- Andrew will deliver this presentation live on the booth with Q&A, Tuesday 8 December at 15:15 CET. Click here to join with your EAGE 2020 login details.
PGS will showcase the integrated value of its marine seismic solutions at EAGE, with no better example than the GeoStreamer X platform presented by Carine Roalkvam. GeoStreamer X is a multisensor streamer survey design that uses wide-tow multi-sources to improve near offset sampling, variable streamer lengths to enable deep velocity model updates with FWI, and multi-azimuth shooting to optimize target illumination at a much lower cost than OBN acquisition. When combined with industry-leading full wavefield imaging technologies, tremendous flexibility exists to tailor precise marine seismic solutions for our customers.
Three related topics are correspondingly addressed by other PGS presentations:
- Improving the ability of reflection FWI to robustly recover high-resolution deep model updates
- Dense spatial sampling of the source wavefield, and
- Exploiting the power of Least-Squares Migration to enable flexible survey designs, as well as optimizing Quantitative Interpretation (QI) studies.
FWI and Imaging
A key area of FWI development involves the improved use of reflections to provide very deep model updates without a dependency upon ultra-long offsets. This remains something of an industry challenge, as has been part of the motivation to use expensive Ocean Bottom Seismic acquisition to record ultra-long offset diving wave information. Dan Whitmore introduces a new method to pursue remarkably detailed synthetic modeling without any knowledge of the density model-a key historical assumption that can create problems for reflection FWI. Yang Yang then applies this new platform to demonstrate a reduced dependence upon long offsets for deep FWI model updates using multisensor GeoStreamer data from the Orphan Basin in offshore Canada. Continuing the FWI theme, PGS will also have Andrew Long participate in an EAGE ‘Hot Topics’ debate on the future directions of FWI.
In the manner of GeoStreamer X, PGS has acquired several surveys in the last couple of years with significantly increased lateral separation of the sources being towed. This enables high-resolution shallow imaging, and the sampling of the source wavefield is starting to approach the density of OBN surveys. One necessary practical consequence of wide-tow multi-sources is that smaller sources are towed, and blended shooting is inevitable. PGS can, however, turn these challenges into an advantage. Stian Hegna shares results of the latest testing of the continuous wavefield acquisition, or eSeismic method, in shallow water, offshore Malaysia. The continuous firing of individual air gun elements with random firing intervals greatly reduces the environmental sound footprint, improves the source-side spatial sampling, and is particularly suitable for scenarios where many sources are operated simultaneously.
Paradoxically perhaps, despite the much coarser receiver-side spatial sampling of OBN surveys by comparison to towed streamer surveys, Didier Lecerf will explain how Full Wavefield Migration (FWM) using both primaries and multiples can enable a radical reduction in the density of OBN acquisition. The methodology has been assessed using a real OBN dataset from deep water West Africa and is a research collaboration between Total and PGS. Reducing the cost of OBN acquisition is the highest priority for many operators.
Automation and Machine Learning
FWM is typically implemented within a Least-Squares framework, and PGS again expands its portfolio of flexible Least-Squares Migration (LSM) solutions when Elena Klochikhina presents a new (iterative) data domain LSM implementation that provides high-resolution angle gathers, corrects for angle-dependent illumination effects, and therefore provides more reliable amplitudes for AVA analysis.
Faster turnaround and project delivery is increasingly desired within E&P timeframes, which has contributed to the explosion of renewed interest in machine learning methods; and Deep Learning implementations in particular. PGS has two contributions to the EAGE based upon forms of Convolutional Neural Networks. In the pre-processing stage before velocity estimation, Bagher Farmani applies a U-net image segmentation model to improve the classification of swell noise, thereby paving the way forward to automated noise removal. And Elena Klochikhina trains a U-net-like convolutional network to remove uncancelled migration isochrones without affecting the integrity of geological structures. Alternatively, the PGS hyperModel presentation by Marcus Bell is a timely reminder that not all processes need to be abstracted by Deep Learning, and a series of case studies illustrate how complex velocity model building can be reduced from months to weeks using simpler machine learning strategies. Simplification of traditional workflows is also a fundamental element of automation, and Maiza Bekara proposes a new and elegant methodology to estimate a mixed phase wavelet from seismic data with no a priori assumption about the phase or use of well-log information.
Finally, Marta Wierzchowski demonstrates the merits of well-sampled multisensor streamer acquisition and advanced 4D processing over the Gullfaks field and sets a path towards future high-resolution 4D seismic.