MARK-RECAPTURE POPULATION ESTIMATES OF COHO, PINK, AND CHUM SALMON RUNS TO UPPER COOK INLET IN 2002

MARK-RECAPTURE POPULATION ESTIMATES OF COHO, PINK, AND CHUM SALMON RUNS TO UPPER COOK INLET IN 2002 by T. Mark Wllette Robert DeCno Nany Gove Regonal Informaton Report No. 2A03-20 Alaska Department of Fsh and Game Commeral Fsheres Dvson 333 Raspberry Rd. Anhorage, Alaska 9958-599 June 2003 The Regonal Informaton Report Seres was establshed n 987 to provde an nformaton aess system for all unpublshed dvsonal reports. These reports frequently serve dverse ad ho nformatonal purposes or arhve bas unnterpreted data. To aommodate tmely reportng of reently olleted nformaton, reports n ths seres undergo only lmted nternal revew and may ontan prelmnary data; ths nformaton may be subsequently fnalzed and publshed n the formal lterature. Consequently, these reports should not be ted wthout pror approval of the author or of the Commeral Fsheres Dvson.

AUTHORS T. Mark Wllette s a Researh Projet Leader for the Alaska Department of Fsh and Game, Commeral Fsheres Dvson, Regon II, Upper Cook Inlet, 4396 Kalfornsky Beah Road, Sute B, Soldotna, AK 99669. Robert DeCno s a Researh Bologst for the Alaska Department of Fsh and Game, Commeral Fsheres Dvson, Regon II, Upper Cook Inlet, 4396 Kalfornsky Beah Road, Sute B, Soldotna, AK 99669. Nany Gove s a Bometran for the Alaska Department of Fsh and Game, Commeral Fsheres Dvson, Regon II, 333 Raspberry Rd., Anhorage, AK 9958. ACKNOWLEGEMENTS The authors would lke to thank Mr. Kenny Rogers and rew aboard the F/ V Hstoran, staff from Wally H. Noerenberg hathery, Cook Inlet Aquaulture Assoaton and Ile Seafoods for help n performng short term mortalty experments. Thanks to the aptans and rews of the F/V Agave, F/V Infamous Glory, F/V Mllenum and F/V Just n Case for workng endless hours n all weather ondtons to apture and tag fsh. Thanks to all seafood proessors nvolved n ths study for allowng unfettered aess nto ther faltes for our rews to perform numerous tests. We would also lke to thank Jose DeCreeft (Northwnd Avaton) for hs very ompetent support n ondutng aeral surveys to loate rado telemetry tags. And, we thank all ADF&G employees nvolved n taggng, olletng data, and tunng systems to perform to the best of ther abltes. Fnally, we appreate omments on the manusrpt and other tehnal assstane provded by Terrane J. Qunn II, Danel J. Reed, Bran Bue, and Patrk Shelds.

TABLE OF CONTENTS LIST OF TABLES...v LIST OF FIGURES...v LIST OF APPENDICES... v ABSTRACT...x INTRODUCTION... OBJECTIVES...3 METHODS...3 Estmaton of short-term tag mortalty...3 Applaton of PIT and rado tags...3 Estmaton of rate of PIT tag loss...4 Reovery of PIT tags at proessors and estmaton of PIT tag deteton rates...5 Estmaton of salmon populaton szes and evaluaton of soures of error...6 Rado telemetry study on oho salmon...0 RESULTS...3 Estmaton of short-term tag mortalty...3 Applaton of PIT and rado tags...3 Estmaton of rate of PIT tag loss...3 Reovery of PIT tags at proessors and estmaton of PIT tag deteton rates...3 Estmaton of salmon populaton szes and evaluaton of soures of error...4 Rado telemetry study on oho salmon...5 DISCUSSION...7 LITERATURE CITED...22 FIGURES...26 TABLES...33 APPENDICES...5 Page

LIST OF TABLES Table Page. Statsts olleted from a stratfed mark-reapture experment... 33 2. Populaton parameters from a stratfed mark-reapture experment... 33 3. Expeted value of statsts n Table... 33 4. Short-term tag mortalty of oho and hum salmon estmated from net pen studes... 34 5. Total number of net sets and geometr mean ath per net set for sokeye, oho, pnk, and hum salmon durng fve weekly tag release strata, 2002. Numbers n parentheses ndate the standard error of the mean... 34 6. Total ath and number of oho, pnk, and hum salmon tagged wth PIT and rado tags durng fve weekly tag release strata, 2002... 34 7. Mean dfferene between PIT tag deteton rates estmated usng dummy versus atual salmon heads by proessor and lne, 2002. Numbers n parentheses ndate the standard error of the mean... 35 8. Mean PIT tag deteton rate by proessor and proessng lne durng sx weekly tag reovery strata, 2002. Numbers n parentheses ndate the standard error of the mean... 35 9. Summary statsts for oho salmon PIT tagged n Upper Cook Inlet and reovered n proessors, 2002. The number of tagged fsh released ( n ) has been adjusted for shortterm tag mortalty and tag loss. Number of tagged fsh reovered (m j ) has been adjusted for tag deteton rate... 36 0. Detaled results from a maxmum lkelhood Darroh estmate of the populaton sze of oho salmon returnng to Upper Cook Inlet, 2002 (fnal poolng) and test results for ompletng poolng. In ths analyss, the number of tagged fsh released ( n ) has been adjusted for short-term tag mortalty and tag loss. Number of tagged fsh reovered (m j ) has been adjusted for tag deteton rate.... 37. Detaled results from a maxmum lkelhood Darroh estmate of the populaton sze of oho salmon returnng to Upper Cook Inlet, 2002 (fnal poolng) and test results for ompletng poolng. In ths analyss, the number of tagged fsh released ( n ) has been adjusted for long-term tag mortalty and tag loss. Number of tagged fsh reovered (m j ) has been adjusted for tag deteton rate.... 38 2. Summary statsts for pnk salmon PIT tagged n Upper Cook Inlet and reovered n proessors, 2002. The number of tagged fsh released ( n ) has been adjusted for shortterm tag mortalty and tag loss. Number of tagged fsh reovered (m j ) has been adjusted for tag deteton rate... 39 v

LIST OF TABLES (ontnued) Table Page 3. Detaled results from a maxmum lkelhood Darroh estmate of the populaton sze of pnk salmon returnng to Upper Cook Inlet, 2002 (fnal poolng) and test results for ompletng poolng. In ths analyss, the number of tagged fsh released ( n ) has been adjusted for short-term tag mortalty and tag loss. Number of tagged fsh reovered (m j ) has been adjusted for tag deteton rate.... 40 4. Summary statsts for hum salmon PIT tagged n Upper Cook Inlet and reovered n proessors, 2002. The number of tagged fsh released ( n ) has been adjusted for shortterm tag mortalty and tag loss. Number of tagged fsh reovered (m j ) has been adjusted for tag deteton rate... 4 5. Detaled results from a maxmum lkelhood Darroh estmate of the populaton sze of hum salmon returnng to Upper Cook Inlet, 2002 (fnal poolng) and test results for ompletng poolng. In ths analyss, the number of tagged fsh released ( n ) has been adjusted for short-term tag mortalty and tag loss. Number of tagged fsh reovered (m j ) has been adjusted for tag deteton rate.... 42 6. Results from a h-square test of the null hypothess that the probablty of reapturng PIT tagged oho, pnk, and hum salmon dd not dffer for fsh that were aptured durng ebb, flood, or slak tdes.... 43 7. Results from a h-square test of the null hypothess that the probablty of reapturng PIT tagged oho, pnk, and hum salmon dd not dffer among three groups that were held on the taggng vessels for <30 mns., 30-60 mns, and >60 mns.... 44 8. Results from a h-square test of the null hypothess that the probablty of reapturng PIT tagged oho, pnk, and hum salmon dd not dffer among sx length lasses (<50 m, 50-55 m, 55-60 m, 60-65 m, 65-70 m, >70 m). The length dstrbuton for eah spees tagged s also ndated for omparson.... 45 9. Ratos of the number of tagged and untagged oho, pnk and hum salmon reovered at seven plants proessng salmon returnng to Upper Cook Inlet, 2002. Tag ratos adjusted for tag deteton rates at eah proessor are nluded for omparson.... 45 20. Results from a h-square test of the null hypothess that there was no dfferene n the rato of the numbers of rado tagged to untagged oho salmon returnng to fve streams flowng nto northern Cook Inlet for whh salmon esapement estmates were avalable n 2002. The date at whh 50% the total esapement passed the wer s nluded for omparson.... 46 v

LIST OF TABLES (ontnued) Table Page 2. Estmated total esapement (wth 95% onfdene ntervals) of oho salmon nto 33 streams and fve areas around Upper Cook Inlet, 2002. The number of tags weghted by the ath per unt effort of oho salmon n eah release stratum s also ndated, as well as, the weghted perent of total tags (reovered n freshwater) found n eah stream or area....47 22. Perent of total rado tags reovered (n freshwater) for seven stoks of oho salmon n Upper Cook Inlet, 2002 by release strata.... 48 23. Perent of total rado tags reovered (n freshwater) for seven stoks of oho salmon n Upper Cook Inlet, 2002 by reovery strata.... 49 24. Perent of total rado tagged oho salmon reovered n streams flowng nto Upper Cook Inlet, 2002 by release and reovery strata (all stoks ombned)... 49 25. Geometr mean resdene tme and mgraton rate for seven stoks of rado tagged oho salmon n Upper Cook Inlet... 49 26. Geometr mean resdene tme and mgraton rate for rado tagged oho salmon n Upper Cook Inlet by release strata.... 50 27. Estmated populaton szes (mllons), esapements, and explotaton rates on oho, pnk and hum salmon returnng to Upper Cook Inlet n 2002 derved from mark-reapture studes.... 50 v

LIST OF FIGURES Fgure Page. Loatons of rado tagged oho salmon found durng surveys of the lower portons of Upper Cook Inlet streams n 200. These fsh were tagged along the offshore test fshery transet west of Anhor Pont before July 20, 200. Numbers n boxes ndate perent of total reoveres (n freshwater) ourrng n eah stream. Numbers along test fshery transet ndate statons... 26 2. Loatons of rado tagged oho salmon found durng surveys of the lower portons of Upper Cook Inlet streams n 2002. These fsh were tagged along the offshore test fshery transet west of Anhor Pont before July 20, 2002. Numbers n boxes ndate perent of total reoveres (n freshwater) ourrng n eah stream... 27 3. Loatons of rado tagged oho salmon found durng surveys of the lower portons of Upper Cook Inlet streams n 2002. These fsh were tagged along the offshore test fshery transet west of Anhor Pont after July 20, 2002. Numbers n boxes ndate perent of total reoveres (n freshwater) ourrng n eah stream... 28 4. Loatons of rado tagged oho salmon (sold rles) found durng surveys of the Upper Cook Inlet dranage basn n Otober, 2002. Streams n the Anhorage area and those south of Bg Rver on the west sde of the nlet were not flown durng these surveys. The fsh were tagged along the offshore test fshery transet west of Anhor Pont n 2002... 29 5. Hstoral effort (number of delveres x hours fshed) n the drft gllnet fshery (dstrt wde openngs only), 972-2002...30 6. Relatonshp between ommeral fsheres explotaton rate on oho salmon (estmated from oded wre tags) and effort (number of delveres x hours fshed) n the drft gllnet fshery (dstrt wde openngs only), 993-998.... 3 7. Relatonshp between explotaton rate and the unertanty n estmated salmon populaton szes assumng a harvest of 0.25 mllon (example for oho and hum salmon)... 32 v

LIST OF APPENDICES Appendx Page. Summary statsts for oho, pnk, and hum salmon PIT tagged n Upper Cook Inlet and reovered n proessors, 2002 wthout any adjustments for tag mortalty, tag loss, or tag deteton rate... 5 2. Perent of total rado tags reovered by release and reovery strata for seven stoks of oho salmon...53 3. Summary of hstoral oded-wre tag estmates of oho salmon explotaton rates n Upper Cook Inlet, 993-998.... 56 v

ABSTRACT Ths projet estmated the total populaton szes, esapements, and explotaton rates for oho, pnk, and hum salmon returnng to Upper Cook Inlet (UCI) n 2002 as a frst step toward determnng esapement levels needed to aheve sustaned yelds for these spees. Mark-reapture tehnques were used to estmate the total populaton szes for eah spees returnng to UCI as a whole. Salmon were tagged along a transet runnng from Anhor Pont to the Red Rver delta on the west sde of Cook Inlet durng July and early August. Total populaton szes for eah spees were estmated from reoveres of passve ntegrated transponder (PIT) tags n ommeral fshery harvests. Reoveres of rado telemetry tags were used to estmate the total esapement of oho salmon nto all UCI streams for omparson to the estmate derved from PIT tags. Rado telemetry tag data were also used to estmate oho salmon esapements nto 33 streams and 5 areas around UCI. Our best PIT tag estmate of the total populaton sze of oho salmon returnng to UCI was 2.52 mllon (95% CI: 2.6-2.87 mllon). Gven a ommeral harvest of 0.25 mllon, the total esapement of oho salmon nto all UCI streams was 2.27 mllon (95% CI:.9-2.62 mllon), and the explotaton rate n the ommeral fshery was about 0%. Our rado tag estmate of the total esapement of oho salmon nto all UCI streams was.36 mllon (95% CI: 0.98-.96 mllon). Thus, our PIT taggng experment estmated a populaton sze for oho salmon enterng UCI streams that was hgher than the estmate obtaned from rado taggng. Although, the 95% onfdene ntervals around the two estmates overlapped slghtly, the z-test statst ndated the two estmates were sgnfantly dfferent. Of the total oho salmon esapement nto all UCI streams, 56% (0.76 mllon) returned to the Sustna and Lttle Sustna Rver dranages, 9% (0.26 mllon) returned to streams along the west sde of UCI, 7% (0.24 mllon) returned to streams along Knk Arm, 5% (0.07 mllon) returned to streams along Turnagn Arm, and 3% (0.04 mllon) returned to streams on the Kena Pennsula. However, these estmates for Turnagn Arm and Kena Pennsula streams do not nlude the entre esapement, beause we stopped taggng before the runs to these areas were omplete. Our PIT tag estmate of the total populaton sze of pnk salmon returnng to UCI was 2.28 mllon (95% CI:.60-40.96 mllon). However, ths estmate was of questonable value due to ts very low preson resultng from problems wth tag reovery. Therefore, we estmated a maxmum explotaton rate on pnk salmon n the ommeral fshery by smply summng esapements that were atually enumerated n 3 streams. Gven a ommeral harvest of 0.45 mllon, the maxmum explotaton rate n the ommeral fshery was about 2%. However, the atual explotaton rate must have been muh lower, beause we dd not nlude esapements nto numerous other streams around UCI. Our PIT tag estmate of the total populaton sze of hum salmon returnng to UCI was 3.88 mllon (95% CI: 3.30-4.47 mllon). Gven a ommeral harvest of 0.24 mllon, the total esapement of hum salmon nto all UCI streams was 3.64 mllon (95% CI: 3.06-4.23 mllon), and the explotaton rate n the ommeral fshery was about 6%. Despte unertanty n our salmon populaton estmates, t s reasonable to onlude that explotaton rates on oho, pnk, and hum salmon n the UCI ommeral fshery were substantally below optmal rates n 2002. KEY WORDS: Coho salmon, Onorhynhus ksuth, pnk salmon, O. gorbusha, hum salmon, O. keta, markreapture, passve ntegrated transponder tags, rado telemetry tags, total populaton sze, esapement, explotaton rate. x

INTRODUCTION Commeral salmon fsheres n Upper Cook Inlet (UCI) generally target sokeye salmon, but oho, pnk, and hum salmon are taken ndentally. In ts 999 meetng, the Board of Fsheres (BOF) dreted the Alaska Department of Fsh and Game (ADF&G) to develop a management plan for pnk salmon and management strateges for hum salmon n UCI. Untl that tme, the BOF dreted that no targeted pnk salmon fshng would be allowed n UCI. The BOF further dreted that no addtonal fshng perods would be allowed for the drft gllnet fshery outsde the Kena and Kaslof setons of the Upper Subdstrt untl sgnfant harvestable surpluses of hum salmon were avalable. The ommeral sokeye salmon fshery n UCI n 2000 experened a run falure. In August 2000, ommeral fshermen pettoned the BOF to open fshng for pnk salmon. Ther request for an extended ommeral fshery was dened, beause of lak of esapement nformaton for pnk salmon and onservaton onerns for oho salmon. At present, the ADF&G does not have a omprehensve program to estmate esapement, explotaton, and sustanable yelds for oho, pnk, and hum salmon n UCI. Although, esapements of these spees are enumerated or partally enumerated at several wers throughout the area, t s not known to what extent esapements n these systems represent overall produton n the area. The goal of ths projet was to estmate the total populaton sze, esapement, and explotaton rates for oho, pnk, and hum salmon returnng to UCI. Ths projet was a frst step toward determnng esapement levels needed to aheve sustaned yelds for these spees. Several methods have been used to assess stoks of salmon returnng to UCI, but eah has ts lmtatons. Wers have been used to enumerate salmon esapng to spawnng grounds on numerous streams around UCI. Whle these projets an provde aurate estmates of stok sze for ndvdual small streams, esapement estmates from wers on a small number of streams may not be representatve of trends over the entre nlet. Moble hydroaoust surveys have been used to estmate salmon populaton sze n UCI (Tarbox and Thorne, 996), but these surveys only provde an estmate of the populaton sze at the tme of the survey, so multple surveys would be requred to estmate total run sze and resdene tme would also need to be estmated. Aeral surveys provde a ost effetve means to estmate salmon esapements over large areas, but the large number of oluded glaal streams n UCI prelude use of ths tehnque n many systems. Sde-san sonars have been used to enumerate salmon mgratng n several large glaal streams around UCI, but aurate estmates are dffult to obtan when spees are mxed and mgratng throughout the rver ross seton. Marne mark-reapture experments an provde total populaton estmates for ndvdual salmon spees enablng esapements to be estmated after subtraton of the ommeral harvest. The methods used to estmate salmon populaton sze by mark-reapture were ntally developed n the 930 s and 940 s, but the orret oneptualzaton of analyss proedures were largely developed by Seber (962, 982). Hstorally wthn UCI, Thompson (930) used mark reapture to nvestgate salmon mgraton patterns n the nlet. Lkewse, Tyler and Noerenberg (967) studed salmon mgraton and noted that nearly all salmon tagged north of Anhor Pont were reaptured n UCI. Tarbox (988) orroborated these fndngs. Sne the late 970 s, the ADF&G has onduted an offshore test fshng (OTF) projet to estmate the populaton sze of sokeye salmon returnng to UCI durng the fshng season. The test fshng vessel fshes a drft gllnet eah day durng July at 6 statons along a transet runnng from Anhor Pont to the Red Rver delta on the west sde of Cook Inlet (Fgure ). Durng of the past 4 years, the ath per unt effort from the test fshng vessel has foreast the sze of the sokeye salmon run nto UCI to wthn 20% of the atual value (Shelds 2003). Although, none of these studes used mark-reapture to estmate the sze of salmon populatons returnng to UCI, they dd lay the groundwork for mark-reapture populaton

experments by demonstratng that nearly all salmon mgratng past Anhor Pont were destned for streams n UCI. Marne mark-reapture methods have been used suessfully to estmate the sze of salmon populatons returnng to Puget Sound and Kodak Island. Eames et al (98, 983) tagged oho and hum salmon n northern Puget Sound to estmate returns to partular rver systems n the regon. They demonstrated approprate use of stratfed populaton estmators when multple stoks were present and doumented that short-term mortalty assoated wth taggng these spees n saltwater was nsgnfant. Lkewse, Bevan (962) estmated the sze of sokeye salmon populatons returnng to Kodak Island, Alaska, notng that the majorty of the sokeye salmon returned to Karluk Lake. Bevan (962) found that tag loss was about 0%, and that tagged fsh exhbted a 48-hour lag before returnng to the populaton. Ths fndng was onsstent wth results from subsequent ultrason taggng studes whh demonstrated that tagged salmon ntally dve and reman at depth for about 48 hours before returnng to the surfae layer (Candy et al. 996). Most mark-reapture studes have used vsble tags, but ths approah an ntrodue an unknown bas nto populaton estmates f fshermen dsard tagged fsh. To avod ths problem, we used Passve Integrated Transponder (PIT) tags that were njeted nto the fsh and were not externally vsble. These tags an also be deteted usng eletron equpment, so tag reovery n proessng plants ould be automated and made muh less ntrusve to proessor operatons. PIT tags are onstruted wth an ntegrated rut hp onneted to a tghtly wound opper hoop antenna. The tags an be nterrogated by 25 khz sgnal from a sannng deve. When the sannng deve frequeny extes the PIT tag, the tag emts a sgnal bak to the reever wth a unque ode (0-dgt hexdemal ode dsplayed alphanumerally). The PIT tags are enapsulated n glass and are typally 2-mm long by 2.-mm wde. PIT tags have been used extensvely n researh on salmond survval (Prente 990; Skalsk et al. 998), movement (Prente et al. 990, Hldebrand and Krshner 2000) and behavor (Brannas et. al. 994), as well as, rustaean researh (Prente et al. 985; Penglly and Watson 994). When properly njeted n the body avty, PIT tags have hgh retenton rates (Prente et al. 990a) and mortalty rates of tagged fsh are low. Prente et al. (990a) found that tag retenton rates n males (00%) were slghtly hgher than n females (99.7%) f egg skens were not strpped from the fsh. Prente et al. (990b) desrbed a taggng method developed for Columba Rver salmond researh, and Prente et al. (990a) noted that all wounds were losed and healng by the thrd day after maturng Atlant salmon were PIT tagged. Prente (986) ompared juvenle hnook salmon and steelhead trout that were PIT tagged wth old brandng, oded wre taggng, old brandng and oded wre taggng, and a ontrol group (handlng but not tagged) at dams on the Columba rver. He noted no sgnfant mortalty of PIT tagged fsh when ompared to these other taggng methods. Smlarly, Qunn and Peterson (996) found no sgnfant mortalty of juvenle oho salmon that ould be attrbuted to PIT taggng. The fundamental assumptons of a mark-reapture experment are: () the populaton s losed, (2) all fsh have equal probablty of beng marked durng the frst samplng event, (3) tagged fsh do not suffer greater mortalty than untagged fsh, (4) fsh do not lose ther marks, (5) no marks are overlooked, and (6) ether marked and unmarked fsh are unformly mxed or the reaptures are a random sample (Seber 982). Volaton of these assumptons may not nvaldate estmaton of populaton sze by mark-reapture, f the magntude of the errors s known. We onduted several studes to estmate the magntude of these soures of error and orreted for ther effets on our populaton estmates. We also appled rado tags to oho salmon mgratng nto UCI. Ths omponent of the projet provded () a seond estmate of the sze of the total oho salmon populaton enterng UCI streams for omparson to our estmate derved from reovery of PIT tags, (2) an estmate of the populaton sze of oho salmon enterng eah major stream flowng nto UCI, (3) estmates of the tmng of varous stoks of oho salmon mgratng 2

past the OTF transet, (4) estmates of the tmng of varous stoks of oho salmon enterng ther natal streams, and (5) estmates of the resdene tme and mgraton rate of oho salmon n UCI. OBJECTIVES. Estmate short-term tag mortalty. 2. Apply PIT and rado tags. 3. Estmate rate of PIT tag loss. 4. Reover PIT tags at proessors and estmate PIT tag deteton rate. 5. Estmate salmon populaton szes and evaluate soures of error. 6. Estmate esapements of oho salmon usng rado telemetry. METHODS Objetve : Estmaton of short-term tag mortalty In 200, oho and hum salmon were aptured by a hartered purse sene vessel and tagged to estmate short-term tag mortalty. Dummy rado tags (n=200) were appled to oho salmon, and PIT tags (n=200) were appled to hum salmon. All tags were unquely numbered, and the tme eah fsh was tagged was reorded. The study on oho salmon was onduted n a lagoon near the Homer spt, and the study on hum salmon was onduted near the Wally H. Noerenberg hathery n Prne Wllam Sound. The methods used to handle and tag fsh were smlar to those used on the taggng vessels n UCI (objetve 2). Tagged fsh were mmedately released to a floatng net pen seured along sde the taggng vessel and held for 48 hours. All mortaltes were retreved and the tme eah fsh was held n the net pens pror to taggng was reorded. Mortaltes were enumerated for 4 lots of 50 sequentally tagged fsh,.e. 0-50, 50-00, 00-50, and 50-200. Lots of 50 fsh orresponded to holdng tmes of about 60 mns eah, sne ths was the tme requred to tag ths number of fsh. No ontrol group was nluded n the study, beause our goal was to estmate the mortalty assoated wth handlng and taggng. The survval of tagged fsh n eah of the lots was estmated from S t =m t /T t, where m t was the number of lve tagged fsh from lot t at the end of the experment, and T t was the total number of fsh tagged n lot t. The standard error of the estmate was alulated as desrbed by Zar (984). Objetve 2: Applaton of PIT and rado tags In 200, an approxmately 52 purse sene vessel (F/V Agave) was hartered from July to July 5. Ths vessel fshed an approxmately 200-fm sene (3.5 mesh, 375 meshes deep) to apture salmon for taggng. However, the harter was termnated before the end of the projet, so a seond 58 purse sene vessel (F/V Infnte Glory) was hartered from July 28 to August 4. Ths vessel fshed an approxmately 250-fm sene (3.5 mesh, 375 meshes deep). Lak of vessel support durng the entre salmon run preluded our estmaton of salmon populaton szes n 200. However, we were able to obtan useful nformaton regardng ath rates, fsh handlng and taggng methods, tag retenton rates, and some prelmnary oho salmon esapement dstrbuton data from rado tags. The methods used to apture, handle and tag fsh n 200 were generally smlar to those used n 2002. In 2002, two approxmately 58 purse sene vessels (F/V Just-n-Case and F/V Mllenum) were hartered (July 2 August 7) to apture salmon for taggng n UCI. Eah vessel fshed an approxmately 250-fm 3

sene (3.5 mesh, 375 meshes deep). All salmon were tagged wthn about 5 km of the OTF transet that runs from Anhor Pont to the Red Rver delta. Sne, our goal was to tag a representatve sample of salmon mgratng nto UCI, we attempted to tag fsh near eah of the sx OTF statons eah day. But, we also foused most of our fshng effort n areas along the OTF transet where salmon athes were hghest, beause the preson of our populaton estmate was dependent on the number of fsh tagged and reovered. The sene was generally set n an approxmate sem-rle, open nto the urrent for 20 mnutes at eah staton. After the sene was pursed, all jellyfsh and other debrs were removed from the bunt end of the sene. On board the F/V Just-n-Case, aptured fsh were rolled out of the sene nto -2 totes along the port sde of the vessel, and fsh to be tagged were sent down a hute to a seond set of totes on the starboard sde of the vessel and sorted by spees. On board the F/V Mllenum, aptured fsh were generally braled from the sene onto the dek and fsh to be tagged were qukly sorted by spees nto a set of totes on the starboard sde of the vessel. All totes used to hold fsh pror to taggng were suppled wth re-rulatng seawater. Salmon aptured n eah net set were generally enumerated by spees, but f a large number of a partular spees was aptured the number of that spees was vsually estmated. The start and stop tme of eah net set, oordnates (lattude, longtude), wnd veloty, and stage of tde were reorded for eah net set. Immedately before taggng, fsh were dp-netted from the re-rulatng seawater tote nto a love ol bath. Clove ol was used as an anesthet, beause anesthetzed fsh ould be harvested and onsumed on the same day (Pre and Powell 2000). The number of fsh held n the totes and the tme they were held was kept to a mnmum to redue mortalty. Pror to taggng, eah fsh was removed from the love ol bath, nspeted to nsure t had not already been tagged, measured (total length), and tagged wth an ndvdually dentfable PIT or rado tag. The tme at whh eah fsh was tagged was also reorded. PIT tags were appled to oho, pnk and hum salmon, and rado tags were appled to a subsample of oho salmon eah day. Rado tags were appled to oho salmon before fsh were PIT tagged. We used 25 khz ylndral glass enased PIT tags (20 mm x 3.2 mm). A hypoderm needle was used to njet eah PIT tag nto the fshes heek musle. The needles were perodally sterlzed by mmerson n a betadne soluton. Rado transmtters (20 mm x 55 mm) were mounted externally on oho salmon about 3-4 m below the dorsal fn. Two wres were passed through the fsh, and the tag fxed by rmpng a 2-m dameter plast Petersen ds tag (unquely numbered) onto the wre. We used 729 unque transmtter odes wth frequenes rangng from 50.054 50.963 mhz and 5 pulse odes wthn eah frequeny. Eah transmtter weghed about 5 g and had a battery lfe of about 80 days. Eah rado tag was sanned by a reever to establsh that t was transmttng before beng attahed to a fsh. Eah PIT tagged fsh was sanned pror to release to establsh that the tag was retaned and detetable. Tagged fsh were mmedately returned to the sea. We alulated the geometr mean ath per net set (CPUE ) for sokeye, oho, pnk, and hum salmon durng fve weekly (July -6, July 7-3, July 4-20, July 2-27, and after July 28) tag release strata () to evaluate the relatve abundane of eah spees and ther run tmng aross the OTF transet. Objetve 3: Estmaton of rate of PIT tag loss In 200, a double-markng experment was onduted wth sokeye salmon to estmate the rate of PIT tag loss. The sokeye salmon used n ths experment were aptured, handled, and PIT tagged usng methods desrbed n objetve 2, but T- bar anhor/dart tags were also appled to these fsh approxmately 3-4 m below the dorsal fn. Double-marked sokeye salmon were reovered by tehnans n fsh proessng plants and by ommeral and sport fshermen. An eletron PIT tag reader was used to san eah of these fsh for the presene of a PIT tag. If a tag was not deteted, the head was dsseted to determne f 4

the tag had been damaged and to evaluate how the tag may have been lost. The proporton of fsh that retaned a readable PIT tag was estmated from L =m p /m d, where m p was the number of double marked fsh that retaned a readable tag, and m d was the number of double-marked fsh examned for PIT tags. The standard error of the estmate was alulated as desrbed by Zar (984). Objetve 4: Reovery of PIT tags at proessors and estmaton of PIT tag deteton rates Eletron PIT tag readers were nstalled at eah major plant that proessed salmon from UCI. The readers were most often nstalled on hutes mmedately below the salmon header mahnes. These hutes were usually onstruted of approxmately 25-m dameter PVC ppe ut longtudnally nto half setons. Two hand-held raket antennas were attahed to eah hute usng zptes to provde for redundany n the deteton of PIT tags. The two antennas were attahed to the hute at dfferent angles, beause tag deteton s a funton of the angle of the tag n the eletromagnet feld reated by the antenna. The antennas were also attahed as far as possble away from eah other and from any metal or eletr motors to redue nterferene that mght redue tag deteton. A PIT tag reader was attahed to eah antenna by a able. The two readers needed for the nstallaton on eah hute were housed n a tote mmedately below the proessng lne. An external 2V battery was used to power both readers. The onfguraton of the nstallaton vared among proessng plants dependng on the desgn of the proessng equpment. We made every effort to maxmze tag deteton rate gven the onstrants of the envronment at eah plant. Tehnans mantaned the PIT tag readers and onduted tag deteton tests at eah proessng plant on most days durng the fshng season. Upon eah vst to the plant, the tehnans nspeted the readers for any problems wth the nstallaton (e.g. loose antenna, error messages on the reader, water damage, et.). The voltage on the external batteres was tested and the battery replaed wth a newly harged one f the voltage dropped below 2V. Upon eah vst, the tehnans reorded date, tme, proessor, lne number, PIT tag reader seral number, any problems wth the reader, battery voltage, and whether the battery had been replaed. In addton, tehnans onduted tag deteton tests upon eah vst to eah proessor. These tests nvolved passng 50 dummy or atual salmon heads that had been prevously PIT tagged past the antenna array attahed to eah hute. Dummy heads were onstruted of styrofoam gllnet floats ut laterally n half and shaped lke a salmon head. Atual salmon heads were also perodally retreved from the headng mahnes, PIT tagged n the heek and used for deteton tests. Deteton tests wth dummy heads were onduted to montor relatve tag deteton rates. Tests wth atual heads were used to albrate relatve rates to atual rates. These tests were generally onduted wth the proessng equpment operatng to replate atual ondtons durng the headng operaton. Deteton tests were not onduted wth atual heads at all tmes due to the extra work nvolved n perodally reylng these heads as they deomposed. Eah set of heads used for deteton tests was sanned by a PIT tag reader to reate a fle of the tag odes n the set. The tagged heads were tossed down the hute past the blade of the headng mahne to smulate the atual headng proess. After eah tag deteton test, the data from the two PIT tag readers attahed to eah hute was downloaded to a hand held omputer. Later n the laboratory, the data from the hand held omputer was downloaded to a desktop or laptop omputer and an algorthm run to alulate deteton rate. The algorthm ompared the tag odes n the deteton test set to the tag odes deteted by the reader durng the test. Tag deteton was estmated for eah day at eah proessor from d= m d /m t, where m d was the number of deteted tags, and m t was the number of known tagged dummy or atual heads sanned. The algorthm alulated deteton rate for eah reader and for both readers ombned,.e. f a tag was deteted by one reader but not the other. The algorthm wrote these three deteton rates and a lst of tag odes that were not deteted to a fle. Lsts of 5

undeteted tag odes were perodally nspeted to determne f spef odes were onsstently not deteted ndatng damage to the tag. Deteton tests were onduted wth dummy and atual salmon heads on the same proessng lnes at eah plant on several dfferent days. These data were used to alulate the dfferene between deteton rates estmated usng dummy versus atual heads. The Wloxon sgned ranks test was used to test whether the mean dfferene of ranks was sgnfantly dfferent from zero (Conover 999). The atual deteton rate at eah plant and proessng lne on eah day of the season was estmated from tests onduted wth atual heads when avalable. But, when only tests wth dummy heads were onduted, the atual rate was estmated by adjustng the relatve deteton rate obtaned usng dummy heads by the mean dfferene between the rates measured usng atual versus dummy heads. An analyss of varane (ANOVA) was onduted to test whether deteton rates dffered among proessng lnes and among fve tag reovery strata (July -6, July 7-3, July 4-20, July 2-27, July 28- August 3, and after August 4). An nteraton term was nluded n the model, and the least-squares mean deteton rate (d kj ) and standard error was estmated for eah of k proessors\lnes and j reovery strata. Objetve 5: Estmaton of salmon populaton szes and evaluaton of soures of error A smple Petersen estmate of the sze of the salmon populaton returnng to UCI s gven by ˆ n n2 N = () m 2 where n s the number of vald tagged fsh released by the purse sene vessel at tme, n 2 s the number of fsh sanned for tags at tme 2, and m 2 s the number of tagged fsh reovered at tme 2. The Peterson estmator provdes an unbased estmate of populaton sze when the followng ondtons are met: () all fsh n the populaton have the same probablty of beng tagged, or all fsh have the same probablty of beng aught n the seond reovery sample, or tagged fsh mx unformly wth untagged fsh, (2) losed populaton, (3) no tag loss, (4) no tags overlooked, and (5) taggng has no effet on fsh behavor. In the present study, we expet that assumpton s volated, beause at a mnmum fsh would have to be tagged n proporton to ther abundane as they ross the OTF transet, or ommeral harvests would have to be randomly dstrbuted, or fsh tagged at the begnnng of the run would have to mx equally wth fsh from the end of the run. One soluton to ths problem s to stratfy by tme. A stratfed Petersen method (Darroh estmator) was used to estmate the populatons of oho, pnk and hum salmon returnng to UCI. We used a Stratfed Populaton Analyss System (SPAS) software pakage developed spefally for analyss of data from stratfed mark-reapture experments (Arnason et al. 996). Ths software allows researhers to defne strata n spae or tme or both wth the s strata n whh markng took plae dfferng, f neessary, from the t reovery strata. Arnason et al (996) provded the followng notaton for mark-reapture experments. The number of strata at taggng and reapture are denoted by s and t, and statsts or parameters assoated wth these events are denoted by and r. The statsts are as follows: n r n number of fsh marked n release stratum, = s number of fsh taken n reovery stratum j, j = t. 6

m j the number of the n reovered n stratum j u j number of unmarked fsh reovered n stratum j. The parameters are as follows: N populaton sze at ntal (release) stratum, = s r N j p r p j θ j populaton sze n fnal (reovery) stratum j, j = t. probablty that a fsh n the ntal stratum at apture tme s aptured n that sample; = s. probablty that an fsh n fnal stratum j at reovery tme s reaptured n that sample; j = t. probablty that a fsh n stratum at apture s n stratum j at reovery tme. µ j expeted number of fsh tagged n strata that are reovered n strata j. The above statsts and parameters an be arranged nto a matrx (Table ) wth assoated populaton parameters (Table 2). The total populaton at tme of taggng ( N. ) s then gven by And, total populaton at tme of reovery ( r N. s N. = N (2) = ) s gven by t r r N. = N j (3) j= It s assumed that no part of the populaton enters reovery strata wthout beng part of one of the taggng strata. To ouple tables &2, the usual assumptons assoated wth mark-reapture experments are requred, and t s also assumed that: () fsh behave ndependently of one another wth respet to movement among strata, (2) all tagged fsh released n a stratum have the same probablty dstrbuton of movement to reovery strata, (3) all fsh n a reovery stratum behave ndependently n regard to beng aught and all have equal probablty of beng aught, (4) no tags are lost, and (5) tags are reorded properly and orretly upon deteton (Shwarz and Taylor 998). In addton, one or both of the followng assumptons are made dependng on whether the goal of the study s to estmate the number of fsh n the taggng or reovery strata: (6a) movement pattern, death, mgraton rates for both tagged and untagged fsh are the same n eah taggng stratum (requred to estmate the total populaton n the taggng strata), and (6b) the populaton s losed wth respet to movement among strata (requred to estmate the total populaton n the reovery strata). Gven these assumptons the expeted values of the statsts n table an be wrtten n terms of the followng parameters (Table 3). Let θ j equal the probablty that a fsh aptured n taggng stratum wll survve and mgrate to reovery stratum j, and let N j be the orrespondng number of fsh. If the populaton s losed, θ = for =,,s. and by defnton N j θ j =, =,..., s, j =,..., t (4) N There are a total of st + s + t parameters, the movement parameters, the ntal apture probabltes, and the reovery probabltes. Wth these parameters ertan funtons an be estmated under two dfferent senaros (Shwarz and Taylor 998). 7

Frst, the number of taggng strata may be less than or equal to the number of reovery strata (s t). Gven assumpton 6a, (same movement patterns of tagged and untagged fsh, but not neessarly losure over reovery strata), Banneheka et al. (997) showed that fsh n the populaton at tme of taggng ould be estmated. Gven ths senaro, the above models an be parameterzed wth st +2s parameters. The expeted number of fsh movng from taggng strata to reovery strata j that are tagged and reovered (st parameters) s gven by µ = N pθ p ; (5) j the odds that a fsh wll not be aptured at taggng stratum (s parameters), p β = ; (6) p and the expeted number of fsh tagged n stratum and never reovered (s parameters), t j= j r j r ( p ) γ = N p θ. (7) j One an desrbe the expeted values of the observed statsts and the number of fsh not seen (Shwarz and Taylor 998). The{β }are then essentally weghts that an be used to onstrut a lnear ombnaton of the rows of the E[m j ] that equals the E[u j ]. Thus, we an solve for the {β } to mnmze the sum of squares of the predtons,.e. t s ˆ u j β mj. (8) j= = However, we used an alternatve teratve maxmum-lkelhood tehnque to estmate the {β }, beause ths approah allows unertanty n the m j to be nluded n the estmaton proedure (Plante 990; Plante and Rvest 995). Ths proedure fnds estmates of the {β } that best predt the {u j } whle allowng the {m j } to vary around ther observed values n a way that s onsstent wth observed data but also mproves the ft (Shwarz and Taylor 998). We also alulated the effetve number of tags released n eah strata (Arnason et al. 996) by orretng for taggng-ndued mortalty {S t } and tag loss { L },.e. n _ eff = n S. (9) Two estmates of the oho salmon populaton were omputed. The frst estmate was orreted for short-term tag mortalty, and the seond estmate was orreted for long-term tag mortalty. Short-term tag mortalty was estmated from our net pen studes (objetve ). Long-term tag mortalty was estmated from the rato of the total number of rado tags reovered and the total number appled to oho salmon. Ths method provdes an estmate of the mnmum fraton of tagged oho salmon that survved and mgrated through the reovery area (ommeral fshng dstrts). We further alulated the effetve number of tags reovered by orretng for tag deteton d kj at eah proessor (k) durng eah reovery strata (j),.e. mj m _ effj =. (0) d kj t L j 2 8

Note that orretons for tag mortalty, tag loss, and tag deteton were made to mnmze bas n the populaton estmates. However, these orretons add varaton that was not aounted for n the standard errors and the onfdene ntervals for the populaton estmates (Arnason and Mlls 98). We ntally establshed weekly taggng and reovery strata (July -6, July 7-3, July 4-20, July 2-27, July 28-August 3, and after August 4). One the model was ft, goodness-of-ft tests were onduted to test whether any of the followng ondtons were satsfed:. the reovery probabltes were onstant aross strata, 2. the (expeted) rato of marked to unmarked fsh was onstant aross all reovery strata. Ths ould have been aheved n one of several ways. Two possbltes were: (a) the proporton of eah ntal stratum marked was onstant aross all apture strata and marked and unmarked anmals experened the same mgraton patterns, or (b) the mgraton pattern of marked and unmarked anmals aross fnal strata was ndependent of ther ntal strata (Arnason et al. 996). A omplete mxng test was used to test the hypothess that the probablty of resghtng a released anmal was ndependent of ts stratum of orgn. An equal proportons test was used to test the hypothess that the rato of marked to unmarked anmals was onstant aross reovery strata (Arnason et al. 996). If ether test passes (.e. p>0.05), t should be possble to pool strata, but ths s unusual n prate (Arnason et al. 996). In ether ase, falure to pass these tests does not prelude poolng, other fators must be onsdered (Arnason et al. 996). Poolng strata an nrease the preson of the estmate but wll ntrodue bas f done mproperly. Other than goodness-of-ft statsts, there are no formal tests to determne f one should pool or drop strata. 2 The χ and G 2 goodness-of-ft statsts were omputed to evaluate model ft,.e. s t ( m t s j mˆ j ) 2 ( u j uˆ) 2 ( n m 2. γˆ 2 ) χ = + + () mˆ u γˆ = j= j j= j = s t t s 2 j j or the = + G 2 mj ln u j ln + ( n m. ) m mˆ u uˆ = j= j j= j = n ln m γˆ. (2) The followng fators were onsdered when dentfyng strata to pool: () elmnaton of strata wth E[m j ]<5, (2) poolng of adjaent strata wth smlar ntal apture or reapture probabltes, and (3) mnmzaton of the standard error of the estmate. Poolngs that resulted n a large hange n the G 2 statst or standard error of the populaton estmate (greater than SE) were onsdered questonable (Arnason et al. 996). In addton, strata were dropped f the number of tags released or reovered was very small. Ths was neessary to mnmze the number of ells wth E[m j ]<5. Fnally, we onduted 5 analyses to evaluate soures of error n our populaton estmates. The frst 2 analyses were foused on whether the salmon tagged n our study were exlusvely mgratng north nto UCI. We frst onduted a h-square test of the null hypothess that the probablty of reapturng PIT tagged salmon dd not dffer for fsh that were aptured north versus south of 59.852 o N lattude. Approxmately, one half of the salmon PIT tagged n our study were tagged north of ths lattude. Seond, we onduted a h-square test of the null hypothess that the probablty of reapturng PIT tagged salmon dd not dffer for fsh that were aptured durng ebb, flood, or slak tdes. Next, we onduted a h-square test of the null hypothess that the probablty of reapturng PIT tagged salmon dd not dffer among three groups that were held on the taggng vessels for <30 mns, 30-60 mns, and 9

>60 mns. Sne tme was reorded when eah fsh was tagged, we were able to nlude all of our PIT tag data n ths analyss. In these frst 3 analyses, separate tests were onduted for eah spees and for all spees ombned. A fourth h-square analyss was onduted to test the null hypothess that the probablty of reapturng PIT tagged salmon dd not dffer among sx length lasses (<50 m, 50-55 m, 55-60 m, 60-65 m, 65-70 m, >70 m). Ths analyss was onduted wth all spees ombned, and the length dstrbuton of eah spees was also alulated for omparson. A fnal h-square analyss was onduted to test the null hypothess that the rato of the number of tagged to untagged salmon dd not dffer among seven proessors n UCI. Separate analyses were onduted for eah spees. Objetve 6: Rado telemetry study on oho salmon Rado tagged oho salmon were traked from a fxed-wng arraft usng a reever nterfaed wth a data olleton omputer (Advaned Telemetry Systems) and ontrolled by an external hand-held omputer nterfaed wth a global postonng system. The oordnates and alttude of the arraft were ontnuously logged at user defned dstane ntervals usually between 50 and 00 m. Ths system allowed tags to be qukly nterrogated wth data regardng frequeny, pulse ode, number of hts, date, tme of day, and oordnates of eah tag easly logged to a data fle n flght. The data olleton omputer was set to yle between frequenes at ntervals from -2 seonds per frequeny. In 2002, streams flowng nto UCI were surveyed one eah week from md July through September. On August 22 & 29, streams south of the OTF transet were surveyed one to determne f any rado tagged oho salmon mgrated southward. Ths survey overed streams south to Cottonwood Bay on the west sde of the nlet and on the east sde from Port Graham nto Kahemak Bay. Only the lower portons of eah watershed were surveyed durng these flghts to mnmze ost and survey tme. Later n Otober 2002, most of the UCI dranage basn was surveyed to doument the loaton of tagged salmon wthn eah watershed. Anhorage area streams and streams south of Bg Rver were not nluded n these surveys of the entre dranage basn. In 200, only one aeral survey was onduted to loate any tags that had entered the lower portons of streams flowng nto the nlet. In 200 and 2002, fxed reevers were operated on the Sustna Rver near Sustna Staton and on the Yentna Rver approxmately 3 mles above the Yentna sonar ste. Reevers operated by the Sport Fsh Dvson of ADF&G sanned for tags on the Kena, Kaslof, and Swanson rvers. All of the analyses desrbed below were onduted usng data from 2002 exept that a map of the dstrbuton of rado tag reoveres around UCI was onstruted usng data from 200. We ntally used our rado tag data to estmate the total populaton sze of oho salmon enterng UCI streams for omparson to our PIT tag estmate. Rado tag reoveres and oho salmon wer ounts were avalable from fve streams flowng nto northern Cook Inlet (Deshka R., Lttle Sustna R., Fsh Creek, Cottonwood Creek, and Waslla Creek). We ntally onsdered those portons of the fve streams above the wers as fve reovery strata wth a sngle release stratum. The statsts were the total number of rado-tagged oho salmon loated n all UCI streams nludng those aught n rereatonal fsheres (n ), the number of oho salmon ounted through eah of the j wers (n 2j ), and the number of rado-tagged oho salmon loated above eah of the j wers (m 2 j ). Rado tags not loated n freshwater (.e. aptured n the ommeral fshery, et.) were exluded from ths analyss, beause we were estmatng only the populaton sze of oho salmon that entered freshwater. We next onduted a h-square test of the null hypothess of equal marked proportons among reovery strata. The pooled-petersen method was then used to estmate the total populaton sze of oho salmon enterng all UCI streams (N Rado ) derved from rado tag reoveres. Sne, the sample sze was relatvely small, an nverse ube root transform of the estmate was used to alulate the onfdene nterval (Arnason et al. 99). 0

Our PIT tag estmate of the total populaton of oho salmon returnng to UCI was then used to alulate the populaton sze of oho salmon enterng all UCI streams (N PIT ) by subtratng the ommeral harvest from the total populaton. The PIT and rado tag estmates of the populaton of oho salmon enterng all UCI streams were then ompared. The z-test statst was used to test whether the two estmates dffered,.e. N PIT N Rado z =, (3) Var N N where ( ) PIT Rado ( N N ) = Var( N ) Var( N ) Var +. (4) PIT Rado Ths test assumes that the two abundane estmates are ndependent and normally dstrbuted. The esapement of oho salmon nto eah of 33 major streams (N k ) was estmated from N = p N, (5) k where p k was the weghted proporton of the total number of reovered rado tags (m k ) from taggng strata found n freshwater n eah (k) stream,.e. w mk p k =. (6) w m k k k To orret for apparent unequal taggng proportons among release strata, the number of rado tags (m ) reovered n eah stream was weghted (w ) by the mean CPUE n eah () release stratum and the nverse of the proporton of tags used n release strata,.e. CPUE CPUE w =. (7) m m The varane of the estmated esapement of oho salmon nto eah stream, Var(N k ), was estmated from 2 2 Var( N ) = N Var( p ) + p Var( N ) Var( p ) Var( N ) (8) k T k (Goodman 960). An estmate of the varane of p k was derved from m k m k Var( mk ) = M, M M (9) m k m k Var( wmk ) = w2 M, M M (20) PIT k T T Rado m k m = k Var wmk w2 M, (2) M M w mk Var ( pk ) = Var, (22) w mk k k T

where M = m k k Var 2 m k m k w M M M p, (23) wmk k ( k ) = 2 = number of rado tags from strata reovered n freshwater. The same method was used to estmate the esapement of oho salmon nto 5 areas (Westsde, Sustna, Knk Arm, Turnagn Arm, and Kena Pennsula) around UCI by smply poolng the data from streams wthn eah area. The area alled Westsde nluded all streams flowng nto UCI west of the Sustna Rver. Poolng tags reovered n these fve areas nreased the number of tag reoveres and narrowed the onfdene ntervals around the estmated populaton szes. We then examned the dstrbuton of rado tag reoveres among the 33 streams flowng nto the nlet by ther date of release from the taggng vessels. The weghted proporton of the total number of reovered rado tags (n freshwater) found n eah stream was plotted on a map of UCI usng data for releases pror to and after July 20. Ths analyss was onduted usng data from 200 as well as 2002 for omparson of dstrbutons between years, but proportons were not weghted n the 200 analyss due to lak of CPUE data throughout the entre run. Next, we examned the tmng of seven stoks of oho salmon mgratng past the OTF transet by estmatng the proporton of total rado tag reoveres n eah area by ther date of release from the taggng vessel. The seven stoks were defned by the fve areas prevously desrbed exept Sustna R., Yentna R., and Lttle Sustna R. were treated as separate stoks, beause there were suffent tag reoveres n these streams for the analyss. We onduted a h-square test of the null hypothess that the proporton of total tags reovered for eah stok dd not dffer by ther date of release. The frst release stratum and the Kena Pennsula stok were omtted from the h-square analyss, beause the small number of tag reoveres n these ells resulted n expeted values less than fve. We also examned the run tmng of these seven stoks of oho salmon nto freshwater usng the date eah rado tag was frst deteted n eah stream. Rado tags returned by rereatonal fshermen were not nluded n ths analyss, beause the date of entry of these fsh nto freshwater ould not be presely determned. We onduted a h-square test of the null hypothess that the proporton of total tags reovered for eah stok dd not dffer by ther date of entry nto freshwater. Sx reovery strata were establshed for ths analyss (July 4-20, July 2-27, July 28-Aug. 3, Aug. 4-0, Aug. -7, after Aug. 8). The frst and last reovery strata and the Kena Pennsula and Turnagn Arm stoks were omtted from ths h-square analyss, beause the small number of tag reoveres n these ells resulted n expeted values less than fve. We further examned the mgraton patterns of oho salmon through UCI by estmatng the proporton of total rado tags reovered by ther date of release and ther date of entry nto freshwater. Ths analyss was onduted for all stoks pooled and for eah of the seven stoks separately. Resdene tmes and mgraton rates of oho salmon were examned n relaton to stok of orgn and mgraton tmng aross the OTF transet. Resdene tme was estmated by the dfferene between the date eah fsh was frst deteted n freshwater and ts date of release from the taggng vessel. The straght-lne dstane from the OTF transet to the mouth of eah stream was used as a measure of the mnmum dstane eah fsh traveled n the nlet. Mgraton rate was estmated by the rato of the mnmum dstane traveled and resdene tme n the nlet. Two ANOVAs were onduted to test the null hypotheses that mean resdene tme and mean mgraton rate dd not dffer by stok of orgn or date of release. Eah dependent varable was natural-logarthm transformed pror to the analyss and an nteraton term was ntally nluded n the model. Fnally, we examned travel tmes for oho salmon between our fxed rado tag reevers at Sunshne Staton and Yentna R. An ANOVA was onduted to test the null hypothess that 2

travel tmes (natural-logarthm transformed) dd not dffer by the date eah fsh was frst deteted by the reever at Sunshne Staton. RESULTS Objetve : Estmaton of short-term tag mortalty Survval (S t ) of tagged oho salmon delned from 88% to 56% as holdng tme nreased from less than 83 mns to 25 mns (Table 4). Survval of tagged hum salmon was onsstently hgh and not learly related to holdng tme. Sne our study on oho salmon was onduted n a shallow lagoon, tagged salmon may have been exposed to anox mud near the bottom of the net pen. It s not lear whether ths affeted our results, but ths was not a fator n our study on hum salmon, beause t was onduted n a deep bay. Objetve 2: Applaton of PIT and rado tags The number of net sets made durng fve weekly tme perods ranged from 34 to 75 (Table 5). The number of sets made eah week was lowest durng late July, beause CPUE peaked at ths tme so fewer sets were requred to ath the fsh needed for taggng. Also, we restrted the number of PIT tags appled eah day durng ths tme to avod exhaustng our supply of tags. The CPUE for all 4 spees of salmon peaked the thrd week of July, and t was hghest for sokeye salmon (Table 5). The CPUE for sokeye and hum salmon delned at a greater rate n late July than t dd for oho and pnk salmon. PIT tags were appled to 4,925 oho salmon, 5,338 pnk salmon, and 5,07 hum salmon (Table 6). Rado tags were appled to 729 oho salmon. The total ath of oho, pnk, and hum salmon delned at a slower rate n late July than dd the CPUE. The number of net sets made eah day was nreased durng ths perod to mantan the number of tags released. Objetve 3: Estmaton of rate of PIT tag loss One hundred and sxty eght double-marked sokeye salmon were reovered to estmate PIT tag loss. Seventy nne perent of these fsh were reovered at proessors and the remander n the esapement or rereatonal fshery. One hundred and ffty three ( L =0.9, SE=0.02) of these fsh retaned a readable PIT tag. We dd not fnd any PIT tags that ould no longer be deoded by the eletron PIT tag reader, and we found no dfferene between the lengths of those fsh that retaned versus lost the PIT tag. Objetve 4: Reovery of PIT tags at proessors and estmaton of PIT tag deteton rates PIT tag readers were nstalled at seven plants that proessed salmon harvested n UCI. The onfguraton of proessng equpment at Oean Beauty and Snug Harbor prevented an effetve nstallaton of PIT tag readers pror to July 27. Modfatons were made to the equpment at these plants allowng readers to be nstalled and operated after that date. We sanned 73% of the ommeral harvest of oho salmon, 42% of the pnk salmon harvest, and 75% of the hum salmon harvest n UCI n 2002. The fraton of the pnk salmon harvest that we sanned was relatvely low, beause several proessors dd not pass pnk salmon through the headng mahnes. The mean dfferene between deteton rates estmated usng dummy versus atual salmon heads ranged from 0-0.47 (Table 7). These mean dfferenes were used to adjust deteton rates estmated usng dummy heads at the four plants lsted n Table 7. The relatvely large adjustment fator at Salamantof was only 3

appled pror to July 23 when all tests were onduted wth dummy heads. On that date, the onfguraton of the antenna array at Salamantof was modfed, and all subsequent tests were onduted wth atual salmon heads. No adjustments were neessary at Oean Beauty and Snug Harbor, beause all deteton tests were onduted wth atual salmon heads at these plants. An ANOVA ndated that mean deteton rates dffered sgnfantly (p<0.00) among proessors and reovery strata. At Ile Seafoods and Oean Beauty, deteton rates also dffered (p<0.05) among proessng lnes. Mean deteton rates (d kj ) ranged from 0.37 on lne 3 to 0.98 on lne 2 both at Ile Seafoods (Table 8). The low rate on lne 3 was due to the onfguraton of the proessng equpment. Ths lne was only used to proess pnk salmon. Objetve 5: Estmaton of salmon populaton szes and evaluaton of soures of error Of the 4,925 PIT tags appled to oho salmon, we deteted 67 at the 7 salmon proessors nluded n our study (Appendx ). When the total number of tags appled was adjusted for short-term tag mortalty and tag loss, the effetve number of tags released was redued to 3,944 (Table 9). A short-term survval rate of 0.88 (SE=0.05) was used n ths analyss, beause ths was the survval of oho salmon held less than 83 mns pror to taggng n our net pen study, and most of the oho salmon tagged n UCI were held for less tme. When the number of tags reovered was adjusted for tag deteton, the effetve number of reovered tags was nreased to 24. In every ase, the peak number of reoveres from eah release stratum ourred one week after release, and tags from eah release stratum were reovered over a 3-4 week perod after release. No tags were reovered from the frst release stratum durng the frst week of July, and no tags were deteted at proessors durng the frst two reovery strata. These strata were dropped from the analyss. The remanng strata nluded 98% of the harvest that was sanned for tags. We attempted several dfferent poolngs. The fnal model, whh produed the lowest standard error of the populaton estmate, nvolved poolng reovery strata for the weeks begnnng July 4 and 2 (Table 0). Ths model resulted n of 2 ells wth E[m j ]<5. The G 2 statst for ths model ndated no sgnfant dfferene (p=0.08) between observed and ftted reoveres (m j ). The estmated populaton sze was 3.22 mllon wth a 95% onfdene nterval from 2.76-3.68 mllon. The estmated populaton sze was greatest durng the mddle of July. For omparson, the pooled Petersen populaton estmate was 3.9 mllon. We also estmated the oho salmon populaton after adjustng the number of tags released for long-term tag mortalty and tag loss. Long-term tag mortalty was estmated from reoveres of rado-tagged oho salmon. We loated 58 of 729 rado-tagged oho salmon released resultng n a long-term mnmum survval of 0.7 (SE=0.02). The strata retaned and the fnal poolng were the same as n the prevous analyss. The G 2 statst also ndated no sgnfant dfferene (p=0.08) between observed and ftted reoveres (Table ). The estmated populaton sze was 2.52 mllon wth a 95% onfdene nterval from 2.6-2.87 mllon. The estmated populaton sze was greatest durng the mddle of July. For omparson, the pooled Petersen populaton estmate was 2.58 mllon. Of the 5,333 PIT tags appled to pnk salmon, we deteted only 45 at proessng plants (Appendx ). When the total number of tags appled was adjusted for short-term tag mortalty and tag loss, the effetve number of tags released was redued to 4,809 (Table 2). When the number of tags reovered was adjusted for tag deteton, the effetve number of reovered tags was nreased to 85. Ths relatvely large adjustment to the tag reoveres for pnk salmon resulted n large part, beause the greatest numbers of pnk salmon were proessed at Ile Seafoods, and all of these fsh were proessed on lne 3, whh had a farly low tag deteton rate. The peak number of reoveres from most release strata ourred one week after release wth one exepton. The peak number of reoveres from the last release strata ourred durng the same week the fsh were released. Also, the perod of tme over whh tags were reovered was less for pnk than oho salmon. Tags from eah release strata were reovered over a -3 week perod after release. As wth oho salmon, no tags were reovered from the frst release stratum, and no tags were deteted at proessors 4

durng the frst two reovery strata. These strata were dropped from the analyss. The remanng strata nluded 99% of the harvest that was sanned for tags. Several dfferent poolngs were attempted, the fnal model nvolved poolng reovery strata for the weeks begnnng July 2 and 28 (Table 3). Ths model resulted n 6 of 2 ells wth E[m j ]<5. The G 2 statst for ths model ndated no sgnfant dfferene (p=0.6) between observed and ftted reoveres (m j ). The estmated populaton sze was 2.28 mllon, but the preson was poor wth a 95% onfdene nterval from.60-40.96 mllon. The estmated populaton sze was greatest durng the frst week of August. For omparson, the pooled Petersen populaton estmate was 3.92 mllon. Of the 5,07 PIT tags appled to hum salmon, we deteted 54 at the 7 salmon proessors nluded n our study (Appendx ). When the total number of tags appled was adjusted for short-term tag mortalty and tag loss, the effetve number of tags released was redued to 4,568 (Table 4). When the number of tags reovered was adjusted for tag deteton, the effetve number of reovered tags was nreased to 97. Tags were reovered n all reovery strata. Smlar to pnk salmon, the peak number of reoveres from most release strata ourred one week after release wth one exepton. The peak number of reoveres from the last release strata ourred durng the same week the fsh were released. Reovery strata begnnng July and August 4 were dropped from the analyss, beause of the relatvely small number of hum salmon sanned for tags and small number of tags reovered n these strata. The remanng strata nluded 92% of the harvest that was sanned for tags. We attempted several dfferent poolngs. The fnal model nvolved poolng release strata for weeks begnnng July and 7, and July 2 and 28. Also, reovery strata were pooled for weeks begnnng July 7 and 4, and July 2 and 28 (Table 5). Ths model resulted n no ells wth E[m j ]<5. The G 2 statst for ths model ndated no sgnfant dfferene (p=0.95) between observed and ftted reoveres (m j ). The estmated populaton sze was 3.88 mllon wth a 95% onfdene nterval from 3.30-4.47 mllon. The estmated populaton sze was greatest durng early July. For omparson, the pooled Petersen populaton estmate was 3.74 mllon. The probablty of reapturng PIT tagged oho, pnk, and hum salmon was not sgnfantly related to the lattude where the fsh were aptured. However, the probablty of reapturng PIT tagged hum salmon was sgnfantly greater (p<0.0) when the fsh were aptured durng a flood or slak tde (Table 6). When the data from all spees were pooled, reapture probabltes were stll sgnfantly related to stage of tde (p<0.0). For all 3 spees of salmon, the probablty of reapturng PIT tagged salmon nreased wth the tme fsh were held on the taggng vessel, but the dfferenes were only sgnfant for hum salmon (p=0.02) and when data from all spees (p=0.0) were pooled (Table 7). Results from a h-square test also ndated that the probablty of reapturng PIT tagged salmon was sgnfantly dfferent (p<0.0) among sx length lasses of salmon (Table 8). Comparson of reovery probabltes and salmon length dstrbutons ndated that the numbers of tags reovered from the smaller pnk salmon were lkely redued due to the seletve nature of gllnet harvests. The tagged-to-untagged rato for oho salmon dd not dffer (p>0.05) among seven proessors, but ths rato dd dffer (p<0.05) among proessors for pnk and hum salmon (Table 9). Ths result dd not hange when the number of tag reoveres was adjusted for tag deteton rates measured at eah proessor. Tagged-to-untagged ratos were onsstently hgher at Ile Seafoods and Oean Beauty. Objetve 6: Rado telemetry study on oho salmon In 200, 67 oho salmon were rado tagged and 4 (68%) were later loated n the UCI area. Nne perent of these fsh were returned from ommeral fshery and 54% were found n streams. In 2002, 729 oho salmon were rado tagged and 58 (7%) were later loated n the UCI area. Seven perent of these fsh were returned from the ommeral fshery, 4% were returned from the rereatonal fshery, 69% were loated n freshwater by ether an arraft or fxed reever, 7% were loated by arraft n the 5

ntertdal zone but were not later loated n freshwater, and 3% were ether returned to ADF&G wthout any addtonal nformaton or were mpresely loated by other means. The fates of the tagged salmon were somewhat related to ther dates of release from the taggng vessel. Sxty-four perent of the tags returned by ommeral fshermen were tagged after July 20, and 63% of the tags found only n the ntertdal zone were tagged after July 20. We frst used our 2002 rado tag data to estmate the total oho salmon populaton enterng all UCI streams. Ch-square analyss ndated that we ould not rejet the null hypothess (p=0.2) of equal marked proportons of oho salmon returnng to the fve streams flowng nto UCI (Table 20). There was also no apparent relatonshp between the run tmng of oho salmon nto eah stream and ther marked proportons. Thus, we used the pooled-petersen method to estmate the total populaton sze of oho salmon enterng all UCI streams. The pont estmate was.36 mllon wth a 95% onfdene nterval of 0.98-.96 mllon. When the 2002 ommeral harvest of oho salmon n UCI (0.25 mllon) was subtrated from the total oho salmon populaton estmated usng PIT tags (Table ), the pont estmate for the oho salmon populaton enterng all UCI streams was 2.27 mllon wth a 95% onfdene nterval of.9-2.62 mllon. Thus, our PIT taggng experment estmated a populaton sze for oho salmon enterng UCI streams that was hgher than the estmate obtaned from rado taggng. Although, the 95% onfdene ntervals around the two estmates overlapped slghtly, the z-test statst ndated the two estmates were sgnfantly (p=0.002) dfferent. We next parttoned our estmate of the total oho salmon esapement to 33 streams flowng nto the nlet. The numbers of rado tags reovered n eah stream were frst weghted (w ) by the mean CPUE n eah () release stratum: July, w =0.25; July 7, w =0.39; July 4, w =.78; July 2, w =.42; July 28, w =0.58. Estmated numbers of oho salmon esapng nto the 33 streams ranged from 2,05 n several small streams to 357,99 n the Sustna Rver (Table 2). Due to the small number of tag reoveres n ndvdual streams, the 95% onfdene ntervals around these estmates overlapped zero n about 66% of the ases. But, when the data were pooled nto 5 areas, the 95% onfdene ntervals around the estmates dd not overlap zero. Coho salmon mgratng past the OTF transet before July 20 returned prmarly to the Sustna dranage, whle those mgratng later n the season returned prmarly to other streams around the nlet on both the west and east sdes. Of the 67 oho salmon tagged before July 20, 200, 4 were later found n 7 streams around the nlet and 68% of these were found n the Sustna Rver dranage (Fgure ). Of the 372 oho salmon tagged before July 20, 2002, 99 were later found n 2 streams around the nlet and 60% of these were found n the Sustna Rver dranage (Fgure 2). Of the 358 oho salmon tagged after July 20, 2002, 78 were later found n 29 streams around the nlet and only 34% of these were found n the Sustna Rver dranage (Fgure 3). Two hundred and seventy one tagged oho salmon were loated durng aeral surveys of the entre UCI dranage basn n Otober, 2002 (Fgure 4). Tagged oho salmon were found throughout many parts of the Sustna, Lttle Sustna, and Beluga Rver watersheds. In the Lttle Sustna Rver, 9 tagged oho salmon were found above the wer loated near the Parks Hghway and 9 were found below the wer. The tmng of oho salmon mgratng aross the OTF transet was sgnfantly (p<0.00) dfferent among 7 stoks. Greater than 50% of the oho salmon returnng to the Westsde, Turnagn Arm and Kena Pennsula mgrated aross the OTF transet after July 20 (Table 22). The mgraton of oho salmon returnng to the Sustna dranage, Lttle Sustna Rver, and Knk Arm peaked durng the week of July 4. The tmng of entry nto freshwater also dffered sgnfantly (p<0.00) among these 7 stoks of oho salmon. The mgraton of oho salmon enterng freshwater along the Westsde, Knk Arm and the Lttle Sustna Rver peaked the week of Aug. 4, whle the peak of the mgraton nto freshwater was earler for salmon returnng to the Sustna dranage, and later for salmon returnng to Turnagn Arm and Kena Pennsula (Table 23). Examnaton of the mgraton patterns of oho salmon through UCI (all stoks ombned) ndated that 6

ther mgraton aross the OTF transet peaked the week of July 4 whle entry nto freshwater peaked from July 28 through Aug. 0 (Table 24). A smlar desrpton of the ndvdual mgraton patterns of these 7 oho salmon stoks s provded n Appendx 2. An ANOVA ndated that the resdene tme of oho salmon dffered sgnfantly (R 2 =0.260, df=0, p<0.00) among 7 stoks and 5 release strata. Smlarly, ANOVA ndated that the mgraton rate of oho salmon also dffered sgnfantly (R 2 =0.44, df=0, p<0.00) among 7 stoks and 5 release strata. The nteraton terms were not sgnfant n ether of these models. Coho salmon returnng to the Sustna dranage exhbted shorter resdene tmes and hgher mgraton rates through UCI than the other 5 stoks nluded n the analyss (Table 25). The mgraton rate of oho salmon through UCI nreased from 6.7 km/day n early July to 4.9 km/day n late July (Table 26). Fnally, the travel tmes for oho salmon between our fxed reevers at Sustna Staton and Yentna Rver dd not dffer by ther date of arrval at the Sustna Staton reever. The mean travel tme between the 2 reevers was 3.5 days and the dstane between the 2 stes was 20.5 km. DISCUSSION The auray of mark-reapture estmates of populaton sze s dependent on the degree to whh the underlyng model assumptons are satsfed. The pooled Peterson estmator s only vald f all ndvduals have equal probablty of beng tagged and reaptured. In our PIT taggng study, ths assumpton was not satsfed, beause fsh probably were not tagged n proporton to ther relatve abundane and reapture probabltes vared over tme due to hangng explotaton rates n the ommeral fshery. Therefore, we used the stratfed Darroh estmator to redue bas resultng from varable ntal apture and fnal reapture probabltes. In our analyss, we also appled orreton fators for taggng-ndued mortalty, tag loss and tag deteton. Ths was done to mnmze bas n our populaton estmates that ould otherwse result from volaton of model assumptons. Estmatng taggng-ndued mortalty s problemat due to the dffulty of desgnng holdng studes that smulate natural ondtons. Our estmates of short-term mortalty were lkely a mnmum estmate of atual taggng-ndued mortalty, beause net pen studes of ths knd annot measure delayed mortalty that may result from the stress of handlng. Candy et al. (996) estmated mortalty of purse sene aught hnook salmon usng ultrason telemetry. They doumented a delayed mortalty of 23% ourrng 8-2 hrs after release and attrbuted t to stress-related physologal hanges ndued by hyperatvty durng apture. Laboratory studes have shown that the stress of apture auses blood lat ad levels to nrease for up to 4 hrs after apture wth mortalty ourrng f rtal levels of latate are reahed (Parker and Blak 959; Parker et al. 959; Farrell et al. 2000). Candy et al. (996) found that delayed mortalty of hnook salmon nreased from zero to 50% for fsh held <5 mns versus > 30 mns. To evaluate whether delayed mortalty was related to holdng tme on the taggng vessel, we tested for a dfferene n the probablty of reapture for groups of PIT tagged salmon held for dfferent lengths of tme. Holdng tme was not sgnfantly related to probablty of reapture for oho and pnk salmon. But, we were surprsed to fnd that the probablty of reapture nreased slghtly wth holdng tme for hum salmon (Table 7). Perhaps the stress of handlng aused these fsh to beome more vulnerable to apture n the gllnet fshery wthout ausng dret mortalty. We also used reoveres of rado tags to estmate the maxmum long-term mortalty of oho salmon. Applaton of ths estmate of taggng-ndued mortalty produed a mnmum PIT tag populaton estmate for oho salmon (Table ) sne atual mortalty was lkely not hgher. Although, we do not know whether mortalty dffers between fsh that were rado tagged versus PIT tagged, the dfferene f any may be small sne mortalty of oho (Farrell et al. 2000) and hnook salmon (Candy et al. 996) was not strongly related method of handlng or obvous njures. 7

We were also surprsed to fnd 7% of our rado tagged oho salmon n the ntertdal zone near the mouths of several rvers. These fsh were never loated n freshwater. The transmtters attahed to many of these fsh emtted a mortalty ode ndatng that the fsh were dead or had not moved reently. Some of these fsh may have moved nto freshwater undeteted and later washed downstream after spawnng, or they may have ded, beause they ould not osmoregulate suessfully n freshwater. If so, t s not lear whether ths ould have resulted from the stress of taggng, but t has been amply demonstrated that stress nterferes wth osmoregulaton (Clarke and Hrano 995). We used PIT tags to estmate the populaton sze of oho, pnk, and hum salmon n part beause ths method elmnated the potental problem of under reportng of tags by fshermen. However, use of PIT tags requred orretng for tag deteton rates at salmon proessng plants. Our approah nvolved estmatng deteton rates daly on eah proessng lne at eah plant. PIT tag deteton rates were affeted by the onfguraton of the proessng equpment at eah plant. The best deteton rates were aheved at plants where the tag reader antennas were not n lose proxmty to the salmon header mahnes, beause the vbraton of these mahnes sometmes affeted tag deteton. Durng the early part of the season, we were unable to effetvely san for tags at three proessng plants due to problems wth the onfguraton of the proessng lnes (Table 8). Ths redued the fraton of the total harvest that was sanned for tags. Dfferenes n unorreted marked proportons among proessng plants an also be used to evaluate whether tag deteton rates dffered among plants. Our h-square test ndated no dfferene n the marked proportons among proessors for oho salmon but there was a sgnfant dfferene for pnk and hum salmon (Table 9). However, when orretons for measured deteton rates at eah proessor were appled, the results dd not hange. Ths suggests that the dfferent marked proportons among proessors were due to somethng other than varable tag deteton rates. The hghest marked proportons ourred at Ile Seafoods and Oean Beauty. We examned whether marked proportons were related to numbers of fsh proessed from set versus drft gllnet harvests at eah proessor, but there was no apparent relatonshp. It may be that dfferent marked proportons among proessors were related to loatons n the nlet where fsh were harvested. But, we were unable to effetvely evaluate ths, beause data on loatons of harvests n the drft fshery are not very aurate. Our PIT and rado tag estmates of the oho salmon populaton sze lkely braket the atual populaton sze. Both methods nvolved taggng fsh usng the same gear type n the same area, but the reovery methods were very dfferent. Commeral fshng vessels reovered PIT tagged salmon n saltwater, whle rado tagged salmon were loated n freshwater by fxed reevers and arraft. Bas n our pooled- Petersen estmate derved from rado tag reoveres may have been mnmal, beause any taggngndued mortalty lkely ourred before the fsh entered freshwater, and there was lkely onsderable mxng of tagged and untagged fsh between ther release from the taggng vessel and entry nto freshwater. Mxng of oho salmon n the nlet was evdent from ther relatvely long resdene tmes (Tables 25 & 26) and the upper trangular struture n the reovery matres (Shwarz and Taylor 998) onstruted from our PIT (Table 9) and rado tag data (Table 24). Our oho salmon populaton estmate ould have been based f the probablty of loatng rado tags above the wers was dfferent from the probablty of loatng all other rado tags found n freshwater. Our last survey to loate rado tags above wers on streams east of the Sustna Rver was not onduted untl late Otober due to poor weather earler n the month. Loss of voltage n the transmtter batteres ould have affeted our probablty of loatng tags durng ths later survey. The battery manufaturer spefed a 60-day lfe for the batteres used n our study, and Advaned Telemetry Systems warrantes these batteres for 80 days of operaton. About 05 days elapsed between the tme these fsh were tagged and the last survey. Prevous experene wth these transmtters has ndated the lfe of most of the batteres s about two tmes the warranted lfe (pers. omm., Jay Carlon, ADF&G Sport Fsh Dvson, Soldotna, Alaska). To further evaluate ths queston, we onduted a h-square analyss to test whether marked proportons dffered between 8

Deshka Rver, whh was surveyed n early Otober, and those streams loated east of the Sustna Rver, whh were surveyed n late Otober. There was no dfferene (p>0.0). Our PIT tag oho salmon populaton estmate ould have been based upward, beause we dropped the July release stratum to mnmze the number of ells wth E[m j ]<5. When release strata are dropped, estmates of â (and stratum populaton estmates) an be based upwards tryng to aount for the untagged reovered fsh (Shwarz and Taylor 998). Although our CPUE data ndated low relatve abundanes of oho salmon mgratng aross the OTF transet durng the July release stratum (Table 5), our rado tag data ndated that these fsh mgrated relatvely slowly through the nlet and thus lkely ontrbuted to ommeral harvests n later reovery strata (Table 25). The populaton estmate for the July 4+2 reovery stratum (Tables 0 & ) may have been most affeted by ths bas, beause fsh from the frst release stratum were most lkely to have ontrbuted to the ommeral harvests durng ths tme. Nevertheless, the bas resultng from droppng the frst release stratum was lkely small. Our estmate of the populaton sze of pnk salmon was of questonable value. As wth oho salmon, the estmate may have been based upwards, beause we dropped the July release stratum to mnmze the number of ells wth E[m j ]<5. Sze-dependent tag loss may have also aused an upward bas n our pnk salmon populaton estmate. Although, we dd not fnd that tag loss was sze dependent n our study usng sokeye salmon, many pnk salmon were muh smaller than sokeye salmon. Although, we do not know whether these smaller fsh lost tags at a hgher rate, our observatons on the taggng vessel suggest that ths probably ourred. The preson of our pnk salmon populaton estmate was also substantally redued, beause many proessors dd not pass pnk salmon through ther headng mahnes, and our tag deteton rate was low on the one lne at Ile Seafoods where most of the pnk salmon harvested n UCI were proessed. Fnally, we found the probablty of reapturng PIT tagged salmon n the ommeral gllnet fshery was strongly sze dependent (Table 8). Although, gllnet seletvty aused lower reapture probabltes for small tagged pnk salmon, t also resulted n lower apture probabltes for small untagged pnk salmon. Our pnk salmon populaton estmate was lkely not based sgnfantly by gllnet seletvty, beause we used a relatvely non-seletve gear type to obtan the ntal taggng sample, and the soures of seletvty between the apture and reapture samples were ndependent (Seber 982). We dd not attempt to stratfy our pnk salmon populaton analyss by sze beause of the small number of tags reovered. Our estmate of the populaton sze of hum salmon may be based upward, beause we dd not aount for delayed mortalty, and hum salmon aptured on the ebb tde exhbted a lower tag reapture probablty. Other fators do not appear to have based the estmate. We dd not drop any release or reovery strata, our estmate of tag loss from sokeye salmon was lkely representatve of ths rate n hum salmon, and the G 2 statst ndated a good model ft to the data (Table 5). However, n ths analyss we used an estmate of short-term tag mortalty obtaned from net pen studes. Our studes wth oho salmon and others wth hnook salmon (Candy et al. 996) ndate that delayed tag mortalty probably ours. If so, our hum salmon populaton estmate ould be based upward, but the magntude of the bas, f any, lkely does not exeed that found for oho salmon,.e. about 28% (Tables 0 & ). Fnally, hum salmon aptured on the ebb tde exhbted a lower tag reapture probablty (Table 6) suggestng that fewer of these fsh mgrated nto UCI. It s unlear whether ths was a taggng effet, or f salmon mgratng to areas outsde UCI may have been aptured at a hgher rate on the ebb tde. Burbank (977) desrbed a ylon gyre south of our OTF transet and a northward flowng urrent along the east sde of the nlet n sprng and summer. Salmon mgratng to areas south of UCI may orent to freshwater flowng nto the nlet along the east sde (Hasler and Sholz 983). We attempted to san athes of salmon harvested n lower Cook Inlet for PIT tags, but were unable to do so, beause totes of fsh from the entre nlet were mxed together when they were proessed. Nevertheless, prevous studes 9

have ndated that the majorty of hum salmon tagged west of Anhor Pont mgrated north nto Cook Inlet, only 8% mgrated to other areas outsde of the nlet (Tyler and Noerenberg 967). Fnally, we used our populaton estmates for oho, pnk, and hum salmon to evaluate the probable ranges of explotaton rates on these spees n the ommeral fshery and ther esapements n 2002 (Table 27). Ths was done as a frst step toward determnng esapement levels needed to aheve sustaned yelds. Our best PIT tag estmate of the total populaton sze of oho salmon returnng to UCI was 2.52 mllon (95% CI: 2.6-2.87 mllon). Gven a ommeral harvest of 0.25 mllon (Fox and Shelds 2003), the total esapement of oho salmon nto all UCI streams was 2.27 mllon (95% CI:.9-2.62 mllon), and the explotaton rate n the ommeral fshery was 0% (95% CI: 9-%). However, gven the lower range of our rado tag esapement estmate for oho salmon (95% CI: 0.98.96 mllon), the explotaton rate ould have ranged as hgh as 20%. Ths relatvely low explotaton an be explaned by a derease n effort (no. of delveres x hours fshed) n the drft gllnet fshery over the past 20 years (Fgure 5). Prevous nvestgators estmated explotaton rates on hathery-reared oho salmon usng reoveres of oded-wred tagged fsh n the ommeral fshery (Hasbrouk and Hoffman 994, Stratton et al. 996, Cyr et al. 997, 998, 999, 200). Ther estmates have ranged from 6-93% (Appendx 3). We onduted a regresson analyss to test whether these oded-wre tag estmates of explotaton rate were related to effort n the drft gllnet fshery, whh typally harvests over 70% of the oho salmon n the nlet. We omtted the estmate from Waslla Creek n 997, beause the wer was removed due to hgh water before the end of the oho salmon run. Explotaton rate was sgnfantly orrelated (R 2 =0.367, df=20, p=0.003) wth effort n the drft gllnet fshery (Fgure 6). Interestngly, effort n 998 (28,932 boat-hours) was very smlar to that n 2002 (30,504 boat-hours), and explotaton rates estmated usng oded-wre tags n 998 (0.5-0.2) were very smlar to those estmated n our study. Sne our populaton estmate for pnk salmon was of questonable value, we estmated a maxmum explotaton rate on ths spees by smply summng esapements that were atually enumerated: Kena Rver - 2,353,786, Deshka Rver 946,255, Yentna Rver 44,658 (Westerman and Wllette 2003). We used sde-san sonar to roughly estmate the esapement of pnk salmon nto the Kena Rver above rver mle 9. The sonar was operated on the south bank of the rver untl August 29. Sonar ounts of pnk salmon mgratng along the north bank were not onsdered relable due to mllng fsh wthn the sonar beam, so we assumed the passage rate on the north bank was equal to that on the south bank. Cathes of pnk salmon n fsh wheels operated by the ADF&G Sport Fsh Dvson at rver mle 26 through September 26 were used to estmate that the pnk salmon run was 43% omplete by August 29. We appled ths fraton to our sonar ount to estmate the pnk salmon esapement above the sonar ste. A large but unknown number of pnk salmon spawned below our sonar ste. Summng the esapements from these three rvers and gven a ommeral harvest of 0.45 mllon (Fox and Shelds 2003), the maxmum explotaton rate on pnk salmon n the ommeral fshery was about 2%. However, the atual explotaton rate must be muh lower, sne we dd not aount for pnk salmon esapements nto numerous other streams around the nlet. A relatvely low explotaton rate on pnk salmon may be expeted sne the probablty of apture was substantally redued for small pnk salmon that omprsed more than one half of the populaton (Table 7), and fshermen lkely avoded ths spees due to ts very low value. Our PIT tag estmate of the total populaton sze of hum salmon returnng to UCI was 3.88 mllon (95% CI: 3.30-4.47 mllon). Gven a ommeral harvest of 0.24 mllon (Fox and Shelds 2003), the total esapement of hum salmon nto all UCI streams was 3.64 mllon (95% CI: 3.06-4.23 mllon), and the explotaton rate n the ommeral fshery was 6% (95% CI: 5-7%). Tarbox (988) tagged hum salmon n the mddle of UCI n 983 and 52% of these tags were aptured n the ommeral fshery. Sne under reportng of tags by fshermen was lkely, Tarbox (988) estmated that the atual explotaton rate may have been as hgh as 75%, but ths estmate was based on an assumpton regardng hum esapements outsde of the Sustna Rver. Typally, 87% of the ommeral harvest of hum salmon has been taken n 20

the drft gllnet fshery (Fox and Shelds 2003). Sne 983, effort (no. of delveres x hours fshed) n ths fshery has delned by nearly 5-fold (Fgure 5). In 2002, effort was 28% of that n 983. Assumng ondtons n the fshery (other than the amount of effort) were smlar n these 2 years, we alulated an expeted explotaton rate on hum salmon n 2002 by applyng ths rato (28%) to the fraton of reaptures and the explotaton rate Tarbox (988) estmated for the 983 season. The expeted explotaton rate ranged from 4-2%. Although, ths estmate s hgher than the one obtaned n our study, the dfferene s relatvely small onsderng the unertanty n both estmates. Ths analyss supports the noton that the dfferene n explotaton rates estmated n these 2 years was largely due to a 5-fold delne n effort n the fshery. Relatvely low explotaton rates on hum salmon may be expeted sne ommeral gllnets n UCI extend only about 4 m deep n the water olumn. Ultrason trakng studes have shown that hum salmon spend a sgnfant amount of tme deeper n the water olumn durng ther nshore mgraton (Ishda et al. 988). The offshore areas of the nlet are about 25-80 m deep, so hum salmon may be less vulnerable to apture n surfae drft gllnets. Further studes are needed to determne the vertal dstrbuton of hum salmon mgratng though UCI and the dstrbuton of hum salmon esapements around the nlet. We wll be ntatng studes n 2003 to begn nvestgatng vertal and horzontal dstrbutons of salmon mgratng nto the nlet and whether nterannual hanges n ther vertal dstrbuton affet athablty n drft gllnets. Despte unertanty n our salmon populaton estmates, t s reasonable to onlude that explotaton rates on oho, pnk, and hum salmon n the UCI ommeral fshery were substantally below optmal rates n 2002. Our populaton estmates for oho and hum salmon ranged between.23 and 4.23 mllon, and the ommeral fshery harvested about 0.25 mllon of eah spees. Unertanty regardng atual populaton szes wthn ths range resulted n lttle hange n estmated explotaton rates (range 6-20%), beause explotaton rate was an nverse funton of estmated populaton sze (Fgure 7). Gven that optmal explotaton rates typally range from 50-80% (Chapman 986), a severe bas n our populaton estmates for oho and hum salmon would be neessary to approah the optmal range. Our assessment of unertantes n these data ndates that ths level of bas was unlkely. Fnally, the explotaton rate on pnk salmon n the ommeral fshery was ertanly far below the optmal rate n 2002, beause n our alulaton of the maxmum rate, we only aounted for pnk salmon atually enumerated n 3 streams, whle ths spees was known to esape nto numerous other streams around the nlet. 2

LITERATURE CITED Arnason, A.N. and K.H. Mlls. 98. Bas and loss of preson due to tag loss n Jolly-Seber estmates for mark-reapture experments. Can. J. Fsh. Aquat. S. 38: 077-095. Arnason, A.N., C.J. Shwarz, and J.M. Gerrard. 99. Estmatng losed populaton sze and number of marked anmals from sghtng data. J. Wldl. Mmgt. 55: 76-730. Arnason, A.N., C.W. Krby, C.J. Shwarz, and J.R. Irvne. 996. Computer analyss of Markng data from stratfed populatons for estmaton of salmond esapements and the sze of other populatons. Can. Teh. Rep. Fsh. Aquat. S. No. 206. Banneheka, S.G., R.D. Routledge, and C.J. Shwarz. 997. Stratfed two-sample tag reovery ensus of losed populatons. Bometrs 53: 22-224. Bevan, D.E. 962. Estmaton by taggng of the sze of mgratng salmon populatons n oastal waters. In The researh program of the Fsheres Researh Insttute n Brstol Bay, 945-958. W.F. Thompson (ed.), Unversty of Washngton, College of Fsheres, Contrbuton no. 05, pp 373-449. Brannas, E. and fve oauthors. 994. Use of the passve ntegrated transponder (PIT) n a fsh ndentfaton and montorng system for fsh behavoral studes. Trans. Amer. Fsh. So. 23: 395-40. Burbank, D.C. 977. Crulaton studes n Kahemak Bay and lower Cook Inlet. In Envronmental Studes of Kahemak Bay and Lower Cook Inlet, vol. III. L.L. Traskey et al. (eds.), Marne Coastal Habtat Management Report, Alaska Dept. Fsh and Game, Anhorage, AK. Candy, J.R., E.W. Carter, T. Qunn, and B.E. Rddell. 996. Adult hnook salmon behavor and survval after ath and release from purse-sene vessels n Johnstone Strat, Brtsh Columba. N. Amer. J. Fsh.Mgmnt. 6: 52-529. Chapman, D.W. 986. Salmon and steelhead abundane n the Columba Rver n the nneteenth entury. Trans. Am. Fsh. So. 5: 662-670. Clarke, W.C., and T. Hrano. 995. Osmoregulaton. In Physologal Eology of Paf Salmon, C. Groot, L. Margols, and W.C. Clarke (eds.). Unversty of Brtsh Columba Press, Vanouver, BC. Conover, W.J. 999. Pratal nonparametr statsts. John Wley & Sons, In. New York. Cyr, P., B. Stratton, and J.J. Hasbrouk. 997. Estmates of ommeral harvest and esapement of oho salmon stoked n Northern Cook Inlet streams, 995. Alaska Department of Fsh and Game, Fshery Data Seres No. 97-3, Anhorage. Cyr, P., B. Stratton, and J.J. Hasbrouk. 998. Estmates of ommeral harvest and esapement of oho salmon stoked n Northern Cook Inlet streams, 996. Alaska Department of Fsh and Game, Fshery Data Seres No. 98-5, Anhorage. Cyr, P., B. Stratton, and J.J. Hasbrouk. 999. Estmates of ommeral harvest and esapement of oho salmon stoked n Northern Cook Inlet streams, 997. Alaska Department of Fsh and Game, Fshery Data Seres No. 99-7, Anhorage. 22

Cyr, P., B. Stratton, and J.J. Hasbrouk. 200. Estmates of ommeral harvest and esapement of oho salmon stoked n Northern Cook Inlet streams, 998. Alaska Department of Fsh and Game, Fshery Data Seres No. 0-6, Anhorage. Eames, M., T. Qunn., K. Rednger, and D. Harng. 98. Northern Puget Sound 976 adult oho and hum taggng studes. Wash. Dept. Fsh. Teh. Rpt. 64, 27p. Eames, M., T. Qunn., and M. Hno. 983. Northern Puget Sound 977 adult oho and hum taggng studes. Wash. Dept. Fsh. Teh. Rpt. 75, 239p. Farrell, A.P., P. Gallaugher, C. Clarke, N. DeLury, H. Kreberg, W. Parkhouse, and R. Routledge. 2000. Physologal status of oho salmon (Onorhynhus ksuth) aptured n ommeral nonretenton fsheres. Can. J. Fsh. Aquat. S. 57: 668-678. Fox, J., and P. Shelds. 2003. Upper Cook Inlet ommeral fsheres annual management report, 2002. Alaska Department of Fsh and Game, Regonal Informaton Report no. 2A03-4. Goodman, L.A. 960. On the exat varane of produts. J. Amer. Stat. Asso. 55: 708-73. Hasler, A.D., and A.T. Sholz. 983. Olfatory mprntng and homng n salmon. Investgatons nto the homng and mprntng proess. Sprnger-Verlag, New York.. Hderbrand, R. H., and J.L. Kershner. 2000. Movement patterns of stream-resdent utthroat trout n Beaver Creek, Idaho Utah. Trans. A. Fsh. So. 29:60 70. Hoffman, A.G. and J.J. Hasbrouk. 994. Estmates of ommeral harvest and esapement of oho salmon stoked n Northern Cook Inlet streams, 993. Alaska Department of Fsh and Game, Fshery Data Seres No. 94-45, Anhorage. Ishda, K., S. Nagahora, Y. Inoue, and T. Watanabe. 988. Behavor of adult hum salmon (Onorhynhus keta) homng to the oast of Sanrku. Nppon Susan Gakkash 54(8):279-287. Parker, R.R., and E.C. Blak. 959. Musular fatgue and mortalty n troll-aught hnook salmon (Onorhynhus tshawytsha). J. Fsh. Res. Bd. Can. 6: 95-06. Parker, R.R., E.C. Blak, and P.A. Larkn. 959. Fatgue and mortalty n troll-aught Paf salmon (Onorhynhus). J. Fsh. Res. Bd. Can. 6: 429-448. Penglly D., and L.J. Watson 994. Automated deteton of red kng rab at rab proessng faltes. Fshery Researh. 9: 293-300. Plante, N. 990. Estmaton de la talle d une populaton anmale a l ade d un modele de apturereapture ave stratfaton. M. S. thess, Unverste Laval, Quebe, Que. Plante, N., and L.-P. Rvest. 995. Stratfed apture-reapture estmaton of the sze of a losed populaton. Teh. Rep. 95-03, Departement de mathematques and de statstques, Unverste Laval, Quebe, Que. 23

Prente, E.F. 986. A new nternal telemetry tag for fsh and rustaeans. In Marne farmng and enhanement: Proeedngs of the 5 th U.S.-Japan meetng on aquaulture, Kyoto Japan, NOAA Teh. Rep. NMFS 85: -9. Prente, E.F., 990. A new nternal telemetry tag for fsh and rustaeans. In Marne farmng and enhanement; proeedngs of the ffteenth U. S.-Japan meetng on aquauture, Kyoto, Japan, Otober 22-23. 986. Sparks, A.K. edtor. NOAA tehnal report NMFS 85, Marh 990. Prente, E.P., C.W. Sms, and D.L. Park. 985. A study to determne the bologal feasblty of a new fsh taggng system. Report to Bonnevlle Power Admnstraton (Projet 83-9), Portland, Oregon. Prente, E.F., T. A. Flagg, and C.S. MCutheon. 990a. Feasblty of usng passve ntgegrated transponder (PIT) tags n salmonds. Am. Fsh. So. Symp. 7: 37-322.; Prente, E.F., T. A. Flagg, and C.S. MCutheon. 990b. PIT-tag montorng systems for hydroeletr dams and fsh hatheres. Am. Fsh. So. Symp. 7: 323-334. Prente, E.F., T. A. Flagg, C.S. MCutheon, D.F. Brastow, and D.C. Cross. 990. Equpment, methods, and an automated data-entry staton for PIT taggng. Am. Fsh. So. Symp. 7: 335-340. Pre, A., and C. Powell. 2000. Clove ol as an anesthet for nvasve feld proedures on adult ranbow trout. N. Amer. J. Fsh. Mgmt. 20: 029-032. Qunn, T.P., and N.P. Peterson. 996. The nfluene of habtat omplexty and fsh sze on overwnter survval and growth of ndvdually marked juvenle oho salmon (Onorhynhus ksuth) n Bg Beef Creek, Washngton. Can. J. Fsh. Aquat. S. 53: 555-564. Shwarz, C.J., and C.G. Taylor. 998. Use of the stratfed-petersen estmator n fsheres management: estmatng the number of pnk salmon (Onorhynhus gorbusha) spawners n the Fraser Rver. Can. J. Fsh. Aquat. S. 55: 28-296. Seber, G.A.F. 962. The mult-sample sngle reapture ensus. Bometrka 49(3-4): 339-350. Seber, G.A.F. 982. The estmaton of anmal abundane and related parameters. Charles Grffn and Co. Ltd., London. Shelds, P.A. 2003. An estmate of the mgratory tmng and abundane of sokeye salmon nto upper Cook Inlet, Alaska. Alaska Dept. of Fsh and Game, Regonal Informaton Report no. 2A03-0, 46p. Skalsk, J.R., S.G. Smth, R.N. Iwamoto, J. GT. Wllams, and A. Hoffmann. 998. Use of passve ntegrated transponder tags to estmate survval of mgrant juvenle salmonds n the Snake and Columba rvers. Can. J. Fsh. Aquat. S. 55: 484-493. Stratton, B., J.J. Hasbrouk, and P. Cyr. 996. Estmates of ommeral harvest and esapement of oho salmon stoked n Northern Cook Inlet streams, 994. Alaska Department of Fsh and Game, Fshery Data Seres No. 96-4, Anhorage. Tarbox, K.E. 988. Mgratory rate and behavor of salmon n Upper Cook Inlet, Alaska, 983-984. Fshery Researh Bullentn 88-05, 56p. 24

Tarbox, K.E., and R.E. Thorne. 996. Assessment of adult salmon n near-surfae waters of Cook Inlet, Alaska. J. Mar. S. 53: 397-40 Thompson, S.H. 930. Salmon taggng experments n Alaska, 929. U.S. Bureau of Fsheres Do. 084. Tyler, R.W., and W.H. Noerenberg. 967. Salmon taggng n Cook Inlet. U.S. Fsh and Wldlfe Serve, Bureau of Commeral Fsheres, Bologal Laboratory, Auke Bay, Alaska, Manusrpt Report no. 37, 20p. Westerman, D.L., and T.M. Wllette. 2003. Upper Cook Inlet salmon esapement studes, 2002. Alaska Dept. of Fsh and Game, Regonal Informaton Report no. 2A02-7. Zar, J.H. 984. Bostatstal Analyss. Prente-Hall, In., Englewood Clffs, NJ. 25

ALASKA Cook Inlet N Sustna Rver Ivan Rver Beluga Rver Theodore Rver Chutna Rver Tyonek Creek Nkola Creek MArthur Rver Kustatan Rver Swanson Rver Bg Rver Montana Bll Creek Drft Rver Harret Creek Kena Yentna Rver 22 46 5 2 Bshop Creek Rabbt Cottonwood Slough Creek Matanuska Bg Lake 5 Rver Lttle Knk Sustna 2 Rver Fsh Creek Rver Peters Creek 5 Chkaloon Rver Eagle Rver Shp Creek Campbell Creek Rabbt Creek Twentymle Rver Sxmle Creek Resurreton Creek Hghway System Cresent Rver Redoubt Cr. Kalgn Island Kena Rver Sklak Lake Kaslof Rver Kena Lake Chntna Bay Tuxedn Bay 8 7 6.5 6 5 4 Tustumena Lake Nnlhk Rver Deep Creek Starsk Creek Anhor Rver Seward Anhor Pont Homer Offshore Test Fshery Transet 50 klometers Gulf of Alaska Fgure. Loatons of rado tagged oho salmon found durng surveys of the lower portons of Upper Cook Inlet streams n 200. These fsh were tagged along the offshore test fshery transet west of Anhor Pont before July 20, 200. Numbers n boxes ndate perent of total reoveres (n freshwater) ourrng n eah stream. Numbers along test fshery transet ndate statons. 26

ALASKA Cook Inlet N Sustna Rver Ivan Rver Beluga Rver Theodore Rver Chutna Rver Tyonek Creek Nkola Creek MArthur Rver Kustatan Rver Bg Rver Montana Bll Creek Drft Rver < 2 Kena Yentna Rver < 29 3 Bg Lake Lttle Sustna Fsh Creek Rver 8 8 Bshop Creek Chkaloon Rver Matanuska Rabbt Rver Cottonwood Slough Creek 2 2 Rabbt Creek < Knk Rver Peters Creek Eagle Rver Shp Creek Campbell Creek Sxmle Swanson Rver Creek 2 9 Twentymle Rver Resurreton Creek Hghway System Cresent Rver Harret Creek Redoubt Cr. Kalgn Island Kena Rver Sklak Lake Kaslof Rver Kena Lake Tuxedn Bay Tustumena Lake Nnlhk Rver Seward Deep Creek Starsk Creek Chntna Bay Anhor Rver Anhor Pont Homer Offshore Test Fshery Transet 50 klometers Gulf of Alaska Fgure 2. Loatons of rado tagged oho salmon found durng surveys of the lower portons of Upper Cook Inlet streams n 2002. These fsh were tagged along the offshore test fshery transet west of Anhor Pont before July 20, 2002. Numbers n boxes ndate perent of total reoveres (n freshwater) ourrng n eah stream. 27