Researchers in Canada are using artificial intelligence to monitor the ongoing mass extinction of insects, hoping to collect data that can help reverse species collapse and avert catastrophe for the planet.

"Of all the mass extinctions we have experienced in the past, the one affecting insects is happening a thousand times faster," said Maxim Larrivee, director of the Montreal Insectarium.

The decline is occurring so quickly it can't be properly monitored, making it impossible "to put in place the necessary actions to slow it down," he told AFP.

For the Montreal-based project, called Antenna, some of the data collection is happening inside the insectarium under a large transparent dome, where thousands of butterflies, ants and praying mantises are being studied.

Solar-powered camera traps have also been installed in several regions, from the Canadian far north to Panamanian rainforests, snapping photos every 10 seconds of insects attracted to UV lights.

Larrivee said innovations like high-resolution cameras, low-cost sensors and AI models to process data could double the amount of biodiversity information collected over the last 150 years in two to five years.

"Even for us, it sounds like science fiction," he said, a grin stretched across his face.

Tip of the iceberg

Scientists have warned the world is facing its biggest mass extinction event since the dinosaur age.

The drivers of insect species collapse are well understood -- including climate change, habitat loss and pesticides -- but the extent and nature of insect losses have been hard to quantify.

Better data should make it possible to create "decision-making tools for governments and environmentalists" to develop conservation policies that help restore biodiversity, Larrivee said.

There are an estimated 10 million species of insects, representing half the world's biodiversity, but only a million of those have been documented and studied by scientists.

David Rolnick, a biodiversity specialist at the Quebec AI Institute working on the Antenna project, noted that artificial intelligence could help document some of the 90% of insect species that remain undiscovered.

"We found that when we went to Panama and tested our sensor systems in the rainforest, within a week, we found 300 new species. And that is just the tip of the iceberg," Rolnick told AFP.

Public education

At Antenna, testing to advance AI tools is currently focused on moths.

With more than 160,000 different species, moths represent a diverse group of insects that are "easy to identify visually" and are low in the food chain, Rolnick explained.

"This is the next frontier for biodiversity monitoring," he said.

The Montreal project is using an open-source model, aiming to allow anyone to contribute to enriching the platform.

Researchers eventually hope to apply their modelling to identify new species in the deep sea and others harmful to agriculture.

Meanwhile, the Montreal Insectarium is using its technology for educational purposes. Visitors can snap pictures of butterflies in a vivarium and use an app to identify the exact species.

French tourist Camille Clement sounded a note of caution, saying she supported using AI to protect ecology provided "we use it meticulously."

For Julie Jodoin, director of Espace Pour La Vie, an umbrella organization for five Montreal museums including the Insectarium: "If we don't know nature, we can't ask citizens to change their behaviour."