Evaluating the impacts of fixing or estimating natural mortality, across life histories and data availability.
Claudio A. Castillo Jordan
November 12th, 2019 9:00 AM (PST): FSH 203
In integrated, age-structured stock assessment model, common practice gives us two alternatives on how to use natural mortality. First, to use a value coming from an indirect method and not estimate it. Second, natural mortality is estimated in the model simultaneously with other parameters. We used simulation testing with models simplified from real stock assessments to generate simulated data, using the ss3sim framework. The operating model “truth” included various levels of complexity, such as sex- and age-specific, and time-varying M. Simulated data sets were used to investigate the bias and uncertainty in estimates of natural mortality and management quantities when assuming and estimating various levels of complexity of M in stock assessments, as well as the data sources that inform natural mortality by varying the type and amount of data available to the model. Given a large number of assessment models available and the range of data, life histories, and other characteristics that they cover, the results were evaluated to determine what characteristics are essential to be able to reliably estimate natural mortality. The uncertainty in the estimates of M was contrasted with the variation in the different indirect methods to evaluate in what context estimation of M inside the stock assessment model is an improvement over the indirect methods and in which cases it is worse. This study provides information about the overall uncertainty in management quantities given the uncertainty in M.