Recent summer diet of hooded Cystophora cristata and harp Pagophilus groenlandicus seals in the drift ice of the Greenland Sea

Multi-decadal trends in biomarkers in harp seal teeth from the North Atlantic reveal the influence of prey availability on seal trophic position

Arctic food webs are being impacted by borealisation and environmental change. To quantify the impact of these multiple forcings, it is crucial to accurately determine the temporal change in key ecosystem metrics, such as trophic position of top predators. Here, we measured stable nitrogen isotopes (δ15 N) in amino acids in harp seal teeth from across the North Atlantic spanning a period of 60 years to robustly assess multi-decadal trends in harp seal trophic position, accounting for changes in δ15 N at the base of the food web. We reveal long-term variations in trophic position of harp seals which are likely to reflect fluctuations in prey availability, specifically fish- or invertebrate-dominated diets. We show that the temporal trends in harp seal trophic position differ between the Northwest Atlantic, Greenland Sea and Barents Sea, suggesting divergent changes in each local ecosystem. Our results provide invaluable data for population dynamic and ecotoxicology studies.

Automatic Fish Age Determination across Different Otolith Image Labs Using Domain Adaptation

The age determination of fish is fundamental to marine resource management. This task is commonly done by analysis of otoliths performed manually by human experts. Otolith images from Greenland halibut acquired by the Institute of Marine Research (Norway) were recently used to train a convolutional neural network (CNN) for automatically predicting fish age, opening the way for requiring less human effort and availability of expertise by means of deep learning (DL). In this study, we demonstrate that applying a CNN model trained on images from one lab (in Norway) does not lead to a suitable performance when predicting fish ages from otolith images from another lab (in Iceland) for the same species. This is due to a problem known as dataset shift, where the source data, i.e., the dataset the model was trained on have different characteristics from the dataset at test stage, here denoted as target data. We further demonstrate that we can handle this problem by using domain adaptation, such that an existing model trained in the source domain is adapted to perform well in the target domain, without requiring extra annotation effort. We investigate four different approaches: (i) simple adaptation via image standardization, (ii) adversarial generative adaptation, (iii) adversarial discriminative adaptation and (iv) self-supervised adaptation. The results show that the performance varies substantially between the methods, with adversarial discriminative and self-supervised adaptations being the best approaches. Without using a domain adaptation approach, the root mean squared error (RMSE) and coefficient of variation (CV) on the Icelandic dataset are as high as 5.12 years and 28.6%, respectively, whereas by using the self-supervised domain adaptation, the RMSE and CV are reduced to 1.94 years and 11.1%. We conclude that careful consideration must be given before DL-based predictors are applied to perform large scale inference. Despite that, domain adaptation is a promising solution for handling problems of dataset shift across image labs.

Hooded seal Cystophora cristata foraging areas in the Northeast Atlantic Ocean-Investigated using three complementary methods

Identifying environmental characteristics that define the ecological niche of a species is essential to understanding how changes in physical conditions might affect its distribution and other aspects of its ecology. The present study used satellite relay data loggers (SRDLs) to study habitat use by Northeast Atlantic hooded seals (N = 20; 9 adult females, 3 adult males, and 8 juveniles). Three different methods were used in combination to achieve maximum insight regarding key foraging areas for hooded seals in this region, which have decline by 85% in recent decades: 1) first passage time (FPT); 2) vertical transit rate and; 3) change in dive drift rate. Generalized additive mixed models (GAMM) were applied to each method to determine whether specific habitat characteristics were associated with foraging. Separate models were run for the post-molting and the post-breeding seasons; sex and age classes were included in the GAMMs. All three methods highlighted a few common geographic areas as being important foraging zones; however, there were also some different areas identified by the different methods, which highlights the importance of using multiple indexes when analyzing tracking and diving data to study foraging behavior. Foraging occurred most commonly in relatively shallow areas with high Sea Surface Temperatures (SST), corresponding to continental shelf areas with Atlantic Water masses. All age and sex classes overlapped spatially to some extent, but the different age and sex groups showed differences in the bathymetry of their foraging areas as well as in their vertical use of the water column. When foraging, pups dove in the upper part of the water column in relatively deep areas. Adult females foraged relatively shallowly in deep water areas too, though in shallower areas than pups. Adult males foraged close to the bottom in shallower areas.