Bioacoustic analysis made easier: Q&A with Rainforest Connection CEO Bourhan Yassin

For scientists and ecologists studying wildlife, listening to the sounds of nature is all the rage now. And with good reason.

The exponential growth in the use of audio to study, monitor and track wildlife and biodiversity can be attributed to its very many advantages. For one, it’s noninvasive; audio data can be gathered without capturing animals and by virtue of installing audio recorders across forests. Passive acoustic monitoring, or bioacoustics, is also capable of collecting large amounts of data that can be used to monitor entire landscapes, detect species and understand the behavioral and communications patterns of animals. This makes the methodology more efficient than traditional camera-trapping and remote-tracking techniques.

But large amounts of data mean more audio to parse. Despite its advantages, the labor-intensive analysis of audio data makes it difficult to sift through the data and derive meaningful conclusions from it. In recent years, artificial intelligence and machine learning have made the analysis work easier. However, not everyone has the budget and technical expertise to use these tools.

Arbimon, a free online platform by conservation technology nonprofit Rainforest Connection, aims to fill that gap. The platform allows scientists and ecologists to upload their audio data and enables them to run sophisticated analyses on it. More recently, Arbimon has introduced tools that allow users to detect sounds from audio data and group them based on various factors, such as range and frequency.

“The platform was designed to specifically address the gap between research and conservation, so that we can get to a point where we can drive conservation action on the ground,” Bourhan Yassin, CEO of Rainforest Connection, told Mongabay in a video interview.

Bourhan Yassin spoke with Mongabay about the new developments in Arbimon, the impact the platform has had and the gaps that continue to exist. This interview has been lightly edited for length and clarity.

Mongabay: How would you explain Arbimon to someone who doesn’t know about it?

Bourhan Yassin: Arbimon started out as an acoustic platform. The primary focus was to put some of the acoustic analysis tools in a way that scientists and researchers could benefit from them as they analyze acoustic data. The platform then grew to focusing on providing insights beyond just the analysis parts. We started releasing features that are focused on showing insights in the data, like occupancy and richness of species.

Mongabay: Can you explain how Arbimon works to make conservation using bioacoustics easier?

Bourhan Yassin: As it stands today, Arbimon is focused primarily on science and research. We have a science team and a deployment team on the ground that is focused on answering very specific scientific questions from partners around the world. At the moment, we have around 60 projects that are happening in different countries. Our intention is always to keep it open and free. So Arbimon is available with all of its tools for everyone to use.

Mongabay: What is the gap Arbimon is trying to fill?

Bourhan Yassin: With bioacoustics, managing large amounts of data has always been a difficult thing. A typical scientist, even an experienced ecologist, will take 5-6 minutes to process one single minute of audio, if not using AI or machine learning. With the introduction of something like Arbimon, you can run simple analysis tools and some of the built-in functionalities, and you can process 10,000 files in a matter of a few seconds. Additionally, we are trying to bridge that gap between scientific research and conservation on the ground. A lot of science and research is super focused on publications and releasing papers. Oftentimes, it doesn’t make its way down the stream to the conservation organizations. Since we work very closely with our conservation partners, we know exactly what they are looking for. The platform was designed to specifically address that gap between research and conservation so that we can get to a point where we can drive conservation action on the ground.

Mongabay: Have you started seeing encouraging results?

Bourhan Yassin: As we collect more and more data and as we get more and more scientists contributing to the platform, we are starting to map what biodiversity looks like on a holistic level. That’s going to allow us to hopefully predict biodiversity health in areas where there is very little or no data. It could also lead to more concentrated efforts in places to collect more data, and it could allow organizations on the ground to drive more efforts toward it.

Mongabay: Can you explain how Arbimon came to be? What is the backstory?

Bourhan Yassin: Arbimon was launched, I want to say, over 10 years ago. It started as a project out of the University of Puerto Rico. Somewhere around 2019 is when Rainforest Connection got involved with Arbimon. We took on Arbimon as this tool that we would offer our partners and our projects, and we have grown it from there since. It went from being a software tool that’s available online for acoustic data uploading to being essentially an organization with a team of scientists and researchers that are focused on answering scientific questions.

Mongabay: Rainforest Connection has long worked in bioacoustics. What is it about Arbimon that made the organization want to collaborate with them?

Bourhan Yassin: At Rainforest Connection, most of our work in bioacoustics was primarily focused on threat detection and tracking illegal activity. So a lot of our work was on chainsaws, vehicles, trucks, gunshots and any signs of illegal activities. We have always collected the entire soundscape by virtue of trying to find these illegal activities. So biodiversity monitoring was always on top of our list. But honestly, we didn’t have the scientific knowledge to do anything with it. The science element of biodiversity is extremely important. It’s a critical component. So that’s where Arbimon came about. So having that scientific knowledge and being able to build a team around some really experienced people was important.

Mongabay: How do you measure Arbimon’s impact?

Bourhan Yassin: Arbimon now adds somewhere between 2-3 million recordings every few days. There are entire research institutions that spend years to collect that much amount of data that’s being added every couple of days to Arbimon. Arbimon, as it stands right now, has somewhere around 114 million recordings uploaded to it. In terms of analysis, there are different types of tools that run on Arbimon. Somewhere around 3,000 scientists have used it or are using it. So the scale is quite massive. To be honest, it’s a daunting task to make sure this is supported for years to come.

Mongabay: Could you briefly explain the workflow in Arbimon?

Bourhan Yassin: The workflow essentially starts by data collection. So your first interaction will be getting the data up to the platform. The moment that the data is there, you will see it immediately in Arbimon. And then you can start creating playlists — you can create a playlist of 100,000 files, or you can create a playlist of 5,000 files; however you want to do it. They are smart playlists where you could put it by date or filter it by location or by site or whatever the case may be. The next step is the analysis. That’s where you can use the variety of tools to analyze the audio data.

Mongabay: Can you walk me through the analysis tools in Arbimon?

Bourhan Yassin: Pattern matching is one really popular analysis mechanism where you provide one single example or sound signature, and you say, ‘Go ahead and find me all similar sound signatures in this playlist.’ It will then run an analysis and come back and give you all the possible examples that it thinks matches this particular pattern. You can then validate it and say, ‘Yes, this is correct,’ or ‘No, this is incorrect.’ And essentially, what that leads to is a labeled dataset.

You can also use an unsupervised learning approach called cluster analysis, which is also available. Basically, you can give it a dataset or a playlist, and it will come back and it will put it into different clusters — all the groups of sounds that have similar frequency, range and other characteristics. Then, you can start identifying what these clusters are and label them. So, that’s an unsupervised way where the computer comes back and clusters all that information, and you can do an entire soundscape analysis.

The labeled data could be used as is. Or you could use the labeled data to create a CNN [convolutional neural network]. The workflow of a CNN creation is not part of Arbimon. But it’s something that you can create outside of Arbimon and then run it over any new data that comes in, and it automatically extracts information.

The next step is the insights. Here, you can start seeing a holistic view of the data. So we start mapping this data out with the IUCN Red List. We extract species information and taxonomy information. You’ll see occupancy maps, you’ll see species richness maps, and you’ll see what species have been detected.

Mongabay: What is the importance of a tool like Arbimon at this point?

Bourhan Yassin: As we see more and more data come into the platform, we see how great eco-acoustics has been in recent years and how more people are adopting it. And so, I think it becomes very important to create these tools to facilitate that process. Because honestly, the one thing about acoustics is that the large amount of data can be daunting. So I think continuing to facilitate that process for people is very important.

Mongabay: What does the future of Arbimon look like? What new developments are in the works?

Bourhan Yassin: I think the future of what we are looking at is in getting better insights. So you are going to see us combining other data sets. We have already started doing this. So combining bioacoustics with eDNA [environmental DNA], camera-trap images and GIS information to answer a lot of the scientific questions that we get from partners. We are planning to introduce more of these integrations into Arbimon.

For us, the key is going to be to predict what biodiversity health looks like in areas where there aren’t necessarily data that’s available, but we could predict what that looks like based on information that we have available from all the other places that are of similar biomes. So being able to predict biodiversity health in a way that allows us to also be able to update it on the fly. I don’t think generating a static report of what biodiversity health looks like in a particular region is enough. Being able to create a dynamic system that, as you feed it more data, will automatically update and improve, I think that is going to be where the focus will be.

Mongabay: Given that your organization does a lot of work in bioacoustics, I am curious to know how you think the field is faring.

Bourhan Yassin: Bioacoustics is one of the most profound and important ways to understand biodiversity. I can’t think of another technology that is as insightful as bioacoustics. So I think the future is very bright for bioacoustics. I think it’s also going to start getting the attention of carbon project developers, which we are seeing ourselves. Acoustics gives you more capability because you can listen over large distances and you can deploy so many more devices much more quickly.

Mongabay: What are the gaps that continue to exist in the field of bioacoustics?

Bourhan Yassin: I think that we need to come together at some point and figure out how we all collaborate. Most of the time, we are essentially spending time and resources re-creating the same stuff, which I see happen sometimes a little too often in the conservation space. Lots of people are essentially doing different versions of the same thing. So I think it’s important that we come together so that we are not creating this world where we are slowly advancing, but we could be advancing much faster if we are collaborating versus sort of working in silos.

Mongabay: Where do you see Arbimon and Rainforest Conservation 10 years from now?

Bourhan Yassin: We know the struggles of conservation organizations. So if we can create something that enables anyone, literally anyone, with little technical skills and good scientific chops to sit and analyze data at scale and extract information, then somebody from the conservation organization comes in, logs in and says, ‘I understand what that is, I can take that information and I can extract meaningful action on the ground from it.’ That is what we want to achieve. If we can achieve that, I think that makes a world of difference because that bridges that gap, and it enables more conservation action to happen on the ground. That’s where we see ourselves.

Banner image: Bourhan Yassin in the field in Chile. Image courtesy of Bourhan Yassin.