FISHERY BULLETIN: VOL. 84, NO. 1 



800 r 



700 - 



600 



c 

 a 



M 



3 



o 



£ 500 - 



z 



400 



O 



a 



"- 300 

 O 



K 

 Ul 



CD 



2 

 z 



200 



100 



J l_ 



_l_ 



_l_ 



JL 



_1_ 



1959 60 61 62 63 64 65 66 67 68 69 70 71 72 

 YEAR 



Figure 1.— Estimated numbers of dolphins killed in the east- 

 ern tropical Pacific tuna purse seine fishery from 1959 to 1972. 

 Standard errors of the estimates shown as vertical bars. From Table 

 4. 



from 32% in 1959 to 15% from 1966 to 1970, and 

 then increase only slightly in 1971 and 1972. The 

 covariances are large (upper triangular matrix, Table 

 5). They are all positives, and tend to be smaller for 

 pairs of estimates widely spaced in time The 

 covariances can be examined more easily in terms 

 of correlation coefficients (lower triangular matrix, 

 Table 5). The correlations range from 0.31 to 0.99. 

 The CV of the estimated total is 18%. This is 

 substantially higher than the corresponding value 

 of 6% obtained when the covariances are ignored. 

 Because the total is the sum of 14 numbers, an ap- 

 proximate 95% confidence interval, obtained by add- 



ing and subtracting two standard errors, is 3.1-6.5 

 million dolphins. 



The variation in the estimated numbers of dolphins 

 killed over the period 1959-72 is due to several fac- 

 tors: 1) The number of sets made involving dolphins 

 varied from year to year depending on the number 

 of sets of tuna schooling in the absence of dolphins; 

 such tuna are apparently preferred when available 

 2) The use of the backdown dolphin-release pro- 

 cedure increased rapidly from 1959 to 1964. How- 

 ever, the development of the backdown dolphin- 

 release procedure is not well known. The available 

 data reflect the tendency of captains to use the 

 technique once it was known. There is little infor- 

 mation on how rapidly the procedure became known 

 to other captains and no information on how rapid- 

 ly they learned to use it effectively. Our assumption 

 of a linear increase probably overestimates the use 

 of backdown initially, but may or may not overesti- 

 mate its subsequent use 3) The proportion of suc- 

 cessful sets made by small vessels increased from 

 about 50% from 1959 to 1964, to >75% from 1965 

 to 1972 (Table 1). The higher dolphin kill rate for suc- 

 cessful sets results in an increase in estimated 

 dolphin kills as the proportion of successful sets in- 

 creased. 4) The increase in the proportion of sets 

 which were made by large vessels starting in 1968 

 results in a decrease in estimated dolphin kill rates 

 due to the lower dolphin kill rate of these vessels. 



Several factors which may have affected the 

 numbers of dolphins killed in this period have not 

 been accounted for because of the assumptions made 

 by incomplete data. Chief among these assumptions 

 were 1) the relatively small samples are represen- 

 tative of the fleet as a whole 2) the kill rates on un- 

 successful sets are not affected by the use of back- 

 down, 3) the ratio of kill-per-set in successful sets 

 without backdown to that with backdown is constant 



Table 5.- 



-Covariances (upper triangular matrix, x10 10 ) and correlation coefficients (lower triangular matrix) for the 

 estimated total dolphins killed by year, from 1959 to 1972. 



32 



