Search

The combination of ever-increasing computational power and more robust algorithms have made it possible to run full-waveform inversion (FWI) to higher frequencies and, also, offer more possibilities to take advantage of the reflections in the inversion. Through a process known as FWI Imaging, the detailed velocity models produced can be used to generate a reflectivity normal to the reflector plane. We outline the methodology and advantages of FWI Imaging, and introduce the concept of a dip-coherency image as an additional interpretation tool, using information parallel to the reflector plane. We show examples from the densely sampled source-over-spread Greater Castberg survey in the Barents Sea, demonstrating the uplift in the FWI Image over conventional imaging methods in terms of more balanced illumination and richer low frequencies. We performed decimation tests to assess the acquisition geometry impact on FWI imaging. Although the benefit of FWI imaging can still be observed on less well-sampled data, the best result remains with the original, densely sampled source-over-spread acquisition.

The Messinian interval in the West Nile Delta, offshore Egypt, is a thin evaporite layer characterized by highly irregular velocities with rapid spatial variations. Its complexity poses unique challenges for sub-Messinian reservoir imaging, resulting in erratic gather curvatures, distorted structures, and nonuniform illumination. Conventional ray-based tomography suffers from unreliable curvature picking due to poor gather quality, complex gather move-out, and inaccurate ray tracing through the fast and complex Messinian layer. Previous full-waveform inversion (FWI) had little success in this area because of strong amplitude mismatch between recorded data and modelled data at the Messinian layer and the limitations of the existing multi narrow azimuth (multi-NAZ) streamer data, which lacks good low frequencies and has limited offsets. We present a case study that utilizes a model building flow driven primarily by Time-lag FWI (TLFWI) starting from a tomography model with a reasonable long-wavelength velocity. TLFWI using both refraction and reflection data resolved the velocity errors in and below the Messinian interval and provided good uplifts to sub-Messinian reservoir imaging. In addition, least-squares Q-Kirchhoff (LSQ-Kir) migration with the improved velocity model compensated for irregular illumination and earth attenuation without over-boosting noise, thus further improved the S/N and resolution of the reservoir image.

The North Sea is a mature basin where it is increasingly challenging to make new and economically viable discoveries. Even after discoveries have been made, it is vitally important to have high confidence in the reservoir’s extent and spatial location before committing to a development plan. Here we detail the case history of a multi-azimuth pre-stack depth migration re-processing and re-imaging project over the Oda (formerly Butch) discovery in the Norwegian sector of the North Sea. The Oda structure is a complex salt diapir making it crucial to have as much confidence as possible in the position of the reservoir before making a drilling commitment. This study documents how the use of the latest imaging technologies, which include source and receiver deghosting, multi-azimuth full waveform inversion, multi-layer tomography and tomographic uncertainty analysis can aid in de-risking and improving reservoir definition, thereby increasing confidence in development well planning.

The geology of northern Oman presents significant challenges for land velocity model building. We show in this paper that these challenges can be overcome by using an integrated high-resolution velocity model workflow, through the combination of different types of waves, that allow resolving different parts of the velocity model. This dedicated workflow consists of Multi-Wave Inversion (MWI) for the near-surface, followed by Optimal Transport Full Waveform Inversion (OT-FWI) and then by ray-based joint reflected and diving wave tomography inversion. It resolves challenges imposed by complex shallow geology and allows for proper imaging of deeper structures. Compared to a conventional ray-based only model building flow, the integrated high-resolution workflow enabled generating a geologically plausible velocity model which minimizes depth positioning errors and greatly enhances structural and stratigraphic trends.

Inversion has become the standard procedure to quantify elastic properties using seismic. PP seismic is commonly used for this purpose, but in areas where PP seismic is affected by gas pockets or mud volcanoes, the reflectivity can be compromised for reservoir characterization. PP-PS inversion can step in to improve the determination of elastic properties when PP seismic inversion alone is challenging. However, the biggest challenge of this multi-component inversion lies in the registration between PP and PS information. This abstract illustrates the application of PP-PS elastic inversion using an improved technique that includes an innovative correction of travel time to rectify for residual time registration. This case study from a deep-water field in West Africa was carried out on acquired ocean-bottom nodes (OBN) dataset. The seismic preconditioning, PP inversion, registration approach and the implementation of the advanced PP-PS inversion helped in the better characterization of elastic properties of the reservoir.

Pure S-wave (SS) seismic data have the potential to bring significant uplift to seismic imaging and reservoir characterization. Combining PP, PS and/or SS data in a joint inversion for seismic reservoir characterization presents some theoretical advantages for the estimation of shear-velocity and density related reservoir properties. The inversion stability of the various combinations of seismic modes is compared using a condition number analysis. The biggest challenge of joint inversion comes from the difference in travel times between the seismic modes. To overcome this issue, the joint inversion is performed in the native time domain where the travel times difference between the seismic modes is optimized during the inversion.

Land seismic presents more and different challenges for Full-Waveform Inversion (FWI) than marine data. Among these challenges, the most fundamental ones are the irregular topography, strong nearsurface effects, and common FWI difficulties such as cycle-skipping and amplitude issues. In this work, we propose to deploy three strategies to deal with these difficulties respectively: curvilinear topography modelling to effectively model the irregular topography for the earth’s surface, mitigation of near-surface effects to reduce the negative impacts of strong near-surface noise, and a stable cost function as the foundation for land FWI to alleviate cycle-skipping and amplitude issues. We demonstrate the effectiveness of our strategies with two field data examples representing different geological settings. Based on learnings from these studies, we believe that land FWI is becoming more stable and consistent than before, and success can be expected on more land datasets.

The migrated dip-angle domain provides a powerful opportunity to distinguish 4D noise from signal based on similarity filtering applied to data decomposed by position, frequency, and geological dip. However, 4D signal protection is problematic when the signal itself forms from differences between baseline and monitor. A set of dip-angle similarity filtering methods applied to towed streamer and OBN data from South Arne field show that 4D signal preservation is possible even with strong timeshift signals between baseline and monitor. Signal protection can be achieved with wrap-around timewarping applied within the filtering methods. A better approach detects lack of coherent signal rather than similarity of coherent signal when present. Using this method it is possible to attenuate significant levels of migration noise without appreciably altering the 4D signal. Dip-angle filtering with workflows that preserve surface offset also allow the similarity filtering to be combined with least-squares Kirchhoff migration using single-iteration migration deconvolution. Results show noise attenuation via similarity filtering complementing the illumination compensation achieved by the least-squares method.

Marine seismic surveys in shallow water regions typically suffer from acquisition striping and poor shallow resolution. Multiple imaging has been discussed in the literature for several years as a processing-based approach to this problem. We compare least-squares wave-equation multiple migration (LS-WEMM) results for towed-streamer and ocean bottom node (OBN) data sets co-located in the Central North Sea. With either data type, LS-WEMM results reduce acquisition striping and improve spatial resolution, when compared with primary imaging. Further, we observe that the towed-streamer LS-WEMM result provides better lateral resolution than the OBN LS-WEMM result, whereas the OBN LS-WEMM result provides better image illumination overall. We propose a LS-WEMM method constrained jointly by towed-streamer and OBN data, which is shown to combine the benefits offered by both data types.