A Bayesian interpretation of target strength data from the Grand Banks

Knowledge of the target strength (TS) distribution of the fish being surveyed is necessary for accurate acoustic biomass estimates. Although dual-beam and split-beam transducers can measure TS in situ, these measurements are usually misleading because fish with higher TS are detected more readily, measurements are coming from several fish species, and measurements are made on the total echo from several unresolved targets. A Bayesian method of analysis was developed to extract information from TS data despite these problems. The first problem was addressed using a model of fish detection that was calibrated with data from a survey of northern cod (Gadus morhua) on the Grand Banks. The second problem, species composition, was dealt with by incorporating information about the relative abundance of species groups. Prior information about the concentration profile of the species assemblage was used to take into account the third problem, multiple detection. Results suggested that the difficulties mentioned above are serious sources of bias that should not be ignored. The method also indicated how research effort can be directed to improve acoustic surveys.