Inversion Of1-Hz GPS Data For Rupture Process Of The2011Great Northeast Japan Earthquake And The2008Wenchuan Earthquake

It has lasted several decades for using seismic waves to infer the nature of source byearthquake magnitude, inversion of seismic moment tensor and inversion of rupture process.Inversion for the rupture process is the basis topic of seismology. It’s not only helpful to knowthe dynamics character earthquake but also important in earthquake mitigation.With thedevelopment of GPS positioning techniques, Larson et al.(2003) demonstrated that1-Hz GPSdata could be analyzed with sufficient precision to measure seismic waves caused byearthquake. Then GPS data was not only used as the constraint for seismic data but also usedin the single inversion to infer the rupture process. Although there were few researches inrupture process using1-Hz GPS waveform data,1-Hz GPS waveform data has high spatialresolution, sensitive to the variation of displacement, not saturated in amplitude and so on. Italso supplied the new data type to infer the rupture process.We used the inversion method of Yoshida et al.(1996) to infer the rupture process of the2011Great Northeast Japan Earthquake using1-Hz GPS waveform data only, and the2008Wenchuan Earthquake using1-Hz GPS waveform data and strong motion records. TheGreen’s functions were computed by the frequency-wavenumber method developed by Zhuand Rivera (2002).In chapter one, we introduced the significance of research papers, and summarized theprogress in the inversion of rupture process, especially in using GPS data. The four aspects asfollows: prior information, relative weight ratio, optimization algorithm and stability ofinversion result have been discussed to state the possible problems in the inversion of ruptureprocess.In chapter two, we introduced the theories of the method of Yoshida et al.(1996) andinversion procedures in detail. We also deduced the computing formulas offrequency-wavenumber method developed by Zhu and Rivera (2002).In chapter three, we imposed the dislocation not reverse as a constraint to determine thevariation of the slip vectors with prior main slip direction. The inversion was also subject to asmoothness constraint with a discrete Laplacian in time and space, and the constraint formulaswere used in the inversion. The velocity structure model proposed by Wu et al.(2008) was used in calculating both dynamic and static displacement responses.We selected the faultgeometry in the inversion, based on the published fault models, distribution of aftershocks in24hours after initial rupture and strike/dip of Pacific plate (Hasegawa et al.,1994). We usedthree-component GPS ground motion time series to infer the rupture process of the2011Great Northeast Japan Earthquake by the method of Yoshida et al.(1996), with the optimalmodel selected by Akaike’s Bayesian Information Criterion(ABIC)(Akaike,1980).The inversion result showed the concentrations of slip upper the hypocenter to the trench,with maximum slip of~70m northeast to the hypocenter, yielding the estimate of the seismicmoment (3.8×1022Nm). The facts that the resulted main rupture region had the leastaftershocks, most of the fatalities caused by the earthquake was from tsunami damage, andmain rupture region of single inversion result of tsunami data was near the trench, supportedthe compact shallow slip distribution. We took snapshots of slip distribution every10s toshow the rupture character in time and space for the whole rupture process, including initialrupture (0-60s), main rupture (60-100s) and slip decreasing largely (100-120s) within120s.After the initial break, the rupture propagated outside from the hypocenter and generateda small compact slip region,30km northeast to the hypocenter. In this stage, the rupturepropagated to the trench (0-60s). Later, the main rupture concentrated near and upper thehypocenter along with large energy released and slip increasing, generating three largecompact slip region near the hypocenter and the trench (60-100s). At this time, the rupturebegan to propagate to the south and then the rupture reached the eastmost area at the end ofthis stage. The rupture was focus on the southern part of the fault. We discussed the selectionand influence of smoothing factor, fault model, rupture velocity and source time function, anddetermined the stability for various choices of these parameters. We also briefly discusseddifferent inversion results using the stations of different epicentral distance.In chapter four, we tried to do the simultaneous inversion with1-Hz GPS waveform dataand strong motion records to infer the slip distribution for the Wenchuan earthquake. Theinversion result showed that the rupture was concentrated on Yingxiu-Beichuan fault northto the hypocenter with small slip on north fault, and the maximum slip was about14m. Inconsideration both of the time windows only50s for eight1-Hz GPS stations and propagationtime from hypocenter to station, the available time windows was only30s. Due to thecomplex tectonic structure of the Longmen Shan region, it was difficult to choose the accurate fault model for the inversion of near-field datasets, sensitive to the fault model. We haven’tobtained the well source model to show the source process.