Determination of Rupture Propagation for Large Earthquakes from Back-Projection Analyses Using Large Arrays

Dun Wang

With the deployment of large regional seismic arrays, such as Hi-net in Japan and the USArray in the USA, seismologists have begun to look at rupture front propagation of large earthquakes (e.g. Ishii et al. [2005] and Kruger and Ohrnberger [2005]) using back-projection methods. This approach becomes increasingly popular and is now widely used to track the rupture process of moderate to large earthquakes. This thesis mainly focuses on the rupture speed and extent of recent large earthquakes using array methods such as back-projection and envelope inversion based on empirical Green functions.

I have also made improvements to the back-projection method which can empirically correct for the effects of the array smear. This enables more accurate estimate of the rupture area and rupture pattern. I applied this new method for the case of the 2012 Mw 8.6 Sumatra earthquake, in which four conjugate faults are clearly identified. Compared to regular finite slip inversions and past back-projection results, this new method gives much better estimate of the rupture propagation for complicated earthquakes.

I tested the location uncertainties for the back-projection method considering heterogeneities in source region, frequency bands, directivity effect, depth phases, location uncertainty of epicenter, and stacking window length. The results indicate a location uncertainty of 20 to 40 km for Hi-net and European data for the case of the 2012 Mw 8.6 Sumatra earthquake, with the main errors coming from the structure heterogeneities around the source area. Depth phases in the waveforms, cause offsets of about 10 km around the source with significantly smaller amplitudes, for the results derived from Hi-net data.

I also applied the back-projection methods to trace the rupture propagation for a number of recent large earthquakes where array data are available. Mainly Hi-net, Japan, USArray, and European array are used in these studies. Estimates of rupture speeds for earthquakes with mode II (strike-slip) and III (thrust and normal) ruptures suggest that earthquakes with faster than local S wave rupture speed always have mode II faulting mechanism. Natural supershear earthquakes have been previously reported but the list is still very short, with 3 earthquakes. I have identified two more supershear earthquakes- the 2012 Mw 8.6 Sumatra earthquake and the 2010 Mw 6.9 Qinghai earthquake. The analyses for earthquakes that occurred in shallow region of subduction zones indicates slow rupture speeds of up to 1.5 km/s, which is an important feature for these tsunami earthquakes. I also show a strong frequency dependence of the rupture pattern for the 2010 Maule, Chile and 2011 Tohoku, Japan earthquakes using back-projections of seismograms filtered in different frequency bands.