ADAPTED ELBOW METHOD FOR SPHERICAL CLUSTERING OF FRACTURE DATA FROM DIGITAL OUTCROP MODEL

Abstract

Exposed rock formations, also know as outcrops, are a great source of geological data, enabling the characterization of fracture networks that are less visible when analyzing reservoir data from seismic and well logs. Remote sensing techniques like LiDAR and UAV Photogrammetry allow the generation of Digital Outcrop Models (DOM) that provide accurate outcrop geometry enabling further investigation of fracture attributes, necessary when creating Discrete Fracture Network (DFN) models and reservoir simulation models. Besides the manual fracture interpretation in DOMs, automatic fracture plane segmentation and clusterization can provide significantly more data. However, the identification of family sets is still done visually in most cases. In this work, we develop a modification to the elbow method using spherical Fisher statistics of dispersion to determine the ideal number of clusters or fracture family sets. We employed this method to cluster fracture families in a point cloud DOM obtained from UAV-SfM processing of a flight over a carbonate outcrop in the Jandaira formation, Northeast Brazil, obtaining clusters with low dispersion values.