POPULATION DYNAMICS OF THE MAIN PELAGIC SPECIES EXPLOITED lN THE JAVA SEA : STOCK EVALUATION

This study was based on length composition data collected from purse seine ffeets operating in the Java Sea, and aimed to elaborate the state ofthe stock of the main species. Two approaches, i.e cohort analysis and dynamic pool model were employed in order to simulate the effect of fishing modality being generated by mini purse seine fishery on the yield, as well as to describe the population structure by size. The impact of exploitation of the young fishes by small scale lishery on lhe the total yield was considerably not important. However, ths accuracy of lhe analysis seemed to be hampered by pseudo estimates of mortality parameters, due to the condition that the structure of data input were strongly influenced by migration phenomenon and fishing strategy.

length composition data can be used in the same fashion.In this case, the growth parameter estimates are absolutely required for converting length into age.With assuming Von Bertalanffy growth model, we perform the yield per recruit model of Beverton & Holt (1957) and Jones' length cohort analysis (Jones, 1981) for 2 species of Decapterus, Amblygaster s,7m, and Sardinella g,bbosa.ln this study, length based models are applied with particular caution due lo incomplete data reouired for modelization and mortality estimation, In fact, the accuracy of this approaches may be hampered by insufficient information from other fisheries exploiting the Java Sea stock, while our main data were collected from the big and medium purse seine fishery landing in the main fishery harbor in the north coast of Java.This constraint relates to the migratory behavior of the pelagic fishes, as well as unreliability of the statistical data in the regions outside Central Java province The pelagic fisheries in the Java Sea consist of different type of fishing gear having different selectivity by species and size.However, the large and medium purse seine contribute more than 6070 of Decapterus spp and A. sirm, and around 3070 of S. gibbosa to the total landing of the same taxonomic group in the Java Sea area (included south of Kalimantan).For this reason, we assume that the catch data are coming from the same DoDulation and no other fishery (except of the Javanese fishery) exploit the same stock But orecaution should be taken for possible bias generated by inaccurate official statistical data.
Since the origin version of the last approach requires that the catch is classified into ages, Many fishing gears are used to catch the same Jones (1974; 1981); Pope (1984) pointed out that species beyond the fishing areas of the Javanese T-ffi;arctG;6tor caplure risheries, Muara Baru-Jakarta lnd Fish Res.J. VoL12 No.1 June-2OO6: 65-90 purse seine, but only three type of gears probably have significant contribution to the total removal.The first one is the mini purse seine that widely adopted by fishermen in most of the region surrounding the Java Sea, i.e. all part of Malacca Strait, southern part of Sumatra region (around Bangka and Belitung lsland), along north coast of Java, some subarea in the Lesser Sunda archipelago, and the south and south east of Sulawesi.The second one is another type of seaning gear, namely as payang.The total catch of Rastrelligar kanaguna landed by payang boats in the southeast area of Kalimantan was enormous, reported to be around 20,000 ton per year (DGF, 1994;1995).Along the north coast of Java this gear catches sardine and anchovy as complemenlary one, but around Bawean and Kangean lsland payang boats also catch Decapterus as reported in the statistical data of Brondong fishing port (Luong, 1997).

Length Cohort Cohort Analysis
Derlvation yearly basis catch at given class of age may be used to estimate population parameter, biomass, etc.This method provides an estimate of original number of individual of a year class being based on number of removal from that year class throughout its life.Analysis associates lo this appjoach is known as virtual population analysis, Gulland's method or Murphy's method.Pope (1972) developed a simple approximation to virtual population technique by pfoviding the basic formulae for this analysis.The basic procedure for this method then was called as cohort analysis.
. Jones length cohort analysis (Jones, 1981) is based on the extension of pope's cohort analysis formula to time variable interval.The method uses common deterministic assumption that within any one age group the decline in number with time variable follow an exponential decay function, which can be expressed as: i:::-l:ll (1 where: Nr+r = DUmber in the sea at age t+1 Z = instantaneous total mortality (in exponential basis) In cohort analysis the exponential curve within any age group is repl?ced by a slep function which is based on the following assumptions: 1.The whole catch for that age group is taken at middle of age intervat.
2. Natural loss is constant through all age (instantaneous natural mortality, M is constant) Detail of the derivation was presented in Pope (1972); Jones ('1981).
In this study, we apply lhe basic model of Jones length cohort analysis.Fishing mortality and cohort size (biomass of the survivors) are estimated from the catch at length data alone without any auxiliary data (such as fishing effort).For the length cohort analysis applied in this study, we use catch in numbers from beginning of recruitment to largest group following the tendency of modal progression.
The periods of data set are defined as the same period of those used for growth estimation, In fact, thjs method needs the data lo perform to a steady state condition.lt is impossible to know this condition without elucidating the historical data of catch and effort.In practice, it can be approximated by averaging the catch at length data for couple of years (Pope, 1980), For this reason we use the catch at length data of D.
russelrT and D. macrosoma from three years periods (i.e, 1993 to 1995).White for s. gibbosa we apply to one year period (1994 or 1995)   because of the lack of measuremenl data from other fishery, but assumption of steady state condition can be taken as this species is more coastal and considerably undergo heavy fishing pressure.
Diagrams showing the range of age exploited by two types of fisheries (Marcille, 1978).
The first step is to compute yield per recruit is for F2 and L62 var! at arbritary Value of Fr and presenting the result in isopleth diagrams.
The second step is done in the similar manner, but with L": are fixed at given values.lt is mainly aimed to determine the optimum fishing mortality (as so called as F optimum and denoted as Foer).
Optimum value of fishing mortality of second fishery can be calculated in 2 ways: a. F6e1 is equal to the value of F giving maximum yield or yield per recruit, i.e.F6p1=F63,.
b.In case that M is relatively large, the Fopt may be very large and tends to be unreasonable or may not exist.The concept of Fo.r as another crileria to define the optimum value of F can be adopted for this case.Fool is defined as fishing mortality rate corresponding lo a rate of increase in yield with F being 0.10 of the initial rate of increase at the start of fjshing (Gulland & Structure of Catch and Length at First Capture Boerema, 1973).This value is denoted as Fo1 that can be mathematically derived as the first partial derivate of yield function (with subject to F) at conditional restriction F=For lt can be expressed as (Gulland & Boerema, 1973) In practice, it can be illustrated graphically by taking tangent line at the initial rate of increase in yield (line a) and measuring the slope to the absisca (as angle of d).Then taking a tangent line (line c) which is parallel to the line b having slope ot 0.1.This is a tangential to the yield curve at the point having coordinate (Fo.riYor), where Ye.1 is yield per recruit at F6.1 Figure 2).Two important oarameters relaled to the population are estimated using length composition data, ,.e. the length at first capture and mortality.A brief discussion on repartition of the catch by type of fishery is given, in order to describe possible bias of our estimates.
Length at which 50% of fish captured can be reflected by length at probability 50% derived from cumulative frequency distribution.We estimate the length at first capture by category of fishing boats of the purse seine, as they tend to catch fish in different size (Sadhotomo & potier, 1995).
Quarterly based estimations are performed as well as the last 2 years periods of observations (1993  or 1994 and 1994 or 1995).The estimates are -discriminated according to boat size category: large, medium and mini purse seine.General This^wo|k-is^a part of the more comprehensive study carried out by the Java Sea pelagic Assessment project during the period of 1994-J995 ob explanation on the catch structure is also reviewed in relation to the operation of fishing fleets.

Selne
Rough estimates can be performed by interpolating some points from cumulative frequency (Figure 3 and Table 1).lmmediate conclusion can be drawn concerning with the different vulnerability of the species to the size category of fishing boat (also for the fishing gea|. Generally speaking, large seiners tend to catch slightly bigger size of fishes (Figure 4,5), even though those type of boat use the same fishing melhods.Similar tendency has been presented in the previous studies using length data of 1991 to 1992 (Sadhotomo & Potier, 1995).
In this part we focus our analysis on the three main species, e.g.Decapterus russellii, D. macrosoma, and Sarding//a glbbosa.We noticed that the two type of boats may catch the same cohort with some exception during certain periods, i.e. for D. russe/fi and R. kanagufta, dwing quarters January to March 1995, and October to December 1995, respectively.In the peak of fishing season (August to November) the differences seem to decrease, particularly for migrant species.However, different fishing ground can be regarded as source of this variability that the medium seiner tend to prospect the tishing ground near main island, ie.North Coast of Java waters, Karimunjawa and Bawean Bank that usually inhabited by younger fish.Characteristic of the three fishing grounds, where medium seiner usually operated, can be considered as mole coastal than other zones, As shown by average fish length sampled from medium seiner, a general Dhenomenon of the occurrence of smaller fish in the near coastal area may be evident.
Comparison of the fish length structure of caught by the 2 categories of gears (e.9.during May or June '1993 to April or May1994) indicates significant ditference of fish size of the migrant species, such as Decapterus macrosoma and R.
kanagurTa.However, it should not be interpreted as possibly impacl of difference size of the mesh of nel, as the seleclivity is mostly influenced by the operation.Two sources of influehces may be inferred lrom the change characteristic of the medium seiner after 1994.
Before that year, most of the medium seiner used to be equipped by lower intensity of light for aggregating the fish ie.mercury bulp lamps.After this period, more proportion of new boats classified as this category utilize the same lighting apparatus and fishing strategy as those of the large seine.The better autonomy and more powerful engine seem to enable the medium purse seine to operate in the same strategy and to use the same lighting (ie.lhe use of metal halide lamps replacing the mercury bulps) as the larger category.Reason why, in the successive year, the length at first caoture derived from the medium seine catch does not indicate any significant difference.But, in order to avoid any bias may be generated by this fact, we treat the data separately, especially in raising up the length frequency.

Mini Purse Seino
Mini purse seine fleets also exploit the pelagic species along coastal area of Java lsland, from Kangean lsland until Sunda Strait.The catch composition may vary with the location of fishing zones, as the fleets tend to move systematically following aggregations of the most profitable aggregation of fish.In the fishing zone around Bawean and Karimunjawa lsland the monthly fluctuation of catch shows a similar trend with ones of the medium seiner.
It may be demonstrated by catch composition samples from Sarang fishing port (its latitude is between Karimunjawa and Bawean lslands).In this subarea scads seem to be the main target during the south east monsoon, while in other subarea the mini purse seine tend to prospect other specles (Auxls spp., Thunnus tonggol, sardine, and black pomfret).
Hrgh contribution ol Decapterus lo the total landing during peak of fishing season (August to November) is obviously observed, on the contrary, the sardine catch is more important during the opposite season (Figure 6), In Pekalongan harbor where the mini purse seine operating in north coast of central Java unload their catch, lne Decapterus russ€/lll category has never been recorded in the auction data base, although low percentage of this category occasionally exist in the catch Nevertheless, the length of this species caught by mini purse seine appears to be similar with that of medium one, in their pattern and size (unfortunately, only 2 samples of length measurement taken in this fishing port consisl of Decapterus russellii).
We used official statistics published annually (DGF, 1996) for estimating the total catch of S. gibbosa coming from the Java Sea stock, due to invariability of better statistical data.The length composition of the sardine species are based on In this case, we perform two ways of data poofing for Decapterus.In the first way, the data of the years of 1993 or 1994 and 1994 or.1995 are treated separately (not presented in the above table ), and the second one, we use average of the two periods in order to have an input for more equilibrium condilion.Fot Amblygaster sirm, Dooled data of 1991 to 1995 are used with assuming that no another tishery exploiting this population.While for Sardinella glbbosa, only the data of the oeriod 1993 or 1994 are used with taking into consideration that this species is heavily of this species is done by multiplying the length frequency data with two factors corresponding to the origin of the samples.The first factor is for the length composition measured from the large and medium purse seiner that calculated as ratio of total catch of this semi industrial fishery divided by calculated weight of samples.
The second one is for raising up the length frequency measured from the mini purse seiner landed in Pekalongan during the period 1994 or 1995 which is defined as ratio of total catch of sardine in the Java Sea as reported officially minus  Mar Apr MaY Jun Jul Aug SeP Oct Nov Dec Monthly catch composition of mini purse seine landed in Sarang Source: auclion samgl€ dals of Sa.ang fishing port the total catch of large and medium seiner' divided by weight of samples.In this case, we consider that the estimated catch of the large and medium seine fishery is more reliable.mortality Total Mortality Total instantaneous mortality (denoted as Z) is estimated by means of catch curve.lt is derived from the exponential decay equation of population chanoe.This method commonly employs age com;srtion data of the catch with assuming that the catches represent or proportional to the abundance at sea (Beverton & Holt;Ricker, 1975) Since the method requires catoh at age data input, and the fact that aging data are not available' the catch at length dala coming from non selective fishino oears can be used in similar manner as age como;;ition ones to derive a catch curve (Gulland, 1986).Converting length into age using-Von Bertilantfy growth formula with previously estimated paAmeters (Sadhotomo, 1998), then Z In this estimation, an attention should be taken on the assumption underlying the model that population has stable age structure (Le. as represented by catch slructure or catch at length) and constant recruitment, as well as the same mortality for all chsses.For this reason, we use average of several years data for avoiding any possibility of unstable age structure as listed in the Table 2. .In regression, only points representing the decreasing abundance by successive length class are selected, while the others (smaller classes and sometime the largest class) are excluded due to the lower selectivity of fishing gear to these classes.
. The estimate values of Z (Table 3.) appear to be higher than the results of other studies.Widodo (1988) estimated the total mortality using the same method to be 2.15 and 3.13 for D. rusieltii and D. macrosoma respectively.Other estimation conducted by Dwiponggo ef a/.(1986) gave almost the same results.
Another method, such as mean length method which derived lrom ZIK quotient (Beverton & Holt, 1956) also tend to give unreasonable results, due to mathematical artifact as the mean length of the dala are very close to the length at first of capture (see also Table 1 for estimates values of [.c).
The formula of this method can be exoressed z (L--L-.."\ R=lC;;-LJ' ( 6where: Lm6an is lhe mean length of catch composition and Lc is length at first capture.

S.aNirnophthalfiut
Arrll-.ljn.9l8.3 8.5 10,5 12.5 ra,5 16.5 1E.3 20.5 22.6 Fork l.noth (cm) Histogram of length frequency of some of the main species caught by mini purse seine.   . sirm 1993-1994 1994-1995 1993-'1995 1993-1994 1994-1995 1993-1995 1993-1995 I  The simplest way is to apply the empirical equation of Pauly (1990) that expressed as log M=-0.0066-0.279.log L"+9.6543.log K+0.4634.log T, where T is average seawater temperature in Celsius degree that is approximated to be 29 Celsius degree.The results of this estimation are listed in the Table 6.3.However, bias certainlv existing as this equation is derived from differeni ecological areas.In this case we use this formula as rough approximatton since the other methods do not give reasonable results.

POPULATION STRUCTURE
We perform Jones' length cohort analysis to Investigate the structure of population by length cfasses for 2 species of scad (Decapfe rus russe i and D. macrosoma), and 2 species of sardine (Anblygaster sirm and Sardinella glbbosa) using the Input data listed in Table 2.

Survlvorc and Steady State Biomass
As commonly produced by this method, the average number of fishes attaining successive length classes (denoted as NL) exhibit a common pnenomenon that is a tendency to follow an exponential function.No oscillation of these figures rs observed for the selected species (Figure g).
Analysis can be focused on the average number at sea (survivors) and its converted val;e In oromass (r.e. as defined as steady state biomass).Hence, we can evaluate the pattern of catch profile and the steady stale of biomass for each species.In general, the most frequent catch by length class are not in the same pulie with that of survivors biomass, also it can be indicated that the total biomass tends to be lower than the fisherv yield (Figure 9).lt is clear that structure populations may not be represented by the structure of the catch and no appropriate ratio value that can be defined to indicate the representation of the population at sea from catch Again, a comparison of ratio of biomass over catch of 2 Decapterus (the most dominant species in the catch) do not lead to an indication on relationship between biomass structure and their bulk catch in the fishery (Figure 10).But similar patterns are more explained by type of species, l.e.
between D. macrosoma and A. sirm (within more oceanic species) and between D. russe/r7 and S.
glbbosa (within more neritic species).The patterns of the ratio are very different, for the first group of species, very low values exist jn the size ranqe representing the most frequent size in the catc-h.
While, the patterns of another group tend to be characterized by oscillation at the common length classes and by higher value of the biomass caich ratio.For all species the biomass-catch ratio of the smaller length class are generally very high (not shown in the figures) and nothang can be deicribed from this tendency.
Fishing Mortatity . Variability of fishing mortality by tength ctasses ol the 2 species ol Decapterus appears to tndicate different pattern.For D. macrosoma, the highest mortality value occurs in the length class of 16 to 19 cm, while the peak of removala tie in the smaller classes.While, the pattern of D. russe/r7 is quile different with osciltation in the 16 to t7 cm and tZ to l8 cm length class during the 1993 or 1994 and 1994 or 1995 periods respectively.In order to know the influence of interval size, we attemDt to enlarge it from I cm to 2 cm class interval (not presented in the text), but the result shows that ihis oscillatjon still exists although in lower intensity.8.5 11.5 14.5 17.5 20.5 23.5 Fork length (cm) 510 15 20 25 Fork length (cm) 8.5 11.5 14.5 17.5 Fork length (cm) Figure 8a.
DecaDterus russe/r7: structure of the catch and population, biomass of survival, and fishing mortality as computed by using length cohort analysis 510 15 20 25 Fork length (cm) exploitation rute, E (E=F lZ, and average value of F (Table 4).
In general the rapidity of convergence of F would indicate the robustnesF of the estimates.lt can be shown that the estimates are well fitted to the data at leasl for three species (D-macrosoma, D.russeltii, and A. s,7m) of which their convergent ooint lie before the peak of the most frequent length.
Similar pattern is also shown by variability of fishing mortality of S. gibbosa (Figure 1 1) with oscillation point in 12 to 13 cm length class-In this case, the profile of catch at length data influence on the Dosition of the highest value of F or its variation.Pope (1972);Jones (1981) implicitly indicated that the value of M alsd determining the oattern of fishing mortality by length class We tonflrm it by introducing some values of M (other than the esiimates) which changing the mean of Fod( lenglh (cm) 8.5 '11.5 14.5 17.5 20,5 Fork length (cm) 14 000 12 000 ^ 10 000 -8 000 f 6 000 4 000 2 000 Fort l€ngb (cfi|) 8,5 11.5 1,a,6 17.5 20.5 Fo* length (cm) Figure 8b.

YIELD PER RECRUIT
In this part, the Beverton and Holt model is employed by exerting two fisheries in the model.
Natural mortality is assumed to be constant with knife edge recruitmenl as well as that the first fishery catches smaller size of fish than that of the second one.Furlher interpretation can be done by considering the fish at this size as juvenile stage.
This scenario can be aimed to predict unsuccessful recruitment or high removal of young fishes, rather than to investigate the effects in the change of exploitation pattern on the yield and to define the current status of exploitation.However this type of 78 25.5 5.5 10.5 15.5 20.5 Fork tength (crn) Decapterus macrosoma.
analysis appear to be very ambitious and ambiguous, as the model is very sensitive to the change of value of M.
The procedure can be simply implemented by simulation rather than actual data inout.The first step is to employ this model with fishing mortality and age at first capture is variable at various values of F1.Any change of Fl can be expected to impact the yield of second fishery as the large part of younger stage being caught by the first fishery.
The second one is to find the oDtimal values of fishing mortality corresponding the maximum yield per recruil at the fix values of length at first Poputahon Dynamcs of .Stock Evalualon (Eambang Sadhotomo) 1993/1994 a4 E.
dl 0 o 12 000 10 000 ?I ooo € 6 ooo t l ooo 2 000 0 7.5 9.5 11.5 13.5 '15.5 17.5 Fork length (cm) 9.5 1'1.5 13.5 15.5 17.5 Fork length (crn) caoture.The concept of Foi is applied if optimal F lf 'at tne yietO per recruit maximum) appears to be unreasonable or too large lnteraction of Two Fisherieg: A Cla3sic Simulation 9.5 11.5 13,5 15,5 17.5 Fork length (cm) lsopleth diagrams of yield per recruit of four main species -are performed for the second fisheries (e.g.large purse seine fishery) First step is to araivri the diagrams by setting the fishing mortalitv of the first iishery equal to zero and the ""i"rt"tion will return to the ordinary model with single fishery (the first fishery is still in the model)' Th; result indicates that the current status (in term of length at first capture) of four main species are. in the;ptimum condition.lt means that the value of lenoth at first capture are in favorable size as the pos]tion of currerit Lc lie in lhe optimum area in the isopleth diagrams (,.e.coordinate representing- combination -of F and Lc giving highest value of yield per recruit).
Considering that the first fishery (1.e.small scale fishery) exploit the same stock at lower length.at first caoture than that of the large purse selne fishery.Assuming F1 about half of the mean F in computing yield per recruit seem to be more Sardinella gibbosa.realistic although this procedure violates the validitv of the niodel (i e. mean F is computed in differ;nt assumption underlying the Jon€s cohort analvsis).Due tb unreliable estimates of length at first ' capture of mini purse seine fishery' we- ""rur!'tn"t" values are 3 cm less than that of JeconO fistrery.We could use the estimate values of lenqth at flrst capture of mini purse seine (as listed -in Table 1), but they seem to be over estimate.Also thise figures are reasonable as many small-scale fishery frequently catch these sDecies in smaller size This Drocedure is clearly able to explained an impact oi other fishery catching smaller fish size on thd vield of the lar6e purse seine fishery (that capturing larger size).The results show that an JxLrting-ot tlie additional Ft in moderate level is abte td stritt the optimum area to lower values' wittrout cnanging significantly their shape ln fact' iit" i*tea."6r it is-followed by slightly indease.ofiot"t ti"tO per recruit, but lowbring the yield of ,the i""oiJitn"tv (right hand side of the Figure 12 to 14]-.

Optimum F
Ootimum F usually is defined as a value of F, which give the maximum yield lt can be estimated Su"ri"o* lnd.Fish Res.J. Vol.12No 1 June-2006: 6190 Decapferus russe/t (1993 to 1995) 3 500 3,* Decapterus macrosoma (1993 to 1995) .D 3 000 5.5 7.5 9.5 11.5 13.5 15.5 17.5 19.521.5 t\,4 id tegth (cm) Amblygaster sirm (1992   Mid length (cm) 6.5 8.5 10.5 12.s 14.5 16.5 18.5 20.5 22.5 Mid lerEffi (cm) ? 4 000 93000 .9z ooo z @-1 ooo Eô u 5.5 ?.5 9.5 11.5 13.5 15 5 17.5 Mid length (cm) 5.5 7.5 9.5 11.5 13.5 .t5.5  r4.5 16.5 18.5 20.5 Mid l.ngth (cm) d.  1994-1995   1993-1995   '1993-1994   1994-1995  1993-1995  I 99'1-1995   1993-1995    by performing the model at given length at first cioiure and M. In our case, the input.parameteruaiues are considerably high (,.eM)' although the current length at first capture lies in the optimum area (in th; isopleth diagram) ln this case, the maximum value of yield does not exist or not reasonable.because the yield curve is very flat with very high value of F at maximum yield (F'*) (Figure 15 and '16).For this reason, we apply the ioi"ept of Fe 1 to overcome this mathematical conseouence of the use of high input value of M and Lc.This concept of F rate at Fe 1 has become widelv aoplied as management criteria in some reoions.such as in the Atlantic fisheries in the .The advantage of lhis criteria is that Foj is always lower than Fmrx and applicable to th; flat curve phenomenon being generated by high value or M and Lc 0.e.Fmlx would be very large).we simply estimate lhe Fo.1 by eye from the yieid curve rather than applying theoretical formula.We perform this procedure to calculate Fo.l fof two species of Decapterus.Approximate values of Fs I of the two Decapterus arc listed in the Table 5.
.4 and 2.O yeat 1 for D. russe//rl and D. macrosoma respectively).These rough approximations appear to be in good agreement with the results of Widodo (1988).He found the values of Eo.1 are slight higher than that derived from the lowest approximate values of For.The methods used are ditferent, in this study an approximation is applied rather than perfoiming direct calculation of Fo.j or Eo 1.We believe that the variation of estimates is strongly influenced by the input value of M. As the value of M can not be precisely estimated, the estimate value of Fe 1 and Eo.1 should be aimed to give a general description on the state of exploitation.
It is reason why we evoid evaluating the stock being based on the inappropriate input.However, the use of a range of the input of M may be promisable, and can be hoped to describe general condition of the stocks at low accuracy.Instead of using the range of M, the rough approximate of Fo 1 would be analogous to the use M as variable.This would reslJlt more and less the same variation of yields estimates.
The analysis performed in this part has two drawbacks, since the assumptions geem to be difficult to achieve.The first one, whicli is one oart of the stocks (of the same population), is exploited by the Javanese fisheries of which the data were Table 5. collected from.The second one, the validity is obscured by ecological phenomena that is not included in the model such as false growth oscillation induced by seasonal migration activity.
In the analysis we use conversion formula of oscillation growth that is facilitated in the Fisat software and in our computation of yield per recruit, but the models do not include lhe removal being caused by emigration of the adult fishes.
The unclear result as mentioned above may be caused by several factors.Recall to the previous chapter, that migration of the dominant species (i.e.D. russellii and D. macrosoma) play an important role in determined the repartition of size and shoal pattern formed by the species ln relatron to the seasonal pattern ol seawater properties (especially, salinity) and the east west gradient of salinity, the migration tends to follow this environmental gradient pattern.lt will give an immediate impact on spatial distribution of fish size.This oattern has been known by the fishermen, and their knowledge on the migration has well developed, but it is valid only for the stock inside the Java Sea, while the data from other area are not available at this time.Finally, the length composition data represent only the structure of stock exploited in the Java Sea area.lgnoring the accuracy on the natural mortality estimation, the validity of model applied in this study may need an adjustment due to this Dhenomenon.Theoretically (e.g based on the mathematical formula), the drop of the number of frequency of the big size in the catch precisely cause a'lower estimate of the number of fish attaining this range size, as manifested by the lower biomass-catch ratio.As described in the previous chapter, an emigration of the most of large size class to other areas beyond flshing ground of purse seine is obvious.
In this case, the modification of the VPA model should involve an effect of the emigration and immigration during southeast monsoon period.This affect would be strong because of its influence on the inDut data structure and input parameter (e.g.growth rate).Inserting a function replacing, ixp 1lltZ1 term for representing the effect oJ seasonal catches (Mertz & Myers, 1996)' but it seems to be an unclear approximation since the Droblem in quantification of that faclor has not 6een solved yet in this study Another solution has been proposed by regarding the model as an aDproximation to more general conditions when fiihery varies continuously during the year (Kizner & Vasilyev, 1997).These amended models, however, never consider the effect of migration, although the last one gives an option of zero catch that may be analogous to the undeterminable catch.lt would arrive to another question related to the definition of catches as removal.In this case, removals from the Java Sea equal to the real catch plus emigrants.Our data shows that the majority of the catch consists of progeny of different class exist in the last year period.Empirically, there is no agreement between the age derived from oack calculation of the pulse of recruitment and calculated of date of birth from the period of the occurrences of the highest value of lGS.However' no method has been developed to model this type of emigration process.
Other constraint of application of this method is related to the data acquisition in the official level.
Until recent year there is no reliable statistical data available from the provinces out side the Central Java Province.The basic system of statistical data collection is the same for all provinces but the tradition of data recording is really different.
Unfortunately, this study is not concerned lo this problem as the project objectives are more focused bn the Java Sea.Citing from the report of Directorate General of Fisheries may arrive at a substantial biases, as the tenddncy of an optimistic increase of landing at certain level (/.e around 4 to 6% per year) whatsoever the condition of the fisheries.
Evaluation using another approach (yield per recruit model of Beverton and Holt) faces the .problem of the input value of natural mortality (M).ihe influence of different M on the calculated yield (yield per recruit) appears to be important, due to the validity of the estimate M, a general description on the optimum fishing mortality maybe calculated at given range of value M.
A solution by applying the concept of Fo r to avoid the undeterminable optimum F may be an artificial even though this technique successfully applied in this study.Unfortunately, the accuracy oJ eiiimate maybe very low, as the range of F ootimum at different value of Lc and M are wide for the two Decapferus, although these values cover the results of previous study of Widodo (1988) But, it may pose a question relating to development of the exploitation ls the state of exoloitation in 1993 to 1995 lower than that in 1986?This question should be in relation with historical development of lhe javanese purse seine fishery.As indicated by other studies (Potier & Sadhotomo, 1995), the dramatic increase of the develoDment (/.e.investment) of the purse seine fishery was begun in the year 1986 as marked by new investment on the bigger boats.This be triggered by lhe enormous means that fishing mortality in the recent years are abundance .ofperagic fish disponibre.at the year gr""t".' ih"n thosi of previ6us p"iiooi." ' 1_985 being induced by interannual varlability oi the climatic phenomena.However, one year wourd be Again, it wi back to the accuracy of estrmate, In too short ro generate drasticalv incr6age of fishing ttris ?se, the moder used ind the Inpur effort, and from this oeriod ihe inveitment njE ;;Fr;A we can notice ihai ;ur' samprrng been going on until 1gg3 to 1995.
scheme is more comprehensive than other studtes rr is crear thar rhe removar,g_r?J:lg.
lf the values of Fo.r estimates can be assumed to be well estimated r.e.no violation lo the model' it v/ill be unfortunated by the lack of conversion D. macrosoma: lsopleth diagram of yield per recruit for large seine fishery' The left.side is.a absence of other fishery; tne rrght'hino'side is showing-the impact of othef fishery with assumed to be 12,5 cm.Yield per recruit of D. russetlii at various values of M and Lc, based on estimate paiameters of 1993/1994.Ll and L2 denote lines paralel to line of Fo.r at Lc=10.5 and i5.5 cm respectively.Shadow area is showing the approximate range of points at tangents lines.At the current slate, more than b0 percent of catch consist of young fish or in other word, length at tirst capture of the main species are lesser thin .. ... . .. ... ... ... .(s Figure 2.
fished by coastal tishery.Raising up the samples Figure 3.
standard etfort.However, information on the fishing effort are not available for this study' detailed description are presented in Potie n Dress.Nevertheless, this model has an advantage in modeling the impact of exploitation on the juvenile stages.

Figure 16 .
Figure 16.Yield per recruit of D. macrosoma at various values of M and Lc, based on estimate parameters of 1993 or 1994.Ll and L2 denote lin€s parallel to line of F 0.1 at Lc=10.0 and 15 0 cm respectively.Shadow area is showing the afproximate range ot poinis at iangents tines.

Table 2 .
Estimates of average catch at length of four pelagic spedes be derived by plotting or regressing the natural logarithm of relative abundance of consecutive length classes against relative age (calculated age with to to be assumed equal to zero). can