On 19 May 2025 in Cañete, Peru, the sirens sounded even before the tremors: a first for the country. This early-warning system, designed by the IRD and the Instituto Geofísico del Perú, uses just a single seismic station — as soon as the first seismic waves are detected, AI algorithms estimate the magnitude and location within seconds, then issue an automatic alert. Just a few seconds, but enough time to get out of a building or reach a safe area, and thus to save lives.

detectors, earthquakes, tsunamis

The project DIG-AI for Deciphering plant genotype-phenotype Interactions using knowledge Graphs and AI, aims to leverage AI to improve crop production capacities in developing countries through a better understanding of genotype-phenotype relationships in a context of climate change.

agroecology, developmental genetics in plants, artificial intelligence

This project aims to generate ecological indicators and predictive models of biodiversity in disturbed ecosystems by combining artificial intelligence methods with known assessment approaches. It will thus enable the creation of the first marine biodiversity knowledge base and, subsequently, the development of prediction and interpretation models. It also aims to reveal intelligent solutions for nature in order to ensure the sustainability of coastal socio-ecological systems.

marine biotechnology, marine ecology, biological oceanography

Predomics is a new machine learning approach designed for metagenomic data, which addresses the lack of interpretability in current predictive models for microbiome biomarkers. Inspired by interactions within microbial ecosystems, it bases its decisions on a simple score, obtained by adding, subtracting or dividing the cumulative abundances of microbiome measurements.

nutrition, environmental biotechnology, genomics

The AIME project utilises artificial intelligence to integrate heterogeneous data — satellite imagery, underwater images and videos, environmental DNA, GPS coordinates and telemetry — and to produce marine biodiversity indicators at different scales. The methods combine deep learning, ontologies and knowledge bases, natural language processing, machine learning and dynamic Bayesian networks, with a hybridisation of symbolic and sub-symbolic approaches aimed at enhancing the robustness and accuracy of the indicators.

marine biology, bioindicators, oceanography, artificial intelligence

DISoft uses innovative technology based on high-resolution melting (HRM) analysis of DNA using artificial intelligence. This technology was developed through a long-standing collaboration between Innovative Diagnostics, the French National Research Institute for Sustainable Development (IRD) and the University of Montpellier.

statistical data processing, dna, other medical technologies for diagnosis, health care research

"The rise of machine learning allows us to consider a completely different approach, without any physical assumptions, based solely on empirical data. The idea is to feed the algorithms with all the quantitative observations of natural phenomena that we can muster, to see if they can identify factors that have so far escaped the attention of specialists."

geophysics, geodynamics, artificial intelligence

There is an important gap between biodiversity research and the management of natural areas. This research project aims to reduce this gap by proposing spatial planning methods that robustly and accurately integrate socio-ecological issues. Artificial intelligence will play a central role and will make it possible to remove the methodological obstacles that prevent us from properly addressing the complexity and heterogeneity of sustainability issues in the management of ecosystems.

biology, artificial intelligence

The management of snapper stocks is key to the sustainability of New Caledonia's fisheries. To assess their abundance and diversity, scientists from the IRD and their partners have deployed baited stereo cameras on the seamounts where these fish live, at depths of between 47 and 557 metres. Artificial intelligence then takes over to analyse the hours of recorded video.

feed resources, agricultural economics, biological oceanography, fisheries