Selected work

Representative Papers

Statistical Tapers for Correlation-Based Localization in Ensemble Data Assimilation

May 28, 2026

This study addresses the limitations of conventional distance-based localization in subsurface reservoir data assimilation, which often introduces spurious updates and excessively damps ensemble variance by neglecting flow dynamics, nonlinear observation operators, or prior structural information. The authors reformulate localization as a shrinkage problem in correlation space and, for the first time, incorporate statistical reliability into taper function design. They propose three novel approaches: a generalized power-law taper, a Bayesian spike-and-slab logistic taper, and a discrepancy-based taper derived from Morozov’s discrepancy principle. Numerical experiments demonstrate that the proposed correlation-based localization effectively suppresses spurious correlations while preserving genuine parameter–data relationships. Among the methods, the logistic taper best maintains posterior ensemble variance and consistently outperforms traditional distance-based localization, particularly in scenarios where spatial distance is ineffective or misleading.

0 citationsRead paper

Deploying Self-Supervised Learning for Real Seismic Data Denoising

May 11, 2026

This study addresses the challenge of denoising real seismic data, where supervised methods are hindered by the absence of clean reference signals. Building upon the Noisy-as-Clean (NaC) self-supervised learning framework, the work abandons conventional synthetic Gaussian noise and instead injects controllable real seismic noise during training. A systematic experimental design with unified hyperparameters is employed to rigorously compare self-supervised and supervised strategies. The results demonstrate that synthetic Gaussian noise is ill-suited for real seismic denoising. In contrast, the proposed method effectively fine-tunes models without requiring clean labels, exhibiting high efficiency, strong generalization, and architecture independence across multiple network backbones, thereby significantly improving denoising performance.

0 citationsRead paper

Hybrid Context-Fusion Attention (CFA) U-Net and Clustering for Robust Seismic Horizon Interpretation

Nov 28, 2025

To address inaccurate segmentation of complex geological structures and poor continuity under sparse annotations in seismic horizon interpretation, this paper proposes a Context-Fusion Attention (CFA) U-Net. The model innovatively couples Sobel-based geometric edge priors with spatial attention mechanisms and integrates DBSCAN density clustering to optimize horizon topological connectivity. Additionally, a multi-directional prediction fusion strategy is employed to enhance robustness in fault-proximal zones and folded regions. Evaluated on the Mexilhao field dataset, the method achieves an IoU of 0.881 and a mean absolute error of 2.49 ms. On the North Sea F3 block, using only 5% sparse annotations, it attains 97.6% horizon coverage—significantly outperforming baseline models including U-Net++ and Attention U-Net. The proposed approach establishes a new state-of-the-art in automated seismic horizon interpretation.

0 citationsRead paper

Deep mineralogical segmentation of thin section images based on QEMSCAN maps

May 22, 2025

Manual mineralogical interpretation of carbonate thin sections suffers from subjectivity, while automated techniques like QEMSCAN are costly and time-consuming. Method: This paper proposes a low-cost, high-efficiency semantic segmentation framework: (1) leveraging QEMSCAN-derived mineral distribution maps as weak supervision; (2) training a multimodal U-Net that jointly processes orthogonal and plane-polarized light images; (3) incorporating cross-resolution image registration to accommodate real-world data variability; and (4) adopting a cross-facies generalization training strategy. Contribution/Results: The model achieves accurate segmentation of six classes—Calcite, Dolomite, Mg-Clay, Quartz, Pores, and Others. Validation shows R² scores of 0.97 on seen facies and 0.88 on unseen facies for mineral abundance estimation. Segmentations exhibit sharp boundaries, high texture fidelity, and reliable pore and mineral volumetric quantification, significantly advancing the practical deployment of intelligent rock image interpretation.

0 citationsRead paper

Entropy-based measure of rock sample heterogeneity derived from micro-CT images

Feb 01, 2025

Traditional methods for quantifying rock heterogeneity are time-consuming, costly, and highly subjective. To address these limitations, this paper proposes an end-to-end, segmentation-free entropy-based quantification method that directly computes grayscale and gradient entropy from sliding volumetric sub-blocks of micro-CT images, capturing local structural uncertainty as an objective, automated heterogeneity metric. The method adaptively accommodates diverse rock sample characteristics, eliminating bias introduced by manual segmentation. Validated on 4,935 micro-CT images from Brazilian reservoir core samples, a single entropy feature achieves statistically significant differentiation between homogeneous and heterogeneous samples across all four expert annotations (p < 0.01), outperforming conventional texture features and expert visual interpretation in inter-rater consistency. The approach is highly reproducible, low-cost, and provides a standardized, generalizable framework for heterogeneity quantification—enhancing rock physics modeling and reservoir characterization.

0 citationsRead paper

Recent publications

Latest Papers

Statistical Tapers for Correlation-Based Localization in Ensemble Data Assimilation

May 28, 2026

This study addresses the limitations of conventional distance-based localization in subsurface reservoir data assimilation, which often introduces spurious updates and excessively damps ensemble variance by neglecting flow dynamics, nonlinear observation operators, or prior structural information. The authors reformulate localization as a shrinkage problem in correlation space and, for the first time, incorporate statistical reliability into taper function design. They propose three novel approaches: a generalized power-law taper, a Bayesian spike-and-slab logistic taper, and a discrepancy-based taper derived from Morozov’s discrepancy principle. Numerical experiments demonstrate that the proposed correlation-based localization effectively suppresses spurious correlations while preserving genuine parameter–data relationships. Among the methods, the logistic taper best maintains posterior ensemble variance and consistently outperforms traditional distance-based localization, particularly in scenarios where spatial distance is ineffective or misleading.

0 citationsRead paper

Deploying Self-Supervised Learning for Real Seismic Data Denoising

May 11, 2026

This study addresses the challenge of denoising real seismic data, where supervised methods are hindered by the absence of clean reference signals. Building upon the Noisy-as-Clean (NaC) self-supervised learning framework, the work abandons conventional synthetic Gaussian noise and instead injects controllable real seismic noise during training. A systematic experimental design with unified hyperparameters is employed to rigorously compare self-supervised and supervised strategies. The results demonstrate that synthetic Gaussian noise is ill-suited for real seismic denoising. In contrast, the proposed method effectively fine-tunes models without requiring clean labels, exhibiting high efficiency, strong generalization, and architecture independence across multiple network backbones, thereby significantly improving denoising performance.

0 citationsRead paper

Hybrid Context-Fusion Attention (CFA) U-Net and Clustering for Robust Seismic Horizon Interpretation

Nov 28, 2025

To address inaccurate segmentation of complex geological structures and poor continuity under sparse annotations in seismic horizon interpretation, this paper proposes a Context-Fusion Attention (CFA) U-Net. The model innovatively couples Sobel-based geometric edge priors with spatial attention mechanisms and integrates DBSCAN density clustering to optimize horizon topological connectivity. Additionally, a multi-directional prediction fusion strategy is employed to enhance robustness in fault-proximal zones and folded regions. Evaluated on the Mexilhao field dataset, the method achieves an IoU of 0.881 and a mean absolute error of 2.49 ms. On the North Sea F3 block, using only 5% sparse annotations, it attains 97.6% horizon coverage—significantly outperforming baseline models including U-Net++ and Attention U-Net. The proposed approach establishes a new state-of-the-art in automated seismic horizon interpretation.

0 citationsRead paper

Deep mineralogical segmentation of thin section images based on QEMSCAN maps

May 22, 2025

Manual mineralogical interpretation of carbonate thin sections suffers from subjectivity, while automated techniques like QEMSCAN are costly and time-consuming. Method: This paper proposes a low-cost, high-efficiency semantic segmentation framework: (1) leveraging QEMSCAN-derived mineral distribution maps as weak supervision; (2) training a multimodal U-Net that jointly processes orthogonal and plane-polarized light images; (3) incorporating cross-resolution image registration to accommodate real-world data variability; and (4) adopting a cross-facies generalization training strategy. Contribution/Results: The model achieves accurate segmentation of six classes—Calcite, Dolomite, Mg-Clay, Quartz, Pores, and Others. Validation shows R² scores of 0.97 on seen facies and 0.88 on unseen facies for mineral abundance estimation. Segmentations exhibit sharp boundaries, high texture fidelity, and reliable pore and mineral volumetric quantification, significantly advancing the practical deployment of intelligent rock image interpretation.

0 citationsRead paper

Entropy-based measure of rock sample heterogeneity derived from micro-CT images

Feb 01, 2025

Traditional methods for quantifying rock heterogeneity are time-consuming, costly, and highly subjective. To address these limitations, this paper proposes an end-to-end, segmentation-free entropy-based quantification method that directly computes grayscale and gradient entropy from sliding volumetric sub-blocks of micro-CT images, capturing local structural uncertainty as an objective, automated heterogeneity metric. The method adaptively accommodates diverse rock sample characteristics, eliminating bias introduced by manual segmentation. Validated on 4,935 micro-CT images from Brazilian reservoir core samples, a single entropy feature achieves statistically significant differentiation between homogeneous and heterogeneous samples across all four expert annotations (p < 0.01), outperforming conventional texture features and expert visual interpretation in inter-rater consistency. The approach is highly reproducible, low-cost, and provides a standardized, generalizable framework for heterogeneity quantification—enhancing rock physics modeling and reservoir characterization.

0 citationsRead paper