Repository logo
Home
Research Outputs
Collections
Statistics
Shared Repository Homepage
  1. Home
  2. Cultural Heritage Shared Repository Service
  3. Royal Botanic Gardens, Kew
  4. Article
  5. A scalable transfer learning workflow for extracting biological and behavioural insights from forest elephant vocalizations.

A scalable transfer learning workflow for extracting biological and behavioural insights from forest elephant vocalizations.

Thumbnail Image
Download
Name

Remote_Sens_Ecol_Conserv_-_2025_-_Pickering_-_A_scalable_transfer_learning_workflow_for_extracting_biological_and.pdf

Description
visibility:open
Size

1.48 MB

Format

Adobe PDF

Checksum (CRC64NVME)

e+Leaabl6NI=

Resource type
Journal article
Creator (person)
Pickering, Alastair
ORCIDORCID logo
Martinez Balvanera, Santiago
ORCIDORCID logo
Jones, Kate E.
ORCIDORCID logo
Hedwig, Daniela
ORCIDORCID logo
Date published
April 25, 2025
Abstract
Animal vocalizations encode rich biological information—such as age, sex, behavioural context and emotional state—making bioacoustic analysis a promising non‐invasive method for assessing welfare and population demography. However, traditional bioacoustic approaches, which rely on manually defined acoustic features, are time‐consuming, require specialized expertise and may introduce subjective bias. These constraints reduce the feasibility of analysing increasingly large datasets generated by passive acoustic monitoring (PAM). Transfer learning with Convolutional Neural Networks (CNNs) offers a scalable alternative by enabling automatic acoustic feature extraction without predefined criteria. Here, we applied four pre‐trained CNNs—two general purpose models (VGGish and YAMNet) and two avian bioacoustic models (Perch and BirdNET)—to African forest elephant ( ) recordings. We used a dimensionality reduction algorithm (UMAP) to represent the extracted acoustic features in two dimensions and evaluated these representations across three key tasks: (1) call‐type classification (rumble, roar and trumpet), (2) rumble sub‐type identification and (3) behavioural and demographic analysis. A Random Forest classifier trained on these features achieved near‐perfect accuracy for rumbles, with Perch attaining the highest average accuracy (0.85) across all call types. Clustering the reduced features identified biologically meaningful rumble sub‐types—such as adult female calls linked to logistics—and provided clearer groupings than manual classification. Statistical analyses showed that factors including age and behavioural context significantly influenced call variation ( < 0.001), with additional comparisons revealing clear differences among contexts (e.g. nursing, competition, separation), sexes and multiple age classes. Perch and BirdNET consistently outperformed general purpose models when dealing with complex or ambiguous calls. These findings demonstrate that transfer learning enables scalable, reproducible bioacoustic workflows capable of detecting biologically meaningful acoustic variation. Integrating this approach into PAM pipelines can enhance the non‐invasive assessment of population dynamics, behaviour and welfare in acoustically active species.
Funder
Funder nameAwards
National Geographic Society, United States
28718
Natural Environment Research Council, Unite Kingdom
NE/S007229/1
Consejo Nacional de Humanidades, Ciencias y Tecnologías, Brazil
2020‐000017‐02EXTF‐00334
Journal title
Remote Sensing in Ecology and Conservation
Article number
rse2.70008
Publisher
Published by John Wiley & Sons Ltd on behalf of Zoological Society of London.
Place of publication
UK
ISSN
2056-3485
eISSN
2056-3485
Date accepted
March 18, 2025
Official URL
https://doi.org/10.1002/rse2.70008
Rights statement
In Copyright
Licence
https://creativecommons.org/licenses/by/4.0/
DOI
10.1002/rse2.70008
Keywords
Loxodonta cyclotis
Bioacoustics
Transfer learning
Passive acoustic monitoring
Behavioural ecology
Population ecology
African forest elephant
Additional information
IF = 2.846 (2023)
Managed by the British Library and supported by the AHRC

Built with DSpace-CRIS software - Extension maintained and optimized by 4Science

  • Cookie settings
  • End User Agreement
  • About
  • Contact
  • Help
Repository logo COAR Notify