Control of assessment for demersal fish stocks in ICES area: analysis for 36 stocks and investigation of some potential bias sources.

Transcription

1 ICES CM 004/K:17 Control of ssessment for demersl fsh stocks n ICES re: nlyss for 36 stocks nd nvestgton of some potentl bs sources. Mre Lesueur, Dder Gscuel nd Trstn Rouyer Abstrct We nlyzed here the relblty of ssessments performed between 1997 nd 001 for 36 demersl stocks evluted by ICES. Tkng s strtng pont the metrcs of bs nd vrblty defned by Jónsson nd Hjörlefsson (000), we hghlghted stocks for whch prmeters of fshng mortlty or spwnng stock bomss were sgnfcntly over or underestmted. Then we explored the possble cuses of these ssessment bses by exmnng three hypotheses: ) ssessments were bsed when the explotton rte or stock bomss ws vrble; ) ssessments were bsed when prmeters F or SSB presented pronounced trend over the studed perod; ) the yers of ncresng trend of F or SSB (recproclly decresng) led to underestmte (recproclly to overestmte) ths mortlty or ths bomss. For lrge mjorty of stocks we observed n underestmton of F nd n overestmton of SSB. On verge, dgnostcs were therefore "excessvely optmstc". A logcl negtve correlton between the two types of bses ws found. Sgnfcnt errors on F were observed for the thrd of the studed stocks nd lso the thrd of the stocks were found to show sgnfcnt errors on SSB. Averge bses ndcted tht ssessment errors reched 50% over the 5-yers perod for the ffth of the studed stocks. Moreover, certn stocks presented extremely vrble bses reflectng sometmes mportnt but non-systemtc bses n the ssessment. Keywords : Stock ssessment qulty, fshery mngement Mre Lesueur, Dder Gscuel nd Trstn Rouyer. Agrocmpus Rennes, UPR MESH, Déprtement Hleutque, 65 rue de Snt Breuc CS 84 15, 3504 Rennes cedex, Frnce. Correspondng uthor: trstnrouyer@yhoo.fr 1

2 Introducton Stock ssessment s the bss for scentfc dvce used n fsheres mngement. The ICES workng groups yerly gther to ssess the dfferent Europen stocks. Ths ssessment conssts n complex process tht requres lrge mount of dt. The qulty of these ssessments s questoned nd n prtculr by fshermen. Despte ther mportnce, very few studes del wth ther relblty nd put forwrd the orgn of the potentl error sources. Ths note comes under the Work Pckge "evluton of performnce of ssessments" of the Europen Advsory System Evluton project (EASE). We were nterested n the dgnoss performed between 1997 nd 001 for 36 demersl stocks evluted by the workng groups of ICES. Severl ndctors were clculted tht mde t possble on the one hnd, to nlyze the relblty of ssessments, nd on the other hnd llowed the explorton of vrous cuses lkely to expln the demonstrted bses. Dt nd methods Dt sources The dt used cme from the Qulty Control Dgrms (QCD) provded by the vrous EASE prtners. For both stocks of Megrm (dvsons VIIIc nd Ix), ws supplemented by ACFM dt. Stocks QCD s were vlble for the 36 followng stocks (Tble 1). Tble 1: Lst of the stock consdered n the present nlyss nd correspondng ICES subdvsons Group 3 ICES subdvsons Group 4 ICES subdvsons NEA Cod I; II Anglerfsh (L. psctorus) VII b-h, j, k ; VIII, b, d NS Cod III ; IV, b, c; VII d Anglerfsh (L. budegss) VII b-h, j, k ; VIII, b, d VI Cod VI Celtc Cod VII e-h, j, k Bltc Cod W, 3, 4 Hke North III ; IV; VI; VII; VIII, b Bltc Cod E 5, 6, 7, 8, 9, 30, 31, 33 Hke South VIII c; IX NEA Hddock I; II Megrm North VII b-h, j, k; VIII,, b, d NS Hddock III ; IV, b, c Megrm South (L. bosc) VIII c; IX VI Hddock VI Megrm South (L.whffgons) VIII c; IX VI b Hddock VI b VIIe Plce VII e Norwy pout III ; IV, b, c Irsh Plce VII III Plce III Celtc Plce VII f, g NS Plce IV, b, c Irsh Sole VII VIId Plce VII d VIIe Sole VII e NEA Sthe I; II Bscy Sole VIII, b NS Sthe IV, b, c; III ; VI, b Celtc Sole VII f, g Sndeel IV, b, c III Sole III NS Sole IV, b, c VIId Sole VII d NS Whtng IV, b, c; VII d VI Whtng VI

3 Nottons j j F nd SSB represent the fshng mortlty nd the spwnng stock bomss of the yer estmted the yer j respectvely. Prmeters were noted n generc wy. Conventonlly, prmeters estmtes from the ssessments were used s reference. They were regrded s true vlues nd were compred wth the sme prmeter estmtes from the prevous yers. j Relblty of ssessments The relblty of ssessments estblshed for the yer durng the yer +1 ws nlyzed by tkng nto ccount two prmeters: the fshng mortlty (F) nd the spwnng stock bomss (SSB). Two ndctors clled ssessment rtos were estmted, ccordng to the ssumpton tht the estmtes could be consdered s true vlue: 1 F + nd SSB 1 + F SSB Rtos hgher thn 1 ndcte n overestmton of the prmeter, those less thn 1 n underestmton. For ech rto, two ndctors were clculted on the fve yers perod before. Thus these ndctors took nto ccount ssessments crred out from 1997 to 001 (nd therefore concernng the estmtons) 1. These two ndctors were smlr to those defned by Jónsson nd Hjörlefsson (000): Averge bs: ( ) 1 b = ln( ) Averge stndrd devton: = 1 sd ( ) ln( ) b ( ) 5 The frst ndctor mesures the verge bs n the prmeter estmte +1, the second ts vrblty. Vlues of the b ndctor bove 0 ponted out n overestmton of the prmeter ll over the perod consdered, those below 0 ponted out n underestmton. In order to dentfy the stocks whose prmeters were sgnfcntly over or underestmted, sttstcl test of comprson to the 0 vlue ws crred out on the verge bs b for three rsk levels: 1, 5 nd 10 % Men level nd vrblty of prmeters In order to explore the hypothess consderng tht ssessments were bsed due to the vrblty of the prmeters studed, severl sttstcs were clculted for the prmeters estmtes from the ssessment : the men, the stndrd devton nd the coeffcent of vrton. They llowed us to consder the men vlue nd the vrblty of these prmeters over the perod We therefore nlyzed the reltonshp between the verge bs b of prmeter nd ts coeffcent of vrton. Internnul evoluton of prmeters 1. For 5 stocks nd for fshng mortlty, these clcultons were performed over 3 or 4 yers. Dt were mssng for one or two yers over the selected perod. Smlrly, for stocks nd for the spwnng stock bomss, these clcultons were performed over 3 or 4 yers. Dt were mssng for one or two yers over the selected perod. 3

4 In order to check whether ssessments were bsed when prmeters F or SSB presented pronounced trend over the perod, the evolutons of fshng mortlty nd spwnng stock bomss were nlyzed. The evoluton of these prmeters between yer nd yer -1 ws ssessed by exmnng the two followng ndces clled evoluton rtos: F F 1 SSB nd SSB 1 Rtos hgher thn 1 ponted out tht the prmeter ncresed between yer -1 nd yer. Those less thn 1 showed tht t decresed. As for the ssessment rtos, we clculted both e nd sde ndctors for these new ones : 000 Averge evoluton: ( ) 1 e = ln( ) Stndrd devton of evoluton: = 1 sde ( ) ln( ) b ( ) 1 n A vlue below 0 for e ponted out tht on verge the prmeter decresed over the lst fve yers, those bove 0 tht t ncresed. In order to dentfy stocks whch presented sgnfcnt decresng or ncresng trend n the prmeters vlues, sttstcl test of comprson of the verge evoluton e to 0 ws crred out for three rsk levels, 1, 5 nd 10%. 000 Relton between estmton nd evoluton of prmeters In order to exmne the hypothess tht yers of ncresng fshng mortlty or ncresng spwnng stock bomss (recproclly decresng) led to underestmte (to overestmte recproclly) these prmeters, we tred to sought for negtve correlton between the ssessment rtos nd the evoluton rtos. We therefore clculted the slope of the regresson lne between the ssessment rto nd the evoluton rto, nd the coeffcent of determnton R². Ths lst one s equl to 1 the squre of the lner correlton coeffcent of Person between the vrbles nd Clcultons were mde on the totl number of yerly vlble dt, whch rnged between 3 nd 14 ccordng to stocks. To dentfy the stocks for whch there were correltons between the ssessment rto nd the evoluton rto, sttstcl test of sgnfcnce of the Person coeffcent for severl rsks (1, 5 nd 10%) ws crred out, we checked whether the Person coeffcent ws sgnfcntly below

5 Results Relblty of ssessments For the fshng mortlty, b rnged from to 0.44 ccordng to stocks, whch reflected sometmes mportnt verge errors rechng for exmple, -54% for the Southern Hke stock nd 56% for the North Est Artc Sthe stock. The men of ths ndctor ws on the perod. On verge, the fshng mortlty ws underestmted for the studed stocks. For the spwnng stock bomss, b ws lso consderble for some stocks. It rnged from to 0.4 ccordng to stocks. Errors spred out from 70% for the Southern Hke stock to 5% for the North Est Arctc Sthe stock. The vlue of 0.09, estmted s men for ll stocks, showed tht the spwnng stock bomss ws on verge overestmted on the perod. For lrge mjorty of stocks (7%), we observed n underestmton of fshng mortlty nd n overestmton of the spwnng stock bomss (Fg. 1). On verge, dgnostcs were therefore "excessvely optmstc". Logclly, we notced negtve correlton between the two types of bs. The more F ws overestmted, the more SSB ws underestmted nd recproclly. Fgure 1 shows tht verge bs could be mportnt. Only 6 stocks were nsde the 10% crcle of the trget (verge bs less thn 10% for F nd SSB), 11 were between 10 nd 15%, 1 between 5 nd 50%. For 7 stocks, the verge bs ws bove 50%. b(f +1 /F ) F underestmted F overestmted SSB SSB overestmted 4 stocks 3 stocks 50% 3 stocks 5% 10% b(ssb +1 /SSB ) 6 stocks Other Stocks Celtc Plce VIIe Plce Norwy Pout Irsh Sole Fg. 1. Estmte of the verge bses over the perod , for F nd SSB prmeters of 36 demersl stocks. For 1 stocks,.e. 33% of the studed stocks, the fshng mortlty ws bdly estmted over the perod (Fg..), b ws sgnfcntly dfferent from 0 wth 95% of sgnfcnce. For 11 stocks, the fshng mortlty ws underestmted. Moreover, extremely vrble bses were observed for some stocks (hgh sd) reflectng sometmes mportnt but non-systemtc bses n ssessments. It ws the cse for the stocks of VIb Hddock, Megrm South, III Sole nd III Plce. For 1 stocks,.e. 33% of stocks, the spwnng bomss ws bdly estmted over the perod, b ws dfferent from 0 wth 95% of sgnfcnce (Fg. 3.). For 10 stocks, the spwnng stock bomss ws overestmted nd for the others t ws underestmted. 5

6 % 0.4 5% Other stocks b(f +1 /F ) % VIb Hddock Megrm South w -0.4 III Sole sd(f +1 /F ) Fg.. Averge bses over the perod for F ccordng to ts vrblty. The strght lnes represent y = t α/ * x / 5 nd y = - t α/ * x / 5 for α = 1, 5 or 10%. Very vrble bs on the rght of the dotted lne. sd(ssb +1 / SSB ) % 5% 10% Other stocks Megrm North w VI Cod III Sole VIb Hddock sd(ssb +1 / SSB ) Fg. 3. Averge bses over the perod for SSB ccordng to ts vrblty. The strght lnes represent y = t α/ * x / 5 nd y = - t α/ * x / 5 for α = 1, 5 or 10%. Very vrble bs on the rght of the dotted lne. 6

7 Tble : Clssfcton of the 36 stocks ccordng to ther verge bs on F nd SSB nd ther vrblty. The 7 stocks underlned, hd n verge bs bove 50%. Sgnfcnt verge bs on F Sgnfcnt verge bs on SSB NS Cod Bltc Cod E NEA Hddock NS Hddock VI Hddock Hke South Irsh Plce NEA Sthe NS Sole VIIe Sole 10 Averge bs on SSB non sgnfcnt nd very vrble Averge bs on SSB non sgnfcnt nd not very vrble Anglerfsh North b VIId Sole Averge bs on F non sgnfcnt nd very vrble VIb Hddock III Sole Megrm South w III Plce Averge bs on F non sgnfcnt nd not very vrble NEA Cod, VI Whtng VI Cod Megrm North w Celtc Cod Bltc Cod W Hke North Megrm South b NS Plce Norwy Pout VIId Plce VIIe Plce Celtc Plce NS Sthe Sndeel Irsh Sole Bscy Sole Celtc Sole NS Whtng Anglerfsh North p 16 As for fshng mortlty, certn stocks presented extremely vrble bses (hgh sd) reflectng sometmes mportnt but non-systemtc bses n ssessments. It ws the cse for the stocks of VIb Hddock, Megrm North w, III Sole nd VI Cod. The results showed tht stocks cn be clssfed ccordng to the degree of error on F nd on SSB (Tble ). We could thus dentfy 14 stocks, for whch t lest one of the two prmeters presented n verge bs sgnfcntly dfferent from 0. For 10 of these stocks, the two prmeters consdered were regulrly bdly estmted. For 6 stocks, bses were not sgnfcntly dfferent from 0, but the very hgh stndrd error ndcted sometmes lrge nd not systemtc errors n ssessments. Fnlly, only 16 stocks presented non-sgnfcnt bses nd low stndrd errors. Relton between relblty of estmtes nd vrblty of prmeters Among the 1 stocks whose verge bs ws dfferent from 0 wth 95% of sgnfcnce, 6 were found not to hve very vrble fshng mortlty wth CV below 15% (Fg. 4 nd Tble 4). For 4 stocks the ndctors clculted on F dd not revel bses, b ws not sgnfcntly dfferent from 0.7. These stocks showed low vrblty n F, wth CV under 15%. 7

8 Logclly, n cse of hgh vrblty n F, CV bove 0%, no systemtc error on the F estmtes ws found excepted for the Irsh Plce stock. On the other hnd the hgh sd ndctes tht n some yers, mportnt errors cn be done. When fshng mortlty ws reltvely stble (CV below 15%), the proporton of stocks for whch F ws sgnfcntly bsed, 6 stocks out of 13 (.e. 54%), dd not seem dfferent from wht ws generlly observed for ll stocks Stocks 1 Stocks 11 Stocks 0.3 b(f +1 / F ) % 5% 10% 15% 0% 5% 30% 35% 40% 45% CV(F ) Fg. 4. Dstrbuton of verge bses on F ccordng to ts vrblty for the 36 stocks Tble 3: Clssfcton of stocks ccordng to the verge bs on F nd ts vrblty CV > 0 % 15 < CV < 0 % CV< 15 % Sgnfcnt verge bs on F Irsh Plce 1 NEA Hddock NS Hddock Hke South VIId Sole VIIe Sole Anglerfsh North b NS Cod Bltc Cod E VI Hddock NEA Sthe NS Sole 5 6 Averge bs on F non sgnfcnt nd very vrble VIb Hddock Megrm South w III Plce III Sole 4 Averge bs on F non sgnfcnt nd not very vrble Megrm South b Norwy Pout VIId Plce NS Sthe 6 Sndeel Irsh Sole Anglerfsh North p Megrm North w Celtc Plce NS Plce 7 Bscy Sole Celtc Sole NS Whtng VI Cod NEA Cod Bltc Cod W Celtc Cod 7 Hke North VIIe Plce VI Whtng For the SSB, 4 stocks presented n b not dfferent from 0 wth 95% of confdence. For hlf of them, the vrblty of the SSB ws hgh, wth CV bove 15%. On the contrry, for the 1 stocks whose verge bs ws sgnfcnt, the two thrds were those whose vrton on SSB ws hgh, wth CV bove 15% (Fg. 5 nd Tble 5). 8

9 stocks stocks 18 stocks 0.4 b(ssb + 1/SSB) % 5% 10% 15% 0% 5% 30% 35% 40% 45% CV(SSB ) Fg. 5. Dstrbuton of the verge bses on SSB ccordng to ts vrblty for 36 stocks. Tble 4: Clssfcton of stocks ccordng to ther bs on SSB nd ts vrblty. Sgnfcnt verge bs on SSB Non sgnfcnt bs on SSB but very vrble Non sgnfcnt bs on SSB nd not very vrble CV > 0 % NEA Cod NS Cod Bltc Cod E NEA Hddock NS Hddock VI Hddock NS Sole VI Whtng 8 VI Cod VIb Hddock III Sole 3 Anglerfsh North p Bltc Cod W Celtc Cod Norwy Pout III Plce NS Plce Sndeel 7 15 < CV < 0 % Bscy Sole Celtc Sole CV< 15 % Hke South Irsh Plce NEA Sthe VIIe Sole 4 Megrm North w 1 Anglerfsh North b Hke North Megrm South w Megrm South b VIId Plce VIIe Plce Celtc Plce NS Sthe VIId Sole Irsh Sole NS Whtng 11 9

10 For the spwnng stock bomss, results dd not present ny systemtc relton between the relblty of the SSB estmton nd ts vrblty. Errors were not found to be sgnfcnt for hlf of the stocks wth hgh vrton n SSB (CV bove 0%). As prevously, the stocks tht presented dgnostcs vrble n qulty (hgh sd) were those wth vrble level of SSB. The results concernng the SSB, dffered from those relted to F on two ponts: systemtc bses, b sgnfcntly dfferent from 0, tht corresponded to SSB overestmtons were frequently observed for vrble stocks. n contrst, when the SSB ws more stble, CV below 15%, SSB estmtes seemed to be generlly relble (11 stocks out of 16,.e. 69%). Relblty of ssessments nd trend-settng evoluton of the prmeters Regrdng to the evoluton rtos, the men of the e ndctor for F ws 0.01, whch mens tht for the studed stocks F ws on verge stble. On the other hnd the SSB ws decresng for the studed stocks, ths ndctor ws on verge For 19 stocks,.e. 5% of the studed stocks, the fshng mortlty ws underestmted n cse of ncresng trends (Fg. 6). For 3 stocks,.e. 64% of the studed stocks, the spwnng stock bomss ws overestmted n cse of decresng trends (Fg. 7). For hlf of the stocks (5%), we observed n ncrese n F nd reducton n SSB (Fg. 8). We notced negtve correlton between the two types of evoluton. The more F ncresed, the more SSB decresed (nd recproclly). b(f +1 / F ) SSB underestmted SSB overestmted 4 Stocks 10 Stocks Decresng F Incresng F 3 Stocks 19 Stocks e(f +1 / F ) Fg. 6. Averge bses on mortlty F ccordng to ts evoluton over the perod

11 0.6 Decresng SSB Incresng SSB b(ssb +1 / SSB ) F underestmted F overestmted 3 Stocks 6 Stocks 4 Stocks 3 Stocks e(ssb +1 / SSB ) Fg. 7. Averge bses on spwnng bomss SSB ccordng to ts evoluton over the perod Decresng SSB Incresng SSB Incresng F 19 stocks 3 stocks e(f +1 / F ) Decresng F 6 stocks 8 stocks e(ssb +1 / SSB ) Fg. 8. Evoluton over the perod of the prmeters F nd SSB for 36 demersl stocks 11

12 Tble 5: Clssfcton of stocks ccordng to ther verge bs on SSB nd ts evoluton Sgnfcnt verge bs on SSB Averge bs on SSB non sgnfcnt nd very vrble Averge bs on SSB non sgnfcnt nd not very vrble Sgnfcnt decrese n SSB Bltc Cod E VIIe Sole VI Cod VIb Hddock Celtc Plce Bscy Sole No sgnfcnt trend n SSB NEA Cod NS Cod NEA Hddock NS Hddock VI Hddock Hke South Irsh Plce NEA Sthe NS Sole VI Whtng 10 III Sole Megrm North w Anglerfsh North b Hke North Megrm South w Megrm South b VIId Plce VIIe Plce NS Sthe VIId Sole Irsh Sole NS Whtng Celtc Sole Anglerfsh North p Bltc Cod W Celtc Cod Norwy Pout III Plce NS Plce Sndeel 18 However the logrthmc men e ws not sgnfcntly dfferent from 0 for mjorty of stocks, F nd SSB dd not hve sgnfcnt ncresng or decresng trend. For ech stock, the e ndctor ponted out sgnfcnt evoluton n F over the perod Trends were more pronounced for SSB. 6 stocks were dentfed for whch SSB sgnfcntly decresed over the perod (Tble 5). Although relton seemed to pper clerly on the fgures (Fg. 6 nd 7), the sttstcl tests dd not llow us to estblsh ny sgnfcnt relton between the trend-settng evoluton of the F nd SSB prmeters nd the ssessment errors for these prmeters. Processes hve probbly to be regrded here s prelmnres nd wll hve to be tken gn therefter. 1

13 Relton between relblty of ssessments nd nter nnul evolutons of the prmeters For 1 stocks,.e. 33% of stocks (tble 6),we demonstrted negtve correlton between the ssessment rtos nd the evoluton rtos concernng the fshng mortlty. For these stocks, the more the F ncresed, the more underestmted. Ths relton ppered prtculrly cler for the stocks of Celtc Cod, VIb Hddock, North Se Cod nd VIId Sole. For these lttes, the yers of ncrese n F or n SSB (recproclly decrese) resulted n underestmtng (recproclly n overestmtng) F or R respectvely. The dgnoss ws lwys lte on the rel evoluton. For ech stock, the correlton exsted for both prmeters t the sme tme. Tble 6: Clssfcton of stocks ccordng to the verge bs on F nd the exstence of correlton between ssessment rtos nd evoluton rtos for the fshng mortlty Sgnfcnt verge bs on F Averge bs on F non sgnfcnt nd very vrble Averge bs on F non sgnfcnt nd not very vrble Negtve correlton wth t lest 95% of sgnfcnce NS Cod Bltc Cod E VIId Sole 3 VIb Hddock III Sole NEA Cod Bltc Cod W Celtc Cod VIId Plce Sndeel Celtc Sole VI Whtng 7 Negtve correlton between 90 nd 95% of sgnfcnce NS Hddock VI Hddock NEA Sthe NS Sole 4 III Plce 1 Megrm South b 1 Non sgnfcnt negtve correlton Anglerfsh North b NEA Hddock Hke South Irsh Plce VIIe Sole 5 Megrm South w 1 Anglerfsh North p VI Cod Hke North Megrm North w Norwy Pout VIIe Plce Celtc Plce NS Plce NS Sthe Irsh Sole Bscy Sole NS Whtng 1 13

14 Tble 7: Clssfcton of stocks ccordng to the verge bs on SSB nd the exstence of correlton between ssessment nd evoluton of SSB Sgnfcnt verge bs on SSB Averge bs on SBB non sgnfcnt nd very vrble Averge bs on SSB non sgnfcnt nd not very vrble Negtve correlton wth t lest 95% of sgnfcnce VIb Hddock 1 Bscy Sole NS Plce Negtve correlton between 90 nd 95% of sgnfcnce NS Sole 1 III Sole 1 VIId Sole 1 Non sgnfcnt negtve correlton Bltc Cod E NEA Cod NS Cod NEA Hddock NS Hddock VI Hddock Hke South Irsh Plce NEA Sthe VIIe Sole VI Whtng 11 VI Cod Megrm North w Anglerfsh North b Hke North Megrm South w Megrm South b VIId Plce VIIe Plce Celtc Plce NS Sthe Irsh Sole NS Whtng Celtc Sole Anglerfsh North p Bltc Cod W Celtc Cod Norwy Pout III Plce Sndeel 17 14

15 Dscusson The fshng mortlty estmte n stock ssessment s produced by Vrtul Populton Anlyss (VPA). The spwnng stock bomss s lso result produced by the VPA s t s derved from the numbers t ge mtrx. Thus ll the dt used by ths process nclude potentl source of error. The dt for ths prt of the stock ssessment re vrous: ctch t ge mtrx, weght t ge mtrx, totl lndngs, tunng fleet mtrces, mturty mtrx nd nturl mortlty mtrx These dt used by the workng groups for stock ssessment re ncremented ech yer wth the new yer nformton, nd revews re done when new consstent nformton s ccepted. When ths occurs, the chnges occsoned (mnly n the number of ge groups nd n tunng fleets) could be s well cuse of errors n the prmeters estmtes. The results presented here re synthetc overvew tht gther lrge rry of potentl errors. It surely mxed dfferent error sources but gves cler sght of the ssessment relblty for stocks. The method used here emphszes stocks whose prmeters F nd SSB re generlly spekng msestmted. More prtculrly, ths study underlnes the underestmton of the fshng mortlty for lrge prt of the stocks, ledng to n optmstc pont of vew tht could hve gret mpct on mngement optons. The reference prmeters used were tken on the ssumpton tht the ssessment qulty mproves ech yer. Ths s lnked to the convergence property of the VPA (Jones 1961) nd to the fct tht the dt become more ccurte wth the ncrese of stock knowledge. Thus the lst ssessment done could be consdered s more relble thn the prevous ones. The prmeters estmted from ths lst ssessment re the best nd they cn be tken s the truth. Despte these consdertons, dt mprovement s not systemtc ech yer nd prtculrly consderng the dscrds. Gret chnges cn occur n only fve yers regrdng the bologcl percepton of stock, nd the revews mde could hevly ffect the ssessment results. A lck of dscrd dt could led to underestmte the recrutment of the current ssessment nd to overestmte the fshng mortlty frst ges of the next ones. Dscrd dt remn n mportnt ssue on stock ssessment tht could be nfluenced by stock mngement optons s well s economy. For exmple f fsh under commercl sze re bnned from lndngs, dscrd dt become nccessble nd ths wll produce n overestmte of the fshng mortlty n the next ssessments. A hgh vrblty n bs estmtes could be relted to the vrblty n recrutment estmte, tht s known to be hghly vrble prmeter, nd thus the type of errors observed could be ted to the stock prtculrtes. The recrutment estmte hs n effect on the fshng mortlty nd prtculrly on the spwnng stock bomss, ths could crete n mplfcton of the msestmte f t exsts n nconsstency wth the young ges number estmte. For exmple f hgh recrutng estmte s not notced becuse of lck of dscrd dt, ths cn produce n mportnt flse ncrese n the fshng mortlty the next yers, not reflected by the F ndctor. We thus obtn stble F ndctor wth n overestmte of F t young ges nd very low SSB due to the ncrese n F. In ths cse the ssessment would hve led to pessmstc pont of vew corrected the followng yers. The VPA the most used by ICES workng groups s tuned. It requres Ctch Per Unt Effort dt from scentfc nd professonl fleets (tunng fleets). Ths mnly llows to mprove the lst yer fshng mortlty estmte. Ths lst yer estmte s the most unrelble one s no convergence occurred, but t s often the most used of the seres nd n terms of comprson t wll led to greter msestmte thn prevous ones. We here used F tht ws men of the fshng mortlty vector clculted over the ges the more exploted. Ths hd the dvntge not to tke nto ccount too much vrblty n F tht mght be cused by cohort effects. However ths could nduce some nconsstency wth the 15

16 spwnng stock bomss. For exmple chnges n the young ge ctches could hve no mpct on the SSB s these ges re not mture, but ths could ffect the fshng mortlty. The softwre used to compute the VPA hs lot of dfferent optons tht mght ffect the results nd cuse errors. Here we worked on stock ssessment models prmeters. But they re not the only ones, the fshng mortlty nd the spwnng stock bomss re not two ndependent vrbles nd n ncrese n F should produce decrese n SSB n short whle. An underestmton n F obvously leds to n overestmton of the SSB f ll prmeters of the ssessment models re not chnged. The fct tht ths relton ws not systemtclly found could be the occurrence of trnston sttes, where fshng mortlty underestmton dd not mpct yet hevly the populton ge structure. An other rgument could be the evoluton of the other prmeters s recrutment tht could effcently ncrese the SSB. References Gscuel D., Fontenelle G., Mguer C., Bseu A., Consstency of stock ssessment nd evluton of fsheres mngement for Europen fsh stocks, Fshery stock ssessment models. Alsk se grnt college progrm. AK-SG-98-01, pp Jones, R The ssessment of the long term effects of chnges n ger selectvty nd fshng effort. Mr. Res. Scotl. : 19p. Jonsson S.T., nd Hjorlefsson E., 000.-Stock ssessment bs nd vrton nlysed retrospectvely nd ntroducng the PA-resduls. ICES CM 000/X:9. 13 p. 16