Search

Driven by the notion that blending noise may materially increase the background noise level and obscure the interpretation of weak time-lapse (4D) signals related to subtle reservoir changes, the industry has not yet seen any simultaneous-source (sim-source) surveys acquired for reservoir monitoring. Thus, whether sim-source acquisition is feasible for 4D remains a long-standing question. In 2019, BP took the step forward to acquire the first sim-source ocean bottom node (OBN) monitor survey at the Atlantis field in the Gulf of Mexico (GoM). With 4D-friendly deblending and matching of sources from different vessels, we were able to mitigate the challenges associated with the 2019 independent simultaneous source (ISS) OBN survey and obtain a similar level of 4D signal-to-noise ratio (S/N) and valuable 4D signals comparable to what we could achieve with conventional OBN data. Further, meaningful subsalt 4D signals were revealed for the first time in the areas with fairly poor illumination even with the 2019 ISS OBN survey, partly due to the larger reservoir changes from a longer production history and a more accurate velocity built from full-waveform inversion (FWI).

Shear wave velocity model building (S-wave VMB) is a critical and difficult processing step for converted wave imaging. Conventional S-wave VMB depends on PP-PS joint interpretation-based image registration and PP-PS joint tomography-based residual moveout flattening which have certain advantages and drawbacks. We introduce PS reflection FWI (PS-RFWI) for S-wave VMB in the presence of a complex, heterogeneous subsurface, which can address some of the concerns with conventional approaches. This method assumes that we have previously obtained sufficiently accurate pressure wave (P-wave) velocities and reflectivity. PS-RFWI solely updates S-wave velocity by minimizing the kinematic difference between the modeled and recorded PS reflections, while leaving the P-wave parameters unchanged. In conjunction with conventional methods, PS-RFWI can provide a superior PS image, as we will demonstrate with a field data example from offshore Malaysia.

Although the resolution of a seismic image is ultimately bound by the spatial and temporal sampling of the acquired seismic data, the seismic images obtained through conventional imaging methods normally fall far short of this limit. In addition to attenuation in the Earth, factors such as velocity errors, illumination holes, residual noise and multiples, source and receiver ghost notches, and migration swings can prevent conventional imaging methods from obtaining a high-resolution image of good signal-to-noise ratio (S/N) and well-focused details as promised by the maximum migration frequency. Recently, FWI Imaging, which uses the full-wavefield data to iteratively invert for the reflectivity together with velocity through full-waveform inversion (FWI), has shown to be a superior method for providing seismic images of greatly improved illumination, S/N, focusing, and thus better resolution, over conventional imaging methods. Here, we push FWI Imaging to a frequency close to the temporal resolution limit of seismic data (100 Hz) and demonstrate that FWI Imaging at a very high frequency can provide seismic images of unprecedented resolution from the recorded data, which has been impossible to achieve by other seismic imaging approaches.

Full-waveform inversion has been established as a standard tool for building high-resolution velocity models. To take full advantage of such models, the migration algorithm must be capable of handling fine-scale geo-bodies and sharp contrasts while affordably producing high-frequency migration stacks and gathers. Even though ray-based Kirchhoff migration can efficiently generate high-resolution migration stacks and gathers, the calculation of traveltimes becomes inaccurate and unstable near large velocity variations, sharp contrasts, and complex structures. Reverse[1]time migration (RTM), on the other hand, can accurately handle complex velocity models with fine details and sharp contrasts due to its deployment of full-wavefield propagation. However, the cost of RTM becomes prohibitive when high-frequency stacks and gathers are required. Following this idea of wave-equation-based traveltimes, we propose a wave-equation Kirchhoff (WEK) scheme that performs Kirchhoff migration using maximum-amplitude traveltimes and amplitudes from the wavefield. These traveltimes and amplitudes are computed through affordable low-frequency full-wavefield propagation. WEK not only partly inherits the benefit of full-wavefield propagation for high-resolution models, but it also maintains the affordability of ray-based Kirchhoff migration. We use synthetic and field data to evaluate this method and compare the WEK results with those from ray-based Kirchhoff migration and RTM.

High-quality depth images typically require accurate high resolution representations of the earth model. Full-waveform inversion (FWI) has recently justified its value throughout the industry in providing high-resolution velocity models. However, obtaining an accurate FWI velocity model using narrow-azimuth streamer data can still be challenging in complex geologic environments. The uncertainties, often caused by relatively low resolution perpendicular to the shooting direction and weaker illuminations areas, instill less confidence for reservoir delineation and depth mapping. In this case study from offshore Senegal, we present a joint velocity (Vp) and epsilon (?) multi-azimuth FWI workflow to construct a high-resolution model to overcome severe image distortion. The updated model improved event focusing and gather flatness and demonstrated significant imaging uplifts consistent with our understanding of the geology in the area.

Recently, land full-waveform inversion (FWI) has shown great potential in resolving near-surface complexity in the Delaware Basin, providing significant imaging uplift and useful information for shallow hazard identification. However, deep section updates beyond diving wave penetration remain challenging. We present an application of land FWI in the Central Basin Platform (CBP) for both shallow and deep updates. Results show that, with a time-lag cost function, a fine spatially sampled data set with proper preconditioning, and a good starting model for regions beyond diving wave penetration, land FWI was able to produce a high-resolution velocity model to resolve small-scale anomalies in the deep sections as well as detailed velocities in the near surface, leading to improved seismic images at reservoir levels. Furthermore, the impact of the FWI input data spatial sampling and the starting model in inversion are studied respectively.

Machine learning (ML) has garnered great attention within the field of seismic processing due to its vast achievements for quality and efficiency in the area of computer vision. Recent academic papers have demonstrated some potential for the use of machine learning in processing seismic signal, such as random and coherent noise removal, deblending, and interpolation. In this paper, we illustrate some uses of ML on real 3D seismic data and discuss the common challenges that need to be addressed in order to fulfill the promises of the deep neural network (DNN) for seismic processing. We also point out that, in some cases, the result of ML could be good enough for some fit-for-purpose applications. Finally, we summarize a few learnings based on our research and experiences in both the seismic processing and ML worlds.

A new method is outlined that allows you to use isotropic seismic modeling and inversion technology in an anisotropic setting. The method is based on the mapping of the isotropic elastic parameters into effective elastic parameters, appropriate for the expected type of anisotropy. This paper outlines the transforms for VTI and HTI, but the method is extensible to TTI and other types of anisotropy. Examples show that the isotropic modeling with the effective elastic parameters produces the same results as full anisotropic modeling. Results from isotropic pre-stack seismic inversion can be predicted and analyzed using this technology.

CGG has always been at the forefront of industrial High Perfor­mance Computing (HPC) architectures: we were operating vector supercomputers (Convex, Cray and NEC) in the early 1990s, and large parallel supercomputers (Convex SPP, IBM SP, Sgi Origin) by the end of that decade. At the turn of the millennium, we were pioneering the use of commodity clusters, and started to add accelerators a couple of years later, even before GPGPU programming languages formally emerged.