Seismic noise

Abstract

Seismic noise is an integral part of the seismic record and is defined as all unwanted seismic energy on data. It can be divided into two categories: random and coherent noises. Random noise is not correlated among traces and is easier to attenuate compared to coherent noise that is spatially and/or temporally correlated. Multiples and geologic noise that are coherent noise are more difficult to attenuate and often interfere with seismic signal and makes seismic analysis challenging. Strategies for seismic noise attenuation are needed to preserve the seismic signal of interest and to improve signal-to-noise ratio (SNR). The success in noise attenuation lies in identification and then separation or prediction of signal and noise. Transformation of data to different data domains (X-T, F-K, t-p, curvelet, wavelet domains) have helped in better separating noise from signal. Wave-equation extrapolations, inverse scattering methods, surface related multiple elimination, deconvolution, etc., model the noise and/or data in the process of noise attenuation. Advances in seismic data acquisition and processing have been made to improve SNR. Recently various efforts have been made to use noise as signal and are an active topic of research. This includes using multiples as well as primaries in seismic migration and inversion, and using low and high frequency passive seismic data in imaging subsurface.