Decoding the sounds of the rainforest

When Neal Digre graduated from Western in 2016 majoring in computer science and linguistics, he was continuing a journey that started in his tiny hometown of Hendricks, Minnesota. But even Digre could never foresee that path winding its way into the triple-canopy jungle of the Brazilian Amazon rainforest.

Ahead of him lies the opportunity of a lifetime—working with a team of scientists trying to save the world’s rainforests by competing in the finals of the $10 million XPRIZE Rainforest competition this July in Manaus, Brazil.

“I wanted to find a way for my training and education to mean more than just having a good job,” Digre says. “My goal is to help solve problems that go beyond my job and start impacting the world in a more altruistic way. But I had no idea that goal would lead me to where I am now, in the way that it has. It’s been an amazing ride.”

WWU Professor of Linguistics Ed Vajda remembers Digre well.

“Neal was an extraordinary student with a deep curiosity about the world. He was a voracious learner with a creative and original spirit,” Vajda says. He was unsurprised that Digre had found his way to an opportunity like the XPRIZE.

Digre credits WWU Linguistics Professor Vajda for steering him onto the path that has now led into the Amazon.

“Ed was such an amazing teacher,” Digre says. “I took his class on the languages of J.R.R. Tolkien’s books and was just hooked. But even then I didn’t realize how much my linguistics work would inform the research I was doing in computer science, and vice–versa.”

After immersing himself in Tolkien, Digre worked in Professor Brian Hutchinson’s computer science lab, diving into a different—but related—set of language skills.

Hutchinson’s expertise lies in Artificial Intelligence, specifically the subset of AI known as machine learning, where researchers build algorithms that allow computer networks to process data and learn how to process tasks without instruction.

One specific offshoot of machine learning is called Natural Language Processing, where networks analyze and predict human language models—and this research fuels many of today’s most talked-about AI efforts such as ChatGPT. Digre’s fascination with the structure and foundations of languages was a perfect fit for Hutchinson’s lab.

"As a student and scholar, Neal learned quickly, worked hard and was a great communicator. It is no surprise that he has continued to have success since graduating from WWU,” Hutchinson says.

Combining the symbiotic fields of study from both departments, Digre left Western as the Linguistics Department’s Outstanding Graduate and spent a year in Edinburgh, Scotland getting his master’s degree. After a 6-month internship with the National Institute for Informatics in Tokyo, Japan, Digre began his career putting his expertise in machine learning to good use with a tech startup in Seattle.

The next radical turn in Digre’s life path happened one Christmas when he was home for the holidays in Minnesota and catching up with family. 

Digre ran into his cousin, Johanna Varner, and her partner, Tom Walla, both professors at Colorado Mesa University. Walla was the leader of a team that had just entered the new $10 million XPRIZE Rainforest competition, and Varner was leading the team’s communications.  

“Tom and Johanna had some questions on the coding and machine-learning side of things that I was able to help them with,” Digre says. “And I was for sure intrigued by what they were working on.”

Not long after Digre joined the team in time to prepare for the semifinal round of the competition  in the jungles near Singapore in June 2023.

“I was excited, but at the same time immediately felt the pressure of not wanting to let the team down,” Digre says. “But first I had to get up to speed on what the competition really was, and how I could be a positive impact .”

XPRIZE Rainforest is a global five-year competition that convenes innovators and experts across disciplines—from conservationists and Indigenous scientists to engineers and roboticists—and challenges them to use novel technologies to expedite the monitoring of tropical biodiversity.

 The organizers of the XPRIZE hope that if the world knew the true value of the Amazon, we’d work harder to preserve it instead of rapidly destroying it. Key to valuing the rainforest’s rare ecosystem is fully understanding its biodiversity.

Digre’s team, Limelight Rainforest, is focused on collecting, building, and cataloging a database of the rainforest ecosystem. They take their team name from their signature technology, the Limelight, a remote-sensing platform delivered by drone to the top of the jungle canopy where it spends the night listening in on the Amazon’s insects, birds and tree-dwelling animals.

Overnight, Limelight uses lights to attract insects and record the sounds they make. The drone then collects the platform and returns it to the forest floor.

Digre’s task in Singapore was vitally important: He needed to help design and build a computer model that could take the nighttime recordings of jungle sounds and reliably match them to sound files of existing wildlife, especially bugs and birds.  Digre’s coding needed to pull the audio from Limelight, run it through his custom machine-learning module, and run predictions to get positive matches from the database.

Thankfully, Digre’s experience with both languages and machine learning was a perfect match for the task—what are the chirping of crickets or the calls of birds but their own languages? Still, tweaking and fine-tuning the software continued right up until the team left for Singapore.

“It worked well in the lab, using test sound files. As with every part of our device and its delivery, testing it here in the States was one thing, but how would it perform in the jungle? That was what we had yet to find out,” Digre says. “In the end, everything from a machine-learning standpoint under the hood of the Limelight—it is all just math. Lots and lots of math.”

Once they got to Singapore, the teams took their devices to their assigned sampling areas of the nearby jungle and picked them up in the morning. Then they had 48 hours to download their data, and present their report.

The competition environment was stressful—but fascinating and exhilarating at the same time, Digre says.

“It was incredible. And we learned so much about how we could make things better if we made it to the finals. One thing that leaped out at us was how much louder and full of activity the sound files were ... the jungle was SO much richer and more full of sounds than even our test ‘jungle’ files had prepared us for,” he says.

Digre helped with all facets of the competition,  but he became the true focus of the team once Limelight was brought back from its night in the treetops.

“My work really begins once the data starts to come in,” he said.

Despite the competition’s obstacles, Limelight Rainforest got its data, and Digre’s machine-learning module worked like a charm. After the 24-hour scramble to build the 100+ page report required by XPRIZE, based on the strength of its results and verified matches provided by Digre’s module, Limelight Rainforest is one of six teams going to the finals in July in the Amazon.

“Well, the first thing I had to do was request the time off,” says Digre with a laugh. “There was no way I was going to miss the finals.” This offhand comment highlighted a reality: All of the Limelight Rainforest team—more than 30 members—are volunteers with day jobs, from professors and teachers to researchers and engineers.

Digre says Limelight Rainforest  knows it can’t perform at the same level  and expect to win the finals in Brazil.

“Everything needs to be better, because we know the other five teams are going to have made improvements too,” he says.

One new feature for the finals: Limelight Rainforest has arranged to incorporate Cornell University’s award-winning Merlin birdsong-recognition software into its detection algorithms.

“It was so great that they allowed us to use that module. It will really help,” Digre says. “And we promised to share all of our data with them, so hopefully we will help them make Merlin even better and more accurate.”

Of course, any competitor wants to win—and Limelight Rainforest is no exception. But Digre says it is easy to see the “$10 million” that accompanies the competition’s name and assume it is all about the money.

“Nothing about why any of us has invested our time and effort in this project has anything to do with money, to be honest,” he says. “I think I accepted Tom and Joanna’s invitation before I even knew there was a monetary prize.”

After the competition is concluded, $5 million will go to the winning team, with second and third place also getting funds—but Digre doesn’t have the time or energy to think about that right now, not when there is fine-tuning still to be done before they leave for Manaus in the heart of the Brazilian Amazon.

“I kind of feel like we’ve already won, because of the connections we have made, and the exposure that the contest has given to the plight of the rainforest,” he says. “In the end, this is about what I could do to help the planet—and it is easy to feel good about that.”