Skip to main content
Using accelerometer-based behavioral classification to enhance scavenger conservation

Using accelerometer-based behavioral classification to enhance scavenger conservation

This is a Preprint and has not been peer reviewed. This is version 1 of this Preprint.

Add a Comment

You must log in to post a comment.


Comments

There are no comments or no comments have been made public for this article.

Downloads

Download Preprint

Supplementary Files

Authors

Gideon Vaadia, Marta Acácio, Tal Agassi, Nili Anglister, Yigal Miller, Ohad Hatzofe, Patricia Mateo-Tomás, Jorge Rodríguez-Pérez, María Fernández-García, Pedro P. Olea, Ignacio Otero, Noa Pinter-Wollman, Moni Shahar, Orr Spiegel

Abstract

Human activities are endangering animal species globally and implementing effective conservation strategies requires understanding animal behavior and ecology. Technological advancements in GPS tracking technology, accelerometry, and machine learning algorithms are now making it possible to study animal movement and behavior remotely. However, due to the challenge of building supervised machine learning algorithms and collecting the large datasets required to train them, the use of these algorithms is still not common practice. Additionally, after building the algorithms, their reliability in classifying unobserved behaviors is rarely validated, resulting in possible classification errors. Here, we built a supervised accelerometer-based behavioral classification model for griffon vultures (Gyps fulvus). This scavenger is critically endangered in Israel and neighboring countries, mostly due to mass poisonings at carcass feeding events. In fact, poisoning is one of the main threats to scavenger populations worldwide. Thus, identifying this scavenger’s feeding behavior and foraging areas is crucial for their conservation. We trained a random forest model on acceleration data of 14 captive and 17 free-roaming griffons. We collected 5783 behavioral observations grouped into 6 distinct classes: feeding, lying, standing, other ground behaviors, flapping and soaring flight. The classification model performed well (0.96 accuracy, 0.89 precision and 0.82 recall) and, importantly, feeding behaviors were accurately classified (0.87 precision, 0.92 recall). Importantly, we calculated an observation-specific confidence score and demonstrated its effectiveness (for all but one of the behavioral classes) in identifying true- and false-positive classifications, in both captive and free-roaming individuals. Further, our classification model enables us to identify vulture feeding hotspots, potentially aiding the implementation of conservation actions related to carcass management. Finally, our training dataset and model are provided in a user-friendly platform and accompanied by a conceptual framework, to encourage use by ecologists and conservation practitioners overcoming the data-analysis challenges involved in this powerful approach.

DOI

https://doi.org/10.32942/X2WK9Z

Subjects

Behavior and Ethology, Life Sciences

Keywords

accelerometer, Behavior classification, Random Forest, Griffon Vulture, poisoning, conservation, Biotelemetry, Supervised machine learning

Dates

Published: 2025-07-07 18:48

Last Updated: 2025-07-07 18:48

License

CC-By Attribution-NonCommercial-NoDerivatives 4.0 International

Additional Metadata

Conflict of interest statement:
None

Data and Code Availability Statement:
All the training data, the code necessary to train and build the algorithm and a tutorial are publicly available on GitHub (www.github.com/Orrslab/ACC_behavior_classification).

Language:
English