arxiv: 2604.18598 · v1 · submitted 2026-04-09 · 📊 stat.AP · cs.CE· cs.NA· math.NA

Recognition: 2 theorem links

· Lean Theorem

Bathymetry Reconstruction by Bayesian Inference

Lars Stietz , Sebastian G\"otschel , Peter Schleper , Daniel Ruprecht

Authors on Pith no claims yet

Pith reviewed 2026-05-10 18:06 UTC · model grok-4.3

classification 📊 stat.AP cs.CEcs.NAmath.NA

keywords Bayesian inferencebathymetry reconstructionwater height measurementsuncertainty quantificationadjoint optimizationwave flume experimentsshallow water modelinverse problem

0 comments

The pith

Bayesian inference reconstructs bathymetry from water height measurements with lower error than adjoint optimization while quantifying uncertainty.

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

The paper presents a Bayesian inference method to recover bathymetry from sparse point measurements of water height. It evaluates the approach first on synthetic data using both parameterized and discretized bathymetry representations. On experimental data from a wave flume, the Bayesian reconstructions achieve lower normalized root mean squared error than an adjoint optimization baseline, show improved qualitative agreement with the true shape, and supply posterior uncertainty estimates. Accurate bathymetry supports oceanographic modeling and environmental monitoring, so the ability to obtain both a reconstruction and its uncertainty matters for downstream predictions.

Core claim

The authors demonstrate that a Bayesian framework, which combines a forward model mapping bathymetry to water height with appropriate priors, yields robust reconstructions from point measurements. Tests on synthetic cases establish performance for both low-dimensional parameterized bathymetries and higher-dimensional discretized ones. Application to real wave-flume measurements shows the Bayesian posterior mean improves normalized root mean squared error relative to adjoint optimization, produces better visual fidelity, and supplies uncertainty quantification absent from the deterministic baseline.

What carries the argument

Bayesian posterior inference over bathymetry parameters or fields, conditioned on water-height observations through a forward model that solves the shallow-water equations or equivalent.

If this is right

The method supplies not only a point estimate but also credible intervals on the reconstructed bathymetry.
Reconstruction quality remains usable even when the bathymetry is represented on a fine spatial grid.
The same framework can be applied directly to new measurement campaigns without retraining an optimizer.
Uncertainty maps can highlight regions where additional measurements would most reduce ambiguity.

Where Pith is reading between the lines

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

The approach could be combined with real-time sensor networks to update bathymetry estimates as new water-height data arrive.
Similar Bayesian formulations might apply to other inverse problems in shallow-water flows where only surface observations are available.
If the forward model is replaced by a more detailed physics solver, the same inference structure would still produce uncertainty-aware reconstructions.

Load-bearing premise

The forward model that predicts water height from a given bathymetry is accurate enough that errors in the model do not dominate the reconstruction.

What would settle it

On the same wave-flume experimental data set, if the Bayesian reconstruction yields a higher normalized root mean squared error or visibly worse qualitative match than the adjoint optimization result, the claim of improvement would be falsified.

Figures

Figures reproduced from arXiv: 2604.18598 by Daniel Ruprecht, Lars Stietz, Peter Schleper, Sebastian G\"otschel.

**Figure 2.** Figure 2: Difference between the reconstructed and simulated position (left) and width [PITH_FULL_IMAGE:figures/full_fig_p011_2.png] view at source ↗

**Figure 3.** Figure 3: Difference between the reconstructed and simulated position (left) and width [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗

**Figure 5.** Figure 5: The Gaussian prior reduces the probability of the spurious local [PITH_FULL_IMAGE:figures/full_fig_p012_5.png] view at source ↗

**Figure 4.** Figure 4: The non-normalized log-posterior distribution with uniform priors on the po [PITH_FULL_IMAGE:figures/full_fig_p013_4.png] view at source ↗

**Figure 5.** Figure 5: The non-normalized log-posterior distribution with prior on position [PITH_FULL_IMAGE:figures/full_fig_p014_5.png] view at source ↗

**Figure 6.** Figure 6: Sampled mean parameterized Gaussian bathymetry and 95% credible interval [PITH_FULL_IMAGE:figures/full_fig_p015_6.png] view at source ↗

**Figure 7.** Figure 7: The orange line shows the mean of the MCMC samples after burnin using a [PITH_FULL_IMAGE:figures/full_fig_p018_7.png] view at source ↗

**Figure 8.** Figure 8: The orange graph shows the mean of the MCMC samples after burnin using [PITH_FULL_IMAGE:figures/full_fig_p019_8.png] view at source ↗

read the original abstract

Bathymetry reconstruction is an important problem in various fields, including oceanography and environmental monitoring. This paper presents a Bayesian inference approach to reconstructing bathymetries from point measurements of the water height. We test the method for parameterized and discretized bathymetries with synthetic data to evaluate its performance and limitations. Our results indicate that the Bayesian framework provides a robust approach to bathymetry reconstruction. Finally, we use the framework to reconstruct a real-world bathymetry in a wave flume from experimental measurements and compare its performance to an adjoint optimization method. The Bayesian approach improves the normalized root mean squared error (NRMSE) of the reconstruction and provides better qualitative features, while also quantifying uncertainty.

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.

Desk Editor's Note private letter to a colleague

The paper gives a Bayesian workflow for bathymetry from water height points that beats adjoint optimization on one wave-flume case while adding uncertainty, but the gains rest on forward-model accuracy that gets limited real-data checking.

read the letter

The one thing to know is that this paper develops a Bayesian inference method for bathymetry reconstruction using water height point measurements. It validates on synthetic parameterized and discretized bathymetries, then tests on real wave flume data against an adjoint method, claiming better NRMSE and qualitative match plus uncertainty. What the paper does well is lay out a complete workflow and provide that direct comparison on experimental data. The synthetic tests are good for understanding performance limits, and adding uncertainty quantification is a clear advantage over deterministic optimization approaches like adjoint. Where it is softer is in the reliance on the forward model being accurate. The stress-test note is on point here: without explicit checks for model discrepancy or more than one real case, the reported improvements and uncertainty estimates could be overstated if the model misses turbulence, boundaries, or calibration offsets. The abstract mentions improved results but lacks details on prior choices, sampling method, or calibration of the uncertainty, which makes it harder to assess how solid the claims are. For a reader working on similar inverse problems in geophysics or fluid dynamics, this could be a useful reference for how to set up Bayesian reconstruction in this domain. It engages honestly with the literature on inverse problems by building on established Bayesian tools for this application. I think it deserves peer review. The empirical results on real data give it enough substance for referees to evaluate, though they will likely ask for expanded validation.

Referee Report

2 major / 2 minor

Summary. The manuscript presents a Bayesian inference approach for reconstructing bathymetry from point measurements of water height. It evaluates the method on synthetic data for both parameterized and discretized bathymetries, then applies it to experimental measurements from a single wave-flume setup, claiming improved normalized root mean squared error (NRMSE) relative to an adjoint optimization baseline along with the ability to quantify uncertainty.

Significance. If the forward model is shown to be sufficiently accurate and the NRMSE gains prove robust to model discrepancy, the work would provide a useful probabilistic framework for bathymetry reconstruction that also delivers uncertainty estimates, which is relevant for oceanography and environmental monitoring applications.

major comments (2)

[Abstract] Abstract: the central robustness claim rests on the forward model (bathymetry to water-height mapping) being sufficiently accurate for both synthetic and real cases, yet no model-error analysis, posterior predictive checks on synthetic ground truth, or sensitivity tests to unmodeled effects (turbulence, boundary conditions, sensor offsets) are described; this is load-bearing because any shared model discrepancy could artifactually inflate the reported NRMSE improvement and uncertainty quantification.
[Abstract] Abstract and real-data section: only a single wave-flume experiment is used for the adjoint comparison, with no explicit statement of data exclusion rules, cross-validation strategy, or generalization across multiple real cases; this limits the strength of the claim that the Bayesian method provides better qualitative features and lower NRMSE.

minor comments (2)

Clarify the exact form of the likelihood and prior used in the Bayesian formulation, including any hyperparameters and how they are chosen or marginalized.
Add a brief discussion of computational cost and scalability for the discretized bathymetry case relative to the adjoint method.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for their detailed and constructive comments on our manuscript. We address the major comments point by point below and outline the revisions we intend to make to strengthen the paper.

read point-by-point responses

Referee: [Abstract] Abstract: the central robustness claim rests on the forward model (bathymetry to water-height mapping) being sufficiently accurate for both synthetic and real cases, yet no model-error analysis, posterior predictive checks on synthetic ground truth, or sensitivity tests to unmodeled effects (turbulence, boundary conditions, sensor offsets) are described; this is load-bearing because any shared model discrepancy could artifactually inflate the reported NRMSE improvement and uncertainty quantification.

Authors: We agree that the absence of explicit model-error analysis and posterior predictive checks limits the strength of the robustness claims. The synthetic experiments do involve comparison to known ground truth bathymetries, providing some validation of the overall pipeline. In the revised version, we will add posterior predictive checks for the synthetic cases and a sensitivity analysis to unmodeled effects such as sensor offsets and boundary conditions. This will help confirm that the reported NRMSE improvements are not artifacts of model discrepancy. revision: yes
Referee: [Abstract] Abstract and real-data section: only a single wave-flume experiment is used for the adjoint comparison, with no explicit statement of data exclusion rules, cross-validation strategy, or generalization across multiple real cases; this limits the strength of the claim that the Bayesian method provides better qualitative features and lower NRMSE.

Authors: We acknowledge that the real-data evaluation is based on a single wave-flume experiment. In the revision, we will include a clear description of the data used, including any exclusion rules or preprocessing steps. Cross-validation and testing on multiple independent real cases would provide stronger evidence of generalization; however, we currently lack access to such additional datasets. The synthetic results across parameterized and discretized bathymetries offer supporting evidence for the method's applicability. We will revise the abstract and real-data section to more precisely state the scope of the claims. revision: partial

standing simulated objections not resolved

We do not have additional real-world experimental datasets beyond the single wave-flume setup to perform cross-validation or demonstrate generalization across multiple cases.

Circularity Check

0 steps flagged

No significant circularity; Bayesian reconstruction relies on independent physics forward model and external measurements

full rationale

The derivation applies standard Bayesian inference with a forward model (bathymetry to water-height mapping) grounded in physical equations, not constructed from the inference target itself. Synthetic test cases generate data from known ground-truth bathymetries via the same forward model for validation only, while the real wave-flume experiment uses independent experimental measurements compared against an adjoint optimization baseline. No load-bearing step reduces by definition to a fitted parameter, self-citation chain, or renamed input; uncertainty quantification and NRMSE gains are assessed against these external benchmarks, rendering the chain self-contained.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review provides no details on specific priors, likelihood models, discretization choices, or forward solvers; therefore no free parameters, axioms, or invented entities can be identified.

pith-pipeline@v0.9.0 · 5419 in / 1030 out tokens · 32212 ms · 2026-05-10T18:06:51.881982+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

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

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

We use a Markov chain Monte Carlo (MCMC) method to sample from the posterior distribution... likelihood function... prior distributions, particularly Gaussian and Cauchy... shallow water equations (SWEs)
IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

NRMSE... relative L2 and relative L∞ error... 95% credible interval

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

30 extracted references · 24 canonical work pages · 1 internal anchor

[1]

Behrens, A

J. Behrens, A. Androsov, A. Babeyko, S. Harig, F. Klaschka, L. Men- trup, A new multi-sensor approach to simulation assisted tsunami early warning, Natural Hazards and Earth System Sciences 10 (6) (2010) 1085–1100.doi:10.5194/nhess-10-1085-2010

work page doi:10.5194/nhess-10-1085-2010 2010
[2]

R.A.Dalrymple, J.H.MacMahan, A.J.Reniers, V.Nelko, Ripcurrents, Annual Review of Fluid Mechanics 43 (1) (2011) 551.doi:10.1146/ annurev-fluid-122109-160733

2011
[3]

J. Neal, L. Hawker, J. Savage, M. Durand, P. Bates, C. Sampson, Es- timating river channel bathymetry in large scale flood inundation mod- els, Water Resources Research 57 (5) (2021) e2020WR028301.doi: 10.1029/2020WR028301

work page doi:10.1029/2020wr028301 2021
[4]

J. He, S. Zhang, X. Cui, W. Feng, Remote sensing for shallow bathy- metry: A systematic review, Earth-Science Reviews 258 (2024) 104957. doi:10.1016/j.earscirev.2024.104957

work page doi:10.1016/j.earscirev.2024.104957 2024
[5]

Collins, K

A. Collins, K. Brodie, A. S. Bak, T. Hesser, M. Farthing, J. Lee, J. Long, Bathymetric Inversion and Uncertainty Estimation from Synthetic Surf- Zone Imagery with Machine Learning, Remote Sensing 12 (20) (2020) 3364.doi:10.3390/rs12203364

work page doi:10.3390/rs12203364 2020
[6]

E. M. Oades, R. P. Mulligan, M. L. Palmsten, Evaluation of nearshore bathymetric inversion algorithms using camera observations and syn- thetic numerical input of surface waves during storms, Coastal Engi- neering 184 (2023) 104338.doi:10.1016/j.coastaleng.2023.104338

work page doi:10.1016/j.coastaleng.2023.104338 2023
[7]

Vasan, B

V. Vasan, B. Deconinck, The inverse water wave problem of bathymetry detection, Journal of Fluid Mechanics 714 (2013) 562–590.doi:10. 1017/jfm.2012.497. 20

2013
[8]

Angel, J

J. Angel, J. Behrens, S. Götschel, M. Hollm, D. Ruprecht, R. Seifried, Bathymetry reconstruction from experimental data using pde-constrained optimisation, Computers & Fluids 278 (2024) 106321. doi:10.1016/j.compfluid.2024.106321

work page doi:10.1016/j.compfluid.2024.106321 2024
[9]

(2025) Integrated Information Theory: A Consciousness-First Approach to What Exists.https://doi.org/10.48550/arXiv.2510

L. Noureddine, C. Rosier, A Direct Approach for Detection of Bot- tom Topography in Shallow Water (2025).doi:10.48550/arXiv.2510. 03505

work page doi:10.48550/arxiv.2510 2025
[10]

Kaipio, E

J. Kaipio, E. Somersalo, Statistical and Computational Inverse Prob- lems, no. v. 160 in Applied Mathematical Sciences, Springer, 2005. doi:10.1007/b138659

work page doi:10.1007/b138659 2005
[11]

Sellier, Inverse problems in free surface flows: A review, Acta Me- chanica 227 (3) (2016) 913–935.doi:10.1007/s00707-015-1477-1

M. Sellier, Inverse problems in free surface flows: A review, Acta Me- chanica 227 (3) (2016) 913–935.doi:10.1007/s00707-015-1477-1

work page doi:10.1007/s00707-015-1477-1 2016
[12]

A. M. Stuart, Inverse problems: A Bayesian perspective, Acta Numerica 19 (2010) 451–559.doi:10.1017/S0962492910000061

work page doi:10.1017/s0962492910000061 2010
[14]

J. Wu, X. Hao, T. Li, L. Shen, Adjoint-based high-order spectral method of wave simulation for coastal bathymetry reconstruction, Journal of Fluid Mechanics 972 (2023) A41.doi:10.1017/jfm.2023.733

work page doi:10.1017/jfm.2023.733 2023
[15]

Enabling and scaling the HPCG benchmark on the newest generation sunway supercomputer with 42 million heterogeneous cores,

L. Seelinger, A. Reinarz, L. Rannabauer, M. Bader, P. Bastian, R. Sche- ichl, High performance uncertainty quantification with parallelized mul- tilevel markov chain monte carlo, in: Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis, SC ’21, 2021.doi:10.1145/3458817.3476150

work page doi:10.1145/3458817.3476150 2021
[16]

Muldoon, C

Lasith Adhikari, Charnelle Bland, Lopamudra Chakravarty, Wenbin Dong, Olaniyi Samuel Iyiola, G. Muldoon, C. Seinen, L. Jenkins, T. Hesser, M. Farthing, Inverting for Near Shore Bathymetry from Sur- face Wave Properties (2016).doi:10.13140/RG.2.2.30230.14409

work page doi:10.13140/rg.2.2.30230.14409 2016
[17]

P.Ramesh, J.-M.Lueckmann, J.Boelts, Á.Tejero-Cantero, D.S.Green- berg, P. J. Gonçalves, J. H. Macke, GATSBI: Generative Adversarial 21 Training for Simulation-Based Inference (2022).doi:10.48550/arXiv. 2203.06481

work page internal anchor Pith review doi:10.48550/arxiv 2022
[18]

2026).doi:10.5281/zenodo.19472936

L.O.Stietz, D.Ruprecht, Code: Bathymetryreconstructionbybayesian inference (Apr. 2026).doi:10.5281/zenodo.19472936

work page doi:10.5281/zenodo.19472936 2026
[19]

Angel, J

J. Angel, J. Behrens, S. Götschel, M. Hollm, D. Ruprecht, R. Seifried, Data artefact: Bathymetry reconstruction from experimental data with pde-constrained optimisation (2024).doi:10.15480/882.9403

work page doi:10.15480/882.9403 2024
[20]

K. J. Burns, G. M. Vasil, J. S. Oishi, D. Lecoanet, B. P. Brown, Dedalus: A flexible framework for numerical simulations with spec- tral methods, Physical Review Research 2 (2) (2020) 023068.doi: 10.1103/PhysRevResearch.2.023068

work page doi:10.1103/physrevresearch.2.023068 2020
[21]

Joergensen, B

J. Kaipio, E. Somersalo, Statistical inverse problems: Discretization, model reduction and inverse crimes, Journal of Computational and Ap- plied Mathematics 198 (2) (2007) 493–504.doi:10.1016/j.cam.2005. 09.027

work page doi:10.1016/j.cam.2005 2007
[22]

H. Ge, K. Xu, Z. Ghahramani, Turing: A language for flexible prob- abilistic inference, in: Proceedings of the Twenty-First International Conference on Artificial Intelligence and Statistics, Vol. 84 of Proceed- ings of Machine Learning Research, 2018, pp. 1682–1690. URLhttps://proceedings.mlr.press/v84/ge18b.html

2018
[23]

T. E. Fjelde, K. Xu, D. Widmann, M. Tarek, C. Pfiffer, M. Trapp, S. D. Axen, X. Sun, M. Hauru, P. Yong, W. Tebbutt, Z. Ghahramani, H. Ge, Turing.jl: A general-purpose probabilistic programming language, ACM Trans. Probab. Mach. Learn. (2025).doi:10.1145/3711897

work page doi:10.1145/3711897 2025
[24]

W., Rosenbluth M

N. Metropolis, A. W. Rosenbluth, M. N. Rosenbluth, A. H. Teller, E. Teller, Equation of state calculations by fast computing machines, The Journal of Chemical Physics 21 (6) (1953) 1087–1092.doi: 10.1063/1.1699114

work page doi:10.1063/1.1699114 1953
[25]

W. K. Hastings, Monte Carlo sampling methods using Markov chains and their applications, Biometrika 57 (1) (1970) 97–109.doi:10.1093/ biomet/57.1.97. 22

1970
[26]

S. L. Cotter, G. O. Roberts, A. M. Stuart, D. White, MCMC Meth- ods for Functions: Modifying Old Algorithms to Make Them Faster, Statistical Science 28 (3) (2013) 424–446.doi:10.1214/13-STS421

work page doi:10.1214/13-sts421 2013
[27]

Duane, A

S. Duane, A. D. Kennedy, B. J. Pendleton, D. Roweth, Hybrid Monte Carlo, Physics Letters B 195 (2) (1987) 216–222.doi:10.1016/ 0370-2693(87)91197-X

1987
[28]

M. D. Hoffman, A. Gelman, et al., The no-u-turn sampler: adaptively setting path lengths in hamiltonian monte carlo., J. Mach. Learn. Res. 15 (1) (2014) 1593–1623.doi:10.48550/arXiv.1111.4246

work page Pith review doi:10.48550/arxiv.1111.4246 2014
[29]

G. O. Roberts, R. L. Tweedie, Exponential convergence of Langevin distributions and their discrete approximations, Bernoulli 2 (4) (1996) 341–363

1996
[30]

G. O. Roberts, J. S. Rosenthal, Optimal scaling for various Metropolis- Hastings algorithms, Statistical Science 16 (4) (2001).doi:10.1214/ ss/1015346320

work page arXiv 2001
[31]

Bhagavan, B

S. Bhagavan, B. de Koning, S. Maddhashiya, C. Rackauckas, Datain- terpolations.jl: Fast interpolations of 1d data, Journal of Open Source Software 9 (101) (2024) 6917.doi:10.21105/joss.06917. 23

work page doi:10.21105/joss.06917 2024