Probabilistic Performance-Assessment Model

1. Develop and screen senarios based on regulatory requirements (performance objectives) and relevant features, events, and processes
A scenario is identified as a well-defined sequence of features, events and processes that describes possible future conditions at the disposal site. An example of a scenario is the release of radionuclides from a landfill via the vadose zone to the aquifer, where water is pumped from a well and ingested by an individual. Another scenario might be the inadvertent intrusion of a person digging for natural resources, which disrupts the repository and causes a direct release of radionuclides to the surface. The decision to evaluate or not evaluate various scenarios depends, in part, on relevant performance objectives set forth by regulatory requirements. In addition, scenarios should be chosen that represent features, events, and processes (FEPs) that are relevant to the specific site being evaluated. It is through the FEPs process that the analyst demonstrates that all events and processes that may cause releases to the biosphere are addressed.

2. Develop models of relevant features, events, and processes
The models that are used vary in complexity, and a hierarchy of models can exist. An overarching conceptual model of each scenario is developed to guide the development of more detailed mechanistic models of individual features, events, and processes that comprise the scenario. These detailed models are then integrated into a total-system model of the entire scenario. The integration of the more detailed models may include the models themselves or a simplified abstraction of the model results.

3. Develop vaules and/or uncertainty distributions for uncertain input parameters
After the models are developed, values must be assigned to the parameters to populate the model. If the parameter is well characterized, a single deterministic value may be assigned. However, uncertainty and/or variability in the parameter may require the use of distributions (e.g., log-normal, uniform, etc.) to define the values. Experimental data, literature sources, and professional judgment are often used to determine these distributions.

4. Preform calculations and sensitivity/uncertainty analyses
Calculations are performed using the integrated total-system model. Because stochastic parameters are used, a Monte Carlo approach is taken to create an ensemble of simulations that use different combinations of the input parameters. For each run (realization), a value for each input parameter is sampled from the uncertainty distribution, and the simulation is performed. The results of each realization are equally probable, and the collection of simulation results yields an uncertainty distribution that can be compared to performance objectives to assess the risk of exceeding those performance objectives or metrics. Sensitivity analyses can also be performed to determine which parameters the performance metrics are most sensitive to.

5. Document results and provide feedback to previous steps and associated areas to improve calculations, as needed
Document the findings, typically as cumulative distribution functions that present the probability (or risk) of exceeding a performance objective. These findings may be used to evaluate alternative designs, where performance objectives, cost, and schedule comprise some of the criteria in choosing the most suitable cover for a site.