Accounting for noise improves the generalization of acoustic indices when estimating biodiversity

2026-08-04

Samuel Hugo Suárez-Ronay *

Katarina Biljman

Jonathan Belmaker

Yossi Yovel

DOI：10.1016/j.ecolind.2026.115304

Original

Original request for help

Save

Abstract

En 中文

• Noise limits the generalizability of acoustic index-based biodiversity estimation. • Selecting noise-invariant indices improves generalization across datasets. • Training across noise types improves generalization to unseen soundscapes. • Soundscape segmentation improves biodiversity estimation across noise types.

Keywords:

Artificial simulations

Background noise

Bioacoustics

Field recordings

Generalization

Soundscapes

Transfer-learning

Accounting for noise improves the generalization of acoustic indices when estimating biodiversity

Abstract

Journal

Organization

Cited Papers

Citing Papers