Thèse De Doctorat - Prédiction De La Vitesse D'écoulement Des Glaciers Par Apprentissage Profond Job Boling Texas USA,Science

Position: THÈSE DE DOCTORAT - PRÉDICTION DE LA VITESSE D'ÉCOULEMENT DES GLACIERS PAR APPRENTISSAGE PROFOND
Location: Boling

Doctoral Thesis Position In Deep Learning For Glacier Movement Tracking

The main objective of this doctoral work is to develop a robust deep learning architecture for similarity, tailored to track glacier movements from multimodal imagery. To achieve this, the research will focus on:

Designing a deep similarity model architecture, with a particular emphasis on creating a dedicated module for extracting image features, generating discriminative spatiotemporal embeddings;

A physics-informed training strategy: designing a training, validation, and testing pipeline with contrastive learning that incorporates knowledge of the physics of ice movement (e.g., flow direction), so that the model learns significant and physically plausible surface flows;

Data generation and annotation: obtaining well-annotated matching samples between two images for training is a crucial but challenging task. The goal here will be to produce training data consisting of annotated pairs of pixels in standardized image patches of potentially different sizes, derived from multimodal data: optical/optical, optical/SAR, and SAR/SAR pairs. To do this, several strategies for generating training data will be explored;

Multimodal integration: enhancing the model by incorporating multimodal data sources, relying on both empirical observations and physical principles. Additionally, inspired by recent cross-modal autoencoder techniques, modality-invariant contextual information will be passed to the network to aid in the identification of corresponding features that may be difficult to detect based on image intensities alone.

Required Profile

This work requires a genuine interest and curiosity in Earth sciences (particularly glaciology and climate science). Strong skills in statistical mathematics, deep learning, computer vision, and remote sensing are expected. Proficiency in one or more Python machine learning libraries (PyTorch, Tensor Flow, Keras) is required. Good knowledge of scientific computing with Python (scipy, scikit-learn, numpy) is also necessary.

Additional Information

The National Institute of Geography and Forestry (IGN) is a French public institution under the Ministry of Ecology and Forests. Its main role is to produce and disseminate reference data and representations (paper and online maps, geovisualizations) relevant to the understanding of the French national territory, its forests, and their evolution. Through its engineering school, Géodata Paris, and its multidisciplinary research laboratories, the institute encourages a strong and high-level culture of innovation in several fields (geodesy, forest management, photogrammetry, artificial intelligence, spatial analysis, visualization, etc.).

The candidate will work within the LASTIG laboratory, and more specifically within the Strudel team, specialized in the study of spatiotemporal structures for territorial analysis. Short trips or stays at the Institute of Geosciences of the Environment in Grenoble are envisaged. Advantages:
Flexible teleworking after a period of onboarding; sports and cultural equipment available on site; sports and cultural associations accessible within the institute; access to the campus cafeteria; 75% transport pass coverage by the employer; bicycle purchase assistance.

Specific Working Conditions

Supervisory team:
Alexandre Hippert-Ferrer and Ewelina Rupnik, both experts in remote sensing and image matching algorithms, are the main directors of this doctoral thesis and members of the LaSTIG laboratory. The work will be carried out in close collaboration with two glaciologists from the IGE (Institute of Geosciences of the Environment) in Grenoble, Antoine Rabatel and Romain Millan. Finally, Loïc Landrieu from the Imagine laboratory will provide support thanks to his expertise in deep learning.

Job

Status

Vacant as of 22/06/2026

Reference Job Title

Researcher