Assessing the Accuracy of Light-Based Underwater Positioning Algorithms Using Data from ARGOS-Tracked Leatherback Sea Turtles

Underwater positioning of tagged animals currently relies largely on light level measurements. Longitude is classically given by the estimated noon time and latitude is deduced from sunrise/sunset duration and resultant estimated day length. Light-based positioning comes with large uncertainties, especially for the measurement of latitude. Sea surface temperature and bathymetric information are typically used to help reduce this uncertainty.

This paper presents a new state-space model based positioning software. Its inputs are the raw position estimates provided by most tag manufacturers plus records of the daily maximum depths and water temperatures. The resolution method relies on a state-of-the-art Grid Filter. All model parameters are automatically adjusted using the Expectation-Maximization algorithm.

The accuracy of this new underwater positioning tool is assessed using data from seven ARGOS-tracked leatherback turtles for which light-level, temperature and depth data are available. This unique data set, including nearly 1400 days of tracking, allows us to precisely study the impact of the different measurement errors on the actual positioning accuracy.