Spatial Variability and Parametric Uncertainty in Performance Assessment Models

The problem of defining an appropriate treatment of distribution functions (which could represent spatial variability or parametric uncertainty) is examined based on a generic performance assessment model for a high-level waste repository. The generic model incorporated source term models available in GoldSim^®, the TDRW code for contaminant transport in sparse fracture networks with a complex fracture-matrix interaction process, and a biosphere dose model known as BDOSE™. Using the GoldSim framework, several Monte Carlo sampling approaches and transport conceptualizations were evaluated to explore the effect of various treatments of spatial variability and parametric uncertainty on dose estimates. Results from a model employing a representative source and ensemble-averaged pathway properties were compared to results from a model allowing for stochastic variation of transport properties along streamline segments (i.e., explicit representation of spatial variability within a Monte Carlo realization). We concluded that the sampling approach and the definition of an ensemble representative do influence consequence estimates. In the examples analyzed in this paper, approaches considering limited variability of a transport resistance parameter along a streamline increased the frequency of fast pathways resulting in relatively high dose estimates, while those allowing for broad variability along streamlines increased the frequency of “bottlenecks” reducing dose estimates. On this basis, simplified approaches with limited consideration of variability may suffice for intended uses of the performance assessment model, such as evaluation of site safety.