TRENT and PRISTAS: SELECTIVITY OF GILL NETS 



basic models, for describing selectivity of gill nets 

 were proposed by Baranov (as described by 

 McCombie and Fry 1960), Olsen (1959), McCom- 

 bie and Fry (1960), Gulland and Harding (1961), 

 Ishida (1962), Holt (1963), Regier and Robson 

 (1966), Hamley (1972), and Hamley and Regier 

 (1973). Ten methods of describing selectivity were 

 used by the above authors. Except for the DeLury 

 method described by Hamley (1972), the 

 mathematics and details of application of these 

 methods were discussed by Regier and Robson 

 (1966). 



A comprehensive review of gill net selectivity 

 was presented by Hamley (1975). All basic models, 

 applications and shortcomings of these models, 

 and the variety of factors (thickness, materials, 

 and color of net twine, hanging of net, and methods 

 of fishing) that must be considered in determining 

 selectivity were discussed. 



The method proposed by Holt (1963) was used to 

 evaluate selectivity on species that were caught in 

 this study. Holt assumed that: 1) the selectivity 

 curve would take the form of a normal frequency 

 distribution; 2) the efficiencies of two nets with 

 different mesh sizes would be similar for fish of 

 their respective mean lengths; and 3 ) the standard 

 deviations of the distributions for two different 

 mesh sizes would be equal. The equations for 

 evaluating the above assumptions and for de- 

 scribing selectivity have been given by Holt 

 (1963), Regier and Robson (1966), and Hamley 

 (1975). 



If Holt's three asssumptions are analyzed and 

 deemed acceptable, points of the selectivity curve 

 for mesh size m, can be computed by 



5 <; = ex p[-^ ,/ '- 7 < )2 ] 



2s, 



where /,- = length offish in length stratum j 

 7, = mean selection length 

 s, = standard deviation of the selectivity 



curve 

 n y = number of fish of length /, caught in 

 net m, . 



Then nJs^ can be used to estimate abundance of 

 fish for each / ; and therefore, the length-frequency 

 distribution in the fished population can be es- 

 timated from the length-frequency distribution 

 obtained from fishing a particular mesh size on the 

 population. 



An additional assumption is necessary if 



catches from a series of nets with different mesh 

 sizes are combined and used to estimate the 

 length-frequency distribution of the fished 

 population. The assumption is that the selectivity 

 curves for all meshes have the same shape (each s, 

 is an estimate of a commons) and amplitude (each 

 net fishes with equal efficiency on the length at 

 which the net is maximally efficient). This as- 

 sumption was questioned by Ricker (1947), Ishida 

 (1964), Regier and Robson (1966), and Hamley 

 (1972). The assumption can be tested only if the 

 length-frequency distribution of the fished 

 population is known. Hamley and Regier (1973) 

 tested this assumption on walleye, Stizostedion 

 vitreum vitreum, which were tagged prior to being 

 recaptured with gill nets, and found that the 

 shapes and amplitudes of their selectivity curves 

 changed with mesh size. This assumption could 

 not be tested in our study. 



Information derived from a selectivity study has 

 various uses depending upon the validity of the 

 mathematical model used to describe selectivity 

 and on the accuracy and precision required. The 

 model can be useful for some purposes even if all 

 the assumptions are not met or even if the model is 

 not the most accurate and precise one for describ- 

 ing the empirical data. 



The objective of most selectivity studies has 

 been to determine the most appropriate model for 

 describing gill net selectivity for a single species of 

 fish (Regier and Robson 1966). In this study we 

 have attempted to provide as much information as 

 possible about gill net selectivity on 22 species. To 

 10 of these we applied a single mathematical 

 model and either accepted or rejected the model in 

 relation to each of several potential applications. 

 By accepting the model we do not infer that it is 

 the most accurate or precise model but that the 

 approximation to the data is sufficiently close and 

 accurate to be useful. 



NUMBERS AND MEAN LENGTHS OF 

 FISHES SELECTED FOR ANALYSES 



Of the 76 species that were caught in the study 

 area during 1973 (May et al. 1976; Pristas and 

 Trent 2 ), 22 species had catches exceeding 100 

 specimens. Of the 22 species, 15 were commer- 

 cially important in gill net fisheries in one or more 

 states along the south Atlantic and Gulf of Mexico 



2 Pristas, P. J., and L. Trent. 1976. Seasonal abundance, size, 

 and sex ratio of fishes caught with gill nets in St. Andrew Bay, 

 Florida. (Unpubl. manuscr.) 



187 



