A Bayesian hierarchical modelling approach to estimating landings- and discards-at-age

From SFI

A Bayesian modelling approach has been implemented in the COST project to estimate landings and discards at age, along with length and weight at age. The method takes the various different kinds of data commonly available from sampling of commercial fisheries, ie length only samples, length and age samples and length stratified age samples, both for discards and landings, and combines them into a single analysis. Because length-given-age is modelled explicitly, there is no need to create age-length keys, and modelling of the covariates means that empty “cells” (ie combinations of covariates) do not present a problem. The hierarchical nature of the analysis allows data at different stratifications (hauls or trips) to be included, and the uncertainty is fully described in the posterior distributions of the parameters. The modelling approach also allows virtual populations to be simulated, which mean that the results can be compared with known true values and a realistic assessment of the performance made.

This is joint work with Hanne Rognebakke, Geir Storvik and Magne Aldrin.