YUEWANG'S THESIS

VARIABLE GRID-SIZE AND TIME-STEP FINITE DIFFERENCE METHOD FOR SEISMIC FORWARD MODELING AND REVERSE-TIME MIGRATION

(Ph.D. Dissertation)

Yue Wang, University of Utah


Click here to view the HTML version of the thesis.
Click here to view the Microsoft PowerPoint presentation of the oral defense.
Click here to download the PDF file of the thesis. (8.5MB)
Click here to download the PostScript file of the thesis. (25MB)
Click here to download the Latex texts and the PostScript figures for the thesis. All the Latex texts and PS files are tared into a tar file ywang_thesis.tar (32MB)


ABSTRACT

A new variable grid-size and time-step finite-difference (FD) method is developed and applied to three different geophysical problems: simulation of tube waves in boreholes, three-dimensional (3-D) ground-motion simulation in sedimentary basin models, and reverse-time migration of multicomponent data. Unlike the conventional FD method, which uses a fixed grid-size and time-step for the entire model region, spatially variable grid-sizes and time-steps are used to achieve the optimal computational efficiency. For tube wave simulations, a fine grid-spacing is used for simulation inside the borehole region, while a coarse grid is used in the exterior region. While the stability condition requires a very fine time step for the fine grid, a variable time-step method provides coarse time steps for simulation in the coarse grid. Variable grid-size and time-step changes are used to achieve both accuracy and efficiency in the simulations.

Numerical tests are performed for the Bayou Choctaw salt-flank model with different borehole models. The results show the important borehole effects on the seismic wavefield for a realistic source bandwidth. The combination of variable grid-size and time-step methods reduces computational costs by several orders of magnitude for the borehole models.

Viscoelastic 3-D simulations are performed for a three-layer Salt Lake basin model. The near-surface unconsolidated layer is modeled with a fine grid, and the deep part of the model is modeled by a coarse grid. Simulation results show that the 3-D basin features and the shallow layer significantly affect the amplitude and duration time of the ground motion. In the elastic case, the approximation by 2-D modeling is insufficient to simulate the 3-D ground motion response. A basin model without a shallow low-velocity layer underestimates the ground motion duration and cumulative kinetic energy by 50$\%$ or more. The simulation of a Bingham Mine blast suggests that a lower S-velocity should be used to get a good match between numerical results and observed field data.

For ocean-bottom or land survey data associated with a low shear-velocity unconsolidated layer near the geophone locations, the variable grid FD method can be used to extrapolate wavefields using a fine grid for the shallow part and a coarse grid for the deep part. It is found that a staggered-grid reverse-time migration scheme can image both primary and multiple energy to their correct reflection positions by using both pressure and particle-velocity data. This is a new result in that migration can now be used to simultaneously image both primary and receiver-side pegleg reflections. The new variable time-step method can be used for the staggered-grid FD scheme and provides optimal computational savings. The combination of variable grid-size and time-step methods speeds up the reverse-time migration by more than ten times for the multicomponent data set in this thesis, compared to a standard reverse-time migration method.