A simulation model for designing groundfish trawl surveys

Abstract

This paper describes a convenient simulation model, based on the compound binomial-gamma distribution, to assist the planning and design of groundfish trawl surveys. The analysis uses swept-area density measurements from stratified tows to give a simple nonparametric biomass estimate. A parametric simulation model requires only three input parameters for each stratum, which can be estimated initially from past surveys or commercial fishery data. Analytical results provide intuitive algorithms for estimating variances, investigating tow allocation strategies, and exploring potential survey results. Simulations make it possible to compare the estimated biomass with its true value and to assess coverage properties of confidence intervals obtained from bootstraps. Bias correction and acceleration both improve the results, but small samples taken from populations with highly variable densities tend to produce underestimates of available biomass. The simulation framework allows easy adaptation to address broader issues, such as the design of a multispecies survey.