Cause-specific temporal and spatial trends in green sea turtle strandings in the Hawaiian Archipelago ( )

Transcription

1 Mr Biol (2008) 154: DOI /s ORIGINAL PAPER Cuse-specific temporl nd sptil trends in green se turtle strndings in the Hwiin Archipelgo ( ) Milni Chloupk Æ Thierry M. Work Æ George H. Blzs Æ Shwn K. K. Murkw Æ Robert Morris Received: 16 October 2006 / Accepted: 11 April 2008 / Published online: 22 April 2008 Ó Springer-Verlg 2008 Abstrct We investigted cuse-specific temporl nd sptil trends in se turtle strndings in the Hwiin Archipelgo. Five species of se turtle were recorded in 3,861 strndings over 22-yer period ( ). Green turtles comprised 97% of these strndings with size nd gender composition reflecting the demogrphic structure of the resident green turtle popultion nd reltive green turtle bundnce in Hwiin wters. The cuse of strndings ws determined by necropsy bsed on complete gross externl nd internl exmintion. Totlly 75% of the 3,732 green turtle strndings were from Ohu where strndings occur yer-round. The most common known cuse of the green turtle strndings ws the tumour-forming disese, fibroppillomtosis (28%) followed by Communicted by P.W. Smmrco. M. Chloupk (&) Ecologicl Modelling Services Pty Ltd, University of Queenslnd, P.O. Box 6150, St Luci, QLD 4067, Austrli e-mil: m.chloupk@uq.edu.u T. M. Work USGS-Ntionl Wildlife Helth Center, Honolulu Field Sttion, P.O. Box 50167, Honolulu, Hwii 96850, USA G. H. Blzs S. K. K. Murkw Pcific Islnds Fisheries Science Center, NOAA Ntionl Mrine Fisheries Service, 2570 Dole Street, Honolulu, Hwii , USA R. Morris 348 Iliin Street, Kilu, Hwii 96734, USA hook-nd-line fishing ger-induced trum (7%), gillnet fishing ger-induced trum (5%), bot strike (2.5%), nd shrk ttck (2.7%). Miscellneous cuses comprised 5.4% of strndings wheres 49% of green turtle strndings could not be ttributed to ny known cuse. Green turtle strndings ttributble to bot strike were more likely from Kui nd Ohu while fibroppillom strndings were more likely from Ohu nd Mui. Hook-nd-line ger strndings were more likely from Ohu due to higher per cpit inshore fishing effort. The specific mortlity rte (conditionl probbility) for fibroppillomtosis ws 88%, 69% for gillnet ger nd 52% for hook-nd-line ger. The probbility of ded green turtle strnding incresed from 1982 but levelled off by the mid-1990s. The declining mortlity risk ws becuse the prevlence nd severity of fibroppillomtosis hs decresed recently nd so hs the mortlity risk ttributble to gillnet ger. Despite exposure to disese nd inshore fishing gers, the Hwiin green turtle stock continues to recover following protection since the lte 1970s. Nevertheless, mesures to reduce incidentl cpture of se turtles in costl Hwiin fisheries would be prudent, especilly since strndings ttributble to hook-nd-line fishing ger hve incresed stedily since Introduction The green se turtle (Cheloni myds) is the most bundnt lrge mrine herbivore in the world nd hs been subject to long history of humn exploittion (Frzier 1980; Prsons 1962) leding some popultions to be extirpted or depleted (Prsons 1962). Declining

2 888 Mr Biol (2008) 154: popultions re of concern becuse green turtle bundnce is considered to be n indictor of segrss nd corl reef ecosystem helth nd resilience (Jckson et l. 2001). However, green turtle conservtion is severely hindered by the lck of relible informtion on popultion sttus nd trends due to smpling difficulties of lrge mobile nimls in costl nd ocenic hbitts (Chloupk nd Limpus 2001; Chloupk nd Blzs 2007). However, importnt informtion on se turtle helth, ge, size composition, diet, reproductive sttus, popultion trends nd cuse-specific mortlity cn be derived from strndings of moribund or ded turtles (Limpus nd Reed 1985; Bjorndl et l. 1994; Cillouet et l. 1996; Boulon 2000; Schwrtz 2000; Bugoni et l. 2001; Zug et l. 2002; Work et l. 2004; Foley et l. 2005). Strndings re lso commonly used to infer geogrphic distribution nd bundnce (McLeod et l. 2004; Mldini et l. 2005) or the primry cuse of declining popultions exposed to nthropogenic hzrds (Pinedo nd Polcheck 1999; Lopez et l. 2002; Kreuder et l. 2003). Five species of se turtle re recorded from the Hwiin Archipelgo with the green turtle being the most bundnt nd resident yer-round (Blzs nd Chloupk 2004). The min nesting region for Hwiin green turtles is French Frigte Shols in the Northwestern Hwiin Islnds, where turtles nest minly from June to August (Blzs nd Chloupk 2004). The Hwiin green turtle genetic stock is sptilly disjunct metpopultion resident in mny corl reef nd costl forging grounds dispersed throughout the Archipelgo (Blzs nd Chloupk 2004b). The Hwiin green turtle ws seriously depleted by over-hrvesting nd nesting hbitt destruction but is now on the wy to recovery following protection under the US Endngered Species Act since the lte 1970s (Chloupk nd Blzs 2007). Green turtle strndings in the Hwiin Archipelgo hve incresed over the pst 20 yers or more, trend consistent with the incresing bundnce of the recovering Hwiin stock (Chloupk nd Blzs 2005). However, there re mny pprent cuses of mortlity ttributed to these green turtle strndings other thn nturl mortlity tht wrrnts further investigtion. So estimting cuse-specific mortlity trends could be useful for identifying potentil impediments to the ongoing recovery of the Hwiin green turtle stock. We present comprehensive nlysis of 22 yers of green turtle strndings in the Hwiin Archipelgo ( ) bsed on necropsy dt. Specificlly, the min purpose of our nlysis ws to (1) define the reltive risk of nthropogenic hzrds such s costl fisheries nd vessel trffic nd (2) identify ny temporl or sptil trends in these cuse-specific mortlity risks. Mterils nd methods Dt set Since 1982 the US Ntionl Mrine Fisheries Service (NMFS, Pcific Islnds Fisheries Science Center, Honolulu, Hwii) mintins comprehensive progrm of reporting nd slvging strnded se turtles in the Hwiin Archipelgo, especilly round the min Hwiin Islnds (Aguirre et l. 1998; Murkw et l. 2000; Work et l. 2004). Strnded turtles re reported to the NMFS vi well-dvertised telephone hotline with the specimens then recovered by NMFS personnel. Moribund turtles re cliniclly evluted by two veterinrins nd one biologist. Those judged to hve poor prognosis for survivl re euthnised (Work et l. 2004) with the reminder rehbilitted nd relesed. Loctions of strnding were recorded by islnd (Kui, Ohu, Mui nd Hwii) dt from the two smll islnds of Moloki nd Lni ner Mui were combined with Mui nd referred to here s Mui. The primry cuse of strnding bsed on the most significnt nd severe gross lesion ws determined vi gross necropsy (Work et l. 2004). All strnded turtles were then grouped into eight strnding ctegories bot strike, gillnet fishing ger, nets other thn gillnets, hook-nd-line fishing ger, fibroppillomtosis, shrk ttck, miscellneous nturl cuses nd unknown. The fishing ger relted hzrd ctegories were bsed on the fisheries known to interct with se turtles in Hwiin wters (Nitt nd Henderson 1993). Bot strike strndings were turtles with gross evidence of liner to prllel crpce frctures indictive of propeller, skeg or hull strikes (Kreuder et l. 2003). Gillnet fishing ger, nets other thn gillnets nd hook-nd-line fishing ger strndings were those where there ws documented interctions or entnglement with the specific fishing ger. Fibroppillomtosis were those nimls with gross evidence of externl tumors (Work et l. 2004). Shrk predtion strndings were those turtles with recent evidence of trumtic mputtion of body prts with or without irregulr liner brsions (Kreuder et l. 2003) or semicirculr bltion of the crpce or plstron reveling contents of the coelomic cvity. Miscellneous nturl cuses included ll non-nthropogenic cuses of mortlity other thn fibroppillomtosis or shrk ttck. Unknown were those cuses for which cuse of deth could not be determined. In the smll number of cses (\2%) where strnding hd multiple possible cuses only the primry cuse ws used in the nlysis for instnce if strnding ws ttributed to gillnet entnglement/shrk ttck then only gillnet entnglement ws used in the nlysis. Gender of strnded turtles ws determined by visul inspection of gonds t necropsy. Stright crpce length (cm SCL) or curved crpce length (cm CCL) ws mesured for ech

3 Mr Biol (2008) 154: turtle so ll size mesurements were converted to SCL using the following robust liner regression function derived by us from n independent smple of 1,323 Hwiin green turtles: SCL = *CCL, R 2 = Sttisticl methods Mesures of ssocition for count dt (or probbilities) in the form of contingency tbles were crried out using v 2 tests (Bishop et l. 1975). Mntel Henszel tests (Bishop et l. 1975) were used for multidimensionl tbles (k contingency tbles) while exct binomil tests were used for single proportions such s strndings sex rtio (Fleiss 1981). Cuse-specific mortlity rtes (conditionl probbilities) were derived using cuse-specific hzrd probbilities nd Byes theorem (Fleiss 1981). Long-term trends nd sesonlity in the monthly strndings were modelled using nonprmetric regression smoothing known s STL (Clevelnd et l 1990), which is time domin filtering procedure tht decomposes series into frequency components of vrition (1) trend, (2) sesonl nd (3) residul. Nonprmetric logistic regression (Hstie nd Tibshirni 1990) ws then used to model the probbility of hzrd-specific strnding or mortlity given informtive covrites or risk fctors such s yer, seson, stright crpce size (cm SCL) nd strnding loction. All the logistic regression models comprised logit link, binomil error nd cubic smoothing splines to model ny nonliner functionl form (Hstie nd Tibshirni 1990). All dt nd grphicl nlyses were crried out using the sttisticl modelling progrm R (Ihk nd Gentlemn 1996). Results Strnding summry Five species of se turtle occurred in the 3,861 strndings reported in the Hwiin Archipelgo over the 22-yer period from 1982 to ,732 green turtles (97%), 47 hwksbills (1.2%), 31 olive ridleys (0.8%), 5 letherbcks, 1 loggerhed nd 45 of unknown species identity. The reltive occurrence of these five species in the strndings closely reflects the reltive bundnce in Hwiin wters. Totlly 74% of ll the strndings were recorded from Ohu (Fig. 1), 12.6% from Mui (including Moloki nd Lni), 6.6% from the islnd of Hwii, 4.2% from Kui nd the reminder (2.6%) from other islnds throughout the Archipelgo. Totlly 99% of green turtle strndings recorded in the Archipelgo occurred round the four min islnds of Kui, Ohu, Mui (including Moloki nd strnded turtles yer Lni) nd Hwii. However, most green turtle strndings occurred round Ohu (75%) with 50% of these reported from the northest cost of Ohu (including Kneohe By). Very few green turtle strndings were reported from the southwest cost of Ohu (4%). Temporl trends in green turtle strndings Strndings of green turtles incresed since 1982 but levelled off by the lte 1990s (Fig. 1). There were significnt fluctutions in the monthly strndings of green turtles recorded for both Ohu nd Mui but little temporl synchrony between the two islnds (Fig. 2). Strndings on Ohu (men of ll yers) peked between Mrch nd June with secondry pek nd in September (Fig. 2b). It ppers tht monthly strndings decresed over the 22 yers in Jnury nd August but incresed for most of the other months. Hence, there ws min historicl sesonl pek but incresingly more green turtle strndings were recorded throughout the yer in more recent times perhps reflecting chnging pttern in reporting or chnging sesonlity of the underlying hzrds tht cuse the strndings. Hzrd-specific strndings nd mortlity The tumour-forming disese, fibroppillomtosis, ws the most frequent known cuse ccounting for c. 28% of the green turtle strndings while hook-nd-line fishing ger ccounted for 7% (Fig. 3). The cuse of 49% of the green turtle strndings ws undetermined becuse no lesions indictive of cuse of deth were seen or becuse crcsses were too decomposed precluding determintion of cuse of deth (Hrt et l. 2006). Totlly 75% of the green turtle strndings were ded nd strnding sttus (live or ded) ws significntly hzrd-specific (v ,21 = 964.7, P \ 0.005). ll greens greens (ohu) greens (mui) Fig. 1 Shows trends in reported nnul strndings since 1982 for ll five species (solid curve) nd for greens throughout the Archipelgo (dshed curve) or for greens round Ohu nd Mui

4 890 Mr Biol (2008) 154: bstrnded turtles sesonl component (STL scle) Jn 1985 Jn 1990 Jn 1995 Jn 2000 yer Jn Feb Mr Apr My Jun Jul Aug Sep Oct Nov Dec The joint probbility of being ded nd hving given known strnding cuse ws highest for fibroppillomtosis nd lowest for nets other thn gill net (Fig. 3b). Bot strike nd shrk ttck were the hzrds most likely to result in ded strnded green turtle (Fig. 3c). On the other hnd, if strnded green turtle ws ded nd the cuse of deth ws known, then it most likely due to fibroppillomtosis thn ny of the fisheries-relted hzrds (Fig. 3d). Size- nd gender-specific green turtle strndings ohu mui Fig. 2 Cheloni myds. Pnel shows trends in monthly strndings since Jnury 1982 for green turtles round Ohu nd Mui. Pnel b shows sesonl cycle monthly subseries plot of estimted number of monthly green turtle strndings round Ohu from period Horizontl line is fitted mid-men vlue of sesonl component for ech month. Fitted vlues for ech yer (ends of verticl lines) ssocited with ech midmen show pttern of internnul vrition of monthly subseries Strnded green turtles SCL rnged from c. 20 cm to c. 100 cm (Fig. 4), which spns the full rnge of size from recruitment from n ocenic hbitt to dult size in the neritic hbitt (Blzs nd Chloupk 2004b). Strnded green turtles were predomintely smll juveniles rnging from 40 to 60 cm SCL, which corresponds to n estimted ge rnge of c yer-old (Blzs nd Chloupk 2004b). Green turtle strndings from shrk ttck were significntly lrger thn strndings from other ctegories (Fig. 4), perhps becuse lrger green turtles re more likely to be tken by shrks or tht the remins of lrger turtle re more likely to be found wshed up on bech. On the other hnd, medin size of fisheriesrelted strndings (gillnet, hook-nd-line) ws significntly smller thn for ll other hzrds (Fig. 4). There ws no significnt trend over the 22-yer period ( ) in the medin size of strnded green turtles in the Archipelgo nor ws there ny sesonl trend in the size of strnded green turtles (dt not shown). The medin size of green turtles tht strnded (live or ded) in either Mui or Kui ws generlly lrger thn strnded greens round either Ohu or Hwii (Fig. 4b). There ws no gender bis in the Ohu green turtle strndings [exct binomil sex rtio test for P = 0.50 (or 50%), n = 1009, P = 0.71] but strnded mle green turtles were significntly smller thn strnded femle turtles. Strnding sttus (live, ded) for green turtle strndings round Ohu ws not function of gender or sector (Mntel 2 Henszel test: v 0.05,1 = 0.028, P = 0.87). Hzrd-specific green turtle strnding probbility Figure 5 shows the probbility of fibroppillomtosis (tumour-forming disese) relted strnding s function of four informtive covrites or cofctors strnding yer, seson, green turtle size (cm SCL) nd the min Hwiin islnds. The probbility of fibroppillom relted strnding incresed significntly from the erly 1980s, levelled off by the mid-1990s nd begn declining in recent yers (Fig. 5). Moreover, such strndings were significntly more likely to occur during winter (Fig. 5b). There ws lso size-dependent effect with the probbility of fibroppillom strnding incresing with turtle size (Fig. 5c). Fibroppillom strndings were more likely to occur round Ohu nd Mui thn round either Kui or Hwii (Fig. 5d). The probbility of gillnet ger relted strnding s function of the four covrites (strnding yer, seson, size, islnd) incresed significntly from the erly 1980s nd then declined significntly from the mid-1990s (Fig. 6). There ws no significnt sesonl effect in gillnet strndings (Fig. 6b). Gillnet strndings were more likely for smll turtles thn for lrge turtles (Fig. 6c) nd there ws no significnt sptil effect in gillnet strnding probbility (Fig. 6d). The probbility of hook-nd-line ger relted strnding s function of the covrites incresed significntly from the erly 1980s (Fig. 7) nd ppered more likely during the summer (Fig. 7b). Hook-nd-line strndings were more likely for smll (\40 cm SCL) or lrger turtles [80 cm SCL (Fig. 7c) nd there ws no significnt sptil effect (Fig. 7d).

5 Mr Biol (2008) 154: unkn fp hline misc gillnet bot shrk net b unkn fp misc hline gillnet bot shrk net strnding probbility (live or ded) probbility (ded nd hzrd) c bot shrk fp misc gillnet unkn net hline d unkn fp misc hline gillnet bot shrk net probbility (ded hzrd) Fig. 3 Cheloni myds. Hzrd-specific strndings of 3,732 green turtles in the Hwiin Archipelgo ( ). Pnel shows hzrd-specific strnding frequency for green turtles (live or ded). Pnel b shows joint probbility of green turtle being both ded nd ttributed to specific hzrd. Pnel c shows probbility of strnded green turtle being ded given specific hzrd. These hzrd-specific mortlity rtes derived from probbilities in (, b) s follows: P(ded/ hzrd) = P(ded nd hzrd)/p(strnding probbility). Pnel d probbility (hzrd ded) shows conditionl probbilities of green turtle strnding being ttributble to specific hzrd given tht turtle ws ded. Conditionl probbilities in (d) derived from probbilities shown in () conditioned on probbility of ded turtle strnding, which ws P(ded) = Conditionl probbilities in (c, d) relted by Byes theorem. fp fibroppillomtosis (tumour-forming disese), hline hooknd-line, misc vrious pthologicl cuses, bot bot strike, net other nets besides gillnets Hzrd-specific green turtle strnding mortlity Around 75% of ll green turtles strnded were ded. Figure 8 shows the probbility of ded turtle strnding s function of four informtive covrites or fctors strnding yer, turtle size, min islnd in the southestern chin of the Hwiin islnds where nerly ll strndings were recorded nd the min hzrd ssumed to be the cuse of the ded strnding. The probbility of ded green turtle strnding incresed significntly from the erly 1980s nd then levelled off by the lte-1990s (Fig. 8). Menwhile, the probbility of ded green turtle strnding decresed significntly with incresing turtle size (Fig. 8b). A ded strnded green turtle ws more likely to be recorded from Mui nd less likely to be recorded from Hwii (Fig. 8c). When the dt re conditioned on strnding yer, strnding size, nd strnding loction then the hzrds lest likely to be the cuse of ded green turtle strnding re hook-ndline nd gillnet fishing ger (Fig. 8d). Bot strike nd shrk ttck were the hzrds most likely to result in ded strnded green turtle round the min Hwiin islnds (Fig. 8d), which is finding tht is consistent with the hzrd-specific strnding mortlity rtes shown in Fig. 3c. Discussion Strndings cn be useful supplementry source of informtion on mortlity trends nd for helping to understnd the helth sttus of recovering mrine popultions (Kreuder et l. 2003). Most strnding mortlity trend studies del with mrine mmml (Pinedo nd Polcheck 1999; Lopez et l. 2002; Kreuder et l. 2003) or se birds (Cmphuysen 1998). There hve been few nlyses of se turtle strnding

6 892 Mr Biol (2008) 154: crpce size (cm SCL) b crpce size (cm SCL) bot fp gill hline misc net shrk unk hzrd Kui Ohu Mui Hwii islnd Fig. 4 Cheloni myds. Box plot summries of the hzrd- or islndspecific crpce size (cm SCL) distribution of the green turtles strnded (live or ded) in the Hwiin Archipelgo from 1982 to Pnel shows the distribution for the eight hzrd ctegories. Pnel b shows crpce size distribution of the strndings for the four min islnds. The white horizontl br in ech box shows 50th percentile (medin) of crpce size for ech hzrd. Upper nd lower boundries of ech box show 75th nd 25th percentiles, respectively. Top nd bottom cp of dotted verticl lines show 90th nd 10th percentiles, respectively. bot bot strike, fp fibroppillomtosis, gill gillnet ger, hline hook-nd-line ger, misc miscellneous pthologies, net other fishing net ger besides gillnets, shrk shrk ttck, unk unknown or inconclusive cuse of strnding trends (Cillouet et l. 1996; Boulon 2000; Schwrtz 2000; Foley et l. 2005) nd even fewer tht were bsed on systemtic postmortem exmintion (Work et l. 2004; present study). Strnding dt re often overlooked s source of demogrphic informtion on mrine wildlife becuse of problems with smple coverge nd difficulty in identifying the primry cuse of strnding (Cillouet et l. 1996; Cmphuysen 1998; Kreuder et l. 2003; McLeod et l. 2004; Work et l. 2004). The dt sets presented here re unique in tht they cover long time period ([20 yers) nd re bsed on systemtic exmintion of ech strnding cse. An importnt cvet is tht our strnding ctegories were primrily bsed on gross necropsy findings, nd it is possible given sufficient resources, tht more detiled exmintions of the crcsses might hve reveled underlying pthology not identified in this study. Regrdless, this would not hve helped in cses where crcsses were decomposed. The reltive bundnces of the five species of mrine turtle strndings recorded in the min Hwiin islnds re consistent with known forging nd nesting ctivities in the Archipelgo. Green turtles re the most bundnt se turtle in Hwiin wters with nesting occurring minly in the Northwestern Hwiin Islnds (Blzs nd Chloupk 2004). Hwksbills re more common in the inshore wters of the islnds of Hwii nd Mui (Blzs 1982), nd nesting occurs round some of the min Hwiin islnds (Ellis et l. 2000). There re few records of olive ridleys nesting in the Hwiin islnds (Blzs nd Hu 1986) nd this turtle is rre in the inshore Hwiin wters (Blzs 1982). Letherbck nd loggerhed strndings re extremely rre since neither species nest or forge round the Hwiin islnds nd so re probbly just migrtory trnsients. However, it is not known whether the green turtle strndings re relible indictor of the reltive mortlity risks ttributble to vrious nthropogenic hzrds. Whether most moribund or ded turtles wsh shore is unknown nd clerly the reported strndings cn only be some unknown frction of ll green turtle mortlities. Epperly et l. (1996) found tht winter bech strndings might not be relible indictor of fishery-induced mortlities in North Crolin becuse the previling winter currents were offshore nd so ded loggerhed turtles were less likely to wsh shore. Hrt et l. (2006) speculted tht perhps c. 20% of se turtle mortlities due to ger-induced injury might strnd long the North Crolin cost, depending on sesonl fctors. It is unlikely tht such findings pply yer-round or to broder rnge of costl geogrphic settings (Lugo-Fernndez et l. 2001), such s round the min Hwiin islnds. Strndings of green turtles round the min Hwiin islnds incresed significntly since 1982 but levelled off by the lte 1990s to c. 200 per nnum or c per month. The slowing of the strndings rte for Ohu greens reflects the recovering trend in the Hwiin green se turtle stock since the 1970s, where the stock is now pproching crrying cpcity (Chloupk nd Blzs 2007). However, the initil rpid increse in reported strndings rte from 1982 to 1988 my lso reflect incresing public wreness during this period. The Hwiin se turtle strndings progrm ws estblished in the lte 1970s but the sptil nd temporl coverge ws considered indequte until 1982 so tht reporting bis t lest for Ohu is considered minor source of error in the green turtle strndings record. Moreover, the Ohu green turtle strndings rte from 1988 onwrds ws 5.4% p (95% CI: %), which is consistent with the Hwiin nesting popultion increse estimted 5.7% p (95% CI:

7 Mr Biol (2008) 154: b prob(fp strnding, centred scle) yer prob(fp strnding, centred scle) month c prob(fp strnding, centred scle) crpce length (cm (SCL) d prob(fp strnding, centred scle) Kui Ohu islnd Mui Hwii Fig. 5 Cheloni myds. Grphicl summry of logistic regression model fit for probbility of fibroppillom fflicted green turtle strnding conditioned on four predictors (yer, month, crpce length, nd islnd where strnding occurred). The response vrible (probbility of fibroppillom fflicted strnding) is shown on y-xis in ech pnel s centred scle to ensure vlid pointwise 95% confidence bnds nd comprison between predictors cross pnels. The expected predictor-specific function is shown in ech pnel by the solid curve while the 95% pointwise confidence bnds re shown by the dshed curves. Pnel shows response vrible s function of strnding yer conditioned on the other three predictors. Pnel b shows response vrible s function of seson conditioned on predictors. Pnel c response vrible s function of size conditioned on predictors. Pnel d response vribles s function of where strnding ws recorded in the min Hwiin islnds conditioned on predictors. The verticl brs on the topside of the lower x-xis of (, c) re known s rug, which shows dt distribution within ech pnel. Width of ISLAND ctegories in (d) re proportionl to smple size with 95% confidence intervl for ech ctegory shown by top nd bottom cross brs with expected response shown by middle cross br 5.3% 6.1%). Another plusible explntion for ny increse in reported green turtle strndings round the min Hwiin islnds would be n incresing trend in exposure to vrious nthropogenic hzrds. Most strndings were recorded from Ohu presumbly reflecting the greter probbility of strnding being reported due to the lrger humn popultion on Ohu. Mui is the next most populous islnd nd ccounted for the next highest number of green turtle strndings record in ccordnce with previous findings (Work et l. 2004). The reltively lower strndings reported from the islnd of Hwii reltive to the mount of costline, popultion nd recretionl fishing effort might be due to the much nrrower costl shelf round this islnd thus decresing the probbility of ded or debilitted turtle wshing onshore. We suspect tht the strnding pttern round Ohu ws not function of reporting bis due to humn popultion density but probbly reflected the previling winds nd surfce currents (Work et l. 2004; Mldini et l. 2005). The size clss structure of strnded green turtles ws consistent with the expected reltive size-clss structure of the Hwiin green turtle metpopultion (Blzs nd Chloupk 2004b). The decresed probbility of strnding with incresing turtle size could reflect fewer lrge turtles t se, decresed probbility of lrge turtles wshing up on shore, or decresed probbility of mortlity from the fctors listed here s turtles ge. The smller size of mles suggests tht femle Hwiin green turtles might grow fster thn mles t comprble sizes (or ges), s shown for other green turtle popultions where gender is known

8 894 Mr Biol (2008) 154: prob (gillnet strnding, centred scle) yer prob (gillnet strnding, centred scle) b month prob (gillnet strnding, centred scle) c crpce size (cm SCL) prob (gillnet strnding, centred scle) d Kui Ohu islnd Mui Hwii Fig. 6 Cheloni myds. Grphicl summry of logistic regression model fit for the probbility of gillnet ger relted green turtle strnding conditioned on four predictors (strnding yer, strnding month, crpce length, islnd) see Fig. 5 for bbrevited codes using lproscopic exmintion of gonds (Chloupk et l. 2004). However, this remins to be confirmed for Hwiin green turtles (Blzs nd Chloupk 2004b). On the other hnd, there ws no gender-bis in the green turtle strndings since mles were s frequent in the strndings s femles. This is consistent with Wibbels et l. (1993) who found 1:1 primry sex rtio for the Hwiin green turtle stock. We conclude then tht se turtle strndings round the min Hwiin islnds reflects the reltive species bundnce in costl Hwiin wters s well s the generl sizend gender-composition of the resident green turtle stock. Similrly, Mldini et l. (2005) found tht the pttern of odontocete cetcen strndings round the min Hwiin islnds reflects well the species composition nd reltive bundnce trends derived from live-niml surveys. So it is pprent tht mrine niml strndings cn be useful supplementry source of informtion bout the geogrphic distribution nd reltive bundnce of green turtles nd some cetcens in costl Hwiin wters. Importntly, the green turtle strndings in our study provide insight into the vrious sources of mortlity for green turtles in costl Hwiin wters nd the reltive risks of exposure to nthropogenic hzrds such s fishing gers nd bot strike. Fibroppillomtosis ws the most common known cuse of green turtle strndings nd strndings were more likely round Ohu nd Mui thn the other min Hwiin islnds, which is consistent with the restricted geogrphic prevlence of this disese in the Hwiin Archipelgo (Blzs nd Chloupk 2004b; Work et l. 2004). Importntly, this disese does not pper to be ffecting the recovery of the Hwiin green turtle stock (Blzs nd Chloupk 2004; Chloupk nd Blzs 2005). It hs lso been suggested tht protubernt fibroppilloms increse the risk of entnglement nd fishery-induced mortlity (Foley et l. 2005). We found no such ssocition between strnding sttus (live or ded) nd fibroppillom-fflicted strndings with or without fishery-induced injuries (gillnet, hook-nd-line, nets), irrespective of the min Hwiin islnd where the strnding ws recorded (Mntel Henszel 2 k test: v 0.05,1 = 2.78, P [ 0.09; or collpsed over 2 the 4-islnd fctor for single tble v 0.05,1 = 2.39,

9 Mr Biol (2008) 154: prob(hook/line strnding, centred scle) yer prob(hook/line strnding, centred scle) b month prob(hook/line strnding, centred scle) c crpce length (cm SCL) prob(hook/line strnding, centred scle) d Kui Ohu islnd Mui Hwii Fig. 7 Cheloni myds. Grphicl summry of logistic regression model fit for the probbility of hook-nd-line ger relted green turtle strnding conditioned on four predictors (strnding yer, strnding month, crpce length, islnd see Fig. 5 for bbrevited codes P [ 0.12). In other words, we found no evidence tht fibroppillomtosis increses the risk of fishery-induced mortlity for Hwiin green turtles. Interction with hook-nd-line fishing ger ws the second most common cuse of green turtle strnding round the min Hwiin islnds. Hook-nd-line ger strndings were lso more likely round Ohu thn the other min Hwiin islnds probbly reflecting higher per cpit inshore fishing effort. Shrimp trwling (Cillouet et l. 1996) nd gillnets (FAO 2004) re considered to be the mjor fishing ger hzrds for se turtles in costl wters round the world. However, there re no costl trwl fisheries in Hwiin wters nd we found tht gillnet interctions were less likely thn hook-nd-line interctions in costl Hwiin wters. Moreover, most fishing effort in inshore Hwiin wters is relted to hook-ndline ctivities rther thn net-relted ctivities (Nitt nd Henderson 1993). Nonetheless, gillnet fishing ger poses significnt risk to green turtles in Hwiin wters becuse of the much higher specific mortlity rte thn hook-ndline ger. Three of the hzrds were found to hve significnt temporl trend in the probbility of strnding fibroppillomtosis, gillnet fishing ger nd hook-nd-line fishing ger. Despite incresing initilly, the probbility of fibroppillomtosis strndings hs declined since the mid- 1990s. Moreover, Chloupk nd Blzs (2005) hve shown tht the severity of this disese hs declined since the lte 1990s for green turtle popultion resident round Moloki ner Mui. The probbility of gillnet ger strndings hs lso declined significntly since the mid-1990s. Menwhile, the probbility of hook-nd-line ger strndings hs incresed significntly since the erly 1990s. Even though this hzrd hd the lowest specific mortlity rte, if the risk of hook-nd-line strndings continues to increse then mngement intervention might be wrrnted. The probbility of strnding ttributble to gillnet nd hook-nd-line ger decresed significntly with incresing turtle size. Such fishing ger could be sizeselective or juvenile turtles could be more likely to become debilitted from fishing ger interction (resulting

10 896 Mr Biol (2008) 154: b prob (ded, centred scle) yer prob (ded, centred scle) crpce size (cm SCL) c Kui Ohu Mui Hwii d bot fp gill hline net misc shrk unk prob (ded, centred scle) islnd prob (ded, centred scle) hzrd Fig. 8 Cheloni myds. Grphicl summry of logistic regression model fit for the probbility of ded green turtle strnding (response vrible) conditioned on four predictors (strnding yer, crpce length, islnd where strnding occurred, nd specific hzrd). Pnel shows response vrible s function of strnding yer conditioned in strnding). However, more likely explntion is tht ger-induced strndings reflect the higher reltive bundnce of juveniles expected for this recovering stock. Recll tht most cuses of strnding were unknown nd tht the criteri for ttributing fishing ger s cuse of strnding demnded physicl evidence of ger interction. It is likely tht some turtles die from fishing ger interction through drowning leving no externl mrks. However, drowning is n extremely difficult cuse of deth to determine for strnded mrine nimls (Kreuder et l. 2003). So strnded se turtles tht drown due to incidentl cpture in fishing ger my hve no conclusive cuse of deth or could be mistkenly ttributed to some other cuse. Therefore, our study my underestimte the impcts of fishing ger on the Hwiin green turtle stock. However, ssuming tht there ws no hzrd-specific strnding bis (for instnce bot strike strnding ws no more likely to wsh shore nd be found thn gillnet strnding), then green turtle strndings ttributble to fishing ger interctions were not common in the on the other three predictors. Pnel b shows response vrible s function of turtle size conditioned on predictors. Pnel c shows response vrible s function of where strnding ws recorded in the min Hwiin islnds conditioned on predictors. Pnel d shows response vrible s function of hzrd conditioned on predictors Archipelgo. Moreover, while the specific mortlity rte for fibroppillomtosis ws 88% it ws significntly lower t 69% for gillnet fishing ger nd 52% for hook-nd-line fishing ger. These conditionl probbilities or specific mortlity rtes provide bsis for evluting the reltive risks of exposure to specific nthropogenic hzrds nd to help ssess the cost-benefits of imposing hzrd-specific mitigtion strtegies. So while hook-nd-line ger interctions ws one of the more common cuses of green turtle strndings, it ws the lest likely hzrd to cuse ded strnded green turtle in Hwiin wters. Furthermore, bot strike rther thn fisheries interctions ws the most likely humn-relted cuse of ded green turtle strnding round the min Hwiin islnds. It is possible tht sick turtle or turtle tht died from some other cuse ws then subsequently struck by bot nd sorting this out would require more detiled postmortem exms on freshly ded turtles (Work et l. 2004). Bot strike ws lso reported s mjor cuse of se turtle strndings round the US Virgin Islnds lthough smple sizes in tht study were smll (Boulon 2000).

11 Mr Biol (2008) 154: Sesonl trends were lso evident for some hzrd-specific strndings, which might be due to environmentl fctors (Hrt et l. 2006). For instnce, temporl fluctutions in se surfce temperture nd food supply were proposed to ccount for trends in loggerhed se turtle strndings long the US northest Atlntic cost (Schwrtz 2000). On Ohu there ws pek historicl seson of green turtle strndings from Mrch to June nd secondry pek in September perhps reflecting sesonl bundnce of turtles or sesonlity of onshore winds or currents. Fibroppillom strndings were more likely during winter. This could be due to winter storms wshing debilitted turtles onshore or to temperturedependent immunosuppression in ffecting tumoured turtles more severely (see Zpt et l. 1992). Foley et l. (2005) lso found tht strnded green turtles with fibroppillomtosis were more common during winter in Florid. There ws lso summer pek in fishing ger relted strndings (gillnets, hook-nd-line, nets) tht could be due to incresed fishing ctivity during tht time of yer. The identifiction of the vrious sources of green turtle mortlity in Hwiin wters is essentil for understnding the helth nd long-term popultion recovery of the Hwiin green turtle stock (Chloupk nd Blzs 2005; Chloupk nd Blzs 2007). Temporl nd sptil trends in green turtle strndings cn provide importnt informtion on disese threts nd trends in mortlity risk ttributble to nthropogenic hzrds such s costl fisheries. For instnce, the declining risk of exposure to fibroppillom disese nd gillnet fishing ger bodes well for the long-term helth nd recovery of the once-depleted Hwiin green turtle stock (Chloupk nd Blzs 2007). However, more comprehensive informtion on geclss- nd sex-specific green turtle bundnce in Hwiin wters, nd direct estimtes of costl fishing effort, re now needed to support improved risk ssessment of such hzrds in costl Hwiin wters. Acknowledgments For extensive long-term support of nd contributions to the Hwiin se turtle strnding reserch progrm we thnk the following individuls, gencies nd orgniztions: Bridget McBride, Donn Brown, Shndell Brunson, Sllie Bevers, John Coney, Skippy Hu, Leon Hllcher, Don Hecock, Cody Hooven, Joy Oliveir, Erin Siebert, Sherwood Mynrd, Bill Pulelo, Mrc Rice, Denise Prker, Stte of Hwii Division of Aqutic Resources nd Division of Conservtion nd Resources Enforcement, Mrine Option Progrm t the University of Hwii t Mno nd t Hilo nd Mui Community College, the Hwii Preprtory Acdemy, nd the NOAA Hwiin Islnds Humpbck Whle Ntionl Mrine Snctury. This work ws supported by NOAA Fisheries Contrct to the first uthor. References Aguirre A, Sprker T, Blzs G, Zimmermn B (1998) Spirorchidisis nd fibroppillomtosis in green turtles of the Hwiin Islnds. J Wildl Dis 34:91 98 Blzs G (1982) Sttus of se turtles in the centrl Pcific Ocen. In: Bjorndl KA (ed) Biology nd conservtion of se turtles. Smithsonin Institution Press, Wshington, pp Blzs G, Hu S (1986) Geogrphic distribution: Lepidochelys olivce in Hwii. Herpetol Rev 17:51 Blzs G, Chloupk M (2004) Thirty-yer recovery trend in the once depleted Hwiin green se turtle stock. Biol Conserv 117: Blzs G, Chloupk M (2004b) Sptil nd temporl vribility in somtic growth of green se turtles resident within the Hwiin Archipelgo. Mr Biol 145: Bishop Y, Fienberg S, Hollnd P (1975) Discrete multivrite nlysis: theory nd prctice. MIT Press, Cmbridge Bjorndl K, Bolten A, Lgueux C (1994) Ingestion of mrine debris by juvenile se turtles in costl Florid hbitts. Mr Pollut Bull 28: Boulon R (2000) Trends in se turtle strndings, US Virgin Islnds: 1982 to In: Abreu A, Briseno R, Mrquez R, Srti L (eds) Proceedings of the eighteenth interntionl se turtle symposium. NOAA Technicl Memorndum NMFS-SEFSC-436, pp Bugoni L, Kruse L, Petry M (2001) Mrine debris nd humn impcts on se turtles in southern Brzil. Mr Pollut Bull 42: Cillouet C, Shver D, Tes W, Nnce J, Rever D, Cnnon A (1996) Reltionship between se turtle strnding rtes nd shrimp fishing intensities in the northwestern Gulf of Mexico: versus Fish Bull 94: Cmphuysen C (1998) Beched bird surveys indicte decline in chronic oil pollution in the North Se. Mr Pollut Bull 36: Chloupk M, Limpus C (2001) Trends in the bundnce of se turtles resident in southern Gret Brrier Reef wters. Biol Conserv 102: Chloupk M, Blzs G (2005) Modelling the effect of fibroppillom disese on the somtic growth dynmics of Hwiin green se turtles. Mr Biol 147: Chloupk M, Blzs G (2007) Using Byesin stte-spce modelling to ssess the recovery nd hrvest potentil of the Hwiin green se turtle stock. Ecol Modell 205: Chloupk M, Limpus C, Miller J (2004) Green turtle somtic growth dynmics in sptilly disjunct Gret Brrier Reef metpopultion. Corl Reefs 23: Clevelnd R, Clevelnd W, McRe J, Terpenning P (1990) STL: sesonl-trend decomposition procedure bsed on Loess. J Officil Stt 6:3 73 Ellis D, Blzs G, Gilmrtin W, Murkw S, Kthir L (2000) Short-rnge reproductive migrtions of hwksbill turtles in the Hwiin islnds s determined by stellite telemetry. In: Abreu A, Briseno R, Mrquez R, Srti L (eds) Proceedings of the eighteenth interntionl se turtle symposium. NOAA Technicl Memorndum NMFS-SEFSC-436, pp Epperly S, Brun J, Chester A, Cross F, Merriner J, Tester P, Churchill J (1996) Bech strndings s n indictor of t-se mortlity of se turtles. Bull Mr Sci 59: FAO (2004) Report of the Expert Consulttion on interctions between se turtles nd fisheries within n ecosystem context. FAO Fisheries Report Number 738. Food nd Agriculture Orgniztion of the United Ntions, Rome, Itly, 9 12 Mrch 2004, p. 37 Fleiss J (1981) Sttisticl methods for rtes nd proportions, 2nd edn. Wiley, New York Foley A, Schroeder A, Redlow A, Fick-Child K, Tes W (2005) Fibroppillomtosis in strnded green turtles (Cheloni myds) from the estern United Sttes ( ): trends nd ssocitions with environmentl fctors. J Wildl Dis 41:29 41

12 898 Mr Biol (2008) 154: Frzier J (1980) Exploittion of mrine turtles in the Indin Ocen. Humn Ecol 8: Hrt K, Mooreside P, Crowder L (2006) Interpreting the sptiotemporl ptterns of se turtle strndings: going with the flow. Biol Conserv 129: Hstie T, Tibshirni J (1990) Generlized dditive models. Monogrphs on sttistics nd pplied probbility 43. Chpmn nd Hll, London Ihk R, Gentlemn R (1996) R: lnguge for dt nlysis nd grphics. J Comput Grp Stt 5: Jckson J, Kirby M, Berger W, Bjorndl K, Botsford L, Bourque R, 13 others (2001) Historicl overfishing nd the recent collpse of costl ecosystems. Science 293: Kreuder C, Miller M, Jessup D, Lowenstine L, Hrris M, Ames J, Crpenter T, Conrd P, Mzet J (2003) Ptterns of mortlity in southern se otters (Enhydr lutris nereis) from J Wildl Dis 39: Limpus C, Reed P (1985) Green se turtles Cheloni myds strnded by Cyclone Kthy on the southwestern cost of the Gulf of Crpentri, Austrli. Aust Wildl Res 12: Lopez A, Sntos M, Pierce G, Gonzlez A, Vleirs X, Guerrs A (2002) Trends in strndings nd by-ctch of mrine mmmls in north-west Spin during the 1990s. J Mr Biol Assoc UK 82: Lugo-Fernndez A, Morin M, Ebesmeyer C, Mrshll C (2001) Gulf of Mexico historic ( ) surfce drifter dt nlysis. J Cost Res 17:1 16 McLeod C, Pierce G, Sntos M (2004) Geogrphic nd temporl vritions in strndings of beked whles (Ziphiide) on the costs of the UK nd the Republic of Irelnd from J Cetcen Res Mnge 6:79 86 Mldini D, Mzzuc L, Atkinson S (2005) Odontocete strndings ptterns in the min Hwiin Islnds ( ): how do they compre with live niml surveys? Pc Sci 59:55 67 Murkw S, Blzs G, Ellis D, Hu S, Emes S (2000) Trends in fibroppillomtosis mong green turtles strnded in the Hwiin Islnds, In: Klb H, Wibbels T (eds) Proceedings of the nineteenth nnul symposium on se turtle biology nd conservtion. NOAA Technicl Memorndum NMFS-SEFSC- 443, pp Nitt E, Henderson J (1993) A review of interctions between Hwii s fisheries nd protected species. Mr Fish Rev 55:83 92 Prsons J (1962) The green turtle nd mn. University of Florid Press, Ginsville Pinedo M, Polcheck T (1999) Trends in frnciscn (Pontopori blinvillei) strnding rtes in Rio Grnde do Sul, Southern Brzil ( ). J Cetcen Res Mnge 1: Schwrtz F (2000) Body-orgn weight reltionships of strnded loggerhed se turtles, Crett crett, with n exmintion of the effects of the environment cusing their demise. J Elish Mitchell Sci Soc 116:86 90 Wibbels T, Blzs G, Owens D, Amoss M (1993) Sex rtios of immture green turtles inhbiting the Hwiin Archipelgo. J Herpetol 27: Work T, Blzs G, Rmeyer R, Morris R (2004) Retrospective pthology survey of green turtles Cheloni myds with fibroppillomtosis in the Hwiin Islnds, Dis Aqut Orgn 62: Zpt A, Vrs A, Torrob M (1992) Sesonl vritions in the immune system of lower vertebrtes. Immunol Tody 13: Zug G, Wetherll J, Blzs G, Prker D, Murkw S (2002) Age nd growth in Hwiin green seturtles (Cheloni myds): n nlysis bsed on skeletochronology. Fish Bull 100: