Improving Reservoir Characterization with Seismic Data Preconditioning - A Case Study of Saudi Arabian Channel Detection

Al-Dossary, Saleh (Saudi Aramco)

OnePetro 

Abstract

Random seismic noise, present in every authentic seismic data set, hampers both geoscientists’ manual interpretation of data and computerized delineation and analysis of seismic features. Therefore, many noise suppression techniques have the goal of preserving image quality. For channel detection, accurately suppressing seismic noise without damaging image detail is crucial. Theoretically, channel patterns can be automatically detected owing to their unique spatial footprint, which differentiates them from other three-dimensional seismic features. One notable characteristic of channels is their local linearity: Their spatial extent is much greater in one direction than in any other direction. A variety of techniques, such as spatial filters, can enhance the “slender” characteristic of channels. Unfortunately, most of these techniques reduce noise by smoothing the data, resulting in a loss of edge definition. During the past few years the literature has revealed several new edge-preserving noise reduction techniques, including edge-preserving smoothing and complex wavelet transforms. In this case study I illustrate the performance of edge-preserving smoothing based on the redundant wavelet transform (RWT) and demonstrate its usefulness before running an edge detection algorithm to reveal channel patterns in seismic data from Saudi Arabia. Our examples demonstrate that RWT can successfully preserve, enhance, and delineate channel edges that are otherwise not readily visible on conventional seismic amplitude displays.

Introduction

Channel detection is an important part of seismic interpretation for oil and gas exploration. In theory, channel patterns should be possible to detect automatically owing to their unique spatial footprints, which differentiate them from other features encountered in 3D seismic data. One notable characteristic of channels is their local linearity; that is, their extent is much greater in one direction than in any other direction. Traces of a channel can typically be seen on a seismic time slice.

Spatial filters are commonly used to remove noise from seismic data. Unfortunately, most smooth the data to reduce noise (Jervis 2006). During the past few years researchers have developed several new 3D seismic noise reduction techniques that preserve edges, including edge-preserving smoothing (Blumentritt et al. 2003) and complex wavelet transforms (Jervis 2006).

Here I use the same mechanism Al-Dossary and Ananos (2012) used to demonstrate the effects of applying the redundant wavelet transform (RWT) (Shensa 1992) before edge detection and subsequent channel identification.

Al-Dossary and Ananos (2012) demonstrated the usefulness of RWT before edge detection to enhance channel patterns in seismic data. In this paper I apply the concept to data collected from Saudi Arabia. The RWT is a type of discrete wavelet transform (DWT). However, it differs from the standard DWTs in that it does not carry out decimation or subsampling at successive resolution levels; rather, RWT decomposes the data into low-frequency information (approximation) and high-frequency information (wavelet coefficients) to obtain a projective decomposition of the data into different scales. RWT can be used for noise reduction in image processing, texture classification, and image fusion. RWT’s advantage in feature characterization lies in its pixel-wise analysis of images without performing image decimation. After application of RWT, locally linear features appear at adjacent scale levels, whereas non-significant features such as random noise slowly decrease as the scale level increases. I demonstrate RWT’s effectiveness by applying it to seismic data over a channel from Saudi Arabia. The input data are a time slice extracted from the Hawtah field in central Saudi Arabia.