Comparison of methods for the estimation of gillnet selectivity to tilapia, cyprinids and other fish species in a Sri Lankan reservoir

Abstract

Methods based on the models of Holt (1957) and Sechin (1969a, b) were compared for the estimation of gillnet selectivity for two tilapiine, five cyprinid, two catfish, one goby and one halfbeak species, inhabiting a Sri Lankan reservoir. Holt's model was extended to the case where the standard deviation of the selection curve increases in proportion to the optimum selection length. Sechin's model was extended to the case where standard deviation of body girth increases in proportion to the body girth. Selection curves from extended Holt and Sechin models were compared with length frequency distributions of the catch (catch LFD) for the largest samples per species. A method for rapid estimation of the position and shape of selection curves was proposed. This method is based on the most common position of retention in gillnets, the length-girth relationship at that position and the ratio between mesh perimeter and body girth. A constant coefficient of variation of 9% was considered to be a good estimate for the standard deviation of the selection curves for non-tangled fish. Tangling percentages were low for all but one toothy species, Glossogobius giuris (Hamilton), which showed bimodality in the catch LFD owing to high tangling percentages. The selectivity for different species was compared using the girth at the position of retention as the independent variable. Differences between species were small and a rough estimation for the most efficiently selected retention girth could be calculated as 2.10 times the stretched mesh size, or 1.05 times the mesh perimeter, independent of fish species. The Holt and Sechin models are only suitable to reconstruct population structures of species with low tangling percentages. The extended Holt model is preferred over the Sechin model, because the Sechin model results in selection curves which are systematically more narrow than the catch LFD and Holt selection curves.