Proceedings are already available for sponsors download.

Dear Colleagues, dear Friends,

It is a pleasure to introduce this new volume of proceedings of the RING Consortium Meeting and to look back at a year of Research in Integrative Numerical Geology. As each year, this volume combines papers presenting new methods and applications to integrate not only data, but also concepts in subsurface models. These papers were written by the students and researchers of the RING Team, most of them supported by the RING consortium members, and papers contributed by sponsors and guests. I would like to express my sincere appreciation to all the authors for their contributions and to the industrial and associate RING Consortium members for their invaluable support.

In a geomodeling study, geological structures often play a large role in defining the subsurface compartments and controlling the physical parameter fields. For a given structural interpretation of the data at hand, forward geophysical modeling can be used to identify inconsistent areas, as illustrated by the Canadian REE deposit case study by Katerina Poliakovska et al. For the structural modeling itself, a variety of implicit modeling approaches have been proposed. The benchmark study presented by Guillaume Caumon et al. reviews and compares these methods on several standard data sets which display specific geological challenges. Gaining insights about implicit structural modeling methods is important, because they are becoming key components of modern structural uncertainty quantification approaches. This is exemplified by Anne-Laure Tertois et al., who present a workflow to generate many possible structural geometries reflecting average velocity uncertainty. However, efficiently building and processing geological models also calls for managing the level of structural detail in data and models. For this, Ranee Joshi et al. present an important work on automatic geological map subsampling. Finally, Gautier Laurent and Lachlan Grose show, using discrete model reduction, that data perturbation is not sufficient to capture structural uncertainty and advocate uncertainty models driven by geological concepts.

This calls for defining and manipulating such concepts in a numerical way. Stratigraphy is one such example, where many geological and physical concepts are used to define and correlate stratigraphic units. Assisted stochastic stratigraphic correlation has been a playground for RING for about a decade. It provides a new way to investigate about key components and to assess the impact of stratigraphic uncertainties. This year, Christophe Antoine and Guillaume Caumon present improvements in the multi-well correlation code WeCo, with the introduction of nested stratigraphic units. This is expected to enable a more efficient exploration of correlation uncertainties. On the geological side, Paul Baville et al. propose a new correlation rule based on paleotopography estimation combining facies and dipmeter data.

Among the various concepts that we manipulate in geology, the notion of rock type is certainly key, but it often calls for significant data curation by experts. To automate and account for classification uncertainty, Antoine Bouziat et al. share some first results of the RockNetTM project, which appraises several computer vision and machine learning algorithms on a rock picture data base.

In salt tectonics, the existing geological concepts are difficult to integrate with data-driven approaches. This has led RING to propose methods combining level sets and random fields to address salt-related uncertainty. This year, Sacha Görne et al. adapt the pseudo-distance field stochastic modeling approach to a context where salt top has been interpreted from a set of cross sections. On the same topic, Nicolas Clausolles et al. use stochastic salt modeling to automatically propose multiple velocity models in regions which are poorly constrained by seismic imaging, and share some results toward automatic velocity model updating in salt tectonics contexts. Another possible avenue to validate or reduce uncertainty about stochastic salt interpretations can be to use structural restoration. This is the goal of the paper of Melchior Schuh-Senlis et al., which shows a nice progress to deal with large deformation using viscous restoration and a meshless numerical scheme.

Modeling realistic rock unit boundaries often serves the purpose of numerical forecasting of physical processes in the subsurface. For this, meshing the geological volumes is often needed. Pierre Anquez and Arnaud Botella present a general workflow and some results on three-dimensional mesh repair and remeshing. Capucine Legentil et al. also address this problem from a model updating perspective. To illustrate the approach with a concrete use case, they consider the effects of oil/water contact uncertainty on wavefield simulation. As unstructured meshes are not necessarily accepted in mainstream simulators, Enrico Scarpa et al. also presents the first steps of a method to easily discretize channel-based deposits in a conventional cornerpoint grid.

For a given mesh, several discretization schemes may be used to numerically solve the flow equations and account for specific fault properties. The workflow and benchmark proposed by Margaux Raguenel et al. is, therefore, fundamental to gain insights on the impact of these modeling choices on the forecast. In complement to such detailed and bottom-up modeling, it can be useful to use flow information directly to interpret the behavior of subsurface reservoirs. This is the approach taken by Christophe Reype et al., who propose a Bayesian and spatial statistics model to simulate fluid sources based on a set of punctual geochemical measurements.

Geomodeling is an interpretive process, which mobilizes very diverse data, principles, objectives and computer programs. Interface is, therefore, very important, starting with data transfer between various applications. For this purpose, Jean-Marie Léonard presents a new RESQML 2.2 management to bridge the gap between geomodeling and the AUTODESK building information modeling and computer-aided design software. Interfacing geomodels with the human visual system is also a useful but challenging task. Björn Zehner et al. highlights the need to incorporate uncertainties in this visualization, and he reviews and discusses the possible options to visualize uncertainties on web browsers. Another possibility for innovative interfaces is to give geomodels the ability to talk. Antoine Bouziat et al present some interesting work towards this, by evaluating how natural language processing can be used to extract formal petroleum system information from documents, and to analyze basin simulation results using a conversation engine.

Among the fascinating geological features faced by geoscientists, fractures and karsts are among the most complex to characterize, as classical data is generally too sparse or too coarse to precisely map these features. To generate karst networks from sparse observations, Yves Frantz and Pauline Collon propose a new stochastic simulation method which controls the degree of network node connectivity. For fracture characterization, François Bonneau and Dietrich Stoyan present metrics derived from stochastic geometry to rigorously characterize 2D fracture networks and their spatial properties using marked point processes and fiber processes. Etienne Lavoine et al. introduce a new computational suite to simulate fracture networks, characterize their connectivity and simulate flow in these networks. For non-linear mechanical modeling, discrete element model (DEM) has shown its complementarity to finite element formulations, and coupled hydromechanical modeling has recently been introduced. However, DEM simulations in three dimensions can be computationally challenging, and the discretization of pre-existing mechanical interfaces is not straightforward. François Bonneau et al. show some new results to address these limitations, by looking at the effects of particle assembly on the DEM simulations, which is an important step to increase geological and physical accuracy.

In some cases, the DFN representation is not applicable for large scale physical modeling, but rather is an intermediate step to compute equivalent properties. As shown by Anaïs Ibourichène et al., non-periodic homogenization can be used in such a context: it yields accurate effective elastic coefficients of fractured media for seismic wave propagation. Fractures not only affect waves but they also localize seismic rupture. To account for interactions between seismic events, Loubna Ben Allal et al. infer the parameters of a Hawkes point process from seismicity catalogs, and show the appropriateness of this model on a Guadeloupe data set. For seismic event location, time reversal can be used to focalize the wavefield onto the source Zoé Renat et al. propose a new implementation to do this, and make a first sensitivity analysis on the impact of possible time reversal simplifications. In complement to source location, seismic hazard assessment relies on assessing the impact of a seismic event of a given magnitude in terms of peak ground acceleration. Such a relationship is generally defined by a ground motion prediction equation (GMPE). As this function depends on the geological medium between the event and the earth surface, Corentin Gouache et al. use spectral element simulations to compare GMPE and waveform results and enrich an existing GMPE data base. For such seismic simulations, Paul Cupillard et al. demonstrate the computational value of non-periodic homogenization on the SEG-EAGE Overthrust model, and highlight that significant tilted transverse anisotropy may appear in the equivalent medium even when the fine-scale medium is isotropic.

As I am writing these lines, I know and regret that many of you will not be able to be physically present in Nancy because of the sanitary crisis. Of course, the persons present will use masks and distancing, and video conferencing will be set up for remote participants. As each year, recordings will also be posted on the RING sponsor’s corner after the conference. Whether you can safely make it to Nancy or attend remotely this vintage of the RING Meeting, I wish you an enjoyable and productive RING Meeting, and hope to see you back for a vibrant live meeting in 2021!

Thanks very much for your support,
Guillaume Caumon.

Please DO NOT FORGET to register!!

Proceedings are already available for sponsors download.