Southwest Fisheries Science Center National Marine Fisheries Service 8604 La Jolla Shores Dr. La Jolla, California 92037

Transcription

1 233 Abstrat We estimated the total number of pantropial spotted dolphin (Stenella attenuata) mothers killed without their alves ( alf defiit ) in all tuna purse-seine sets from and in the eastern tropial Paifi. Estimates were based on a tally of the mothers killed as reported by olor pattern and gender, several olor-pattern-based frequeny tables, and a weaning model. Over the time series, there was a derease in the alf defiit from approximately 2800 for the western-southern stok and 5000 in the northeastern stok to about 60 missing alves per year. The mean defiit per set dereased from approximately 1.5 missing alves per set in the mid-1970s to 0.01 per set in the late-1990s. Over the time series examined, from 75% to 95% of the latating females killed were killed without a alf. Under the assumption that these orphaned alves did not survive without their mothers, this alf defiit represents an approximately 14% inrease in the reported kill of alves, whih is relatively onstant aross the years examined. Beause the alf defiit as we have defined it is based on the kill of mothers, the total number of missing alves that we estimate is potentially an underestimate of the atual number killed. Further researh on the mehanism by whih separation of mother and alf ours is required to obtain better estimates of the unobserved kill of dolphin alves in this fishery. Manusript approved for publiation 7 January 2004 by Sientifi Editor. Manusript reeived 20 January 2004 at NMFS Seintifi Publiations Offie. Fish. Bull. 102: (2004). Annual estimates of the unobserved inidental kill of pantropial spotted dolphin (Stenella attenuata attenuata) alves in the tuna purse-seine fishery of the eastern tropial Paifi Frederik Arher Tim Gerrodette Susan Chivers Alan Jakson Southwest Fisheries Siene Center National Marine Fisheries Servie 8604 La Jolla Shores Dr. La Jolla, California address (for F. Arher): eri.arher@noaa.gov In the eastern tropial Paifi (ETP), yellowfin tuna (Thunnus albaares) are frequently found swimming under shools of pantropial spotted (Stenella attenuata) and spinner (S. longirostris) dolphins. For the past four deades, the ETP yellowfin tuna fishery has made use of this assoiation by hasing the more visible dolphins at the surfae and using purse-seines to enirle the shools arrying the tuna (NRC, 1992). The large byath of dolphins in this fishery has beome widely known as the tuna-dolphin issue (Gerrodette, 2002). During the 1960s, the number of dolphins killed by the fishery was estimated to be 200, ,000 per year (Wade, 1995), and two stoks of spotted and spinner dolphins were redued to frations of their previous sizes (Smith, 1983; Wade et al. 1 ). A long history of tehnologial innovations by fishermen, laws and fishing regulations, dolphin quotas, eo-labeling of dolphin-safe tuna, and a omprehensive international observer program (Gosliner, 1999; Hall et al., 2000; Gerrodette, 2002) has redued the dolphin byath to less than 1% of its former level. The reported byath in reent years is less than 2000 dolphins per year for all speies ombined (IATTC, 2002). Although the reported kill has dramatially dereased, reent studies suggest that there is little evidene that the stoks are growing lose to expeted rates (Wade et al. 1 ). One hypothesis for this lak of reovery has been that there are unobserved kills of dolphins during tuna purse-seine sets. Arher et al. (2001) presented evidene of an under-representation of sukling spotted and spinner dolphin alves in a sample of tuna purse-seine sets in the eastern tropial Paifi. Given that some of these missing alves are still dependent on their mothers for nutrition, it is likely that one separated they would die and this under-representation represents some degree of unobserved kill. In Arher et al. (2001), the sample of sets examined was limited to those sets in whih all of the animals killed had biologial data olleted by tehniians aboard the tuna vessel. Calves still dependent on their mothers in the kill were identified by five intervals of body length, hosen to over a range of 1 Wade, P. R., S. B. Reilly, and T. Gerrodette Assessment of the population dynamis of the northeastern offshore spotted and the eastern spinner dolphin populations through National Oeanographi and Atmospheri Administration Administrative Report LJ-02-13, 58 p. Southwest Fisheries Siene Center, 8604 La Jolla Shores Dr., La Jolla, CA

2 234 Fishery Bulletin 102(2) alf sizes. Beause of this approah, it was not possible to derive a single estimate of the number of missing alves or to extrapolate their estimate to sets not used in this analysis. In the urrent study, we present a different method of estimating the number of missing alves in eah set where offshore spotted dolphins (S. attenuata attenuata) were killed. For brevity, we all the shortage of alves in the kill in relation to the number of latating females in the kill the alf defiit. We examined the western-southern and northeastern offshore stoks separately aording to the geographi boundaries desribed by Dizon et al. (1994). As they age, spotted dolphins hange olor through five olor phases (Perrin, 1970). We used the olor-phase frequeny distribution of the kill in onjuntion with ageand olor-based frequeny distributions from a sample of the kill to estimate the total number of missing alves in eah stok, along with onfidene intervals derived from bootstrap repliations. This method also allowed us to examine the alf defiit from sets in reent years from whih we did not have biologial samples and to examine the time series of available years for evidene of a trend in the alf defiit. Methods Sine 1973, observers have been randomly plaed on tuna purse-seine vessels. For eah spotted dolphin killed during an observed set, observers attempted to reord the sex and the olor phase of the dolphin (neonate, two-tone, spekled, mottled, and fused, see Perrin, 1970). From the National Marine Fisheries Servie (NMFS) set log database, we obtained the number of northeastern and western-southern offshore spotted dolphins (by gender and olor phase) killed in every observed set from 1973 to The Inter- Amerian Tropial Tuna Commission (IATTC) provided the same data from 1996 to Proration In eah set, olor phase or gender (or both) may not have been reorded for some dolphins. Assuming that the distribution of the demographi omposition of this missing data is equivalent to the overall demographi omposition of the kill, we alloated the number of dolphins of unknown olor phase (nu) to unknown gender in eah olor phase (ngu ) aording to the following formula, N ngu = ngu + nu N, (1) where = one of the five olor phases (neonate to fused); N = the total number of dolphins in eah olor phase in the entire data set; and nguʹ = the new number of dolphins in eah olor phase where gender is unknown, inluding the individuals of prorated unknown olor phase The number of male (nmʹ) or female (nf ʹ) dolphins in a olor phase was alulated as Nm nm = nm + ngu Nm + Nf, (2) Nf nf = nf + ngu Nm + Nf, (3) where Nm and Nf are the total number of males and females, respetively, observed in that olor phase in the entire data. Table 1 gives the sample size of sets for both stoks by year, as well as the fration of the kill of unknown gender and olor phase that were prorated as desribed above. Number of sukling alves As time permitted, NMFS observers would also ollet biologial data from a subset of the kill. For this study, we used ages estimated from teeth olleted for a study of spotted dolphin growth and reprodution (Myrik et al., 1986). The speimens used were a random sample of all male and female spotted dolphins olleted between 1973 and 1978 for whih total body length was reorded and teeth were olleted. However, additional speimens with lengths less than 150 m were seleted in order to math as losely as possible the length distribution of the aged sample to the underlying length distribution of the spotted dolphins in the kill. This was neessary beause observers did not generally ollet teeth from smaller, younger animals. Later, another sample of female spotted dolphins was seleted from speimens olleted in Speimens were aged as desribed in Myrik et al. (1986). The final data set used in our analyses inluded age estimates for 1094 female spotted dolphin speimens and 798 male speimens. Of these, 649 females and 457 males belonged to the northeastern stok and had olor phase reorded. These 1106 dolphins were used to generate the age frequeny distribution for eah olor phase (F a, Table 2), F a Sa =, (4) S a where S a = the number of samples of age a in olor phase. The oldest age reorded was 36 years. To derive an age distribution for the dolphins killed in eah tuna set, we estimated the number of dolphins in eah age lass (n a ) as a na = ( Fa n ), (5) where nʹ = the sum of nmʹ and nf ć (the number of males and females in eah olor phase after proration from Equations 2 and 3).

3 Arher et al.: Estimates of the inidental kill of Stenella attenuata attenuata alves in the tuna purse-seine fishery 235 Table 1 Sample sizes of NMFS ( ) and IATTC ( ) observed sets with spotted dolphin kill made on two stoks of pantropial spotted dolphins (Stenella attenuata) by year. Northeastern stok Western-southern stok Fration of Fration of Fration of Fration of kill of kill of kill of kill of Number of Observed unknown unknown Number of Observed unknown unknown Year sets with kill kill olor phase gender sets with kill kill olor phase gender In Equation 4, an age distribution was generated for eah olor phase, and then the number of dolphins in eah age lass was summed aross all olor phases. To estimate the number of alves in eah set, we used this age distribution in onjuntion with a weaning model developed from a study of the stomah ontents and ages of alves (Arher and Robertson, in press). The model predits the probability that an animal of a given age (a) will be sukling: 1 P( milk) a = a 1 e The estimated number of alves (N alf ) in a set is then 3 Nalf = ( na P( milk) a). a= 0 In our estimate of N alf we hose to use only the first four age lasses (0 to 3) beause P(milk) 4 was extremely small ( ). These age lasses allowed us to derease omputational time without signifiantly affeting the estimates. (6) (7) Number of latating females Observers visually examined the mammaries of the 649 females used in the age distribution above (Eq. 4) for the presene of milk as part of the suite of biologial data olleted. Using these data in onjuntion with the olor phase of these females, we alulated the fration of latating females in eah olor phase (Fla ), Fla Sla =, (8) Sfem where Sla and Sfem = the number of females that were latating and the total number of females in olor phase of the samples examined. Fla was 0.00, 0.01, 0.04, 0.22, and 0.50 for neonate, twotone, spekled, mottled, and fused speimens, respetively. The estimated number of latating females (N la ) in a set was then N = ( nf Fla ). (9) la

4 236 Fishery Bulletin 102(2) Calf defiit As desribed in Arher et al. (2001), the alf defiit (D) in eah set was alulated by subtrating the number of alves (N alf ) from the number of latating females (N la ). If this value was zero or less, then D was set to zero to indiate that there were enough alves to aount for all latating females killed (Fig. 1), D = N N if N > N 0 if N N la alf la alf la alf. (10) We alulated three defiit-based frations: 1) the mean defiit per set (D s ); 2) the mean defiit per dolphin killed (D k ); and 3), the mean defiit per latating female killed (D l ): D D =, s (11) ObsSets D D =, k (12) ObsKill D D = l, (13) EstLaKill where ΣD = the total observed alf defiit in eah year; ObsSets = the number of observed sets used in the analysis, inluding those sets without a dolphin kill; ObsKill = the number of dolphins killed in the observed sets; and EstLaKill = the total estimated number of latating females killed. The above analysis was onduted eah year. Estimation error was evaluated with 20,000 bootstrap repliates for eah year. For eah repliate, the sets within that year were randomly resampled. The frequeny tables F a and Fla were also realulated by resampling the list of biologial speimens. The parameters for the weaning model, P(milk) a, were estimated again by resampling the 29 alves and by fitting the logisti model to the new data set as desribed in Arher and Robertson (in press). All resampling was done with replaement. N alf, N la, and D were estimated as desribed above for eah set, and D s, D k, and D l were alulated for the repliate. The 95% onfidene intervals for N alf, N la, D, D s, D k, and D l were estimated from the 2.5% and 97.5% quantiles of the distributions of the bootstrap repliate values. The total alf defiit (D total ) was estimated as the defiit per dolphin killed (D k ) multiplied by the total number of dolphins killed (N killed ) by stok eah year, D total = D k N killed. (14) Table 2 Age-lass frequeny distribution for eah olor phase (F a ). Age Two- (yr) Neonate tone Spekled Mottled Fused For the period , annual values of N killed for eah stok were provided by the IATTC (Joseph 2 ). For and , values were published by IATTC (2002). In the bootstrap estimation of the 95% CI around D total, for the period, eah repliate was randomly sampled from a normal distribution by using the estimated total kill standard error. For , the total kill was reported to be exat; therefore the total kill was used without variane in all repliates. 2 Joseph, J Letter of September 6 to Mihael Tillman, 2 p. Southwest Fisheries Siene Center, 8604 La Jolla Shores Dr., La Jolla, CA

5 Arher et al.: Estimates of the inidental kill of Stenella attenuata attenuata alves in the tuna purse-seine fishery 237 Fration of females latating, by olor ( , 1981): Fla Number of latating females killed: N la Tally of females, by olor: nf Tally of dolphins killed, by olor and sex, from set log ( , ) Calf defiit: D Tally of dolphins, by olor: n Fration of dolphins in age lass, by olor ( , 1981): F a Number of dolphins killed, by age: N a Probability of sukling, by age ( ): P(milk) a Number of sukling alves killed: N alf Figure 1 Diagram of the analytial method used to estimate the spotted dolphin (Stenella attenuata attenuata) alf defiit in eah set as desribed in the text. Boxes identify original data that were bootstrapped to produe onfidene intervals. Values in parentheses are years for whih data were available. In a subset of the sets that we examined, every individual killed had been examined and biologial samples had been olleted from it; therefore, we knew the atual number of latating females killed. There were 1108 of these 100% sampled sets on the northeastern stok, and 697 on the western-southern stok from 1973 to We evaluated the auray of our frequeny-based method by onduting a paired t-test between our estimate of the number of latating females and the number observed in eah of these sets. Stomah-ontent data were not available for every animal in these 100%-sampled sets; therefore, we did not know the atual number of sukling alves. However, we also used paired t-tests to ompare our estimate of the number of sukling alves in eah set with the number of animals smaller than 122 m, whih was the estimated length at whih the probability of milk in the stomah was 0.5, given the weaning model of Arher and Robertson (in press). Likewise, our estimate of the alf defiit was ompared with the defiit as estimated by using a utoff length of 122 m. These tests were done to determine if the method in the present study would produe signifiantly different results from the method used in the previous study. Paired t-tests were onduted for eah year separately, as well as for all years ombined. A power analysis was also performed for these paired t-tests to determine the minimum detetable differene at whih we ould rejet the null hypothesis of no differene between methods given observed sample sizes and variability. Results The alf defiit as a fration of the number of dolphins killed (D k ) inreased slightly during the mid-1970s but remained relatively onstant throughout the rest of the time series at approximately 0.14 missing alves per dolphin killed for both stoks (Fig. 2). The total alf defiit (D total ) as estimated from the annual kill dereased from highs of approximately 5000 in the mid-1970s down to by the early 1980s (Fig. 3). In the late 1980s, this value inreased to approximately 5000 in northeastern spotted dolphins (Table 3A) and approximately 2800 in the western-southern stok (Table 3B), refleting an inrease in the reported kills. In the last five years of the time series ( ), the estimated total defiit was approximately 60 missing alves. The mean defiit per set (D s ) for northeastern spotters over all years was 1.03 missing alves per set, and the median was 0.30 (Fig. 4). For western-southern spotted dolphins, the mean was 1.28 missing alves per set, and the median was The estimated mean defiit per set was approximately 1.5 in the mid-1970s and dereased over time to at the end of the time series (Fig. 4). For both stoks, 75 95% of latating females killed were not killed with their alf (Fig. 5). In the sets that were 100%-sampled, for all years ombined, there was no signifiant differene between the observed and the estimated number of latating females killed in either stok (Table 4). The results of paired t-tests

6 238 Fishery Bulletin 102(2) 0.3 Northeastern Northeastern Defiit per dolphin killed: D k Western-southern Total alf defiit: D total Western-southern Year Figure 2 Calf defiit per spotted dolphin (Stenella attenuata attenuata) killed (D k ) by year. Vertial lines indiate 95% onfidene intervals Figure 3 Year Total estimated alf defiit (D total ) by year. Vertial indiate 95% onfidene intervals. by year indiated that the observed number of latating females in eah set was signifiantly greater (P 0.05) than the estimated number in 1977 for the northeastern and the western-southern stoks and in 1979 for the western-southern stok. The differene was signifiantly less in 1984 for the western-southern stok. Using 0.1 as our type-2 error level, we determined through power analysis that the minimum detetable differene (α=0.05) between the mean observed and estimated number of latating females per set aross all years was approximately 0.08 and 0.09 in the northeastern and western-southern stoks respetively. The observed number of alves per set, defined as the number of dolphins less than 122 m, was signifiantly greater for both stoks, for all years ombined, than the values estimated in this paper (Table 5). The overall mean differene was 0.17 alves per set for the northeastern stok and 0.12 for the western-southern stok. About half of the years showed a signifiant differene for eah stok. In the omparison of the alf defiit by year, only a few years showed signifiant differenes in either stok (Table 5). However, the estimated defiit tended to be larger than the observed defiit. The paired t-test for all years ombined was signifiant for the northeastern stok, although the mean differene was only 0.06 missing alves per set. The minimum detetable differene from the power analysis for the mean number of alves per set and mean alf defiit per set aross all years was 0.06 and 0.08 respetively for both stoks. Disussion In the present study, we present an estimate of the number of missing dependent northeastern and western-southern offshore spotted dolphin alves in the tuna purse-seine kill from 1973 to 1990 and from 1996 to The total number of missing alves dereased through the time series, whih, beause we estimated the alf defiit as a funtion of the size of the kill, was a diret result of the large redution in the annual dolphin kill by the fishery. Between 1973 and 2000, the shortage of alves in the kill remained at a relatively onstant fration of the kill, about 14%, for both stoks of pantropial spotted dolphins (Fig. 2). On the assumption that sukling alves do not survive separation from their mother (Arher et al., 2001;

7 Arher et al.: Estimates of the inidental kill of Stenella attenuata attenuata alves in the tuna purse-seine fishery Northeastern Mean alf defiit per set: D s Western-southern Defiit per latating female killed: D l Northeastern Year 0.6 Western-southern Year Figure 4 Mean alf defiit per set (D s ) by year. Vertial lines indiate 95% onfidene intervals. Figure 5 Calf defiit per latating female killed (D l ) by year. Vertial lines indiate 95% onfidene intervals. Edwards 3 ), the estimated alf defiit represents an approximately 14% underestimate of the reported kill. The alf defiit in the present study was estimated from the number of dependent alves and latating females killed by using age-olor frequeny tables and data on the stomah ontents of weaning alves. Speimens used to derive the age and olor table were olleted from 1973 to 1978 and 1981, and speimens used for the weaning model were olleted between 1989 and If the distributions of these samples were not representative of all years that we examined, then our results may be biased. However, the results of a study to onstrut the annual age distribution of the kill (Arher and Chivers 4 ) indiated that there is no signifiant differene in the age-olor frequeny table aross years. The sample size for the stomah data (29 alves) was too small to examine differenes between years. Our finding of no signifiant differene between our estimates of the number of latating females and the observed tally of latating females in sets where the entire kill was sampled validates this portion of our estimation proedure. However, beause the number of sukling alves present in these 100%-sampled sets was not reorded, we were unable to validate the method used to generate these estimates in a similar manner. The results of our paired t-tests indiated that the observed number of animals smaller than 122 m tended to be greater than the number we estimated. This is most likely a result of the differene between how alves were ounted in eah method. Arher et al. (2001) onsidered all animals under a series of utoff values to be alves that were dependent on sukling for survival. In the present study, the weaning model that we used (Arher and Rob- 3 Edwards, E. F Behavioral ontributions to separation and subsequent mortality of dolphin alves hased by tuna purse-seiners in the eastern tropial Paifi Oean. National Oeanographi and Atmospheri Administration Administrative Report LJ-02-28, 34 p. Southwest Fisheries Siene Center, 8604 La Jolla Shores Dr., La Jolla, CA Arher, F., and S. J. Chivers Age struture of the northeastern spotted dolphin inidental kill by year for 1971 to 1990 and 1996 to National Oeanographi and Atmospheri Administration Administrative Report LJ-02-12, 18 p. Southwest Fisheries Siene Center, 8604 La Jolla Shores Dr., La Jolla, CA

8 240 Fishery Bulletin 102(2) Table 3 Estimated alf defiit per kill (D k ) and total alf defiit. Total number of spotted dolphins killed reported by the IATTC (2002) and Joseph (footnote 2 in the general text). Values in parentheses are 95% lower and upper onfidene intervals. Mean alf Total number Estimated alf defiit of NE spotted Estimated Stok and defiit in Observed per kill dolphins killed total alf year observed sets dolphin kill (D k ) (±SE) defiit A Northeastern (NE) stok ± (464,964) (3947,6820) (0.10,0.16) (3972,9532) ± (583,1027) (4943,6916) (0.10,0.16) (3573,6708) ± (618,1269) (6578,9965) (0.08,0.14) (3297,8254) ± (196,408) (1786,3079) (0.09,0.15) (1284,3903) ± (249,416) (1743,2622) (0.13,0.18) (656,1167) ± (83,209) (684,1431) (0.11,0.16) (336,836) ± (96,226) (680,1629) (0.11,0.17) (428,963) ± (107,239) (724,1637) (0.12,0.18) (622,1300) ± (84,173) (560,1122) (0.12,0.18) (753,1774) ± (155,347) (1126,2395) (0.11,0.17) (833,1840) ± (7,59) (35,410) (0.11,0.23) (169,678) ± (26,57) (191,417) (0.10,0.17) (608,1602) ± (235,508) (1839,3529) (0.11,0.16) (2447,4748) ± (119,478) (859,3440) (0.10,0.17) (4496,9935) ± (397,667) (2777,4002) (0.13,0.18) (3949,7106) ± (122,215) (880,1462) (0.12,0.17) (2825,4907) ± (120,217) (871,1371) (0.12,0.17) (3186,5492) ± (53,90) (421,632) (0.11,0.17) (2176,4085) (76,142) (568,926) (0.12,0.17) (96,139) (42,69) (331,461) (0.11,0.17) (81,121) (26,41) (230,296) (0.10,0.16) (30,46) (30,48) (282,357) (0.10,0.15) (35,53) (32,58) (247,342) (0.12,0.18) (35,54) ontinued

9 Arher et al.: Estimates of the inidental kill of Stenella attenuata attenuata alves in the tuna purse-seine fishery 241 Table 3 (ontinued) Mean alf Total number Estimated alf defiit of NE spotted Estimated Stok and defiit in Observed per kill dolphins killed total alf year observed sets dolphin kill (D k ) (±SE) defiit B Western-southern (WS) stok ,712 ±10, (110,229) (836,1638) (0.10,0.17) ( ,633) ,499 ±10, (100,318) (939,2733) (0.07,0.15) (1554,6890) ,837 ±10, (123,322) (1104,2434) (0.09,0.15) (3285,9121) ,206 ± (524,1036) (4925,7860) (0.09,0.15) (3833,9223) ,260 ± (345,563) (2860,3906) (0.11,0.16) (1098,1959) ,610 ± (342,949) (2508,5922) (0.12,0.18) (932,2614) ,254 ± (104,216) (939,1643) (0.09,0.15) (438,1138) ,200 ± (59,344) (411,2542) (0.10,0.17) (831,2320) ,512 ± (90,340) (577,2416) (0.11,0.17) (1010,2682) ± (198,474) (1337,2734) (0.13,0.19) (1156,2088) ± (15,33) (99,206) (0.12,0.20) (418,1087) ,018 ± (80,224) (526,1513) (0.12,0.18) (712,2352) ± (32,79) (225,570) (0.11,0.17) (781,1524) ,074 ± (143,373) (1065,2784) (0.10,0.17) (1776,3617) ,298 ± (147,374) (1068,2661) (0.11,0.16) (1754,3695) ,916 ± (51,96) (368,605) (0.12,0.18) (1453,2785) ,560 ± (242,589) (1819,4277) (0.11,0.16) (2861,4977) ,578 ± (78,160) (582,1128) (0.11,0.17) (1283,2236) (42,71) (308,448) (0.12,0.18) (64,97) (72,132) (598,931) (0.11,0.16) (112,165) (25,42) (192,288) (0.11,0.17) (38,58) (16,32) (123,209) (0.11,0.18) (28,44) (22,44) (189,330) (0.10,0.15) (42,67)

10 242 Fishery Bulletin 102(2) Table 4 Annual mean observed and mean estimated number of latating females per set in 100% sampled sets. Values in parentheses are 95% lower and upper onfidene intervals assuming a normal distribution of differenes. Bold type indiates signifiant differene from zero (P 0.05) in the paired t-tests. Northeastern stok Western-southern stok No. of Differene No. of Differene Year sets Observed Estimated (95% CI) sets Observed Estimated (95% CI) ( 0.17,0.05) ( 0.63,0.42) ( 0.13,0.07) ( 0.36,0.24) ( 0.05,0.22) ( 0.46,0.77) ( 0.18,0.02) ( 0.119,0.115) (0.01,0.15) (0.03,0.21) ( 0.14,0.08) ( 0.19,0.13) ( 0.13,0.14) (0.01,0.33) ( 0.11,0.13) ( 0.09,0.20) ( 0.81,0.38) ( 0.16,0.19) ( 1.40,0.27) ( 0.37,0.10) ( 7.89,6.28) ( 1.57,0.77) ( 0.67,0.18) ( 0.96, 0.23) ( 0.36,0.06) ( 0.49,0.20) ( 0.04,0.51) ( 0.01,0.38) ( 0.18,0.11) ( 0.44,0.06) ( 0.59,0.35) ( 0.59,0.44) ( 0.68,0.64) ( 0.25,0.44) ( 0.94,0.42) All ( 0.08,0.02) ( 0.053,0.051) ertson, in press) estimated the probability that a alf of a given age lass was still sukling. Given that body length has a near linear relationship with age in these young age lasses (Perrin, 1976), this meant that for any hosen length of independene, eah individual smaller than that utoff value would only be ounted frationally, in effet orreting for the probability that an animal of a given age is not sukling. This proedure aused the method in this paper to tally fewer alves in eah set than in the previous study. A seondary result of this effet was that the mean defiit per set estimated in the present study tended to be slightly higher than that presented by Arher et al. in We estimated the total number of missing alves as a funtion of the number of dolphins killed in eah stok (Table 3). Prior to 1995, only a fration of the purse-seine trips arried sientifi observers. To estimate the number killed in eah stok, kill rates from the observed trips were applied to unobserved trips, stratified by area and stok (IATTC, 2002; Joseph, ). Sine 1995 it has been reported that all dolphin sets have been observed, and that the number of dolphins killed is therefore known without error (IATTC, 2002). The total alf defiit ould also be estimated as a funtion of the number of sets by multiplying the total number of sets made on eah stok by D s (Fig. 4). In the only study to estimate the number of sets made on eah stok annually, Arher et al. 5 used a relatively simple proration sheme of unobserved sets derived from ratios of the number of sets made on eah stok in observed sets. However, beause Arher et al. 5 did not stratify unobserved sets by area, basing the total alf defiit on these estimates would produe a different result from that presented in Table 3. Beause the estimates of the kill by stok inluded stratifiation by area, estimates of the total alf defiit alulated by multiplying the kill estimates by D k are likely to be more aurate. It is important to realize that the defiit that we present is diretly related to the kill observed in the sets that we used. In other words, if proration shemes for unobserved sets were the same for the number of sets made and the number of dolphins killed, estimates of the total alf defiit with either D s or D k would be equivalent. Wade et al. 1 explored the effets of 50% and 100% additional fisheries-related mortality on the assessment of the northeastern spotted dolphin stok. The assumption of additional mortality led to higher estimates of maximum 5 Arher, F., T. Gerrodette, and A. Jakson Preliminary estimates of the annual number of sets, number of dolphins hased, and number of dolphins aptured by stok in the tuna purse-seine fishery in the eastern tropial Paifi, National Oeanographi and Atmospheri Administration Administrative Report LJ-02-10, 26 p. Southwest Fisheries Siene Center, 8604 La Jolla Shores Dr., La Jolla, CA

11 Arher et al.: Estimates of the inidental kill of Stenella attenuata attenuata alves in the tuna purse-seine fishery 243 Table 5 Annual mean number of dolphins killed 122 m (alves killed based on length) and estimated number of sukling alves (alves based on weaning model) per set in 100% sampled sets (first line for eah year). Mean defiit per set using 122 m as utoff length (alf defiit based on length) and alf defiit as estimated in this artile (alf defiit based on weaning model) on seond line for eah year. Values in parentheses are 95% lower and upper onfidene intervals assuming a normal distribution of differenes. Differenes in bold indiate signifiant differene from zero (P 0.05) in the paired t-test. Northeastern stok Western-southern stok Calves killed Calves killed Calves killed Calves killed based on based on based on based on length weaning model length weaning model No. Calf defiit Calf defiit No. Calf defiit Calf defiit of based on based on Differene of based on based on Differene Year sets length weaning model (95% CI) sets length weaning model (95% CI) (0.18,0.50) (0.01,0.55) ( 0.26, 0.03) ( 0.79,0.29) (0.20,0.47) ( 0.53,1.47) ( 0.26, 0.03) ( 0.61,0.25) (0.20,0.64) ( 0.03,0.39) ( 0.05,0.16) ( 0.45,0.66) ( 0.01,0.15) (0.001,0.13) ( 0.14,0.06) ( 0.10,0.15) (0.02,0.12) (0.06,0.16) ( 0.01,0.14) (0.03,0.20) (0.05,0.23) (0.04,0.23) ( 0.14,0.07) ( 0.22,0.09) ( 0.02,0.13) (0.04,0.32) ( 0.11,0.13) ( 0.03,0.31) (0.02,0.25) ( 0.02, 0.003) ,0.119) ( 0.08,0.21) ( 0.194,0.185) ( 0.04,0.07) ( 0.57,0.31) ( 0.17,0.20) ( 0.10,0.55) ( 0.04,0.12) ( 1.27,0.14) ( 0.37,0.08) ( 0.64,0.36) ( 0.31,0.57) ( 7.25,5.91) ( 1.56,0.76) ( 0.08,0.04) ( 0.01,0.21) ( 0.67,0.18) ( 1.13, 0.26) (0.02,0.15) ( 0.06,0.10) ( 0.39,0.03) ( 0.48,0.21) (0.01,0.17) ( 0.04,0.06) ( 0.04,0.44) ( 0.02,0.38) (0.05,0.20) (0.02,0.30) ( 0.22,0.07) ( 0.41,0.09) ( 0.23,0.22) ( 0.62,0.28) ( 0.11,0.29) (0.05,0.49) ( 0.60,0.43) ( 0.66,0.48) ( 0.18,0.38) ( 0.18,1.62) ( 0.20,0.51) ( 1.15,0.27) All (0.13,0.20) (0.09,0.16) ( 0.10, 0.01) ( 0.08,0.03)

12 244 Fishery Bulletin 102(2) growth rates and lower estimates of the urrent size of the population in relation to arrying apaity. Wade et al. 1 did not model the alf defiit estimated in our present study, but the effet of 14% additional mortality would probably be less than the 50% additional mortality that was modeled. The 50% mortality was spread over all age lasses, and additional mortality due to missing alves should be assigned to the first two year lasses only. The important question is whether the alf defiit in the kill represents the main effet of mother-alf separation by the fishing proess. As outlined in Arher et al. (2001), the mehanism by whih sukling alves are separated from their mothers is unknown. If separation is simply a funtion of the number of latating females killed, then the defiit presented here is an aurate representation of the number of missing alves. However, there is some evidene that separation an our without the mother being killed. In the early days of the bakdown proedure, purse-seine skippers reported that Babies swim around the outside of the net pushing to get bak in probably beause their mothers are still inside (Gehres 6 ). It is unlear whether these alves were separated prior to enirlement or were released early during bakdown, prior to their mothers. Regardless, given that dolphins exhibit some of their fastest swimming during a set immediately upon release from the net (Chivers and Sott 7 ), separated alves waiting immediately outside the net may risk separation if their mothers join the rest of the shool rapidly swimming away from the net. If this, or any of the other senarios regarding the manner in whih permanent separation an our without the mother being killed (Arher et al., 2001), then the alf defiit underestimates the atual number of orphaned alves. Future researh should fous on the mehanism of alf separation beause a better understanding of this proess is the only way we will be able to estimate the magnitude of the unobserved alf mortality and its subsequent effets on the population. Aknowledgments The authors wish to thank Mihael Sott and Nik Vogel of the IATTC for providing data as well as Jay Bar- 6 Gehres, L. E Letter of July 2 to Alan R. Longhurst, 2 p. Southwest Fisheries Siene Center, 8604 La Jolla Shores Dr., La Jolla, CA Chivers, S. J., and M. D. Sott Tagging and traking of Stenella spp. during the 2001 Chase Enirlement Stress Studies ruise. National Oeanographi and Atmospheri Administration Administrative Report LJ-02-33, 21 p. Southwest Fisheries Siene Center, 8604 La Jolla Shores Dr., La Jolla, CA low and Bill Perrin for helpful reviews and analytial suggestions. Literature ited Arher, F., T. Gerrodette, A. Dizon, K. Abella and S. Southern Unobserved kill of nursing dolphin alves in a tuna purse-seine fishery. Mar. Mamm. Si. 17: Arher, F., and K. M. Robertson. In press. Age and length at weaning and development of prey preferenes of pantropial spotted dolphins, Stenella attenuata, from the eastern tropial Paifi. Mar. Mamm. Si. Dizon, A. E., W. F. Perrin, and P. A. Akin Stoks of dolphins (Stenella spp. and Delphinus delphis) in the eastern tropial Paifi: a phylogeographi lassifiation. NOAA Teh. Rep. NMFS-119, 20 p. Gerrodette, T Tuna-dolphin issue. In Enylopedia of marine mammals (W. F. Perrin, B. Wursig, and J. G. M. Thewissen, eds.), p Aademi Press, San Diego, CA. Hall, M. A., L. A. Dayton, and K. I. Metuzals Byath: problems and solutions. Mar. Poll. Bull. 41: IATTC (Inter-Amerian Tropial Tuna Commission) Annual report of the Inter-Amerian Tropial Tuna Commission 2000, 171 p. IATTC, La Jolla, CA. Gosliner, M. L The tuna-dolphin ontroversy. In Conservation and management of marine mammals (J. R. Twiss Jr., and R. R. Reeves, eds.), p Smithsonian Institution Press, Washington and London. Myrik, Jr., A. C., A. A. Hohn, J., Barlow, and P. A. Sloan Reprodutive biology of female spotted dolphins, Stenella attenuata, from the eastern tropial Paifi. Fish. Bull. 84: NRC (National Researh Counil) Dolphins and the tuna industry, 176 p. National Aademy Press, Washington, D.C. Perrin, W. F Color patterns of the eastern Paifi spotted porpoise Stenella graffmani Lönnbeg (Cetaea, Delphinidae). Zoologia (NY) 54: Perrin, W. F., J. M. Coe, and J. R. Zweifel Growth and reprodution of the spotted porpoise, Stenella attenuata, in the offshore eastern tropial Paifi. Fish. Bull. 74: Smith, T Changes in size of three dolphin (Stenella spp.) populations in the eastern tropial Paifi. Fish. Bull. 81:1 13. Wade, P. R Revised estimates of inidental kill of dolphins (Delphinidae) by the purse-seine tuna fishery in the eastern tropial Paifi, Fish. Bull. 93: