Auszug aus

International Workshop on

How to integrate geochemistry at affordable costs into reactive transport for large-scale systems

February 5 - 7, 2020

Abstract Book

The focus of the workshop was:

 To provide and discuss existing geochemical concepts in reactive transport modelling to describe sorption and related retardation processes of contaminants on a variety of sediments and rocks.

 To explicitly set focus on large-scale natural systems as experienced, e.g., in nuclear waste disposal, carbon capture & storage, environmental remediation, or geothermal applications.

 To explore how the discussed approaches can be integrated at affordable costs into current paradigms in THMC models and long-term safety assessments in general.

Bounding Computation and Complexity for Reactive Transport on Supercomputers: A Perspective from the Nuclear Waste Repository Performance Assessment Community

Glenn Hammond*, Peter Lichtner+, Paul Mariner*, David Sevougian*, Emily Stein*

*Sandia National Laboratories

+OFM Research, University of New Mexico

Simulation of geologic repositories for nuclear waste pushes computational resources to their limits. To be effective and predictive, performance assessment (PA) calculations must couple multiple processes, such as thermal, hydrologic, mechanical, and chemical (THMC), over spatial scales ranging from pore to field scale. Large statistical ensembles of these million-year simulations are generally required to account for uncertainties in the available data, future scenarios and conceptual representation.

The US Department of Energy leverages a balance of mechanistic and reduced-order modeling (ROM) and high-performance computing (HPC) to address these demanding computational requirements for PA in its development of the GDSA (Geologic Disposal Safety Assessment) Framework. The use of reduced-order approximations to THMC process models can lessen computational burden by reducing memory and runtime requirements, while supercomputing can expand available memory and processing power, enabling increased complexity and sophistication. Both ROM and HPC make ensemble-based, large-scale, long-time performance assessment more tractable.

In developing simplified geochemical process models, it is advantageous to first delineate the HPC resources required to simulate mechanistic reactive transport on a large, realistic problem domain. Reduced-order geochemical process models can then be employed on the same problem domain with similar HPC resources in order to quantify the computational savings produced by using the simplified model. For instance, comparing the parallel performance (runtime and scalability) of solute transport with linear sorption (or even solute transport alone) to that of multicomponent reactive transport with aqueous speciation and surface complexation bounds the computational requirements for sorption processes. To this end, we present a comparison of computational performance for reactive transport modeling on a large HPC cluster – specifically, a comparison of PFLOTRAN performance for representations of the sorption process over a large spatial domain: equilibrium and multi-rate kinetic surface complexation vs. linear sorption.

Sandia National Laboratories is a multimission laboratory managed and operated by National Technology & Engineering Solutions of Sandia, LLC, a wholly owned subsidiary of Honeywell International Inc., for the U.S. Department of Energy’s National Nuclear Security Administration under contract DE-NA0003525. SAND2019-13626 A

LAB-LYSIMETER MODELLING

Daniel Jara Heredia1, Marcus Böhm1, Arno Märten1, Volker Hormann2, Dirk Merten1, Thorsten Schäfer1

1) Friedrich-Schiller-Universität Jena (FSU), Institut für Geowissenschaften, Angewandte Geologie, Jena, Germany

2) Universität Bremen, Institut für Umweltphysik (IUP), Bremen, Germany

The complex soil-liquid system is not trivial to model due to several reasons such as 

• limited spatial resolution of data, 
• lack of knowledge on the interplay between the different components (e.g. microbes), 
• mathematical ill-conditioned systems, 
• the difference heterogeneities among soils, etc. 

We approach such complex system by using a sorption component additive approach (Hormann 2015; Hormann and Fischer 2013), which is made up of 

• illite as main active component in clay (Bradbury and Baeyens 2000; Bradbury and Baeyens 2009a, 2009b), 
• hydrous ferric oxides (Dzombak and Morel 1990) and 
• humic substances (Tipping, Lofts, and Sonke 2011; Marsac et al. 2017). 

Although, this chemical system can be solved within the classical reactive transport approach at the continuum scale where each representative element volume is in general treated as a batch reaction, the exploration of new techniques such as smart-Kd (Stockmann et al. 2017) or surrogate models (Jatnieks et al. 2016) is nowadays a must for two main reasons: 

1. reduction of the computational time, since usually the computational time of chemistry is higher than the one of transport (van der Lee et al. 2003),
2. reduction of the complexity of the chemical system, e.g. Tipping model VII requires more than 50 mass action laws for each cation that interacts with the humic substance, making it prone to multiple assumptions.

----------------------------

Version 4.1.2024

Address of this page

Home

Joachim Gruber