Analyzing in situ stress: challenges in quantifying stress domains

ABSTRACT: Accurately characterizing the state of in situ stress in rock is important for the design of all underground engineering projects, but is crucial for safety‑critical projects such as deep geological repositories for nuclear waste. However, designers continue to be confronted with the challenging task of characterizing the significant variability and uncertainty found in the in situ stress state across the project volume, and particularly identifying separate stress domains. One routine approach is to partition, or group, stress data on the basis of depth below ground. In this paper we discuss a customary approach to identifying stress domains, and illustrate challenges in its application. We go on to present a novel approach that uses Bayesian linear segmented regression of Cartesian stress components to statistically characterize the variability and uncertainty in the depth of stress domain boundaries. Synthetic data are used to demonstrate the suitability and efficacy of this method, and it is then applied to stress measurements in crystalline rock obtained at the Forsmark site in Sweden. We conclude that use of a Bayesian approach is beneficial as it is able to formally augment stress measurement data with other valuable geological information.