- From thermal cameras to deep learning AI, researchers are reinventing how they study primates in the wild.
- What began with Jane Goodall’s observational notes has evolved into artificial intelligence that identifies chimpanzees and decodes their social lives.
- A custom-built “dronequi” with a thermal and a high-definition camera is helping scientists spot Brazil’s elusive and endangered muriquis from above the trees.
- Hidden microphones across Borneo’s rainforests capture haunting gibbon duets, revealing clues about hybridization and habitat loss.
Riding in a helicopter, Fabiano Melo, with the Federal University of Viçosa, Brazil, searched for the elusive northern muriqui (Brachyteles hypoxanthus), a large monkey in the northern Atlantic Forest of Brazil. Flying over evergreen canopies with trees as tall as buildings, the researchers tried to spot light brown monkeys moving between branches. Finally, after almost one hour, they found one.
Seeing the helicopter, the monkey started shaking the branches fiercely. It then tried to move out of sight, jumping from branch to branch as if doing parkour. At the same time, Melo took video of the monkey’s agile movements as it disappeared into the canopy. Even though the animal got away easily, the researchers now had an idea of where to look for them.
“The problem is that they have very low densities … and if you go inside the forest and try to track animals [by] walking, it’s almost impossible to find them,” Melo told Mongabay.
Primatologists — scientists who study nonhuman primates such as apes, monkeys and lemurs — once considered the northern muriqui among the world’s 25 most endangered primates. Habitat loss, isolated populations and hunting have made this elusive species critically endangered. If Melo wanted to find the muriquis to understand their distribution, he needed to improve his tactics fast.
New technologies gave Melo a possible solution: fewer helicopters, more drones. In 2017, Melo started to build the “dronequi,” a customized drone with high-definition and thermal cameras, allowing him to find and follow the monkeys more effectively at a lower cost.
“The cost-benefit is much better than helicopters. … For the first drone, I used to fly like 10 minutes. Now I can fly 40 minutes using one battery,” Melo said. “The technology is improving so fast, and it’s helping us [so] much that we are getting much better results than in the last five years.”
Scientists have taken high-tech gadgets to remote areas of the world to study the secret lives of primates, making a transition from pure observational studies to new approaches that are revolutionizing the field of primatology.
Old-school primatology
In his book The Expression of the Emotions in Man and Animals, Charles Darwin was one of the first naturalists to compare monkeys’ and apes’ behaviors with humans. Feelings such as joy, pain and grief could be seen in both.
“Any one who has watched monkeys will not doubt that they perfectly understand each other’s gestures and expression, and to a large extent … those of man,” Darwin wrote. However, it was anthropologists, not biologists, who first studied our closest relatives.
Legends of the field such as Jane Goodall, Frans de Waal and Dian Fossey started to rise in the mid-1900s. They immersed themselves in the middle of forests around the world, studying the behavior and social structures of their chosen species for years. The researchers discovered behaviors such as reconciliation, cooperation and empathy among primates, redefining our understanding of primates. They also discovered the use of tools and self-cognition in nonhuman primates, challenging human-centric views of intelligence.
Today, young conservationists following the steps of their heroes are using brand-new tools that allow them to keep pushing the limits of knowledge in the field.
Here come the drones
The rapid development and accessibility of technology have opened up new ways of solving long-standing questions about primates. Melo can now look to a versatile market to update his drones and their gadgets, improving the data he collects. Through high-definition resolution thermal cameras, Melo can more accurately count individuals, recognize sex, identify faces and capture footage. Melo can also study the primate’s favorite fruit and leaves.
“The technology is very nice today, but we still have many challenges,” Melo said.
Finding the elusive muriquis with drones in the remote mountains of the Atlantic Forest is still no easy task. The montane terrain can interrupt the signal connection between Melo and the drone, leaving him to fly blindly in autopilot.
“You have to be very confident … sometimes spend like hours looking, walking or driving to find the best [takeoff] place to fly,” Melo said, “This is our main challenge today, to have a drone that you can trust and the drone can go far.”
Nowadays, assembling a dronequi costs Melo around $40,000. While expensive, Melo’s dream is far more ambitious. He hopes one day to be able to look for muriquis using an eVTOL (electric vertical takeoff and landing) aircraft, a new generation of electric helicopter made for short-range travel.
“I don’t like to fly autonomously, because when you are looking for animals, you have to pay attention. … It’s much better to find animals flying manually,” Melo said. “But we already have the solutions to fly … with many drones working together and searching in big areas.”
Using AI to ID chimpanzees
Daniel Schofield, Schmidt AI in Science Fellow at the Visual Geometry Group at the University of Oxford, U.K., is also using groundbreaking technology to study primates. From his office, Schofield, looks on his laptop at videos of wild chimpanzees from Bossou, Guinea. The chimps are a renowned population famous for using stone tools to crack open nuts of palm trees.
Instead of working in the field like Goodall, Schofield writes code to train AI in tagging recognized individuals and behaviors.
“My fieldwork was working in a room in Japan, digitalizing tapes and collaborating with researchers for many, many hours,” Schoffield joked. “So very fortunately, the data had been collected in the field over many years.”
Countless hours of video data from camera traps — spanning more than 30 years — awaited analysis at Kyoto University, leaving a bottleneck in manually identifying individuals and tracking their behaviors. So, Schofield teamed up with Arsha Nagrani, currently a senior research scientist at Google AI Research, and started to develop methods for detecting chimpanzees, recognizing individuals and the actions they do as they move in a video frame.
“The challenge is, you need enough training data, you need a lot of images and you need lots of variation,” Schofield told Mongabay.
To make a robust model, Schofield extracted images from the videos, made notes on them, called annotations, and labeled the information on the images. The vision algorithm takes this information as a “ground truth” in the learning process, which will try to perform under more challenging and dynamic scenarios in the wild.
“You have lighting problems, you have variation in the background, fast-moving animals. … So, we had a number of techniques to try and get past those problems,” Schofield said.
When Schofield started working on this in 2016, he did not realize how much this field would take off.
“I think part of why I find it so exciting is it’s the challenge of deploying it in the real world. You have to be very creative in how you’re using these systems and the limitations,” he said.
Today, the algorithm designed by Schofield and collaborators can make automated deep face recognition of individuals from video footage and assign individual-level networks inside the social structure of four chimpanzee groups. The data can help highlight how the individuals interact with each other, helping primatologists understand sociality in wild apes at a low cost and high efficiency.
To Schofield, this is just the beginning. Obtaining data from a place like Bossou, where the population is habituated and field researchers know the individuals, makes his work a lot easier.
“But new information is coming all the time, so there might be infants born, individuals you haven’t seen come into your study area,” Schofield said. He is now mixing video and audio inputs into his algorithms to make the systems more robust. For example, if a chimpanzee moves behind a tree, the system is still able to make notes on any call or sound that it makes, despite not having direct visual contact.
Schofield is moving toward more open data that make the model more adaptable and powerful. He hopes one day the algorithm can extract new information from the images to self-learn and train itself, helping it, for example, to recognize new individuals that arrive in the group.
“There’s still a way to go, I think, for so much potential and growth in this area,” Schofield said enthusiastically.
Rethinking the past with bioacoustics
Jorian Hendriks, a former master’s student at Wageningen University, Netherlands, has walked the forests of Barito Ulu in Central Kalimantan, Indonesian Borneo, in search of gibbons, a rare ape to see but very easy one to hear. Hendriks started installing sound recorders in their habitat to capture what primatologists have dubbed “the great call,” a unique vocalization, specific to each of the 20 species of gibbons.
“The great call is known as a duet where the male and the female sing together. … You can almost see it as … sexual foreplay,” Hendriks told Mongabay.
Since 1976, primatologists have reported sounds that were something like an intermediate call in the shared distribution of the Bornean white-bearded gibbon (Hylobates albibarbis) and Müller’s gibbon (Hylobates muelleri). Hendriks is applying bioacoustics to test whether the two species are crossbreeding to produce hybrid offspring.
No research to date has proven there are hybrids. Genetics would be able to solve the mystery by comparing the DNA of the so-called hybrids with their related parent species, but gibbons are elusive and live in small groups, making it difficult to have a representative number of samples to test the hypothesis. That’s where bioacoustics comes in.
“Literature suggests this great call is genetically inherited,” Hendriks said. “Calls can still vary quite significantly within one species, within one population, but the idea is this call is genetically structured.”
So, Hendriks tied sound recorders all over the forests. He distributed these devices so they overlap, allowing him to triangulate the origin of the calls he detects, just as a criminal investigator would use communication towers to track a mobile phone. Using these sound recorders, Hendriks collected around 120 gigabytes of data, for which he could use only the 40% of the data he collected.
“Hybridization is quite an interesting topic, especially trying to understand why it happens,” Hendriks said. “I feel quite confident to say that … it is unlikely that what we’re seeing is a hybrid because there’s not enough uniqueness in the population for there to be so.”
Hendriks’ devices also allowed him to estimate the gibbon densities in these forests. He discovered the density decreases rapidly after 200 meters (650 feet) above sea level, meaning gibbons are at a greater risk than expected, as much of Borneo’s lowland forests have been cut down for oil palm plantations.
“That’s really cool that you could do that with just bioacoustics,” Hendriks said.
Regardless of whether it is working with muriquis, chimpanzees or gibbons, primatologists around the world keep taking advantage of the accessibility of new technologies that help them push the limits of the field to new horizons.
“The technology is very important for us. I think that people need to know that. … We can use that to enhance our action for protecting endangered species,” Melo told Mongabay.
Banner image: Jorian Hendricks in the forest of Brito Ulu, Indonesian Borneo, taking field notes in his notebook while using a directional microphone to record the sounds of the forest. Image courtesy of Jorian Hendriks.
Citations:
De Melo, F. R. (2021). Drones for conservation: New techniques to monitor muriquis. Oryx, 55(2), 171-171. doi:10.1017/s0030605321000028
Darwin, C. (1897). The expression of the emotions in man and animals. New York: D. Appleton and Company.
Schofield, D. P., Albery, G. F., Firth, J. A., Mielke, A., Hayashi, M., Matsuzawa, T., … Carvalho, S. (2023). Automated face recognition using deep neural networks produces robust primate social networks and sociality measures. Methods in Ecology and Evolution, 14(8), 1937-1951. doi:10.1111/2041-210x.14181
Bain, M., Nagrani, A., Schofield, D., Berdugo, S., Bessa, J., Owen, J., … Zisserman, A. (2021). Automated audiovisual behavior recognition in wild primates. Science Advances, 7(46). doi:10.1126/sciadv.abi4883
Marshall, J. T., & Marshall, E. R. (1976). Gibbons and their territorial songs. Science, 193(4249), 235-237. doi:10.1126/science.193.4249.235
Credits
This story first appeared on Mongabay
This article is licensed under a Creative Commons Attribution-NoDerivatives 4.0 International License.
You may republish this article, so long as you credit the authors and Mongabay, and do not change the text. Please include a link back to the original article.
