Choosing the migration velocity

In principle, we don't need to migrate the data with the correct velocity for the primaries, just a velocity that satisfies the inequality $V_{\mbox{mul}}\le V_{\mbox{mig}}\le V_{\mbox{prim}}$. This will guarantee that the multiples are overmigrated and the primaries undermigrated. Since we are more willing to tolerate some leftover multiple energy rather than to lose primary energy, the second part of the inequality is more critical. Unless there are very strong lateral velocity variations, a simple V(z) will probably suffice since both migration and adjoint-migration are performed with the same velocity model. Migrating with a V(z) model has the obvious advantage that the migration and adjoint-migration will run faster.