Optimized suppression of coherent noise from seismic data using the Karhunen-Loeve transform

Abstract

Signals obtained in land seismic surveys are usually contaminated with coherent noise, among which the ground roll (Rayleigh surface waves) is of major concern for it can severely degrade the quality of the information obtained from the seismic record. Properly suppressing the ground roll from seismic data is not only of great practical importance but also remains a scientific challenge. Here we propose an optimized filter based on the Karhunen--Loéve transform for processing seismic data contaminated with ground roll. In our method, the contaminated region of the seismic record, to be processed by the filter, is selected in such way so as to correspond to the maximum of a properly defined coherence index. The main advantages of the method are that the ground roll is suppressed with negligible distortion of the remanent reflection signals and that the filtering can be performed on the computer in a largely unsupervised manner. The method has been devised to filter seismic data, however it could also be relevant for other applications where localized coherent structures, embedded in a complex spatiotemporal dynamics, need to be identified in a more refined way.