RESEARCHARTICLE Positive Catch & Economic Benefits of Periodic Octopus Fishery Closures: Do Effective, Narrowly Targeted Actions ‘ ’ Catalyze Broader Management? ThomasA.Oliver1,2‡*,KirstenL.L.Oleson1,3‡,HajanainaRatsimbazafy1, DanielRaberinary1,SophieBenbow1,AlasdairHarris1,4 1 BlueVenturesConservation,39–41NorthRoad,London,N79DP,UnitedKingdom,2 Departmentof Biology,UniversityofHawaiʻiMānoa,2853McCarthyMall,EdmondsonHall309,Honolulu,Hawaii,96822, UnitedStatesofAmerica,3 DepartmentofNaturalResourcesandEnvironmentalManagement,University ofHawaiʻiMānoa,1910EastWestRoad,Sherman101,Honolulu,Hawaii,96822,UnitedStatesofAmerica, 4 EnvironmentalChangeInstitute,UniversityofOxford,SouthParksRoad,Oxford,OX13QY,United Kingdom ‡Theseauthorsareco-primaryauthorsonthiswork. * [email protected] OPENACCESS Citation:OliverTA,OlesonKLL,RatsimbazafyH, RaberinaryD,BenbowS,HarrisA(2015)Positive Abstract Catch&EconomicBenefitsofPeriodicOctopus FisheryClosures:DoEffective,NarrowlyTargeted Actions‘Catalyze’BroaderManagement?PLoSONE 10(6):e0129075.doi:10.1371/journal.pone.0129075 Overview AcademicEditor:DennisM.Higgs,Universityof EightyearsofoctopusfisheryrecordsfromsouthwestMadagascarrevealsignificantposi- Windsor,CANADA tiveimpactsfrom36periodicclosureson:(a)fisherycatchesand(b)villagefisheryincome, Received:February14,2014 suchthat(c)economicbenefitsfromincreasedlandingsoutweighcostsofforegonecatch. Accepted:May4,2015 Closurescovered~20%ofavillage’sfishedareaandlasted2-7months. Published:June17,2015 FisheryCatchesfromEachClosedSite Copyright:©2015Oliveretal.Thisisanopen accessarticledistributedunderthetermsofthe Octopuslandingsandcatchperuniteffort(CPUE)significantlyincreasedinthe30daysfol- CreativeCommonsAttributionLicense,whichpermits lowingaclosure’sreopening,relativetothe30daysbeforeaclosure(landings:+718%, unrestricteduse,distribution,andreproductioninany p<0.0001;CPUE:+87%,p<0.0001;n=36).Open-accesscontrolsitesshowednobefore/ medium,providedtheoriginalauthorandsourceare credited. afterchangewhentheyoccurredindependentlyofothermanagement(“noban”,n=17/36). Ontheotherhand,open-accesscontrolsitesshowedmodestcatchincreaseswhenthey Funding:Fundingforthisstudywasprovidedbythe WaterlooFoundation,theJohnD.andCatherineT. extendeda6-weekseasonalfisheryshutdown(“ban”,n=19/36).Theseasonalfishery MacArthurFoundation,theUSNSFOISE0853086, shutdownaffectstheentireregion,soconfoundallpotentialcontrolsites. andtheNetworkforSocialChange.Thefundershad noroleinstudydesign,datacollectionandanalysis, FisheryIncomeinImplementingVillages decisiontopublish,orpreparationofthemanuscript. CompetingInterests:Theauthorshavereadthe Invillagesimplementingaclosure,octopusfisheryincomedoubledinthe30daysaftera journal'spolicyandhavethefollowingpotential closure,relativeto30daysbefore(+132%,p<0.001,n=28).Controlvillagesnotimple- conflicts:Theauthorsareeitheradvisorsto(Oliver, mentingaclosureshowednoincreaseinincomeafter“noban”closuresandmodestin- Oleson)oremployeesofBlueVentures creasesafter“ban”closures.Villagesdidnotshowasignificantdeclineinincomeduring Conservation,theNGOthatco-managesthe VelondriakeLocallyManagedMarineArea(LMMA). closureevents. PLOSONE|DOI:10.1371/journal.pone.0129075 June17,2015 1/24 Catch&EconomicBenefitsofPeriodicOctopusFisheryClosures Thepresentedworkanalyzesfisherymanagement NetEconomicBenefitsfromEachClosedSite techniquesperformedintheVelondriakeLMMA,by Landingsinclosuresitesgeneratedmorerevenuethansimulatedlandingsassumingcon- theco-managingpartners:(1)theVelondriake Committee,amanagementbodyrepresentingthe25 tinuedopen-accessfishingatthatsite(27/36showpositivenetearnings;mean+$305/clo- villagesoftheLMMAand(2)BlueVentures sure;mean+57.7%monthly).Benefitsaccruedfasterthanlocalfishers’timepreferences Conservationstaff.Thesefactsdonotalterthe during17-27ofthe36closures.Highreportedratesofillegalfishingduringclosurescorre- authors'adherencetoPLoSONEpoliciesonsharing dataandmaterials. latedwithpooreconomicperformance. BroaderCo-Management Wediscusstheimplicationsofourfindingsforbroaderco-managementarrangements,par- ticularlyforcatalyzingmorecomprehensivemanagement. Introduction Asover-exploitationandglobalchangethreatenreefsworldwide,sustainablymanagingcoral reefsiscrucialtoprotectingbothreefbiodiversityandthefoodsecurityofhundredsofmillions ofcoastalpeople[1–4].Becausetwo-thirdsofallreefslieindevelopingcountries[5],thegoal ofconservingreefsgloballyrequiresmanagementstrategiesthatcaneffectivelybalanceboth conservationanddevelopmentgoals.Thisdevelopingworldsettingfrequentlyincludeshigh populationgrowthrates,lowincomes,andweaknational-scalegovernance[6–8].Inthiscon- text,localcommunities’supportformanagementactionsiscrucialtoeffectivelyprotectbiodi- versityandhumanlivelihoods[9–13]. Thereisagrowingbodyofresearchreportingoncoastalmanagementeffortsdesigned,en- forced,andmaintainedbycommunitiesorcommunitieswithanexternalpartner(co-manage- ment)[9,14–16].Employingabroadarrayofmeasures,communityandco-management arrangementsaroundtheworldhaveproducedpositiveoutcomesforbothconservationand developmentgoals[9,13,16].Wheneffective,sucharrangementscanhelpcommunitiesbetter managetheirresourcesoverthelongterm,helpingthembreakfromthetragedyofthecom- mons,whereopenaccessleadstooverexploitation,andfromresource-dependentpoverty traps,wherenaturalresourcedepletionanddependencereinforceeachother[17–19].Howev- er,whilecommunityandco-managementmodelsarebecomingmorecommon,quantitative impactassessmentsremainuncommonandmanymanagementfailuresareunder-reported. Theseresearchgapshinderrobustgeneralizationsabouttheeffectivenessofcommunityand co-managementapproaches[9,11]. Theperiodicfisheryclosure,inwhichfisherstemporarilyrefrainfromharvestinginspecific areas[20,21]isanincreasinglypopularcommunity-basedtoolwithagrowingbaseofempirical support[22–24].Periodicclosureshavelongbeenapartoftraditionalfishingculturesacross theIndo-Pacific[20,25–27],andstillplayanactiveroleincommunitymanagementofmarine resourcesintheregion[22–24,28].Periodicharvest,orpulsefishing,alsohasbeenacommonly discussedstrategyinthewesternfisheriesliterature[29],andhasbeensuggestedasaviableal- ternativetoconstant,orstationary,fishingyieldssinceatleastthe1970s[30,31]. Manyperiodicharvestregimeshavebeendesignedwithasingle-speciesinmind[32].Prac- ticalexamplesfrombothmodelsandfielddatagenerallytargetsedentarymarineinvertebrates, andhighlightthaturchins[33],seascallops[34,35],andabalone[36]makegoodcandidates foraperiodicregime.PeriodicclosureregimesinthetropicalIndo-Pacifichaveshownpositive effectsonabundanceingiantclams(Tridacnaspp.)[22]andvariedresultsfortrochus(aka topshell,Tectusniloticus)[22–24,37,38]. PLOSONE|DOI:10.1371/journal.pone.0129075 June17,2015 2/24 Catch&EconomicBenefitsofPeriodicOctopusFisheryClosures Periodicharveststrategiesinartisanalcontextsfrequentlyapplynottosingletargetspecies butinsteadtomulti-speciesassemblages,includingrelativelylong-livedreeffishes[22– 24,28,39,40].Thefewstudiesthathaveshownpositiveeffectsofperiodicclosuresonmixed reeffishfisheriesnotedincreasesinfishbiomassinperiodicallyclosedareasrelativetoopenac- cesssites[22,23].Allthreefocalareasinthesestudies,however,werecharacterizedbysmall humanpopulationsexertinglowfishingpressureonresourcestowhichtheyhaveexclusive tenure[22,23];theresultsdonotholdinareassubjecttohigherfishingpressures[28,39–41], perhapsbecauseclosureperiodsweretooshort,openperiodstoolong,orfishingintensitydur- ingopenperiodswastoointensetosupportrobustrecoveryfromfishingmortality[28,39–41]. Anotherreasonmaybethatinareaswithhigherpressureandcompetition,fisherpopulations preferimmediaterewardfromlandingasmallercatchtodayoveralargerandmoreuncertain futurecatch[42]. Whileresultsfromthefieldhavebeenvariable,modelsoffisherieseconomicssuggestthat incertaincasesaperiodicharvestcanprovideabettereconomicyieldthanstationaryharvest [30,31],specificallywhenthefisheryhaslowselectivity[31,43].Afishery’soptimalopening/ closingcycle(i.e.,thepulse-length)isafunctionofboththetargetspecies’biology(i.e.,specifi- callythetargetspecies’growthrateandlife-span)andthefishery’seconomics(i.e.,landing pricesandthelocalfisher’sdiscountrate,theirtimepreferenceforimmediateversusdelayed reward)[43].Thetimebetweenopeningsvariesdramaticallydependingonthetargetspecies’ biology[32],andhigherdiscountratesleadtoeithershorteningtheoptimalclosuredurations orshiftingtheeconomicoptimumtostationary,ratherthanperiodicharvest[31,44]. Modelsandexperiencesuggestthatthesuccessorfailureofaperiodicclosureregimede- pendsonthegovernancesystem’sabilitytomatchfishingpatternstoafishery’s“optimal”peri- odicharvestschedule[23,24,29].Factorsshowntoimprovetheoddsofmatchingactualand optimalharvestinthecontextofperiodicclosuresinclude:exclusivetenuretotheresourcein question,respectedandlegitimateleadership,highsocialcapital,lowfishingpressure,loweffi- ciencygears,androbustecologicalknowledge[22–24,45].Notsurprisingly,thesegovernance factorsmirrorthosethatmoregenerallycorrelatewithsuccessfulcommunity/co-management [13,16,46]. Experiencewithsuccessful,targetedmanagementmightalsoserveasacatalystforbroader communitymanagement[24,45,47].InVanuatu,government-sponsoredmanagementefforts employingarangeofinterventions,includingperiodicclosures,ledtocommunityengagement withmanagersandco-managementofmanyotherspeciesoffishandinvertebrates[47].InIn- donesia,villageswithactiveorlapsedperiodicclosuretraditionsshowedbroader,moreactive marinemanagementthanvillageswithnosuchtradition[45].Commonstheorysuggeststhat communitiesaremorelikelytoengageinmanagementwhenexpectedbenefitsoutweighthe perceivedcostsofmanagement[17].Inthesecases,successfuldemonstrationsofdesirableben- efit:costratioslikelyinformedexpectations,whileofferinganopportunitytobuildgovernance capacityandsocialcapitalneededtobroadenmanagementefforts[17]. Herewepresentananalysisofthefisheryandeconomiceffectsofperiodicoctopusfishery closuresintheVelondriakeLocallyManagedMarineArea(LMMA)insouthwestMadagascar (Fig1).Thisworkservestofillresearchgapsbyprovidingempiricalimpactassessmentsofco- managementoutcomesforaspecificperiodicfisheryclosureregime.Establishingthebaseline efficacyoftheseinterventionsisparticularlytimely,astheuseofperiodicclosuresasafisheries managementtoolisproliferatingacrossthewesternIndianOcean[48]. Todoso,first,wequantifyeffectsonsite-specificlandingsandcatchperuniteffort(CPUE) frommultipleperiodicclosureeventscomparedtopairedcontrols.Second,weexamineocto- pusfishery-generatedincomeaccruedatthevillagelevel.Third,weassesswhetherindividual closedsitesgenerateneteconomicbenefits,andcomparetheratesatwhichthesebenefitsare PLOSONE|DOI:10.1371/journal.pone.0129075 June17,2015 3/24 Catch&EconomicBenefitsofPeriodicOctopusFisheryClosures Fig1.MapsofStudyArea.(A)Large-scalemapofMadagascarandtheAfricancontinent,(B)Insetofthe 25villagesoftheVelondriakeLocallyManagedMarineAreainsouthwesternMadagascar.Vertical boxextentis~75km.(C)Representativeexampleofaperiodicoctopusfisheryclosure.Indicatedinthemap aretwovillages,AndavadoakaandAmpasilava,withtheirrespectiveoctopusfishingsitesmappedinorange andyellow.Ingreen,youcanseethesitesAmagnahitseandNosinkara,inwhichthesetwovillageshave repeatedlyco-implementedaperiodicoctopusfisheryclosure. doi:10.1371/journal.pone.0129075.g001 generatedtolocalfishers’timepreferences(Fig2).Finally,asbroaderco-managementefforts intheLMMAfollowedthewidespreadadoptionoftheoctopusclosureregime,wediscussa fertileareaforfutureresearchtestingthehypothesisthateffectiveperiodicclosurescanserve asacatalystforbroadercommunitymanagement. Methods 2.1.MarineResourceManagementintheVelondriakeLMMA Startingin2003,thenon-governmentalorganizationBlueVentures,withlocalandinterna- tionalpartners(InstitutHalieutiqueetdesScièncesMarines,WildlifeConservationSociety, COPEFRITO)beganaseriesofmeetingswiththecommunityofAndavadoakainsouthwest Madagascartodiscussapproachestomanaginglocalmarineresources.Ininitialconversations, thecommunitydemurredfromengaginginpermanentno-takeareas,butwaswillingtoat- tempta7-monthclosureofoctopusfishingonashallowoffshorereefbeginningNovember1, 2004[49]. Afterafavorableinitialreception,thisclosureregimespread.Locally,the25villagesthat nowcomposethe~1,000km2VelondriakeLMMAoversaw69differentoctopusclosuresbe- tween2004and2011[48,50].AnAfricanDevelopmentBankprojectsupported50additional closuresaroundsouthwestMadagascarbetween2009and2013[49].Further,beginningin 2005thenationalgovernmentformalizedthecommunityinitiativebyshuttingdowntheentire southwestregionoctopusfisheryforsixweeksbetweenmid-DecemberandlateJanuary[49]. Themodelalsospreadinternationally,withtheneighboringislandstateofMauritiusenacting similarlegislationin2012[51].Followingthespreadoftheoctopusclosureregime,theVelon- driakeregionalmanagementcommitteetookbroadermanagementstepswithintheLMMA, institutingperiodicmangroveclosurestargetedatalocalcrabfishery,banningdestructivefish- ingpractices,engaginginecologicalmonitoring,and,fiveyearsafterrefusingtheidea,institut- ingthefirstofnowsixpermanent,community-enforcedno-takeareas[48,49]. TheoctopusfisheryinVelondriaketargetsagroupoffourshallow-waterspecies:Octopus cyanea(95%oflocalcatches),Callistoctopusmacropus(~4%),Amphioctopusaegina(~1%), Callistoctopusornatus(rare)(D.Raberinarypers.comm,[52]).Thesefouroctopusspecies PLOSONE|DOI:10.1371/journal.pone.0129075 June17,2015 4/24 Catch&EconomicBenefitsofPeriodicOctopusFisheryClosures Fig2.Experimentaldesignandsamplessizesusedtoinvestigateeffectsofperiodicfisheryclosures.(A)Sitefisherycatches,(B)Villagefishery income,and(C)Siteneteconomicbenefts.Colorshighlightthedistinctionsamong“no-ban”and“ban”closures,andbetweenclosuresites/villagesand eitheropen-accesscontrols(A&B)orsimulatedlandings(C). doi:10.1371/journal.pone.0129075.g002 eachhavealifecycleofaboutoneyear,dispersingasparalarvaefor2–3months,thengrowing over6–9monthsfrom<1gatsettlementtocommonlyabove3kg[53–55].Theyappeartobe year-roundspawners,althoughrecentstudiessuggestthatrecruitmentfluctuatesthroughout theyear[52]. Thebulkofoctopusiscaughtduringspringlowtidesbygleaners,predominatelywomen. Theygenerallysellanyoctopusoverthenationallyregulatedminimumsizeof350gtooutside buyers[56,57].Thoughonly180kmnorthofToliara,thisregionlackstransportinfrastructure, renderingtheisolatedvillagesdependentonprivateexportingcompaniesformarketaccess. Uponinstitutinganoctopusclosure,villagerstypicallycloseaboutonefifthoftheirvillage’s octopusharvestarea(~124ha+/-45CI95),foraperiodbetween2–7months,sometimesre- peatedly(Fig1C;KeyInformantInterviews,ShawnPeabody,co-manager).TheVelondriake Committee,anelectedmanagementbody,selectsclosuresites,choosesclosuredurations,and coordinatesmanagement.Communitiesself-enforcetheclosures,sanctionsareprescribedby locallaw(dina),andenforcedbyconsensusatcommunitymeetings[58].BlueVenturesCon- servation,theco-managingnon-governmentalorganization,providestechnicalandfunding supportformanagementeffortsincooperationwiththepartnersmentionedabove[49]. PLOSONE|DOI:10.1371/journal.pone.0129075 June17,2015 5/24 Catch&EconomicBenefitsofPeriodicOctopusFisheryClosures 2.2.Socioeconomicbackgroundandsurveys Mostoftheapproximately7,500peoplelivinginVelondriake(S1andS2Tables,S1File)are Vezo,asubgroupoftheSakalavaethnicgroupwhoseculturalidentityistiedtoafishingand gleaninglifestyle[59].Consistentwiththeirlivinginaverypoornation,Vezopopulationsare frequentlycharacterizedbylowincomes,highresourcedependence,andrapidpopulation growth(~3%annually)[60]. Toexpandthelocalinformationavailable,asocio-economichouseholdsurveywascon- ductedbetweenAugustandSeptember2010across16villagesand301households(alsosee [61]).The35-questionhouseholdsurveycollecteddataonhouseholddemographics,income sources,fishingpractices,wealth,andresourceextractionhabits.Thesurveydesignwasbased onregionalsocio-economicmonitoringguidelines[62]andvalidityrecommendations[63].To ensurevalidity,atrained,localVezosurveyteamundertookthesurveyinVezo;abi-lingual fieldmanagersupervisedtheteams.Pilotsurveysinthreevillageshelpedinformthefinalsur- vey.AllsurveydataweredoubleenteredinExcelandqualitycontrolled. Villageswerestratifiedaccordingtogeographicregion(north,central,south)toaccountfor proximitytomarket,andsurroundinghabitat(island,coast,mangrove,inland)toaccountfor differencesinfishinghabits.Inlandvillageswereeliminatedfromthestudyduetotheirgreater dependenceonfarmingratherthanfishingandnosouth-islandvillagesexist.Theeightre- mainingstrata(north-island,north-coastal,north-mangrove,central-island,central-coast, central-mangrove,south-coast,south-mangrove)allowforextrapolationtonon-sampledvil- lages.Whenpossible,fortyrandomhouseholdsweresampledineachstratum,andfemaleand maleheadsofhouseholdwerealternatelyinterviewed(S1Table).Uponenteringavillage,each memberofthesurveyteampickedarandomnumberbetween1and12representingthedirec- tionheorshehadtowalk(e.g.,3meant3o’clock).Walkinguptoeachhouseholdalongthat trajectory,surveyorsconsultedalistofpreviouslygeneratedrandomnumbersbetween1and 100;ifthenumberwasbelow“X”thenthehouseholdwassampled.“X”wasdifferentforeach village,andisthe(#ofsampledhouseholdsdesired)/(thetotal#ofhouseholdsinthatvillage). FocusgroupswithfishersandgleanersinthemainvillageofAndavadoakaprovideddata onmarketpricespaidforcatch,quantityandcostofgearused,andseasons(totalof12focus groups).Interviewswithfishers,commercialbuyers,localmiddlemen(“souscollectors”),fish mongers,managers,andvillagersacrossVelondriake(totalof26interviews)providedinforma- tionon:marketprices,patternsofcommunitydecisionmaking,localengagementwithman- agement,etc.Focusgroupparticipantsandkeyinformantswereopportunisticallysampled, andsnowballsamplingidentifiedadditionalparticipants/informants.Marketsurveysanddi- rectobservationscorroboratedinformationsuchasmarketpricesoffishandgear. 2.3.LandingsData,SalePrice,ParticipatoryMapping Since2004,traineddatacollectorsrecordedoctopuslandingsacrossVelondriakeatthepoint ofsale.Eachday,collectorswaitedatthepointofsaleineachparticipatingvillage,whichal- lowedeasycollectionofalargeproportion(ifnotcompletecoverage)oftheday’scatch.Collec- torsrecordedeachfishingtripincludingthenumberoffishersinthegroup,numberof octopuscaught,weightofeachindividualoctopus,fishingsite,date,villagename,geartype, fishernames,fisherages,andfishergenders.Datacollectedafter2008includesoctopussexas well.Thedataset’s258,108individualweighedoctopusesfrom67,990tripsweredouble-en- tered,cross-checked,andqualitycontrolledin2010–2011. Thepriceperkilogramoctopuswasassessedthroughdirectobservationsatpointsofsalein June-August2009,andtrendsinthe“beachprice”overtimewereconfirmedthroughfocus PLOSONE|DOI:10.1371/journal.pone.0129075 June17,2015 6/24 Catch&EconomicBenefitsofPeriodicOctopusFisheryClosures groupsandkeyinformantinterviews(S2Table).Priceswereadjustedforinflationandpur- chasingpower[8]. In2009–10,weconductedparticipatorymappingexerciseswithfishersinallVelondriake’s villagestodefineboundariesofeachfishingsite.Thisexerciseimprovedthemanagers’andre- searchers’abilitytotranslatebetweenlocalsitenamesandspecificfishingareas.Afocusgroup ofeachvillage’soctopusfisherswasaskedtodelineatetheirvillage’snamedsitesonlarge,lami- natedpostersshowingthesatelliteimagery(GoogleEarth)ofthevillage’scoast.Thesemaps werethendigitizedandtransferredintoGISshapefiles.Eachsitewasgivenauniqueidentifier, cross-referencedtothelocalsitenames,andconfirmedwithvillagers[64]. 2.4.IllegalCatchRate Asameasureofcompliancewithclosures,weassessedtheseverityofreportedillegalfishingin theclosedsitesas“low,”“moderate,”or“high.”Toassignthesecategories,weassessedlevelsof illegalfishingreportedinthelandingsdataduringaclosurerelativetobaselinecatches,defined astotallandingsfromtheclosuresiteinthe30daysbeforeaclosure.Here,“low”indicatesthat octopuscatchesrecordedduringtheclosureequaled0–5%ofbaseline“before”catches;“mod- erate”,5–50%ofbaseline;and“high”atleast50%.Fishersreadilyreportedthisactivity,but neverthelessweconsiderthesereportsasaminimumestimateofillegalactivity. 2.5.Fisheryeffectsanalysis:Landings,effort&catchperuniteffort WeusedaBefore,After,Control,Impact(BACI)statisticaldesignandmixedmodelANOVA totesttheeffectsofoctopusclosuresintheVelondriakeLMMAon(a)octopusfisherylandings (kgoctopus/30days),(b)octopusfishereffort(totalfisher-days/30days),and(c)catchper uniteffort(kgoctopus/fisher-day).Asubsetof36closureeventshadadequatebaselinedata, definedasatleast5fishersand10octopusesrecordedineachofthe30-dayperiodsbeforeand aftertheclosure(Fig2A). Seventeen(17)ofthese36closuresoccurredindependentlyofothermanagementmeasures, whiletheother19extendedthesix-weekgovernmentally-imposedregionaloctopusfishery shutdownthatwasineffectinaustralsummereachyearbeginningin2005.Wereferthrough- outthepapertothe17independentlyoccurringclosuresas“no-ban”closures,andthe19clo- suresthatextendedtheshutdownas“ban”closures(Fig2). ControlSiteSelection. Wematchedeachofthe36focalclosuresiteswithasimilarcon- trolsitethat(a)neverhadalocalclosure,(b)showedtrendsinbaselinedatasimilartothoseof theclosedsite(i.e.,thesite’smonthlyoctopuslandings;seebaselinelandingscorrelation(r) below),and(c)hadadequatedataavailableduringthefocalperiods(seerelativedataavailabil- ityindexbelow).Wetookfivestepstoestablishasetofimpact-controlpairings.(1)Wepre- paredabaselinedatasetfreeof‘closureeffects’byremovingdatafromeachknownclosuresite fortheperiodfromclosureto60daysafterreopening.(2)Generatebaselinelandingscorrela- tion:Tohighlightsitepairsthatpresentedcorrelatedbaselinetrends,weaggregatedthetotal baselinelandingsofoctopusbymonthforeachfishingsite,andthencomparedthefullrecords ofthe36focalclosedsitesto318potentialcontrolsitesusingPearson’scorrelationofmonthly catchtotals(i.e.,r).(3)Generaterelativedataavailabilityindex:Toassessrelativeavailability ofdataatpotentialcontrolsites,wecountedtheminimumnumberoffisher-daysavailablein eitherthe30daysbeforeoraftertheclosure,dividedbythevaluefromthecontrol-closure pairingwiththehighestrecordedfisher-days.Minimumavailablefisher-daysduringfocalperi- odsinselectedcontrolsitesrangedfrom6to120,withameanof27.6fisher-days.(4)Wethen rankedeachofthe11,448potentialcontrolsite-closuresitepairingsbasedonasuitabilityscore thatwascomposedoftheaverageofthepair’sbaselinelandingscorrelationandrelativedata PLOSONE|DOI:10.1371/journal.pone.0129075 June17,2015 7/24 Catch&EconomicBenefitsofPeriodicOctopusFisheryClosures availabilityindexduringfocalperiods.(5)Finally,giventhesuitabilityscorerankings,werana “draft-pick”algorithm,allowingeachclosuretoselect(andexclude)itstop-rankedcontrol; thenrandomizedtheselectionorderofthe“draft-pick”setof36control-closurepairsover 10,000times,takingthebestglobalcontrol-closuresolution. Normalityandhomoscedasticityofresponsevariableswereassayedusingq-qplotsand Levene’stest.Uponfailureofeitherconditionwelog-transformedthevariableinquestion, whichmettheseassumptionsineachcase.Eachanalysiswasperformedusingthelme4pack- ageinR[65],usingPeriod(i.e.,BeforeorAfterclosure),Control/Impact,andco-occurrenceof theregionalfisheryshutdown(“ban”/”noban”)asfixedeffects,and,becausemultipleclosure eventscouldhappenwiththesameclosuresite(atdifferenttimes),weincludedClosureSiteas arandomeffect.Allreportedsignificanceprobabilitiesderivefromindependentcontrasts withinthismixed-modelframework[65–66]. 2.6.VillageIncome Forallvillagesthatimplementedaclosureevent,weusedamixedmodeldesignsimilartothat describedabovetotestfordifferencesinthreevariables:(a)totalvillageoctopusincome(alloc- (cid:1) topuslandedinavillage(kg) beachprice($/kg);$),(b)totalfishingeffort(fisher-days),and (c)incomeperuniteffort(2011$PPP/fisher-day)acrossthreeperiods:(i)30dayspre-closure (before);(ii)closureperiod(during;normalizedto30-daymeasure);(iii)30dayspost-reopen- ing(after).Weanalyzedincomeandeffortfromvillagesimplementing28closureevents, whichrepresentthesubsetofthe36closuresanalyzedaboveforwhichwehaddatacoverageto pairvillagesimplementingaclosurewithcontrolvillagesthathadnoclosureatthesametime. Ofthese28,14were“no-ban”closures,and14were“ban”closures(Fig2B). 2.7.Stochasticmodelingofsite-specificclosureneteconomicbenefits Toassessthesite-specificneteconomicbenefitsofeachclosuresite,wecomparedlandings fromclosuresitestostochasticallymodeledlandingsassumingcontinuedopen-accessfishing atthesamesite.Todoso,wemodeledboththeforegoneearningsforperiodsduringand60 daysafter36closuresandthencomparedthesemodeledearningstotheactualcatchdatafrom 36closureevents(Fig2C). Ourdataprovideuswithtwoobserveddistributionsrequiredforoursimulations:(1)V,the numberoffishersvisitingthefocalsiteonagivenday,and(2)C,thebeachvalueofoctopus caughtbyonefisherononeday.Eachofthesedistributionsaredrawndirectlyfromthefishery data,andstochasticallysampledtogenerateoursimulateddatacomparison(S7Fig).Tobuild thevisitationdistribution,V,wefirstrecordedthenumberoffishersthatwenttothefocalsite eachdayitwasvisitedduringtheentirestudyperiodexcludingclosureperiodsandsix-weeks afteraclosurereopened.Ondaysthatasitereceivednorecordedlandings,therearethreepos- sibilities:(1)therewasnofishingactivity,(2)therewasactivitybutnoonevisitedthesite,or (3)atleastonefishervisitedthesite,buttheycaughtnooctopus.Toaccuratelyestimatecase2 (actualzero-visitdays)wefirstexcludedcase1(nofishingdays)byonlycountingzerodayson which(a)therewasfishingrecordedinavillagethathadeverfishedthefocalsite,and(b)no landingswererecordedfromthesiteinquestion.Tofurthercorrecttheestimatedzero-visit daysforcase(3)wefittheparameterZ,whereasite’smodeled#ofzero-visitdays=(Z(cid:1)zero dayswithactivefishing).Weran100iterationsofourlandingssimulationmodelondatafrom 36open-accesscontrolsitesforeachpotentialestimateofZ(from0to2,by0.025)andthen calculatedthedifferencebetweenourmodeledfisherylandingsandtheactuallandingsin those36sites.Byfindingtheminimummediandivergencebetweenactualandmodeledvalues, PLOSONE|DOI:10.1371/journal.pone.0129075 June17,2015 8/24 Catch&EconomicBenefitsofPeriodicOctopusFisheryClosures wedeterminedthattheminimalmodelbiaswasgeneratedwithavalueofZ=0.525.Allsimu- lationsthereafterusedthatparametervalue. Thiszero-binmultipliertuningmadeourneteconomicbenefitcriterionmoresevere.That is,byroughlyhalvingtheprobabilityof“novisit”,thisprovidedahigherestimateofcounter- factualcatches(or“cost”),andthereforemadeour“profitability”criterionmoreconservative. Thevalueperuniteffortdistribution,C,derivesfromCPUEat150controlsitesduringthepe- riodinquestion(duringorafter),usingbeachpricesonthedayandvillageofsale. Tosimulatethe“noclosure”catchvalue,foreachdayinwhichfishingisreportedinarele- vantvillage(i.e.,onethathasfishedinthefocalsiteinthepast),wesampledfromthefocal site’sV,returninganumberoffishersthatvisitedthatsitethatday:V .ThenwesampledV d d timesfromCtogenerateadistributionofsinglefisher’sdailycatchvalues(C).Thedailysum f ofC overallfishersonagivendaygeneratedatime-seriesoftotaldailycatchvalues,L .The f d sumofdailyvalueeachday(L )overalldayssampled(N ),generatedatotallandingsL. d FD X L ¼ Vd C d f¼1 f X X X L¼ NFDL ¼ NFD Vd C d¼1 d d¼1 f¼1 f BytreatingL,thesimulatedcatchvalue(i.e.,whatpeoplewouldhaveearnedhadtheynot institutedaclosure)asthecost,andA,thetotalactualrecordedlandingsvaluefromtheclosure siteoverthesametimeperiodasthebenefits,weestimatedthenetearnings(NE)oftheclosure relativetoitscounter-factualcontrol(NE=A-L).Foreachofthe36modeledclosures,weran ourstochasticmodel1000timesresultinginadistributionofnetearningsvaluesforeach closure. 2.8.Internalrateofreturn Theinternalrateofreturnofaparticularclosureisthediscountrateatwhichthenetpresent valueofthenetearningsisequaltozero: X 1 NPV ¼0¼ T NE NE t¼0 tð1þ@Þt whereNPV isthenetpresentvalueofthenetearnings,tisthedayssinceclosure,NE isthe NE t dailynetearnings,[email protected]’spercentagere- turnoninvestment(ROI)bydividingthenetbenefitsbythecosts. 2.9.SeasonalityofsettlementandCPUE Weassessedseasonalpatternsofsettlementandcatchperuniteffortfrom2004–2011(S9Fig). First,toassesspatternsunaffectedbyclosureeffects,weremovedallclosuresitesfromthefish- erydataset.Thenusinganoctopus’massatcaptureandagrowthcurveforO.cyanea[67]we back-calculatedanestimateofthatoctopus’settlementdate.Fromthiscollectionofdates,we reporttherelativefrequencyofestimatedsettlementeventsoccurringonagivenJulianday (S9Fig).ThenweassesstheCPUEacrosstheentiredatasetoneachdayfrom2004–2011,and presentaLOESSfitofthesedata(with95%confidenceintervals).Wethencalculatethelagged cross-correlationbetweensettlementfrequencyandsubsequentseasonalCPUEshift. PLOSONE|DOI:10.1371/journal.pone.0129075 June17,2015 9/24 Catch&EconomicBenefitsofPeriodicOctopusFisheryClosures Results 3.1.Socioeconomics TheVezowithintheVelondriakeLMMAhaveameanincomeof$1.72perpersonperday, belowthe$2perdaypovertystandard,andtheyrelyheavilyonseafoodproteinfortheirfood security(all$figurespresentedarein2011internationaldollars,whichadjustsforpurchasing powerparity(PPP);S1File,S3andS4Tables).Gleaningcontributedatleasthalfofhousehold incomefor62%ofhouseholds[S1File],thoughindividualfishersearnmorefromseacucum- bersandfinfish[61]. 3.2.FisherylandingsandCPUE–BACIAnalysis The36closuresitesforwhichwehadadequatebaselinedatashowedsignificantincreasesafter re-openinginbothoctopuslandingsandCPUEofoctopus,regardlessoftheirtimingwiththe annualregionaloctopusfisheryshutdown(Figs3,S1andS2).Acrossthe36closures,median octopuslandingsincreasedfrom49.5(±22.8CI95)kginthe30daysbeforeclosureto404.8 (±119.9)kginthe30daysafterreopening,a717.8%increase(p<0.0001,S2Fig).Thissignifi- cantincreaseisrobusttothetimingoftheregionalshutdown,appearingbothinthe17“no ban”closures,thatoccurredindependentlyoftheregionalshutdown(+550%,p<0.0001),and the19“ban”closures,thatextendedtheshutdown(+821%,p<0.0001;Figs2andS1). Controlsiteshadmedianlandingsof44.5(±35.5)kginthe30daysbefore,and74.6(±46.6) kgafterre-opening(+67%;S2Fig).Thoughthisincreaseincontrolsitesisstatisticallysignifi- cant(p<0.01),itisten-foldsmallerthaninclosedsites(S2andS3Figs).Moreover,theeffectin controlsisdrivenbythe19“ban”closuresthatextendedtheshutdown(+97%,p<0.05)andab- sentinthecontrolsforthe17“noban”closures(+67.7%,p=0.36;S1Fig). Aclosure’sreopeningattractedmanyfishers.Againcomparing30-dayperiodsimmediately beforeeachclosureandafterre-opening,the36closedsiteshadamedian477.8%increasein effort(fisher-days;p<0.0001,S1andS2Figs).Thereisalsoaweaklysignificantefforteffectin thecontrols(median+74%,p=0.05;S2Fig),however,oncesplitbytimingrelativetothere- gionalshutdown,neithercontrolsforthe17“noban”closuresnorthoseforthe19“ban”clo- suresshowedsignificanteffortincreases(median+88%,p=0.37;+117%,p=0.28;Figs2,3 andS1). Catchperuniteffort(kg/fisher-day,CPUE)showedlargeandsignificantincreasesatclo- suresiteswhilecontrolsitesshowedonlyaminorboost,whichwasagainrestrictedtothose “ban”closuresco-occurringwiththeregionalshutdown(Figs2andS2).Inclosuresitesfishers caughtamedianof2.37(±0.33)kgofoctopusperfisher-daybeforeclosure,butaftertheclo- surere-opened,fisherscaught4.42(±0.51)kg/fisher-day,aCPUEincreaseof86.6% (p<<0.0001;S2Fig).ThesesignificantCPUEincreaseswerepresentbothin“ban”and“no ban”closures(Figs2and3).Incontrolsitesforthe17“noban”closures,medianCPUE showednosignificantchange(p=0.93;Fig3),whilecontrolsforthe19“ban”closuresshowed amoderateboost(+49%,p<0.01;Fig3). BoththelandingsandCPUEboostsweregreatestimmediatelyaftertheclosure’sreopening, anddiminishedwithindaystoweeksaftertheopening(S3andS4Figs).Landingstendedtore- turntobaselinelevelsafterthefirstorsecondtidalseries,generallywithinabout7–10days afterreopening(S3Fig).CPUEeffectswerealsostrongestinthefirstsetofspringlowtides aftertheopening,butcontinuedintothesecondorthirdsetofspringlowtides(i.e.,14–25 days;S4Fig).Asthisfisheryisdepthlimited,mostactivefishingoccursduringthelowesttides (i.e.spring)tides. PLOSONE|DOI:10.1371/journal.pone.0129075 June17,2015 10/24