One among the ecosystems that are subjected to experience threats from climate change is montane forests, also called biodiversity hotspots.
Automatic recorders have been set up in Yushan National Park, Taiwan, by scientists to understand the possible effects of climate change on birds in such forests. Also, they have come up with an AI tool that has been specifically developed for species identification with the help of bird sounds. Their aim is to examine the status and trends in animal activity via acoustic data.
Professor Hsueh-Wen Chang and Ph.D. Candidate Shih-Hung Wu from National Sun Yat-Sen University, Taiwan, Dr. Ruey-Shing Lin, Assistant Researcher Jerome Chie-Jen Ko from the Endemic Species Research Institute, and Ms. Wen-Ling Tsai from Yushan National Park Headquarters has reported a paper in the open access journal Biodiversity Data Journal, detailing their use of AI to detect 6 million bird songs.
In comparison to conventional observation-based techniques, passive acoustic monitoring with the help of automatic recorders to capture wildlife sounds offers an economical, long-lasting, and systematic alternative for long-term biodiversity tracking.
The authors employed six recorders in Yushan National Park, Taiwan, a subtropical montane forest habitat with elevations varying from 1,200 to 2,800 m. From 2020 to 2021, they recorded almost 30,000 hours of audio files with ample biological information. But examining this huge dataset is a difficult task and hence cannot be done only by humans.
For this difficulty to be addressed, the authors made use of deep learning technology to come up with an AI tool known as SILIC that has the potential to determine species by sound. SILIC could rapidly identify the accurate timing of every animal call inside the audio files.
Following various optimizations, the tool is currently capable of identifying 169 species of wildlife native to Taiwan, such as 137 bird species, as well as mammals, frogs, and reptiles.
In this study, authors made use of SILIC to withdraw 6,243,820 vocalizations obtained from seven montane forest bird species with a high accuracy of 95%, thereby making the first open-access AI-examined species occurrence dataset available on the Global Biodiversity Information Facility.
This is known to be the first open-access dataset with species occurrence data withdrawn from sounds in soundscape recordings through artificial intelligence.
The dataset reveals elaborate acoustic activity patterns of wildlife throughout both short and long temporal scales. For example, in diel patterns, the authors determine a morning vocalization peak for all species.
On a yearly basis, the majority of the species have a single breeding season peak, but a few, like the Gray-chinned Minivet, exhibit a secondary non-breeding season peak, probably concerning flocking behavior.
Since the monitoring projects continue, the acoustic data might help to comprehend changes and trends in animal behavior and population throughout the years in an affordable and automated way.
The authors predict that this broad wildlife vocalization dataset will not be beneficial only for the National Park’s headquarters as far as decision-making is concerned.
“We expect our dataset will be able to help fill the data gaps of fine-scale avian temporal activity patterns in montane forests and contribute to studies concerning the impacts of climate change on montane forest ecosystems.”
Wu, S-H., et al. (2023) An acoustic detection dataset of birds (Aves) in montane forests using a deep learning approach. Biodiversity Data Journal. doi.org/10.3897/BDJ.11.e97811.