GEM (Earthquake Science) Computational Infrastructure

Investigator: G.Fox (with Group of Computer and Earth Scientists from Academia and Government). The importance of simulating earthquakes is intuitively obvious. For instance, the recent January 16, 1995 Kobe, Japan earthquake was only a magnitude 6.9 event and yet produced an estimated $200 billion loss. Despite an active earthquake prediction program in Japan, this event was a complete surprise. Drastic scenarios similar to those reported in 1999 for Turkey and Taiwan are possible and indeed eventually likely in Los Angeles, San Francisco, Seattle, and other urban centers around the Pacific plate boundary.

Over the last three years, Fox has worked with a team of Computer and Earthquake scientists from academia and government to initiate a program GEM "General Earthquake Models” aimed at applying the latest computational technology in this area. This thrust contributes to the nationally identified importance of developing new approaches to Geoscience involving advanced instrumentation (EarthScope) and computing. The GEM group includes representatives of over a dozen universities, multiple government agencies and laboratories (DoE, NASA, NSF, USGS) and is coordinated with the major NSF Southern California Earthquake Center (SCEC). Fox leads the design of the computational infrastructure GEMCI for this effort and we will use the proposed CSIT resources as the site as an experimental system on which to test GEMCI without the limiting requirements (for computer science research) of production use. We note that as Fox’s move to FSU is still in progress, we have not integrated many existing CSIT faculty capabilities into GEMCI and we will naturally work on this during the next months. CSIT expertise in Problem Solving Environments, data assimilation, visualization and basic numerical algorithms are all relevant to GEMCI.

GEM is an attractive target for computer science due to the intrinsic richness of the applications and the societal importance and also because the use of computers is not yet too extensive and so modern approaches and infrastructure can be used without major distraction from existing legacy approaches. The field is naturally distributed with sensors, scientists and earthquakes scattered around the globe. Thus there is an immediate application of emerging concepts such as computational grids to link large-scale simulations, data and people in a distributed fashion. Our computer science research focuses on the issues on building an integrated information infrastructure that can support collaborative distributed scientific research over a range of time scales and computational needs. The work of our collaborators has been divided into three application area timeframes characterized by time scales of hours (post earthquake analysis), 6-12 months (data assimilation and development of new earthquake forecasting approaches) and ten years (fundamental theory). GEMCI work is divided into five thrust areas; distributed collaborative (shared) scientific objects, HPCC simulations including new uses of fast multipole techniques, multi-sensor metadata, data and simulation visualization, and interactive scientific datamining for earthquake pattern analysis. Fox’s initial activities focused on use of multipole techniques where GEM shows interesting differences from previous applications to areas like astrophysics and CFD. Now we are concentrating on prototype science portals and work on collaborative systems to support this range of application requirements. Further as mentioned in parallel Java discussion, much of the code for this field can be developed from scratch and we expect to experiment with the use of Java as the coding language.

GEMCI will make use of the full range of proposed resources in this proposal. As well as central simulation engines, we need to support hand-held devices for workers in the field and experts on call for real-time participation in analysis of an earthquake in process. Further as indicated by our list of thrust areas, visualization resources will be an important capability.