High-frequency Surface-wave Method For 2D Inhomogeneous Materials

Surface waves, commonly known as either Love waves or Rayleigh waves, are characterized by relatively low velocity, low frequency, and high amplitude. Rayleigh waves, which are the result of interfering P and SV waves, have both longitudinal (movement parallel to the direction of travel) and transverse motion (movement is perpendicular to the direction of travel). Love waves, which are formed from the constructive interference of multiple reflections of SH waves, only have transverse motion. In near-surface seismic shot gathers, surface waves dominate the energy of near-surface wavefield, which contains abundant information about near-surface shear (S)-wave velocity. By analyzing high-frequency surface-wave signals, high-frequency surface-wave methods have been widely used to estimate S-wave velocity of near-surface materials.The Multichannel Analysis of Surface Waves (here the surface wave refers to Rayleigh waves, MASW) and Multichannel Analysis of Love Waves (MALW) are the most widely used high-frequency surface-wave methods. They are getting increasingly attention in near-surface geophysics and geotechnique communities for its non-invasive, high precision, and high efficiency. MASW and MALW share similar data processing and inversion steps, including:acquisition of multichannel Rayleigh-(or Love-) wave field data; transforming field data into the frequency-velocity (f-v) domain and pick its phase velocities (dispersion curve) along the continuous energy peaks; inverting picked phase velocities to estimate 1D S-wave velocities; and finally get a pseudo 2D S-wave velocity image (if needed) by repeating previous steps. This thesis mainly focuses on improving high-frequency surface-wave method for delineating near-surface 2D inhomogeneous structures.One of the key steps in MASW and MALW is to calculate correct theoretical dispersion curves of an earth model. When the earth model contains a low-velocity layer (LVL), however, calculation of dispersion curves by existing algorithms may fail. If the half-space is an LVL, its dispersion equation turns out to be a complex matrix at some frequencies, which means no real root (phase velocities) can be obtained in these frequencies. If there is an LVL within a layered model and its S-wave velocity is lower than that of the surface layer, the calculated dispersion curve approaches the S-wave velocity of the LVL at a high-frequency range, rather than a value related to the surface layer. However, according to numerical modeling results based on wave equation, trends of the Rayleigh-wave dispersive energy approach about a 91% of the S-wave velocity of the surface layer at a high-frequency range when Rayleigh-wave wavelength is much short than the thickness of the surface layer, and cannot be fitted by the dispersion curve calculated by existing algorithms. In the first situation, we proposed to use real parts of dispersion equation to calculate Rayleigh-wave phase velocities. In the second situation, we proposed a method to calculate Rayleigh-wave phase velocities by considering its penetration depth. We built a substituted model that only contains the LVL and layers above it. We used the substituted model to calculate phase velocities when the Rayleigh-wave wavelength is not long enough to penetrate this LVL. Several synthetic models were used to verify the fitness between the dispersion curves calculated by our proposed methods and the trends of highest dispersive energy modeled by numerical method. A synthetic inversion test demonstrated high stability of using our method as the forward calculation method during the inversion.Because Love waves are only formed by SH waves and is independent of P-wave velocity, Love-wave dispersion curves are simpler than Rayleigh waves, which make MALW an appealing way to estimate near-surface S-wave velocity. A transverse source is usually used to generate Love waves in MALW. The using of a transverse source, however, is more difficult compared to a vertical impacting source, especially when encountering tough ground. It is regarded as one of the limitations to applying MALW. We proposed to use a vertical impacting source to generate Love waves in MALW. We performed four groups of field experiments in which both a transverse source and a vertical impacting source are used. By comparing the waveform, amplitude, and dispersion energy of the data generated by two kinds of sources, we found that Love waves generated with a vertical impacting source are of similar quality and energy compared to the one generated with a transverse source. It demonstrated that vertical impacting sources can also generate Love waves, which provides an alternative source type to MALW and increases the applicability of MALW.Based on 2D geometry spread in which source and receivers are placed along a same line, current MALW fails to work in three-dimensional (3D) seismic acquisition system. This is because that Love-wave particle motion direction is perpendicular to its propagation direction, which makes it difficult to record Love-wave signal in 3D geometries. We proposed a method to perform MALW in 3D geometry (MALW3D). We recorded two orthogonal horizontal components (in-line and cross-line components) at each receiver point at the same time. By rotating and composing those two components, we recovered Love-wave data at each receiver point. In order to achieve Love-wave dispersion curve, the recovered Love-wave data were transformed into a conventional receiver offset domain, and then transformed into the f-v domain. A synthetic model and a real-world example verified the validity of our proposed method.MASW and MALW are based on inversions of dispersion curves, and only suitable to laterally homogeneous or smoothly laterally varying heterogeneous earth models due to the layered-model assumption involving in the calculation of dispersion curves. It is regarded as a limitation of current high-frequency surface-wave method. Another problem that current high-frequency surface-wave method faces is the picking of correct modal dispersion curves, especially in the cases of complicated structures such as encountering low-velocity embedded layer and rough surface topography. Waveform inversion directly fits the waveform of observed data, and can be applied to most earth models. Besides, waveform inversion does not depend on the picking of dispersion curves, which makes waveform inversion an appealing way to overcome the limitations of current dispersion-analysis-based high-frequency surface-wave methods.We proposed to use surface-wave waveform inversion in the time domain to estimate near-surface S-wave velocity. Before performing the inversion, the model was divided into several blocks, the sizes of which increase with depth. The block sizes were determined by the resolution of surface waves, which is tested by checkerboard models. A finite-difference method was used as the waveform forward modeling algorithm. We updated the S-wave (and P-wave) velocity of each block via a conjugate gradient algorithm to minimize the misfit function.One of the most important parts in surface-wave waveform inversion is how to handle the source wavelet effect. We proposed two ways to deal with it. The first way is to use Virtual Real Source algorithm to obtain the source wavelet from two (or multi) shot gathers with a same geophone spread. Another way is to remove the source wavelet from observed waveform by deconvolution technique, which makes our method independent of source wavelet. We used several synthetic models to test the effectiveness of our methods. Some real-world cases were shown to verify the validity of using our method.