Pre-Show Focus | Improving Today For a Better Tomorrow

Andrew Long provides a preview of PGS technical talks in the context of the wider EAGE technical program. 

wave imagewave image

Less Carbon in More Ways Than One: EAGE 2022 Preview

The annual EAGE conference returns to Madrid in a more compact form than the pre-Covid heyday, with no more than six simultaneous sessions at any given time. Abstract submissions were reportedly only about half the volume of bygone years, but the content remains strong, and the EAGE is clearly the professional geoscience organization doing the most to stay on top of several seismic shifts in the global energy landscape. The online program can be found here.

Curiously, for a world transitioning to more near-field exploration, the exploitation of proven oil and gas assets, and the proliferation of engineering-led sources of renewable energy, it feels like the engineering component of the technical program has been cut back the most. The theme that has grown as expected is that of low carbon / CCS / hydrogen / geothermal. After something of a period of consolidation and rather mundane case studies in the last couple of years, machine learning and AI applications have also made a strong comeback with several sessions being explicitly devoted to productive augmented and automated workflows and solutions. Still strongly represented on equal footing with various ‘traditional’ subsurface interpretation and characterization sessions, FWI / velocity model building / high-end seismic imaging is again prominent, with FWI increasingly now playing a role in quantitative interpretation and subsurface characterization workflows.

I will provide some commentary from the EAGE 2022 event once the show begins, but in the meantime, I summarize several excellent PGS contributions below, including links to the PDF technical abstracts.


FWI: Removing Obstacles and Creating Opportunities

PGS will give two presentations on ways to reconstruct the low-frequency content of FWI (Full Waveform Inversion) so that traditional risks of cycle skipping are reduced, and model convergence is faster and more stable. First, Maiza Bekara will present “A new look at autoregressive low-frequency reconstruction of seismic data”, a computationally efficient solution that follows a conventional signal processing approach rather than relying upon cost offline network training. Second, Ramzi Djebbi will present “Full Waveform Inversion with low frequency reconstructed data”, which describes how a recursive filter estimated from the high-frequency content of the data is used to reconstruct the low frequencies. Both methods have proven robust.

Two case studies showcase the efficacy of modern FWI solutions in large-scale 3D projects. Jyoti Kumar will present “Imaging pre-Messinian targets in the East Mediterranean Sea - A case study using FWI”, wherein a velocity model building workflow using FWI and RTM gathers has successfully captured historically-elusive post-Messinian and-Messinian complexities. Antonio Castiello will present “Integrating FWI and reflection tomography to rejuvenate legacy seismic data: an example from the Faroes-Shetland Basin”, a robust velocity model building workflow that leverages FWI updates and geologically constrained reflection tomography to quite seamlessly rejuvenate 38 seismic datasets, resolve sharp lateral and vertical velocity variations in the overburden, and enhance the imaging quality of the targets associated intra-volcanic and intrusive features.

PGS launched a breakthrough integration of FWI and Least-Squares Migration (LSM) in 2021, and Øystein Korsmo will demonstrate the power of the solution in “Imaging by seismic inversion based on the adjoint state method”. A new nonlinear data-domain LS-RTM, implemented using the adjoint state method, can invert for the reflectivity while simultaneously refining the FWI velocity model. Critically, reflectivity changes caused by density variations are not erroneously mapped as velocity updates, and the reflectivity imaging results show significant structural improvements, more focus, and better fault imaging compared to traditional RTM.


More (and New) Sources, and More Azimuths

The most significant development for towed-streamer seismic in recent years has been the commercialization of wide-tow multi-source shooting, with up to six sources towed with a combined separation of almost 450 m. Martin Widmaier will present a workshop (W12) talk titled “Recent advances with wide-tow multi-sources in marine seismic acquisition and imaging”, which charts the engineering achievements that have enabled highly efficient acquisition configurations to also deliver remarkable spatial sampling and near-surface image resolution and quality previously unimaginable. Now that wide-tow multi-source shooting is standard across the PGS fleet, novel streamer configurations are starting to emerge too. Correspondingly, Martin Widmaier will also present “Combining wide-tow multi-sources with a non-uniform streamer configuration: A case study from the Sarawak Basin”, wherein wide-tow multi-sources with a non-uniform streamer separation were deployed for the first time in a seismic survey in offshore Malaysia.

Stian Hegna dispenses with traditional seismic sources altogether in “The acoustic wavefield generated by a vessel sailing on top of a streamer spread”, and instead uses the broadband (4-250 Hz) acoustic wavefield generated by a seismic vessel to image the Earth. The test was part of a world-record source-over-streamer survey towing six sources in wide-tow mode. Encouragingly for the future of this novel approach in areas where active seismic sources are not permissible, the acoustic wavefield emitted by the vessel appears to be very broadband and almost omnidirectional with only minor variations related to emission angle.

Multi-Azimuth (MAZ) shooting has been applied for two decades now, benefits from the applicability of traditional imaging workflows, and the rejuvenation of legacy datasets via the acquisition of a new survey azimuth(s) is becoming commonplace in many regions. Roberto Ruiz and Cyrille both co-author two quantitative interpretation case studies that also showcase the value of jointly inverting the pre-stack data from each azimuth to create opportunities for near-field exploration. First, Roberto will present “Mature Paleocene South Viking Graben play derisked with multi-azimuth Seismic data, a Norwegian case study”, and demonstrates how the integration of MAZ broadband seismic data in AVA analyses can help discriminate the response of volcanic tuffs facies from deep-water marine sandstone facies at the Paleocene level in the prolific South Viking Graben. In a complementary presentation, Cyrille Reiser will use “Data-driven reservoir properties estimation using MAZ towed multisensor streamer seismic: A Norwegian case study” to go a step further by estimating reservoir properties such as volume of shale and porosity. Of other significance, no low-frequency model is necessary to achieve absolute reservoir properties.

Automation Creating Value for Subsurface Characterization

The First Break paper by Per Avseth and others from Dig Science and PGS in 2020 titled “Exploration workflow for real-time modeling of rock property and AVO feasibilities in areas with complex burial history — a Barents Sea demonstration” showcased the value of burial-constrained rock physics modeling to generate AVO feasibility cubes. The workflow included low-frequency model building based upon FWI velocities and enabled efficient and geologically consistent DHI derisking of leads and prospects in any geological context. In another collaboration with PGS, Per Avseth will use “Combined burial and rock physics modeling to explore the complex velocity and AVO depth trends offshore Canada” to showcase how such a workflow in the Tablelands and North Tablelands areas of offshore Canada can be adapted to accommodate a natural ambiguity between the temperature and burial histories of the region. The multiple rifting history and complex tectonics during late Jurassic/early Cretaceous make it challenging to separate burial from thermal effects on elastic properties. Nevertheless, AVO scenario cubes were successfully generated to screen and de-risk leads and prospects in the area from calibrated inversion data. Such flexibility is particularly applicable when high-quality petrophysical data calibrations are available. In a workshop (W4) presentation Cyrille Reiser will show the successful application of machine learning to predicting incomplete data in “Data mining for prediction of petrophysical properties from well logs”.

The use of integrated G&G screening workflow to consider both the container and containment aspects of CCS will be presented by Noemie Pernin in “Integrated workflow for characterization of CO2 subsurface storage sites”. The site assessment workflow allows the validation of various technologies on a local scale, with the option and feasibility to be expanded regionally. As the industry profile of CCS grows rapidly in coming years, efficient and robust platforms will be essential to manage the proliferation of data becoming available courtesy of the ways that digital transformation is starting to change everything we do.

Accordingly, two PGS papers will address fundamental ways to use deep learning / dictionary learning to streamline and optimize traditional signal processing workflows. Bagher Farmani will illustrate how deep learning models trained from multisensor streamer data can be used to automatically classify various noise types in noise attenuation workflows and pursue automated and targeted noise removal in “Stepping towards automated multisensor noise attenuation guided by deep learning”. The advantages to turnaround and quality that arise are abundantly clear in the examples shown: under most circumstances, there is no need for denoise testing ahead of production, the automation of the noise classification reduces the workload on the traditional user, and the user more productively uses their time to detect and resolve any ambiguities. At a more fundamental level, Mohammed Faouzi Zizi will use “Simultaneous dual-sensor wavefield separation and seismic data compression using constrained dictionary learning” to show an automated method that pursues multisensor wavefield separation process in a compressed domain. Advantages include being robust to aliasing without the need for data preconditioning, involves significant data compression of the data, and requires less time and human resources.

The Power of Integration to Resolve Small Details

Finally, Didier Lecerf will use “Water column corrections, joint water velocity inversion for 4D surveys”  to present a new methodology for correcting the effect of the water layer variability on time-lapse seismic datasets that is significantly beneficial for deepwater ‘4D friendly’ data. The main difference from conventional approaches is the simultaneous use of both 4D datasets to estimate the water velocity changes and therefore minimize the seismic difference in the overburden. Inversion is performed in the image domain using cross-image CDP gathers built from cross-correlations of acquisition sequences of migrated data.