Title of Presentation: QuakeSim: Efficient Modeling of Sensor Web Data in a Web Services Environment
Primary (Corresponding) Author: Andrea Donnellan
Organization of Primary Author:
Co-Authors: Jay Parker, Robert Granat, Geoffrey Fox, Marlon Pierce, John Rundle, Dennis McLeod, Rami Al-Ghanmi, Lisa Grant, Walter Brooks
Abstract: QuakeSim is a project to develop a modeling environment for studying earthquake processes using a web services environment. The multi-scale nature of earthquakes requires integrating many data types and models to fully simulate and understand the earthquake process. QuakeSim focuses on modeling interseismic processes and the multiple data types that must be ingested include spaceborne GPS and InSAR data, geological fault data, and seismicity data. QuakeSim federates data from these multiple sources and integrates the databases with modeling applications. Modeling applications include various boundary element, finite element, and analytic applications, which run on a range of platforms including desktop and high end computers. Because the models are complex and compute intensive we are using the Columbia computer located at NASA Ames to integrate and run software programs to improve our understanding of the solid Earth and earthquake processes. The complementary software programs are used to simulate interacting earthquake fault systems, model nucleation and slip on faults, and calculate run-up and inundation from tsunamis generated by offshore earthquakes. QuakeSim also applies pattern recognition techniques to real and simulated data to elucidate subtle features in the processes.
The QuakeSim earthquake forecasting methodology is proving extremely successful. In the last six months five earthquakes above magnitude 5 have occurred in identified hotspots. The identified hotspots make up only 1.2% of the total map area of the forecast or the state of California. The approach is to minimize the forecast area, which is essentially the false alarm rate, while still detecting all the large earthquakes (maximizing the hit rate). Our other pattern recognition technique, RDAHMM, which analyzed GPS time series data identified a reference frame error in the data processing introduced by a large earthquake in Siberia. The GeoFEST finite element software is being used for supporting NASAís decadal survey DESDynI mission to establish requirements such as the need for 3D vector deformation. We are working solutions with both NASA Ames and JPL to add grid services so that QuakeSim can interface with Columbia and Cosmos. QuakeSim represents the first major user of Columbia that has identified Grid services and Condor G as a requirement for job launch. The QuakeTables database now includes InSAR interferograms. Analysis of Virtual California interacting fault systems shows that events on the southern San Andreas fault typically follow, but do not precede, events on the Eastern California Shear Zone. Other modeling work shows the need for adding complexity to the models and indicates that there is 1 mm/yr of postseismic motion still occurring from the 1906 San Francisco earthquake. QuakeSim is also being used in the classroom in geophysics and tectonics classes.