A smartphone sits on a rickety, uneven piece of wood. There is nothing but silence in the forest.
Cell towers don’t make it this far down the valley. Identifying animals based on their sounds was laborious before.
In order to identify an animal’s call, you had to carry around recording equipment, field guides, and reference books.
After returning to your desk computer, it took hours upon hours of manual research to analyze each recorded file.
Developed by the Cornell Lab of Ornithology and Chemnitz University of Technology, this new mobile app does its processing entirely offline.
The app listens and identifies hundreds of different types of animal vocalizations within seconds, all without ever connecting to the Internet.
The forest no longer needs a signal
Using machine learning algorithms directly on mobile hardware instead of sending information to cloud-based servers to be processed makes a huge difference in field research.
With local processing, researchers have access to instant results in the field as opposed to waiting weeks or months after they leave for an expedition.
The app converts the raw environmental sound into a visual representation called a “spectrogram.”
Thousands of species’ vocal patterns were previously recorded and are used by the neural network to determine which species produced specific vocal patterns.
Since there is no need to upload data via the internet, researchers have more time in the field.
They can focus on tracking and studying migratory patterns through dense jungles, alpine heights, and remote wetlands.
The app presents identification results directly in the field, allowing researchers to review detections while surveys are still underway.
Researchers now have the ability to accurately distinguish single calls from multiple calls made by many different animals in proximity.
Naturalists no longer have to rely exclusively on guesswork when trying to identify unfamiliar animal sounds.
Tracking behavioral patterns across diverse landscapes
The use of machine learning and local processing also allows researchers to gather data for large-scale conservation projects developed by the Cornell Lab of Ornithology.
Automated point counts allow researchers to document species activity while recording the exact location where detections occur.
By utilizing local processing capabilities, researchers can build a clearer picture of how multiple species use landscapes over time.
With access to larger amounts of data than ever before, researchers can spend less time managing recordings and more time reviewing results.
Researchers can continue collecting information in places where reliable connectivity is unavailable.
Field researchers are able to review results immediately instead of waiting for later analysis.
Continuous monitoring allows researchers to observe how animal activity changes over time.
The resulting data documents vocal activity across different times of day and night.
Ecologists now have the opportunity to view animal interactions over longer periods than traditional daytime spot-checks alone.
Every recording adds to a bigger picture
Bioacoustics enables monitoring across broad regional expanses.
Historically, most species tracking has relied on visual detection.
However, these traditional methods fail to account for nocturnal and cryptic or highly canopy-dense species that exclusively utilize sound as their primary communication tool.
Acoustic monitoring greatly expands the scope of conservation monitoring by allowing ordinary mobile devices to contribute useful field observations.
When conservation organizations use tools that operate independently of cellular connections, remote locations become much easier to monitor.
The technology opens doors for grassroots volunteer communities. It allows them to use consistent documentation methods during field deployments across protected reserve lands.
Once recordings are reviewed and shared, they can contribute to broader biodiversity research efforts.
Conservation trends documented longitudinally become increasingly reliable.
Putting high-level intelligent capabilities into every day mobile devices closes the gap between observational study and rigorous scientific investigation.
Read the whole thing?
Get the week's signal, not the noise
Our sharpest reporting on energy, climate and nature — free, once a week.
