Search

Full-waveform inversion (FWI), proposed by Lailly and Tarantola in the 1980s, is considered the most promising data-driven tool to automatically build velocity models. Many successful examples have been reported using FWI to update shallow sediments, gas pockets, and mud volcanoes. However, successful applications of FWI to update salt structures had almost only been seen on synthetic data until recent progress at the Atlantis field, Gulf of Mexico (GoM). We revisited some aspects of FWI algorithms to minimize cycle-skipping and amplitude discrepancy issues and derived an FWI algorithm that is able to build complex salt velocity models. We applied this algorithm to a variety of data sets including WAZ (wide-azimuth) and FAZ (full-azimuth) streamer data as well as OBN (ocean bottom node) data with different geologic settings in order to: 1) demonstrate the effectiveness of the method for salt velocity updates, and 2) examine some fundamentals of the salt problem. We observe that in multiple cases, salt velocity models from this FWI produce subsalt images of superior quality. We demonstrate with one FAZ streamer data example in Keathley Canyon that we probably do not need very high frequency in FWI for subsalt imaging purposes. Based on this observation, we envision that sparse node for velocity (NFV) acquisition may provide appropriate data to handle large and complex salt bodies with FWI. We believe that the combination of advanced FWI algorithms and appropriate data acquisition will bring a step-change to subsalt imaging.

The paper presents a review of classic rock physics models used for clastic sedimentary rocks and how they have been combined into extended models through the introduction of a few parameters associated with a compositional or textural property of the rock. The models are used on a variety of real data sets to showcase how rock properties can be inferred from elastic properties.

In this paper, we present a high-resolution FWI case study from the foothills of the Southern Carpathians, in Romania. The input data was acquired in 2022, mostly with dynamite source, and designed for optimal subsurface wide azimuthal illumination. The varied rough terrain and the presence of a complex thrust body represent the main imaging challenges. Firstly, we will show how the combination of Multi-Wave Inversion (MWI), and Full-Waveform Inversion (FWI) enabled us to construct a high-resolution near-surface velocity model that solved some important imaging distortions. Next, we will outline the benefit of inverting data to 50Hz to attain a high resolution FWI model which fully captures the strong lateral and vertical velocity contrasts of the upper folded structures and the underlying stratigraphy. Finally, the migrated image and derived reflectivity from this 50Hz FWI model has helped to de-risk the reservoir uncertainties in both size and positioning of the crest. These enhance the imaging to better define the spill point of the prospect.

Well drilling and abandonment work offshore Brunei is very challenging due to the presence of complex shallow geological features, such as corals, river channels, gas clouds/chimneys, and small faults. Therefore high-resolution seismic images, which can reveal those features, are desired to help avoid potential geohazards. In this study, Least-Squares Wave-Equation Multiple Migration (LSWEMM) was used to produce 3D high-resolution shallow images. The new results have both high spatial and temporal resolution. The near water bottom geological features are more clearly revealed than on conventional primary images. The resolution of fine channels and small faults is also greatly improved compared to nearfield hydrophone (NFH) images. The new mapping of shallow channels is very helpful for ongoing geohazard assessment work near a planned well. LSWEMM was applied on existing OBN data, hence is much more cost effective compared with an additional survey, such as 3D P-cable acquisition. Moreover, decimation studies show great value in optimizing future OBN acquisition and 2D high-resolution survey design, which could further reduce the cost associated with geohazard assessment work.

Deriving an accurate high-resolution seismic quality factor (Q) model is necessary to both compensate for the phase dispersion and amplitude attenuation effects of the Earth’s anelasticity during imaging, and to reduce parameter cross-talk during velocity model building (VMB). Various methods exist for deriving Q, each with potential inherent limitations that can restrict their suitability for scalable applications, such as over the ?14,000 km2 area presented here in the Norwegian North Sea. We describe a multi-process Q derivation workflow, including dual-azimuth Q full-waveform inversion (Q-FWI), time-lag FWI guided ray-based Q tomography, and well-synthetic validations using Q sensitive metrics to derive a regional high-resolution Q model. When incorporated into VMB and subsequent Q migrations, the regional Q model is shown to both locally and regionally compensate for amplitude attenuation and phase dispersion, and providing an associated improvement in imaging.

We present a 4D pre-salt reservoir monitoring study from two wide-azimuth towed-streamer (WATS) surveys. Advanced flows were implemented to mitigate complex salt-related challenges and WATS repeatability issues in the image domain. Three models of converted waves interfering at reservoir level were generated via dual salt-flood Kirchhoff demigrations, and then subtracted at the post-migration stage. Repeatability issues were also addressed in the image domain via a novel 4D Least-Squares Wave-Equation Kirchhoff flow. The results reveal an unexpected level of 4D signal in such a complex geological setting.

Much recent work has been done on comparing machine learning and geophysical inversion techniques to the extraction of model parameters from seismic reflection data. In our profession we are used to analyzing the physics of geophysical problems in detail. However, in many of the recent studies the machine learning algorithms are treated almost as “black boxes”. In this study l I will use a straightforward numerical example to illustrate the difference between geophysical inversion and machine learning inversion. In doing so I will try to “demystify” machine learning algorithms and show that, like inverse problems, they have a definite mathematical structure that can be written down and understood. The example used is this tutorial is the extraction of the underlying reflection coefficients from an overlapping wavelet response that was created by convolving a reflection coefficient dipole with a symmetric wavelet. In discussing the solution to this problem I will cover the topics of deconvolution, recursive inversion, linear regression and nonlinear regression using a feedforward neural network. I will present both the full inverse approach as well as gradient descent algorithms, which can be applied to both linear and nonlinear problems. This will lead to a description of the backpropagation algorithm, which is used to train a feedforward neural network. In the final section of the tutorial I will look at the impact of local minima in the search for a global minimum in the backpropagation algorithm.

The hunt for less obvious deeper targets below the Base Cretaceous Unconformity (BCU) within the Norwegian North Sea greatly relies on the accuracy of the velocity model as it impacts the definition of the structural traps. The presence of the limestone/carbonate sequence with high velocities overlying the targeted lower velocity mudstone units represents the main challenge in term of velocity model building and is usually out of the reach of diving waves for the full waveform inversion (FWI) application when using streamer-based data with offset limited to 8km. The recent shift of acquisition industry toward hybrid acquisition combining streamer and sparse node opens the road for deeper application of FWI and even more. In this paper, we show how this new hybrid acquisition design can help to build a reliable high-resolution velocity model down to the Brent level. Moreover, we also present how elastic FWI enables us to better explain the elastic effects induced by the large impedance contrast at BCU level.

Quantifying and classifying pore systems in carbonates is notoriously challenging, particularly in rocks associated with complex diagenetic histories. Here we report novel computed tomography (CT) core and thin section image analyses through key reservoir intervals in the Buah Formation and Khufai Formation, both part of the Precambrian (Ediacaran) Nafun Group, from two wells located onshore Oman. Our primary objective is to constrain the volume, shape, connectivity, and distribution of vugs down-core in two and three dimensions – this is a key control on reservoir quality. We combine classic sedimentological core descriptions with image analysis on a range of data types. In this paper we focus on the analysis of core CT scan data, but we also introduce a high-level analysis of thin section images from discrete samples from the same boreholes, where available (sidewall core plugs, ditch cuttings and conventional cores). Finally, we integrated the results to (1) provide a holistic understanding of pore systems in the Buah and Khufai Formations; (2) identify the key uncertainties and weaknesses in our approach, and (3) plan for further reservoir assessment.