The program that generated Figure 24 could be run in reverse to do a migration. All the energy from all the interesting rays would march back to the impulsive source. Would this be an effective migration program in a field environment? It is unlikely that it would. The process is far too sensitive to quantitative knowledge of the lateral velocity jump. It is the quantitative value that determines the reflection coefficient and ultimately the correct recombination of all the wavefronts back to a pulse. To see how an incorrect value can result in further error, imagine using the hyperbola-summation migration method. Applied to this geometry this method implies weighted summation over all the raypaths in the figure. The incorrect value would put erroneous amplitudes on various branches. An erroneous location for the fault would likewise mislocate several branches.
The lesson to be learned from this example is clear. Unnecessary bumps in the velocity function can create imaginary fault-plane reflections. Consistent with known information, a presumed migration velocity should be as smooth as possible in the lateral direction. Unskilled and uninformed staff at a processing center remote from the decision making should not have the freedom to introduce rapid lateral changes in the velocity model.