Stepping towards automated Multisensor noise attenuation guided by deep learning
Author: Bagher Fermani, Yash Pal, Morten W. Pedersen, Edwin Hodge
EAGE - 25 May 2022
Despite technological advancements in marine seismic multisensor acquisition and processing, noise attenuation remains a fundamental step in the early processes for producing high-quality upgoing pressure wavefield data. If we assume the main shortcoming of traditional methods is in the noise detection step, deep learning can be used in only the detection step and the selected noise attenuation engine can be automatically guided by the deep learning noise classification. We have created different deep learning models to detect a variety of noise types present in both marine hydrophone and geophone records. These models are used to automatically classify the samples in the noise attenuation workflows and pass the samples to the appropriate noise attenuation steps. Targeted noise detection lets us perform a better targeted noise attenuation with appropriate levels of harshness without undue concern over possible signal loss. Models can also be used at any step of the processing to classify the samples in both hydrophone and geophone records. The improvement in noise attenuation and its impact on the PUP generation is presented for a real dataset. The advantages to turnaround and quality that arise from the use of these workflows are discussed.
Simultaneous dual sensor wavefield separation and seismic data compression using parabolic dictionary learning.
Author: Mohammed Ouhmane Faouzi Zizi, PGS / University of Oslo (UiO), Pierre Turquais, Anthony Day, Morten W. Pedersen
EAGE - 25 May 2022
Areas covered by seismic exploration surveys are continuously increasing and the data recorded are becoming of very large size. In addition, conventional seismic processing and imaging is a long
meticulous workflow which is time consuming and very expensive in terms of human and computational resources. Consequently, allowing seismic processing steps in a compressed domain can play a key role in the marine seismic industry as it would be a faster and cheaper alternative to the standard seismic processing sequence and would save cost on storage and data transfer. Parabolic dictionary learning has the ability to compress the seismic data by transforming them into an appropriate sparse domain, and also to extract local parameters which relate to the kinematics of the wavefield. In this paper, we use these kinematic parameters to correct for the geophone obliquity problem and thus enable the dualsensor wavefield separation processing step in the compressed domain. Without the need for data interpolation, our method succeeds in reconstructing aliased events and shows comparable results to an industry-standard FK based algorithm in terms of up- and down-going pressure fields reconstruction. It also comes with the advantage of compressing the data by a factor higher than 10.
A new look at autoregressive low frequency reconstruction of seismic data
Author: Maiza Bekara, Ramzi Djebbi, Nizar Chemingui
EAGE - 25 May 2022
This abstract proposes a novel, easy to deploy and computationally fast solution to reconstruct the low frequency content of seismic data for FWI application. This solution is needed when the SNR is very poor or when denoise fails to enhance sufficiently the low frequency content of the data. The method follows a signal processing approach and does not need to train an offline model on auxiliary data, as is the case for solutions based on machine learning. The reconstruction is done from the higher frequencies using a recursion filter which is estimated from the data itself. The novelty of the proposed method is that it transforms the data locally from the time-space domain to the time-slowness domain where the reconstruction is performed. This transformation enforces the time domain sparsity needed to justify the use of recursion modeling in the frequency domain and exploits the spatial coherency in the data. The method also implicitly uses the signal cone limits of the seismic wavefield to provide a physically constrained solution. The proposed method was tested on many datasets to condition it for FWI and proved to be successful to help the inversion to mitigate cycle skipping and to improve the final model.
Imaging by seismic inversion based on the adjoint state method
Author: Øystein Korsmo, Yang Yang, Nizar Chemingui, Andrey Pankov, Antonio Castello
EAGE - 25 May 2022
Least-squares migration (LSM) has the ability to compensate for limitations in the imaging system and estimate an image that is closer to the earth reflectivity. These favourable properties over conventional imaging have made it the method of choice in complex geological settings. In this paper, we discuss the different properties of LSM and Full Waveform Inversion (FWI) and suggest a new nonlinear least-squares reverse time migration (LS-RTM) implemented in the framework of FWI. The new inversion solution iteratively estimates the earth reflectivity while simultaneously updating the velocity model. We demonstrate the effectiveness of the method through a case study from
the Norwegian Sea.
Full waveform inversion with low frequency reconstructed data
Author: Ramzi Djebbi, Maiza Bekara, Nizar Chemingui, Jaime Ramos-Martinez, Amir Asnaashari
EAGE - 25 May 2022
Full Waveform Inversion (FWI) requires the minimization of a highly non-linear objective function which makes the inversion suffer from cycle-skipping. To overcome this issue, we need to have an accurate initial velocity model and/or input seismic data with good low-frequency content. However, low frequencies can be very noisy, and conventional noise attenuation tools fail to recover the useful signal. In this abstract, we propose to apply a novel low-frequency reconstruction method to condition the data for Full Waveform Inversion. The reconstruction is done from the higher frequencies using a recursive filter which is estimated from the data itself. We apply the low-frequency reconstruction method on synthetic data to show its high accuracy even in presence of strong noise extracted from field data. We successfully performed FWI using low frequency reconstructed field data starting from simple velocity models. The reconstructed low frequencies help to mitigate the cycle-skipping observed when these frequencies cannot be utilized in the inversion and the initial models are not accurate. Results demonstrate the effectiveness of the novel data reconstruction method and show its benefits in reducing the turnaround time for building accurate velocity models by FWI, when starting from less suitable initial velocity models.
A Path to the Reduction of Marine Seismic CO2 Emissions
Author: Andrew Long
Industry Insghts - 7 April 2022
PGS is committed to advancing several of the United Nations Sustainable Development Goals (SDGs), including SDG 13: Take urgent action to combat climate change and its impacts. Management are maintaining the target to reduce relative CO2 emissions (t CO2 per CMP km) by 50% compared to 2011 within 2030, and leverage digitalization to identify emission reduction opportunities and meet future greenhouse gas (GHG) monitoring and reporting obligations.
As described below, DataOps optimization of vessel and survey management has been applied to the four Titan-class Ramform seismic vessels since 2020 and allows tracking of fuel consumption and CO2 emissions per CMP km. Drag reduction initiatives and novel survey designs to reduce CO2 emissions per CMP km and improve survey efficiency are also being considered.
Low-Frequency Marine Seismic Source Considerations
Author: Andrew Long
Industry Insights - 13 December 2021
Driven largely by the significance of Full Waveform Inversion (FWI) in many seismic imaging workflows, several marine seismic source concepts have been developed over the years that share a common ambition of displacing a large volume of water (hundreds of liters) per cycle to yield high amplitudes in the 1-8 Hz frequency range where the output from traditional air guns decays rapidly. Most low-frequency source concepts are either large-volume pneumatic devices that variously operate at low or high pressure, or large-volume mechanical resonators or vibrators that displace the surrounding water with a flexible external surface. For reasons of practicality and to reduce cost, most low-frequency source concepts are likely to be used with sparse source lines and large ‘shot’ intervals. Nevertheless, it can be demonstrated that dense 3D spatial sampling of both the source and receiver wavefields will often be beneficial to multi-channel signal processing or wave equation-based imaging workflows, including FWI.
I provide a simple framework to understand the comparative merits of marine seismic low-frequency source concepts recently published at EAGE 2021 and elsewhere. Overall, finding an efficient solution that generates high-amplitude low-frequency data remains a key historical challenge, but some recent progress is evident. I briefly consider the comparative elements of two low-frequency pneumatic source concepts (the Tuned Pulse Source concept of Sercel, and the Gemini concept of ION), the Wolfspar mechanical resonator of bp, and the relevance of the eSeismic method of PGS to acquire continuous wavefields from individually triggered air guns. I also consider methods to 'manufacture' additional low-frequency amplitude content using either ambient noise interferometry or some form of machine learning and conclude with a consideration of low-frequency source deployment factors that may in fact contaminate FWI efforts and present a challenge to model convergence.
Development of a rock physics atlas in the Talara-Progreso Basin, Peru
Author: Roberto Ruiz, Cyrille Reiser, Anna Roubíčková, Neelofer Banglawala
EAGE Rock Physics Workshop - 26 November 2021
The geological provinces on the Pacific Coast of Peru have a long and complex geological history. The petroleum system has undergone more than a hundred years of exploration and exploitation (de Souza et at.), however its full potential remains to be understood. Integration of seismic and well data is key for building a consistent geological framework that can be used to explore its prospectivity. A case study with a robust petrophysics and rock physics workflow implemented in eleven wells from the North-West Pacific region of Peru is presented here. This workflow allowed us to predict the in-situ elastic response of the well logs, as well as investigate, in real-time, how potential geological scenarios, such as changes in porosity, mineral volume and fluid properties, can affect the response in elastic well logs and by extension in the seismic amplitudes.
Maximizing quality and efficiency with wide-tow multi-source configurations
Author: Martin Widmaier, Carine Roalkvam, Julien Oukili, Rune Tønnessen
First Break - 9 November 2021
The authors present the latest achievements in multi-sensor streamer acquisition with wide-tow sources and how these have optimized high-resolution imaging of the shallow subsurface.
EAGE 2021: Decarbonization the Catalyst for a New Geoscience Era?
Author: Andrew Long
Industry Insights - 26 October 2021
Held in Amsterdam on October 17-22, the annual EAGE conference provided a particularly interesting industry forum as the world moves into a lower carbon future. Carbon Capture and Storage (CCS) featured strongly in the program, and the proportional CCS content of future conferences is expected to grow substantially. I comment on short-term applications of surface seismic methods and geological paradigms to CCS and note that much R&D into both geophysical and geochemical aspects is necessary to support the likely scale of CCS for global net-zero goals.
Synergies between towed streamer and ocean bottom node (OBN) acquisition had a high profile—with particular emphasis upon wide-tow multi-source developments, low-frequency seismic considerations, and various continuous wavefield source concepts. For seismic imaging, Full Waveform Inversion (FWI) has progressed beyond a velocity model building tool to now yield seismic interpretation deliverables of various sophistication. The most complete realization combines model building and full-wavefield least-squares migration into an abbreviated workflow for rapid project turnaround. Overall, it is evident that greater seismic acquisition and imaging effort, combined with better integration of geoscience and engineering methods, is necessary to solve long-standing conventional hydrocarbon discovery and recovery challenges, and to meet the unique subsurface resolution and characterization requirements for the transition to a net-zero carbon emissions. This may seem familiar, but a clear urgency exists to accelerate access to better data—augmented of course by machine learning and other automation platforms—and to throw everything at previously unassailable problems on quite grand scales.
An elephant in the room is whether the challenge of decarbonizing the planet with sustainable, affordable and accessible energy sources can motivate a new generation and boost recruitment to the geosciences and engineering – and on what timescale. Several forum discussions attempted to address these challenges.