Submarine canyons: Velocity replacement by wave‐equation datuming before stack

Abstract

Submarine canyons incised into the continental slope interfere with the quality of common‐midpoint (CMP) stacked seismic data obtainable from reflectors beneath the sea floor. The interference problem is caused by rough topography in conjunction with the contrast between the acoustic velocity of sea water and the velocity of the exposed rock layers. Geophysicists have long recognized that part of the solution is to replace the traveltimes of raypaths through the water by their traveltimes through an identical thickness of rock. However, use of wave‐equation datuming to effect velocity replacement yields an additional correction for the change in raypath direction that occurs in crossing from rock to sea water; the wave‐equation datuming implementation of velocity replacement is more comprehensive and complete. The wave‐equation datuming method requires an accurate sea‐floor profile as part of the input, along with values of replacement velocity; it does not require knowledge of geology or velocities at depths much greater than the sea floor. Unstacked common‐source and common‐receiver records are processed to appear as if sources and receivers were moved to the water bottom; the velocity of water is replaced; and the sources and receivers are moved back to the sea surface through the replacement medium. The computational method is well‐suited to the irregular surfaces and laterally variable velocities inherent in the problem of submarine canyons. The advantage of this method is that the corrected seismic records accurately emulate the data that would actually be observed if the acoustic velocity of water could be changed physically. The normal‐moveout (NMO) velocity for optimum CMP stacking becomes the root mean square of the layer velocities, including the velocity substituted for that of water. The spurious lateral variation of stacking velocity in the original data is eliminated. Processing of the corrected data through velocity analysis, stacking, migration, and conversion to depth is therefore more reliable.