##### Velocity model building using residual moveout-based wave-equation migration velocity analysis

* by Yang Zhang *

**Downloads**

- Thesis + Computation recipe tar.gz

- Thesis pdf

**Table of contents**

- Chapter 1: Introduction
- Chapter 2: 2-D RMO-based WEMVA
- Chapter 3: 3-D extension of RMO-based WEMVA
- Chapter 4: 3-D field data test —- a target-oriented approach
- Bibliography

**Abstract**

Wave-equation-based velocity estimation is a set of powerful techniques for robust ve- locity model building for complex subsurface regions, in which ray-based methods are usually ineffective or even unsuccessful. However simply switching from ray-based tomography methods to wave-equation-based ones does not fully solve the prob- lem. In the area of wave-equation migration velocity analysis (WEMVA), although some promising results have been shown, several issues are still not well solved in today’s WEMVA methods, preventing them from becoming the industry standard. Specifically, the main issues include: 1) severe nonlinearity, which causes the cycle- skipping problem under large velocity error; and 2) imprecise objective functions, which wrongly penalizes residuals that are not caused by velocity error but other factors such as uneven subsurface illumination and incomplete acquisition geometry.
In this dissertation, I address these issues by developing a new WEMVA method that uses the residual-moveout (RMO) information of the angle-domain common- image gathers (ADCIG) to quantify the velocity model error. In this RMO-based WEMVA approach, I combine the strengths of the wave-equation and the ray-based tomography by replacing the ray-based tomographic operator with a wave-equation- based one, while keeping the conventional ray-based tomography workflow. In con- trast to other WEMVA methods that build their objective functions directly based on the common-image gather amplitdues, this method defines a purely kinematics- based objective function that links to the velocity model through an residual-moveout (RMO) parameter. Since the RMO parameter scales almost linearly with the veloc- ity error, this approach greatly reduces the risk of cycle-skipping in the absence of low-frequency data. Moreover, focusing on the gather kinematics makes this method insensitive to spatial and angular variations of the gather amplitudes, thus leads to high-quality model gradients. In addition, this method does not require explicit picking of the moveout parameters because it uses the derivative over the velocity- scanning semblances to calculate the moveout perturbation. With promising results, my 2-D examples demonstrate that this RMO-based WEMVA method is very robust against cycle-skipping, can effectively flatten the angle gathers, and does not require moveout parameters picking.

Furthermore, I extend the RMO-based WEMVA method to the 3-D case. To deal with multiple azimuths 3-D ADCIG, I augment my method’s formulation by assign- ing independent moveout parameters to each azimuth. A simple synthetic example verifies that the 3-D extension of the RMO-based WEMVA is able to invert simulta- neously velocity information from multiple azimuths. Finally, I apply my RMO-based WEMVA to a 3-D WATS (Wide Azimuth Towed Streamers) field dataset from GOM (Gulf of Mexico). To make applying WEMVA methods to this large industrial scale dataset computationally affordable on the academic computing resources I have in the school of Earth, Energy and Environmental Sciences, I adopt a target-oriented in- version approach that concentrates on a relatively small target area of interest inside the full physical domain of the dataset. The target-oriented RMO-based WEMVA inversion of this field dataset yields geophysically more consistent models. The inver- sion results show convincing imaging improvements and enhancements in the flatness of the 3-D ADCIG universally across the target domain and all azimuths.