BirdNET is a popular machine learning tool for automated recognition of bird sounds. Here we evaluate how BirdNET settings affect the model performance both at vocalization and species levels, using 4,225 one-minute recordings from 67 recording locations worldwide.

Giving equal importance to recall and precision, a low confidence score threshold (0.1-0.3) appears optimal for detecting bird vocalisations, whereas higher thresholds (around 0.5) are more suitable for characterising bird communities.

Based on our findings, we recommend increasing the Overlap parameter from its default value of 0 seconds to 2 seconds, as this consistently improves BirdNET performance in detecting both individual vocalisations and species presence.

The effect of the Sensitivity parameter varied across regions. However, a value of 0.5 maximises global performance for community-level analyses across all confidence thresholds, while a value of 1.5 generally yields better results for vocalisation-level studies, particularly at low confidence thresholds.