A seismic survey located at Hero Ridges tectonic zone is overwhelmed by refraction multiple, which makes attenuation of multiples critical to obtain meaningful subsurface image. P & S Refraction multiples, or later refractions, are a common, difficult type of land seismic noise that are often strong and hard to remove. They occur frequently in foothills or karsted terrain around the world and other areas with surface heterogeneity and strong refractors. S-refraction multiples can be stronger than P refraction multiples and are often confused with surface waves.

The problem of refraction multiples is that they 1) are often strong, 2) are fast energy that is only a little slower than reflection energy, 3) often arrive at the same time as reflections, and 4) have some irregularity so that linear radon or FK filtering are only partially effective.

We test a new approach to remove these refraction multiples on a difficult land dataset from the study area with some success. We compare the new approach against conventional approaches, such as linear radon noise removal. The new approach uses a pattern-based machine learning method that is able to resolve the irregularity of the refraction multiples and aid in their removal.

This content is only available via PDF.
You can access this article if you purchase or spend a download.