METHODS OF SENSITIVITY ANALYSIS IN GEOLOGIC DISPOSAL SAFETY ASSESSMENT (GDSA) FRAMEWORK

摘要

Probabilistic simulations of the post-closureperformance of a generic deep geologic repository forcommercial spent nuclear fuel in shale host rock provide atest case for comparing sensitivity analysis methodsavailable in Geologic Disposal Safety Assessment (GDSA)Framework, the U.S. Department of Energy’s state-of-thearttoolkit for repository performance assessment.Simulations assume a thick low-permeability shale withaquifers (potential paths to the biosphere) above and belowthe host rock. Multi-physics simulations on the 7-millioncellgrid are run in a high-performance computingenvironment with PFLOTRAN. Epistemic uncertain inputsinclude properties of the engineered and natural systems.The output variables of interest, maximum I-129concentrations (independent of time) at observation pointsin the aquifers, vary over several orders of magnitude.Variance-based global sensitivity analyses (i.e.,calculations of sensitivity indices) conducted with Dakotause polynomial chaos expansion (PCE) and Gaussianprocess (GP) surrogate models. Results of analysesconducted with raw output concentrations and with logtransformedoutput concentrations are compared. Usinglog-transformed concentrations results in larger sensitivityindices for more influential input variables, smallersensitivity indices for less influential input variables, andmore consistent values for sensitivity indices betweenmethods (PCE and GP) and between analyses repeatedwith samples of different sizes.