Multiple-model stock assessment frameworks for precautionary management and conservation on fishery-targeted coastal dolphin populations off Japan.
Stock assessment approaches are often oversimplified due to lack of biological knowledge and insufficient data. In spite of world-wide attention, fishery-targeted coastal dolphin species in Japan have lacked in-depth quantitative stock assessments because of limited time series of population size and an absence of associated biological information. We consequently developed integrated population models that analysed multiple sources of data simultaneously with published biological information within a Bayesian framework. We estimate population status and trends for three main species targeted by fisheries, bottlenose dolphins Tursiops truncatus, Risso's dolphins Grampus griseu, and short-finned pilot whales Globicephala macrorhynchus, using single-species age-aggregated and age-structured population dynamics models. Models were fit to absolute abundance estimates from systematic line-transect surveys, four series of abundance indices calculated from fisher's logbooks, and historical catch records. Published biological information was used to develop prior distributions for the biological parameters. We assessed the possible effects of ecological disturbance and competition using state-space and multispecies models. The multispecies model was selected by the model selection both for the age-aggregated and the age-structured approaches. Single-species assessments found that population size declined 4.2% (Risso's dolphin) to 8.0% (bottlenose dolphin) for three species since the late 1800s based on median posterior values, while the state-space and multispecies models found that environmental disturbance and an interaction among species could have reduced population size more substantially. Policy implications. Simple single-species models are often used to provide conservation and management advice for wild animals but, in this case, results from such models are overly optimistic because they overlook important ecological process. In contrast, current population status could be less favorable if environmental disturbance and interspecific competition actually control population dynamics. Even if that is not the case, considering ecological process in the model will provide more precautionary ways. Our approach of combining multiple modelling frameworks is applicable to many other management systems, and offers increased confidence in estimated status and trends over assessments that consider only a single model.