Les offres
Si tu es intéressé·e par une offre de stage, rapproche-toi de l’équipe pédagogique de ta formation, pour faire valider l’offre de stage proposée.
Retrouve toutes les offres de stage et d’emploi.
17-11-2026
Se ferme:
14-01-2027
Vues: 8
The IASI (Infrared Atmospheric Sounding Interferometer) instruments onboard the Metop satellites have provided continuous infrared observations of the atmosphere since 2007, enabling the detection of trace gases emitted by extreme events such as wildfires, volcanic eruptions and pollution episodes. Its successor, IASI-NG (New Generation), launched in August 2025, offers a finer spectral resolution and a better signal-to-noise ratio, opening the way to a new generation of atmospheric monitoring systems.
This internship is part of a PhD thesis that aims to detect extreme events automatically and in near real time, directly from IASI spectra, using artificial intelligence (AI). Supervised neural-network models (ResNet, Baseline) have been trained on an atlas of fire events built from anomalies of IASI Level-2 products (CO, NH₃, PAN, CH₃OH, HCOOH). They give promising results, including on events not seen during training. The code is currently being extended to volcanic eruptions (SO₂ atlas) and dust outbreaks (dust anomalies). However, the models rely on a limited set of species, those for which Level-2 products exist.
Hyperspectral Range Index (HRI) data are available for the entire IASI period and give access to additional species. The HRI is a dimensionless index computed from the IASI radiances in a spectral range specific to a target gas. It quantifies the strength of the gas spectral signature in each observation, without a full retrieval. The aim of the internship is to exploit these data to extend the supervised detection code to additional biomass-burning tracers (HCN, C₂H₄, C₂H₂, CH₃COOH). This extension is expected to improve the detection and characterization of fire events. It will also allow training and evaluation over a longer period, and prepare the transfer of the detection models to IASI-NG.
Internship tasks
- Get familiar with the event atlases, the current multi-event supervised code and the HRI data.
- Build the HRI-based database over the IASI period
- Integrate the additional species into the supervised models and retrain them.
- Compare the performance of the extended models with the current ones (event intensity, false detections, day/night).
- Evaluate the models on selected known events, including events outside the training database.
Prerequisites
- Master’s level (M2) in data science, atmospheric physics, remote sensing, or a related field.
- Background in machine learning, with experience of supervised neural networks
- Good programming skills in Python and familiarity with Linux.
- Experience with satellite or atmospheric data is a plus.
- Good scientific reading and writing skills.