Annotated Bird Soundscapes from Sierra Nevada, 33 Hours of Recordings

33 hour-long soundscape recordings from the Sierra Nevada, California, annotated with 20,147 bounding box labels for 56 bird species. The data were recorded in 2018 by researchers from Cornell University and partially featured in the 2021 BirdCLEF competition. Passive acoustic surveys were conducted in the Lassen and Plumas National Forests using SWIFT recording units.

Use Cases

• Training bird call detection models based on time-frequency bounding box annotations.
• Evaluating species classification algorithms using the 56 distinct eBird species codes.
• Studying avian population density and diversity in relation to forest management based on the described survey methodology.
• Benchmarking audio event detection systems on a dataset used in the BirdCLEF competition.

Strengths

• Contains 20,147 high-precision bounding box annotations for bird calls.
• Covers 56 different bird species from a defined geographic region.
• Audio recordings are 33 hours long, captured with detailed technical specifications (e.g., 32 kHz sampling, -44 dB sensitivity).
• Annotations use a standardized taxonomy (2021 eBird/Clements list).

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Data may reflect geographic and temporal bias, being limited to the Western U.S. in 2018.
• Annotation was performed by a single annotator, which may introduce consistency bias.

Provenance

Source

Cornell University

Collection Method

Passive acoustic surveys using SWIFT recording units deployed in randomly selected grid cells within national forests.

Time Range

2018 (May-August)

Geography

Lassen and Plumas National Forests, Sierra Nevada, California, USA (39.41 – 40.71°N, ~840 – 1700 m elevation)