We use energy flux conservation of one tube of rays in conjunction with first order diffraction theory to compute Fréchet kernel (or sensitivity kernel) for finite-frequency wave traveltime in 2D acoustic media. Destructive interference among adjacent broad-band pulse renders the finite-frequency wave traveltime delay sensitive only to the velocity structure in a banana-shaped region surrounding the unperturbed geometrical ray, in which case fast computation of Fréchet kernel with paraxial approximation is valid. In this abstract, we investigate slope tomography assisted by finite-frequency sensitivity kernel. Slope tomography here referring to stereotomography and normal incident point (NIP) wave tomography basing on ray perturbation theory. A 2D numerical test of NIP-wave tomography assisted by finite-frequency sensitivity kernel is presented to validate the proposed method.
In petrophysical-properties inversion, elastic parameters underground can be obtained by seismic inversion. Petrophysical-properties determine elastic properties and this relation can be depicted by rock-physics model. Since petrophysical-properties estimation is always associated with uncertainties, stochastic inversion plays an important role in petrophysical-properties estimation. The stochastic sampling method provides us a tool for predicting rock and fluid properties from probability distributions function (PDF) of petrophysical-parameters inverted based on rock-physics model. In this paper, we combine geostatistics and a new statistical rock-physics inversion method under Bayesian framework to compute the PDF of petrophysical properties. Then, a Markov-Chain-Monte-Carlo (MCMC) algorithm is used to sample posterior PDF to get multiple realizations thus uncertainty of both petrophysical-properties and elastic parameters inverted can be evaluated. The synthetic data example and real application to seismic data near ODP1144 demonstrated the effectiveness of presented method.
The CRS-OIS is a common reflection surface (CRS) stack method with an output imaging scheme. Differed from conventional 3D CRS stack, 3D CRS-OIS operator is a double-smearing operator based on 3D local coherent events searched in common-offset section. The so-called double-smearing means, for any a Ph in CO section, firstly smearing Ph along its 3D local coherent events in 3D CO volume; secondly, for each sample being smeared out from Ph, smearing it again along its own 3D NMO/DMO responses. The outplanat in the first step is a 3D local coherent events searched in CO section and the outplanat in the second step is their 3D NMO/DMO responses. If we just perform the first step of 3D CRS-OIS, a superior 3D pre-stack dataset with a higher signal/noise (S/N) ratio and a better regularity can be generated as a by-product of 3D CRS-OIS. The synthetic and real data example demonstrates that 3D CRS-OIS can be used as a robust tool for noise suppression and data interpolation.
The time-frequency domain divergence and absorption compensation (TFCOMP) is an algorithm to compensate the differences of spherical divergence and absorption among different shots in time-frequency domain. This paper presents a practical land data preconditioning scheme for AVO inversion based on TFCOMP algorithm. Firstly, the selection of pilot shot is not arbitrary but depends on if it is near enough with a well-drilling location. Thus the velocity, density and attention parameters (Q) for the pilot shot can be estimated from logging data and VSP data with enough accuracy. Secondly, visco-acoustic plane wave equation is used to compensate the spherical divergence and attenuation of the pilot shot along relevant ray paths. Thirdly, the trend of divergence and attenuation of all nonpilot shots are adjusted as the same with the trend of pilot shot in time-frequency domain. Since the compensation of pilot shot is based on accurate velocity, density and Q information, the data after TFCOMP processing will not only show a good consistency in wavelet, amplitude and frequency but also preserve its AVO characteristics correctly. The above scheme can be seen as a practical land data preconditioning scheme for AVO inversion
The common reflection surface stack with the output imaging scheme (CRS-OIS) is an important revision for conventional CRS stack method. It can provide both a superior zero-offset (ZO) image and a better pre-stack dataset with higher signal/noise ratio (SNR) simultaneously, which expands the applicability of ZO CRS imaging theory. The 3D-CRS-OIS reduces the computational cost dramatically compared with conventional 3D CRS stack. However, it is still a computationally intensive task when dealing with real 3D seismic data. In this paper, considering the advantages of graphic processing unit (GPU) platform in the aspects of high memory bandwidth, multi-registers and multi-processors, we presented a GPU-based 3D-CRS-OIS algorithm. The synthetic and field data examples show that the proposed method is robust and much more efficient than CPU-based algorithm in dealing with 3D real data especially when a large CRS-stack aperture is used.
The accuracy of near-surface velocity construction always plays an important role when dealing with overthrust data. Under the assumption of equivalent layered medium, we combine micro well-log information and first-arrival of surface seismic survey as prior information to estimate near-surface velocity structure based on Monte Carlo optimization scheme. The simulated annealing algorithm is taken to search the global optimum. Compared with conventional least-square (LS) based first arrival tomography, a clearer base of low-velocity-layer can be delineated thus it will surely be helpful to either wave-equation datuming or statics correction. The numerical tests on two synthetic and one real data examples demonstrated the effectiveness of this scheme.