Pith. sign in

REVIEW 3 major objections 3 minor 15 references

Advancing Geological Carbon Storage Monitoring With 3d Digital Shadow Technology

T0 review · 3 major / 3 minor · reviewed 2026-08-08 · deepseek-v4-flash

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

Pith's one-line read A conditional normalizing flow trained on simulated 3D reservoir and seismic data estimates CO2 plume state and uncertainty from 4D seismic measurements, marking the first generative model applied to realistically sized 3D volumes.

desk verdict A real but incremental 3D extension of the authors' 2D Digital Shadow; however, the single synthetic test with no quantitative metrics or baselines is too thin to support the claims of well-calibrated uncertainty. read the letter →

arxiv 2502.07169 v1 pith:V3V4GWGT submitted 2025-02-11 physics.comp-ph cs.LGphysics.geo-ph

Abhinav Prakash Gahlot , Rafael Orozco , Felix J. Herrmann This is my paper · ORCID
classification physics.comp-phcs.LGphysics.geo-ph MSC 86A2268T0762F15
keywords geologicalcarbonstorageCO2plumemonitoring4DseismicconditionalnormalizingflowsBayesianinverseproblemsuncertaintyquantificationdigitalshadowtime-lapse
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

This paper addresses a practical need in geological carbon storage: knowing, with quantified uncertainty, where injected CO2 is spreading underground. The authors extend an existing 2D 'digital shadow' monitoring framework to three dimensions, replacing a 2D generative model with a conditional normalizing flow that maps 4D seismic images to a posterior distribution over a 128x128x128 CO2 saturation volume. The network is trained entirely on simulated data--128 pairs of forecasts from multiphase flow simulations and the corresponding synthetic time-lapse seismic surveys--and then tested on an unseen simulated plume. The paper claims that the assimilated posterior mean matches the true plume's shape and size, and that the posterior uncertainty is well calibrated, meaning the model's error is high where its uncertainty is high. If this approach transfers to real field data, it would provide a practical tool for monitoring CO2 storage sites and certifying containment.

What carries the argument

The load-bearing object is the conditional normalizing flow: an invertible neural network that transforms a latent Gaussian distribution into the plume state while conditioning on the seismic observation. Its memory-frugal invertible layers are what make a 128x128x128 generative model tractable, and its exact log-density enables direct maximum-likelihood training on the simulated pairs. The surrounding simulation-based Bayesian filtering loop generates the training data: multiphase flow simulation advances the plume state, time-lapse seismic simulation produces observations with colored noise, and the flow is trained to approximate the posterior p(plume | seismic). The permeability fields, which drive the forecast ensemble, come from a probabilistic full-waveform inversion of a baseline survey, converted to permeability through an empirical relationship.

What would settle it

Take the trained 3D normalizing flow and apply it to a plume produced by a different multiphase flow simulator, or to a permeability field drawn from a prior distribution that lies outside the training range, and compare the posterior mean and uncertainty to the ground truth. A simpler and stronger test would be to condition the network on field 4D seismic data from an operating CO2 storage site, such as Sleipner, and check whether the posterior uncertainty brackets the independently known plume extent. If the error-vs-uncertainty correlation breaks in either test, the calibration claim would be falsified.

Watch

Extended reading notes

Core claim

The central claim is that a 3D conditional normalizing flow, trained on simulation pairs, solves the Bayesian inverse problem of reconstructing CO2 saturation from 4D seismic data at reservoir scale. Each training pair couples a forecasted saturation plume, produced by multiphase flow simulation under randomly sampled permeability fields, with the corresponding synthetic 4D seismic image at 24 Hz with added colored noise. After training, the network turns a seismic observation into a full posterior distribution over the plume volume, from which a mean map and a standard-deviation map are computed. On the unseen test plume, the posterior mean closely tracks the ground truth, and the posterior standard deviation is strongly correlated with the absolute error, which the authors take as evidence of well-calibrated uncertainty. This constitutes, in the paper's words, the first known application of generative modeling to realistically sized 3D volumes.

Load-bearing premise

The validation plume is generated by the same simulation pipeline (flow simulator, seismic simulator, and observation model) used to create the 128 training pairs, so the demonstrated accuracy and calibration are only guaranteed with respect to the authors' own synthetic world; if the simulator or the underlying velocity-permeability transform misrepresents a real storage site, the performance claims would not transfer.

Editorial extensions

If this is right

  • Operators could turn each repeated 3D seismic survey into a volumetric CO2 saturation map with explicit uncertainty, directly supporting leak detection and storage certification.
  • The demonstration that normalizing flows scale to realistically sized 3D volumes opens the same conditional generative approach to other 3D subsurface inverse problems, such as reservoir characterization or monitoring of other injected fluids.
  • If the uncertainty is genuinely well calibrated, the posterior standard deviation can be used as a quantitative error estimate, which is a prerequisite for risk-based decisions in carbon storage.
  • Because the training data are purely simulated, the framework can in principle be rebuilt for a new storage site by re-running the same simulation pipeline on that site's geological prior, without needing labeled field data.

Reading between the lines

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

  • An unreported stress test would be to train on permeability samples from one geological trend and evaluate on samples from a different trend; this would reveal how quickly the posterior calibration degrades when the prior is misspecified.
  • The fixed, sparse acquisition geometry used in the synthetic surveys (16 sources, 16,000 receivers) leaves open whether the learned posterior remains calibrated when the acquisition geometry changes between surveys, a common field condition that could be tested with simulated geometry variations.
  • The fact that 128 training pairs suffice for a 3D generative model suggests the simulator encodes a strong physical prior; this could be exploited to build site-specific digital shadows with even smaller ensembles, or to fine-tune from a pre-trained base model.
  • The stated path from 'digital shadow' to 'digital twin' requires a control loop that adjusts injection operations based on the posteriors; the paper demonstrates the statistical input to that loop, but not the loop itself.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 3 minor

Summary. The paper proposes a 3D Digital Shadow framework for monitoring CO2 plumes in geological carbon storage, using conditional normalizing flows to approximate the posterior distribution of CO2 saturation given 4D seismic data. The authors simulate plume dynamics with JutulDarcy, model 4D seismic responses with JUDI and the Patchy saturation model, and train a 3D conditional normalizing flow in InvertibleNetworks.jl on 128 synthetic ensemble pairs. The method is tested on a single unseen synthetic plume from the Compass model, with results shown in Figure 1. The authors conclude that assimilating 4D seismic data significantly improves plume estimates and that the framework provides well-calibrated uncertainty quantification.

Significance. If fully supported, the work would be a valuable advance: it extends a 2D digital shadow to realistically sized 3D volumes, uses open-source and reproducible components (JutulDarcy, JUDI, InvertibleNetworks.jl), and addresses an application of societal importance. The stated novelty of a scalable 3D conditional normalizing flow and the potential for practical GCS monitoring are noteworthy. However, as presented, the evidence is insufficient to substantiate the headline claims of improved accuracy and well-calibrated uncertainty, so the significance is conditional on a substantially stronger evaluation.

major comments (3)
  1. [Section 4, Figure 1] The central claim of improved plume estimates rests on a single visual comparison. No quantitative accuracy metric (e.g., RMSE, structural similarity, or pixel-wise error norm) is reported for the inferred plume in Figure 1d against the ground truth in Figure 1b, and there is no comparison against a forecast-only prior or an alternative assimilation method such as an ensemble Kalman filter. The statement in Section 5 that incorporating 4D seismic data produced 'significant improvements' is therefore not supported by the reported evidence; please add quantitative error metrics over multiple test plumes and a baseline.
  2. [Section 3 and Figure 1b] The evaluation is in-distribution: the 128 training pairs and the single test plume are all generated by the same flow simulator (JutulDarcy), the same seismic simulation package (JUDI), and the same Patchy saturation observation model, and the permeability prior is produced by the authors' own full-waveform inversion approach. This experiment only demonstrates that the network can invert the exact forward model used to create the data. To support generalization claims, please test on out-of-distribution scenarios, such as different geological realizations, different injection rates or times, or perturbations to the observation model.
  3. [Section 5, Conclusions] The claim of 'well-calibrated uncertainty quantification, with strong correlations between inferred uncertainty and ground-truth errors' is based on a visual similarity between Figure 1e and Figure 1f. A calibration claim requires quantitative verification, such as empirical coverage of credible intervals, interval scores, or a correlation coefficient between uncertainty and absolute error computed across the volume or across an ensemble of test cases. Please include such diagnostics.
minor comments (3)
  1. [Equation (3)] The L2 norm notation in the training objective, written as '||...|| 2 2 2', is garbled and should be cleaned up to a standard norm notation for a normalizing flow negative log-likelihood.
  2. [Figure 1 caption] The caption for panel (d) says 'A posterior sample', while Section 4 states that the mean of 128 posterior samples is shown; clarify which quantity is displayed in panel (d).
  3. [Section 1] The text says 'This abstract extends...' but the document is a full preprint; change 'abstract' to 'paper' or 'work'.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's synthetic monitoring experiment is an in-distribution simulation study, but the CNF posterior is not fitted to the test plume and no derivation reduces to its inputs.

full rationale

The paper's derivation chain is a simulation-based inference workflow: equations (1) and (2) define the plume dynamics and seismic observation models, and equation (3) trains a conditional normalizing flow on 128 ensemble pairs. The test plume in Figure 1b is generated by the same JutulDarcy flow simulator, JUDI seismic simulator, and Patchy saturation observation model used for the training pairs, which limits external validity but does not make the experiment circular. The network is not fit to the test plume; the posterior is evaluated on a sample that is not in the training set, and the posterior mean is a genuine function of the conditioning seismic data. The self-citations to Gahlot, Orozco, et al. (2024) and Orozco, Siahkoohi, et al. (2024) supply the method, the empirical velocity-permeability relationship, and the permeability prior, but the central claim of improved plume estimates and calibrated uncertainty rests on the synthetic experiment reported here, not on a theorem imported from those citations. The calibration statement is based on a visual correlation between error and uncertainty rather than a quantitative coverage metric, which is a weakness in evidence strength, not a circularity. No equation is defined in terms of the target result, no fitted parameter is renamed a prediction, and no uniqueness argument is used to force a choice. Accordingly, no specific circular step can be identified.

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

The central claim rests on a chain of simulation and modeling choices: the flow simulator, the seismic simulator, the empirical permeability transform, and the small training ensemble. No new physical entities are introduced, but the method's validity depends heavily on the fidelity of these models and on the representative-ness of the 128-sample prior.

free parameters (4)
  • injection rate (limited to 3% storage capacity) = not reported, chosen to respect 3% capacity
    Modeling choice in Section 3.1 that determines the plume dynamics and training data; the central demonstration depends on this value but it is not reported.
  • velocity-to-permeability empirical transform = not reported; from prior paper
    Section 3 states permeability samples are obtained by converting FWI velocity models using an empirical relationship from Gahlot, Orozco, et al. (2024); the conversion coefficients are not given here and are taken from the authors' prior work.
  • seismic noise level (18 dB SNR) = 18 dB
    Section 3.2 sets colored Gaussian noise to 18 dB SNR; the calibration of uncertainties is likely sensitive to this simulation setting.
  • training set size (128 ensemble members) = 128
    Section 3 uses only 128 members for both the prior ensemble and CNF training; with such a small effective dataset, the trained posterior may be overfit to the simulator.
assumptions (4)
  • domain assumption Multi-phase flow simulator JutulDarcy accurately represents CO2 migration in a saline aquifer.
    Section 3.1 uses JutulDarcy to produce the synthetic states; if the flow model is wrong, all training and test plumes are unrepresentative.
  • domain assumption Seismic observations relate to CO2 saturation through the Patchy saturation model and JUDI wave simulations.
    Section 3.2 maps flow outputs to acoustic changes and simulates 4D seismic data; the posterior is only as good as this observation model.
  • domain assumption The permeability prior from probabilistic full-waveform inference covers the range of plausible subsurface models.
    Section 3 establishes the probabilistic baseline by FWI on the site; if the prior is misspecified, the posterior and the synthetic test are biased.
  • ad hoc to paper Conditional normalizing flows can approximate the true posterior with 128 training pairs.
    Section 3.3 trains CNFs on only 128 pairs; the paper provides no convergence analysis or evidence that the learned posterior is close to the true posterior.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Advancing Geological Carbon Storage Monitoring With 3d Digital Shadow Technology." pith.science (2026). https://pith.science/paper/V3V4GWGT

@misc{pith2026250207169,
  author       = {Pith},
  title        = {Pith review of: Advancing Geological Carbon Storage Monitoring With 3d Digital Shadow Technology},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/V3V4GWGT}},
  note         = {Machine review of arXiv:2502.07169}
}
read the original abstract

Geological Carbon Storage (GCS) is a key technology for achieving global climate goals by capturing and storing CO2 in deep geological formations. Its effectiveness and safety rely on accurate monitoring of subsurface CO2 migration using advanced time-lapse seismic imaging. A Digital Shadow framework integrates field data, including seismic and borehole measurements, to track CO2 saturation over time. Machine learning-assisted data assimilation techniques, such as generative AI and nonlinear ensemble Bayesian filtering, update a digital model of the CO2 plume while incorporating uncertainties in reservoir properties. Compared to 2D approaches, 3D monitoring enhances the spatial accuracy of GCS assessments, capturing the full extent of CO2 migration. This study extends the uncertainty-aware 2D Digital Shadow framework by incorporating 3D seismic imaging and reservoir modeling, improving decision-making and risk mitigation in CO2 storage projects.

Figures

Figures reproduced from arXiv: 2502.07169 by the authors.

Figure 1
Figure 1. (a) Time-lapse seismic measurements used as conditioning input to the trained conditional Normalizing Flow. (b) The ground-truth unknown plume that is the target to monitor (c) An ensemble member for the state of the plume used for training. (d)* A posterior sample conditioned on the seismic observation (e) Error between the inferred posterior mean and the ground truth (f) The variance between posterior samples. 4 … view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

15 extracted references · 12 canonical work pages

  1. [1]

    Jones, C., J

    E. Jones, C., J. A. Edgar, J. I. Selvage, and H. Crook. 2012. ``Building Complex Synthetic Models to Evaluate Acquisition Geometries and Velocity Inversion Technologies.'' In 74th EAGE Conference and Exhibition Incorporating EUROPEC 2012, cp--293. https://doi.org/https://doi.org/10.3997/2214-4609.20148575

    arXiv 2012

  2. [2]

    Gahlot, Abhinav Prakash, Haoyun Li, Ziyi Yin, Rafael Orozco, and Felix J Herrmann. 2024. ``A Digital Twin for Geological Carbon Storage with Controlled Injectivity.'' arXiv Preprint arXiv:2403.19819

    arXiv 2024

  3. [3]

    An uncertainty-aware Digital Shadow for underground multimodal CO2 storage monitoring

    Gahlot, Abhinav Prakash, Rafael Orozco, Ziyi Yin, and Felix J. Herrmann. 2024. ``An Uncertainty-Aware Digital Shadow for Underground Multimodal CO2 Storage Monitoring.'' https://doi.org/10.48550/arXiv.2410.01218

    work page Pith review arXiv doi:10.48550/arxiv.2410.01218 2024

  4. [4]

    Kingma, Diederik P., and Jimmy Ba. 2014. ``Adam: A Method for Stochastic Optimization.'' CoRR abs/1412.6980. https://api.semanticscholar.org/CorpusID:6628106

    arXiv 2014

  5. [5]

    Louboutin, Mathias, Philipp Witte, Ziyi Yin, Henryk Modzelewski, Kerim, Carlos da Costa, and Peterson Nogueira. 2023. ``Slimgroup/JUDI.jl: V3.2.3.'' Zenodo. https://doi.org/10.5281/zenodo.7785440

    work page doi:10.5281/zenodo.7785440 2023

  6. [6]

    Lumley, David. 2010. ``4D Seismic Monitoring of CO 2 Sequestration.'' The Leading Edge 29 (2): 150--55

    work page 2010

  7. [7]

    Møyner, Olav, Grant Bruer, and Ziyi Yin. 2023. ``Sintefmath/JutulDarcy.jl: V0.2.3.'' Zenodo. https://doi.org/10.5281/zenodo.7855628

    work page doi:10.5281/zenodo.7855628 2023

  8. [8]

    ASPIRE: Iterative Amortized Posterior Inference for Bayesian Inverse Problems

    Orozco, Rafael, Ali Siahkoohi, Mathias Louboutin, and Felix J. Herrmann. 2024. ``ASPIRE: Iterative Amortized Posterior Inference for Bayesian Inverse Problems.'' https://arxiv.org/abs/2405.05398

    work page Pith review arXiv 2024

Show all 15 references
  1. [9]

    Herrmann

    Orozco, Rafael, Philipp Witte, Mathias Louboutin, Ali Siahkoohi, Gabrio Rizzuti, Bas Peters, and Felix J. Herrmann. 2024. ``InvertibleNetworks.jl: A Julia Package for Scalable Normalizing Flows.'' Journal of Open Source Software 9 (99): 6554. https://doi.org/10.21105/joss.06554

    2024 doi

  2. [10]

    Panel on Climate Change), IPCC (Intergovernmental. 2018. Global Warming of 1.5° c. An IPCC Special Report on the Impacts of Global Warming of 1.5° c Above Pre-Industrial Levels and Related Global Greenhouse Gas Emission Pathways, in the Context of Strengthening the Global Resp...

    2018

  3. [11]

    Papamakarios, George, Eric Nalisnick, Danilo Jimenez Rezende, Shakir Mohamed, and Balaji Lakshminarayanan. 2021. ``Normalizing Flows for Probabilistic Modeling and Inference.'' J. Mach. Learn. Res. 22 (1)

    2021

  4. [12]

    Ringrose, Philip. 2020. How to Store CO\(_ 2 \) Underground: Insights from Early-Mover CCS Projects . Vol. 129. Springer

    2020

  5. [13]

    ---------. 2023. Storage of Carbon Dioxide in Saline Aquifers: Building Confidence by Forecasting and Monitoring. Society of Exploration Geophysicists

    2023

  6. [14]

    Spantini, Alessio, Ricardo Baptista, and Youssef Marzouk. 2022. ``Coupling Techniques for Nonlinear Ensemble Filtering.'' SIAM Review 64 (4): 921--53

    2022

  7. [15]

    Gorman, and Felix J

    Witte, Philipp A., Mathias Louboutin, Navjot Kukreja, Fabio Luporini, Michael Lange, Gerard J. Gorman, and Felix J. Herrmann. 2019. ``A Large-Scale Framework for Symbolic Implementations of Seismic Inversion Algorithms in Julia.'' Geophysics 84 (3): F57--71. https://doi.org/10...

    2019 doi

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.