Comparison of surface and subsurface data for the construction of a comprehensive DFN model

Stochastic discrete fracture network (DFN) models are commonly employed to represent fractured media, improving the reliability of hydro-mechanical coupled models. DFNs have been established as an essential asset to ensure post-closure safety of nuclear waste deposits in crystalline host rocks [1]-[5].
However, a comprehensive understanding of fracture patterns and statistical properties often requires more than borehole investigations. Due to their quasi one-dimensional nature and limited availability, well-log fracture data present an incomplete, biased picture, which requires assumptions on spatial distribution and clustering of fractures, as well as geometrical properties such as fracture length and aperture distribution to include into a stochastic framework for the modeling process [4]. Furthermore, they do not provide information on structural features, such as fracture length, clustering, and spatial distribution.
To address these limitations, surface data is commonly used as supplementary information to calibrate DFN models [3]. Yet, while two-dimensional surface data offers accessible fracture information, it remains underutilized compared to subsurface data. This is partly due to the influence of weathering and glaciation, as well as surficial variations in stress and strain, which can distort fracture statistics derived from surface observations [2][5].
Here, we explore how fracture properties and patterns evolve with depth through both analytical and numerical approaches. We systematically assess a comprehensive dataset covering remote sensing and surface data, borehole data up to a few hundred meters, and data compiled from the Grimsel Test Site (GTS) plus adjacent tunnels of the Kraftwerke Oberhasli AG (KWO) in Switzerland.
By statistically comparing these datasets and correlating them with known stress conditions and structural models of the region, we aim to mathematically explain the variation of fracture statistics with depth. We aim to clarify when and under what conditions surface data can be reliably used, while also providing a methodological framework for its effective integration into DFN modeling. Our findings may improve the value of surface data, provided there is a robust understanding of the regional structural inventory.

 

[1] Lavoine Etienne, Davy Philippe, Darcel Caroline, Munier Raymond, A Discrete Fracture Network Model With Stress-Driven Nucleation: Impact on Clustering, Connectivity, and Topology, Frontiers in Physics 8 (2020).

[2] Lee Hartley, Simon Libby, Tomas Bym, James Carty, Mark Cottrell, Kyle Mosley, Baseline Forsmark – A discrete fracture network (DFN) model applying grown fractures and hydromechanical (HM) coupling, SKB Report R-23-01 (2024).

[3] Qinghua Lei, John-Paul Latham, Chin-Fu Tsang, The use of discrete fracture networks for modeling coupled geomechanical and hydrological behavior of fractured rocks, Computers and Geotechnics 85 (2017).

[4] Weiwei Zhu, Xupeng He, Ryan Kurniawan Santoso, Gang Lei, Tad Patzek, Moran Wang, Enhancing Fracture Network Characterization: A Data-Driven, Outcrop-Based Analysis, Earth Space Sci. Open Arch. 35 (2021).

[5] Kevin Bisdom, Bertrand Gauthier, Giovanni Bertotti, Nico Hardebol, Calibrating discrete fracture-network models with a carbonate three-dimensional outcrop fracture network: Implications for naturally fractured reservoir modeling, AAPG Bulletin 98 (2014).