Ward et al.: The effect of soak time on pelagic longlme catches 



181 



Seabirds 

 (-0.06) 



p 

 c\i 



(XX) 



5 10 15 20 25 



5 10 15 20 25 



5 10 15 20 25 5 10 15 20 25 



Soak time (h) 



Figure 1 



Mean catch rates plotted against soak time for skipjack tuna, long-nosed lancetfish. and 

 swordfish in the South Pacific yellowfin tuna fishery and for "other seabirds" in the South 

 Pacific bluefin tuna fishery. To reduce variability, the estimates were limited to longline 

 segments with more than 25 hooks and soak times of 5-20 hours. Vertical bars are 95% 

 confidence intervals for the mean hourly catch rate. In parentheses are the soak-time coef- 

 ficients from random effects models ( note that the soak-time coefficient is not the same as 

 the slope coefficient of a regression of the data presented in this graph). 



Our model is similar to the parabolic catch model exam- 

 ined by Zhou and Shirley (1997). It is simpler than catch 

 equations developed by other authors because it does not 

 include specific terms for the loss of baits, for fish competi- 

 tion, and gear saturation. 



Preliminary plots of observer data indicated a variety of 

 patterns in the relationship between catch rates and soak 

 time (e.g. Fig. 1). By varying the values of P (probability 

 of capture), a (capture rate), and p (loss rate), our simple 

 catch equation (Eq. 4) can mimic the observed patterns 

 (Fig. 2). However, estimates of P , a , and /3 are not avail- 

 able. Instead, we used the empirical approach described 



in the following section to model the effect of soak time 

 on catch rates. The relationship of soak time to catch rate 

 represents the product of the probability of capture and the 

 probability of being retained. 



One approach to investigating the effects of soak time 

 on catch rates is to fit linear regressions to aggregated 

 data like those presented in Figure 1. Such an approach, 

 however, would violate assumptions of independence 

 (within each longline operation, catch rates in consecutive 

 segments will be related), normality (these are binomial 

 data), and homogeneity of variance (for binomial data, the 

 variance is dependent on the mean). 



