Machine Learning for Bat Acoustics
Description
Improving the accuracy and timeliness of species status assessments and science to support deregulation efforts.
Detailed example
This model will be refined and used for prediction and decisions.
AI / analytics pattern
Other
Automation level / stage
b) Pilot – The use case has been deployed in a limited test or pilot capacity.
Expected benefit
This effort will result in significant cost savings related to environmental review for permitting.
Controls / human review
ATO: No; PIA: Not published
Data needed
This dataset contains audio files of bat echolocation calls used to develop NABat ML algorithm V1.0, excluding the test (holdout) set. Recordings were collected by monitoring partners across North America using ultrasonic recorders for stationary and mobile surveys, then post-processed to remove noise and assign species labels. Labeling typically involves automated classification followed by manual review (see NABat guides). Files were submitted in WAV format and include 35 classes (34 species + noise), with 4 species excluded for low sample size. From this pool, recordings were randomly selected and split into training and validation sets; the test set is not included. Audio files are grouped by four-letter species codes, with a reference dataset providing Family, Genus, Species, and Common name definitions.