Comparing Geologic Repository Assessment Tools

There is broad international scientific consensus that a geologic repository is the safest and most secure method of permanently disposing of this spent nuclear fuel, Stein said. Sandia’s framework is designed to assess the safety of future geologic repositories and assist in licensing efforts. It can model everything from what happens to the spent nuclear fuel rods and the canisters that contain them, to the rock mere inches from the canisters or rock miles away.

“All of us in the international nuclear community are in the same boat — we need a safe place for our waste,” said Paul Mariner, a lead Sandia engineer on the project. “The framework is a system of codes that we can use to build a total-system model of a potential repository for a safety assessment. It really comes down to probabilistic risk assessment, which is a way to carefully and methodically address all of the questions regulators and citizens have about a repository by accounting for the probabilities of various events that could cause the release of radioactive material.”

The Geologic Disposal Safety Assessment framework is built upon two core Sandia software packages: PFLOTRAN and Dakota. PFLOTRAN is open-source software maintained and developed at Sandia and several other national labs that models how chemicals react, and how liquids and gases flow, underground, Stein said. “This is an important toolset because the primary thing we’re concerned with for a deep geologic repository is the movement of radioactive atoms out of the repository.”

Dakota is Sandia software that specializes in uncertainty sampling and sensitivity analysis, in other words determining which input values have the largest impact on the final results. Since no one can predict the future, Dakota assures researchers that the results are trustworthy, even if, for example, their estimate for the likelihood of a big earthquake occurring at the location over the next million years is uncertain. Laura Swiler, a Sandia computer scientist, has been instrumental in ensuring that the team has the sensitivity analysis tools needed for the safety assessment framework, Stein said.

“Uncertainty quantification calculations help us better characterize the repository system and also help us assess the significance of certain input values and processes that affect the final results,” Mariner said.

Stein added that the framework is massively parallel and runs on supercomputers, which is important for modeling what happens at a repository over a million years.

Recently a Sandia team, led by Michael Nole, made some significant improvements to PFLOTRAN to speed up calculations of certain challenging processes and improve the representation of how dry soil acts like a sponge, Stein said.

International Competition Improves Confidence
In late April, 10 different teams of repository scientists from across the globe, including teams from Canada, Germany and Taiwan, will compare the safety-assessment models they developed for two different reference cases. These reference cases are hypothetical repositories in two different kinds of rock that would be relevant for every team. One reference case is for a repository in a salt formation. The other reference case is for a repository in crystalline rock. Each case comes with a set of key questions, which took about two years of discussion to fully define along with the details of the hypothetical repository.

Comparing the results of Sandia’s software against the results from other teams’ software will build confidence in the software and models, Stein said. Specifically, the teams will compare the results of their models to see if there is a consensus reached by multiple teams. Then the models that produced different results will be studied to try to determine why their results were different. This will allow repository scientists from across the globe to learn from one another and improve their models. This comparison is part of an international collaboration called Development of Coupled Models and their Validation against Experiments, or DECOVALEX.

“The DECOVALEX initiative creates an important framework for experts in repository sciences from around the world to test and improve simulation models that are important to assessing the safety of geologic disposal,” said Jens Birkholzer, the chairman of the initiative and a senior scientist at Lawrence Berkeley National Laboratory. “Seventeen partner organizations, 50 modeling teams, and more than 100 simulation experts are currently involved in this global effort. Through collaborative model comparison, DECOVALEX demonstrates that long-term prediction of complex subsurface processes associated with geologic disposal can be achieved with a high level of confidence.”

Specifically for the two DECOVALEX reference cases, the Sandia team expanded the framework to model how salt mined out of a salt-based repository and then replaced around waste containers can heal itself back into solid, impermeable blocks over millennia, Stein said. They also refined their model on how water can flow around cracked crystalline rock and possibly move radioactive atoms within a geologic repository. These two improvements will likely be helpful in answering safety assessment questions about future U.S. spent nuclear fuel repositories.

In the future, Stein and her team hope to add more machine learning algorithms to the framework to speed up modeling some complex processes without impacting accuracy or transparency, she said.

“This type of comparison, where we’re all starting with the same two problem descriptions and are developing numerical models around them is one important way to create confidence in safety assessment models and the tools we’re using to do the simulations,” Stein said. “We can’t actually take this big, full million-year simulation and compare it to a real-world experiment of the same scale. It’s a very important exercise.”

Other DOE national laboratories involved in developing the framework and the DECOVALEX initiative include Argonne, Idaho, Lawrence Berkeley, Lawrence Livermore, Los Alamos, Oak Ridge and Pacific Northwest national laboratories.