Incorporating Uncertainty into Fishery Models

Incorporating Uncertainty into Stock Assessments: A Case Study of Atlantic Striped Bass

Gary R. Shepherd

doi: https://doi.org/10.47886/9781888569315.ch5

Abstract.— A stock assessment of Atlantic striped bass Morone saxatilis was presented to illustrate potential sources of uncertainty in application of an age-based population model. Erroneous conclusions in stock assessment can result from incorrect model selection, input data that are not representative of the target population, and improper configuration of the selected model. Influence of incorrect input data and model configuration was investigated using striped bass catch-at-age data analyzed with a tuned virtual population analysis model (ADAPT VPA). Variations in model configurations were explored in addition to sensitivity to input parameters such as natural mortality. Violation of the assumption of constant natural mortality-at-age had a significant influence on the resulting estimates of F and stock size. Discard losses, particularly from the commercial fishery, were the largest source of uncertainty in the catch-at-age. Uncertainty due to process error in the VPA model was characterized by bootstrap realizations of the nonlinear least-squares estimates of fishing mortality. The implications associated with fishing at various F s were also examined using a stochastic projection model. A comparison of fishing mortality estimates derived from two independent models, an age-structured population model and a tag-recovery model, indicated that both methods produced equivalent results. Evaluation of the striped bass stock assessment demonstrates that uncertainty could result from a variety of sources but this variability was only partially captured within the model framework. Understanding the possible sources of uncertainty and implications in interpreting model results should benefit the analyst in providing assessment advice to managers.