Rock Properties and Modelled Stress State Uncertainties: A Study of Variability and Dependence

Abstract

The safety and sustainability of subsurface applications requires a profound knowledge of the local stress state which is frequently assessed using 3D geomechanical-numerical models. Various factors lead to generally large uncertainties in these models. The variabilities in the rock properties as one of the sources of uncertainties and their influence on the modelled stress state is addressed herein. A generic 3D geomechanical-numerical model is used to investigate the influence of different distributions of variability and their effect on different stress states. The variability in rock properties clearly affects the uncertainties in the stress state in a positive correlation with differences that depend on the affected component of the stress tensor. The basic observation is that largest uncertainties are observed in the normal components of the stress tensor where the variabilities apparently are most effective. The same rock property variabilities affect the shear components uncertainties to a significantly lesser extent. Variabilities in the Young’s modulus and the Poisson’s ratio chiefly affect the uncertainties in σxx and σyy. The density variability, however, leads to highest uncertainties in σzz. In general, variabilities in the Young’s modulus are most effective, followed by the Density and then the Poisson’s ratio. Furthermore, an influence of the tectonic stress regime on how the variability in the rock properties affects the stress state is observed. At the same time only a small effect is observed for different stress magnitudes. The eventual uncertainties in a modelled stress state depend not only on the uncertainties in the rock properties but also whether the uncertainties are found mainly in the Young’s modulus, the Poisson’s ratio or the Density. These findings indicate the importance to regard variabilities in rock properties as a source for significant uncertainties in geomechanical-numerical models. It is proposed to use the derived relations for an inexpensive quantification of uncertainties by means of a post-computation assignment of uncertainties to a stress model.