PhD Position F/M Next-generation Bayesian inference for Earth and Space Science

5 Feb 2026
Job Information

Organisation/Company

Inria, the French national research institute for the digital sciences

Research Field

Computer science
Mathematics

Researcher Profile

First Stage Researcher (R1)

Application Deadline

5 Mar 2026 - 00:00 (UTC)

Country

France

Type of Contract

Temporary

Job Status

Full-time

Hours Per Week

38.5

Offer Starting Date

1 Oct 2026

Is the job funded through the EU Research Framework Programme?

Not funded by a EU programme

Reference Number

2026-09787

Is the Job related to staff position within a Research Infrastructure?

No

Offer Description

Thanks to advances in space observation, large, high-resolution datasets are now available for studying Earth’s climate, as well as the environments of Mars and the Moon. Extracting physical information from these datasets involves solving complex inverse problems that link measurements to their underlying causes. This PhD interdisciplinary programme focuses on Bayesian methods for estimating physical parameters from high-dimensional remote sensing data. Rather than relying on traditional assumptions, it uses modern generative AI models, particularly diffusion models, to better represent priors and sample the posterior probability distribution. The work also aims to improve inference efficiency by reducing dimensionality, reusing computations, and combining multiple measurements. The PhD candidate will develop, test and validate these methods. They will then be applied by him (her) to selected case studies for mapping the surfaces of Mars, the Moon and Earth using data from current space missions (MRO NASA, TGO ESA, EnMAP DLR).

Context

The spectacular development of space systems and sensors, for observing the Earth and other Planets, provides access to numerous geophysical, geochemical and biophysical parameters over vast areas with increasingly high spatial resolution and revisit frequency. Such infrastructures are crucial to understand our changing world and climate but also the Martian or the Moon environments and their degree of habitability. This involves theory but also the integration of observational information into models through data assimilation and model inversion.  In this domain as in many fields of applied science, researchers face high-dimensional non-linear inverse problems. The fundamental challenge is to estimate physical parameters (the “causes,” denoted as x) from observed signals (the “effects,” denoted as y).

Bibliography

Douté, F. Forbes, S. Borkowski, S. Heidmann, and L. Meyer. Massive analysis of multidimensional astrophysical data by inverse regression of physical models. In GRETSI 2023 - XXIXème Colloque Francophone de Traitement du Signal et des Images, 2023.

Douté, S., Forbes, F., Borkowski, S., Meyer, L., Heidmann, S., 2024. Massive analysis of multi-angular images by inverse regression of reflectance models for the physical characterization of planetary surfaces., in: European Planetary Science Congress. pp. EPSC2024-535. https://doi.org/10.5194/epsc2024-535

Haggstrom, P.L.C. Rodrigues, G. Oudoumanessah, F. Forbes, U. Picchini. Fast, accurate and lightweight sequential simulation-based inference using Gaussian locally linear mappings. Transactions on Machine Learning Research, 2024

Kugler, F. Forbes, and S. Douté. Fast Bayesian Inversion for high dimensional inverse problems. Statistics and Computing, 2021.

Nguyen, D.T., Jacquemoud, S., Lucas, A., Douté, S., Ferrari, C., Coustance, S., Marcq, S., Meygret, A., 2025. Mapping the surface properties of the Asal-Ghoubbet rift by massive inversion of the Hapke model on Pleiades multiangular images. Remote Sensing of Environment 322, 114691. https://doi.org/10.1016/j.rse.2025.114691

Eric Tatull, Silvere Gousset, Sylvain Douté, Fast inversion of hyperspectral observations using Gaussian Locally Linear Mapping, proceedings of the Société Française de Photogrammétrie et Télédétection (section hyperspectrale) SFPT – GH 2025

 

The PhD work focuses on Bayesian methods to estimate physical parameters from large volume and high dimensional remote-sensing measurements. Because these inverse problems are ill-posed, prior information is required to obtain meaningful solutions. This research aim at replacing traditional hand-crafted priors with learned priors based on modern generative AI, in particular diffusion models that can capture complex data distributions. These models are further extended to sample efficiently from the full Bayesian posterior, which is usually computationally expensive. Conditional diffusion models will be investigated as fast, data-driven alternatives to classical sampling methods. Key objectives of the PhD include implementing dimension reduction strategies, reusing computations across many observations and efficiently combining multiple complementary measurements. This will enable scalable and accurate inference for large Earth and planetary observation datasets. The development of a diffusion-based framework will also pave the way to Bayesian Optimal Experimental Design (BOED) to optimize compact spectral imagers such as ImSPOC (Imaging SPectrometer On Chip) for embedded E&S applications, including atmospheric CO₂ and CH₄ monitoring.

This PhD work will also play a central role in developing, testing, and validating the proposed statistical and machine learning methods, and in applying them to concrete case studies in planetary and Earth observation. These applications will focus on Mars and Moon exploration, using large, multi-dimensional remote-sensing datasets product by current space missions to characterize and map surface materials (such as mineralogical composition, texture, and micro-roughness) and to quantify uncertainties. In collaboration with partner laboratories, the PhD candidate will work on selected case studies where the performance of the algorithms and the quality of the products they generate will be systematically evaluated.

Keywords

Bayesian statistics, AI-assisted inverse problems, planetary remote sensing, and environmental monitoring.

 

 

Where to apply

Website

https://jobs.inria.fr/public/classic/en/offres/2026-09787

Requirements
Skills/Qualifications

Strong proficiency in at least one of these domains: astrophysics, data science (statistics, inference, and machine learning), or physical remote sensing/Earth observation. Strong skills in scientific programming and digital techniques.

Scientific publications in conference proceedings or international journals. Integration of code into the xLLiM toolbox jointly developed by INRIA/Statify and IPAG (https://gitlab.inria.fr/xllim/xllim )

 

 

Specific Requirements

The thesis will be co-supervised by Sylvain Douté (CNRS Research Director, HDR) from the PLANETO team at IPAG and Florence Forbes, Research Director and head of the Statify team at INRIA Grenoble. 

IPAG is an internationally recognized research institute in planetology and astrophysics. The doctoral student will be provided with a modern laptop equipped with a GPU and funding to participate in summer schools and national and international conferences. The thesis work will also be supervised by a CNES research engineer as part of the Planetary Surfaces Data and Services Center (PDSSP).  In terms of computation, we will strongly rely on the GRICAD computing infrastructure (CPU, GPU, and high performance storage).

Languages

FRENCH

Level

Basic

Languages

ENGLISH

Level

Good

Additional Information
Benefits

• Subsidized meals
• Partial reimbursement of public transport costs
• Leave: 7 weeks of annual leave + 10 extra days off due to RTT (statutory reduction in working hours) + possibility of exceptional leave (sick children, moving home, etc.)
• Possibility of teleworking (after 6 months of employment) and flexible organization of working hours
• Professional equipment available (videoconferencing, loan of computer equipment, etc.)
• Social, cultural and sports events and activities
• Access to vocational training
• Social security coverage under conditions

2300 euros gross salary /month

Selection process

Applications must include a CV, covering letter, copy of diploma and valid proof of disabled worker status.

Applications must be submitted online via the Inria website. Processing of applications submitted via other channels is not guaranteed.

Website for additional job details

https://jobs.inria.fr/public/classic/en/offres/2026-09787

Work Location(s)

Number of offers available

1

Company/Institute

Inria

Country

France

City

Montbonnot

Contact

City

LE CHESNAY CEDEX

Website

http://www.inria.fr

Street

Domaine de Voluceau - Rocquencourt

Postal Code

78153

STATUS: EXPIRED

• X (formerly Twitter)
• Facebook
• LinkedIn
• Whatsapp
• 
  More share options
  • E-mail
  • Pocket
  • Viadeo
  • Gmail
  • Weibo
  • Blogger
  • Qzone
  • YahooMail