We have developed the theory and synthetic tests of elastic least-squares reverse time migration (ELSRTM). In this method, a least-squares reverse time migration algorithm is used to image multicomponent seismic data based on the first-order elastic velocity-stress wave equation, in which the linearized elastic modeling equations are used for forward modeling and its adjoint equations are derived based on the adjoint-state method for back propagating the data residuals. Also, we have developed another ELSRTM scheme based on the wavefield separation technique, in which the P-wave image is obtained using P-wave forward and adjoint wavefields and the S-wave image is obtained using P-wave forward and S-wave adjoint wavefields. In this way, the crosstalk artifacts can be minimized to a significant extent. In general, seismic data inevitably contain noise. We apply the hybrid [ ] misfit function to the ELSRTM algorithm to improve the robustness of our ELSRTM to noise. Numerical tests on synthetic data reveal that our ELSRTM, when compared with elastic reverse time migration, can produce images with higher spatial resolution, more-balanced amplitudes, and fewer artifacts. Moreover, the hybrid [ ] misfit function makes the ELSRTM more robust than the [ ] misfit function in the presence of noise.

Elastic reverse time migration (ERTM) is a state-of-the-art imaging technique used for determining complicated subsurface structures. However, the migrated images often suffer from low spatial resolution, low signal-to-noise ratio (S/N), and unbalanced amplitudes because the theoretical hypothesis of ERTM cannot be satisfied in practice. Although elastic least-squares reverse time migration (ELSRTM) has been proposed to address the issues of ERTM, the resulting images are generally represented by parameter perturbations such as P- and S-velocity perturbations, which have the different physical meanings from the ERTM images. To produce improved ERTM images, we used a least-squares RTM method for elastic data in isotropic media by applying least-squares inversion to ERTM. In the least-squares ERTM method, the forward operator generates multicomponent seismic data from the migrated images by applying elastic wavefield decomposition, scalar wavefield extrapolation, and wavefield recomposition operators. Additionally, the adjoint operator generates PP and PS images using ERTM, at which point the wavefield decomposition operator and scalar imaging condition are applied in the imaging process. Compared to conventional ERTM, our least-squares ERTM method enables us to produce improved ERTM images with higher resolution, more balanced amplitudes, and fewer artifacts. Several synthetic and field data examples were used to validate the effectiveness of the proposed least-squares ERTM method.

Anelasticity of the earth subsurface medium, which is quantified by the quality factor $Q$ , causes the dissipation of seismic energy. Strong attenuation effect resulting from geology such as gas clouds (gas-filled sandstone) is a challenging problem for high-resolution imaging. To compensate the attenuation effect, first we need to accurately estimate the attenuation parameter. However, it is difficult to directly derive a heterogeneous attenuation Q model. This research letter proposes a method to derive a Q model corresponding to strong attenuative media from marine reflection seismic data using convolutional neural network (CNN), a popular deep learning framework. We treat Q anomaly detection problem as a semantic segmentation task and train a network to perform a pixel-by-pixel prediction to invert a pixel group that belongs to the strong level of attenuation probability. The proposed method uses a volume of marine 3-D reflection seismic data for network training and validation, which needs only a small part of real data as the training set due to the feature of U-Net. In the final stage, to evaluate the attenuation model, we validate the predicted heterogeneous Q model using deabsorption prestack depth migration (Q-PSDM), a high-resolution imaging result in depth domain with appropriate compensation is obtained.

Elastic angle-domain common-imaging gathers (ADCIGs) extracted from elastic reverse time migration (ERTM) play a pivotal part in elastic migration velocity analysis, elastic amplitude variation with angle, and attribute interpretation. In practice, however, elastic ADCIGs often suffer from unbalanced amplitude behavior, poor resolution, and low-wavenumber artifacts because of insufficient velocity information, limited recording aperture, uneven illumination, and other inaccuracies of the migration operator. We have developed a new method to improve the quality of elastic ADCIGs extracted from ERTM by posing ERTM imaging as an inverse problem whose misfit function measures the difference between simulated and observed data. The misfit function can be minimized by updating elastic offset-domain common-imaging gathers (ODCIGs) using an optimization method. Based on the transformation between ADCIGs and ODCIGs, the forward operator generates multicomponent seismic data from elastic ODCIGs by applying a scattering condition, and the adjoint operator generates elastic ODCIGs from ERTM using a subsurface space-shift imaging condition. Compared with elastic ODCIGs extracted from ERTM, our method effectively improves the focusing of elastic ODCIGs to produce elastic ADCIGs with higher resolution, fewer artifacts, and improved amplitude coherency across different reflection angles. Several synthetic examples were used to validate the effectiveness of the method.

Elastic-wave migration in anisotropic media is a vital challenge, particularly for areas with irregular topography. Gaussian-beam migration (GBM) is an accurate and flexible depth migration technique, which is adaptable for imaging complex surface areas. It retains the dynamic features of the wavefield and overcomes the multivalued traveltimes and caustic problems of Kirchhoff migration. We have extended the GBM method to work for 2D anisotropic multicomponent migration under complex surface conditions. We use Gaussian beams to calculate the wavefield from irregular topography, and we use two schemes to derive the down-continued recorded wavefields. One is based on the local slant stack as in classic GBM, in which the PP- and PS-wave seismic records within the local region are directly decomposed into local plane-wave components from irregular topography. The other scheme does not perform the local slant stack. The Green’s function is calculated with a Gaussian beam summation emitted from the receiver point at the irregular surface. Using the crosscorrelation imaging condition and combining with the 2D anisotropic ray-tracing algorithm, we develop two 2D anisotropic multicomponent Gaussian-beam prestack depth migration (GB-PSDM) methods, i.e., using the slant stack and nonslant stack, for irregular topography. Numerical tests demonstrate that our anisotropic multicomponent GB-PSDM can accurately image subsurface structures under complex topography conditions.

Processing pressure shear (PS)-wave data is more challenging than PP-wave data because of the asymmetry of the source-to-receiver ray paths, particularly for irregular surfaces and sparse acquisition. Gaussian beam summation (GBS) migration is an effective method for imaging seismic data from irregular surfaces. In this letter, we introduce a converted PS-wave migration method for irregular surfaces using GBS, in which a scalar wavefield is used for wavefield propagation imaging. Cross correlation imaging is performed using forward-continued source wavefields calculated with the P-wave and reverse-continued wavefields calculated using the S-waves from the receivers. The corresponding Green's function is constructed as a superposition integral of the Gaussian beams emitted from the source and the receivers on the irregular surfaces, respectively. Numerical tests demonstrate that the method is a flexible and effective alternative for accurate imaging of converted PS-wave data from irregular surfaces.