Ranbow trout survval and capture probabltes n the upper Rangtke Rver, New Zealand Rchard J Barker Department of Mathematcs and Statstcs Unversty of Otago P.O. Box 56 Dunedn, New Zealand Peter H Taylor and Stephen Smth 1 Wellngton Fsh and Game Councl P.O. Box 1325 Palmerston North, New Zealand December 13, 1999 Summary Because of concerns about the effect of ncreasng anglng pressure, survval and capture probabltes of ranbow trout Onchorhyncus mykss tagged n the headwaters of the Rangtke Rver between March 1993 and Aprl 1999 were estmated usng smultaneous analyss of recapture data obtaned durng taggng operatons and tag return data reported by anglers. In October 1995 an upper sze lmt was mposed on the fshery, wth all fsh over 550mm n length requred to be released alve by anglers. We found strong evdence that the upper sze lmt ncreased the release rate of fsh 550mm but no evdence that the release rate changed durng the study for fsh <550mm. The probablty a fsh was caught by an angler durng a 6-month perod was hgher n summer than wnter but otherwse there appeared to be lttle varaton through tme. Nether angler capture or release probabltes appeared to depend on sex of fsh. Survval probabltes ncreased wth ncreasng length of fsh, and ths relatonshp strengthened followng the ntroducton of the upper sze lmt. Our study provdes evdence that the upper sze lmt regulaton was effectve n changng both the behavor of the anglers and the populaton dynamcs of the fsh. Ths s renforced by lengthfrequency data showng an ncrease n the average length of fsh followng ntroducton of the sze lmt regulaton. Introducton The Rangtke Rver (Fgure 1) rses n the Kamanawa Ranges, central North Island, New Zealand 22 km South East of Turang and flows 241 km to reach the Tasman Sea 40 km south east of Wanganu. The ranbow trout (Onchorhyncus mykss) fshery n the headwaters of the Rangtke Rver provdes anglers wth the opportunty to catch large ranbow trout n a wlderness settng. A sample of 149 angler caught trout averaged 604mm (se = 14.2) P.H. Taylor, unpubl. data), and ths large sze s beleved to reflect the presence of relatvely large numbers of old fsh, rather than fast growth rates. Jellyman and Graynoth (1994) reported the mean age of 15 Rangtke headwater angler caught ranbow trout to be 6.9 years (se = 2.1) and 7.6 (se = 1.9) years for females and males respectvely. 1 Present address: Eastern Fsh and Game Councl, Prvate Bag 3010, Ngongotaha, Rotorua, New Zealand
Access to the headwaters s dffcult wth no formed road and s manly reached by hkng, helcopter, or four-wheeled drve vehcle. Recently anglers have expressed concern about an apparent declne n the number of large fsh present n the headwaters beleved to result from an ncrease n anglng pressure. There s lttle nformaton on changes n the fsh populaton over tme and no relable nformaton on changes n anglng use over tme. However anecdotal evdence from fshng gudes and helcopter operators suggests that anglng pressure has ncreased several-fold n the past 5 years. The Wellngton Fsh and Game Councl responded to these concerns by mposng a one fsh bag lmt n 1992. Angler harvest was further restrcted n 1995 by the mposton of a maxmum sze lmt of 550mm. In ths paper we report results from a study usng recently developed mark-recapture analyss whch allow smultaneous analyss of recapture data obtaned by several methods (Barker 1997). Results from ths study are used to obtan nsght nto the possble effects of anglers on the populaton dynamcs of the Rangtke Rver headwaters ranbow trout fshery and the effectveness of the sze lmt regulaton ntroduced n 1995. We thank Ross Baley, Jm Greeks, Jack McKenze, Glenn McLean, Steve Plkngton, and Les Scown for assstng durng taggng operatons. The New Zealand Army allowed access across Army Tranng Group land. Helpful comments on earler drafts of the manuscrpt were provded by John Hayes, Maurce Rodway and Megan Ogle- Mannng. Methods Trout were tagged n a 30km secton of rver located 15km above Sprngvale between March 1993 and Aprl 1999. Each taggng operaton occurred over a two-week nterval never startng before 15 March and never fnshng after 15 Aprl for the autumn taggng, and never startng before 7 October and never fnshng after 31 October n sprng. Fsh were caught usng natural bat fshed upstream wth spnnng gear, fshed at a dead drft by Wellngton Fsh and Game staff and volunteer expert anglers. Almost all trout were hooked n the mouth by ths method, and only these fsh were tagged and returned to the water. To mtgate aganst the effects of possble tag loss, each fsh was tagged wth two tags placed ether sde of the dorsal fn, approxmately 10mm below the lne of the back usng ndvdually numbered and addressed yellow plastc T-bar tags. To compare fsh sze wth hstorcal data, and for use as an ndvdual covarate n analyses, the length of each fsh was recorded each tme t was captured. Data on the average length of trout between 1971 and 1992 were obtaned from records of the Wellngton Fsh and Game Councl and were based on a combnaton of angler ntervews and fled samples obtaned durng an ntensve study of the rver between 00 and 00 (Rodway 1985). Followng ntal capture and release, data from re-encounters of marked fsh were obtaned n two ways. Frst, lve-recapture data obtaned by Fsh and Game staff durng the sprng and autumn taggng operatons. Second, tag-returns obtaned from anglers fshng the Rangtke Rver whch has an open trout season throughout ts length between 1 October and 20 Aprl each year. Tag-returns reported by anglers were both of fsh that had been released by the angler and fsh that had been klled.
Inttal screenng of mark recapture models was carred out usng program Mark (Whte and Burnham 1999) wth model selecton based on a small sample verson of Akake s Informaton Crteron (AICc) for choosng between models (Burnham and Anderson 1998). We also adopted the model selecton strategy recommended by Burnham and Anderson (1998) for selectng the best approxmatng model from a set of canddate models n models. In ths approach, models wthn 2 AICc unts ( AICc 2) of the model mnmsng AICc are consdered to have substantal support and should be used for makng nferences (Burnham and Anderson 1998). Models havng AICc of about 4 to 7 have consderably less support, and models wth AICc > 10 have essentally no support. For each model we also computed Akake weghts, normalzed to sum to 1, whch provde the relatve weght of evdence n favor of a partcular model beng the best model, gven that one of the complete set s best (Burnham and Anderson 1998). The model was ftted to data from lve recapture of tagged fsh durng sprng and autumn taggng operatons and to reports of tagged fsh caught by anglers ncludng those released alve followng angler capture. Analyss was based on the jont lve recapture/tag recovery/lve resghtng model developed by Barker (1997) and extended by program MARK to allow for ndvdual effects on parameters. Fsh length was ncorporated n the analyss by modelng the logt of model parameters as a lnear functon of length and other effects, where the logt of the parameter θ s defned to be θ logt ( θ) = ln. 1 θ Unlke a conventonal Jolly-Seber model based on an assumpton that any emgraton s permanent (Pollock et al. 1990) the model we used n our analyss allows three dstnct forms of movement of fsh n and out of the study populaton: 1. Random temporary emgraton where the probablty that a brd s at rsk of capture at tme s the same regardless of whether or not t was at rsk of capture at tme 1. 2. Permanent emgraton where a brd whch leaves the study area between tme 1 and can never return 3. Markov emgraton where the probablty that a brd s at rsk of capture at tme depends on whether or not t was at rsk of capture a tme 1. The relaxaton of restrctve assumptons about fsh movement means that estmates of survval probabltes wll be vald n a wder range of bologcal condtons. There was evdence from ths ntal screenng that parameters n the mark-recapture model depended on the length of the fsh, however the length of the fsh can only be measured when the fsh has been captured or recaptured durng taggng operatons. Because tme-varyng covarates cannot presently be ncluded n mark-recapture models usng exstng software (Nchols and Kendall 1995) we ftted a model n whch parameters dffered for fsh n ther frst 6-month perod followng ntal taggng compared to later perods. Usng ths approach, parameters for fsh n the frst 6-months followng taggng can be correctly modeled usng length at frst capture as a covarate. We also used length at frst capture to model parameters for fsh n subsequent perods, however ths approach s only approxmate as true length n these subsequent perods s greater than or equal to length at frst capture.
Due to the apparent nfluence of fsh length on parameters of the mark recapture model, the effect of length on the probablty of angler capture (referred to as tag recovery) and the probablty that a fsh s released gven capture was assessed drectly. For each fsh captured at, we assgned the bnary response varable 0 f the fsh was not captured by an angler n,+1 and 1 f t was captured. Smlarly, f the fsh was caught we can assgn a second bnary response varable that takes the value 0 f the fsh was released and 1 f t was klled. These data were analyzed usng logstc regresson to determne the effect of length, season and tme (tag recovery) and length and tme for the probablty of release gven capture. Because the regulatons requred that all fsh over 550mm be returned, the length covarate was recoded so that t was 0 f the fsh was smaller than 550mm and 1 otherwse n the second analyss. Results There were 575 ndvdual ranbow trout tagged durng the study, 339 females and 236 males. Release, recapture and angler recovery data are summarzed n Table #summary.statstcs usng the defntons of statstcs and parameters n Table #defntons. Of 653 fsh released (ncludng recaptures) 246 were encountered agan, ether through recapture durng taggng operatons or by anglers durng the open fshng season. The average sze of ranbow trout caught durng the study declned between Autumn 1992 and Sprng 1995 and then ncreased (Fgure #average.length). Over the longer perod 1971-1999, eght of the 10 samples that were below the long-term average of 586mm occurred durng the mark-recapture study perod. An excepton to the trend for ncreasng average length snce sprng 1995 was the result for autumn 1998 when the average length of 514mm was the lowest recorded over the perod 1971-1999. However, the autumn of 1998 sample followed perod of record low flows n the Rangtke Rver. Of 92 fsh caught n ths sample, 18 were under 400mm, but of the 534 fsh measured over the perod March 1992 to Aprl 1999, excludng March 1998, only one other fsh under 400mm was recorded. Drect recovery probabltes Reported angler recaptures wthn the frst 6-months followng release for fsh released durng sprng taggng operatons ranged between 4/42 (11.1%) n October 1995 and 15/69 (21.7%) n October 1994 (Table 2). The best fttng model for drect recovery of tagged fsh had an Akake weght of 0.999). In ths model, the probablty that a fsh captured durng taggng operatons was caught by an angler at least once n the followng 6-month perod dffers between summer and wnter but s otherwse constant through tme. No other model was wthn 14 AICc unts of ths best model therefore there was no evdence that drect recovery probabltes depended on sex or fsh length (Table #drect recovery model selecton). The estmated drect recovery rates of tags from ths model were 0.024 (95% CI: 0.011, 0.052) followng autumn taggng and 0.178 (95% CI: 0.140, 0.223) followng sprng taggng. Effect of fsh length on probablty a fsh caught by an angler s released The three best fttng models, wth 0 AICc 4.975 and a combned Akake weght of 0.947, were ones n whch the probablty a fsh caught by an angler was released depended on ts length and whether or not the upper sze lmt regulaton was n place (Table #nu model selecton). In the best of these three models (Akake weght = 0.752) the release probablty was the same for all fsh pror to the ntroducton of the
regulaton ( ν ˆ = 0.628, se = 0.074), was unchanged for fsh <550mm after the ntroducton of the regulaton, but ncreased to 1.0 for fsh 550mm followng the ntroducton of the regulaton. Because the estmated release probablty of 1.0 s at the boundary of the parameter space, the asymptotc standard error s not vald. However, an approxmate profle lkelhood confdence nterval (McCullagh and Nelder 1989) s gven by (0.825,1.0). Mark recapture model selecton The three best fttng models had a combned Akake weght of 0.988 and 0 AICc 4.220. No other model had AICc < 10. Each of the best three models had no sexeffect, Markovan emgraton wth movement probabltes constant through tme, and (1) R dependng only on season. Also, n each of these models the logt of (0) (1) (0) parameters r, r and R were modeled usng addtve effects of season, sze-lmt and length. In the best model (Akake weght = 0.645) the logt of the parameter R was modeled usng addtve effects of just season and sze-lmt but n the (1) (1) next best model (Akake weght = 0.265) R was modeled wth an addtve effect of length n addton to season and sze-lmt. In both these models, havng a combned (0) Akake weght of 0.910, the logt of S was modeled as a lnear functon of length that was dfferent before and after the ntroducton of the upper sze lmt regulaton. The thrd best model (Akake weght = 0.078) was dentcal n all respects to the second best model, but the ntercept of the the straght-lne relatonshp between length and the logt (0) of S was dfferent before and after the ntroducton of the sze lmt but the slope was the same. Overall, our analyss provded strong evdence that fsh survval was nfluenced by the length of fsh and that ths relatonshp changed followng the ntroducton of the sze lmt regulaton. Parameter estmates averaged across the best three models by modelaveragng (Burnham and Anderson 1998) ndcated that both before and after the ntroducton of the sze lmt, survval rate tended to ncrease wth sze of fsh. Also, followng the ntroducton of the sze lmt survval rate tended to be hgher and the dependency between survval and sze of fsh even stronger (Table #logt survval estmates and Fgure #logt survval v length). Dscusson Data on length of trout between 1971 and 1999 ndcates that average sze of fsh decreased durng the early 1990 s, but then ncreased followng the ntroducton of the upper sze lmt at the start of the 1995 anglng season. Wthout a manpulatve experment t s mpossble to establsh a drect causal lnk between the changes n fsh sze and the mposton of the upper sze lmt unequvocally, however the results from our analyss provdes crcumstantal evdence that there s such a lnk. A surprsng aspect of ths study was the relatvely hgh estmates for the probablty that a fsh released n sprng was reported caught at least once n the followng summer. In the summers of 1993/94 and 1994/95 the estmates exceeded 20%, and a 95% confdence nterval based on all the drect recovery data was between 14% and 22%. Because of non-reportng of tags by anglers, the estmates provde lower bounds on an
estmate of the probablty that a fsh was caught by an angler at least once durng the 6- month perod. It s not known what the rate of tag-reportng was durng ths study. Evdence from waterfowl bandng studes ndcates that for gamebrds non-reportng of bands n North Amerca exceeds 50% (Nchols et al. 1995). Anecdotal evdence durng ths study provded by angler ntervews suggest that non-reportng of tagged Rangtke Rver ranbow trout occurred. Thus, the true rate of capture by anglers s lkely to be well n excess of 20%, at least n some years, ndcatng potental for anglng pressure to have an effect on fsh populaton dynamcs. Modelng of releases of angler-caught trout provded evdence that the upper sze lmt was effectve n modfyng angler behavor. Therefore, an upper sze lmt s a potentally useful tool for management of the Rangtke Rver trout fshery. Our estmates of the probablty that an angler-caught fsh was returned alve ndcated that pror to the ntroducton of the upper sze lmt about 37% of fsh caught by anglers durng a 6-month perod were klled. Followng the ntroducton of the upper sze lmt the probablty appeared to reman at about 37% for fsh under the 550mm upper sze lmt, but almost all fsh over the upper sze lmt were reported as released alve. Therefore, the true probablty that a fsh alve at the start of sprng s caught and klled durng the 6-month perod n the absence of an upper sze lmt s lkely greater than about 5%-10%, dependng on the rate of non-reportng of tags. The relatvely hgh angler capture rate suggests that anglers have the potental to exert consderable pressure on the Rangtke Rver headwaters trout fshery. For such pressure to result n a change n fsh populaton dynamcs, angler-nduced mortalty must add to the natural mortalty occurrng n the fsh populaton. Evdence from our study that survval rates of larger fsh were hgher followng the ntroducton of the upper sze lmt suggests that anglng mortalty adds to natural mortalty. In fact, our estmates (Fgure 2) provde evdence that anglng-nduced mortalty s the major source of mortalty for larger fsh. An nterestng aspect of ths study s that survval probabltes appear to ncrease wth fsh sze even for fsh over 450mm n length. It s well known that at least n frst 3 years, survval probablty of trout ncreases wth age (Horokw study etc). However our study provdes evdence that ths effect perssts throughout the lfe of the fsh. (I need a bt more nformaton here ths s an nterestng aspect of the study) Another mechansm by whch angler harvest may nfluence the fshery s through removal of larger fsh leadng to a shft n age structure favorng younger and smaller fsh. The appeal of the Rangtke Rver headwaters ranbow trout fshery s for large, trophy-szed trout. These trout are the oldest members of the populaton and the most dffcult to replace. The apparent effectveness of the upper sze lmt n ncreasng the rate at whch anglers released larger suggests that the upper sze lmt contrbuted to the ncrease n average sze of fsh noted followng ts ntroducton. Despte the fact that evdence s crcumstantal t certanly provdes an after-the-fact justfcaton for the acton taken by the Wellngton Fsh and Game Councl n mposng lmts on the removal of fsh longer than 550mm mposed at the start of the 1995 fshng season.
The use of taggng studes to estmate fsh survval rate and anglng capture probabltes, provdes a useful mechansm for montorng angler pressure. An appealng aspect of ths approach s that unlke tradtonal measures of anglng pressure such as the number of vsts by anglers to a rver, data obtaned from the reportng of tags can be used to derve measures of anglng pressure from the pont of vew of the fsh. One dffculty however s non-reportng of tags. Where possble taggng studes should also be combned wth studes that measure non-reportng of tags. Currently two approaches are possble. Frstly, the use of reward tags as n the study of Nchols et al. (1995) and secondly the use of nformaton obtaned from streamsde ntervews of anglers (Green et al. 1983, Pollock et al. 1991). References Barker, R.J. 1997. Jont modelng of lve-recapture, tag-resght, and tag-recovery data. Bometrcs 53: 147-158. Barker, R.J. In press. Jont analyss of mark recapture, resghtng and rng-recovery data wth age-dependence and markng-effect. Brd Study. Burnham, K. P., and Anderson, D. R. 1998. Model Selecton and Inference: A Practcal Informaton Theoretc Approach. New York: Sprnger-Verlag. Burnham, K.P., and Anderson, D.R., Whte, G.C., Browne, C., and Pollock, K.H. 1987. Desgn and analyss of fsh survval experments based on release-recapture. Amercan Fsheres Socety Monograph 5. Burnham, K.P., Whte, G.C. and Anderson, D.R. 1995. Model selecton strategy n the analyss of capture-recapture data. Bometrcs 51: 888-898. Jellyman, D.J. and Graynoth, E. 1994. Headwater trout fsheres n New Zealand. N.Z. Freshwater Research Report No. 12. 87p. Green, A. W., Matlock, G. C., and Weaver, J. E. 1983. A method for drectly estmatng the tag-reportng rate of anglers. Transactons of the Amercan Fsheres Socety 112:412-415. McCullagh, P. and Nelder, J. A. 1989. Generalzed Lnear Models. Chapman and Hall, London 511pp. Nchols, J. D., and Kendall, W. L. 1995. The use of mult-state capture-recapture models to address questons n evolutonary ecology. Journal of Appled Statstcs 22: 835-846. Nchols, J.D., Reynolds, R.E., Blohm, R.J., Trost, R.E., Hnes, J.E., and Bladen, J.P. 1995. Geographc varaton n band reportng rates for mallards based on reward bandng. J. Wldl. Manage. 59: 697-708. Pollock, K.H., Hoeng, J.M., and Jones, C.M. 1991. Estmaton of fshng and natural mortalty when a taggng study s combned wth a creel survey or port samplng. Am. Fsh. Soc. Symp. 12:423-434.
Rogers, J.P., Nchols, J.D., Martn, F.W., Kmball, C.F., and Pospahala, R. S. 1979. An examnaton of harvest and survval rates of ducks n relaton to huntng. Trans. North Am. Wldl. Nat. Resour. Conf. 44: 114-126. Seber, G.A. F. 1982. The estmaton of anmal abundance and related parameters. Grffn, London. 654 pp. Whte, G. C., and Burnham, K. P. 1999. Program MARK: survval estmaton from populatons of marked anmals. Brd Study
Table #summary.statstcs: Summary statstcs for results from taggng ranbow trout n the upper Rangtke Rver between March 1993 and Aprl 1999. Sample Year Month R,c r,c R,r r,r m o d 1993 Aprl 35 18 1 0 0 2 2 1993 October 66 38 15 8 1 18 14 1994 Aprl 89 38 2 0 6 3 2 1994 October 69 33 27 7 6 43 15 1995 Aprl 15 5 1 0 4 2 0 1995 October 42 22 20 7 5 24 4 1996 Aprl 12 5 1 1 6 2 0 1996 October 79 34 28 11 8 31 10 1997 Aprl 33 12 2 1 4 4 0 1997 October 69 19 44 14 15 47 11 1998 Aprl 92 16 4 0 13 5 2 1998 October 44 6 24 0 14 27 5 1999 Aprl 8 0 3 0 1 4 0
Table #defntons: Statstcs and parameters Statstcs R,c = the number of fsh released at followng capture. R,r = the number of fsh released followng capture by an angler n (, +1). r,c = the total number of fsh released followng capture durng taggng at tme that are subsequently encountered. r,r = the total number of fsh released followng angler capture at tme at are subsequently encountered. m = the total number of marked fsh captured durng taggng at tme. o = the total number of marked fsh captured by an angler at tme. d = The number of fsh released at that were reported caught by an angler n,+1. Parameters (0) S = the probablty that a fsh alve and newly tagged at tme s alve at tme + 1. (1) S = the probablty that a fsh tagged before and alve at tme s alve at p = tme + 1. the probablty that a fsh s captured durng taggng at tme gven that t s alve and at rsk of capture at tme. (0) r = the probablty that a fsh alve and newly tagged at tme, and that des n +1, s caught by an angler and klled between and + 1. (1) r = the probablty that a fsh tagged before and alve at tme, and that des n +1, s caught by an angler and klled between and + 1. (0) R = the probablty that a fsh alve and newly tagged at tme, and alve at tme +1, s caught by an angler and released between and + 1. (1) R = the probablty that a fsh tagged before and alve at tme, and alve at tme +1, s caught by an angler and released between and + 1. (0) R = the probablty that a fsh alve and newly tagged at tme, but that s dead at +1, s caught by an angler and released between and + 1. (1) R = the probablty that a fsh tagged before and alve at tme, but that s dead at +1, s caught by an angler and released between and + 1. v = the probablty a fsh s released gven that t s captured by an angler n, +1.
Table #drect recovery model selecton Model fttng summary for logstc regresson models ftted to drect tag recovery data. Delta Akake No. Model AICc AICc Weght parameters Devance S(season) 360.985 0.000 0.999 2 356.964 S(tme) 375.424 14.440 0.001 13 348.769 S(tme+length) 376.139 15.150 0.001 14 347.382 S(sex*t) 393.863 32.880 0.000 26 339.277 S(tme*length) 395.342 34.360 0.000 24 345.140 S(sex) 398.245 37.260 0.000 2 394.224 S(.) 398.988 38.000 0.000 1 396.981 Table #nu model selecton Model fttng summary for logstc regresson models ftted to angler release data, where ν s the probablty that a fsh captured by an angler n,+1 s released. Akake No. Model AICc AICc weght parameters Devance ν(pre(small)=post(small)=pre(large),post(large)) 60.965 0.000 0.752 2 56.765 ν(reg*sze) 64.426 3.461 0.133 4 55.736 ν(reg+sze) 65.940 4.975 0.062 3 59.533 ν(reg) 67.142 6.177 0.034 2 62.942 ν(season*reg) 69.084 8.119 0.013 4 60.394 ν(tme) 72.067 11.102 0.003 9 50.671 ν (.) 73.464 12.499 0.001 1 71.398 ν (sze) 73.792 12.827 0.001 2 69.592 ν (sex*tme) 84.484 23.519 0.000 16 40.658
Table #logt survval estmates Parameter estmates averaged across the three best fttng models for the relatonshp between fsh length and the logt of 6-month survval for fsh n ther frst 6 months followng taggng, before and after the ntroducton of the upper sze lmt. Parameter Estmate SE Intercept before regulaton change 2.251 0.326 Slope before regulaton change 1.141 1.155 Intercept after regulaton change 5.133 3.043 Slope after regulaton change 3.881 1.985
Fgure 1: The Rangtke Rver, North Island, New Zealand.
Fgure #average.length: Average length (wth 95% confdence nterval bars) of Rangtke Rver ranbow trout captured captured durng taggng operatons between March 1992 and Aprl 1997. Data provded by anglers between 1970 and 1991 are ncluded for reference. The horzontal bar represents the average length recorded between 1971 and 1999. 700 650 Mean sze (mm) 600 550 500 450 70 75 80 85 90 95 Year
Fgure #logt survval v length Estmated 6-month survval probablty of Rangtke Rver ranbow trout n ther frst 6-month perod followng release as a functon of fsh sze. 1 Estmated survval 0.8 0.6 0.4 0.2 No lmt lmt 0 450 500 550 600 650 700 750 Length (mm)