We are advertising a 2-year fully funded postdoc position to work at IGE - Université Grenoble Alpes , in Grenoble (France), on physics-informed machine learning modelling of glaciers. The goal will be to contribute to the development and application of ODINN.jl, a new global glacier model leveraging the Scientific Machine Learning ecosystem of the Julia programming language. The scientific focus of this position will be on improving our understanding of basal sliding and creep for glacier ice thickness inversions at large scales from novel remote sensing datasets.
Project objectives
The recruited postdoc will develop and use hybrid modelling methods, based on Universal Differential Equations, combining partial differential equations describing ice flow with neural networks, in order to parametrize different glacier processes (e.g. basal sliding, creep) from globally available proxies and novel remote sensing observations. These developments will be performed in both the ODINN.jl model and in the SciML stack. We will explore the use of symbolic regression to interpret the nonlinear functions learnt by the neural networks. These inferred parametrizations will then be used to improve our understanding of these physical processes at large scales, which in turn will be used to improve regional ice thickness estimates. Once good ice thickness inversions have been achieved, we will perform glacier evolution projections under different scenarios of climate change in selected catchments. This will build the necessary first steps towards a fully coupled differentiable glacio-hydrological modelling framework.
Due to the interdisciplinary nature of this position, we will offer some flexibility to tailor the project to the candidate’s profile and interests.
Preferred qualifications
- PhD in Earth sciences, computer science, or machine learning.
- Experience in inversions with numerical methods for differential equations and/or machine learning will be necessary.
- Good proficiency in programming, ideally in Julia and/or Python, with previous experience in developing geoscientific models.
- Experience in differentiable programming will be an advantage (e.g., in Julia and/or JAX-Pytorch-Tensorflow).
- A good level of English will be required to successfully interact with the international project collaborators and to properly disseminate the scientific results of this project.
- A high level of autonomy and a great degree of curiosity about the research topic.
We highly encourage underrepresented groups of our community to apply to this position.
Contract and location
The position will have a duration of 2 years, and it will take place in person at the Institut des Géosciences de l’Environnement (IGE), Université Grenoble Alpes, in Grenoble, a highly dynamic city located in the French Alps. Salary starts at €3020 gross per month, depending on experience, and includes an additional 12,500€ of financial support for all related research activities (e.g. conferences, papers, scientific visits…). The contract includes 45 days of paid holidays, social security benefits, and the possibility to work in a highly interdisciplinary institute and university, with extremely easy outdoor access from your doorstep.
How to apply and find more details
You can find all the details and apply for this position here. The source code of the ODINN.jl modelling framework can also be found here.
For any additional questions, contact Romain Millan and myself through our email addresses: romain.millan@univ-grenoble-alpes.fr and jordi.bolibar@univ-grenoble-alpes.fr .
Looking forward to many interesting applications!
