The amplitude-variation-with-angle (AVA) inversion for seismic data has been widely used in hydrocarbon detection of exploration seismology. Traditional AVA analysis estimates the elastic parameters, i.e., P-wave velocity, S-wave velocity and density, using the AVA intercept, gradient, and higher-order AVA terms by solving either a linear or nonlinear inverse problem. Recently, the deep learning has been introduced to the AVA inversion to describe the complicated nonlinear relation between seismic data and elastic parameters. But because of sparse well locations, the application of deep-learning based AVA inversion is limited by few well-log label sets in production. To mitigate this problem, we present a convolution neural network (CNN) trained by pseudo-well data for AVA inversion. By considering spatial correlation and cross-correlation of elastic parameters, we first generate a large number of realistic pseudo-well logs. Then, the angle-domain common-image gathers are computed as source wavelet convolved with reflectivity series, and then are used to train a CNN to predict elastic parameters. Numerical tests for both synthetic and field data demonstrate that the pseudo-well based CNN AVA inversion not only can accurately and efficiently estimate P-wave velocity, S-wave velocity and density, but also has a potential to alleviate the inter-parameter crosstalk artifacts, in comparison with traditional linear and nonlinear AVA inversion methods.

A detailed introduction can be downloaded and viewed as a pdf, or click this link.

We present a viscoacoustic modeling method to study seismic wave phenomena in the Loess Plateau. A viscoacoustic wave equation is first derived based on a non-linear optimization for the frequency-independent Q effect within a frequency band. To conform to the topographic surface, a vertically deformed grid is adopted to transform the irregular domain to a regular computational coordinate. This strategy does not introduce as many additional partial derivatives as the curvilinear grid method that deforms along three axes. In addition, to accurately compute the spatial derivatives, we apply a fully staggered-grid finite-difference scheme to numerically solve the viscoacoustic wave equation on a vertical deformed grid. Numerical experiments for 3D LoessPlateau models demonstrate the proposed method can accurately simulate wave propagation in the Loess Plateau with thick loess layers.

A detailed introduction can be downloaded and viewed as a pdf, or click this link.

Denoising is always an important step in seismic processing, in order to produce high-quality data for subsequent imaging and inversion. Different types of noise can be suppressed using targeted denoising methods. For outlier noise with singular amplitudes, many classical denoising methods suffer from signal leakage. To mitigate this issue, we developed a statistics-based mask method and incorporated it into the compressive sensing (CS) framework, in order to remove outlier noise. A statistical analysis for seismic data amplitudes was first used to identify the locations of traces containing outlier noise. Then, the outlier trace locations were compared with a mask matrix generated by jitter sampling, and we replaced the sampled traces of the jitter mask that had the outlier noise with their nearby unsampled traces. The optimized sampling matrix enabled us to effectively identify and remove outliers. This optimized mask strategy converts an outlier denoising problem into a data reconstruction problem. Finally, a sparsely constrained inverse problem was solved using a soft-threshold iteration solver to recover signals at the null locations. The feasibility and adaptability of our approach were demonstrated through numerical experiments for synthetic and field data.

A detailed introduction can be downloaded and viewed as a pdf, or click this link.

In exploration seismology, the reflections have been extensively used for imaging and inversion to detect hydrocarbon and mine resources, which are generated from subsurface continuous impedance interfaces. When the reflector is not continuous and its size reduces to less than half wavelength, the reflected wave becomes scattering, which is also known as the diffraction. Both reflection and diffraction can be used to image subsurface structures, and the latter is helpful to resolve small-scale discontinuities, such as fault plane, pinch-out, Karst caves and salt edge. However, the amplitudes of diffractions are usually much weaker than that of reflections. This makes it difficult to directly identify and extract diffractions from common-shot gathers. On the other hand, they have different geometrical characteristics in the dip-angle common-image gathers (DACIGs), which provides one opportunity to separate diffractions and reflections. In this study, we present an efficient and accurate diffraction separation and imaging method using convolutional neural network (CNN). The labeled data of DACIGs are generated using one pass of seismic modeling and migration for velocity models with and without artificial scatterers. Then, a simplified end-to-end CNN is trained to identify and extract reflections from the DACIGs that contain both reflections and diffractions. Next, two adaptive subtraction strategies are presented to compute diffraction DACIGs and stacked images.

A detailed introduction can be downloaded and viewed as a pdf, or click this link.

We propose a method based on the eigenform analysis to completely separate the fast P wave, slow P wave, and S wave in the poroelastic medium. Because the propagation characteristic of a slow P wave is different from that of a fast P wave and S wave, the wavefield separation methods in the elastic medium cannot be directly applied to produce a complete wave-mode separation. We first use the Helmholtz decomposition to compute the P- and S-mode potential wavefields to separate the two wave modes. Then, the eigenvalues and eigenvectors of the P-wave potentials for the solid and fluid phases are calculated. Using the eigenvectors, we can construct the separation operators to extract fast and slow P waves. Numerical experiments illustrate that the fast and slow P waves not only can be separated completely, but also have the same amplitudes and phases as the coupled P-wave potentials.

A detailed introduction can be downloaded and viewed as a pdf, or click this link.

Wave field simulation in part of the Shengli model

Sensitive kernel analysis

● 5km reception, penetration depth of about 2.8km

● 15km reception, penetration depth of about 7.8km

● 20km reception, penetration depth of about 9.8km

● 25km reception, penetration depth of about 10km

We present an elastic least-squares imaging method for tilted transversely isotropic media and apply it to land multicomponent and marine pressure data. Unlike traditional RTM, we use the relative perturbations to the product of density and squared axial (compressional/shear) velocities as reflectivity models (Δln C33 and Δln C55), and estimate them by solving a linear inverse problem. Numerical experiments illustrate that subsurface reflectors can be well resolved in adjoint images for land multicomponent data, because of the presence of both P- and S-waves in seismograms. Least-squares migration helps to further improve spatial resolution and image amplitudes. Since there are no direct S-waves in marine streamer data, adjoint RTM images of Δln C55 are mainly resolved with the converted S-waves and are not as good as those in Δln C33 images. By approximating the Hessian inverse, least-squares migration allows us to take advantage of the weak converted PSP-waves and improve the Δln C55 image quality.

A detailed introduction can be downloaded and viewed as a pdf, or click this link.

We parameterize the wave equation with an angle-dependent reflectivity, and derive the corresponding linearized forward modeling and adjoint migration operators. Because Gaussian Beam migration naturally incorporates propagation directions in wavefield extrapolation, we compute the Green’s function using the Gaussian beam summation method. To improve the common-image gather (CIG) quality for low-fold and low-SNR data, a shaping regularization over the half-opening angles is introduced in the conjugate gradient scheme to iteratively update the angle-dependent reflectivity model. A flattening-enhanced summation is used to compute the stacked images by accounting for the depth moveout of CIGs caused by velocity errors, and produces constructive stacking results.

A detailed introduction can be downloaded and viewed as a pdf, or click this link.

We present an integrative analysis for the 2020 Stanley earthquake using state-of-the-art methods in seismology and remote sensing. An opposing-dip two-fault model is prescribed to reconcile the observations of aftershock distribution, teleseismic records and near-field InSAR deformation, which includes an unmapped northern subfault with predominantly strike-slip faulting and a southern subfault subparallel to the Sawtooth fault with predominantly normal faulting. The converging fault geometry allows the rupture to traverse a surficial 10-km-wide step, which is greater than the limiting dimension(3-4 km) that commonly ceases earthquake ruptures. Considering tectonic settings, we infer that the composite ruptures involving both strike-slip and normal faulting appears to be typical for earthquakes located near the northern boundary of the CTB, which is different from the predominantly normal faulting in the central CTB.

A detailed introduction can be downloaded and viewed as a pdf, or click this link.

Based on the eigenform analysis, we develop an efficient pseudo-Helmholtz decomposition method for the VTI and TTI media, which produces vector P and S wavefields with the same amplitudes, phases and physical units as the input elastic wavefields. Starting from the elastic VTI wave equations, we first derive the analytical eigenvalues and eigenvectors, then use the Taylor expansion to approximate the square-root term in the eigenvalues, and finally obtain a zero-order and a first-order pseudo-Helmholtz decomposition operator. Because the zero-order operator is the true solution for the case of = δ, it produces accurate wavefield separation results for elliptical anisotropic media. The first-order separation operator is more accurate for non-elliptical anisotropy. Since the proposed pseudo-Helmholtz decomposition requires solving an anisotropic Poisson’s equation, we propose two fast numerical solvers. One is based on the sparse lower-upper (LU) factorization, which can be repeatedly applied to the input elastic wavefields once computing the lower and upper triangular matrices. The second solver assumes the model parameters are laterally homogeneous within a given migration aperture. This assumption allows us to efficiently solve the anisotropic Poisson’s equation in the z − k**x domain, where k**x and z denote the horizontal wavenumber and depth, respectively. Using the coordinate transform, we extend the pseudo-Helmholtz decomposition to the TTI media. The separated vector wavefields are used to produce PP and PS images by applying a dot-product imaging condition.

A detailed introduction can be downloaded and viewed as a pdf, or click this link.

During seismic wave propagation, intrinsic attenuation inside the earth gives rise to amplitude loss and phase dispersion. Without appropriate correction strategies in migration, these effects degrade the amplitudes and resolution of migrated images. Based on a new time-domain viscoacoustic wave equation, we have developed a viscoacoustic reverse time migration (RTM) approach to correct attenuation-associated dispersion and dissipation effects. A time-reverse wave equation is derived to extrapolate the receiver wavefields, in which the sign of the dissipation term is reversed, whereas the dispersion term remains unchanged. The difference between the forward and time-reverse wave equations is consistent with the physical insights of attenuation compensation during wavefield backpropagation. Due to the introduction of an imaginary unit in the dispersion term, the forward and time-reverse wave equations are complex valued. They are similar to the time-dependent Schrödinger equation, whose real and imaginary parts are coupled during wavefield extrapolation. The analytic property of the extrapolated source and receiver wavefields allows us to explicitly separate up- and downgoing waves. A causal imaging condition is implemented by crosscorrelating downgoing source and upgoing receiver wavefields to remove low-wavenumber artifacts in migrated images. Numerical examples demonstrate that our viscoacoustic RTM approach is capable of producing subsurface reflectivity images with correct spatial locations as well as amplitudes.

A detailed introduction can be downloaded and viewed as a pdf, or click this link.

We present a novel passive-source monitoring approach using vector-based elastic time-reversal imaging. By solving the elastic wave equation using observed multicomponent records as boundary conditions, we first compute back-propagated elastic wavefields in the subsurface. Then, we separate the extrapolated wavefields into compressional (P-wave) and shear (S-wave) modes using the vector Helmholtz decomposition. A zero-lag cross-correlation imaging condition is applied to the separated pure-mode vector wavefields to produce passive-source images. To capture the propagation of microseismic fracture and earthquake rupture, we modify the traditional zero-lag cross-correlation imaging condition by summing the multiplication of the separated P and S wavefields within local time windows, which enables us to capture the temporal and spatial evolution of earthquake rupture.

A detailed introduction can be downloaded and viewed as a pdf, or click this link.

We present a novel passive-source monitoring approach using vector-based elastic time-reversal imaging. By solving the elastic wave equation using observed multicomponent records as boundary conditions, we first compute back-propagated elastic wavefields in the subsurface. Then, we separate the extrapolated wavefields into compressional (P-wave) and shear (S-wave) modes using the vector Helmholtz decomposition. A zero-lag cross-correlation imaging condition is applied to the separated pure-mode vector wavefields to produce passive-source images. To capture the propagation of microseismic fracture and earthquake rupture, we modify the traditional zero-lag cross-correlation imaging condition by summing the multiplication of the separated P and S wavefields within local time windows, which enables us to capture the temporal and spatial evolution of earthquake rupture.

A detailed introduction can be downloaded and viewed as a pdf, or click this link.

We derive a time-domain complex-valued viscoacoustic wave equation for modelling seismic wave propagation in constant-Q media. The advantages of the proposed viscoacoustic wave equation include (1) that dispersion and dissipation effects are separated, which allows compensating amplitude loss in the migration and inversion by flipping the sign of dissipation term, (2) that quality factor Q is explicitly incorporated in the wave equation, which makes it easier to derive its sensitivity kernel in comparison with constant-Q and generated standard linear solid (GSLS) approaches, and (3) that it can be solved in the time-domain using time matching, which does not require to solve the inverse of impedance matrix and hence saves computational cost compared with the frequency-domain approach.

Download the data.