
SPECIAL TOPIC
SPECIAL TOPIC
EAGE NEWS Make room for a space adventure
CROSSTALK Industry challenges lost in transition
TECHNICAL ARTICLE A model to generate pseudo sonic logs
The world is in an ‘ energy addition ’ phase. Oil and gas will be needed for decades to come, with demand continuing to be robust beyond 2035. Meeting this demand will only be possible by utilising the latest technology and leveraging strategic partnerships . The geoscience and engineering community has much to contribute.
FIRST BREAK ®
An EAGE Publication www.firstbreak.org
ISSN 0263-5046 (print) / ISSN 1365-2397 (online)
CHAIR EDITORIAL BOARD
Clément Kostov (cvkostov@icloud.com)
EDITOR Damian Arnold (arnolddamian@googlemail.com)
MEMBERS, EDITORIAL BOARD
• Philippe Caprioli, SLB (caprioli0@slb.com) Satinder Chopra, SamiGeo (satinder.chopra@samigeo.com) Anthony Day, NORSAR (anthony.day@norsar.no)
• Peter Dromgoole, Retired Geophysicist (peterdromgoole@gmail.com)
• Kara English, University College Dublin (kara.english@ucd.ie)
• Hamidreza Hamdi, University of Calgary (hhamdi@ucalgary.ca)
• Fabio Marco Miotti, Baker Hughes (fabiomarco.miotti@bakerhughes.com)
• Roderick Perez Altamar, OMV (roderick.perezaltamar@omv.com)
• Susanne Rentsch-Smith, Shearwater (srentsch@shearwatergeo.com) Martin Riviere, Retired Geophysicist (martinriviere@btinternet.com)
• Angelika-Maria Wulff, Consultant (gp.awulff@gmail.com)
EAGE EDITOR EMERITUS Andrew McBarnet (andrew@andrewmcbarnet.com)
PUBLICATIONS MANAGER
Hang Pham (publications@eage.org)
MEDIA PRODUCTION
Saskia Nota (firstbreakproduction@eage.org) Ivana Geurts (firstbreakproduction@eage.org)
ADVERTISING INQUIRIES corporaterelations@eage.org
EAGE EUROPE OFFICE
Kosterijland 48
3981 AJ Bunnik
The Netherlands
• +31 88 995 5055 eage@eage.org
• www.eage.org
EAGE MIDDLE EAST OFFICE
EAGE Middle East FZ-LLC Dubai Knowledge Village PO Box 501711 Dubai, United Arab Emirates
• +971 4 369 3897
• middle_east@eage.org
• www.eage.org
EAGE ASIA PACIFIC OFFICE
EAGE Asia Pacific Sdn. Bhd.
UOA Centre Office Suite 19-15-3A No. 19, Jalan Pinang 50450 Kuala Lumpur
Malaysia
• +60 3 272 201 40
• asiapacific@eage.org www.eage.org
EAGE LATIN AMERICA OFFICE
EAGE Americas SAS Av Cra 19 #114-65 - Office 205 Bogotá, Colombia
• +57 310 8610709 americas@eage.org
• www.eage.org
EAGE MEMBERS’ CHANGE OF ADDRESS
Update via your MyEAGE account, or contact the EAGE Membership Dept at membership@eage.org
COPYRIGHT & PHOTOCOPYING © 2025 EAGE
All rights reserved. First Break or any part thereof may not be reproduced, stored in a retrieval system, or transcribed in any form or by any means, electronically or mechanically, including photocopying and recording, without the prior written permission of the publisher.
PAPER
The publisher’s policy is to use acid-free permanent paper (TCF), to the draft standard ISO/DIS/9706, made from sustainable forests using chlorine-free pulp (Nordic-Swan standard).
29 A robust empirical model to generate pseudo sonic logs from neutron porosity logs
James Corey Morgan, Kevin Chesser and Theodore Stieglitz
37 Modelling and inversion of superconducting quantum interference device transient electromagnetic survey data for geothermal resource exploration
Michael S. Zhdanov, Alexander Gribenko, Leif Cox, Keiichi Tanabe, Tsunehiro Hato, and Akira Tsukamoto
43 Assessing petroleum system risks in deep Middle Eastern gas plays using regional screening
Chris Gravestock, Owen Sutcliffe, Thomas Jewell, Mike Simmons and Joseph Jennings
49 Sand-clay distribution and best quality sand thickness in the Våle and Lista Formations of Rogaland Group: Comparison of stratigraphic reference maps and AI-based inversion results in the Norwegian North Sea Elephant database
Vita Kalashnikova, Rune Øverås, Tatiana Nekrasova, Carl Fredrik Gyllnehammar and Ivar Meisingset
59 Interactive seismic stratigraphic analysis: user-guided visual enhancement and AI-driven depositional element extraction
Julien Razza, Remi Leblond, Nasser Olleik, Étienne Legeay, Marie Etchebes and Laurent Souche
69 Super Nova Scotia: It’s time for an old sun to reignite
Karyna Rodriguez, Lauren Found, George Kovacic and Neil Hodgson
73 Enhancing FWI convergence through self-supervised low-frequency extrapolation of legacy marine data: A case study from the Asri Basin, Indonesia
Sonny Winardhi, Asido Saputra Sigalingging and Ekkal Dinanto
81 From data to discovery: Exploring East Java’s subsurface with the Facies Map Browser
David Little
86 Calendar
cover: Seismic survey data visualised on a digital map used by geologists to identify oil reserves in various geological formations.
Environment, Minerals & Infrastructure Circle
Andreas Aspmo Pfaffhuber Chair
Florina Tuluca Vice-Chair
Esther Bloem Immediate Past Chair
Micki Allen Liaison EEGS
Martin Brook Liaison Asia Pacific
Ruth Chigbo Liaison Young Professionals Community
Deyan Draganov Technical Programme Representative
Madeline Lee Liaison Women in Geoscience and Engineering Community
Gaud Pouliquen Liaison Industry and Critical Minerals Community
Eduardo Rodrigues Liaison First Break
Mark Vardy Editor-in-Chief Near Surface Geophysics
Oil & Gas Geoscience Circle
Johannes Wendebourg Chair
Timothy Tylor-Jones Vice-Chair
Yohaney Gomez Galarza Immediate Past Chair
Wiebke Athmer Member
Alireza Malehmir Editor-in-Chief Geophysical Prospecting
Adeline Parent Member
Jonathan Redfern Editor-in-Chief Petroleum Geoscience
Robert Tugume Member
Anke Wendt Member
Martin Widmaier Technical Programme Officer
Sustainable Energy Circle
Giovanni Sosio Chair
Benjamin Bellwald Vice-Chair
Carla Martín-Clavé Immediate Past Chair
Emer Caslin Liaison Technical Communities
Sebastian Geiger Editor-in-Chief Geoenergy
Maximilian Haas Publications Assistant
Dan Hemingway Technical Programme Representative
Carrie Holloway Liaison Young Professionals Community
Adeline Parent Liaison Education Committee
Longying Xiao Liaison Women in Geoscience and Engineering Community
Martin Widmaier Technical Programme Officer
SUBSCRIPTIONS
First Break is published monthly. It is free to EAGE members. The membership fee of EAGE is € 85.00 a year including First Break, EarthDoc (EAGE’s geoscience database), Learning Geoscience (EAGE’s Education website) and online access to a scientific journal.
Companies can subscribe to First Break via an institutional subscription. Every subscription includes a monthly hard copy and online access to the full First Break archive for the requested number of online users.
Orders for current subscriptions and back issues should be sent to First Break B.V., Journal Subscriptions, Kosterijland 48, 3981 AJ Bunnik, The Netherlands. Tel: +31 (0)88 9955055, E-mail: subscriptions@eage.org, www.firstbreak.org.
First Break is published by First Break B.V., The Netherlands. However, responsibility for the opinions given and the statements made rests with the authors.
The call for abstracts will open this month for the 87th EAGE Annual Conference & Exhibition, taking place from 8 to 11 June 2026 in Aberdeen. This will be the first time the event is held in the city, long known as a global hub for oil and gas and now increasingly at the centre of energy transition efforts in Europe.
The theme for this year is Maximising recovery: Unlocking value through technology and partnerships, and we’re looking for contributions that show how geoscience and engineering can deliver real, practical solutions across sectors. Whether you’re focused on hydrocarbons, geothermal, CCS, or emerging technologies, we want to hear how your work is driving innovation, improving efficiency, and building collaboration.
This year’s event is hosted by bp. The company is also leading the Local Advisory Committee (LAC). The committee brings together experts from across industry and academia, including bp, TotalEnergies, Viridien, Heriot-Watt University, and the University of Aberdeen, to help shape a programme that reflects both local strengths and global relevance.
We’re aiming for a broad and balanced Technical Programme that brings together the full EAGE community: from geophysics, geology and reservoir engineering, to renewables, infrastructure and
data science. We also want to highlight the growing number of crossover technologies and projects, those that don’t fall neatly into one category, but push the boundaries of our disciplines.
If you’re working on something new, testing ideas, or bringing different fields together, now is the time to share it. This is a chance to present your work to peers from around the world, take part in tech-
nical discussions that matter, and connect with others in your field.
We’re especially keen to see strong involvement from early-career professionals and students, those who are just starting out and already helping to shape the future of the industry.
The deadline for submissions is 15 January 2026. Visit www.eageannual.org for full details and to submit your abstract.
Two special optional events have been added to our EAGE Digital 2026 being held in Stavanger, Norway on 9-12 March next year. One is a hackathon Wind power revolution: Harnessing AI and simulation for optimal wind farm design and energy integration , the other is a short course Language models for geoscience applications
The hackathon aims to empower participants to explore and implement AI-driven models to forecast and optimise renewable energy production. Participants will get hands-on experience using various tools, such as using NVIDIA Earth-2 and FourCastNet to forecast wind patterns. Participants will be invited to combine in a unified workflow machine learning-based weather predictions, wind farm simulation with PyWake, and energy system optimisation with PyPSA.
The short course is being presented by Dr Thomas Bartholomew Grant, domain expert of Cegal, and will explore the potential of generative AI (Gen-AI) for geoscience. By examining the key concepts of large language models and
real-world applications of them, participants can gain insights into how these technologies are being used to solve complex geoscience challenges. The course material is aimed at geoscientists
who are looking to use AI applications and want a better understanding of how they work, how to get the best out of them and how to critically evaluate their performance.
Abstracts welcomed
The side activities are designed to enrich your conference experience. For those joining us in Stavanger, Norway and interested in being part of shaping the future of digitalization, there is the opportunity to submit your abstract for the Technical Programme by 1 November 2025.
We’re excited to announce a webinar that takes our geoscience discussions out of this world – literally. It also includes a special incentive for becoming an EAGE member for 2026.
As part of World Space Week 2025 (4-10 October), led by Bruno Pagliccia, a veteran geophysicist and space exploration enthusiast, we will explore how geophysics is contributing to the next frontier: space exploration. Geophysical activities in space exploration – Moon, Mars and beyond will bring a fresh perspective to the ways our community’s expertise is being applied beyond Earth including analysis of seismic activity on the Moon and Mars and examination of the potential for using underground 3D imaging in planetary exploration.
As space missions ramp up and interest in off-Earth resources intensifies, geophysics is emerging as a critical enabler of sustainable, science-based space initiatives. The webinar will reflect on past seismic studies and explore the tools, strategies, and questions surrounding the search for and extraction of resources in space.
The timing couldn’t be better. With the membership renewal period opening on 1 October, this webinar is a great early benefit for returning members, and a perfect welcome for new ones. By joining, you will not only gain access to this webinar planned on 7 October, but also enjoy full membership benefits for the rest of 2025 and all of 2026, including unlimited free webinars and learning resources, discounts on all EAGE courses and events, and more.
DUG Elastic MP-FWI Imaging provides accurate models of of Vp, density, Vp/Vs, S-impedance and P-impedance ratio directly from raw field data. These quantities were derived without the need to generate angle stacks for an AVA inversion workflow.
The DUG Elastic MP-FWI Imaging derived quantities are geologically conformable and show a significant increase in resolution. We can readily identify the reservoir location and fluid effects — a beautiful example of a flat spot!
info@dug.com | dug.com/fwi
At the upcoming EAGE Masterclass on Geothermal Energy (17-20 November 2025, Paris), Professor Denis Voskov of TU Delft will lead a course on Geothermal reservoir engineering of energy production. Here he tells us what is involved.
What can participants learn from the course?
The main thing they will learn are the key physical concepts behind geothermal energy production, understanding the basic principles of reservoir simulation for geothermal applications, obtaining some practical grasp of geothermal operations through simulation exercises, and evaluating how different numerical and physical properties affect the dynamics of geothermal production.
How does reservoir engineering integrate with other disciplines in geothermal energy production?
Reservoir engineering allows us to integrate different geological concepts and hypotheses, a large variety of geophysical observations, environmental
constraints, and production data into a single simulation model. This model can help in further adjustment and improvement of geological concepts, suggest which geophysical observations can improve our understanding of subsurface reservoirs, prevent environmental hazards, and finally optimise the production of geothermal resources.
Does the course incorporate real-world cases to reinforce theoretical concepts?
Besides learning the theory behind geothermal energy production and its modelling, participants will improve their understanding with several buildfor-purpose simulation exercises using open-source software solutions.
Are practical exercises involved?
There are several practical exercises in Jupyter Notebooks using openDARTS (https://darts.citg.tudelft.nl/). The exercises will allow participants to learn how various physical and numerical parameters affect geothermal systems.
Summarise why participants might attend
Are you missing reservoir engineering experience in your work? Do you want to understand how geothermal systems perform and what are the most important properties affecting energy production? Are you interested to learn how to model geothermal reservoirs with an open-source simulator? Join our course and feel yourself as a true reservoir engineer in the energy transition world.
As the energy sector undergoes a profound transformation, the next generation of geoscientists and engineers must be ready to lead. This is what the upcoming Energy Transition Student Days is all about.
As part of the EAGE Global Energy Transition Conference (GET 2025) in Rotterdam, students will be able to join a three-day programme 28-30 October 2025 addressing geoscience-related project planning in a teambased environment.
The Geosciences and the Energy Transition challenge provides an opportunity to experience the building of a real-world case study set in NW Europe. Participants will be able to collaborate to assess the feasibility of developing either a carbon storage site, geothermal energy project, or hydrocarbon field, navigating technical, economic, and societal considerations along the way.
Adopting a problem-based learning approach, the programme is intended to push students beyond the classroom, from interpreting geological data to building production forecasts and cashflow models, The hands-on activities simulate the complexity of subsurface energy development in the context of the Paris climate goals. A highlight of the final day includes a stakeholder role-play to examine competing interests in energy planning.
Expert guidance will be provided by Drs Raymond Franssen, Manuel Willemse, and Eilard Hoogerduijn Strating, who bring extensive experience in geoscience and energy transition.
For a fee of €100, participants also gain full access to the GET 2025 conference, including technical sessions, field trips, networking events, and more. For students ready to explore real challenges and contribute to the energy future, this is an opportunity not to be missed.
With just a month to go, anticipation is building as we prepare to welcome the global geoscience and engineering community to GET 2025, the 6th EAGE Global Energy Transition Conference and Exhibition from 27-31 October 2025 at the Convention Centre WTC in Rotterdam. Once again, the event brings together professional communities involved in carbon capture and storage, geothermal, hydrogen, energy storage, and offshore wind.
GET 2025 comes at a key moment when the need for innovation, collaboration and practical solutions has never been greater as the demand for both traditional and low-carbon energy continues to rise. Delegates can look forward to nine focused tracks, some 25 panel discussions, 350 presentations and over 400 expert speakers.
Whether you’re coming from industry, academia, government or a related sector, this is an opportunity to engage with leading voices shaping the future of energy. The event offers the chance to stay up to date on both technical advancements and policy developments, ensuring you remain ahead in a fast-moving landscape.
Beyond the sessions, GET 2025 is a place to connect with peers, build new relationships and collaborate across sectors. The exhibition floor will showcase the latest tools, technologies and solutions, offering hands-on insight into where the energy transition is heading.
Free visitor passes and early registration discounts are still available. Make sure to register by 1 October to secure reduced
rates. If you want the full GET experience, opt for an all-access pass, which includes workshops, field trips and short courses that dive deeper into key topics.
Visit eageget.org for the full programme, registration details and travel information.
‘This is an opportunity to explore the technologies enabling the energy transition, from seismic advancements to cost-effective monitoring and renewable integration. Collaboration across disciplines will spark new solutions to urgent challenges.’ –Mike Branston, new energy domain lead – EXD, SLB
‘We’re covering the full spectrum of geothermal innovation - from advanced geophysics and reservoir modelling to real-world applications like district heating and power. Both low- and high-temperature resources are in focus, reflecting geothermal’s expanding global role.’
– Ghazal Izadi, COO, XGS Energy
‘With strong momentum from both industry and academia, hydrogen is emerging as a key pillar in a secure, low-carbon energy future. Across nine sessions, around 40 experts will address storage in salt caverns, depleted reservoirs, and aquifers, along with microbial effects and natural hydrogen.’ –Bahman Bohloli, senior specialist, NGI
‘Our offshore wind sessions will span technical innovation in geophysics and ground modelling, frontier technologies like geohazards and seismic sources, and strategic synergies with CCS and hydrogen. This is where global experts shape what’s next in offshore renewables.’ –Maarten Vanneste, technical expert, NGI
Here’s what was discussed at the successful 1st EAGE/SBGf workshop on marine seismic held on 21-22 May 2025 in Rio de Janeiro, Brazil.
The event was inspired by the marine seismic acquisition workshops that have been taking place in Norway since 2018. Now Brazil is providing a complementary event as a hub for innovation and high-quality seismic data to image and monitor the prolific pre-salt reservoirs under strict environmental constraints.
Exceeding all expectations, the event attracted more than 110 participants, with a very strong representation from Petrobras as the main customer for the technologies under discussion, and proportional representation from other oil companies with operated or non-operated ventures in Brazil, including Equinor, Shell, TotalEnergies, Chevron and ExxonMobil.
There was also broad participation from service companies which support seismic acquisition, including TGS, Sercel, BGP, PXGEO, Shearwater, SAE, Alcatel, InApril, Ocean Infinity, Toveri, and Gaia. Also represented were seismic processing
companies, including DUG, SLB, Bluware, Searcher Seismic. Companies offering environmental and logistical services also participated, including Sonardyne, Seiche, Seisintel, GeoMain, and lesBrasil. Other organisations represented included research agencies funded by the Brazil R&D obligation, including UFRN, UFF, Aqualie, and UTFPR. Finally, there was important attendance by the environmental regulator IBAMA.
Topics spanned innovative seismic sources and sensors, case studies, and the latest technologies designed to enhance efficiency while minimising environmental impact. Key areas included advances in permanent reservoir monitoring systems, broadband and simultaneous source techniques, marine vibrators, and innovative ocean-bottom nodes, including the integration of fibre optics for data acquisition. Also on the agenda were AI-driven applications, drones, and satellite technology for wildlife monitoring, alongside sustainable approaches to seismic acquisition through real-world case studies and innovative survey designs.
The topics were organised into eight sessions, with 25 speakers, addressing advances in permanent reservoir monitoring (PRM), seismic sources, seismic sensors, and sustainable acquisition.
We were reminded of the four PRM systems in operation by Equinor in shallow waters in Norway, two older electrical systems and two newer fibre optic systems, and the efficiency of 4D processing delivering images in six weeks from surveys acquired every six months. Inspired by the Jubarte PRM pilot (2011-2013), a technological pioneer ultra deepwater and proofof-concept in Brazil, the newest fibre optic PRM deployment will be in deepwater Brazil in the Mero field, with one system monitoring two platforms starting in late 2025. A second system is planned for two other platforms from 2027. Considering a lifespan of 20 years or more, there was a debate about incorporating improvements in source technology over time, for improved
data quality or lower cost operations. Justification for such large upfront investments was discussed, bearing in mind the return on investment together with the remaining oil reserves and field development plans.
The upcoming technology of on-demand OBN (OD OBN) was discussed as a middle ground between OBN and PRM. The first deployment of such a system is planned in Brazil starting in Q4 2025. PRM systems open the door for testing innovative technology, like distributed acoustic sensing (DAS), where fibre optic cables are deployed along the PRM cables (surface DAS or S-DAS) and record seismic data. Also having a continuous stream of recorded data allows for using passive seismic records either for imaging or wildlife monitoring.
The various 4D OBN campaigns in execution in Brazil are being used to evaluate recent technologies and innovative monitoring concepts. For example, during the latest monitor survey over the Búzios field, Petrobras experimented with the concept of Focus 4D (also known as i4D) to determine the minimum time for 4D signals to be detectable (weeks or months) over an area with a fluid switchover during a water-alternating-gas (WAG) operation.
Continuing development of marine vibrators including field trials plus detailed processing by Shearwater and the MVJIP were discussed to demonstrate the benefits of the technology. Demands from the business and the path to commerciality remain less clear.
There was broad interest on lower-impact operations, including smaller airgun arrays, marine vibrators, and a variety of sources focused on low frequencies. The latter especially has targeted the needs of full waveform inversion (FWI). Lower footprint operations with denser node arrays (and sparser shot
grids), lean-crewed vessels and autonomous vessels were also discussed. An important context was provided by the UN Sustainable Development Goals (SDGs), particularly the item #14 Life Below Water.
A key perspective in the Brazilian seismic acquisition context was provided by the regulator IBAMA. One key point was its concern that the emphasis on lower-frequency sources needed a proactive approach to investigate the impact that they may have
Informal exchange of views.
on marine life. Another specific concern related to the speculative seismic market, which inundated their analysts with numerous environmental licensing requests, frequently in overlapping areas, and which were often discontinued or not used if granted. These requests consume a lot of time and should be addressed by the seismic operators, the agency said.
31 OCT
•
ASSESSMENT OF CO2 STORAGE INTEGRITY BY UNDERSTANDING COUPLED THERMO-HYDRO-CHEMICAL-MECHANICAL PROCESSES, BY ANDREAS BUSCH
• BOREHOLE SEISMIC MONITORING FOR SUSTAINABLE ENERGY SOLUTIONS, BY SEBASTIEN SOULAS
• UNDERGROUND HYDROGEN STORAGE IN ROCKS: PORE-TO-CORE SCALE FLOW PROCESSES, X-RAY IMAGING AND MODELLING, BY KAMALJIT SINGH
• SEISMIC DATA PROCESSING FOR OFFSHORE WIND FARM DEVELOPMENT, BY SHAJI MATHEW
ROTTERDAM, THE NETHERLANDS DURING
ROTTERDAM, THE NETHERLANDS
DURING GET 2025
The journey of the EAGE Local Chapter of Kuwait has been nothing short of remarkable. Since its establishment, the Chapter has steadily grown into an active and inspiring platform that brings together geoscientists, engineers, and students from across the country. In just two years, we have already been honoured with the Best Local Chapter Newcomers award in 2024 and proudly secured the prestigious title of Best Local Chapter 2025. This global recognition serves as a testament to the commitment, energy, and vision shared by our members and leadership.
Since the beginning the Chapter has embraced a variety of initiatives aimed at engaging members and creating value for the geoscience and engineering communities in Kuwait. Some of the most notable trends we have observed among our local members include a strong interest in field trips, which continue to be one of the most anticipated activities. Additionally, our in-person technical sessions on cutting-edge topics in geoscience and engineering have witnessed high levels of attendance and interaction, reflecting the eagerness of our community to stay informed and connected. Complementing these activities, social events have played a key role in strengthening the sense of community among members, fostering collaboration beyond the professional realm.
Looking ahead, we have ambitious and thoughtful aspirations for the future. We aim to increase the level of involvement of engineers, as geologists currently form the majority of our membership base. Reaching out to a broader and more diverse audience
is also a priority, particularly through initiatives like our bilingual Geology Handbook of Kuwait, which is currently in preparation and designed for school students, non-geologists, and government entities. We believe that incorporating more social and environmental events will help attract a wider segment of society and contribute to the inclusive growth of our Chapter.
In addition, we seek to deepen our collaboration with governmental
sectors and local societies, supporting the expansion and sustainability of our initiatives. We also aspire to enhance diversity within our chapter by encouraging non-Kuwaiti professionals and students to take part in our activities. Recognising the importance of holistic development, we plan to introduce programmes focused on soft skills, career development, and community service, ensuring that our Chapter not only supports technical growth but also empowers individuals to make meaningful contributions to society.
We will soon be preparing for the first EAGE Offshore Workshop titled Seismic to simulation and organising an international geological field trip to Tuwaiq Mountain in Saudi Arabia in collaboration with Aramco. We will also be organising the second edition of the Advances in Carbonate Reservoirs Workshop, building on the success of the inaugural event. These efforts aim to elevate our Chapter’s impact and create opportunities for our members to engage in transformative learning experiences.
To celebrate the journey that led to winning the Best Local Chapter award for 2025 board members gathered over dinner, reflecting this is not just a title but represents the dedication, creativity, and teamwork that have shaped us into a vibrant, engaging, and forward-thinking community.
Every month we highlight some of the key upcoming conferences, workshops, etc. in the EAGE’s calendar of events. We cover separately our four flagship events – the EAGE Annual, Digitalization, Near Surface Geoscience (NSG), and Global Energy Transition (GET).
EAGE Workshop on Enhancing Subsurface Practices using AI/ML 10-11 November 2025 – Perth, Australia
Innovate and inspire! This workshop is your platform to showcase how you’re tackling one of humanity’s greatest challenges: understanding the hidden structure of the Earth. AI/ML has evolved from a concept to a powerful tool for geoscientific inquiry. From delineating faults to refining seismic images, these technologies allow us to infer, classify, and quantify uncertainty in ways never before possible. We want to hear how you’re using ML across the vast spectrum of geophysical data – from classification to interpretation and beyond. Join the conversation, and help shape the future of sub surface exploration and recovery by registering for the event.
Early fee until 30 September
EAGE Rock Physics Workshop 10-12 November 2025 – Cape Town, South Africa
Hosted in a city renowned for its stunning geological setting, the workshop begins with a field trip (included in delegate fee) on 9 November covering The Geology of Cape Town & Cape Peninsula offering a hands-on look at the region’s remarkable rock formations. A core display will also be featured during the event.
Third EAGE Workshop on Geothermal Energy in Latin America
12-14 November 2025 – Guancaste, Costa Rica
As Latin America moves toward greener and more sustainable economies, geothermal energy emerges as a key alternative – offering high resource potential, exceptional reliability with capacity factors above 80%, and resilience to climate variability. However, harnessing this resource requires a deeper understanding of complex geological environments. The third edition of the workshop will serve as a dynamic forum to exchange knowledge, share lessons learned, and explore strategies for reducing development risks. The programme will also feature a field trip to ICE’s geothermal
The workshop will explore how cutting-edge advancements, such as full waveform inversion (FWI), extended FWI (EFWI), and deep learning, are being synergistically integrated with rock physics principles. This integration is now playing a critical role in enhanced reservoir characterisation.
Early bird fee until 1 October (50% discount for African residents)
Early bird fee until 20 September
EAGE/FESM Conference on Petrophysics
Meets Geoscience: Unlocking Reservoir Potential in a Dynamic Energy Landscape 18-20 November 2025 – Kuala Lumpur, Malaysia
The conference brings together global experts to advance integrated reservoir characterisation by aligning petrophysics, geology, and geophysics to explore multi-scale workflows – from nano-pore rock-fluid interactions to basin-scale reservoir development. Technical discussions will cover quantitative interpretation (QI), 4D time-lapse techniques, AI-driven analysis, advanced logging-while-drilling (LWD), and high-resolution data acquisition. With the energy landscape shifting toward decarbonisation, key themes also include carbon capture and storage (CCS), net-zero strategies, and cross-disciplinary innovations supporting the energy transition.
Early fee until 17 October
Dong Zhang (Fugro), Eric Verschuur (TU Delft), Eric Cauquil (TotalEnergies) and Gwenaëlle Salaün (Independent consultant, formerly Ørsted) report on EAGE Annual 2025 workshop on Elastic inversion of 2D/3D UHR/EHR seismic data for offshore wind.
The workshop was a resounding success with over 60 participants from across the global offshore wind industry and beyond fostering vibrant discussions and knowledge exchange.
The objective was to gain insights and explore current practice and innovative workflows related to UHR seismic elastic inversion, and its application to offshore renewables as part of the energy transition. This was done by identifying best practices, potential pitfalls and causes of uncertainties in the required input data from hands-on experience and lessons learned throughout the industry and academia.
The workshop was structured around three key themes (organised in three presentation sessions), being: (1) the stateof-the-art UHR seismic acquisition and processing; (2) borehole geophysics and seismic CPT; (3) seismic inversion techniques, ending with a thought-provoking panel discussion.
From the first session it became clear that pre-processing of UHR seismic data – compared to the more traditional deep seismic data – has some extra challenges regarding extreme feathering and other positional limitations like source and streamer depth variations, requiring advanced regularisation methods. In addition, the sources – typically sparker sources – give less predictable source signatures with strong angle-dependent variations. In addition, the variations of the water waves are in the wavelength range as the subsurface resolution. This requires extra effort in removing ghost effects and multiples. Tidal variations and spatially varying water speed are additional complicating factors. Altogether, data pre-processing needs to be done very carefully to make the data useful for elastic inversion.
For the second session two methods used to acquire in situ Vp and Vs were discussed, i.e., seismic CPT and PSSL. As demonstrated during the workshop, each method has its own strengths and limitations, but they can also be complementary – especially in the unconsolidated layers of the shallow subsurface. The use of (seismic) CPTs is considered as a ground truth experiment to calibrate the seismic inversion. However, traditional cone-penetration tests (CPTs) do not always provide reliable results and the connection with the elastic parameters coming from seismic inversion is not obvious and based on some a priori models. Furthermore, the so-called seismic CPTs, where shear wave speeds are measured during a CPT survey with an acoustic logging tool, cannot be considered ground truth, as they have quite a bit of inaccuracy in them, especially in the unconsolidated part of the shallow subsurface. The only alternative would be to directly measure surface waves and derive the shear velocities from dispersion-curve analysis. However, this needs a multi-channel seismic set-up with sources emitting much lower frequencies than the ones used for the UHR data.
In the third session the actual elastic inversion of the seismic reflection data was considered. From described cases studies it was shown that success was not guaranteed: low quality data did not allow to access shear wave velocities from the AVO information. Furthermore, typical streamer lengths of 100 m are not considered long enough to access the S-wave velocity information via AVO inversion. Therefore, calibration with reliable ground truth measurements is indispensable, while going to larger offsets seems a must for future data acquisition projects. Also, more advanced inversion methodologies like elastic FWI, combining both kinematic and dynamic information from seismic data, seems a way forward for more accurate inversion results.
For the panel discussion Gwenaëlle Salaün (independent consultant), Eric Cauquil (TotalEnergies), Alistair Robertshaw (bp) and Joek Peuchen (Fugro) were invited to the stage enabling the complete wind farm site characterisation process to be reviewed from a larger perspective. The aim of geophysical data, like seismic surveys is to limit, but not eliminate, the use of hard measurements such as (seismic) CPTs. At the same time a typical wind farm site investigation project takes about 30% of the total costs for windfarm development, where a balance between costs and payback, e.g., leading to the use of more cost-effective monopiles, should be found. At the same time regulators play a role in their requirements for site investigation, which may limit the freedom of geophysical acquisition. An additional factor is that relationships between elastic parameters (seismic P-wave and S-wave velocities and densities) and the required geotechnical soil properties are largely unknown and also unexplored. A lot of work is still to be done, in which machine learning could play a role.
At the end of the workshop the key takeaways were summarised as follows: 1) Stronger integration between UHR/EHR
seismic acquisition and processing is essential. Questions such as how processing can inform acquisition design, the optimal streamer length for AVO analysis, and the necessity of near-field hydrophones and sparker source characterisation were raised as critical areas for collaboration; 2) Borehole geophysics and CPTs are often considered the ground truth, but the workshop highlighted significant uncertainties and limitations in borehole measurements, which may not always reflect true subsurface conditions; 3) Improved inter-disciplinary understanding is needed between geophysicists and geotechnical engineers. A notable disconnect exists regarding resolution expectations: while geophysicists aim for ultra-high resolution (~0.5 m), geotechnical engineers typically focus on averaged soil stiffness over areas of approximately 20 m x 20 m. This raises questions about the practical value of UHR data for seismic inversion, though it may still be valuable for boulder detection; and 4) Enhanced communication among operators, engineers, and geophysicists is crucial for the success of offshore wind projects. Bridging these gaps will lead to more effective project planning and execution.
Time to start planning for the EAGE Near Surface Geoscience Conference and Exhibition 2026 set to take place in Thessaloniki, Greece on 20-24 September 2026.
Thessaloniki’s rich archaeological history provides the perfect meeting place for professionals, researchers, academics and industry experts from around the world to come together and discuss the latest developments, trends, and innovations in the near surface geoscience and engineering sector.
Four parallel conferences will be held at the event. These include the 32nd Meeting on Environmental and Engineering Geophysics exploring near surface challenges the world is currently facing and how they are being tackled. The 7th Conference on Mineral Exploration and Mining will focus on recent advances, trending topics, and novel geophysical methods in mineral exploration and mining. The 2nd Conference on Geohazards Assessment and Risk Mitigation will address the growing impact of geohazards, such as earthquakes, landslides, volcanic eruptions, floods, and coastal erosion, risks that have become more frequent and severe with climate change and rapid
urbanisation. Lastly, the 3rd Conference on Hydrogeophysics will explore the structure and processes of the subsurface environment to identify key properties and variables related to water flow and solute transport, particularly in a rapidly changing climate.
The conferences will offer attendees a unique opportunity to explore a
wider range of topics and innovative methods, encouraging cross-disciplinary collaboration and knowledge exchange to expand the boundaries of geoscience.
Stay tuned on www.eagensg.org for further updates, as more detailed information about the conference and exhibition will be announced soon.
EAGE welcomes the opportunity to recognise outstanding young professionals who have become recipients of the Petroleum Geoscience and Basin Research Early Career Awards 2024. These accolades are presented annually for the best papers published in the two journals in the past year by authors in the early stages of their careers.
The Petroleum Geoscience Early Career Award is for a paper that offers a novel approach or interpretation of a compelling and widely relevant topic in petroleum geoscience. This year, we congratulate André Gondim Brandão, research geologist at Laboratory of Sedimentary Geology (Lagesed), Department of Geology, Institute of Geosciences, Federal University of Rio de Janeiro. His winning paper ‘Fracture analysis in borehole images from BM-C-33 area, outer Campos Basin, Brazil’ is published in Petroleum Geoscience, volume 30, issue 2.
André Gondim Brandão says: ‘Carried out as part of my master’s research, the study investigates the characteristics of natural fractures and how they have influenced reservoir properties in the Raia area, offshore Brazil. The research crowns years of dedication and collaboration between Lagesed and Equinor, and the Early Career Award represents a key achievement in my career. I am deeply honoured by this recognition and excited to use it as motivation to contribute to future projects in the energy industry.’
The reservoirs in the BM-C-33 area, located within the Raia Manta and Raia Pintada development zones, are pre-salt
limestones situated on volcanic sequences. These reservoirs have a complex geological history, including post-depositional silicification that altered mineral composition and led to intense fracturing.
Brandão’s study using borehole image logs, wireline data, and seismic surveys analysed the acoustic and resistivity properties of these fractures. The pre-salt section was divided into three informal stratigraphic units, and it was found that major fracturing was caused by regional tectonic stress, with local positioning also playing a role. A direct relationship was identified between fracturing and silicification.
The analysis further indicated that the distribution of fracture density, vug volumes, and dissolution features, which were limited to specific units, suggests a stratigraphic control on fluid percolation. Brandão’s study concluded by highlighting how the unique structural features of the BM-C-33 area likely influenced the intensity and extent of these diagenetic alterations.
Early Career Award 2024
The Basin Research Early Career Award recognises research published in the journal that marks a significant step forward in our understanding of sedimentary basins and published within three years of thesis completion. We congratulate Ziqiang Zhou (PhD researcher, Department of Earth Science and Engineering, Imperial College London) on winning the 2024 award for his paper ‘Unravelling tectonic and lithological effects on transient landscapes in the Gulf of Corinth, Greece’, published in Basin Research, volume 36, issue 5.
Ziqiang Zhou explains the approach of his study: ‘Our paper is motivated by the fact that although many tectonic active areas have heterogenous bedrock, the effect of lithology is often not considered in tectono-geomorphologic analysis. This matters as spurious tectonic interpreta-
tions can emerge if the effect of lithology is not accounted for. Focusing on the Gulf of Corinth, one of the best-studied rifts in Europe, we constrain the erodibility of multiple contrasting lithologies. In particular, we show that lithology-calibrated topographic metrics provide tectonic constraints consistent with independent geological evidences. These calibrated metrics also extend previously scattered constraints on the timing and magnitude of tectonic events to the entire region, allowing us to refine the fault growth and linkage history of the Gulf of Corinth. By doing so, this work demonstrates tectonic and lithological effects on topography can be reasonably isolated through our relatively simple topographic analyses. Excitingly, our approach can be readily applied to many other regions as long as digital elevation models and geological maps are available.’
Zhou adds: ‘I’m truly honoured to receive the Early Career Award for this work – something I couldn’t have achieved without the incredible guidance and support from my supervisors every step of the way. Seeing the many hours we poured into this research recognised by the community means a lot to me. It feels like an important milestone in my PhD journey, and has given me a big boost of confidence and motivation to dive into even more exciting research ahead.’
With 2 award categories for students and young professionals, EAGE champions talent from the start - fostering excellence, innovation, and shaping the future of geoscience and engineering.
MYRTO PAPADOPOULOU
Arie van Weelden
Award recipient 2025
Receiving the EAGE van Weelden Award for my work in reflection seismic and surface-wave methods marked a personal journey of collaborations, research and industry partnerships, and active community engagement. To young professionals, I say: stay curious, be fearless and let your passion, curiosity, and collaboration guide you. Near-surface geoscience is a catalyst for building a sustainable future, and you are an essential part of making that future a reality.
6 CELEBRATING EXCELLENCE SINCE 1955
UNLOCK OPPORTUNITIES FOR GROWTH
STUDENTS CAN APPLY FOR THEIR FIRST YEAR OF MEMBERSHIP FOR FREE (RE)ACTIVATE YOUR MEMBERSHIP TODAY!
Over 200 participants attended the fifth edition of the EAGE workshop on well injectivity and productivity in carbonates (WIPIC 2025), held in April in Doha, Qatar. This is the report.
Under the theme Innovative Technology for Reservoir Optimisation, the four-day workshop served as a robust platform to bridge theory and practice, facilitating knowledge exchange between academia and industry, and offering an arena for discussing novel technologies, field applications, and future-forward strategies.
A recurring narrative throughout WIPIC 2025 was the intersection of digital innovation and human expertise. The integration of artificial intelligence (AI), machine learning (ML), and digital transformation into reservoir management was widely discussed, revealing a sector increasingly reliant on data-driven tools without losing sight of the critical role of human judgment.
Several sessions emphasised cross-disciplinary approaches, integrating geomechanics, geophysics, and digital rock physics.
A standout example was the QASR simulator, a Qatar-initiated innovation that successfully transitioned from academic research to commercial deployment underscoring the workshop’s focus on regional contributions and applied innovation.
Sustainability and energy transition also emerged as dominant themes. Topics such as CO2 storage, fault reactivation, and water control were explored from both technical and environmental lenses.
Audience engagement was notably high, with active participation during short courses, panel discussions, and poster sessions. Day One’s hands-on short courses attracted professionals eager to sharpen their skills in machine learning, reservoir simulation, and uncertainty quantification.
Attendees showed particular interest in AI/ML applications, not only in technical modelling but also in their broader implications for workforce development and digital workflows. The 10th anniversary celebration of the event added a human dimension, reinforcing the workshop’s sense of community and continuity.
WIPIC 2025 highlighted the evolving identity of the energy sector, a field actively shaping, not just responding to, technological shifts. The workshop confirmed a collective move toward data-augmented decision-making, where models and algorithms complement, but do not replace, professional insight.
Panel discussions brought forth debates on the role of AI in empowering versus replacing human expertise, maintaining production performance while aligning with sustainability goals, and the value of regional innovations like the QASR simulator. Education, upskilling, and inter-disciplinary cooperation were highlighted as crucial to harnessing these innovations.
As the curtain closed on this fifth edition, the message was clear: technology and talent will continue to shape the future of energy.
The EAGE Student Fund supports student activities that help students bridge the gap between university and professional environments. This is only possible with the support from the EAGE community. If you want to support the next generation of geoscientists and engineers, go to donate.eagestudentfund.org or simply scan the QR code. Many thanks for your donation in advance!
We have now published over 50 Personal Record interviews , allowing us a glimpse into the varied careers and private lives of professionals working in our geoscience and engineering community. Here is just a brief selection of some memorable quotes from the archive.
Advice for young geoscientists
Give, expect, and demand respect. If you ever witness or experience something drastic as discrimination or sexual harassment, seek help and speak out. Don’t shy away from the larger challenges. Be curious and follow your passions. Explore the idea of being paid for what could be your hobby.
- Adriana Citali Ramirez, chief geophyscisit, TGS and artist/poet
Personal inspiration
On my first night in hospital, blinded with shrapnel, a nurse read the following passage to me from The Greatest Salesman in the World by Og Mandino: ‘I will persist until I succeed. I was not delivered unto this world in defeat, nor does failure course in my veins. I am not a sheep, waiting to be prodded by my shepherd. I am a lion and I refuse to talk, to walk, to sleep with the sheep ….’
- Andrew Long, PGS (now TGS) on recovering from survey accident
Lifetime conclusion
After years of exhausting effort, I have reached a conclusion .... My message is that the idealism of the young age has surrendered to the realism of the old age.
- Oz Yilmaz, CTO, GeoTomo, author of seismic processing reference works
Challenging beginnings
I was born and raised in Ndwedwe, one of the poor rural villages in South Africa. I had a tough early childhood. My father died when I was two years old, leaving my mother to raise me and my five siblings from her low income as a domestic worker. Though my mother did not have an opportunity to go to school she worked hard to instil discipline and the importance
of education in my life. So she was my greatest strength and true inspiration.
- Musa Manzi, associate professor in geophysics, University of Witswatersrand
Invention(s) most proud of?
It’s hard to pick one of my children …
- Doug Crice, owner, Geostuff
Working in oil industry
It did take me almost 10 years to feel good about what I do. As a biologist, I felt like I was working for the dark side! What changed was understanding how energy access is essential to pull people out of extreme poverty and feeling like I was able to make a real change …
- Daniella Bordon, global ESG manager, BGP Offshore
Lessons from sport for geoscience
To be successful in sport you need to work hard, and be dedicated and persistent over long periods of time with no instant gratification. You should never be content with past glory but always strive for improvement. Another lesson is that no matter how talented and hard-working you are, you will never succeed alone.
- Vetle Vinje, CGG (now Viridien) and Olympic rowing silver medallist
Everyday life in Kyiv
I witnessed the aftermath of an attack drone just 200 metres from my home. Eleven flats were damaged serving as a stark reminder of the proximity of danger. From our shelters, we hear the reverberations of hundreds of explosions in various parts of the city.
- Dmitry Bobheza, EAGE office, Kyiv
Prospects for students
I always encourage students to study what they think is fun. What they actually will work with will sort itself out eventually. Doing what you think is exciting will take you there.
- Johan Robertsson, professor applied geophysics, ETH-Zurich
Can activism affect change?
It is the only thing that can affect change.
- Richard Pancost, professor of biochemistry, University of Bristol
Advice to young entrepreneurs
Gain experience, do a job, work for a boss. From a practical perspective, not everyone can drop out of college and be Mark Zuckerberg, as we often have family and financial responsibilities that have to be balanced with entrepreneurship.
- Nina Hernandez, Iraya Energies
Do geology and music relate?
I’ll spare you the obvious plays on the word ‘rock’. But really, I don’t separate them in my mind. Both require creativity and imagination…
- Tony Dore, Ex-Equinor and professional musician
Pole position
To me pole (dance) is a sport that challenges my body every day just like petrophysics challenges my brain every day. In my job interview the pole was behind me and I think I introduced myself as ‘I’m, Zoë, I love rocks and pole’. Don’t knock it until you’ve tried it – life is better upside down.
- Zoë Cumberpatch, petrophysicist, Equinor
BY ANDREW
There’s little more perplexing than trying to assess exactly where the world is going with energy transition, so much so that sometimes it feels as though we have set sail on a journey in which even the destination is unclear. Best we can do is judge which way the wind is blowing.
This may not be a conclusion that delegates to our flagship Global Energy Transition (GET 2025) next month in Rotterdam will want to contemplate. Nor, one suspects, will this community of geoscientists dedicated to energy transition and the ultimate goal of a decarbonised world be entirely happy with the progress of needed technology subject to a world buffeted by political and economic crosswinds.
Undeniably there has been a shift in the conversation post-Covid and the advent of two wars. In summary, energy security, industrial/commmercial considerations, near stalemate of COP-type meetings, and a rising tide of populist/anti-regulation movements in many countries have all taken a toll on the priority being accorded to climate change (even in a summer of raging fires around the world).
It is becoming the orthodoxy to talk about energy addition rather than energy transition, which is indeed the theme of EAGE’s Annual Conference & Exhibition next year in Aberdeen. It was also a feature of future energy scenarios offered by Dr Scott Tinker in his Opening Session presentation at this year’s meeting in Toulouse.
The mindset now is that energy addition is about meeting the unavoidable growth in demand factoring in rapid industrialisation and population increases in the developing world. Substantial oil and gas production for the foreseeable future is going to be needed, while at the same time we somehow do not lose sight of climate change mitigation/decarbonisation in whatever form that might take, e.g. building renewable capacity, cleaner energy production, carbon capture and storage, etc.
For some naysayers, even this strategy is too great a concession to the energy transition. From day one in office the Trump Administration proclaimed its hostility, notably withdrawing
the US from the 2015 Paris Climate Agreement, rolling back clean energy provisions in the Biden Inflation Reduction Act, and waging war on wind power whilst enacting measures to boost US domestic oil and gas production. It remains to be seen exactly how much impact these policies have, given the US is second only to China in its energy transition investments (World Econmic Forum). Whether the US can afford to overlook some green energy sources may be an issue. For example, the Energy Information Administration (IEA) forecasts US electricity sales to the commercial sector to rise by 3% in 2025 and 4.5% in 2026, driven largely by demand from data centres, while electricity sales to industrial consumers are expected to rise by 2% in 2025 and 3.5% in 2026.
We may wonder whether following the energy addition credo is putting off some hard choices on cooling the planet. We can all agree that the whole process was never likely to be a straightforward switch and is going to be much slower than optimists anticipated. In a recent Foreign Affairs essay on ‘The troubled energy transition’, Daniel Yergin reminded readers of the historical precedents from the moment in January 1709 that Abraham Darby, a village foundryman and ironmaster in England, hastened the transition from wood to coal and helped to initiate the subsequent Industrial Revolution by pioneering the use of coke as the fuel to smelt iron (instead of charcoal). Astonishingly it was not until the beginning of the 20th century that coal overtook wood as the world’s number one energy source. Similarly oil discovered in the mid-19th century did not overtake coal as the world’s number one energy source until the 1960s, and even today the world is using three times as much coal as it did in the 1960s.
Yergin makes the point on energy addition: ‘In 2024, the world used more wind and solar energy than ever before. But it also used more oil and coal than ever before’. He also cautions on the sheer scale of the decarbonising project in a $115 trillion-plus world economy that continues to grow making 2050 Net Zero increasingly unrealistic. He cites the Energy Institute’s Statistical Review of World Energy finding that between 2022 and 2023 the
world’s dependence on conventional energy – oil, natural gas and coal – declined by less than 0.5% from 81.9% to 81.5%.
Yet, for all the negativity, the commitment to energy transition worldwide is gigantic and obviously unstoppable, as the numbers show. One headline figure from the International Energy Agency’s 2025 annual World Energy Investment report suggests that global energy investment is set to increase this year to a record $3.3 trillion of which the share of clean technologies – renewables, nuclear, grids, storage, low-emissions fuels, efficiency and electrification – is on course to hit a record $2.2 trillion with oil, natural gas and coal reaching $1.1 trillion.
There are some concerning portents. Earlier this year Energy Transition Investment Trends 2025 compiled by research provider Bloomberg New Energy Finance found low-carbon energy transition worldwide grew 11% to hit a record $2.1 trillion in 2024 (close to the IEA figures). Growth was driven by electrified transport, renewable energy, and power grids, which all reached new highs last year, along with energy storage investment. However, the pace of growth was slower than the previous three years, when investment jumped by 24-29% annually.
Electrified transport remained the largest investment driver, reaching $757 billion in 2024, according to BNEF when spending on passenger EVs, electric two- and three-wheelers, commercial electric vehicles, public charging infrastructure and fuel cell vehicles is taken into account. Renewable energy hit $728 billion, including on and offshore wind, solar, biofuels, biomass and waste, marine, geothermal and small hydro. Investment in power grids totalled $390 billion, which includes investment in transmission and distribution lines, substation equipment, and the digitalization of the grid. Not specifically mentioned is IEA modelling that projects data centres will use 945 terawatt-hours (TWh) in 2030, roughly equivalent to the current annual electricity consumption of Japan. By comparison, data centres consumed 415 TWh in 2024, roughly 1.5% of the world’s total electricity consumption.
BNEF’s report also reveals a marked difference between investment in mature and emerging sectors of the clean energy economy. Technologies that are proven, commercially scalable and have established business models, like renewables, energy storage, electric vehicles, and power grids, accounted for the vast majority of investment in 2024. These sectors drew $1.93 trillion, growing 14.7%, despite perceived hindrance from policy decisions, higher interest rates and expected slower consumer purchasing.
In contrast, investment in emerging technologies, like electrified heat, hydrogen, carbon capture and storage (CCS), nuclear, clean industry and clean shipping, reached only $155 billion, for an overall drop of 23% year-on-year. Factors that discourage investment in these sectors include affordability, technology maturity, and commercial scalability. ‘In order to scale these industries, the public and private sectors need to do more to
de-risk these technologies, otherwise, they are not likely to have any meaningful impact on emissions by the end of the decade’, BNEF concludes.
Regionally the largest market for investment was China, which alone accounted for $818 billion of investment, up 20% from 2023. China’s investment growth was equivalent to twothirds of the total global increase in the year.
The EU, US, and UK, which drove growth in 2023, saw different results in 2024. Investment was stagnant in the US, reaching $338 billion, and down in both the EU and UK, hitting $375 billion and $65.3 billion, respectively. This made China’s total investment last year greater than the combined investment of the US, EU and UK.
A similar story is told by analyst company Bridgewater Associates in its March publication. Is the green energy transition dead? The short answer is no, it is not. However there are caveats. ‘To compete on costs and economics … companies can no longer rely on a “policy backstop” to make overambitious investments’. The company has previously estimated that about 40-50% of global emissions reductions required to achieve net-zero goals can come from scaling technologies that are already mature but the technology is not yet there to address the remaining 50-60% of emissions. The rollback of subsidies and other policy supports is likely to make the path more challenging. Already more than 90% of climate investment flows go to mature technologies, which are cost-competitive with fossil fuel-based options and thus likely to continue growing. The remaining 5-10% goes toward emerging or immature technologies that are unlikely to be profitable in the absence of subsidies and whose share has already been shrinking.
None of these observations should come as a surprise to those intending to attend GET 2025. The geoscience and engineering community is focused on technology solutions, many of which are extraordinary. But, technology advance is constrained by the investment imperatives of the market economy, political will and community acceptance.
Carbon capture and storage (CCS), due to garner a lot of attention at GET, illustrates the point. The trajectory of CCS deployment remains a long way off where it must be to deliver net zero by 2050, according to DNV’s Pathway to Net Zero study published in June. CCS will progress, growing from 41 MtCO2/ yr captured and stored today to an estimated 1,300 MtCO2/yr in 2050, 6% of global emissions, but six times less that what is said to be. ‘Economic headwinds have put pressure on this capital-intensive technology and corrective action will need to be taken by government and industry if we are to close the gap between ambition and reality’, DNV warns.
This is just one of many mammoth challenges ahead. The ever quotable Albert Einstein once noted: ‘A ship is always safe at shore but that is not what it’s built for’. At least the energy transiton ship has set sail.
Views expressed in Crosstalk are solely those of the author, who can be contacted at andrew@andrewmcbarnet.com.
WWW.EAGE. ORG
10-12 NOVEMBER 2025 • CAPE TOWN, SOUTH AFRICA
Register for the Seventh EAGE Rock Physics Workshop before 2 October to enjoy the early bird rate, which includes full access to both the workshop and field trip! Plus, we’re excited to offer a 50% discount on registration fees for African residents — don’t miss this opportunity to connect, explore, and learn at a great value!
Early bird Deadline: 2 October 2025
RPW25 V3H.indd 1 13/06/2025 08:11
SEABED SEISMIC: EVOLUTION THROUGH INNOVATION
24-26 NOVEMBER 2025 • MANAMA, BAHRAIN
Dive into the future of marine geophysics — register now for the Third EAGE Seabed Seismic Today Workshop and short courses , and join us in vibrant Manama for cutting-edge insights and collaboration !
WWW.EAGE. ORG Register now!
TGS has reported a second quarter net loss of $60 million on operating revenues of $334 million, compared to a net profit of $35 million on revenues of $224 million in the second quarter of 2024. It reported a second quarter operating loss of $18 million compared to an operating profit of $55 million in Q2 2024.
The company said that its results were ‘negatively impacted by several factors’, with several data licensing deals being postponed and ‘low client commitment to ongoing projects’.
To reduce operating expenses TGS has announced it is in the process of selling the vessels Ramform Explorer and the Ramform Valiant, and stacking the Ramform Vanguard
Challenging operational conditions for large contract projects and lower than expected JV partner participation for certain multi-client programmes further negatively affected contract revenues, TGS said.
Order inflow was a disappointing $133 million during Q2 2025, with total order backlog of $425 million, compared with an order inflow of $368 million and a total order backlog of $600 million in Q2 2024.
Organic multi-client investments were $115 million compared with $52 million in Q2 2024. The multi-client investment level was higher than expected due to ‘lower than expected joint venture partner participation for certain projects’.
Multi-client sales were $136 million compared with $115 million in Q2 2024. Contract sales were $172 million compared with $100 million in Q2 2024.
Net cash flow of $11 million in Q2 2025, compared with -$13 million in Q2 2024.
Gross operating costs for 2025 are expected to be approximately $950 million compared to previous guidance of approximately $1 billion – a reduction driven by further efficiency gains and vessel scheduling. Another piece of good news is that TGS’ external Imaging revenues were $19.4 million compared to $5.5 million in Q2 2024, as a result of TGS
strengthening its focus on data imaging on behalf of third party clients.
Kristian Johansen, CEO of TGS, said: ‘The Q2 2025 results were negatively impacted by several factors. End-of-quarter data licensing came in below expectations, Further, we encountered challenging operational conditions on one of our streamer projects, negatively impacting revenue recognition. Finally, lower-than-expected partner participation in certain multi-client projects resulted in lower recognition of contract revenues and higher multi-client investments.
‘Although significant macroeconomic uncertainty and high oil price volatility during Q2 caused our clients to be more cautious in the short term, the long-term need for more exploration remains intact. With falling remaining reserve life, many large E&P companies will face declining production rates unless more reserves are added and brought on stream. As a result, we remain optimistic for the long-term opportunities for TGS.’
In its outlook TGS said that ‘global energy demand is expected to increase over the coming decades, and we believe oil and gas will continue to play a vital role in the global energy mix. At the same time, the rapid decline of existing production, combined with rising costs, environmental concerns, and political and regulatory challenges underscore the need for more exploration in both mature and frontier basins. High-quality subsurface data is critical for optimising production from existing assets and enabling effective exploration in both well-known and underexplored areas.
‘During Q2 2025, macroeconomic uncertainty and OPEC’s accelerated production reversion of the production cuts implemented in Q4 2023 have contributed to significant oil price volatility. Although most of our key customers have reiterated their capex plans for 2025, the increased uncertainty has prompted a more cautious approach to data purchases in the short term.’
President Trump has moved to limit wind energy projects in America as a series of orders come into force.
The US Bureau of Ocean Energy Management (BOEM) has rescinded regulations that required the Secretary of the Interior to publish a five-year schedule of anticipated offshore renewable energy lease sales at least every two years.
‘After reviewing this regulation, the Department of the Interior has determined this provision is not mandated under the Outer Continental Shelf Lands Act and unnecessarily limits the Secretary’s discretion over scheduling renewable lease sales.’
Meanwhile, the US Department of the Interior is launching a full review of offshore wind energy regulations to ensure alignment with the Outer Continental Shelf Lands Act and America’s energy priorities under President Trump.
This effort includes reviewing the Renewable Energy Modernization Rule, as well as financial assurance requirements and decommissioning cost estimates for offshore wind projects, ‘to ensure federal regulations do not provide preferential treatment to unreliable, foreign-controlled
energy sources over dependable, American-made energy’.
The review, led by the Bureau of Ocean Energy Management and the Bureau of Safety and Environmental Enforcement will support Secretary’s Order 3437, ‘Ending Preferential Treatment for Unreliable, Foreign-Controlled Energy Sources in Department Decision-Making,’ and President Trump’s memorandum on wind energy signed on 20 January 2025.
‘The Department is fully committed to making sure that offshore energy development reflects President Trump’s America First Energy Dominance agenda and the real-world demands of today’s global energy landscape,’ said Secretary of the Interior Doug Burgum. ‘We’re taking a results-driven approach that prioritises reliability, strengthens national security and upholds both scientific integrity and responsible environmental stewardship.’
The Department has paused new approvals for offshore wind projects— including leases, permits, rights-of-way and loans — in compliance with the Presidential Memorandum on wind energy, while it conducts a review of offshore wind energy projects and their impact on
the environment, national security and the economy.
In support of these presidential directives, BOEM rescinded all Designated Wind Energy Areas on the Outer Continental Shelf on 30 July 2025.
Meanwhile, the US Department of the Interior has reversed the Biden administration’s approval of the Lava Ridge Wind Project, a 1000-megawatt wind facility in southern Idaho.
STRYDE has sold its Mini and Nimble seismic systems to the National Observatory of Brazil to support its geophysical research and exploration activities across Brazil.
The first deployment will focus on a pioneering natural hydrogen exploration project. ‘The National Observatory’s project will leverage STRYDE’s lightweight and agile seismic acquisition technology to acquire high-density seismic data to gain unprecedented insight into Brazil’s subsurface structures,’ said STRYDE in a statement.
‘This acquisition reflects STRYDE’s continued expansion into the Latin American market and underscores the growing global recognition of the company’s innovative approach to seismic acquisition for energy transition applications,’ the company added.
Sergio Luiz Fontes, senior researcher at the National Observatory of Brazil, said: ‘The use of STRYDE’s seismic systems will enable the Observatory to conduct advanced geophysical surveys with minimal logistical footprint and lower operational costs.’
The National Observatory of Brazil is one of Brazil’s oldest scientific research institutions. Founded in 1827, it operates under the Ministry of Science, Technology, and Innovation, based in Rio de Janeiro, is actively involved in monitoring seismic activity across Brazil, and partnering with national and international organisations on natural hydrogen, geothermal energy, and carbon capture and storage (CCUS).
‘This campaign will be the first of several research initiatives,’ added Fontes.
Viridien reported second quarter net profit of $6 million on segment revenues of $274 million, compared with 2024 Q2 net profit of $35 million on segment revenues of $258 million.
Operating profit was $52 million compared to $15 million in Q2 2024. Net cash flow of $30 million compared to a net cash deficit of $6 million in Q2 2024.
The company’s net debt has increased to $997 million from $941 million at the end of Q2 2024.
Data, Digital and Energy Transition revenue was $181 million, up 3% year on year, driven by a 10% increase in Geoscience revenue to $115 million.
‘For the past few years, Viridien has seen growing demand for advanced, high-quality, high-end subsurface imaging, especially in the US Gulf, Middle East, North Africa, and South America,’ said Viridien.
Earth Data revenue of $66 million was down 8%, but OBN projects started in Norway and the US Gulf segment.
Sensing and Monitoring revenue of $93 million was up 14% year-on-year. ‘Activity is mostly driven by the Land segment, with strong deliveries of nodal
system in South America and cabled systems in the MENA region, in particular. The Marine segment remains subdued. In New Businesses, Infrastructure monitoring is showing double-digit growth, while our Marlin Offshore Logistics solution achieved encouraging initial commercial success, with a contract signed with ONGC,’ said the results statement.
Sophie Zurquiyah, chief executive officer of Viridien, said: ‘Viridien delivered a solid performance in the second quarter of 2025. Despite a volatile environment, the group demonstrated resilience, driven by its primary focus on offshore markets and on leading oil companies. Combined with ongoing internal performance improvements, this resulted in robust year-on-year growth in both segment revenue and margins. From a cash perspective, Viridien generated a solid $30 million in net cash flow during the quarter, reinforcing our confidence in reaching our full-year target of $100 million. The combination of a healthy Geoscience backlog and expected licensing activity toward year-end supports our confidence in maintaining momentum on our deleveraging path.’
ExxonMobil has signed an agreement with Libya’s National Oil Corporation (NOC) to explore four blocks offshore Libya. A memorandum of understanding states that ExxonMobil will conduct a detailed technical study of four offshore blocks near the northwest coast and the Sirte Basin.
This MoU establishes a geological and geophysical study to identify the hydrocarbon resources in these blocks. It also paves the way for the resumption of the partnership between NOC and ExxonMobil, which aims to restart its activities in Libya after a decade-long hiatus.
The NOC said it was committed to ‘expanding partnerships with major American energy companies, particularly ExxonMobil’ and that contract terms are more favourable than in the past, reflecting global changes in the energy sector.
ExxonMobil has expressed interest in participating in the public bidding round initiated by the NOC, which includes 22 offshore and onshore blocks.
TotalEnergies has sold its 45% operated interest in two unconventional oil and gas blocks in Argentina, Rincon La Ceniza and La Escalonada, to YPF for $500 million. Total’s partners in these concessions are Gas y Petroleo de Neuquen (10%) and Shell (45%).
Tullow Oil has completed the sale of its assets in Gabon to the Gabon Oil Company (GOC) for $307 million. The sale of its Gabon assets marks Tullow’s exit from its licences in Gabon after 21 years. Tullow has also sold its assets in Kenya to Auron Energy Limited for $120 million. The transaction proceeds will be used to strengthen Tullow’s balance sheet by materially reducing Tullow’s net debt.
The US Bureau of Land Management New Mexico State Office has leased 16 parcels totalling 7501.76 acres for $58,260,939 in total receipts for its quarterly oil and gas lease sale. This is the third highest value for highest bid/ acre for a parcel for BLM. The lease sale is the first conducted under the One Big Beautiful Bill Act, which reset the royalty rate for new federal onshore oil and gas production to a minimum of 12.5%, reversing the 16.67% rate set by the Biden administration.
Norway’s preliminary production figures for June 2025 show an average daily production of 1,854,000 barrels of oil, NGL and condensate. Total gas sales were 8.8 billion Sm3 (GSm3), which is 0.5 (GSm3) less than the previous month. Average daily liquids production in June was 1,675,000 barrels of oil, 174,000 barrels of NGL and 5000 barrels of condensate.
Stone Ridge Energy has signed a deal to acquire ConocoPhillips’ Anadarko Assets for $1.3 billion. The agreement was for ConocoPhillips’ Lower 48 assets in the Anadarko Basin, Oklahoma, US, and the transaction is expected to close at the beginning of the fourth quarter.
TGS has completed of the Dawson Phase III 3D multi-client seismic survey in the Western Canadian Sedimentary Basin, covering 141 km2 and merged with the existing Dawson Phase II 3D to the north and west.
The project has utilised TGS’ Phase and AVO-compliant processing flow to enable precise subsurface imaging and valuable insights for operators developing Montney resources.
‘The Dawson III 3D survey marks an important step as our first new multi-client seismic project in British Columbia since 2019,’ said David Hajovsky, executive vice-president of multi-client at TGS. ‘Through the application of modern seismic imaging techniques and close collaboration with First Nations to minimise
environmental impact, we continue our commitment in delivering high-quality data that supports informed decision-making in one of Canada’s most active formations.’
The Dawson III 3D seismic survey has been integrated with existing TGS data to offer greater subsurface understanding in the region. The project incorporates 291 wells and 191 LAS logs, further refining the dataset and offering valuable insights for operators in the Montney Formation.
Meanwhile, TGS has won a streamer acquisition contract in the East Mediterranean. Acquisition is scheduled to commence in Q3 this year and the contract has a duration of approx. 30 days.
Germany’s current tender for the 10.1 gigawatt offshore wind farms N-10.2 and N-2,5 have failed to attract any bids.
As a result, the German Offshore Wind Energy Association (BWO) has called on the German government to fundamentally reform the auction design.
Stefan Thimm, managing director of the German Offshore Wind Energy Association, said: ‘The industry has been warning for years against burdening companies with too many risks. The current auction design forces developers to bear risks beyond their control without any protection.’
North Sea sites N-10.1 and N-10.2, with a total area of approx. 182 km2, were scheduled to go into operation in 2030 and 2031. Following a tender process that failed to produce any bids and thus failed to award contracts, the Federal Network
Agency is obligated to conduct a new tender process for the same site.
Thimm said that legislators must ensure better conditions for the new tenders. ‘The federal government must finally pave the way for a reliable Contracts for Differences (CfD) system alongside long-term electricity supply contracts. CfDs lead to a reduction in electricity generation costs of up to 30%. Without this reform, further auctions could fail –and with them the energy transition.’
Fugro has won a contract for the 700 MW Youde offshore wind farm by Shinfox Energy in Taiwan’s Round 3.2 offshore wind tender. It will deliver comprehensive geotechnical services to support the development of the Youde project, located off the west coast of Taiwan.
Fugro’s Taiwan-flagged vessel, Pacific Hornbill , an IOVTEC-managed DP2 vessel, is equipped with an advanced C30 marine geotechnical drilling rig and Fugro’s proprietary specialist WISONMkV EcoDrive downhole in situ testing and
sampling system. It also features specialised seabed geotechnical equipment, including the SEACALF MkV Deep Drive, enabling the acquisition of high-quality geotechnical data in both seabed and downhole mode throughout the project.
The geotechnical fieldwork campaign has started and is expected to be completed in the fourth quarter of 2025. ‘By leveraging our ISO-accredited local laboratory, we will deliver integrated high-quality services’, said Vincent Tsai, chairman of Fugro IOVTEC.
Floating liquefied natural gas (FLNG) capacity is expected to triple by 2030 according to research from Rystad Energy.
Once hindered by technical and operational challenges, FLNG projects are now achieving utilisation rates comparable to onshore terminals. With LNG demand rising alongside the growing viability of smaller gas fields, FLNG is emerging as a faster, more flexible and cost-effective solution capable of adapting to shifting market dynamics while unlocking previously stranded reserves.
Rystad Energy estimates global FLNG capacity will reach 42 million tonnes per annum (Mtpa) by 2030, climbing to 55 Mtpa by 2035, almost four times the 14.1 Mtpa recorded in 2024. Terminals commissioned before 2024 achieved an average utilisation rate of 86.5% in 2024 and 76% to date in 2025, figures comparable to global onshore LNG facilities.
‘FLNG has come a long way in less than a decade. The only real roadblocks were early teething issues that come with any new technology, as seen with projects like Shell’s Prelude, which faced cost overruns and unstable output. Since then, the industry has matured significantly, including Prelude itself. Utilisation rates are improving, the technology is proving reliable across a range of environments, and the economics are starting to make more sense,’ said Kaushal Ramesh, vice-president, gas & LNG research, Rystad Energy.
Early FLNG projects, such as Shell’s Prelude, built in South Korea by the Technip–Samsung consortium, exemplified FLNG’s early limitations. Costs ballooned to $2114 per tonne for liquefaction alone. However, as the industry gained operational and construction experience, capital expenditure per tonne has declined significantly, bringing costs in line with onshore LNG projects.
Proposed developments along the US Gulf Coast now average around $1054 per tonne. Delfin FLNG, a proposed project in the US, sits just above that average at $1134 per tonne, while Coral South FLNG in Mozambique, which is similar in scale, reports a comparable liquefaction cost of $1062 per tonne.
In parallel, FLNG developers are increasingly turning to vessel conversions as a cost-efficient alternative to newbuild facilities. Projects such as Tortue/Ahmeyim FLNG, Cameroon FLNG and Southern Energy’s FLNG MK II have achieved notably lower capex levels of $640, $500 and $630 per tonne, by repurposing Moss-type LNG carriers. With several Moss-type LNG tankers expected to retire in the coming years, more could be repurposed, expanding the pipeline of lower-cost FLNG solutions, said Rystad.
Rystad Energy data also shows that FLNG units can be delivered significantly faster than onshore liquefaction facilities, enabling quicker final investment decisions and more agile execution. On average, newbuild FLNG projects are completed in approximately three years, compared to about 4.5 years (capacity-weighted) for operational onshore plants. For FLNG vessels currently under construction, the average projected build time is even lower at 2.85 years.
Mosman Oil and Gas has contracted Sproule to provide an independent estimate of helium resources at the Sagebrush helium project in Colorado, US. The company’s internal assessments recently estimated 205 million cubic feet (mmcf) of net helium contingent resources, and 1.683 billion cubic feet (bcf) of net hydrocarbon contingent resources (based on an initial gas in place of 18.4 bcf and recoverable gas of 11 bcf).
Sproule will review geological and engineering data, including historic gas sample analyses from the Sagebrush-1 well, which indicated a gas composition of 2.76% helium, 19.5% methane, and other gases.
The company will shoot a 3D seismic survey over the Sagebrush project area this quarter. This seismic program, which is expected to cover approximately 16 square miles, will provide detailed subsurface mapping, enabling Mosman
to better define high-potential drilling targets.
Mosman interim CEO Howard Mclaughlin, said: ‘Engaging Sproule, acquiring the 3D seismic survey, and advancing the extended production test at Sagebrush-1 are pivotal steps in our helium strategy. These initiatives are designed to validate our resource estimates, derisk the project, and provide the data needed to advance towards commercialisation.’
JERA and bp have launched JERA Nex bp, a 50:50-owned joint venture global offshore wind developer, owner and operator. The company’s portfolio of operating assets and development projects has a net potential generating capacity of 13GW. This includes around 1GW of installed net generating capacity, a 7.5GW development pipeline and an additional 4.5GW of secured leases.
The Scottish Government has granted consent for SSE’s 4.1GW Berwick Bank offshore wind farm located 38 km east of the Scottish Borders coastline. If fully delivered, Berwick Bank would become the world’s largest offshore wind farm, capable of generating enough clean energy to power more than six million homes annually.
The UK government has confirmed that 10 projects from the first phase of its flagship green hydrogen programme – Hydrogen Allocation Round (HAR1) – can begin construction after long-term contracts were signed to fuel heavy industry with clean energy in South Wales, Bradford (North West), North Scotland and Teesside (North East).
Offshore wind developers Flotation Energy and Cobra have been granted planning approval for the Celtic Sea offshore wind project White Cross. The White Cross Offshore Windfarm project applied to North Devon Council and the Marine Management Organisation (MMO) in 2023 to construct and operate a 100MW floating offshore windfarm 52 km off the Devon coast, UK.
SLB has won a technologies and services contract for carbon storage site development in the North Sea from the Northern Endurance Partnership (NEP) of bp, Equinor and TotalEnergies. SLB will deploy its Sequestri carbon storage solutions portfolio to construct six carbon storage wells. The project scope includes drilling, measurement, cementing, fluids, completions, wireline and pumping services. NEP will store upto to 1 billion metric tons of CO2. ENERGY TRANSITION BRIEFS
Operators in New Zealand will be able to apply for new petroleum exploration permits this month under the country’s Crown Minerals Amendment Bill. The bill removes the ban on oil and gas exploration beyond onshore Taranaki and signals the coalition government’s intent to reinvigorate investment in Crown-owned minerals.
‘This Government is pragmatic about the vital role natural gas will play in our
energy mix in the decades ahead and we have set a course for greater energy security backed by our own indigenous reserves,’ said resources minister Shane Jones.
‘The ill-fated exploration ban in 2018 has exacerbated shortages in our domestic gas supply by obliterating new investment in the exploration and development needed to meet our future gas needs. Reserves are also falling faster than anticipated
TGS and offshore wind technology developer Entrion Wind have released the results of a joint market study confirming the commercial viability of Entrion Wind’s Fully Restrained Platform (FRP) monopile foundation for offshore wind farms in water depths of 60 to 120 m.
TGS used its 4C Offshore market intelligence database, the world’s largest offshore wind project dataset, covering more than 3400 projects globally, to assess the pipeline of projects planned in deeper waters. The analysis identified more than 44GW of capacity under development through to 2040 where deepwater monopile-based foundations could be feasible. Site characteristics such as water depth, seabed conditions and turbine configurations were evaluated to determine the technical suitability of the FRP monopile.
The study indicates that the FRP monopile design offers a ‘technically and economically competitive solution in transitional depth ranges where conventional monopiles are no longer feasible and jackets or floating foundations have traditionally been required,’ said TGS in a statement. ‘TGS research suggests the FRP’s ability to simplify installation processes and reduce foundation capital costs could support broader project viability in deeper waters.’
Bjørnar Eide, CFO of Entrion Wind, added: ‘Working with TGS allowed us to ground our innovation in real-world data. The breadth of insight from the 4C Offshore database provided clarity on where the FRP monopile can have the greatest impact globally. By analysing the global project pipeline, policy conditions, and cost trends, Entrion Wind sees an opportunity for this technology to fill a critical gap in the foundation market.’
‘As the offshore wind sector moves into deeper waters, solutions like the FRP monopile, supported by detailed market analysis, could streamline project design, reduce costs and expand development opportunities,’ said TGS. ‘The study highlights how collaborative research like this helps to reduce uncertainty around new technologies and supports the continued growth of offshore wind globally.’
After a few years of increasing offshore rig activity, the market began experiencing a dip in demand in 2024, according to research by Westwood. The dip is not the start of a new down cycle, but rather a mid-cycle correction in a longer-term upcycle, the company added.
Demand for jackups, semisubs and drillships combined stands at 520 units as of mid-2025, with the marketed committed utilisation rate at 86%.
In 2023, while demand was still rising, only two units – one semisub and one drillship – were retired from the global offshore rig fleet. Both were converted for use outside the industry. In 2024, the number of retirements rose to seven, all of which were semisubs sold for recycling.
The focus on removing semisub supply is the result of this segment taking the hardest hit in terms of a drop in demand. Semisub demand finished 2024 at about 62 units, a drop from 68 at the end of the previous year, and as of mid-2025, semisub demand has fallen to about 58 units. Only one new rig order was made last year – ARO Drilling jackup Kingdom 3.
As a result of the aforementioned attrition and no new supply, the global supply of jackups, semisubs and drillships finished 2024 at 717 units, down from 724 at the end of 2023. This count includes 26 units deemed under
Globally, the number of working rigs at the end of 2024 was 2% lower than at the end of 2023. However, the number of working rigs managed by the top 10 contractors at the end of 2024 was up 3% over the previous year. Looking ahead to the end
construction. By rig type, there were 512 jackups, 94 semisubs and 111 drillships in the global fleet at the end of 2024. For comparison, at the end of 2023, the count was 512 jackups, 101 semisubs and 111 drillships.
ExxonMobil Corporation has announced second-quarter 2025 earnings of $7.1 billion. Cash flow from operating activities was $11.5 billion and free cash flow was $5.4 billion.
Equinor has reported adjusted operating income $6.53 billion and $1.74 billion after tax in the second quarter of 2025. The company reported a net operating
income of $5.72 billion and a net income of $1.32 billion.
Shell has reported adjusted earnings of $4.3 billion despite lower trading contribution in a weaker margin environment. Cash capex outlook was unchanged at $20-22 billion. The company also achieved $0.8 billion of structural cost reductions in the first half of 2025; cumulative reductions
of 2025, Westwood anticipates a decrease in the global working count and in the total offshore rig supply ahead of demand picking up once again for long-term work starting in the second half of 2026 and beyond.
since 2022 are $3.9 billion, against CMD25 target of $5-7 billion by end of 2028.
ConocoPhillips reported cash provided by operating activities of $3.5 billion and cash from operations of $4.7 billion in the second quarter.
TotalEnergies reported second quarter adjusted net income of $3.6 billion and cashflow of $6.6 million.
bp has reported second quarter underlying replacement cost profit of $2.4 billion, compared with $1.4 billion for the previous quarter. Reported profit for the quarter was $1.6 billion, compared with $0.7 billion for the first quarter 2025.
TGS is collaborating with Equinor by providing the Prediktor Data Gateway solution for the Empire Wind offshore wind project.
Empire Wind 1, located off the Atlantic coast of New York, will have a total installed capacity of 810 MW, providing clean energy to approximately 500,000 homes. Prediktor Data Gateway will deliver data management services to
bp has announced a big oil and gas discovery at the Bumerangue prospect in the deepwater offshore Brazil. It drilled at the Bumerangue block in the Santos Basin, 404 km from Rio de Janeiro, in a water depth of 2372 m. The well was drilled to a total depth of 5855 m. The well penetrated an estimated 500 m gross hydrocarbon column in high-quality pre-salt carbonate reservoir with an areal extent of greater than 300 km2
OGDCL (95%) and GHPL (5%) have made an oil discovery at the Tay Exploration Licence in District Tando Allah Yar in Sinde Province, Pakistan. An exploratory well was drilled down to a total depth of 1926 m into the Upper Shale of the Lower Goru Formation. The well flowed 275 barrels of oil per day (BOPD). Furthermore, during formation testing using RES, oil was encountered in the Lower Ranikot formation.
Central European Petroleum has made a significant oil discovery (33.4 API oil) in the 100% owned and operated Wolin East 1 (WE1) well, located in the Baltic Sea 6 km offshore from the city of Swinoujscie. The WE1 well was drilled in waters 9.5 m deep and reached a vertical depth of 2715 m. Tests confirmed a 62 m hydrocarbon column and excellent reservoir properties for oil and gas production in the Main Dolomite geological formation. The Wolin East oil discovery is estimated to contain mean recoverable oil, sales gas and natural gas liquids totalling 200 million barrels of oil equivalent (mmboe). There is also
support standardisation, security and efficient handling of data.
‘Already deployed in major offshore wind projects such as Dogger Bank, Prediktor Data Gateway enables seamless data integration across assets, streamlining operations and maintenance (O&M) applications while allowing operators to scale their digital infrastructure efficiently,’ said TGS.
significant further low-risk exploration, appraisal and secondary recovery potential in the Main Dolomite as well as in the deeper Rotliegend formation. The Wolin licence in total is estimated to contain more than 400 mmboe of recoverable hydrocarbon resources. It is the largest conventional hydrocarbon field yet discovered in Poland, and one of the largest conventional oil discoveries in Europe in the past decade, said Central European Petroleum.
Equinor and its partners have made a gas discovery in the ‘Skred’ prospect in the Barents Sea. The well was drilled about 23 km north of discovery well 7220/8-1 on the Johan Castberg field and 210 km northwest of Hammerfest. The discovery is estimated to between to 1.93.1 million barrels of oil equivalent. The objective of the well was to prove petroleum in reservoir rocks from the Middle Jurassic. The primary exploration target was the Stø Formation, and the secondary exploration target was the Nordmela Formation. The well encountered a 14-m gas column in the Stø Formation, in sandstone totalling 70 m and with good reservoir quality. In the lower part of the Nordmela Formation, the well also encountered gas in a 3-m thick isolated sandstone layer with moderate-to-good reservoir quality. Above the primary exploration target, the well encountered a 14 m thick sandstone layer from the Cretaceous, where a 1-2 m thick zone in the lower part of the reservoir was filled with oil. The well was drilled to a vertical depth of 2144 m
below sea level, and was terminated in the Fruholmen Formation from the Late Triassic.
A consortium led by ExxonMobil has encountered natural gas at a prospect off the coast of Cyprus. Drilling resulted in preliminary indications of 350 m of a gas-bearing reservoir at a depth of 1.9 km in the Pegasus-1 well.
Equinor has struck oil in exploration well 7720/7-DD-1H, Drivis Tubåen, at the Johan Castberg field in the Barents Sea. According to preliminary estimates the size of the discovery is 9-15 million barrels of oil. The oil was proven in a new segment called the Tubåen formation 1769 m below the seabed in 345 m of water. To further increase John Castberg reserves from 450-650 million barrels by adding another 250-550 million barrels, Equinor is planning six new exploration wells.
bp has signed a deal with Libya’s National Oil Corporation (NOC) to evaluate redevelopment opportunities in the mature giant Sarir and Messla oilfields in Libya’s Sirte basin, including the exploration potential of adjacent areas, and to understand the wider unconventional oil and gas potential within the country.
Vår Energi has announced a commercial gas and condensate discovery in the Fenja field in the Norwegian Sea. The discovery on the Vidsyn ridge has recoverable resources in the range of 25 to 40 mmboe gross. The well encountered very good quality reservoirs with more than 200 m of hydrocarbon column.
James Corey Morgan1*, Kevin Chesser 2 and Theodore Stieglitz3
Abstract
Good well ties are an essential part of any seismic interpretation or reservoir characterisation workflow. This requires accurate and complete density and sonic curves in order to generate a synthetic seismogram. There are a wide variety of published empirical models based upon the work of Gardner et al. (1974) and Faust (1951) that are routinely used to calculate replacement sonic or density curves with varying degrees of accuracy. Using a large and globally distributed set of well logs, we develop a novel empirical model that relates neutron porosity and slowness (compressional sonic) that may be used to calculate a reasonable and accurate pseudo sonic estimate. The method relies on the availability of a neutron porosity log, a type of log that is commonly found in a wide range of log vintages dating back to the late 1940s. Results from multiple basins will be shown to demonstrate the robustness of this method and the corresponding model. We believe that this very simple and effective method will provide the working geophysicist with another useful tool to create pseudo sonic logs suitable for generating synthetic seismograms.
Introduction
Missing data is one of the most common problems we face as scientists. There are a variety of ways that we can fill those blanks. Interpolation or extrapolation work well for small gaps in data that are changing slowly or very predictable between two known points. For instance, multi-dimensional 5D interpolation has now become a standard practice in most seismic processing workflows as a means to further regularise data and remove noise and artifacts (Trad, 2014). However, when large amounts of data are missing, empirical modelling is sometimes the best option. Modelling requires that we have a sufficiently large dataset and that a relationship exists within that dataset which makes physical sense.
Well logs are a vital link between geological and geophysical information and the calibration of log and seismic data is necessary for the success of almost any seismic interpretation workflow. The range, quality, and precision of modern well log data have never been better, surpassing what previous generations could have imagined. Despite these advances, we still encounter issues with incomplete data from wells of any vintage. This data insufficiency can stem from various issues, such as a narrow focus on specific reservoirs, reluctance to log dry holes, lax state regulations on data release, borehole stability problems during drilling, and poor preservation of data in commercial databases. Accurate log data is essential to creating diagnostic models of reflectivity, in situ stress, pore fluid substitutions and many other rock properties from which we rely on borehole measurements to calibrate with seismic data. Any of these require both a sonic and density log.
There are two commonly used empirical models that are useful for generating sonic logs. Gardner et al. (1974) used core data from clastic rocks to derive a relationship between compressional velocity and bulk density by use of a power law as shown in equation 1:
(1)
1 Land Seismic Noise Specialists | 2 Ank Geosciences, LLC | 3 Collier Geophysics
* Corresponding author, E-mail: jcoreymorgan@gmail.com, corey@landnoise.com DOI: 10.3997/1365-2397.fb2025065
where ρ is bulk density in g/cm3 and V p is the compressional velocity in ft/s. Other parameters may be derived that are basin or lithology specific. Castagna (1993) derived new Gardner coefficients and reported results in specific lithologies using laboratory data. Figure 1 shows a crossplot of recorded slowness versus slowness calculated using the Gardner model. The data shown in this and subsequent crossplot tests is from a large and geographically diverse set of well data that is further described later in the modelling section. This crossplot shows that most data points are clustered symmetrically around the diagonal, where the model and recorded data are in exact agreement (in red), but it is important to also note that there are a number of points with high error. This may be attributable to bad hole conditions such as washouts that would prevent the density tool from making contact with the borehole wall. Areas of bad borehole conditions are a common problem for a method that relies on a tool requiring sidewall contact.
Faust (1951), on the other hand, relates compressional velocity to formation resistivity and the age or depth of the rocks as shown in equation 2:
good summary of these methods may be found in Mavko (2003). These methods require more a priori knowledge of mineralogy and fluid types and are useful for establishing theoretical bounds for doing rock physics work. Additionally, we know that velocity (VP) and porosity ( are related through equation 3:
where R is the formation resistivity in ohm-ft and Z is the depth in feet. Both methods can be useful in the right geological conditions. Figure 2 shows a similar crossplot of recorded slowness versus slowness calculated using the Faust method. The plot shows that Faust’s method consistently overcorrects the slowness value. Since Faust’s method uses formation resistivity, it is also prone to errors due to problems with the borehole that would prevent the tool from making contact with the borehole wall or the physical response of formation fluids causing an erroneous change in resistivity.
There are well-established relationships between velocity and porosity such as Wylie et al. (1963), and Raymer et al. (1980). A
where K is the bulk modulus, μ is the shear modulus, and ρfluid and ρmatrix are the mean densities of the fluid and matrix respectively. Porosity (φ) cannot be measured directly and is usually obtained by combining a porosity log derived from density measurements () and the neutron porosity log (). Acoustic (sonic porosity) or NMR logs may be used as well. A description of these techniques can be found summarised in Asquith and Krygowski (2004).
Using these relationships to model the petrophysical response of idealised rocks over a range of porosity and fluid types (Figure 3) demonstrates the relationship between porosity and slowness as it relates to lithology.
Anecdotal reports have suggested that rescaled neutron logs were used to generate synthetic seismograms (without density) in the Permian Basin as early as 1978 (Chesser, pers. comm.). Prior work by Chesser (1997) established a quasi-linear correlation, over moderate depth intervals between P-wave travel time (slowness or DTC) and a ‘porosity vector’ defined from a crossplot of the neutron porosity (N) and the porosity calculated from the density log (D):
(4)
The length of this vector is controlled by the largest of the two porosity terms in (4), typically from N
Neutron porosity is a nuclear tool that is routinely acquired in many wells and various forms of such tools have been around
since approximately 1940. These logs measure how fast neutrons, which are slowed down by hydrogen atoms, are absorbed by larger atoms that emit detectable gamma radiation. Since all pore fluids (gas, oil, brine, fresh water) contain hydrogen, the neutron absorption can be linked to porosity. Belknap (1959) demonstrated that raw neutron log measurements could be calibrated to porosity in rocks with known lithology, such as sandstone or limestone, which led to the development of the neutron porosity curve that has been standard since the 1960s. Neutron porosity logs are advantageous in their simple design as they do not rely on contact with the borehole walls and require little calibration. Therefore, they remain accurate in a wide variety of conditions. Raw count rates are approximately related to porosity () by:
(5)
where N is the neutron count in the zone and K and C are constants related to borehole size, tool model, and lithology. Unlike bulk density and sonic logs, issues with borehole integrity typically have less impact on neutron porosity measurements. Furthermore, neutron tools can be utilised in cased holes, unlike most other open hole logs.
The relationship between porosity and slowness and the anecdotal use of neutron porosity as a proxy for slowness in the past in the Permian basin gives us encouragement to explore whether this method might work in other geographic regions. These facts suggest that it may be possible to expand on this use to construct a more generalised relationship between neutron porosity and slowness across a variety of geological conditions. Figure 4 shows a plot of neutron porosity versus slowness coloured by geographical origin. It is clear from this Figure that the data does not show a great deal of regional variation, implying that it might be possible to create a model that would be useful worldwide.
Given that it is always useful to have more tools in our computational toolbox, we seek a simple empirical method. Density or resistivity are not always available and both are susceptible to problems due to bad hole conditions. Our aim is to create a single variable empirical model that can be used to facilitate better seismic interpretation with the use of existing well logs in areas where new data is either too expensive or too difficult to acquire. In light of this goal, we adhered to two ‘golden rules’ as quoted
by George Box and Winston Churchill respectively; ’All models are wrong but some are useful’ and ‘Perfection is the enemy of progress’.
The data for this study consisted of a large and geographically diverse set of logs from a variety of geological settings. Wells from the Gulf of Mexico, the North Sea, NW Shelf Australia, Wyoming, Texas, Louisiana, and Kansas were used to form a dataset of almost 1500 wells with approximately eight million individual data points contributing to the model. Kansas data is prevalent in the full dataset due to the fact that the Kansas Geological Survey has released a very complete dataset of logs going back several decades. We limited the wells we used from the Kansas data archive to those newer than 2010 to ensure that the best quality curves went into the model. Wells were selected that included compressional sonic, density, and neutron porosity. Density was not included in any of the final models, but was used to compare results to the commonly used Gardner approximation. A subset of these wells with resistivity logs was also used for comparison to Faust’s method.
Initial analysis of the data indicated that noise could be a significant issue, necessitating further data conditioning. Statistical models require the removal of outliers, bad data, and other contaminants to avoid skewing results. With a dataset this large, editing each individual well would have been prohibitively time consuming. To address noise and objectively bad data, we first applied a series of filters to reject unrealistic values for sedimentary rocks. In addition to this, we used a 41-point median filter on all curves to further remove spikes and other anomalous data points.
Various methods and forms of the fit equation were tried, including linear, second and third order polynomial, and power law fits. A simple linear fit came out with the lowest R² of approximately 0.65. The power law fit yielded an R² value of 0.72. Both the second and third order polynomial fit provided an overall R² value of 0.76. While the second and third order polynomial had similar correlation values, we found that the third order polynomial had trouble matching the overall shape of the point cluster at higher porosity values. Therefore, we decided to move forward with a second order polynomial of the form:
for the final model where y would be slowness (DTC) in μs/ft and x would be neutron porosity (ϕΝ). The model parameter values (a, b, and c) are shown in Table 1 for the second order polynomial fit.
Figure 5 illustrates a crossplot test of this model similar to our tests of Gardner and Faust in Figures 1 and 2. While the crossplot is slightly less symmetric about the diagonal than Gardner, it has notably less spread in the point cloud.
In order to test our model, we will take a more in depth look at the Equinor North Sea Volve data (Equinor, 2018). This data is a subset of the larger modelling dataset. It is the most complete dataset we have and includes a lithology log in each well. The data contains 171 usable wells with about 900,000 data points. Figure 6 shows us neutron porosity vs. P-sonic (similar to Figure 4) limited to the Volve data and coloured by the lithology log. We see a very distinct set of lithology trends evident in the figure. At the lower values of neutron porosity and slowness, i.e. higher velocities, we see that there are somewhat distinct lithological trends, which merge at higher values of neutron porosity. We believe this points to a relationship with compaction as a driving element in the accuracy of this empirical model. As overburden pressure increases, rocks lose porosity at varying rates. Sand grains are rearranged into a more efficient packing fraction whereas clay platelets better align as overburden pressure increases. As they compact with burial, sands lose porosity much faster than the shales, which are already at a much lower porosity. Geochemical processes such as cementing due to diagenesis can also decrease porosity in sand.
In shallower, under-consolidated clastic rocks, the shales have a higher velocity than the sands and those velocities converge and will cross over each other at some point on the compaction trend. At the same time, the neutron porosity ‘sees’ this loss of apparent porosity in the sands but due to sensitivity to–OH and clay-bound water in the shales, sees higher apparent porosity in those shales. This gives the shales a similar apparent porosity to the sands in the system.
Carbonates, on the other hand, are less affected by compaction but usually have lower porosity than sands as a function of depth (Croizé et al. 2013). However, they are much more prone to geochemical effects like diagenesis. Relative to a shale ‘baseline’, the neutron porosity log sees the carbonate as having lower total porosity while the sonic log sees it as having a higher velocity (lower slowness). It therefore seems logical that the highest correlations for the neutron porosity relationship will occur in basins dominated by carbonate sequences or in older basins where the clastic rocks have undergone compaction. Additionally, due to the fact that Gardner’s relationship is based on clastic core data without carbonates, these may also be the basins where Gardner is likely to be less accurate.
In Figure 7, we crossplot both our neutron porosity calculated slowness values on the left and the Gardner calculated slowness values (right) against the actual slowness values. Once again, both are coloured by lithology. The shales are slightly under-corrected,
although not nearly as much as the Gardner shales are over-corrected, while the sands and carbonates are slightly over-corrected. The grouping of the points is, overall, much tighter using neutron porosity than it is using Gardner.
While these are good tests of the overall statistical effectiveness of the model, it is useful to show how it directly compares to actual well data. Figure 8 shows two of the Volve wells with both the Gardner sonic and the neutron porosity sonic overlain on the actual P-sonic logs. In the 24_4-5 well, we note that the shallower trend, down to approximately 1900 m, is better matched by Gardner; however, below 1900 m, there is a far better match to the recorded sonic using the neutron porosity method. It should be noted that the neutron porosity-based models do match the shape of the recorded sonic curve in the shallow parts of the well, suggesting that some adjustment for the compaction trend might correct the calculation in the shallow data. In the 16_10-3 well, we show a similar comparison with the neutron porosity model and Gardner against the as recorded sonic log. In this case,
the neutron porosity model provides a very good match to the well with the exception of one area between 2300 m and 2375 m Measured Depth (MD).
A second pair of wells (Figure 9) from the Netherlands, available from the Dutch Oil and Gas Authority (Amerada Hess, 1999) and Elf Petroland (1999), shows similar effects. Well A-0501 was drilled offshore Netherlands in 1999. Gardner matches it well down to approximately 3100 m. However, below 3100 m, the Gardner calculation error increases with depth. The neutron porosity method once again appears to be the better performer overall but seems to read a slightly higher velocity (lower slowness value) below approximately 2900 m. The second well, Blesdijke-01, was drilled in the Netherlands onshore in 1998. Once again, we see a slightly better match with Gardner down to about 1640 m. Below that, the neutron porosity model is much better for the remainder of the log.
Finally, we present two wells (Figure 10) from the Northwest Shelf of Australia. Both are offshore exploration wells drilled in
8 Neutron porosity calculated sonic logs (red) and Gardner calculated sonic logs (light-blue) compared to recorded sonic logs (black) for two Volve wells.
9 Neutron porosity calculated sonic logs (red) and Gardner calculated sonic logs (light-blue) compared to recorded sonic logs (black) for two wells from the Netherlands.
the Carnavaron Basin. The Legendre South well (Apache, 2010) was drilled in 2010 and the Forrestier well (Woodside, 1986) was drilled in 1986. As in previous examples, using neutron porosity to calculate pseudo sonic logs gives a better overall result than using Gardner in both wells.
Since our desired end result is a model that aids in generating synthetic seismograms for use in seismic interpretation workflows, then perhaps it would be useful to show how this model performs on an actual well tie. For this, we are using the Teapot Dome (US Department of Energy, 2001) 3D seismic dataset released by the US Department of Energy and the Rocky Mountain Oilfield Testing Center. The field is located in Wyoming, USA and the seismic data was acquired in 2001. This data was reprocessed by Land Seismic Noise Specialists in 2022 with a modern prestack time migration workflow for marketing and internal research use.
Geologically, this area is, in many ways, ideal for the purposes of demonstrating this method. There are a number of thick clastic and limestone beds in the sequence and the rocks are older and more consolidated. There are also a number of well bores in the area and as it was used as a research and test site. Many of these wells have been logged over the years providing us with a useful dataset for comparative analysis.
For this paper, we will show the tie in the 25-1-X-14 well using the recorded sonic log, a sonic log calculated from Gardner, and one calculated from neutron porosity. The well tie was made using the recorded sonic log synthetic seismogram. Subsequently the two generated pseudo sonic logs were substituted in using the same time-depth relationship. It is worth noting here that we used a time-depth relationship we felt was trustworthy and that was based on the acquired sonic log. However, that may not always be the case, especially if the only velocity information one has is a pseudo sonic curve such as we propose in this work. We rarely have perfect well ties and even with check shots, we often require a bit of stretching and squeezing to optimise the fit. In the case where these generated pseudo sonic logs are our only velocity information, it may be worth exploring options such as using seismic velocities or nearby
10 Neutron porosity calculated sonic logs (red) and Gardner calculated sonic logs (light-blue) compared to recorded sonic logs (black) for two wells from the Carnavaron Basin, offshore Australia.
checkshot values as a constraint on the time-depth relationship. For this paper, we used the same time-depth relationship for each of the sonic logs in order to make the synthetic correlation comparison more meaningful. We plan to further explore the quality of well ties without the benefit of a priori velocity information in a future work.
The original tie, using the recorded sonic log, has a seismic-synthetic correlation of 0.738, a respectable correlation for onshore seismic data over almost 800 ms of the time section; when we substitute in the Gardner pseudo sonic, the correlation drops to 0.575. However, when we substitute in the neutron porosity pseudo sonic, we increase the seismic-synthetic correlation up to 0.684. Figure 11 shows the comparison of the three synthetic seismograms resulting from each of the sonic/pseudo sonic logs.
Figure 11 Comparison of synthetic seismograms (in red) for Teapot Dome well 25-1X-14 with the collocated seismic trace (in black). From left to right, recorded P sonic log, pseudo sonic calculated from neutron porosity, and pseudo sonic calculated with Gardner.
In this paper, we propose a novel method for generating pseudo sonic logs using neutron porosity curves. While this method is not perfect, it has the advantage of being simple and elegant as well as utilising a log type that is commonly found in wells going back as far as the 1940s. This method is not a replacement for a proper sonic measurement, but we believe this method is particularly useful for evaluation of older fields, new exploration in more mature provinces, or energy transition projects such as geothermal or CCUS where new subsurface data may be impossible or uneconomic to acquire. We have shown that it generally produces results that are often as good or better than alternative methods for generating pseudo sonic logs, such as Gardner or Faust. Under a wide range of geological conditions, this method produces a reasonably accurate pseudo sonic log that is adequate for creating usable synthetic seismograms for well tie and quantitative interpretation applications. Those adopting this method are encouraged to use their own well logs to generate localised, custom-fit parameters to test against our more global model.
In addition to the methods explained here (which are partially physics-based), we have also been experimenting with machine learning (ML) methods. A simple ML method using linear regression on bulk density and neutron porosity data achieved a slightly higher R² value of 0.81. The higher correlation value for the machine learning model has prompted us to investigate some of these possibilities in further study. Future work will compare the merits and pitfalls of using less complicated, single variable models to more complex multivariable regression or machine learning models. For our purposes, the single variable models were still highly correlative, simple to deploy, and applicable to sparse data sets. We believe they provide a useful and elegant solution without a large sacrifice of accuracy.
The data that support the findings of this study are available from the corresponding author upon reasonable request. Volve data is courtesy of Equinor and is used in accordance with the Equinor Open Data Licence: https://cdn.equinor.com/files/h61q9gi9/global/de6532f6134b9a953f6c41bac47a0c055a3712d3.pdf?equinorhrsterms-and-conditions-for-licence-to-data-volve.pdf
The authors wish to thank our colleagues at Land Seismic Noise Specialists, Ank Geocsciences LLC, and Collier Geophysics for the resources to complete this work. Mr Chesser wishes to additionally thank Neil Peake who has allowed him to continue dabbling in seismic petrophysics, well into what otherwise would
be retirement as well as former employers and supervisors who let him develop novel workflows at the risk of daily profits, including Fred Hilterman, Gabino Castillo, Rob Mayer, Brad Bacon, and the late J.W Humbard. Mr. Morgan extends his gratitude to his wife Kerry for editorial help and encouragement.
Amerada Hess Netherlands Ltd. [1999]. https://www.nlog.nl/ nlog-mapviewer/brh/106531513?lang=en.
Apache Energy. [2010]. https://public.neats.nopta.gov.au/nopims/wells.
Asquith, G. and Krygowski, D. [2004]. Porosity Logs in Basic Well Log Analysis. AAPG Methods in Exploration Series, 16, p. 37-76.
Belknap, J.T. [1959]. API Calibration Facility for Nuclear Logs, Drilling and Production Practice. American Petroleum Institute in Gamma Ray, Neutron and Density Logging, SPWLA Reprint Volume, March, 1978.
Castagna, J.P., Batzle, M.L. and Kan, T.K. [1993]. Rock Physics – The link between rock properties and AVO response. Offset Dependent Reflectivity – Theory and Practice of AVO Analysis, Investigations in Geophysics, 8, p. 135-171.
Chesser, K. [1997]. Missing Data Problems in Forward AVO Modeling. SEG Annual Meeting, Expanded Abstracts.
Croizé, D., Renard, F. and Gratier, J.P. [2013]. Compaction and Porosity Reduction in Carbonates: A Review of Observations, Theory, and Experiments. Advances in Geophysics, 54, p. 181-238.
Elf Petroland B.V. [1999]. https://www.nlog.nl/nlog-mapviewer/ brh/106520549?lang=en.
Equinor. [2018]. Volve Field Dataset. https://www.equinor.com/energy/ volve-data-sharing.
Faust, L.Y. [1951]. A Velocity Function Including Lithologic Variation. Geophysics, 18, p. 271-288
Gardner, G.H.F., Gardner, L.W. and Gregory, A.R. [1974]. Formation Velocity and Density – The Diagnostic Basics for Stratigraphic Traps. Geophysics, 39, p. 770-780.
Mavko, G., Mukerji, T. and Dvorkin, J. [2003]. The Rock Physics Handbook. Cambridge University Press.
Raymer, L.L, Hunt, E.R. and Gardner, J.S. [1980]. An improved sonic transit time-to-porosity transform. Transactions of the Society of Professional Well Log Analysts, 21st Annual Logging Symposium, Paper P.
Trad, D. [2014]. Five-dimensional interpolation: New directions and challenges. CSEG Recorder, 39(3), p. 22-29.
United States Department of Energy. [2001]. https://wiki.seg.org/wiki/ Teapot_dome_3D_survey. Woodside Energy. [1986]. https://public.neats.nopta.gov.au/nopims/wells. Wylie, M.R.J, Gardner, G.H.F. and Gregory, A.R. [1963]. Studies of Elastic Wave Attenuation in Porous Media. Geophysics, 27, p. 569-589.
Geoscience modelling and interpretation techniques are vital for finding the hydrocarbons that the world still needs but also for modelling reservoirs for carbon capture and storage and other renewable energy sites.
Artificial intelligence and machine learning are leading to advances. The latest high-performance computing and algorithmic innovations, aimed at increasing the quality and quantity of modelling and interpretation are presented here –demonstrating how plays can be opened up for explorers.
Michael S. Zhdanov et al present an approach based on the contraction integral equation (CIE) method of numerical electromagnetic modelling and adaptive regularised inversion.
Chris Gravestock et al present the exploration potential of six play fairways in the Middle East, which span the Paleozoic and earliest Mesozoic.
Vita Kalashnikova et al demonstrate the seismic-driven identification of possible sandstone distribution within the Rogaland Group of the Norwegian North Sea.
Julien Razza et al introduce an interactive framework that directly addresses both through an innovative three-step workflow.
Karyna Rodriguez et al present the geoscience evidence that suggests there is huge potential in blocks that are currently in the 2025-26 Nova Scotia licensing round.
Sonny Winardhi et al integrate extrapolated low-frequency data into a sourceindependent FWI workflow, significantly improving the resulting velocity model and enabling reliable inversion without requiring accurate source waveform estimates.
David Little presents datasets matched and merged into a single interpretable volume, called 2Dcubed[VA1], and used to gain a regional overview enabling structural framework and consistent regional interpretation that was calibrated to 86 exploration and appraisal wells.
First Break Special Topics are covered by a mix of original articles dealing with case studies and the latest technology. Contributions to a Special Topic in First Break can be sent directly to the editorial office (firstbreak@eage.org). Submissions will be considered for publication by the editor.
It is also possible to submit a Technical Article to First Break. Technical Articles are subject to a peer review process and should be submitted via EAGE’s ScholarOne website: http://mc.manuscriptcentral.com/fb
You can find the First Break author guidelines online at www.firstbreak.org/guidelines.
January Land Seismic
February Digitalization / Machine Learning
March Reservoir Monitoring
April Underground Storage and Passive Seismic
May Global Exploration
June Navigating Change: Geosciences Shaping a Sustainable Transition
July Reservoir Engineering & Geoscience
August Environment, Minerals and Infrastructure
September Modelling / Interpretation
October Energy Transition
November Marine Acquisition
December Data Management and Processing
More Special Topics may be added during the course of the year.
Michael S. Zhdanov1,2, Alexander Gribenko1, Leif Cox1, Keiichi Tanabe3, Tsunehiro Hato3, and Akira Tsukamoto 3 present an approach based on the contraction integral equation (CIE) method of numerical electromagnetic modelling and adaptive regularised inversion.
Introduction
The Superconducting Quantum Interference Device (SQUID) Transient Electromagnetic (TEM) survey is a cutting-edge geophysical method for exploring the Earth’s subsurface, particularly for mineral, geothermal, and groundwater resources. It consists of an electric bipole or induction loop transmitter and multiple SQUID magnetic field receivers located over the study area. The advantage of the SQUID receivers is that they can detect signals much weaker and at later times than conventional induction coils or fluxgate sensors. For example, SQUID measures the components of the transient magnetic field, B(t), generated by the transmitter with high precision. In contrast, the induction coils measure the time derivatives dB(t)/dt. It is well known that the magnetic field decays much more slowly with time than its time derivative; thus, recording the data at a later time corresponds to a greater depth of investigation. In addition, using the electric bipole or an induction loop as a controlled source makes it possible to conduct repeated sounding of the subsurface and stack the recorded response. This results in a significant increase in the signal-to-noise ratio and, therefore, the method’s sensitivity. Thus, due to its high sensitivity and late-time response, SQUIDTEM can distinguish deeper and lower-conductivity targets more effectively than the conventional TEM method.
The SQUID-TEM survey can be applied in mineral exploration (particularly for deep ore bodies, massive sulphides, and nickel deposits), geothermal reservoir mapping, groundwater exploration, CO2 sequestration, and oil and gas reservoir monitoring.
The SQUID-TEM survey leverages the unmatched sensitivity of superconducting quantum magnetometers to extend the depth, resolution, and effectiveness of transient electromagnetic methods for deep and subtle geophysical targets. The proper modelling and inversion techniques for SQUID-TEM data should be developed to utilise the benefits of the SQUID-TEM method fully. This paper presents an approach to solving this problem based on the contraction integral equation (CIE) method of numerical electromagnetic modelling and adaptive regularised inversion. The
1 TechnoImaging | 2 The University of Utah | 3 SUSTEC
method is illustrated by the case study of modelling and inversion of SQUID-TEM survey data collected in the Geothermal Project Area in Southern Sumatra, Indonesia.
In the framework of the SQUID-TEM method, the array of SQUID sensors is located on the ground over the study area. It records the three magnetic field components of the EM field generated by the electric bipole or induction loop transmitter (Zhdanov et al., 2025). The recorded data are then transformed into a 3D model of the resistivity distribution in the subsurface. The 3D resistivity model can provide a volume image of the geothermal reservoir.
Finite-difference time-domain (FDTD), finite element (FE) methods, and integral equation methods are commonly used for time-domain EM modelling (Zhdanov, 2018). This paper considers the 3D numerical modelling based on the contraction integral equation (CIE) method (Hursán and Zhdanov, 2002; Zhdanov, 2002). This method offers several key advantages over the finite-difference (FD) and finite-element (FE) methods in modelling electromagnetic (EM) fields for geophysical exploration, particularly in cases involving highly heterogeneous or complex geological structures. The IE methods are particularly well-suited for modelling localised conductivity anomalies (e.g., mineral ore bodies, geothermal reservoirs) embedded in a homogeneous or layered host. The background can be treated analytically, and only the anomalous region is discretised, reducing computational effort.
This advantage can be extended to models with arbitrary backgrounds (e.g., those with topography) using an approach based on the formulation of the integral equation method with inhomogeneous background conductivity (IE-IBC) (Zhdanov et al., 2006). The IE IBC method is based on separating the effects related to the excess electric current induced in the inhomogeneous background domain from those associated with the anomalous electric current in the location of the anomalous
* Corresponding author, E-mail: mzhdanov@technoimaging.com DOI: 10.3997/1365-2397.fb2025066
conductivity, respectively. As a result, we arrive at a system of integral equations that uses the same simple Green’s functions for the layered model as in the original IE formulation. However, the new equations take into account the effect of the variable background conductivity distribution. This approach is instrumental in solving an inverse problem, where we must maintain some known geologic structures unchanged during the iterative inversion.
The conductivity is represented as a superposition of the background conductivity model and anomalous conductivity. The observed EM field is divided into anomalous and background components as follows: E=Ea+Eb; H=Ha+Hb. The background fields are computed for the model with the background conductivity using the semi-analytical method. The anomalous fields are calculated separately, which increases the method’s accuracy and decreases computation time. The anomalous electric and magnetic fields satisfy the following operator equations: (1)
where GE and GH are integral electric and magnetic Green’s operators, and ED is the total electric field inside the domain with a 3D distribution of anomalous conductivity ∆σ.
Note that Green’s operators become matrices after the discretisation of equations (1), and anomalous fields can be obtained by simple matrix-vector multiplication. Equations (1) hold everywhere, including the anomalous domain D. The first step of the IE solution is to find a domain field, ED by solving the system of equations formulated for the anomalous domain. In numerical dressing, the domain equation for the total electric field, ED, takes the form:
where m is a vector of the anomalous conductivities, ∆σ of the model determined on some discretisation grid, I is the identity matrix, GE is the matrix of volume-integrated Green’s tensors for the background conductivity model.
As mentioned above, we use the contraction form of equation (2), which can be obtained by applying some linear transformation to Green’s operator, GE, to create a new contraction Green’s operator with a norm of less than one. The specific form of this linear transformation is motivated by the energy inequality for the anomalous electromagnetic field, which expresses a fundamental physical fact that the energy flow of the anomalous field outside the domain of anomalous conductivity is always non-negative. In particular, the contraction integral equation for the total electric field, ED, takes the form:
(3)
where the diagonal contraction preconditioner matrices, M1 and M2 are defined by the anomalous and background conductivity distributions.
The system of linear equations above is solved using the Complex Generalised Minimal Residual (CGMRM) method, which always converges, ensuring a robust performance of the
forward modelling algorithm (Zhdanov, 2015, 2018). After ED is found, one can compute electric and magnetic fields anywhere, including receiver locations, using equations (1).
We should also note that the CIE method is based on Green’s function solutions, which exactly satisfy Maxwell’s equations for the background medium. This ensures high physical fidelity and accuracy near source or field singularities, which FD/FE methods may struggle with unless refined meshes are used. Since only the anomalous region is discretised, fewer computational resources (memory and CPU time) are needed compared to FD/FE methods that discretise the entire domain. This can be critical for 3D EM modelling, where mesh sizes can become prohibitive. Finally, CIE formulations with inhomogeneous background domains enable fast-forward solvers, which are particularly well-suited for iterative nonlinear inversion methods commonly used in geophysics.
The CIE method is performed in the frequency domain. The corresponding frequency-domain EM fields must be transformed to the time domain to model the field survey system. An optimal set of frequencies must be selected for the modelling. The frequencies must contain the entire spectrum of the instrument response. The frequency-to-time domain transform is based on a cosine transformation of the imaginary part of the frequency spectrum to a step response. The step response is then convolved with the time derivative of the transmitter current waveform.
The SQUID-TEM inverse problem is ill-posed, as many different subsurface conductivity models can equally well explain the SQUID data, and the solution may be unstable or non-unique. Regularisation introduces additional information or constraints (a ‘stabiliser’ or ‘stabilising functional’) to select a unique, stable, and geologically plausible solution.
The regularised inversion is carried out by minimising the Tikhonov parametric functional, Pa(m):
Pa(m) = ϕ(m) +aS(m) → min, (4)
where m is a vector of the unknown parameters (anomalous conductivities), ϕ(m) is a least square misfit functional between the observed and predicted SQUID-TEM data for a given model m, S(m) is a stabilising functional, and a is the regularisation parameter that balances the misfit and stabilising functional. Data and model weights are introduced to equation (4) through data and model weighting matrices, which are selected based on their integrated sensitivity. As a result, they provide equal sensitivity of the observed data to cells located at different depths and at different horizontal positions. Thus, our weighting functions automatically introduce appropriate corrections for the vertical and horizontal distribution of the resistivity.
The geological constraints can be enforced through a choice of data and model weights, model upper and lower bounds, an a priori model, and the type of stabilising functional. The latter incorporates information about the class of models used in the inversion. The choice of stabilising functional is based on the user’s geological knowledge and prejudice.
The standard stabilisers are typically minimum-structure (smoothness) or damping (smallness) functionals, promoting
smooth, distributed models. As an example of a smoothness stabiliser, we can refer to a minimum norm (MN) stabiliser that seeks to minimise the norm of the difference between the current conductivity model m, and an a priori conductivity model, mapr:
The minimum norm stabiliser usually produces a relatively smooth model.
Focusing stabilising functionals can be used to recover models with sharp boundaries, compact anomalies, or blocky features — more geologically realistic for many mineral and structural targets, including geothermal reservoirs.
For example, the minimum support (MS) stabiliser, SMS(m) , minimises the volume where the anomalous conductivity can be distributed:
approach can be briefly described as follows: for a given receiver, compute and store the Fréchet derivative for those inversion cells within a predetermined horizontal distance from this receiver, i.e., the sensitivity domain. The radius of the sensitivity domain is based on the rate of sensitivity attenuation. The application of the moving sensitivity domain approach to the inversion of SQUID-TEM data enables consideration of large-scale models, if necessary.
Case study: SQUID-TEM survey in the Raja Basa Geothermal Project Area in South Sumatra, Indonesia
where e > 0 is a small number called a focusing parameter. The smaller the focusing parameter is, the more focused the inverse image of the anomalous domain becomes.
In order to produce an image with sharp boundaries, one can use the minimum gradient support stabilising functional:
The focusing stabilisers promote blocky, focused, or sparse models (as opposed to smooth models), which helps us to recover sharp boundaries and compact features, matching realworld geological structures more closely than traditional smooth regularisation.
Minimisation of the parametric functional (4) requires calculating its variation on every iteration step. We base our solution on the re-weighted regularised conjugate gradient (RRCG) method (Zhdanov, 2002, 2015). This method iteratively updates the vector of model parameters m to minimise the vector of residual errors, ri, (difference between the predicted and observed data on the current iteration, i), akin to: (8)
where ki is a step length, SF is the derivative of the corresponding stabiliser, and FT is the conjugate transpose of the Fréchet (sensitivity) matrix, which is computed using quasi-Born approximation. The inversion proceeds to iterate in a manner similar to equation (8) until the residual error reaches a preset threshold or a maximum number of iterations is reached. Upon completion, the quality of the inversion is appraised by the data misfit and visual inspection of the model.
The Raja Basa Geothermal Project Area is situated within the tectonically active Sunda Arc region, located on the southern tip of Sumatra, Indonesia. The area is dominated by the Raja Basa volcano, a prominent Quaternary stratovolcano that forms the backbone of the geothermal system. The Raja Basa volcano is the principal geological structure, composed mainly of andesitic to dacitic lava flows, pyroclastic deposits, tuffs, and volcanic breccias. The summit area and flanks are marked by craters, fumaroles, and active hydrothermal manifestations (hot springs, steaming ground). The area is influenced by the Sumatra Fault System — a major strike-slip fault that runs parallel to the west coast of Sumatra. This tectonic activity has facilitated deep fracturing and the ascent of magmatic fluids, which play a crucial role in the development of geothermal reservoirs. The main geothermal reservoir is hosted within fractured volcanic rocks (andesites, tuffs, breccias) beneath a relatively low-permeability cap rock of altered volcanic material.
The Raja Basa Geothermal Project Area represents a young, active volcanic system with abundant surface and subsurface hydrothermal activity, significant tectonic fracturing, and well-developed geothermal alteration and reservoir zones, making it a promising target for geothermal energy development.
Figure 1 presents a schematic diagram of the geothermal system structure, which is similar to other high-temperature volcanic geothermal systems in the Indonesian volcanic arc. The system is driven by magmatic heat associated with volcanic activity. The typical electrical properties of the system rocks can be summarised as follows. At the surface lies fresh volcanic rock, exhibiting high resistivity — generally above 100 Ohm-m — representing a dry, unaltered andesitic cap.
We also use the concept of the moving sensitivity domain for 3D inversion of the SQUID-TEM data (Cox and Zhdanov, 2008). Within this framework there is no need to calculate responses or sensitivities beyond the receiver’s sensitivity domain. The sensitivity matrix for the entire 3D earth model could be constructed as the superposition of the sensitivity domains for all receivers. This Figure 1
Conductive alteration zone (clay cap) formed by hydrothermal alteration (argillic alteration zone) lies beneath. This zone traps heat and fluids. The main geothermal reservoir is located at depths ranging from 200 m to 1500 m, where temperatures often exceed 220°C. The reservoir rocks are composed of fractured andesite, dacite, tuff, and breccia, often with chlorite, epidote, and sometimes adularia as alteration minerals. This is the main geothermal fluid-bearing zone. Older volcanic, intrusive, or metamorphic rocks form the deeper, unaltered basement rocks, which usually exhibit low porosity and high resistivity (>100 Ohm-m).
Exploration at Raja Basa includes geological, geochemical, and geophysical surveys aimed at delineating the extent of the
geothermal resource and estimating its potential for power generation. Superconducting Sensor Technology Corporation (SUSTEC) conducted a SQUID-TEM survey in the Raja Basa geothermal field, which consisted of a long transmitting electric bipole and 40 observation points. Figure 2 presents satellite imagery of the Raja Basa SQUID-TEM survey. The red line represents the path of the electric current transmitter, while the yellow crosses indicate the locations of the SQUID receivers. The collected data included amplitudes of the three components of the magnetic B-field at 40 locations, along with their standard deviations at 33 time gates ranging from 2e-05 to 1.4 seconds.
We have applied the 3D modelling and inversion methodology outlined above to the observed SQUID-TEM survey data. A time range between 0.0001 and 1.4 seconds was used for inversion. The data within this time range are most sensitive to the subsurface resistivity model and contain little noise. The dimensions of the inversion domain were 7000 by 8000 by 3500 m in X (Easting), Y (Northing), and Z (Elevation) directions. A uniform grid of 25 m by 25 m by 25 m cells was used in the modelling and inversion. Figure 3 shows the inversion domain dimensions and the discretisation.
The inversion was performed using the moving sensitivity domain approach (Cox and Zhdanov, 2008), with a domain radius of 4 km centered around each receiver. We have used the upper and lower resistivity boundaries of 10,000 Ohm-m and 0.1 Ohmm, respectively, for the allowed resistivity variations. The inversions converged to a global RMS (root mean square) misfit of about 3.6, a statistically reasonable value indicating a relatively low misfit level between the observed and predicted data.
Figure 4 shows the station-by-station distribution of the RMS misfit. For most stations, the RMS is below 2, confirming the accuracy of the data fitting by the geoelectric model produced through inversion.
Figures 5 and 6 represent 3D views of the inversion result. One can identify two strong, connected conductive structures
Figure 4 Selected station-by-station distribution of the local RMS misfit. The bold dots show the positions of the SQUID stations. The colour of the dots indicates the value of the RMS misfit between the observed and predicted data. The black line schematically shows the position of the transmitter.
in the South-Central area of the survey at depths between 500 m and 1200 m, as shown by the gold colour in these images. These structures could represent geothermal reservoirs with generally low resistivity (below 15 Ohm-m), a characteristic typical of high-temperature geothermal fields associated with volcanic systems. Magnetotelluric (MT) and electrical resistivity surveys in similar volcanic geothermal systems worldwide, as well as in Indonesian analogs, consistently demonstrate that productive geothermal reservoir zones exhibit resistivity values in the 5 to 15 Ohm-m range. This low resistivity range primarily reflects the presence of high-temperature saline fluids and pervasive hydrothermal alteration, particularly the conversion of primary minerals to conductive clay assemblages such as illite and chlorite below the argillic (smectite-rich) clay cap. The 5-15 Ohm-m range is consistent with resistivity structures imaged in MT surveys at other high-temperature volcanic geothermal systems, such as Rotokawa (New Zealand) and Xinzhou (China), where the main reservoir also coincides with a moderate-resistivity anomaly reflecting the interplay of hot saline fluids and alteration zones (Heise et al., 2008; Han et al., 2021).
Figures 7 and 8 show the horizontal section at 500 m below sea level and the vertical section, respectively, which outline the horizontal and vertical extent of the conductive reservoirs.
Thus, the SQUID-TEM survey reveals that the Raja Basa geothermal system displays a classic resistivity signature characteristic of volcanic environments: a high-resistivity volcanic cap, conductive clay alteration, and deeper conductive, fluid-bearing reservoir rocks. These conductive zone positions appear consistent with the previous geothermal reservoir’s conceptual model, based on the region’s geological structure and drilling results (Figure 1). The low resistivity of geothermal reservoirs primarily reflects the presence of high-temperature saline fluids and pervasive hydrothermal alteration.
The SQUID-TEM survey method is an advanced electromagnetic geophysical technique that can be used for high-resolution geothermal resource exploration, particularly effective in resistivity imaging of deep and conductive geothermal reservoirs.
This paper presents a modelling and inversion methodology for a new method of geothermal exploration based on SQUID magnetic receivers and controlled-source transmitters. This method has significant advantages over the traditional magnetotelluric method, providing better resolution of geoelectrical structures due to the use of a controlled source (electrical
current bipole or induction loop) and the high sensitivity of SQUID receivers. The ultra-high sensitivity of SQUID sensors enables the detection of very weak signals, allowing for exploration to greater depths and in high-conductivity environments. High dynamic range and bandwidth are also essential for accurately resolving shallow and deep resistivity features. By using the controlled source, we avoid the limitations of the MT method related to the plane-wave structure of the primary field.
The developed 3D modelling and inversion methodology uses the contraction integral equation (CIE) method, which is particularly advantageous in geophysical EM modelling when targeting localised anomalies in a known background, offering computational efficiency, high accuracy, and natural treatment of unbounded domains — features that make it especially attractive in mineral and geothermal exploration applications. The method generates detailed 3D geoelectrical images of the subsurface from the SQUID data.
These advances provide practising geoscientists with new geophysical tools for exploring not only geothermal but also mineral, oil, and gas deposits.
The authors acknowledge TechnoImaging, SUSTEC, the Consortium for Electromagnetic Modeling and Inversion (CEMI) at the University of Utah, and the Supreme Energy and Sumitomo Corporation for their support of this project.
Cox, L.H. and Zhdanov, M.S. [2008]. Advanced computational methods of rapid and rigorous 3-D inversion of airborne electromagnetic data. Communications in Computational Physics, 3(1), 160- 179.
Han, Q., Kelbert, A. and Hu, X. [2021]. An electrical conductivity model of a coastal geothermal field in southeastern China based on 3D magnetotelluric imaging. Geophysics, 86( 4), B265-B276.
Heise, W., Caldwell, T., Bibby, H. and Bannister, S. [2008]. Three-dimensional modeling of magnetotelluric data from the Rotokawa geothermal field, Taupo Volcanic Zone, New Zealand. Geophysical Journal International, 173, 740-750.
Hursán, G. and Zhdanov, M.S. [2002]. Contraction integral equation method in three-dimensional electromagnetic modeling. Radio Science, 37(6), 1089.
Zhdanov, M.S. [2002]. Geophysical inverse theory and regularization problems. Elsevier.
Zhdanov, M.S. [2015]. Inverse theory and applications in geophysics. Elsevier.
Zhdanov, M.S. [2018]. Foundations of geophysical electromagnetic theory and methods. Elsevier.
Zhdanov, M.S., Gribenko, A., Cox, L., Tanabe, K., Hato, T., Tsukamoto, A. and Wakana, H. [2025]. Application of Superconducting Quantum Magnetic Sensors Based Transient Electromagnetic Method (SQUID-TEM) for Geothermal Resource Exploration. 86th EAGE Annual Conference & Exhibition, Extended Abstracts
Zhdanov, M.S., Lee, S.K. and Yoshioka, K. [2006]. Integral equation method for 3D modeling of electromagnetic fields in complex structures with inhomogeneous background conductivity. Geophysics, 71, G333–G345.
Chris Gravestock1*, Owen Sutcliffe1, Thomas Jewell1, Mike Simmons1 and Joseph Jennings1 present the exploration potential of six play fairways in the Middle East, which span the Paleozoic and earliest Mesozoic.
Introduction
The Mesozoic and Cenozoic petroleum systems of the Middle East are renowned for producing large volumes of hydrocarbons to supply global demand. Thus far, considerable exploration and production activity has been focused on these shallower petroleum systems, helping to locate many of the giant and supergiant hydrocarbon fields, particularly within the Gulf region and adjacent to it. However, the contributions from the Paleozoic are still significant. Vast reserves of gas in the Permo-Triassic Khuff carbonates, for example, are charged from the world-class Silurian Qusaiba Formation which is widespread across Arabia. Within the public domain, there are large volumes of multi-disciplinary data to assess, understand, analyse, and determine the distribution of petroleum system elements associated with the Mesozoic and Cenozoic plays. The availability of such valuable data begins to fade beyond the Permian, further reducing to isolated ‘pockets’ of data for the Precambrian-Cambrian. This lack of subsurface penetrations into the Paleozoic stratigraphy results in uncertainty in the distribution of petroleum system elements with increasing depth. To date, successful Paleozoic discoveries are clustered in Central Arabia (Ghawar area), the Arabian Gulf, Zagros, and Oman.
To further illuminate the exploration potential of this underexplored Paleozoic interval in the Middle East, a plate-wide screening approach has been conducted using the data and insights within the Neftex® solution from Halliburton Landmark to provide a valuable first-pass contribution to help identify
‘zones’ of exploration potential with multiple stacked opportunities for plays. This article will focus on six play fairways which span the Paleozoic and earliest Mesozoic, namely:
• Middle-Late Ordovician Qasim Formation
• Latest Ordovician (Hirnantian) Sarah Formation
• Early-Middle Devonian Jauf Formation
• Permo-Carboniferous Unayzah Formation
• Permo-Triassic Khuff Formation
• Middle-Late Triassic Jilh Formation
Over the next 30 years (to 2050 and beyond), the energy needs of global society will continue to create demand for a certain amount of oil and perhaps more importantly, a significant amount of gas as part of the overall energy mix within the energy transition (Davies and Simmons, 2021). Crucially, ongoing production from existing gas assets will be insufficient to meet predicted estimates, requiring new exploration to assist in locating additional gas reserves in the subsurface (Figure 1). Yet where will these required reserves come from? Wood Mackenzie (2022) have categorised the produced and remaining reserves from several highly petroliferous basins across the globe. Within the top five are three Middle Eastern basins, namely the Rub al Khali, Widyan, and Zagros basins with huge yet-to-find values. Remaining reserves in these basins are vast with approximately 600 Billion barrels of oil equivalent (Bboe) in the Rub al Khali Basin, ~450 Bboe in the Widyan Basin, and ~150 Bboe in the Zagros Basin (Wood
1 Halliburton
* Corresponding author, E-mail: Chris.Gravestock@halliburton.com
DOI: 10.3997/1365-2397.fb2025067
Figure 1 Projections of future gas demand versus supply in typical slow and rapid energy transition scenarios. In either scenario, for the period to 2050 the gap between total gas demand and supply from existing producing assets is substantial (54,577 Bcm in a slow energy transition scenario, 29,334 Bcm in a rapid energy transition scenario). This requires the gap in supply to be met by a combination of increased recovery, commercialisation of stranded assets, and –most notably – new exploration success. Methodology follows Davies and Simmons (2021) and uses a range of industry forecasts summarised as means.
Mackenzie, 2022). This highlights the importance of the Middle East being one of the main contributors to future supply of oil and gas with perhaps a significant proportion of the gas supply being within Paleozoic stratigraphy, given the prevalence of gas-mature source rocks within this stratigraphy associated with under-explored reservoirs and seals. Therefore, the work discussed herein provides a valuable first-pass contribution to help with the effort of locating ‘zones’ of exploration potential.
The Paleozoic stratigraphy of the Middle East represents a significant and underexplored succession. Aside from the Precambrian and early Cambrian stratigraphy which contains significant petroleum carbonates (especially in Oman), much of the stratigraphic succession is dominated by siliciclastics (see Sutcliffe, 2016 and Figure 2) and is related to the gradual movement of Arabia from equatorial latitudes toward southern high latitudes (see Figure 3.3 of Sharland et al., 2001) which induced important glacial periods such as the Late Ordovician (Hirnantian) glaciation.
The Paleozoic successions contain multiple petroleum systems with some of the most significant located in Oman charged by the Precambrian-Cambrian Huqf Supergroup. Elsewhere across the Arabian Plate, the dominance of gas is derived from the world-class, organic-rich shales of the early Silurian Qusaiba Formation and its equivalents. Its distribution across the Arabian Peninsula relates in part to its preservation beneath the Late Paleozoic ‘Hercynian’ unconformity, the occurrence of major structural depressions and paleobathymetric variations related to the Late Ordovician (Hirnantian) glaciation. Depending on its location in the subsurface, it can charge both younger and older Paleozoic stratigraphy (see Sutcliffe and Cousins, 2018). At present-day depths, large areas of the Silurian are overmature and therefore require an understanding of charge timing, the nature of migration pathways either laterally, vertically or multiple ‘mobile’ phases of charge related to structural development of the region.
Subsurface expression of the Paleozoic stratigraphy
Across the Middle East, the expression and preservation of the Paleozoic stratigraphy is varied, which is related to the impact of the ‘Hercynian’ orogeny (see Sutcliffe and Cousins, 2018). Additionally, the Paleozoic stratigraphy is impacted by later significant events such as Permo-Triassic rifting in the Palmyrides. These key events in Arabian Plate history have led to a geologically complex Paleozoic succession beneath the ‘Hercynian’ unconformity, resulting in multiple episodes of basin lineament reactivation.
The ‘Hercynian’ Orogeny, which is the result of the collision of Gondwana and Laurussia, formed the Pangea supercontinent. This collision resulted in the uplift of Paleozoic stratigraphy, non-deposition, and differential erosion across the plate. As expressed in Figure 2, in some places, the ‘Hercynian’ unconformity has completely eroded many of the Paleozoic petroleum system elements, incising down into the underlying Cambrian stratigraphy. Therefore, the expression of the preserved stratigraphy beneath the ‘Hercynian’ unconformity, i.e., the subcrop, must be considered in any plate-wide screening effort to determine
Figure 2 Simplified NW-SE Neftex® chronostratigraphic chart across the Arabian Plate to highlight the dominance of Paleozoic siliciclastic deposition across the Middle East and the impact of differential erosion on the preservation of Paleozoic stratigraphy from the ‘Hercynian’ Unconformity.
exploration potential in places where petroleum system elements may have been eroded.
In the identified six fairways spanning the Paleozoic to earliest Mesozoic, a plate-wide screening methodology is applied, in which the highest petroleum system element risk is carried forward through to the final Combined Common Chance Map (CCCM). This approach illuminates ‘zones’ across Arabia where an individual play has maximum potential. Each of the input parameters e.g., source rock presence, is derived from the integration of multi-disciplinary subsurface datasets that are tied to a stratigraphic framework. Each of the identified play fairways were assessed on an individual basis with a focus on charge from the Silurian Qusaiba Formation and its equivalents, with an arbitrary migration of 80 km from the source kitchen. The results from this individual play fairway analysis approach were then stacked to allow the final output to be engineered, which highlighted areas of stacked play potential (Figure 3). Within the results and conclusion section, the outcomes of this process and associated petroleum systems risk for each fairway will be discussed in more detail.
By reviewing the final CCCM of each fairway, ‘zones’ of potential can be quickly located across the Arabian Plate (Figure 4). These ‘zones’ can then undergo more detailed investigation utilising proprietary datasets to identify exploration potential, undergo risk analysis, and develop a portfolio of opportunity.
The Middle-Late Ordovician Qasim Formation and its equivalents represent a set of alternating sandstone and shales deposited on a stable shelf setting that is well preserved in the subsurface
across central and northern Arabia (Figure 2). Preserved equivalents of the Qasim Formation in Oman are largely restricted to the axis of the major salt basins due to erosion from the late Carboniferous ‘Hercynian’ unconformity. Plate-wide, there is considerable potential represented in the Qasim fairway, with widespread potential across Saudi Arabia, Bahrain, Qatar, western and northern Iraq, Syria, and southern Turkey. However, as this screening analysis considers charge from the overlying Silurian Qusaiba Formation, areas where the Qasim Formation and its equivalents are preserved on major paleo-highs could be more favourable, where charge from the onlapping Silurian in the adjacent topographic lows can migrate up-dip and along major basement faults. Other favourable ‘zones’ for the Qasim are areas that have undergone rifting such as the Palmyrides which may favour the juxtaposition of source rock and reservoir as well as generating potential trapping mechanisms. Other considerations for charge may include downward-migration where pressure differences exist between the overlying source rock (being over pressured) and atmospheric pressure existing in the underlying reservoir (Yangfang et al., 2009). An effective seal is not a risk for this play due to the alternation of sandstones and thick shale sequences (Figure 2). However, in northern Iraq (Omer and Friis, 2014) and the Rub al Khali Basin (Hui et al., 2016), quartz cementation represents a major risk for this play due to depth of burial reducing reservoir quality, requiring chlorite coating of sand grains or early charge to preserve primary porosity.
Latest Ordovician (Hirnantian) Sarah Formation
Potential in the Latest Ordovician (Hirnantian) Sarah Formation, compared to the Qasim Formation, is limited to northern Saudi Arabia and western Iraq. As this reservoir is glacially influenced, related to the movement of Arabia to southern high latitudes (Figure 2.3 of Sharland et al., 2001), it is vitally important to understand glacial maxima and minima as this will have important implications on the sedimentological model to predict the distribution and presence of potential reservoir facies (e.g. Huuse et al., 2012; Kurjanski et al., 2020). Additionally, the
glaciers tended to erode deeply into very mud-prone underlying stratigraphy (Figure 2) and it is likely that the facies associated with the Sarah Formation will be very heterogeneous and therefore pose limited connectivity between reservoir bodies. As discussed above in the Qasim Formation, charge (whether up-dip, downward migration or juxtaposition of source rock) will be the key risk to play potential.
Early-Middle Devonian Jauf Formation
Moving through the stratigraphic succession to the Early-Middle Devonian Jauf Formation, much of the Devonian stratigraphy has been eroded (Figure 2) due to the impact of the ‘Hercynian’ unconformity and the earlier Mid-Devonian unconformity across Saudi Arabia, related to plate reorganisation associated with subduction and back-arc rifting of the Proto-Tethys (Sharland et al., 2001). This uplift affected the long-lived Ghawar High, causing it to be exposed and eroded (Al-Husseini, 2000; Sharland et al., 2001). The Early and Middle Devonian stratigraphy was eroded during this event, particularly in southern Saudi Arabia, but the amount of stratigraphy removed diminishes towards the east. The presence of reservoir is a major risk due to the complex geology and preservation of the Jauf Formation in the subcrop beneath the ‘Hercynian’ unconformity (Figure 4). Notably, the Devonian stratigraphic package is absent across many of the northward paleo-topographic highs across Arabia and is only preserved in topographic lows and the flanks of structures (Figure 4).
Permo-Carboniferous Unayzah Formation
The deposits of the Permo-Carboniferous Unayzah Formation and its equivalents vary from glacially influenced like that of the latest Ordovician (Hirnantian) Sarah Formation to fluviatile and/ or aeolian and overly the ‘Hercynian’ unconformity. The Unayzah formation is conventionally split into three members (Unayzah A, B and C), but a fourth between the A and B members was identified by Melvin and Sprague (2006) named the Unnamed Middle Unayzah Member. The oldest member is assigned to the Unayzah
Figure 3 A worked example of the individual final CCCM for the fairways (B, C, and D). These are then stacked on top of one another to illuminate ‘zones’ of stacked play potential (A).
Figure 4 A simplified worked example of the Final Combined Common Chance maps for the Early-Middle Devonian Jauf Formation. Note, lateral and up-dip migration from the Qusaiba Formation source kitchens to a distance of 80 km indicated by the black lines.
C, consisting of a monotonous succession of quartzose sandstones deposited in a glacio-fluvial braided plain. The overlying Unayzah B Member contains a diverse and complex suite of depositional elements ranging from terminal moraines, glacially-influenced lacustrine deposits, and diamictites with less common braided stream and aeolian deposits (Melvin and Sprague, 2006). The Unnamed Middle Unayzah Member consists of very shallow lacustrine silts and sandstones deposited in an extensive alluvial braided plain with the Unayzah A Member indicative of aeolian sandstones (Macdonald et al, 2010) with braided fluvial deposits also recorded. This highlights the complexity in the Unayzah Formation reservoir and therefore requires a strong understanding of depositional and sedimentological models to identify where the best reservoir packages may be in the subsurface.
Unlike the Sarah Formation, the ‘zones’ for exploration potential across Arabia associated with the Unayzah are vast in comparison. In Omani equivalent stratigraphy, additional risks associated with this play are associated with biodegradation and the formation of heavy oil due to shallow burial and the infiltration of meteoric waters. It is important to note that in Oman almost all the Silurian stratigraphy has been eroded and therefore, the oil will be associated with charge from the Precambrian-Cambrian Huqf Supergroup.
Khuff and Jilh formations
Moving into the early Mesozoic signals a dramatic change in facies from the dominant siliciclastic deposits that comprise the Paleozoic to shallow-marine carbonates. This is related to the gradual movement of the Arabian Plate northward to equatorial latitudes and favourable climatic conditions for carbonate deposition (Figure 2.3 of Sharland et al., 2001). The Khuff Formation represents broad carbonate deposition over a low-angle ramp associated with a vast epeiric sea across an area of 3.7 million km2. The low-angled dip of this carbonate platform made the Arabian Plate extremely susceptible to eustatic sea-level change resulting in cyclic carbonate
deposition. Whilst the presence of reservoir is not a major risk for this fairway, it is important to determine the distribution of grainy/oolitic facies alongside the degree of karstification related to exposure of the carbonate platform during periods of lower sea-level for enhanced reservoir quality characteristics. Within the public domain, there is very little data constraint on the presence of grainy/oolitic facies. This highlights the need to integrate datasets and methodologies such as hydrocarbon occurrences, tectonic lineaments, and sequence stratigraphy with a holistic Earth System Science approach (see Jennings, 2021) to understand tidal and bed shear stresses to predict the likelihood of grainy/oolitic facies. A major risk to the carbonates of the Khuff Formation and its equivalents is gas souring. Within the Khuff Formation, there are multiple significant anhydrite beds that, at subsurface temperatures of >130–140°C and in the presence of methane, have undergone thermochemical sulphate reduction (TSR), which has resulted in H2S concentrations of up to 98% (e.g. Worden and Smalley, 1996), destroying hydrocarbon accumulations. Therefore, understanding the distribution of anhydrite, migration pathways and the subsurface thermal regime are important parameters to determine in the exploration workflow. Similar risks (reservoir quality and gas souring) are associated with the Jilh Formation, equivalent to the Kurrachine and Geli Khani formations, with other petroleum system risks related to the patchy nature of the seal for the Jilh carbonate reservoir. Seal lithology and age varies across the plate from interbedded transgressive and highstand shales, to evaporites in the Palmyrides and Early Jurassic-aged evaporites through Iraq.
Whilst in each fairway ’zones’ of maximum potential have been identified, it is important to highlight that this first-pass, plate-wide screening approach has focused on charge from the Silurian Qusaiba Formation and its equivalents. Aside from the Silurian, there are multiple source rock horizons that have the potential to charge these deep gas plays. Whilst proven in Oman, could equivalents to the Huqf Supergroup source rocks be preserved elsewhere across Arabia? Available seismic
5 Showing areas across the Arabian Plate where ‘zones’ of more than four plays exist to help illuminate areas where multiple play potential may exist.
from Stewart et al. (2016), interprets a deep Precambrian rift structure within the Rub al Khali Basin, whilst wells have penetrated Precambrian sections in Yemen e.g. Qinab-1 (Petroleum Exploration and Production Authority, 2013), Jordan e.g Wadi Sirhan-3 (Andrews et al., 1991; Al-Husseini, 2010; and Luning et al. 2005) as well as in Israel, the Sinai Peninsula and southern Turkey. However, in Jordan, the Precambrian-Cambrian succession is dominated by coarse siliciclastics and is likely to be diluting any organic enrichment. Several authors e.g., Mahmoud et al. (1992) and Al Ghamdi et al. (2024) have investigated the potential of the Ordovician Hanadir Member and Ra’an Member shales. Reported Total Organic Carbon (TOC) values in these members are low (<1 wt.%) and have not been directly linked to charging hydrocarbon accumulations across Arabia. In isolated areas, could these members have elevated TOC and represent source rock potential? Zooming out of the Arabian Plate, the Frasnian Awaynat Wanin Formation is a known source rock in North Africa with up to 11% TOC; could similar values be present in the ‘pre’-Mesopotamian basin with speculative organic enrichment proposed in the Abu Safah-29 well? (Al-Hajri and Owens, 2000). Additionally, a recent publication from Sfidari et al. 2025, identifies a new upper Devonian (and lowermost Permian) source rock in the offshore Gulf comprised of gas-prone Type III kerogen. Finally, isolated ‘pods’ of organic enrichment (Aqrawi et al., 2010) in the Permo-Triassic carbonate platform related to differential bathymetry on a low-angle carbonate ramp have been speculated to generate organic-rich facies, whilst in the Zagros, gas and condensate accumulations in the Dalan and Kangan formations have been tied to source rock facies within the Kangan Formation (Ahanjan et al., 2017).
Migration from the source kitchen in the screening outputs are based on lateral migration and does not account for vertical migration or the utilisation of faults to juxtapose stratigraphically older reservoir in proximity with stratigraphically younger source rock to act as migration pathways. This juxtaposition via faults for older reservoir and younger source rock is particularly important for the Middle-Late Ordovician Qasim Formation. First-pass assessments of this geometry can be assessed using the Neftex® solution, from Halliburton Landmark, by integrating depth structure grids and tectonic elements within geospatial software. By removing or buffering the distance between the faults, the depth surfaces can be ‘re-stitched’ to illuminate zones where the younger source rock can be juxtaposed next to, or near, stratigraphically older reservoir and fluids can migrate.
Maturity of the Silurian is based on present-day depth and therefore the source kitchens are extensively overmature. A more thorough investigation in timing of charge, migration pathways, and the potential for multiple ‘mobile’ phases of charge related to structural development of the region is an important consideration. Finally, understanding differential topography across the region to locate areas where the Silurian is onlapping topographic highs and therefore, the potential to charge reservoirs up-dip, whilst combining this with a strong understanding of the preserved subcrop stratigraphy, is key to help further delineate play potential.
Understanding deep gas systems within the Middle East requires a regional approach to fully appreciate and contextualise the distribution of petroleum system elements and processes. We can use a play screening workflow to not only assess the likely presence and effectiveness of each petroleum system element but understand what the key risks for each area will be and how this will influence each petroleum systems development. This methodology has identified multiple ‘zones’ across the Arabian Plate where deep gas plays could be viable provided a sufficient trapping mechanism is in place. By stacking each of the individual fairways (Figure 5), we can high-grade areas across the Arabian Peninsula which may have multiple, stacked play potential areas in which proprietary datasets can then be used to test and validate identified ‘zones’. This approach has identified stacked play potential on and in proximity to basin highs such as the Qatar Arch and multiple play potential in rifted basin margins such as the Palmyrides (Figure 5). Areas across the Arabian Plate which have limited stacked play potential are a result of erosion from the ‘Hercynian’ Unconformity (removal and/or non-deposition of stratigraphy) and where reservoir effectiveness and charge pose a significant risk due to depth of burial (Figure 5).
This approach has identified that the Paleozoic stratigraphy of the Middle East has potential plate-wide and perhaps the ability to replicate the exploration and production success of the Mesozoic and Cenozoic intervals.
Andy Davies (Halliburton) is thanked for his help with future gas demand analysis. Halliburton is thanked for permission to publish.
Ahanjan, A., Rabbani, A.R. and Kamali, M.R. [2017]. An improved understanding of the origin and mechanism of Permian-Triassic natural gas-condensate accumulations in the Gavbendi High, Southwest Iran: An integrated approach. Journal of Natural Gas Science and Engineering, 37, 217-233.
Al Ghamdi, F., Babalola, L., Alqubalee, A., Abu-Mahfouz, I.S., Alotaibi, B. and Abouelresh, M. [2024]. Source rock potential, thermal maturity, and depositional environment of the Ordovician Ra’an Shale Member, central Saudi Arabia. Journal of Asian Earth Sciences, 11, 1-18.
Al-Hajri, S. and Owens, B. [2000]. Sub-surface palynostratigraphy of the Palaeozoic of Saudi Arabia. In S. Al-Hajri and B. Owens (Eds.), Stratigraphic Palynology of the Palaeozoic of Saudi Arabia. GeoArabia Special Publication, 1, 10-17.
Al-Husseini, M.I. [2000]. Origin of the Arabian Plate structures: Amar Collision and Najd Rift. GeoArabia, 5(4), 527-542.
Al-Husseini, M.I. [2010]. Middle East Geologic Time Scale 2010: Early Cambrian Asfar Sequence. GeoArabia, 15(1), 137-160.
Andrews, I.J. [1991]. Palaeozoic lithostratigraphy in the subsurface of Jordan. Natural Resources Authority. The Hashemite Kingdom of Jordan, 1-81 p.
Aqrawi, A.A.M., Goff, J.C., Horbury, A.D. and Sadooni, F.N. [2010]. The Petroleum Geology of Iraq. Scientific Press Ltd (Wiley-Blackwell), 1-424 p.
Davies, A. and Simmons, M.D. [2021]. Demand for ‘advantaged’ hydrocarbons during the 21st century energy transition. Energy Reports, 7, 4483-4497.
Hui, J., Shengli, F., Yingquan, Z., Shaonan, Z., Jin, Z. and Zhiqiang, L. [2016]. Initial porosity reduction phase of ultra-deep tight sandstone reservoirs: A case study of the Rub’ Al Khali Basin in the Middle East. Petroleum Geology and Experiment (Shiyou Shiyan Dizhi), 38(2), 219-223.
Huuse, M., Redfern, J., Le Heron, D.P., Dixon, R.J., Moscariello, A. and Craig, J. [2012]. Glaciogenic Reservoirs and Hydrocarbon Systems. Geological Society of London - Special Publications, 368, 1-401 p. Jennings, J. [2021]. Predicting Reservoir Variation and Sour Gas Presence to Better Constrain the Remaining Potential within the Khuff Play. Subsurface Insights Magazine. Subsurface Insights, October, p. 11-18.
Kurjanski, B., Rea, B.R., Spagnolo, M., Cornwell, D.G., Howell, J. and Archer, S. [2020]. A conceptual model for glaciogenic reservoirs: From land systems to reservoir architecture. Marine and Petroleum Geology, 115
Luning, S., Shahin, Y.M., Loydell, D., Al-Rabi, H.T., Masri, A., Tarawneh, B. and Kolonic, S. [2005]. Anatomy of a World-Class Source Rock: Distribution And Depositional Model Of Silurian Organic-Rich Shales In Jordan And Implications For Hydrocarbon Potential. AAPG Bulletin, 89(10), 1397-1427.
MacDonald, R., Hardman, D., Sprague, R., Meridji, Y., Mudjioni, W., Galford, J., Rourke, M., Dix, M. and Kelton, M. [2010]. Using elemental geochemistry to improve sandstone reservoir characterization: A case study from the Unayzah A interval of Saudi Arabia. Proceedings of SPWLA 51st Annual Logging Symposium
Mahmoud, M.D., Vaslet, D. and Husseini, M.L. [1992]. The Lower Silurian Qalibah Formation of Saudi Arabia: An Important Hydrocarbon Source Rock. AAPG Bulletin, 76(10), 1491-1506.
Melvin, J. and Sprague, R.A. [2006]. Advances in Arabian stratigraphy: Origin and stratigraphic architecture of glaciogenic sediments in Permian-Carboniferous lower Unayzah sandstones, eastern central Saudi Arabia. GeoArabia, 11(4), 105-152.
Omer, M.F. and Friis, H. [2014]. Cathodoluminescence investigations on quartz cement in the sandstones of Khabour Formation from Iraqi Kurdistan Region, Northern Iraq. Journal of African Earth Sciences, 91, 44-54.
Petroleum Exploration and Production Authority 2013. Block 12 (North Sanau), Petroleum Exploration and Production Authority (PEPA), 5 p.
Sfidari, E., Sharifi, M., Maroufi, K., Zamanzadeh, S.M., Amraie, J., Mehrabi, H. and Spina, A. [2025]. Source rock characterizations of the Devonian to Permian strata in the Persian Gulf Basin, SW Iran Journal of African Earth Sciences, 231, 1-17.
Sharland, P.R., Archer, R., Casey, D.M., Davies, R.B., Hall, S., Heward, A., Horbury, A. and Simmons, M.D. [2001]. Arabian Plate Sequence Stratigraphy. GeoArabia Special Publication, 2, Gulf PetroLink, 1-387 p.
Stewart, S.A. 2016. Structural geology of the Rub’ Al-Khali Basin, Saudi Arabia. Tectonics, 35, 2417-2438.
Sutcliffe, O. [2016]. Regional Scale Stratigraphic Heterogeneities in the Lower Paleozoic of the Middle East. Exploration Insights Magazine, February, 16-21.
Sutcliffe O.E and T.D Cousins [2018]. Analogue insights for super basins: Silurian unconventional plays in the Middle East. Exploration Insights Magazine, July, 4-14.
Wood Mackenzie. [2022]. Half of traditional super basins are future-fit and they have three things in common. In https:// www.woodmac.com/press-releases/half-of-traditional-super-basins-are-future-fit-and-they-have-three-things-incommon/.
Worden, R.H. and Smalley, P.C. [1996]. H2S-producing reactions in deep carbonate gas reservoirs: Khuff Formation, Abu Dhabi. Chemical Geology, 133(1-4), 157-171.
Yanfang, L., Jianmin, L., Xiaofei, F., Guang, F., Yougong, W. and Changli, X. [2009]. Geologic conditions of oil-gas downward discharging and exploration direction in the Sanzhao Sag, Songliao Basin. Chinese Journal of Geology (Scientia Geologica Sinica), 44(2), 525-533.
Vita Kalashnikova1*, Rune Øverås1, Tatiana Nekrasova1, Carl Fredrik Gyllnehammar 2 and
Ivar Meisingset3 demonstrate the seismic-driven identification of possible sandstone distribution within the Rogaland Group of the Norwegian North Sea.
Abstract
The authors analysed the regional distribution of Rogaland Group sands, which were predicted using AI-based Rune Inversion and the Gyllenhammar equation for clay volume estimation applied to the ultra-large Elephant 2.0 database (constructed in 2022). The results were compared with the stratigraphic framework established by Brunstad et al. (2013). Drawing upon publicly available information and insights from personal comments made by Harald Brunstad, the study demonstrates general compliance with the work of Brunstad et al. and reveals potential new areas and thicknesses for sandstone accumulations. Considering the challenges in estimating volume of sand from post-stack inversion, the results show a high degree of correlation, revealing more details between the wells.
Introduction
In this paper, we demonstrate the seismic-driven identification of possible sandstone distribution within the Rogaland Group of the Norwegian North Sea, focusing specifically on the Lista and Våle Formations (Fms). This is done using data from the Elephant 2.0 project (2022-2024) — an ultra-large, AI-based inverted seismic dataset for rock properties. The results are then compared with the sandstone members outlined by Brunstad and colleagues in their comprehensive stratigraphic work, which we will refer to as the reference work and reference maps/outlines
The inverted Elephant 2.0 volumes were originally designed to be recomputed into volumes of clay and sand, relevant to the approximately Jurassic-Cretaceous interval. However, with the introduction of the Gyllenhammar formula for each formation individually, the Paleocene interval also became accessible for analysis (Nekrasova et al., 2025).
The study also tackles the essential problem of carbonates being misclassified as sandstone in predictive
models. We demonstrate the ability to filter out these false positives when generating maps of the best quality sandstone attribute, and we identify areas where carbonates can still be misinterpreted, thus requiring additional filtering techniques.
Stratigraphic description of the Lista and Våle Fms (Based on Brunstad et al., 2013)
The stratigraphic framework established by Brunstad et al. for the Rogaland Group (Paleocene epoch) in 2013 provides a solid reference for modern subsurface analysis. It integrates lithological, biostratigraphic, geophysical, and depositional insights into a coherent model that supports both academic research and applied exploration. This work forms a regional key basis for comparing
1 Pre Stack Solutions-GEO AS | 2 CaMa Geo AS | 3 Model-Geo AS
* Corresponding author, E-mail: vita@pss-geo.com
DOI: 10.3997/1365-2397.fb2025068
the predicted sandstone distribution using seismic inversion methods. It is worth noting that Harald Brunstad has made significant progress since that time in clarifying the members’ outlines using new wells and modern 3D seismic data. Based on his comments, we will highlight several areas outside the reference outlines of 2013.
The Lista and Våle Fms of the Rogaland Group (Figure 1) represent an important part of the stratigraphic record on the Norwegian Continental Shelf. These formations, deposited during the Paleocene epoch, are particularly significant in both regional basin analysis and petroleum exploration. Brunstad and colleagues contributed substantially to refining the lithostratigraphy of the Rogaland Group, especially by clarifying the characteristics and regional mapping of the Lista Fm, its subunits and stratigraphic foundation, the Våle Fm. We follow the formation and member classification of Brunstad et al. in this paper, who classify time-equivalent sub-units of the Våle and Lista formations as members, instead of following the Norwegian Offshore Directorate (SODIR), which sometimes classifies them as formations.
Figure 2 Key reference maps used to compare the inversion-based predicted sand and clay distribution for the Lista and Våle Formations. The figures are sourced from the electronic version of the Brunstad et al. work. (a) – Extract from Figure 12: a sketch showing the approximate distribution of sandstone members within the four shale formations of the Rogaland Group. (b) – Extract from the same electronic resource, the page with the update in the publication: ‘Stratigraphic Guide to the Rogaland Group, Norwegian North Sea’ by Harald Brunstad, Felix M. Gradstein, Jan Erik Lie, Øyvind Hammer, Dirk Munsterman, Gabi Ogg, and Michelle Hollerbach. Published in Newsletter on Stratigraphy, Vol. 46/2, pp. 137-286, 2013 (Figures 48 and 72).
The Våle Fm (Early Rogaland Group) underlies the Lista Fm and serves as a reliable seismic marker due to its higher density and acoustic velocity, identifiable in both seismic and well-log datasets. This is due to its compact and relatively homogeneous lithology, composed largely of dense claystones and argillaceous sediments, suggesting that deposition occurred in a relatively low-energy marine environment. Its thickness generally ranges between 20 and 60 m. This contrast helps to constrain the base of sand-rich packages in the Lista Fm and also serves as a key horizon for structural and stratigraphic modelling. Våle Fm contains four members: the Borr, Ty, Maureen, and Egga (Figure 2).
The Lista Fm, a marine shale unit, is characterised by its homogeneity, intense bioturbation, and consistent log and seismic responses, and has three subdivisions: L1, L2, and L3. Lithologically, it comprises predominantly soft, non-calcareous mudstones and shales. Intense bioturbation, especially in the middle and upper parts, often leads to near-total homogenisation, with common trace fossils.
Four members, namely Heimdal, Mey (likely originating from the Shetland Platform), Siri, and Sotra (Scandinavian-sourced), are identified. These sand-rich intervals are crucial for hydrocarbon exploration, appearing seismically as thickened lenses or sheets with internal reflection continuity. The Lista Fm’s thickness varies regionally up to 500 m. Its broad distribution makes it an essential unit for regional mapping and correlation.
We constructed maps for the Våle and Lista Fms of the 1−Vclay (where Vclay is the volume of clay) attribute, best quality sand distributions, and sandstone thicknesses. The 1−Vclay attribute maps were overlaid with the referenced outlines (from Figure 2). To validate the results of our analysis, we utilised the SODIR website (www.sodir.no) to access well reports, lithology descriptions, and top depths. We also incorporated publicly available logs, such as Gamma Ray, and the volume of clay curve recomputed using conventional petrophysical log interpretation. Then, we performed the analysis of the results.
The research by Brunstad and colleagues employs an integrated approach, combining the correlation of seismic and well-log markers across multiple wells with comprehensive regional seismic interpretation. This methodology facilitates the identification, clarification, and introduction of distinct internal members within the formation. The study incorporates detailed stratigraphic, sedimentological, and biostratigraphic analyses to refine the subdivisions. The outcomes of this extensive work are conveniently summarised in maps, designed for ease of use in subsequent comparative studies and quality control workflows.
The Brunstad et al. maps with outlines (Figure 2) have substantial implications for hydrocarbon exploration, as the identified sand members provide potential locations for sand reservoirs, offer the possibility of studying their distribution patterns, and suggest possible traps.
Due to the extremely large nature of the Elephant 2.0 database (Figure 1), the only detailed surfaces interpreted are the top of the
Rogaland Group and the top of the underlying Shetland Group. To exclude Sele and Balder Fms (late Rogaland), we calculated a midpoint in time between the Shetland and Rogaland surfaces (defined as [Rogaland + (Shetland – Rogaland)/2]) to represent the ‘Inner horizon’. In areas where the Balder and Sele Fms are thin or absent this horizon will cut into the Lista Fm, and we risk losing some potential sand layers which are present in the uppermost L3 Lista unit. In addition, given that the Våle Fm and its sand units are relatively thin, and the Lista Fm is characterised by significantly thicker clay/mud intervals, we anticipate that calculating average rock property values for mapping may not adequately capture the full internal distribution of sands within these formations. Therefore, we compute a Best Quality Sand attribute and construct the map of the thickness of only the best quality sand layers. Also, the Shetland surface was picked as a peak (positive amplitude, indicating an increase in impedance) throughout the ultra-large cube. Consequently, especially in the southern part of the cube, some soft, porous chalk can be merged into the analysis, masking a clear distinction for the sandstones. These rocks may possess Vp and density properties similar to the best porous sand we are looking for, complicating the analysis of a two-component clay-sand cube.
Since our primary aim is to test the informativeness of the entire inverted volume in a scanning approach, this preliminary analysis provides sufficient insight to assess the database’s potential.
The seismic surveys were conditioned and then merged in the Elephant 2.0 ultra-large cube. It consists of more than 200 seismic surveys of true amplitude data. 722 wells were tied for inversion and, where they are available in reasonable quality, used to construct the regional low-frequency models for P-wave velocity (Vp) and Density (Den). Ten prior interpreted surfaces were used in the modelling process. As no seismic gathers were available, the Rune Inversion method was applied to estimate Vp and Density from the post-stack seismic data (Øverås and Kalashnikova, 2021).
The inversion process applies an optimisation technique to identify the global minimum by iteratively comparing synthetic gathers with those generated from the stacked seismic trace. The
Figure 3 Volume of clay calculation from DT and RHOB logs cross plot for 39 key wells from the North Sea Elephant database, Rogaland Group. The key wells are indicated by red circles on the map with the Elephant 2.0 data coverage.
Figure 4 A comparison of conventional clay volume estimates (VCL, black) and Gyllenhammar formula-derived estimates (VCL_DS, red) for selected wells within the Rogaland Group, serving as a quality control example. The green area between these curves highlights the difference in the prediction of the approaches. RT — True Resistivity, RHOB — Bulk Density, DTp — Compressional Delta-T, NPHI — Neutron Porosity Index.
synthetic gathers are produced iteratively using initial models of Vp and Den and are updated randomly at each iteration. The resulting inverted Vp and Den volumes are then compared with the corresponding well logs. The inverted velocity and density generally show a very good match with well logs; however, relying solely on these properties to compute others is quite challenging. We used the Gyllenhammar formula to calculate the volume of clay (Vclay), taking 1−Vclay as the volume of sand (Vsand) (Gyllenhammar, 2020). This result does not differ from conventional petrophysical Vclay estimation. By knowing Vclay and Den, we computed porosity, and the product of (1−Vclay)×porosity served as our Best Quality Sand attribute.
The Gyllenhammar formula offers a seismic-compatible method for estimating Vclay using Vp and bulk density (RHOB), thereby overcoming the limitations of traditional clay indicators, such as Gamma Ray (GR) and Neutron–Density (ND), which cannot be directly derived from seismic data. Developed by Dr Carl Fredrik Gyllenhammar (2020 and 2022), the method involves cross-plotting acoustic (DT) and density data, with calibration performed using well logs. In this study, the formula was applied to the Rogaland Group using seismic inverted data from the Elephant 2.0. The methodology involves identifying three reference points on the Sonic–Den (DT–RHOB) cross-plot (Figure 3) and solving equation 1.
(1),
Vclay_DS = a/Velocity + b·RHOB + c. (2)
The next step is to adjust the equation’s parameters and present it in m/s and g/cm³ units. About 30 wells distributed across the Norwegian North Sea were used as a reference, Figure 3. For the Rogaland Group, the calibrated coefficients are: a=4124.03936, b=1.033210332, c=−3.4600246m (Nekrasova et al. 2025). The cross-plots and track-by-track comparisons confirmed that Vclay by Gyllenhammar formula (VCL_DS) matched traditional clay volume estimates from Gamma Ray (VCL on Figure 4) in most wells for the Rogaland Group when using these zone-specific coefficients. Then, Formula 2 will be expressed as follows:
Vclay=4124.03936/Vp+1.033210332*Den-3.4600246. (3)
Assuming a two-component system of clay and sand based on the cross-plot in Figure 3, the sand volume can be expressed as:
Vsand=1-Vclay. (4)
Mitigating errors in carbonate and complex mineralised reservoirs
It is important to note that this formula allows us to estimate the volume of clay. As a counterpoint to the approximate clay line (represented by blue and light blue dots on the cross-plot in Figure 3), we assume that 1−Vclay equates to Vsand (Formula 3). This two-component estimation will not allow us to distinguish sandstones from carbonates (which often plot around point B on the cross-plot in Figure 3) and other rock types. Therefore, to help identify sandstone reservoirs, we compute the Best Quality Sand Attribute. First, we estimate porosity using the following formula: (5)
Where Den is the inverted seismic density, j is porosity, ρclay is the density of clay minerals, ρsand is the density of quartz, ρfl is the density of fluid equal to 1.0 g/cm3, assuming water saturation 100%, ρsand is 2.65 g/cm3, ρclay is 2.58g/cm3, Vclay is computed by Gyllenhammar formula 1. It is assumed the media is of normal compaction with no erosion.
Afterwards, we compute the Best Reservoir Quality Sand attribute (Res_att) as the multiplication of porosity and volume of clay:
Res_att = (1-Vclay)*j. (6)
where:
• Point A – clean, compact sandstone (RHOB ~ 2.55 g/cm3; wDT ~ 61 μs/ft)
• Point B – porous clean rock (RHOB ~ 2.00 g/cm3; DT ~ 103 μs/ft)
• Point C – 100% clay (RHOB ~ 2.30 g/cm3; DT ~ 154 μs/ft)
By defining these points for the formation, the Gyllenhammar equation (1) can be simplified to the following:
Depending on the porosity in the area, the attribute is typically scaled from about 0 to 0.3 v/v, representing the fraction of sandstone volume. Higher values indicate better quality sand, while values close to 0 are considered ‘no sand’. It must also be noted that by analysing the Res_att, we filter out a significant amount of carbonates. However, in cases where carbonates have similar Vp and density properties to sandstone, carbonates with the highest porosity might be among the predicted sandstones. It is also interesting to note that while gas sandstone and coal can exhibit similar seismic responses, most known coals are filtered out by the Res_att and presented as ‘no sand’.
We divided the Norwegian North Sea full-size (FNS) into three parts: the Northern North Sea (NNS), the Central North Sea (CNS), and the Southern North Sea (SNS) (Figure 1). For each part, we constructed a map of summed average amplitudes between the Inner and Shetland horizons. Firstly, maps were generated from the 1−Vclay volume, which, as previously discussed, is assumed to represent the potential for identifying sands (Vsand). Then, we created maps using the Res_att attribute. This approach allows us to eliminate sandstones with undesirable properties, primarily low porosity, and also possibly other types of rocks (Figure 15). For depth conversion, including for thickness estimation, we use the inverted P-wave velocity field converted to average. As the aim of this work was to assess the predictability of regional inversion in the Elephant 2.0 database, this preliminary ‘rough’ analysis serves as a sufficient scanning tool.
Figure 5 illustrates the distribution of the Våle and Lista sandstone members from the Elephant 2.0 database (constructed using the 1−Vclay volume), with overlaid formation member outlines from reference maps (Figure 2). Colours towards blue
indicate a predominance of the clay component; other colours can be referred to as ‘no clay’. We assume that within these colours (white-yellow-red), we can analyse the presence of sandstone. We can observe that the outlined Sotra Mbr (Lista Fm) generally shows a good match with the predicted sand distribution. In contrast, the northern part of the Heimdal Mbr (Lista Fm), and especially the Ty Mbr (Våle Fm), at the Tampen Spur, appears shifted in comparison with the referenced outlines. The contours were transferred to the constructed map with inverted sandstones using degree indicators from the reference maps (Figure 2). We anticipate some inconsistencies in the map projections, and we also account for the approximate nature of the line sketching the members’ outlines and the limited well data penetrating particular formations. However, this mismatch on the Tampen Spur is observed as the most noticeable.
Figure 6 shows a cross-section along the arbitrary line WE through Ty and Heimdal Members where sandstones would be expected. Wells are displayed with their 1−Vclay logs (Vclay was computed conventionally). It is difficult to distinguish thin layers from the seismic, but the well log 1-Vclay and the inverted section of 1-Vclay match the observation. Our interest is to distinguish sandstones.
Figure 5 NNS: Average summed amplitudes of the 1− Vclay attribute within the Inner and Shetland horizons, derived from inverted Vp and Density. Overlaid are the reference outlines for the Våle (orange) and Lista (green) Formation Members (Ty, Egga, and Heimdal), based on the work of Brunstad et al., 2013. The yellow quadrant outlines the map shown in Figure 7.
Figure 6 NNS. A 1−Vclay and Res_att attribute cross-sections, overlaid with seismic data, in depth along the black arbitrary line WE (location shown in Figure 5), passing through several wells. Inserted curves are 1-Vclay (where Vclay is conventionally computed), and the tops are sourced from www.sodir.no.
7
Despite the section being taken within reference outlines, no Våle Fm tops are defined for these wells, except 30/9-12. Averaging values in the Inner to Shetland interval should yield a blue-white colour, as seen in Figure 5. We also observe on the inverted section the ‘sandstone’ in the Balder (latest Lista Fm).
In general, the well log data support the inverted result for the Rogaland Group. However, lithology descriptions may differ, which was not fully captured in the 1-Vclay section. In well 34/7-12 (publication release 2008), Figure 6, the Våle Fm consists of marlstones, grading into the Lista Fm mudstones
with occasional carbonate and tuff layers. The earliest Balder Fm contains a high-quality sandstone. In 34/7-26 A (2008), the Lista Fm shows similar lithology, while the Balder Fm is dominated by tuffaceous shales with local sandstone beds. Well 34/7-21 (2008) mainly encountered clay/claystone with minor sand in the upper Rogaland Group, with a Lista Fm sandstone at the base. In well 34/7-31 (2006), a thin Balder Fm of claystones and tuffs is present, overlying the Sele Fm with laminated organic-rich shales and glauconitic silt/sand interbeds. Petrophysical interpretation in the Balder Fm is difficult due to the presence of volcanic ash (tuff) that appears like sand on logs with low GR. The Lista Fm contains claystone with limestone stringers and rare sand, with a distinct sand-prone layer seen in Res_Att at the top, and two more down to the Shetland Group. Considering the limitations imposed by seismic resolution, we observe a generally satisfactory prediction for a scanning tool.
To the east of well 34/7-31, along the Shetland slope, sand layers clearly appear on the Res_Att. These fall within the area designated as the Ty Mbr of the Våle Fm by reference outlines, suggesting a likely extension of this Våle sandstone Mbr. However, in well 30/9-12 (2006), the Våle Fm (~8 m thick) comprises calcareous claystone, the Lista Fm shows claystone with limestone, and the upper Rogaland section did not penetrate sandstone, according to the well report. Although the logs and inversion data show good agreement, a clear sandstone identification remains inconclusive.
We reviewed all wells in quadrants starting with 33 and 34 on the SODIR website for the presence of the Våle Fm. On the map (Figure 7), wells with sandstone are marked with yellow circles, those with carbonates/mudstone/clay with purple, and wells with no lithology descriptions available are marked in black.
Some wells inside the outlines, such as 33/9-12, 33/12-7, 33/9-3, 33/9-4, and 33/9-5 (numbered towards the North) near the arbitrary line (letter W), as well as wells 34/8-1 and 34/8-9 S at the edge of the outlines, do not have the Våle Fm defined, and the Lista Fm sandstone is not often observed. These observations confirm the need for a northward shift in the outlines of the reference maps, which leads to a better match with the inverted clay-sandstone prediction map. The inversion also indicates sandstone in the Magne Sub-basin. This is consistent with Harald Brunstad’s confirmation that the modern, revised stratigraphic sequences for these formations also have more northward coverage and extend toward the Magne Sub-basin, making the inversion prediction reasonable.
Another inconsistency with reference map outlines can be observed within the mud and silt-dominated area of the Lista Fm in the Fensal Sub-Basin. The underlying Våle Fm is expected to have sandy shelf deposits only towards the west. Figure 8 shows an arbitrary line (blue line WE, Figure 5) going through this area. All wells along this line have Lista and Våle Fms defined (Figure 8). Well 30/10-2, as reported, penetrated a thick Paleocene sand; no per-formation info is available. Well 30/10-6 has predominantly shale with stringers of sandstones, and the Res_att shows a better match with the well logs. The Lista Fm in well 30/7-3 has a good sandstone layer, while its Våle Fm has the thickest sandstone stringers at an early stratigraphic level. Below the Våle surface, there are chalky limestones. Well 30/8-4
8 NNS. A 1−Vclay and Res_att attribute cross-sections, overlaid with seismic data, in depth along the blue arbitrary line WE (location shown in Figure 5), passing through several wells. Inserted curves are 1-Vclay (where Vclay is conventionally computed), and the tops are sourced from www.sodir.no.
S has no description on the SODIR, but it is expected to be more shaly, consistent with both the reference maps and predictions from the inverted results. Thus, the reference outline for the Lista Fm sandy shelf can be extended toward the east, yielding a better match with computed results.
The single well 31/8-1 in the Stord Basin, although placed in a possible sandstone area computed from the 1−Vclay cube in Figure 5, exhibits no sands and no thickness for best quality sands on the last Figure 15, which is correctly filtering other types of rocks.
Figure 15 summarises the results of the predicted sandstones from the inversion. The left panel shows the Best Quality Sand Attribute, indicating high-quality sand occurrences (yellow to red). The right panel displays its thicknesses; summed amplitude values range from 0.17 (Res_att, light yellow to dark-red), with grey areas representing zero thickness. The combined view of sand distribution and thickness shows a better match with the wells’ descriptions.
The CNS part is generally consistent with reference outlines. The Stord Basin, which falls within sand-clay areas of referenced outlines, is well filtered on Figure 15, where Best Quality Sands (Res_att) are shown, indicating possible sand accumulation only in some areas. Line WE (Figure 10) illustrates how carbonates can be filtered using Res_att. The inserted well, 24/6-2, had a carbonate layer within the Heimdal sands, but this was effectively filtered out on Res_att. In addition, possible areas with sandstone on Heimdal
Figure 9 CNS. Average summed amplitudes of the 1−Vclay attribute within the Inner and Shetland horizons, derived from inverted Vp and Density. Overlaid are the reference outlines for the Våle (orange) and Lista (green) Formation Members (Ty and Heimdal), based on the work of Brunstad et al., 2013.
and Gudrun Terraces that are outside of the reference lines were also filtered out on Res_att (Figure 15). Harald Brunstad modern works also suggests possible extensions with good sands in the Stord Basin, which is positively supported by our inversion.
Predictions for the Southern North Sea show more deviation in the outlines. We will look into the details for the Åsta Graben, along the west border, and the Egersund Basin.
In the Åsta Graben, well 8/12-1 is shaly in the target interval. Nearby wells that are not covered by inversion, e.g., to its north-east, well 8/9-1 has silty sandstone very rich in glauconite. Towards the south-west, wells are predominantly shaly, with well 8/10-2 having dolomite and shaly marl in the early Rogaland Group. To the north-west, well 7/3-1, for example, shows shale/ clay predominance. The result is as expected if one looks at the Res_att (Figure 12.), which better reflects the correct lithology. Similar to the previously presented CNS case (Figure 10), we see that carbonates were properly filtered out, leaving a satisfactory
Figure 10 CNS. A 1−Vclay and Res_att attribute cross-sections, overlaid with seismic data, in depth along the black arbitrary line WE (location shown in Figure 9), passing through the 24/6-2 well. Inserted curves are 1-Vclay (where Vclay is conventionally computed), and the tops are sourced from www. sodir.no.
prediction. Harald Brunstad commented that an extended area of carbonates is present in this area. The Vidar Fm olistolith and chalk were rafted from the east into the Central Graben. This is notably tracked in our inverted results.
A similar situation is present in the western Southern North Sea (SNS) (blue arbitrary NS line, from Cod Terrace
to Ål Basin). The Lista Fm is underlain with carbonates. 7/11-1 has sandstone layers and claystone. 1/3-6 has ‘Cod’ sands underlain by interbedded shale and sands, and the Våle Fm transitioning to carbonates. 1/3-1 has claystone, tuffs and tuffaceous sandstone, underlain with thick Danian/Late Cretaceous chalk. 1/6-1 has claystone, underlain with muddy
Figure 11 SNS. Average summed amplitudes of the 1− Vclay attribute within the Inner and Shetland horizons, derived from inverted Vp and Density. Overlaid are the reference outlines for the Våle (orange) and Lista (green) Formation Members (Maureen, Borr, and Siri), based on the work of Brunstad et al., 2013.
Figure 12 SNS. A 1−Vclay and Res_att attribute cross-sections, overlaid with seismic data, in depth along the black arbitrary line WN (location shown in Figure 11), passing several wells. Inserted curves are 1-Vclay (where Vclay is conventionally computed), and the tops are sourced from www.sodir.no.
limestone and then limestone. 2/7-29 has mudstone in both the Våle and Lista Fms. And, last in the line, 2/11-8 has Lista Fm with clay, sandstone and chert, Våle Fm has claystone and marl.
Figure 13 SNS. The Res_att attribute cross-section, overlaid by seismic data, in depth along the blue arbitrary line NS (location shown in Figure 11), passing through several wells. Inserted curves are 1-Vclay (where Vclay is conventionally computed), and the tops are sourced from www.sodir.no.
Figure 14 SNS. The Res_att attribute cross-section, overlaid with seismic data, in depth along the red arbitrary line WE (location shown in Figure 11), passing several wells. Inserted curves are 1-Vclay (where Vclay is conventionally computed), and the tops are sourced from www.sodir.no.
Figure 15 Map summarising the sandstone and clay distribution predicted by inversion for Våle and Lista Fms (a), along with moderate-to-best quality sand thicknesses (b).
In the Egersund Basin, where the Lista Fm reference outline is silt- and mud-dominated, just above the Shetland surface, chalk and limestones have fallen into the mapping zones. The maps on Figure 11 show small zones with a yellowish colour
and little thickness; only the 10/7-1 well has Lista Fm with claystone and thin sandstone and limestone layers in the early Rogaland. Wells in block 9/2 have mainly clays/shales with 9/2-5 and 9/2-4 S underlain by marl/chalk and limestone. The formation consists of chalky limestone. Thin beds of chert and claystone laminae are common in this section. The Tor and earlier Hod Fms mainly contain limestone that failed to be separated by inversion and Vclay pre-computation.
We are using the Gyllenhammar formula for each formation: one for the entire Rogaland Group and one for the underlying Shetland Group. It is quite possible that carbonates in the Paleocene-Cretaceous transition should also be split into a separate group, and the Gyllenhammar formula should be defined for each. For example, in Nekrasova et al.’s (2025) work, it was noted that Upper Jurassic rocks must be split into two formulas — depending on depth before and after 1700 m — to properly define gas sandstone.
As described earlier, a mis-mapping issue above the Shetland Group may arise because the regional Shetland surface is picked as a peak (positive amplitude). The carbonate rocks (often chalk) just above out picked Shetland surface, with similar properties to sandstone, can get incorporated into the mapping. This makes it difficult to separate them. The soft, porous chalk may exhibit Vp and density properties similar to those of the best porous sandstones, which complicates the analysis of a two-component clay-sand cube. Therefore, accurate surfaces picking at the Paleocene-Cretaceous transition are crucial to eliminate these carbonates from the analysis.
Lastly, predicted sandstone can be observed in wells 3/4-1 and 3/5-2 in the Søgne Basin. Well 3/4-1 contains claystone that becomes calcareous and grades into marl. Well 3/5-2 has sandstone that is gradually calcareous and glauconitic and is generally shaly. However, in his modern analysis, Harald Brunstad also outlines thin lines of potential sandstone there.
Conclusion
We have discussed the regional Våle and Lista Fms sandstone distribution predicted by AI-based seismic inversion in comparison with the reference stratigraphical work of Brunstad and colleagues (2013) in the Norwegian North Sea. The post-stack Rune Inversion and Gyllenhammar formula for volume of clay computation, adjusted for the Rogaland Group, enabled quite accurate mapping of clay. Sand distribution within the Lista and Våle Fms can be better identified using the Best Quality Sand Attribute; however, some misinterpretation of mainly carbonates can occur. This misinterpretation can happen in carbonate rocks that have similar Vp and density properties to sandstones. It is evident that chalk is often misinterpreted, and it is suggested that formations predominantly composed of these types of carbonates should be treated separately by
the Gyllenhammar formula and probably excluded from the analysis through accurate seismic interpretation. The results for sand prediction are generally consistent with the sketched sand outlines defined in the work of Brunstad et al. (2013) and positively match with the modern reference outlines of Harald Brunstad at the quick analysis. The fact that we obtain reasonable results from the proposed procedure is a significant achievement, demonstrating the fundamental soundness of our inversion technique and representing a substantial step forward compared with traditional methods. The proposed approach is effective across large inter-well distances, significantly improving stratigraphic understanding, while internal seismic markers were limited.
The authors thank Harald Brunstad for reviewing our results against his modern stratigraphical outlines for the Rogaland sandstones. We saw great progress in his work and only hope that it will eventually be published as an update to his and his colleagues’ already comprehensive work of 2013.
The Elephant database is the proprietary product of Pre Stack Solutions-GEO and Lime Petroleum, which is freely available for academia and educational institutions and can be licensed by commercial organisations.
Brunstad, H., Gradstein, F., Vergara, L., Lie, J.E. and Hammer, Ø. [2008]. The Rogaland Group – Stratigraphic Units of North-West Europe. Electronic resource. Time Scale Foundation.
Available at: https://timescalefoundation.org/resources/NW_Europe_Lex/litho/rogaland.php https://timescalefoundation.org/resources/NW_Europe_Lex/litho/vale.php https://timescalefoundation.org/resources/NW_Europe_Lex/litho/lista.php
Brunstad, H., Gradstein, F., Lie, J.E., Hammer, Ø., Munsterman, D., Ogg, G. and Hollerbach, M. [2013]. Stratigraphic Guide to the Rogaland Group, Norwegian North Sea. Newsletters on Stratigraphy, 46(2), 137–286. DOI: 10.1127/0078-0421/2013/0032
Gyllenhammar, C.F. [2020]. Calculating a synthetic density curve using volume of clay and velocity. First Break, 38(6), 45–49.
Gyllenhammar, C.F. [2022]. Calculating Volume of Clay and/or Shale From Wireline Logs. First Break, 40(9), 59–65.
Nekrasova, T., Kalashnikova, V., Finogenova, A., Gyllenhammar, C.F. and Sopivnik, V. [2025]. Clay Volume Calculation from Acoustic and Density Curves using the Gyllenhammar Equation. 86th EAGE Annual Conference and Exhibition, Extended Abstracts
DOI: 10.3997/2214-4609.202510624
Øverås, R. and Kalashnikova, V. [2021]. Method of seismic inversion using artificial kinematic constraints. Norwegian Patent NIPO 20210133.
Julien Razza¹*, Remi Leblond¹, Nasser Olleik¹, Étienne Legeay¹, Marie Etchebes¹ and Laurent Souche¹ introduce an interactive framework that reduces 3D seismic stratigraphic interpretation time from weeks to days through AI-guided depositional element extraction.
Abstract
While crucial to understand the subsurface, interpreting 3D depositional elements from seismic data is often challenged by issues such as effectively visualising subtle stratigraphic features and accurately transforming them into 3D objects that reflect geological reality. This paper introduces an interactive framework that directly addresses both through an innovative three-step workflow. The process begins with chronostratigraphic colour blending on stratal slices, enabling rapid visualisation of depositional patterns and sequence relationships. Next, to enhance feature separation, interpreters interactively place seed points within target elements and background facies, allowing for the automatic determination of optimal spectral decomposition frequencies. Finally, the interpreter guides an image segmentation foundation model with simple point prompts, facilitating the precise extraction of depositional elements without the bottleneck of geological-specific training data. Application of the method to fields, namely channel systems from the Maui Field, New Zealand, and carbonate build-ups and reefs from the Poseidon Field, Australia demonstrates a reduction in interpretation time from weeks to just days, all while maintaining high accuracy. This chronostratigraphic-centric approach guarantees that extracted objects represent true depositional architecture rather than arbitrary seismic anomalies.
Introduction
Stratigraphic interpretation of 3D seismic data forms the bedrock of understanding subsurface depositional architecture. The precise mapping of features such as channel systems, turbidite fans, carbonate build-ups, and deltaic complexes is not just academic, it drives crucial decisions in hydrocarbon exploration and reservoir modelling, providing the essential geometric framework for understanding reservoir compartmentalisation and predicting fluid flow.
Traditionally, interpretation adheres to sequence stratigraphic principles, placing depositional elements within their chronostratigraphic context. Interpreters extract depositional surfaces and volumes of sedimentary bodies deposited during specific
1 Eliis
* Corresponding author, E-mail: julien.razza@eliis.fr DOI: 10.3997/1365-2397.fb2025069
time intervals. Yet, this process remains largely manual, often demanding weeks as experts pick depositional boundaries slice by slice through seismic volumes. While skilled interpreters certainly achieve accurate results, inherent subjectivity and significant time constraints often limit the scope and depth of analysis, particularly when dealing with complex depositional systems (Wu et al., 2023).
These limitations have naturally prompted the search for more efficient, consistent, and objective methods, driven by two primary challenges. First, depositional elements frequently exhibit subtle seismic signatures that blend with surrounding facies. For example, channel fills often share acoustic properties with overbank deposits, and thin turbidite beds can push the limits of seismic resolution (Chopra and Marfurt, 2005). Spectral decomposition attempts to enhance stratigraphic contrasts using frequency analysis and RGB colour blends, but it typically demands tedious trial-and-error frequency selection (Partyka et al., 1999). Even interactive approaches, which allow users to analyse frequencies at multiple locations, show improved results but still rely heavily on manual picking (Durot et al. 2017).
The second challenge lies in accurately translating stratigraphic observations into comprehensive 3D interpretations. Geological boundaries follow intricate patterns dictated by sediment supply, accommodation space, and depositional processes. Channels meander and migrate, carbonate platforms display complex progradational patterns, and deltas develop diverse geometries controlled by waves, tides, and river discharge. While deep learning models, such as U-Net (Ronneberger et al., 2015), offered the promise of automating stratigraphic segmentation, they have faced a critical drawback: their effectiveness is confined to the specific depositional environments on which they were trained. A model trained on fluvial systems, for instance, simply cannot extract carbonate build-ups without extensive retraining on thousands of carbonate platform examples. Moreover, even data-centric methodologies that rely on interactive labelling in 2D or 2.5D domains remain susceptible to human errors during annotation, still demand tedious manual work, and introduce a significant (and often uncontrolled) risk of overfit-
ting. This combination of domain specificity, high training costs, and inherent data labelling challenges has significantly hindered their widespread adoption (Milosavljević, 2020; Pratama and Latiff, 2022).
Recent advances in foundation models introduce a fundamentally different approach to tackling these challenges. These systems, pre-trained on millions of diverse images, understand boundaries, shapes, and spatial relationships without ever needing specialised geological examples. General-purpose visual foundation models, like the Segment Anything Model (SAM) family, can segment objects effectively using simple prompts, eliminating the need for extensive, domain-specific training (Ravi et al., 2024). This capability has opened new possibilities for stratigraphic interpretation.
We will present an interactive framework that integrates established geological knowledge with the capabilities of foundation models for stratigraphic interpretation. Our approach
maintains core sequence stratigraphic principles while amplifying an interpreter’s ability to map depositional architecture efficiently and consistently. We demonstrate how combining chronostratigraphic visualisation, user-guided spectral enhancement, and foundation model segmentation creates a practical workflow that can reduce interpretation time by an order of magnitude, all the while maintaining geological consistency.
Our framework guides interpreters through a structured process focused on depositional element extraction. This involves three key stages: (1) chronostratigraphic visualisation, (2) user-guided feature enhancement, and (3) interactive stratigraphic segmentation (Figure 1). Each stage builds upon the previous one, with the entire workflow operating on stratal slices that represent surfaces of constant geological time, ensuring interpretation stays within the proper geological context.
Chronostratigraphic architecture construction
The foundation of our approach is the computation of a Relative Geological Time (RGT) model from the seismic volume. This model assigns relative ages to every sample, effectively transforming the vertical axis into depositional time. The RGT model is built using a global optimisation process that propagates seismic events based on waveform similarity, ensuring the resulting horizons accurately follow the stratigraphy (Pauget et al., 2009).
This RGT volume serves as the base for all subsequent processing, generating a dense stack of conformable stratal slices that represent isochronous surfaces throughout the seismic volume.
Initial visualisation: Chronostratigraphic colour blending
With the chronostratigraphic framework established, understanding depositional architecture begins with visualisation techniques
that reveal stratigraphic relationships quickly and intuitively. Chronostratigraphic colour blending accomplishes this by highlighting depositional relief and stratigraphic boundaries within a proper temporal framework (Leblond and Souche, 2025).
The method operates on RGT-derived stratal slices. For each geological time of interest, we select three adjacent chronostratigraphic slices: one representing the geological time of interest, plus surfaces from slightly older and younger time slices. The distance between slices is variable, and proportional to the inverse of the RGT gradient. The same seismic attribute, typically seismic envelope, is then mapped onto these three strata slices in an RGB viewer. Data from the shallower slice is assigned to the red channel, data from the central slice to the green channel, and data from the deeper slice to the blue channel (Figure 2). This unique colour combination immediately reveals
depositional architecture often hidden in conventional displays. In areas where depositional surfaces lie parallel with minimal relief, such as condensed sections or uniform sheet deposits, the three time slices show similar seismic character, creating grayscale tones. Where significant amplitude contrast or relief exists, such as channel cuts, clinoforms, or carbonate buildups, the slices intersect different parts of the depositional architecture, creating distinct colour fringes that outline boundaries. These colour variations provide immediate visual cues: a red fringe might indicate where a feature first appears in the stratigraphic sequence while a blue fringe shows where a depositional element terminates.
Because colour-coded slices closely follow seismic reflectors, the distance separating them depends on the thickness of the deposits. As a result, the generated RGB image is influenced not only by amplitude contrasts but also by subtle frequency changes related to depositional thickness variations. This visualisation method proves particularly powerful for channel system analysis, where colour fringes clearly delineate channel margins, point bar migration patterns, and avulsion points.
User-guided feature enhancement: Optimised spectral decomposition
Once key depositional elements are identified through chronostratigraphic visualisation, interpreters move to the enhancement step. This step automates spectral decomposition to maximise contrast between depositional elements and their stratigraphic context, eliminating the tedious trial-and-error manual frequency selection (Figure 3A). This user-guided approach extends the work of Schmidt et al. (2013) by combining frequency analysis with geological understanding.
The process begins when interpreters place a small number of seed points on a stratal slice: ‘in’ points mark targets such as channel fill, or turbidite sandy lobes, while ‘out’ points mark contrasting facies like overbank mudstones, or background pelagic sediments (Figure 3B). This step ensures that the enhancement
targets true depositional contrasts rather than arbitrary amplitude differences. Typically, only a few points of each type are required to capture the representative seismic character of the target and its surroundings.
Once these points are placed, a data-driven algorithm takes over. The algorithm analyses the seismic frequency spectra at all seed locations and computes the difference between ‘in’ and ‘out’ spectra. To maximise visual contrasts, it identifies the local maxima of the residual spectrum while ensuring sufficient separation between selected frequency bands. If too few maxima are identified, the algorithm reverts to a percentile-based technique to find frequency bands that maximise the statistical separation between the ‘in’ and ‘out’ populations.
These optimal frequencies are then used to drive spectral decomposition across the volume, generating custom-tuned 3D RGB images where targets appear bright against muted background facies (Figure 3C). A key advantage of this technique over chronostratigraphic colour blending is its ability to generate colours that directly relate to the thickness and acoustic properties of the interpreted elements, maintaining consistency across stratigraphic slices. This consistency is crucial for the subsequent segmentation stage.
Having optimally enhanced the visual contrast of depositional elements, the final stage of the workflow leverages a general-purpose vision foundation model trained on millions of non-geological images. These models inherently understand object boundaries and spatial relationships without requiring geological training, enabling the extraction of diverse features ranging from sinuous channel systems and carbonate buildups to salt bodies and volcanic intrusions.
Foundation models require strong visual contrast for effective segmentation, which directly explains why the spectral enhancement step is so critical. Interpreters use these enhanced images to guide the model through point prompts (Razza and Sarcy, 2025): ‘positive’ prompts mark target depositional bodies (channel fills,
sand lobes, or progradational units), while ‘negative’ prompts explicitly exclude background facies (Figure 4). The model architecture employs a vision transformer backbone that processes
image patches through self-attention mechanisms, enabling it to understand complex spatial relationships across the entire stratal slice. As interpreters add prompts, the model generates real-time
segmentation masks for iterative refinement until the output matches their geological understanding.
To ensure stratigraphic consistency, the process typically repeats on key chronostratigraphic slices, ensuring that extracted objects maintain proper stratigraphic relationships and honour depositional sequence boundaries. The model then automatically propagates interpretation across adjacent, time-equivalent slices using prompt guidance from these key levels, maintaining continuity while reducing manual work.
Two types of constraints can be applied to improve segmentation accuracy. Spatially, constraints can be used by drawing a simple polygon to define an area of interest (e.g. Figure 6C). This focuses the segmentation on a specific region, saving computation time and preventing the model from generating results in geologically irrelevant areas. Stratigraphically, constraints can be applied by incorporating sequence stratigraphic surfaces (e.g. Figure 7A). Maximum flooding surfaces, sequence boundaries, and other chronostratigraphic surfaces can effectively bound the extraction, ensuring elements respect established frameworks and exclude facies from different time intervals.
The final output is a 3D probability volume where each voxel has a value from 0 to 1, indicating the model’s confidence that the voxel belongs to the target depositional element. Final depositional elements are extracted by selecting a probability threshold based on geological understanding and project requirements, generating 3D meshes representing true depositional architecture for subsequent volume calculations, connectivity assessment, and integration within basin or reservoir models.
To demonstrate the practical application and effectiveness of this workflow, we present two case studies from distinctly different depositional environments. Both examples followed the entire workflow explained previously, including the creation of a RGT model and the stack of stratal slices necessary to perform chronostratigraphic colour blending and optimised spectral decomposition. These case studies clearly illustrate the method’s performance on both clastic and carbonate systems.
Deep-water turbiditic channel systems: Maui field, New Zealand
Deep-water channels share many morphological characteristics with their fluvial counterparts, including sinuosities, cutoffs, and downstream sweeps of channel bends. However, deepwater channel systems are generally much larger in scale. The stratigraphic variation within deep-water slope channel complexes depends heavily on the processes that operate to erode, transport, and deposit sediment.
The Taranaki Basin (Figure 5A) provides an excellent case study for demonstrating the capacity to detect and segment various channel morphologies (Harishidayat et al., 2023). The basin contains up to 9 km of Cretaceous to recent sediments. For this study, our focus is on the Miocene interval, which is particularly rich in sands, with a large number of turbidite systems with contrasted morphologies (Figures 5B and C). To demonstrate workflow efficiency without overwhelming the presentation, we selected five representative channel complexes from the numerous channels and incisions present in the area.
The workflow handles different channel morphologies with equal effectiveness. For example, Channel #1 represents a narrow system (approximately 250m wide), with moderate sinuosity and narrow or absent margins. Despite its small size, a consistent 3D geobody can be obtained by establishing only a few ‘positive’/ ‘negative’ points on the main stratal slice and adjacent ones, resulting in clear and well-defined segmentation. The method also successfully extracts larger incisions. For instance, channel complexes #3 and #4 (500–1000m wide), though relatively linear, show amalgamated channels within a confined channel belt with more significant margins. These complex channels, lacking significant sinuosity, are likely to have developed on a relatively steeper slope.
The foundation model’s capability truly shines when extracting more complex and sinuous geometries (Channel complexes #1, 5 and Figure 4). These complexes often display less internal homogeneity due to channel stacking, with strong sinuosity and large meanders within belts spanning several km. For these highly complex systems, the interpreter may need to focus more on prompting, depending on the seismic data resolution and channel stacking rate. This involves placing the necessary number of ‘positive’ points within the structure and ensuring that the prediction on adjacent stratal slices corresponds to the desired structure.
All depositional bodies extracted from this seismic volume required no more than 10 minutes each, including the time for prompting and then saving the geobody and its confidence. By comparison, manual interpretation of similar features typically requires 2-3 hours per channel complex, representing a time reduction factor of 10-15×. Furthermore, the user can adjust the level of detail based on project requirements, either performing quick reconnaissance or detailed analysis, while still achieving significant time savings compared to manual interpretation. This workflow ultimately enables interpreters to easily and quickly visualise all the desired structures in 3D (Figure 5D).
Carbonate environments: Poseidon field overburden, Australia
Carbonate depositional environments present different interpretation challenges than clastic systems but are equally important as exploration targets. The spatial and temporal evolution of carbonate environments results from a combination of tectonic processes (e.g., faulting and accelerated subsidence), sea-level changes, oceanographic processes, and local sedimentary processes. This interplay generally results in a wide diversity of geometries and structures, which can be efficiently extracted when seismic resolution is adequate.
Our second case study focuses on the Browse Basin (Figure 6A), located on the Australia’s Northwest Shelf. Here, the inherited Jurassic rift-related topography was buried by an early Cretaceous passive margin sequence. The Cenozoic comprises mainly carbonate sequences, specifically documenting the transition from a ramp to a rimmed platform from the Eocene to the Miocene (Van Tuyl et al., 2018). We focus on this latter interval to demonstrate the workflow’s effectiveness on carbonate systems (Figure 6B).
The progressive evolution of the Miocene paleo-landscape becomes visible through the multiple stratal slices computed dur-
ing the previous steps. We present two representative slices after performing spectral decomposition with the selected frequencies. These two stratal slices effectively highlight the transition from a rimmed platform with a prograding reef barrier (stratal slice 1, Figure 6C) to a rimmed platform dominated by carbonate buildups (stratal slice 2, Figure 6C).
In stratal slice 1, the barrier reef forms an elongated and continuous structure along a SW-NE axis (Figure 6C), easily recognisable within several stratal slices. Its continuity makes segmentation straightforward and efficient with only a few ‘positive’ points. To accurately constrain its boundary geometry, a few ‘negative’ points are placed along its edges, preventing segmentation overflow. Additionally, at the barrier front, we identified a mega-slump structure (up to 25 km long, 15 km wide), including mega-clasts (up to 1 km) in the up-dip part. This gravity-driven collapse structure displays a basin-ward semicircular shape. Its strong contrast with the surrounding sediments and its presence on multiple stratal slices enables extraction with only a few ‘positive’ points. For additional control, a spatial constraint can be applied by embedding a polyline to limit the segmentation within an area, as shown in stratal slice 1.
Moving upward stratigraphically within the stratal slice 2 (Figure 6C), the progressive coalescence of isolated build-ups (ranging from a few hundred metres to kilometres in diameter) results in the formation of build-ups of significant size (more than 10-km wide). Spectral decomposition effectively highlights these features, which exhibit curved-to-circular shapes and distinctive patterns of progradational geometries. For these features, the extraction of 3D geobodies becomes almost trivial. By adding a single ‘positive’ point, the model immediately recognises the shapes, and its limits can be refined by adding a few ‘negative’ points to achieve near-perfect delineation.
Similar to the clastic example, this entire operation was completed in less than 45 minutes for all features, compared to several days typically required for manual editing. The significant
time savings allow the user to focus on geological interpretation and lateral variability rather than tedious manual picking.
While our proposed workflow primarily focuses on extracting depositional elements, the segmentation capabilities of the foundation model allow for broader application. This section presents its utility for post-depositional features like salt bodies, igneous intrusions, and injectites (Figure 7). For these features, which often exhibit sharp seismic contrasts, direct application of the segmentation algorithm is possible. This flexibility means that while chronostratigraphic visualisation and optimised spectral decomposition are crucial for subtle stratigraphic features, the segmentation step can be leveraged more directly when sufficient visual contrast already exists.
Salt bodies, due to their massive and homogeneous nature, can be extracted directly from the seismic image using the segmentation algorithm. Here, image segmentation works best along the inline or crossline seismic orientation, where the salt bodies’ texture clearly differs from the background. Traditional structural seismic attributes (e.g., chaos) can further enhance salt boundaries, simplifying extraction. Additionally, incorporating a pre-existing horizon as a constraint ensures the geobody accurately follows the salt-sediment interface.
Injectite complexes represent another compelling application, where post-depositional intrusion of sediments affects pre-existing stratigraphic sequences. While they inherently disrupt original depositional architecture, the segmentation algorithm effectively extracts their complex 3D geometries once enhancement creates adequate contrast between the injected materials and their host sediments.
Throughout all these applications, the key principle remains maintaining geological integrity. Whether targeting depositional elements within sequence stratigraphic sequences or extracting post-depositional features, the interpreter must ensure that visual-
isation, enhancement, and segmentation strategies consistently align with the geological processes responsible for feature formation and preservation.
Discussion
The success of this workflow relies on aligning general-purpose visual segmentation models with the specific requirements of seismic interpretation. These models demonstrate zero-shot learning capabilities, meaning they can delineate objects they have never ‘seen’ before, purely based on differences in texture and colour. Furthermore, video segmentation models naturally handle series of related images (like video frames) that exhibit geometrical continuity, similar colour palettes, and consistent textures – properties that align well with seismic stratal slices.
To fully leverage these models, seismic data must be conditioned to: (1) make depositional objects clearly identifiable, (2) maximise texture and colour contrast, and (3) maintain colour consistency throughout the stratigraphic pile. Our workflow achieves these requirements through its three-stage design.
First, using high-resolution stratal slices derived from an RGT model ensures sedimentary deposits are properly delineated along geological time surfaces. Second, the user-guided identification of optimal spectral decomposition frequencies maximises the colour-contrast between the interpreted objects and their surroundings. Third, the intrinsic link between spectral decomposition’s RGB colours and layer thickness ensures colour consistency from one stratal slice to the next.
The proposed framework therefore optimises the accuracy of the final segmentation while maximising the interpreter’s control. However, in the context of seismic interpretation, general purpose visual segmentation faces two main limitations:
Limited to RGB channels: the models can currently only process three different colour channels, whereas seismic data contains richer information across more frequency bands or from multiple stratal slices that could further enhance characterisation.
Sequential processing: the models are designed to work on a single ordered set of images (akin to a video), preventing simultaneous interpretation across multiple orientations such as inlines, crosslines, and stratal slices concurrently.
Fortunately, these limitations are partially offset because the segmentation process outputs a probability volume. This facilitates the combination and merging of results from various segmentations performed on different attributes or along different orientations, offering flexibility in post-processing.
Our approach also avoids several major pitfalls common in other AI-driven interpretation systems. It eliminates the need for large, pre-labelled training datasets, which is a significant practical obstacle for many active learning strategies (Lowell et al., 2019). It also avoids the problem of ‘catastrophic forgetting,’ where a model trained incrementally on new data loses knowledge of previously learnt patterns, because our foundation model is never retrained, only prompted (Evano and Cubizolle, 2023). This strategy aligns with a broader, emerging trend of using powerful, pre-trained foundation models in specialised scientific domains, including remote sensing (Osco et al., 2023) and medical imaging (Zhang et al., 2024b).
Despite these advantages, quality control remains critical. While foundation models provide remarkable segmentation capabilities, geological expertise must guide the entire process. Interpreters need to continually evaluate results against sequence stratigraphic principles and validate outputs using available well data and regional geological knowledge. Moreover, the probability outputs from the foundation model provide a quantitative measure of interpretation uncertainty, which can be directly incorporated into reservoir modelling workflows and risk assessment.
We have developed an interactive framework for stratigraphic interpretation that automates depositional element extraction while preserving geological accuracy. This three-stage workflow combines chronostratigraphic visualisation, user-guided spectral enhancement, and foundation model segmentation, ensuring geological principles are maintained throughout the process.
By operating on time-equivalent surfaces, our framework guarantees that extracted objects represent true depositional architecture rather than arbitrary seismic artifacts. The probability outputs further quantify interpretation uncertainty, providing valuable input for reservoir modelling and risk assessment decisions. Most significantly, the framework advances stratigraphic interpretation by automating spectral optimisation while adhering to sequence stratigraphic principles, enabling consistent and rapid mapping of complex depositional architectures for energy exploration.
The demonstrated time savings of 10-15× compared to manual interpretation, combined with maintained accuracy, make this approach immediately applicable to both exploration and development projects. As foundation models continue to improve and seismic data quality increases, we anticipate that this type of workflow will become a standard tool for stratigraphic interpretation, empowering geoscientists to focus on geological understanding rather than manual digitization.
We thank New Zealand Petroleum and Minerals, Geoscience Australia, the US Geological Survey, Equinor and the Volve licence partners for providing the datasets used in this study.
Chopra, S. and Marfurt, K.J. [2005]. Seismic attributes — A historical perspective. Geophysics, 70(5), 3SO-28SO.
Durot, B., Mangue, M.M., Luquet, B.L., Adam, J.P.A. and Daynac, N.D. [2017]. Innovative and Interactive Methods Emphasizing Geological Events through Spectral Decomposition New Zealand Case Study. 79th EAGE Annual Conference and Exhibition, Extended Abstracts
Evano, L. and Cubizolle, F. [2023]. Manual Active Learning for Salt Interpretation: An Empirical Study to Avoid Forgetting During Incremental Trainings. Third EAGE Digitalization Conference and Exhibition, Extended Abstracts Larsen, C., Harishidayat, D. and Omosanya, K.O.L. [2023]. Geomorphologic control on the evolution of Middle-Late Miocene submarine channels in the Southern Taranaki Basin, New Zealand. Marine and Petroleum Geology, 156, 106447.
Leblond, R. and L. Souche. [2025]. Chronostratigraphic Color Decomposition: A Fast Method for Identifying Sedimentary Features in Seismic Data. Fifth EAGE Digitalization Conference and Exhibition, Extended Abstracts
Lowell, D., Lipton, Z.C. and Wallace, B.C. [2019]. Practical obstacles to deploying active learning. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing, 21-30.
Milosavljević, M. [2020]. Early integration of machine interpretation and structural frameworks. First Break, 38(7), 45-51
Osco, L.P., Wu, Q., de Lemos, E.L. et al. [2023]. The Segment Anything Model (SAM) for Remote Sensing Applications: From Zero to One Shot. International Journal of Applied Earth Observation and Geoinformation, 124, 103516.
Partyka, G., Gridley, J. and Lopez, J. [1999]. Interpretational applications of spectral decomposition in reservoir characterization. The Leading Edge, 18(3), 353-360.
Pauget, F., Lacaze, S. and Valding, T. [2009]. A global approach in seismic interpretation based on cost function minimization. SEG Technical Program, Expanded Abstracts, 2592-2596.
Pratama, H. and Latiff, A.H.A. [2022]. Automated Geological Features Detection in 3D Seismic Data Using Semi-Supervised Learning. Applied Sciences, 12(13), 6723.
Ravi, N., Gabeur, V., Hu, Y.-T. et al. [2024]. SAM 2: Segment Anything in Images and Videos. arXiv preprint arXiv:2408.00714.
Razza, J. and Sarcy, L. [2025]. Interactive Geobody Extraction: Combining Expert Knowledge with Deep Learning Segmentation. Fifth EAGE Digitalization Conference and Exhibition Extended Abstracts
Ronneberger, O., Fischer, P. and Brox, T. [2015]. U-Net: Convolutional Networks for Biomedical Image Segmentation. Proceedings of the International Conference on Medical image computing and computer-assisted intervention
Schmidt, I., Lacaze, S. and Paton, G. [2013]. Spectral Decomposition combined with geo-model interpretation: Creating new workflows by integrating advanced technologies for seismic imaging and interpretation. 75th EAGE Annual Conference and Exhibition, Extended Abstracts
Van Tuyl, J., Alves, T.M. and Cherns, L. [2018]. Geometric and depositional responses of carbonate build-ups to Miocene Sea level and regional tectonics offshore northwest Australia. Marine and Petroleum Geology, 94, 144-165.
Wu, X., Sheng, H., Si, X. et al. [2023]. Seismic Foundation Model (SFM): a new generation deep learning model in geophysics. arXiv preprint arXiv:2309.02791.
Zhang, Y., Shen, Z. and Jiao, R. [2024b]. Segment Anything Model for Medical Image Segmentation: Current Applications and Future Directions. Computers in Biology and Medicine, 167, 107669.
Explore the latest innovations in site characterisation at the 3rd EAGE/ SUT Workshop for Offshore Renewable Energy. Join industry experts in Australia next October 2025 as they share insights, tools, and techniques driving the future of sustainable offshore development
Karyna Rodriguez1*, Lauren Found1, George Kovacic1 and Neil Hodgson1 present the geoscience evidence that suggests there is huge potential in blocks currently in the 2025-26 Nova Scotia licensing round.
Supernovae are the natural evolution of massive stars as they release phenomenal amounts of energy whist transitioning to neutron stars. These piercing lights shine, when their moment comes, out across the galaxy just as super Nova Scotia’s bright prospectivity is shining at exactly the moment when the evolution of the industry needs big, simple, repeatable success. This beacon cutting through the fog of perception will be a bellwether for our industry in 202526, as a licensing round has opened to invite explorers to re-evaluate their perceptions of both the low-risk shelf and the more uncertain but potentially huge prospectivity on the lower slope.
Offshore Nova Scotia has had its first moment in the sun already from the oil and gas discoveries of the Sable Island area southeast of Nova Scotia (Figure 1). Exploration around Sable Island started in 1967 and continued to the mid-1980s, in Cretaceous and Jurassic sandstone plays (largely tilted fault blocks as this was the play you could image). The success rate reached one in three despite wells being drilled on (by today’s standards) pretty frightful 2D seismic data.
3D seismic (a number of small footprint acquisitions) was acquired over the fields in the 1990s to early 2000s, as drilling became primarily development-oriented. Gas production commenced in 1999, and the gas and oil fields developed on the shelf produced over 2 TCF of gas and 45 MMBbls of oil. Gas production facilities are now abandoned, despite the discovery in the early 2000s of several still undeveloped gas fields, and the recognition of multiple untested plays.
So, the shelf now lingers on the liminal shores, waiting for a new phase of exploration and interest. Yet, if you have to look for new low-risk accumulations, you can do worse than examine adjacent to where it has been found before. All parts of the hydrocarbon story are derisked, and today’s explorer will be armed with cheap, available, reprocessed 3D seismic to revisit this arena of past successes and surpass them. The potential below Sable Island is considerable as the proprietary 3D seismic has not been reprocessed since it was acquired, the imaging then was moderate in the shallows and poor at depth, and 25 years of processing technology is well capable of shining a new light on this area. Where structure was barely mappable, sedimentology
1 Searcher Seismic
* Corresponding author, E-mail: k.rodriguez@searcherseismic.com
DOI: 10.3997/1365-2397.fb2025070
Figure 1 The historical discoveries, blocks in the 2025-26 Licensing round, and key prospects offshore Nova Scotia.
and DHI technology will give explorers with new eyes, powerful tools for igniting a supernova of brilliant success, in a new wave of low-risk exploration on this margin.
A light shining in the salty deep
However, in Nova Scotia, there is another, bigger game afoot, where just as gravity crushes stars into Supernovae, so during the Jurassic and Early Cretaceous gravity dragged turbidites stuffed with coarse-grained quartz-rich sands off the shelf and down slope toward the basin floor. It is possible that the slope was so steep up-dip that sands were just not deposited and the few wells near the slope crest have indeed been mud-rich. They might have continued far down to the basin floor, however, under the northeastern part of Nova Scotia’s slope lurked turbulent halokenetically hyper-active turbidite traps, ready to pounce. Breaks of slope, topologically induced tortuosity in the confined turbidite conveying channel systems were generated as sediments sunk into the underlying halofer creating walls and diapirs through the Jurassic. In the central region of the Tangier 3D survey, these diapirs rose near to the seabed when either salt feed from the halofer ceased due to pod grounding or the diapirs achieved buoyant equilibrium. Yet the Early Cretaceous sediments still poured off the shelf until the eastern parts of the lower slope salt basin were so overloaded that salt was ejected out onto the seabed as salt lavas. When this lava was self-buried, and became a salt canopy, the overlying sediment again began podologically sink-
ing creating new Late Cretaceous and Tertiary pods that grounded onto underlying Early Cretaceous pods (Figure 2).
And so, in what is known as the ‘fourth exploration cycle’ (Wach and Brown, 2021), the exploration of the salt basins on the Nova Scotian lower slope began and ran between 2015 and 2018. Just three wells were drilled, the last of which bp Canada’s Aspy-D11 well was drilled in a sub-canopy play and encountered good shows of oil and was drilled on the most technically advanced wide azimuth data acquired anywhere in the world at that time. The Tangier 3D is an extraordinary piece of technology. Searcher has interpreted the full 8502 km² survey in detail, particularly focusing on the central and eastern 5000 km² reprocessed portions of the survey. Working with our technical partner Lyme Bay, using a version of the Paleoscan technology, we have sought to find the transport route for sands off the shelf, where the sands have encountered topological complexity and been deposited in channels or fans in mini-basin settings.
It is into this extraordinary environment that the Aspy-D11 well was drilled. Drilling through the upper salt canopy in a non-crestal position and encountering the good oil shows just below the salt, the well drilled on into a phase 1 pod. The well was drilled into a thick pod of Early Cretaceous and Jurassic sediment wrapped up in salt like a Beef Wellington. The concept of sands caught in the inter-pod mini-basins has been well explored in a number of salt basins globally and has proven very effective. However, in this case, whilst the seismic defines the structure,
Figure 2 Fully stack PSDM and Sweetness attribute W-E line through the Tangier 3D. To the east (right) the complexity of podological halokinesis with a welldeveloped salt canopy above the Early Cretaceous, makes imaging sand fairways complex. To the west however, imaging is so good the sand fairways down the slope have been revealed.
Figure 3 RMS amplitude of the Hecate and Nemesis
Upper Cretaceous sand channel (similar in age to the thin sands encountered in Aspy-D11 well). This channel linking a series of pods between salt walls / diapir stocks has several depth consistent Amplitude vs Offset (AVO) shut-offs.
and the well encountered good oil shows, the evidence for sand as derived from RMS amplitude maps within the Aspy-D11 pod, was limited.
In fact, during this time in the early Cretaceous, a lot of sand was coming into the basin and down slope yet it was being channelled through the western and central part of the Tangier 3D, rather than the east where the Aspy well was drilled. This area was the focus of prospectivity evaluation (see Figure 3).
The main sand channel fairways that brought Early Cretaceous sands down the slope actually pass through the Tangier 3D between 30 and 50 km west of the Aspy-D11 well. These sand fairways are defined using RMS amplitude extractions from the 3D, and they show that not only are there very wide, thick channel systems on the block but that they have tangled with the halokinetic palimpsest of a slope topology and sands have found themselves in inter salt
Figure 4 Example of RMS amplitude maps derived from the numerous stacked plays that border the Elektra Diapir. The surfaces come from a shallow group of Upper Early Cretaceous sand fans, a deeper Lower Cretaceous set of ponded fans and an underlying Upper Jurassic channel feature.
Figure 5 Geothermal modelling in the western and eastern portions of the Tangier 3D. There is considerable uncertainty on the geothermal gradient. Within this uncertainty the Callovian source rock (J160) in the west may be in the oil window (LO case) (supported by data from the Shelbourne area to the west). A higher, less controversial geothermal gradient of 25 oC/km is used for resource estimates calculated on the basis of gas rather than oil.
wall pods, onlapping large diapirs, on the flanks of salt walls. In the central portion of the Tangier 3D, seven substantial channel bypass/ truncation type traps have been mapped with RMS amplitude, dip conformant shut off, AVO and even flat spot support, with multiple stack targets ranging from Late Early Cretaceous to Upper Jurassic in age (one such example seen in Figure 4). The stacking of targets is key as some of these prospects have up to three discrete targets each of which have multiple stacked pay and these can be tested from a single well.
For estimation of prospect resources, parameters and methodology was similar to the Scotian Basin Integration Atlas 2023 estimation method (see reference below). Gross Rock Volume (GRV) was calculated on the basis of assumed individual gross and net sand thicknesses. However, net sand volumes calculated are consistent with those in the Scotian Basin Integration Atlas 2023 study.
Fan sandstones are assumed to be 30-m thick with 50% Net to Gross (NTG), whilst channel systems are assumed to be 50-m thick with effective Net to Gross of that unit of 80%. We assumed sand porosity averages 18%, hydrocarbon saturation 60% and
Formation Volume Factor (FVF or 1/Bg) as 300. These values are similar to those used by Scotian Basin Integration Atlas 2023 in their recent resource assessment.
Due to the areal extent of the anomalies mapped, even quite thin sands would generate substantial resources. However, multiple sands and stack targets make for some very large potential resources, both in individual prospects and in summary located in the central portion of the Tangier 3D. These resource estimates are unrisked, yet the potential for repeated success in a seismically driven DHI play fairway, leading to really game-changing resources would appear low risk. Strong seismic indicators of both abundant sand in the central portion of the Tangier 3D allow us to complement a traditional risking method based on estimates of charge risk, reservoir risk and trap risk, with an approach that allows us to just express our confidence in the seismic DHIs.
Supernovae are the universe’s factories, creating and distributing heavy elements necessary for life. So too will the building blocks for economic independence be produced by
Figure 6 Highest ranked prospects located the Central portion of the Tangier 3D.
the explosion of new exploration offshore Nova Scotia. By targeting the key risk of the slop / salt basin – this is finding the sand that fell off the shelf in the Upper Jurassic and Early Cretaceous – huge potential in blocks have been identified that are currently in the 2025-26 licensing round. The Nova Scotian sun started to shine on the shelf – but the supernova journey to vast resources on the low slope is igniting in the 2025-26 bid round.
Beicip-Franlab and Nova Scotia Department of Natural Resources and Renewables*. 2023. Scotian Basin Integration Atlas 2023, A collaborative study of stratigraphy, architecture, evolution, and geologic prospectivity. - https://oera.ca/research/scotian-basin-integration-atlas-2023. (*Now Department of Energy).
Wach. G.D. and Brown, D.E. [2021]. Petroleum Exploration on the Scotian Margin. Proceedings of Geoconvention Virtual Event 2021
E-Lectures, Distinguished Lecturer Webinars, How-to-Videos
EAGE COURSES
Extensive, Self-Paced, Interactive, EAGE Education Tours
PARTNER COURSES
Online courses provided by EAGE partners
Explore all the courses in our catalogue!
Sonny Winardhi1*, Asido Saputra Sigalingging2 and Ekkal Dinanto1 integrate extrapolated low-frequency data into a source-independent FWI workflow, significantly improving the resulting velocity model and enabling reliable inversion without requiring accurate source waveform estimates.
Abstract
Marine seismic data often lacks low-frequency content, limiting its usefulness for advanced modelling and inversion techniques such as Full Waveform Inversion (FWI). In this study, we applied a self-supervised deep learning (SSL) framework for low-frequency extrapolation, tested on field data from the Asri Basin, Indonesia. Our approach uses a modified U-Net architecture and a two-stage learning scheme, namely: synthetic warm-up followed by iterative data refinement (IDR), to train directly on unlabelled band-limited seismic data. When used as input for a source-independent FWI (SI-FWI) method, the low-frequency extrapolated data significantly improves inversion results, demonstrating better recovery of deep velocity structures and reduced cycle skipping compared to inversions using band-limited data as inputs. The integrated SSL and SI-FWI workflow provides practical value for reprocessing legacy datasets lacking both low-frequency energy and source wavelet information. This method offers a data-driven, label-free approach for expanding the spectral bandwidth of field seismic data and further improving inversion convergence.
Introduction
Low-frequency seismic signals play a critical role in Full Waveform Inversion (FWI), particularly in addressing issues such as cycle skipping and non-uniqueness in the inversion process (Bunks et al., 1995). However, these low-frequency components, typically below 10 Hz, are often missing in older seismic datasets due to limitations in acquisition tools, source bandwidth, or historical survey designs. The absence of low frequencies significantly reduces the reliability and resolution of inversion results, making it difficult to produce accurate subsurface velocity models.
Many researchers have adopted deep learning for signal processing, particularly in the seismic domain, and have demonstrated the capabilities of deep learning in addressing various challenges (Sun & Demanet, 2020; Sigalingging et al., 2024; Winardhi et al., 2024). To address the issue of low-frequency
component loss, we adopt a self-supervised learning (SSL) framework for low-frequency reconstruction, inspired by recent developments in SSL for geophysical applications, such as Cheng et al. (2024). SSL offers the advantage of generating training pairs (inputs and pseudo-labels) directly from observed seismic data, eliminating the need for manually labelled datasets or synthetic labels. This approach is based on the Noisier2Noise strategy (Moran et al., 2020), where neural networks are trained to recover the original signals using only corrupted versions of the data. By intentionally applying additional high-pass filtering, the model learns to recover low-frequency components without requiring clean or complete training targets.
In this study, we applied a modified U-Net architecture (Ronneberger et al, 2015) within the SSL framework to reconstruct low-frequency signals from band-limited marine seismic data. This method was first verified on a synthetic dataset using the Marmoussi2 velocity model. We then applied the trained model to real seismic data obtained from a 2D marine survey line in the Asri Basin, Java Sea, Indonesia. Legacy data from this region is characterised by limited spectral bandwidth and poor low-frequency content due to acquisition constraints (Ralanarko et al, 2021).
Finally, we show that integrating extrapolated low-frequency data into a source-independent FWI workflow significantly improves the resulting velocity model. This integrated approach enables reliable inversion without requiring accurate source waveform estimates, making it well-suited for our case study.
Background
Self-Supervised Learning for Seismic Data
Self-supervised learning (SSL) has emerged as a promising alternative to supervised methods in seismic processing, particularly when labelled data is unavailable. Unlike conventional approaches that require clean target data, SSL generates pseudo-labels directly from the observed data. This study adopts the LessLowto-Low (L2L) strategy, inspired by the Noisier2Noise framework
1 Institute of Technology Bandung | 2 Institute of Technology Sumatera
* Corresponding author, E-mail: swinardhi@itb.ac.id
DOI: 10.3997/1365-2397.fb2025071
(Moran et al., 2020), where neural networks learn to recover signals using more degraded versions as input. In L2L, we apply an additional high-pass filter to the already band-limited seismic data, treating the less-degraded version as the pseudo-label. This allows the model to learn to reconstruct low frequencies without requiring external training labels, making it highly suitable for real-world field data applications.
Source-Independent FWI
Source wavelets are often difficult to model accurately due to limitations in data acquisition. However, accurate source wavelets are essential for Full Waveform Inversion (FWI) convergence. To overcome these limitations, Source-Independent Full Waveform Inversion (SI-FWI) has been developed. Choi and Alkhalifah (2011) and Zhang et al. (2016) successfully applied SI-FWI to marine seismic data, demonstrating that this method can effectively eliminate artifacts caused by incorrect or unknown source wavelets.
The source independent objective function consists of the convolution of the modelled wavefield with a selected reference trace from the observed seismogram, subtracted from the convolution of the observed wavefield with a selected reference trace from the modelled seismogram. We implement the source-independent FWI formula in the time domain following Zhang et al, (2016); the objective function is defined by:
(1)
The back-propagated source of the adjoint wave equation based on the above objective function is defined by:
Where * denotes the convolution operator, is the objective function, is the synthetic seismic data generated through forward simulation, is the observed data, , and are the synthetic and observed reference traces, defined by the average of n traces stacked around the source location.
Self-supervised learning (SSL) Workflow
Self-supervised learning (SSL) for low-frequency extrapolation essentially consists of two components: a warm-up phase and iterative data refinement (IDR). The warm-up phase involves supervised learning using a synthetic dataset. In this phase, the model is trained on data generated from a simulated subsurface model. A high-pass filter is applied to the original data to create the input dataset, while the unfiltered original data serves as the labelled target, thereby forming the ‘less-low-to-low’ dataset. The model is trained for a predetermined number of epochs, and the optimal model from this phase is then used as the backbone for the subsequent IDR phase.
In the IDR phase, the model is iteratively refined while the dataset is built by predicting band-limited seismic data. These predictions serve as initial pseudo-labels for the IDR phase, with the corresponding input obtained by applying a high-pass filter to the predictions. In the first iteration, the previously trained model from the warm-up phase is used to predict the original seismic data. The SSL workflow is illustrated in Figure 1.
Deep learning architecture
The self-supervised learning (SSL) framework introduced by Chen et al. (2024) for low-frequency tasks utilises a conventional U-Net architecture. In this study, we modified the architecture to better suit our problem. The details of the modified architecture are illustrated in Figure 2. Our model consists of five scales, employing 2×2 down-sampling and 2×2 up-sampling operations. Each convolutional layer uses a 3×3 kernel without batch normalisation, and the Leaky Rectified Linear Unit (Leaky ReLU) activation function is applied. Each block in the model comprises two consecutive convolutional layers. The number of filters in both the encoder and decoder stages is set to 96, while the final layer contains 48 filters. The complete filter distribution is depicted in Figure 2.
We employ a hybrid loss function that combines data loss and amplitude spectrum loss. The hybrid loss function is formulated using the mean absolute error (MAE) and can be expressed as:
is the selection of the reference trace, which must be defined for both the observed and modelled datasets.
In this study, we assume a flat surface model, which simplifies the process of selecting an optimal time window just below the first arrival for constructing the reference trace. To ensure stability and signal quality, we select an average reference trace from near-offset receivers close to the source location. These traces typically preserve the source wavelet with higher fidelity than those at distant offsets, where waveforms are often affected by scattering, attenuation, and geometrical spreading.
For model updates during the inversion, we employ a spectral gradient projection method combined with a line search algorithm, which ensures efficient and stable convergence across frequency stages. We utilised Julia Frameworks called JUDI to perform FWI (Witte et al, 2018).
Result and discussion
Synthetic benchmark results
where represents the label data, and O denotes the output of the models.
For the inversion, we adopt an iterative multiscale strategy by progressively increasing the dominant frequency of the source wavelet in stages (5, 7, 15 and 20 Hz). At each frequency level, we perform up to 10 iterations using source-independent Full Waveform Inversion (SI-FWI). A critical parameter in SI-FWI
We validated the self-supervised learning (SSL) workflow using synthetic shot gathers generated from the Marmoussi2 velocity model. The model was tested with input data filtered at 5, 10, and 15 Hz cutoff frequencies to simulate varying degrees of low-frequency loss. Results show that the SSL model successfully reconstructs missing low-frequency components, particularly for inputs with cutoffs at 5 and 10 Hz, as depicted in Figure 3. Residual analysis indicates that prediction errors increase as the cutoff frequency rises, with the largest discrepancies observed for inputs filtered at 15 Hz. This increase in residuals at higher cutoffs is expected, as the model receives less information
Figure 3 The results of testing the low-frequency data are presented as follows. We evaluate the model's accuracy by predicting low-frequency components using input data with different high-pass filter cutoff frequencies: (a) 5 Hz, (d) 10 Hz, and (g) 15 Hz. The corresponding predictions obtained using the self-supervised learning (SSL) model are shown in (b) 5 Hz, (e) 10 Hz, and (h) 15 Hz. To assess the prediction quality, the residuals-computed as the diIerence between the predicted data and the original shot data shown in Figure 4(a) are displayed in (c) 5 Hz, (f) 10 Hz, and (h) 15 Hz.
about the low-frequency trend from the input data. With fewer low-frequency cues available, the extrapolation task becomes more underdetermined, making it more difficult for the model to accurately reconstruct the missing components.
Waveform comparisons on selected traces demonstrate that the predicted signals closely match the original data, with near-perfect phase alignment even as low-frequency content diminishes, as shown in Figure 4. This is critical for Full Waveform Inversion (FWI), as accurate phase reconstruction directly reduces the risk of cycle skipping. Spectral analysis further confirms the model’s ability to restore energy below 5 Hz, with reconstructed spectra closely matching the original, even when input data lacked substantial low-frequency information. The model recovers frequency content down to 2 Hz, outperforming conventional methods, which tend
to underestimate energy in this range. These results highlight the robustness and accuracy of the proposed SSL-based extrapolation method, confirming its suitability for enhancing legacy seismic data in preparation for inversion workflows.
To evaluate the inversion performance, we implemented Source-Independent Full Waveform Inversion (SI-FWI) using both the band-limited and SSL-enhanced datasets. Figure 5 shows (a) the original Marmoussi2 velocity model, (b) the inversion result using high-pass-filtered data above 7 Hz, and (c) the inversion result using data enhanced by SSL low-frequency extrapolation. The inversion result in Figure 5b shows limited recovery of deeper and low-wavenumber features due to the lack of low-frequency content, and noticeable cycle skipping artifacts. In contrast, the result in Figure 5c, using SSL-enhanced data,
Figure 4 Comparison of (a) the original trace waveform with the predictions obtained using input data processed with different HPF cutoff frequencies (5, 10, and 15 Hz), and (b), (c), and (d) the corresponding frequency spectra of the predictions at 5, 10, and 15 Hz, respectively.
Figure 5 Comparison of Marmoussi2 velocity models: (a) original reference model; (b) inversion result using high-pass filtered data with missing frequencies below 7 Hz; (c) inversion result using data extended with low-frequency extrapolation (LFE) via self-supervised learning (SSL). The result in (c) demonstrates improved convergence and more accurate recovery of deep and low-wavenumber structures.
demonstrate significantly improved model recovery, especially in deeper and structurally complex regions. The inverted model shows much better alignment with the ground-truth velocity model, confirming the effectiveness of the extrapolated low frequencies in supporting stable and accurate inversion. These results demonstrate that SSL not only improves spectral completeness but also enhances the performance of SI-FWI by reducing inversion artifacts and improving convergence. The ability to reconstruct reliable low-frequency data directly from band-limited input has significant implications for legacy seismic datasets, where accurate source wavelets and low-frequency energy are often absent.
Field data results – Asri Basin
We applied the proposed self-supervised learning (SSL) method to a legacy 2D marine seismic dataset acquired in the Asri Basin,
Figure 6 (a), (d), and (f) present test shot datasets obtained by applying high-pass filters with cut-off frequencies of 5 Hz, 10 Hz, and 15 Hz, respectively, to the original seismic data. These filtered datasets serve as inputs to our self-supervised learning (SSL) model, with the corresponding prediction results displayed in Figures 8(b), (e), and (g). In contrast, Figure 8(c) shows the full-band seismic data processed using conventional methods.
Java Sea. The primary objective was to enhance the dataset’s spectral bandwidth by reconstructing missing low-frequency components, which are critical for effective inversion workflows. Like many legacy surveys, the Asri data lacked usable frequencies below 10 Hz, limiting its suitability for full waveform-based imaging.
Using the SSL framework trained previously, we processed the field data without requiring any additional labelling or synthetic augmentation. The extrapolated results showed significant enhancement in the low-frequency range between 2 and 8 Hz, while preserving the original mid- and high-frequency components, as shown in Figure 6. Spectral comparisons before and after extrapolation, as depicted in Figure 7, revealed a substantial extension of the usable bandwidth, recovering information that is typically inaccessible through conventional deghosting or spectral shaping methods.
This enhanced dataset was then used as input for Source-Independent Full Waveform Inversion (SI-FWI). The inversion was conducted using the same convolution-based misfit function applied in the synthetic tests. Figure 8 is the resulting velocity
Figure 7 shows a comparative spectral analysis of field data from the Asri Basin. In each subplot, the spectral response of a single trace is displayed using three lines: the blue line represents data reconstructed by conventional extended frequency methods, the orange line shows the prediction by the SSL model, and the green line denotes the input data processed with a specific low-frequency cut-oI. Subplot (a) corresponds to a 5 Hz cut-oI, subplot (b) to a 10 Hz cut-oI, and subplot (c) to a 15 Hz cut-oI.
model of SI-FWI in the Asri basin, showing strong consistency with geological expectations and reflects an improved reconstruction of deep structures compared to the inversion using unprocessed band-limited data.
Notably, the improvements observed in the Asri basin SI-FWI results closely align with those from the synthetic Marmoussi2 experiments. In both cases, the inclusion of SSL-extrapolated low frequencies led to better convergence, more stable inversion behaviour, and reduced cycle skipping. In the Asri case, the inversion revealed clearer reflector geometries and enhanced lateral continuity, particularly in deeper intervals where conventional inversion failed to resolve structures accurately.
These findings demonstrate that the proposed SSL-based extrapolation can be effectively applied as a preprocessing step to enhance legacy seismic data. When combined with SI-FWI, it offers a powerful, data-driven workflow for broadband velocity model-building, even when both low-frequency content and source information are limited or unavailable.
In this study, we presented a self-supervised learning (SSL) framework for seismic low-frequency extrapolation, applied to both synthetic and real-world marine seismic datasets. Using a modified U-Net architecture and a data-driven training strategy, the model was able to reconstruct missing low-frequency components down to 2 Hz, even when input data lacked significant frequency content below 10 or 15 Hz. The extrapolated data preserved phase alignment and mid-frequency integrity, which are crucial for reducing cycle skipping in Full Waveform Inversion (FWI).
When combined with Source-Independent FWI (SI-FWI), the SSL-enhanced datasets produced significantly improved inversion results. In the synthetic Marmoussi2 model, the extrapolated data enabled more accurate recovery of deeper and low-wavenumber structures compared to inversions performed on high-pass-filtered input. This improvement is reflected in the real-world application to the Asri basin dataset, where the inversion model demonstrated better structural continuity and geological plausibility, especially in deeper zones.
The combined SSL and SI-FWI approach offers a practical solution for enhancing legacy datasets lacking low-frequency content and source wavelet information. While the results are promising, limitations remain in terms of training data diversity, sensitivity to residual noise, and adaptability across different datasets. Future work will address these aspects to further improve robustness and applicability.
We thank PT Pertamina’s Upstream Research & Technology Innovation for data support and infrastructure. We also acknowledge our institutional collaborators at Institut Teknologi Bandung and Institut Teknologi Sumatera.
References
Bunks, C., Saleck, F.M., Zaleski, S. and Chavent, G. [1995]. Multiscale seismic waveform inversion. Geophysics, 60(5), 1457–1473. https:// doi.org/10.1190/1.1443880.
Cheng, S., Wang, Y., Zhang, Q., Harsuko, R. and Alkhalifah, T. [2024].
A self-supervised learning framework for seismic low-frequency extrapolation. Journal of Geophysical Research: Machine Learning and Computation, 1, e2024JH000157, https://doi.org/10.1029/ 2024JH000157.
Choi, Y. and Alkhalifah, T. [2011]. Source-independent time-domain waveform inversion using convolved wavefields: Application to the encoded multisource waveform inversion. Geophysics, 76(5), R125–R134, doi:10.1190/geo2010-0210.1.
Moran, N., Schmidt, D., Zhong, Y. and Coady, P. [2020]. Noisier2Noise: Learning to denoise from unpaired noisy data. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Proceedings, pp. 12061-12069. https://doi.org/10.1109/CVPR42600.2020.01208.
Ralanarko, D., Wahyuadi, D., Nugroho, P., Rulandoko, W., Syafri, I., Abdurrokhim, A. and Nur, A. [2021]. Seismic expression of Paleogene Talangakar Formation - Asri & Sunda Basins, Java Sea, Indonesia. Berita Sedimentologi, 46, 21-43. https://doi.org/10.51835/ bsed.2020.46.1.58.
Ronneberger, O., Fischer, P. and Brox, T. [2015]. U-Net: Convolutional networks for biomedical image segmentation’, in Medical Image Computing and Computer-Assisted Intervention – MICCAI 2015. 18th International Conference, Munich, Germany, 5-9 October 2015, Proceedings, Part III, pp. 234-241.
Sigalingging, A.S. et al. [2024]. Application of deep learning for low-frequency extrapolation to marine seismic data in the Sadewa Field, Kutei Basin, Indonesia. In Bezzeghoud, M. et al. (eds), Recent Research on Geotechnical Engineering, Remote Sensing, Geophysics and Earthquake Seismology. MedGU 2022. Advances in Science, Technology & Innovation, Springer, Cham, viewed, https://doi. org/10.1007/978-3-031-48715-6_51.
Sun, H. and Demanet, L. [2020]. Extrapolated full-waveform inversion with deep learning. Geophysics, 85(3), R275–R288. https://doi. org/10.1190/geo2019-0195.1.
Tarantola, A. [1986]. A strategy for nonlinear elastic inversion of seismic reflection data. Geophysics, 51(10), 1893–1903. https://doi. org/10.1190/1.1442046.
Winardhi, S., Sigalingging, A.S., Triyoso, W., Sukmono, S., Dinanto, E., Hendriyana, A., Wardaya, P.D., Septama, E. and Raguwanti, R. [2024]. Deep learning-based low-frequency extrapolation: Its implication in 2D full waveform imaging for marine seismic data in the Sadewa Field, Indonesia. First Break, 42(7), 65–72. https://doi. org/10.3997/1365-2397.fb2024059.
Witte, P.A., Louboutin, M., Kukreja, N., Luporini, F., Lange, M., Gorman, G.J. and Herrmann, F.J. [2019]. A large-scale framework for symbolic implementations of seismic inversion algorithms in Julia. Geophysics, 84(3), F57–F71. https://doi.org/10.1190/geo2018-0174.1.
Zhang, Q., Zhou, H., Li, Q., Chen, H. and Wang., J. [2016]. Robust source-independent elastic full waveform inversion in the time domain. Geophysics, 81, R29-R44, doi:10.1190/GEO2015-073.1
Zhang, Q., Mao, W., Zhou, H., Zhang, H. and Chen, Y. [2018]. Hybrid-domain simultaneous-source full waveform inversion without crosstalk noise. Geophysical Journal International, 215(3), 1659-1681. https://doi.org/10.1093/gji/ggy366.
To serve the interests of our members and the wider multidisciplinary geoscience and engineering community, EAGE publishes a range of books and scientific journals in-house. Our extensive, professional marketing network is used to ensure publications get the attention they deserve.
EAGE is continually seeking submissions for both book publishing and articles for our journals. A dedicated and qualified publishing team is available to support your publication at EAGE.
David Little1* presents datasets matched and merged into a single interpretable volume, called 2D cubed, that is used to gain a regional overview enabling structural framework and consistent regional interpretation calibrated to 86 exploration and appraisal wells.
Indonesia is emerging as a key region for frontier and Infrastructure-Led Exploration (ILX). To support exploration efforts, TGS has completed a comprehensive regional study that integrates publicly available data with the latest seismic and well information. This integration enables the creation of high-resolution, calibrated facies maps across key petroleum system intervals, helping to identify play potential and assess prospectivity in underexplored parts of the basin. The results are delivered through an interactive online platform known as the Facies Map Browser (FMB).
Indonesia has a well organised national data repository where vintage data can be accessed. The country has a large volume of 2D data but this varies widely in vintage and quality. TGS matched and merged the available datasets into a single interpretable volume. We called this transformation a 2Dcubed and use this 3D volume to gain a regional overview. This was the study baseline that enabled structural framework and consistent regional interpretation that was calibrated to 86 exploration and appraisal wells.
Detailed well interpretation, including sequence, chronostratigraphic, and lithostratigraphic picks, as well as gross depositional environment (GDE) and facies, are part of the study. With inputs derived from cutting descriptions, core data, well correlations, seismic facies interpretations, and biostratigraphic reports. Specific attention was paid to the lithological interpretation, particularly the variation and gradation within the dominant carbonates of the region, feeding into the wider depositional model, building upon the hydrocarbon story that already exists.
Maps derived from the latest seismic data enable the generation of regional GDE maps that delineate the geological evolution and sedimentary history of the region, providing insights into the petroleum system at key intervals, but also have multiple applications for Carbon Capture Storage (CCS) site screening, through overburden and potential aquifer distribution mapping.
The FMB helps users to high-grade areas for prospectivity and also includes a block in the 2024 bid round, with the Kojo block in the Makassar Strait that was awarded to Bumi Armada and will likely include areas in future rounds. New seismic data continues to be acquired throughout this study area and is being used as an interpretation QC. This work suggests that there is
still plenty of potential for conventional exploration as well as for CCS applications within this region.
Exploration in the East Java Sea began in the 19th century, with the first commercial oil production recorded around 1887. While early activity laid the groundwork, substantial progress has been made in recent decades, particularly in exploring Cenozoic petroleum systems. These efforts primarily target structural traps commonly linked to carbonate build-ups.
This FMB study harnesses the latest seismic data alongside a detailed sequence stratigraphic analysis of 86 exploration and appraisal wells. By integrating these datasets, it enhances our understanding of the basin’s tectonic and depositional history.
1 TGS
* Corresponding author, E-mail: David.Little@tgs.com
DOI: 10.3997/1365-2397.fb2025072
This refined perspective supports more accurate play fairway evaluations, identifies potential future exploration hotspots, and provides critical input for early stage carbon capture and storage (CCS) screening workflows.
The East Java Sea is located between the islands of Java, Kalimantan, and Sulawesi is part of the tectonically dynamic Southeast Asian region. It sits on the eastern edge of the Sunda Shelf, shaped by the interactions between the Eurasian, Indo-Australian, and Pacific tectonic plates. These plate movements have influenced the basin’s evolution from early rifting through to complex deformation and subsidence phases.
The basin’s geological history is marked by multiple tectonic episodes. Initial rifting during the Late Jurassic to Early Cretaceous periods, laying the structural framework of the region. Major basin development occurred during the Paleogene (Eocene–Oligocene) as back-arc extension formed deep grabens and half-grabens. This was followed by Miocene compressional events, which caused inversion of earlier extensional structures, forming numerous fault-bounded traps that are favourable for hydrocarbon accumulation.
This study can be broken out into 2 separate phases that run in tandem across the project cycle that ultimately ends with a fully integrated product.
2Dcubed is a method of guiding interpolation of 2D seismic data along 3D geological horizons, generated from 2D dip fields and layer models. This allows distances of kilometres to be bridged, something that would otherwise not be possible with conventional interpolation techniques.
The input data must be a grid of intersecting 2D lines. These are subjected to a post-stack demigration. This ultimately results in the creation of a volume that can have a full 3D post-stack migration applied to it, correctly positioning events that would not be possible to position in a 2D sense.
The input data must be matched as closely as possible in terms of amplitude, time and spectral character (Whiteside et al., 2013). This is a two-pass process. First, global adjustments are made. These can be line or survey specific. Secondly, windowed adjustments to time and overall amplitude are made at each intersection to make the individual horizons tie. This demigrated
input data must be matched so that the geological horizons can be tracked between 2D lines, forming a continuous network of horizons that can be used as the framework for the 3D layer model.
The 3D geological time model is used to guide the seismic interpolation. 2D input amplitudes from around the 3D output point (x,y,t) are drawn together along the 3D layers to form a gather. This gather is then processed to form the output sample (Whiteside et al., 2013).
3D post-stack migration is the final step in 2Dcubed generation. A 3D velocity field is generated from the available 2D migration velocities. The 3D migration gives the 2Dcubed a distinct advantage over other interpretation tools in that, although other geological horizon generators may be able to estimate the position of horizons in 3D space, they cannot correct for out of plane events present in 2D data.
With a final 2Dcubed output delivering a regionally consistent volume a seismic interpretation was applied for the whole project area in order to capture all the mega-sequence intervals relevant to the petroleum systems, with a primary focus on the Late Miocene to Early Eocene.
This interpretation formed the structural framework and key control for tying project wells consistently to the regional volume.
As part of the FMB project, a subset of strategically selected wells underwent a comprehensive subsurface analysis. (Figure 1) This included detailed sequence stratigraphy, chronostratigraphic and lithostratigraphic tops picking, and the characterisation of gross depositional environments (GDEs) and associated facies.
Lithological modelling was completed using an integrated dataset comprising cuttings descriptions, core analysis, data derived from biostratigraphic reports, and petrophysical log responses. This multi-disciplinary approach enabled robust lithological classification and highlighted heterogeneities within the basin.
Special emphasis was placed on interpreting the carbonate lithologies, which dominate much of the region’s stratigraphic record. Carbonate classification followed the foundational scheme of Dunham (1962), later refined by Embry and Klovan (1971), where carbonates are categorised according to depositional texture — specifically the balance between mud-supported and grain-supported structures — and their inferred depositional environment. This framework allowed for a more nuanced interpretation of carbonate facies and their spatial distribution within the basin.
Lithological interpretation at a sequence stratigraphic level of detail (3rd and 4th Order) permitted the identification of GDE specific facies, which, when combined with 2Dcubed seismic interpretation, enabled the mapping of these facies away from well control into more underexplored frontier areas of the basin.
Key stratigraphic intervals from the Late Miocene to the Early Eocene were prioritised for mapping, as they represent the more prolific petroleum plays within the region.
The objective was to delineate potential source rocks, seal intervals, and reservoir-quality units, thereby enhancing the regional hydrocarbon prospectivity model and refining the playbased exploration framework.
With a fully integrated dataset, merging high-quality regional 2Dcubed seismic with consistent, sequence-based stratigraphic interpretations across 86 exploration and appraisal wells, the FMB reveals valuable insights into the basin’s geological evolution.
For example, sediment input pathways during the Bartonian to Ypresian interval were interpreted, shedding light on clastic dispersal mechanisms and carbonate platform development. Additionally, the FMB helps to refine the timing and nature of major carbonate generation phases, offering a clearer picture of the region’s stratigraphic architecture and petroleum system evolution.
A subset of the facies maps created for this study is presented. The earliest map (Figure 5A) comprises of alluvial, lacustrine and fluvio-deltaic sediments that were deposited during the Early
Eocene rifting. Rapid sedimentation rates and increasing water depths within the half-graben and graben depocentres indicated rapid extension during or immediately prior to deposition. Observed tilted fault blocks formed structural highs and led to the development of carbonates in the Mid-Late Eocene as marine conditions prevailed (Figure 5B and 5C).
Platform and pinnacle reef type carbonate development continued into the Oligocene, especially in the east of the basin as bathymetry continued to decrease. The deposition of deep marine sediments came after a hiatus marking the end of the rift phase and high subsidence (Figure 5D).
Understanding the depositional regime as well as the timing of events provides a lot of the parameters required for prospecting in a region.
The Kojo block shown in (Figure 1) was one of six blocks that were awarded in 2024. It is centrally located over the study area, and tying into a regionally consistent interpretation allows observations to be made on the big picture (Figure 6 and Figure 7).
Interpretation at the mega-sequence level offers a broad regional perspective and serves as a key input for understanding the structural controls within the project area. Figure 6 is an example northwest-southeast seismic line extending from the Paternoster High, through the South Makassar Basin, and into the Salayar Basin to the east.
In the geo-seismic section (Figure 6), platform carbonate plays are evident at the base of the Miocene, highlighting
potential reservoir targets. Near the shelf edge, during the Eocene to Oligocene periods, there is evidence of combination traps involving both stratigraphic and structural elements. As the section progresses eastward, carbonate buildup plays become more prominent over structural highs. Additionally, the presence of possible Eocene-aged coaly shale packages as possible source rock could be preserved within the depocentres, suggesting favourable conditions for hydrocarbon generation and entrapment.
Indonesia is positioning itself as a regional leader in carbon capture and storage (CCS), leveraging its energy expertise, skilled workforce, and infrastructure to support emissions reduction across the Asia Pacific.
In a significant policy shift, the Indonesian government recently issued a presidential regulation enabling CCS operators to allocate a portion of their storage capacity for CO2 from neighbouring countries. Building on this momentum, Indonesia and Singapore signed a Letter of Intent (LOI) earlier this year to explore cross-border CCS collaboration.
ExxonMobil is partnering with Indonesia’s state-owned energy company, Pertamina, to evaluate the development of a major CCS hub. The proposed site, located beneath the Java Sea, is estimated to hold up to three gigatons of CO2, potentially making it the largest storage facility in Southeast Asia. Through strategic initiatives like these, Indonesia is actively advancing its role in enabling a lower-carbon future for the region and establishing itself as a key hub for both domestic and regional CCS efforts.
Building on the outputs from this FMB study, we begin to explore how these results can inform early-stage regional carbon capture and storage (CCS) screening — not only within the East Java Basin, but in any comparable sedimentary basin.
The foundational datasets and interpretations generated through this FMB work offer broad applicability for identifying and evaluating potential CCS sites.
Effective CCS site screening relies on several key geological parameters. Central among them is the ability to clearly define potential aquifers, understand their spatial extent, and character-
ise variability in thickness. Equally important is the quantification of total overburden, as well as an assessment of internal geomorphological features and the integrity of sealing intervals that can act as caprocks.
Facies mapping within the study has enabled the delineation of potential aquifer zones. Lithological variations across these aquifers can be visualised using lithology pie charts plotted spatially on maps, helping to assess heterogeneity and connectivity at a regional scale (Figure 8). Seismic interpretation further contributes by identifying zones where aquifers are buried beneath a minimum of 800 m of overburden — considered a necessary threshold for long-term CO2 containment.
Additionally, well data provides direct observations confirming aquifer intervals exceeding 50 m in thickness, along with insight into both the lithology and depositional facies of the overburden and potential sealing units.
Many of these essential inputs — facies distributions, thickness maps (Figure 9), overburden characteristics, and lithological profile — are already available as part of this FMB study. When integrated, these layers of information can be synthesised into a preliminary common risk map, supporting high- and low-grading of areas based on CCS suitability.
While this represents a strong foundation for early-stage screening, further refinement can be achieved through more advanced workflows. These include containment analysis, petrophysical and rock physics evaluation of both aquifer and seal units, detailed volumetric assessments of aquifers, and expanded sedimentological studies to better understand reservoir quality and caprock effectiveness.
Nonetheless, as a first-pass tool for regional CCS screening, this study provides a comprehensive and scalable framework for identifying viable storage sites, laying the groundwork for more detailed site-specific assessments in the future.
This integrated study of the East Java region has advanced the understanding of the area’s geological evolution and sedimentary history. By combining stratigraphic interpretation, structural analysis, and seismic integration, the study offers valuable insights into the development of the petroleum system across key stratigraphic intervals.
Figure 9 Cenozoic Isochron (seabed down to top Cretaceous) grid for the study highlighting potential depocentres for CCS aquifers and extended continuous 2D cubed volume into the Kutei Basin.
Combining data and interpretation into cohesive, user-friendly interpretation tools, such as FMB, allows rapid analysis and geological prediction. This integrated approach accelerates the identification of high-quality reservoirs and source rock potential across large underexplored regions, significantly streamlining exploration decisions and collaborative analysis.
Importantly, the results of this study align with a resurgence of industry interest, as evidenced by recent a bid round block within the study area. This renewed attention underscores the exploration potential and relevance of the East Java Basin in the current energy landscape.
Beyond hydrocarbon exploration, the FMB provides critical datasets for evaluating the region’s suitability for carbon capture and storage (CCS). Detailed mapping of overburden characteristics and potential aquifer distributions forms a solid foundation for initial CCS screening and feasibility assessments.
Looking ahead, the value of this work can be further enhanced by incorporating additional and more recent seismic datasets, such as the 6552 km West Sulawesi 2D survey and the South Makassar 3D volume (Figure 1).
In summary, the strategic use of advanced subsurface data is fundamental to the continued success of exploration efforts in the Asia Pacific region. The integration of modern seismic acquisition, advanced data processing techniques, and comprehensive interpretation platforms greatly improves geological insight and exploration outcomes, underscoring the critical importance of data-driven methodologies in discovering new oil and gas reserves.
Dunham, R.J. [1962]. Classification of carbonate rocks according to depositional texture. In: Classification of Carbonate Rocks (Ed. W.E. Ham). Am. Assoc. Pet. Geol. Mem., 1, 108-121.
Embry, A.F. and Klovan, J.E. [1971]. A late Devonian reef tract on northeastern Banks Island, N.W.T. Bulletin of Canadian Petroleum Geology, 19(4), 730–781.
Whiteside, W., Wang, B., Bondeson, H. and Li, Z. [2013]. Enhanced 3D Imaging from 2D Seismic Data and its Application to Surveys in the North Sea. 75th EAGE Annual Conference & Exhibition incorporating SPE EUROPEC, Extended Abstracts.
September 2025
9-11 Sep Second EAGE Conference and Exhibition on Guyana-Suriname Basin www.eage.org
14-18 Sep Seventh International Conference on Fault and Top Seals www.eage.org
15-17 Sep Fifth EAGE Workshop on Assessment of Landslide Hazards and Impact on Communities www.eage.org
15-18 Sep AOW Energy 2025 www.aowenergy.com
16-18 Sep The Middle East Oil, Gas and Geosciences Show (MEOS GEO) www.meos-geo.com
17-18 Sep First EAGE Workshop on Energy Transition in Latin America’s Southern Cone www.eage.org
21-24 Sep Sixth EAGE Borehole Geology Workshop www.eage.org
29 Sep1 Oct Second AAPG/ EAGE Mediterranean and North African Conference (MEDiNA) medinace.aapg.org
29 Sep1 Oct Eighth EAGE Borehole Geophysics Workshop www.eage.org
October 2025
6-8 Oct Second EAGE Data Processing Workshop www.eage.org
6-8 Oct Empowering the Energy Shift - The Role of HPC in Sustainable Innovation: Ninth EAGE High Performance Computing Workshop www.eage.org
6-9 Oct GEOTERRACE-2025: International Conference of Young Professionals www.eage.org
14-16 Oct First EAGE Conference on the Future of Mineral Exploration: Challenges and Opportunities www.eage.org
15-16 Oct 3 rd EAGE/SUT Workshop on Integrated Site Characterization for Offshore Renewable Energy www.eage.org
21-22 Oct First EAGE Workshop on Geophysical Techniques for Monitoring CO2 Storage www.eage.org
21-23 Oct First EAGE/ AAPG/ SEG Carbon Capture Utilisation and Storage Workshop (CCUS) www.eage.org
27-31 Oct 6th EAGE Global Energy Transition Conference & Exhibition (GET 2025) - EAGE Carbon Capture and Storage Conference - EAGE Geothermal Energy Conference - EAGE Hydrogen and Energy Storage Conference - EAGE Offshore Wind Energy Conference www.eageget.org
28-30 Oct 2 nd Edition AAPG/EAGE Petroleum Systems of the Middle East GTW www.eage.org
30-31 Oct GEOSCIENCE International Symposium www.geosymposium.org
November 2025
2-5 Nov EAGE/AAPG Workshop on Tectonostratigraphy of the Arabian Plate: Structural Evolution of the Arabian Basins www.eage.org
10-11 Nov EAGE Workshop on Enhancing Subsurface Practices using AI/ML www.eage.org
10-12 Nov Seventh EAGE Rock Physics Workshop www.eage.org
12-14 Nov First EAGE Workshop on Surface Logging www.eage.org
12-14 Nov Third EAGE Workshop on Geothermal Energy in Latin America www.eage.org
12-14 Nov EAGE/Aqua Foundation Third Indian Near Surface Geophysics Conference & Exhibition www.eage.org
Santiago de Chile Chile
Melbourne Australia
Toronto Canada
Al Khobar Saudi Arabia
Rotterdam The Netherlands
Kuwait City Kuwait
Bucharest Romania
Riyadh Saudi Arabia
Perth Australia
Cape Town South Africa
Paris France
Guanacaste Costa Rica
New Delhi India
500+ ATTENDEES
100+ TECHNICAL PRESENTATIONS
80+ STRATEGIC SPEAKERS
20+ EXHIBITORS