October 9 - Anna Kelbert, Ph.D., Oregon State University. Earth System Bridge: NSF's EarthCube entry point for solid Earth geosciences
November 13 – Jed Brown, Ph.D., Argonne National Lab. Software Design and packaging for extensibility, provenance, and sharing
January 15 – Professor Louise Kellogg & Pierre Arrial, Ph.D., UC Davis. Influence of numerical discretization on preferred thermal convection patterns in a 3-D spherical shell
February 12–Eric Heien, Ph.D., & Hiro Matsui, Ph.D, UC Davis, Accuracy and Performance Benchmarks for Geodynamo Simulation
March 12 – Cedric Thieulot, Ph.D.; Anne Glerum, and Menno Fraters, University of Utrecht, ASPECT: from benchmarking to 3D subduction applications
April 9 – Professor Lucy Flesch, Purdue University, Work flows and 3-D geodynamic simulations of the India-Eurasia collision zone
May 28 - Arben Pitarka, Ph.D., Douglas Dodge, Ph.D, Steven Magana-Zook, Ph.D. & Stanley Ruppert, Ph.D., Lawrence Livermore National Lab, Ground motion simulation, seismic imaging, large-scale time series processing, and Big Data technology for solving earth science problems
NSF’s EarthCube is a relatively new Earth science knowledge integration initiative. It has a grand ambition to develop a common cyberinfrastructure for all of the environmental sciences in the United States. Even though the EarthCube initiative is actively seeking community engagement, it could easily win the prize for the most misinterpreted program in the National Science Foundation. In this talk, I give an overview of the aims and current state of the EarthCube program, and explain why and how, in my opinion, it could well be successful. I then describe my role in a funded EarthCube Building Blocks project, the “Earth System Bridge” (ESB), which intends to provide pathways for communication between existing modeling frameworks.
The early developments of the ESB project insofar as they are of relevance to the CIG community, are focused on providing a controlled vocabulary and a model metadata format that together would allow unambiguous description of numerical models for geodynamics, surface dynamics, seismology, magnetotellurics, and petrology. In addition to the obvious value of providing a comprehensive metadata format, this would also allow for model coupling, e.g., between tectonic and landscape evolution models, thus providing a computational bridge between the CIG and the Community Surface Dynamics Modeling System (CSDMS). I provide a progress update on these initial efforts, which I view as a possible CIG entry point into the cyberinfrastructure initiative, allowing us to help shape the outcome of the EarthCube effort. I explain how you could help by providing direct feedback, as well as the procedures for getting involved through the EarthCube governance, and discuss pathways to obtaining EarthCube funding.
There is more to developing successful scientific software than the core numerical implementation. Slapping an open source license on the code does not mean an army of talented developers will swoop in and turn your work into a wonderful package that everyone loves. This talk will discuss techniques to improve extensibility so that users and developers can extend your software to solve problems you never imagined; provenance so that published results can be understood, reproduced, and extended; and facilitate sharing of enhancements, configurations, and benchmarks to advance the software capability and encourage a vibrant and insightful community.
3-D numerical simulations of thermal convection in a spherical shell have become a standard for studying the dynamics of pattern formation and its stability under perturbations to various parameter values. The question arises as to how the discretization of the governing equations affects the outcome and thus, any physical interpretation. Motivated by numerical simulations of convection in the Earth’s mantle, we consider isoviscous Rayleigh–Bénard convection at infinite Prandtl number. We show that the subtleties involved in developing mantle convection models are considerably more delicate than has been previously appreciated, due to the rich dynamical behavior of the system. Two codes with different numerical discretization schemes – an established, community developed, and benchmarked finite-element code (CitcomS) and a novel spectral method that combines Chebyshev polynomials with radial basis functions (RBF) – are fully compared. A third code (ASPECT), using finite-element, is also tested for the comparison and reproduces partially results with other methods. This work demonstrates the impact of numerical discretization on the observed patterns, the value at which symmetry is broken, and how stability and stationary behavior is dependent upon it.
Numerical simulations of planetary dynamos have revealed many scientific insights over the past several years. However, due to the limitations of computing power current models are unable to properly resolve turbulent dynamos or approach simulation of realistic liquid metal dynamos. In order to better understand the best path forward to a realistic dynamo simulation, we have performed accuracy and performance benchmarks using 17 codes which use different numerical methods (spectral, finite difference, finite element, hybrid methods) and different domain decompositions. We examine the convergence rates of different methods, parameters necessary to resolve different physical regimes, and scalability of different schemes up to large numbers of processes. The goal of this work is to understand which methods work best in preparation for high resolution runs on O(10^5) cores.
ASPECT, the Advanced Solver for Problems in Earth's Convection, is an extensible open source, community supported code. The code is being applied to a broad range of problems in geodynamics. This talk will explore the community's efforts in benchmarking, implementation of visco-plastic rheologies as applied to 3D instantaneous modelling of subduction in the Aegean, and investigations of 'real free surface' vs. 'sticky air’ models of 2D thermo-mechanically coupled subduction.
The theory of plate tectonics tends to breakdown at continental collisional boundaries where deformation is diffuse and highly spatially variable. Large-scale continental deformation has traditionally been modeled as 2-D averaged over the thickness of the lithosphere or in 2-D cross section. Due the large increase in both seismic and geodetic observations within these regions it is obvious both vertical and lateral heterogeneity need to be considered. Thus, several groups are addressing the 3-D nature of continental lithosphere.
This talk will focus on the workflow and benchmarking of building a 3-D geodynamic lithospheric model. We will first discuss generating and benchmarking numeric simulations to analog models. Then discuss the expansion of these results to include the curvature of the Earth, lateral and vertical variable model parameters to begin simulations of the India-Eurasia collision zone.
Ground motion simulation, seismic imaging, large-scale time series processing, and Big Data technology for solving earth science problems
Arben Pitarka, Ph.D., Douglas Dodge, Ph.D, Steven Magana-Zook, & Stanley Ruppert, Ph.D.
Lawrence Livermore National Lab
We begin with an overview of current projects at LLNL on source and ground motion simulation using HPC. In the second part of the talk, we will present the phenomenology that supports large-scale time series processing for signal detection and identification. In the third part, we will present options of Big Data technology for use in solving earth science problems. We will talk about several technologies, present preliminary results, and discuss a roadmap for future investigations.