On December 11, 2025, the WILDLABS Geospatial and AI subgroups hosted a joint panel on AI developments for geospatial applications. The panel included Dr. Mahya G.Z. Hashemi (NASA Goddard), Dr. Isla Duporge (Princeton), Valerie Zermatten (EPFL), and Dr. Li Mi (ETH Zurich). The panel was attended by nearly 100 WILDLAB members worldwide, it was great to see so many of you!
Mahya presented on NASA’s new Hydrology Co-pilot, a cloud-native AI system that enables drought monitoring and hydrological data analysis. Isla presented on the ability to track animals with high precision, using AI-based methods to count animals and analyze their movement patterns. Valerie presented her research on integrating text data with Earth observations to predict environmental variables. Li presented GeoExplorer, a system that uses “active geo-localization” that guides drones in search and rescue scenarios using multiple modalities.
The event led to great discussions, and we hope the community continues to find new ways to link AI for geospatial efforts. The recorded version is available here!
11 February 2026 11:08am
Thanks to all the organisers and moderators (@annavallery , @cbreen and @ViktorDo), speakers (@Isla, Mahya, Valerie and Li) and participants for attending! We had a very interesting conversation with Isla on her research during the break-out groups and also addressed some of the questions that came up during her presentation:
- What imagery did you use?
Digital Global Foundation provides imagery at no cost for research purposes only. Isla could get access to satellite imagery such as GeoEye-1. Cost is still a limiting factor for operational use cases outside of research projects. Back in December there was an opportunity to get VHR imagery (Pleiades Neo) from Airbus Foundation through Connected Conservation (this is an annual opportunity that they offer).
- How does forest canopy or clouds impact imagery use for research purposes and results?
Clouds are an issue. When you put in a request for the imagery tasking you can specify % cloud cover threshold. Unfortunately, you cannot see forest animals using optical data if they are under the canopy.
Also posting here her papers in case anyone is interested:
- AI-based satellite survey offers independent assessment of migratory wildebeest numbers in the Serengeti
- Automated rhinoceros detection in satellite imagery using deep learning
- Deep learning enables satellite-based monitoring of large populations of terrestrial mammals across heterogeneous landscape
- Using very-high-resolution satellite imagery and deep learning to detect and count African elephants in heterogeneous landscapes.
We also touched on the use of spaceborne/airborne radar for detecting animal migration and we would be keen to hear about any research that is being done on that topic (if any). If anyone has any papers or is working on an R&D project please add the to this thread!
Elsa
Fauna & Flora