Ines Görz

Workflows for generating tetrahedral meshes for finite element simulations on complex geological structures

Subsurface processing numerical simulations require accurate discretization of the modeling domain such that the geological units are represented correctly. Unstructured tetrahedral grids are particularly flexible in adapting to the shape of geo-bodies and are used in many finite element codes. In order to generate a tetrahedral mesh on a 3D geological model, the tetrahedrons have to belong completely to one geological unit and have to describe geological boundaries by connected facets of tetrahedrons. This is especially complicated at the contact points between several units and for irregular sharp-shaped bodies, especially in case of faulted zones. This study develops, tests and validates three workflows to generate a good tetrahedral mesh from a geological basis model. The tessellation of the model needs (i) to be of good quality to guarantee a stable calculation, (ii) to include certain nodes to apply boundary conditions for the numerical solution, and (iii) support local mesh refinement. As a test case we use the simulation of a transient electromagnetic measurement above a salt diapir. We can show that the suggested workflows lead to a tessellation of the structure on which the simulation can be run robustly. All workflows show advantages and disadvantages with respect to the workload, the control the user has over the resulting mesh and the skills in software handling that are required. HighlightsWe developed, tested and validated three workflows to generate a good tetrahedral mesh from a geological basis model.We used a simulation of a transient electromagnetic measurement as a testcase.We can show that the suggested worklfows lead to a tessellation on which the simulation can be run robustly.

Multi-method virtual electromagnetic experiments for developing suitable monitoring designs: A fictitious CO2 sequestration scenario in Northern Germany

We present a numerical study for 3D time-lapse electromagnetic monitoring of a fictitious CO2 sequestration using the geometry of a real geological site and a suite of suitable electromagnetic methods with different source/receiver configurations and different sensitivity patterns. All available geological information is processed and directly implemented into the computational domain, which is discretized by unstructured tetrahedral grids. We thus demonstrate the performance capability of our numerical simulation techniques. The scenario considers a CO2 injection in approximately 1100 m depth. The expected changes in conductivity were inferred from preceding laboratory measurements. A resistive anomaly is caused within the conductive brines of the undisturbed reservoir horizon. The resistive nature of the anomaly is enhanced by the CO2 dissolution regime, which prevails in the high-salinity environment. Due to the physicochemical properties of CO2, the affected portion of the subsurface is laterally widespread but very thin. We combine controlled-source electromagnetics, borehole transient electromagnetics, and the direct-current resistivity method to perform a virtual experiment with the aim of scrutinizing a set of source/receiver configurations with respect to coverage, resolution, and detectability of the anomalous CO2 plume prior to the field survey. Our simulation studies are carried out using the 3D codes developed in our working group. They are all based on linear and higher order Lagrange and N´ed´elec finite-element formulations on unstructured grids, providing the necessary flexibility with respect to the complex real-world geometry. We provide different strategies for addressing the accuracy of numerical simulations in the case of arbitrary structures. The presented computations demonstrate the expected great advantage of positioning transmitters or receivers close to the target. For direct-current geoelectrics, 50% change in electric potential may be detected even at the Earth’s surface. Monitoring with inductive methods is also promising. For a well-positioned surface transmitter, more than 10% difference in the vertical electric field is predicted for a receiver located 200 m above the target. Our borehole transient electromagnetics results demonstrate that traditional transient electromagnetics with a vertical magnetic dipole source is not well suited for monitoring a thin horizontal resistive target. This is due to the mainly horizontal current system, which is induced by a vertical magnetic dipole.

Rasterizing geological models for parallel finite difference simulation using seismic simulation as an example

3D geological underground models are often presented by vector data, such as triangulated networks representing boundaries of geological bodies and geological structures. Since models are to be used for numerical simulations based on the finite difference method, they have to be converted into a representation discretizing the full volume of the model into hexahedral cells. Often the simulations require a high grid resolution and are done using parallel computing. The storage of such a high-resolution raster model would require a large amount of storage space and it is difficult to create such a model using the standard geomodelling packages. Since the raster representation is only required for the calculation, but not for the geometry description, we present an algorithm and concept for rasterizing geological models on the fly for the use in finite difference codes that are parallelized by domain decomposition. As a proof of concept we implemented a rasterizer library and integrated it into seismic simulation software that is run as parallel code on a UNIX cluster using the Message Passing Interface. We can thus run the simulation with realistic and complicated surface-based geological models that are created using 3D geomodelling software, instead of using a simplified representation of the geological subsurface using mathematical functions or geometric primitives. We tested this set-up using an example model that we provide along with the implemented library. Graphical abstractDisplay Omitted HighlightsConverting 3D surface based geological models into 3D structured grids for simulation.Suitable for complex structural situations with diapirs, faults and overturned folds.Parallel computation - conversion is done on the fly and no limitation on grid size.Integration of the library into parallel seismic simulation software as proof of concept.3D simulation of seismic wave propagation using a model of a salt diapir as input.

Database versioning and its implementation in geoscience information systems

Many different versions of geoscience data concurrently exist in a database for different geological paradigms, source data, and authors. The aim of this study is to manage these versions in a database management system. Our data include geological surfaces, which are triangulated meshes in this study. Unlike revision/version/source control systems, our data are stored in a central database without local copies. The main contributions of this study include (1) a data model with input/output/manage functions, (2) a mesh comparison function, (3) a version merging strategy, and (4) the implementation of all of the concepts in PostgreSQL and gOcad. The software has been tested using synthetic surfaces and a simple tectonic model of a deformed stratigraphic horizon. The versions of geoscience data is managed in database systems.The data model is simple.Mesh comparison algorithm is based on the k-d tree sliding-median-widest-spread.Version merging strategy considers the conflict situations.The software is implemented in PostgreSQL and gOcad.