|
Getting your Trinity Audio player ready...
|
Researchers from the Department of Biodiversity, Conservation, and Attractions and Murdoch University have employed artificial intelligence (AI) to analyse an extensive collection of data in order to reveal the utilisation of habitat by a vulnerable turtle species in Yawuru Nagulagun Roebuck Bay, located in Broome.
The flatback turtle (Natator depressus) is native to the northern regions of Australia, and although female turtles come ashore to lay their eggs, there is limited knowledge about their behaviour in the sea.
According to the lead author and a PhD candidate, the research demonstrates the effectiveness of machine learning, a form of AI, in deciphering extensive datasets obtained from multi-sensor data loggers. This enables the researchers to establish connections between animal behaviour and their temporal and spatial patterns.
Motion sensors, similar to those found in smartwatches, were attached to the turtles to capture high-resolution data at a rate of 50 times per second. This data provided a comprehensive understanding of the turtles’ underwater activities, which are difficult to observe directly.
Additionally, some of the data loggers were equipped with video cameras, allowing the researchers to train the AI system to accurately identify resting and foraging behaviours solely based on data, utilising observations from 83 hours of video footage.
Through this innovative approach, the researchers made significant discoveries about the behaviour of flatback turtles who tend to rest and forage in consistent locations. However, due to variations in seasonal water temperatures and the region’s notable large tides, the specific foraging locations change throughout the year.
The research has revealed previously unknown patterns of habitat use by flatback turtles in Roebuck Bay. The turtles were observed resting on the sea floor near intertidal areas and adapting their foraging activities in response to tidal patterns rather than following a strict day/night cycle.
A scientist from the Department of Biodiversity, Conservation, and Attractions (DBCA) and a co-supervisor of the PhD, highlighted the significance of this study for conservation planning and the establishment of protected areas such as marine parks.
Understanding how turtles use different parts of their habitat throughout the year is crucial for effective conservation strategies and prioritising conservation actions, she explained. By comprehending the resting and feeding preferences of the turtles, it becomes possible to identify areas that need protection from potential disturbances caused by vessel traffic or other human activities at sea.
A senior lecturer at Murdoch University and the senior author of the study expressed enthusiasm about the application of machine learning techniques as a valuable tool for enhancing the comprehension and conservation of marine fauna.
It was noted that direct observation of marine species is challenging, and the use of multi-sensor data loggers often generates an overwhelming volume of data. Therefore, this new technique holds great potential in unravelling the behaviours of numerous endangered species.
The findings of this study, titled “Behaviour-specific spatiotemporal patterns of habitat use by sea turtles revealed using biologging and supervised machine learning,” have been published in the Journal of Applied Ecology.
The research was a collaborative effort between the North West Shelf Flatback Turtle Conservation Program of the Department of Biodiversity, Conservation, and Attractions (DBCA) and Nyamba Buru Yawuru. Conducted since 2018 within the jointly-managed Yawuru Nagulagun Roebuck Bay Marine Park, the project received field support from the Yawuru Rangers and Yawuru Country Managers.
Funding for the research was provided by the DBCA and Murdoch University, with additional support from the Holsworth Wildlife Research Endowment of the Ecological Society of Australia.
