NASA HPCC Earth and Space Sciences Project
Science Team Symposium
 |
NASA Headquarters
|
 |
|
NASA Headquarters
|
|
| 8:30 | Gather/coffee |
| 9:00 | Welcoming remarks/objectives Joe Bredekamp, NASA Headquarters/STI ESS Project Manager - Jim Fischer, NASA Goddard Space Flight Center ESS Project Scientist - George Lake, University of Washington |
|
|
| 9:15 | Simulating Origins ESS Project Scientist - George Lake, University of Washington |
| 9:45 | Modeling Merging Neutron Stars: A Testbed for Developing a General Purpose Relativistic Hydrodynamics Code for Astrophysical Simulations PI - Paul Saylor, University of Illinois at Urbana-Champaign Speaker - Doug Swesty, University of Illinois at Urbana-Champaign |
| 10:15 | Coffee |
| 10:30 | Turbulent Convection and Dynamos in Stars PI and Speaker - Andrea Malagoli, University of Chicago |
| 11:00 | Modeling Solar Activity PI - John Gardner, Naval Research Laboratory Speaker - Spiro Antiochos, Naval Research Laboratory |
| 11:30 | Multiscale Modeling of Heliospheric Plasmas: Development of a 3-D AMR MHD Code for Massively Parallel Computers PI - Tamas Gombosi, University of Michigan Speaker - Ken Powell, University of Michigan |
| 12:00 | Lunch |
|
|
| 1:00 | Scalable Parallel Finite Element Computations of Rayleigh-Benard-Marangoni Problems in a Microgravity Environment PI and Speaker - Graham Carey, University of Texas at Austin |
|
|
| 1:30 | Probing the Geodynamo PI and Speaker - Peter Olson, Johns Hopkins University |
| 2:00 | SAR Imaging and Interferometric Science Applications Project Dave Curkendall, Jet Propulsion Laboratory |
| 2:30 | Coffee |
| 3:00 | Development of an Earth System Model: Atmosphere/Ocean Dynamics and Tracers Chemistry PI - Roberto Mechoso, University of California, Los Angeles Speaker - John Farrara, University of California, Los Angeles Speaker - Yi Chao, Jet Propulsion Laboratory |
| 3:30 | High Performance Computing at NASA's Data Assimilation Office PI and Speaker - Peter Lyster, University of Maryland |
| 4:00 | Short break |
|
|
| Principal Investigator statements of top accomplishment and recommendations | |
| 5:00 | Adjourn |
George Lake (speaker), Thomas Quinn, Derek Richardson, Joachim Stadel, Jeff Gardner, Department of Astronomy, University of Washington
NASA has set ambitious goals to understand the origins of planetary systems, galaxies, and large-scale structure. To create accurate models of the formation and evolutionary processes, we need simulations to connect inherently incomplete observations to true physical structures, and simple physical theories to today's nonlinear structures.
Using spatially and temporally adaptive N-body codes that perform well on the HPCC ESS testbed, we have attacked a wide range of problems stimulated by the data from the Great Observatories:
Recently, we've been adapting our code to perform simulations of the evolution of the terrestrial planets starting with a million planetesimals. We've also formulated and tested parallel methods to study the stability of planetary systems (such as our nine-planet system) that should scale well to systems with thousands of processors.
Web link:
http://www-hpcc.astro.washington.edu/HQ.html
Doug Swesty (speaker), Department of Astronomy and National Center for Supercomputing Applications, University of Illinois at Urbana-Champaign
The field of astrophysics is experiencing an unprecedented phase of growth in the quantity and quality of observational information about high-energy astrophysical phenomena from orbiting astronomical observatories. This wealth of data presents new challenges to the theorist who would attempt to model phenomena such as stellar accretion disks, active galactic nuclei, supernovae, and compact object mergers. The modeling of such astrophysical objects requires the simulation of relativistic fluid flow in 3-D and over a wide variety of spatial scales. In many cases both the equations of relativistic fluid dynamics and the Einstein field equations (general relativity) must be solved. The solution of the Einstein field equations poses a large memory and floating-point requirement thus requiring the use of cutting-edge HPCC technology.
As a testbed problem for the development of a general purpose 3-D relativistic hydrodynamics code, we are modeling the merger of binary neutron star systems. I will discuss some of the aspects of design and development of this code. In addition to its excellent attributes as a testbed problem, the merger of neutron stars is of tremendous scientific interest. Such mergers are thought to be the possible source of the mysterious gamma-ray bursts seen by the NASA Compton Gamma-Ray Observatory and are thought to be sources of observable gravitational radiation. I will briefly discuss how our models impact both of these subjects.
Web link:
http://jean-luc.ncsa.uiuc.edu/nsnsgc/
Andrea Malagoli (speaker), Department of Astronomy and Astrophysics, University of Chicago
Strong magnetic activity in the Sun and Sun-like stars originates many spectacular and well observed phenomena like sunspots, solar flares, the solar cycle, and the solar wind. Such activity, known generically as "solar activity," is believed to originate from the interaction between magnetic fields and highly turbulent convective motions taking place below the surface of the sun. The wide range of spatial and temporal scales of the motions leads to the continuous regeneration and evolution of the magnetic fields by means of the "dynamo" process.
In this presentation, I will describe recent results on large-scale simulations of magneto-convection and dynamo processes in the sun obtained by members of this team. These models make use of the application codes developed to meet the technical milestones of our NASA cooperative agreement, while their scientific rationale originates from research conducted under NASA's scientific programs like SPTP and SR&T. The results are novel and unique in that they explore complex physical regimes where the underlying equations are strongly nonlinear, which is in close agreement with the phenomenology suggested by observations. The results also demonstrate how future progress in this field will be strongly determined by progress in large-scale computer simulations and in the high-performance computing technologies enabling them.
Web link:
http://astro.uchicago.edu/Computing/HPCC
Spiro Antiochos (Speaker), Russ Dahlburg, Rick DeVore, John Gardner, Judith Karpen, Jim Klimchuk, Laboratory for Computational Physics and Fluid Dynamics and E. O. Hulburt Center for Space Research, Naval Research Laboratory
With the advent of the revolutionary observations of solar activity by the NASA/ESA SOHO mission and the great advances in numerical technology brought about by the NASA/DoD HPCC program, it has been possible to achieve a new level of understanding of the major solar drivers of the Sun-Earth connection. With the EIT and LASCO telescopes on SOHO we have been able to observe in unprecedented detail the structure and dynamics of the whole solar atmosphere, from the chromosphere out to 30 solar radii. With the numerical technology provided by the HPCC program at NRL, we have been able to develop three new highly-optimized parallel codes: CRUNCH3D -- a spectral code for investigating basic processes such as 3-D reconnection and turbulence generation, MHD3D -- an FCT-based code for modeling observed phenomena with complex global structure, and MESH3D -- a fully adaptive grid FCT code for investigating the interaction between micro- and macro-scales.
We present results of the application of these codes to two examples of solar activity -- the formation of solar prominences/filaments and the initiation of coronal mass ejections (CME's). These phenomena are closely related in that it is often the eruption of a massive filament (up to 1016 gm) that is the major component of the CME event. It is believed that the prominence mass accompanying a strong CME in January 1997 may have been responsible for the failure of a telecommunications satellite. We describe our theoretical models and numerical simulations of prominences and CME's and show that the key features of the phenomena lie in the 3-D complexity of the solar magnetic field. We conclude that, due to the increasing simulation power afforded by the HPCC program, the essential physics of solar activity is now beginning to be understood.
Web links:
http://www.lcp.nrl.navy.mil/hpcc-ess/
http://www.lcp.nrl.navy.mil/hpcc-ess/crunch3d.7.html
http://www.lcp.nrl.navy.mil/hpcc-ess/fctmhd3d.7.html
http://lasco-www.nrl.navy.mil/
http://www-istp.gsfc.nasa.gov/istp/cloud_jan97/
Tamas Gombosi, Ken Powell (speaker), Quentin Stout, Darren De Zeeuw, Clinton Groth, Hal Marshall, College of Engineering, University of Michigan
In this talk, a 3-D AMR MHD code designed to model heliospheric plasmas will be described. The code has achieved a sustained rate of 64.4 gigaFLOPS on a coronal mass ejection problem. This is 23 percent peak performance on the CRAY T3E at GSFC. The code has achieved the highest performance to date of the NASA HPCC ESS computational Grand Challenge codes.
The code uses solution-adaptive mesh refinement (AMR); the mesh on which the governing equations are being solved adapts, on the fly, to the plasma flow. In this way, under-resolution of high-gradient regions and over-resolution of low-gradient regions are avoided, leading to highly efficient use of computer resources.
In this work, a "start from scratch" approach was taken. The data structure, the AMR algorithm and the governing-equation solution algorithm were designed with maximum parallel performance in mind. The resulting code has high single-processor performance and scales to 512 processors with very high efficiency.
The code has been and is being applied to the following space science application areas:
In the talk, the philosophy of the solution adaptation and the parallelization will be discussed, and results from several application areas will be shown.
Web link:
http://hpcc.engin.umich.edu/HPCC
Graham Carey (speaker), Department of Aerospace Engineering and Engineering Mechanics, University of Texas at Austin
This research project concerns the analysis and simulation of coupled viscous flow and transport processes, including free surface thermocapillary effects. Of particular interest are microgravity applications for manufacturing and life support processes relevant to the U.S. space station, space shuttle program, and future space projects. The main thrust of the present research is directed to the development of a highly parallel distributed simulation capability for analyzing 3-D transient and steady state coupled viscous flow and transport processes in this class. However, the simulator will also be more broadly applicable to other problems of interest to NASA, U.S. manufacturing, and science and engineering. E.g., it can be applied more generally to terrestrial manufacturing processes and will be particularly important to problems involving thin films and coatings since surface tension is also dominant (over gravity) in these applications. Part of the research work also concerns supporting experimental studies on thermocapillary effects in thin films.
The approach is based on a Galerkin finite element formulation using hexahedral elements with piecewise continuous triquadratic approximation for the velocity, temperature (or species) fields, and trilinear pressure finite element basis. Parallelism is achieved using a domain decomposition of the flow domain and mesh. The resulting discretized problem leads to a sequence of sparse algebraic systems that are solved using a parallel form of generalized conjugate gradient solver. Special attention has been paid to parallel performance issues to meet the performance milestones specified in the contract. We have currently run the code at 50 gigaFLOPS on the GSFC platform and 113 gigaFLOPS on the 1,280-processor CRAY T3E at Silicon Graphics/Cray Research in Eagan, MN. The experimental work has been directed to fundamental studies of free surface phenomena for thin films being carried out in the Experimental Flows Lab of the Center for Nonlinear Dynamics.
Web link: http://zoyd.ae.utexas.edu/nasa_hpcc/
Peter Olson (speaker), Department of Earth and Planetary Sciences, Johns Hopkins University
The Earth is among the majority of planets within the solar system possessing magnetic fields that originate from self-sustaining dynamos. The number of planetary dynamos continues to grow: the Galileo mission has discovered that Io and Ganymede also have magnetic dynamos, and new evidence from the Global Surveyor suggests an ancient dynamo may have operated in Mars. Planetary magnetism is a common phenomenon because only three basic ingredients are needed for a self-sustaining dynamo: a sufficiently large volume of an electrically conducting, fluid core, such as the Earth's iron-rich outer core; an energy source such as convection to circulate the fluid; and planetary rotation to impart helicity to the fluid motions.
Although seemingly simple, the details of planetary dynamos have long remained elusive. But, recently, breakthrough advances have come using direct numerical simulations of 3-D, time-dependent, convection-driven dynamos. These numerical simulations not only reproduce many of the observed characteristics of the Earth's magnetic field but also offer new insights into the process of magnetic field generation in other planets.
In this talk I present the results of several numerical dynamo models which explain many of the important characteristics of the geomagnetic field, including its dominantly axial dipole, westward drift of the non-dipole field, and the recently-discovered super-rotation of the solid inner core. We find that these dynamo models exhibit occasional reversals in magnetic polarity, when the heat flow on the boundary between the liquid core and solid mantle is spatially heterogeneous. These results suggest that thermal coupling between the core and the mantle may be the ultimate cause of geomagnetic polarity reversals.
Web links:
http://curie.eps.jhu.edu/
http://www.igpp.lanl.gov/Geodynamo.html
http://www.psc.edu/science/glatzmaier.html
Dave Curkendall (speaker), Center for Space Microelectronics Technology, Jet Propulsion Laboratory
The overall Synthetic Aperture Radar (SAR) Imaging and Interferometric Science Applications Project is a complex one, being composed of a computational team and three separate science sub-teams.
The computational team is guided by and is largely responsible for the main computational milestones of the project: 10, 50, and 100 gigaFLOPS In addition, the objectives of the computational area include the design and implementation of the Scalable SAR Software Suite (S4), a publicly available set of software for high-performance SAR image and interferometry processing (InSAR). Finally, this team is charged with giving the science teams computational support and for developing the visualization techniques used by the Project. Recent accomplishments include achievement of the 50 gigaFLOPS milestone, a new Digital Light Table software system for the display of very large SAR mosaics, and the extension of (S4) for multi-platform InSAR support.
Scripps Institution of Oceanography SAR Interferometry: The long-range objectives of our research are to understand the kinematics and dynamics of active plate boundaries through precise geodetic and geophysical measurements. To pursue this our specific objectives are to i) generate a highly accurate, uniformly sampled digital elevation model for the Southern California region, ii) understand the short-term transients associated with the delay of the radar echo through the troposphere and ionosphere and mitigate the contamination of the interferograms by these effects, and iii) use the repeat images to attempt to unravel the co-seismic, post-seismic, and inter-seismic components of crustal deformation within the plate boundary zone. Recent accomplishments include the development of a new "phase gradient technique" that allows the interpretation of SAR interferograms without phase unwrapping and permits averaging of multiple interferograms for noise reduction. The application of this technique to the recovery of Southern California topography using ERS images will be shown.
UC Santa Barbara Snow Measurement: The objective of the Snow Measurement Project is to develop and prove techniques using SAR data for the modeling and forecasting of snow melt runoff. The prime vehicle for this research has been the multi-band/polarization properties of the SIR-C data. The phenomenologies employed have recently been broadened to include repeat orbit interferometric data as well.
JPL ARVORES: The objective of the ARVORES (Amazon Rainforest Visualization/Classification by Orbiting Radar, Enabled by Supercomputers) project has been to use the computational facilities at JPL to process the large Amazon SAR data set and study the large-scale phenomenology related to the Amazon basin. The data set features dual season (low flood: October 1995 and high flood: June 1996) L-band SAR coverage obtained by the Japanese JERS-1 satellite. The total land mass imaged by the radar was on the order of 1,500 scenes for each season's campaign covering approximately 12 million square kilometers. Recent accomplishments include a 100 m low flood mosaic and preliminary classification of the entire land mass.
Web links:
http://topex.ucsd.edu/SAR/sar.html
http://www.icess.ucsb.edu/hydro/sirc/sircpg.html
http://alphabits.jpl.nasa.gov/SAR/images/lghtbox.gif
John Farrara (speaker), Department of Atmospheric Sciences, University of California, Los Angeles; Yi Chao (speaker), Jet Propulsion Laboratory
We are developing a state-of-the-art model that describes the coupled global atmosphere-global ocean system, including chemical tracers that are found in, and may be exchanged between, the atmosphere and the oceans. Such an Earth System Model can be used to study nonlinear interactions and feedbacks between different components of the climate system, leading to a better understanding of phenomena such as the El Niño-Southern Oscillation (ENSO), the role of the oceans in moderating the greenhouse warming effect of carbon dioxide, and the Antarctic ozone hole.
The components of our model are: an atmospheric general circulation model (AGCM), an oceanic general circulation (OGCM), an atmospheric chemistry model (ACM), and an oceanic chemistry model (OCM). Parallel versions of each of the AGCM, OGCM, and ACM are currently operational.
Our model is currently being applied in the following investigations: i) the simulation and prediction of ENSO events, ii) the impact of ENSO events on world climate, and iii) ocean modeling at eddy-resolving resolutions. A 50-year simulation of the coupled atmosphere-ocean model, which shows a substantial improvement in the seasonal cycle and interannual variability, was recently completed in support of i). A high-resolution version of the AGCM has been run on the CRAY T3E for the equivalent of 10 simulated years in support of research on ii). In support of iii), the OGCM has been integrated for 40 years using the CRAY T3D/E at 1/6 degree resolution, the longest such simulation conducted at eddy-resolving resolutions.
The component codes are being optimized for the CRAY T3E and coupled using a distributed Data Broker software developed as part of this project. Our initial optimization efforts centered on minimizing the impact of substantial static and weakly dynamic load imbalances among processors through load redistribution schemes. Recent optimization efforts have centered on single-node optimization. Strategies employed include loop unrolling, the use of an optimized assembler-code library for special function calls, and restructuring of parts of the code to improve data locality. Currently, the model code runs four times faster on the T3E than on the T3D.
Web links:
http://www.atmos.ucla.edu/~drummond/esm/
http://uniblab.atmos.ucla.edu/~vwk206/fcst.html
http://lochness.jpl.nasa.gov/hpcc.html
Peter Lyster (speaker), Earth System Science Interdisciplinary Center, University of Maryland
Atmospheric data assimilation is a method of combining actual observations with model simulations to produce a more accurate description of the Earth system than the observations alone provide. The output of data assimilation, sometimes called "the analysis," are accurate, regular, gridded data sets of observed and unobserved variables. This is used not only for weather forecasting but is becoming increasingly important for climate research. For example, these data sets may be used to assess retrospectively energy budgets or the effects of traced gases such as ozone. Such uses allow researchers to understand processes driving weather and climate, which have important scientific and policy implications. The work carried out under this Principal Investigator project is in support of software used at NASA's Data Assimilation Office (DAO). The primary goal of the DAO is to provide data sets for climate research and to support NASA satellite and aircraft missions. This presentation will cover the principles of data assimilation, discuss the relevance of high-performance computing, and finally show some examples where computing plays a key role. Available are a discussion of the scientific and computational complexity of atmospheric data assimilation and general information on the Data Assimilation Office.
Web links:
http://dao.gsfc.nasa.gov/DAO_people/lys
http://dao.gsfc.nasa.gov
