Variable water depth can cause severe degradation of marine seismic data. This paper presents a technique for correcting the effects of water depth variation and is a case history of applying the technique to a line of data from the Philippines offshore. The line crosses a deep submarine valley. It will be shown that when the water depth changes rapidly relative to the cable length, the timing variations introduced will not be static. They are dynamic, not static, because they differ for different event times of a single trace. To compensate for these dynamic timing variations, a two‐stage technique was used. A ray‐trace modeling program calculated the traveltimes to several depths, both for where the valley is present and where it is absent. A second program used the model results to shift the samples on all seismic traces to the time they would have if the valley were not present. The most difficult part of this project was finding a good model. The model is composed of two parts: the depth of the sea floor and the velocity‐depth relationships below the sea floor. The depth of the sea floor was estimated from the first arrivals on the near‐offset traces of the seismic data. This was difficult because of the shallowness of the normal sea floor (about 80 m) and the large offset between the shot and the first group (255 m). The first arrivals were head waves, not reflections, off the sea floor. The reflections from the valley had to be migrated to obtain accurate depths. The subsea velocity‐depth relations also had to be estimated from the seismic data. However, the results of applying the corrections calculated from this model to the data show a definite enhancement of reflector continuity; velocity semblance contour plots show the same enhancement. These results are contrasted with the results of applying purely static corrections. The static corrections also improve reflector continuity, but the dynamic corrections do a better job of it. Although the dynamic corrections improve a brute stack of the data, more importantly they allow additional processing to produce a much better final stack. Thus, the data were further processed to produce an optimal final stack. The dynamic corrections in particular allowed a much better choice of normal moveout (NMO) velocities near the valley. Also, a zone of near‐surface, high‐velocity material near the valley was detected by distortion of reflections on 100 percent shot records. Compensation for the zone was effected with a set of localized, static corrections. The data were also muted, band‐pass filtered, and dip filtered. Although the final stack is greatly improved, there is still a serious degradation of the data under the valley. This is because the valley not only introduces timing errors, but it also reduces the amplitude of the reflections returned from below it. The valley also introduces coherent noise in the form of scattering off its sides and enhanced multiples. These additional problems not only affect the final stack, but limit the accuracy with which the model can be built to correct the timing errors. Thus, corrections for the effects of highly variable water depth, preferably dynamic, are required in order to obtain the optimal stack of seismic data recorded over such a sea bottom. The difficulty in obtaining the corrections would be greatly reduced if accurate, closely spaced, fathometer measurements of water depth were made an integral part of marine seismic data recording.
Simultaneous sources: The inaugural full-field, marine seismic case history from Australia