HR: 0830h
AN: S21F-0404 [PDF]
TI: Differential Semblance Velocity Analysis by Wave Equation Migration, Applications to High Resolution
HAFB Dataset
AU: * Shen, P
EM: pshen@rice.edu
AF: Rice University, 6100 Main Street, Houston, TX 77005 United States
AU: Symes, W W
EM: symes@rice.edu
AF: Rice University, 6100 Main Street, Houston, TX 77005 United States
AU: Morton, S
EM: Morton@hess.com
AF: Amerada Hess Corporation, One Allen Center
500 Dallas Street, Houston, TX 77002 United States
AU: Gao, F
EM: fcgao@rice.edu
AF: Rice University, 6100 Main Street, Houston, TX 77005 United States
AU: Levander, A
EM: alan@rice.edu
AF: Rice University, 6100 Main Street, Houston, TX 77005 United States
AB:
In pre-stack migrated seismic reflection data, common image gathers in angle are flat at correct velocity, and gathers in
offset are concentrated at zero offset. Differential semblance measures the deviation from flatness or concentration. The
objective function of differential semblance optimization (DSO) responds smoothly to velocity changes. Therefore gradient
descent methods are attractive for DSO velocity updates. Because of their kinematic fidelity, wave equation (depth
extrapolation) migration methods are a natural means for velocity analysis in complex structure. We demonstrate the
computation of the differential semblance gradient formulated through the adjoint of the linearized wave equation migration
and its use in constructing velocity updates by optimization. A synthetic example with a low velocity lens has been tested
and shows satisfactory velocity reconstruction by differential semblance optimization. We apply this algorithm to the high
resolution land surface data collected at Hill Air Force Base. Images and velocity reconstruction are compared with results
of waveform type of analysis, and with more conventional data processing results.
UR: http://terra.rice.edu/~pshen
DE: 0689 Wave propagation (4275)
DE: 0900 EXPLORATION GEOPHYSICS
DE: 0902 Computational methods, seismic
SC: Seismology [S]
MN: 2003 Fall Meeting