A State Space Age-Structure Model with Variation in Natural Mortality

Age structured assessment models are commonly used for stock assessment: e.g., statistical catch-at-age (SCAA) model.  The advantage of age-structured models lies in high resolution of the demography of a fish population, but those models suffer from over-parameterization where some parameters are not estimable, given fishery and survey data.  One of the common practices to overcome the over-parameterization is to assume a constant force of natural mortality and then estimate the other parameters.  However, natural mortality rates vary over time, and thus the practice can lead to substantial bias in estimates of parameters.  Also in the SCAA model, ‘arbitrary’ weights are allocated to likelihood components for different types of observations for fitting.  The major objective of our study is to develop a ‘state-space’ age structured model where natural mortality is expressed as a function of external information such as fish body size, and arbitrary relative weights of different likelihood components (e.g., from fishery catch, survey indices, age compositions, etc.) are avoided.  We will demonstrate our new model with data on a New England groundfish population.