Land Seismic Noise Suppression: Tough Challenge, Intelligent Implementation

Khalid Al-Rufaii* and Weihong Zhu
Saudi Aramco
*[email protected]

Land data, unlike marine data, are associated with complex surface and near-surface geology, which generate a suite of coherent and non-coherent noises. The suppression of such types of seismic noise has long been a great challenge for both 2-D and 3-D seismic data processing. This is because signal and noise share the same range of seismic characteristics such as velocity and frequency. In this presentation, we introduce a new methodology for noise suppression that is based on the principle of localized transforms. We demonstrate that utilizing localized transforms, unlike other conventional methods, ensures preserving the integrity of the seismic signal in terms of its amplitude and frequency contents. Our methodology focused on the suppression of mainly two types of seismic noise: linear and strong-energy randomly generated noise. Our approach is model-based and operates in 2-D and 3-D pre-stack modes. A key characteristic is that it tapers and scales-down the noise rather than muting it out. The linear noise is first suppressed by implementing a localized 3-D filter in the frequency-wave number (F-Kx-Ky) domain. Next, for each localized filtered data, an average amplitude spectrum is computed. Finally, a certain threshold value is assigned for trimming or scaling-down the linear noise, followed by an inverse transform back to the time-offset domain. These steps are repeated continuously with overlaps in time and space. The strong energy noise, being narrow band-limited, is tapered in a slightly different manner. The processes involve 1-D Fast Fourier Transform (FFT), decomposition via band-pass filtering, and median-filtering. The effectiveness of our proposed noise-suppression methodology is illustrated with both synthetic and field land data example.

AAPG Search and Discovery Article #90077©2008 GEO 2008 Middle East Conference and Exhibition, Manama, Bahrain