Soil properties from seismic intrinsic dispersion

Theoretical and experimental studies in the past have shown the sensitivity of seismic waves to soil/rock properties, such as composition, porosity, pore fluid, and permeability. However, quantitative characterization of these properties has remained challenging. In case of unconsolidated soils, the inherently loose and heterogeneous nature complicates the task of obtaining the in situ properties and spatial variations. In this thesis, we investigate the possibility of exploiting the information of seismic intrinsic dispersion in the low frequency band (10-200 Hz), which is relevant to onshore field data, in order to quantify these physical properties, with special focus on soil porosity and permeability. In situ values of these properties are crucial in many different projects. We first investigate the frequency-dependent seismic velocity and attenuation caused by inelastic losses at grain contacts and wave-induced fluid flow at different scales (from grain size to seismic wavelength), using the theory of poroelasticity first proposed by Biot and many subsequent extensions and modifications. Several pertinent models of poroelasticity are looked at in order to find out their applicability in explaining the observed seismic dispersion. The observed dispersion can vary greatly between various unconsolidated, fully-saturated soils. Further, we develop a stress-dependent Biot (SDB) model in order to study the behaviour of seismic waves propagating through a fully-saturated porous medium subjected to different stress conditions. This is achieved by combining the mechanics of granular soils with the effective-stress laws, finally coupling with Biot's theory. Careful analyses of the underlying soil/rock physics that relate geophysical observations to the physical properties reveal an interesting feature in the property domain among several different measurements. This is an extention to some recent work done by others. We have found that it is possible to find two or more measured quantities, showing contrasting (sometimes quasi-orthogonal) behaviour in the common parameter space, such that a combination of those measured quantities leads to a physics-based uniqueness in the property estimation. This quasi-orthogonality in the common property domain among different measured quantities is advantageously used for estimation of porosity, permeability, water saturation, and effective stress. Several numerical examples are presented where P- and S-wave velocity and attenuation are efficiently integrated in order to obtain soil properties. In addition to seismic waves, electromagnetic waves are briefly considered for extracting extra soil properties. In this research, considerable attention has been paid to the investigation of S waves travelling through a porous medium, since S waves have well-known significance in the context of shallow subsurface characterization. Twelve selected datasets of frequency-dependent S-wave velocity and attenuation from various soft-soil sites are used in this study. Data for fully-saturated, unconsolidated soils from land/onshore environment are only considered. It is found that the behaviour of seismic intrinsic dispersion can vary greatly with the soil-type. One of the main challenges in property estimation using intrinsic dispersion relates to reliable extraction of the information of intrinsic dispersion from the recorded seismic data. The difficulty lies in the quantification of scattering attenuation, the effect of which is always present in the recorded seismograms due to the wavelength-scale and smaller heterogeneities in the subsurface. Scattering has an absorption-like effect on the transmitted seismic energy. Accordingly, determining and subtracting the scattering attenuation from the total (or apparent) attenuation is critically important. We have discussed and successfully tested an approach, achieving this goal. Several shallow vertical seismic profiling (VSP) measurements are conducted in the field using a recently developed digital, array-seismic cone penetrometer (CPT) system. CPT provides information on cone-tip resistance, sleeve friction, and pore pressure, thus offering direct, additional knowledge on geological layering, that is used to calculate the scattering attenuation. To obtain the soil properties, an inversion algorithm is presented based on simulated annealing and the poroelasticity theory. We study the sensitivity of different parameters involved in the cost function to be minimized. The most and the least sensitive parameters are discriminated based on the eigenanalyses of the covariance matrix of the gradient of the cost function. The eigenvectors and the corresponding eigenvalues of the covariance matrix are used to navigate efficiently the search algorithm in the multidimensional space and find a relatively stable, global solution of the cost function. Finally, we apply the methodology developed in this research to a VSP dataset acquired in a layered sequence of siltstone, shale and sandstone. The porous sandstone contains hydrocarbon accumulations. The influence of fluid mobility (permeability-to-viscosity ratio) on the estimated P-wave intrinsic dispersion is distinctly observed. Using optimization by simulated annealing together with VSP and well-log measurements, the Biot and squirt flow (BISQ) model is found to provide one possible mechanism for the observed dispersion. The layer-specific fluid mobility values are estimated using our approach; they are found to be close to the independent measurements of mobility using Stoneley waves and from dynamic formation-tests carried out at the same borehole. The depth distribution of fluid mobility matches well between our estimate and the independent measurements. The methodology developed and the results obtained in this research pave the way to a new direction for in situ, quantitative soil/rock characterization using seismic waves