Speaker: Joseph Delbreil
Date: Thursday 5th of March 2026, 1:15pm
Speaker: Nour Alawieh
Date: Thursday 29th of January 2026, 1:15pm
Abstract:
Geothermal reservoirs are often hosted in highly fractured porous rocks, where fractures control fluid flow while the surrounding matrix acts as the main heat storage. Modeling coupled fluid flow and heat transfer in such systems is challenging due to the strong contrast between fracture and matrix properties and the geometric complexity of fracture networks. Existing modeling approaches involve a trade-off between accuracy and computational efficiency: implicit models rely on upscaled properties and are computationally efficient but may overlook local fracture–matrix interactions, whereas explicit models such as Discrete Fracture Matrix (DFM) provide high accuracy at the expense of high computational cost.
In this seminar, I present an efficient hybrid modeling framework for flow and heat transfer in fractured porous media, with applications to geothermal energy. A fully explicit DFM model is first introduced as a reference. Then, a Discrete Fracture Network–Dual Porosity (DFNDP) model is proposed in this work, in which fluid flow is restricted to the fracture network while heat exchange with the surrounding matrix is represented through a semi-empirical exchange coefficient. The DFNDP model is validated against the DFM reference under various flow conditions and fracture densities. Results show that the hybrid approach accurately reproduces DFM heat transfer simulations, particularly in advection-dominated and highly fractured systems, while reducing computational cost significantly. These results indicate that the DFNDP model provides a reliable and efficient alternative for simulating heat transfer in fractured geothermal reservoirs.
Speaker: Charles Dapogny
Date: Thursday 22nd of January 2026, 1:15pm
Abstract:
L’optimisation de formes est une discipline au confluent des mathématiques, de la physique et du calcul scientifique, qui suscite un engouement croissant au sein des milieux académique et industriel. En quelques mots, il s’agit d’optimiser une fonction objectif, dépendant de la “forme" (qui suivant les applications peut représenter une structure mécanique, un domaine fluide, etc.), sous certaines contraintes. Dans les applications, ces fonctions dépendent de la physique en jeu par l’intermédiaire de la solution d’équations aux dérivées partielles posées sur la forme. Ces problèmes soulèvent de nombreuses difficultés spécifiques, liées par exemple au calcul des “dérivées" des critères d’optimisation par rapport au domaine, à la représentation numérique de la forme, etc.
L’objectif de cette présentation est de brosser un panorama succinct (et biaisé) de ce domaine en pleine ébullition. Une première partie traitera de quelques généralités et présentera les principaux paradigmes d’optimisation de formes, à savoir l’optimisation paramétrique (ou contrôle optimal), l’optimisation géométrique, et l’optimisation topologique. On introduira ensuite quelques ingrédients théoriques pour l’étude des problèmes d’optimisation de formes, et notamment la méthode de l’état adjoint permettant de calculer les dérivées des critères d’optimisation par rapport au design et de leur donner une structure “exploitable" en pratique. Dans une troisième partie, on discutera de quelques points essentiels touchant à l’implémentation numérique de ces méthodes. Notamment, on expliquera comment décrire la forme optimisée et son évolution au cours du processus d’optimisation. On présentera finalement quelques applications récentes de ces techniques en mécanique des structures ou bien en mécanique des fluides.
Speaker: Michel Jébrak
Date: Thursday 15th of January 2026, 1:15pm
Speaker: Zhixiang Guo
Date: Thursday 8th of January 2026, 1:15pm
Abstract:
In recent years, deep learning has achieved transformative progress across many domains, with foundation models demonstrating remarkable generalization under few-shot and even zero-shot settings. However, training such models from scratch is often impractical in geophysics due to limited labeled data and constrained computational resources. To address this, we investigate an efficient adaptation strategy that fine-tunes foundation models pretrained in other domains for geophysical tasks, aiming for a cost-effective pathway to leverage large models. Within our adaptation framework, the adapted models consistently exhibit stronger generalization and improved performance over conventional deep learning baselines on multiple small, labeled datasets. These findings provide new insights for deploying foundation models in geophysics. This seminar also surveys recent advances in deep learning for seismic interpretation and inversion, and reports my ongoing progress on diffusion-based implicit structural modeling jointly constrained by faults and horizons.
Speaker: Ayoub Belhachmi
Date: Thursday 13th of November 2025, 1:15pm
Abstract:
Generating a valid mesh from geological interfaces is a difficult task due to the complexity of the geometrical configurations encountered in geosciences. Furthermore, uncertainties regarding the locations or even the existence of some geological discontinuities call for robust local mesh updating techniques. These methods allow for the local update of the model, instead of remeshing the entire model.
This seminar will explore techniques for local updating of tetrahedral meshes, with a particular focus on the insertion of finite surfaces such as faults in existing multi-material meshes. I will present recent progress in fault insertion using the open-source remeshing library MMG. In this approach, the finite surface to be inserted is defined as the intersection of two level set functions: one describing the surface itself, and another describing its boundary. The two level set functions can be obtained via interpolation, subject to a smoothness criteria. In the second part of the seminar, I will introduce an anisotropic smoothing energy for geological data interpolation. This high-order smoothing energy is discretized using linear finite elements via a mixed finite element formulation.
Speaker: Erwan Gloagen
Date: Friday 7th of November 2025, 1:15pm
Abstract:
Le site des lagunes de Mercier, Québec, Canada, est un site contaminé géré par pompage et traitement. Le gouvernement du Québec souhaite faire une rénovation de l’usine et en profiter pour optimiser les débits de pompage. Pour ce faire, nous avons effectué une analyse des données existantes et proposé ensuite une série de mesures ciblées : sismique réflexion, forages, diagraphies et slug tests multi-niveaux. L’ensemble de ces données a permis de réalise une modélisation stochastique des propriétés hydrauliques de l’aquifères et de le caler sur les données en régimes permanent et transitoire. Ces jumeaux numériques permettent de tester différents scénarios et de les classer selon des critères environnementaux, économiques et sociétaux.
Speaker: Christophe Antoine
Date: Thursday 23rd of October 2025, 1:15pm
Speaker: Roberto Cilli
Date: Thursday 25th of September 2025, 1:15pm
Abstract:
The seminar will first discuss a downstream application of a pretrained SentenceBERT + Random Forest (RF) for the lithological classification of texts from borehole logs. A custom lightweight architecture relying on a single Transformer module is proposed to handle contextual and positional relationships of each lithological text description. The proposed method achieves an accuracy gain of approximately 10% compared to a RF fed with SentenceBERT embeddings. Finally, a comparison between benchmark uncertainty quantification (UQ) algorithms, including Bayesian NN (Blundell et al. 2015), MAPIE Conformal Learning (Taquet et al. 2023) and Deep Ensemble (Lakshminarayanan et al. 2016) is shown. Preliminary results indicate that the Deep Ensemble UQ method seems the most reliable while still feasible in low-resource computing environments.
Speaker: Bastien Morin
Date: Thursday 09th of October 2025, 1:15pm
Abstract:
L’exploitation de ressources minières souterraines implique des systèmes d’exhaure destinés à maintenir les travaux au sec. Après l’arrêt de l’activité, la remontée de nappe provoque l’ennoyage des vides et la création de réservoirs artificiels d’eau souterraine. Ces réservoirs « miniers » constituent des hydrosystèmes originaux, fortement anthropisés mais en interaction directe avec les aquifères naturels environnants. Ils s’inscrivent dans la problématique de la zone critique, où pressions anthropiques et changements climatiques affectent le cycle de l’eau. L’étude de leur dynamique requiert des approches de modélisation spécifiques, capables d’évaluer leur évolution quantitative et d’éclairer les enjeux liés à ces nouveaux réservoirs souterrains.
La présente étude porte sur la mine de charbon de Gardanne (Bouches-du-Rhône), ennoyée depuis l’arrêt de l’exhaure en 2003. En raison de la qualité chimique défavorable du réservoir, un pompage a repris en 2010 pour éviter tout débordement par une galerie de drainage, où l’eau s’oxyderait et se colorerait, rendant impossible un rejet direct. L’objectif est de valoriser les données de volume d’eau exhaurée pendant l’exploitation pour mieux comprendre l’évolution post-ennoyage et reproduire le fonctionnement hydrologique actuel du réservoir. Pour cela, nous avons construit, avec le logiciel Gardénia (BRGM), un modèle global à réservoirs capable de simuler des chroniques de débit et de niveau de nappe.
Un premier modèle est calibré sur la période 1993-2003, afin de reproduire les débits d’exhaure à partir des chroniques de précipitations et d’évapotranspiration potentielle. Une bonne calibration a été obtenue en répartissant l’infiltration entre deux réservoirs souterrains, traduisant une composante rapide (Q1) et une composante lente (Q2) des débits d’exhaure. Ensuite, un second modèle, « pluie-niveau », a été élaboré sur la période post-ennoyage 2008-2024, en conservant le schéma conceptuel et les paramètres de sol issus du modèle d’exhaure. Le modèle montre une très bonne performance avec sept années de validation.
Ce modèle reproduit les variations de niveau du réservoir et constitue un cadre robuste pour explorer des scénarios climatiques. En suivant l’approche narrative proposée par DRIAS, des simulations prospectives permettent d’anticiper l’évolution future des niveaux et d’optimiser la gestion opérationnelle du pompage dans un contexte de changement climatique.
Speaker: Jeff Boisvert
Date: Monday 21st of July 2025, 2pm
Recording (Restricted access - Sponsors only) : Video
Abstract:
The earth sciences are being transformed by advances in machine learning (ML) and artificial intelligence. From optimizing mineral estimation and hydrocarbon production to improving wildfire prediction and management, these methods offer exciting opportunities for modeling and decision-making. However, these advances bring challenges with model validation, which is critical for ensuring that predictions are robust, reasonable, and actionable.
This lecture will delve into the evolving role of ML in the mining, hydrocarbon, or wildfire industry, highlighting successes, pitfalls, and future prospects. "The Good" will explore case studies and implementations where ML has significantly improved modeling, decision making, and inference. "The Bad" will examine common pitfalls, including data biases, overfitting, and the misuse of algorithms without understanding domain constraints. Finally, "The Ugly" will confront the ethical and operational risks posed by poorly validated models, emphasizing the importance of transparency and domain experts.
This lecture will not only focus on ML methods, but will also consider how to validate all types of earth science models including estimates, simulations, and decision making. We will discuss best practices for integrating ML models into traditional workflows while addressing the complexities of model validation.
Speaker: Amandine Fratani
Date: Thursday 26th of June 2025, 1:15pm
Abstract:
The construction of geological models in sedimentary basins is largely constrained by the interpretation of faults and horizons in seismic and drillhole data and by associating observations into distinct entities (e.g., forming a single fault or one horizon). Due to the sparsity and incompleteness of data, several fault networks can usually be drawn from a given set of observations. This problem has been considered using graph formalism with nodes carrying the fault observation and the edges carrying information on the potential that they are associated. This potential has previously been proposed to be computed using machine learning, specifically the application of a Random Forest. However, the lack of open access structural models limits the use of machine learning. Therefore, this methodology has only been tested on partially interpreted cases. To generalise the approach, this work presents a database under development comprising synthetic structural models featuring normal faults. A random geological history and model generation code, Noddy, has been modified to include more realistic fault events. Faults are grouped into families where fault from a family have similar orientation for fault surfaces. Each family is defined as a mean dip and a mean dip direction, select randomly. A fault orientation is defined by sampling a Kent distribution centred on the dip and dip direction of the family to which it belongs. The resulting geological models are imported into geological modelling software where the surfaces are smoothed. Fault observations are then sampled in these models and will be used to train a Random Forest to retrieve the potential associations.
Speaker: Mohamed Sherif Mahrous
Date: Thursday 19th of June 2025, 1:15pm
Abstract:
A significant production of hydrogen is expected inside geological nuclear waste repository. The gaseous phase is foreseen to modify the flows and mechanical conditions of the rock and engineered material. As a part of EURAD 2 project (2024-2028), this work plans to implement the coupling between the different physical (THMC) processes involved in gas migration and to upscale results into the continuum scale. Towards this end, a pore scale Hydro-Mechanical-gas code has been developed in EURAD1 project, based on Smoothed Particle Hydrodynamics, to study water-gas migration accounting for drying within a deformable solid (elastic with thermodynamical damage model). The main objectives of this work are:
1. Apply the THMC couplings in the near- and far-field at the micro and meso scale (order of tens of pores) in the already existing SPH code.
2. Investigate the temperature dependency of multiphase flow parameters.
3. Compute effective properties (e.g., saturation curve, relative permeability, poromechanical properties).
4. Explore the consequences of usual simplifications.
5. Provide benchmarking and training data for predictive surrogate models.
6. Deliver results for code comparisons and experimental benchmarking.
Speaker: Imadeddine Laouici
Date: Thursday 12th of June 2025, 1:15pm.
Abstract:
Building structural models of geological entities is generally addressed as an interpolation problem that requires human experts to interpret input data and use knowledge (Wellmann and Caumon, 2018). Although experts can effectively interpret, their interpretations can be subjective and occasionally prone to error (Bond, 2015). This is largely due to under-sampling of data, requiring experts to make choices in the selection and preparation of these data and knowledge (Bond et al., 2012), and selection and configuration of modeling algorithms (Caumon et al., 2009). Modeling algorithms also do not reflect the complex expert interpretation process, as they incorporate only a portion of the knowledge typically held by experts and have limited ability to directly interact with experts during the interpretation process itself. This makes it challenging to build geologically complex models and systematically identify and address inconsistencies in a model. A crucial step toward resolving these issues is the formalization of the interpretation process and the explicit use of formalized knowledge. In this work we develop and prototype such a formalization. A prototype algorithm and tool (Figure 1) are presented and applied to simple folding structures, and the results are favorably compared to existing approaches. This comparison highlights the potential of the proposed approach to reduce the need for expert involvement and increase the range of knowledge utilized.
Speaker: Oussama Larkem
Date: Thursday 22nd of May 2025, 11am.