Water-bottom multiples have long been recognized as a problem in marine seismic prospecting. This problem becomes serious in an area such as the Arctic where the seafloor has much higher reflectivity than that of deeper geological horizons. In this case multiple interference is often strong enough to completely overpower primary reflections and thus makes interpretation impossible.
Water-bottom multiples are those multiples whose raypaths lie entirely within the water layer. Because water velocity is much lower than the seismic velocities of sediments, velocity discrimination methods are commonly used to suppress the water-bottom multiples. Examples of such methods are velocity filtering (Thorson and Claerbout, 1985) and f-k filtering (Yilmaz, 1987).
Here I present a new algorithm that, in theory, can accomplish the goal of eliminating the water-bottom multiples on common midpoint (CMP) gathers, provided that the differences between the stacking velocities of primaries and multiples are sufficiently large. As does velocity filtering, my algorithm consists of three steps: forward transformation, filtering, and backward transformation. But the transformation operators in my algorithm are completely different from the one used in velocity filtering. In the following sections, I will first explain the principles of the algorithm and then show synthetic examples and field-data examples. In the end, I will discuss several limitations of the algorithm that may arise in practice.