Seismic oil exploration requires the processing of large scale data sets. One of the main goals is to convert the seismic data (time histories at receivers) into depth maps of the underground in the search for stratigraphic oil traps and other likely places for the precise drilling of producing wells.

In complex geological areas the most advanced technique is depth reverse time migration using full wave equation techniques. All migration techniques are as good as the velocity model that is used and therefore are iterative in nature: from coarse models to more precise ones. In general, that requires considerable computing time and manual interaction.

Seismic travel time ray tracing or eikonal based tomography showed some early promise in helping automate the velocity improvement process but it was dimmed to be too hard to implement practically, because of the need to pick interpreted travel times from unprocessed seismic data. In the last few years a new set of techniques have arisen that uses depth residuals in depth migrated common offset sections that are considerably easier to identify and that can be picked automatically.

We show in this talk how to convert a travel time tomography code into one that can process depth residuals and exemplify its use in some large scale real data sets.