Abstract

We present a mixed model analysis to compare the relative importance of catch and fishing effort as factors in a model to predict turtle bycatch for gillnet fisheries in the Atlantic. We therefore test the assumption that landings, which are universally available, are a sufficient metric to estimate probability of bycatch. Mixed models are useful for modeling observations which are complex, correlated or incomplete and represent a novel approach to bycatch estimation. Our objective is to simulate data-deficient situations where measures of effort might be lacking or effort may be under-reported to determine if catch are suitable proxy for use in predicting sea turtle bycatch. Turtle bycatch were treated as binomially distributed data (1 occur, 0 did not occur) and modeled using a logistic regression analysis with a logit link function. We next compared the fit of models with fixed effects and no fixed effects by plotting ROC curves. Fixed effects for both catch and fishing effort were significant and indicated that the odds of catching a turtle increased, as expected, with more fishing effort or catch. Catch and fishing effort explained a comparable but a much smaller amount of the variation in turtle bycatch. In the Mid-Atlantic gillnet fishery sea turtle bycatch is not significantly different if modeled as a function of mean catch or mean effort. In the absence of detailed information on location and effort, a mixed model approach does offer substantially similar bycatch probability from catch as from effort metrics. We discuss the potential of these techniques to assess sea turtle bycatch in Caribbean fisheries

PDF Preview