This award will fund continued development of methods to process huge volumes of seismic waveform data. This will lead to a great increase in the number of located and characterized seismic events of various kinds and will potentially identify patterns in earthquake occurrence that could inform hazard and near-term rupture forecasting. The initial "Mining Seismic Wavefields" NSF Geoinformatics grant has led to significant progress dealing with data volumes that would have been impossible to process when the project began. This award will fund an additional year of that effort and will maintain this momentum in technique development to complete the analysis of proof-of-concept projects on vast waveform data sets, and to deploy the cyberinfrastructure for wider use by the seismological community.
The premise of the research is that continuous and/or densely recorded data coupled with high performance computing and scalable algorithms can enable a network-based approach to earthquake detection that greatly improves the detection of weak and unusual events that would be difficult or impossible to detect using traditional approaches. Numerous seismological observations confirm that proximal earthquake sources generate similar signals. Exploiting the discriminative power of this similarity has led to many fundamental discoveries; however, most similarity-based detection methods require prior knowledge of the source waveform, or template. Blind/uninformed search for signals having unknown signatures based on pair-wise or multiple matches has seen some success, but naïve implementations of this approach suffer from quadratic scaling of computation with time such that problems of interest are inaccessible even for the most capable computers. Similarly, for dense networks, the availability of continuous waveform data motivates alternative detection schemes based on waveform similarity at adjacent stations. This project will further develop efficient data-mining techniques to enable scalable similarity search of seismic wavefields. Technical challenges to be addressed as part of the research for spatially sparse recording are to develop improved similarity-preserving compression for repeating signals detected over a network, and to improve post-processing of search output that will both isolate signals of seismological interest and minimize false detections. For spatially dense recording, this would extend recently developed wavefield matching techniques to similarity across adjacent stations, which would enable similarity search across unaliased elastic wavefields in four dimensions.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
