From signals to species: Improving acoustic classification reliability for North Pacific false killer whales.

This study presents an automated pre-classification data filtering method that increases confidence in the classification of cetacean acoustic events lacking visual verification, while minimizing bias and reducing the need for manual review. The study tests the approach on false killer whale (Pseudorca crassidens) acoustic data from the North Pacific Ocean. A minimum-data requirement derived from sensitivity analyses, combined with filters based on species-specific acoustic characteristics, defines objective criteria for excluding acoustic events before classification. This framework is a practical tool for systematic, dataset-specific quality control to support the acoustic classification workflow and, ultimately, cetacean population assessment and conservation efforts.