Summary

Delivering high-resolution seismic images is an important objective for both land as well as marine data processing. Often, we tend to compromise bandwidth in return for better signal-to-noise ratio, however, this has a direct negative impact on the resolution of the final seismic image. In this abstract, we develop and discuss a new approach using wavelet deconvolution for noise attenuation that delivers a better signal-to-noise ratio without compromising the bandwidth. Our approach is adaptive, automatic, non-stationary and exploits the signal characteristics in a complete 3D sense. We use forward and inverse normal move-out (NMO) in conjunction with a wavelet transform to effectively distinguish signal from noise and subsequently remove the undesired components. Our approach is well suited for dealing with high amplitude scattered noise such as ground roll and other forms of noise that may be spatially aliased. We demonstrate the effectiveness of our algorithm using multiple examples from various land data-sets acquired around the world in different geological settings.

Introduction

Land seismic data is often dominated by noise. Contamination from both coherent and incoherent noise is an often unavoidable and pervasive problem for seismic data. In order to produce reliable interpretation and analysis of the subsurface, the elimination of noise is essential. Noise attenuation is not new to industry. However, despite much effort and great progress, some challenges still remain. One of these challenges is to achieve signal-noiseratio (SNR) improvement without affecting the bandwidth such that we don’t compromise the resolution of the final migrated images.

In standard land data processing, any recorded wavefields that don’t provide information in the form of a subsurface reflection seismic image, such as, surface waves or ground roll, multiples and direct arrivals, are often considered to be coherent noise. Can we use the properties of the noise types to eliminate them? It is known in seismic data processing that coherent noise or signal in one domain may become incoherent in another domain. Signal and noise that appear inextricably tangled can often be separated by transforming to a different domain. This produces one of the mantras of noise elimination techniques, "find a domain where we can distinguish signal and noise." One example is the cross-spread domain and there are many successful examples in literature (e.g., Stein and Langston, 2007).

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