Speaker: Paul Baville

Date: Thursday 1st of April 2021, 1:20 pm.


To evaluate the difference between a three-dimensional Bézier triangle cubic interpolation and a parametric three-dimensional surface, I compute the absolute volume in between. The Bézier triangle is defined within the Bézier space (u,v,w are the barycentric coordinates) and the parametric surface is defined in the three-dimensional space (x,y are the Cartesian coordinates).
The Bézier triangle is defined by three points in between a three-dimensional cubic interpolation is computed. This triangle is meshed in 100 sub-triangles (each edge is divided into 10 parts) defined by 66 nodes. These nodes have barycentric coordinates which have a Cartesian equivalent and the absolute volume is computed by multiplying the vertical difference between nodes and the parametric surface and the planar surface of the Bézier triangle.
The goal of this seminar is to compare two ways to compute the vertical difference (at nodes or at sub-triangle barycenters) and to discuss another way to obtain similar results by limiting the computational time.

Speaker: Marius Huber

Date: Thursday 18th of March 2021, 1:20 pm.


Deep seated gravitational slope deformations (DSGSDs) are a commonly observed type of slope instability process found in mountain ranges all over the world. Their rates of movement differ fundamentally from more superficial rapid mass movements known in high relief terrain, with displacements characterized by long periods of dubious activity or time spans of inactivity. If some DSGSDs do not induce catastrophic failures but instead cease activity, others might be in a preparatory stage before a large volume catastrophic landslide. This last evolutionary stage could be favored by geological and fracturing conditions, as well as topographical, climatic, or seismo-tectonic conditions. However, due to the scarcity of very large rockslides, this possible catastrophic evolution and the reasons for it are still poorly documented and understood.

In order to explore the conditions that lead or not to an unfavorable catastrophic failure, we propose to use discrete element models to investigate DSGSDs from a mechanical viewpoint. After validating our modeling approach based on comparisons with Limit Equilibrium Analysis of simple slope geometries, we explore preconditions and mechanisms acting towards failure considering the influence of mechanical rock properties, slope topography, and pre-existing structures. Furthermore, we can implement our method on real existing case studies by shaping our simulations with digital terrain models (DTMs).

Comparison of our modeling results with those DSGSDs for which long-term and recent activity has been documented (e.g. La Clapière case in the South Western Alps of France) provides insight into their prevailing patterns of deformation (continuous or stepwise over time) and the conditions (geological, fracturing or topographic) for a catastrophic landslide or progressive stabilization. This may lead to better understanding and prediction of behavior for large scale deep seated deformation in mountainous areas, especially relevant for estimating natural hazards.

Speaker: Mustapha Zakari

Date: Thursday 25th of February 2021, 1:20 pm.


Unstructured grids can offer flexible and accurate geometrical approximations of reservoir models. Upscaling methods like Darcy flow-based numerical approaches can produce realistic and accurate reservoir grids. Initial flow-based upscaling approaches were developed for cartesian grids. Flow simulations were computed on groups of fine-scale grid elements covering one or many coarser elements (local, extended local, global methods). The main difficulty in flow-based methods is to define accurate fine-scale boundary conditions allowing to recover realistic coarse-scale fluxes. While cartesian grids use orthogonal directions to define vertical or horizontal pressure gradients across the simulation domain in unstructured meshes no more vertical or horizontal coarse grid alignment can be used. We investigate here fluid flow immersed boundary methods to allow incorporating cartesian-like vertical and horizontal boundary conditions for Cartesian like upscaling on unstructured meshes. I will briefly present cartesian-like flow-based upscaling approaches and then give a short review of immersed boundary methods from cartesian grids to unstructured mesh ones.

Speaker: Kou Du

Date: Thursday 18th of February 2021, 1:20 pm.


The aim of this presentation is to extend recent works devoted to the study of the effect of 3D pores of concave shape embedded in isotropic matrix to the case of transversely isotropic (TI) matrix. In the first part, we'll present a brief review of numerical calculation of compliance contribution tensors (which have been introduced in this seminar last year). Then, the approximate relations for the compliance contribution tensor of pores of two reference shapes, supersphere, and axisymmetrical superspheroid, are developed on the basis of 3D Finite Element Modelling, and known exact solutions for the limiting cases of spherical pores and circular crack. In the last part application to effective elastic coefficients of transversely isotropic materials in the frame of homogenization theory is presented to illustrate the impact of concavity parameter on overall properties.

Speaker: Guillaume Caumon

Date: Thursday 11th of February 2021, 1:20 pm.


In this seminar, I will present some work recently done in collaboration with Equinor on the use of magnetostratigraphic data to better constrain the ages of formations in the Central Graben. For this, we used individual correlation to 5 wells to the geomagnetic polarity time scale (GPTS) using the Cupydon code (Lallier et al, EPSL 2013). As the recorded series are relatively short, we also started to use magnetic chrons to constrain log-based multi-well correlation. The ultimate goal is to use these correlation results to create a composite magnetostratigraphic series that could be matched to the GPTS. Significant uncertainty yet exists about the results, as no external dating constraints are available. I will suggest some avenues for improving the methods and discuss some sources of uncertainty in the process.


Speaker: Ahmad Mostafa

Date: Thursday 4th of February 2021, 1:20 pm.


Coal beds are dual permeability systems characterized by a porous matrix enclosed within sets of orthogonal fractures known as cleats. Production of coalbed methane (CBM) consists of desorbing methane from the low permeable coal matrix to the high permeable cleat system. Unlike in conventional reservoir exploitation, sorption mechanisms cause shrinkage and swelling of the matrix which increase the complexity of the phenomena at stake, leading to complex reservoir behaviors in terms of production. A 3D discrete element method (DEM) coupled to a pore scale finite volume method (PFVM) is used here to better understand the different mechanisms at stake. The model, implemented in the open-source software Yade Open DEM (Smilauer et al., 2015), is an offspring of the hydro-mechanical model proposed by Catalano et al. (2014). The coal matrix is treated as an assembly of bonded particles interacting one with another through elastic-brittle contact laws. The pore space is discretized into tetrahedra, generated from a regular triangulation of the particle assembly. Both Knudsen and surface diffusion, as well as sorption processes, are modeled considering the coal matrix as a microporous material. The method is hydro-mechanically coupled in the sense that changes in pore pressure produce hydrostatic forces that deform the solid skeleton, while deformation of the pore space induces pore pressure changes that promote interporal flow. In addition, sorption induced deformations are taken into account by considering an additional pressure term related to the concentration of gas within the medium (the so-called solvation pressure). In this work, we first present the model and its constitutive equations. We assess its capabilities by comparing its predictions to well established solutions describing diffusive flow in porous media as well as to classic poroelasticity concepts. In particular, we focus on the influence of sorption induced deformations on the Biot coefficient estimation. Finally, we compare the model predictions to swelling test data from the literature to illustrate its consistency.

Speaker: Enrico Scarpa

Date: Thursday 28th of January 2021, 1:20 pm.


Nowadays, karst aquifers are strategic groundwater resources for the provision of drinking water all around the world. One-fourth of the population worldwide drinks water stored in karst aquifers, which are extremely vulnerable to various pollution sources. However, due to the complexity of karstification patterns and geometries, it is not easy to simulate groundwater flow in karstic aquifers. Over the last years, different techniques have been proposed to model karstic conduit networks (e.g., MODFLOW-CFP, GROUNDWATER). Those techniques include methods based on purely statistical approaches, object-based methods, genetic algorithms that mimic the physico-chemical processes that bring to conduit formation. Consequently, new possibilities have been introduced to evaluate the impact of karstic network architectures on groundwater flow accurately.
This work focuses on modeling the karst network stochastically using a geostatistical technique known as Multiple-Point Statistics (MPS) to characterize the karst architecture in the Dogger aquifer at the hydrogeological experimental site (HES) of Poitiers, France. In particular, here we apply the Direct Sampling (DS), an MPS simulation method, with the recent implementations proposed from the literature. Field data include core description, hydrogeological tests, and seismic surveys. A critical review of the available data set suggests the occurrence of cave patterns and several, well-identified, karst layers. The first step consists of reviewing all the (available data) collected in the last ten years and then studying the correlation between the geological, geophysical, and hydrogeological information. In the second step, DS simulations are performed to obtain several equiprobable realizations. These different DS outcomes can be compared and can estimate a global 3D karstic framework architecture of the Dogger aquifer. The resulting representation of the geological heterogeneity can support flow modeling at the HES and explore the impact of the karstic framework on the dynamic flow. 

Speaker: Fabrice Gbewade

Date: Thursday 21st of January 2021, 1:20 pm.


Prévoir l’endommagement hydraulique dans les systèmes géologiques constitue un challenge majeur dans l’ingénierie de subsurface. Ces processus très couplés hydro-mécaniquement (HM) qui naissent à différentes échelles, de l’échelle de la fissure (centimètre) à celle de la fracture ou de la faille (mètre ou décimètre), constituent des problèmes majeurs en géo-ingénierie qui ne sont pas encore complètement compris. L’objectif principal de ce projet est donc d’améliorer notre compréhension des processus HM qui prennent place en subsurface pour une meilleure évaluation des risques associés et de proposer des méthodes non intrusives permettant d’améliorer l’évaluation de ces processus. La présentation portera sur une synthèse bibliographique des essais de fracturations hydrauliques avec écoute acoustiques présents  dans la littérature en fonction de paramètres hydromécaniques.

Speaker: Capucine Legentil, Paul Baville & Zoé Renat

Date: Thursday 14th of January 2021, 1:20 pm.


Ce séminaire sera l'occasion de faire découvrir aux 3As l'IAMG (International Association for Mathematical Geosciences) et son Student Chapter. L'IAMG est l'association scientifique la plus proche des thématiques de recherche de l'équipe, et possède un Student Chapter à Nancy, qui est géré par les doctorants de l'équipe.

Le bureau de cette année est composé des doctorants de et 3ème année (Capucine Legentil, Zoé Renat et Paul Baville).