Using accelerometers for tracking loggerhead and green sea turtle behaviour

Understanding animal behaviour is critical for the design of effective conservation and management strategies. Animal-borne tri-axial accelerometers constitute a type of biologging device which have the potential to provide continuous high-resolution behavioural data. For marine animals, device attachment position may influence both the accuracy of behavioural predictions and the hydrodynamic profile of the animal. We present a case study on the use of accelerometers for the behavioural classification of two sea turtle species in captivity: the loggerhead (Caretta caretta) and green (Chelonia mydas) turtle. Accelerometers were placed on the first and third scute to represent extreme placement scenarios. We trained Random Forest (RF) models to classify behaviour and assessed the impact of placement and sampling frequency on accuracy. In addition, we assessed the impact of device position on carapace drag coefficient using Computational Fluid Dynamics (CFD).

We achieved a high accuracy for behavioural classification (0.86 for loggerhead and 0.83 for green turtles). We determined that overall RF accuracy for both species is significantly higher for devices positioned on the third scute compared to the first scute (P < 0.001) and with a smoothing window of 2 s compared to 1 s (P < 0.001). We found no significant effect of sampling frequency and therefore recommend the use of 2 Hz in future work to optimise battery life and device memory. CFD modelling indicated an increase in drag coefficient from a maximum of 0.028 without a device to a maximum of 0.064 with a device for an isolated turtle carapace. Attachment to the first scute significantly (P < 0.001) increased drag coefficient relative to the third scute.

Moving forward, the attachment and sampling protocols we present here may be adopted in future studies involving captive sea turtles. Further research is needed to assess their applicability and effectiveness under free-ranging conditions to enable their use in wild populations.

The online version contains supplementary material available at 10.1186/s40317-025-00415-3.