https://doi.org/10.1140/epjs/s11734-026-02301-2
Regular Article
Assisted seismic wavefield interpretation with recurrent neural networks: the role of statistical data conditioning
1
Measuring chemical-analytical laboratory, M.S. Institute of Geotechnical Mechanics of the National Academy of Sciences of Ukraine, Simferopolska St., 2a, 49005, Dnipro, Ukraine
2
Department of Geophysical Prospection, Dnipro University of Technology, D. Yavornytskyi Ave., 49000, Dnipro, Ukraine
a
This email address is being protected from spambots. You need JavaScript enabled to view it.
Received:
4
November
2025
Accepted:
27
March
2026
Published online:
6
April
2026
Abstract
The interpretation of deep seismic wavefields, critical for energy resource exploration, is increasingly utilizing artificial intelligence (AI) tools such as recurrent neural networks (RNNs). However, the reliability of these tools is fundamentally limited by signal distortions originating in the complex near-surface layer. This paper researches a context-aware Long Short-Term Memory (LSTM) architecture designed to improve horizon identification by incorporating spatial information. We validate its performance on geologically realistic synthetic datasets, demonstrating its ability to generate probabilistic data of potential horizons under ideal conditions. We highlight the model’s inherent vulnerability to signal disruptions common in real-world data due to its context-dependent nature. We present and validate a statistical technique for characterizing near-surface amplitude distortions by decomposing the seismic wavefield. We conclude that robust, physics-grounded data conditioning, such as the statistical correction demonstrated, is a critical prerequisite for alleviating artifacts and enabling the reliable application of AI-based methods in subsurface signal interpretation.
Copyright comment Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
L. Anisimova, O. Piskunov, and O. Tiapkin have contributed equally to this work.
© The Author(s), under exclusive licence to EDP Sciences, Springer-Verlag GmbH Germany, part of Springer Nature 2026
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
