Animal tracking with particle algorithms for conservation

Edward Lavender, Andreas Scheidegger, Carlo Albert, Stanisław W. Biber, Jakob Brodersen, Dmitry Aleynik, Georgina Cole, Jane Dodd, Peter J. Wright, Janine Illian, Mark James, Sophie Smout, James Thorburn, Helen MoorbioRxiv 2025 · DOI

TL;DR

This paper introduces advanced particle algorithms for analyzing animal movement data from acoustic telemetry, significantly improving the accuracy of space use and residency estimates compared to traditional methods. This breakthrough offers a more robust framework for understanding animal behavior and designing effective marine conservation areas.

What Problem Does This Paper Address?

Marine Protected Areas (MPAs) are crucial for conservation, but their effectiveness hinges on understanding the movement patterns of the species they aim to protect. For many aquatic animals, especially those that rarely surface, tracking data is sparse and often limited to acoustic detections from receivers and ancillary sensors like depth tags. Reconstructing accurate movement trajectories from this limited data is challenging, hindering efforts to assess MPA design and efficacy. Existing heuristic methods often struggle with sparse data, leading to inaccurate maps of space use and residency, which can misinform conservation decisions.

Methods

The authors pioneer the use of state-of-the-art particle filtering-smoothing algorithms for animal tracking. These algorithms, specifically the acoustic-container (AC), depth-contour (DC), and acoustic-container depth-contour (ACDC) variants, build upon Bayesian state-space models. They represent an individual's location probabilistically using weighted "particles." The key innovation lies in their ability to integrate diverse datasets (acoustic detections and depth data) within a unified probabilistic framework. This contrasts with heuristic methods like Center-of-Activity (COA) and Refined Shortest Path (RSP), which rely on simpler statistical summaries and interpolations, often struggling with data gaps. The study also includes a simulation-based analysis to calibrate and evaluate these algorithms against heuristic methods and a null model.

Datasets

The study utilizes data from a passive acoustic telemetry system deployed in Scotland, focusing on tagged flapper skate (Dipturus intermedius). This includes:

  • Acoustic telemetry data (detections of tagged animals by receivers).
  • Archival tag data (depth measurements).

The paper also uses simulated data generated from a best-guess movement model for performance and sensitivity analyses.

Key Insights and Ideas

  • Particle algorithms offer a significant leap in accuracy: They move beyond simple interpolation to a probabilistic representation of animal location, leading to more refined and reliable movement maps.
  • Data integration is key: Combining acoustic and depth data provides a more comprehensive understanding of movement, especially for benthic species like the flapper skate. Depth data, in particular, helps to constrain movements to relevant habitats.
  • Heuristic methods have inherent limitations: They are often driven by receiver locations, struggle with data gaps, and can overestimate residency, potentially leading to a false sense of security regarding MPA effectiveness.
  • Understanding movements beyond receivers is crucial: The particle algorithms can reveal important habitat use and movement patterns in areas not covered by the acoustic array, including fished zones and areas outside MPAs that might warrant protection.

Key Results

  • 5-fold improvement in space use maps: Particle algorithms generated significantly more accurate and precise maps of where flapper skate utilized their habitat compared to heuristic methods.
  • 30-fold improvement in residency estimates: The mean error in estimating how long skate spent in specific areas was drastically reduced (by a factor of 30) using particle algorithms.
  • Improved precision: Residency estimates achieved a precision (standard error) of 7.6% (AC), 3.7% (DC), and 2.8% (ACDC), representing substantial improvements over heuristic methods.
  • Identification of critical habitats: The study identified specific habitat patches beyond MPA boundaries that may warrant protection, including areas used by skate for foraging and potentially for egg-laying.
  • Insight into MPA effectiveness: The analysis suggests that the studied MPA is sufficiently large to protect flapper skate for much of the year, but also highlights the need for broader fisheries management measures.

Limitations and Future Work

  • Sensitivity to parameterization: While generally robust, particle algorithms can be sensitive to parameter mis-specification, particularly the depth observation model. The authors suggest that future studies should conduct similar sensitivity analyses.
  • Computational demands: While particle algorithms are becoming more accessible, they can still be computationally intensive, especially with large datasets.
  • Data quality: The precision of the results is still dependent on the quality of the input data.
  • Future work: The authors propose several avenues for future research:
  • Incorporating additional sensor data (e.g., accelerometers) for more detailed behavioral insights.
  • Expanding acoustic observation models with more receiver deployments and improved detection range testing.
  • Investigating the benthic versus pelagic behavior of skate in complex bathymetric environments.
  • Applying these methods to evaluate existing array designs, explore alternative designs, and determine optimal receiver numbers for specific accuracy targets.
  • Further analyses of skate-fisheries interactions.

Summary updated

Abstract

Abstract The movements of aquatic animals affect their exposure to threats and the efficacy of conservation measures, such as Marine Protected Areas (MPAs). However, many species’ movements remain poorly understood and difficult to reconstruct from available datasets, hampering conservation efforts. This is especially the case for species that rarely surface, for which data are often limited to observations from acoustic telemetry (detections) and ancillary sensors, such as archival tags. Here, we pioneer the use of state-of-the-art particle algorithms to model animal movement, integrate datasets and assess MPA design, using a case study of the Critically Endangered flapper skate ( Dipturus intermedius ) in Scotland. Our algorithms led to 5-fold improvements in maps of space use and 30-fold improvements in residency estimates (lower mean error) compared to prevailing heuristic methods. By formally integrating tracking datasets, we were uniquely able to examine movements beyond receivers into fished zones, MPA-scale residency and specific habitats beyond protected areas that may warrant protection. This work showcases a probabilistically sound modelling framework that is sufficiently fast, flexible and accessible to meet the demands of modern animal-tracking datasets in acoustic telemetry systems. This represents a marked advance for analyses of animal movements and MPA efficacy worldwide.
BibTeX
@misc{lavender2025animal,
  author = {Edward Lavender and Andreas Scheidegger and Carlo Albert and Stanisław W. Biber and Jakob Brodersen and Dmitry Aleynik and Georgina Cole and Jane Dodd and Peter J. Wright and Janine Illian and Mark James and Sophie Smout and James Thorburn and Helen Moor},
  title = {{Animal tracking with particle algorithms for conservation}},
  year = {2025},
  doi = {10.1101/2025.02.13.638042},
  howpublished = {bioRxiv},
}
Figure