Embrace the uncertainty – Geomechanical example for the value of uncertainties

The initial in-situ stress state is a key parameter for the evaluation of siting regions for deep geological repositories. To achieve a 3D description of the stress field geomechanical-numerical models are used to extrapolate between a usually small number of in-situ stress measurements and finding a best fit in a model calibration process. However, if all in-situ data points cannot be fitted equally well by the model, the predictive value decreases. Furthermore, this approach neglects that the data records of in-situ stress magnitudes are not data points, but ranges. Each data record is a range of possible values or even a probability distribution of the physical value. As a consequence, the modelled stress field has to be a range as well. 

Considering the inherent uncertainty of in-situ stress magnitude data a single fit of a model to the observed stress state is usually not meaningful as it can only explain a subset of the data records used for the model calibration. To quantify these uncertainties a range of model results can be used that contain extreme and average cases resulting in a range of possible stress states. However, this range of results also includes extreme and therefore unlikely scenarios and probably overestimates the range of modelled stress states resulting in an unspecific prediction of the stress field. If the in-situ stress magnitude data is provided as a range, this modelled stress range can be refined. Data records that come as a range can be fitted in a model scenario that also agrees with other ranges of in-situ stress magnitude data. At the same time, extreme model scenarios can be identified since they only agree with very few data ranges. This allows to narrow down the range of modelled stress states.

The concept is exemplified and its applicability demonstrated using a case study of the siting region Zürich Nordost located in northern Switzerland. Here in the context of the site-selection process for a deep geological repository a 3D geomechanical models has been built and a unique data set of 45 in-situ stress magnitude data records is used for the model calibration. Our results show that using the measurement uncertainties of the in-situ stress magnitude data narrows the modelled stress state range compared to an approach where data records are used as data points only. Thus, we propose that using the ranges of the in-situ stress magnitude data instead of treating them as data points using e.g. their mean value, will increase the significance of 3D geomechanical models.