Informing U.S. Caribbean fisheries management through simulation modeling: a case of length-based mortality estimation models

Abstract

Length-based stock assessment models estimating mortality rates are attractive choices for assessing fisheries with data deficiencies. The U.S. Caribbean is exploring using these models and trying to optimize their commercial sampling program for such a model.

A simulation model was constructed to compare two length-based mortality estimators, the Beverton-Holt and Gedamke-Hoenig models. The simulations also tested aspects of the Gedamke-Hoenig model previously not thoroughly addressed, such as the effects of varied life history parameters, violating the assumptions of constant growth and recruitment, sample sizes (n), and sampling program length (Ys) on total mortality rate estimates (Z).

Given the scenarios investigated, the Beverton-Holt model produced consistently biased, but more stable results when n was low, variation was high for both growth and recruitment, and sampling began after the change in Z took place. The Gedamke-Hoenig model was generally less biased and detected changes in Z, but produced variable results of the current Z, especially with low sample sizes and high variability. In those situations, both models can be carefully interpreted together for management advice.

In the Gedamke-Hoenig model results, a clear pattern emerged in the mean accuracy and precision of the model where after an asymptote was reached, increasing n did not improve the means. The variance of the model improved with both increasing n and increasing Ys. Outliers were predictable and could be accounted for on a case-by-case basis.

The model developed here can be a tool for guiding future stock assessment model choice and sample design in the U.S. Caribbean and other regions.