Pith. sign in

REVIEW 5 major objections 6 minor 83 references

Pix2Geomodel: A Next-Generation Reservoir Geomodeling with Property-to-Property Translation

T0 review · 5 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The paper claims that Pix2Pix, applied for the first time to reservoir geomodeling, directly translates facies maps into porosity, permeability, and water-saturation maps of the Groningen gas field, with pixel accuracy up to 0.98.

desk verdict Leaky augmentation split likely inflates the reported accuracies; the application is plausible but the evaluation needs major rework. read the letter →

arxiv 2506.17747 v1 pith:S44OANLC submitted 2025-06-21 physics.geo-ph cs.CEcs.CVcs.LGcs.NE

Abdulrahman Al-Fakih , Ardiansyah Koeshidayatullah , Nabil A. Saraih , Tapan Mukerji , Rayan Kanfar , Abdulmohsen Alali , SanLinn I. Kaka This is my paper · ORCID
classification physics.geo-phcs.CEcs.CVcs.LGcs.NE
keywords ConditionalGANPix2PixReservoirgeomodelingImage-to-imagetranslationFacies-to-propertypredictionPorosityandpermeabilityWatersaturationGroningengasfield
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper claims that a standard paired image-to-image network, Pix2Pix, can serve as a reservoir geomodeling engine: given rasterized maps of one geological property from the Groningen gas field's Rotliegend reservoir, it directly generates maps of another property, translating facies into porosity, permeability, and water saturation and working in reverse from porosity to facies. This would matter because traditional geomodeling builds such property maps through multi-step geostatistical workflows that struggle with complex heterogeneity and data conditioning; a single trained network that maps property images to each other would be a shorter, data-driven path. The authors report pixel accuracies of 0.88 for masked facies reconstruction, 0.96 for water saturation, and 0.98 for the facies-to-water-saturation translation, with variogram analysis showing generated maps reproduce the observed spatial continuity of the reservoir. They present this as the first application of the Pix2Pix framework to reservoir geomodeling.

What carries the argument

The load-bearing object is the paired training dataset together with the Pix2Pix objective acting on it. Each sample is a 256-by-256 pair of property images, a source (facies or a mask) and a target (porosity, permeability, or water saturation), rendered by replicating Petrel's color visualization, normalized to a range of -1 to 1, and augmented from 235 original layers to 2,350 images per property through constrained rotations within plus or minus 10 degrees, zooms from 0.9 to 1.1 times, translations within plus or minus 5 percent of image dimensions, flips, and crops. The generator is a U-Net, eight downsampling convolutions to a 1-by-1 feature map followed by transposed-convolution upsampling with 50 percent dropout in the first three decoder blocks and a tanh output, whose skip connections preserve spatial detail; the discriminator is a PatchGAN scoring 70-by-70 patches. The training objective is the Pix2Pix loss, GAN loss plus L1 reconstruction loss weighted by a factor of 100, optimized with Adam at learning rate 0.0002 over 19,000 steps, and the choice of input-output pair specializes each model to one translation. Horizontal variograms of generated versus real maps supply the spatial-continuity check that grounds the claim of geological realism.

What would settle it

Decode a trained generator's porosity output back to porosity numbers by inverting the colormap from Section 2.1 and compare those values cell-by-cell with the original Gslib porosity grid at matching I-J-K locations, or with porosity logs from wells held out of training; alternatively, retrain without the color-matching and sharpness preprocessing. If continuous error (root-mean-square error, correlation) is far worse than the reported PA 0.70 and FWIoU 0.55 suggest, the metrics are tracking image fidelity rather than property accuracy.

Watch

Extended reading notes

Core claim

On its own terms, the paper's discovery is that a conditional GAN, the Pix2Pix architecture with a U-Net generator, a PatchGAN discriminator, and an adversarial-plus-L1 objective, learns to translate between reservoir property maps at field scale. Trained on 2,350 augmented rasterized layers per property derived from a 7.6-million-cell model of the Groningen field, each dedicated model maps a source property image to a target property image. In the masked-property task, the network reconstructs complete maps from color-coded masks and reports pixel accuracy of 0.88 for facies, 0.70 for porosity, 0.74 for permeability, and 0.96 for water saturation, with frequency-weighted IoU of 0.85, 0.55, 0.60, and 0.95. In the property-to-property task, facies-to-water-saturation translation reaches PA 0.98 and FWIoU 0.97, and generated maps reproduce horizontal variograms of the real data, with porosity semivariance rising from about 0.02 to about 0.07 across lags 1-25 in both real and generated data.

Load-bearing premise

The whole evaluation assumes that the 256x256 color images, made by mimicking Petrel's visualization with color-matching and sharpness adjustments, faithfully encode the numerical property values, and that slicing continuous properties into classes for IoU scoring is meaningful, so if that is wrong, the reported accuracies measure image similarity rather than reservoir-property prediction.

Editorial extensions

If this is right

  • Reservoir characterization gains a direct route: one trained network maps a facies map straight to porosity, permeability, or water-saturation maps, skipping explicit geostatistical property modeling for that target.
  • The strongest signal is in fluid-related and categorical properties, facies-to-water-saturation translation (PA 0.98, FWIoU 0.97) and masked water-saturation reconstruction (PA 0.96), so early practical use should target saturation and facies mapping rather than porosity or permeability.
  • Masked-input prediction works under incomplete data: the same network fills missing property regions from color-coded masks, offering a route for data-scarce geomodel conditioning.
  • Translation is reversible, porosity-to-facies reaches PA 0.94, so different property models can be cross-checked or converted into one another without retraining the workflow.
  • Continuous properties are the hard cases, porosity PA 0.70 and permeability PA 0.74 with the highest L1 and total generator losses, so users should expect the method's fidelity to degrade as microstructural variability increases.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The reported scores are segmentation metrics on discretized classes of color-rendered images; a decisive extension the paper does not perform is decoding generator outputs back to continuous property values and comparing them cell-by-cell with the original grid, and if continuous error is much larger than PA and FWIoU suggest, the metrics are measuring color-image fidelity rather than property acc
  • The gap between PA 0.98 and mPA 0.33 on facies-to-water-saturation suggests the high accuracies are driven largely by the dominant class, since the reservoir is about 80 percent sandstone, so readers should weigh class-balanced metrics rather than headline PA.
  • Because the machinery encodes no reservoir physics, the same paired-translation setup should transfer to other subsurface volumes such as CO2 plume concentration, temperature, or stress fields, and the natural baseline to beat is a joint-histogram or kriging-based transform, which the paper does not compare against.
  • The stated 3D sequel (Pix2Geomodel v2.0) is the key open test: the current evaluation is layer-wise 2D, so it says nothing yet about vertical continuity, and a volumetric GAN is where image-translation geomodeling would actually replace full 3D geostatistical simulation.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 6 minor

Summary. The paper presents Pix2Geomodel, a Pix2Pix-based conditional GAN that translates masked property inputs or facies maps into reservoir property maps (facies, porosity, permeability, water saturation) using 2D slices extracted from the Groningen gas field Rotliegend reservoir model. The authors augment 235 original layers to 2,350 images per property, train a U-Net generator with a PatchGAN discriminator, and evaluate on masked-property prediction and property-to-property translation tasks using pixel accuracy (PA), mean pixel accuracy (mPA), mean intersection over union (mIoU), frequency-weighted IoU (FWIoU), and variogram comparison. They report, for example, facies PA 0.88/FWIoU 0.85, water saturation PA 0.96/FWIoU 0.95, and facies-to-Sw translation PA 0.98/FWIoU 0.97, and claim this is the first application of Pix2Pix to reservoir geomodeling. The core technical approach is a standard supervised image-to-image translation setup; however, several methodological gaps in data splitting, metric definition, and image encoding affect the validity of the reported quantitative results.

Significance. If the evaluation were sound, the paper would provide a useful benchmark for applying Pix2Pix-style cGANs to direct property mapping on a real field-scale reservoir model, and the promised open dataset and code would be a community resource. The authors also include a spatial-continuity check via variograms, which is a constructive validation step, and they attempt to compare several tasks with multiple quantitative metrics. However, the current manuscript does not support its headline accuracy claims because of potential train/test leakage through augmentation, unspecified discretization of continuous properties for segmentation metrics, unvalidated raster encoding of quantitative values, and a novelty claim contradicted by the paper's own references. The strengths—the use of a real reservoir dataset, the variogram analysis, and the open-science intention—are outweighed by these evaluation issues at present.

major comments (5)
  1. [Section 2.2] The dataset augmentation and split procedure as described threatens all headline accuracy figures. The authors state that 10 augmented versions of each of the 235 original layers produce 2,350 images per property, and that the dataset is split 70/15/15 per property. It is not stated whether the split was performed at the original-layer level or at the image level. If the split is random over augmented images, then augmented near-duplicates (flipped, cropped, rotated copies) of the same original layer appear in both training and test sets, so the reported PA and FWIoU in Tables 4 and 5 would measure near-duplicate reconstruction rather than generalization to unseen reservoir locations. This is a load-bearing concern for the central claim of the paper. I request a grouped split by original layer (for example, leave-one-layer-out) with re-evaluation, and an explicit statement of the split counts. Note also that the stated counts (1,809 training + 389 validation + 390 testing = 2,588) do not match the stated 2,350 images per property.
  2. [Sections 2.6 and 3.3] The manuscript applies semantic segmentation metrics (PA, mPA, mIoU, FWIoU) to continuous properties—porosity, permeability, and water saturation—but it never explains how the continuous values were discretized into classes for the computation of these metrics. There is no description of class thresholds, bin widths, or the number of classes. Without this information, the metrics are not reproducible, and it is unclear whether they reflect petrophysical accuracy or arbitrary binning artifacts. Please specify the discretization procedure and, in addition, report continuous regression metrics (e.g., MAE, RMSE, R²) computed directly on the original property values.
  3. [Section 2.1] The rasterized 256×256 images used for training and evaluation are generated by 'replicating Petrel's visualization style through Python scripts that applied color-matching, contrast enhancement, and sharpness adjustments.' The paper provides no evidence that the resulting pixel color values map linearly or uniquely to the quantitative property values (porosity, permeability, Sw). If the colormap or the contrast/sharpness adjustments distort the encoding, then the reported PA, mIoU, and FWIoU measure image similarity rather than reservoir-property prediction accuracy, as this encoding is the foundation of every quantitative result in the paper. The authors should document the exact colormap-to-value mapping and validate it against the original Gslib/CSV values on a random sample of pixels.
  4. [Abstract and Section 1] The claim that Pix2Geomodel is 'the first application of the Pix2Pix framework in reservoir geomodeling' is contradicted by the paper's own references. Reference [55] (Pan et al., Stochastic Pix2Pix) applies Pix2Pix to conditioning channelized reservoir models, and reference [56] (Lü et al., 3D-Pix2Pix) applies Pix2Pix to history matching in reservoir geomodeling. The novelty statement should be revised to a defensible and specific formulation, for example, 'the first application of Pix2Pix to direct property-to-property translation of a full-field reservoir model' or another claim that is actually supported by the literature review.
  5. [Section 3.3, Table 4] The characterization of the results as 'high accuracy' is not supported by the overlap metrics for class-imbalanced properties. Facies PA=0.88 with mIoU=0.31 and water saturation PA=0.96 with mIoU=0.37 indicate that the majority class dominates PA and FWIoU, while minority classes are predicted poorly. Without per-class IoU values and a discussion of the class imbalance, the headline numbers are misleading. Please report per-class IoU, or evaluate with balanced metrics, and temper the accuracy claims accordingly.
minor comments (6)
  1. [Abstract] The abstract mentions 'facies-to-facies PA 0.98, FWIoU 0.97,' but Table 5 reports 'Facies to Sw PA 0.98, FWIoU 0.97'; these should be harmonized to the correct translation direction.
  2. [Table 2] The 'Number of steps' entry in Table 2 says '1,000 (example from the provided code...)' while Section 2.4 and Figure 11 report training over 19,000 steps; this is a direct contradiction that must be resolved.
  3. [Figure 9] The caption for Figure 9 lists only 'porosity, permeability, and facies,' but the text in Section 3.1 describes panels for water saturation (a), permeability (b), porosity (c), and facies (d); the caption and the in-text reference for the facies variogram (shown as Fig. 9c when it should be Fig. 9d) need correction.
  4. [Keywords] The keywords listed in the manuscript ('Time series generative adversarial networks models; Sequence generative adversarial networks models; Well log data imputation; Synthetic well log data generation') do not match the content of the paper; replace them with keywords relevant to reservoir geomodeling and image-to-image translation.
  5. [Section 2.6] The text refers to discriminator heatmaps as 'Fig. ??', which is an unresolved cross-reference; please provide the correct figure number.
  6. [Data availability] The data availability statement gives a generic account URL (https://github.com/ARhaman) rather than a direct repository link; please provide the specific repository URL or state that the code will be made available upon request.

Circularity Check

1 steps flagged · score 6.0 of 10

Masked-property prediction metrics are self-referential: the input 'mask' is a discretization of the target property, so class-level PA/FWIoU can be satisfied by copying the input classes; property-to-property translation remains independent.

  1. self definitional [Section 2.2 (masked-task construction); Section 2.6 (metrics); Section 3.3 and Table 4 (masked-task results)]
    "an automated annotation process generated color-coded masks to simulate incomplete data scenarios, mapping pixel intensities to discrete classes and pairing these masks with corresponding property images ... to form the input-output datasets required for training the Pix2Pix-based Pix2Geomodel framework (Fig. 5a). ... For water saturation prediction, the model achieved a PA of 0.96, an mPA of 0.4579, a mean IoU of 0.3710, and FWIoU of 0.95."

    The masked-task input is generated by discretizing the target property ('mapping pixel intensities to discrete classes'). The evaluation then scores the model with PA/mIoU/FWIoU, which compare discretized predicted values against ground-truth class labels (Section 2.6). Because the ground-truth class map is exactly the class map encoded in the input mask, a generator that preserves the input color classes attains near-perfect PA/FWIoU without recovering any withheld information. The reported masked-prediction numbers (e.g., Sw PA 0.96, FWIoU 0.95) therefore measure consistency with the input's own discretization rather than reconstruction of missing data; the class-level prediction target is contained in the input by construction.

full rationale

The core supervised-learning pipeline (U-Net generator + PatchGAN discriminator trained with L1+GAN loss on paired property maps) is a standard Pix2Pix application with an external architecture reference; no uniqueness theorem or load-bearing self-citation chain appears. GANSim and other references by co-authors are used for comparison, not as premises that force the results. The property-to-property translation numbers (Table 5) have independent content, although facies-to-Sw may be strongly correlated in this field. The principal circularity is confined to the masked-property task: the 'mask' is constructed by thresholding/discretizing the target property, so the class labels that PA/mIoU/FWIoU compare against are already present in the input. This makes the masked-task headline metrics partly self-referential, while leaving the translation results as genuine empirical claims. Separately, the augmentation-before-split protocol (Section 2.2: 2,350 images from 235 layers, then 70/15/15 split) risks near-duplicate leakage between train and test, and the class thresholds for continuous properties are never specified; these are correctness/validity threats, not circularity, and are noted here for completeness rather than added to the circularity score.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The paper introduces no new physical entities. It relies on a proprietary geological model as ground truth, a rasterization pipeline that is not quantitatively validated, and standard deep learning assumptions. The main unquantified choices are the class discretization thresholds, the mask generation process, and the augmentation settings.

free parameters (3)
  • Class discretization thresholds for continuous properties (porosity, permeability, Sw) = not specified
    mIoU and mPA require discrete classes; the thresholds or bins are not described, so the reported IoU values are unverifiable.
  • Masking ratio and pattern for masked property prediction = not specified
    The automated mask generation process (Section 2.2) is not quantified; results depend on mask difficulty.
  • Augmentation parameters = rotations +/-10 deg, zoom 0.9-1.1, translations +/-5%, random crop 286 to 256
    These choices are hand-set and influence generalization; no comparison or justification is given for these specific values.
assumptions (4)
  • domain assumption The NAM model of the Groningen field accurately represents the subsurface properties used as ground truth.
    The paper treats the 7.6 million-cell model from NAM as ground truth without independent validation (Section 2.1).
  • domain assumption The rasterized 256x256 images preserve the quantitative property relationships of the underlying grid.
    Image generation via Petrel-style color rendering may introduce artifacts or loss of precision; the paper does not verify that pixel values map invertibly to property values (Section 2.1).
  • domain assumption The Pix2Pix architecture (U-Net generator, PatchGAN discriminator) can learn the mapping between property images from the given dataset.
    This is a standard assumption for supervised deep learning; not proven but consistent with prior applications.
  • domain assumption Train/test split after augmentation does not leak information between sets.
    If split is at the image level rather than the original layer level, augmented copies of the same layer appear in both train and test, inflating scores (Section 2.2).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Pix2Geomodel: A Next-Generation Reservoir Geomodeling with Property-to-Property Translation." pith.science (2026). https://pith.science/paper/S44OANLC

@misc{pith2026250617747,
  author       = {Pith},
  title        = {Pith review of: Pix2Geomodel: A Next-Generation Reservoir Geomodeling with Property-to-Property Translation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/S44OANLC}},
  note         = {Machine review of arXiv:2506.17747}
}
read the original abstract

Accurate geological modeling is critical for reservoir characterization, yet traditional methods struggle with complex subsurface heterogeneity, and they have problems with conditioning to observed data. This study introduces Pix2Geomodel, a novel conditional generative adversarial network (cGAN) framework based on Pix2Pix, designed to predict reservoir properties (facies, porosity, permeability, and water saturation) from the Rotliegend reservoir of the Groningen gas field. Utilizing a 7.6 million-cell dataset from the Nederlandse Aardolie Maatschappij, accessed via EPOS-NL, the methodology included data preprocessing, augmentation to generate 2,350 images per property, and training with a U-Net generator and PatchGAN discriminator over 19,000 steps. Evaluation metrics include pixel accuracy (PA), mean intersection over union (mIoU), frequency weighted intersection over union (FWIoU), and visualizations assessed performance in masked property prediction and property-to-property translation tasks. Results demonstrated high accuracy for facies (PA 0.88, FWIoU 0.85) and water saturation (PA 0.96, FWIoU 0.95), with moderate success for porosity (PA 0.70, FWIoU 0.55) and permeability (PA 0.74, FWIoU 0.60), and robust translation performance (e.g., facies-to-facies PA 0.98, FWIoU 0.97). The framework captured spatial variability and geological realism, as validated by variogram analysis, and calculated the training loss curves for the generator and discriminator for each property. Compared to traditional methods, Pix2Geomodel offers enhanced fidelity in direct property mapping. Limitations include challenges with microstructural variability and 2D constraints, suggesting future integration of multi-modal data and 3D modeling (Pix2Geomodel v2.0). This study advances the application of generative AI in geoscience, supporting improved reservoir management and open science initiatives.

Figures

Figures reproduced from arXiv: 2506.17747 by the authors.

Figure 1
Figure 1. Usage over time of generative models (2018-2025) [source: [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. High-level architecture of the proposed Pix2Geomodel framework for subsurface reservoir characterization, [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Visualization of porosity distribution in a geological reservoir shown in two forms: the left-hand side illustrates [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (10 more)
Figure 4
Figure 4. Figure 4: Augmented property images generated for porosity using random transformations such as rotations, flipping, [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: Paired datasets for Pix2Geomodel tasks: (a) Masked property prediction with color-coded property [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 6
Figure 6. Figure 6: Architecture of the Pix2Geomodel framework, showcasing the overall architecture, including the interaction [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 7
Figure 7. Figure 7: Schematic of the Pix2Geomodel for property-to-property translation. Top: U-Net generator with en [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]
Figure 8
Figure 8. Figure 8: Training procedure of the Pix2Geomodel framework (adopted from Isola et al., 2017). [PITH_FULL_IMAGE:figures/full_fig_p013_8.png]
Figure 9
Figure 9. Figure 9: Horizontal experimental variograms comparing real and Pix2Geomodel-generated spatial continuity for [PITH_FULL_IMAGE:figures/full_fig_p015_9.png]
Figure 10
Figure 10. Figure 10: Pix2Geomodel model’s progressive improvement in predicting permeability and facies across training [PITH_FULL_IMAGE:figures/full_fig_p017_10.png]
Figure 11
Figure 11. Figure 11: Training loss curves for Pix2Geomodel: (a) Discriminator loss, (b) Generator GAN loss, (c) Generator [PITH_FULL_IMAGE:figures/full_fig_p018_11.png]
Figure 12
Figure 12. Figure 12: Pix2Geomodel predictions for masked inputs. (a) Facies, (b) Porosity, (c) Permeability, and (d) Water [PITH_FULL_IMAGE:figures/full_fig_p019_12.png]
Figure 13
Figure 13. Figure 13: Property-to-property translation using Pix2Geomodel. (a) Facies-to-porosity; (b) Porosity-to-facies; (c) [PITH_FULL_IMAGE:figures/full_fig_p021_13.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

83 extracted references · 55 canonical work pages

  1. [55]

    W. Pan, C. Torres-Verdín, and M. J. Pyrcz. Stochastic pix2pix: A new machine learning method for geophysical and well conditioning of rule-based channel reservoir models. Natural Resources Research, 30:1319–1345, 2021. https://doi.org/10.1016/j.nrr.2021.03.005

    doi:10.1016/j.nrr.2021.03.005 2021

  2. [56]

    Y . Lü, W. Liu, K. Zhang, B. Liu, Y . Wang, and J. Wang. 3D-pix2pix generative adversarial network for history matching with complex geologies. In Proc. SPIE 12059, Tenth International Symposium on Precision Mechanical Measurements, 120591C:19–23, 2021. https://doi.org/10.1117/12.2612149

    work page doi:10.1117/12.2612149 2021

  3. [1]

    M. A. Celia and J. M. Nordbotten. Practical modeling approaches for geological storage of carbon dioxide. Groundwater, 47(5):627–638, 2009. https://doi.org/10.1111/j.1745-6584.2009.00590.x

    work page arXiv 2009

  4. [2]

    M. J. Pyrcz and C. V . Deutsch. Geostatistical Reservoir Modeling. Oxford University Press, USA, 2014

    2014

  5. [3]

    A. Singh. Groundwater resources management through the applications of simulation modeling: A review. Science of The Total Environment, 499:414–423, 2014. https://doi.org/10.1016/j.scitotenv.2014.05.048

    work page doi:10.1016/j.scitotenv.2014.05.048 2014

  6. [4]

    Pandey, A

    Y .N. Pandey, A. Rastogi, S. Kainkaryam, S. Bhattacharya, and L. Saputelli. Geomodeling. InMachine Learning in the Oil and Gas Industry, Apress, Berkeley, CA, 2020. https://doi.org/10.1007/978-1-4842-6094-4-5

    work page doi:10.1007/978-1-4842-6094-4-5 2020

  7. [5]

    D. Garner. Closing the gap III: Advances in applied geomodeling for hydrocarbon reservoirs. Bulletin of Canadian Petroleum Geology, 67(4):215–216, 2019. https://doi.org/10.2113/bcpg.67.4.215

    work page doi:10.2113/bcpg.67.4.215 2019

  8. [6]

    Zhang, Y

    B. Zhang, Y . Tong, J. Du, S. Hussain, Z. Jiang, S. Ali, I. Ali, M. Khan, and U. Khan. Three-Dimensional Structural Modeling (3D SM) and Joint Geophysical Characterization (JGC) of Hydrocarbon Reservoir.Minerals, 12(3):363, 2022. https://doi.org/10.3390/min12030363

    doi:10.3390/min12030363 2022

Show all 83 references
  1. [7]

    S. J. C. Cannon. Reservoir modelling: A practical guide. 2018. https://api.semanticscholar.org/CorpusID:86600890

    2018

  2. [8]

    Cavero, N

    J. Cavero, N. H. Orellana, I. Yemez, V . Singh, and E. Izaguirre. Importance of conceptual geological models in 3D reservoir modelling. First Break, 34(7), 2016. https://doi.org/10.3997/1365-2397.2016010

    2016

  3. [9]

    Solanke, F

    B. Solanke, F. B. Onita, O. J. Ochulor, and H. O. Iriogbe. Techniques for improved reservoir characterization using advanced geological modeling in the oil and gas industry. International Journal of Applied Research in Social Sciences, 2024. Retrieved from https://api.semantic...

    2024

  4. [10]

    A. N. Okon, S. E. Adewole, and E. M. Uguma. Artificial neural network model for reservoir petrophysical properties: Porosity, permeability and water saturation prediction. Modeling Earth Systems and Environment, 7:2373–2390, 2021. https://doi.org/10.1007/s40808-020-01012-4

    2021 doi

  5. [11]

    A. Zare, M. Bagheri, and M. Ebadi. Reservoir facies and porosity modeling using seismic data and well logs by geostatistical simulation in an oil field. Carbonates and Evaporites, 35:65, 2020. https://doi.org/10.1007/s13146- 020-00605-5

    2020 doi

  6. [12]

    Obradors-Prats, E

    J. Obradors-Prats, E. E. Calderon Medina, S. J. Jones, M. Rouainia, A. C. Aplin, and A. J. L. Crook. Integrating petrophysical, geological and geomechanical modelling to assess stress states, overpressure development and compartmentalisation adjacent to a salt wall, Gulf of Me...

    2023

  7. [13]

    D. Grana. Joint facies and reservoir properties inversion. Geophysics, 83(3):M15–M24, 2018. https://doi.org/10.1190/geo2017-0670.1

    2018 doi

  8. [14]

    Goovaerts

    P. Goovaerts. Geostatistics for Natural Resources Evaluation. Oxford University Press, 1997

    1997

  9. [15]

    Zhang, C

    Y . Zhang, C. Gai, B. Song, et al. The influence of permeability and heterogeneity on chemical flooding efficiency and remaining oil distribution—based on NMR displacement imaging. Scientific Reports, 13:14316, 2023. https://doi.org/10.1038/s41598-023-39535-2

    2023 doi

  10. [16]

    J. L. Hernandez Mejia. Enhanced subsurface estimation and uncertainty modeling through data science and engineering physics. Doctoral dissertation, The University of Texas at Austin, 2024. https://

    2024

  11. [17]

    Khalili and M

    Y . Khalili and M. Ahmadi. Reservoir modeling & simulation: Advancements, challenges, and future perspectives. Journal of Chemical and Petroleum Engineering , 57(2):343–364, 2023. https://doi.org/10.22059/jchpe.2023.363392.1447

    2023

  12. [18]

    Catinat, B

    M. Catinat, B. Brigaud, M. Fleury, H. Thomas, M. Antics, and P. Ungemach. Characterizing facies and porosity- permeability heterogeneity in a geothermal carbonate reservoir with the use of NMR-wireline logging data. Geothermics, 115:102821, 2023. https://doi.org/10.1016/j.geot...

    2023

  13. [19]

    Yousefzadeh, A

    R. Yousefzadeh, A. Kazemi, M. Ahmadi, and J. Gholinezhad. Uncertainty management in reservoir engineer- ing. In Introduction to geological uncertainty management in reservoir characterization and optimization . SpringerBriefs in Petroleum Geoscience & Engineering. Springer, Ch...

    2023 doi

  14. [20]

    Kupfersberger, C

    H. Kupfersberger, C. V . Deutsch, and A. G. Journel. Deriving Constraints on Small-Scale Var- iograms due to Variograms of Large-Scale Data. Mathematical Geology , 30:837–852, 1998. https://doi.org/10.1023/A:1021726609413

    1998 doi

  15. [21]

    Strebelle

    S. Strebelle. Conditional Simulation of Complex Geological Structures Using Multiple-Point Statistics. Mathe- matical Geology, 34:1–21, 2002. https://doi.org/10.1023/A:1014009426274

    2002 doi

  16. [22]

    J. Wang, Y . Cao, K. Liu, J. Liu, and M. Kashif. Identification of sedimentary-diagenetic facies and reservoir poros- ity and permeability prediction: An example from the Eocene beach-bar sandstone in the Dongying Depression, China. Marine and Petroleum Geology, 82:69–84, 2017...

    2017 doi

  17. [23]

    Goodfellow, J

    I. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville, and Y . Bengio. Generative Adversarial Nets. In Advances in Neural Information Processing Systems (NeurIPS), 27:2672–2680,

  18. [24]

    Goodfellow, Y

    I. Goodfellow, Y . Bengio, and A. Courville. Deep Learning. MIT Press, 2016

    2016

  19. [25]

    Arjovsky, S

    M. Arjovsky, S. Chintala, and L. Bottou. Wasserstein generative adversarial networks. In Proceedings of the International Conference on Machine Learning , 2017. Retrieved from https://api.semanticscholar.org/CorpusID:2057420

    2017

  20. [26]

    Radford, L

    A. Radford, L. Metz, and S. Chintala. Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks. arXiv Preprint arXiv:1511.06434, 2015. https://doi.org/10.48550/arXiv.1511.06434

    2015 doi arXiv

  21. [27]

    Isola, J.-Y

    P. Isola, J.-Y . Zhu, T. Zhou, and A. A. Efros. Image-to-image translation with conditional adversarial networks. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 1125–1134, IEEE, 2017. https://doi.org/10.1109/CVPR.2017.632

    2017 doi

  22. [29]

    A. A. Efros and T. K. Leung. Texture synthesis by non-parametric sampling. In Proceedings of the Sev- enth IEEE International Conference on Computer Vision , Kerkyra, Greece, pages 1033–1038 vol.2, 1999. https://doi.org/10.1109/ICCV .1999.790383

    1999

  23. [30]

    A. A. Efros and W. T. Freeman. Image quilting for texture synthesis and transfer. In Seminal Graphics Papers: Pushing the Boundaries, Volume 2, pages 571–576, 2023

    2023

  24. [31]

    L. A. Gatys, A. S. Ecker, and M. Bethge. Image Style Transfer Using Convolutional Neural Networks. In 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV , USA, pages 2414–2423,

    2016

  25. [32]

    M. Chen, P. Xu, and R. Zhao. Bridging paired and unpaired medical image translation. Medical Image Analysis, 34:123–140, 2021. https://doi.org/10.1016/j.media.2021.101813

    2021

  26. [33]

    T. Chen, M. M. Cheng, P. Tan, A. Shamir, and S. M. Hu. Sketch2photo: Internet image montage. ACM Transactions on Graphics (TOG), 28(5):1–10, 2009

    2009

  27. [34]

    Y . Shih, S. Paris, F. Durand, and W. T. Freeman. Data-driven hallucination of different times of day from a single outdoor photo. ACM Transactions on Graphics (TOG), 32(6):1–11, 2013

    2013

  28. [35]

    T. Gao, Y . Liu, and Q. Zhang. A survey of generative adversarial networks for image-to-image translation.arXiv,

  29. [36]

    Henry, T

    J. Henry, T. Natalie, and D. Madsen. Pix2pix GAN for image-to-image translation. Research Gate Publication, 1–5, 2021

    2021

  30. [37]

    R. K. Senapati, R. Satvika, A. Anmandla, G. Ashesh Reddy, and C. Anil Kumar. Image-To-Image Translation Using Pix2Pix GAN and Cycle GAN. In Data Intelligence and Cognitive Informatics. ICDICI 2023. Algorithms for Intelligent Systems , edited by I.J. Jacob, S. Piramuthu, P. Fal...

    2023 doi

  31. [38]

    J. Wang, Y . Sun, and M. Lee. Multi-scale gradient U-Net for high-resolution image translation. arXiv, 2021. https://arxiv.org/abs/2105.13067

    2021 arXiv

  32. [39]

    Z. Li, B. Guan, Y . Wei, Y . Zhou, J. Zhang, and J. Xu. Mapping new realities: Ground truth image creation with Pix2Pix image-to-image translation. arXiv, 2024. https://arxiv.org/abs/2404.19265

    2024 arXiv

  33. [40]

    Z. Li, Z. Wei, and H. Zhang. Pix2Pix-Zero: Image editing with structure-preserving cross-attention. arXiv, 2023. https://arxiv.org/abs/2302.03027

    2023 arXiv

  34. [41]

    Zhong, A

    Z. Zhong, A. Y . Sun, and H. Jeong. Predicting CO2 plume migration in heterogeneous formations using condi- tional deep convolutional generative adversarial network. Water Resources Research, 56(12):e2020WR028400,

  35. [42]

    B. Choi, S. Pyun, W. Choi, and Y . Cho. Synthetic seismic data generation with pix2pix for enhanced fault detec- tion model training. Computers & Geosciences, 197:105879, 2025. https://doi.org/10.1016/j.cageo.2025.105879

    2025

  36. [43]

    Parasyris, L

    A. Parasyris, L. Stankovic, and V . Stankovic. Synthetic Data Generation for Deep Learning-Based Inversion for Velocity Model Building. Remote Sensing, 15(11):2901, 2023. https://doi.org/10.3390/rs15112901

    2023 doi

  37. [44]

    B. T. Weldemikael, G. Woldetinsae, and G. Neshir. Generating land gravity anomalies from satellite gravity observations using PIX2PIX GAN image translation. Applied Computing and Geosciences , 2024. https://doi.org/10.1016/j.acags.2024.100205

    2024

  38. [45]

    Baharuddin and C

    S. Baharuddin and C. M. John. Using generative adversarial networks to translate microresistivity image logs of carbonates into synthetic core images with accurate Dunham textures. Journal of Geophysical Research: Machine Learning and Computation, 2:e2024JH000533, 2025. https:...

    2025 doi

  39. [46]

    M. H. Shahnas, O. M. Alile, and R. N. Pysklywec. Application of Generative Adversarial Networks in Geoelectrical Field Data Processing: Innovative Approach to Solving Inverse Problems. Pure Appl. Geophys.,

  40. [47]

    Chan and A

    S. Chan and A. H. Elsheikh. Parametric generation of conditional geological realizations using generative neural networks. Computational Geosciences, 22(5):925–952, 2017. https://doi.org/10.1007/s10596-017-9670-7

    2017 doi

  41. [48]

    Mosser, O

    L. Mosser, O. Dubrule, and M. J. Blunt. Reconstruction of three-dimensional porous media using generative ad- versarial neural networks. Physical Review E, 96(4):043309, 2017. https://doi.org/10.1103/PhysRevE.96.043309

    2017 doi

  42. [49]

    Mosser, O

    L. Mosser, O. Dubrule, and M. J. Blunt. Conditioning of three-dimensional generative adversarial networks for pore and reservoir-scale models. arXiv, 2018. https://arxiv.org/abs/1802.05622

    2018 arXiv

  43. [50]

    Mosser, O

    L. Mosser, O. Dubrule, and M. J. Blunt. DeepFlow: History Matching in the Space of Deep Generative Models. arXiv, 2019. abs/1905.05749. 27 A PREPRINT - SEPTEMBER 2, 2025

    2019 arXiv

  44. [52]

    Eigen and R

    D. Eigen and R. Fergus. Predicting Depth, Surface Normals and Semantic Labels with a Common Multi-scale Convolutional Architecture. In 2015 IEEE International Conference on Computer Vision (ICCV) , Santiago, Chile, pages 2650–2658, 2015. https://doi.org/10.1109/ICCV .2015.304

    2015 doi

  45. [53]

    Hertzmann, C

    A. Hertzmann, C. E. Jacobs, N. Oliver, B. Curless, and D. H. Salesin. Image analogies. In Proceedings of the 28th Annual Conference on Computer Graphics and Interactive Techniques, pages 327–340, Association for Computing Machinery, 2001. https://doi.org/10.1145/383259.383295

    2001

  46. [54]

    Q. Liu, W. Liu, J. Yao, Y . Liu, and M. Pan. An improved method of reservoir facies modeling based on generative adversarial networks. Energies, 14(13):3873, 2021. https://doi.org/10.3390/en14133873

    2021 doi

  47. [57]

    S. Song, T. Mukerji, and J. Hou. GANSim: Conditional Facies Simulation Using an Improved Progressive Growing of Generative Adversarial Networks (GANs). Mathematical Geosciences, 53:1413–1444, 2021. https://doi.org/10.1007/s11004-021-09934-0

    2021 doi

  48. [58]

    Zhang, P

    T.-F. Zhang, P. Tilke, E. Dupont, L.-C. Zhu, L. Liang, and W. Bailey. Generating geologically realistic 3D reservoir facies models using deep learning of sedimentary architecture with generative adversarial networks. Petroleum Science, 16:541–549, 2019. https://doi.org/10.1016...

    2019 doi

  49. [59]

    M. Chen, S. Wu, H. Bedle, P. Xie, J. Zhang, and Y . Wang. Modeling of subsurface sedimen- tary facies using Self-Attention Generative Adversarial Networks (SAGANs). Petroleum Science, 2022. https://doi.org/10.1016/j.petrol.2022.110470

    2022

  50. [60]

    S. Song, T. Mukerji, J. Hou, D. Zhang, and X. Lyu. GANSim-3D for Conditional Geomodeling: Theory and Field Application. Water Resources Research, 58(7):e2021WR031865, 2022. https://doi.org/10.1029/2021WR031865

    2022 doi

  51. [61]

    Zhang, Y

    T. Zhang, Y . Li, and S. Sun. Geologically consistent facies modeling using adversarial neural net- works: a hybrid training approach. Journal of Petroleum Science and Engineering , 220:111252, 2023. https://doi.org/10.1016/j.petrol.2022.111252

    2023

  52. [62]

    S. Song, D. Zhang, T. Mukerji, and N. Wang. GANSim-surrogate: An integrated framework for stochastic condi- tional geomodelling. Journal of Hydrology, 620:129493, 2023. https://doi.org/10.1016/j.jhydrol.2023.129493

    2023

  53. [63]

    Fossum, S

    K. Fossum, S. Alyaev, and A. H. Elsheikh. Ensemble history-matching workflow using interpretable SPADE- GAN geomodel. First Break, 42(2):57–63, 2024. https://doi.org/10.3997/1365-2397.fb2024014

    2024 doi

  54. [64]

    P. H. Ranazzi, X. Luo, and M. A. Sampaio. Improving the training performance of generative adversarial networks with limited data: Application to the generation of geological models. Computers & Geosciences, 193:105747, 2024. https://doi.org/10.1016/j.cageo.2024.105747

    2024

  55. [65]

    M. M. Morales, C. Torres-Verdín, and M. J. Pyrcz. Stochastic pix2vid: A new spatiotemporal deep learning method for image-to-video synthesis in geologic CO2 storage prediction. Comput Geosci, 28:733–754, 2024. https://doi.org/10.1007/s10596-024-10298-7

    2024 doi

  56. [66]

    Merzoug and M

    A. Merzoug and M. Pyrcz. Conditional Generative Adversarial Networks for Subsurface Modeling: How Good They Really Are? Preprint, 2024. https://doi.org/10.21203/rs.3.rs-4219533/v1

    2024 doi

  57. [67]

    Merzoug and M

    A. Merzoug and M. Pyrcz. Checking Generative Artificial Intelligence-Based Subsurface Models. In IntechOpen,

  58. [68]

    Matenco and F

    L. Matenco and F. Beekman. Integrating the geological database of the subsurface of the Nether- lands, for efficient visualization and interpretation (Version 3.4). Utrecht University, EPOS-NL, 2023. https://doi.org/10.24416/UU01-7M15N6

    2023 doi

  59. [69]

    Ioffe and C

    S. Ioffe and C. Szegedy. Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift. arXiv, 2015. abs/1502.03167

    2015 arXiv

  60. [70]

    Ronneberger, P

    O. Ronneberger, P. Fischer, and T. Brox. U-Net: Convolutional Networks for Biomedical Image Segmentation. In Medical Image Computing and Computer, 2015

    2015

  61. [71]

    de Jager and C

    J. de Jager and C. Visser. Geology of the Groningen field – an overview. Netherlands Journal of Geosciences, 96:3–15, 2018. https://doi.org/10.1017/njg.2017.22. 28 A PREPRINT - SEPTEMBER 2, 2025

    2018 doi

  62. [72]

    A. G. and C. J. Huijbregts. Mining geostatistics. Academic Press, London, 1978

    1978

  63. [73]

    https://doi.org/10.5772/intechopen.115623

    doi

  64. [74]

    Bai and P

    T. Bai and P. Tahmasebi. Hybrid geological modeling: Combining machine learning and multiple-point statistics. Computers & Geosciences, 142:104519, 2020. https://doi.org/10.1016/j.cageo.2020.104519

    2020

  65. [75]

    Alqahtani, M

    H. Alqahtani, M. Kavakli-Thorne, and G. Kumar. Applications of generative adversarial networks (GANs): An updated review. Archives of Computational Methods in Engineering , 28(2):525–552, 2021. https://doi.org/10.1007/s11831-019-09388-y

    2021 doi

  66. [76]

    Raissi, P

    M. Raissi, P. Perdikaris, and G. E. Karniadakis. Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations.Journal of Computational Physics, 378:686–707, 2019. https://doi.org/10.1016...

    2019 doi

  67. [77]

    Karras, S

    T. Karras, S. Laine, and T. Aila. A style-based generator architecture for generative adversarial net- works. IEEE Transactions on Pattern Analysis and Machine Intelligence , 43(12):4217–4228, 2021. https://doi.org/10.1109/TPAMI.2020.2970919

    2021

  68. [78]

    H. Wang, T. Fu, Y . Du, J. Huang, and H. Wang. Scientific discovery in the age of artificial intelligence.Nature, 620(7972):47–60, 2023. https://doi.org/10.1038/s41586-023-06221-2

    2023 doi

  69. [79]

    Aljohani and N

    A. Aljohani and N. Alharbe. Generating Synthetic Images for Healthcare with Novel Deep Pix2Pix GAN. Electronics, 11(21):3470, 2022. https://doi.org/10.3390/electronics11213470

    2022 doi

  70. [85]

    Degen, D

    D. Degen, D. Caviedes V oullième, S. Buiter, H. J. Hendriks Franssen, H. Vereecken, A. González-Nicolás, and F. Wellmann. Perspectives of physics-based machine learning for geoscientific applications gov- erned by partial differential equations. Geoscientific Model Development...

    2023 doi

  71. [2014]

    https://doi.org/10.48550/arXiv.1406.2661

    doi arXiv

  72. [2016]

    https://doi.org/10.1109/CVPR.2016.265

    2016 doi

  73. [2020]

    https://doi.org/10.1029/2020WR028400

    doi

  74. [2021]

    https://arxiv.org/abs/2112.12625

    arXiv

  75. [2025]

    https://doi.org/10.1007/s00024-025-03710-5

    doi

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.