Search

High-resolution (HR) site survey acquisitions are traditionally utilized for shallow geohazard investigations and infrastructure planning. However, due to cost reasons, these are often acquired in a sparse 2D manner, and supplemented with conventional multi-streamer 3D seismic imaging aimed at deeper targets. We show how products derived from 100 Hz dual-azimuth (DAZ) full-waveform inversion (FWI) using multi- streamer 3D data, including FWI Imaging, provide a superior uplift in spatial resolution and illumination compared to a combination of both conventional 3D imaging using the same data, and 2DHR site survey data. This is shown to be the case for data and attributes compared against those from a site survey report at various near surface intervals where known geohazards occur. The improved spatial resolution with 100 Hz DAZ FWI can reduce uncertainties in predicting hazards and act as rapidly available supplementary information to a sparser 2DHR survey with less need to acquire a denser 3DHR survey.

Time-lapse (4D) surveys have traditionally been reliant on baseline (base) surveys being well repeated by the monitor to decrease 4D noise. In this case study the monitor was acquired independently from the 3D narrow-azimuth, towed-streamer, hydrophone-only base, and “mixes” two different types of acquisition: a multi-sensor, towed-streamer acquisition for prime coverage and a multi-sensor towed-streamer infill. We describe technologies used to overcome the limitations in 4D repeatability and intra-monitor consistency. 3D Ghost Wavefield Elimination and a novel blind signature inversion method were crucial to create a seamless monitor across the target and to reconcile the signature and ghost differences between the base and monitor. Inconsistent azimuth content between base and monitors introduced complexities for the demultiple process; nevertheless, multiples were successfully attenuated using wave equation deconvolution with joint base and monitor reflectivity imaging. Residual non-repeated 4D noise was attenuated using a curvelet domain 4D co-operative denoise workflow.

Short period multiple prediction for land data is challenging due to poor imaging of the shallow multiple generators as well little information about the down-going reflection at the weathering layer. Based on multiple imaging of the shallow section, surface-related wave-equation deconvolution has been used in recent years to improve multiple prediction in such areas. We improve the accuracy of wave-equation deconvolution to include an angle dependency of the multiple generator reflectivity. In addition, we modify the approach to handle internal multiple predictions where the lower-generator is provided by surface-related wave-equation deconvolution, and the upper-reflectivity is derived through least-squares inversion. The combined benefit of these two approaches is demonstrated on a land dataset from south Oman.

In recent years, the Brazilian presalt characterization has been a challenging task, either because of the imaging problems associated with complex salt layer geometry (Penna et al., 2019) or due to large heterogeneity in the presalt carbonates reservoir (Oliveira et al., 2018). Some efforts to address these challenges include full-azimuthal acquisitions technologies (nodes acquisition) and new processing techniques and studies to better understanding the geological model. Recent multi-azimuthal seismic data acquisition provides not only better seismic imaging, but also fractures properties through velocity and amplitude changes with azimuth variation, providing a better spatial characterization of the fracture system. We present an azimuthal elastic seismic inversion (Roure et al., 2012) in a Nodes presalt reservoir data, addressing the local fracture system characterization. Fracture parameters are described in terms of the normal and tangential weakness plus fracture strike. Prior to the inversion, an azimuthal seismic preconditioning has also performed on the seismic dataset to attenuate the imaging problems in the reservoir interval and consequently improving the signal-to-noise ratio providing better estimates of the properties of interest. Fracture characterization studies commonly integrates borehole image logs, core data and well test, usually restrict to specific areas around drilled wells. In some cases, tridimensional fracture models are created through geostatistical modelling constrained by seismic attributes. However, those are indirect measures of fractures, with high associated uncertainties. Recent multi-azimuthal seismic data acquisition provides not only better seismic imaging, but also fractures properties through velocity and amplitude changes with azimuth variation, providing a better spatial characterization of the fracture system. We present an azimuthal elastic seismic inversion (Roure et al., 2012) in a Nodes presalt reservoir data, addressing the local fracture system characterization. Fracture parameters are described in terms of the normal and tangential weakness plus fracture strike. Prior to the inversion, an azimuthal seismic preconditioning has also performed on the seismic dataset to attenuate the imaging problems in the reservoir interval and consequently improving the signal-to-noise ratio providing better estimates of the properties of interest.

The imaging of the complex fault system plays an important role in hydrocarbon exploration in Wenchang field since the fault system forms a bridge between the source rocks and reservoirs. However, it is challenging to obtain a high quality depth image of the fault system due to the complex depth velocity and Q absorption effect. In this paper, we demonstrate how a combination of Fault Constraint Tomography (FCT) model building flow and Q-compensated High Fidelity Controlled Beam Migration (QHFCBM) work together to provide a step change in the imaging quality and bring significant impact to the reservoir delineation.

The merits of Total and Effective porosity approaches have always been a source of discussion within the Petrophysical community globally. In general, an operating company adopts a single approach (Total or Effective) in their modelling workflows and ignores the alternative method. This is normally to have a consistent approach across the company so that the end users know what they are receiving into their subsequent workflows. In the proposed method, both Total and Effective Porosity Methods have been applied. The differences are then used to minimise and improve the resulting porosity and saturation calculations such that the results are mutually comparable. Total Porosity based water saturation equations are dependent on the ‘shale/clay’ volume and porosity to compensate for the ‘shale/clay’ bound water resistivity. Effective Porosity is based on water saturation equations on the shale volume and resistivity. The difference is that the Effective Porosity Water Saturation approach is not directly dependent on the ‘shale/clay’ porosity and can be used as a fitting parameter via the dry clay density (that doesn’t exist in-situ) in addition to compensating for the invaded fluid volume. Examples will be presented in a wider range of geological environments will be discussed.

In 4D (time-lapse) imaging, the primary objective is to obtain a 4D image that is sufficiently free of noise associated with acquisition and processing in order to understand the changes at the reservoir interval. To obtain accurate 4D products, all seismic processing steps must effectively mitigate the differences arising from inconsistencies in acquisition setup and recording conditions of the time-lapse surveys. One of the key steps of the 4D processing sequence is deghosting, which is commonly used to remove the ghost variations between the baseline and monitor vintages. Deghosting can eliminate the variations in source and receiver tow depths and produce normalized broadband datasets at the same datum (mean sea level), which serve as input for further co-processing. We present a 4D multi-sensor deghosting algorithm as an extension of recent deghosting technologies and apply it to a deep-tow multi-sensor streamer time-lapse survey acquired over the Liza field located offshore Guyana. The 4D image, which uses the proposed approach, shows significantly reduced 4D noise compared to result obtained with 3D deghosting of each vintage separately.

Geological bodies like sand injectites are found in numerous geodynamic and geological contexts around the world, and they can manifest as low- or high-amplitude seismic responses with complex structures. In this paper, we demonstrated an amplitude preserving DNN-based workflow for injectite detection using customized U-Net. Our workflow addresses the challenges inherent in the limited number of training datasets and produces a pretrained model that delineates injectite events on migrated seismic images. To address the issue of domain shifting, we proposed a transfer learning approach that avoids mis-predicting faults and other diffraction events as injectites. Finally, we discussed how this result could benefit the seismic processing workflow.

Seismic wavefields traveling through the subsurface lose energy due to various phenomena, such as geometric divergence and intrinsic attenuation. Imaging algorithms that account for these effects are well established. However, an often overlooked aspect of seismic imaging is transmission loss, which occurs due to propagating wavefields losing energy to back-scattered reflections. In this study, we examine the impact of transmission loss on various imaging algorithms and argue that it is not compensated for in standard applications of RTM and LS-RTM. Moreover, we find that transmission loss is naturally corrected for in the high-resolution reflectivity models derived from FWI Imaging, in which the effect is automatically encoded via sharp contrasts in velocity or density added to the earth model by the inversion. We assess the significance of transmission loss in different geological scenarios and evaluate the relative importance of this effect compared to other mechanisms that dissipate energy in the subsurface.