Two SEG multiple suppression workshop data sets were used as input to a multiple suppression algorithm based on upward continuation, originally formulated by Berryhill and Kim 1986. The recorded data are upward continued a distance equal to twice the water depth, with water velocity, in order to add a lag equal to the multiple period. A primary in the upward continued data should then line up with its first multiple in the original recorded data, a first order multiple in the upward continued data with a second order multiple in the original, and so on. The upward continued data and the original are cross-equalized using a scheme similar to Rickett 1997, and subtracted, leaving (hopefully) only primaries.
The first of the two data sets was recorded in deep water in the Gulf of Mexico, and shows individually identifiable multiples, but is complicated because there are multiple trains with different interfering periodicities. A shallow subsurface reflector brighter than the seafloor gives rise to a second set of multiple reflections. I find that both multiple trains can be removed simultaneously, but the large number of filter coefficients raises the danger that primaries will be predictable from the upward continued data.
The second data set was recorded in shallower water, off the coast of Western Australia, where a hard water bottom gives rise to short period, high amplitude multiples. These may present an intractable problem. At any rate, the results produced by this attempt at suppression are uninspiring to this point.