The reviewed record of science sign in
Pith

arxiv: 2606.05448 · v2 · pith:TQKIVWJ5 · submitted 2026-06-03 · physics.geo-ph · physics.comp-ph

Learning and Inferring Multiphase Flow Dynamics in Porous Media using Scientific Machine Learning: Application to the "FluidFlower" CO2 Injection Experiment

Hannah Lu , Lluis Salo-Salgado , Yun-Ting Chou , Ehsan Haghighat , Ruben Juanes This is my paper

Reviewed by Pith2026-06-28 02:21 UTCgrok-4.3pith:TQKIVWJ5open to challenge →

classification physics.geo-ph physics.comp-ph
keywords multiphase flowporous mediaCO2 storagesurrogate modelingBayesian inferenceconvolutional neural networkparameter estimationFluidFlower experiment
0
0 comments X

The pith

A convolutional neural network surrogate trained on simulations enables efficient Bayesian inference of multiphase flow parameters from full-field FluidFlower observations.

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

The paper develops a framework that combines a convolutional neural network surrogate with Bayesian inference to predict and identify parameters governing CO2-brine flow in porous media. The surrogate is trained on high-fidelity numerical simulations to reproduce saturation and dissolved concentration fields across a wide range of properties. It is then placed inside a Markov chain Monte Carlo sampler so that unknown parameters can be inferred from the detailed spatial and temporal data collected in the FluidFlower laboratory experiment. This setup makes it practical to explore high-dimensional parameter spaces and to use the entire observed fields rather than summary plume measurements. The resulting models match experimental CO2 migration and dissolution more closely than earlier manual calibrations.

Core claim

By training a convolutional neural network on a wide range of high-fidelity multiphase flow simulations, the authors create a surrogate model that reproduces CO2 saturation and dissolved concentration fields at a fraction of the computational cost. Embedding this surrogate in a Bayesian Markov chain Monte Carlo sampler allows inference of the underlying flow parameters conditioned on the full time-resolved, spatially resolved observations from the FluidFlower CO2 injection experiment. The resulting posterior distributions identify both well-constrained and degenerate parameter sets, and the inferred models achieve substantially closer agreement with experimental plumes than previous calibrat

What carries the argument

Convolutional neural network surrogate for the evolution of saturation and concentration fields, embedded within Markov chain Monte Carlo for Bayesian parameter inference.

If this is right

  • The surrogate accelerates predictions by orders of magnitude relative to direct numerical simulation.
  • Full-field observations make more parameters identifiable than plume-scale metrics alone.
  • Parameter inference becomes tractable for high-dimensional spaces in multiphase flow problems.
  • Observations gain informativeness once the CO2 plume interacts with faults and sealing layers.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Similar surrogate-plus-Bayesian workflows could be applied to field data from actual geological storage sites.
  • Quantifying the effect of surrogate approximation error on posterior uncertainty would strengthen confidence in the inferred parameters.
  • Extending the training data to include additional physical processes such as capillary pressure hysteresis might improve robustness.

Load-bearing premise

The convolutional neural network surrogate reproduces the true multiphase flow physics accurately enough that its approximation errors do not systematically shift the inferred parameter distributions.

What would settle it

A direct comparison of saturation and concentration fields produced by the surrogate against an independent set of high-fidelity numerical simulations run at parameter values sampled from the inferred posterior.

read the original abstract

Accurate prediction and parameter identification of multiphase flow in porous media remain central challenges in geological carbon dioxide storage due to strong nonlinearities, high-dimensional parameter spaces, and limited observational data. We present a machine learning framework that integrates surrogate modeling and Bayesian inference to enable efficient forward prediction and inverse parameter estimation for CO2-brine flows in geological media. The approach is demonstrated using the "FluidFlower" experimental rig, a controlled laboratory system that provides high-resolution, time-resolved observations of CO2 migration in heterogeneous porous media. A convolutional neural network surrogate is trained on high-fidelity numerical simulations to learn the evolution of CO2 saturation and dissolved CO2 concentration fields over a wide range of multiphase flow properties. The trained surrogate is embedded within a Markov chain Monte Carlo framework for parameter inference conditioned on experimental observations. Results show that the surrogate accurately captures large-scale CO2 plume migration, dissolution dynamics, and multiphase flow behavior while providing orders-of-magnitude acceleration compared to traditional simulations. Embedding the surrogate within a Bayesian framework enables computationally tractable exploration of the parameter space and reveals both identifiable and non-identifiable parameter combinations that produce similar plume behavior. By leveraging spatially and temporally resolved full-field observations, the framework substantially improves agreement between simulations and experiments compared to previous manual calibrations based on limited plume-scale metrics. Analysis using progressively increasing observation horizons further shows that observations become more informative once the plume interacts with geological features such as faults and sealing layers.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper presents a scientific machine learning framework that trains a convolutional neural network (CNN) surrogate on high-fidelity multiphase flow simulations and embeds it in an MCMC procedure for Bayesian parameter inference. The method is applied to the FluidFlower CO2 injection experiment, with the central claim that the surrogate accurately reproduces large-scale plume migration and dissolution while the use of spatially and temporally resolved full-field observations yields substantially better agreement with experiment than prior manual calibrations based on limited plume metrics.

Significance. If the surrogate fidelity and lack of parameter-dependent bias are demonstrated, the work would supply a practical route to tractable inversion with rich experimental data in porous-media flows, together with reproducible acceleration of forward modeling. The identification of identifiable versus non-identifiable parameter combinations is also potentially useful for experimental design.

major comments (2)
  1. [Abstract] Abstract and results section: the assertions that the CNN 'accurately captures large-scale CO2 plume migration' and that the framework 'substantially improves agreement' are not supported by any reported quantitative error metrics (e.g., L2 norms, saturation RMSE on held-out simulations, or posterior predictive checks). Without these, the comparison to manual calibrations cannot be evaluated and the central claim remains unverified.
  2. [Methods] Methods (surrogate training) and results (inference): no cross-validation statistics, sensitivity analysis, or residual-correlation tests are described that would confirm the CNN approximation error is uncorrelated with the parameters varied in MCMC (relative permeability exponents, dissolution rates, etc.). Parameter-dependent surrogate bias would systematically shift the reported posteriors even if the surrogate matches the experimental images perfectly.
minor comments (2)
  1. Notation for the CNN architecture, loss function, and training hyperparameters should be collected in a single table or subsection for reproducibility.
  2. Figure captions should explicitly state whether the displayed fields are experimental data, high-fidelity simulation, or surrogate output.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed comments. We agree that quantitative support for the surrogate accuracy claims and checks for parameter-dependent bias are necessary to substantiate the central results. We address each point below and will incorporate the requested analyses and metrics in a revised manuscript.

read point-by-point responses

  1. Referee: [Abstract] Abstract and results section: the assertions that the CNN 'accurately captures large-scale CO2 plume migration' and that the framework 'substantially improves agreement' are not supported by any reported quantitative error metrics (e.g., L2 norms, saturation RMSE on held-out simulations, or posterior predictive checks). Without these, the comparison to manual calibrations cannot be evaluated and the central claim remains unverified.

    Authors: We agree that explicit quantitative error metrics are required to support the statements in the abstract and results. The current manuscript relies on visual comparisons and qualitative descriptions of plume migration. In the revision we will add L2 norms and saturation RMSE computed on held-out high-fidelity simulations, as well as posterior predictive checks against the experimental images. These additions will also allow a direct quantitative comparison with the earlier manual calibration results. revision: yes

  2. Referee: [Methods] Methods (surrogate training) and results (inference): no cross-validation statistics, sensitivity analysis, or residual-correlation tests are described that would confirm the CNN approximation error is uncorrelated with the parameters varied in MCMC (relative permeability exponents, dissolution rates, etc.). Parameter-dependent surrogate bias would systematically shift the reported posteriors even if the surrogate matches the experimental images perfectly.

    Authors: We acknowledge that the manuscript does not present cross-validation statistics, sensitivity analysis of the surrogate error with respect to the MCMC parameters, or residual-correlation tests. Such checks are indeed necessary to rule out systematic bias. In the revised version we will include k-fold cross-validation error statistics, a sensitivity study of surrogate residuals versus the varied parameters (relative permeability exponents, dissolution rates, etc.), and correlation analysis between residuals and parameter values to demonstrate that any approximation error is uncorrelated with the inference parameters. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation self-contained against external benchmarks

full rationale

The paper trains a CNN surrogate on independent high-fidelity numerical simulations (not on the target experiment), then embeds it in MCMC for parameter inference conditioned on external FluidFlower experimental observations. Reported improvements are quantified against prior manual calibrations that used limited plume-scale metrics on the same external data. No equation or claim reduces a 'prediction' to a quantity defined by the fitted parameters themselves, and the provided text contains no load-bearing self-citation chains or ansatzes smuggled via prior work. The derivation chain is therefore self-contained and externally falsifiable.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 0 invented entities

The central claim depends on the accuracy of the high-fidelity simulations used as training targets and on the assumption that the laboratory observations contain no significant unmodeled biases.

free parameters (1)
  • CNN architecture and training hyperparameters
    Network depth, width, learning rate, and regularization choices are selected to fit the simulation dataset.
axioms (2)
  • domain assumption High-fidelity numerical simulations accurately represent the true multiphase flow physics for the parameter ranges explored
    The surrogate is trained directly on these simulations as ground truth.
  • domain assumption Experimental observations from the FluidFlower rig are representative and free of significant unaccounted measurement or setup errors
    These observations are used to condition the Bayesian posterior.

pith-pipeline@v0.9.1-grok · 5819 in / 1356 out tokens · 54909 ms · 2026-06-28T02:21:00.192478+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

54 extracted references · 4 canonical work pages

  1. [1]

    Benson and David R

    Sally M. Benson and David R. Cole. Underground geological storage. In Bert Metz, Ogunlade Davidson, Heleen de Coninck, Manuela Loos, and Leo Meyer, editors,IPCC Special Report on Carbon Dioxide Capture and Storage, chapter 5, pages 195–276. Cambridge University Press, 2005

    2005

  2. [2]

    CO 2 sequestration in deep sedimentary formations.Elements, 4(5):325–331, 2008

    Sally M Benson and David R Cole. CO 2 sequestration in deep sedimentary formations.Elements, 4(5):325–331, 2008

    2008

  3. [3]

    Sequestration of CO 2 in geological media: criteria and approach for site selection in response to climate change.Energy conversion and management, 41(9):953– 970, 2000

    Stefan Bachu. Sequestration of CO 2 in geological media: criteria and approach for site selection in response to climate change.Energy conversion and management, 41(9):953– 970, 2000

    2000

  4. [4]

    Subsurface carbon dioxide and hydrogen storage for a sustainable energy future.Nature Reviews Earth & Environment, 4(2):102–118, 2023

    Samuel Krevor, Heleen De Coninck, Sarah E Gasda, Navraj Singh Ghaleigh, Vincent de Gooyert, Hadi Hajibeygi, Ruben Juanes, Jerome Neufeld, Jennifer J Roberts, and Floris Swennenhuis. Subsurface carbon dioxide and hydrogen storage for a sustainable energy future.Nature Reviews Earth & Environment, 4(2):102–118, 2023

    2023

  5. [5]

    The CO 2 CRC Otway Project: Leveraging experience and exploiting new opportunities at Australia’s first CCS project site.Energy Procedia, 4:5447–5454, 2011

    Sandeep Sharma, Peter Cook, Charles Jenkins, Tony Steeper, Mal Lees, and Namiko Ranasinghe. The CO 2 CRC Otway Project: Leveraging experience and exploiting new opportunities at Australia’s first CCS project site.Energy Procedia, 4:5447–5454, 2011

    2011

  6. [6]

    IEA GHG Weyburn CO 2 monitoring and storage project.Fuel Processing Technology, 86(14-15):1547–1568, 2005

    C Preston, M Monea, W Jazrawi, K Brown, S Whittaker, D White, D Law, R Chalaturnyk, and B Rostron. IEA GHG Weyburn CO 2 monitoring and storage project.Fuel Processing Technology, 86(14-15):1547–1568, 2005

    2005

  7. [7]

    Measuring permanence of CO 2 storage in saline formations: the Frio experiment.Environmental Geosciences, 13(2):105–121, 2006

    Susan D Hovorka, Sally M Benson, Christine Doughty, Barry M Freifeld, Shinichi Sakurai, Thomas M Daley, Yousif K Kharaka, Mark H Holtz, Robert C Trautz, H Seay Nance, et al. Measuring permanence of CO 2 storage in saline formations: the Frio experiment.Environmental Geosciences, 13(2):105–121, 2006

    2006

  8. [8]

    Geophysical monitoring of the injection and postclosure phases at the Ketzin pilot site

    Stefan L¨ uth, Jan Henninges, Monika Ivandic, Christopher Juhlin, Thomas Kempka, Ben Norden, Dennis Rippe, and Cornelia Schmidt-Hattenberger. Geophysical monitoring of the injection and postclosure phases at the Ketzin pilot site. InActive geophysical monitoring, pages 523–561. Elsevier, 2020

    2020

  9. [9]

    Auli Niemi, Jacob Bensabat, Saba Joodaki, Farzad Basirat, Maryeh Hedayati, Zhibing Yang, Lily Perez, Stanislav Levchenko, Alon Shklarnik, Rona Ronen, et al. Characterizing CO 2 residual trapping in-situ by means of single-well push-pull experiments at Heletz, Israel, pilot injection site–experimental procedures and results of the experiments.International...

    2020

  10. [10]

    Impact of relative permeability hysteresis on geological CO2 storage

    Ruben Juanes, EJ Spiteri, FM Orr Jr, and MJ Blunt. Impact of relative permeability hysteresis on geological CO2 storage. Water resources research, 42(12), 2006

    2006

  11. [11]

    Gravity currents with residual trapping.Journal of Fluid Mechanics, 611:35–60, 2008

    Mark A Hesse, FM Orr Jr, and HA Tchelepi. Gravity currents with residual trapping.Journal of Fluid Mechanics, 611:35–60, 2008

    2008

  12. [12]

    CO 2 migration in saline aquifers

    CW MacMinn, ML Szulczewski, and R Juanes. CO 2 migration in saline aquifers. Part 2. Capillary and solubility trapping.Journal of Fluid Mechanics, 688:321–351, 2011

    2011

  13. [13]

    Lifetime of carbon capture and storage as a climate-change mitigation technology

    Michael L Szulczewski, Christopher W MacMinn, Howard J Herzog, and Ruben Juanes. Lifetime of carbon capture and storage as a climate-change mitigation technology. Proceedings of the National Academy of Sciences, 109(14):5185–5189, 2012

    2012

  14. [14]

    Maturing global CO 2 storage resources on offshore continental margins to achieve 2DS emissions reductions.Scientific reports, 9(1):17944, 2019

    Philip S Ringrose and Tip A Meckel. Maturing global CO 2 storage resources on offshore continental margins to achieve 2DS emissions reductions.Scientific reports, 9(1):17944, 2019

    2019

  15. [15]

    Petroleum reservoir simulation.Applied Science Publishers, 476, 1979

    Khalid Aziz. Petroleum reservoir simulation.Applied Science Publishers, 476, 1979

    1979

  16. [16]

    TOUGH2 user’s guide, version 2.0

    Karsten Pruess, Curt Oldenburg, and George Moridis. TOUGH2 user’s guide, version 2.0. 1999

    1999

  17. [17]

    Springer, 2021

    Dirk Scheer, Holger Class, and Bernd Flemisch.Subsurface environmental modelling between science and policy. Springer, 2021

    2021

  18. [18]

    Room-scale CO 2 injections in a physical reservoir model with faults.Transport in Porous Media, 151(5):913–937, 2024

    MA Fernø, Malin Haugen, Kristoffer Eikehaug, Olav Folkvord, Benyamine Benali, Jakub W Both, Erlend Storvik, Casey W Nixon, Robert L Gawthrope, and Jan Martin Nordbotten. Room-scale CO 2 injections in a physical reservoir model with faults.Transport in Porous Media, 151(5):913–937, 2024

    2024

  19. [19]

    Final benchmark description: FluidFlower International Benchmark Study

    JM Nordbotten, M Fernø, B Flemisch, R Juanes, and M Jørgensen. Final benchmark description: FluidFlower International Benchmark Study. Zenodo, 2022

    2022

  20. [20]

    The fluidflower validation benchmark study for the storage of CO2.Transport in Porous Media, 151(5):865–912, 2024

    Bernd Flemisch, Jan M Nordbotten, Martin Fernø, Ruben Juanes, Jakub W Both, Holger Class, Mojdeh Delshad, Florian Doster, Jonathan Ennis-King, Jacques Franc, et al. The fluidflower validation benchmark study for the storage of CO2.Transport in Porous Media, 151(5):865–912, 2024

    2024

  21. [21]

    Physical variability in meter-scale laboratory CO 2 injections in faulted geometries.arXiv preprint arXiv:2301.07347, 2023

    Malin Haugen, Lluis Salo-Salgado, Kristoffer Eikehaug, Benyamine Benali, Jakub W Both, Erlend Storvik, Olav Folkvord, Ruben Juanes, Jan Martin Nordbotten, and Martin A Ferno. Physical variability in meter-scale laboratory CO 2 injections in faulted geometries.arXiv preprint arXiv:2301.07347, 2023

    work page arXiv 2023

  22. [22]

    Llu ´ ıs Sal´ o-Salgado, Malin Haugen, Kristoffer Eikehaug, Martin Fernø, Jan M Nordbotten, and Ruben Juanes. Direct comparison of numerical simulations and experiments of CO2 injection and migration in geologic media: value of local data and forecasting capability.Transport in Porous Media, 151(5):1199–1240, 2024

    2024

  23. [23]

    A history matching study for the fluidflower benchmark project.Transport in Porous Media, 151(5):1113–1139, 2024

    Xiaoming Tian, Michiel Wapperom, James Gunning, Samuel Jackson, Andy Wilkins, Chris Green, Jonathan Ennis-King, and Denis Voskov. A history matching study for the fluidflower benchmark project.Transport in Porous Media, 151(5):1113–1139, 2024. Draft 14 PNAS Nexus, YEAR, Volume XX, Issue x

    2024

  24. [24]

    Real-time high- resolution CO 2 geological storage prediction using nested fourier neural operators.Energy & Environmental Science, 16(4):1732–1741, 2023

    Gege Wen, Zongyi Li, Qirui Long, Kamyar Azizzadenesheli, Anima Anandkumar, and Sally M Benson. Real-time high- resolution CO 2 geological storage prediction using nested fourier neural operators.Energy & Environmental Science, 16(4):1732–1741, 2023

    2023

  25. [25]

    Deep convolutional encoder- decoder networks for uncertainty quantification of dynamic multiphase flow in heterogeneous media.Water Resources Research, 55(1):703–728, 2019

    Shaoxing Mo, Yinhao Zhu, Nicholas Zabaras, Xiaoqing Shi, and Jichun Wu. Deep convolutional encoder- decoder networks for uncertainty quantification of dynamic multiphase flow in heterogeneous media.Water Resources Research, 55(1):703–728, 2019

    2019

  26. [26]

    Predicting CO 2 plume migration in heterogeneous formations using conditional deep convolutional generative adversarial network.Water Resources Research, 55(7):5830–5851, 2019

    Zhi Zhong, Alexander Y Sun, and Hoonyoung Jeong. Predicting CO 2 plume migration in heterogeneous formations using conditional deep convolutional generative adversarial network.Water Resources Research, 55(7):5830–5851, 2019

    2019

  27. [27]

    Uncertainty quantification of fluid leakage and fault instability in geologic CO 2 storage.Water Resources Research, 61(10):e2024WR039275, 2025

    Hannah Lu, Llu ´ ıs Sal´ o-Salgado, Youssef M Marzouk, and Ruben Juanes. Uncertainty quantification of fluid leakage and fault instability in geologic CO 2 storage.Water Resources Research, 61(10):e2024WR039275, 2025

    2025

  28. [28]

    Lithological controls on the permeability of geologic faults: Surrogate modeling and sensitivity analysis.arXiv preprint arXiv:2511.09674, 2025

    Hannah Lu, Lluıs Salo-Salgado, and Ruben Juanes. Lithological controls on the permeability of geologic faults: Surrogate modeling and sensitivity analysis.arXiv preprint arXiv:2511.09674, 2025

    work page arXiv 2025

  29. [29]

    A deep-learning-based surrogate model for data assimilation in dynamic subsurface flow problems.Journal of Computational Physics, 413:109456, 2020

    Meng Tang, Yimin Liu, and Louis J Durlofsky. A deep-learning-based surrogate model for data assimilation in dynamic subsurface flow problems.Journal of Computational Physics, 413:109456, 2020

    2020

  30. [30]

    Stochastic reconstruction of an oolitic limestone by generative adversarial networks.Transport in Porous Media, 125(1):81–103, 2018

    Lukas Mosser, Olivier Dubrule, and Martin J Blunt. Stochastic reconstruction of an oolitic limestone by generative adversarial networks.Transport in Porous Media, 125(1):81–103, 2018

    2018

  31. [31]

    Reliable amortized variational inference with physics-based latent distribution correction.Geophysics, 88(3):R297–R322, 2023

    Ali Siahkoohi, Gabrio Rizzuti, Rafael Orozco, and Felix J Herrmann. Reliable amortized variational inference with physics-based latent distribution correction.Geophysics, 88(3):R297–R322, 2023

    2023

  32. [32]

    Inferring fault frictional and reservoir hydraulic properties from injection-induced seismicity.Geophysical Research Letters, 45(3):1313–1320, 2018

    Jayanth Jagalur-Mohan, Birendra Jha, Zheng Wang, Ruben Juanes, and Youssef Marzouk. Inferring fault frictional and reservoir hydraulic properties from injection-induced seismicity.Geophysical Research Letters, 45(3):1313–1320, 2018

    2018

  33. [33]

    National Academies Press Washington, DC, USA, 2023

    Karen Willcox, D Bingham, C Chung, J Chung, C Cruz- Neira, C Grant, J Kinter, R Leung, P Moin, L Ohno- Machado, et al.Foundational research gaps and future directions for digital twins. National Academies Press Washington, DC, USA, 2023

    2023

  34. [34]

    Performance of an open- source image-based history matching framework for CO 2 storage.Transport in Porous Media, 153(2):21, 2026

    David Landa-Marb´ an, Tor H Sandve, Jakub W Both, Jan M Nordbotten, and Sarah E Gasda. Performance of an open- source image-based history matching framework for CO 2 storage.Transport in Porous Media, 153(2):21, 2026

    2026

  35. [35]

    Curtis M Oldenburg. Are we all in concordance with the meaning of the word conformance, and is our definition in conformity with standard definitions?Greenhouse Gases: Science and Technology, 8(2):210–214, 2018

    2018

  36. [36]

    Learning Mesh-Based Simulation with Graph Networks

    Tobias Pfaff, Meire Fortunato, Alvaro Sanchez-Gonzalez, and Peter W Battaglia. Learning mesh-based simulation with graph networks.arXiv preprint arXiv:2010.03409, 2020

    work page Pith review arXiv 2010

  37. [37]

    Learning CO2 plume migration in faulted reservoirs with graph neural networks.Computers & Geosciences, 193:105711, 2024

    Xin Ju, Fran¸ cois P Hamon, Gege Wen, Rayan Kanfar, Mauricio Araya-Polo, and Hamdi A Tchelepi. Learning CO2 plume migration in faulted reservoirs with graph neural networks.Computers & Geosciences, 193:105711, 2024

    2024

  38. [38]

    Neural operator: Learning maps between function spaces with applications to pdes.Journal of Machine Learning Research, 24(89):1–97, 2023

    Nikola Kovachki, Zongyi Li, Burigede Liu, Kamyar Azizzadenesheli, Kaushik Bhattacharya, Andrew Stuart, and Anima Anandkumar. Neural operator: Learning maps between function spaces with applications to pdes.Journal of Machine Learning Research, 24(89):1–97, 2023

    2023

  39. [39]

    U-FNO–an enhanced fourier neural operator-based deep-learning model for multiphase flow.Advances in Water Resources, 163:104180, 2022

    Gege Wen, Zongyi Li, Kamyar Azizzadenesheli, Anima Anandkumar, and Sally M Benson. U-FNO–an enhanced fourier neural operator-based deep-learning model for multiphase flow.Advances in Water Resources, 163:104180, 2022

    2022

  40. [40]

    A digital twin for geological carbon storage with controlled injectivity.arXiv preprint arXiv:2403.19819, 2024

    Abhinav Prakash Gahlot, Haoyun Li, Ziyi Yin, Rafael Orozco, and Felix J Herrmann. A digital twin for geological carbon storage with controlled injectivity.arXiv preprint arXiv:2403.19819, 2024

    work page arXiv 2024

  41. [41]

    Darsia: An open-source python toolbox for two-scale image processing of dynamics in porous media.Transport in Porous Media, 151(5):939–973, 2024

    Jan Martin Nordbotten, Benyamine Benali, Jakub Wiktor Both, Bergit Brattek˚ as, Erlend Storvik, and Martin A Fernø. Darsia: An open-source python toolbox for two-scale image processing of dynamics in porous media.Transport in Porous Media, 151(5):939–973, 2024

    2024

  42. [42]

    Courier Corporation, 2013

    Jacob Bear.Dynamics of fluids in porous media. Courier Corporation, 2013

    2013

  43. [43]

    Cambridge University Press, 2019

    Knut-Andreas Lie.An introduction to reservoir simulation using MATLAB/GNU Octave: User guide for the MATLAB Reservoir Simulation Toolbox (MRST). Cambridge University Press, 2019

    2019

  44. [44]

    ECLIPSE technical description (2014.1 ed.)

    Schlumberger Ltd. ECLIPSE technical description (2014.1 ed.). 2014

    2014

  45. [45]

    Three-dimensional simulation of geologic carbon dioxide sequestration using mrst.Advances in Geo- Energy Research, 14(1):34–48, 2024

    Llu ´ ıs Sal´ o-Salgado, Olav Møyner, Knut-Andreas Lie, and Ruben Juanes. Three-dimensional simulation of geologic carbon dioxide sequestration using mrst.Advances in Geo- Energy Research, 14(1):34–48, 2024

    2024

  46. [46]

    MRST-AD–an open-source framework for rapid prototyping and evaluation of reservoir simulation problems

    Stein Krogstad, Knut-Andreas Lie, Olav Møyner, Halvor Møll Nilsen, Xavier Raynaud, and B˚ ard Skaflestad. MRST-AD–an open-source framework for rapid prototyping and evaluation of reservoir simulation problems. InSPE Reservoir Simulation Conference, page D022S002R004. Spe, 2015

    2015

  47. [47]

    Cambridge University Press, 2021

    Knut-Andreas Lie and Olav Møyner.Advanced modelling with the MATLAB reservoir simulation toolbox. Cambridge University Press, 2021

    2021

  48. [48]

    Influence of texture on porosity and permeability of unconsolidated sand.AAPG bulletin, 57(2):349–369, 1973

    DC Beard and PK Weyl. Influence of texture on porosity and permeability of unconsolidated sand.AAPG bulletin, 57(2):349–369, 1973

    1973

  49. [49]

    Thermal conductivity of sands under varying moisture and porosity in drainage–wetting cycles

    Kathleen M Smits, Toshihiro Sakaki, Anuchit Limsuwat, and Tissa H Illangasekare. Thermal conductivity of sands under varying moisture and porosity in drainage–wetting cycles. Vadose Zone Journal, 9(1):172–180, 2010

    2010

  50. [50]

    Investigation of mechanisms of supercritical CO 2 trapping in deep saline reservoirs using surrogate fluids at ambient laboratory conditions

    Luca Trevisan, Abdullah Cihan, Fritjof Fagerlund, Elif Agartan, Hiroko Mori, Jens T Birkholzer, Quanlin Zhou, and Tissa H Illangasekare. Investigation of mechanisms of supercritical CO 2 trapping in deep saline reservoirs using surrogate fluids at ambient laboratory conditions. International Journal of Greenhouse Gas Control, 29:35–49, 2014

    2014

  51. [51]

    Colorado State University, 1965

    Royal Harvard Brooks.Hydraulic properties of porous media. Colorado State University, 1965

    1965

  52. [52]

    U-net: Convolutional networks for biomedical image segmentation

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U-net: Convolutional networks for biomedical image segmentation. InInternational Conference on Medical image computing and computer-assisted intervention, pages Draft PNAS Nexus, YEAR, Volume XX, Issue x 15 234–241. Springer, 2015

    2015

  53. [53]

    Pot: Python optimal transport.Journal of Machine Learning Research, 22(78):1–8, 2021

    R´ emi Flamary, Nicolas Courty, Alexandre Gramfort, Mokhtar Z Alaya, Aur´ elie Boisbunon, Stanislas Chambon, Laetitia Chapel, Adrien Corenflos, Kilian Fatras, Nemo Fournier, et al. Pot: Python optimal transport.Journal of Machine Learning Research, 22(78):1–8, 2021

    2021

  54. [54]

    emcee: the MCMC hammer

    Daniel Foreman-Mackey, David W Hogg, Dustin Lang, and Jonathan Goodman. emcee: the MCMC hammer. Publications of the Astronomical Society of the Pacific, 125(925):306–312, 2013. Draft 16 PNAS Nexus, YEAR, Volume XX, Issue x Supplementary Information Figure 10Surrogate model predictions of forward multiphase flow dynamics for another unseen test case (diffe...

    2013