Implications of the maximum modelled age on the estimation of natural mortality when using a meta-analytic prior: The example of eastern Australian orange roughy (Hoplostethus atlanticus)

Abstract

The eastern Australian stock of orange roughy (Hoplostethus atlanticus) is a deep-water, long-lived species with a history of considerable exploitation during the late 1980s and early 1990s before being reduced to around 10% of unfished spawning biomass only a few years later, resulting in the closure of the fishery in 2006. Recent assessments have shown an increase in biomass, and the fishery was re-opened to targeted fishing in 2015. The stock is an example of the consequences of over-exploitation and of managed recovery. As such, conservation groups and fisheries managers are very interested in the status and future of the stock. Consequently, the assessment is both contentious and highly scrutinised. The current assessment uses the Stock Synthesis platform, with key inputs being annual catches, occasional acoustic surveys and age-composition data. Natural mortality (M) has been fixed at 0.04 yr−1 in the model on which management advice was based for several assessment cycles, but the maximum likelihood estimate of M is closer to 0.03 yr−1. Reducing the assumed value for M would lead to large reductions in catch, as determined by the Australian harvest control rule. A prior for M was developed based on assessments of orange roughy stocks in New Zealand and included in a Bayesian analysis in which M was treated as an estimable parameter. The median of the posterior for M is 0.0353 yr−1 when the maximum age in the assessment (i.e., the ‘plus-group age’) is set to 80 years, but setting the plus-group age to 80 is not based on analysis of data. Increasing the plus-group age to 100 and 120 years leads to posterior medians for M of 0.0381 and 0.0393 yr−1 respectively. This has consequences for catch limits under the Australian harvest control rule, with models that have older plus-group ages having higher estimated productivity and recommended biological catches. The results highlight the value of making use of the results from assessments of similar species to develop priors for M, especially for species that are poorly represented in the data sets on which current meta-analyses are based, and for the need to consider the choice of plus-group age when conducting assessments, particularly for long-lived species.