![]() |
|
| |
|
|
|
Variable grid Finite-difference modeling including surface topographyby Submitted to the Department of Earth, Atmospheric, and Planetary Sciences on August 6, 1999 in partial fulfillment of the requirements for the degree of Master of Science ABSTRACT
Several methods of free surface condition are bench marked, and an accurate
and simple condition is proposed. In the proposed free surface condition, stresses
are calculated so that the shear and normal stresses perpendicular to the boundary
are zero. The calculation of particle velocities does not involve any specific
calculations, and the particle velocities are set to zero above the free surface.
A stable variable grid method is introduced, where we use a three times finer
grid in the near surface or low velocity region compared to the rest of the
model. In order to reduce instability, we apply averaging or weighting to the
replacement of the coarse grid components within the fine grid. The method allows
us to avoid any limitation of the shape of the grid size boundary. Numerical
tests indicate that approximately ten grid-points per shortest wavelength with
the variable grid method results in accurate calculations. The method requires
a stair-shaped discretization of a free surface. We investigated the stair-shaped
structures, and found that the cause of the dispersion from irregular free surface
is mainly a numerical error due to the large grid sizes rather than the Rayleigh
waves scattering due to the stair-shaped boundary. The finite-difference modeling
is applied to the investigation of near surface wave propagation. Several numerical
simulations are performed to show the characters of wave propagation in the
near surface region. The simulations show that the low velocity thin layers
just below the surface and anelastic attenuation have significant effect on
surface seismic record. The 2-D modeling of near surface structure beneath a
2-D refraction survey line is carried out. The comparison of the observed data
with theoretical waveforms is performed. The characters in the observed data
can be explained by a subsurface model constructed by P-wave traveltime tomography. Return to Theses Return to ERL Home Updated: August, 1999
|