Time domain internal multiple prediction on synthetic and field vertical seismic profile data

Matthew Eaid, Kristopher A. Innanen

Surface seismic methods represent the only technology capable of large scale 3D characterization of conventional and unconventional oil and gas reservoirs, monitoring of enhanced oil recovery, and water and CO₂ injection. When seismic energy is injected into the earth, and the earth response recorded, two main type of events are seen, primaries and multiples. Primary energy arises from waves that have reflected once and returned to the surface to be recorded; multiples arise from waves that reflect more than once. Most seismic interpretation and processing workflows treat primary events as signal and multiples as undesirable noise. It is for this reason that the prediction and subsequent removal of multiple energy has become a popular research topic.

Multiple energy is classified into two main types, surface related or free surface multiples, and internal multiples. Free surface multiples are multiples in which at least one of the downward reflections occurs at the free surface of the earth. Internal multiples are multiples in which all downward reflections are restricted to the subsurface. Many successful schemes have been investigated to accurately predict and remove free surface multiples. The same cannot be said for internal multiples, although some important strides have been made in recent years. Weglein et al. (1997) proposed a wave equation based method of internal multiple prediction that is fully data driven, based on the inverse scattering series. The algorithm proposed by Weglein et al. was originally derived in the frequency-wavenumber domain. This paper reviews the time domain algorithm proposed by Innanen (2015) and then applies it to both synthetic and land zero offset VSP datasets. We also introduce a method of improving our predictions by converting the VSP data into zero offset sections.