The Vrije Universiteit Brussel (VUB) is innovating in urban green space management with new artificial intelligence technology. Researcher Robbe Neyns has developed a system capable of identifying trees down to species level using satellite and aerial imagery. This advancement aims to help cities better map their tree populations and biodiversity, thereby making the urban environment more liveable, especially in the face of a warming climate.
Trees play a crucial role in our cities: they provide shade and cooling, filter the air, and serve as habitats and food sources for numerous animals. However, precise tree inventories are often incomplete or outdated, and their identification on the ground is a time-consuming and labour-intensive task.
To overcome these challenges, Robbe Neyns explored the use of AI to analyse aerial images. For the Brussels-Capital Region, he combined satellite images taken at different times of the year with highly detailed aerial photographs. The system, trained using deep learning, learns to recognise different tree species from this visual data.
“Recognising a tree crown is one thing, but in a densely built-up city, crowns overlap, buildings cast shadows, and you have to deal with different background materials,” explains Neyns. “By combining different types of images, we provide the model with sufficient information to distinguish between tree species, as each species has its own characteristics and follows a different cycle throughout the year.” This creates a sort of digital tree expert capable of identifying where each tree is located on a large scale.
The technology was then applied to ecological research. In Braunschweig, Germany, Neyns mapped willow trees to study Andrena vaga, a wild bee species heavily dependent on these trees for pollen. By cross-referencing the tree map with other environmental factors and observations of bee nests, it was possible to predict which areas of the city offered a suitable habitat.
In a second application, the researcher investigated the health of the city's trees themselves. He linked information on tree species to urban heat, air pollution, and paving. This allowed for an analysis of the impact of these various urban stressors on the annual growth cycle of different tree species.
This research demonstrates that satellites, aerial photographs, and AI can do more than just count a city's greenery. By knowing the species involved, researchers can better understand which trees thrive where, how they respond to a changing climate, and what role they play for other species.
This information is crucial for cities aiming to plant more trees to safeguard against increasingly hot summers. “Cities currently undertaking large-scale reforestation in response to climate change face a choice: not just how many trees, but which species to plant where,” concludes Neyns. “We hope to make this information more accessible using this technology.”
The PhD thesis title is: “Beyond the Canopy: Deep Learning for Urban Tree Species Classification Applications in Pollinator Ecology and Tree Phenological Responses to Urban Stressors”. Robbe Neyns studied Geography at VUB and Artificial Intelligence at KU Leuven, beginning his PhD at VUB in 2020. He received the Young Scientist Award for his research at the international EARSeL conference in 2024.
