FISHERY BULLETIN: VOL. 84, NO. 1 



backdown, a dolphin-release procedure, is used 

 (Green et al. 1971; Barham et al. 1977; Smith and 

 Lo 1983). lb account for these factors affecting rates 

 of dolphin kill, Equation (1) can be reexpressed with 

 the rates and numbers of sets stratified by vessel 

 tuna carrying capacity, catch of fish, and use of back- 

 down procedure: 



ft-ZIZ 



1=1 y=i a-=i 



Ktijk Xfijk 



(2) 



where t 



i 



J 

 k 



year 



1 for vessel capacity >600 tons; 2 for 



vessel capacity <600 tons 



1 for yellowfin tuna catch > l k ton; 2 for 



yellowfin tuna catch < l k ton 



1 backdown is used; 2 backdown is not 



used. 



Data on the number of dolphins killed during 

 fishing trips in the period from 1964 to 1968 are 

 given in Smith and Lo (1983). Similar but more ex- 

 tensive data (eg, backdown information) are avail- 

 able in NMFS records for 1971 and 1972. Estimates 

 on the number of sets involving dolphins from 1959 

 to 1972 are given by Punsley (1983). These data 

 sources have certain limitations which do not allow 

 for the use of the complete stratification scheme in 

 Equation (2). Assumptions are made based on sam- 

 ple sizes and on apparent lack of changes in rates 

 over time to accommodate these limitations. 



The mean numbers of dolphins killed (kill-per-set) 

 are shown in Table 1 for each year in which data are 

 available, stratified by vessel size and by catch of fish 

 (successful, > l k ton of yellowfin tuna; unsuccessful, 

 < l k ton of yellowfin tuna). The definition of suc- 



cessful set follows that of Perrin and Zweifel (fn. 3). 

 The vessel class stratification was based on the 

 vessel's fish carrying capacity. The 1964-74 kill data 

 indicate that kill-per-set was different for vessels 

 with <600 tons carrying capacity and vessels with 

 >600 tons for unsuccessful sets. For successful sets 

 the optimal vessel class stratification was not clear; 

 either 400, 600, or 800 tons can be used as division 

 points for stratification. For consistency, we adopted 

 the same stratification used for unsuccessful sets. 

 (The results were similar with alternative stratifica- 

 tion schemes.) Other factors such as the age of the 

 vessel and the experience of the captain could af- 

 fect kill rates but were not considered in the 

 stratification because these factors could not be 

 isolated for analysis. 



The mean number of dolphins killed varied 

 markedly over the years but without any obvious 

 trends (Table 1). A two-way analysis of variance with 

 the data pooled over years showed statistically 

 significant differences in kill rates in sets made by 

 small and large vessels (P < 0.01) and in successful 

 and unsuccessful sets (P < 0.01). Thus Equation (2) 

 was simplified by eliminating the time stratification 

 for kill rates, whereas the vessel size and catch strata 

 were retained. 



Few observations are available for sets where 

 backdown was not used. In successful sets, backdown 

 was used more than 90% of the time; thus, we have 

 observations on kill rates in only 20 sets where back- 

 down was not used. Thirteen of these sets were made 

 by large vessels and seven by small vessels, and the 

 mean kill rates within vessel size class are highly 

 variable and not significantly different. The overall 

 ratio of the kill rates, pooled over vessel size, when 

 backdown was not used and when it was used is 

 significantly greater than unity, and the annual 



Table 1. — Average numbers of dolphins killed (M) in purse seine sets in the eastern tropical Pacific by 

 year, for small and large vessels making successful (> 1 A ton tuna) and unsuccessful (< 1 A ton tuna) net 

 sets. Standard deviation (SD), number of sets (A/), and number of trips are given. 



'From table 5 of Smith and Lo (1983), omitting incomplete data collected in 1966. 



28 



