Andrey Bakulin

Building tilted transversely isotropic depth models using localized anisotropic tomography with well information

Tilted transverse isotropyTTI is increasingly recognized as a more geologically plausible description of anisotropy in sedimentary formations than vertical transverse isotropy VTI .A lthough model-building approaches for VTI media are well understood, similar approaches for TTI media are in their infancy, even when the symmetry-axis direction is assumed known. We describe a tomographic approach that builds localized anisotropic models by jointly inverting surface-seismic and well data. We present a synthetic data example of anisotropic tomography applied to a layered TTI model with a symmetry-axis tilt of 45 degrees. We demonstrate three scenarios for constraining the solution. In the first scenario, velocity along the symmetry axis is known and tomography inverts for Thomsen’s and parameters. In the second scenario, tomography inverts for, , and velocity, using surface-seismic data and vertical check-shot traveltimes. In contrast to the VTI case, both these inversions are nonunique. To combat nonuniqueness, in the third scenario, we supplement check-shot and seismic data with the profile from an offset well. This allows recovery of the correct profiles for velocity along the symmetry axis and. We conclude that TTI is more ambiguous than VTI for model building. Additional well data or rock-physics assumptions may be required to constrain the tomography and arrive at geologically plausible TTI models. Furthermore, we demonstrate that VTI models with atypical Thomsen parameters can also fit the same joint seismic and check-shot data set. In this case, although imaging with VTI models can focus the TTI data and match vertical event depths, it leads to substantial lateral mispositioning of the reflections.

Localized anisotropic tomography with well information in VTI media

We develop a concept of localized seismic grid tomography constrainedbywellinformationandapplyittobuildingvertically transversely isotropic VTI velocity models in depth. The goal is to use a highly automated migration velocity analysis to build anisotropic models that combine optimal image focusing withaccuratedepthpositioninginonestep.Welocalizetomographytoalimitedvolumearoundthewellandjointlyinvertthesurface seismic and well data. Well information is propagated into thelocalvolumebyusingthemethodofpreconditioning,whereby model updates are shaped to follow geologic layers with spatial smoothing constraints. We analyze our concept with a synthetic data example of anisotropic tomography applied to a 1D VTI model.We demonstrate four cases of introducing additional information. In the first case, vertical velocity is assumed to be known, and the tomography inverts only for Thomsen’s and profiles using surface seismic data alone. In the second case, tomography simultaneously inverts for all three VTI parameters, including vertical velocity, using a joint data set that consists of surface seismic data and vertical check-shot traveltimes. In the thirdandfourthcases,sparsedepthmarkersandwalkawayverticalseismicprofilingVSPareused,respectively,tosupplement the seismic data. For all four examples, tomography reliably recovers the anisotropic velocity field up to a vertical resolution comparable to that of the well data. Even though walkawayVSP hastheadditionaldimensionofangleoroffset,itoffersnofurther increase in this resolution limit. Anisotropic tomography with well constraints has multiple advantages over other approaches anddeservesaplaceintheportfolioofmodel-buildingtools.

Shallow virtual source redatuming by multi-dimensional deconvolution

Recently, new on-shore acquisition designs have been presented with multicomponent sensors deployed in the shallow sub-surface (20 m–60 m). Virtual source redatuming has been proposed for these data to compensate for surface statics and to enhance survey repeatability. In this paper, we investigate the feasibility of replacing the correlation-based formalism that undergirds virtual source redatuming with multi-dimensional deconvolution, offering various advantages such as the elimination of free-surface multiples and the potential to improve virtual source repeatability. To allow for data-driven calibration of the sensors and to improve robustness in cases with poor sensor spacing in the shallow sub-surface (resulting in a relatively high wavenumber content), we propose a new workflow for this configuration. We assume a dense source sampling and target signals that arrive at near-vertical propagation angles. First, the data are preconditioned by applying synthetic-aperture-source filters in the common receiver domain. Virtual source redatuming is carried out for the multi-component recordings individually, followed by an intermediate deconvolution step. After this specific pre-processing, we show that the downgoing and upgoing constituents of the wavefields can be separated without knowledge of the medium parameters, the source wavelet, or sensor characteristics. As a final step, free-surface multiples can be eliminated by multi-dimensional deconvolution of the upgoing fields with the downgoing fields.

Amplitude Radiation Pattern of a Virtual Source - Estimation and Correction

In the Virtual Source (VS) method we cross-correlate recordings at two receiver locations to create data as if one of these receivers is a Virtual Source and the other is a receiver. We study the amplitude radiation pattern of Virtual Sources. This pattern can be estimated by autocorrelation of the spatial Fourier transform of the downgoing wave field that is used for the VS creation in the special case of a laterally invariant medium. The generated VS data can be improved by deconvolution with the estimated amplitude radiation pattern in the FK-domain. The methodology is tested on a 1D elastic model, where it is shown that almost perfect amplitude retrieval is possible within a limited aperture of VS radiation. In general heterogeneous media the spatial Fourier transforms are not laterally invariant and the VS amplitude radiation pattern has to be estimated in a different way. We argue that Wigner distribution functions can serve this purpose. We show how these functions can be used for diagnosis of the spatial distributions of the wave fields that are used for VS creation, with a VS synthetic data example generated by the Peace River 2D elastic model.

A two-phase automatic static correction method

ABSTRACT Statics are an effective approach to correct for complex velocity variations in the near surface, but so far, to a large extent, a general and robust automatic static correction method is still lacking. In this paper, we propose a novel two‐phase automatic static correction method, which is capable of handling both primary wave statics (PP statics) and converted‐wave statics (S‐wave statics). Our method is purely data driven, and it aims at maximizing stacking power in the target zone of the stack image. Low‐frequency components of the data are analysed first using an advanced genetic algorithm to estimate seed statics and the time structure for an event of interest, and then the original full‐band data are further aligned via the back‐and‐forth coordinate descent method using the seed statics as initial values and the time structure for event alignment guidance. We apply our new method to two field datasets, i.e., one for 2D PP static correction and the other for 3D S‐wave static correction.

Improving repeatability of land seismic data using virtual source approach based on multidimensional deconvolution

We present a new redatuming workflow developed for improving the repeatability of seismic data and designed specifically to account for changes in the source signatures or variations in downgoing fields in general. The new approach is based on the virtual source method with the same potential for reducing nonrepeatability, associated with acquisition geometry changes and variations in the near surface. To correct for changes in the source wavelet between surveys, we suggest deconvolving the virtual source gather of the monitor survey with the point-spread function (PSF) of the same survey, and immediately convolving with the PSF of the base or reference survey. The PSF governs the radiation pattern of the virtual source. Trying to completely deconvolve the effects of individual PSFs on each virtual source response may degrade repeatability due to possible amplification of noise. Instead, we try to equalize radiation patterns of the virtual sources across all repeat surveys by reassigning a new reference PSF to all surveys. We apply the deconvolution-convolution method to a field 4D dataset with buried receivers and demonstrate significant improvement in repeatability.