An indirect estimate of bimodal gillnet selectivity of smelt

This study showed that it is possible to model a bimodal gillnet selectivity curve without knowing how individual fish have been caught. A bimodal model based on a modified normal density distribution was fitted over the length ranger of smelt Osmerus eperlanus from 8.9 to 17.0 cm. The model had eight parameters, and the height of the second peak was allowed to change with fish length. The coefficient of determination (r(2)) was 0.915 and the residuals were distributed symmetrically against variable axis. Corrected relative length distribution did not differ statistically from the relative length distribution of the trawl catch. In earlier studies of other species, pooled relative efficiency of multimesh gillnets composed by a geometric series mesh-size combination has been constant on the average. Because the model increased along fish length, the pooled relative efficiency was increasing also with fish length. Therefore, it is suggested that the gillnet catches of smelt always have to be corrected for gillnet selectivity. (C) 1998 The Fisheries Society of the British Isles.