SCEC Award Number 19154 View PDF
Proposal Category Collaborative Proposal (Integration and Theory)
Proposal Title Understanding the Seismic Ground Motion Spatial Variability Using Network Analysis - Community Detection
Investigator(s)
Name Organization
Qingkai Kong University of California, Berkeley Gregory Beroza Stanford University
Other Participants 1 graduate student
SCEC Priorities 4a, 4c, 4d SCEC Groups Seismology, GM, CS
Report Due Date 04/30/2020 Date Report Submitted 03/14/2022
Project Abstract
 This project is to explore ground motion spatial distribution using a new approach - network analysis. In this method, each seismic station will be modeled as a node and the similarities of the waveform between two stations are modeled as edges. Using community detection, we can group the stations into groups that have similar waveforms. Currently, we are testing this method on 52 events recorded on San Jacinto Seismic Nodal Array. The initial results are compared with the current engineering way to study the ground motion spatial distribution, and we are trying to understand the detected communities.
Intellectual Merit This project uses a new approach (graph-based method) to study the spatial ground motion correlations and directly investigates some of the factors that control the spatial variations of the ground motions. The results have the potential contributions to the SCEC priorities 4a, 4b and 4c.
Broader Impacts This project has a geophysics Ph.D. student from Stanford University to work on the analysis, and help to train future researchers. The results are also presented at conferences like AGU and published in peer-reviewed journals to increase the impact.
Exemplary Figure Figure 3. Waveform similarity analysis results from three groups of earthquakes. Panel (a) shows the selected similarity edges projected onto the geophone array from earthquakes in Group S. Panels (b) and (c) show the same results from earthquakes in Group N and Group W, respectively. Panels (d), (e) and (f) present the community detection results for Group S, N and W. Each community is colour-coded for clarity. Communities with less than 10 nodes are excluded for visualization purposes.

Credit: from figure 3 - https://academic.oup.com/gji/article/225/3/1704/6134073