INTRODUCTION

Reflection seismology gives us magnificent volumetric images of the earth's interior. Image construction is a two-stage process. The first stage estimates the three-dimensional velocity. The second stage uses it to make the reflectivity cube. Both reflectivity estimation and velocity estimation are inverse problems. For most datasets, reflectivity would be better determined than velocity, if somebody gave us the velocity function. But nobody does. Velocity estimation remains an art, a mixture of every skill we have and more. It requires ideas we haven't coded yet and it requires ideas we haven't yet concocted. We have no chance to estimate velocity reliably if we do not honor the data complexity when migrating. We need better migration methods when the data are poorly sampled and the velocity function is rapidly varying. I claim that SEP needs to avoid being distracted by the ultimate goals of the petroleum community. We need to solve particular problems in estimating the earth's 3-D velocity (and reflectivity). Everyone else depends on us to get it right.