sammlapp/Alberta_SBT_2016_REVI_Localized

Red-eyed Vireo localized songs

Creators: Sam Lapp (sam.lapp@pitt.edu) [1], Scott J. Wilson [2], Erin Bayne [3], and Justin Kitzes [1]

Affiliations: [1] University of Pittsburgh, [2] Government of Alberta, [3] University of Alberta

Version 1.1

Date Updated: 2026-09-22

DOI: not yet assigned

General characteristics

audio format: 10 second .FLAC clips starting 4 seconds before localized events dimensions localized: 2
number of localization arrays: 13
array geometry: 5x10 square grid with 33 m spacing
sounds localized: Red-eyed Vireo songs
number of audio files: 910
size: 88 Mb

Study description

This dataset includes spatially localized Red-eyed Vireo (Vireo olivaceus) songs. The data were collected to study acoustic distance estimation and habitat attenuation.

We used automated sepecies detection (HawkEars v0.1.0) and acoustic localization (OpenSoundscape) to localize Red-eyed Vireo singing events from microphone arrays in Alberta, Canada.

Personnel: Wilson and Gregroire collected the data, which was published in Wilson and Bayne 2018 [2]. Lapp created this dataset of localized sounds.

Files

Format follows the Data Loca standard in formatting, column names, and file structure

In addition to the ./localized_events.csv table with one row per localized song, we also include localized_clips.csv, which contains the localized events in a 'long' format with one row per clip rather than one row per event.

audio clips are contained in the /audio/ folder and referenced in /localization_metadata/audio_file_table.csv.

Scripts are included for reference, rather than to reproduce the localization results. In particular, (1) the original PAM data is not provided; (2) included clips are only those which passed the review process and had the target sound unmasked by other sounds; as a result, the localization process cannot be reproduced with the audio provided in this dataset.

conda_env.yml contains the package specifications for a conda env which can be used in associated with python scripts and notebooks

Sites

Restored well pad sites across Alberta, Canada (see details in [1])

Hardware

• Recorder source: Wildlife Acoustics
• Recorder model: SongMeter 3 with additional wired external microphone recording to 2nd channel
• Firmware version: N/A

Recording properties

Date range of data: [2016-05-26] to [2016-06-29]

Recording schedule description:

Times of recording: 05:00-08:30 local time Sleep-wake schedule: 29 minutes starting every half-hour Sample rate: 48000 Hz

The gain setting was 19.5 dB. Combined with a microphone sensitivity of the SMM A1 microphone of -11 dB (V/Pa) and an ADC with .775V rms = 0 dBFS (=1V peak), the expected end-to-end sensitivity is 8.5 dBFS @ 1 Pa (=94dB SPL re 20 uPa), in other words dB SPL re 20 uPA = dBFS + 85.5 dB.

Recorder positioning

Placement: typically 5x10 m arrays with ~33m spacing, but varies by array; see details in [1] Spatial pattern or geometry: 10x5 square grid of recorders Range of spacing between adjacent mics: avg 33m Dimensions of array: approximately 130x300

Coordinate Reference System:
Microphone locations and localized events are provided in meters relative to the Western-most and Southern-most coordinates of any microphone in the array (./localization_metadata/point_table.csv). The origin of each array (./localization_metadata/array_origin_coordinates.csv), and a version of the point coordinates as absolute coordinates (./localization_metadata/point_table_absolute.csv), are provided in the local UTM zone. Note that the arrays spand both UTM Zones 11N (NAD 83, EPSG 26912) and 12N (NAD 83, EPSG 26911).

Deployment: See [1]

Position measurement: See [1]

Synchronization

The SM3 recorders perform on-board precise time alignment, so that post-processing for accurate temporal synchronization across files is not required.

Sound detection

Red-eyed vireo songs were detected using the HawkEars regional classifier version 0.1.0 [2], via the Bioacoustics Model Zoo.

Post-processing detector outputs: - threshold score of -1.0 for Red-eyed Vireo class - only clips with detection were included in localization process

Localization

OpenSoundscape localization module, with hyperbolic TDOA localization.

Localization algorithm: SoundFinder Time delay calculation algorithm: GCC-PHAT References: Freeland-Haynes et al [3]

Parameters: localization algoritnm: soundfinder min n receivers for event: 5 max receiver distance from reference receiver: 80 m cross correlation threshold: 0.01
Bandpassed audio to 2.4-3.4 kHz prior to cross correlation to target Red-eyed Vireo song

We also performed an automated error-rejection post processing procedure (/scripts/4_minspec_filtering.py)

Manual review

Each localized event was reviewed to confirm the localized acoustic event was a Red-eyed Vireo song, and that cross correlation had correctly aligned the song across microphones. Additionally, the manual review process involved annotating the section of audio for which the song was not obscured by louder non-target sounds in each clip. Clips for which the Red-eyed vireo song was masked by louder sounds were excluded from this dataset.

Observational data

None

Acknowledgements

Scott Wilson, Jocelyn Gregoire, and Erin Bayne collected the data. Tessa Rhinehart, Elly Knight, Dan Yip, and Brandon Edwards participated in methodological development, conceptualization, and data analysis.

License

CC-BY 4.0 https://creativecommons.org/licenses/by/4.0/

Work Cited and Links

[1] Wilson, Scott J., and Erin M. Bayne. "Songbird community response to regeneration of reclaimed wellsites in the boreal forest of Alberta." Journal of Ecoacoustics 3.1 (2019): 1-11.

[2] Huus, Jan, et al. "HawkEars: A regional, high-performance avian acoustic classifier." Ecological Informatics 87 (2025): 103122.

[3] Freeland-Haynes, Louis, et al. "A fully automated framework for acoustic identification and localization of terrestrial wildlife at scale." Communications Biology (2026).

Version log:

v1.1: updated audio from .mp3 to .FLAC, generated from the original uncompressed .WAV files.