From point-wise stress data to a continuous description of the 3D crustal in situ stress state

The in situ stress is a key parameter for the safe and sustainable management of geo-reservoirs or storage of waste and energy in deep geological repositories. It is also an essential initial condition for thermo-hydro-mechanical (THM) models that investigate man-made induced processes e.g. seismicity due to fluid injection/extraction, reservoir depletion or storage of heat producing high-level radioactive waste. Without a reasonable assumption on the initial stress condition it is not possible to assess if a man-made process is pushing the system into a critical state or not. However, modelling the initial 3D stress state on reservoir scale is challenging since data are hardly available before drilling in the area of interest. This is in particular the case for the stress magnitude data which are a prerequisite for a reliable model calibration. Here, we present a multi-stage 3D geomechani­cal-numerical model approach to estimate for a reservoir-scale volume the 3D in situ stress state. First, we set up a large-scale model which is calibrated by stress data and use the modelled stress field subsequently to calibrate a small-scale model located within the large-scale model. The local model contains a significantly higher resolution representation of the subsurface geometry around boreholes of a projected geothermal power plant. This approach incorporates two models and is an alternative to the required trade-off between resolution, computational cost and calibration data which is inevitable for a single model; an extension to a three-stage approach would be straight forward. We exemplify the two-stage approach for the area around Munich in the German Molasse Basin. The results of the reservoir-scale model are presented in terms of values for slip tendency as a measure for the criticality of fault reactivation. The model results show that variations due to uncertainties in the input data are mainly introduced by the uncertain material properties and missing estimates for the magnitude of the maximum horizontal stress SHmax, needed for a more reliable model calibration. This leads to the conclusion that at this stage the model's reliability depends only on the amount and quality of input data records such as available stress information rather than on the modelling technique itself.