Chenlong Wang

Elastic Wave Mode Decoupling for Full Waveform Inversion

Elastic full waveform inversion (EFWI) is coming into use to obtain more accurate parameter estimation in the subsurface. However, it seriously suffers from trade-off between the elastic parameters. To mitigate this effect, we present a strategy for gradients calculation by taking account of elastic wave mode decoupling. Through the analysis of diffraction patterns, we modify the gradients by applying cross correlation with the forward wavefields emitted from the source and the back-propagated residual wavefields after P-S decoupling. Synthetic examples on a constant-density isotropic model show that our strategy provides new opportunity to reduce the trade-off effects and thus enhance the validity of EFWI.

Local Angle Domain Elastic Reverse Time Migration in TI Media

Local angle-domain elastic reverse time migration (ERTM) can provide more opportunities for complex structure imaging and elastic parameter inversion. However, its implementation still has many challenges, especially in anisotropic media. We present an accurate local angle-domain ERTM algorithm for heterogeneous TI media on the base of the cross correlation imaging condition and vector decomposition of the elastic wave modes. At the mean time, the computational cost is significantly reduced by applying fast algorithms of wave mode separation and decomposition using low-rank approximation of the mixed-domain operators. Synthetic example shows that, compared with the Helmholtz decomposition, the efficient wave mode separation/decomposition based on polarization projection promotes the quality of angle-domain common-image gathers for TI media.

Numerical Pure Wave Source Implementation and Its Application to Elastic Reverse Time Migration in Anisotropic Media

For isotropic media, a P- or S-wave source can be easily implemented by adding a disturbance (wavelet) on its related stress components. However, it does not work for anisotropic media. Inspired by the idea of elastic wave mode decoupling, we propose a general approach to numerically simulate the pure wave source for elastic wave propagation. This is achieved by calculating the spatial counterpart of the normalized polarization for the given mode in the anisotropic media where the actual source is triggered. This approach may be helpful for seismic imaging, waveform inversion and micro-seismic monitoring when the source locates in an anisotropic rock, such as gas shale. Synthetic examples demonstrate the feasibility and application to elastic reverse time migration in a VTI medium.

Automatic Shear-wave Velocity Analysis with Elastic Reverse Time Migration

Converted-wave has the potential to be a positive complement of the compressional imaging in some special cases. However, converted wave imaging is more difficult because the need of estimating the background velocity of S-wave. By applying differential semblance to measure the focusing error in the imaging domain, we automatically estimate the background S-wave velocity using reverse-time migration with a given estimated P-wave velocity. Compared to the P-wave, the S-wave velocity estimation via PS imaging gather does not need gradient from the source side. A simple synthetic result demonstrates the feasibility of the proposed method.

P/S Separation of Multi-component Seismogram Recorded in Anisotropic Media

Summary The multi-component receiver records complete wavefields, which include both P- and S-wave modes. The combination of P- and S-wave characteristics yields better insight into rock properties in the sub-surface. We propose a new approach to decoupling different wave modes in multi-component gather in anisotropic media. By solving the quartic dispersion relation and eigenvalue problem of the Christoffel equation, the polarization vectors of different wave modes can be obtained on the acquisition surface. Accordingly, one can decouple different wave modes both into scalar and vector forms by projecting elastic wavefields onto polarization-related directions in the frequency-wavenumber domain. Synthetic examples demonstrate the validity of the proposed method.

3D Vector Imaging of Converted Waves for Fractured Reservoirs

Summary The fractured system widely exists in the earth media. It causes anisotropy on the seismic wave scale. Passing through such media, a shear wave splits into two mutually orthogonal waves with different propagation velocities. Unlike the well-behaved qP mode, the qS1- and qS2-wave do not consistently polarize as a function of the propagation direction. Therefore, continuous polarization directions of qSV and qSH mode are used for accurate wave mode decoupling. Combined with vector imaging condition, the proposed method properly utilizes two converted shear waves in elastic reverse time migration (ERTM), without polarity reversal. Sensitivity analysis of wave mode decoupling demonstrates the potential of converted wave imaging in analyzing fracture attribute. Synthetic examples demonstrate the validity of this approach.

Elastic Wave-equation Reflection Traveltime Inversion Using Dynamic Warping and Wave Mode Decomposition

Elastic full waveform inversion (EFWI) provides high-resolution parameter estimation of the subsurface but requires good initial guess of the true model. The traveltime inversion only minimizes traveltime misfits which are more sensitive and linearly related to the low-wavenumber model perturbation. Therefore, building initial P and S wave velocity models for EFWI by using elastic wave-equation reflections traveltime inversion (WERTI) would be effective and robust, especially for the deeper part. In order to distinguish the reflection travletimes of P or S-waves in elastic media, we decompose the surface multicomponent data into vector P- and S-wave seismogram. We utilize the dynamic image warping to extract the reflected P- or S-wave traveltimes. The P-wave velocity are first inverted using P-wave traveltime followed by the S-wave velocity inversion with S-wave traveltime, during which the wave mode decomposition is applied to the gradients calculation. Synthetic example on the Sigbee2A model proves the validity of our method for recovering the long wavelength components of the model.