In a study of the Sterling-Triangle area of Alaska, U.S.A., we initiated prestack depth migration (PSDM) to improve imaging on a prospect initially identified on a prestack time-migrated (PSTM) volume. Under the isotropic media assumption, the first few iterations of the reflection tomography had difficulty in converging to the proper velocity model. Upon further investigation, a very-high-velocity conglomerate layer was identified in the middle of the section across the whole survey area. We adopted the salt-flood practice, routine in depth-imaging salt provinces such as the Gulf of Mexico. The strategy was to focus on the shallow section above the conglomerate first, followed by a constant-velocity flood for picking the conglomerate base. The finalisotropic PSDM result showed that significant residual moveout differences existed on gathers along different azimuths. The net anisotropic effect on the isotropic PSDM was a degraded final PSDM volume. In the subsequent anisotropic PSDM work, azimuthally variant horizontal velocities were allowed in the model building. Common-image-point (CIP) gathers were created along different azimuths using sectored input gathers. Residuals picked on the sectored CIP gathers were used in joint tomography to invert different horizontal velocities. Incorporating significant well information, we built an anisotropic velocity model such that the azimuthal moveout on the butterfly gathers was essentially flat. The resulting anisotropic PSDM was consistent with well data and could be interpreted with much higher confidence.