Interpretation of structural domains in discontinuity data from Nirex deep boreholes at Sellafield

Knowledge of the fracture network is fundamental to the fluid transport and geotechnical modelling necessary to assess the suitability of a rock mass to host a potential underground repository for the disposal of radioactive waste. However, in the early stages of a site investigation programme, site-specific fracture data are frequently available only from boreholes. This was the case at Sellafield, where characterization of borehole fracture data presented a number of inter-related problems including the volume of data, two data sets for each borehole and borehole sampling effects. A solution to the presentation and interpretation of such data was found in simple graphical display which allows rapid appraisal of a range of fracture attributes in relation to depth, lithostratigraphy and structure or in any other context. The succession sampled by the boreholes in the immediate vicinity of the proposed repository at Sellafield comprises three major units. The Borrowdale Volcanic Group, of Ordovician age, is unconformably overlain by a Permo-Triassic cover sequence com prising up to 100 m of sedimentary breccia (Brockram) succeeded by c. 450 m of the fluvial St Bees Sandstone to rockhead, Fracture data from these are available from logging of the borehole core and from interpretation of resistivity and acoustic images of the borehole wall, providing accurate frequency and orientation information respectively. These show that the discontinuity character of each of the major units is distinctive, indicating overall lithological control on the fracture network. Fracture frequency in the sedimentary cover rocks is relatively low, especially in the Brockram, and orientation patterns are dominated by bedding. The fracture frequency in the volcanic rocks is generally high and orientation typically very variable in the volumetrically dominant welded ignimbrites. However, interbedded units of non-welded lapilli tuff and breccia have lower fracture frequency and simpler orientation patterns. In all units, variations in fracture frequency and orientation character on a scale of a few metres to 100 m can be correlated with fault intersections indicating a structural overprint. Borehole sampling effects are illustrated by comparison of data from differently orientated boreholes and the underlying causes reviewed. Orientation bias is a well known problem usually addressed by weighting according to the probability of the borehole intersecting a planar feature in a particular orientation. However, data from borehole wall imagery are subject to additional sampling effects which reduce the effectiveness of such weighting. A different approach is suggested, using the probability density function of the population from non-censored parts of the sample.