Protected Species Bycatch Estimating Approaches: Estimating Harbor Porpoise Bycatch in U.S. Northwestern Atlantic Gillnet Fisheries

Abstract

Harbor porpoise bycatch estimates for federally managed gillnet fisheries in northwestern Atlantic US waters were calculated for a nine-year period (1999–2007) using two new methods, and the results were compared to the traditionally used stratified ratio estimation method. The aims of this research were to improve on the existing methods for estimating harbor porpoise bycatch for the Northeast and Mid-Atlantic US gillnet fisheries, to provide insight into the causes of harbor porpoise bycatch, and to compare bycatch estimation techniques that could be applied to other fisheries and species. The new methods included a model approach, and a ratio estimation approach that incorporated variables from both the regression model and the existing ratio estimation method. Initially, bycatch was modeled using a GAM forward stepwise process and included testing numerous variables describing the time, duration, and location of the fishing gear, the fishing gear configuration, and the environmental characteristics of the fished waters. The final model was simplified to a GLM and included variables describing port groupings, seasons, bottom depth, stretched mesh size, and year categories. The new mixed variable ratio estimation approach calculated both the Northeast and Mid-Atlantic bycatch estimates within the same framework by using the port groupings and mesh size variables from the model, and the season variable from the traditional ratio estimating approach. Bycatch estimates for the entire study area were similar between modeling and ratio estimator approaches, though estimates for the two new techniques were more stable from year to year in areas with less observed bycatch. The CVs for the model based estimates were much lower than ratio based estimates, and CVs for the two ratio estimation approaches were similar to each other. However, the model CVs may have been artificially low, as the model may have been over-parameterized in an attempt to accurately calculate annual estimates. Despite some differences, the estimates were not significantly different between approaches for the majority of comparisons.