pith. sign in

arxiv: 2509.12861 · v3 · submitted 2025-09-16 · ⚛️ physics.flu-dyn

Augmenting a pure and hybrid vertical equilibrium scheme via data-driven surrogate modelling

Ivan Buntic , Bernd Flemisch This is my paper

Pith reviewed 2026-05-18 16:37 UTC · model grok-4.3

classification ⚛️ physics.flu-dyn
keywords vertical equilibriumdata-driven surrogateshybrid modelingporous media flowgas plume distancemass conservationcomputational efficiency
0
0 comments X

The pith

Data-driven surrogates augment hybrid vertical equilibrium schemes to cut simulation times with negligible errors

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

The paper aims to speed up hybrid vertical equilibrium models for fluid flow in porous media by replacing expensive parts of the calculation with data-driven surrogate models. These surrogates predict the gas plume distance and coarse-level mobilities inside the VE region and also accelerate the interface quantities that couple the VE model to full-dimensional regions. A sympathetic reader would care because the coupling step often makes hybrid methods slower than pure traditional simulations, so removing that overhead could deliver both the accuracy of full models where needed and the speed of VE models elsewhere. The approach keeps mass conservation intact and reports only small overall errors.

Core claim

By training and deploying surrogate models to predict gas plume distance, coarse-level mobilities in the VE zone, and the coupling-scheme quantities at model interfaces, the augmented hybrid scheme achieves substantially lower runtimes than traditional mass-and-momentum simulations while introducing only negligible errors and preserving physical properties such as mass conservation.

What carries the argument

Data-driven surrogate models that replace direct computation of gas plume distance, coarse-level mobilities, and coupling terms inside the hybrid vertical equilibrium framework.

Load-bearing premise

The surrogate models must accurately predict gas plume distance, coarse-level mobilities, and coupling quantities over the relevant parameter space without violating physical constraints such as mass conservation.

What would settle it

A benchmark run in which the surrogate-augmented hybrid model either produces a mass-conservation error larger than the non-surrogate hybrid or fails to run faster than a traditional simulation would falsify the performance claim.

read the original abstract

Vertical equilibrium (VE) models have been introduced as computationally efficient alternatives to traditional mass and momentum balance equations for fluid flow in porous media. Since VE models are only accurate in regions where phase equilibrium holds and traditional simulations are computationally demanding, hybrid methods have been proposed to combine the accuracy of the full-dimensional approach with the efficiency of VE model. However, coupling both models introduces computational overhead that can make hybrid simulations slower than fully traditional ones. To address the computational overhead introduced by coupling interfaces in hybrid models, we utilize data-driven surrogates to accelerate the overall scheme. To this end, we predict the gas plume distance and coarse-level mobilities in the VE model, and also enhance the computation of the coupling scheme via surrogates. We focus on surrogate models with short inference times to minimize computational overhead during frequent function calls. The proposed approach preserves key physical properties, such as mass conservation, despite the deployment of data-driven models, while substantially reducing simulation runtimes. Overall, combining data-driven methods with the hybrid VE scheme yields an enhanced model that outperforms traditional simulations in speed while introducing only negligible errors.

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 manuscript proposes augmenting pure vertical equilibrium (VE) and hybrid VE/full-dimensional schemes for porous-media flow with data-driven surrogate models. Surrogates are trained to predict gas-plume distance and coarse-level mobilities inside VE regions and to accelerate the computation of coupling quantities at VE–full-dimensional interfaces. The authors claim that the resulting scheme retains mass conservation, produces only negligible errors relative to the unsimplified hybrid model, and delivers substantial runtime reductions compared with traditional full-dimensional simulations.

Significance. If the conservation and accuracy claims are rigorously verified, the work would demonstrate a practical route to mitigating the coupling overhead that currently limits hybrid VE models, thereby extending the applicability of reduced-order methods to larger-scale reservoir simulations. The emphasis on short-inference-time surrogates is well-aligned with the requirements of explicit time-stepping schemes.

major comments (2)
  1. [§4.2 and §5.1] §4.2 and §5.1: The central claim that mass conservation is preserved rests on the assertion that the surrogates for coarse mobilities and interface quantities reproduce the discrete divergence-free property of the underlying finite-volume scheme. No description is given of a projection step, physics-informed loss term, or post-processing correction that would enforce this algebraic identity; standard regression surrogates minimize pointwise error but do not automatically satisfy the local flux-balance condition required at coupling interfaces. Without such a mechanism, accumulated mass drift over hundreds of time steps would invalidate the performance claims.
  2. [Table 2 and Figure 7] Table 2 and Figure 7: The reported error metrics and runtime speed-ups are presented without baseline comparisons to the original hybrid VE scheme (i.e., without surrogates) or to fully traditional simulations on the same grids. The absence of these controls makes it impossible to quantify whether the observed errors remain negligible once the surrogate-induced perturbations propagate through the coupled system.
minor comments (2)
  1. [§3.1] Notation for the surrogate inputs (e.g., the precise definition of the coarse mobility vector and the interface flux vector) is introduced inconsistently between §3.1 and the appendix; a single consolidated table of symbols would improve readability.
  2. [§4.1] The training-data generation procedure (parameter ranges, number of full-order simulations, and sampling strategy) is described only qualitatively; explicit ranges and sample counts should be stated so that reproducibility can be assessed.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We are grateful to the referee for the thorough review and valuable feedback on our manuscript. The comments have helped us identify areas where additional clarifications and comparisons can strengthen the presentation. Below, we address each major comment in detail.

read point-by-point responses

  1. Referee: [§4.2 and §5.1] §4.2 and §5.1: The central claim that mass conservation is preserved rests on the assertion that the surrogates for coarse mobilities and interface quantities reproduce the discrete divergence-free property of the underlying finite-volume scheme. No description is given of a projection step, physics-informed loss term, or post-processing correction that would enforce this algebraic identity; standard regression surrogates minimize pointwise error but do not automatically satisfy the local flux-balance condition required at coupling interfaces. Without such a mechanism, accumulated mass drift over hundreds of time steps would invalidate the performance claims.

    Authors: We appreciate the referee pointing out the need for explicit justification of mass conservation in the surrogate-augmented scheme. The surrogates are trained directly on data obtained from the original finite-volume discretization, which satisfies the discrete divergence-free condition by construction. The predicted coarse mobilities and interface quantities are therefore approximations to quantities that already fulfill the local flux balance. Our numerical results indicate that the approximation errors are sufficiently small that no significant mass drift occurs over the simulated time horizons. To strengthen the manuscript, we will add a dedicated paragraph in §4.2 explaining this training-based preservation of the algebraic property and include a supplementary figure demonstrating the evolution of global mass error for the surrogate model. revision: partial

  2. Referee: [Table 2 and Figure 7] Table 2 and Figure 7: The reported error metrics and runtime speed-ups are presented without baseline comparisons to the original hybrid VE scheme (i.e., without surrogates) or to fully traditional simulations on the same grids. The absence of these controls makes it impossible to quantify whether the observed errors remain negligible once the surrogate-induced perturbations propagate through the coupled system.

    Authors: We agree that including baseline comparisons is important for a complete assessment. The current Table 2 and Figure 7 primarily contrast the surrogate-enhanced hybrid model with traditional full-dimensional simulations to emphasize the achieved speed-up. We will revise these to also report results from the original hybrid VE scheme without surrogates. This will allow readers to evaluate the surrogate-induced errors relative to the unsimplified hybrid model and confirm that they remain negligible while still providing substantial runtime improvements over both the baseline hybrid and the full-dimensional approaches. revision: yes

Circularity Check

0 steps flagged

Data-driven surrogates trained independently on simulation outputs accelerate hybrid VE without reducing claims to self-definition or fitted inputs by construction.

full rationale

The paper trains surrogate models on data generated from the underlying VE and hybrid schemes to predict plume distance, coarse mobilities, and interface quantities. These predictions are evaluated empirically against full simulations for speed and error, with mass conservation preserved through the overall scheme structure rather than by redefining the target quantities in terms of the surrogates themselves. No load-bearing step equates a derived result to its own training inputs or relies on a self-citation chain that assumes the target outcome. The approach remains self-contained against external numerical benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Based solely on the abstract, the central claim rests on the premise that trained surrogate models can substitute for traditional calculations while preserving mass conservation. No explicit free parameters, axioms, or invented entities are described. The approach implicitly assumes that the training data distribution covers the operating regimes of interest and that short-inference surrogates do not accumulate errors that break physical invariants.

pith-pipeline@v0.9.0 · 5722 in / 1230 out tokens · 53074 ms · 2026-05-18T16:37:25.213301+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

39 extracted references · 39 canonical work pages

  1. [1]

    Technical report, Gas Infrastructure Europe (2021)

    Europe, G.I.: Picturing the value of underground gas storage to the euro- pean hydrogen system. Technical report, Gas Infrastructure Europe (2021). https://www.gie.eu/wp-content/uploads/filr/3517/Picturing%20 the%20value%20of%20gas%20storage%20to%20the%20 FINAL 140621.pdf

    work page 2021

  2. [2]

    Energy Reports 9, 6251– 6266 (2023) https://doi.org/10.1016/j.egyr.2023.05.236

    Al-Shafi, M., Massarweh, O., Abushaikha, A.S., Bicer, Y.: A review on underground gas storage systems: Natural gas, hydrogen and carbon sequestr ation. Energy Reports 9, 6251– 6266 (2023) https://doi.org/10.1016/j.egyr.2023.05.236

    work page doi:10.1016/j.egyr.2023.05.236 2023

  3. [3]

    Energy Storage M aterials 63, 103045 (2023) https://doi.org/10.1016/j.ensm.2023.103045

    Liu, W., Li, Q., Yang, C., Shi, X., Wan, J., Jurado, M.J., Li, Y., Jiang, D., C hen, J., Qiao, W., Zhang, X., Fan, J., Peng, T., He, Y.: The role of underground salt c averns for large- scale energy storage: A review and prospects. Energy Storage M aterials 63, 103045 (2023) https://doi.org/10.1016/j.ensm.2023.103045

    work page doi:10.1016/j.ensm.2023.103045 2023

  4. [4]

    Internatio nal Journal of Coal Science & Technology 11(1), 29 (2024) https://doi.org/10.1007/s40789-024-00676-y

    Wan, J., Sun, Y., He, Y., Ji, W., Li, J., Jiang, L., Jurado, M.J.: Developm ent and technology status of energy storage in depleted gas reservoirs. Internatio nal Journal of Coal Science & Technology 11(1), 29 (2024) https://doi.org/10.1007/s40789-024-00676-y

    work page doi:10.1007/s40789-024-00676-y 2024

  5. [5]

    In: SPE Energy Tr ansition Symposium

    Aluah, R., Fadairo, A., Opeyemi, O., Ni, R., Foerster, I.: An experime ntal study on the caprock integrity of reservoirs to assess the repurposing deplet ed bakken formation oil and gas fields for underground hydrogen storage. In: SPE Energy Tr ansition Symposium. SPE Energy Transition Symposium, vol. SPE Energy Transition Symposium , pp. 012–002002 (2024)...

    work page doi:10.2118/221398-ms 2024

  6. [6]

    Natural Gas Industry B 9(4), 383–393 (2022) https://doi.org/10.1016/j.ngib.2022.07.002

    Luo, A., Li, Y., Chen, X., Zhu, Z., Peng, Y.: Review of co2 sequestra tion mechanism in saline aquifers. Natural Gas Industry B 9(4), 383–393 (2022) https://doi.org/10.1016/j.ngib.2022.07.002

    work page doi:10.1016/j.ngib.2022.07.002 2022

  7. [7]

    Water Resources Research 51(9), 6846–6892 (2015) https://doi.org/10.1002/2015WR017609 https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1002/2015WR017609

    Celia, M.A., Bachu, S., Nordbotten, J.M., Bandilla, K.W.: Status of co2 storage in deep saline aquifers with emphasis on modeling approaches and pract ical simulations. Water Resources Research 51(9), 6846–6892 (2015) https://doi.org/10.1002/2015WR017609 https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1002/2015WR017609

    work page doi:10.1002/2015wr017609 2015

  8. [8]

    Das, M., Mukherjee, P., Muralidhar, K.: Modeling Transport Pheno mena in Porous Media with Applications, p. 241. Springer, ??? (2018). https://doi.org/10.1007/978-3-319-69866-3

    work page doi:10.1007/978-3-319-69866-3 2018

  9. [9]

    Water Resources Research 47(5) (2011) https://doi.org/10.1029/2010WR009075

    Gasda, S.E., Nordbotten, J.M., Celia, M.A.: Vertically averaged appr oaches for co2 migration with solubility trapping. Water Resources Research 47(5) (2011) https://doi.org/10.1029/2010WR009075

    work page doi:10.1029/2010wr009075 2011

  10. [10]

    Geological Storage of CO2: Modeling Approaches for Large -Scale Simulation, (2011) https://doi.org/10.1002/9781118137086.fmatter

    Nordbotten, J., Celia, M.: Geological storage of co: Modeling app roaches for large-scale sim- ulation. Geological Storage of CO2: Modeling Approaches for Large -Scale Simulation, (2011) https://doi.org/10.1002/9781118137086.fmatter

    work page doi:10.1002/9781118137086.fmatter 2011

  11. [11]

    International Journal of Hydrogen Energy 106, 790–805 (2025) https://doi.org/10.1016/j.ijhydene.2025.01.201

    Vahabzadeh, E., Buntic, I., Nazari, F., Flemisch, B., Helmig, R., Nias ar, V.: Applicability of the vertical equilibrium model to underground hydrog en injection and withdrawal. International Journal of Hydrogen Energy 106, 790–805 (2025) https://doi.org/10.1016/j.ijhydene.2025.01.201

    work page doi:10.1016/j.ijhydene.2025.01.201 2025

  12. [12]

    Water Resources Research 58(1), 2021–030990 (2022) https://doi.org/10.1029/2021WR030990 https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1029/2021WR030990

    Becker, B., Guo, B., Buntic, I., Flemisch, B., Helmig, R.: An adaptive hybrid vertical equilibrium/full-dimensional model for compositional multiph ase flow. Water Resources Research 58(1), 2021–030990 (2022) https://doi.org/10.1029/2021WR030990 https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1029/2021WR030990. e2021WR030990 2021WR030990

    work page doi:10.1029/2021wr030990 2021

  13. [13]

    Computational Geosciences 23(1), 1–20 (2019) https://doi.org/10.1007/s10596-018-9775-z

    Møyner, O., Nilsen, H.M.: Multiresolution coupled vertical equilibrium model for fast flexible simulation of co2 storage. Computational Geosciences 23(1), 1–20 (2019) https://doi.org/10.1007/s10596-018-9775-z

    work page doi:10.1007/s10596-018-9775-z 2019

  14. [14]

    Computational Geos ciences 29(2), 11 (2025) https://doi.org/10.1007/s10596-025-10351-z

    Buntic, I., Schneider, M., Flemisch, B., Helmig, R.: A fully-implicit solvin g approach to an adaptive multi-scale model - coupling a vertical-equilibrium and full-dime nsional model for compressible, multi-phase flow in porous media. Computational Geos ciences 29(2), 11 (2025) https://doi.org/10.1007/s10596-025-10351-z

    work page doi:10.1007/s10596-025-10351-z 2025

  15. [15]

    Computer Metho ds in Applied Mechanics and Engineering 361, 112732 (2020) https://doi.org/10.1016/j.cma.2019.112732

    Sun, L., Gao, H., Pan, S., Wang, J.-X.: Surrogate modeling for fluid flows based on physics- constrained deep learning without simulation data. Computer Metho ds in Applied Mechanics and Engineering 361, 112732 (2020) https://doi.org/10.1016/j.cma.2019.112732

    work page doi:10.1016/j.cma.2019.112732 2020

  16. [16]

    In: AISyS 2024 : The First International Conference on AI-based Systems and Services

    Kavana, S., Kulkarni, K., K¨ ostler, H.: An advanced surrogate m odel approach for enhancing fluid dynamics simulations. In: AISyS 2024 : The First International Conference on AI-based Systems and Services. AISyS 2024, p. 6 (2024)

    work page 2024

  17. [17]

    Computers & Chemica l Engineering 158, 107639 (2022) https://doi.org/10.1016/j.compchemeng.2021.107639

    Wen, H., Khan, F., Amin, M.T., Halim, S.Z.: Myths and misconceptions o f data-driven meth- ods: Applications to process safety analysis. Computers & Chemica l Engineering 158, 107639 (2022) https://doi.org/10.1016/j.compchemeng.2021.107639

    work page doi:10.1016/j.compchemeng.2021.107639 2022

  18. [18]

    Schultzendorff, P.V., Sandve, T.H., Kane, B., Landa-Marb´ an, D ., Both, J.W., Nord- botten, J.M.: A machine-learned near-well model in opm flow 2024(1), 1–23 (2024) https://doi.org/10.3997/2214-4609.202437033 22

    work page doi:10.3997/2214-4609.202437033 2024

  19. [19]

    Water Resources Research 54(7), 4347–4360 (2018) https://doi.org/10.1029/2017WR022303

    Becker, B., Guo, B., Bandilla, K., Celia, M.A., Flemisch, B., Helmig, R.: An adap- tive multiphysics model coupling vertical equilibrium and full multidimens ions for mul- tiphase flow in porous media. Water Resources Research 54(7), 4347–4360 (2018) https://doi.org/10.1029/2017WR022303

    work page doi:10.1029/2017wr022303 2018

  20. [20]

    Hydrology Papers 3 (1964)

    Brooks, R.H., Corey, A.T.: Hydraulic properties of porous media. Hydrology Papers 3 (1964)

    work page 1964

  21. [21]

    In: Zhang, D

    Zhang, D.: 6 - two-phase flow. In: Zhang, D. (ed.) Stochastic Methods for Flow in Porous Media, pp. 262–296. Academic Press, San Diego (2002). https://doi.org/10.1016/B978-012779621-5/50007-3 . https://www.sciencedirect.com/science/article/pii/B9780127796215500073

    work page doi:10.1016/b978-012779621-5/50007-3 2002

  22. [22]

    Disse rtation, Universit¨ at Stuttgart, Institut f¨ ur Wasser- Umweltsystemmodellierung, St uttgart (June 2021)

    Becker, B.: Development of efficient multiscale multiphysics models accounting for reversible flow at various subsurface energy storage sites. Disse rtation, Universit¨ at Stuttgart, Institut f¨ ur Wasser- Umweltsystemmodellierung, St uttgart (June 2021). https://doi.org/10.18419/opus-11753 . http://dx.doi.org/10.18419/opus-11753

    work page doi:10.18419/opus-11753 2021

  23. [23]

    Water Resources Research 47(2) (2011) https://doi.org/10.1029/2009WR008958 https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1029/2009WR008958

    Nordbotten, J.M., Dahle, H.K.: Impact of the capillary fringe in vertically integrated models for co2 storage. Water Resources Research 47(2) (2011) https://doi.org/10.1029/2009WR008958 https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1029/2009WR008958

    work page doi:10.1029/2009wr008958 2011

  24. [24]

    Computers & Geosciences 169, 105207 (2022)

    Wen, G., Li, Z., Azizzadenesheli, K., Anandkumar, A., Durlofsky, L .J.: Fourier neural operator for multiphase flow simulation in carbon sequestration. Computers & Geosciences 169, 105207 (2022)

    work page 2022

  25. [25]

    Journal of Computational Physics 337, 275–288 (2017) https://doi.org/10.1016/j.jcp.2017.02.041

    Voskov, D.V.: Operator-based linearization approach for mode ling of multiphase multi- component flow in porous media. Journal of Computational Physics 337, 275–288 (2017) https://doi.org/10.1016/j.jcp.2017.02.041

    work page doi:10.1016/j.jcp.2017.02.041 2017

  26. [26]

    Journal of Petroleum Science and Engineering 157, 990–998 (2017) https://doi.org/10.1016/j.petrol.2017.08.009

    Khait, M., Voskov, D.V.: Operator-based linearization for gener al purpose reser- voir simulation. Journal of Petroleum Science and Engineering 157, 990–998 (2017) https://doi.org/10.1016/j.petrol.2017.08.009

    work page doi:10.1016/j.petrol.2017.08.009 2017

  27. [27]

    Springer, Berlin, Heidelberg (2006)

    Bishop, C.M.: Pattern Recognition and Machine Learning (Inform ation Science and Statis- tics). Springer, Berlin, Heidelberg (2006)

    work page 2006

  28. [28]

    The MIT Press, ??? (2012)

    Murphy, K.P.: Machine Learning: A Probabilistic Perspective. The MIT Press, ??? (2012)

    work page 2012

  29. [29]

    Springer, New York (1978 )

    Boor, C.: A Practical Guide to Splines. Springer, New York (1978 )

    work page 1978

  30. [30]

    http://www.iapws.org/IF97-Rev.pdf , Lucerne, Switzerland (1997)

    Water, I.T.I.A., Steam): Revised Release on the IAPWS Industria l Formulation 1997 for the Thermodynamic Properties of Water and Steam. http://www.iapws.org/IF97-Rev.pdf , Lucerne, Switzerland (1997)

    work page 1997

  31. [31]

    D aRUS (2023)

    Buntic, I., Coltman, E., Flemisch, B., Ghosh, T., Gl¨ aser, D., Gr¨ un inger, C., Hommel, J., Keim, L., Kelm, M., Koch, T., Kostelecky, A.M., Lipp, M., Oukili, H., Schneide r, M., Utz, M., Wang, Y., Weishaupt, K., Wendel, K., Winter, R., Wu, H.: DuMux 3.8.0. D aRUS (2023). https://doi.org/10.18419/darus-3788

    work page doi:10.18419/darus-3788 2023

  32. [32]

    Computers & Mathemat ics with Applications 23 81, 423–443 (2021) https://doi.org/10.1016/j.camwa.2020.02.012

    Koch, T., Gl¨ aser, D., Weishaupt, K., Ackermann, S., Beck, M., Becker, B., Burbulla, S., Class, H., Coltman, E., Emmert, S., Fetzer, T., Gr¨ uninger, C., Heck, K., Hommel, J., Kurz, T., Lipp, M., Mohammadi, F., Scherrer, S., Schneider, M., Seitz, G., Stadler, L., Utz, M., Weinhardt, F., Flemisch, B.: Dumux 3 – an open-source simulator for solving flow and...

    work page doi:10.1016/j.camwa.2020.02.012 2021

  33. [33]

    Computers & Mathematics with Applications 81 (Jan

    Bastian, P., Blatt, M., Dedner, A., Dreier, N.-A., Engwer, C., Fritz e, R., Gr¨ aser, C., Gr¨ uninger, C., Kempf, D., Kl¨ ofkorn, R., Ohlberger, M., Sander, O.: The dune fra mework: Basic concepts and recent developments. Computers & Mathematics with Applicatio ns 81, 75–112 (2021) https://doi.org/10.1016/j.camwa.2020.06.007

    work page doi:10.1016/j.camwa.2020.06.007 2021

  34. [34]

    Journal of Machine Learning Research 12, 2825–2830 (2011)

    Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., Blondel, M., Prettenhofer, P., Weiss, R., Dubourg, V., Vanderplas, J., Passo s, A., Cournapeau, D., Brucher, M., Perrot, M., Duchesnay, E.: Scikit-learn: Machine learn ing in Python. Journal of Machine Learning Research 12, 2825–2830 (2011)

    work page 2011

  35. [35]

    https://onnxruntime.ai/

    developers, O.R.: ONNX Runtime. https://onnxruntime.ai/. Version: 1.18.0 (2021)

    work page 2021

  36. [36]

    http://www.gnu.org/software/gsl/

    Galassi, M.e.a.: GNU Scientific Library Reference Manual (3rd Ed.) . http://www.gnu.org/software/gsl/

    work page

  37. [37]

    Water Resources Research 52(8), 6490–6505 (2016) https://doi.org/10.1002/2016WR018714

    Guo, B., Bandilla, K.W., Nordbotten, J.M., Celia, M.A., Keilegavlen, E., D oster, F.: A multiscale multilayer vertically integrated model with vertical dynamic s for co2 sequestra- tion in layered geological formations. Water Resources Research 52(8), 6490–6505 (2016) https://doi.org/10.1002/2016WR018714

    work page doi:10.1002/2016wr018714 2016

  38. [38]

    McGraw-Hill Inc., ??? (1987)

    Reid, R.C., Prausnitz, J.M., Poling, B.E.: The Properties of Gases an d Liquids. McGraw-Hill Inc., ??? (1987). https://www.osti.gov/scitech/biblio/6504847

    work page arXiv 1987

  39. [39]

    Journal f¨ ur die reine und angewandte Mathematik 1, 65–84 (1826) 24

    Abel, N.H.: M´ emoire sur les ´ equations alg´ ebriques o` u on d´ emo ntre l’impossibilit´ e de la r´ esolution de l’´ equation g´ en´ erale du cinqui` eme degr´ e. Journal f¨ ur die reine und angewandte Mathematik 1, 65–84 (1826) 24

    work page