Earthquake source images from high frequency waveforms – application and uncertainty analysis
About the Event:
In the last decade, back-projecting high-frequency (HF) seismic waveforms have been evidenced of great value to image HF radiation and the spatial and temporal evolution of great earthquakes. However, so far few comprehensive error analyses for back-projection (BP) methods have been conducted, although it is evident that HF seismic waves are strongly affected by source depth, focal mechanisms and the Earth’s 3D velocity structures. Here we perform 1D and 3D synthetic tests in which we know the exact source information to investigate the impacts of these factors on the uncertainties in BP results, by using two high resolution BP methods, MUltiple SIgnal Classification (MUSIC) [Meng et al., 2011] and Compressive Sensing (CS) [Yao et al., 2011]. We generate synthetic data for virtual sources with different depths, focal mechanisms by embedding them in the 1D and 3D velocity models. We then back-project the synthetics using MUSIC or CS based on the array configurations. Our synthetic tests for source depth show that the depth phases can be back-projected as pseudo-sources swimming towards the array. For instance, when the source is placed at a depth of 10km, BP images the depth phases as strong signals at ~8km away from the true location. Such bias increases with depth, e.g., the error could be about 20km for a depth of 30km. For complex rupture process that involves different focal mechanisms, the seismograms display significant discrepancies in different arrays. Using the 2016 Mw7.8 Kaikoura earthquake as a scenario, we show that the differences in BP results from South America array and Australian array are primarily due to the change of focal mechanism at the final stage of the rupture, which largely explains the results derived from the real data. Finally, we also test the impact of 3D velocity structures by generating synthetic data for an earthquake in Java subduction zone at shallow depth. We show that the strong and long-lasting coda waves due to 3D near trench structure can be mirrored as artificial sources far from the true source. In summary, our analyses indicate that the impact of various factors, such as those tested in this study, should be considered when interpreting any detailed BP image to infer the earthquake rupture kinematics and dynamics.
About the Speaker:
Hongyu received his Bachelor and Master degree in Nanjing University, China, in 2013 and 2016, respectively, and now is a Research Associate at EOS working in Shengji’s group. His research interests focus on the earthquake source study using high-frequency seismic waveforms. Also, he is working on 3D waveform simulation.