Broadening the horizon of size selectivity in trawl gears

(1)

ContentslistsavailableatScienceDirect

Fisheries Research

jo u r n al h om ep ag e :w w w . e l s e v i e r . c o m / l o c a t e / f i s h r e s

Broadening the horizon of size selectivity in trawl gears

Daniel Stepputtis

^a,∗,1

, Juan Santos

^a,1

, Bent Herrmann

^b,c

, Bernd Mieske

^a

aThünen-InstituteofBalticSeaFisheries,AlterHafenSüd2,18069Rostock,Germany

bSINTEFFisheriesandAquaculture,FishingGearTechnology,Willemoesvej2,9850Hirtshals,Denmark

cUniversityofTromsø,Breivika,N-9037Tromsø,Norway

a r t i c l e i n f o

Articlehistory:

Received17January2015

Receivedinrevisedform14August2015 Accepted31August2015

Availableonline12September2015

Keywords:

Sizeselectivity trawl

Exploitationpattern Bell-shape S-shape Selectivitycurve

a b s t r a c t

Thediscussionofalternativeharvestpatternsincommercialfisherieshasbeenraisedbystockassessment andfisherymodelers,especiallyinthewidercontextofbalancedharvesting.Butoften,thesetheoretical approachesproposealternativeexploitationpatternsthataredifficulttoachievewithinthecurrent limitationsintheselectivitycharacteristicsoffishinggears,suchastrawlgears.Theaimofthepresent studyistobroadenthehorizonforsizeselectivityintrawlgearsbydemonstratingthefeasibilityof alternativeselectivitypatternsfortrawls.Asacasestudy,wecombinedtwowell-knownselectiondevices toobtainabell-shapedselectivitycurveintrawlswithlowcatchabilityofbothsmallandlargeindividuals fromthetargetspecies.WehavesuccessfullytestedthisgearintheBalticSeacodfishery.Theresults revealedthatcompletelydifferentexploitationpatternsfortrawlgearscanbeachievedbymeansofgear technology.

1. Introduction

Selectivity can be defined as the dependence of a fishing gear’scaptureefficiencyonfactorssuchassize,age,andspecies (MacLennan,1992).Adaptingtheselectivityoffishinggearsisthe mostimportantstrategyusedinmanyfisheriesaroundtheworldto achievethedesiredexploitationpatterns.Sofar,awidelyaccepted paradigmisthat“Improvingselectivityleadstoamoreefficient exploitationofthestock’sgrowthpotential”(Macheretal.,2008), andthatgoodfisherymanagementrequiresfishinggearstocatch largeadultfishwhileallowingsmalljuvenilestoescape(Armstrong etal.,1990).Accordingtoclassicaltheory,lengthatfirstcatchisthe keyparametertooptimizingastock’syield.(Armstrongetal.,1990;

BevertonandHolt,1957).

Thesizeselection offishinggears isdescribed byselectivity curves,which quantifytheprobabilitythat agiven lengthclass ofagiven fishspecieswillbecaught,assumingthatitis avail- abletothegear.Selectivitycurvesdifferbetweengeartypesand configurationsofgears(Dicksonetal.,1995;HovgårdandLassen, 2000;Wilemanetal.,1996).Passivegears,suchasgillnets,have sizeselectionpropertiesusuallydescribedasbell-shapedcurves (Dicksonetal.,1995;HovgårdandLassen,2000;Huse,2000;Millar

∗Correspondingauthor.

E-mailaddresses:daniel.stepputtis@ti.bund.de(D.Stepputtis), juan.santos@ti.bund.de(J.Santos),Bent.Herrmann@sintef.no(B.Herrmann), bernd.mieske@ti.bund.de(B.Mieske).

1 Equalauthorship

andFryer,1999;MillarandHolst,1997).Theyarecharacterizedby lowretention probabilitiesatsmalllengthclasses, aswellasat largelengthclasses,withtheresultthatgillnetscatchprimarily medium-sizedlengthclasses.

Historically,theselectivepropertiesoftrawlsandotheractive gearswereadaptedbyaltering thesizeselectionin thecodend (Glass,2000).Thisstrategy assumesthat mostfishenteringthe geardrifttowardthecodend,whereasimplesize-selectionpro- cessoccurs:smallerfishwithspecificmorphologicalcharacteristics haveagreaterprobabilityofpassingthroughthemeshesandescap- ing,whereaslargerfishhaveagreaterprobabilityofbeingretained inthecodend.Incontrasttopassivegears,theselectioncurvein trawlgearsisS-shaped.Thus,theretentionprobabilityincreases withthesizeoffish(Dicksonetal.,1995;Gulland,1983;Huse, 2000;MacLennan,1995;MillarandFryer,1999;Reevesetal.,1992;

Wilemanetal.,1996).Toreduceunwantedbycatch,theclassical codendselectionisoftensupplementedwithadditionalselectivity approaches,suchasgrids(HeandBalzano,2012;Sistiagaetal., 2010),escape windows(Armstronget al.,1998;Bulloughetal., 2007;CatchpoleandRevill,2008;Madsen,2007),andotherstrate- gies(Herrmann etal., 2015).Currently, theselectiveproperties ofthesetypesofdevicesareoptimizedbychangingtheS-shaped selectivitycurve,resultinginachangeinthepositionofthecurve alongthelengthrangeofthespecies(oftendescribedastheL50- value,lengthof50%rejection/retention)and/orinthesteepness ofthecurveoftendescribedastheSR-Value,L25–L75;(Dickson et al., 1995; Wileman et al., 1996). A good example of such a limitedapproachisthedevelopmentofgearregulationsforcod-

http://dx.doi.org/10.1016/j.ﬁshres.2015.08.030

0/).

(2)

directedfisheriesintheBalticSea(Feekingsetal.,2013;Madsen, 2007).Since1999,fisherymanagementandfisherysciencehave triedrepeatedlytoadapt thesizeselectivityoflegalcodendsto accomplishspecificmanagementgoals.Thiseffort,mostlylimited todiscardreduction,hasbeencarriedoutwithoutconsideringa broadersetoffisherymanagementobjectives,suchasoptimalpop- ulationdynamicsandhealthypopulationstructure.Nevertheless, owingtoalackofalternativeselectivityoptions,thestandardS- shapedtrawlselectivitycurve was“only”moved left andright (Fig.1).

ThelackofpossiblealternativestotheS-shapedtrawlselectiv- itycurvesalsonarrowstherangeofpotentialexploitationpatterns tobeinvestigatedinﬁsherymodels,inthesearchforoptimalhar- veststrategies.Typically,suchstudiesonlyconsideredS-shaped selectivityscenarios (Kronbaketal.,2009;Macher etal.,2008).

Withthedebateaboutbalancedharvesting(Garcia etal.,2012;

Jacobsenetal.,2013;Zhouetal.,2010),additionalselectivitypat- ternsarebeingdiscussedandusedformodelingpurposes(Jacobsen etal.,2013).However,itoftenremainsunclearhowthealternative harvestpatternscouldbeimplementedtechnicallyintheﬁsheries.

Apartfrom thefundamental concept of balanced harvesting andunderlyingaims,otherrationalesofferthemselvesasalterna- tiveharveststrategiesfortrawlfisheries:Althoughtheimportance ofagestructureforrecruitmentsuccessisstillunderdiscussion (Brunel, 2010; Morgan et al., 2011), there are arguments for a healthyagestructure,includinglargeandoldindividuals(Berkeley et al., 2004; Hixon et al., 2014; Law et al., 2015). For several stocks,thepositiveinfluenceonpopulationdynamicscausedby olderindividualshasbeenpostulated,withvaryingdrivingfactors, includingparentaleffects(CardinaleandArrhenius,2000;Cervi ño etal.,2013;MarteinsdottirandBegg,2002;Trippeletal.,2005) and enhanced resilience against excessive fishing pressure and againstclimatevariation(Ottersenetal.,2006).Theextentofsuch effectsisstillbeingdebated(Marshalletal.,2010;O’Farrelland Botsford,2006).Inaddition,age-structureindicesarealsoimpor- tanttoecosystem-basedfisherymanagement.

Inlinewiththeabovearguments,weaiminthisstudytoreduce thecatchabilityoftrawlgearsforbothtailsofthelengthdistribu- tion(juvenilesandolderfish)foragiventargetspecies.Achieving thisthroughfishingtechnologywouldrequirefindingwaystoshift

Fig.1. Selectioncurvesof legalizedcodendsfor theBalticcod trawl ﬁshery, 1999–2015.Verticallinesrepresentthecorrespondingminimumlanding/reference sizes(MLS;35cm,1999–2002and2015;38cm,2003–2014).Codendsare(a)T0 120mm(1999–2001);(b)T0130mm(2002–2003);(c)Exit-window(1999–2001);

(d)Bacoma110mm(2003–2009);Bacoma120mm(2001–2003and2010–2015);

T90 110mm (2006–2009); T90120mm (2010–2015). Selectivitycurveswere derivedfrompersonal,unpublishedselectivityexperimentsconductedbetween 1999and2010.AdescriptionofthelegislativedevelopmentcanbefoundinFeekings etal.(2013).

thetraditionalS-shapedtrawlselectioncurvestowardbell-shaped selectioncurves,commonlyassociatedwithpassivegearssuchas gillnets(Dicksonetal.,1995).Thestrategyadoptedhereemulates gillnet-likebell-shapedselectivitybyaddingtherejectionoflarger individualsduringtheselectivityprocessinastandardtrawlgear.

Thetechnologicalapproachissimpleandisbasedonthecombi- nationoftwowell-knownandwidelyusedselectiondevices.The proofof conceptwascarried out intheBaltic Seacod-directed ﬁshery.

Fig.2.Illustrationofthegridandcodendselectionsystemusedtoobtainbell-shapedtrawlselectivity.Inadditiontotechnicaldetails,thedifferenttraitsoffishenteringthe extensionpieceareillustrated:(a)fishnotcontactingthegridandescapingthroughtheMEO;(b)fishcontactingthegrid,butnotabletopassthrough;(c)fishcontacting thegrid,passingthrough,andenteringthecodend;(d)fishescapingthroughthecodendmeshes;(e)fishfinallycaughtwithinthetestcodend.

(3)

Theoverallaimofthestudyistodemonstratethefeasibility ofalternativeselectivitypatternsfortrawlsingeneral.Basedon thisdemonstration,itishopedthatthestudywillstimulatefurther discussionanddevelopmentthatwillbroadenthescopeofﬁshery management.

2. Materialandmethods 2.1. Selectivityconcept

Toachieve a bell-shapedsize selectivitypatternfor a target speciesintrawlfisheries,twoselectiondevices—agridsystemand astandardcodend—weremountedsequentially(Fig.2).Thefirst selectiondevice,asteelgrid,wasmountedintheextensionpiece betweenthebellysectionofthetrawlandthecodend.Thepur- poseofthegridwastochangethepopulationstructureentering thecodendbyrejectinglargefishandallowingsmallandmedium- sizedfishtopassthroughitandcontinuetheselectionprocess.

Largefishunabletopassthroughthegridwouldbeexcludedfrom thegearthroughtheescapeoutletplacedintheupperpanelin frontofthegrid.Ideally,allfishshouldcontactthegridintheir normalswimmingorientationandbesortedaccordingtosizeby thegrid.However,notallfishenteringthegearwillnecessarily contactthegrid,andsomemaysubsequentlyescapethroughthe outlet,regardlessoftheirsize(MillarandFryer,1999;Sistiagaetal., 2010).Consequently,thisstudyfacedthechallengeofensuringthat alargeproportionoffishmadepropercontactwiththegridtobe sortedbysizebeforeencounteringtheescapeoutlet.Tostimulate gridcontact,weattachedarectangularpieceofnettingatthefront oftheescapeoutlet.Thenettingwasmountedovertheoutletto maketheoutletlessvisibletofish(Fig.2).Theresultingmasked escapeoutletisdenotedhereafterasMEO.

Thesmallandmedium-sizedfishnotrejectedinthegridzone aresortedbythesecondsize-selectionprocessdeterminedbythe selectivitypropertiesofthecodend.Atthisstage,onlysmallfish haveanyprobabilityofescapingbypassingthroughthecodend meshes.Theprofileoftheresultingcatchisthereforedetermined bythecombinationof twosize-selection processes,differing in purposesandactingsequentiallyalongthegear.Becausecodend sizeselectionactsonlyonfishthatcontactandpassthroughthe gridinthefirstselectionprocess,thesecondselectionprocessis conditionedbythefirst.

2.2. Experimentalsetup

Toestimatetheindividualandcombinedselectivityproperties

ofbothselectiondevices,itishelpfultouseathree-compartment setup(Jørgensenetal.,2006;KvammeandIsaksen,2004;Sistiaga etal.,2010(Fig.3)todirectlyquantifyfishescapingthroughthe MEO(fish rejectedby thegridor fish thatdid notcontact the grid),fish retainedinthecodend,and fishthat passedthrough thecodendmeshes.We used anexperimentaldesign basedon thecovermethod(Wilemanetal.,1996)tocollecttheexperimen- taldata.Inadditiontothecommonsetup,basedoncoveringthe codendwithasmallmeshnetcover,thisexperimentalsetupuses atopcovertocollectthefishusingtheMEOtoescapefromthe gear.Consequently,theexperimentaldesignincludesthreecom- partments:

(a)TC=topcover tocollectallindividualsescapingthroughthe MEO(n_TC,l)

(b)CD=codend,containingthegear’sﬁnalcatch(n_CD,l)

(c)CC=covercodendtocollectallindividualsescapingthroughthe codendmeshes(n_CC,l)

2.3. Modelfordescribingbell-shapedselectioncurves

Theprobability thata ﬁsh willbecaught(r(l), overall reten- tionprobabilityofthegear)uponenteringtheexperimentalgear dependsontheprobabilitythatitpassesthroughthegrid(p_grid(l), passageprobabilitythroughthegrid)towardthecodend,andthatit issubsequentlyretainedinthecodendthroughsizeselectionthere (r_codend(l),retentionprobabilityinthecodendconditionedentry).The overallsizeselectionofthegearcanbedescribedbythefollowing model:

r(l)=pgrid(l,Cgrid,L50_grid,SRgrid)×rcodend(l,L50_codend,SRcodend) (1)

EachofthepartialselectivityfunctionsontherightsideofEq.(1) hasaspeciﬁcstructureandthereforemustbedescribedseparately.

Thefirstistheprobabilitythatafishwillpassthroughthegrid towardthecodend(p_grid(l)).Thisisthecombinedprobabilitythat afishefficientlycontactsthegrid(C_grid,contactprobabilitywithgrid) and,onceitcontactsthegrid,itissmallenoughnottoberejected bytheselectivepropertiesofthegrid(1-r_grid(l));therefore:

pgrid(l,Cgrid,L50_grid,SRgrid)=Cgrid×(1−rgrid(l,L50_grid,SRgrid)) (2)

Second,r_codend(l)inEq.(1)referstotheprobabilitythataﬁsh willbe retainedin thecodend,presupposingthat it entersthe codend.Theprobabilitiesr_grid(l)andr_codend(l)canbedescribedby standardS-shapedsize-selectionmodelsfortrawlgears.Wecon- sideredfourdifferentS-shapedmodels:Logit,Probit,Gompertz,and Richard.Detailsofthesefunctionsandtherespectivecalculationsof theselectivityparametersL50(lengthof50%rejection/retention) andSR(L75–L25)canbefoundinWilemanetal.(1996).

2.4. Modelestimationandselection

Thevaluesfortheparametersfortheoverallselectionmodel(1) –C_grid,L50_grid,SR_grid,L50_codend,andSR_codend—wereobtainedusing maximumlikelihoodestimationbasedontheexperimentaldata, pooledoverhaulsj(1tom)byminimizing:

−

l

m

j=1

n_TC,l,j×ln

1.0−pgrid

l,Cgrid, L50_grid_, SRgrid

+

n_CC,l,j+n_CD,l,j

×ln

pgrid

l,Cgrid,L50_grid_,SRgrid

+nCC,l,j×ln(1.0−rcodend

l,L50_codend, SRcodend

)+nCD,l,j×ln

rcodend(l,L50_codend, SRcodend)

(3)

Intotal,16modelswereconsideredtodescribetheoverallsize selectivityinthetrawl,basedonthenumberofcombinationsofthe fourdifferentS-shapedfunctionsconsideredforbothr_grid(l)and r_codend(l)(Section2.3).The16competingmodelswereevaluated basedontheirAIC-values(Akaike,1974);themodelwiththelowest valuewasselected.Thediagnosisofgoodness-of-fitoftheselected modeltodescribetheexperimentaldatawasbasedonthep-value, modeldeviancevs.degreeoffreedom,andfinallytheinspectionof themodelcurve’sabilitytoreflectthelength-basedtrendsinthe data.

ThemaximumlikelihoodestimateusingEq.(3)withEq.(1)and (2) and requirestheaggregation oftheexperimentaldata over

(4)

Fig.3. Illustrationoftheexperimentalsetupwiththreecompartments.Foradescriptionofdifferentnumbers,seeFig.2.

hauls.Thisresultsinstrongerdatatoestimatetheaveragesize selectivity,attheexpenseofnotconsideringexplicitvariationin selectivitybetweenhauls(Fryer,1991).Toaccountcorrectlyforthe effectofbetween-haulvariationinestimatinguncertaintyinsize selection,weusedadoublebootstrapmethodtoestimatetheEfron percentileConﬁdenceIntervalsforboththeestimatedparameters inEquation(1)andtheresultingcurvesforp_grid(l),r_codend(l),and r(l).WeusedthesoftwaretoolSELNET(Herrmannetal.,2012)for theanalysisandapplied1000bootstrapiterationstoestimatethe conﬁdenceintervals.

2.5. Speciﬁcsetupofthetrawl

TheexperimentaltrawlwasaTV300/60(300meshescircum- ferencebehindthesquarewitha 120mmmeshopeninginthe bellyand60mmintheextensionpiece),astandard trawlused intheBalticcod-directedtrawlﬁshery.Thetrawlandthecodend weretwo-panelconstructions,whereastheextensionpiecewasa four-panelconstruction(Fig.3).Theextensionpieceincludedsmall transitionsectionsthatallowedthetwo-panel(bellyandcodend) andfour-panel(extensionpiece)constructionstobejoined.

Toachievetheintendedbell-shapedselectioncurvebyusingthe proposedsequentialselectionsystem,itwasnecessarytodeﬁne thegrid’sbarspacingandcodendcharacteristics,consideringthe lengthstructureofthepopulationavailableatthemomentofthe experiment(obtainedfromBalticInternationalTrawlSurvey,ICES SD24,ﬁrstquarter2014).Theinformationaboutthepopulation structurerevealedverylowabundanceoflargecod(above50cm, Fig.4).WeusedSELNET’sbuilt-inparametricsimulationfacilities topredicttheselectioncurvesofagridcombinedwithacodend.

Thissimulation(Fig.4left)indicatedthatitwouldprobablynot leadtosufﬁcientcoverageofthebell-shapedselectioncurvewhen combiningahighlyselectivegrid(forexamplewithbarspacing of70mm)andacodend(forexamplethemandatoryT90120mm codend).Therefore,itwasproposedtocombineagridwithreduced barspacing(50mm)anda less-selectivecodend(T90105mm).

Thegridwasinstalledatanangleof75^◦andaguidingpanelwas installedinfrontofthegridtofurtherencourageﬁshcontactwith thegrid,inadditiontotheuseofMEO(Fig.3).Thecodendwasmade of4mmPEdoubletwinewithanactualmeshsizeof107mmand 50meshesalongand50meshesaround.

Thetopcoverandcovercodendweredesignedfollowingrecom- mendationsofWilemanetal.(1996)(Fig.3).Thecovercodendand

thelastpartofthetopcoverweremadeofPEsingletwine2.5mm nettingwithameshsizeof60mm.Thecovercodenddimensions were 570meshes in circumference and 275meshes in length.

Thetopcover constructionfollowedthedesign guidelinesfrom Wilemanetal.(1996),thereforeitcomprisestheassemblyofnet pieceswithdifferentdimensionsandcuttingedges.Toavoidmask- ingeffects,11ﬂoatswithabuoyancyof∼800geachwereattached tothetopcover,whilethecombinationof5kiteswithleadweights wereusedtoseparatethecovercodendfromthecodend.

Tounderstandtheoperationoftheselectivitydevicesandthe behaviorofﬁshnearsuchdevices,weusedGoProcameras(GoPro Hero3HDcameraswithoutartiﬁciallight),installedatseveralposi- tionsonthetrawl.

3. Results

TheexperimentalﬁshingwasconductedonboardtheGerman FisheryResearchVessel(FRV)“Solea”(totallength=42m,950kW, sterntrawler)overaperiodof3days(21–23March2014)inthe WesternBalticSea(Table1).Thewaterdepthvariedbetween14 and46m.Theaveragetowingspeedwas3knots.Thehaulduration waseither90or120min.

Inall,eightvalidhaulswereachievedbytheexperimentalﬁsh- ing(Table1).Allcodobservedinthedifferentcompartmentswere measuredtothenearesthalfcentimeterbelowtheirtotallength.

Atotalof12514cod(5371.28kg)werecaughtinexperimen- talhaulsusedintheanalysis(Table1).Allthreecompartments containedenoughcodforproperanalysis.

The16 differentmodels(Section2.4)weresuccessfully esti- mated, and the best model (considering the AIC-value) was determined to bethe one that used the Gompertz function to describeboththegridandthecodendselectivity(Table2).The estimatedcurvesforgridpassageprobability,conditionedcodend retention,andoverallselectiontogetherwiththeir95%confidence intervalsare shown inFig.5 (left).Inspectingthep-valuesand deviancevs.DOF-from-the-fitstatistics(Table2)couldhaveindi- catedlackoffitforthemodel.Butinspectingtheabilityofthemodel curvestoreproducethetrendsintheexperimentaldatarevealed nosystematicpatternofdeviancesforanyofthecurves(Fig.5).

Therefore,weconsiderthepoorfitstatisticsaresultofoverdisper- sioninthedataand,basedonthis,weareconfidentinapplyingthe modeltodescribethetrendsinthedata.Theprobabilitythatafish efficientlycontactedthegridwasestimatedasC_grid=0.73(Table2),

(5)

Fig.4.Apriorisimulationofexpectedselectivityofthesequentialselectivitysystemtobeusedduringexperimentalfishing.Greyshadedarea:expectedlengthdistribution forcodinthefishingarea(derivedfromBalticInternationalTrawlSurvey,ICESSD24,firstquarter2014).Greylines:simulatedselectivitycurvesforgridandcodend(assuming 100%contactprobabilitywiththegrid).Blackline:resultingretentionprobabilityoftheentiretrawl.Dotsindicatethedistributionoflengthclassesinpopulationalongthe simulatedretentioncurve.Left:combinationofgridspacing70mmwithT90120mmcodend;right:combinationofgridspacing50mmandT90105mmcodend.

Table1

Operationalinformationoftheexperimentalﬁshinghauls.TC=topcover,CC=covercodend,CD=codend.

Haul Towduration(min) Latitude Longitude Depth

(m)

Codcatchindifferentcompartments Number(Catchweightinkg)

TC CD CC

1 120 54^◦12,227N 012^◦00,860E 14 321(137.65) 835(402.33) 657(202.11)

2 120 54^◦12,568N 011^◦47,101E 23 375(396.62) 1151(433.63) 751(246.03)

3 90 54^◦12,254N 012^◦00,422E 15 953(176.15) 839(526.22) 741(238.89)

4 90 54^◦45,378N 013^◦29,785E 41 38(16.40) 364(148.26) 138(32.09)

5 120 54^◦50,315N 013^◦27,635E 46 608(239.47) 966(418.59) 396(115.05)

6 120 54^◦52,660N 013^◦15,529E 45 197(80.44) 649(254.39) 634(147.39)

7 120 54^◦52,610N 013^◦15,166E 45 742(331.47) 424(167.88) 268(63.46)

8 120 54^◦52,540N 013^◦30,885E 47 647(225.76) 487(266.01) 333(104.99)

Total 3881(1608.44) 4715(2617.31) 3918(1150.01)

meaningthat73%ofﬁshenteringthetrawleffectivelycontacted thegridandweresortedbyit,basedonsize.Therefore,anum- berofindividualsthatcouldhavepassedthroughthegridescaped throughtheMEOandwerereleasedtothetopcover(Fig.5,top left).Theunderwatervideorecordingsrevealedthatmanyﬁshhit thegridsoonafterenteringthetrawl,whileotherswereactively swimminginfrontofthegridandnotmakingimmediatelyuseofit.

Forthoseﬁsh,thechancesincreasedtoﬁndthewayoutthroughthe escapementopeningabovethegrid—evenwhencoveredbyanet

panel.Thisgrid-avoidanceresponsebycodcouldhavecontributed tothereductioninC_grid.

OwingtothevalueobtainedforC_grid,whichimpliesthelossof someﬁshbelongingtothedesiredlengthclasses,thebell-shaped selectioncurvedidnotreachthefullcatchability(retentionprob- ability)atthetargetedmid-sizedlengthclasses.Nevertheless,the overallgearselectivitycurve(Fig.5,bottomleft)clearlydemon- stratesthepossibilityofobtainingbell-shapedsizeselectivityin trawls.

Table2

Selectivityparametersforthebestmodelsdescribingthesizeselectionsofthetwoselectivedevicesinthetestgearduringtheexperimentalseatrials;95%conﬁdencelimits showninparentheses;DOF:degreeoffreedom.

Selectiondevice Model Parameter Value

Grid Gompertz Cgrid 0.73(0.64–0.83)

L50grid 47.93(46.45–49.46)

SRgrid 8.40(5.72–12.14)

Codend Gompertz L50codend 29.70(28.22–30.94)

SRcodend 11.05(10.17–11.82)

p-Value 0.0093

Deviance 217.57

DOF 171

Numberofhauls 8

AIC 27060.56

(6)

Fig.5.Left:sizeselectioncurvesofcodindifferentselectivitydevices(includingexperimentaldata(points)and95%conﬁdencelimits).Top:gridwithverticalbarsand 50mmbardistance;Middle:T90105mmcodend;Bottom:selectivitycurvesofgridandcodend(greylines)andresultingcombinedselectivitycurve.Right:catchwithina givencompartment(stippledcurve)inrelationtothelengthdistributionencounteringtherelevantselectiondevice(greyshadedarea).

Ourresultsdonotindicateanybiasresultingfromcoverselec- tion, because the model we applied was able to describe the fullrange of thedatawithoutanysystematic pattern ofdevia- tion.

4. Discussion

Thediscussionofalternative harvestpatternsin commercial ﬁsherieshasbeenraisedbystockassessmentscientistsandﬁshery modellers,especiallyinthewidercontextofbalancedharvesting

(7)

(Garciaetal.,2012;Jacobsenetal.,2013;Zhouetal.,2010).These theoreticalapproachesoftenproposealternativeexploitationpat- terns, which cannot be achieved under the currently assumed paradigmforsize-selectioncharacteristicsfortargetspecieswith activeﬁshinggears,suchastrawlgears.

Theaimofthepresentstudywastobroadenthehorizonforsize selectivityintrawlgearsbydemonstratingthefeasibilityofalterna- tivesize-selectionpatternsfortrawls,inadditiontothetraditional S-shapedpattern.

Therefore,wehavechosenoneexampletodemonstratethat completelydifferentexploitationpatternscanbeachievedintrawl ﬁsheriestoaccomplishalternativeﬁshery-managementobjectives.

Thepracticalexercisewastosimultaneouslyobtainlowcatchprob- abilityofthesmallerandlargerindividualsavailableinthetargeted fishpopulation.Theunderlyingideaisbasedonthehypothesis that,inadditiontoshort-andmedium-termeffectsofthelossof reproductivepotentialofolderandlargerfish,sizeselectivityof trawlsalsohasalong-termeffect.Itisknownthatthefishingpres- sureincombinationwithtraditionalS-shapedselectivitypatterns oftrawlscanresultinfishery-induced evolution (Andersen and Brander,2009;Jørgensenetal.,2007;KuparinenandMerilä,2007;

Law,2000).Althoughtherateofevolutionisassumedtobelower thanpreviouslypublished(AndersenandBrander,2009),analter- nativeharvestpattern—targetingnotonlylargeindividuals—may helptoreducetheevolutionaryeffectsoftrawlselectivity.

Thetechnologicalstrategyadoptedtoachieveourgoalwasthe combinationoftwowell-knownsize-selectiondevicesinfishing- geartechnology,integratedsequentiallyinthetrawltoestablish adualselectionsystem.Thishasbeentestedforthecod-directed trawlfisheryintheBalticSea.Weusedagridtospecificallysortout thelargeindividualsofthetargetspecies,whileallowingsmaller fishtoenterthecodend.Theuseofagridforthispurposeisnew forthetargetspecies.Untilnow,gridshavebeenusedtosupple- mentcodendsizeselectivitybyallowingsmallindividualstoescape (Heand Balzano,2012;Herrmannet al.,2013;Jørgensen etal., 2006;KvammeandIsaksen,2004;Sistiagaetal.,2010;Wileman etal.,1996)ortoexcludetheentirelengthrangeofspecificbycatch speciesfromthecatch(HeandBalzano,2011;Isaksenetal.,1992;

Salaetal.,2011).Insomecases,bothgridapplicationsarecombined inthesamegear(HeandBalzano,2013).

Inexcluder-grid-basedselectivitysystems,itisalsolikelythat selectivitypatterns canbe foundthat differ from thestandard S-shapetrawlselectivitycurve.Possibleexamplesareshrimpfish- eries,wheretrawlsareusedtoavoidcatchofunwantedroundfish species(HeandBalzano,2011;Isaksenetal.,1992).Ifthegrid- barspacingallowedthepassageofindividualsofroundfishspecies withinthelengthrange,whichisalsorelevanttocodendselectiv- ity,itmayalsobepossibletofindbell-shapedselectivityforthese species.Thisbell-shapedselectivitycurvereleasesthelargeindi- vidualsinfrontofthegridandthesmallindividualsinthecodend.

Incontrasttothedesign usedinthis study,this potentialbell- shapedselectivitycurveisderivedbyaccidentandisnotobtained onpurpose,andcertainlynotforthetargetspecies.

Theexperimentalresultspresentedheredemonstratethatitis possibletoobtaincompletelydifferentexploitation patternsfor trawlgearsbymeansofgeartechnology.

Basedonthe lengthdistributionof codavailable duringthe experiments,theselectivepropertiesoftheselectiondevicesused didnotnecessarilyresultinanoptimizedharvestpatternforcodin theBalticSea,butwerechosenbasedonexperimentalconsidera- tions(seeSection2.5)and,followingtheaimofthisstudy,toactas afeasibilitystudy.Optimalcombinationsofgridandcodendselec- tivityforavarietyofﬁsheriescanbeidentiﬁedinfuturemodelling studies.Toimprovetheproposedselectivitypattern,attentionhas tobepaidtoincreasingtheprobabilityofcontactwiththegridby specimensenteringthetrawl.

Asmentionedabove,theuseofmultipleselectiondevicesgives moreﬂexibilitytoobtaindesiredharvestpatterns.Ontheother side,thecomplexityofthetrawlhaseffectsoncostsandhandling ofthegear.Suchaspectsalsohavetobetakenintoaccountwhen identifyingoptimalharveststrategiestoobtainasustainableuseof apopulationandasustainableﬁshery.

Itwasshownthatitispossibletoachieveabell-shapedselec- tivityintrawlfisheries,whichissimilartotheselectivitycurveof gill-nets.Nevertheless,itisnotclearwhetherthepopulationeffect ofbothfisheriesisidenticalwhenusingbellshapedcurves.For instance,itcouldbeinfluencedbypotentialdifferencesinsurvival ofescapeesinbothfisheries.

Wehopethisstudywillinitiatefurtherdiscussionanddevel- opmentthat willbroaden thescopeand possibilitiesof ﬁshery management.Modelersareencouragedtoenlargethescopeoftheir modelstoincludealternativeselectivitypatternsandtodiscuss withﬁshinggeartechnologistshowtobringthemintopractice.

Acknowledgements

WethankthecrewmembersoftheFRV“Solea”fortheirvaluable helpduringtheseatrials.We extendspecial thankstoourcol- leagueswhohelpedusatsea:PeterSchael,KerstinSchöps,Susann Diercks,StefanieHaase,ValerieHofman,FriederPfaff,andJulian Hofmann.Also,specialthankstoAnnemarieSchützandMartina Bleilfortheirhelpinpreparingthemanuscript.

Additionally,wethankthetworeviewers,whosevaluablecom- mentsimprovedthemanuscriptsigniﬁcantly.

References

Akaike,H.,1974.Anewlookatthestatisticalmodelidentiﬁcation.IEEETrans.

Autom.Control19,716–723,http://dx.doi.org/10.1109/TAC.1100705.

Andersen,K.H.,Brander,K.,2009.Expectedrateofﬁsheries-inducedevolutionis slow.Proc.Natl.Acad.Sci.106,11657–11660.

Armstrong,D.W.,Ferro,R.S.T.,MacLennan,D.N.,Reeves,S.A.,1990.Gearselectivity andtheconservationofﬁsh.J.FishBiol.37,261–262,http://dx.doi.org/10.

1111/j.1095-8649.1990.tb05060.x.

Armstrong,M.J.,Briggs,R.P.,Rihan,D.,1998.Astudyofoptimumpositioningof square-meshescapepanelsinIrishSeaNephropstrawls.Fish.Res.34, 179–189.

Berkeley,S.A.,Hixon,M.A.,Larson,R.J.,Love,M.S.,2004.Fisheriessustainabilityvia protectionofagestructureandspatialdistributionofﬁshpopulations.

Fisheries29,23–32,http://dx.doi.org/10.1577/1548-844629[23:FSVPOA]2.0.

CO;2.

Beverton,R.J.H.,Holt,S.J.,1957.Onthedynamicsofexploitedﬁshpopulations.

Fish.Investig.

Brunel,T.,2010.Age-structure-dependentrecruitment:ameta-analysisappliedto NortheastAtlanticﬁshstocks.ICESJ.Mar.Sci.J.Cons.67,1921–1930.

Bullough,L.W.,Napier,I.R.,Laurenson,C.H.,Riley,D.,Fryer,R.J.,Ferro,R.S.T., Kynoch,R.J.,2007.Ayear-longtrialofasquaremeshpanelinacommercial demersaltrawl.Fish.Res.83,105–112,http://dx.doi.org/10.1016/j.ﬁshres.

2006.09.008.

Cardinale,M.,Arrhenius,F.,2000.Theinﬂuenceofstockstructureand

environmentalconditionsontherecruitmentprocessofBalticcodestimated usingageneralizedadditivemodel.Can.J.Fish.Aquat.Sci.57,2402–2409, http://dx.doi.org/10.1139/f00-221.

Catchpole,T.L.,Revill,A.S.,2008.GeartechnologyinNephropstrawlﬁsheries.Rev.

FishBiol.Fish.18,17–31,http://dx.doi.org/10.1007/s11160-007-9061-y.

Cervi ˜no,S.,Domínguez-Petit,R.,Jardim,E.,Mehault,S.,Pi ˜neiro,C.,Saborido-Rey,F., 2013.ImpactofeggproductionandstockstructureonMSYreferencepoints anditsmanagementimplicationsforsouthernhake(Merlucciusmerluccius).

Fish.Res.138,168–178,http://dx.doi.org/10.1016/j.ﬁshres.2012.07.016.

Dickson,W.,Smith,A.,Walsh,S.,1995.MethodologyManual:Measurementof FishingGearSelectivity.CanadaDept.ofFisheriesandOceans,Ottawa.

Feekings,J.,Lewy,P.,Madsen,N.,Marshall,C.T.,2013.Theeffectofregulation changesandinﬂuentialfactorsonAtlanticcoddiscardsintheBalticSea demersaltrawlﬁshery.Can.J.Fish.Aquat.Sci.70,534–542,http://dx.doi.org/

10.1139/cjfas-2012-0273.

Fryer,R.J.,1991.Amodelofbetween-haulvariationinselectivity.ICESJ.Mar.Sci.J.

Cons.48,281–290,http://dx.doi.org/10.1093/icesjms/48.3.281.

Garcia,S.M.,Kolding,J.,Rice,J.,Rochet,M.-J.,Zhou,S.,Arimoto,T.,Beyer,J.E., Borges,L.,Bundy,A.,Dunn,D.,Fulton,E.A.,Hall,M.,Heino,M.,Law,R.,Makino, M.,Rijnsdorp,A.D.,Simard,F.,Smith,A.D.M.,2012.Reconsideringthe consequencesofselectiveﬁsheries.Science335,1045–1047,http://dx.doi.org/

10.1126/science.1214594.

(8)

Glass,C.W.,2000.Conservationofﬁshstocksthroughbycatchreduction:areview.

NortheastNat.7,395–410.

J.A.GullandFishstockassessment:amanualofbasicmethods,FAO/WileySeries onFoodandAgriculture1983.

He,P.,Balzano,V.,2013.Anewshrimptrawlcombinationgridsystemthatreduces smallshrimpandﬁnﬁshbycatch.Fish.Res.140,20–27,http://dx.doi.org/10.

1016/j.ﬁshres.2012.11.009.

He,P.,Balzano,V.,2012.Theeffectofgridspacingonsizeselectivityofshrimpsin apinkshrimptrawlwithadual-gridsize-sortingsystem.Fish.Res.121–122, 81–87,http://dx.doi.org/10.1016/j.ﬁshres.2012.01.012.

He,P.,Balzano,V.,2011.Ropegrid:anewgriddesigntofurtherreducefinfish bycatchintheGulfofMainepinkshrimpfishery.Fish.Res.111,100–107, http://dx.doi.org/10.1016/j.fishres.2011.07.001.

Herrmann,B.,Sistiaga,M.,Larsen,R.B.,Nielsen,K.N.,2013.Sizeselectivityof redﬁsh(Sebastesspp.)intheNortheastAtlanticusinggrid-basedselection systemsfortrawls.Aquat.LivingResour.26,109–120,http://dx.doi.org/10.

1051/alr/2013051.

Herrmann,B.,Sistiaga,M.,Nielsen,K.N.,Larsen,R.B.,2012.Understandingthesize selectivityofredﬁsh(Sebastesspp.)inNorthAtlantictrawlcodends.J.

NorthwestAtl.Fish.Sci.44,1–13.

Herrmann,B.,Wienbeck,H.,Karlsen,J.D.,Stepputtis,D.,Dahm,E.,Moderhak,W., 2015.UnderstandingthereleaseefﬁciencyofAtlanticcod(Gadusmorhua) fromtrawlswithasquaremeshpanel:effectsofpanelarea,panelposition, andstimulationofescaperesponse.ICESJ.Mar.Sci.72,686–696,http://dx.doi.

org/10.1093/icesjms/fsu124.

Hixon,M.A.,Johnson,D.W.,Sogard,S.M.,2014.BOFFFFs:ontheimportanceof conservingold-growthagestructureinﬁsherypopulations.ICESJ.Mar.Sci.71, 2171–2185,http://dx.doi.org/10.1093/icesjms/fst200.

Hovgård,H.,Lassen,H.,2000.Manualonestimationofselectivityforgillnetand longlinegearsinabundancesurveys.In:FAOFisheriesTechnialPaper.FAO, Rome.

Huse,I.,2000.Relativeselectivityintrawl,longlineandgillnetﬁsheriesforcodand haddock.ICESJ.Mar.Sci.57,1271–1282,http://dx.doi.org/10.1006/jmsc.2000.

0813.

Isaksen,B.,Valdemarsen,J.W.,Larsen,R.B.,Karlsen,L.,1992.Reductionofﬁsh by-catchinshrimptrawlusingarigidseparatorgridintheaftbelly.Fish.Res.

13,335–352,http://dx.doi.org/10.1016/0165-7836(92)90086-9.

Jacobsen,N.S.,Gislason,H.,Andersen,K.H.,2013.Theconsequencesofbalanced harvestingofﬁshcommunities.Proc.R.Soc.BBiol.Sci.281,20132701,http://

dx.doi.org/10.1098/rspb.2013.2701.

Jørgensen,C.,Enberg,K.,Dunlop,E.S.,Arlinghaus,R.,Boukal,D.S.,Brander,K., Ernande,B.,Gårdmark,A.G.,Johnston,F.,Matsumura,S.,Pardoe,H.,Raab,K., Silva,A.,Vainikka,A.,Dieckmann,U.,Heino,M.,Rijnsdorp,A.D.,2007.Ecology:

managingevolvingﬁshstocks.Science318,1247–1248,http://dx.doi.org/10.

1126/science.1148089.

Jørgensen,T.,Ingólfsson,Ó.A.,Graham,N.,Isaksen,B.,2006.Sizeselectionofcodby rigidgrids—isanythinggainedcomparedtodiamondmeshcodendsonly?

Fish.Res79,337–348,http://dx.doi.org/10.1016/j.ﬁshres.2006.01.017.

Kronbak,L.G.,Nielsen,J.R.,Jørgensen,O.A.,Vestergaard,N.,2009.Bio-economic evaluationofimplementingtrawlﬁshinggearwithdifferentselectivity.J.

Environ.Manage.90,3665–3674,http://dx.doi.org/10.1016/j.jenvman.2009.

07.008.

Kuparinen,A.,Merilä,J.,2007.Detectingandmanagingﬁsheries-induced evolution.TrendsEcol.Evol.22,652–659,http://dx.doi.org/10.1016/j.tree.

2007.08.011.

Kvamme,C.,Isaksen,B.,2004.Totalselectivityofacommercialcodtrawlwithand withoutagridmounted:gridandcodendselectivityofnorth-eastArticcod.

Fish.Res.68,305–318,http://dx.doi.org/10.1016/j.ﬁshres.2003.11.011.

Law,R.,2000.Fishing,selection,andphenotypicevolution.ICESJ.Mar.Sci57, 659–668,http://dx.doi.org/10.1006/jmsc.2000.0731.

Law,R.,Kolding,J.,Plank,M.J.,2015.Squaringthecircle:reconcilingﬁshingand conservationofaquaticecosystems.FishFish.16,160–174,http://dx.doi.org/

10.1111/faf.12056.

Macher,C.,Guyader,O.,Talidec,C.,Bertignac,M.,2008.Acost-beneﬁtanalysisof improvingtrawlselectivityinthecaseofdiscards:theNephropsnorvegicus ﬁsheryintheBayofBiscay.Fish.Res.92,76–89,http://dx.doi.org/10.1016/j.

ﬁshres.2007.12.021.

MacLennan,D.N.,1995.Gearselectivityandthevariationofyield.ICESJ.Mar.Sci.

52,827–836,http://dx.doi.org/10.1006/jmsc.1995.0097.

MacLennan,D.N.,1992.Fishinggearselectivity:anoverview.Fish.Res.13, 201–204,http://dx.doi.org/10.1016/0165-7836(92)90076-6.

Madsen,N.,2007.SelectivityofﬁshinggearsusedintheBalticSeacodﬁshery.Rev.

FishBiol.Fish.17,517–544,http://dx.doi.org/10.1007/s11160-007-9053-y.

Marshall,D.J.,Heppell,S.S.,Munch,S.B.,Warner,R.R.,2010.Therelationship betweenmaternalphenotypeandoffspringquality:dooldermothersreally producethebestoffspring?Ecology91,2862–2873,http://dx.doi.org/10.1890/

09-0156.1.

Marteinsdottir,G.,Begg,G.A.,2002.Essentialrelationshipsincorporatingthe inﬂuenceofage,sizeandconditiononvariablesrequiredforestimationof reproductivepotentialinAtlanticcodGadusmorhua.Mar.Ecol.Prog.Ser.,235.

Millar,R.B.,Fryer,R.J.,1999.Estimatingthesize-selectioncurvesoftowedgears, traps,netsandhooks.Rev.FishBiol.Fish.9,89–116,http://dx.doi.org/10.1023/

A.1,008,838,220,001.

Millar,R.B.,Holst,R.,1997.Estimationofgillnetandhookselectivityusing log-linearmodels.ICESJ.Mar.Sci.54,471–477,http://dx.doi.org/10.1006/

jmsc.1996.0196.

Morgan,M.J.,Perez-Rodriguez,A.,Saborido-Rey,F.,2011.Doesincreased informationaboutreproductivepotentialresultinbetterpredictionof recruitment?Can.J.Fish.Aquat.Sci.68,1361–1368,http://dx.doi.org/10.1139/

f2011-049.

O’Farrell,M.R.,Botsford,L.W.,2006.Theﬁsheriesmanagementimplicationsof maternal-age-dependentlarvalsurvival.Can.J.Fish.Aquat.Sci.63,2249–2258, http://dx.doi.org/10.1139/f06-130.

Ottersen,G.,Hjermann,D.Ø.,Stenseth,N.C.,2006.Changesinspawningstock structurestrengthenthelinkbetweenclimateandrecruitmentinaheavily ﬁshedcod(Gadusmorhua)stock.Fish.Oceanogr.15,230–243,http://dx.doi.

org/10.1111/j.1365-2419.2006.00404.x.

Reeves,S.A.,Armstrong,D.W.,Fryer,R.J.,Coull,K.A.,1992.Theeffectsofmeshsize, cod-endextensionlengthandcod-enddiameterontheselectivityofScottish trawlsandseines.ICESJ.Mar.Sci.49,279–288,http://dx.doi.org/10.1093/

icesjms/49.3.279.

Sala,A.,Lucchetti,A.,Affronte,M.,2011.EffectsofTurtleExcluderDeviceson bycatchanddiscardreductioninthedemersalﬁsheriesofMediterraneanSea.

Aquat.LivingResour.24,183–192,http://dx.doi.org/10.1051/alr/2011109.

Sistiaga,M.,Herrmann,B.,Grimaldo,E.,Larsen,R.B.,2010.Assessmentofdual selectioningridbasedselectivitysystems.Fish.Res.105,187–199,http://dx.

doi.org/10.1016/j.ﬁshres.2010.05.006.

Trippel,E.A.,Kraus,G.,Köster,F.W.,2005.Maternalandpaternalinﬂuenceson earlylifehistorytraitsandprocessesofBalticcodGadusmorhua.Mar.Ecol.

Prog.Ser.303,259–267.

Wileman,D.,Ferro,R.S.T.,Fonteyne,R.,Millar,R.B.,1996.Manualofmethodsof measuringtheselectivityoftowedﬁshinggears.ICESCoop.Res.Rep.

Zhou,S.,Smith,A.D.M.,Punt,A.E.,Richardson,A.J.,Gibbs,M.,Fulton,E.A.,Pascoe,S., Bulman,C.,Bayliss,P.,Sainsbury,K.,2010.Ecosystem-basedﬁsheries managementrequiresachangetotheselectiveﬁshingphilosophy.Proc.Natl.

Acad.Sci.107,9485–9489,http://dx.doi.org/10.1073/pnas.0912771107.