Antoine Caron-Guy
1 credit
Will we be able to use drones and artificial intelligence to map invasive exotic plant species in the future?
This approach to detecting changes in biodiversity seems promising, according to Antoine Caron-Guay, a master’s student in biological sciences at the University of Montreal. In his thesis, written under the supervision of Etienne Laliberte, professor at UdeM, and Michael Germain, professor at the University of Sherbrooke, he was interested in the early discovery of common reed on the Boucherville Islands.
Common reed, invasive exotic species.
The rapid spread of common reed threatens biodiversity. Particularly robust, common reed threatens native plants in surrounding wetlands such as cattails, thereby threatening the habitat of vulnerable wildlife species such as the least bittern.
This invasive exotic plant, which has spread throughout southern Quebec, is found in the Ile de Boucherville National Park, where a research team from the University of Montreal went to study ways to limit its population.
Early detection of common reed using drones
Early detection of common reed is of great importance. Once the reed forms a large colony, managing and eradicating it becomes more difficult. Indeed, this plant has the ability to firmly establish itself thanks to its significant underground biomass. To eradicate it, intervention requiring excavation becomes inevitable.
Action before the species is fully established, or after initial treatment, will facilitate control, especially for a few isolated stems per square meter. In these cases, it is advantageous to be able to pinpoint the affected areas so that the herbicide can be applied locally. Some biologists then use glyphosate, selectively applied to the cane leaves using a sponge.
Although the presence of common cane can be detected by satellite, the limited resolution of each pixel makes this method inaccurate for detecting individual stems. On the other hand, drones provide greater spatial resolution, allowing isolated stems to be distinguished from large reed colonies. This high spatial resolution results in very accurate results in detecting and managing this species.
Images analyzed using artificial intelligence
Antoine Caron-Guay annotated images captured by a drone to find the location of common reeds. He then used an artificial intelligence model to detect these canes. This model analyzed the shapes, colors and textures of the images to distinguish common canes. Once trained on a large amount of data, the model can be applied to previously unpublished photographs to test its ability to systematically identify common canes.
“For invasive exotic species, we especially want to minimize false negatives, that is, places where the model did not detect a common cane when it was actually present. These unmarked areas allow species to continue to reproduce without treatment, which can result in large colonies that are then difficult to manage in the long term. We obtained almost perfect results with virtually no false negatives,” explains Antoine Caron-Guay.
Aim for the best time
The team collected data six times during the plant’s growing season, examining environments with varying levels of cane infestation. The results showed an improvement in model performance that peaked after flowering in September. Lighting conditions also affected the results, causing the model’s performance to decrease on sunny days.
“This identification will depend each time on the life cycle of the species,” explains Antoine Caron-Guay.
Approach to other invasive alien species
This approach can be extended to other invasive alien species such as water chestnut. However, some plants have different cycles. For example, buckthorn, a plant that grows primarily under the canopy, may require earlier detection, perhaps in the spring, before tree leaves have grown. “This plant grows quite early, so it can be seen before the crown is fully formed. Acting too late in the season, when the canopy has already formed, poses a problem because drones cannot see through dense vegetation,” concludes Antoine Caron-Guay.
Related Posts
Read on Tech&Co: Claude, a ChatGPT competitor available for iOS, by Willem Gay
The James-Webb Telescope reveals extreme weather on the planet driven by hypersonic winds!
