Morandi, I. | Search bioRxiv

bioRxiv2026

BioDCASE: Using data challenges to make community advances in computational bioacoustics

Computational bioacoustics has seen significant advances in recent decades. However, the rate of insights from automated analysis of bioacoustic audio lags behind our rate of collecting the data - due to key capacity constraints in data annotation and bioacoustic algorithm development. Gaps in analysis methodology persist: not because they are intractable, but because of resource limitations in the bioacoustics community. To bridge these gaps, we advocate the open science method of data challenges, structured as public contests. We conducted a bioacoustics data challenge named BioDCASE, within the format of an existing event (DCASE). In this work we report on the procedures needed to select and then conduct useful bioacoustics data challenges. We consider aspects of task design such as dataset curation, annotation, and evaluation metrics. We report the three tasks included in BioDCASE 2025 and the resulting progress made. Based on this we make recommendations for open community initiatives in computational bioacoustics.

animal behavior and cognition↗

bioRxiv2025

Assessing the Effective Range for Individual Acoustic Identification: Comparison of Manual and Automatic Methods

Individual Acoustic Monitoring (IAM), especially when combined with passive acoustic monitoring (PAM), offers a non-invasive alternative to traditional mark-recapture methods to gain insights into species demography. Few studies have examined how identification decreases over distance along with degradation of identity cues. In this study, we conducted a song transmission experiment to quantify the range at which individual Yellowhammers (Emberiza citrinella) can be reliably identified. Songs from ten males (20 songs per individual, covering their full repertoire) were broadcast and re-recorded along a 200 m transect using AudioMoth recorders. Comparing manual classification by human observers with BirdNET classifier adapted for individual identification, we assess how well individuals could be distinguished at increasing distances. Humans were confident in assigning identity to a larger proportion of songs. Where identity has been assigned, both human and BirdNET were highly reliable at short distances (up to 50 m) discriminating even between very similar song types shared among males. At moderate distances (100 - 150 m), simple augmentation boosted BirdNETs performance remarkably, almost matching human classification accuracy. Our results indicate that individual recognition remains reliable up to 100 meters where both accuracy and agreement between assignment methods were high. Our results confirm that automated systems offer promising tools for large-scale, non-invasive individual monitoring, but challenges in accuracy and robustness persist at greater distances. We highlight the difference between signal detectability and individual identification range (much shorter) and its importance for optimizing PAM array designs.

zoology↗