2016 SCEC Undergraduate Studies in Earthquake Information Technology (UseIT): Earthquake Forecasting Through Physics-Based Simulations

Zhenyu Fu, Morgan T. Bent, Hernan M. Lopez, Spencer P. Ortega, & Kevin C. Scroggins

Submitted September 4, 2016, SCEC Contribution #6524, 2016 SCEC Annual Meeting Poster #329 (PDF)

Poster Image: 
As part of the 2016 Undergraduate Studies in Earthquake Information Technology (UseIT) internship program, students worked in collaborative groups to tackle unsolved problems in earthquake information technology presented in the form of a Grand Challenge. Earthquake forecasting was the overall theme of this year’s Grand Challenge. The UseIT High Performance Computing (HPC) Team was challenged to simulate long catalogs of California’s seismic activity on a supercomputer. The team used a physics-based earthquake simulator, the Rate-State earthQuake Simulator (RSQSim). The students ran the simulations on the Blue Waters system at the University of Illinois, one of the most powerful open-science supercomputers in the world. To configure an RSQSim simulation, a series of initial physical parameters must be specified, including initial normal and shear stresses, rate- and state-friction parameters, and earthquake slip rate. The HPC Team studied the effects of these parameters on simulated catalogs. Each team member ran several short simulations (25,000 simulated years) with different input parameters and then compared the catalogs with the Uniform California Earthquake Rupture Forecast Version 3 (UCERF3) to determine which parameter set produced the best match. The HPC team also utilized the R Programing Language, a language commonly used by statisticians as well as data miners, in order to analyze the results of each catalog. Varying the parameter sets produced drastic changes in the catalogs generated. With the help of the Probabilistic Forecasting Team, the HPC Team compared each of the short catalogs with multiple aspects of the UCERF3 data. Of the seventeen short catalogs generated, Sigma High had less than a 2% difference from UCERF3 in the recurrence interval of events M ≥ 7 on the Southern San Andreas Fault, which indicated that the parameter set was the best representation of an earth-like system. Based on this information, the team extended the catalog to 530,000 years in order to create a more comprehensive dataspace for earthquake forecasting. The UseIT interns used these catalogs to answer probabilistic questions posted in the Grand Challenge and generate simulator-based forecasts for the San Andreas Fault System.

Key Words
earthquakes, RSQSim, UCERF3, HPC, earthquake forecasting, models,

Citation
Fu, Z., Bent, M. T., Lopez, H. M., Ortega, S. P., & Scroggins, K. C. (2016, 09). 2016 SCEC Undergraduate Studies in Earthquake Information Technology (UseIT): Earthquake Forecasting Through Physics-Based Simulations. Poster Presentation at 2016 SCEC Annual Meeting.


Related Projects & Working Groups
Communication, Education, and Outreach (CEO)