A new subsampling methodology to optimize the characterization of two-dimensional bedrock fracture networks

This paper introduces a new subsampling method to determine the empirical relationships between the areal sampling coverage and the topological-geometric parameters resulting from characterization of two-dimensional outcrop fracture networks. We further developed these relationships into correlations between the total sampling coverage and variance of the characterization result, hence providing insight to the objective uncertainties related to fracture network characterization. The analyses were conducted using two new open-source Python packages: fractopo and fractopo-subsampling, designed for fracture network analysis and subsampling, respectively. We conducted the study on the well-exposed crystalline outcrops of Getaberget, Aland Islands, Finland, where a total of 42499 fracture traces were manually digitized from 13 circular target areas. For the purposes of subsampling, we conducted fracture network characterization for randomly located and sized subsample areas, which locate within the larger target areas.& nbsp;Based on our subsampling results we provide recommendations for the preliminary optimization of areal coverage used in outcrop fracture sampling and the use of our subsampling method for assessing the precision related to the areal fracture network characterization in other previously uncharacterized areas. As an example, we recommend using a total sampling area of 8000 m2 with 8 circular sampling areas to define the power-law exponents of fracture traces when conducting outcrop fracture network characterization with drone-based methodology done with similar initial sampling setup in comparable geological environments, as this coverage has shown an acceptable level of precision.

Automated summary

This paper introduces a new subsampling method to determine the empirical relationships between areal sampling coverage and topological-geometric parameters of two-dimensional outcrop fracture networks. The study provides insights into the uncertainties related to fracture network characterization. The results are based on a study conducted in the Aland Islands, Finland using open-source Python packages for fracture network analysis and subsampling.