Hierarchical Clustering and Graph Distance Metrics". The dataset consists of circular subgraphs that are sampled from larger 2D fracture networks using a fixed spacing and diameter. The large-scale fracture networks in the form of spatial graphs can be found in the dataset, "Fracture Patterns from the Lilstock Pavement, Bristol Channel, UK (https://doi.org/10.4121/14039234)".

The subgraphs are extracted from fracture networks corresponding to three regions that are named as Regions 1,2, and 3. Region 1 contains 219 subgraphs (spread over approximately 6017 sq. m), Region 2 contains 212 subgraphs (spread over approximately 6749 sq. m), and Region 3 contains 117 subgraphs (spread over approximately 1473 sq. m). The subgraphs are sampled with diameters of 7.5 m. The spacing between sampling circle centers is 5 m for Regions 1 and 2, and is 3 m for Region 3, so that there is some degree of overlap between the subgraphs.

The subgraph data is in the form of *.mat files. Each subgraph sample consists of a MATLAB graph object and a positioning matrix. The positioning is relative and may be georeferenced to UTM Zone 30 N or Coordinate Reference System EPSG: 32630 by shifting origin of the spatial positioning matrices using the 'xmin','ymin' variables corresponding to each region. %I 4TU.ResearchData