Facial Recognition Analyses Reveal Social Networks of Co-Occurrence at Harbor Seal Haul-Out Sites

Understanding the behavior and population dynamics of harbor seals (Phoca vitulina), an ecologically significant and widespread coastal pinniped, is vital to effectively manage vulnerable coastal marine ecosystems. A novel, non-invasive facial recognition technology was used to identify harbor seals at haul-out sites across two consecutive molting seasons in Middle Bay, Maine. Using a gallery of 672 individual seals found in the bay over a 6-year period, we recorded the presence of seals at target sites over 30 days across 2 years and constructed social network analyses based on repeated co-occurrences at these sites within the seasons and across years. Our results revealed strong site fidelity among a subset of seals across years and persistent co-occurrences between individual seals within seasons and across years, suggesting non-random haul-out behaviors within the region. A core group of seals central to the networks were regularly observed at the sites across the span of the molting season, indicating strong site co-fidelity of seals during this season. Module membership of co-occurring seals varied across years, but the overall structure of the network (graph density, average degree, average clustering and average path length) remained consistent across years. These findings underscore the robust site fidelity of harbor seals to specific haul-out locations in Middle Bay and demonstrate that harbor seals are not transient visitors to this region. This study also demonstrates the utility of facial recognition as an effective, non-invasive method for long-term monitoring of wild seal movements, distribution, habitat use, and social interactions. Using this approach can provide novel insights into how individuals and populations of wild seals respond to environmental or anthropogenic changes and can inform conservation efforts to mitigate threats to coastal ecosystems.