- As scientists and the Chinese government ramp up efforts to protect the critically endangered Hainan gibbon, technology is playing an important part in helping track and monitor the species better.
- In recent years, bioacoustics, infrared technology and machine learning are among the tools that have been used to make data collection and analysis easier in the study of Hainan gibbons.
- According to estimates, there are only 35 or 36 individuals of the species left, limited to Bawangling National Nature Reserve in China’s Hainan province.
Emmanuel Dufourq had a gargantuan task at hand in mid-2019. He had the job of listening to 6,000 hours of audio data collected from Bawangling National Nature Reserve in the southern Chinese province of Hainan, thousands of miles away from his home base in Cape Town, South Africa.
The audio had been collected in 2016 over the course of six months by researchers at the Zoological Society of London in a bid to study and conserve Hainan gibbons (Nomascus hainanus), the rarest primates on Earth and a critically endangered species. Dufourq spent the ensuing months listening to the audio and training a machine-learning algorithm to enable automated identification of gibbon sounds.
“It’s like teaching a child about the world, you show examples saying, ‘this is a cat, and this is a dog,’” Dufourq, resident researcher and the AIMS-Canada junior research chair at the African Institute of Mathematical Sciences, told Mongabay in a video interview. “It’s this repetitive process of showing information to a computer software so that it can learn to identify the gibbons on its own.”
His work seems to have yielded results. According to a 2021 study published by his team in the journal Remote Sensing in Ecology and Conservation, the algorithm was able to identify nearly 80% of the audio segments that contained gibbon calls. A more recent study, published in September 2022 in the journal Ecological Informatics, improved upon the model; now, the algorithm can recognize false positives — sounds mistakenly thought to be gibbon calls — by virtue of being fed human insights gathered through years of field research. For example, the latest algorithm has been taught that Hainan gibbons don’t call at night, enabling the system to realize that a call it detected during night hours probably doesn’t emanate from the species.
Bioacoustics and machine learning, as used by Dufourq, are some of the technological tools that have been deployed in recent years to monitor and study Hainan gibbons, a task that has become especially urgent now. From being widespread across Hainan Island at one point, the population of the species dropped to around seven or eight individuals in the 1970s, with rampant poaching and deforestation the major factors for the decline. Although there are believed to be 35 or 36 individuals currently, the species remains in a precarious position. Conservation interventions led by the Chinese government have ramped up in recent years, but drastically increasing the gibbons’ population size is a challenging task.
“They are really on the precipice of extinction,” Paul A. Garber, professor emeritus at the Department of Anthropology at the University of Illinois, told Mongabay in a video interview. “All apes have very slow periods of development and reproduction, and so we are talking about really long generation times for these primates, which means that, under the best of circumstances, populations can increase only very slowly.”
Technology is now playing a renewed role in keeping track of their small population and studying their habitats in a non-invasive manner. Technological tools have long been important in helping map the habitats and distribution of the species, and these have in turn been used to develop conservation strategies. But methods such as field research, drones, and callback surveys (where researchers play back recordings of gibbon calls to elicit a response) pose a higher risk of disturbing the species in their environment. That’s why more automated and remote-sensing tools are being increasingly deployed.
Machine learning, and broadly artificial intelligence, have also made it easier to identify Hainan gibbons from images captured using infrared camera traps. Previously, the foggy conditions in the forest combined with the black and yellow fur color of Hainan gibbons often led to poor image detection, causing scientists to miss images with gibbons in them and making it difficult for them to spot the species. A 2022 study published in the journal Global Ecology and Conservation laid out how computer vision, a subfield of artificial intelligence, helped defog the images and filter them to improve the accuracy of detecting Hainan gibbons.
Much like in Dufourq’s work, bioacoustics forms the foundation of a collaboration by the IUCN, the global wildlife conservation authority, and Chinese technology giant Huawei to study and protect Hainan gibbons. In a press release, Huawei said the combination of acoustic monitoring and artificial intelligence has led to a recognition accuracy of 89.2%. The team is now working to set up a framework to develop a system that will “establish a unique voice ID for individual gibbons.”
Dufourq is also working on improving his own algorithm to enable real-time monitoring and analysis.
“Now, the way that a lot of these acoustic studies are working is that someone will go into the field and conduct a survey over six months or one year,” he said. “But the processing happens in the future, and that means that the data is already historical data. So the future of this is to do real-time monitoring.”
All tools on the table, experts say, should be developed and deployed to protect the species. For years, Garber has been working with a team of researchers in China to study Hainan gibbons. His colleagues use tools such as infrared cameras and DNA testing to track the gibbons and understand more about their habitat and behavior. While he said he strongly believes there’s no substitute for being out in the field and observing the animals, in the case of Hainan gibbons, however, technology is the next best tool out there. Apart from enabling scientists to streamline the tedious data-gathering processes and analyze copious amounts of data, Garber said, technology also plays a role in raising awareness about the need for conservation.
“It’s critical because even if we don’t get a lot of newer, important information using some technologies, using them somehow highlights the research, and gives it a profile,” he said. “The impact is higher than if you are just walking in the forest with a notebook.”
Abhishyant Kidangoor is a staff writer at Mongabay. Find him on Twitter @AbhishyantPK.
Dufourq, E., Durbach, I., Hansford, J. P., Hoepfner, A., Ma, H., Bryant, J. V., … Turvey, S. T. (2020). Automated detection of Hainan gibbon calls for passive acoustic monitoring. Remote Sensing in Ecology and Conservation, 7(3), 475-487. doi:10.1002/rse2.201
Dufourq, E., Batist, C., Foquet, R., & Durbach, I. (2022). Passive acoustic monitoring of animal populations with transfer learning. Ecological Informatics, 70, 101688. doi:10.1016/j.ecoinf.2022.101688
Wang, X., Wen, S., Niu, N., Wang, G., Long, W., Zou, Y., & Huang, M. (2022). Automatic detection for the world’s rarest primates based on a tropical rainforest environment. Global Ecology and Conservation, 38, e02250. doi:10.1016/j.gecco.2022.e02250