Studies On Frequency-dependent Coseismic Radiation During Large Earthquakes

Since the 21st century, the global surge of large earthquakes and the improvement of the seismological observations provide us an unprecedented opportunity to study the rupture process of large earthquakes. Previous studies on the megathrust earthquakes in subduction zones present a systematic frequency-dependent/depth-vary ing pattern during the coseismic process. This frequency-depth pattern greatly helps us to understand the physical properties on the surface of the subducting slab. However, there are still some unclear questions concerned with this observation, such as:does this frequency-dependence only exist in subduction zones or can it also be observed during the coseismic process of other kinds of earthquakes? What is the physical mechanism leading to the depth-varying distribution of coseismic radiation at different frequencies? What further information can this kind of observation provide us to understand the rupture process of large earthquakes? In order to understand these questions, we apply the compressive sensing (CS) method to study 3 representative large earthquakes in the recent 5 years.First, we improve the conventional CS method for better applying it to the studies of great earthquakes. (1) The spatial resolution and computational efficiency of the conventional CS method are limited by the grid size in the source region. We develop an auto-grid refinement CS method and with this method we can get the coseismic radiation distribution from CS with very high spatial resolution at the cost of much less computational time than the conventional method. (2) In our problem, the aligned teleseismic P wave will, especially from the late stage of the large earthquake rupture, be sloping and this sloping effect will introduce systematic data errors into our inversion due to the use of the sliding window in order to achieve time resolution. To suppress this effect, we apply a sliding window shifting technique during the inversion. To test the reliability of our improvements, we design a series of synthetic tests to validate the proposed method and its resolution.Then with the improved method we study 3 different kinds of earthquakes in the recent 5 years for their spatial and temporal distribution of coseismic radiation. The first one is the 2012 Sumatra Mw8.6 great strike-slip earthquake, which is the largest strike-slip event ever recorded. Our results indicate that this event is a very complex one with multistage rupture processes and this event is closely related to the complex regional conjugate fractures. The second one is the 2015 Nepal Mw7.8 earthquake. It starts from the epicenter and propagates southeastward along the Main Himalaya Belt. The third one is the 2015 Chile Mw8.3 event. This earthquake is a typical megathrust event in the subduction zone. Its rupture propagates mainly toward the shallow part of the subduction zone. Our coseismic radiation distribution results of these large earthquakes from the CS method present clear frequency-dependence in all these events.To better understand the physical mechanism of this frequency-dependence, we compare our coseismic radiation distribution with the coseismic slip distribution from finite fault inversion as well as the coseismic static shear stress change, which is calculated from the slip models. These comparisons indicate that the low frequency coseismic radiation corresponds to the high slip patches and negative shear stress changes (stress-releasing) on the fault. The low frequency radiation is systematically located in the updip part and probably related to the large scale asperity. However, the high frequency radiation is consistent with the boundary of large coseismic slip region and positive shear stress changes (loading). These regions have been loaded during the earthquake and the high frequency radiation is possibly due to the breaking of the smaller scale asperity. Our results are consistent with previous ones of subduction zone earthquakes and yet make significant contributions in understanding such frequency-dependent coseismic radiation. We propose that the frequency-dependence of coseismic radiation is not only for certain type of earthquakes or in certain regions, but is a common phenomenon due to variation of physical properties of the faults. Our method can greatly help to better understand the rupture dynamics as well as the mechanism of large earthquakes.