EGU General Assembly 2021
© Author(s) 2021. This work is distributed under
the Creative Commons Attribution 4.0 License.

Now you see me... : Impact of sample representativity in fracture network characterization.

Fatemeh Nazari Vanani and Oscar Fernandez
Fatemeh Nazari Vanani and Oscar Fernandez
  • University of Vienna, Althanstrasse 14, 1090 Vienna, AUSTRIA

Adequately characterizing the properties of a fracture network is the first step in accurately modelling its behavior, be it mechanically or hydraulically. Characterizing fracture networks requires determining fracture frequency, orientation, connectivity, and fracture properties. This becomes particularly challenging in subsurface systems, where hard data on fracture networks comes mainly from boreholes, that are samples of very limited volume in relation to the fracture network. Because of this scale relationship between sample dimension and the dimension of natural fracture networks, boreholes capture a very partial picture of the fracture network. This is particularly relevant when attempting to estimate fracture frequency and network connectivity from borehole data. Corrections are normally used to account for sampling bias related to fracture size and orientation. Whereas these corrections are valid for the sample itself, the topology and heterogeneity of fracture networks means that measurements obtained in any given borehole are not necessarily representative of the broader fracture network.

To determine how “wrong” single-borehole analyses can be, we have conducted experiments on synthetic datasets to quantify how representative borehole samples are of entire fracture networks. Results show that properties that have an impact on the anisotropy of the fracture network (orientation, number of fracture sets) can be accurately resolved even in low data-density scenarios. On the contrary, accurately determining fracture frequency (which also impacts connectivity) for the entire fracture network is strongly dependent on the ratio between fracture frequency and the sampled volume. Measurements of fracture frequency in individual boreholes indicate that it frequency easily be overestimated or underestimated by a factor of 2 relative to the real network’s fracture frequency. The application of sampling bias corrections has a limited impact on reducing this error.

Based on the results from our experiments, we present methods to assess how representative of a fracture network a single borehole is. Representativity can be translated into uncertainty in fracture frequency, a metric that can be used in fracture modelling.

How to cite: Nazari Vanani, F. and Fernandez, O.: Now you see me... : Impact of sample representativity in fracture network characterization., EGU General Assembly 2021, online, 19–30 Apr 2021, EGU21-16341,, 2021.