Reimagining insect research: Interview with Roel van Klink and Leandro Nascimento

They might comprise close to 75% of all known animal species. But when it comes to the monitoring and protection of insects, global efforts have been inadequate.

The numbers tell a harrowing story.

A global review of insect populations published in 2019 in the journal Biological Conservation found that 40% of insect species are threatened with extinction, which could lead to “a catastrophic collapse” of ecosystems worldwide.

“Most insects are relatively uninteresting for humans,” Roel van Klink, senior researcher at the Biodiversity Synthesis Research Group at the German Centre for Integrative Biodiversity Research, told Mongabay in a video interview. “It’s very laborious work, and society has not been willing to pay for the time that it actually needs to do this well.”

Even as technology is being increasingly deployed to study and monitor insects globally, van Klink said he realized that groups developing these tools were often unaware of developments in other spheres. “It seemed that they were not talking to each other,” he said. “There were parallel developments in different countries.”

As a first step toward addressing this concern and to get people to collaborate, van Klink brought together a group of scientists, researchers and technologists working on insect biodiversity. Working with them, he and a team collected, compiled and edited a special issue of studies titled “Towards a toolkit for global insect biodiversity monitoring” that was published in the journal Philosophical Transactions of the Royal Society.

The issue comprises research from 142 authors in 27 countries with a focus on four cutting-edge technological developments that could potentially reshape insect biodiversity monitoring in the years ahead. They include acoustic monitoring to identify insects from sound; computer vision to identify them from images; radar technology to detect insects remotely; and the use of DNA to identify species.

Roel van Klink spoke with Mongabay’s Abhishyant Kidangoor about the state of global insect biodiversity monitoring and how technology might help fill the gaps. He was also joined by bioacoustics scientist Leandro Nascimento, who was part of the editorial team for the special issue. This interview has been edited for length and clarity.

Mongabay: How is insect biodiversity monitoring faring globally?

Roel van Klink: The status is actually really bad. It’s all being done with traditional methods, which means you need to catch the insects and then you need to identify and count morphologically. Everything is done manually. Most of that is done in Europe. We have a very long tradition going back to maybe the 1920s. In the U.S., too, they’ve been doing that for a very long time. But it’s usually for very specific groups of insects. For example, we have a good network for butterfly counting that’s well established in Europe, but there are still European countries that don’t have it. Even in the best case, it goes back only to the ’70s. But again, it’s only for butterflies. In the U.S., they have been measuring mosquitoes going back to the ’60s, but that’s only mosquitoes. Everything else is just random depending on whether there was a person that was willing to or interested in identifying one specific group of insects at one place.

Eventually, the situation is that we may be monitoring only ground beetles in one place, butterflies in another place and mosquitoes in another. And that’s for only some countries. Most of the rest of the world, there’s just almost nothing. It’s really bad for most groups of insects. We don’t even know how they’re doing. We don’t even know which species there are in the tropics.

Mongabay: Why is it this way?

Roel van Klink: Most insects are relatively uninteresting for humans. We can do pest monitoring or disease vector monitoring, and track butterflies, because they’re easy. You can get lots of people to do it. But for all the other obscure little flies and wasps that are out there, it’s very hard even to identify them. For specialists, the literature is huge. There are so many species. Most species have not been described, and it is very laborious work. Society has not been willing to pay for the time that it actually needs to do this well.

Mongabay: What are the gaps in the traditional methods of monitoring and studying insects?

Roel van Klink: The usual methods to catch insects would be something like pitfall trapping. In that, you will only catch the stuff that’s walking on the ground surface and can fall into the trap. If you want to catch flying insects, you will do that with a different method. For example, with a net Malaise trap. Everything that flies in goes up, and there they get caught. If you want to get stuff below the ground, there are different methods that you need. If you want to get insects in vegetation, you would need different methods. Every group of insects needs a different way of getting them.

And then you need so much specialist knowledge to identify all those species because you have a large group to choose from. Here in Germany, for example, we have 33,000 species. Nobody knows all of them. It’s impossible. The people that have most species knowledge can maybe identify 4,000 beetle species. That seems to be really the maximum. If you go to tropical countries, the diversity is even larger. So you have to specialize in one specific group to identify, and therefore you cannot get everything identified. Therefore, for most of the things, we just don’t know what it is.

That’s where these modern methods can help us to monitor many more species, because artificial intelligence essentially can learn unlimited amounts of species in contrast to a human. We can use these methods to identify more species for more locations over longer time periods with much less human effort.

Leandro Nascimento: If monitored more efficiently, we don’t need to be sending people all the time to the field to collect these animals and to kill them. That’s another problem some people are starting to discuss. Most of these traditional methods are lethal. That means you kill insects. And, who knows, maybe you are capturing an insect that is already endangered or is at risk of extinction. With these other technologies, you don’t necessarily have to kill them. For example, in acoustic monitoring, we’ll be recording the sound passively without killing them. That’s another big advantage.

Mongabay: What was the spark for getting this issue on global insect biodiversity monitoring together?

Roel van Klink: We started this coming from a working group. Me and a colleague had been working on traditional insect monitoring data for quite a while, and we were noticing that lots of different groups of people were trying to develop these modern methods. But it seemed that they were not talking to each other. There were parallel developments in different countries. Radars were being developed. There was acoustics being developed better and better. Then there was the molecular stuff in its own right. We figured we needed to get all these people to start talking to each other. So we organized a workshop in 2021. It had to be virtual, but it was quite well-attended.

Then, in 2022, we could finally do an in-person workshop. That’s where the idea came up to do this as a special issue. We had a very interdisciplinary group. We have ecologists, but we also have computer-vision experts, laser physicists and geneticists. You can do this only in an interdisciplinary manner because of the specialized knowledge that you need, both on insects and the methodology. No one person can do all of that.

Leandro Nascimento: I was called in because my expertise is in bioacoustics. I was working in acoustics for over 10 years. I realized that most of the people working in bioacoustics didn’t have the tools to monitor insects, like training data. They were using bioacoustics and doing mostly classification of birds and mammals, but the insects were again being neglected. In places like the Amazon, India and most tropical countries, insects are the most biodiverse group. Even if you just walk in the forest, you hear all these sounds. We might be recording their sounds, but why are we not studying them? Why are we not classifying them and trying to collect data on them? That was the main motivation for me.

Mongabay: Could you tell me the types of technology that interested you and why you thought they were worth mentioning in your issue?

Roel van Klink: These were the four technologies we started with because they are the best developed. Since then, only two have really started to become established. But these four are the ones that have probably the biggest potential and the widest application. They have different strengths and weaknesses.

With molecular methods, you can probably get the best data on the species identity, but you cannot really get reliable estimates of how many individuals there were in your sample or any environment. But it’s becoming quite an established method. Lots of people are working on it, with barcoding and meta barcoding. And now it has even become possible to do large-scale individual-level barcoding. All of these are established.

The only ones that are really quite new are the lidar stuff. So it’s a laser that gives you very detailed information about anything flying through the laser. We have only very little about that in this issue because it’s fairly new and very few people are working on it.

The use of substrate vibrations is also relatively new. It is the sound that insects make when they knock on wood, or they are gnawing inside the wood or in a plant or in the soil. So it’s related to acoustics in some way. But you need very specific machinery to detect these signals, either with a laser or with the microphone.

Leandro Nascimento: That’s right. Vibration is quite new. But so is everything else about the hardware. For tracking animals, for example, visually, the way we do for large species is by using camera traps. But we cannot use a camera that is activated by motion sensors to study insects. They’re too tiny for that. So people have been developing new hardware to try to identify insects. And that brings a whole new set of challenges. For visual identification, software is more advanced than hardware.

Roel van Klink: Yeah, software is working quite well. For example, all these apps for identifying stuff like iNaturalist and other observation apps, they work amazingly well with a picture of one insect. If you have enough training data, it works really well. But if you really want a standardized sample of insects, and you need to attract them in some way to a screen with light or a yellow screen and then take a photo of them, that hardware is still being developed. People have been working on it for years, but it’s a slow process. And, basically, you start from scratch because nobody has ever done anything like that before. It’s happening and, hopefully, in a few years, we will have something that can really do this well.

Leandro Nascimento: You can deploy it, for example, in large-scale monitoring programs. The molecular methods, the acoustic methods and even the lidar, you have national-level programs where people are trying to monitor or count insects. But with just camera traps for insects, that’s not still not at that level yet even though the software is quite well developed for recognizing patterns and visual identification.

Roel van Klink: As for the other methods that we cover, radar is quite well developed. And we’ve known actually for a very long time that a radar can detect insects. They did that in the U.K. for a very long time starting in the ’90s. There were some good papers coming out and now there is actually a radar network that can detect anything flying through the beam. But because the radar waves are relatively long, you do not get a whole lot of detail on the organism flying through the beam. And whenever there’s a bird flying through your beam, it gives such a bright signal that everything else cannot be seen anymore. It’s a very interesting method, and it needs more development to really get more information about the insects flying through the beam. But at least there is a network. There are probably 20 radars deployed in Europe now.

In the U.S., they’re working a lot on the weather radar systems. These are horizontal radars that try to capture the rainfall and stuff. They will also detect insects, but that wavelength is even bigger. So the resolution of what you get is much coarser and you can probably only detect the biggest insects. But it covers such a large area that it becomes very interesting because it can measure at places where we have no data. And so that’s why we believe that the combination of all of these different methods gives us the best overview of what’s happening to insects.

Mongabay: How can we better bring all these tools together and make them work seamlessly?

Roel van Klink: What we are hoping for is to maybe establish a network of sensors. But then, we do need to get good sensors first. The computer-vision stuff will be absolutely crucial to have. And then what we hope is that we can combine the observations at the ground level from the camera traps and the molecular data with what we observe higher in the sky using the radar, and maybe there will be a lidar in between. Eventually, we hope to scale it up.

What we need to know is how the information is related between all of these different strata. If you have all these different sensors and stations that are local, and you have at the same time the radar going above it at a very high level, then what we hope is that we can extrapolate what we detect at the local level. From that, we hope that the radar detects things where we do not have stations by extrapolating down to what’s happening there. But first, we need proof of principle that these methods can be combined, and more details on what information is being transported between the different levels.

Mongabay: How are you planning to get that moving?

Roel van Klink: What we are hoping to do is get some money to do this. We have a very international team of people. Even in Europe, this is not easy. We have European grants that can work internationally, but they’re extremely competitive. Most of the grants that you can actually get are at the national level. We can maybe get national funding in Germany for developing computer-vision traps. But if we want the best acoustic experts, then we probably need two people from France because the French have developed their methods especially for bush crickets in France. The best radar people are located in Switzerland. The amount of money you need and having all these different countries involved is very challenging. And it’s not straightforward to get funding for all of these different institutes.

Leandro Nascimento: Especially if it is purely for research. I think companies in the private sector can also play a role in getting the fund because then you can get funding not only from the government agencies, but also from private sector investors to try to invest in these technologies. Because many of them are working with these different technologies to get an overall biodiversity assessment of the area. It’s just that they use the technology in isolation and so they’re not combining it. Either they are doing acoustic monitoring, or they’re doing some molecular monitoring, or just using traditional methods, but then not combining them. I also think that the traditional methods are always going to be necessary to describe new species to do the taxonomy work, but it can be integrated with these newer technologies.

I have a lot of taxonomist friends working in Brazil. When they go to collect new species or describe a new species, they don’t take, for example, recorders because it’s not part of their agenda. But why not also record these animals so we can have a reference library of how those animals sound? Then people across science can train machines based on that data.

Roel van Klink: This library issue is really, really important. It’s not just for the acoustics, but also for computer vision, which does not have enough images. You cannot take a machine that was trained in Denmark and put it to use even in Switzerland, let alone in a tropical country. So you’re going to need local experts to identify what the species are. And then you need to have enough photos or sounds to actually train the classifier. This is the challenging part.

For radar and lidar, we probably just have to accept that we’re not going to know everything. Whereas for acoustics, we can get really far in associating a sound with a species. Since we still have 80% of the species on Earth undescribed, it’s going to be a lot of work to also record the sounds of these things as they get described.

What we actually need is a large campaign. There have been big campaigns to record all the DNA sequences of all the species. We’re making progress in that field, but the same should be done for sounds and images so that we can actually start using AI to recognize all these things.

Banner image: A dragonfly perched on a leaf in Thailand. Image by Andre Mouton via Pexels (Public domain).

Abhishyant Kidangoor is a staff writer at Mongabay. Find him on 𝕏 @AbhishyantPK.

Bioacoustics in your backyard: Q&A with conservation technologist Topher White

See more