372 



Fishery Bulletin 91(2). 1993 



aO° 100° 120° 140° 160° 180° '60° .140° 120° 100° 80" 60° 



80° 100° 120° 140° 160° 180° 160° 140° 120° 100° 80° 



Figure 1 



Areas sampled for albacore Thunnus alalunga by drift gillnets. (A) JAMARC 1980 

 survey (JAMARC 1983); (B) JAMARC 1982 survey (JAMARC 1985); (C) JAMARC 1983 

 survey (JAMARC 1986); (D) NMFS 1985 experiment. 



comparison. Sample sizes of al- 

 bacore caught by mesh size for 

 each experiment, as well as the 

 quantity of gear fished, are 

 shown in Table 1. 



Results 



JAMARC 1 980 data 



Gear selectivity can only be esti- 

 mated because there is no infor- 

 mation on the actual size struc- 

 ture of the North Pacific albacore 

 population (Hamley 1975). Data 

 from several mesh sizes fished 

 simultaneously and assumptions 

 about shape, variance, and effi- 

 ciency of the selectivity curves 

 are required. The JAMARC 1980 

 experiment provided the only 

 published dataset containing this 

 information. Length-frequencies 

 obtained in the 1980 experiment 

 for mesh sizes 130, 150, 160, 170, 

 180, and 200 mm are shown in 

 Fig. 2 (A-F). In all these mesh 

 sizes, albacore captured had size 

 modes at 53, 62, and 78cmFL. 



Two options are available for 

 calculating an indirect selectiv- 

 ity curve: (1) Fitting a pre- 

 determined distribution function 

 to data, using a method similar 

 to that of Ishida (1962); or (2) 

 estimating the selectivity of dif- 

 ferent meshes to one common 

 size-class (mode) of fish and ex- 

 trapolating to other mesh sizes 

 (Regier & Robson 1966). Lack of 

 an objective mathematical func- 

 tion in determining the shape of 

 the selectivity curve in the first 

 method, and the need for fitting 

 the selectivity curve "by eye," re- 

 quired that a combination ap- 

 proach was employed. 



A basic assumption of gillnet 

 selectivity analysis is that selec- 

 tivity is the same for those com- 

 binations of length interval L(i) 

 and mesh size M(j ) where the ra- 

 tios of L( i l/M(j ) are equal ( Regier 



