Network Analysis to Characterize Seismic Ground Motion Spatial Variability: A Case study on the San Jacinto Seismic Array

Yixiao Sheng, Qingkai Kong, & Gregory C. Beroza

Accepted December 11, 2019, SCEC Contribution #11807

Spatial variability of seismic ground motion is important in earthquake hazard analysis. It has particularly strong impacts on large structures, because during an earthquake, different parts of the foundations may experience different levels of shaking. It also has a strong influence on the spatial distribution of damage. A better understanding of such spatial correlation would help the guide policy to quantify and mitigate the risks posed by earthquakes.

Previous studies have computed the spatial correlation of peak ground acceleration (PGA) caused by large earthquakes. However, the ground motion in those studies was only sparsely sampled and the inter-station distance was considered as the only variable. These limitations impose difficulties on constraining important aspects of the spatial correlation, such as earthquake source effect, wave propagating path effect as well as site amplification effect.
In our study, we combine network analysis and large-N seismic array to study such spatial variability. We treat each station as a node and the whole network as a graph while the link between each pair of nodes is weighted by the similarity between the recorded seismograms. Communities, within which the nodes are strongly correlated, are determined from the graph using community detection algorithms. Community detection allows us to consider the connections beyond individual station pairs, extending to multiple stations that are distributed over wide areas. For a given earthquake, by applying this algorithm, seismic stations are grouped into different clusters while in each cluster, stations record resembling waveforms.

We test the method on the ZG dense array deployed across the San Jacinto fault. ZG array has about 1000 geophones with 10-meter inter-station spacing. We performed our algorithm on 65 local events listed in the Southern California earthquake catalog. The station clustering results match well with the local structure. We further group the events based on the station clustering patterns. The outcome shows that source location is critical in determining these observed spatial patterns, suggesting a major influence from the path effect. Though distinct from previous studies based on intensity measure, our waveform-similarity-based results also show consistency with those given by PGA analysis.

Sheng, Y., Kong, Q., & Beroza, G. C. (2019, 12). Network Analysis to Characterize Seismic Ground Motion Spatial Variability: A Case study on the San Jacinto Seismic Array. Oral Presentation at American Geophysical Union.