Least-squares reverse time migration (LSRTM) is an advanced seismic imaging technique that reconstructs subsurface models by minimizing the residuals between simulated and observed data. Mathematically, the LSRTM inversion of the sub-surface reflectivity is a large-scale, highly ill-posed sparse inverse problem, where conventional inversion methods typically lead to poor imaging quality. In this study, we propose a regularized LSRTM method based on the flexible Krylov subspace inversion framework. Through the strategy of the Krylov subspace projection, a basis set for the projection solution is generated, and then the inversion of a large ill-posed problem is expressed as the small matrix optimization problem. With flexible preconditioning, the proposed method could solve the sparse regularization LSRTM, like with the Tikhonov regularization style. Sparse penalization solution is implemented by decomposing it into a set of Tikhonov penalization problems with iterative reweighted norm, and then the flexible Golub–Kahan process is employed to solve the regularization problem in a low-dimensional subspace, thereby finally obtaining a sparse projection solution. Numerical tests on the Valley model and the Salt model validate that the LSRTM based on Krylov subspace method can effectively address the sparse inversion problem of subsurface reflectivity and produce higher-quality imaging results.
10 April 2025 in IEEE Transactions on Geoscience and Remote Sensing
This study addresses the challenges in seismic interferometry (SI), particularly the deghosting of virtual-shot gathers affected by nonphysical reflections in marine seismic exploration. SI’s effectiveness largely depends on the distribution of seismic sources, which in field situations often leads to complex wavefield interactions and nonphysical reflections, adversely impacting data processing. Traditional methods are limited with these nonphysical reflections, and similarity-based approaches have shown potential in overcoming this challenge. In this paper, our study introduces an innovative deghosting framework based on Regularization by Denoising (RED). This approach effectively mitigates the influence of nonphysical reflections and enhances the accuracy of deghosting. We propose a simultaneous multi-shot deghosting framework that integrates RED, focusing on optimizing data similarity and computational efficiency. This is complemented by a novel denoising engine based on an enhanced BM3D method, specifically tailored for virtual-shot gathers. The efficacy of our framework is demonstrated through synthetic and field data, showcasing significant improvements in seismic data processing and offering a new paradigm for the application of seismic interferometry. This research contributes not only to the advancement of deghosting techniques but also to broadening the scope of seismic exploration.
10 August 2026 in Geophysical Journal International
SUMMARY: Full waveform inversion (FWI) is a powerful tool for constructing high-resolution subsurface models but remains fundamentally ill-posed due to sparse and noisy data, modelling errors, and the severe nonlinearity of the forward modelling. Because this inherent nonlinearity creates complex uncertainty structures that traditional deterministic optimization cannot easily resolve. To fully quantify this inversion uncertainty and explore all plausible solutions, Bayesian FWI seeks to sample directly from the posterior probability density function using Markov chain Monte Carlo (MCMC) algorithms. However, traditional MCMC methods suffer from inefficient exploration (poor mixing) in high-dimensional model spaces. In this study, we adopt an efficient sampler based on underdamped Langevin diffusion (ULD). The stochastic differential equation governing ULD is approximated using the shifted ordinary differential equation (ODE) method, which converts the stochastic dynamics into a deterministic ODE that is easier to solve numerically. Building on this formulation, we introduce the SORT (Shifted ODE with Runge–Kutta Three) sampling method. The SORT method employs a third-order Runge–Kutta scheme to discretize the ODE, which effectively reduces discretization errors and enables highly efficient sampling. We validate the performance of SORT through synthetic 2-D acoustic FWI experiments in both low- and high-dimensional model spaces. The numerical results show that SORT efficiently explores high-dimensional posterior distributions and produces stable and reliable mean and uncertainty estimates for complex subsurface structures.
Accurate subsurface velocity modeling is of fundamental scientific and practical significance for seismic data processing and interpretation. However, conventional depth-domain methods still face limitations in physical consistency and inversion accuracy. To overcome these challenges, this study proposes a deep learning-based seismic velocity modeling approach in the time domain. The method establishes an end-to-end mapping between seismic records and velocity models directly in the time domain, reducing the nonlinear complexity of mapping time-domain data to depth-domain models and improving prediction stability and accuracy. Synthetic aquifer velocity models were constructed from representative stratigraphic features, and multi-shot seismic records were generated through forward modeling. A U-Net network was employed, taking multi-shot seismic records as input and time-domain velocity fields as output, with training guided by a mean squared error (MSE) loss function. Experimental results show that the proposed strategy outperforms conventional depth-domain approaches in aquifer structure identification, velocity recovery, and interlayer contrast depiction. Quantitatively, significant improvements in MSE, peak signal-to-noise ratio, and structural similarity index indicate higher reconstruction reliability. Overall, the results confirm the effectiveness and potential of the proposed time-domain framework for aquifer velocity inversion and its promise for intelligent seismic velocity modeling.
The Apex-Shifted Hyperbolic Radon Transform (ASHRT) is a variant of the Radon Transform. In the field of seismic exploration, it can be applied to simultaneous source separation, diffraction- and reflection-wave separation, seismic data reconstruction, among other purposes. This paper primarily investigates the application of ASHRT in the separation of diffraction and reflection waves. Detailed exploration of complex structures using diffraction wave imaging has become a new trend, thereby necessitating the separation of diffraction wave fields. The conventional ASHRT based on the Stolt operator, due to its weak sparsity, increasingly struggles to meet current separation requirements. Compared to conventional ASHRT, the Stolt-based ASHRT enables fast, efficient computation; however, the Stolt operator exhibits relatively weaker sparseness and fidelity. To address this issue, replacing the Stolt operator with the PS operator for performing ASHRT allows the transform to achieve both high sparseness and high fidelity simultaneously. In this study, synthetic data were used to investigate the advantages of the PS operator over the Stolt operator. Furthermore, both operators were applied to separate diffraction and reflection waves in marine seismic data and land seismic data, respectively. The research demonstrates that, in the separation of diffraction and reflection waves using the ASHRT method, the PS operator provides significant advantages over the Stolt operator in terms of both sparseness and fidelity.
9 February 2026 in IEEE Transactions on Geoscience and Remote Sensing
Deep unfolding networks (DUNs), which offer strong interpretability, have gradually attracted increasing attention in seismic data reconstruction. Most existing DUN approaches are limited to unfolding in the time-space (t–x) domain, making it difficult to fully exploit the multiscale features of seismic signals, thereby decreasing reconstruction performance and often leading to the loss of fine details. To overcome the above limitation, a dual-domain optimization framework is proposed, in which the traditional iterative process is replaced by neural networks, yielding a dual-domain deep unfolding network (D3U-Net) based on implicit regularization. The proposed method combines prior information from both the t–x domain and the wavelet domain, guiding the network to simultaneously capture spatial structural information and frequency characteristics at each stage, thus achieving cooperative reconstruction in both domains and enhancing detail recovery. Meanwhile, an implicit regularization mechanism is introduced, where a learnable neural network substitutes the conventional proximal mapping operator and does not require manual specification of the weights of the regularization term, simplifying the tuning process and endowing the network with greater flexibility and adaptability. Experiments conducted on synthetic data, the BP2004 dataset, and field marine seismic data demonstrate that D3U-Net achieves superior reconstruction accuracy and stability under both regular and random seismic-trace missing scenarios compared with current state-of-the-art methods, confirming overall advantages in reconstruction fidelity, generalization capability, and interpretability.