A Macquarie University researcher has tested two prominent marine animal-tracking models to determine which is better at inferring the movements of Weddell seals – a large species of seal that reside on and under the ice around Antarctica.
Currently, many problems arise when scientists try to correctly locate and map the movements of an animal using satellite data, which often contains location errors. Dr Ian Jonsen, the author of the study, looked at how these two models cope with satellite errors when inferring the movements of marine predators, such as Weddell seals.
“It is a huge challenge to learn how marine predators make their living in the ocean because we can't watch them directly,” Dr Jonsen explained.
“In the case of tracking Antarctic marine animals, scientists overwhelmingly use the Argos satellite tracking system, which works better at the poles than GPS tracking. However, Argos satellite tracking comes at a price, as it introduces more errors into the data. In this study, I used these large satellite tracking datasets, collected by the Integrated Marine Observing System (IMOS), with maths to unravel the movement behaviours of the seals.”
To test how well each of these models were able to deduce the animals’ movement behaviours, Dr Jonsen fed each model simulated data that introduced satellite location errors and introduced known behavioural switching patterns, such as when a seal will move from a feeding ground to a place of rest. He then fit the two models to real data collected by Argos satellite on the movements of 10 female Weddell seals.
“I firstly tested each model by giving it simulated data where the true movement pattern was known, and compared the ability of the models to recover the true pattern. Then I did the same thing for the real movement data on the seals.”
When the two models were compared, Dr Jonsen found that the first model was more precise and better at predicting the true switches in movement behaviour from the simulated data. It was also better able to identify apparent switches in behaviour of the seals.
“The findings really show that if you want to understand the movements of Weddell seals tracked by Argos satellite, or most other types of marine predator tracked this way, a joint estimation model, which looks at the grouped animal movements, is the way to go. This model performed much better in the tests than the second model which only looks at the individual movements of each animal,” Dr Jonsen said.
Dr Jonsen hopes that the results will be used by scientists looking to improve their understanding of animal movement behaviour from location data collected by satellite tags.
“These models are useful in identifying where and when Weddell seals and other animals engage in different activities like foraging, resting and migrating. The more we refine these models, the better we can apply them to studies looking at how animal behavior may shift in the face of current and future environmental changes,” he concluded.
Currently, many problems arise when scientists try to correctly locate and map the movements of an animal using satellite data, which often contains location errors. Dr Ian Jonsen, the author of the study, looked at how these two models cope with satellite errors when inferring the movements of marine predators, such as Weddell seals.
“It is a huge challenge to learn how marine predators make their living in the ocean because we can't watch them directly,” Dr Jonsen explained.
“In the case of tracking Antarctic marine animals, scientists overwhelmingly use the Argos satellite tracking system, which works better at the poles than GPS tracking. However, Argos satellite tracking comes at a price, as it introduces more errors into the data. In this study, I used these large satellite tracking datasets, collected by the Integrated Marine Observing System (IMOS), with maths to unravel the movement behaviours of the seals.”
To test how well each of these models were able to deduce the animals’ movement behaviours, Dr Jonsen fed each model simulated data that introduced satellite location errors and introduced known behavioural switching patterns, such as when a seal will move from a feeding ground to a place of rest. He then fit the two models to real data collected by Argos satellite on the movements of 10 female Weddell seals.
“I firstly tested each model by giving it simulated data where the true movement pattern was known, and compared the ability of the models to recover the true pattern. Then I did the same thing for the real movement data on the seals.”
When the two models were compared, Dr Jonsen found that the first model was more precise and better at predicting the true switches in movement behaviour from the simulated data. It was also better able to identify apparent switches in behaviour of the seals.
“The findings really show that if you want to understand the movements of Weddell seals tracked by Argos satellite, or most other types of marine predator tracked this way, a joint estimation model, which looks at the grouped animal movements, is the way to go. This model performed much better in the tests than the second model which only looks at the individual movements of each animal,” Dr Jonsen said.
Dr Jonsen hopes that the results will be used by scientists looking to improve their understanding of animal movement behaviour from location data collected by satellite tags.
“These models are useful in identifying where and when Weddell seals and other animals engage in different activities like foraging, resting and migrating. The more we refine these models, the better we can apply them to studies looking at how animal behavior may shift in the face of current and future environmental changes,” he concluded.
