Correction filter use in finite-difference elastic modelling

Peter Malcolm Manning

Correction filtering of the finite-difference elastic wavefield has been found to be a practical and efficient process. In a particular case shown, the cost of the beneficial effects was obtained by using a minimal convolution filter with an overall size of 3 by 3 points, and this resulted in a 80 percent increase in run times. Comparable results obtained by reducing sampling intervals required a one third reduction, which cost a 180 percent increase in run times. Further tests showed that the particular corrections filter set used was still quite effective when used on models with velocities 25% lower than the design velocities.