindered by strong river flows and tur-bidity in the flood season [4,10–12] It has been assumedthat, after hatching from eggs affixed to the roots of trees,the larvae gradually swim downstream to locate suitablemicrohabitats [38] Our behavioral observations reveal thatsutchi catfish larvae responded to water flow and showedactive swimming in vertical directions from the beginningstages of notochord flexion The sensory organs of theyolk-sac larvae developed rapidly, within 2–3 days afterhatching The early development of the inner ears and freeneuromasts may enable the survival of these fish in theriver environment

According to Nguyen et al [11], sutchi catfish larvae(estimated to be postflexion stage) drift downstream in theMekong River during the flood season Nguyen et al [11]observed five peaks of postflexion stage larval drifting fromMay to July The total lengths of the larvae in the firstgroup were 11.6–17.7 mm The size of the first group issimilar to that of postflexion larvae at 10 and 20 days oldmeasured in this study (10 days old, TL 11.9 ± 0.6 mm;

20 days old, TL 15.7 ± 2.0 mm) The traditional culture ofsutchi catfish utilizes this habit for collecting seeds in theMekong River [10] Sutchi catfish larvae in the postflexionstage had developed free neuromasts, and 20-day-old lar-vae had canal organs on the head and showed schoolingbehavior and strong rheotaxis According to a previousfield study [11, 38], larger numbers of larvae are caughtnear the bank of the Mekong River than in the middle, andthe larvae drift to the surface of the river at night As aresult, the larvae of sutchi catfish seem to be able to expandtheir habitat during the flood season The reason fortheir drifting in the river is not well understood but couldrepresent a strategy for enhancing survival in thisenvironment

Previous studies have demonstrated that African catfishlarvae, juveniles, and fingerlings can be reared undercontinuous dark conditions, and that survival rates in darkconditions are superior to those in continuous light oralternating light and dark conditions [39–44] This is due tolower rates of cannibalism in the dark In sutchi catfish, thelarvae had many taste buds at the commencement offeeding behavior Seed production of sutchi catfish isrequired to reduce cannibalism, thus there is a possibilitythat dark conditions could be used to reduce cannibalism[24]

Acknowledgments We would like to express our sincere thanks to Assoc Prof Dr Abdul Hamid Ahmad, Director of Institute for Tropical Biology and Conservation of Universiti Malaysia Sabah for his cooperation in using SEM This study was supported by the Fundamental Research Grant (FRG 0002-ST-1/2006) of the Ministry

of Education of Malaysia and the e-Science Fund (05-01-10-SF0054)

of the Ministry of Science, Technology, and Innovation of Malaysia.

References

1 Roberts TR, Vidthayanon C (1991) Systematic revision of the

Asian catfish family Pangasiidae, with biological observations

and descriptions of three new species Proc Acad Nat Sci Phila

143:97–144

2 Rohul Amin AKM, Bapary MAJ, Islam MS, Shahjahan M,

Hossain MAR (2005) The impacts of compensatory growth on

food intake, growth rate and efficiently of feed utilization in Thai

Pangas (Pangasius hypophthalmus) Pak J Biol Sci 8:766–770

3 Subagja J, Slembrouck J, Hung LT, Legendre M (1999) Larval

rearing of an Asian catfish (Siluroidei, Pangasiidae): analysis of

precocious mortality and proposition of appropriate treatments.

Aquat Living Resour 12:37–44

4 Chheng P, Baran E, Touch BT (2004) Synthesis of all published

information on sutchi catfish Pangasius hypophthalmus (‘‘trey

pra’’) based on FishBase 2004 Worldfish Center and Inland

Fisheries Research and Development Institute, Phnom Penh

5 Ali MZ, Hossain MA, Mazid MA (2005) Effect of mixed feeding

schedules with varying dietary protein levels on growth of sutchi

catfish, Pangasius hypopthalmus (Sauvage) with silver carp,

Hypopthalmichthys molitrix (Valenciennes) in ponds Aquac Res

36:627–634

6 Phan LT, Bui TM, Nguyen TTT, Gooley GJ, Ingram BA, Nguyen

HV, Nguyen PT, De Silva SS (2009) Current status of farming

practices of striped catfish, Pangasianodon hypophthalmus in the

Mekong delta, Vietnam Aquaculture 296:227–236

7 Phuong NT, Oanh DTH (2009) Striped catfish aquaculture in Viet

Nam: a decade of unprecedented development In: De Silva SS,

Davy FB (eds) Success stories in Asia aquaculture Springer,

NACA, and IDRC, Dordrecht, pp 133–149

8 Baras E, Slembrouck J, Cochet C, Caruso D, Legendre M (2010)

Morphological factors behind the early mortality of cultured

larvae of the Asian catfish, Pangasianodon hypophthalmus.

Aquaculture 298:211–219

9 Slembrouck J, Baras E, Subagia J, Hung LT, Legendre M (2009)

Survival, growth and food conversion of cultured larvae of

Pangasianodon hypophthalmus, depending on feeding level, prey

density and fish density Aquaculture 294:52–59

10 Poulsen AF, Hortle KG, Valbo-Jorgensen J, Chan S, Chhuon CK,

Viravong S, Bouakhamvongsa U, Yoorong N, Nguyen TT, Tran

BQ (2004) Distribution and ecology of some important riverine

fish species of the Mekong River basin Technical paper no.10.

Mekong River Commission, Phnom Penh

11 Nguyen TT, Nguyen ND, Vu VA, Truong TT (2005) Monitoring

of fish larvae during the annual flood of the Mekong and Bassac

rivers, Mekong Delta, Vietnam In: Burnhill TJ, Warren TJ (eds)

Proceedings of 7th technical symposium on Mekong fisheries,

Vientiane, pp 71–88

12 Van Zalinge N, Lieng S, Ngor PB, Heng K, Jorgensen JV (2002)

Status of the Mekong Pangasianodon hypophthalmus resources,

with special references to the stock shared between Cambodia

and Viet Nam Technical paper no.1 Mekong River Commission,

15 Kawamura G, Mukai Y, Ohta H (1983) Morphology of the eye

and free neuromasts of larval ayu and pale chub and its relation to

behavior in water flow Nippon Suisan Gakkaishi 49:1527–1532

16 Blaxter JHS (1986) Development of sense organs and behavior of teleost larvae with special reference to feeding and predator avoidance Trans Am Fish Soc 115:98–114

17 Kawamura G, Ishida K (1985) Changes in sense organ phology and behavior with growth in the flounder Paralichthys olivaceus Nippon Suisan Gakkaishi 51:155–165

mor-18 Kawamura G, Washiyama N (1989) Ontogenetic changes in behavior and sense organ morphogenesis in largemouth bass and Tilapia nilotica Trans Am Fish Soc 118:203–213

19 Mukai Y, Kobayashi H (1991) Morphological studies on the cupulae of free neuromasts along with the growth of larvae in cyprinid fish Nippon Suisan Gakkaishi 57:1339–1346

20 Mukai Y, Kobayashi H (1995) Development of free neuromasts with special reference to sensory polarity in larvae of the willow shiner, Gnathopogon elongatus caerulescens (Teleostei, Cyprin- idae) Zool Sci 12:125–131

21 Mukai Y, Yoshikawa H, Kobayashi H (1994) The relationship between the length of the cupulae of free neuromasts and feeding ability in larvae of the willow shiner Gnathopogon elongates caerulescens (Teleostei, Cyprinidae) J Exp Biol 197:399–403

22 Kawamura G, Masaru S, Tezuka N, Koiso M, Jinbo T, Namba K (2003) Morphogenesis of sense organs in the bluefin tuna Thunnus orientalis In: Browman HI, Skiftesvik AB (eds) The big fish bang Proceedings of the 26th annual larval fish conference, Bergen, pp 123–135

23 Mukai Y, Lang CL, Sitti Raehanah MS, Senoo S (2007) opment of free neuromasts in larvae of barramundi Lates calca- rifer (Block) Zool Sci 24:829–835

Devel-24 Mukai Y, Tuzan AD, Lim LS, Yahaya S (2010) Feeding behavior under dark conditions in larvae of sutchi catfish Pangasianodon hypophthalmus Fish Sci 76:457–461

25 Karnovsky MJ (1965) Formaldehyde–glutaraldehyde fixation of high osmolality for use in electron microscopy J Cell Biol 27:137–138

26 Yamamoto KY, Ozawa R, Takata K, Kitoh J (1981) Cell surface changes of the presumptive ectoderm following neural-inducing treatment by concanavalin A Wilhelm Roux’s Arch 190:313–319

27 Yamamoto M (1982) Comparative morphology of the peripheral olfactory organ in teleosts In: Hara TJ (ed) Chemoreception in fishes Elsevier, Amsterdam, pp 39–59

28 Mukai Y (2006) Role of free neuromasts in larval feeding of willow shiner Gnathopogon elongates caerulescens Teleostei, Cyprinidae Fish Sci 72:705–709

29 Mukai Y, Tuzan AD, Lim LS, Sitti Raehanah MS, Wahid N, Senoo S (2008) Development of sensory organs in larvae of African catfish, Clarias gariepinus (Burchell) J Fish Biol 73:1648–1661

30 Ishida K (1987) Studies on the development of sense organs and behavior in marine fish and juveniles Doctoral dissertation, Tokyo University

31 Ueda K, Sato M, Oka Y (1981) Structure of fish olfactory nerve system In: Japanese society of fisheries science (ed) Chemical sense of fish and feeding stimulants Koseisya Koseikaku, Tokyo,

pp 9–25

32 Tanaka M (1969) Studies on the structure and function of the digestive system in teleost larvae-II Characteristics of the digestive system in larvae at the stage of first feeding Japan J Ichthyol 16:41–49

33 Kiyohara S, Caprio J (1996) Somatotopic organization of the facial lobe of the sea catfish Arius felis studied by transganglionic transport of horseradish peroxidase J Comp Neurol 368:121–135

34 Kiyohara S, Tsukahara J (2005) Barbel taste system in catfishes and goatfish In: Reutter K, Kapoor BG (eds) Fish chemosenses Science, Plymouth, pp 175–209

35 Bever MM, Fekete DM (2002) Atlas of the developing inner ear

in zebrafish Dev Dyn 223:536–543

36 Whitfield TT, Riley BB, Chiang MY, Phillips B (2002)

Devel-opment of the zebrafish inner ear A peer reviewed forum Dev

Dyn 223:427–458

37 Becerra M, Anado´n R (1993) Development of the inner ear of the

brown trout (Salmo trutta fario): I Gross morphology and sensory

cell proliferation J Morphol 216:209–223

38 Hortle KG, Chea T, Bun R, Em S, Thac P (2003) Drift of fish

juveniles and larvae and invertebrates over 24-hour periods in the

Mekong river Phnom Penh, Cambodia Proceedings of the 6th

Technical Symposium on Mekong Fisheries, Pakse, pp 19–33

39 Britz PJ, Pienaar AG (1992) Laboratory experiments on the effect

of light and cover on the behavior and growth of African catfish,

Clarias gariepinus (Pisces: Clariidae) J Zool (Lond) 227:43–62

40 Appelbaum S, Mcgeer JC (1998) Effect of diet and light regime

on growth and survival of African catfish (Clarias gariepinus)

larvae and early juveniles Aquac Nutr 4:157–164

41 Hossain MAR, Beveridge MCM, Haylor GSH (1998) The effects

of density, light and shelter on the growth and survival of African catfish (Clarias gaeriepinus Burchell, 1822) fingerlings Aqua- culture 160:251–258

42 Appelbaum S, Kamler E (2000) Survival, growth, metabolism and behavior of Clarias gaeriepinus (Burchell 1822) early stages under different light conditions Aquac Eng 22:269–287

43 Almaza`n-Rueda P, Schrama JW, Verreth JAJ (2004) Behavioral responses under feeding methods and light regimes of the African catfish (Clarias gariepinus) juveniles Aquaculture 231:347–359

44 Adewolu MA, Adeniji CA, Adejobi AB (2008) Feed utilization, growth and survival of Clarias gariepinus (Burchell 1822) fin- gerlings cultured under different photoperiods Aquaculture 283:64–67

O R I G I N A L A R T I C L E Biology

Reproductive cycle of the venerid clam Meretrix lusoria

in Ariake Sound and Tokyo Bay, Japan

Yasuo Nakamura•Tadashi Nakano •

Tatsuya Yurimoto• Yukio Maeno•

Takayoshi Koizumi•Akio Tamaki

Received: 19 April 2010 / Accepted: 8 September 2010 / Published online: 15 October 2010

Ó The Japanese Society of Fisheries Science 2010

Abstract We assessed the reproductive cycle of the

venerid clam Meretrix lusoria by histological analysis of

the gonads Individuals for study were collected from

natural populations on the Shirakawa tidal flat, Ariake

Sound, and from populations that had been transplanted

from the Shirakawa flat to the Oi flat in Tokyo Bay In both

study areas, the reproductive cycle was synchronized

between sexes Gonads of the clam started to develop in

early spring and matured during the summer Mass

spawning occurred in the late summer/early fall The clam

matured at a shell length of 17–20 mm, which is much

smaller than previously considered While trophic

condi-tions and salinity differed considerably in the two study

areas, water temperatures showed similar seasonal changes

(12°C during the winter and around 30°C during the

summer) Thus, temperature probably controlled gonadaldevelopment The coincidence of the period of spawningwith the period of frequent intrusion of hypoxic waters intothe tidal flats in Tokyo Bay suggests that such hypoxicevents interfere with clam recruitment and are at leastpartly responsible for the disappearance of the naturalpopulation at this location

Keywords Meretrix lusoria Reproductive cycle Spawning Ariake Sound  Tokyo Bay

IntroductionThe venerid clam Meretrix lusoria (Roeding, 1798) lives inthe sandy sediments of shallow coastal waters of Japan andKorea, reaching a shell length of approximately 10 cm.This clam used to be an important fishery resource in theestuaries of Japan, such as Ariake Sound, Ise Bay, andTokyo Bay However, the catch of M lusoria in Japan hasdecreased drastically since the 1970s [1], and the naturalpopulation in Tokyo Bay has almost entirely disappeared[2] The cause of these declines is still unclear [1], butefforts are ongoing to determine a way to restore thepopulation of M lusoria to the estuaries of Japan Toaccomplish this, it is essential to understand the biologicaland ecological characteristics of M lusoria, includingsurvival, growth, and life cycle In an earlier study, wecaged M lusoria and other clam species on a hypoxia-prone tidal flat in Tokyo Bay and assessed their survivaland growth in relation to environmental conditions [3] Wefound that the survival of M lusoria was high and com-parable to that of the venerid Mercenaria mercenaria, thedominant species on the flat as well as an exotic one, andhigher than the venerid Ruditapes philippinarum and the

Y Nakamura ( &)

National Institute for Environmental Studies,

Tsukuba, Ibaraki 305-8506, Japan

e-mail: yasuo@nies.go.jp

T Nakano

Graduate School of Science and Technology,

Nagasaki University, 1-14 Bunkyo-machi,

Nagasaki 852-8521, Japan

T Yurimoto  Y Maeno

Seikai National Fisheries Research Institute,

1551-8 Taira-machi, Nagasaki 851-2213, Japan

T Koizumi

Nihon Mikuniya Co., 3-45-10 Mizonokuchi,

Takatsu, Kawasaki, Kanagawa 213-0001, Japan

A Tamaki

Faculty of Fisheries, Nagasaki University,

1-14 Bunkyo-machi, Nagasaki 852-8521, Japan

DOI 10.1007/s12562-010-0289-4

mactrid Mactra veneriformis (abundant species in both

Tokyo Bay and Ariake Sound) during the late summer and

early fall when hypoxic waters often intrude into the flat In

addition, the growth of M lusoria was rapid and

compa-rable to M mercenaria These results strongly suggest that

M lusoria with a shell length of[10 mm (the size used for

caging experiments) is highly tolerant of unfavorable

conditions and, consequently, that the disappearance of the

population in Tokyo Bay is not a result of low survival or

low growth under unfavorable conditions

Species with a narrow spawning period that overlaps

with the occurrence of unfavorable events for recruitment

[e.g., intrusion of hypoxic waters or sulfide-rich waters

(‘‘blue tides’’) to shallow coastal areas] have a higher risk

of recruitment failure than species with a longer spawning

period as the latter can mitigate the risk of recruitment

failure The spawning period of M lusoria in Japan is

reported to be extend from the summer to early fall [4 6]

when hypoxic waters and blue tides often intrude into

tidal flats in Tokyo Bay This spawning period is narrower

than that for R philippinarum [7], Merc mercenaria [8]

and Mac veneriformis (Nakano and Tamaki, unpublished

data, 2008) Given that recruitment failure over

consecu-tive years could lead to a rapid decline in the population,

we hypothesized that the overlap of the spawning period

with hypoxic events is responsible for the decline of the

natural population of M lusoria in Tokyo Bay However,

earlier estimations of the spawning period for M lusoria

in Tokyo Bay were based on the observation of gonads

with the naked eye [4]; histological analysis of the

gonads, a reliable method by which to analyze gonadal

development and spawning period, have yet to be

con-ducted in Japan As such, the estimation of the spawning

period of M lusoria in Tokyo Bay by histological analysis

is required to confirm the above hypothesis In

addi-tion, information on the reproductive characteristics of

M lusoria, such as spawning period and minimum size

for sexual maturation, is indispensable for resource

man-agement For example, using reproductive information we

can provide a rationale for setting a closed season and a

minimum size for fishing to maintain or restore natural

populations To address these knowledge gaps, we

ana-lyzed the reproductive cycle of M lusoria in samples

collected from the Shirakawa tidal flat (Ariake Sound),

where a natural population of the clam is still abundant

but its spawning period has not yet been determined, and

from the Oi tidal flat (Tokyo Bay), using caged clams

transplanted from the Shirakawa flat Using this

method-ology, we estimated the timing of spawning and the

minimum size of sexual maturation in specimens in both

study areas in relation to environmental conditions, as a

first step in the search for a way to restore M lusoria

populations

Materials and methodsStudy sites and monitoring of environmentThe Shirakawa tidal flat (32°470N, 130°360E) is bounded

by the Shirakawa and Tsuboigawa Rivers and has an area

of 4.2 km2 [9] Meretrix lusoria is abundant on thesouthern sandy part of the flat Tides are semidiurnal, withamplitudes averaging 3.9 and 2.0 m at spring and neaptides, respectively Water temperature and salinity 30 cmabove the surface of the flat was monitored at a station

690 m seaward of the uppermost shoreline from September

2006 to October 2007 at 10-min intervals, using a perature and salinity data logger (Compact-CT; JFEAdvantech Co., Hyogo, Japan) The Oi tidal flat (35°350N,139°450E) is located in the Keihin Canal of Tokyo Bay andhas an area of 0.01 km2 [3] Mercenaria mercenaria isabundant there, and no individuals of M lusoria are nat-urally found on the flat Tides are semidiurnal, withamplitudes averaging 1.9 and 1.0 m at spring and neaptides, respectively Cages for rearing M lusoria wereplaced on a sandy part of this flat (Sta P; silt–clay frac-tion 1.1%) at an elevation 0.7 m higher than the mean lowwater spring tide (MLWS) [3] Sediment temperature wasmonitored at Sta P about 5 cm below the surface at 30-minintervals from March 2008 to October 2009, using a tem-perature data logger (Thermochron G; Embedded DataSystems Co., Lawrenceburg, KY) Water temperature andsalinity were monitored at a subtidal station (about 50 mfrom Sta P) at an elevation 0.8 m below MLWS [3].Sampling and identification of the clams

tem-on Shirakawa flatClams for reproductive cycle analysis were collected fromthe Shirakawa flat monthly from September 2006 toOctober 2007 and from May to October 2009 In the firstpart of the sampling period, clams with a shell length[25 mm were collected (n = 21–42 on each samplingoccasion) In the latter period, clams with shell length[30 mm (n = 5–10 on each sampling occasion) and15–25 mm (n = 15–20 on each sampling occasion) werecollected

Clams collected from the Oi tidal flat for reproductivecycle analysis originated from Shirakawa flat Clams with ashell length of 13–20 mm (Group A; n = 61) and7–11 mm (Group B; n = 210) were collected in March

2008, and those with a shell length of 11–13 mm (Group C;

n = 220) were collected in October 2008 These clamswere transplanted to cages placed on the Oi tidal flat within

3 days [3]

A congeneric species of M lusoria, M petechialis, hasbeen reported to have invaded Ariake Sound from China

and/or Korea [10] As it is difficult to differentiate between

the two species by morphology alone, we identified Meretrix

clams from the Shirakawa flat on the basis of partial

sequences of their cytochrome c oxidase I (COI) gene

(GenBank Accession No AB280785 and AB280786) We

analyzed 135 clams (on 10 occasions in 2006 and 2007), and

all were identified as M lusoria [3] In addition, three, seven,

and 30 clams belonging to the size categories of Groups A, B,

and C, respectively, were collected on the same date as the

sampling of the clams of each group These were identified

by a simple molecular technique using specific primer sets

for M lusoria and M petechialis (Nakamura:http://www

nies.go.jp/aquaterra/member/nakamura/hamahama/index

html; accessed 19 April 2010) Electrophoresis of the PCR

products indicated that all could be assigned to M lusoria

Thus, most, if not all, Meretrix clams collected from the

Shirakawa flat were M lusoria

Cage rearing and sampling of clams on the Oi flat

Clams were reared in three cages [40 (length) 9 40 (width)

9 20 cm (height)] made of stainless steel, with a mesh size

of 5 mm and a removable lid The cages were buried in the

sediment of the Oi flat to a depth of approximately 10 cm

Clams assigned to Groups A and B were numbered at the

start of the rearing (March 2008), and those in Group C

were numbered in March 2009 The clams were retrieved

at intervals of approximately 2–4 weeks and their survival

checked and shell sizes (length l, height h, width w)

mea-sured to the nearest 0.1 mm with a digital caliper; live

clams, except for the ones used for sampling, were placed

back into the cages

Clams for reproductive cycle analysis were sampled

monthly from April 2008 to February 2009 for Group A

(n = 4–6 on each sampling occasion), from July 2008 to

October 2009 for Group B (n = 8–10 on each sampling

occasion), and from March 2009 to October 2009 for

Group C (n = 14–20 on each sampling occasion) For each

sampling period, clams were allocated to n ranks based on

shell volume (v = lhw) in the preceding sampling period,

and one clam was randomly sampled from each rank

Histological analysis

Clams collected were first measured for shell size, and then

their valves were opened with a knife and their tissue fixed

with 10% formalin solution The fixed samples were

transferred to 70% ethanol solution within a month and

preserved until processing For processing, the flesh was

separated from the shell, and a central part of the visceral

portion (2–3 mm thick) was removed, dehydrated with a

series of alcohol solutions, embedded into paraffin, sliced

into 5- to 7-lm sections, and stained with hematoxylin and

eosin The gonadal slides were examined with an opticalmicroscope (10–209 objective) Each gonadal sample wassexed and assigned to a developmental stage, as described

by Takashima and Onuma [11], Walker and Heffernan[12], and Chung [13] Scores were used for five serialstages: early active (EA: 2; Fig 1a, b), late active (LA: 3;Fig.1c, d), ripe (R: 4; Fig 1e, f), partially spawned (PS: 1;Fig.1g, h), and spent inactive (S/IA: 0; Fig.1i) Weassigned a clam to the R stage if follicles contained manymature gametes; even if the follicles showed only a sign ofslight spawning (small empty space in the follicles), weassigned the clam to R stage as long as most of thefollicular space was filled with mature gametes [11] Weassigned a clam to the PS stage in cases where over half thegametes seemed to have been released in spawning Thegonadal index (GI) was obtained by averaging the stagescores over the specimens collected

Condition factorTogether with the clams for reproductive analysis, wecollected 25–40 clams (l = 15–25 mm) from the Shirakawaflat in 2009 for estimation of the condition factor (CF) (fd/

v mg/mL; fd = flesh dry weight) [3] The CFs were alsomeasured monthly for clams in group C from March toOctober 2009 (n = 8 on each sampling occasions)

ResultsEnvironmental conditionsTen-day averages of water temperatures on both theShirakawa (in 2007) and Oi flats and sediment tempera-tures on the Oi flat reached around 30°C during the summerand dropped to 12°C during winter (Fig.2) Althoughwater temperatures on Shirakawa flat in 2009 were notmonitored, 10-day averages of air temperatures at Misumi(about 10 km from the flat; Japan Meteorological Agency)

in 2007 coincided within 1°C with water temperatures onShirakawa flat from May to October in 2007 (Fig.2a).Water temperature in 2009 during the corresponding periodcan therefore be approximated by the air temperature in

2009, and were probably 1–2°C lower than those in 2007(Fig.2b) The salinity on the Shirakawa and Oi flats wasgenerally in the range of 25–30 and 15–25, respectively.However, after heavy rainfalls, salinity on both flatsoccasionally decreased below 10

Gonadal development on Shirakawa flatDuring 2006 to 2007, a total of 507 clams were collected forgonadal analysis The sex was skewed toward females