POPULATION DYNAMICS OF THE MAIN PELAGIC SPECIES EXPLOITED lN THE JAVA SEA : BIOLOGICAL PARAMETERS ESTTMATES

several sets of length frequency data (1991 to '199s) of the six species were used to estrmate the groMh parameters and to discuss the results in relation to possible influence of data structure on the parimeter estimates The.von Bertalanfry groMh parameters were well estimated using the Elefan routine and FordWalford method The existing oscillation pattern and the pseudo groMh progression curye were identifled. rr can be noticed that the esiimaled values do not represent the whole stocks. T-he emmigration of aOut nsnli is considered as the more imporlant factor determining the shape of oscillation curve-rather tnan tnat otihe nature somatic groMh pattern. l\4ore cohorts identifed staiing in the Java sea durinq anomalv vear consequently resutt a higher estimate values of K ot certain species. The fish stock available-for tne tiin6rLs mainly consists of young fshes, i.e. the average stzes of the main species are smafler than approximate length_at first mature and very few specimens of adult fish were not availabl€ in th6 fishing grorni'"il ;ith" year. Two groups of recruits were identified, the major and the minor ones entered ttre fiitrlries durino the south-east and northwest monsoon, r€spectively The'major recruits were not the otfspri"g;;";;;;d"d ;"y il" adult tishes stayed in this area in last year period.


INTRODUCTION
Biological study can be considered as the first stanza in understanding the population dynamics of the species.In the context of fishsries problems, these components of the dynamic system may not be analyzed separately.Nevertheless, seoarate analysis of the part of the system would be still valuable in explaining possible influenc6 of the ecological dimension on the biological parameter.
The parameter estimates of these comoonenl.,.e. growth, recruitment pattern and reoroduction.will be evaluated in order to find possible relation to ecological aspects and inieraction with biologioal dimensions.
Due to the need of groMh paramater estimates as input parameters for length based slock assessment models, several studies on the estimalion on growth of pelagic species in the Java Sea have been conducted since the beginning of 1970's. Sujastani (1974) ) estimated the growth parameters of Rastrelliger kanagurta and R. brachysoma in Karimata Strait and Nurhakim (1993; 1995) in the whole Java Sea for R.
kanagufta.Dwiponggo et at (19E6) reanalysed the old data including five pslagic species, with no inspection on lhe validity ot the data We may cite also Suwarso ef a/. (1995); Sadholomo & Aimaja (1985) for four among the main species, and Widodo (1988b) for 2 specios of Oecapterus.Most of those studies were based on one year period of observations and there was no evaluation on the pattern of estimales to be done.In lhis sludv.
several sel of length frequency data (1991 to 1 995) of the main six species caught by purse seine fishery were used in attempting to estimate the growth parameters as input data for the model of stock 6valuation.

Materials
This study was mainly based on the lenglh frequency measurement carried out in the years 1991 to 1995 and reproduction daia of 1992 to lnd.Fish Res.J. Vol.12 No.1 June-2006: 37-63 Length Composition Data Collection sampling scheme was definad within Pelfish proiecl for three main harbors, ie.Pekalongan, Juana, and Tegal (Potier & Sadhotomo, 1991) while general sampling procedure had been presentod before the project implemented (Boely et a,., 1990).The sampling was intended to gather the length frequency data and species composition by fishing zones and by purse seiners type.In this case, an assumDtion was done that catches 0r abundance be proportional to repartition of the fleet operating in the fishing zones.
In the fleld, 2 type of seiners, medium and big size of categories of boats were discriminated, since their operation pattern seemed io be different at the first three years of sampling activities.After 1994 the fishing ground of the two types tended to be similar, due to the increaso of the size 0f the new boat of medium category.But, because the usage of old ones was still existing, separation of the categories could be expected lo increese the accuracy of sampling.The nshing operation of the seiners landed in those harbors showed a similar oattern as well as theif distribution of the lenglh frequency of fish data (Potier & Sadhotomo, 1995).For this reason, the data collected from the largesl landing place (,.e.Pekalongan) would represent all harbours (Pekalongan, Tegal, and Juana).
For implementation ol ihe sampling in the field, four surveyors were employed by the Pelfish worked for collecting thesq data under supervision of a sciontific leam.For clarifying the explanetion about the sampling, detdils of the protobol were presented along the following paragraphs.

Length Composition Sampling
The sampling frame in this procedure was defined as the list of daily landings by soinors (,'.e. 2 frames, one for medium size category and the other for the big size one), whero sampling unit was the sample boet.The list consists of 1) the name of tha seiners landed one day before In the anernoon until early in the morning in the same day; 2) type of boat; 3) the fishing zones; 4) information on the dominant catch; and 5) the type of presgNation used (i,e.; ice or salt).Only the boats disembarking the fresh fish were taken as samplef other wise they are considered as out of the frame.lf these were more than five boats coming from the same fishing zones, then 2 samples were systematically taken from the list with interval of 4, other wise only one sample to be taken.The first sample was taken from the flrsi order of this daily list.The second sample would be the second order if the boat was coming from anolher fishing zones.Otherwise second samole would be the fifth order if all of the boals coming from the same fishing ground.lf the number of landings of the same zone w6re less than five, only one sample taken in a day.This procedure was applied separately io the big size end medium size of seiners.Detail of lhis procedure has been given by Potier & Sadhotomo (19e1).
In the next sleps, one or two sub samples (a basket of fresh fish that is c.a. 30 kg) were randomly taken from a fish hold ot the sampling boat.The specimens were sorted into specaes and counted by species.Then the measurement was done for 50 or more fishes of each species if the soecies consisled of more than 50 fishes, otheMise all of specimens would be measured The fork length measurement was done using measuring paper to 0.5 cm of accuracy.The data then to be transferred to a working sheet.A comoarison on the composition 0f the sample by fishing grounds on the first six month of sampling shows that the samples were well represented in term of the distribution of lhe Jishing zones prospected by the seiners.In this part, we did not perform any formal test to confirm the reDresgntation of the samPles.
Tho monthly l€ngth frequency data then were pooled by species and fishing ground.A simple m€thod of calculation for monthly summary was as follow.Lst x,, is the frequency of species i"', j is the i rh lenolh interval, N is number of fish measured of ine so.-ecies itn and equal to summation of x', denoted as: N,=E x,,, Let c, is number of fish of soecies i'n in a sub sample, lhen calculated frequency per sub sample by species would be fir=Xii.c/Ni This value then to be fllled in the working she6t.Monthly compiletion is done by summing fr by fishing ground, that is grx=I f,;, where k is index for fishing ground,

Reproduction
Ragular samplings were conducted in Pekalongan and Juwana from August 1991 to December 1995.The observations cover the l6ngth, weight of body and gonads, stage of maturity and adiposity, Until end of 1991 no weight measuremeni was done.Only the data from the Deriod 1992 to 1993 were used in this study.JO Subsamples were taken randomly from the fish holds 0f lhe seiners and immediately followed by sorting for separating the species.Special sampling were achieved several times for gathering mature fishes for microscopic description purposes.Specimens were dissecled for gonads stages observation.The gonads were removed, and gross examination of the gonads were carried out by weighting and observing their shape and color visually.All aspects related to the sample are recorded in the working sheet such as name of the vessel, fishing zone, and some remarks concerning with biological observation.

Growth
GroMh somatic parameters were estimated under the model of von Bertalanffy for entjre and cert€in parts of the data.Concernjng the coverage of the samples 0.e.fishing zones of the seineri), lhe usual restriction and assumo on are exerted on this study, that is the growlh estimates are only valid for the range of size or age being exploiled by serners or during thelr stay ln the Java sea and not of their life span.
Two methods of estimations were oerformed.in order to reveal a possible periodical changes of groMh parameters, as described in lhe followino paragraph.

Prog ression
The means of length classes were estimated using a kind of graphical method (Bhattacharya, 1967) and lhen were confirmed through maximum likelihood estimation.The main obiectives of the use of these methods are for deiomposing the polymodal frequency or a mixlure of normal distribulion and estimating their means.Hasselblad (1966) introduced the maximum likelihood estimalion using the log likelihood function for a multinomial distribution for estimating the parameter for a mixture of normal distribution.ln the FISAT package program, it is called as the Normsep (stands for normal separalion), This method usually give good perameter estimates, but the.estimalion procedure has a disadvanlage for testing lhe statistical significance of likelihood ratio test and for eslimaling the covariance matrix of the estimates of parameter (Fournier et at, .1990).In this study, the use of this method was merelv aimed for verifying the results of a preliminarv estimation and not intended to further inference analvsis, Estimations of the mean length for symmetrical unimodal distribution would be easily determined from its mode.First trial using a maximum likelihood based softwaie showed that the number of groups introduced at the starting point tended to be the same as the comoutation result.For this reason, rough estimation using other method (i.e. Battacharya method) prior to this method was aimed for having a more accurate estimates.The progression of lhe mean length of the same cohort could be traced by eyes or by following the grolvth line generated by elefan system.Then the growth parameler could be estimated for a single cohort 0nly by ignoring minor distribution that do not describe a clear progression.Only the samples contribute to mejor cohorts were used in regression.In this step, the pldtting of growlh increment of unequal lime interval were performed manually according to Gulland-Holt Plot (Gulland &. Holt, 1967).

Multiple Class Estimation
ELEFAN routine was performed to estimate groMh parameters.The growlh curves produced by this melhod maybe used to trace modal class progression, and t0 define appropriate samples for being used in the next step which would clearly describe a progression.This procedure has an ability to estimate objectively the grovvth parameters from multiple lenglh class represenled in the data.But, it is based on ad hoc estimation procedure (Fournier ef al, '1990), and its reliabitity could be doubtful in some cases and with certain data types.
This analysis works on restructured samples created by using certain standardized moving average, Restructurisation of the samples is a conversion of the ordinary lenglh frequency data to the restruclured ones which standardized lo a new value being able to explain the strength of its probabiliiy.In lhis procedure, frequency of each class length is expressed in score reflecting lhe strength of its probability.The maximum value would indicate lhe means of distribution or the mode.In a set of length frequency data, sum of all positive scores (they may correspond to several peaks o'f lhe polymodal distribution) is denotes as available sum of peak, and the sum of scores or points associated with those.peaks in a set of length freguency data that are passed by a given von Berlalanffy's groMh cuNe, is named as explained sum of peak.
GroMh parameters were computed iterativety with maximizing the explained sum of peak being passed through by the growth lines.Detailed of the original version of elefan I procedure has been given in Pauly & David (1981).In this study, a revised and enhanced version of Elefan I is used to estimated groMh parameters K and L.. Assumptions underlying this method are (Pauly & David, 1981 ; Pauly, 'l 985): the von Bertalanffy's growth formula describes the average growth of the siocks under investigation, and growth pattern in the population is similar from year to year' Due to the second assumption, a wide number of groMh curve may be applied on the same set of samples to result an optimum value of K and L.. lt means that samoles of some months could be considered as additional samples of the same months for the successive years.However, the algorithm of this method enabling the growth curve to follow lhe above scheme as the optimal criterium is matched, even for a single sample data set.ln order to avoid an unreasonable number of growth lines being produced by this method' the tample sets used were defined from the beginning of the recruitment until the end of growlh progression.The data sets wore composed of regular samples having clear modal progression By this way, growth lines would tend to pass the Deaks of the same cohort as the maxima value of explained sum of peak also tend to follow modal progressron.
Inaccurate estimate maybe resulted by a kind of seasonal oscillation of growth pattern that was possibly generated by biological or other ohenomeni.For this reason, the optional oscillation parameters integrated in the software were introduced during executing the program' Theoritical modification of von Berlelanffy equalion used in the elefan was done by exerting an amplitude of seasonal oscillation factor and a point of which the growth line is oscillating' in the ordinary formula.
, -' 14 ^ K(t-to)-(c k/2') s,n{2rrl-ts)h WP stands for winter point that is posiiion of oscillating growth line.In case of c=1, ihe growth rale is zero when the winter point is reached once a year, For C=0 the equation reverts to the original form of von Bertallanffy's growth formula,

Reproduction Maturity Scale
The scale of maturity was adapted from several studies with differences species and ecological areas, because of inavailibilty of intensive investigation carried out in the Java Sea.
Assumption underlying the determination of the maturity scales, for female fishes can be noted as follow; a.The species are total spawner categories.ll means that for which docytes are nps ano immediately shod in a very short time.
b.The morphologlcal changes of lhe ovaries are linoarly related to thb development of the ova Ths stege of maturity is reflected by the developmen'i of ova slze; ie. the size of ova thal readily:for spawning influence the relative size (in weighi and volume) of the ovaries.So that, the size of ovaries can be used as indicator of the matunty stage.
The different maturity stages of the gonad used in this study are sxpressed as follow: 1 .Stage 0: Undeterminable.
Gonad very small, usually white' and sex is not dotorminable by gross examinati0n 2. Stage l: lmmature.Ovaries: very small, shape oval elongated (carangids), slender (for non carangids species) iolor: white or transparent, ova not visible to naked eyes, but sex determinable by gross examination.
3. Stage ll: Developed virgin li iJincluded the Resting stage.Some clump of the eggs are visible to the naked eye, The strapJot ovaries: elongated, length ol ova.ry 2, to i cm for Decapterus spp.and occupy 1/3 of body capaciiy.

Stage lll:
Early maturing or developing.
The ovaries are enlarged and occupy 2/3 of body cavity, aMomen seems normal, ova are opaque.
5. Stage lv:Late maturing or rhature.ThJ ovary ocoupies almost the whole body cavity, somelime the abdomen distended' ova are still opaque, but individual ova visible to the naked eyes.
Peritoneal cavity filled with ovaries, abdomen distended, some translucent eggs visible lhrough the ovary wall and some of the translucent eggs are easily dislodged from the follicles or loose in lumen of ovary.In facl these criteria may result a bias generated by subjectivity as function of observetions, more over these macroscopic observations were done by three persons in different landing place.In order to avoid this subjectivity, a common index figuring the percentage of gonads size is performed in lhis study.This index, denoted as gonado sometic index would be able to indicate a primary reproductive effort in fish and to rEflect the development of the ova, ie.slage of maturily.lt is calculated as percenlage of the gonads weight over body weight without gonads and stomach or with empty stomach, as expressed in the following formula (Miller, 1984) ws and wb = the gonads and body weight respectively In this case, exclusion of lhe digestive organ could be expocted to improve the accuracy, but during the observation carried out in tha fiold so far, almost all the specimens had empty stomach.
However, the stomach content would not generate wide variation of the measuroment, as the weight of empty stomach would be less lhan 2 g.A microscopic verification of this index was conducted in order lo confirm the relation of maturity staging by mecroscopic examination with the development of ova size, lt was focused on S. crumenophthalmus of which the determination was more dimcult due to higher variation of the shape of ovaries.

Average Size of the Mature Fish
Statistically the length at first maturity would be the same measurement as the average length 0f the sexually matured fish that usually defined as lhe size where 50o/o of the fish to be mature from the distribution fundion like curve.lf the length of the mature fish symmetricelly distribuled follows a normal like probability distribution density, the estimates of the length at first mature stage would be equal to the arithmetic mean.In lhis case, the use of mode or the mean would be efficient enough for estimating the parameter in question that formally defined as the probability oI 50o/o of the distribution function.
The mature fish is defined as those of stage ripe and spawning conditions, which is indicated by a minimum value of gonado somatic index end particular condition of the ovaries.The gonado somatic index will be employed in a grephical presenlation, instead of applying maturity stage criteria.A minimum value of each soecies is defined precisely as confirmed by microscopic examination.An arbitrary value will be used in order to give particular emphasis to cerlain case or simulation.In this case, specimens in spent or partly spent condition are excluded from the analysis since the velue of gonado somatic index already decreases.

Sizc Progrsssion
This part is aimed to obtain a summary of the average of monthly pooled samples, and to describe modal progression of each species for validaling the growth estimation, Figure 1a and 1b show lhe tr6nd of modal progression derived from mean length estimates of the polymodal frequency disiribution samples, As can be seen in the figure, lhe average lengih of the cohort did not exhibit the same pattern at all of the years, However, monlhly growth progression are not as smoolh as exponenlial trend of ihe groMh formula.Variation of longth frequancy distribution and the new r€cruits of D, russellii entering the Java Sea in oeriod of 1991 to 1992 seem lo be more variable lhan in other successive years.,In tracing groMh progression, we ignore the length class strength, as far as the minor distribution can clearly represent any cohort.In many monthly pooled samples more than one length class can be dissected and determined but selection done only for the mean length values being able to indicate general irend of growth line.
It is clear that most of the points belong to the major cohort (cohort thal is clearly visible and gathers most of individuals).Thus, growth es{imation can be performed using the data consist of a cohorl clearly defined without selecting individual samples for tracing modal progression.

Growth estimation ELEFAN method
Grow'th parameters are estimated by performing ELEFAN program and graphicel methods of Gulland-Holt's and Ford-Walford's plots for some case and set of data.In general, estimation crileria or goodness of fit of ELEFAN depend on the Rn's values or ratio ESP/ASP The highest values is considered as an indicator of the best estimate among several trials achieved during estimation.
First attempt to estimale growth parameters was done for D. russe/ri and D mactosoma' applying it to period 1992 to 1995 data.Estimate oarameters resulted by this step are averageo on ihe four cohorts ,.e. cohort 1991, 1992, '1993, and 1994.In this case, the same pattern ot growth cuNes Dass all cohorts with origin on September (as assumed as date of birth) and ended by four or five lines passing unobseNed length' Consequently, some modes and some cohons wers n;t hit -or not properly passed by the grow'th lines.lt could mean that each cohort has different pattern of growth line or the mean length of the bistribution of samples for the same months ere not always equal for different years.In this case, the maih assumption underlying this method cannot be achieved perfeclly and the accuracy of this method is reduced by this diffsrence ln order to increase the precision of fitting, we divide the samDles sets based on the period from the beoinnino of recruitmenl until the last month that a sairote is still being passed by growth line This orocedure can give more reliable estimates as ;hown by the better position of growth lines that hit more precisely the mode of lhe major cohorts (Figure 2).All of the results of computation are lisled in the Table 1.
However, 2 points can be outlined for intepretating the results.Firstly' strong oscillations exist somewhere under the period of sampling as indicated by flat lines drawn by the program and high values of amplitude of oscillation.Variation of or6wth oattern of all main species are clearly 5fro,rn Oy different values of oscillation amplitude of each iohort.Some cohorts show a zero growth during certain periods (as indicaled by the values of C being close or equal lo unity).
The second point is a possible wide range of estimate values because of a presence of combination of K and L. at the same Rn (Kleiber & Pauly, 1991).So, applying this program with the maximum values of Rn's as a criterium would not directly give the best estimates.In prectice.several trials and errors are obviously needed for defining lhe best one.Then, a pattern of maxima values of Rn at function of K for any sample set can be drawn by scanning values of K at given fixed L and other paramete[ estimates (as considered as the best estimate).The combination of K and L" are derived from growlh curve passing certain samples contributing the index of available sum of oeak.This procedur is known as response surface ihat is available in the module of FiSAT system An examDle of this procedur run for D russe//ll of 1 991 to '1993 is presented in Figure 3 As stated by Shepherd ef ai (1987) in Kleiber & Pauly (1991) thet an examination of the goodness of fit resDonse an essential element in any assessmenl technique utilizing size composition data, since this surface will conlain at least qualitative information on the confidence region ol the parameter value and their degree of interdependence.
In this case, a combination of K and Lare usually produced at oach optimum value of Rn, but differeni combination can result lhe same value of Rn's as demonstrated in lhe Table 2. Several trials ' have been attempted with the range of K and L- covering ihe best estimale value being derived from EaEFAN.The results revoal that variability of combination of Loo and K at high value of goodness of fit seem to be evident as vividly indicated by O- russe//li of '1991 to 1992 period Meanwhile' those of olher periods do nol behave similarly (not presented here), with lower scale of variability We notice that the curve patterns (the value of Rn as function of K at given L.) reflects the characteristic of distribution struclure of the data sel generating the growth curve.We adopt this method to identify qualitatively. the growth pattern of the major cohort from different periods.The values of Rn are computed with K variable at given fixed L-(the value L is defined as the best estimate found from several runs) (Figure 3) As shown in this figure, differenl patterns are exhibited by certain periods, particularly, by major cohort of 1991 or 1992 (except for S. gibbosa). .other merhods or estimarion -usins srowth ''fii#ri"Ti'"i| :T"t'ffixifl tHffilrlfT increment of unequal interval of month.. "irrinq b"rrr.nJ-ri"tsplot or ptot of length of successtve successive length at equal interval are performei perioct oi Ford-walfod plot result unreasonabte for rhe besr data avaitabte tn ttris stuo'v ii.e.' b. !iii,iii"i'o"ol" gl.Feb May Aug Nov Feb May AW Nov Fob May Aug l'lov Feb May Aug flov MonttE (1991 -1995) x xxxxxx x MortfB (1991 -199s) Nov Feb MaY Aug Nov May Aw Nov Feb i,lay Aug SardlnelL glbbds.Feb May aug Nov F€b M.y AW ftov Feb May aug l'lov Feb Mav Aug Nov Months (19S1'1995) Monthly mean length derived from length frequency dala' Remarki fs = calculaisd trom minor distrlbutlon oth€rwiee from maior modal The estimates values of L-tend to be too low the large olass) Therefore' it tends to result compared with the maximum -iengih oOserveO t9l?!!Y:|y.. lower value of L estimates that ati.[;tXowevir, there is no stffi mode of the statistic;lly may b9 correcl. (as .derived from Lrg" sir;, (close to maximum l;gthi, used in the regression).lf the largest class is.included in iaiiulationi (excluding few data of-observations on comPutation, the L@ estimate value normally The different eslimates resulted by those msthods can be explained in two points: 1.The ELEFAN has an ability in tracing the oscillated growth progression, while lh; last methods are derived from non oscillated growth curve.David, 1981).We can compare the mean estimates comouted Figure 3.An example of possibles combination of growth parameters (K and L.) of D. i"usse/ri from 1992 to '1993 data.

In
Remarks: the bold line of ihe peaks are optimal combination with the same values of Rn Table 1 .
In general, the elgorithm of this progam tends.todrawn repaatly lhe growth curye and to hit the large groups of fish.Consequently, the L will be ctole 1o the maximum size group.In the G-H methods we ignore lhis group as its mean length is not clear.lt is reason why the estimate Lderived from the ELEFAN are higher than that of G-H method.
Table 3.  Scanning values of K at given L-and other parameter estimates

Comparison of Estimates
A comparison of these results with those from other regions or populations reveal that high variation of groMh parameters eslimate is apparent.The most reasonable seems then to compare with that of the same population (the Java Sea population) and other populations of neighboring regions.
Several previous studies of the growth estimations were performed using the same methods length measurements as applied in this study, ie. in the years 1982 to 1983 (Sadhotomo & Atmaja, 1985), 1985 to 1986 (Atmaja, 1998), and 1991 to 1993 (Suwarso et a/., 1995t.A reesiimate of growth paramelers was also done using first version of ELEFAN | (Dwiponggo et al, 1996), Comparison groMh parameterts of the same area (population) would be more valid because of lhe same fishing method used by the sample boals (purse seiners).In this case we cannot compare the level of accuracy of the estimates, although the goodness of fit for this purpose is available as denoted as Rn.More samples or periods of observation in the data set usually result ln lower value of Rn, because more samDles are not passed by the groMh line, In general, the estimate value of K in this study are higher and the L_ are lower than those of the Philippines waters, but not so much different with the result of lhe same population.Before comparing these estimates with those of other studies, it is necessery to standardize our measuremenl using TL-FL relationshiD'.Once this done, we can see that our estimates of K and L_ lend to lie in the range of the values of othe; studies in lhe same area.But different settino of inpxt data used by other studies (lngle and pa-uly, 1984;Dwiponggo ef al, 1996;Widodo, 1984) could make their growth estimales become incomparable.Two reasons can be pointed out.
Firstly, the data sets used by other studies were not based on appropriate periods but likaly based on calendrical period (from January to the last monln) ot non predetined petiods (from any month to the lasl period of sampling).Secondly, several estimalions were based on shorl Deiods of sampling as indicated by some sets of data used in those studies consisted of less then one year of observalions (as used in Dwiponggo et al, 19E6).
The following paragraphs are focused on annual and interannual variability of parameler estimates with emphazing on the following points a) characteristictic of K pattern; b) the grolvth and oscillalion patterns and their relation to migration; and c) a probable ecological impact on the structure of length composition data.

K's Pattern
Apparently, a single estimalion of groMh perametters is rerely obtained from analysis of lenglh composition data set.lt is more likely that a range of K and L" is produced ovet a plateau on lhe goodness of fit criteria response surface (Shepherd, 1987).In the ELEFAN method, a combination of K and L" is usually produced at each optimum value of Rn, but different combination can result the same value of Rn's.We notice thal the pattern of K at given L intrinsically charecterizes lhe length composition data or is considered as an indicator of characteristics of the length data set.
However, it is beyond the scope of this studv to evaluate in details the relationship between the lype of the date set and the growth pattern.In most gensrel reason of such phenomenon is that if the kurtosis of the distribution is considerably high or if several modes exist in a sample, lhere is a possibility for more than one growth lines to be drawn passing lhat sample.Empirically, a flat histogram lends to be a polymodal frequency dislribution (Harding, 1949: Batacharya, 1967) and restructurization of sample performed by ELEFAN usually results in more than one class lenoth (Pauly & David, 1981).
Usually, the purpose of this scanning and evaluation on the table of response surface (e.9.Table 2) is to find lhe best estimate of K at given value of L, but in this study we use this way for evaluating the characteristic of K pattern as derived from different length composition data.
Compadng the K pattern of different years or cohorts reveals that the data set 199.1 to 1992 exhibit in different pattern (Figure 4).A wider range of K seems to be generated by polymodll distribulion of lerger size of fish at the first half vear and the existing of more that one maior cohorts during that period.lt means that a steeper groMh line can be drawn passing higher value of length without reducing the value of Rn significan y.
Theoretically, there are some possible factors influencing the interannual variability of groMh rale 1 F ot D. tussellii: f L=-0.48 t +1.137 FL (r1:o 987, n=i | 71, t11loe1=216.1 , Scleq=o.005)For O. rraclosoDar TL=-o.194+0.j66 FL (rr=0.85,n=120) (Vvidodo, 196E)  lnd.Fish Res.J. Vol.12 No.1 Juna-2006: 37-63 of pelagic fishes in the Java Sea The first one is related to the stock sbundance, In this case' density dependence of growth rate corresponds to an e;ological adaptation for maintaining high abundance, for instance by extending their feeding ground.The second factor is an interannual ihange in avaibility of food within the limit of adaptability of the species.These phenomena have been proved for japanese sardine (Sard,nops, melanosficfi.rs)by evaluating the long term data of abundance and growth parameters estimate Wada & Kashiwai, 1991) However, our data shqw an inverse ohenomena.In 1991 the abundance was relatively irigher than in other periods as indicated by drlmatic increase of catch per day of fishing and total landing (Potier & Sadhotomo' 1995;Potier, 1998;Sadhotomo & Widodo, 1994) ln rElation t0 exploitation aspect, an expansion of fishing ground wiil clearly add the data of other subareas that may consist oi a new length class to the samples' but lhe more recent data have less length class than that of 1991 to 1992 (eventhough the historical change of fishing strategy indicated thai recent fishirig ground usually more exlensive) So, for the oeriods-of 1991 10 1995, the impact of fishing pattern on variability of length class in ihe data sei can be ignored.
The most probable factor is the change in mioration pattern of migrant species in the Java sei durind lhe year 1991 However.this may be related to-interjnnual changes of hydrographical Daitern inside the Java Sea.Specially, an impact of Et ruif,o can resuli in a longer duration of oceanic water penetration in the Java Sea and a decrease of desalination process (Sadhotomo & Durand, 1997).During this period the migrant species.frominside the Java sea (that grew up in this aroa)-can mix with new immigr'ant from other areas' l e from the norlhern and sestern archipelagos and stay longer than in a normal year.Then thase mixed poplulations would have entered the fishBry with different length structure.

Seasonal Glotvth OEcilations
Almost all ihe growlh patterns obtained through ELEFAN in this ltudv display strong oscillation during certain period (as function of WP) during which the Individual groMh dscreases in an amplitude value (C).The valuos of amplitude are considerably high, estimated to be in the rang€ betwsen 0.1 io 1 (Iable 1).Following the discussion on lhe topic related to seasonal groMh (Pauly, 1990;Soriano & Pauly, 1989; Sparre' igsti, we know that the growth of fish performs 50 seasonal oscilation.Empirically' temperature differences between winter and summer as small as 2'C (as observed in tropical area) are sufficient to induie detectable seasonal groMh oscilation (Pauly & David, 1981).
Evaluation on this aspect would be essential as far as one concerns on the use of length based model in slock assessment.Most of lhe classic model such as dynamics pooled model of Beverton & Holt (1957) are based on non seasonal groMh oattern.However, bies can be generated in computing age conversion from length using non seaional-pirameter estimates Some solution have been proposed, by modifying the basic models such as for length converted catch cuNe (Pauly, 1990) and for yield per recruii (Sparre' 1 991).
Such growth oscilations which are detected in our data -set may be governed by two possible causes a) biological and ecological phenomena' i.e. intrinsic characteristic of the species navlng itreir gro\dn oscilaling' seasonally and migration; and bi an artifacl of the slruclure of length data set inai ciuseo by non bioecological reasons, such as heavy pressure of fishing on certain class slzes' These causes correspond to other factors that can be hypothesized as the sourcs of influence on seasonal growth oscilation: a. tne nist is the salinity or other hydrographlc oarametsr differences of the opposile seasons' We ooserveO that the difference of maximum .nO.initrtsalinity in the middle of the Java Sa" farounO Kirimunjawa lsland) is consiOeiaOtv high (about 3o6o), as well as the dlfferences of transparancy ln this case we ionore tna influonce of lempereture suggested Ul -m"nv authors, due to non significantly oifferencls of temperature during the Northwest and Southeast monsoon o. ihe secono faclor is in relation to competition and predation during iheir stay in the Java.Sea' com'Jetition can bi defined as a population Drocess (nol at level of single organism) though manv of its r€sult are manifestated as effect on oroa'nl"t of which population are composed' ttiih comoetition can be hyphothesized to be ociured in the Java Sea wiih an impact on decroasing of tho individuql growth, during the oerioo oi Southeast monsoon when the lbunoance of pelagic fishes tremendously lncrease.
Analogous the removal by fishery to that by Dredatiot mighi be applicable lt could exerl influence upon individual groMh on the common sizs caught by purse seine (r.e. in the range of t 12-18 cm in FL for five main species, and I 10-14 cln for S. gibbosa).The oscillations (marted by 'flatter growth cuNes) mos y occur in the same size range as lhe dominant size contributing to the purse seine catch.But, we notice that the high removal of th6se slzo is clearly owing to the high abundance of lish, is not caused by selectivity of the ,lshing gear.
c.The third factor is migration effecl on the growth incremenl of some range of length that finally produce a psaudo length classes and individual growih, as well.So far, an indications of the emigration from the Java Sga hav6 been well described (Hardenberg, 193E; Sadhotomo &   Potier, 1995).
The first factor is impossiblB to solve and no reference for these species are found, while for the secono one, an emergent effect of lhis comDetition would not be indicated by an immediate adaptation ,.e. by reducing invfulual growth.Meanwhile deleterious impacl of the catch removal on this size does not exist as indicated by dominant contribulion to catch of purse seine for all years.lf lhere is a high removal of some length classes being due to fishing, the length structure would not be immediately reflected in the samples collected from the purse seine calches.lf thia is th8 cesg.dramatic annual flucluation of length structure oi the catch will exist, due to fluctuation of numbers of progeny.lt is clear that different length classes are present at differont densities as boing roflEc{sd by composition or slructuro of length of thg ssmplei (within the assumption that the catches reDro;ent lhe abundance at s€a).In this case, ther! is no reasonable information for explaining e suddon effect on individual groMh.Moreove[ in case of orsappearenoe of large fish, lhe effocl of competion is impossible to cause an extindlon of a class length (a modal distribution) of species.
Another factor may influence the variabjlity of length of older classes.Hence, 2 possibilities of differ€nce interpretations on the indivldual growth for the old classes.a) as fish become old€r.
individual groMh differences create higher range variability of length within tho same cohort of the older ags (Casselman,19E7); and b) but, on the contrary, if most of individual fishes uniformly grow up, lhe length of the older ones would not virj, for Inslance, lhe difference in length, of 3 and 15 month old may be greater than lhat between on6 and 2 years old fishes.Naturally, there 2 qo-::1ii!'li?." are not capabte ro change rhe tengrh $rudure lhat finally produce low grow{h of cen;in rango of sizo.
The third one is more reasonable although it is not a biological phenomenon bul rather an impaol of diseppearance of some length class on length structure of the fish staying in the Java Sea.In this case, emigration of c€rtain range of size engender 2 possible biases that caused by a mismatch in drawing growth progression line for estimalion purpose: a. Conceming lhe first bias, a length class or cohort totally emigrales from the Java Sea (Figure 5) makes growth line tending lo pass at smeller size in order to achieved the best growth progression following the criteria defined in the algorithm of ELEFAN (Figure 6).Fina y, the astimale of L tends to be smaller and K to be higher, due to steeper growth line as cons€quent of an absence of large size.
b.In the second one, part of fish Q.e. some length class, usually the large size) emigrates lo other area which causes the mean length of the rest lends to be smaller.The high score (explained sum. of peak) of initial length frequency distribution as indicated in the fiist time berioi ot incoming racruits will tend to force the growth line to pass through tength class.And tht droo off in number of large individuals reduces the mean size of the samplos.This gives lower K and L estimates, because the mean sizes of Iarge groups are reduced and the growth line automalically flattens.
Unfortunately, no data of groups of fishes that emigrated trom thg Java Sea are available in this study.However,.accopting a fragilg assumption thet lhe date origineting from whols populetion will arnvs at a conclusion thal seesonally oscilating growth merely ceusod by biologicel phenomenal I n|s assumption wes commonly used in previous studies (Wdodo, 19gg; Suwarso of a/.. 1995: ).In reanslysing the D. russe/// data of lMdodo (1988) using a stochastic modet, Sultivan ef a/ (1990) showed a high different of estimated valu6 refleciing ihe s€asonal growlh pattem of the young recruits.However, lheir results cannot be usbd to extrapolate th6 growth of older fish.

Reproduction and Recruitment
Preliminary sludies on reproductive biology of scaos nad.D€6n pioneered by Delsman (1926) on llre garly life slage, de Jong (1940) on eggs size orstnbution of some p€lagic species.During the penoo after war until recent year, there was no meaningful inv€stigation on reDroduction conducd€d in the Java Sea.In lasl decads.several investigation have b€en conducled with wider scope.Some reproductive aspect with T:1!99mel proposat was given by Widodo (19E8) for both species of Decapterus,-Nurhakim 15 ',f6 17 18 19 20 21 n 23 24 Llngih (cm) case 1 (Left) 'f8 13 20 21 22 n L..{th (cm) Case 2 (Right) Sexual Maturity Distribution of Gonado Somatic Index Structure of maturity stage' of female lishes is evaluated using gonado somatic index and a vorification for its relationship for the main specles is oresented in Table 4.Most of these figures are noi comoarable with results of other studies of the same population.For examples, Nurhakim (1993): nt..ij "t aL (1995) applied another type of index that calculated as: lllustralion of the effect of migration on determinEtion of growth progression (dol hist-ograms lll-iJiir!'olr"remigrator, s-otid lines ar6 growth progression line afler emigration of one or Dart of cohort). (1993)for R. kanagurta.But the most detailed observations was presented by Atmaja ef al (1995) tor four of the main species.
This Dart describes some reproduqtion aspects' r.e.stagi of maturity, spawning season and size at fi|Et malurity of ihe mein species; a part of the.data has been inalysed and presented in Atmaja et a/.(1995).In this study, presentation is more foiused in lhe context to ecolooical point of view rather than those of the population dynamacs.
ffi.1l GsrJrypg!+g$! ,, , , The sverago values of gonado sometic Ind6x by maturity stage in this study are higher than those of  Widodo (1988) that ussd almost the seme criteria of.sfagingfor thg two Decaptarus.The average vatue ol gon8do someuc indox of lhe stags vl of o. russe//li and D. macrosona ara E,24 and i0,7Vo respectively comparlng with less than 8 and S% for th6 result of his study (as €pproximEtad from th6 graphrc presenled in his report), Gjrs6ter & Souse (1983) gave a lowar figure for the highest veluo of gonado sometic index (stage lV of six criteria) for D. russellii of SofalE Benk, Mozambique 0,e, less th_an 5% as shown in their figure compared lo 7,9 or stage v to vll of our study).For R. kanagurta, tne average value of this index is 4.9% for stage Vl compared to -7.2o/o for stage lV (of six criieria) determined by Sousa & cistason (19SS) in Mozambique.These differences could make anv comparison become invalid due to different criteri; of staging.The studies in Sofala Bank used six cnteria of slaging with the maximum index found in stage lV or mature (stage V and Vl are spawning and fully spent).While the difference with the observation of Widodo (1988) is probably due io the different visual determination.However, we could nol infer that different population may have different reproduclive charateristic since the triteria and the composition of samples by length (age) used are not comoarable.
For lhe sake of uniformity of staging criteria used in this study, we emphasize on the use of gonado somatic index rather than the stage of maturity in describing the meturjty of ell main species.However, visual observations were done by team that possibly generates e subjectivity of observation particularly for the stage lV (last maluring) and V (ripe), For this reason we define strictly the mature stage is the condition if the eggs are already lranslucent (stage V to Vll for tive species, stage lV and V for S.gjbDosa).
Based on period of occurrence of the highest 10.86 nta average vElue, monthly evolution of gonado sometic index demonstrate two Dattern of flucluetlons, one peek e year (i.e.D. russalhi.D macrosoma, and R. kanagufta) and irregular pettorn (ie. A. strn, S crumenophthalmus.and S.g,bbosa) (Figurs 7).These pattorn woutd indic€ts trend of spawning season during the porlods of obssrvetlon, although some peaks that cloerly shown in lhe figure came from few speclmens only, unforlunately, we have no Informetion on thE d€velopmant of gonad0 somatic indsx from immeturo stsge to spawnjng stage that enaDlg us to clessify th8 group of spawner for the next spawning.For instance, it is imoossible to classify from similar value of gonado somatic index of A. sirm observed during six successive monlhs We used the lowest limit of length class interval of '15 cm (at mid length of the 14lo 16 cm class) for the first live species (D. russe i4 D. macrosoma, R. kanagurta, A. sirm, and S. crumenophthalmus\ and I cm (at mid length 7 to 9 cm class) for S. gibbosa.
However, these criteria are larger than the common size of these species caught by purse seine fishery and those of the length sample.We relegated lhe specimens below than these criteria because of none of lhem consisted of mature fishes.As shown in the table, lower limit of size where the female begin to be marure can De determined using minimum value of gonado somatic index for a mature fish.Taking Stage V as the minimum stage for the mature fish, we Can see that the midlength of 18 and 13 cm are the border of which the mature female of the first five soecies and of S. grbbosa can be found.

Length at First Maturity
Length at first maturity of six main species are estimated by performing mean length of which the female individual considerabjy mature.This estimator would be analog as the tength at which 50% of the fish are mature.The term of tenoth at tirst maturity is commonly used in the cont;x of population dynamics and estjmated from cummulative frequency distribution.The definition of malure for various literatures were based on 4.78 CJ tnd Fish Res J vol 12 No 1 June-2006. 37'63 different criteria.ln order to avoid subjectivity of staging and ambiguity of criterium used for mature fish, we orefer to use 2 criteria; the minima values of gonado somatic index and stage of maturity.Some of the minima values are approximately defined for giving a comparison with the average value of gonado somelic index of Stages lV and V, Also, we define that stage V (ripe) or above are considered as mature, except for S. crumenophthalmus and O. macrosoma,of which specimens of Stage Vl and Vll were unavailable Deoart from these criteria, estimation of the mean length of the mature female then can be Remarxs: tne ioinrei polrnt!;re the monthly mean' +'s are the maxima and rs are the mlnlma JAN MAR MAY derived from as the frequency distribution of the malure female.The estimate values appear sligh y lower than the result of previous studies from Atmaja et er (1995);Nurhakim (1993).On other hand, Widodo (1991) estimated the length at Rrst maturity to be much lower than o{hers eventhough the criteria of staging are almost same.Apparenfly, JAN FEB I']IAR APR MAY JUN 1992 microscopic veritication is needed for determining the mature condition but this procedure is time consuming.For this reason we used transclucenl eggs for mature criterium of female of all main species.In lhis case, subjectivity of observation plays an importance role in influencihg the final results, more than the methods applied.

Spawning S€ason
Recall to Figure 7, the ultimate of spawning season may be predicted from the peaks of the monthly evolution of gonado somalic index as well as, from back calculation of age of juvenile lish entering the fishery at begining recruitments.More precisely, the occurence of spent specimens would give a direct indication of spawning season.
Prediction of spawning season elso based on summarizing the composition of average gonado somatic index and dala sheet for inspecting the occurence of spent specimens.D. russetlii: High value of gonadb somatic index or ripe stage are normally found May lo July, while spawning period occurs approximately between June to December as indicated by the occurence of partly and fully spent specimens, On December to February, few specimens of resting stage were found in the sample.This stage is merked by residual eggs (dark colour and stuck in ovary wall) left in the-ovary gnd white fat covering the overy surface.
D. macrosoma Ripe stage specimens were found in July 1992 and May to June 1993 in the sub area Mitasiri-sambergeleng Bank, while the spent ones (partly and fully) were frequently found in Julv to Auoust in the more extensive sub areas i.e.Biwean-iumu-lumu Bank in the southernmost of Makassar Strait A preliminary conclusion can be drawn for spawning behaviour of D macrosoma.ltappears that spawning take place in lhe near slope ground or probably beyond the Java Sea.R. kanagutta: similar trend is exhibited with same Deriod of occurence of ripe stage, with the spent specimens found in September to December.
S. crumenophthalrnus: Ripe stages were very rarely found in the Java Sea, only few mature lishei found in March, but in general that did not significantly reflect the occurence of spawning.Also monthly distribution of gonado somatic index performs an irregular pattern without strong peak ivitn nigher average of gonado somatic index appearing in June end JulY.s. glDDosa: Two peaks of average value of gonado somatic index in June and April did not clearly describe the period of spawning There were no clear indication that spawning corresponding to the second period.lrregular pattern was also performed by gonado somatic index distribution by length interval.co Sex Ratio The proportion of female seem to vary Dy month for the four main species.Proportions of females of 2 oceanic species (D macrosoma and A. sirm) tended to pass the male one almost all year (Figure 8).There is no significant conclusron io extract from this flgure without evaluating reproductive behaviour and the relationship between female stock and successive recruits' However, these aspects are beyond the scope of this study, Recruitment Description of recruitment pattern is evaluatqd by projecting length composition of catch of purse seine into time dimension of one year period.We oerform the ELEFAN module in the Fisat software input data are l€ngth frequency by cohort as the sime set of data used for grolvth estmation, length weight relationship (Suwarso at a/, 1995), and growth parameter estimates listed in Table 1 This step needs a raising factor by monthly sample- deiived from ratio catch'calculaled weight 0f sample.We used purse seine catch only with taking an attention that the lenglh frequency do not reorelent the whole size range in the Java Sea Small size of lish usually caught by coastal gear such as liftnet (bagan) were absent in the the purse seine catch.
The results (redrawn from output of the software) are presented in Figure I Different Dattern belween various years are marKly illustrated by main species of pelagic species ln 1992 or 1993 all of species performed a single strong recruilmenl.The time of peak of recrujtment as estimaled using maximum likelihood estimator of Hasselblad method are presented in the Table For most species, special behaviour of the recruitment pattern is performed by the cohon entering Java Sea May to June 1991 (as so called here Cbhort 1991 to 1992).However, different oalterns for each species seem to be evidence for some periods.The pattern can be summarized as follow: 'l. Decapterus russel/ii: Two peak of recruitments, one was a major recruitment and another one was a minor one in the years 1991 or 1992 and 1994 or 1995.single pulse recruitments were observed in 1993 or 1994 and 1992 or 1993 Slightly different peak of recruilment of the periods wore observed.
2. Decapterus macrosoma: Similar tendency with lhose of D. russellii but in different period of oeaKs.The new recruits usually enter the iavanese purse seine fishery around May to June, and being tutty recruiled in the period of September to Oece-mnei The data input used describing the pattern started from May or June to the ne)d vear.

Decapterus russelii
In fact these recruitment patterns did not synchronously follow spawning seasons.Back calculation of age of lhe average size of cohort firstly anter purse seine tishery would be another way t0 trace relationship between recruitmenl and spawning patterns.Calculated age from small size length may be more effective relationship though some of lhem derived from minor distribdtions 1is shown by underline figures).In general.two period or recruitments can be detected, the major recruitment in May to August and lhe minor one in November lo February.
. This would be an auxiliary explanation for the definition of major and minor cohorts as used Recruitment pattern ofthe main species al vanous years' in the beginning of Southeast monsoon (June t0 July) and originated from December to February fUorttrwest m-onsoon period) spawning Whil6 the minor recruils were detected entering the fishery in oeriods ot Novemb€r to March and coming from bther period of spawning.An exception soems-t0 "op".itot S. grDbosa, most of smell or young fish- miy not fully recruit into the Jeva SEe or out of obs;rvation or beyond purse seinB fishing grouno Ho"rever, the ociunence of few ygung fish€s.insamples woud be enough to indicate that the peri6d of recruitment of this species lo be late one month lhan others.

Spawning
Difficulty of finding the mature fish inside the Java Sea in tha area cov€red by this study) would immeOiateiy reveal conclusion that this area is not itre jpawniirg ground of lhe main pelagic species' or at least ihese species do not spawn in the fishino oround of the purse seina ln relation to ,..iriitint and migration scheme, Sadhotomo & poti.ttf eesl imptioitety indicated a possible soiwnini ground of pelagic species (for the iighly niiorant -sp'ecies, such as D. macrosoma) in the ito-oe ot eastem part of the Java Sea without   1992/1993  1993/1994  1994/1995  1991/1992  1992/1993  1 993/1994  1994/1995   '1991/1992   1992/1993  1993/1994  1994/1995   1 991/ 1992 1992t1993 I 993/1994 1 994/1995  1991/1992  1992/1993  1993/1994  1994/1995  1993/1994   4.6   t that the most likely spawning ground would be outside of the Java sea.Delsman (1926) showed an indication of spawning ground of layang (D_ russe//,D near Bawean and Madura lslands in his mission on June 1920 and Octobor 1922. His statement of observed of fullv maturo fish in fishermen catch could be unreliabl; indication for spawning ground as there was no direct obseNation had been made for determining the maturity.Also, eggs and larvals stages specimens found in that subarea might be misidentitied as Caranx kurra and Caranx macrosoma (synonim of D. russellii and D. macrosoma).He speculatively used reference of Trachurus trachurus fot lhe genus Caranx (Dacapterus) and othor genus of Carangidae (that recen y identified as genus of Caranx, Atule, A/epes ).
However, a highly seasonal availability of samples in the period before development of ourse seine could cause a ditficulty in conducting a conlrnous obseNations on pelagic species in the Java Sea.The phenomene of simultaneous spawnrng season as indicated in almosl the same period of the obcurence of the highest average of gonado somatic index of the main species Jould not be used for corroborating the Delsman's specimens.But, in thls case, one could say that the hydrological condition at that time was beter than that of current period.
A question of posslbility of occurence of spawning ground in the Java Sea in past time would relate io disappearence of substock due to heavy exploitation afier 1980's d8cade and shifting of reproduclive regime.Unfortunately, we could not go any further because of the lack of information.
Theoretically, an ideal condition in the past (say 30 years ago) can be said as analog with concept of optimal environmental window of Cury & Roy (1989).Adapting the theory of cargett (1997), the Java Sea area at that time could be regarded as an area with an optimal window during intermediate stabilities that is a range of stability value for which the associated condilion of sea waters (nutrient and others) are sufficient.stimulaling levels of primary production which are significantly larger than those of other period.
Then, we could speculate that before the tremendous increase of fishing, efforl being started in lhe 1970's decade, lhe Java Sea stock might be composed by several subpopulations, more lhan those at recent years.Anyway, lhe Java Sea is not closed to other neighboring areas and migration of larval and nekton stage could be possible following seasonal circulation scheme of sea water.

Recruitment
As shown previously, there are 2 periods where group of young lish enter the Java Sea during the beginning Souiheast and Northwesl monsoon, namely major and minor recruits.For most of species, especially Ioa D. russellii, S. macrosoma and A. s,iryn, the major recruits correspond to the major cohort that possesing clear growlh progression in the Java sea.lt means that most of fishes enter the Java Sea or exactly enter the fishery are born during the period of Northwest spawning season.On other hand, the gonado somatic index data indicate that lhe maiure fishes 60 (stage lV to Vl) exist in June to July of which spawning season is presumably occured in July to October.
Again, we apply an assumption that most of the matuie fishes regularly move to eastward and do spawning outside of the fishing zone Most of the "'oo and |"rv" of the related species are pelagic 156tsman, 1926) that the role of lhe seasonal iunent in transporting the eggs and larvae is obvious.During period of spawning (July to Ociober) part of peiagic eggs and larvae are drifled from the spawning ground to the western area' ano in November current direction is inverse, and most planktonic stage then to be pushed back easterlv.Some parl of those which born in the beginning of spawning season are probably still in the lava Sea and ceught by the fishing gear in November to uecemDer..ln this case, we ignore a possibility of mixure with other progeny from southern of the Soulh ChinaSea that brought by the current during beginning of Norlhwest monsoon in Novemberi December.

coNcLustoN
In general, the von Bertalanffy growth parameter of the six species can be well estimated using the ELEFAN routine.A caution should be taken in relation to the existing oscillation pattern for the .mostspecies and periods that is likely caused by migration of the adult fjsh.The growth estimates derived from this method succelsfullv traces the pseudo growth progression curve, thu; we can notice that the estimated values do not represenl the whole stocks, In this cas6, the emmigration is considered as the more imDortant faclor determining the shape of oscillation curve rather than that of lhe nature somatic growth pattern.
Variability of the Java Sea in 1991 is considered as the the main factor influencing the occurence of additional modal class in the samole sets.More cohorls identified staying in the Java Sea during this anomaly yeai and they consequenfly result a higher estimate values of K ol cerlain soecies.
In general, the tish stock avallable for the fisheries mainly consist of young fishes, i.e. the average size of the main species are smaller than approximate length at flrst mature.Very few specimen of adull fish indicate that these groups are not avaitable in the flshing ground, all of the year.
Two groups of recruits are identified in the Java Sea.The major recruits enter the fisheries durin! the Southeast monsoon (May to August) and th; mrnor ones during December to February.From back calculation of the age of youngest groups of rne maror recruits.we can conclude that these recruils.arenot the offsprings descended by the adult fishes stayed in this area in last year period.
The pe_ak ot maturity of the fishes st,ying in lhe Jave Sea occurs in June to July, and peak spawning season would be on September, while the approximate spawning of the major iecruits is aboul December to January.The scarce of ripe end spawning stage speciemens in the sampl'es indicate that the spawning grounds of the main pelagic species are not in the Java Sea 0.e. at least these are not in the fishing ground of the purse seine fleets).
Figure .la.Monthly mean length derived from length frequency data.R€marksi x's = calculated from mlnor distribu on otherwise fro; malpr modal
JAN FEB MAR APR MAY JUN 1992 JUL AUG SEP OCT NOV DsC 4 ..1, c JUL AUG SEP OCT NOV DEC JAN FEB MAR APR MAY JUN 1993 JUL AUG SEP OCT NOV DEC JAN FEB MAR APR MAY JUN .dexVa|uesofthemainspeciesintheJaVaSea'- Figure 7b.Monthly evolution of gonado somatic index values of the main species in the Java sea.Remarks: the jointed points are th€ monthry mean, +'s are the maxima and ..s aie the minima relation to lhe procedure in defining the mean of length distributon, In ELEFAN.the mean of the distribution is defined,as a mid of length class corresponding to the highest index.This index reflect the sfergth of probability of the class length, and the modal length is indicaled by the highest value.lt is derived bv standardizing the moving average of the obseNed frequency (Pauly &

Table 2 .
Population DynamicAn example of response surface apptied to D. russe/ri data Rema*s: elements are the Rn's values.Shadoll/ed are plaleau on high vatLres ot Rn, bold figures are the Te results of growth parameter estmates of D. russe/lii by applying graphical methods values of gonado somatic index by stage of maturity of the main spectes Average Population Dynamics of ... .. ... Parameters Estmates (Bambang Sadhoiomo) to in relation with bi-modal Oattern of recruilment, as well as for those of R. kanaguta in the year 1993 with recruitment in 1993 oi 1994. appear