Modeling Age-at-Length and Length-at-Age Data Incorporating Multiple Years

Many fishery stock assessments rely on accurate estimates of age composition. For a given year, the age composition of a population can be calculated using age-length keys; the length distribution of the population is converted to an age distribution using age-at-length estimates from a sub-sampled population. This method requires annual estimates for age-at-length since annual fluctuations in recruitment affect size structure. We develop a new method to calculate age composition simultaneously for multiple years. Our model assumes that the length of a fish at a given age does not vary interannually, and thus previous years’ data are used to estimate length-at-age distributions. For a given year, both the age-at-length information for the year (from the classical age-length key) and the length-at-age information for all years are combined in a likelihood function using Bayes’ Rule. The model can also estimate age composition from the length-at-age information for a year when only lengths were recorded, and the performance of the model for this situation is of particular interest. We test this model and assess its efficiency in aging populations using stochastic simulation based on 20 years of striped bass data in which 1000 fish were aged each year.