Image
barn owl

Expand Bioacoustic Data Collection

Scaling data collection of animal sounds in California to improve conservation efforts for key species

The Latest

  • Developed new AI models to detect species of ecological and cultural significance, including coyotes and frogs
  • Launched Jupyter Bioacoustic: an open source toolkit for scientists to build bioacoustics pipelines

 

Our Impact

 

Soundhub

Soundhub is a remarkable web tool that identifies species using bioacoustic sound data (i.e. animal calls). Bioacoustic data is critical for understanding wildlife abundance and distribution, can signal changes in environmental health over time, and informs conservation decisionmaking. The tool can passively collect, sort, and classify terabytes of audio to detect species. This provides an exciting and time-saving approach to data collection at scale rather than relying on field observations. 

 

We are partnering with the UC Berkeley Geospatial Innovation Facility (GIF) and California Department of Fish and Wildlife to expand Soundhub so that it becomes the official home for all bioacoustic data across California, including the California Sentinel Site Network. These sites are in specific geographic areas where long-term, standardized monitoring  and instrumentation is installed and maintained to accomplish state climate and biodiversity monitoring goals. 

 

DSE's roles in this effort are to develop targeted AI models to identify vocalizations from frogs, birds, coyotes, and other species of interest, as well as anthropogenic sounds (i.e. vehicle noise) and natural environmental sounds (i.e. wind and rain). Simultaneously, we are helping to integrate Indigenous Data Sovereignty parameters to ensure the tool protects the privacy of and data from Indigenous users. 

 

Species of Interest

barn owl coyote    frog    wolf

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Above: DSE is training AI models to detect key species within Soundhub. Clockwise from top left: close up of a barn owl looking at the camera, a coyote standing on red and beige shrubs, a gray wolf standing on river rocks, and wood frog on a fallen log.

 

Jupyter Bioacoustic

We also recently developed Jupyter Bioacoustic: new open source toolkit that provides modular tools for building bioacoustics pipelines within the Jupyter ecosystem. Developed as a flexible extension of Soundhub, Jupyter Bioacoustic addresses a key missing piece in bioacoustic workflows: its interface enables an entire analysis workflow in one place. It also allows those concerned with privacy (i.e. Indigenous partners) to run analyses locally without relying on external infrastructure. The tool is complete and entering user testing, with the end goal of making it broadly available to anyone working within the Jupyter ecosystem.

 

screenshot of Jupyter bioacoustic
The Bioacoustic Annotator tool (shown above) functions within the Jupyter lab environment. It allows researchers and scientists to filter, label, and annotate data, and enables model training, validation, and output analysis. This occurs in a single, reusable, maintainable, pipeline so that researchers don't have to bounce back and forth between various coding environments or applications. Image description: a computer screenshot featuring a black background, names of example audio files, and a visual representation of the audio file frequency depicted in purple and orange.

 

 

Future Vision

We are aiming to integrate our new AI models into Soundhub in the coming months. Moreover, we'll continue the Jupyter Bioacoustic user testing phase as we work towards the tool's broader release. 

Partner Feedback

Matthew Toenies headshot

Thanks to Schmidt DSE's work on incorporating new sound identification models into the Wildlife Sound Hub, we will gain even more environmental information from the state’s acoustic monitoring efforts, and can more efficiently and effectively manage California’s diverse nature for the benefit of all.” 

- Matthew Toenies, Senior Environmental Scientist, California Department of Fish and Wildlife

DSE Contributors

  • Image
    Amy Van Scoyoc

    Amy Van Scoyoc

    Postdoctoral Researcher
    Eric and Wendy Schmidt Center for Data Science and Environment at Berkeley
  • Image
    Michael Catchen headshot

    Michael Catchen

    Data Scientist / Research Software Engineer
    Eric and Wendy Schmidt Center for Data Science & Environment
  • Image
    Ciera Martinez

    Ciera Martinez

    Senior Program Manager
    Eric and Wendy Schmidt Center for Data Science & Environment at Berkeley
  • Image
    Justin Brashares

    Justin Brashares

    Faculty Advisor & Professor
    Environmental Science, Policy and Management at UC Berkeley