Maximum likelihood and non-parametric methods for estimating trawl footrope selectivity

Abstract

When a fish comes in contact with a trawl, the probability of its capture may depend on its size, a process that has been called “contact selection”. Estimating capture probability as a function of size is important to properly interpret data from demersal trawl surveys. Previous investigations of contact selectivity of trawls have largely been in the context of mesh retention in commercial fisheries. We extended estimation methods established for retention to account for the more complex capture processes that occur at the footrope. We used a hierarchy of parametric curves, based on the exponential logistic function, to distinguish among forms that represent different fish capture mechanisms. Parameters were estimated using maximum likelihood methods and models were selected based on a sequence of likelihood ratio tests. The cubic spline scatterplot smoother was used as a non-parametric alternative. Scatter plot smoothers are free of three assumptions critical to maximum likelihood estimation of contact selectivity: that the selectivity has a specific functional form, that the fates of individual encountered fish are independent; and that size dependent probabilities of capture are independent among length classes. Mechanisms are suggested that might explain the form of the footrope capture process identified by contact-selectivity estimation. The benefits and limits of each estimation strategy are discussed.