Tracking marine mammals in 3D using DTAGs: limitations and perspectives. Christophe Laplanche – Ecolab

Speaker: Christophe Laplanche (Ecolab)

Abstract

DTAGs, by providing information on animal 3D orientation and displacement, are used contre leur gré to compute 3D animal tracks. The standard approach for this is to process DTAG data with a linear Kalman filter. We have revisited the processing of DTAG data with a Hierarchical Bayesian Model (HBM) which helped us making explicit the hypotheses behind 3D track reconstruction. We used a beaked whale, deep dive DTAG recording as a case study completed with localization results from an independent, passive acoustic survey. We enunciate the limitations when reconstructing 3D tracks from DTAG data and highlight future avenues of research to fulfill this goal with more serenity.

We also show that computing 3D tracks from DTAG is a sensible affair given that underlying hypotheses are clearly enunciated and if it is made clear that results – confidence intervals on estimated locations – are provided given the statistical model being true. Which is not. We finally extend this presentation with the development of the HBM as a non-linear Kalman filter and show how some of the aforementioned limitations are addressed with this approach.

This work has been done in collaboration with Tiago Marques and Len Thomas from CREEM.