ABSTRACT
Radioactive waste may be stored in deep geological disposal facilities. Heat-producing radioactive waste would drive coupled thermo-hydro-mechanical (THM) processes in the host rock mass. Rock mass uncertainty could affect predictions of observable properties that are perturbed by these THM processes. However, the contribution of the individual rock properties to the uncertainty is unknown and is critical for informing modelling and monitoring for reducing uncertainty. Here we rank the rock properties (model input factors) according to their contribution to uncertainty in the observable properties: temperature, pore pressure and displacement (model outputs). We identify non-influential rock properties which have a negligible effect on the uncertainty and should no longer be considered uncertain. We identify the thermal conductivity and permeability as the most influential rock properties which was expected because these are anisotropic. Interestingly we found that closer to the heat source these become more influential relative to the other rock properties. The ranking may be used to inform prioritizing further rock mass characterization efforts for uncertainty reduction. The spatial effect on thermal conductivity and permeability may be used to improve model calibration and back-calculations by counter-intuitively using data proximal to the heat source. We demonstrate that global sensitivity analysis methods can be applied to geomechanical models to improve our understanding of the rock mass uncertainty. We present validated values for the number of models required for results to converge which should be used to estimate appropriate values for future sensitivity analyses on geomechanical models.
