Marine scientific surveys are carried out worldwide to monitor changes in abundance and population structures of fish and zooplankton. Surveys are typically repeated annually in the same area and season using standard sampling equipment and procedures. Such standardization is essential for the quality of stock monitoring, as any change in spatial distribution may affect the relationship between the underlying population density and the survey estimates, which may impact the quality of the stock assessments.
In a changing environment, it is difficult to modify the survey coverage to fully overlap the spatial distribution of the target species. The survey may also be conditioned on weather constraints, vessel costs, and limited access to restricted areas. In areas with poor or no survey coverage, traditional design-based survey estimation techniques are not designed to estimate abundance. To accommodate for time varying survey coverages, the authors of the latest Editor's Choice article, "Predicting abundance indices in areas without coverage with a latent spatio-temporal Gaussian model", have made a statistical spatio-temporal model that utilizes underlying structures in both space-time and between length groups. Importantly, all estimates of abundance are accompanied with a measure of uncertainty making it possible to monitor the effect of poor survey coverage.
The applied model can be adopted to fit all types of catch rate survey data, and is tested here on Northeast Arctic cod trawl survey data collected during winter in the Barents Sea. Since the start of the survey time series in 1994, the Arctic ecosystem has been heavily impacted by global warming where the sea surface temperature has drastically increased transforming areas with thick sea ice into open waters with new areas for cod. In addition, the survey coverage has varied due to weather and other factors causing spatial gaps in the trawl station coverage. With global warming, the authors show how the centre of gravity of this important species has shifted in space over time. It is shown that this shift is length dependent and that the small cod has moved northwards while larger cod have shifted towards northeast. Despite these changes in cod distribution and missing data in some areas, the spatio-temporal model enables us to estimate time series of abundance of cod by length groups with measures of uncertainty. Sun altitude is included to explain catch rates, and it is observed that catch rates of cod varies strongly with time of day.
The applied model is a tool for analyzing spatial-temporal changes in fish and zooplankton population densities and provides insight into what environmental conditions the target species are seeking. It is also a tool for stock assessment and sustainable management as it utilizes underlying structures and correlations in space and time. Theoretically, missing data in an area does not pose an issue for the applied modelling framework; however, poor coverage provides less information about the underlying spatial structures and will result in less certain survey estimates and stock assessments.
Read the full paper, "Predicting abundance indices in areas without coverage with a latent spatio-temporal Gaussian model", in ICES Journal of Marine Science.
Editor's Choice articles are always free to read in ICES Journal of Marine Science.
Shoal of cod (left) and heat map illustrating a snapshot of spatio-temporal cod distribution in the Barents Sea (right).