True 3D Seismic Imaging: Which Acquisition and Processing Approaches are Best?

Abstract

Abstract 3D seismic imaging attempts to reconstruct a continuous image of the subsurface based upon discretely sampled reflectivity traces. The spatial sampling is typically irregular, coarse, and may even be undersampled according to the classic Shannon-Nyquist sampling theory. In addition to primary reflections, unwanted energy in form of coherent noise is recorded in many forms, and is typically more prone to spatial aliasing than the primary signal. As a consequence, traditional processing approaches follow a flow that may be collectively referred to as pre-conditioning. Pre-conditioning removes coherent and incoherent noise, modifies the phase of the data, and attempts to interpolate/regularize/reconstruct the data in a band-limited manner. Velocity model estimation and imaging are then applied, followed by additional noise removal and signal enhancement as required. Attention has moved in recent years to so-called broadband seismic imaging. A variety of deghosting solutions have emerged for removing the effects of both the source-side and receiver-side ghost effects within towed streamer seismic data. Attendant claims are made about whether such solutions are 2D, have azimuth-specific assumptions, are 3D, or even "true 3D". So how do we measure whether the 3D broadband solution obtained by a particular acquisition-processing-imaging methodology is really "true 3D" in a broader imaging context? This question is complicated recently by imaging methods that exploit the illumination from surface multiples in addition to the illumination from primary reflections. I introduce a frame of reference that begins with 3D towed dual-sensor streamer marine seismic acquisition and dual-source shooting. There is no theoretical limit to how many orders of the corresponding surface multiple wavefield may be sampled for a given streamer separation. I demonstrate that by pursuing dual-sensor wavefield separation at the first stage of pre-conditioning and sampling densely enough in the cross-line direction, an uncompromised 3D image of the earth may be achieved for most target depths without any loss of temporal resolution in the frequency range afforded by natural attenuation to each respective target depth. In the case of shallow geology affected by the imprint of the acquisition geometry, imaging with surface multiples will typically yield a better result than any approach based solely on primary reflections. I differentiate between the most significant approaches used for pre-conditioning data during deghosting, and provide a path towards an optimal ghost-free 3D seismic image of the earth. Introduction As seismic interpretation progresses from the exploration to development phases, attention moves from structural interpretation, estimating rock volumes, and optimizing well locations, to more detailed pursuits such as identifying reservoir compartmentalization, mapping fault and fracture details, optimizing well trajectories, estimating reservoir and fluid properties, and even geohazard identification. Collectively, pre-stack seismic gathers become the critical data, both as inputs to pre-stack (depth) velocity model building and imaging, and as inputs to quantitative interpretation (QI). This paper views the data result after pre-stack imaging as the true test of temporal/vertical and spatial/horizontal resolution, and of amplitude and phase fidelity.