Post

A study by Roodbergen et al. in Bird Census News examined the comparability of Passive Acoustic Monitoring (PAM) and automatic classifiers to traditional monitoring methods for tracking common bird species in Dutch agricultural landscapes. More: www.ebcc.info/wp-content/u...

European Bird Census Council

@ebcc.info

Audio data and point counts overlapped significantly (48%), especially for songbirds in closed habitats identified by vocalisations. More species were recorded during point counts (visual). Some species were more in audio,likely omitted as fly-overs/outside radius or missed when focusing on other sp

14/5/26, 4:03 pm0 reposts1 like

European Bird Census Council

@ebcc.info

14 May

Two classifiers (BirdNET, GBVC) performed best but didn't achieve near-perfect accuracy. Despite rapid improvements, manual validation is crucial. The customised model had performance gains, showing the value of tailored training data.

European Bird Census Council

@ebcc.info

14 May

PAM and AI classifiers can be powerful complementary tools, but so far, not a replacement for traditional surveys. The most effective approach would be a hybrid system that combines model predictions with expert validation (“human-in-the-loop”) to improve accuracy over time.