A REVIEW OF GEOLOGICAL MODELING

A. Keith Turner and Carl W. Gable2

1

Colorado School of Mines, Golden, CO 80401, USA, [email protected]; Los Alamos National Laboratory, Los Alamos, NM 87545, USA.

1.

INTRODUCTION

Over the past two decades, a series of sophisticated three-dimensional modeling technologies collectively identified as Geoscientific Information Systems, or GSIS have been developed to address the need for a precise definition of subsurface conditions (Turner, 1991). Geological modeling requires the extension of traditional GIS methods (Turner, 2000; 2006) and the modeling process remains technically challenging. Geologists have always recognized the need to view the world as multidimensional. The requirements for subsurface characterisation and modelling may be defined as follows: The industry requires a system for interactive creation of spatial and spatio-temporal models of the physical nature of portions of the Earths crust. i.e., the capability to effectively model & visualise: the geometry of rock- and time-stratigraphic units, the spatial and temporal relationships between geo-objects, the variation in internal composition of geoobjects, the displacements or distortions by tectonic forces, and the fluid flow through rock units. (Kelk, 1991). Interpolation between widely spaced filed observations requires geological knowledge to successfully replicate actual geological environments. Iterative methods involving assessments and progressive refinements add considerable time and cost to the creation of subsurface models (Turner, 2006). Unlike the older resource-industry-based user-community, many of todays potential users of GSIS models and visualizations cannot interpret basic geoscience data or evaluate the merits of alternative interpretations. They may be unable to distinguish between theories and facts. In short, these new users clearly desire solutions, not data and information in understandable form (Turner, 2006). Users can be classified in terms of their information acceptance capabilities as "Thick" or "Thin" clients. A "Thick" client is one that can accept and interpret or evaluate great deal of raw data; a Thin client is a relatively unsophisticated user desiring relatively simple, concise answers. 2. THE GEOLOGICAL MODELING PROCESS

Figure 1 illustrates the steps in a typical geological modeling process. Raw data collected from various sources can be considered as two types spatial data and properties data. The spatial data are used to create a 3-D geometry model, shown on the left-hand side of Figure 1. Geometry modeling involves two steps first the development of a suitable geometric representation of the fundamental geological framework, and subsequently the subdivision, or discretization of this framework to provide control for the analytical computations within the numerical models used in the predictive modeling. The horizontal arrow linking the discretization and analytical modeling operations in Figure 1 defines this linkage.

Figure 1. Overview of the Geological Modeling Process.

2.1 The Geometry Model

The development of a geometrically accurate representation of the geological framework, suitable for visualization, serves to define and control the spatial distribution and propagation of rock-properties required by modeling. Framework definition is accomplished by applying a variety of data types, including (a) borehole and isolated sample data, (b) surfaces (triangle, quadrilateral, NURB), (c) 2-D grids and meshes, and (d) a variety of iso-volumetric models created from multiple surfaces, cross-sections, and grids and meshes.

2.2 Stratigraphic Layer-cake Models

Sedimentary geologic environments are modeled by creating surfaces defining the strata interfaces, stacking the surfaces in stratigraphic order, and subsequently defining the zones between surfaces as geologic units. Construction of individual surfaces generally proceeds by one of three methods: 1) using the borehole observations to create a triangles defining a surface, 2) applying surface generation and contouring procedures to borehole observations, or 3) developing a series of interpretive cross-sections between boreholes. Regardless of the method used, several problems remain. Surfaces created independently may intersect each other in geologically impossible situations. Careful review and editing of all surfaces is usually required to allow for areas of erosion or non-deposition. While creating volumes that accurately represent the location of known surface contacts is important, a difficult but equally important step is deciding where the zero contour isopach of a stratigraphic unit is located. Since some software systems require all surfaces to exist across the entire model domain, surfaces must be extended with zerothickness units, which in turn cause problems in ensuring correct displays. A limitation of this approach is that geometry can only be represented by single valued functions, z(x,y).

2.3 Non-stratigraphic models

Regions with complex geological structures, or without layered sequences, must be modeled using other techniques. One approach is to develop a series of complex shapes enclosing volumes derived from a series of interpreted cross-sections. The individual volumes must share common bounding surfaces so that there are no voids or overlapped volumes. Several mining GSIS products provide such model building capabilities (Houlding, 1994). An alternative approach begins with an entire regional volume and then progressively subdivides it into regions with a series of intersecting surfaces that represent major discontinuities such as shear zones or faults. Figure 2 demonstrates this concept know as Constructive Solid Geometry (CSG). The various regions may be considered as having distinct material properties with oriented anisotropies or gradients, while the discontinuity surfaces may also be assigned widths and unique properties. A limitation of this approach is that all geometry must define closed , air tight volumes. Nonmanifold geometry, such as a fault plane that terminates within a volume or a well bore represented as a zero volume line, cannot be represented.

Figure 2. Model construction by constructive solid geometry. (Carl Gable, Los Alamos National Laboratory) 2.4 Modeling Faults
Rock strata on opposite sides of a fault may have similar or different thicknesses and characteristics depending on the type of fault and the temporal relationships between the depositional processes and the faulting. A number of modeling tools have been proposed to assist model creation under specific

assumptions. Faults may provide preferential conduits for fluid flow, or they may act as barriers to flow. They typically add anisotropy to property distributions required by the numerical models. Vertical, or nearly vertical, faults and nearly horizontal thrust zones can be defined by adding additional surfaces to the existing stratigraphic models. This increases the complexity of model creation, but otherwise is relatively straightforward. Moderately inclined faults present greater modeling difficulties.

2.5 Discretization
Property distributions are generally modeled by applying discretization methods to subdivide the framework objects into a series of small elements. There is a considerable body of theory concerning the design and construction of meshes appropriate to different modeling requirements (Knupp and Steinberg, 1994). There are two broad classes of meshes structured and unstructured. Available commercial GSIS products depend mostly on structured meshes. While 3D models require volumetric meshes, the differences among the approaches are more easily illustrated by 2D equivalents (Figure 3). 2.6 Structured Meshes A regular cellular mesh (Figure 3B) is the most common among commercially available GSIS. A 3D volume is divided into discrete volume elements, or voxels, that are usually regular cubes. Unless the cell dimensions are very small, important geometric details may be lost, but small cells produce extremely large model files. Quadtree hierarchical cells (Figure 3C) provide greater flexibility in adapting grid resolution to where it is needed. In 3D, the octree representation provides the same functionality. Sedimentary strata typically are much more consistent laterally than vertically; thus, some commercial GSIS offer partly deformable geocellular voxels.

Figure 3. A) Typical two-dimensional geological cross section the rectangular region outlined is used to illustrate a variety of meshes in B-D. B) Orthogonal cellular mesh. C) Quadtree mesh. D) 2D Unstructured triangular mesh. 2.7 Unstructured Meshes
Unstructured meshes (Figure 3D, 4) are not constrained by having to have a constant node and face structure, and can link with finite element models. Three-dimensional unstructured meshes, based on tetrahedrons, hexahedrons, are particularly useful in modeling faults and fracture discontinuities (Figure 4). This provides added flexibility during model development, but this flexibility comes at a price: added computational demands and more effort in model construction and requires use of sophisticated mesh generation software (Gable, et al., 1996).

Figure 4. A 3D unstructured mesh model of a faulted layered sequence. (Source: Carl Gable, Los Alamos National Laboratory) 2.8 The Predictive Model
The primary objective of subsurface geometry modeling is to provide geometric controls and property distributions for some type of numerical modeling, and the purpose of this analytical modeling is prediction. The right-hand side of Figure 1 identifies this predictive modeling step. Prediction has an extrapolative rather than interpretive character; thus it involves risk and uncertainty. Prediction leads to decision-making. Predicted results often require supporting visualizations and interpretations that can be presented to and used by the customer of the modeling results. Both the geometry and predictive models are shown in Figure 1 as passing through visualization and interpretation to delivery to the customer. 3. GEOLOGICAL MODELING WITHIN INFORMATION MANAGEMENT SYSTEMS

Efficient delivery of modeling results, as well as effective modeling, depends on the use of sophisticated information management. Relatively few documented examples of such integrated approaches exist. Figure 5 illustrates the four basic components of an integrated geological data modeling and information management system. The central technical core contains the technical tools supporting geological framework model development and numerical, predictive simulation modeling. The technical core is surrounded by four supporting components. The infrastructure component is required to keep the system operating smoothly. The workflow management component provides tools to access the technical core, including user interfaces, and also links to the risk assessment and decision support component that contains software tools for risk assessment and decision making. The dissemination component manages the production and distribution of results generated by the system.

Figure 5. Integrated geological data and information management.

These concepts and requirements were used to develop a series of systems at Los Alamos and then further expanded to meet the requirements of the U.S. Geological Survey Death Valley ground water project (Rosenbaum and Turner, 2003). 4. LOOKING FORWARD

Turner and Rosenbaum (2003) discuss four major impediments to the greater use of subsurface geological modelling, based on deliberations during a conference sponsored by the European Science Foundation in 2001 (Rosenbaum and Turner, 2003). These impediments are: The lack of fully developed 3D and 4D oriented mathematical, cognitive, and statistical spatial analysis tools. The inability of available systems to encapsulate the natural variability of geological features, or to represent and accurately depict uncertainties related to interpretation of spatial configuration or properties. The broad lack of access to specialised tools for exploring and modelling geological systems. Other than those developed for the oil industry; these are generally too expensive for all but a few organisations to procure. The paucity of comprehensive examples that might overcome what may be termed 'geoscientific inertia'. Many geoscientists are aware of quantitative subsurface characterisation methods, but the majority do not embrace them because they have yet to be convinced that the investment in time, effort, and funds will yield a dividend or result in better science. 5. REFERENCES

Gable, C.W., Trease, H., and Cherry, T. 1996. "Geological Applications of Automatic Grid Generation Tools for Finite Elements Applied to Porous Flow Modeling," in Numerical Grid Generation in Computational Fluid Dynamics and Related Fields, edited by B. K. Soni, J. F. Thompson, H. Hausser and P. R. Eiseman, Engineering Research Center, Mississippi State Univ. Press. Houlding, S. W. 1994. 3D Geoscience Modeling Computer Techniques for Geological Characterization. Springer-Verlag, Berlin-Heidelberg, pp. 75-85, and pp.131-142. Kelk, B. 1991. 3D Modelling with Geoscientific Information Systems: The Problem. In: Three-Dimensional Modeling with Geoscientific Information Systems, [A.K. Turner, Editor], NATO ASI Series C: Mathematical and Physical Sciences, v. 354, Kluwer Academic Publishers, Dordrecht, The Netherlands, pp. 29-38. Knupp, P. and Steinberg, S. 1994. Fundamentals of Grid Generation. CRC Press, Boca Raton, Florida, 286p. Rosenbaum, M.S. and Turner, A.K., [Editors] 2003. New Paradigms in Subsurface Prediction. Lecture Notes in the Earth Sciences, Number 99, Springer-Verlag, Berlin, 397p. Turner, A.K., [Editor] 1991. Three-dimensional Modeling with Geoscientific Information Systems. NATO ASI Series C: Mathematical and Physical Sciences, v. 354, Kluwer Academic Publishers, Dordrecht, The Netherlands, 443p.

Turner, A.K. 2000. Geoscientific Modeling: Past, Present and Future. In: Geographic Information Systems in Petroleum Exploration and Development [T.C. Coburn and J.M. Yarus, Editors], AAPG Computer Applications in Geology, Vol. 4, American Association Petroleum Geologists, pp. 27-36. Turner, A.K. 2006. Challenges and Trends for Geological Modelling and Visualisation, Bulletin of Engineering Geology and the Environment, Volume 65, Number 2, May 2006, pp. 109-127