Chirp Check

Help Us Perfect Birdsong ID with AI!

Birds are fun to watch, lift our mood with their songs, and play critical roles in ecosystems. Together, all of these activities make up the ecosystem services birds provide that benefit humans. Although ecosystem services are incredibly important for human health, well-being, and economies, we know surprisingly little about how birds provide them. This is especially true in cities where ecosystem services are in high demand and birds are one of the few animals providing ecosystem services to humans. The first step to studying how birds provide ecosystem services is understanding which bird species are in certain areas.

The Integrative Ecology Lab at Temple University in Philadelphia, PA, USA, is studying bird species in cities and surrounding areas to better understand how bird species provide ecosystem services. To study bird species, we use data collected from audio-recording devices that capture thousands of hours of birdsong and artificial intelligence (AI) to identify which bird species are present in the recordings. However, AI has its limits, and that’s where you come in! By verifying the species identifications that the AI made, you’ll help us enhance the accuracy of the data and advance bird research.

Research Goals

With your help, we will generate high-quality data to address the following research questions:

• Is the BirdWeather Audio Library usable for research?

• How can we responsibly incorporate AI into avian research?

• How do bird species vary across urban environments?

• Are the bird species that provide ecosystem services present where they’re needed most?

• How can we better design cities to provide high quality habitat for bird species?

Data

The data for this project were collected using BirdWeather PUCs (Portable Universe Codecs). PUCs are small, weatherproof devices that record sounds and collect environmental data. PUCs upload recordings to the internet, where BirdNET, an AI machine learning program, analyzes and identifies the bird songs to species. Alongside each species identification, BirdNET provides a confidence score indicating how likely the identification is accurate.

Photo sourced from BirdWeather's Website

BirdWeather has an amazing open library of bird sounds from around the world, and we will make use of this library in our project. To start, the data in our project originate from the greater Philadelphia region in southeastern Pennsylvania.

To learn more about BirdWeather PUCs, visit BirdWeather’s website.

To learn more about BirdNET, visit BirdNET’s website.

Why We Need You

BirdNET is a powerful tool for identifying bird species, but some identifications—like mistaking a dog bark for a Canada goose—require human verification. By confirming a small percentage of recordings, you’ll help us assess BirdNET’s accuracy for each species. With these data, we’ll develop thresholds to more reliably filter out incorrect identifications, saving time and enhancing the quality of our research.