Assessment of the SNA 8 stock for the fishing year

Transcription

1 Assessment f the SNA 8 stck fr the 24 5 fishing year New Zealand Fisheries Assessment Reprt 213/28 N.M. Davies, J.R. McKenzie, D.J. Gilbert ISSN (nline) ISBN (nline) May 213

2 Requests fr further cpies shuld be directed t: Publicatins Lgistics Officer Ministry fr Primary Industries PO Bx 2526 WELLINGTON Telephne: Facsimile: This publicatin is als available n the Ministry fr Primary Industries websites at: g t Dcument library/research reprts Crwn Cpyright - Ministry fr Primary Industries.

3 Cntents 1. INTRODUCTION Overview Descriptin f the fishery REVIEW OF THE FISHERY TACC Cmmercial landings Freign fishing Illegal catch Nn-cmmercial catch Recreatinal fisheries Mari fisheries SNA 8 stck assessment infrmatin Stck structure Bimass estimates tagging prgramme tagging prgramme Catch-per-unit-effrt (CPUE) pair trawl Catch-per-unit-effrt (CPUE) single trawl Year class strength Catch-at-age Recruitment indices Mean length-at-age Methd-specific selectivity-at-length Catch-at-length Recreatinal bat ramp samples At-sea pair trawl catch samples Bilgical parameters Stck assessment mdel Changes t the SNA 8 assessment mdel New inputs t the mdel Mdel descriptin Mdel develpments fr an updated assessment assessment mdel specificatins Estimated parameters Mdel predictins Mdel fitting Mnte Carl Markv chain (MCMC) prcedure Prjectins and perfrmance indicatrs Mdel results Mdel pint estimates Relative weight f input data MPD mdel sensitivities MCMC results Cmparisn with 24 assessment SNA 8 stck assessment Mdel estimates fr 1931 t Stchastic prjectins and perfrmance indicatrs Yield estimates Estimatin f MSY Estimatin f MCY... 27

4 5.7.3 Estimatin f CAY MANAGEMENT IMPLICATIONS SNA ACKNOWLEDGMENTS REFERENCES Appendix 1: Full mdel (all parameters and interactins) GLM regressin using SNA 8 single bttm trawl data (catch nm -1tw -1) Appendix 2: Predictr variable cefficients frm stepwise GLM regressin using SNA 8 single bttm trawl data (catch nm -1 tw -1 ) Appendix 3: Mean weight-at-age frm the SNA 8 age-length samples cllected frm 1975 t Sampling design Mean weight-at-age... 72

5 EXECUTIVE SUMMARY Davies, N.M.; McKenzie, J.R.; Gilbert, D.J. (25). Assessment f the SNA 8 stck fr the 24 5 fishing year. New Zealand Fisheries Assessment Reprt 213/ p. The SNA 8 stck was assessed fr the 24 5 fishing year using an updated and revised ppulatin mdel. New data inputs t the mdel since the previus assessment in 24 include: mean length-atage, pair trawl catch length frequencies measured at sea, recreatinal catch length frequencies, and single trawl and pair trawl catch-at-age in the 197s. Data updated since the 24 assessment were: catch weights fr the cmmercial fishery in 24 (i.e. 23 4; fr cnvenience fishing years are referred t by the January year); single trawl and pair trawl catch-at-age estimates 24; a revised time series f single trawl standardised CPUE time series The single trawl CPUE was revised by excluding zer catch recrds and including nly vessels that caught a high prprtin f the catch. The indices shwed a shallw decreasing trend frm 1996 t 21 fllwed by a general increase t 24. The data series t which the assessment mdel was fitted included: abslute bimass and ppulatin length cmpsitin derived frm tag-recapture prgrammes in 199 and 22; CPUE time series fr pair trawl and single trawl ; catch-at-age estimates fr pair trawl 1974 t 24 (nt all years), and single trawl 1974 t 1976, and 1991 t 24; research trawl survey recruitment indices fr 1984 t 1997 year classes (excluding 199 and 1995), bat ramp samples f recreatinal catch length frequency in 1991, 1994, 1996, and 2, and pair trawl at-sea catch length frequency in Structural develpments made t the mdel used fr the 24 assessment included: applying fishing mrtality via length-specific selectivity gives; estimating single trawl and pair trawl selectivity-atlength; and annual variability in individual grwth rate. Fr the updated assessment these develpments were extended. The series f bat ramp samples f recreatinal catch length frequency facilitated the estimatin f a stepwise change in the recreatinal selectivity-at-length that takes accunt f the regulated change in minimum legal size fr that fishery in The single trawl and pair trawl catch-at-age data frm the 197s facilitated the estimatin f natural mrtality. Observed mean length-at-age frm the 197s was cnsiderably lwer than in the recent data ( ). Cnsequently, time-varying mean length-at-age was assumed, with slw grwth frm 1931 t 1979, average grwth frm 198 t 1988, and rapid, variable grwth frm 1989 t 24. The statistical assumptins made t reslve cnflicts between bservatin types were examined by testing the sensitivity f the mdel t the relative weight f the tag-recapture abslute abundance estimates. Sensitivity t the assumed histrical cmmercial catch was tested using an ptin with a nnequilibrium initial ppulatin starting in 1974 (ptin: Y1974), and sensitivity t histrical annual recreatinal catch was tested with recent catches f either 3 t r 6 t (ptins: R3 and R6, respectively). A cnsistently gd MPD fit was btained ver the full range f bservatins, with the exceptin f a part f the tag-recapture ppulatin length cmpsitin in 22 and the recreatinal bat ramp length frequency samples. The tag-recapture estimate f ppulatin length cmpsitin in 22 fr lengths ver 55 cm appeared t cnflict with ther data. Mdel ptins assigning high relative weight t this part f the bserved ppulatin length distributin prduced wrse fits t the single trawl CPUE, which indicates increases in recent bimass. Given this cnflict and the level f uncertainty f the ver 55 cm cmpnent f the tag-recapture estimate, the Snapper Wrking Grup cnsidered it reasnable t assign lwer weight t this bservatin. The fit t the bat ramp length frequency samples was pr fr the small length class intervals (25 31 cm) in sme years. This appeared t be independent f the step-wise change in the selectivity-at-length, and may reflect high prcess errr fr this fishery (e.g. annual changes in selectivity). Ministry fr Primary Industries SNA 8 stck assessment

6 It is unlikely that the Bayesian psterir distributin estimate f unfished equilibrium bimass (B ) fr any single ptin captures the full uncertainty f this parameter fr the fllwing reasns. Firstly, the credibility intervals are sensitive t the assumptins made regarding histrical catches as indicated by the Y1974 and R6 mdel ptins. Secndly, instead f estimating natural mrtality (M) as an independent parameter, the marginal psterir distributins fr all the mdels were calculated under a cnstant value fr natural mrtality equivalent t that estimated frm the MPD fit. This was dne t address a technical prblem with the Bayesian estimatin prcedure that is recmmended t be reslved fr future assessments. Thirdly, the mdel bimass must remain abve zer in the mid 198s while als maintaining a fit t the lw abslute abundance estimates frm the later tagging prgrammes. It is therefre cnstrained abve and belw, and hence the credibility intervals fr B fr each ptin are narrw. An indicatin f the true uncertainty in B is given by the range estimated ver all ptins. In cntrast, MPD estimates f B 4 were relatively insensitive t the assumptins tested, and were within a relatively narrw range f 9.6 t 11.5 kt, with B 4 /B having a range f 8.7% t 9.8%. Fr the tw ptins selected by the Snapper Wrking Grup, R3 and R6, B 4 /B was 9.5% and 9.8% respectively, and the 9% credibility intervals ranged between 7.8% and 12.5% f B. Under all the mdel ptins and future cmmercial catch scenaris (5 15 t) the bimass is predicted in stchastic mdel prjectins t increase n average. The rate f increase is cnsiderably lwer fr the ptins fr which a cnstant explitatin rate fr the recreatinal fishery is assumed. Similarly, prjectins were less ptimistic fr the ptin assuming higher histrical recreatinal catch. Under the current TACC, the best perfrming was the R3 ptin under cnstant recreatinal catch, where the prbability f achieving 2%B by 225 is almst 5%. Fr all the mdel ptins the rate f increase is substantially imprved with a reductin in the TACC, but this is strngly ffset by increases in recreatinal catch where a cnstant explitatin rate is assumed fr that fishery. This reprt presents the findings as f 27, and nt at the current time f the reprt s publicatin. N updated infrmatin has been added t the assessment, and nne f the interpretatins have been revised since 27. This wrk was funded by the Ministry f Fisheries under the research prject SNA24/1 Objective 3. 2 SNA 8 stck assessment 24 5 Ministry fr Primary Industries

7 1. INTRODUCTION 1.1 Overview This reprt presents the 25 assessment f the west cast snapper stck, SNA 8, summarised by Sullivan et al. (25). This wrk was carried ut under Specific Objective 3 f the Ministry f Fisheries Prject SNA24/1: T update the assessment f SNA Descriptin f the fishery The SNA 8 fishery has been described by Sullivan (1985), Paul & Sullivan (1988), Davies (1997, 1999) and Davies et al. (1999). The trawl fishery develped frm a small fleet perating frm sheltered harburs with annual landings nt exceeding 1 t until the 195s, when larger Aucklandbased trawlers entered the fleet. Landings frm the trawl fishery gradually increased and, tgether with the intrductin f freign vessels during the 196s, were exceeding 2 t per year by 1973 (Table 1). During the 197s there was a rapid increase in landings t ver 3 t as a result f the transfer f trawl fishing effrt frm the east t the west cast, and the intrductin f pair trawling in It is estimated that ttal landings frm a cmbinatin f pair trawling and Japanese fishing peratins increased t a peak f abut 76 t in After the establishment f the Exclusive Ecnmic Zne (EEZ) in 1978, freign fishing was excluded frm SNA 8. Landings frm the pair trawl fishery declined t abut 3 t by 198. Annual landings cntinued t decline t abut 18 t in With the implementatin f the Quta Management System in , cmmercial catches frm SNA 8 were cnstrained by a ttal allwable cmmercial catch (TACC) f 133 t. In landings were nly 9 t, but they have been very clse t the TACC since then (Table 2). Since 1989 there has been a steady decline in the prprtin f ttal landings derived frm pair trawling with a shift in emphasis t single trawling. In the 23 4 fishing year single trawl vessels accunted fr 79% f ttal landings. There has been a shift t single trawling since the 198s in respnse t market demands fr higher quality snapper, necessitating shrter tws and mre careful fish handling practices that maintain the value f the catch. Histrically, the west cast snapper fishery has been seasnal, with mst f the annual catch ccurring during the spring and summer when fish aggregate fr spawning. This pattern has nt altered markedly in recent years. 2. REVIEW OF THE FISHERY 2.1 TACC The TACC fr SNA 8 was set at 133 t in t permit stck rebuilding and increased t 1594 t by 199 as a result f decisins made by the Quta Appeal Authrity (Table 2). Frm 1 Octber 1992 the TACC was reduced frm 1594 t t 15 t and has nt been changed since. Landings in SNA 8 were less than the TACC in but have increased t clsely match and exceed the TACC in recent years (Table 2). The cnsistent verrun since may be related t the lw deemed value f SNA 8 fish relative t ther snapper stcks. Ministry fr Primary Industries SNA 8 stck assessment

8 2.2 Cmmercial landings A change in the dminant fishing methd in SNA 8 has ccurred since Trawl landings make up n average 95% f ttal annual catch: single trawl has becme mre dminant in recent years and pair trawling less s. Estimates f the ttal reprted cmmercial landings fr SNA 8 are available by calendar year frm 1931 (see Table 1), with reprted freign catches fr (Gilbert & Sullivan 1994) Freign fishing Japanese catch recrds and bservatins by New Zealand naval vessels indicate that significant quantities f snapper were taken frm New Zealand waters frm the late 195s until There are insufficient data t quantify the Japanese catches. Hwever, trawl catches have been reprted by area frm 1967 t 1977, and lngline catches frm 1975 t 1977 (Table 3). The data series is incmplete, particularly fr lngline catches Illegal catch N infrmatin is available t estimate illegal catch. 2.3 Nn-cmmercial catch Recreatinal fisheries The 1987 Natinal Marine Recreatinal Fishing Survey shwed that snapper was the mst imprtant finfish species sught by recreatinal fishers. An estimate f recreatinal catch is available frm the 199 tagging prgramme fr the whle f SNA 8 (239 t, Table 4). The results f telephne and diary surveys in the Central ( ) and Nrth ( ) Fisheries Management Areas estimated recreatinal catch in SNA 8 t be between 3 and 42 t in 1994 (Teirney et al. 1997). This was later revised t be 238 t fllwing a review f the reginal and natinal surveys (Bradfrd 1999). The estimates f recreatinal catch in SNA 8 frm the natinal surveys were 24 t in 1996 (Bradfrd 1998), 661 t in 2 and 1133 t in 21 (Table 4). The latter tw surveys applied a different interview frmat, and the last survey had largely new diarists but used sme parameter estimates frm the 2 telephne survey Mari fisheries Snapper is an imprtant species fr Mari, but the annual catch is nt knwn. 3. SNA 8 stck assessment infrmatin 3.1 Stck structure N infrmatin that wuld alter the accepted stck bundaries f the west cast snapper stck (SNA 8) has becme available since the 24 assessment. SNA 8 is assumed t be separate frm the ther six snapper stcks with bundaries defined by the SNA 8 Quta Management Area (Mana 4 SNA 8 stck assessment 24 5 Ministry fr Primary Industries

9 Island t Nrth Cape). Fr the age-structured mdel, the SNA 8 stck was assumed t be discrete, i.e., n emigratin r immigratin f fish acrss stck bundaries. Recapture sample length, age cmpsitins, and mvements f recaptured fish frm the 22 tagging prgramme have been reviewed (Walsh et al. 26). There were sme spatial differences in length and age cmpsitin between areas, and this was supprted by the bservatin that mvements f tagged fish were relatively lcalised (almst all mvements were less than 5 n. miles within a seasn). Althugh differences in ppulatin length and age cmpsitin were evident at the nrthern and suthern bundaries f SNA 8 relative t central areas (Figure 1), these differences were nt attributed t large-scale mixing with adjacent snapper stcks (Walsh et al. 26). These results supprt the assumptin made f the SNA 8 ppulatin cmprising a relatively discrete unit stck. 3.2 Bimass estimates tagging prgramme An estimate f the abslute bimass f snapper recruited t the SNA 8 fishery is available frm the results f a tag-recapture study in 199. Details f the analysis, including crrectins made fr initial mrtality and grwth, have been described by Davies et al. (1999). The estimate f 955 t was input int the 1999 and 24 assessment mdels. The simulatin estimates f precisin were cnsidered cnservative (c.v. less than.1) and c.v.s f.18 and.3 have been assumed in the updated assessment presented here. The tag-recapture estimates f abslute numbers in each f five length strata were prjected ver 1 cm length intervals by Davies et al. (26). This is expressed as ppulatin prprtins-at-length in Figure tagging prgramme A tag-recapture study was carried ut in SNA 8 in 22 and a descriptin f the cryptic Passive Integrated Transpnder (PIT) tag technlgy used is prvided by McKenzie et al. (26), and the estimatr used t derive ppulatin parameters is described by Gilbert et al. (25). The abslute bimass estimate was t (c.v..12). The ppulatin prprtins-at-length are presented in Figure 2. Variances f the numbers-at-length were derived frm the estimated Hessian matrix, and the c.v.s ranged frm.24 at 25 cm, t a minimum f.6 at 35 cm, and.3 at 66 cm Catch-per-unit-effrt (CPUE) pair trawl Vignaux (1993) presents a time series f pair trawl abundance indices frm 1974 t 1991 that have been an input t SNA 8 assessment mdels in the past ten years. The time series indicates a rapid decline in catch rates frm the mid 197s t mid 198s with increases ccurring fllwing the intrductin f the Quta Management System in Catch-per-unit-effrt (CPUE) single trawl The time series f trawl catch and effrt data recrded n Ministry Catch Effrt Landing Return (CELR) and Trawl Catch Effrt Prcessing Return (TCEPR) begins in September Abundance indices fr nine fishing years ( t 23 4) were derived using this data. This updates the Ministry fr Primary Industries SNA 8 stck assessment

10 analysis carried ut fr the 24 stck assessment with imprved assumptins made regarding the cre vessels included General linear mdel (GLM) The mdel was described by Davies et al. (26) fr the 24 assessment and nly a brief utline fllws. The standardised CPUE analysis used is adapted frm Francis (1999). A general linear mdel prcedure (Equatin 1) was used t derive standardised CPUE abundance indexes fr SNA 8, lg( CPUE ) i j Y j I ij k A k J ik PV i i where Y j is the year cefficient in the jth year, A k is the area cefficient in the kth area, P is the vessel attribute cefficient, V i is the vessel attribute fr the ith bservatin, and i is the errr n the ith bservatin. I ij = 1 if bservatin i ccurs in year j, and zer therwise. J ik = 1 if bservatin i ccurs in area k, and zer therwise. The cefficients Y and A in Equatin 1 represent deviatins frm an arbitrary reference year (r) and area (s) (typically the first year and area in the respective series), therefre Y r = and A s =. The standardised regressin prcedures were then used t derive estimates ( Yˆ j, Â k, Pˆ ) f the cefficients Y j, A k and P. The CPUE index was derived by expnentiating the year cefficient, y j = exp( Yˆ ). j The expected value f y r, the reference year index, is always 1 and therefre the variance is. The variance n the ther CPUE indices (Var(y j )) represents the variatin in the expected value f y j relative t the expected value in the reference year (1). The variance estimates fr each year index will alter depending upn the chice f reference year. In rder t derive variance estimates fr each year index independent f the thers, it is necessary t cnvert the year indices t their cannical frm (Francis 1999). The cannical variances reflect hw well the GLM mdel fits the variatin in the CPUE data. The GLM variances d nt necessarily represent the true variability f the y j as indices prprtinal t bimass, i.e., there may be changes in catchability between years that increase variability arund a prprtinal relatinship. The GLM variances will usually underestimate the true index variance (Francis 1999). A prblem with the GLM apprach arises when zer values are present in the catch data. Althugh a zer catch per unit effrt bservatin can be legitimate, zer is impssible in GLM lg-space. Tw ptins fr dealing with zer values are either t remve them frm the data, r t add a very small value t the bservatin t allw its inclusin. Bth ptins may bias the CPUE index. Because there were relatively few zer values in the SNA 8 CPUE data where snapper was the target species, the apprach taken was t remve these catch recrds frm the analysis GLM fitting criteria The fllwing variables were fitted in the GLM mdels. Year: Mnth: Snapper target: as fishing-year (Octber t September); categrical January t December; categrical yes, n; categrical 6 SNA 8 stck assessment 24 5 Ministry fr Primary Industries

11 Statistical area: Vessel ID: 39, 4, 41, 42, 45, 46, 47; categrical categrical All terms were first entered int the mdel, including all pssible cmbinatins f interactin terms, t estimate the degree f interactin between the year term and the ther predictr variables. A stepwise fitting prcedure was then implemented; a GLM was run fitting first t the year parameter. Statistically significant main effects terms were then sequentially added int the mdel fllwed by statistically significant interactin terms nt invlving the year parameter. Predictr variables were accepted int the mdel nly if they explained at least.5% additinal deviance SNA 8 CPUE data The TCEPR reprting frms require fishers t prvide infrmatin n individual tws, whereas catch data recrded n CELR frms is amalgamated fr each 24-hur perid, i.e., ttal daily catch (in kilgrams), ttal number f tws. Althugh mst SNA 8 trawl skippers switched t the TCEPR frms after 1994, a significant prprtin f SNA 8 CPUE data cllected befre 1994 exists n CELR frms. Fr the 24 assessment, the unit f CPUE was mdified t use an imprved definitin f fishing effrt (Davies et al. 26). Fishing effrt was defined fr each tw in terms f distance (nautical mile), i.e., duratin tw speed. Thus, CPUE was defined as kg n. mile -1. This definitin precluded using the CELR data and TCEPR data fr the early part f the series (pre-1996). Because f the relatively lw number f bservatins in recent years it was cnsidered reasnable t exclude recent pair trawl CPUE data frm the updated analysis. A grming prcess was fllwed that resulted in the remval f unsuitable r sparse recrds frm the cmplete data set (31 66 recrds). Based n the distributins f tw-speed (kt) and twduratin, recrds were excluded if they cntained fields with values that fell utside the tails f the distributins. CELR recrds were ignred, and TCEPR recrds prir t the fishing year are scant, and were therefre excluded. The paucity f recrds available relating t year, mnth and target fr statistical areas: 37, 43, 44, and 48 necessitated the remval f these statistical areas frm the dataset. The fishing pwer f the single trawl fleet increased in the mid 199s with the additin f new and larger vessels. T remve the ptential fr bias caused by differences between vessels ver the time series, vessels that did nt cntribute appreciably t the fishery were excluded frm the GLM. Vessels were retained that displayed cntinuity and lngevity f fishing peratins, and that accunted fr a high prprtin f catch and effrt thrughut Tw ptins fr a vessel criterin were cnsidered: retaining the tp 12, r tp 2 vessels. These criteria significantly reduced the incidences f zer catches. After grming, 21 5 and recrds remained fr the tp 2 r tp 12 vessel ptins, respectively Results The data used fr the 24 assessment was updated t 23 4 and the GLM was undertaken using the same variables as in 24. The data shwed a decreasing trend in the prprtin f zer catches. Varius methds were attempted t include this infrmatin, such as adding a cnstant t the zer catches, r using a cmbined mdel where the prbability f a zer catch was mdelled as a binmial distributin and then the binmial year cefficients were multiplied by the lgnrmal year cefficients. The frmer apprach resulted in unacceptable mdel diagnstics, and fr the latter apprach, the change t the abundance indices was relatively minr cmpared t thse derived frm using the psitive catch data nly. This cnfirmed the decisin made by the Snapper Wrking Grup t exclude zer catch data in the GLM. Ministry fr Primary Industries SNA 8 stck assessment

12 The full GLM based n catch-per-distance-twed prduced significant year interactin terms with target and statistical area predictr variables (Appendix 1), and althugh these explained less than 2% additinal deviance, sme cautin applies. This effectively means that there is a different annual index at each level f the ther predictr variables, e.g. each statistical area has a different abundance signal. It shuld be nted that the year index derived frm the subsequent stepwise GLM (which did nt include year interactin parameters) is effectively a data weighted average ver all the ther predictr variables. N predictr variables were excluded frm the mdel by the AIC stepwise regressin prcedure (Appendix 2). The year indices frm the stepwise GLM fr catch-per-distance-twed fr bth the tp 12 and tp 2 vessel ptins are shwn in Figure 3. The indices shwed a shallw decreasing trend frm 1996 t 21 fllwed by a general increase t 24. These cre vessel analyses were cnsidered mre apprpriate than thse used t generate the 24 series. There was virtually n difference between the year indices based n the data frm the tp 2 r the tp 12 vessels. The series based n the tp 12 vessels was used. Cannical variance estimates were typically lw (Davies & McKenzie 21) with c.v.s generally less than.5 (Table 5). 3.3 Year class strength Catch-at-age Length frequency and tlith samples have been cllected annually frm single trawl and pair trawl landings in SNA 8 since the 1989 fishing year. Catch-at-age data were available fr the fllwing methds and years: - pair trawl 1989, 199, single trawl Details f sample sizes, prprtin at length distributins in catches, and the distributin f length-atage cntained in the age-length keys were presented by Davies & Walsh (1995),Walsh et al. (1995, 1997, 1998, 1999, 2, 21, 22, 23, 24), and Walsh & Davies (24). The data input t the 24 SNA 8 assessment was updated with the single trawl and pair trawl catchat-age estimates fr 24 (Figure 4). Relatively strng recruitments are apparent fr the 1996 and 1998 year classes, hwever, the 1999 and 2 year classes appear t be f average strength (Walsh & Davies 24). Fish in the aggregate age class (2+ years) are virtually absent. A significant lg-lg transfrmed relatinship was fund using a functinal regressin between the c.v.s f trawl prprtins at age and the prprtin at age estimates fr 1989 t 1996 (Davies 1997). This relatinship was updated t include the mst recent data and used t describe the bservatin errr f the trawl prprtins at age fr the years where variance estimates were unavailable (1986 and 1987). Cefficients f variatin fr the years 1989 t 24 were available Histrical Fr the 24 assessment, catch-at-age estimates fr the pair trawl fishery in 1975, 1976, and 1979 were input t the mdel (Table 6). With this infrmatin it was pssible t btain a mdel estimate f natural mrtality (Davies et al. 26). This develpment prmpted a detailed examinatin f the length and age samples cllected during the 197s t add t these histrical data and t ptentially imprve upn the natural mrtality estimates fr the updated assessment. 8 SNA 8 stck assessment 24 5 Ministry fr Primary Industries

13 The sampling design used in the 197s differed frm that emplyed currently fr SNA 8, with direct randm sampling fr age frequency. Generally a fixed number f fish were randmly sampled frm each f a randm selectin f bins within each landing. An estimate f the landing catch-at-age was calculated by scaling up each bin sample frequency t the number f fish in each bin, and then scaling up t the number f bins in the landing. Annual catch-at-age was the mean f the seasnal estimates scaled t the landed weights in each seasn. This was t maintain cnsistency with the recent data having a seasnal distributin where little f the annual catch is taken frm the winter mnths, which are nt sampled. Seasns cmprising a year were defined as spring t winter. The prtcl fr ageing tlith samples frm these landings differed frm that used fr the recent catch-at-age data. T determine the cnsistency between the series, a subset f samples frm the histrical data was aged using the current prtcl. Samples frm fur pair trawl landings (392 tliths) cllected in winter 1974 were aged, and the readings cmpared with thse btained previusly. Fr ages 1 t 1 years 17% f the readings altered, and fr ages lder than 1 years abut 5% f the readings altered. Mst ften the differences were by ne r tw years with the majrity f the riginal readings being lwer. Despite this, the age distributin derived frm the tw readings shwed similar patterns f year class strength (Figure 5). Catch-at-age data were available fr the fllwing methds and years: - pair trawl , ; and - single trawl These estimates indicate that catches in the mid t late 197s cntained a high prprtin f fish ver 19 years f age (Figure 6) Recruitment indices Research trawl survey estimates f recruitment indices are available fr the year classes 1984 t 1997, excluding the 199 and 1995 year classes (Table 7, taken frm Mrrisn & Parkinsn 21). This updates the indices input t the previus assessment with the additinal indices fr the 1996 and 1997 year classes. 3.4 Mean length-at-age Davies et al. (23) described mean length-at-age data fr SNA 8 (1989 t 1998), and shwed interannual variatin in grwth with cnsequent deviatins in annual mean length-at-age frm that predicted frm vn Bertalanffy grwth functins used in SNA 8 assessments prir t 24 (Table 8). In mst years, estimates were lwer than the predicted values, apart frm , i.e. the year frm which data was used t estimate the published vn Bertalanffy functin (McKenzie et al. 1992). Given the imprtance f this parameter fr mdel estimates f ppulatin bimass and catch weights, the available data fr bserved mean length-at-age was used in the mdel fr the 24 SNA 8 stck assessment. This data has been updated t include 23 4, and supplemented with bservatins frm length and age samples cllected frm 1974 t 198 (see Sectin ). The length and age sampling design used in single and pair trawl landings frm 1974 t198 has been described. The analysis f mean length-at-age is described in Appendix 3. Data frm single trawl and pair trawl landings cllected during the spring and summer mnths were used fr calculating mean length-at-age fr cnsistency with data cllected in the recent years. An apprach was develped t take accunt f inadequate sampling f the lder age classes (Davies et al. 23). There is scant data fr age classes 11 years and lder, and therefre grwth f these age classes has prbably nt been well bserved. Fr the years having bservatins, the mean length-atage f fish aged 3 t 1 years was calculated directly frm the length-at-age data. The pled 1989 Ministry fr Primary Industries SNA 8 stck assessment

14 24 length-at-age data was used t derive a vn Bertalanffy grwth curve, frm which mean lengthat-age was predicted fr ages 11 t 19 years. Similarly, the pled data frm was used fr predictins in that perid. Fr the respective perids, the mean length-at-age f the 2+ age class in the cmpnent years was derived directly frm the pled length-at-age data fr that age class. The mean length-at-age fr has been described by Davies et al. (26). The mean vn Bertalanffy grwth functin estimates fr the tw perids and are presented in Table 8, and a cmparisn f the mean size-at-age expressed in terms f weight-at-age is presented in Figure 7. Grwth rates are relatively similar fr ages 3 t 6 years; hwever there is large divergence fr the lder age classes. Slwer grwth is clearly apparent frm samples cllected in the 197s relative t the recent perid, as indicated by the lwer L inf estimate: cmpared t Observatins frm 24, althugh similar t thse frm the 197s fr ages 3 t 7 years, are higher fr age 8 t 1 years, and the bservatins fr lder age classes are cnsistent with the mean vn Bertalanffy functin fr the recent perid. It may be speculated that the slwer grwth rates fr lder age classes bserved in the 197s may be attributable t either density-dependence (Rse et al. 21) r temperature related effects (Davies et al. 22). The 24 assessment mdel f SNA 8 predicted high stck bimass until the 196s, when it was arund 73% f unfished bimass (Davies et al. 26). In cntrast, estimates f bimass in the 198s were cnsiderably lwer, at arund 8% f unfished bimass. This large difference wuld cause faster recent grwth if grwth were density-dependent. Similarly, cler sea surface temperatures arund New Zealand pre-197 wuld have prduced slwer grwth rates if grwth were temperature dependent. It was therefre reasnable t assume a grwth rate fr the perid befre 1975 that was slwer than in recent years. Fr perids withut bservatins ( , and ), mean length-at-age was assumed. A cnstant mean length-at-age vectr fr ages 3 19 years was assumed fr the years and that was derived frm the vn Bertalanffy curve fitted t bservatins frm the perid The mean bserved length-at-age 2+ was assumed. A cnstant vectr, being the mean f the vn Bertalanffy functins fr the tw perids and , was assumed fr the perid The mean f the bserved mean length-at-age 2+ fr the tw perids was assumed. 3.5 Methd-specific selectivity-at-length Estimates f methd-specific selectivity patterns fr SNA 8 were available frm the 199 and 22 tag-recapture studies and were applied in the 24 assessment (Davies et al. 26). Selectivity-atlength parameters fr the pair and single trawl methds were estimated independently in the analysis f the 22 tag-recapture prgramme, using a duble-nrmal functin (Gilbert et al. 25). The 199 estimate f single trawl selectivity indicates higher selectivity fr small fish (less than 3 cm, Figure 8). These estimates prvide independent infrmatin n methd-specific selectivity that has been cnsidered in specifying the prirs fr Bayesian estimatin f selectivity parameters (see Sectin ). 3.6 Catch-at-length Recreatinal bat ramp samples Recreatinal length frequency data are available frm fur bat ramp surveys carried ut in 1991, 1994, 1996 and 2 that include ramps servicing the SNA 8 recreatinal fishery. The first survey, cnducted almst entirely in the first six mnths f 1991, aimed t cllect baseline infrmatin n recreatinal fisheries perating in the Auckland Fisheries Management Area. The survey cnducted in the first six mnths f 1994, examined the validity f mean fish weights reprted by diarists, as part f a reginal telephne/diary survey. The first natinal telephne diary survey was cnducted in 1996, 1 SNA 8 stck assessment 24 5 Ministry fr Primary Industries

15 and a natinal bat ramp survey tk place which estimated mean fish weight fr the recreatinal fishery. A similar survey was cnducted in Mst interviewing has taken place at bat ramps clse t ppulatin centres, predminantly arund Auckland. These catch-at-length data were input t the updated assessment mdel, t allw the estimatin f recreatinal selectivity-at-length. The methd-specific, tempral and spatial cmpsitin f the recreatinal fishery was examined in evaluating the representativeness f bat ramp length frequencies. Byd et al. (24) summarised the recreatinal snapper catch frm SNA 8 using diary data frm the survey, indicating that 8% was taken by trailer bats and dinghies, and arund 8% frm shre-based methds. Althugh frm nly a single year, this infrmatin suggested that bat ramp samples are reasnably representative f recreatinal catches. Hartill et al. (1998) present the balanced design emplyed fr sampling bat ramps with respect t weekdays and weekends. Hwever, the surveys were nt cnducted thrughut the year in sme years (1991 and 1994). Fr the annual surveys the tempral distributin f samples was cmpared t the catch reprted in the diary surveys, revealing similar seasnality with arund 9% f the catch taken frm December t April (B. Hartill, NIWA pers. cmm.). Recreatinal fishers perate either inside r utside f the large west cast harburs. Interviewers recrded the fishing lcatin assciated with each sample that enabled a cmparisn f the lcatin-specific length frequencies. A clear difference was evident with smaller fish taken inside the harburs that was cnsistent in all fur surveys. Diary data cllected cncurrently was used t define the prprtin f annual catch taken frm each lcatin. This was cmpared t the bat ramp prprtins, and fund t be incnsistent within and between years (B. Hartill, NIWA pers. cmm.). Therefre, the lcatin-specific bat ramp length frequencies were scaled by the diarist prprtins f lcatin-specific annual catch t prduce annual estimates f the recreatinal catch length frequency (Figure 9). N diary survey was cnducted in 1991, s the spatial distributin f the catch determined frm the diary survey in 1994 was assumed fr that year. The effects f the increase in the minimum legal size (MLS) frm 25 cm t 27 cm fr the recreatinal fishery in Octber 1994 is evident in the length frequency distributins At-sea pair trawl catch samples Past regulatin changes in SNA 8 are likely t have affected selectivity patterns f the trawl fishery. In 1986, a zne excluding trawlers was impsed fr the waters within ne nautical mile f the cast fr a substantial part f the castline. In the same year regulatins were implemented fr an increase in the cd-end mesh size frm 1 mm t 125 mm. An at-sea catch sampling survey was carried ut at the time t cllect pair trawl length frequencies frm catches using bth cd-end mesh sizes (authrs unpublished data). As expected the larger mesh size reduced the relative selectivity-at-length fr snapper in the small length intervals (under 35 cm, Figure 1). Althugh derived frm a small number f trawl shts, this infrmatin was used t examine the implicatins f the regulatin changes n the relative selectivity-at-length f the trawl methds, with a view t estimating stepwise changes in selectivities. The mean catch-at-length distributin (Figure 1) prvided an bservatin f the catch cmpsitin fr the pair trawl methd in Bilgical parameters The bilgical parameters used in this and previus assessments are given in Table 8 (Gilbert & Sullivan 1994, Davies 1997, 1999, Davies & McKenzie 21, Davies et al. 26). The previus value fr natural mrtality f.75 was used t specify a prir distributin fr this parameter even thugh it had been estimated frm essentially the 197s catch-at-age data used here (see Sectin ). The previus vn Bertalanffy parameters are given fr cmparisn (see Sectin 3.4). Ministry fr Primary Industries SNA 8 stck assessment

16 4. Stck assessment mdel 4.1 Changes t the SNA 8 assessment mdel Bimass and yield were estimated using a revised and updated assessment based n an age-structured ppulatin mdel with length-specific fishing mrtality, develped using the CASAL mdelling framewrk (Bull et al. 24). The mdel was fitted t tw tag-recapture ppulatin estimates, a single trawl CPUE time series, trawl survey recruitment indices, catch-at-length and catch-at-age data. The updated mdel includes a number f new features advanced frm the mdel used in the 24 stck assessment, and these are utlined belw New inputs t the mdel In updating the 24 SNA 8 assessment mdel t 25, in additin t updating the existing bservatins input t the mdel, a range f revised and new bservatins were input including mean length-at-age, pair trawl catch length frequencies measured at sea, recreatinal catch length frequencies, and single trawl and pair trawl catch-at-age in the 197s Updated bservatins Observatins frm SNA 8 t which the assessment mdel is fitted were updated frm that used in the 24 assessment. Updated bservatins included: - catch weights fr the cmmercial fishery in 23 4; - single trawl and pair trawl catch-at-age estimates 23 4; - single trawl standardised CPUE time series Mean length-at-age Sectin 3.4 abve describes the available mean length-at-age data fr SNA 8. Annual variability in mean weight-at-age in the recent time perid ( ) was high, with up t 35% difference between years fr 8-year-ld fish. Grwth rates f age classes 6 years and lder was cnsiderably slwer in the 197s cmpared t the recent perid. The mdel used annual mean length-at-age estimates with a cnstant average bservatin errr (c.v..8) Pair trawl catch-at-length Pair trawl catch-at-length frequencies measured at sea were input int the mdel with a view t estimating pre- and pst-1986 selectivities Recreatinal catch-at-length The 24 assessment mdel assumed cnstant selectivity-at-age fr the recreatinal methd assuming an give equivalent t the 22 tag-recapture estimate f single trawl selectivity-at-length (Davies et al. 26). The bat ramp catch-at-length distributins fr the recreatinal methd were input int the mdel t allw the estimatin f a stepwise change in selectivity crrespnding t the change in MLS fr the fishery in SNA 8 stck assessment 24 5 Ministry fr Primary Industries

17 Trawl catch-at-age The histrical pair trawl catch-at-age data (frm 1975, 1976, and 1979) was amended t include ther years fr which data was available and t add data frm single trawl catches. This expanded the pair trawl series t cver the years 1974, 1975, 1976, 1978, 1979, and 198, and added a single trawl series fr the years 1974, 1975, and This imprved upn the infrmatin available n the prprtin f catches f fish in the age class 19 years and lder. 4.2 Mdel descriptin The underlying mdel structure has been described by Davies et al. (26) fr the 24 SNA 8 stck assessment, s nly a brief utline fllws. Details f the CASAL mdel frmulatins used are described by Bull et al. (24). The mdel assumes a single stck with discrete age classes 3 t 2 years; the final age class is an aggregate fr fish lder than 19 years. Recruitment is defined as the number f 3 year-ld fish entering the ppulatin each year, and n relatinship exists between spawning stck and recruitment. A single sex ppulatin is assumed as grwth in snapper is nt sexually dimrphic and sex ratis are generally even (Paul 1976). Natural mrtality (M) is assumed t be cnstant fr all ages. Methd- and length-specific fishing mrtalities were calculated fr fur fishing methds (single trawl, pair trawl, lngline and recreatinal) using the separability assumptin (Furnier & Archibald 1982, Deris et al. 1985, Metht 199) fr methd-specific selectivity-atlength patterns and the reprted catch weights. An instantaneus explitatin rate catch equatin was used with a maximum annual ttal f.7. Length-specific prcesses fr fishing mrtality with inter-annual variatin in length-at-age were applied. The length distributin f three year-ld fish may typically include fish less than the minimum legal size (MLS 25 cm), with the lwer 95 th percentile in sme years being 23 cm. These fish cmprise a negligible cmpnent f the ppulatin ver all years, and, given the selectin patterns estimated frm the 22 tag-recapture prgramme, are unlikely t be recruited t the trawl fisheries. In effect, the grwth and selectin parameters f the mdel prduce a recruitment prcess that resembles knife-edge recruitment at age three years. Cnsequently, annual ppulatin bimass is defined as including fish in all age classes 3 y and lder. Three years ld fish in SNA 8 are sexually mature (authrs unpublished data). Therefre, a maturity give was nt included in the mdel, and all fish in the ppulatin are assumed t be sexually mature. Hwever this assumptin has n material effect because n spawner-recruit relatinship was included in the mdel. Ppulatin dynamics were mdelled in discrete fishing year time steps frm 1931 t 24. It was assumed that at the beginning f the first fishing year, the ppulatin was in unexplited equilibrium with cnstant mean recruitment. This initial ppulatin estimate defined the maximum unfished, r virgin, bimass, B. The fllwing rder fr within-year transitin prcesses was assumed: ageing, adding annual recruitment, reducing ppulatin numbers due t natural mrtality and methd-specific annual catches. The current estimates derived frm the mdel relate t the ppulatin at the start f the 25 fishing year. Prvisin was made fr under-reprting f the cmmercial catch (2% befre 1987 and 1% since the QMS was intrduced). Japanese lngline catches frm 1965 t 1977 were included in the cmmercial catch histry, but because reprts were unavailable fr the perid 1965 t 1974, an annual catch f 2 t was assumed fr this fishery. Fr the 24 stck assessment, the assumptin regarding histrical recreatinal catches was reviewed n the basis f available estimates. Estimates f 661 t in 2 and 1133 t in 21 were reprted (Table 4) indicating that recreatinal catch may be higher than assumed previusly, althugh the Snapper Ministry fr Primary Industries SNA 8 stck assessment

18 Wrking Grup believes that cnsiderable uncertainty surrunds all f the recreatinal catch estimates. There was anecdtal infrmatin t supprt the implicatin f the surveys that recreatinal catch has increased since 199. Hwever, the nature f this increase was unclear. Fr the updated assessment it was decided t use tw alternative recreatinal catch time series: - C rec,3, = 6 t 3 t linear increase; = 3 t (cnstant) - C rec,6, = 12 t 6 t linear increase; = 6 t (cnstant) T examine ppulatin pint estimates fr the 25 fishing year, the mdel was prjected deterministically frm 24 t the start f 25 by remving the reprted cmmercial catches in 24, assuming cnstant recreatinal catch (C rec,3 r C rec,6 depending upn the mdel), and assuming mean recruitment. Fr 25, cmmercial catches were assumed t equal the TACC f 15 t (plus 1% verrun) allcated prprtinal t the methd-specific catches reprted in Mdel develpments fr an updated assessment In updating the 24 SNA 8 assessment, it was reasnable t adpt the main assumptins and specificatins described by Davies et al. (26). Hwever, certain assumptins were reviewed in light f the new infrmatin input t the mdel. The updated mdel was develped with cnsultatin with the Ministry f Fisheries Snapper Wrking Grup. An utline f the main stages f this prcess fllws t prvide a backgrund f the mdel used fr the 25 assessment. T examine the effects f including new bservatins and parameter mdificatins t the mdel, a stepwise apprach was fllwed. New bservatins were added t the mdel successively, beginning with updating the time series used in the 24 assessment (catch-at-age, CPUE, catch-weights), and then adding the revised catch-at-age estimates fr the 197s, recreatinal catch-at-length (199s t 2), and at-sea pair trawl catch-at-length. Mdificatins were then made where pssible given the new infrmatin available t estimate additinal parameters. Examples include: increasing the maximum age class t 5 years cnsistent with catch-at-age bservatins cllected in the 197s, estimating stepwise changes in trawl methd selectivity-at-length due t changes in cd-end mesh sizes, fitting the 22 tag-recapture estimate as abslute numbers at length, and estimating stepwise changes in selectivity-at-length fr the recreatinal methd by fitting t the catch length frequencies cllected at bat ramps. The vulnerability f 3-year-lds t recreatinal fishing was adjusted t accunt fr the increase in 1994 in the recreatinal MLS (t 27 cm). These develpments were assessed by diagnstic examinatin. In attempting t include these new features within ne base case mdel, prblems were encuntered estimating the recreatinal selectivity-at-length parameters and calculating Bayesian mdel estimates fr natural mrtality. The relative leverage f respective bservatins n mdel estimates was examined by: adjusting the relative weight f the data types, calculating a likelihd prfile fr mean recruitment, calculating the parameter crrelatins, and determining the MCMC acceptance rates fr the varius mdel ptins prepared during the stepwise mdel develpment. This examinatin revealed mean recruitment and natural mrtality t be highly crrelated, which may have cntributed t pr MCMC perfrmance when estimating natural mrtality. Althugh it was thught that a high maximum age (5 years) might imprve the estimatin f natural mrtality, it rendered the mdel cmputatinally intensive t the extent that MCMC calculatins became impractical. These cnsideratins and ther details f the mdel led the Snapper Wrking Grup t agree upn a revised set f ptential base case specificatins. These culd nt be encapsulated in ne single base case but were instead advanced as a set f alternative, equally plausible ptins. Certain ther ptential mdel ptins, were cmpared t the base mdel ptins and discarded frm the assessment as being implausible r requiring unreasnable assumptins. Other sensitivities t base mdel ptins were specified t test rbustness t uncertainty in histrical catches and statistical assumptins regarding the tag-recapture bservatin f abslute abundance. 14 SNA 8 stck assessment 24 5 Ministry fr Primary Industries

19 Bayesian estimates were calculated, and presented t a Plenary meeting. It was determined at this meeting that the statistical assumptin made under the base mdel ptins regarding the fit t the tagrecapture bservatins f abslute numbers at length assigned unduly high weight t these data. The assumptin made under an alternative ptin was cnsidered mre reasnable, which invlved fitting the tag-recapture bservatin as an abslute bimass index instead f abslute numbers at length. Als, lwer relative weight was recmmended fr bth the 199 and 22 tag-recapture bservatins. These recmmendatins were fllwed in specifying five mdel ptins finally used fr the updated assessment and described in detail in Sectin belw. As befre, there was n single base case mdel assessment mdel specificatins The specificatins and assumptins shared by all mdel ptins, sme f which differ frm the 24 assessment mdel, are: stepwise changes in selectivity-at-length fr the recreatinal methd, pre-1994 (MLS 25 cm) and pst-1994 (MLS 27 cm); left-hand limb f the recreatinal selectivity-at-length is assumed t be knife-edge at the MLS; recreatinal selectivity-at-length is maximum at the MLS; the fit t the bat ramp length frequencies (all years) is duplicated fr bth the pre-1994 and pst recreatinal selectivity-at-length functins t ensure that the estimated right-hand limb parameters are the same; single trawl and pair trawl selectivity-at-length are time invariant, with left-hand limbs cnstrained t clsely resemble thse f the 22 tag-recapture estimates; lngline selectivity-at-length is cnstant (equal t 1.); natural mrtality is estimated at the MPD but fixed fr the Bayesian psterir estimatin; C rec is either 3 t r 6 t; mean recruitment crrespnds t the perid 1971 t 2, i.e. mean YCS = 1.; bserved mean weight-at-age is assumed fr the years available: and ; vn Bertalanffy estimates f mean weight-at-age taken frm the perid were assumed cnstant fr all years pre-1975; the average f the tw vn Bertalanffy estimates f mean weight-at-age fr the perids and were assumed cnstant fr all years between 198 and 1988; the mean bserved weight at 2+ years is assumed fr the and perids respectively, with the mean value fr assumed fr the years pre-1975, and the average f bth perids assumed fr the years Estimated parameters The mdel parameters estimated were: R mean annual recruitment f 3 year ld fish; r t annual recruitment indices f year class strength (YCS) fr the 1971 t 2 year classes (n = 3) a 1,s L,s R selectivity at length fr duble nrmal functin; single trawl, pair trawl and recreatinal methds (n = 8, nly the right-hand limbs f the recreatinal functins were estimated) M natural mrtality. Annual year class strengths were the prduct f the mean recruitment and YCS, hence, the year class strength f 3-year-ld fish at the start f year t was R r t. Ministry fr Primary Industries SNA 8 stck assessment