Studies of Mortality Estimation

Instantaneous total mortality rate Z can be partitioned into two components: fishing, F, and natural mortality, M. A number of data-poor methods have been developed to estimate Z, F, and M, and these methods tend to rely on fairly restrictive assumptions as well as on data types that are easy to obtain, e.g., length or other life history information. The overarching goal of this dissertation is to contribute to the advancement of methods for estimating these important and influential stock parameters. The relevant issues and the gaps in knowledge pertaining to these data-poor methods are outlined in the Introduction chapter (Chapter 1). The research papers presented in this dissertation fall into two main categories, namely, the evaluation of existing methods and the improvement of existing methods to estimate mortality rates. In Chapter 2, Monte Carlo simulation is utilized to compare the performance of two length-based methods developed by Beverton and Holt (1957) and Ehrhardt and Ault (1992), for estimating Z. I examine the impact of (1) variability in size at age and (2) the method of handling length truncation on the performance of the estimators. Results show that the Ehrhardt-Ault method exhibits complex patterns of bias and is not unequivocally better than the Beverton-Holt method. In Chapter 3, an existing non-equilibrium, mean length-based estimator of Z is modified to use additional information on fishing effort. The Z parameters are replaced with Z = qft+ M where q is the catchability coefficient and ft is the fishing effort in year t. Thus, only q, M, and the residual error need be estimated. This methodology appears promising for estimating F (= qf) and M, based on simulation studies. Furthermore, even if the estimates of F and M are imprecise and highly correlated, the resultant estimates of Z are year-specific and may be quite precise. The method may serve to bridge the gap between data-poor and data-rich methods to estimate Z. Chapter 4 addresses a long-standing gap in knowledge with respect to the ranking and predictive performance of existing empirical estimators of natural mortality of fish stocks. to address this question, a dataset of over 200 direct M estimates and corresponding life history parameters from unique fish species was compiled. Using this dataset, we were able to definitively quantify the predictive ability and update the equations of four widely used empirical estimators and their variants. Estimators based on maximum age perform substantially better than those based on growth parameters, either with or without consideration of water temperature. Results from this research will provide useful tools and guidelines for stock assessment scientists who need to estimate M and Z for both data-poor and data-rich stocks