Research: Predicting Pack-Ice Seal Occupancy of Ice Floes Along the Western Antarctic Peninsula
Project Overview:
Our research (in review), explores the habitat use of Antarctic pack-ice seals by analyzing their occupancy patterns on pack-ice floes. We employed a combination of high-resolution satellite imagery, computer vision and machine learning technique to predict seal occupancy, providing significant advancements in our ability to monitor these critical marine predators.
Key Findings
High Classification Accuracy: Our Random Forest model, incorporating computer vision convolutional neural networks (CNNs), achieved high accuracy in predicting seal presence based on ice floe characteristics.
Ice Floe Size and Seal Abundance: The size of ice floes was found to be the most significant predictor of seal occupancy, with a positive relationship up to approximately 50 m², plateauing beyond 500 m².
Environmental Correlations: Factors such as ice floe size, fine-scale sea ice concentration, and nearby marine topography were significantly correlated with seal haul-out abundance.
Automated Monitoring Potential: This study represents another step towards fully automating the survey of pack-ice seals using satellite imagery, paving the way for pan-Antarctic monitoring of these key marine predators.
Methods
We utilized very high-resolution (VHR) satellite imagery to detect individual pack-ice seals and examine the sea ice features on which seals are hauled out. The SealNet2 CNN was employed for initial seal detection, followed by a quality control review. We integrated these data into a statistical model incorporating local sea ice concentration, bathymetric topography, and proximity to penguin colonies to predict seal occupancy.
Implications
Our research highlights the intricate relationship between local sea ice conditions and the ecological behavior of Antarctic pack-ice seals. By leveraging advanced computer vision and machine learning techniques, this study enhances our ability to monitor seal populations and understand their responses to environmental changes. This approach is crucial for informing conservation strategies and ensuring the protection of these species in the face of climate change and human activities.
For more detailed information, please refer to the full paper here.
