Assessment of mesh size selectivity under commercial fishing conditions

• Main Author: Perez-Comas, José Antonio
• Format: Thesis
• Language: English
• Published: 1996
• Subjects: Theses > Fisheries; Bering Sea
• Online Access: http://hdl.handle.net/1773/5368

Thesis (Ph. D.)--University of Washington, 1996 The present study is entirely focus on the analysis selectivity from alternate-haul experiments performed on fishing grounds off the coasts of Washington, Oregon and northern California, and in the Bering Sea, under commercial fishing conditions. Participating fishing vessels towed especially designed codends that differed in mesh size and shape, and layers of netting, in predetermined sequence. All other decisions concerning the fishing and hauling operations were left to the skippers of the participant vessels.After a brief introduction describing the main concepts, experimental designs and estimation procedures related to fish size selection and its controlling factors, pooled selectivity data for ten commercially important flat and rockfish species were analyzed for the first time. The logistic, probit, Gompertz, negative extreme value and Richards curves were fitted to data to obtain the best selection curves for four diamond- and two square-mesh codends with different mesh sizes. The negative exponential curve produced best fits in 28 out of, mostly with rockfish data. Length of 50% retention was found to increase with increasing mesh sizes. Square-mesh codends appeared to be less selective than diamond-mesh codends of similar mesh size.Next, a new maximum likelihood estimation procedure based on a multinomial distribution was developed for the analysis of alternate-haul data. The method proved to be better than traditional selection-curve fitting procedures, and a good alternative to the binomial based SELECT method. Alternative non-parametric approaches for the analysis of selection curves are also discussed. Two different approaches, one based on a multinomial likelihood and the other on isotonic regression were used in fitting selection curves for two sets of "uncooperative" data. The parametric maximum likelihood estimation procedure was also used to explore the possibility of a simultaneous estimation of selection curves for experimental and control ...