NOTE Kotas et al.: Incidental capture of Caretta caretta and Dermochelys coriacea by the pelagic longline fishery 



395 



CPUE (number of captured turtles/1000 hooks) was 

 calculated separately for each species. Straight carapace 

 lengths in published data were converted to CCL by us- 

 ing the formula in Teas (1993) to compare the CCL of 

 captured loggerhead sea turtles to carapace length data 

 found in the literature. To assess the significance of the 

 difference in the proportion of dead loggerhead or leath- 

 erback sea turtles among trips, exact tests were applied, 

 because ordinary chi-square tests are not reliable when 

 expected cell frequencies are too small. The test statistics 

 were x 2 = ^[(Observed - Expected )' 2 IExpected] , and exact 

 probabilities were computed for all tables with marginal 

 frequencies fixed at the observed values (Lindgren, 1993, 

 p. 376). These probability calculations were performed by 

 a Turbo Pascal vers. 7 program (Borland International. 

 Scotts Valley, CA). The confidence interval for overall prob- 

 ability of death at capture was calculated by the method in 

 Zar ( 1996, p. 524 ). Ordinary chi-square tests and analysis 

 of variance (ANOVA) tests followed Zar (1996) and were 

 carried out with the software Systat vers. 9 (SPSS Inc., 

 Chicago, IL). In the statistical tests, type-I error a was 

 equal to 0.05. In the construction of Figure 2, to avoid 

 overlapping of data points, the temperatures (but not the 

 CPUEs) were jittered, that is, a small amount of uniform 

 random noise was added to the temperature measure- 

 ments (Cleveland, 1993). 



Results 



From a total of 34 sets and 33,650 hooks, 145 logger- 

 head (CPUE = 4. 31/1000 hooks) and 20 leatherback 

 (CPUE = 0. 59/1000 hooks) sea turtles were captured. 

 There was a significant difference in loggerhead CPUE 

 among the trips (chi-square test, x 2 =137.3, P<0.001), but 

 the proportion of dead loggerhead sea turtles was not sig- 

 nificantly different among the trips (exact test, P= 0.656). 

 The average probability of death at capture for loggerhead 

 sea turtles for the three trips was 0.140 (95% confidence 

 interval= [0.086, 0.210]). For leatherback sea turtles, the 



difference in CPUE among the trips was significant (chi- 

 square test, x 2 =9.76, P<0.01), and the proportion of dead 

 leatherback sea turtles was not significantly different 

 among the trips (exact test, P=1.000). The average prob- 

 ability of death at capture for leatherback sea turtles for 

 the three trips was 0.050 ( 95 r >i confidence interval= [0.001, 

 0.249]). 



The average sea surface temperature (Table 1) was 

 significantly different among the trips (ANOVA, ;? = 34. 

 F=55.37, P<0.001). The average temperature on the first 

 trip was significantly lower than those on the second and 

 third trips, and the average temperature on the second 

 trip was significantly higher than that on the third trip 

 (Tukey's post hoc test). For loggerhead sea turtles, CPUEs 

 were generally higher on the first trip, which had the 

 lowest average temperature (Fig. 2). For leatherback sea 

 turtles, on the contrary, the lowest CPUEs were found on 

 the first trip, on which only one leatherback sea turtle was 

 captured (Table 1). 



CCLs of captured loggerheads were in the range of 

 46-73 cm. Detailed loggerhead CCL data are presented in 

 Table 2. There was a significant difference in average log- 

 gerhead CCL among the trips (Table 2); the average CCL 

 on the third trip was greater than those on the first and 

 second trips (ANOVA, n = 54, F=4.209, P=0.020, Tukey's 



