Chinstrap penguins belong to the brush-tailed penguin group in Antarctica. They are easily identified by the feature that gives them their name – a black band that runs from ear to ear down the chin. The species is commonly found in the West Antarctic Peninsula, on remote islands such as the South Shetland Islands, the South Sandwich Islands and the South Orkney Islands.
Chinstrap penguins are highly specialized predators that feed on marine crustaceans called Antarctic krill. The birds are still very abundant (estimates are between 3 million and 4 million breeding pairs). But most of their colonies are unfortunately experiencing population decline. This trend may be due to climate change, increased populations of other marine predators (such as krill-eating krill whales), and the decline of krill due to commercial krill fishing.
Therefore, it is important to understand how much krill penguins and other marine predators eat. This helps scientists predict future population trends and inform conservation and ecosystem management strategies.
In remote areas of the ocean, it is difficult to directly observe how penguins catch their underwater prey. However, our understanding of their foraging behavior has grown rapidly over the past decade, thanks to innovations in technology that allow for remote monitoring.
Read more: Tracking technology provides new insights into the behavior of migratory birds
We are part of a group of researchers who recently published a study based on such a technological innovation. Using video and motion sensor data from the animals to train machine learning algorithms, we were able to determine how many krill penguins caught. We used “deep learning,” a subset of machine learning, to identify penguin feeding events. In our study, these algorithms not only performed classification tasks faster than human observers, but also identified patterns in data that were difficult to observe visually.
To date, estimates of penguin krill consumption have typically been derived from bioenergetic models based on principles of physiology such as metabolic rate and how energy is assimilated from food. These estimates are often not empirically verifiable. Another old method, gastric bypass is very invasive.
Animal-borne sensors provide continuous, high-resolution data on movement and behavior, enabling the recording of large amounts of data. But all that data needs to be analyzed, which is not an easy task for humans. Machine learning algorithms can quickly process these large data sets.
Searching for penguins
We camped in the South Orkney Islands in January 2022 and January 2023 to collect data for this study. We used waterproof tape to attach miniature video cameras and tags with acceleration and pressure sensors to the backs of the penguins that grew up on the islands. Each penguin spent less than a day at sea collecting data for one foraging expedition. When the penguins returned to their nests to nurse their chicks, we removed the woodcutters.
Read more: New discovery: penguins make sounds underwater when hunting
To attach and remove the devices, we hand-held the nesting penguins, blindfolded them with a soft fabric hood, and held them (in our hands) for several minutes. The short times and the small size of the tags make the possible side effects of this process unlikely.
The videos allowed us to visually confirm each time the penguins caught krill. Other sensors measured the penguins’ diving depths and locomotion dynamics (acceleration in three axes – magnitude, sway and pitch, which allows for determination of body height and rotation at 25 data points per second).
Because the video, acceleration, and depth data were synchronized in time, we were able to use the video observations to identify and mark snippets of acceleration and depth data relevant to prey capture. Machine learning models were then trained on the accelerometer data, using the labels as prey capture instances.
The results showed that the machine learning models we trained with labeled data could identify prey capture events with high accuracy from new velocity and depth data.
Interestingly, the machine learning model can now operate in the absence of video data, identifying prey capture events from new velocity and depth data. In the future, we can use a single speed and depth biological registration tag per bird to obtain information about prey in this species.
For monitoring purposes, video cameras can only record for a few hours before their batteries run out, so it is better to get foraging data without video, while speed and depth can be measured over many days.
Lead conservation
Our hope is that the method we’ve developed can be used to track changes in time and space in how much penguins eat, helping with conservation and ecosystem management in the Southern Ocean around Antarctica.
The research was conducted by Stefan Shumby and Chris Oosthuizen from the University of Cape Town in South Africa in collaboration with experts based at the African Institute of Mathematical Sciences (Lauren Jeantet and Emmanuel Dufourc) and Nelson Mandela University in South Africa (Pierre Pistorius). , Center D’Études Biologiques de Chizé in France (Marianne Cimienti), La Trobe University in Australia (Grace Sutton), and the Norwegian Polar Institute in Norway (Andrew Lowther).
#Diving #penguins #technology #ocean #scientists #birdseye #view #foraging #Antarctic #waters