On 13-14 May 2025, MAMBO, in collaboration with the GUARDEN project, co-hosted the DeepSDM symposium in Montpellier at the Agropolis Fondation. The event brought together researchers, data scientists, and biodiversity experts to explore the potential of deep learning for species distribution modelling (SDM).

With over 65 participants attending in person and more than 120 total attendees including online participants, the symposium focused on the development and application of DeepSDMs - machine learning models designed to map and predict species distributions using complex datasets. These methods offer a step forward in biodiversity science, allowing for more accurate and scalable monitoring of species and ecosystems.

The programme opened with a welcome by Alexis Joly (Inria/Pl@ntNet) and featured contributions from leading experts from EPFL, University of Regensburg, University of Potsdam, and WSL. Presentations addressed both the theoretical foundations and practical implementations of deep learning in ecology and the session was moderated by Pierre Bonnet (CIRAD/Pl@ntNet) A joint session by Alexis Joly and Diego Marcos (Inria) highlighted results and tools developed within both MAMBO and GUARDEN.

The day concluded with a roundtable discussion titled “What can deep learning offer to species distribution models for biodiversity knowledge and its conservation?” moderated by Christophe Botella and Diego Marcos.The panel featured a wide range of perspectives and generated lively discussion with the audience on challenges and opportunities in the field.

The second day shifted focus to an interactive workshop format. Participants engaged in technical deep-dives and group discussions around key research themes such as species-area relationships, plant community structure, and the use of transfer learning to move from presence-only data to abundance modelling. Later sessions spotlighted tools and infrastructure, with demonstrations of platforms like GeoLifeCLEF, Malpolon, GeoPl@ntNet, and the cito R package for training neural networks.

The event wrapped up with open-floor dialogue to share ideas for future research and collaboration, including potential working groups and integration opportunities with existing projects.

The DeepSDM symposium was an opportunity to share recent developments, foster collaboration across disciplines, and position MAMBO’s and GUARDEN's work at the forefront of innovation in biodiversity monitoring.

Find the materials and the recorded sessions from the event here.