Joint inversion of passive seismic datasets for sedimentary basin structures
Imaging sedimentary structures is of great importance to evaluate energy resources and seismic hazards. In petroleum industries, borehole drilling and seismic exploration are the most reliable methods to obtain sedimentary structures. However, due to high cost in field works, those methods are usually only carried out in the reservoir scale (< 25 km×25 km) and it is too expensive to perform a 3D seismic survey on a regional scale. In passive seismology, although basin structure is important for understanding the regional geological or tectonic evolution, basin structures are rarely imaged as they are too thin to be resolved by low frequency signals. However, compared with active seismic data, passive seismic data are more affordable and convenient to acquire. Hence, in this thesis, I develop a joint Bayesian Monte Carlo nonlinear inversion method that jointly inverts the complementary nature of Rayleigh wave phase velocity, Rayleigh wave particle motion and teleseismic P waveform and its coda for sedimentary structures.
In my inversion scheme, Rayleigh wave phase velocity is extracted from ambient noise cross-correlation functions (NCFs). At the present day, the time-frequency domain phase weighted stacking (tf-PWS) technique based on the S transform has been widely employed in stacking empirical Green’s functions (EGFs) derived from ambient noise data, mainly due to its superior power in enhancing signal-to-noise ratios (SNRs) of surface wave signals. However, questions such as if the non-linear method induces waveform distortion and how efficient the non-linear stacking method is, are yet to be thoroughly explored. In this thesis, I investigate these issues by analyzing the mathematical derivation of this method and also conducting extensive numerical tests with both synthetic data and field noise data. I find that the bias of the measured phase velocity associated with waveform distortion caused by the tf-PWS adopting a frequencyinverse S transform method is less than 0.1%, so small enough that the measured phase velocities can be safely used in surface wave tomography. I also find that tf-PWS implemented with a time inversion S transform tends to cause large waveform distortion as the discrepancy can reach as much as ~0.4% at periods longer than 60 s. Therefore, if tf-PWS is used in stacking daily NCFs, then frequency IST is preferred to transform the stacked S spectra back to the time domain of the stacked EGFs.
Then, I apply the tf-PWS and joint inversion techniques to the Wombat array deployed in southeast Australia and construct a 3D model of the shallow and middle crust. To verify the accuracy of the inverted model, sedimentary depths from four boreholes are used for comparisons. On average, the differences between the inverted sedimentary depths and the ground-truth depths are ~198 m. Because the locations of the four boreholes do not exactly overlap with the seismic stations, the differences are reasonable. According to the model constructed by the joint inversion, the average sedimentary depth in the Murray basin is only 713 m, while the deepest sedimentary depth in the Sydney Basin is ~3 km. Besides, we also find three low-velocity zones in the middle crust of southeast Australia. Among them, two are related to the neo-tectonic active region, and one is related to the Begargo Hill volcanic zone.
In a nutshell, my method is a big step forward in better imaging the shallow structures of sedimentary basins. And it can be readily applied to better imaging other basins.