428 



Fishery Bulletin 97(3), 1999 



between observed and predicted returns for model 

 2, particularly in the time periods soon after release. 

 Major discrepancies were eradicated in the model 8 

 fit, which has greater flexibility in fitting the returns 

 for these initial periods. Notice also that model 8 not 

 only better predicts the returns for time periods in 

 which <pi rwas estimated (periods 1-8 for the longline 

 fleet, periods 1-3 and 12-15 for all other fleets) but 

 performs much better in predicting returns from pe- 

 riods of assumed full availability as well (e.g. peri- 

 ods 5 and 6 for U.S. baitboat and troll, periods 10-15 

 for longline). 



Parameter estimates 



Estimates of M and 0, /for the model 8 fit (/3=0.9) are 

 given in Table 5 by way of example. The model 8 es- 

 timate of M (0.608/yr, coefficient of variation 0.09) is 

 similar to estimates obtained by Bertignac et al.' 

 from albacore tagging data in the South Pacific but 



' Bertignac M.. P. Lehodey, and J. Hampton. 1996. An analy- 

 sis of the South Pacific albacore tagging data: estimation of 

 movement patterns, growth and mortality rates. 6th South 

 Pacific Albacore Research Group meeting; 5-7 March 1996. 

 (Available from senior author] 



