An Evaluation of Natural Mortality Assumptions in Stock Assessments

Many stock assessments assume natural mortality (M) is constant among ages and years, but some stock assessments have begun to include age or age and time varying M (collectively referred to below as dynamic M).  The methods used for incorporating dynamic M into assessments can vary among assessments, be ad hoc, and have not been evaluated for the bias and precision of resulting parameter estimates.  The objective of this study was to compare the bias and precision of parameter estimates resulting from assessments that make different assumptions about how and if M varies among ages and years.  Our approach was to model a generic fish population with time and age varying M and simulate the collection of data commonly used in stock assessments.  We then fit statistical catch at age models to these data using a constant M, an age varying M, and an age and time varying M.  The treatment of M that led to the least bias and most precision in stock assessment parameter estimates depended on several factors, suggesting that the appropriate form of M in stock assessments may be case specific.  Stock assessment scientists should consider data quality and the relative contributions of M and fishing mortality to total mortality when choosing an assumption about M.