--> --> Quantification of Fracture Attributes from Terrestrial Laser Scanning to Improve Input Parameters for Discrete Fracture Network (DFN) Modeling

International Conference & Exhibition

Datapages, Inc.Print this page

Quantification of Fracture Attributes from Terrestrial Laser Scanning to Improve Input Parameters for Discrete Fracture Network (DFN) Modeling


Fluid flow in fractured reservoirs is strongly controlled by the connectivity and extent of the fracture network. Information on the fracture network in a hydrocarbon or geothermal reservoir is typically inferred from seismic and borehole data. An additional source of information can be outcrops like quarries or tunnels, provided they belong to the same tectonostratigraphic unit as the subsurface reservoir. Such reservoir analogues can provide comprehensive insights into the geometry and spatial distribution of the fracture network and allow to build a large data base of fracture attributes. This information can be used to set up a Discrete Fracture Network model (DFN) from which fundamental hydraulic properties of the reservoir can be derived. Terrestrial laser scanning (TLS) is a well-established method for outcrop analysis. In comparison to traditional geological survey techniques laser scanning allows for a faster and more efficient collection of fracture attributes. Important input parameters for the DFN model are the distribution functions for the orientation and size of the fractures as well as the fracture density. These data can be extracted from the three-dimensional model of the outcrop obtained from TLS. After preprocessing of the point cloud, including registration of the point clouds and removal of artefacts and vegetation, homogeneously distributed normal vectors on surfaces within the model are calculated. Afterwards individual fracture sets are identified using a cluster analysis or a visual differentiation by the interpreter directly in the model. Patches automatically created from the point cloud allow for determination of fracture density for each set by identifying the number of intersections of an arbitrary path through the point cloud. The shape of these patches is chosen as ellipses to represent natural fracture geometries. The resulting values are expressed in terms of a one dimensional intensity value (P10) and converted to a volumetric intensity (P32). Finally, the geometry and size of the fractures are analyzed through vertices of created patches/polygons on the surface. A best fitting distribution function for the size, most commonly the length of the fracture, is inferred mathematically. This semi-automatic method of fracture parameter extraction based on TLS creates a much larger database than traditional methods and therefore substantially increases the statistical significance of the input parameters for DFN modeling.