The numerical modelling of the in-situ stress state is commonly performed for reservoir management or subsurface applications such as nuclear waste repositories. Modelling is required due to the sparsity of available stress magnitude data records in addition to the uncertain underground structures and heterogeneous material properties. Even though, geomechanical models provide an increase in confidence and allow an interpretation of the in-situ stress state, the associated uncertainties are very high.
We present an approach where, instead of just one model scenario that is a best fit to all available data (but necessarily has a poor fit to some data records), a wide range of model scenarios that each satisfies at least one data record perfectly is estimated. Then, additional indirect data on the stress state or other manifestations are added in order to assess the predictive quality of each model scenario. These can be for example Formation Integrity Tests or Borehole Breakouts. While model scenarios that have a good agreement with indirect data are preferred, those that are largely in contradiction are neglected.
Eventually, this approach can be used to increase the knowledge on the subsurface. Of all things, remaining contradictions help to identify anomalies or geographical and lithological areas where the model is not representative (yet). This allows to improve and evolve the model in a data-based way not only based on geomechanics. Thus, adding uncertainties and further indirect data to the standard toolbox empowers the significance of a model.