Panel Leads
- Site Characterization and Model Calibration
- Tissa Illangasekare - Colorado School of Mines
- Jack Parker - Oak Ridge National Laboratory
- Omar Ghattas - University of Texas at Austin
- Roelof Versteeg - Idaho National Laboratory
- Validation, Verification, Uncertainty Analysis and Decision Optimization
- Dongxiao Zhang - University of Oklahoma
- Jon Helton - Sandia National Laboratory
- Peter Kitanidis - Stanford University
- Don Estep - Colorado State University
- Coupled Phenomena
- Dick Ewing - Texas A&M University
- Steve Yabusaki - Pacific Northwest National Laboratory
- Clint Dawson - University of Texas at Austin
- Mark White - Pacific Northwest National Laboratory
- Carbon Sequestration
- Diana Bacon - Pacific Northwest National Laboratory
- Mary Wheeler - University of Texas at Austin
- Rajesh Pawar - Los Alamos National Laboratory
- Research at Fundamental Scales
- David Dixon - University of Alabama
- Tim Scheibe - Pacific Northwest National Laboratory
- Brent Lindquist - State University of New York at Stony Brook
- Bjorn Engquist - University of Texas at Austin
Panel Charges
We ask each panel to identify what crosscutting computer science, theory, mathematics and information technologies are needed to achieve scientific defensible tera/petascale subsurface science models. Among the topics for discussion are
- Paths forward for hardware and algorithm development;
- New data, knowledge and project management tools;
- Potential uses of advanced scientific visualization and geographic information systems.
We also ask that each panel to identify some of the "grand challenges" in high fidelity subsurface sciences modeling at both the fundamental and applied scales.
- Site Characterization and Model Calibration
- All numerical predictive models of large scale subsurface phenomena require site-specific geologic data and boundary and initial conditions for each modeled quantity. Site characterization typically involves strongly nonlinear (mathematically ill-posed) inverse problems that are often presented as parameter estimation problems. Regularization, linearization and optimization are some mathematical tools used to attack these difficult inverse problems. However, in practice, modelers often must rely on old-fashioned "engineering intuition" to approximate aquifer characteristics and boundary and initial conditions. The charge to this panel is to
- Provide a survey of the current methods for subsurface modeling site characterization.
- Assess the potential role of tera/peta-scale computational science for improving site characterization methodology.
- Validation, Verification, Uncertainty Analysis and Decision Optimization
- Subsurface science problems are today being analyzed using increasingly complex mathematical models. Numerical models are based on conceptual models that seek to describe the basic processes of a physical system. The computer code for a numerical model is a set of statements that provides an algorithm for solving the equations that describe the processes modeled in the system. Code verification is a process designed to establish that a specific code accurately solves the mathematical equations that describe the basic processes of the physical system. Model validation is a process that seeks to determine or to provide assurance that the model accurately reflects the fundamental natural process occurring at a site, and therefore provides a reliable basis for decision-making. Uncertainty analysis seeks to quantify uncertainty in model predictions due to uncertainty in parameter estimates, data employed for calibration, errors in model assumptions, and other variables. Decision optimization is the computational process that seeks to determine optimal decisions for given applied problem solutions that meet specified performance, cost or other criteria. The charge to this panel is to
- Provide a survey of current verification and validation methods.
- Comment on how increased computational power could affect decisions regarding choice of conceptual and numerical models.
- Assess how to quantify model uncertainty and discuss using tera/peta-scale computational science to reduce uncertainty.
- Discuss improved practical decision-making under conditions of prediction uncertainty.
- Coupled Phenomena
- Many subsurface sciences problems are extremely complex and often involve strongly coupled flows of several fluids and transport of multiple reactive substances. In addition, we must often account for thermal effects, fluid phase change and alterations to the porous matrix and fluid composition due to adsorption/desorption or chemical reactions. All these complex processes must be modeled at multiple length and time scales, spanning orders of magnitudes. To obtain computationally tractable algorithms, current subsurface simulation models typically uncouple various physical processes and flow regimes. Lack of sufficient computational capacity is often cited as a reason for not using "first principles based" conceptual and computational models. The charge to this panel is to
- Provide a survey of current modeling practices for dealing with coupled, subsurface physical process.
- Investigate how tera/petascale computational power could support the development of a new generation of conceptual and numerical models, more firmly based on first principles, for understanding coupled, multi-physics subsurface processes.
- Discuss how increased computational power could be used to better deal with multiple space and time scales.
- Carbon Sequestration
- It is widely recognized that, "A much larger science-based CO2 sequestration program should be developed. The aim should be to provide a science-based assessment of the prospects and costs of CO2 sequestration. This is very high-risk, long-term R&D that will not be undertaken by industry alone without strong incentives or regulations, although industry experience and capabilities will be very useful." The charge to this panel is to
- Survey current mathematical models used for understanding carbon capture and carbon storage.
- Recommend ways that high performance computing could be used to new, more powerful simulation models
- Discuss how high fidelity simulations carbon sequestration models could be integrated with decision support systems to enable cost/risk analyses.
- Research at Fundamental Scales
- Our understanding of subsurface phenomena such as fluid flow, thermodynamics, chemical reactions, phase changes and adsoroption and desorption at fundamental scales form the basis for conceptual and numerical models at larger scales. In the subsurface sciences, fundamental scales can span many orders of magnitude, ranging from the nano-scale for some material science problems and upward to the pore scale or the laboratory scale for flow and transport problems. Improved understanding of processes at these fundamental scales can form the basis for more predictive high fidelity macroscopic, field scale models. We ask this panel to
- Survey current computational approaches used for understanding subsurface sciences processes from the nanoscale to laboratory scale
- Address multiscale mathematical issues related to transferring knowledge about a process across scales.
- Assess the potential of tera/petascale computations to advance research at fundamental scales.
