Artificial intelligence helps insects

David Rolnick has always been passionate about insects. As a child, he spent his summers observing moths that he lured to his garden with lights. Now a computer science professor, he does the same thing, this time using artificial intelligence (AI).

In collaboration with the Montreal Insectarium, he launched the Antenna project, which aims to use AI to automate insect monitoring around the world.

We talk about the collapse of insect biodiversity in apocalyptic terms, but the data we have to support this idea is very limitedexplains Professor Rolnick, who is a senior academic member at the Mila Institute.

He cites for example the windshield effect, a phenomenon observable by ordinary people, but little measured with little precision. In some cases, the bugs were found to be gone because people noticed fewer bugs crashing into their windshields. When this is how we measure biodiversity, there is a problemhe maintains.

However, with 1.3 million known species and billions of individuals, counting insects is no easy task. Maxim Larrivee, director of the Insectarium and David Rolnick’s main collaborator in this project, knows something about this.

Insects are hyper diverse, are hyper abundant, and there are few experts who are able to confirm identifications.

Traditional census methods involve capturing insects using traps and recording for each specimen, in addition to the species, information on the date, location and method of collection. A “monk’s work”, according to Mr. Larrivee, which often leads to the death of insects.

An algorithm for moths

To offer a new method, Antenna relies on a computer vision algorithm designed by David Rolnick’s team, which identifies insects from photos.

The team chose to focus on moths first. These insects have the advantage of being attracted to light and being identifiable by the characteristic patterns on their wings, which vary from one species to another.

To photograph the insects, scientists use camera trap stations developed by a team of researchers from Denmark. These stations include an ultraviolet (UV) lamp to attract nocturnal insects, a white screen for them to land on, and a camera that takes photos every few seconds throughout the night.

Open in full screen mode

The photos are then sent to a central computer and analyzed. On each image, the algorithm detects moths and identifies the species of each individual by giving a percentage of certainty of the identification.

He then counts the number of individuals photographed during the night. To avoid double counting, it follows the specimens that move on the screen as the photos take place.

That day, Professor Rolnick explains, showing an example, the system worked for 11 hours. There have been 8000 particular insects found and 54 species.

Open in full screen mode

Such a quantity of information redefines the limits of what is possible in entomology. According to Maxim Larrivee, a station located in Quebec can observe approximately two million insects in one season, all identified, dated and located.

This is the number of observations that an expert will normally collect in his entire career!

I find myself thinking that it’s still science fiction, because it really allows us to go to scales of quantity of information and understanding of diversity that we couldn’t even imagine in the past.he adds.

Dark diversity

The stations are powered by solar panels and deep cycle batteries. It is a system that can operate in the driest and wettest environments autonomously for several months.explains Maxim Larrivee.

This is a perfect opportunity to study insects in hard-to-reach areas. For example, five stations were installed in Nunavik in collaboration with local communities. More than 150 stations have been installed on all continents (except Antarctica) as part of the Antenna project.

Open in full screen mode

Maxim Larrivee and David Rolnick helped install stations in the Barro Colorado rainforest in Panama. We wanted to see how they would withstand extreme conditions of 100% humidityexplains Mr. Larrivee.

This experience revealed Antenna’s unsuspected potential: that of discovering new species. The stations photographed 300 never-before-described species during the test in Panama.

This result is “completely astonishing” for Mr. Larrivee. It took me several months to accept that, in a place where insects have been studied extensively in the last 80 years, 25% of the entomological moth fauna is still unknown to science.

We didn’t expect it at allconfirme David Rolnick.

On the entire planet, it is estimated that 90% of insect species have yet to be discovered. This group of species is sometimes called dark diversitya nickname inspired by dark matter, the presence of which we infer without seeing it.

Species at risk

However, time is running out to catalog this dark diversity. By best estimates, the current insect extinction is occurring at a rate 1,000 times faster than previous major extinctions.

We have species which took thousands of years to evolve, which were sculpted by environmental pressures and which will disappear from this planet without us even being aware of it.laments Maxim Larrivee.

Open in full screen mode

Moths particularly stand to lose in this mass extinction. [Ils] are among the richest groups in biodiversity in the world, underlines David Rolnick. We have 150,000 species of moths on the planet, which represents about 10% of all known species.

And their collapse would have very concrete consequences on human populations. Moths play a crucial pollinating role and are prey for many bird species.

I think people don’t understand how much the biodiversity in their own garden is changing and how much it will affect them, says Professor Rolnick. This will affect other elements of biodiversity: the food we eat, the products we buy that are derived from wood and agriculture.

Tools like Antenna will make it possible to document these population fluctuations in real time with unprecedented precision. They could also measure how our actions contribute to the decline of biodiversity, for example by documenting the impact of certain pesticides on insect populations.

In this type of case, the data collected by Antenna could prove decisive. In the coming years, says Mr. Larrivee, we will have access to reliable information that we can share with decision-makers. This completely changes our understanding and, I hope, our ability to resolve this problem of biodiversity collapse.