The original dataset is considered oversampled. Therefore I decimated the survey to simulate a realistic acquisition geometry. The shot lines from the original survey were spaced at 140 meters, whereas in the decimated experiment they alternate between 140 and 280 m for an average spacing of 210 m. I also extracted every third receiver line to simulate a line spacing of 210 meters instead of the original 70 meter spacing. Figure survey210 shows the layout of the source and receiver lines for the 210 meter grid.
To make the data handling and processing quicker, I only considered the central part of the survey which included the river channels. The model is 2.1 km long and 2.1 km wide sampled at 17.5 meter. I extracted a subset of 20,000 traces whose source-receiver offset is between 600 and 1,000 meters and azimuth between -45o and 45o. Figure fold shows the fold distribution for the 3D subset binned at the nominal CMP spacing, whereas Figure fold-full represents a fold chart for the same offset and azimuth range from the original survey. There is a fundamental problem with the the sub-sampled subset; variations in fold between different bins vary substantially from 0 to a maximum of 10. If not properly accounted for, they can distort the quality of the final image.