Fisheries science JSFS , tập 76, số 5, 2010 6

Comparison of the actual population size with population estimates obtained using the mark/recapture method showed that the percentage of absolute error was \30% in all cases in which th

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

Evaluation of four models for estimating the population size

of largemouth bass in an experimental pond

Osamu Katano

Received: 5 January 2010 / Accepted: 3 June 2010 / Published online: 13 August 2010

 The Japanese Society of Fisheries Science 2010

Abstract The utility of four commonly used models for

estimating population size in teleosts was tested Sixty-five

individually marked largemouth bass, Micropterus

salmo-ides, were introduced into a concrete pond Fishing surveys

were conducted every 2 days for a period of 19 days The

collected data were then used to estimate the population

size under a variety of conditions using the following

models: mark/recapture (Petersen method), DeLury (first

model), and two models of the software program Capture

Comparison of the actual population size with population

estimates obtained using the mark/recapture method

showed that the percentage of absolute error was \30% in

all cases in which the number of fish caught and marked in

the first survey was [30% of the population Using the

DeLury method and Model 1 of Capture, the population

estimates were biased toward underestimation, but the error

was \30% when the number of fish caught in all surveys

was [70% In contrast, in Model 2 of Capture, the error

was relatively small when the percentage of fish caught in

all surveys was \70% These conditions for minimizing

errors should be taken into account by fisheries managers

when estimating the population size of largemouth bass

Keywords DeLury method Largemouth bass 

Mark and recapture Population estimation 

Program Capture

IntroductionLargemouth bass Micropterus salmoides have a significantnegative impact on many inland fisheries and ecosystems[1 4] Considerable effort has been devoted to the eradi-cation of largemouth bass in ponds, lakes, and rivers inJapan However, a number of eradication programs havefailed to evaluate the effectiveness of these efforts incontrolling the population size Instead, most programshave focused on the number of bass that were removed Tosuccessfully evaluate an eradication program, it is critical

to obtain estimates of the population size before and aftereach removal effort

There are several methods for estimating population size

in teleosts [5, 6], some of which have been applied toestimate largemouth bass populations in Japan [7 9].However, there is little basis for determining which ofthese methods is the most appropriate for this species.Generally, the larger size classes of bass become increas-ingly difficult to catch under heavy fishing pressure, pos-sibly due to learned avoidance behavior by these largerfish Individual differences in ease of capture and learningability among largemouth bass have been documented [10].Taken together, these factors suggest that a number of themethods currently in use for estimating population sizemay not be applicable for this species

The objective of the study reported here was to evaluatethe utility of four methods that are commonly used toestimate population size In this study, 65 largemouth basswere introduced into an experimental pond, and anglingsurveys were conducted over a 19-day period The datawere then used to estimate the population size using thefollowing models: mark/recapture (Petersen), DeLury, andtwo models of the software program Capture [11–13] Itwas possible to calculate the exact error of the population

O Katano ( &)

National Research Institute of Fisheries Science,

Fisheries Research Agency, 1088 Komaki, Ueda,

Nagano 386-0031, Japan

e-mail: katano@fra.affrc.go.jp

DOI 10.1007/s12562-010-0276-9

estimates because the actual population size was known.

The results of this paper may be used by fisheries managers

and other interest groups for evaluating the success of

efforts to eradicate largemouth bass

Materials and methods

Fishing experiment

The experiments were conducted at the National Research

Institute of Fisheries Science in the city of Ueda, Nagano

Prefecture, Japan The largemouth bass were captured from

ponds and lakes near Ueda City and acclimated for

[30 days before the experiments were initiated The

experimental pond (length 40.0 m, width 5.3 m) was

con-structed of concrete [10] River water was pumped (3.3 l/s)

into the upper end of the pond and exited from the lower

end The water depth was maintained at 75–80 cm Nine

refuges for the fish, each consisting of four concrete blocks

(39 cm long, 18 cm wide, 15 cm high) were constructed

These refuges were placed at equal intervals (approx

4.9 m) along the center of the pond At the beginning of the

experiment, the fish (n = 65) were anesthetized using

2-phenoxyethanol, and the standard length and body

weight were recorded to the nearest 0.1 cm and 0.1 g,

respectively Each fish was marked with a unique

combi-nation of fin marks by cutting two or three small sections of

the dorsal, caudal, ventral, or anal fins The bass were then

introduced into the experimental pond on 3 September

2007 The water temperature was measured at three sites in

the pond at the same time on each day (1500 hours) The

temperature ranged from 21.0 to 25.6C throughout the

study There was no rainfall on any of the survey days

Angling surveys were conducted using bait or lures

every 2 days between 11 and 29 September 2007 (i.e., 10

surveys) The surveys were conducted during a 6-h period

between 0900 and 1700 hours For bait fishing, an 8.5-m

rod (model H80-85ZT; Shimano, Keihou, Japan) and a

0.205-mm-diameter line with a small float were used The

bait consisted of either live worms or shrimps For lure

fishing, a 1.98-m rod (model 662MRS-S; Daiwa, Tokyo,

Japan) with a reel (model Emblem-S 2000iA; Daiwa) and a

0.235-mm-diameter line were used A soft plastic worm

(3.500 Kut Tail Worm J7S-10-229 10; Gary Yamamoto

Custom Baits, Milam, TX) was attached to a single hook

(Dream hook 15-2; Decoy, Japan) A live Japanese dace

Tribolodon hakonensis [7–10 cm standard length (SL)]

was sometimes attached to this fishing tackle as a bait fish

Each of these methods was used on each survey day to

determine the most effective method, which was then used

until the catch rate declined All fish were released back

into the pond after the fin clip pattern was recorded

In the evening of days when sampling was not ducted, the largemouth bass were fed with 65 live Japanesedace (7–10 cm SL) On 1 October 2009, the pond wasdrained, and all 65 largemouth bass were captured No bassdied during the experiment All experimental procedureswere conducted with permission of the Ministry of theEnvironment of Japan

con-Data analysisThe following methods were used to estimate the popula-tion size: mark and recapture, DeLury, and the programCapture In each instance, the simplest method to derive apopulation estimate was used

To evaluate the mark/recapture model, the Petersenmethod [5,6] was applied The population size was esti-mated using the formula:

where N is the estimated population size, M is the number

of marked fish, m is the number of recaptured fish, and n isthe total number of fish captured in the angling surveys Inthis study, all of the bass were marked prior to theexperiment Therefore, to estimate population size usingthe Petersen method, individuals captured at the time of thefirst survey were considered to be the marked individuals insubsequent recapture surveys Sampling days used for thefirst survey were varied, as shown in Table1 Therecapture rates for these individuals were then calculatedbased on this assumption When the number of fishrecaptured (m) was \10, the following formula [6, 14,

15] was used to calculate the population size:

N¼ M þ 1½ð Þ n þ 1ð Þ= m þ 1ð Þ  1 ð2ÞWhen m C 10, 95% confidence limits were calculatedaccording to DeLury [16] When m \ 10, it is difficult tocalculate 95% confidence limits exactly, and forconvenience the limits were calculated as N 1:96pffiffiffiv

(catch per unit effort at time t) [5,6,16,17] Ntrepresentsthe number of bass that were not captured from time 0 totime t - 1 and is expressed as follows:

where N0is the total number of bass, and Kt is the totalnumber of bass caught between time 0 and time t - 1 The

fishing efficiency is represented by q, and (c/f)tis expressed

as qNt Based on these relationships, the following formula

was used to estimate the population size

c=f

The total number of bass in the pond (N) was calculated

when the regression equation was significant at the 5%

level The 95% confidence limits were calculated followingDeLury [16] The second model of DeLury is expressed asfollows

where Etis the total effort from time 0 to time t [5,6,16,

17] In this study, the second model was not used because

Table 1 Population estimates derived using the mark/recapture model

First survey Number of fish caught

in the first survey (%)

Recapture trials

Number of fish caught

in the recapture trials (%)

Number of fish recaptured

Population estimate Number 95% Confidence limit

1st day 21 (32.3) 2nd–10th day 57 (87.7) 17 70.4b 50.1–118.6

2nd–9th day 53 (81.5) 17 65.5b 46.8–109.1 2nd–8th day 53 (81.5) 17 65.5b 46.8–109.1 2nd–7th day 51 (78.5) 16 66.9b 47.3–114.3 2nd–6th day 48 (73.8) 16 63.0b 44.7–106.5 2nd–5th day 39 (60.0) 16 51.2a 37.0–83.1 2nd–4th day 34 (52.3) 14 51.0a 36.2–86.4 2nd–3rd day 29 (44.6) 11 55.4a 37.5–105.5 2nd day 20 (30.8) 6 65.0b 34.2–95.8 1st and 2nd day 35 (53.8) 3rd–10th day 55 (84.6) 29 66.4b 52.9–89.2

3rd–9th day 48 (73.8) 26 64.6b 51.1–88.0 3rd–8th day 46 (70.8) 24 67.1b 52.3–93.5 3rd–7th day 43 (66.2) 22 68.4b 52.7–97.4 3rd–6th day 38 (58.5) 20 66.5b 50.8–96.1 3rd–5th day 27 (41.5) 18 52.5a 41.3–72.1 3rd–4th day 20 (30.8) 14 50.0a 38.7–70.7 3rd day 13 (20.0) 9 49.4a 37.1–64.1 3rd day 13 (20.0) 4th–10th day 52 (80.0) 10 67.6b 43.1–156.7

4th–9th day 45 (69.2) 10 58.5b 37.6–132.4 4th–8th day 42 (64.6) 9 59.2b 42.5–75.9 4th–7th day 37 (56.9) 7 65.5b 40.2–90.8 4th–6th day 29 (44.6) 4 83.0a 33.8–132.2 4th–5th day 15 (23.1) 1 111.0 1.2–220.8 4th day 8 (12.3) 1 62.0 b 32.3–91.8 1st–3rd day 39 (60.0) 4th–10th day 52 (80.0) 30 67.6 b 54.6–88.6

4th–9th day 45 (69.2) 27 65.0b 52.3–85.9 4th–8th day 42 (64.6) 24 68.3 b 53.8–93.2 4th–7th day 37 (56.9) 20 72.2 a 55.4–103.5 4th–6th day 29 (44.6) 15 75.4 a 55.5–117.6 4th–5th day 15 (23.1) 10 58.5 b 42.9–92.1 4th day 8 (12.3) 6 50.4 a 35.1–65.7 4th day 8 (12.3) 5th–10th day 48 (73.8) 4 87.2 42.6–131.8

5th–9th day 41 (63.1) 4 74.6 a 36.8–112.5 5th–8th day 38 (58.5) 4 69.2 b 34.2–104.2 5th–7th day 33 (50.8) 4 60.2 b 30.1–90.4 5th–6th day 22 (33.8) 1 102.5 3.8–201.2 5th day 8 (12.3) 1 39.5 3.9–75.2 The day on which data were first collected was varied

a Error \30% of the actual number

b

Error \10%

(c/f) equaled 0 on the 9th day in the dataset and, therefore,

N0could not be calculated

Capture is an interactive software program developed by

the Patuxent Wildlife Research Center (PWRC) of the

United States Geological Survey (USGS) [11–13] The

program (release date: 16 May 1994) includes a maximum

likelihood estimation [18] and is available at http://www

mbr-pwrc.usgs.gov/software/capture.html Capture is

com-monly used to investigate animal abundance [19,20] Two

models of Capture were used in this study by inputting

‘‘task read population removal’’ in the first column The

same effort needs to be made in each survey, and the

numbers of newly caught individuals are input

The first (Model 1, mbh) is based on Otis et al [11], and

its result is output first In the first step of this program, the

simplest model, which assumes that all members of the

population are equally at risk of capture on every survey, is

examined This is the null model with no time, behavior, or

heterogeneity effects When this model does not fit the

data, the next model is applied with assumptions that

(q1= q2= q3…) where qk is the average probability of

capture of all individuals in the kth survey This procedure

is continued until the data fit the model However,

popu-lation estimation is not possible when data are poor or

inadequate

The second model of Capture (Model 2, mbh-Pollock)

uses the generalized jackknife statistics [13] and includes

behavior or heterogeneity effects Using specific

assump-tions, population estimation using the second model is

possible for any set of data if the number of investigations

is [1, even when the number of animals captured increases

in the later investigations

In calculations using Capture, to equalize fishing efforts

in each survey, when the number of newly caught bass on

the first and second day was the first input, the second to

third input included data obtained on the following 2 days

(3rd and 4th day, 5th and 6th day, 6th and 7th day, and 8th

and 9th day) Similarly, when the number on the first to

third day was the first input, the second and third day input

included data on the following 3 days (4th–6th day and

7th–9th day)

Results

The total number of bass caught tended to decrease on later

sampling days (Fig.1) However, the number of bass

caught for the first time increased on days 6 and 10 The

number of times an individual bass was caught ranged

between zero and eight (Fig.2) Four individuals were not

caught throughout the experiment

The results of population estimation using the mark/

recapture method are shown in Table1 When the number

of bass marked ranged between 12.3 and 60.0% of the totalnumber, and the number of recaptured fish ranged between12.3 and 87.7%, the estimated population size variedbetween 39.5 and 111.0 The actual number of bass waswithin the 95% confidence limits in 36 (97.3%) of 37cases

In the estimates that used DeLury’s first model(Table2), the regression equation was not significant andthe population size was not estimated in 15 (38.5%) of 39cases When the proportion of bass caught in the firstsurvey ranged between 12.3 and 60.0%, and the proportion

of bass caught in all surveys ranged between 41.5 and

1st 2nd 3rd 4th 5th 6th 7th 8th 9th 10th

Day of sampling

Number of bass newly caughtNumber of bass recaptured25

Fig 1 Number of largemouth bass that were caught for the first time and recaptured on each day of sampling When the same individuals were caught twice on 1 day, the second capture was ignored

93.8%, the population estimates ranged between 38.0 and

71.7, and the actual population size was within the 95%

confidence limits in 22 (91.7%) of 24 cases

The population estimates calculated by Model 1 ofprogram Capture ranged between 37 and 110 in 28 cases,but calculation was not possible in two cases (Table3)

Table 2 Population estimates derived using the DeLury model

First survey Number of fish caught

in the first survey (%)

Number of surveys (n)

Number of fish caught

The day on which data were first collected was varied

NS not significant at the 5% level, – calculation was not possible

a Error \30% of the actual number

b Error \10%

The actual number of bass was within the 95% confidence

limits in 19 (67.9%) of 28 cases

The population estimates calculated by Model 2 of

Capture ranged between 41 and 116 (Table3) The actual

number of bass lay within the 95% confidence limits in 17

(56.7%) of 30 cases When the first survey was on the first,

second, third, or fourth day and the last survey was on the

tenth day, the population estimated by Model 2 ranged

from 94 to 116, which is larger than the estimates produced

by Model 1 and the DeLury model These estimates were

34–63 larger than the estimate obtained when the lastsurvey was on the ninth day

The population estimates by the four methods exactlypredicted the actual number of fish in five (4.2%) cases,overestimated it in 44 cases (37.0%), and underestimated it

in 70 cases (58.8%) (Table4) The ratio of over- tounderestimates differed significantly among the fourmethods (chi-square test, df = 3, v2= 12.51, P = 0.0058).Most of the estimates (21/24) by the DeLury method werelower than the actual population size Underestimates

Table 3 Population estimates derived using the Capture model

First survey Number of fish caught

in the first survey (%)

Number of surveys (n)

Number of fish captured

The day on which data were first collected was varied

– calculation was not possible

a Error \30% of the actual number

b Error \10%

exceeded overestimates by a factor of two when model 1 of

the Capture program was used In contrast, no clear

ten-dency was detected using the mark/recapture method and

model 2 of Capture

The error of estimation was compared among the four

methods (Fig.3) All data are shown in the figure with the

percentages of fish captured in the first survey (X axis) and

the percentage of fish captured in recapture trials or in all

surveys (Y axis)

In the mark and recapture method, the percentage of

absolute error was \30% in all cases in which the

per-centage of fish caught and marked in the first survey

exceeded 30% Errors[30% occurred when the percentage

of fish marked was B20% even when the number of fish

caught in the recapture trials exceeded 70% In the DeLury

method, estimation was not possible in most of the cases

(9/11) in which the percentage of fish caught in the first

survey was B20% The error of estimation was\30% in all

cases in which the number of fish caught in all surveys was

[70%, but the error increased when the proportion

decreased to \65% The same tendency was seen in model

1 of Capture The error was \30% in all cases in which the

number of fish caught in all surveys exceeded 70% In

contrast, the magnitude of the error was not predictable in

model 2 of Capture The error was \30% in all four cases

in which the percentage of fish caught in the first survey

exceeded 50% When the proportion of fish caught in the

first survey was B20%, the error was small compared with

Model 1, but the error increased when the data on the tenth

sampling day were added to the analysis

Discussion

All models for estimating population size have specific

assumptions As no bass died in the experimental pond and

there were no immigrants or emigrants, the four models

used in this study could be applied for the population

estimation

The most appropriate method likely varies depending on

the species of interest In this study, the behavior of

largemouth bass appeared to change over time First, the

number of fish caught on a given day decreased with time.However, there was also a marked increase in the contri-bution of previously uncaught individuals during the laterangling surveys The reason for this increase is unclear, butsuch changes can be common in field investigations.Second, variation in catchability between individual basswas noted The number of times an individual was caughtranged between zero and eight throughout the experiment.Four bass were not captured on any sampling day In aprevious paper [10], 34 of the 65 largemouth bass could beclassified, according to their catchability, as careful (8),learnable (10), or fishable (16) Fishable bass were caughtrepeatedly, whereas careful and learnable bass were rarelycaught or recaptured Individual differences in catchabilityhave been noted previously in some fish species, includingbass [21–24] Such differences may be related to the ability

to learn and the awareness of individuals These differencesappear to be common in animals as individual behavioraldifferences or personalities are now recognized in a variety

of species [25–28]

The existence of learning and individual differences inbehavior may invalidate the assumptions of the method,increasing the error and bias of the estimates However, it isgenerally unknown whether learning and/or individual dif-ferences exist in a population It is therefore important to testpopulation estimates against the actual population size and

to clarify the conditions necessary to minimize errors In thisstudy, the error of estimation was calculated by comparingthe population estimates with the actual population size.The population estimates output by the four methods aresummarized in Table5 Population estimates were leastbiased in terms of over- and underestimation in the mark/recapture method and Model 2 of Capture In contrast, inthe DeLury method and Model 1 of Capture, the outputswere biased to underestimates The tendency for underes-timation in the DeLury method has been pointed byBraaten [29] and Otis et al [11] Such underestimationsmight occur because the catch per unit effort (CPUE) ofnew bass decreased on later sampling days more thanpredicted for random processes

The actual population size generally lay within the 95%confidence limits of the mark/recapture and DeLury

Table 4 Numbers of overestimates and underestimates obtained using the four methods of population estimation listed in Tables 1 , 2 and 3

estimates (Table5) However, in Model 1 and 2 of ture, the real population size lay outside the limits in32–43% of all cases due to the limit being calculated usingthe specific assumptions and statistics of the Capture pro-gram [11–13,30] This problem has been recognized, andthere are now a number of remedies for this calculation[30].

Cap-Compared with the other three methods, the mark/recapture method was excellent in that the error of esti-mation was small when the percentage of fish caught andmarked in the first survey exceeded 30, and the populationsize could be calculated in all cases However, in a number

of prefectures in Japan, it is prohibited to release mouth bass that are caught From the viewpoint of eradi-cation of bass, it is better to remove largemouth bass thatare caught than to mark and release them

large-The DeLury method and Model 1 of Capture requiredsimilar conditions for errors of \30% These small errorswere obtained when the proportion of fish caught in allsurveys exceeded 70% of the population Model 1 of theCapture program includes an assumption that catch prob-abilities differ between different investigations, but a cleardifference between the outputs of the DeLury method andModel 1 of Capture was not detected In the DeLurymethod, the population estimates could not be calculatedwhen data did not fit, but this deficiency is improved inModel 1 of Capture In contrast, in Model 2 of Capture, thepopulation estimation outputs were unpredictable and, inparticular, they seemed to be affected by the number of fishnewly caught in the last survey It is possible to use thismethod even when there are only two sets of catch data,and population estimates can be calculated using any set ofdata, indicating that Model 2 include complex assumptions

in the calculation of population size However, errors werelarge even when the proportion of fish caught in all surveysexceeded 80% When the percentage of fish caught in allsurveys exceeded 70%, it was preferable to use DeLury orModel 1 of Capture to avoid unpredictably large errors ofestimation However, when fewer fish than 70% werecaught in all surveys, Model 2 of Capture was better thanthe DeLury method and Model 1 of Capture Recognition

of the size of the error and of the unpredictability isimportant because investigation plans can be improved Forexample, if the total number of fish caught in the DeLurymethod and Model 1 does not exceed 70% of the totalpopulation, it is advisable to increase the number ofinvestigations to achieve this condition If this is impossi-ble, the accuracy of population estimates needs to bechecked by conducting population estimation sequentially

in another season or year

Among the methods that can be used for populationestimation, the simplest models were analyzed in thisstudy Revised models for the mark/recapture and DeLury

Fig 3 Comparison of the errors of population estimates obtained

using the four methods

methods and Capture have been proposed [5, 6, 15,

29–31] It is possible that there are models that would fit

the present data set for largemouth bass better than the

simple models used here However, the causal factor of the

error of estimation is the behavior and reaction of fish to

the method of investigation, and our knowledge of these

factors is poor Complex models having many assumptions

may not be robust for a variety of data sets collected in the

field More studies are required to compare population

estimates obtained by different methods with the actual

population size in relation to the behavior of individual

animals

Acknowledgments The author thanks F Onuma and A Ishihara for

their help during the experiments and H Hakoyama for advice on the

manuscript This work was supported in part by a grant from the

Fisheries Agency and the Grant-in Aid for Scientific Research (C)

(20570028) from the Japan Society for the Promotion of Science

(JSPS) and the Ministry of Education, Culture, Sports, Science and

Technology (MEXT), Japan.

References

1 Azuma M (2001) Ecological disturbance by two alien fish species

to the indigenous community of a small lake in western Japan.

Verh Internat Verein Limnol 27:3794–3797

2 Maezono Y, Miyashita T (2003) Community-level impacts

induced by introduced largemouth bass and bluegill in farm

ponds in Japan Biol Conserv 109:111–121

3 Takamura K (2007) Performance as a fish predator of largemouth

bass [Micropterus salmoides (Lacepe`de)] invading Japanese

freshwaters: a review Ecol Res 22:940–946

4 Sugiyama H (2005) Forefront in eradicating largemouth bass (in

Japanese) Mumeisha Shuppan, Akita

5 Tanaka S (1985) Outline of fisheries biology (in Japanese).

Kouseisha-Kouseikaku, Tokyo

6 Nose Y, Ishii T, Shimizu M (1988) Fiheries biology (in

Japa-nese) University of Tokyo Press, Tokyo

7 Kuge T, Nobusawa K, Maita M (2004) Abundance and feeding

habits of largemouth bass Micropterus salmoides in Haruna Lake,

Gunma Prefecture, Japan (in Japanese with English abstract).

Suisanzoshoku 52:73–80

8 Ozaki M, Kajiyama M (2006) Estimation of the largemouth bass population in Lake Kameyama, Chiba Prefecture (in Japanese) Bull Chiba Pref Fish Res Ctr 1:1–5

9 Fukuda Y, Azechi K (2007) Ecology of largemouth bass and investigations for developing eradication techniques (in Japa- nese) Bull Oita Pref Fish Res Inst 2007:311–317

10 Katano O (2009) Individual differences in catchability of mouth bass Micropterus salmoides by fishing in an experimental pond (in Japanese with English abstract) Nippon Suisan Gakkaishi 75:425–431

large-11 Otis DL, Burnham KP, White GC, Anderson DR (1978) tical inference from capture data on closed animal populations Wildl Monogr 62:1–135

Statis-12 White GC, Burnham KP, Otis DL, Anderson DR (1978) User’s manual for program Capture Utah State Univ Press, Logan

13 Pollock KH, Otto MC (1983) Robust estimation of population size in closed animal populations from capture–recapture exper- iments Biometrics 39:1035–1049

14 Schaefer MB (1951) Estimation of size of animal population by marking experiments Fish Bull 69:191–203

15 Ito Y, Murai M (1978) Research methods for animal ecology, the first volume (in Japanese) Kokinshoin, Tokyo

16 DeLury DB (1951) On the planning of experiment for the mation of fish populations J Fish Res Bd Can 8:281–307

esti-17 DeLury DB (1947) On the estimation of biological populations Biometrics 3:145–167

18 Moran PAP (1951) A mathematical theory of animal trapping Biometrika 38:307–311

19 Shanker K (2000) Small mammal trapping in tropical montane forests of the upper Nilgiris, southern India: an evaluation of capture–recapture models in estimating abundance J Biosci 25:99–111

20 Katano O, Nakamura T, Abe S, Yamamoto S, Baba Y (2006) Comparison of fish communities between above- and below-dam sections of small streams; barrier effect to diadromous fishes.

J Fish Biol 68:767–782

21 Martin RG (1958) Influence of fishing pressure on bass fishing success Proc Annu Conf Southeast Assoc Game Fish Comm 11:76–82

22 Beukema JJ, De Vos GJ (1970) Experimental tests of a basic assumption of the capture–recapture methods in pond population

of carp Cyprinus carpio L J Fish Biol 6:317–329

23 Ferno A (1983) The effect of experience on the behavior of cod Gadus morhua L towards a baited hook Fish Res 2:19–28

24 Yoneyama K, Yagi S, Kawamura G (1992) Catchability of individual Tilapia Tilapia mossambica in Tank (in Japanese with English abstract) Nippon Suisan Gakkaishi 58:1867–1872

Table 5 Summary of the population estimates obtained using four methods

where the actual number

of fish (65) is within the 95% confidence limit (%)

Percentage of cases

in which calculation was possible (%)

Conditions for error \30%

marked in the first survey [30

surveys [70

surveys [70

relatively small when % of fish caught was \70%

– No clear tendency

25 Clark AB, Ehlinger TJ (1987) Pattern and adaptation in

indi-vidual behavioral differences In: Bateson PPG, Klopfer PH (eds)

Perspectives in ethology, vol 7 Plenum Press, New York,

pp 1–47

26 Katano O (1991) Ecology of personalities (in Japanese) Kyoto

University Press, Kyoto

27 Magurran AE (1993) Individual differences and alternative

behaviors In: Pitcher TJ (ed) Behaviour of teleost fishes,

2nd edn Chapman & Hall, London, pp 441–477

28 Bolnick DI, Svanba¨ck R, Fordyce JA, Yang H, Davis JM, Hulsey

CD, Forister ML (2003) The ecology of individuals: incidence

and implications of individual specialization Am Nat 161:1–28

29 Braaten DO (1969) Robustness of the DeLury population mator J Fish Res Bd Can 26:339–355

esti-30 Rexstad E, Burnham KP (1991) User’s guide for interactive program Capture Colorado Cooperative Fish & Wildlife Research Unit, Colorado State University, Fort Collins

31 Gatz AJ, Loar JM (1988) Petersen and removal population size estimates: combining methods to adjust and interpret results when assumptions are violated Environ Biol Fish 21:293–307

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

Estimation of kelp forest, Laminaria spp., distributions in coastal

waters of the Shiretoko Peninsula, Hokkaido, Japan, using

echosounder and geostatistical analysis

Kenji Minami•Hiroki Yasuma•Naoki Tojo •

Shin-ichi Fukui•Yusuke Ito •Takahiro Nobetsu•

Kazushi Miyashita

Received: 2 September 2009 / Accepted: 18 June 2010 / Published online: 5 August 2010

Ó The Japanese Society of Fisheries Science 2010

Abstract Sustainable management of the kelp forests of

the Shiretoko Peninsula, a World Natural Heritage site, is

necessary due to kelp’s ecological and economic

impor-tance The objectives of this study were to estimate the area

of kelp forests and to clarify their spatial characteristics in

coastal waters of the Shiretoko Peninsula Data on the

presence/absence and thickness of kelp forests were

col-lected via acoustic observation on transects over about

80 km using an echosounder at 200 kHz Acoustic data

were geostatistically interpolated, and the areas covered by

kelp forests were estimated Differences in kelp tion between the eastern and western sides of the peninsulawere compared The total area of kelp forest was 3.88 km2(eastern area: 3.49 km2; western area: 0.39 km2) Therange of thickness of the kelp forests was 34–91 cm Manykelp forests in the eastern area were thick ([78 cm) anddistributed continuously, while kelp forests in the westernarea were sparsely distributed

distribu-Keywords Acoustic observationGeostatistical analysis Kelp forests  Shiretoko Peninsula

IntroductionThe Shiretoko Peninsula, located in northeastern Hokkaido,Japan, was registered by UNESCO as a World NaturalHeritage site in July 2005 because of its unique ecosystemand diverse ecological interactions (Fig.1) In the coastalwaters around Shiretoko Peninsula, Laminaria ochotensisMiyabe and L diabolica Miyabe form dense vegetation ofkelp forests, which are considered to be the ‘‘forests of thesea’’ [1, 2] Kelp forests play important ecological andeconomic roles in the coastal waters of the ShiretokoPeninsula [3] The primary production of kelp forests is ashigh as that of terrestrial vascular plants, such as maturerain forest (approximately 1,300 g C m-2 year-1), andkelp forests are considered to be the high primary pro-ducers in coastal ecosystems [4] The coastline of theShiretoko Peninsula is over 100 km long, so the primaryproduction of the kelp forests along the coastline is pre-sumably considerable Also, kelp forests provide valuablefishing resources in Japan [5] Laminaria spp productsfrom the coastal waters of the Shiretoko Peninsula areregarded as some of the best and most valuable kelp

K Minami  Y Ito

Graduate School of Environmental Science,

Hokkaido University, 3-1-1 Minato-cho,

Hakodate, Hokkaido 041-8611, Japan

H Yasuma

Fisheries Technology Department, Kyoto Prefectural

Agriculture, Forestry and Fisheries Technology Center,

1061 Odashukuno, Miyazu, Kyoto 626-0052, Japan

N Tojo

Field Science Center for Northern Biosphere,

Hokkaido University, Aikappu, Akkeshi-cho,

Akkeshi-gun, Hokkaido 088-1113, Japan

S Fukui  K Miyashita

Field Science Center for Northern Biosphere,

Hokkaido University, 3-1-1 Minato-cho,

Hakodate, Hokkaido 041-8611, Japan

T Nobetsu

Shiretoko Nature Foundation, Shiretoko National Park Nature

Center, 531 Iwaobetsu, Shari-cho, Hokkaido 099-4356, Japan

Present Address:

K Minami ( &)

Field Science Education and Research Center, Kyoto University,

Nagahama, Maizuru, Kyoto 625-0086, Japan

e-mail: k.minami@ky3.ecs.kyoto-u.ac.jp

DOI 10.1007/s12562-010-0270-2

products in Japanese markets because of their superior

quality [6] Due to these ecological and economic

contri-butions, environmental and fisheries management to

sus-tain the kelp forests in the coastal waters of the Shiretoko

Peninsula is needed

Mapping and quantification of kelp forest distribution

along the coast of the Shiretoko Peninsula is a practical first

step toward sustainable kelp forest resource management

However, the extent of the coastal area of the Shiretoko

Peninsula has made kelp surveys difficult, so information

on distribution is limited Spatial variability in kelp

har-vests from the Shiretoko coast has been reported The

annual harvest of Laminaria spp from the eastern side of

the peninsula is approximately 600 tons dry weight, but

harvests from the western side of the peninsula are

approximately 0 [7] It is difficult to quantify the spatial

variability of kelp forests over large areas using traditional

survey designs and analyses Traditional surveys of sea

forests including the kelp of Shiretoko Peninsula have been

conducted mainly by diving or from onboard observations

[8] These direct methods have advantages for species

identification and for obtaining detailed information on the

growing conditions of kelp forests, but such methods are

highly demanding of time and resources when attempting

to cover large and complex areas such as the Shiretokocoast [9]

In recent years, integrated methods using acousticobservations via echosounder and geostatistical analyseshave been suggested as practical survey and quantificationmethodologies for the mapping and ecological study ofseagrass and seaweed beds in coastal waters [9, 10] Anechosounder transmits ultrasonic waves through the waterand continuously measures the reflections of objects (ech-oes) such as sea bottoms and fish schools during surveys[11] Since acoustic observation via echosounder continu-ously measures echoes, researchers are not required tofrequently change the echosounder setting or to take manydirect sample records during surveys Acoustic observa-tions have already been applied to several studies of sea-weed and seagrass beds [12–14] For the estimation ofhorizontal distribution of kelp forests, data obtained fromacoustic observation via echosounder were geostatisticallyinterpolated using kriging [15] Geostatistical interpolationhas been applied in the past to both terrestrial and aquaticplant distribution studies [10, 16] Geostatistical interpo-lations can estimate the abundance of target plants withstatistical references based on values such as the thickness

or density of the target [17] For mapping and precisequantifications with objective validations, the application

of geostatistical interpolation procedures using reasonableamounts of data from acoustic observations offers a prac-tical methodology The combination of acoustic observa-tion and geostatistical interpolation would be an effectivemethod for conducting a quantitative mapping study of thekelp forest distribution along the coast of the ShiretokoPeninsula

In this study, the objective was to estimate the tion of the kelp forests of the Shiretoko Peninsula usingacoustic observation and geostatistical analysis Weobserved the presence or absence and measured the thick-ness (height) of kelp forests using acoustic techniqueobservations, and we then geostatistically interpolated theseacoustic data and estimated the distribution of the forests.The distributions of kelp in eastern and western areas werecompared, and the distribution characteristics of the kelpforests of the Shiretoko Peninsula were evaluated from anecological standpoint

distribu-Materials and methodsData collection

Field surveys were conducted in the coastal waters of theShiretoko Peninsula from 11 to 15 August 2007 beforefisheries harvest and seasonal deterioration caused thekelp forests to decline (Fig.1, [18]) The subareas for

Western area Shiretoko Cape

Fig 1 Study area and survey transect line Thegray area along the

coast is the study region Solid lines indicate where measurements

were made Dashed line indicates the boundary line between the

eastern and western area (145°12.19 0 E, 44°01.14 0 N) The eastern area

is from Shiretoko Cape to Utoro The western area is from Shiretoko

Cape to Rausu Triangles mark the points where underwater video

was used for validation

analysis—the eastern area (23.74 km2) and the western

area (19.58 km2)—were defined by the border at the

Shiretoko Cape (44°01.140N, 145°12.190E; Fig.1) The

survey areas were set in water less than 30 m in depth,

corresponding with the depth limits for L ochotensis and

L diabolica, which are 18 and 25 m, respectively (Fig.1,

[19,20]) The survey cruise ran orthogonal or parallel to

the shoreline at about 400–800 m intervals, unless

evad-ing areas of shallow bottom, set net fisheries, or

aqua-culture The ship’s speed was 4–6 knots to avoid

cavitations around the transducer and to be sure to detect

small (\1 m) kelp forests

Sampling equipment to detect and measure kelp forests

consisted of an acoustic component and a differential GPS

(Trimble) linked to a laptop PC The acoustic component

consisted of a BL550 echosounder (Sonic) with a 200 kHz,

3° single-beam transducer that generated continuous pulses

(pings) every second (Table1) The vertical resolution of

the pulse was 6 cm The transducer was mounted off the

side of the research vessel at a depth of 0.5 m The BL550

digitizes the intensity of the echo with user-defined

parameters from level (Lv.) 1 (weak) to Lv 255 (strong) In

this study, we set the echo of the sea bottom to be equal to

Lv 255 Position report data (latitude and longitude) were

stored on the hard drive of the PC that operated the onboard

system Onboard or underwater video observations

(n = 83) were conducted to distinguish kelp from other

algae Data of other algae were excluded from the analysis

Detection of kelp forest echoes

Detected echoes were categorized into three groups: kelp

forest, sea bottom, and seawater (Fig.2a) Echoes from

solid targets such as the sea bottom are strong, while echoes

from seawater are weak due to the absence of objects to

reflect the transmitted supersonic waves (Fig.2b) Echo

categorization was made by validating acoustic data using

an underwater video camera (Fig.1) Based on these

cate-gorizations, detected echoes with intensities equal to Lv

255 were categorized as sea bottom, and echoes with

intensities of less than Lv 255 were categorized as kelp

forest or seawater All echoes with intensities of less than

Lv 4 were categorized as seawater (Fig.2b) The thickness

of the kelp forest was measured between Lv 255 and Lv 4

We excluded detected objects less than 30 cm in thicknessfrom analyses to avoid possible confounding with theacoustic dead zone, which was calculated based on pulselength and local bathymetry [21,22] For the same reason,abrupt bathymetry changes, such as near large rocks(C30 cm) or steep slopes, were excluded from analyseswhen detected Additionally, acoustic observations in 34randomly selected sites were compared to in situ kelp for-ests using ROV as a post survey validation We confirmedthat mean errors were less than the vertical resolution of theBL550 (6 cm)

Estimation of the distribution of kelp forests usinggeostatistical analysis

We evaluated spatial autocorrelations within processedkelp forest data as horizontal distribution trends, thenestimated the area and thickness of kelp forests by kriging[15, 17] The observed spatial autocorrelations in theeastern and western area experimental semivariograms (c)were calculated as

c hð Þ ¼ 12n

Xn i¼1

This study focused on the spatial aspects of kelp forestover the Shiretoko coastal shelf instead of the dynamics ofindividual plants So, the minimum resolution for semi-variogram analyses corresponded to the unit of distanceamong forest patches The center of each kelp forest patchwas defined based on the statistical quantile marking the

Table 1 Acoustic specifications of the BL550 echosounder

0 Seawater

Kelp Forest

Sea Bottom

Lv 255

Lv 1 4

250

Top of the Kelp Forest

Level 0

Fig 2 a Detected kelp forest with echosounder Black line is the top

of the kelp forest White line is the bottom of the kelp forest The region between the lines is estimated as the kelp forest The color bar indicates echo level b An example of one ping of the kelp forest Dashed lines are the boundaries between the kelp forest and seawater

or between the kelp forest and sea bottom

upper 5th percentile of the measured thickness frequency

distribution The general intervals among the defined

cen-ters of kelp forest were obtained from the average

nearest-neighbor distance among the center points (85 m)

The best-fit theoretical semivariograms were selected

from obtained experimental semivariograms using the

maximum likelihood algorithm for each subarea (Fig.3,

[23]) A spherical model was used as the function because

it makes fewer assumptions in the model parameters and

because it achieved a better fit than other model candidates

in pilot analyses First, to calculate theoretical

semivario-grams with minimum spatial bias from the survey design,

the sampling circles needed to include data from at least

two transect lines Therefore, we used values of c within

1,700 m of each location for model fitting Then, the model

parameters (range, partial sill, and nugget) were obtained

from the selected best-fit models Range is the maximum

distance across which spatial autocorrelation exists in kelp

distribution and is a vector in which the partial sill is

observed The partial sill is the maximum variability,

which depends on the distance between pairs of data at

specific locations The nugget is the value of c, which is the

variability within a lag including random errors Using

these parameters of the theoretical semivariograms, kelp

forest distributions between eastern and western areas were

compared

Kelp forest distribution was analyzed from two different

aspects: (1) presence or absence of the forests or forest

patches and (2) the variation in thickness within or among

the present forest patches We will describe the horizontal

distribution properties of kelp forests and discuss the causal

mechanism of biological distribution [17, 23, 24] We

applied a two-stage approach to predict the presence or

absence of kelp forest and then to estimate the thickness of

present forests in each subarea In the first stage, the

probability of kelp occurrence was predicted using

proba-bility kriging with the best-fit theoretical semivariogram

based on the c of occurrence [23] To calculate c ofoccurrence, the thickness of the subsampled kelp forests(n = 15,000) including absence data (thickness = 0 cm)was first normal-score transformed [23] The presence([0.5 probability) or absence (\0.5 probability) of kelpforest was determined based on the interpolated probability

of occurrence The interpolation of presence or absence ofkelp forests was validated by concordance rate, calcu-lated in the manner of leave-one-out cross-validations(LOOCVs, [23,25]) Then, all of the observed presence orabsence values were compared to predictions In the secondstage, the thickness of the kelp forest was estimated usingordinal kriging based on the best-fit theoretical semivari-ogram of thickness based on the c of thickness [23] Tocalculate c of thickness, the z values of the thicknesssubsamples were natural-log transformed, using onlypresence data (thickness C30 cm) The interpolation ofthickness was validated using LOOCVs, then the root meansquares of errors (RMSEs) were evaluated

Kelp forest areas were calculated and compared betweenthe eastern and western areas The calculations and inter-polations above were made using ArcGIS ver 9.2 (Envi-ronmental Systems Research Institute, ESRI)

ResultsDetected kelp forests

In the eastern area, 2,717 of 15,000 pings were echoes fromkelp forests detected along the survey transect The mea-sured thickness ranged from 30 to 108 cm with an average(±SD) of 53 ± 13 cm (Fig.4) In contrast, in the westernarea, 1,330 pings of 15,000 were echoes from kelp forests.The measured thickness ranged from 30 to 78 cm with anaverage of 44 ± 18 cm (Fig 4) The maximum thickness

of the kelp forest in eastern areas (108 cm) was 30 cmthicker than in western areas (78 cm) Also, 10% of thekelp forests in the eastern area were thicker than themaximum thickness of forests in the western area.Semivariograms

The semivariogram parameters indicated differences inspatial trends in occurrence of kelp forests between theeastern and western areas (Fig.5a; Table2a) The partialsill in the eastern area (5.74 9 10-2) was 22 times largerthan that in the western area (0.26 9 10-2) Conversely, therange in the western area (1,697 m) was three times longerthan that of the eastern area (530 m) These differences insemivariogram parameters indicate that the occurrence ofkelp forest in the eastern area is more horizontally variablewith patches or gaps [26,27] than that in the western area

Fig 3 An example of an experimental semivariogram Circles

comprise the experimental semivariogram The solid line represents

the theoretical semivariogram

Again, the semivariogram parameters of thickness of thekelp forest were different between the eastern and westernareas (Fig.3b; Table2b) The eastern area was character-ized by a smaller partial sill (1.07 9 10-2) and larger range(370 m) than that of the western area (partial sill =1.68 9 10-2, range = 223 m) These results indicate rela-tively uniform thickness among present forest patches in theeastern area compared to the western area On the otherhand, the nugget in the eastern area (9.43 9 10-2) wasapproximately two times larger than that in the western area(5.69 9 10-2) The difference in nugget may indicate thatthe thickness of kelp forests within patches is more sto-chastically variable in the eastern side relative to the westernside, though it may also suggest potential observation errorsduring the survey.

Distribution of the kelp forestsThe interpolated kelp forest is shown in Fig.6 Overall,kelp forests in the eastern area were larger and thicker thanthey were in the western area Kelp forests in the easternarea extended farther offshore compared to those in thewestern area The total area of the kelp forests surveyedwas 3.88 km2 In the eastern area, kelp forests were con-tinuously distributed along the coastline and were espe-cially obvious near the Shiretoko Cape The distributedarea of kelp forests in the eastern region was 3.49 km2,which was 15% of the analyzed area (23.74 km2) Theestimated thickness of the eastern forests ranged between

34 and 91 cm, with most (90%) below 64 cm (Fig.7a).The 10% of kelp forests that was over 64 cm thick made up

an area of 0.34 km2and was patchily distributed over theentire eastern area In the western area, kelp forests weresparsely distributed The area of western kelp forest was0.39 km2, which was 2% of the analyzed area (19.58 km2).The estimated thickness of the western forests ranged from

34 to 75 cm (Fig 7b), with little forest over 64 cm thick.The concordance rates with LOOCVs for the prediction

of presence or absence were 99% for both subareas.The RMSEs of the estimated thickness in eastern andwestern areas were 18 and 12 cm, respectively

Eastern area Western area 0

40 80 120

Fig 4 Box plot of measured kelp forest thickness data in eastern and

western areas The boxes span the first to third quartiles of thickness.

The vertical bar indicates the minimum and maximum values The

cross mark in the box represents the average

0.1

0.2

0.3

(b)

Fig 5 Semivariograms of a the occurrence of kelp forest and b the

thickness of the kelp forest Closed circles comprise the experimental

semivariogram in the eastern area Open circles comprise the

experimental semivariogram in the western area The solid line

represents the theoretical semivariogram in the eastern area The

dashed line represents the theoretical semivariogram in the western

This is the first quantitative study of coastal kelp forests

using acoustic observations and geostatistical analysis The

thickness of kelp forests over a 43.3 km2 area was cessfully measured under a set transect Acoustic obser-vations of various aquatic plants have been conducted[12–14], but detailed observations covering areas of over

N

EstimatedThickness (cm)91

34

(a)

ShiretokoCape

(b)

Fig 6 a Map of kelp forest distribution Thickness estimation was overlapped with presence or absence estimation b Close up view of Shiretoko cape (dashed region)

10 km2are limited to a study of seagrass beds in Ajino Bay

[28] The present study is therefore notable as one of the

pilot applications of the acoustic method Based on the

high-resolution data from across the study area, statistically

valid estimations were made of the kelp forest Spatial

variability in the distribution of the kelp forests was also

statistically analyzed

The kelp forests along the eastern side of the peninsula

were relatively thick and horizontally continuous, while

those along the western side were sparse (Figs.6,7) The

spatial variability in occurrence of forest patches along the

eastern side was larger than it was along the western side, and

the spatial variability in thickness along the eastern side was

more continuous than along the western side, as indicated in

the semivariograms (Fig.5) Structurally complex kelp

forests provide fish and phytal animals with suitable refuges

from advection and predators and also attract a variety of

other fauna [29–33] The horizontal and vertical structure of

the eastern side probably enhances the diversity of fauna and

ecological relationships Conversely, the ecosystem on the

western side might be simpler, but the sparse patches of kelpforests may provide locally valuable habitats to organisms.These spatial characteristics of the kelp forests are prob-ably a consequence of the differences in ecological setupbetween the eastern and western sides of the ShiretokoPeninsula The difference in the thickness of kelp forestsbetween the eastern and western sides was caused by thedifference in species that form the forests The major species

on the eastern side of the peninsula are L ochotensis and L.diabolica, but only L ochotensis is observed in the westernareas [1,2] The different species composition of kelp forests

on the eastern and western sides of the peninsula may be acause of the observed spatial differences in forest thickness,and future field validation of species composition isrecommended

The observed differences in vegetated area between theeastern and western sides of the Shiretoko Peninsula cor-responded with the harvest of Laminaria spp on the pen-insula [7] The grazing impact of benthic herbivores such

as sea urchin (Strongylocentrotus spp.), which influencesthe variability in marine environments, has been discussed

as a possible determinant of change in the area covered bykelp forests in various North Pacific waters [7] On RishiriIsland, located in northern Hokkaido, the distribution ofkelp forests declined due to intense grazing by Strongylo-centrotus nudus, which was influenced by changes in watertemperature [34] The coastal sea off the Shiretoko Pen-insula is one of the major Strongylocentrotus spp habitats,

so the distribution of the kelp forest along the peninsulawould be influenced by Strongylocentrotus spp More than

10 times the number of Strongylocentrotus spp have beenfound on the western side of the peninsula than on theeastern side [7] Grazing impacts associated with envi-ronmental variability are most likely one of the controllingfactors of the continuity of kelp forests in coastal watersalong the Shiretoko Peninsula

One of the most dramatic changes in the marine ronment in the Shiretoko Peninsula comes from drift icereaching the shore from January to March The ice physi-cally scrapes the kelp forest from the rocks along the shore.The harvest of Laminaria spp declines in coastal watersalong Nemuro, located in the east of Hokkaido, in yearsexperiencing intense drift ice [6] From 2000 to 2007, thetotal annual duration for which drift ice reached westernareas was approximately 30 days longer than the durationfor which it reached eastern areas (The Drift Ice ConditionChart of the ICE Information Center, Japan) These spatialdifferences in drift ice distribution potentially influence thefoliaged areas of kelp forests in the study area

envi-As discussed above, various ecological factors will alterthe distribution of kelp forests in the coastal waters of theShiretoko Peninsula The integrated analysis based on thisstudy is probably effective in revealing the causal

Fig 7 Frequency distributions of thickness estimated in a the

eastern and b the western side of Shiretoko Peninsula

relationships that determine the distribution of kelp forests

in this area Quantified information on kelp forest

distri-bution is not only of practical use to local fisheries but is

also important for the sustainability of the marine

ecosys-tems of the peninsula Conducting further studies that

integrate direct species sampling with estimates of kelp

forest primary production along the Shiretoko Peninsula

will provide practical information for the assessment of

ecology Ongoing spatial and temporal monitoring of

ecologically important kelp forests will provide essential

information to fishers and managers for the future

sus-tainability of the coastal waters of the Shiretoko Peninsula

Acknowledgments We thank the Fisheries Cooperation

Associa-tion of Rausu and the Sonic CorporaAssocia-tion for their support in

con-ducting this study and researcher Y Fukuda and boatman K Sudou

for their kind advice and assistance This study was supported by the

Ministry of the Environment, Japan.

References

1 Hasegawa Y (1959) On the distribution of the useful

laminaria-ceous plants in Hokkaido (in Japanese) Hokusuishige 16:201–206

2 Kikuchi S, Uki N (1981) Productivity of benthic grazers,

abalone and sea urchin in Laminaria beds (in Japanese) In:

Yaduka T et al (eds) Seaweed beds Koseisha Koseikaku,

Tokyo, pp 9–23

3 Makino M, Matsuda H, Sakurai Y (2009) Expanding fisheries

co-management to ecosystem-based co-management: a case in the

Shiretoko World Natural Heritage area, Japan Mar Policy

6 Sasaki S (1973) Studies on the life history of Laminaria

angus-tata var longissima (M.) Miyabe Hokkaido Kushiro Fisheries

Experimental Station, Hokkaido

7 Machiguchi Y (2008) Urchin barrens (in Japanese) In: Fujita D

et al (eds) Recovery from urchin barrens Seizando Shoten,

Tokyo, pp 22–28

8 Mukai H, Aioi K, Ishida Y (1980) Distribution and biomass of

eelgrass (Zostera marina L.) and other seagrasses in Odawa Bay,

central Japan Aquat Bot 8:337–342

9 Komatsu T, Igarashi C, Tatsukawa K, Nakaoka M, Hiraishi T,

Taira A (2002) Mapping of seagrass and seaweed beds using

hydro-acoustic methods Fish Sci 68:580–583

10 Holmes KW, Vanniel KP, Kendrick GA, Radford B (2007)

Probabilistic large-area mapping of seagrass species distributions.

Mar Freshw Ecosyst 17:385–407

11 Urick RJ (1979) Principles of underwater sound (translated from

English by A Tuchiya) (in Japanese) Kyoritsu Shuppan, Tokyo

12 Komatsu T, Mikami A, Sultana S, Ishida K, Hiraishi T,

Tatsuk-awa K (2003) Hydro-acoustic methods as a practical tool for

cartography of seagrass beds Otsuchi Mar Sci 28:72–79

13 Sabol BR, Melton E, Chamberlain R, Doering P, Haunert K

(2002) Evaluation of a digital echo sounder system for detection

of submersed aquatic vegetation Estuaries 25:133–141

14 Kitoh H (1983) Seaweed beds research by a scientific under (in Japanese) Seikai Reg Fish Res Inst News 43:2–4

echoso-15 Burrough PA, Mcdonnell RA (1998) Optimal interpolation using geostatistics In: Burrough PA et al (eds) Principles of geo- graphical information systems Oxford University Press, New York, pp 132–161

16 Penn MG, Moncrieff CB, Bridgewater SGM, Garwood NC, Bateman RM, Chan L, Cho P (2009) Using GIS techniques to model the distribution of the economically important xate´ palm Chamaedorea ernesti-augusti within the Greater Maya Moun- tains, Belize Syst Biodiv 7:63–72

17 Wackernagel H (1995) Multivariate geostatistics (in Japanese) Morikita Shuppan, Tokyo, pp 33–119

18 Nabata S, Sakai Y (1996) Annual net production of the second year frond of Laminaria diabolica (in Japanese) Sci Rep Hokkaido Fish Exp Stn 49:1–5

19 Nabata S (2003) Laminaria spp (in Japanese) In: Notoya M (ed) Seaweed beds and creation technology Seizando Shoten, Tokyo,

pro-22 Mitson RB (1994) Fisheries sonar (in Japanese) Koseisha Koseikaku, Tokyo, pp 9–23

23 Johnston K, Jay MH, Krivoruchko K, Lucas N (2001) Using ArcGIS geostatistical analyst ESRI, New York, pp 131–218

24 Burrough PA, Mcdonnell RA (1998) Optimal interpolation using geostatistics In: Burrough PA et al (eds) Principles of geo- graphical information systems Oxford University Press, New York, pp 132–161

25 Manase S, Takeda J (2001) Spatial data modeling (in Japanese) Kyouritsu Shuppan, Tokyo

26 Yamamoto S (2000) Forest gap dynamics and tree regeneration.

J For Res 5:223–229

27 Manabe T, Shimatani K, Kawarasaki S, Aikawa S, Yamamoto S (2009) The patch mosaic of an old-growth warm-temperate for- est: patch-level descriptions of 40-year gap-forming processes and community structures Ecol Res 24:575–586

28 Komatsu T, Tatsukawa K (1998) Mapping of Zostera marina L beds in Ajino Bay, Seto Island Sea, Japan, by using echosounder and global positioning systems J Rech Oceanogr 23:39–46

29 Fuse S (1962) The animal community in the Sargassum belt (in Japanese) Physiol Ecol 11:23–45

30 Umezawa S, Ohno M, Yamamoto T (1977) An ecological study

on the model of a seaweed belt: daily variation of the mental factors with reference to the behavior of fish (in Japanese) Usa Mar Biol Stn Kochi 24:13–26

environ-31 Komatsu T, Ariyama H, Nakahara H, Sakamoto W (1982) Spatial and temporal distribution of water temperature in a Sargassum forest J Oceanogr Soc 38:63–72

32 Jackson GA, Winant CD (1983) Effect of a kelp forest on coastal currents Cont Shelf Res 2:75–80

33 Komatsu T, Murakami S (1994) Influence of a Sargassum forest

on the spatial distribution of water flow Fish Oceanogr 3:256–266

34 Nabata S, Takiya A, Tada M (2003) On the decreased production

of natural kelp, Laminaria ochotensis in Rishiri island, northern Hokkaido (in Japanese) Sci Rep Hokkaido Fish Exp Stn 64:127–136

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

Effect of starvation on biochemical composition

and gametogenesis in the Pacific oyster Crassostrea gigas

Wenguang Liu•Qi Li•Fengxiang Gao •

Lingfeng Kong

Received: 20 December 2009 / Accepted: 3 June 2010 / Published online: 5 August 2010

Ó The Japanese Society of Fisheries Science 2010

Abstract The effects of starvation on biochemical

com-position and gametogenesis were investigated in the Pacific

oyster Crassostrea gigas Histological analysis, combined

with oocyte examination and measurements of protein,

glycogen and lipid levels and RNA/DNA ratio from gonad,

adductor muscle and mantle tissue of each sex were

per-formed In the starved groups, C gigas showed gonad

development, but the progress was delayed during the

experiment Glycogen was the first substrate used by

C gigas for dealing with lack of food While glycogen was

rapidly consumed, protein and lipid contents decreased

gradually A decrease in the RNA/DNA ratio in the starved

groups in all the body components was found during

starvation, illustrating that RNA/DNA ratio was a valid

indicator of nutritional condition in C gigas A significant

increase in water and ash contents and a corresponding

decrease in condition index were observed in the starved

groups, showing that the water and ash content and

con-dition index were related to the usage of glycogen, lipid

and protein reserves in body composition During

starva-tion, energy reserves were mobilized for survival and

gonad development, but spawning was arrested Theinformation obtained in this study is useful for broodstockmanagement in the Pacific oyster industry

Keywords Crassostrea gigas Starvation Biochemical composition  Gametogenesis

IntroductionSeasonal variation in gametogenesis of marine bivalves isclosely related to the energy storage-utilization cycle andenvironmental factors such as water temperature and foodavailability [1] In general, reserves are stored prior togametogenesis, when food is abundant, in the form ofglycogen, lipid and protein, which are used, together withthe available food, according to requirements during thegametogenic process The particular importance of thesesubstrates, where they are stored and the timing of their usevaries among species, as well as among populations of thesame species [2,3] Gametogenesis of marine bivalves is aprocess that requires energy, and the way to obtain energydiffers among species Some species use the recentlyingested energy from the seston (opportunistic species),such as Tellina tenuis and Abra alba [2] Some species usethe energy of substrates stored in various organs and tissuesthrough feeding prior to their gametogenesis (conservativespecies), such as Chlamys opercularis [4], Argopectenirradians concentricus [5] and A purpuratus [6] There arealso species that can adopt either strategy, depending onthe population, or that utilize both strategies, depending onthe time of year [7] The reproductive strategy of marinebivalves can be considered an adaptation to ambientenvironmental factors, such as food availability and watertemperature

W Liu

Key Laboratory of Marine Bio-resources Sustainable Utilization,

South China Sea Institute of Oceanology, Chinese Academy of

Sciences, Guangzhou 510301, China

Q Li ( &)  F Gao  L Kong

Fisheries College, Ocean University of China, Qingdao 266003,

The Pacific oyster Crassostrea gigas is an important

aquaculture species, ranking first in terms of total weight

and second in terms of value of all global aquaculture fish

and shellfish species [8] In China, the production of

oys-ters reached 3.89 9 106metric tons in 2006, accounting for

32.4% of total marine molluscan yield [9] Production of

cultured C gigas seedlings is well developed in China To

a large extent, success of C gigas larvae culture depends

on larval energy reserves, which support embryogenesis

and metamorphosis [10, 11] Therefore, it is essential to

provide broodstock of good quality in order to ensure the

quality of subsequent eggs, larvae and spats [12] As a

consequence, information on the reproductive strategy of

broodstock is important

Several studies have focused on the seasonal variations

in the biochemical composition of C gigas in relation to

the reproduction Kang et al [13] reported that C gigas

cultivated in coastal bays of Korea had a direct dependence

on food availability and could be considered to be an

opportunistic species, whereas Ruiz et al [14] found that in

C gigas suspended in El Grove (Galicia, Spain), the

gly-cogen stored was used in the gametogenesis, and the

pro-tein and lipid were utilized in winter when available food

was scarce Ren et al [15] concluded that glycogen was the

main energy reserve for gametogenesis in C gigas from the

Marlborough Sounds, New Zealand Dridi et al [1]

reported that the amount of energy for gametogenesis

activity in C gigas in Tunisia depended more on the

available food than on nutrient storage sites such as

adductor muscle However, in these studies the utilization

of stored nutrients in C gigas during gametogenesis has

been inferred from monthly field studies, and there is no

direct evidence available on energy sources for

reproduc-tion in C gigas

This study examined the effect of starvation on

game-togenesis and biochemical composition Our objective was

to identify the reproductive strategy of C gigas through

food deprivation A better understanding of reproductive

strategy of the Pacific oyster would be important for

aquaculture practices and lead to optimization of broodstock

management of this commercially important aquaculture

species

Materials and methods

Animal collection and experimental design

In April 2005, Pacific oysters (shell height, 8.4 ± 0.9 cm;

shell length, 4.7 ± 0.6 cm) were collected from Weihai,

Shandong Province, China, and transported live to the

coastal laboratory in Yantai, Shandong Province They

were acclimated in the aquarium at ambient seawater

temperature of 14.8°C for 1 week before beginning theexperiment One thousand oysters were then divided intoten groups and placed in separate aquariums with adensity of 35 individual m-3 Five groups were fed dailywith a microalgae mixture of Chaetoceros muelleri,Nitzschia closterium minutissima, Platymonas helgolan-dica, Chlorella vulgaris and Isochrysis galbana at adensity of 1 9 105cells ml-1for 4 h Another five groupswere used as starvation control The seawater used wasfiltered by absorbent wool and activated charcoal to pre-vent contamination of natural microalgae The seawaterwas aerated and changed daily in the aquariums Duringthe experiment, oysters were reared at ambient seawatertemperature

Oysters from all groups were sampled on days 0, 30, 60and 90 On each sampling day, 30 oysters were randomlyselected from every group They were dissected to obtaingonads, mantles and adductor muscles A 5-mm-thicksection of gonad was fixed in Bouin’s solution for histo-logical examination The remaining tissues were frozen andstored at -80°C until used

Biochemical analysisFor the biochemical characterization of the body com-ponents, the levels of protein, glycogen, lipid and nucleicacids were estimated Total protein was determined by theKjeldahl method The dried, powdered samples werecatalytically digested with sulphuric acid and analyzed by

an automatic Kjeldahl analysis instrument (KjeltecTM2300; Foss Tecator, Sweden) The amount of N wasmultiplied by 6.25 to estimate the amount of proteins[16] For determining the lipid levels, extractions weremade in chloroform-methanol (3:1) using a soxhletapparatus [17] The glycogen content was determinedwith minor modifications to the anthrone-sulfuric acidmethod described by Horikoshi [18] The powdered,freeze-dried samples were suspended in 60 volumes of30% KOH and saponified by heating to 100°C for

30 min After cooling, a portion of the saponified mixturewas treated with the cold 0.2% anthrone-sulfuric acidsolution for 10 min; absorbance of the resulting coloredcomplex was measured at the wave length of 620 nm.After the samples were homogenized in 20 volumes ofdistilled water, 1 ml of each homogenate was used fordetermination of nucleic acid (DNA and RNA) contentsaccording to the modification of the Schmidt-Thamm-hauser-Schnerder method by Nakano [19] Nucleic acidswere precipitated with ethanol and washed with a mixture

of ethanol and ether RNA was separated by alkalinehydrolysis, and DNA was hydrolyzed with perchloricacid DNA and RNA contents were determined by measuringtheir absorbances at 260 nm

Condition index, water and ash content

The condition index of the oyster (CI) was calculated as the

ratio of the dry weight of the soft parts/dry weight of

shell 9 100 [20]

CI¼Dry weight of the soft parts

Dry weight of shell  100

The water content of the soft body was determined by

drying the whole flesh at 105°C to constant weight Ash

content was obtained by burning the whole flesh body of

the oyster at 450°C for 48 h in a muffle furnace

Histology

The gonad was routinely processed for histology, and 6-lm

paraffin-embedded sections were stained with Mayer’s

hematoxylin and eosin The specimens were examined

microscopically to develop a profile of gametogenesis The

diameter of 100 oocytes was microscopically measured in

sections from five animals in each sample

Statistical analysis

All data were tested for homoscedasticity and normality

prior to running statistical tests Two-way ANOVA

fol-lowed by Scheffe test was conducted to assess significant

differences in biochemical composition and indices among

different treatments The software SPSS 13.0 was used for

analyses

Results

Biochemical composition

Dry weights of the mantle, adductor muscle and gonad in

the fed and starved groups of male and female C gigas are

displayed in Table1 Starved oysters had lower dry weightwhen compared to unstarved oysters The glycogen content

in the mantle, adductor muscle and gonad decreased fromday 0 to 60 and was maintained at a constant level on days60–90 in the fed and starved groups (Fig.1) In the mantle,significant differences between the fed and starved groupswere detected on day 30 and 60 in females and males,respectively In the adductor muscle, the glycogen contentshowed significant differences between the fed and starvedgroups on day 30 both in females and males In the gonad,the glycogen content was significantly different betweenthe fed and starved groups on day 30 in females

The lipid content in the mantle did not changemarkedly during the experiment (Fig.2) The lipid con-tent in the adductor muscle showed no change in the fedgroups, but decreased gradually during days 30–90 in thestarved groups The lipid content in the gonad increasedfrom day 60 to 90 in the fed groups both in females andmales, but remained constant in the starved groups ondays 60–90, showing significant differences on day 90both in females and males between the fed and starvedgroups (Fig.2)

The protein content in mantle and adductor muscle didnot show significant variation in the fed groups of femalesand males, but decreased in the starved groups (Fig.3).Significant differences were presented in female and malemantle and adductor muscle on day 60 and 90 between thefed and starved groups In the fed groups of the female andmale gonads, the protein content increased gradually fromday 0 to a maximum value on day 90 In contrast, theprotein content in the gonads of the starved groupsdecreased from day 0 to 90 The protein content was sig-nificantly different between the fed and starved groups ondays 30–90 in the female and male gonads

The RNA/DNA ratio in the mantle remained constantfrom day 0 to 90 in both females and males of the fedgroups (Fig.4) The variations in the RNA/DNA ratio inthe adductor muscle of the fed groups were similar to those

Table 1 Dry weight (mean ± SD, n = 5) of the mantle, adductor muscle and gonad in the fed and starved groups of male and female Crassostrea gigas

of the mantle The RNA/DNA ratio in female and male

gonads of the fed groups increased gradually on days 0–60,

but was not statistically significant The RNA/DNA ratio

decreased continuously in mantle, adductor muscle and

gonad of the starved groups during the starvation period

Significant differences in the RNA/DNA ratio between the

fed and starved groups in both males and females in these

tissues were found on day 90

Water and ash contentThe ash and water contents of the soft body showed noclear change in the fed groups, but increased in the starvedgroups from day 0 to 90 (Fig.5) Significant differencesbetween fed and starved groups were found for the ashcontent during days 60–90 and for water content on day 90,respectively

Fig 1 Glycogen content

(mean ± SD, n = 5) in the fed

and starved groups of male and

female Crassostrea gigas.

Means not sharing the same

superscript are significantly

different (P \ 0.05)

Fig 2 Lipid content

(mean ± SD, n = 5) in the fed

and starved groups of male and

female C gigas Statistical

significance is the same as for

Fig 1

Gametogenesis, oocyte diameter and condition index

For the fed groups, sexes were just identifiable, but very

little gonad was developed on day 0, indicating that the

gonad development was at the start stage of gametogenesis

(Fig.6) [21] On day 30, the follicles were not in contact

with one another, but they had clearly defined folliclewalls In the female, oocytes were beginning to develop Inthe male, follicles filled by spermatozoa were arranged incharacteristic bands On day 60, gametogenesis was com-plete, but spawning has not yet occurred Oocytes wereliberated freely on dissection, and free sperm were

0 20 40 60 80

Male Female

0 20 40 60 80

c

0 20 40 60

a

Fig 3 Protein content

(mean ± SD, n = 5) in the fed

and starved groups of male and

female C gigas Statistical

significance is the same as for

Fig 1

0 1 2

3

b a

0 1 2

Fig 4 RNA/DNA ratio

(mean ± SD, n = 5) in the fed

and starved groups of male and

female C gigas Statistical

significance is the same as for

Fig 1

manifested as streaks On day 90, the male gonads were

characterized by the loss of the radial arrangement of the

spermatozoa, while in females empty spaces were observed

in the follicles, and ripe oocytes at a lower density were

observed, which demonstrated that spawning had started

For the starved groups, gametogenesis began on day 0, but

the progress was delayed compared with the fed groups on

day 30 and 60 The developed gametes were kept

throughout the experiment and spawning did not occur

(Fig.6)

The condition index in the fed groups was maintained at

constant levels from day 0 to 60 and dropped significantly

on day 90 For the starved groups, the condition indexdecreased gradually from 6.6 on day 0 to 5.0 on day 90.The oocyte diameter in the fed groups increased from36.8 lm on day 0 to 53.8 lm on day 60, then slightlydecreased on day 90 (Fig.7) The oocyte diameter in thestarved groups showed a similar pattern, but was lowerthan that in the fed groups There were significant differ-ences between the fed and starved groups in conditionindex and oocyte diameter on day 60 The rearing tem-perature varied from 14.8 to 28.2°C during the starvationperiod

DiscussionStarvation is not an uncommon circumstance for marinebivalves [22,23] It induced both immediate and delayedwhole-body responses, including a reduction in metabolicand excretion rates, and a decline in tissue energy reservesand enzyme activity [24, 25] The depletion of energyreserves resulting from food deprivation brings concomi-tant changes in biochemical composition Storage of gly-cogen, protein and lipid may be differentially affected,which can markedly alter the composition of different tis-sues [26]

In this study, the glycogen content considerablydecreased in all the body components, especially in thegonads with an increase in oocyte diameter during gonad

Fig 5 Water and ash contents (mean ± SD, n = 5) in the fed and

starved groups of male and female C gigas Statistical significance is

the same as for Fig 1

F 0

F 0

F

F

Starved group Fed group

Fig 6 Histological

observations on the gonad

during sexual maturation in

C gigas M male, F female.

Bar 100 lm

development, suggesting that gametogenesis of C gigas

depended largely on the glycogen stored in the tissues It is

generally accepted that glycogen reserves are the main

source of energy in bivalves [13, 27] and also may be

utilized for the formation of gametes under conditions of

nutrient stress [5,28] The glycogen content was lower in

the starved oysters than in the fed oysters, demonstrating

that glycogen reserves were quickly mobilized and

deple-ted because of food deprivation The preferential utilization

of glycogen in proportion to other biochemical components

is probably a means of protecting against the loss of protein

and lipid (the structural components of the animal) [26]

Moreover, a fast decrease in the glycogen content at the

start of starvation but a slow decrease thereafter may be a

response of the enzymatic machinery for cautious

utiliza-tion of glycogen reserves to preserve the valuable reserves

in case fasting is prolonged [29]

Once a large proportion of the glycogen reserves has

been consumed, the lipid and protein become energy

sources In the starved groups, the lipid in adductor muscle

decreased from day 60 to 90, and the protein in all the body

components became significantly lower than those in the

fed groups on days 60–90 This suggests that the lipid and

protein in the starved groups were mobilized as energy

sources during this period This was in agreement with

the result from the starvation experiment of the Manila

clam Tapes phillippinarum, which showed that the total

maintenance calories were supplied by carbohydrate, tein and lipid, respectively [25], but contrast to the primarysupply of energy mainly by carbohydrate in the starvationstudy of Rangia cuneata [26]

pro-Our results indicated that glycogen was the first sourceused by C gigas for dealing with lack of food Whileglycogen was rapidly consumed, protein and lipid contentdecreased Reports about the metabolic requirements ofglycogen, protein and lipid under starvation in bivalveswere different In C gigas, Riley [30] observed that thelevel of glycogen dropped throughout the longest starva-tion period as opposed to a slight decrease for lipid andprotein In the abalone Haliotis discus hannai, carbohy-drate, lipid and protein were used as nutrient reserves inshort starvation periods (20 days; Du and Mai) [31] In thejuvenile clam T philippinarum, 75% of total carbohydrateswere used over a 35-day starvation period [25]

The relative importance of energy reserves and theirorder of utilization varied among species [31] Differences

in energy source usage may be species-specific, and not allthe three energy sources (glycogens, proteins and lipids)were evaluated in the same bivalve at the same time underthe same conditions Strategies of fuel reserve usage maychange depending on the species, starvation time anddevelopmental stage Such strategies were favorablyselected for the ability to prolong the survival of theorganisms and, therefore, to increase their competitiveabilities [32,33]

Many studies have shown that the RNA/DNA ratio is auseful indicator of nutritional condition and growth rate forseveral species of invertebrates [34, 35] Since cellularRNA is essential for the biosynthesis of proteins, theamount of bulk RNA increases rapidly in growing tissues,while the amount of cellular DNA remains fairly constant.The RNA/DNA ratio reflects recent growth and is an index

of the cell’s synthetic capacity [36] Thus, it is expectedthat in a rapidly growing animal the amount of RNA ishigh However, protein synthesis diminishes with foodlimitation, and if this condition persists, attaining starva-tion, there will be degradation of ribosomes and loss ofRNA [36] In the present study, a decline in the RNA/DNAratio in the starved groups in all the body components wasfound from the start of starvation, indicating that starvationclearly affected the RNA content and the RNA/DNA ratiowas a valid indicator of nutritional condition in C gigas Asimilar phenomenon has been reported for C virginica[37] It is likely that the rapid decrease in the RNA/DNAratio during starvation is a common feature in molluscs, asdescribed in fish [38] In the fed group, although the proteincontent in the female gonad increased significantly, theRNA/DNA ratio remained constant during the experiment,suggesting that the RNA/DNA ratio in the female gonad

Fig 7 Condition index (mean ± SD, n = 5), oocyte diameter

(mean ± SD, n = 100) and water temperature in the fed and starved

groups of male and female C gigas Statistical significance is the

same as for Fig 1

was not sensitive enough to monitor the synthetic activity

of protein

One of responses to starvation observed here was the

increase in water and ash content and corresponding

decrease in condition index During starvation, energy was

derived solely from endogenous resources, and tissues

were lost due to catabolic activities To maintain the

nec-essary body volume and internal turgidity during

starva-tion, the lost tissue mass must be replaced by water The

ash content, mainly chloride salts, increased presumably

due to cellular hydration [39]

The influence of environmental conditions on bivalve

reproduction was mentioned by several authors, with the

most important parameters being temperature and food

availability [2 7] In this study, histological analyses of

gametogenesis in C gigas of the fed groups indicated that

initiation of gametogenesis began on day 0 at low water

temperature (14.8°C), and spawning took place on day 90

when water temperature was high (28.2°C) The variations

in the temperature were consistent with the variations in the

oocyte diameters, indicating that temperature may play an

important role in determining the gonadal development and

spawning of C gigas However, in the starved groups,

C gigas showed gonad development from day 0, but the

progress was delayed, and spawning did not occur on day

90 This indicated that energy reserves could be mobilized

for gametogenesis of C gigas during starvation and

sug-gested that gametogenesis was dependent on the reserves

accumulated by C gigas prior to gonad development

In conclusion, our findings indicate that C gigas need

energy reserves (glycogen, lipid and protein) in all tissues

during starvation and spawning is arrested by food

limi-tation The results suggest that C gigas may be considered

a conservative species in gametogenic pattern The

infor-mation obtained in this study is useful for broodstock

management in the Pacific oyster industry

Acknowledgments The study was supported by National High

Technology Research and Development Program (2006AA10A409)

and 973 Program (2010CB126406).

References

1 Dridi S, Romdhane MS, Elcafsi M (2007) Seasonal variation in

weight and biochemical composition of the Pacific oyster,

Crassostrea gigas in relation to the gametogenic cycle and

environmental conditions of the Bizert lagoon, Tunisia Aquaculture

263:238–248

2 Bayne BL (1976) Aspects of reproduction in bivalve molluscs.

In: Wiley ML (ed) Estuarine processes Academic Press, New

York

3 Liu WG, Li Q, Yuan YD, Zhang SH (2008) Seasonal variations

in reproductive activity and biochemical composition of the

Cockle Fulvia mutica (Reeve) from eastern coast of China.

J Shellfish Res 27:1–7

4 Taylor AC, Venn TJ (1979) Seasonal variation in weight and biochemical composition of the tissues of the queen scallop Chlamys opercularis from the Clyde sea area J Mar Biol Assoc

UK 59:605–621

5 Barber BJ, Blake NJ (1981) Energy storage and utilization in relation to gametogenesis in Argopecten irradians concentricus (Say) J Exp Mar Biol Ecol 52:121–134

6 Martinez G (1991) Seasonal variation in biochemical tion of three size classes of the Chilean scallop Argopecten purpuratus Lamarck, 1819 Veliger 34:335–343

composi-7 Thompson RJ (1977) Blood chemistry, biochemical composition, and the annual reproductive cycle in the giant scallop, Placopecten magellanicus, from southeast Newfoundland J Fish Res Board Can 34:2104–2116

8 FAO (2008) World aquaculture production of fish, crustaceans, molluscs, etc., by principal species FAO, Rome

9 DOF (Department of Fisheries) (2007) China fisheries statistic yearbook China Agriculture Press, Beijing

10 Pernet F, Bricelj VM, Cartier S (2006) Lipid class dynamics during larval ontogeny of sea scallops, Placopecten magellani- cus, in relation to metamorphic success and response to antibi- otics J Exp Mar Biol Ecol 329:265–280

11 Pernet F, Tremblay R, Langdon C, Bourget E (2004) Effect of additions of dietary triacylglycerol microspheres on growth, survival, and settlement of mussel (Mytilus sp.) larvae Mar Biol 144:693–703

12 Pronker AE, Nevejan NM, Peene F, Geijsen P, Sorgeloos P (2008) Hatchery broodstock conditioning of the blue mussel Mytilus edulis (Linnaeus 1758) Part I Impact of different micro-algae mixtures on broodstock performance Aquacult Int 16:297–307

13 Kang CK, Park MS, Lee PY, Choi WJ, Lee WC (2000) Seasonal variations in condition, reproductive activity, and biochemical composition of the oyster, Crassostrea gigas (Thunberg), in suspended culture in two coastal bays of Korea J Shellfish Res 19:771–778

14 Ruiz C, Abad M, Sedano F, Garcia-Martin LO, Lo´pez JLS (1992) Influence of seasonal environmental changes on the gamete production and biochemical composition of Crassostrea gigas (Thunberg) in suspended culture in El Grove, Galicia, Spain.

J Exp Mar Biol Ecol 155:249–262

15 Ren JS, Marsden ID, Ross AH, Schiel DR (2003) Seasonal variation in the reproductive activity and biochemical compo- sition of the Pacific oyster (Crassostrea gigas) from the Marl- borough Sounds, New Zealand N Z J Mar Freshw Res 37:171–182

16 Stephen D (1980) The reproductive biology of the Indian oyster Crassotrea madrasensis (Preston) II Gametogenic cycle and biochemical levels Aquaculture 21:147–153

17 Folch J, Lees M, Stanley-Sloane GH (1957) A simple method for the isolation purification of total lipids from animal tissues J Biol Chem 226:497–507

18 Horikoshi H (1958) Glycogen (in Japanese) Chem Field 34:36–39

19 Nakano H (1988) Techniques for studying on the early life history of fishes (in Japanese) Aquabiology 10:23–26

20 Walne PR (1976) Experiments on the culture in the sea of the butterfish Venerupis decussata L Aquaculture 8:371–381

21 Dinamani P (1987) Gametogenic patterns in populations of Pacific oyster Crassostrea gigas in Northland, New Zealand Aquaculture 64:65–76

22 Albentosa MM, Ferna´ndez-Reiriz J, Labarta U, Pe´rez-Camacho

A (2007) Response of two species of clams, Ruditapes decussatus and Venerupis pullastra, to starvation: physiological and biochemical parameters Comp Biochem Physiol B 146:241–249

23 Li Y, Qin JG, Li XX, Benkendorff K (2009) Spawning-dependent

stress response to food deprivation in Pacific oyster Crassostrea

gigas Aquaculture 286:309–317

24 Carefoot TH, Qian PY, Taylor BE, West T, Osborne J (1993)

Effect of starvation on energy reserves and metabolism in the

Northern abalone, Haliotis kamtschatkana Aquaculrure

118:315–325

25 Laing I (1993) The response of Manila clam, Tapes

phillippin-arum, juveniles to nutritive stress J Exp Mar Biol Ecol

173:111–121

26 Lane JM (1986) Allometric and biochemical studies on starved

and unstarved clams, Rangza Cuneata (Sowerby, 1831) J Exp

Mar Biol Ecol 95:131–143

27 Li Q, Osada M, Mori K (2000) Seasonal biochemical variations

in Pacific oyster gonadal tissue during sexual maturation Fish Sci

66:502–508

28 Beninger PG, Lucas A (1984) Seasonal variations in condition,

reproductive activity, and gross biochemical composition of two

species of adult clam reared in a common habitat: Tapes

decussatus L (Jeffreys) and Tapes philippinarum (Adams &

Reeve) J Exp Mar Biol Ecol 79:19–37

29 Sa´nchez-Paz A, Garcı´a-Carren˜o F, Herna´ndez-Lo´pez J,

Muhlia-Almazan A, Yepiz-Plascencia G (2007) Effect of short-term

starvation on hepatopancreas and plasma energy reserves of the

Pacific white shrimp (Litopenaeus vannamei) J Exp Mar Biol

Ecol 340:184–193

30 Riley RT (1976) Changes in the total protein, lipid, carbohydrate,

and extracellular body fluid free amino acids of the Pacific oyster

Crassostrea gigas, during starvation Proc Natl Shellfish Assoc

65:84–90

31 Du SB, Mai KS (2004) Effects of starvation on energy reserves in young juveniles of abalone Haliotis discus hannai Ino J Shellfish Res 23:1037–1039

32 Caers M, Coutteau P, Sorgeloos P (2000) Impact of starvation and of feeding algal and artificial diets on the lipid content and composition of juvenile oysters (Crassostrea gigas) and clams (Tapes philippinarum) Mar Biol 136:891–899

33 Riley RT (1980) The effect of prolonged starvation on the tive free amino acid composition of the extracellular body fluids and protein bound amino acids in the oyster, Crassostrea gigas Comp Biochem Physiol A 67:279–281

rela-34 Wagner M, Durbin E, Buckley L (1998) RNA:DNA ratios

as indicators of nutritional condition in the copepod Calanus fimarchicus Mar Ecol Progr Ser 162:173–181

35 Vidal E ´ AG, DiMarco P, Lee P (2006) Effects of starvation and recovery on the survival, growth and RNA/DNA ratio in loliginid squid paralarvae Aquaculture 260:94–105

36 Clemmesen C (1994) The effect of food availability, age or size

on the RNA/DNA ratio of individually measured herring larvae: laboratory calibration Mar Biol 118:377–382

37 Wright DA, Hetzel EW (1985) Use of RNA:DNA ratios as an indicator of nutritional stress in the American oyster Crassostrea virginica Mar Ecol Progr Ser 25:199–206

38 Ferron A, Leggett WC (1994) An appraisal of condition measures for marine fish larvae In: Blaxter JHS, Southward AJ (eds) Advances in marine biology Academic Press, London

39 Whyte JNC, Englaf JR, Carswell BL (1990) Biochemical position and energy reserves in Crassostrea gigas exposed to different levels of nutrition Aquaculture 90:157–172

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

Distribution patterns of five pleuronectid species

on the continental slope off the Pacific coast

of northern Honshu, Japan

Tsutomu Hattori• Takehiro Okuda•

Yoji Narimatsu•Masaki Ito

Received: 26 October 2009 / Accepted: 10 June 2010 / Published online: 3 August 2010

Ó The Japanese Society of Fisheries Science 2010

Abstract Information on the distributions of demersal

and benthic fishes is fundamental for stock assessment and

management Spatiotemporal changes in the distribution

patterns of five pleuronectid species (flathead flounder

Hippoglossoides dubius, Kamchatka flounder Atheresthes

evermanni, roughscale sole Clidoderma asperrimum, slime

flounder Microstomus achne, and Korean flounder

Glyptocephalus stelleri) off the Pacific coast of northern

Honshu, Japan, at depths of 150–900 m, were examined

using a generalized linear mixed model (GLMM)

Densi-ties of flathead and Korean flounder were highest in the

southernmost area, where the fish were small The body

lengths of both of these species increased from 2003 to

2008, suggesting that an abundant year class was recruited

in 2003 The density of Kamchatka flounder was highest in

the northern area In roughscale sole and slime flounder,

there were no distinctive annual and latitudinal trends in

the density distributions The density distribution of

Kor-ean flounder was bimodal; the peaks were at depths of 210

and 410 m The body length increased as the depth

increased from 150 to 410 m, and then decreased from

410 to 550 m Moreover, ‘‘bigger–deeper’’ trends were

observed in flathead, Kamchatka and slime flounder

Keywords Atheresthes evermanni 

Clidoderma asperrimum Distribution  GLMM 

Glyptocephalus stelleri Hippoglossoides dubius 

Microstomus achne

IntroductionMany demersal and benthic fishes are caught commercially

by Japanese offshore bottom trawlers on the continentalslope (200–1,000 m deep) in the Tohoku area, off thePacific coast of northern Honshu, Japan (Fig.1) For thisarea, stock assessments are carried out by the Japanesegovernment [1] for several offshore species (walleye pol-lock Theragra chalcogramma, Pacific cod Gadus macro-cephalus, threadfin hakeling Laemonema longipes, bighandthornyhead Sebastolobus macrochir, roughscale soleClidoderma asperrimum, and snow club Chionoecetesopilio) Pleuronectid species are one of the main compo-nents of the demersal fish community [2], and flatheadflounder Hippoglossoides dubius, Kamchatka flounderAtheresthes evermanni, roughscale sole, slime flounderMicrostomus achne, and Korean flounder Glyptocephalusstelleri are the dominant pleuronectid species on the con-tinental slope in the Tohoku area They are importantfishery resources; nevertheless, stock assessments and/ormanagement are not conducted regularly for these pleuro-nectid populations, except for roughscale sole

Kitagawa et al [3] reported that flathead flounder in theTohoku area has high densities at depths of 250–550 m andexhibits a ‘‘bigger–deeper’’ trend in October/November.Moreover, Yamada et al [4] suggested that adults migrateseasonally for spawning from depths of 400–600 m inOctober/December to 100–200 m in February However,there is only fragmented information on Kamchatka, slimeand Korean flounder, and roughscale sole; the main dis-tribution depth of roughscale sole is 700–900 m [5], andadults spawn at depths of 500–1,000 m [6], while adultslime flounder migrate from the waters off the Pacific coast

of Hokkaido Island in September to the Tohoku area forspawning in February [7]

T Hattori ( &)  T Okuda  Y Narimatsu  M Ito

Hachinohe Station, Tohoku National Fisheries Research

Institute, Fisheries Research Agency, Hachinohe,

Aomori 031-0841, Japan

e-mail: hmadara@affrc.go.jp

Fish Sci (2010) 76:747–754

DOI 10.1007/s12562-010-0269-8

Information on the distributions of demersal and benthic

fishes is fundamental for stock assessment and

manage-ment For example, the ‘‘bigger–deeper’’ phenomenon has

been observed in many demersal fishes [8 11], and data on

spatial distribution patterns provides useful information for

stock management Therefore, there is a strong need to

elucidate the distribution patterns of Kamchatka flounder,

roughscale sole, slime flounder and Korean flounder Our

objectives were to examine spatiotemporal changes in the

distribution patterns of five pleuronectid species, including

flathead flounder, in the Tohoku area

Materials and methods

Trawl sampling and oceanographic observations

Bottom otter trawl surveys were carried out at depths of

150–900 m in the Tohoku area (Fig.1) Fish samples were

collected from bottom trawls of the R/V Wakataka Maru

of the Fisheries Research Agency, Japan, primarily in

October/November, over a six-year period from 2003 to

2008 Transects for the trawl surveys were set up along

eight latitude lines: 40°500N (Line A), 40°200N (Line B),

39°400N (Line C), 39°000N (Line D), 38°200N (Line E),

37°400N (Line F), 37°000N (Line G), and 36°200N (Line

H) Sampling stations were set at depths of 150, 210, 250,

310, 350, 410, 450, 510, 550, 650, 750 and 900 m alongthese transects, except for two stations set at 150 m depth(Lines C and D)

The bottom trawl had an overall length of 44.1 m and amouth width of 5.4 m; it was only towed during the day.The net had a mesh size of 50 mm (stretched) and a covernet consisting of 8 mm square mesh at the codend All towswere made at an average ship speed of 2.5–3.5 knots, andthe trawling time was about 30 min The arrival anddeparture of the net on the bottom, and the horizontal andvertical openings of the net were measured using the Otterand Net Recorder system (Furuno Electric Co., Nish-inomiya, Hyogo, Japan) The positions at the start and end

of each trawl were recorded using GPS The numbers andweights of the five pleuronectid species were measured.Body length (standard length: BL) was measured to thenearest 1 mm onboard the vessel Density (fish/km2) wasestimated as the number of fish divided by the swept area

of the trawl Bottom water temperatures and salinities wererecorded at all stations using SBE 19 CTD (Sea-BirdElectronics, Bellevue, WA, USA)

‘‘depth’’) were incorporated into each model after beingtransformed to binary variables For these models, weexamined the influence of ‘‘year’’, ‘‘line’’, and ‘‘depth’’ onthe basis of ‘‘2003’’, ‘‘Line A’’, and ‘‘150 m’’, respectively

We also included ‘‘bottom water temperature’’ and

‘‘salinity’’ among the explanatory variables, because thedistribution patterns of the fish could be affected byoceanographic conditions The tolerance values for fourpredictor variables (‘‘latitude’’, ‘‘depth’’, ‘‘water tempera-ture’’, and ‘‘salinity’’) indicated that there was no signifi-cant correlation among the predictor variables (tolerancevalues: 0.474–0.950)

Spatiotemporal changes in the density distributions wereidentified using the following model with a Poissondistribution:

lnðNjtÞ ¼ a0þ a1Ytþ a2Ljþ a3Djþ a4Tjtþ a5Sjt

where Njt is the number of fish caught at stationjin yeart(Yt), and Ljand Djare the line and depth, respectively, atstationj Tjt and Sjt are the bottom water temperature andsalinity, respectively, at stationj in Yt In this model, rjtisthe random effects at stationj in Yt The ak (k = 0–5)are estimated parameters of the linear predictors The

A B C D E F G H

Fig 1 Line transects (A–H) for bottom trawls and oceanographic

observations at depths of 150, 210, 250, 310, 350, 410, 450, 510, 550,

650, 750 and 900 m in the Tohoku area, off the Pacific coast of

northern Honshu, Japan

logarithm of the swept area, ln(Ajt), enters the linear

pre-dictor as an offset term, which normalizes the number of

fish caught per swept area (fish/km2)

Spatiotemporal changes in the body length distributions

were identified using the following model with a Gaussian

0 8 16

0 8 16

0 8 16

0 8 16

0 8 16

0 6

0 6 12

0 6 12

0 6 12

0 6 12

0 0.5

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

Slime flounder Korean flounder

0 2 4

0 2 4

0 2 4

0 2 4

0 2 4

0 2 4

0 6

0 6 12

0 6 12

0 6 12

0 6 12

Fig 2 Mean densities by depth

of flathead flounder, Kamchatka

flounder, roughscale sole, slime

flounder, and Korean flounder

Slime flounder Korean flounder

0 0.1 0.2

0 0.5 1

0 1

0 1 2

0 1 2

0 1 2

0 1 2

0 1 2

0 1 2

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

0 0.1 0.2

0 0.1 0.2

0 0.1 0.2

0 0.1 0.2

0 0.1 0.2

0 0.1 0.2

0 0.2

0 0.2 0.4

0 0.2 0.4

0 0.2 0.4

0 0.2 0.4

0 0.2 0.4

0 0.2 0.4

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

Fig 3 Length–frequency

distributions of the five

pleuronectid species in the

Tohoku area The data were

summed for each species caught

in each transect

(i.e., the error term) associated with the ith individual

caught at stationj in Yt The bk (k = 0–5) are estimated

parameters of the linear predictors

Model selection was conducted to identify the most

parsimonious model by Akaike’s information criterion

(AIC) Analyses were carried out using the statistical

software R, version 2.8.0 (http://cran.r-project.org/) with

the package ‘‘lme4.’’

Results

Density distributions

The mean densities by depth and the length–frequency

distributions in each transect of the five pleuronectid

species are shown in Figs.2 and3, respectively The tiotemporal variations in the density distributions, asderived from the results of the GLMM, are shown inTable1 For flathead flounder, the density was highest atthe southernmost latitude transect line, Line H (Table1).The density increased with increasing depth from 150 to

spa-550 m, and then decreased from spa-550 to 750 m In theGLMM analyses, an annual change in density was foundonly in Kamchatka flounder, which was abundant in2006–2007 The density of this species was highest in thenorthern area between Line A and Line D The distributiondepth ranged from 150 to 650 m, while the high-densityarea was limited to shallow depths of 150–210 m Forroughscale sole, a total of 108 individuals were collectedfrom depths of 150–900 m, though the number of sampleswas relatively small for the analysis Only oceanographic

Table 1 Results of the generalized linear mixed model (GLMM) analyses for the density distributions of five pleuronectid species

Variables Flathead flounder Kamchatka flounder Roughscale sole Slime flounder Korean flounder

Coefficient SE Coefficient SE Coefficient SE Coefficient SE Coefficient SE

n = 316, AIC = 1,082 n = 110, AIC = 224.9 n = 73, AIC = 57.55 n = 81, AIC = 169.8 n = 215, AIC = 612.3 – indicates an explanatory variable that was excluded from the most parsimonious model The influences of ‘‘year’’, ‘‘line’’, and ‘‘depth’’ were examined by categorical predictors transformed to binary variables For ‘‘year’’, we assessed the influence from 2004 to 2008 on the basis of

2003 For ‘‘line’’, we assessed the influence from B to H on the basis of A For ‘‘depth’’, we assessed the influence from 210 m to 900 m on the basis of 150 m The zero-valued coefficients of year (2003), line (A) and depth (150 m) are not shown in this table

variables were selected by AIC; for roughscale sole,

den-sity was negatively affected by temperature and positively

affected by salinity The density of slime flounder was

highest at Lines B, D and G, and at the shallowest depth of

150 m The distribution depth of this flounder was the

narrowest (150–350 m) among the five pleuronectid

spe-cies As with flathead flounder, the density of Korean

flounder was highest at the most southern latitude line

(Line H) The results of the GLMM analysis indicated that

the density distribution had two peaks at depths of 210 and

410 m, respectively In contrast to that of roughscale sole,density was positively affected by temperature and nega-tively affected by salinity

0.5

0 1 1.5

0.5

0 1 1.5

0.5

0 1 1.5

0.5

0 1 1.5

0.5

0 1 1.5

0.5

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

0 0.1 0.2

0 0.1 0.2

0 0.1 0.2

0 0.1 0.2

0 0.1 0.2

0 0.1 0.2

0 0.1 0.2

0 0.1 0.2

0 0.1 0.2

0 0.1 0.2

0 0.1 0.2

0 0.1 0.2

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

Fig 4 Length–frequency

distributions of the five

pleuronectid species in the

Tohoku area The data were

summed for each species caught

Slime flounder Korean flounder

150m 210m 250m 310m 350m 410m 450m 510m 550m 650m 750m 900m

no catch

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

no catch

no catch

150m 210m 250m 310m 350m 410m 450m 510m 550m 650m 750m 900m

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

0 0.1 0.2

no catch

150m 210m 250m 310m 350m 410m 450m 510m 550m 650m 750m 900m

0 0.1 0.2

0 0.1 0.2

0 0.1 0.2

0 0.1 0.2

0 0.1 0.2

0 0.1 0.2

0 0.1 0.2

0 0.1 0.2

0 0.1 0.2

0 0.1 0.2

0 0.1 0.2

0 0.2 0.4

310m 350m 410m 450m 510m 550m 650m 750m 900m

0 0.2 0.4

0 0.2 0.4

0 0.2 0.4

0 0.2 0.4

0 0.2 0.4

0 0.2 0.4

0 0.2 0.4

0 0.2 0.4

0 0.2 0.4

0 0.2 0.4

0 0.2 0.4

0 0.5 1

no catch

no catch

no catch

150m 210m 250m 310m 350m 410m 450m 510m 550m 650m 750m 900m

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

0 0.5 1

Fig 5 Length–frequency

distributions of the five

pleuronectid species in the

Tohoku area The data were

summed for each species caught

at each depth

each year, and by depth, respectively Spatiotemporal

variations in the body length distributions, as derived from

the results of the GLMM, are shown in Table2 In the

GLMM analysis for flathead flounder, body length

increased from 2003 to 2008, and was smallest at Line H,

where the density was highest Body length increased from

a depth of 150 to 410 m, and remained relatively constant

from a depth of 410 to 750 m Thus, a ‘‘bigger–deeper’’

phenomenon was observed in the size–depth relationship of

flathead flounder Fish size was negatively affected by

temperature and salinity

In Kamchatka flounder, the body length increased from

a depth of 150 to 650 m, showing a ‘‘bigger–deeper’’ trend

in the Tohoku area Fish size was negatively affected by

temperature

In roughscale sole, small fish were distributed in shallowdepths from 150 to 210 m, though this species was dis-tributed over a wide depth range Body length was posi-tively affected by salinity, but not by temperature

In the GLMM analysis, the body length of slimeflounder was largest in 2005 and 2008 in the northern tocentral area (Lines A–E) A ‘‘bigger–deeper’’ trend wasfound at depths of 150–350 m The body length was pos-itively affected by temperature but negatively affected bysalinity Unlike the other species, the size of slime flounderwas positively affected by temperature

In Korean flounder, body length increased between 2003and 2008, and body length was smallest at Line H, wheredensity was highest Thus, the distribution pattern ofKorean flounder was similar to that of flathead flounder

Table 2 Results of the generalized linear mixed model (GLMM) analyses for the body length distributions of five pleuronectid species Variables Flathead flounder Kamchatka flounder Roughscale sole Slime flounder Korean flounder

Coefficient SE Coefficient SE Coefficient SE Coefficient SE Coefficient SE

n = 6,191, AIC = 65,624

n = 704, AIC = 7,382

n = 108, AIC = 1,225

n = 501, AIC = 5,241

n = 2,530, AIC = 25,487 – indicates an explanatory variable that was excluded from the most parsimonious model The influences of ‘‘year’’, ‘‘line’’, and ‘‘depth’’ were examined by categorical predictors transformed to binary variables For ‘‘year’’, we assessed the influence from 2004 to 2008 on the basis of

2003 For ‘‘line’’, we assessed the influence from B to H on the basis of A For ‘‘depth’’, we assessed the influence from 210 m to 900 m on the basis of 150 m The zero-valued coefficients of year (2003), line (A) and depth (150 m) are not shown in this table

Body length increased from a depth of 150 to 410 m, and

then decreased from 410 to 550 m The results of GLMM

analysis indicated a dome-shaped size–depth relationship

As with flathead and Kamchatka flounder, body length of

Korean flounder was negatively affected by temperature

Discussion

GLMM is an effective technique for analyzing distribution

patterns because it provides a comprehensive analysis of

the effects of selected predictor variables on fish

distribu-tion patterns We summarized the distribudistribu-tion patterns of

five pleuronectid species using the results of GLMM

analyses (Table3)

The densities of flathead and Korean flounder were

highest in the southernmost latitude line transect (Line H),

where the fish were the smallest (Table3) The body

lengths of these species increased from 2003 to 2008,

suggesting that an abundant year class was recruited in

2003; the fish grew from 2003 to 2008 In Fig.4, B15-cm

BL fish of both species were abundant in 2003 The

co-occurrence of the abundant year classes suggests that the

major factor in the high year-class occurrence of flathead

flounder should be similar to that of Korean flounder in theTohoku area As with roughscale sole, flathead flounderwas distributed over a wide depth range, and they weremost abundant at a depth of 550 m (Table 3) The positivesize–depth correlation for this species supports the findings

of Kitagawa et al [3] The density of Kamchatka flounderwas highest in the northern area (Table 3) This species is aboreal species [12, 13]; the main distribution seems toextend from the northern Tohoku area, northward to thewaters off Hokkaido Island This species was most abun-dant in a narrow depth range of 150–210 m, and it showed

a ‘‘bigger–deeper’’ trend For roughscale sole, there was nodistinctive ‘‘bigger–deeper’’ trend; nevertheless, bodylength was smallest at depths of 150–210 m (Table3).The body length of slime flounder was largest at LinesA–E, and there were no distinctive annual and latitudinaltrends in the density distribution (Table3) Slime floundermigrates southward from the waters off Hokkaido Island tothe Tohoku area [7] Therefore, the density distributionwould be affected by the annual migration of adult fish tothe Tohoku area The highest density was observed at adepth of 150 m, and this species showed a ‘‘bigger–dee-per’’ trend However, body length was positively affected

by water temperature, and bottom water temperature

Table 3 Comparison of the distribution patterns of the five pleuronectid species, as derived from the results of generalized linear mixed model (GLMM) analyses

Flathead flounder Kamchatka

flounder

Roughscale sole Slime flounder Korean flounder

Model for density

Latitudinal distribution

range

Change with latitude High (Line H) High (Lines A–D) – High (Lines B, D, G) High (Line H) Range of distribution depth 150–750 m 150–650 m 150–250, 350–900 m 150–350 m 150–550 m

Change with depth Increase

(150–550 m) Decrease (550–750 m)

Abundant (150–210 m)

– –

Abundant (150 m)

Abundant (210 and 410 m)

Model for body length

Annual change Increase (2003–2008) Small (2006, 2008) Large (2007) Large (2005, 2008) Increase (2003–2008) Change with latitude Small (Line H) Large (Lines A–D) – Large (Lines A–E) Small (Line H) Change with depth Bigger–deeper Bigger–deeper Small (150–210 m)

Not clear (250–900 m)

Bigger–deeper Increase

(150–410 m) Decrease (410–550 m)

– indicates that the obvious influence is not observed in the most parsimonious model

decreases with increasing depth from 150 to 350 m [14].

This apparent contradiction in the results was probably

caused by wide variation in the annual migration of adult

fish [7]

In Korean flounder, the density distribution was

bimo-dal, with peaks at depths of 210 and 410 m (Table3) The

body length increased from a depth of 150 to 410 m, and

then decreased from 410 to 550 m These results show that

most of the small fish were broadly distributed at around

210 m depth, and the density of large fish was highest at a

depth of 410 m Density was positively affected by

tem-perature; the bottom water temperature is stable at depths

of over 350 m in October/November in the Tohoku area

[14] The results of the GLMM seemed to be affected by

the distribution of small fish The distribution pattern of

large fish could not be explained by the variation in

tem-perature There are no ecological studies of the feeding

habits of Korean flounder in the Tohoku area; however,

intraspecific competition often results in high-quality

hab-itats being occupied by dominant individuals, with

subor-dinates relegated to lower quality habitats [15] In Korean

flounder, the density of large fish was highest at a depth of

410 m Hence, it is probable that the sea floor at 410 m

depth is a high-quality habitat for this species

In this study, ‘‘bigger–deeper’’ trends were observed in

flathead, Kamchatka and slime flounder (Table3)

More-over, small flathead flounder was abundant in 2003

(Fig.4), and its body length increased from 2003 to 2008

(Fig.4, Table2), suggesting that this species migrates

from shallow to deep habitats as it grows in the Tohoku

area

In the Tohoku area, Nihira et al [16] reported that the

catch per effort unit (CPUE) of offshore bottom trawlers

was low for flathead and slime flounder in the 1980s, but

that it increased from 1994–1996 to 2001 because of

recruitment success in the 1990s In such cases as this, the

catch of small fish must be controlled to advance the

recovery of the resources In addition, the distribution of

adult fish provides useful information for establishing a

protected area for spawning fish Our results have

impor-tant implications for the management of pleuronectid

species on the continental slope in the Tohoku area

Moreover, long-term monitoring of demersal and benthic

fishes based on trawl surveys should be continued; this will

further our knowledge of the fundamental life history traits

of pleuronectid species and increase understanding of

changes in fishery stock structures for future scientific work

and policy decision making

Acknowledgments We thank the crew of the R/V Wakataka Maru

for help in collecting samples This research was funded by the

Fisheries Agency of Japan The study is contribution number B122

from the Tohoku National Fisheries Research Institute, Fisheries Research Agency, Japan.

References

1 Fisheries Agency and Fisheries Research Agency of Japan (2009) Marine fisheries stock assessment and evaluation for Japanese waters (fiscal year 2008/2009) Fisheries Agency and Fisheries Research Agency of Japan, Tokyo, pp 1–1553 (in Japanese)

2 Fujiwara K (2005) The study of biological product structure of demersal fishes on the continental slope (Ph.D thesis) Tohoku University, Sendai (in Japanese)

3 Kitagawa D, Katayama S, Fujiwara K (2004) Distribution and growth of flathead flounder Hippoglossoides dubius, off Tohoku area Bull Jpn Soc Fish Oceanogr 68:151–157 (in Japanese with English abstract)

4 Yamada M, Ueda Y, Hattori T, Yamanobe A, Yoshida T (2006) Biology and stock assessment of flathead flounder Hippoglosso- ides dubius in the Joban waters Bull Fukushima Pref Fish Exp Stat 13:19–36 (in Japanese)

5 Saeki M (2001) Ecology and stock management of roughscale sole Clidoderma asperrimum (Temminck et Schlegel) caught in the water off Sanriku and Joban Miyagi Pref Rep Fish Sci 1:93–102 (in Japanese)

6 Hattori T, Ueda Y, Narimatsu Y, Ito M (2008) Distributional changes of roughscale sole (Clidoderma asperrimum) off the Pacific coast of northern Honshu, Japan Bull Jpn Soc Fish Oceanogr 72:14–21 (in Japanese with English abstract)

7 Ishito Y (1972) The characteristics of distribution and migration pattern of the slime-flounder, Microstomus achne (Jordan

&Starks), adjusted by the life of their young and adult stages in the north-eastern sea area of Japan Bull Tohoku Reg Fish Res Lab 32:23–46 (in Japanese with English abstract)

8 Macpherson E, Duarte C (1991) Bathymetric trends in demersal fish size: is there a general relationship? Mar Ecol Prog Ser 71:103–112

9 Haedrich RL, Rowe GT (1977) Megafauna biomass in the deep sea Nature 269:141–142

10 Haedrich RL, Rowe GT, Polloni PT (1980) The megabenthic fauna in the deep-sea south of New England, USA Mar Biol 57:165–179

11 Polloni PT, Haedrich RL, Rowe GT, Clifford CH (1979) The size-depth relationship in deep ocean animals Int Revue Ges Hydrobiol 64:39–46

12 Tokranov AM, Orlov AM (2003) On the distribution and biology

of roughscale sole Clidoderma asperrimum (Temminck et Schlegel, 1846) in the Pacific waters off the northern Kuril Islands and southeastern Kamchatka Bull Sea Fish Inst 2:67–80

13 Orlov AM, Tokranov AM (2007) Distribution and some ical features of four poorly studied deep benthic flatfishes (Pleuronectiformes: Pleuronectidae) in the northwestern Pacific Ocean Raffles Bull Zool Suppl 14:221–235

biolog-14 Hattori T, Narimatsu Y, Ito M, Ueda Y, Fujiwara K, Kitagawa D (2007) Growth changes in bighand thornyhead Sebastolobus macrochir off the Pacific coast of northern Honshu, Japan Fish Sci 73:341–347

15 Burns CE (2005) Behavioral ecology of disturbed landscapes: the response of territorial animals to relocation Behav Ecol 16:898–905

16 Nihira A, Suno N, Takahashi M (2003) Regime shift in ground fishes in the Sanriku and Joban waters areas Kaiyo Mon 35:107–116 (in Japanese)

Received: 17 November 2009 / Accepted: 10 June 2010 / Published online: 15 July 2010

Ó The Japanese Society of Fisheries Science 2010

Abstract The production of transparent exopolymer

particles (TEP) by four diatoms, Coscinodiscus granii,

Eucampia zodiacus, Rhizosolenia setigera, and Skeletonema

sp., was examined Most of the TEP in C granii (74% of the

maximum) were produced during the growth phase In

contrast, most of the TEP in E zodiacus (73%), R setigera

(74%), and Skeletonema sp (70%) were produced during the

stationary and declining phases The C granii TEP

pro-duction rate was highest in the growth phase, whereas those

in E zodiacus, R setigera, and Skeletonema sp were highest

in the stationary–decline phase The TEP concentrations per

cell and the cell volume of C granii were 34.97 ± 4.114

(mean ± SD) ng Xeq./cell (xanthan gum equivalents per

cell) and 341.6 ± 56.33 fg Xeq./lm3, and were 23.01 and

4.32 times higher than the values obtained from the other

three diatoms, respectively The results suggest that the

mechanisms of TEP production differ with growth stage and

diatom species Therefore, it is likely that the differences in

TEP production among the diatom species influence the

complexity of TEP dynamics in aquatic environments

Keywords Diatom Growth stage  Production 

Transparent exopolymer particle (TEP)

Introduction

Organic matter, the product of photosynthesis, is the basis

of all ecosystems Much of the suspended organic matter in

the sea exists as macroscopic aggregates [0.5 mm indiameter, called marine snow, which is formed from phyto-plankton, fecal pellets, zooplankton and detritus Theseaggregates are enriched in carbon and nitrogen, and con-centrated communities of microorganisms are found in them

at abundances that are several orders of magnitude greaterthan their abundances when living free in the seawater [1,2]

In addition, they maintain unique chemical environments andare the major form in which particulate organic matter sinks

to the bottom of the sea [2] Marine snow formation in the sea

is closely connected with the presence of transparent polymer particles (TEP) in the environment

exo-TEP are acidic polysaccharides made visible by Alcianblue staining They are spontaneously formed from TEPprecursor substances—such as colloidal and dissolvedpolysaccharides [3 6]—by the processes of gelation,annealing, and aggregation, depending on the environ-mental conditions (such as turbulence, temperature andsalinity) and the precursors present These particles are alsoformed directly from the particulate material released fromorganisms [7,8] The formation of TEP has been studiedextensively, and phytoplankton—especially diatoms—have been found to be the main producers of TEP anddissolved TEP precursor substances Therefore, high con-centrations of TEP are usually associated with phyto-plankton blooms (especially with blooms dominated bydiatoms [9 13]), although not every diatom bloom causeshigh TEP concentrations [14, 15] TEP are utilized as asubstrates and microhabitats by bacteria; they also alter theinteractions between bacteria and their environment, play amajor role in transforming organic substances, influencethe sinking carbon flux from the surface layers of theocean, and serve as a food source for zooplankton [16].Therefore, it is increasingly being recognized that TEP are

an important component of carbon recycling in aquatic

T Fukao ( &)  K Kimoto  Y Kotani

Seikai National Fisheries Research Institute, Fisheries Research

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

Japan

e-mail: fukao@water.ocn.ne.jp

Fish Sci (2010) 76:755–760

DOI 10.1007/s12562-010-0265-z

systems [4,5,8,17,18] Recently, the appearance of

large-scale organic aggregates called marine mucilage [12,13] or

NUTA [19], which is primarily composed of TEP, and its

associated effects on fishing have been observed in Ariake

Sound, Kyushu Island, Japan Therefore, it is necessary to

clarify the mechanism of formation of organic aggregates

in Ariake Sound Previous studies [12, 13,20] have

sug-gested that the diatom species such as Chaetoceros,

Coscinodiscus, Eucampia, Rhizosolenia and Skeletonema

were the dominant species in the large-scale organic

aggregates in Ariake Sound, so these diatoms were

suspected to be the causative agents of the aggregates

However, the occurrence of these organic aggregates does

not always coincide with the peak in phytoplankton

blooms; indeed, it is postponed towards the senescent

phase of the blooms [20] Hence, we speculate that the time

lag between the occurrence of organic aggregates and the

peak in the phytoplankton blooms is due to differences in

TEP production between the growth stages of diatoms

However, little is known about the production of TEP by

the diatoms that caused the blooms in Ariake Sound

In the present study, we evaluated TEP production by

four diatom species, Coscinodiscus granii, Eucampia

zo-diacus, Rhizosolenia setigera, and Skeletonema sp., which

formed blooms in Ariake Sound The experiment was

conducted with axenic batch cultures of diatoms to test the

hypothesis that TEP production differs among species and

depends on the physiological state of the cells

Materials and methods

Establishment of axenic cultures

Four species of diatoms, C granii, E zodiacus, R setigera,

and Skeletonema sp., were used in this study All diatoms

were isolated from seawater samples in Ariake Sound,

Japan, during 2007 The cultures were washed repeatedly

using the micropipette isolation method Axenic checks

were made using STP medium [21] Further checks were

made by 4,6-diamidino-2-phenylindole (DAPI) staining

and epifluorescence microscopy

Stock cultures were grown at 20°C under a 14:10 h

light:dark photocycle Light was provided by a cool-white

fluorescent lamp (150–200 lmol photons/m2/s) The

cul-tures were maintained in f/2 medium [22]

Growth experiments

Stock cultures of C granii, E zodiacus, R setigera, and

Skeletonema sp were inoculated into 4-L flasks containing

3 L of modified f/2 medium The initial concentrations of

NaNO3, NaH2PO42H2O, Na2SiO39H2O and Tris were

0.88, 0.06, 0.5, and 4.1 mM, respectively The pH of themedium was adjusted to 7.8–7.9 The initial cell numbers

of C granii, E zodiacus, R setigera, and Skeletonema sp.were ca 5, 100, 150, and 2500 cells/mL, respectively Theflasks were incubated in triplicate under the same condi-tions as used for the stock cultures The cultures wereshaken once per day During the incubation period, celldensity was counted directly using an inverted lightmicroscope (Nikon TE300, Tokyo, Japan), and the TEPconcentrations were determined Cell volumes weredetermined according to the method of Miyai et al [23].Determination of TEP concentration

We estimated TEP concentrations using the method ofPassow and Alldredge [9] Culture samples (1–20 mL)were filtered using a 0.4 lm Isopore membrane filter(Millipore) Particles retained on the filters were stainedwith 0.02% Alcian blue (8GX, Sigma) in 0.06% aceticacid The filters were then soaked in 80% sulfuric acid for

3 h, and the absorbance was read at 787 nm using aspectrophotometer (Shimadzu, UV-1600) A soaking time

of 3 h was found to be ideal; extended soaking did notincrease absorbance All filters were prepared at least induplicate Acidic polysaccharide xanthan gum was used as

a standard, and TEP concentrations were expressed interms of xanthan gum equivalents (lg Xeq./L)

ResultsThe cell density of C granii increased from 4.77 ± 1.71(mean ± SD) to 3.81 9 102± 2.11 9 10 cells/mL on day

10 (Fig 1) We found that the TEP concentration increasedwith increasing cell density and increased slightly in thestationary–decline (II) phase (day 12) The concentration ofTEP was maximal on day 22 (13.2 ± 1.25 mg Xeq./L).Thus, most of the TEP (74% of the maximum) produced by

C granii were formed during the growth (I) phase.The cell density of E zodiacus increased from 1.36 9

103cells/mL on day 9 and from 2.65 9 103± 2.54 9 10

to 1.26 9 106± 9.29 9 104cells/mL on day 11, tively (Fig.1) The TEP concentrations in both speciesincreased slowly or did not vary during the growth (I)phase In contrast, during the stationary–decline (II) phase,

respec-the TEP concentration increased constantly The

concen-trations of R setigera and Skeletonema sp were maximal

on day 50 (23.5 ± 5.19 and 22.4 ± 1.70 mg Xeq./L,

respectively) Most of the TEP produced by E zodiacus

(73% of the maximum), R setigera (74%), and

Skeleto-nema sp (70%) were formed during the stationary–decline

(II) phase

Discussion

In the present experiments, we found that the production of

TEP depended on growth stage and differed among the

diatom species (Figs.1,3)

The results for TEP production by C granii are in

agreement with those of Abdullahi et al [24] and

Under-wood et al [25], who reported enhanced production of

extracellular polymeric substances (EPS), including TEP,

in the diatoms Phaeodactylum tricornutum and

Cylindrot-heca closterium during the growth phase Microscopic

observations of Alcian blue-stained slides (following the

method described by Fukao et al [12]) confirmed the

presence of abundant TEP; the majority of the TEP in

C granii could have formed from dissolved and/or particulate

polysaccharides released mainly by vigorous cells during

the growth phase (Fig.2a) On the other hand, high TEP

productions of E zodiacus, R setigera and Skeletonema sp

were observed during the stationary–decline phase It

has been shown by Kiørboe and Hansen [26] and Engel

[27] that the amount of TEP generated by Skeletonema

costatum during the growth phase is negligible, and TEP

are only generated during late senescence The particles

were released by cell lysis or breakage in E zodiacus,

R setigera and Skeletonema sp deeply stained with Alcianblue (Fig.2b–d) Accordingly, the majority of the TEP inthese three diatoms were produced by dead cells duringthe stationary–decline phase Many microorganismsrelease TEP by exudation, exocytosis, or autocatalyticprogrammed cell death [7, 28] Berman-Frank et al [28]reported increased TEP production in cultures of thecyanobacterium Trichodesmium after programmed celldeath in response to nutrient and/or oxidative stress.Therefore, the results from E zodiacus, R setigera, andSkeletonema sp provide further evidence that programmedcell death is a key trigger for TEP production TEP con-centrations in the cultures of R setigera and Skeletonema

sp increased during the disappearance (III) phase (Fig.1).TEP are abiotically formed from colloidal and dissolvedpolysaccharides released mainly by phytoplankton [3 5].The increase in TEP during the disappearance (III) phase isprobably caused by abiotic formation from the dissolvedpolysaccharides released by R setigera and Skeletonema

sp cells

Based on our results, the rate of TEP production for each

of the four diatoms during each phase was calculated asfollows:

TEP production rate mg Xeq:=L=dayð Þ

0

5 x 1051.0 x 1061.5 x 106

0

5 x 1031.0 x 1041.5 x 1042.0 x 104

0

1.0 x 1042.0 x 1043.0 x 1044.0 x 104

Eucampia zodiacus

TEP Cell density

0 5 10 15 20 25 30

Fig 1 Changes in cell density

and transparent exopolymer

particle (TEP) concentration in

the cultures of four diatom

species growing in f/2 medium.

Bars represent standard

deviations (n = 3) I, Growth

phase; II, Stationary–decline

phase; III, Disappearance phase

aggregates formed by C granii coincided with the peak in

the blooms [12,13], whereas organic aggregates dominated

by S costatum and Rhizosoplenia were observed at the end

phase of the blooms [7, 26, 27] Therefore, the different

time lags between the occurrences of the various organic

aggregates and the peak in the blooms can be explained by

noting the differences in TEP production among species and

among growth stages in the present study The TEP

con-centrations per cell of C granii, E zodiacus, R setigera,

and Skeletonema sp were 34.97 ± 4.114, 0.52 ± 0.044,

1.52 ± 4.186, and 0.02 ± 0.003 ng Xeq./cell, respectively,

so the value for C granii was 23.01 times larger than that

obtained from any other diatom (Table1) However, there

are considerable differences in cell volume among these

diatom species Thus, TEP production must be evaluated

and compared on the basis of cell volume The TEP centrations per cell volume of C granii, E zodiacus,

con-R setigera, and Skeletonema sp were 341.6 ± 56.33,79.0 ± 4.94, 74.7 ± 8.63, and 68.3 ± 3.28 fg Xeq./lm3,respectively (Table1), so the value for C granii was 4.32times larger than that for any other diatom Blooms due toCoscinodiscus spp sometimes cause larger and more vis-cous organic aggregates than the blooms of other diatoms,and affect fishing activities [29] Field observations andthese results indicate that C granii is a major TEP producerthat easily causes organic aggregates in coastal waters [13].Previous experimental studies have demonstrated theformation of TEP from dissolved polysaccharides [3 6].TEP precursors consist of acidic sulfated polysaccha-rides enriched in fucose, rhamnose and galactose [30]

Fig 2 Alcian blue staining of

transparent exopolymer

particles (TEP) released by the

four diatoms a Coscinodiscus

granii during the growth phase;

b Eucampia zodiacus during the

decline phase; c Rhizosolenia

setigera during the decline

phase; and d Skeletonema sp.

during the decline phase

/ day)

0.0 0.5 1.0 1.5

1.5

Eucampia zodiacus

0.0 0.5 1.0

1.5

Rhizosolenia setigera

0.0 0.5 1.0

1.5

Skeletonema sp.

Fig 3 Production rates of

transparent exopolymer

particles (TEP) by the four

diatom species in the different

phases I, Growth phase;

II, Stationary–decline phase;

III, Disappearance phase.

N.D., no data