Constellation, le dépôt institutionnel de l'Université du Québec à Chicoutimi

Résumé

Three-dimensional computer modelling of geological phenomena is rapidly emerging as a field within the already mushrooming science of computer visualization. In geological applications three-dimensional interpretations are routinely performed through the use of two-dimensional map data and knowledge about the geological history of an area. These interpretations are traditionally depicted with isometric or perspective block diagrams and vertical or horizontal cross-sections. Constructing these three-dimensional snap-shots has been laborious, imprecise and limited to a single viewpoint. The methods presented here automate some of the more laborious tasks and enhance the three-dimensional interpretation environment. Methodology focuses on using field-based structural data, from a variety of scales, to create speculative three-dimensional surfaces that can be useful in addressing geological problems. These methods could help in resolving cryptic early fold geometry, extending stratiform mineralization and the subsurface interpretation of regional thrusts, unconformities or key lithostratigraphic boundaries. Several UNIX based programs are presented for performing the interpolation, extension and conversion tasks required in these approaches. Programs are implemented in conjunction with the commercial three-dimensional visualization and modelling software EarthVision® and gOcad®. Algorithms focus on the densification and variable projection of distributed three-dimensional data which share a common curvilinear geological feature. The result of the various interpolation and extension functions is the conversion of two-dimensional lines to three-dimensional surfaces.

A polynomial and hybrid B-Spline interpolation technique optimizes geometric property components. The automated data-driven technique is applicable for geological problems in which structures are constrained by local linear and planar measurements. Input features are topographic intersections of relatively continuous irregular curved surfaces, which have a near linear known depth predictability at some point along the structure. The local direction cosine estimates derived along surface traces of geological structures are interpolated, and direction vectors linearly projected to depth to form local structural surfaces or 'ribbons'. The program is useful for depicting portions of variably plunging fold geometries as structural ribbons, which in-turn act as visual guides during interpretive fold construction. Idealized and actual field examples of regionally continuous shear zones and brittle faults are presented, along with the development of three-dimensional structural fabric trajectories, horizon propagation, and plutonic boundary geometry evaluation.

Semi-automated techniques are utilized with knowledge-driven interactive graphics. An interpretive or 'design' approach to surface construction is applied to low density data sets which are too sparse for standard global automated interpolation. Bézier curves and surface patches are implemented to act as interpretive construction lines that respect the constraints imposed by structural orientation data. The programs hinge.awk, cast.awk, bspline.awk and bezpatch.awk calculate the interpolated values from the spatial input data. Three-dimensional construction lines are defined by tangents to local planar features, and the projection of key geologic structures. Supporting the interpolation tools, the program trace.awk estimates the local strike and dip of vertices along elevation registered three-dimensional curvilinear map traces. The planar solution method can be applied in highrelief terrains, or to extend three-dimensional curvilinear features from sub-surface mining data.

Techniques are applied on field data from the low-relief and structurally complex Archean Abitibi greenstone belt. Speculative models can be created from such terrains, provided data is respected and appropriate methods are applied at a given scale. The field component focuses on extracting data from maps and optimizing the three-dimensional graphic editing environment for making better interpretations at outcrop, mine and regional scales.

Applied techniques used in this study include:

? Three-dimensional structural symbology: the visualisation of three-dimensional structural symbols representing point observations of bedding, lineations and foliation fabric;

? Structural attribution: the attachment of structural point observations to linear features through the use of a proximity filter. This is done with the program field.awk;

? Variable down-plunge projection: the construction of custom down-plunge projections from surface traces;

? Bézier-based graphics: examples of interactive three-dimensional interpretations with Bézier-based curve;

? Hybrid surface design: a two-step approach to three-dimensional geologic surface design using both Bézier patches and discrete smooth interpolation (DSI), constrained by map traces and local slopes;

? Three-dimensional Map propagation: a method for propagating map elements using two-dimensional map data and field-based plunge models. The program dive.awk is presented as an example of a simple propagation.

The results of this study indicate that a constrained-interpretive approach to three-dimensional visualization is valid for interpreting large to small-scale geological structures, even if the data base is limited to two-dimensional map-based information. This geometric approach provides an initial development path for what could become the routine combination of extracted geological map based information, surface topographic and structural data, and the intuitive knowledge of a geological 'designer'. The developed techniques listed above and presented in this study enhance the field-based geologists ability to create communicable three-dimensional models of complex surfaces. Regardless of the state of visualization technologies, the success of three-dimensional geological modelling is still dependent on the data density, clustering and depth variability of known structural observations. Most important perhaps are the geological relationships of local and regional structures with the bounding surfaces being modelled. New software will be needed to assess the quality of geological models based on these input parameters.