Synthetic seismic data generation for deep learning segmentation of clinoforms: integrating process‐based forward stratigraphic modelling and ray‐based seismic modelling
Author(s)
AlGharbi, Waleed
Bell, Rebecca E
John, Cédric
Type
Journal Article
Abstract
Clinoform identification plays a major role in sequence stratigraphy and applied geosciences, offering key insights into sediment fluxes, basin evolution, and natural resource exploration. However, manual interpretation of clinoform geometries in seismic data is time-consuming and labour-intensive. This study proposes a deep learning framework to automate clinoform detection and segmentation. One of the greatest challenges of seismic interpretation with deep learning is the scarcity of labelled training data. To address this, we generated 20,000 synthetic seismic images using process-based forward stratigraphic modelling followed by ray-based seismic modelling, ensuring both geological realism and variability of data. The generated synthetic seismic images were split into training (70%), validation (15%), and testing (15%) sets. The clinoform binary masks were labelled using the top and base layers from the forward stratigraphic models. Then, a YOLOv8m model was trained on this dataset, achieving 86% precision, 82% recall, and 83% Dice coefficient on segmenting clinoforms in the synthetic testing dataset. Qualitative validation on real seismic images confirmed accurate bounding box detection and segmentation of clinoform bodies. These results highlight the potential of deep learning to accelerate seismic interpretation, particularly during early scanning phases, where rapid clinoform analysis supports stratigraphic assessment and reservoir characterisation. The proposed workflow bridges the synthetic-to-real domain gap in seismic data, offering a scalable approach to train algorithms for advancing subsurface analysis.
Date Issued
2026-07-01
Date Acceptance
2026-06-25
Citation
Basin Research, 2026, 38 (4)
ISSN
0950-091X
Publisher
Wiley
Journal / Book Title
Basin Research
Volume
38
Issue
4
Copyright Statement
© 2026 The Author(s). Basin Research published by International Association of Sedimentologists and European Association of Geoscientists and Engineers and John Wiley & Sons Ltd. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.
License URL
Publication Status
Published
Article Number
e70123
Date Publish Online
2026-08-08