Revisiting Sampling Bias in Fracture Networks Representation

Abstract

Natural and hydraulic fractures are commonly approximated as planar objects for most modelling applications. It is well-recognized that our perception of abundance of planar objects is influenced by the scale of observation, orientation of the objects, spatial arrangement of the objects and the dimensionality of the observation domain in which the objects are analyzed or modeled (well, map, geocellular model). In anisotropic media, such as bedded sedimentary successions, another type of bias is introduced by the physical limits of the volume where fractures are contained; these boundaries typically represent present-day or past geomechanical units. We revisit fracture abundance biases using an example built in a commercially available software package. Sampling biases are illustrated using common sources of fracture data information including well data and maps. The results show that it is important to correct for sampling biases in order to create a mathematically and naturally plausible representation of a fracture network for modelling purposes. In addition, realistic models help identify key parameters that require more understanding.

AAPG Datapages/Search and Discovery Article #90259 ©2016 AAPG Annual Convention and Exhibition, Calgary, Alberta, Canada, June 19-22, 2016