A review of estimation methods for natural mortality and their performance in the context of fishery stock assessment

Natural mortality (M) is one of the most influential parameters in fisheries stock assessment and management. It relates directly to stock productivity and reference points used for fisheries management advice. Unfortunately, M is also very difficult to estimate, and hence very uncertain. Representing the uncertainty in M and how this influences estimates of management quantities is therefore an important component of conducting stock assessments. This paper outlines the range of methods available to estimate M for use in stock assessment. The methods include those based on maximum age, life history theory, relationships between “well-known” values for M (those found in the literature and based on data for the stock being assessed) and covariates, use of tagging data and catch curve analysis, and estimation within a single- or multi-species stock assessment model. All methods are likely subject to bias and imprecision due to incorrect assumptions and incomplete data. Furthermore, M is generally assumed to be constant over time, age, and sex - assumptions that are unlikely to be true for any stock. Based on our review, there is an obvious benefit to directly estimating M using data and within a stock assessment while assigning a prior based on empirical methods. This approach effectively uses all the available information while also representing the uncertainty. Carefully examining diagnostics and checking for model misspecification is required to ensure that the available data and stock assessment model assumptions are appropriately informative about M when it is estimated during the model fitting process. For situations where direct estimation is not possible (a condition found in data-limited to data-rich stock assessments), the use of multiple methods with robust sensitivity exploration is recommended. Even when direct data are integrated into a stock assessment, we recommend using other methods to estimate M and analysing the direct data outside the stock assessment model as diagnostic tools.