pith. sign in

arxiv: 2606.10836 · v1 · pith:RQKQYQX3new · submitted 2026-06-09 · ⚛️ physics.comp-ph · math.OC· physics.geo-ph

Bounding the Null Space: Interval-Based Uncertainty Quantification for Non-Identifiable Groundwater Models

Maximilian Ramgraber , Ksenia Bestuzheva This is my paper

Pith reviewed 2026-06-27 10:57 UTC · model grok-4.3

classification ⚛️ physics.comp-ph math.OCphysics.geo-ph
keywords groundwater modelinguncertainty quantificationoptimization-based bound tighteningnon-identifiable modelsDarcy's lawfinite volume discretizationMcCormick relaxationsinterval methods
0
0 comments X

The pith

Optimization-based bound tightening gives guaranteed outer bounds on uncertain variables in non-identifiable groundwater models without sampling.

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

The paper seeks to establish that uncertainty in groundwater models, where sparse data leave many parameter-state combinations equally plausible, can be quantified directly as tightened intervals rather than by exploring discrete realizations. It does so by casting the problem as an optimization task that extremizes each variable subject to a constraint system encoding Darcy's law, observations, and selected physical restrictions. Finite-volume discretization produces bilinear terms that are relaxed via McCormick envelopes; flow sign prescription and irrotationality constraints are added to restore physical fidelity and enable effective tightening. This approach is shown to work on 1D steady, 2D steady, and 2D transient test cases. A sympathetic reader would care because it replaces the risk of incomplete ensemble exploration with deterministic, conservative bounds that connect directly to the model's null space.

Core claim

Optimization-based bound tightening applied to a finite-volume discretization of Darcy's law, with McCormick relaxations augmented by flow sign prescription and irrotationality constraints, produces guaranteed outer bounds on all uncertain parameters, states, and boundary conditions for non-identifiable groundwater models.

What carries the argument

Optimization-based Bound Tightening (OBBT) that represents uncertainty as intervals and tightens them by solving optimization problems over the discretized physical constraints and data.

If this is right

  • The method supplies conservative bounds on every uncertain variable without any sampling step.
  • It applies unchanged to steady-state and transient problems on Cartesian and hexagonal grids.
  • Flow sign prescription and irrotationality constraints each restore sign coupling and rotational consistency that plain McCormick relaxations lose.
  • The resulting bounds are naturally linked to the null-space structure of the under-determined inverse problem.

Where Pith is reading between the lines

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

  • The same interval formulation could be tested on other under-determined inverse problems such as atmospheric transport or reservoir history matching.
  • The computed outer bounds could serve as a quantitative guide for choosing new observation locations that most reduce the null-space dimension.
  • Scalability on larger three-dimensional grids would depend on the tightness of the chosen relaxations and the cost of the successive optimization solves.

Load-bearing premise

The flow sign prescription and irrotationality constraints, when added to the McCormick relaxations, must keep the resulting bounds both valid and sufficiently tight for the target models.

What would settle it

An admissible solution found by exhaustive sampling or another method that lies outside the interval bounds computed by the OBBT procedure, or a case in which the bounds remain too loose to be useful after the constraints are applied.

Figures

Figures reproduced from arXiv: 2606.10836 by Ksenia Bestuzheva, Maximilian Ramgraber.

Figure 1
Figure 1. Figure 1: (A) Marginal bounds define a (hyper)box in variable space. The hatched area denotes the feasible region. (B) [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: (A) The linear system x1 + x2 = 1 is non-identifiable and thus has an active space and a null space. An SVD reveals the vectors that span both spaces. The null space is aligned with the set of equivalent solutions (black diagonal line), the active space lies orthogonal to it. Moving along the null space dimension v2 increases x1 by the same amount that it decreases x2, or vice versa, preserving the value o… view at source ↗
Figure 3
Figure 3. Figure 3: Schematic illustration of OBBT. (A) Each marginal variable has initial bounds. Together, the feasible region [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: (A) The null manifold (colored surface) of the bilinear product [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Initial (grey) and tightened (blue) interval bounds for hydraulic head (A), Darcy flow (B), recharge (C), and [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Bound-tightening results for the 2D steady-state example. Panel A shows the initial bounds for [PITH_FULL_IMAGE:figures/full_fig_p012_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Conceptual illustration of rotational-flow pathology and its remedies on a 2-by-2 grid with source N and sink [PITH_FULL_IMAGE:figures/full_fig_p014_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Model setup (A), tightened bounds for transmissivity [PITH_FULL_IMAGE:figures/full_fig_p016_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Three different regular grids, with a flow sign described from the top left to the bottom right. Relative to the [PITH_FULL_IMAGE:figures/full_fig_p017_9.png] view at source ↗
read the original abstract

Groundwater models are routinely non-identifiable: sparse subsurface observations leave many combinations of parameters, states, and boundary conditions equally consistent with the available data. Existing uncertainty quantification (UQ) methods address this by exploring a finite set of model realizations, but incomplete exploration can systematically underestimate the true range of admissible solutions. We propose a fundamentally different approach based on Optimization-based Bound Tightening (OBBT), which represents uncertainty directly as intervals and tightens them by extremizing variables over a constraint system encoding physical laws and observations. This yields guaranteed outer bounds on all uncertain variables without sampling, side-stepping the exploration problem entirely. To apply OBBT to groundwater flow, we discretize Darcy's law using a finite-volume scheme and handle the resulting bilinear terms through McCormick relaxations. We show that these relaxations can break the sign coupling between fluxes and head gradients, permitting non-physical rotational flow and failing to provide sufficient information for effective bound tightening. We identify flow sign prescription and irrotationality constraints as effective remedies and characterize their respective strengths and limitations. We demonstrate the framework on three numerical examples - a 1D steady-state model, a 2D steady-state model across four experimental configurations, and a 2D transient model on a hexagonal grid - and discuss computational performance, scalability, and directions for future research. OBBT offers a conservative, deterministic, and physically grounded alternative to ensemble-based UQ, with natural connections to null space theory and data assimilation.

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 / 1 minor

Summary. The manuscript proposes an interval-based uncertainty quantification method for non-identifiable groundwater models via Optimization-based Bound Tightening (OBBT). It discretizes Darcy's law with a finite-volume scheme, relaxes the resulting bilinear terms using McCormick envelopes, and augments the formulation with flow sign prescription and irrotationality constraints to restore physical consistency. The framework is demonstrated on a 1D steady-state model, a 2D steady-state model across four configurations, and a 2D transient model on a hexagonal grid, with the central claim being guaranteed outer bounds on all uncertain variables without sampling.

Significance. If the added constraints produce a valid outer approximation whose optima remain both sound and sufficiently tight, the approach supplies a deterministic, conservative alternative to ensemble UQ that directly encodes the null space of admissible solutions. This would be a useful complement to sampling-based methods in data assimilation contexts, particularly where exhaustive exploration is impractical.

major comments (2)
  1. [Abstract (remedies paragraph)] Abstract (paragraph on remedies): The assertion that flow sign prescription and irrotationality constraints are effective remedies is load-bearing for the guarantee of outer bounds. The manuscript must show that these inequalities are valid for the original nonlinear problem (i.e., every feasible point of the unrelaxed finite-volume system remains feasible after augmentation) and does not merely tighten the relaxation gap at the cost of excluding admissible solutions.
  2. [Numerical examples] Numerical examples section: The demonstrations on the three models report no quantitative metrics of bound tightness (e.g., ratio of obtained interval widths to independently computed ground-truth ranges) or residual relaxation gap. Without such measures or comparison against a reference method, it is not possible to confirm that the remedies leave the bounds both valid and practically useful rather than excessively loose.
minor comments (1)
  1. [Abstract] The abstract mentions characterizing the 'strengths and limitations' of the two remedies; the main text should supply explicit statements of these (e.g., which remedy is more effective on transient vs. steady problems) with supporting numerical evidence.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their thoughtful review and constructive feedback on our manuscript. We address each major comment below, indicating revisions where appropriate to strengthen the presentation of validity and tightness.

read point-by-point responses

  1. Referee: Abstract (paragraph on remedies): The assertion that flow sign prescription and irrotationality constraints are effective remedies is load-bearing for the guarantee of outer bounds. The manuscript must show that these inequalities are valid for the original nonlinear problem (i.e., every feasible point of the unrelaxed finite-volume system remains feasible after augmentation) and does not merely tighten the relaxation gap at the cost of excluding admissible solutions.

    Authors: We agree that an explicit demonstration of validity is required to support the outer-bound guarantee. The original manuscript characterizes the strengths and limitations of these constraints but does not contain a dedicated proof. In the revised version we will insert a new subsection (Section 3.4) that derives the flow-sign and irrotationality inequalities directly from the continuous Darcy and continuity equations, shows that every solution of the unrelaxed finite-volume system satisfies them, and therefore confirms that the augmented relaxation remains a valid outer approximation. This addition will be accompanied by a short remark on the conditions under which each constraint is active. revision: yes

  2. Referee: Numerical examples section: The demonstrations on the three models report no quantitative metrics of bound tightness (e.g., ratio of obtained interval widths to independently computed ground-truth ranges) or residual relaxation gap. Without such measures or comparison against a reference method, it is not possible to confirm that the remedies leave the bounds both valid and practically useful rather than excessively loose.

    Authors: We concur that quantitative assessment of tightness is necessary for readers to judge practical utility. For the 1-D and 2-D steady-state examples, where the admissible set can be enumerated or solved analytically, we will add tables reporting (i) the ratio of our interval widths to the true null-space ranges and (ii) the residual McCormick gap at the obtained optima. For the transient hexagonal-grid case we will report the same gap metric together with a comparison against a modest Monte-Carlo ensemble (10^4 realizations) to illustrate conservatism. These metrics and the associated discussion will be inserted into Section 4. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation self-contained via standard relaxations

full rationale

The paper constructs its OBBT framework from the finite-volume discretization of Darcy's law, McCormick envelopes on bilinear terms, and added flow-sign and irrotationality inequalities drawn directly from the governing physics and observations. No step reduces a claimed prediction or bound to a fitted parameter, self-citation chain, or definitional tautology; the remedies are presented as explicit inequalities whose validity is argued from the original nonlinear model rather than assumed by construction. The central guarantee of outer bounds therefore rests on the soundness of these standard relaxation techniques and domain constraints, which remain independently verifiable outside the paper's fitted values or results.

Axiom & Free-Parameter Ledger

0 free parameters · 3 axioms · 0 invented entities

The framework rests on standard physical and mathematical assumptions rather than new fitted parameters or invented entities; the abstract introduces no ad-hoc constants or new particles.

axioms (3)
  • domain assumption Darcy's law governs steady and transient groundwater flow
    Invoked as the physical constraint system that must be satisfied by all admissible solutions.
  • domain assumption Finite-volume discretization produces a faithful discrete representation of the continuous flow equations
    Basis for converting the PDE system into the algebraic constraints used by OBBT.
  • standard math McCormick envelopes supply valid convex relaxations of bilinear products
    Standard technique invoked to handle the nonlinear terms arising from the discretization.

pith-pipeline@v0.9.1-grok · 5809 in / 1492 out tokens · 40853 ms · 2026-06-27T10:57:21.178288+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

27 extracted references · 3 canonical work pages

  1. [1]

    Journal of hydrology , volume=

    Equifinality, data assimilation, and uncertainty estimation in mechanistic modelling of complex environmental systems using the GLUE methodology , author=. Journal of hydrology , volume=. 2001 , publisher=

    2001

  2. [2]

    Journal of hydrology , volume=

    A manifesto for the equifinality thesis , author=. Journal of hydrology , volume=. 2006 , publisher=

    2006

  3. [3]

    arXiv preprint arXiv:2402.17702 , year=

    The SCIP optimization suite 9.0 , author=. arXiv preprint arXiv:2402.17702 , year=

    work page arXiv

  4. [4]

    Scientific investigations report , number=

    Approaches to highly parameterized inversion: Pilot-point theory, guidelines, and research directions , author=. Scientific investigations report , number=. 2011 , publisher=

    2011

  5. [5]

    Groundwater , volume=

    Decision support modeling: Data assimilation, uncertainty quantification, and strategic abstraction , author=. Groundwater , volume=. 2020 , publisher=

    2020

  6. [6]

    Journal of Global Optimization , volume=

    Three enhancements for optimization-based bound tightening , author=. Journal of Global Optimization , volume=. 2017 , publisher=

    2017

  7. [7]

    Mathematical Programming , volume=

    A solution framework for linear PDE-constrained mixed-integer problems , author=. Mathematical Programming , volume=. 2021 , publisher=

    2021

  8. [8]

    Water resources research , volume=

    Real-time groundwater flow modeling with the ensemble Kalman filter: Joint estimation of states and parameters and the filter inbreeding problem , author=. Water resources research , volume=. 2008 , publisher=

    2008

  9. [9]

    Mathematical Programming Computation , volume=

    Parallelizing the dual revised simplex method , author=. Mathematical Programming Computation , volume=. 2018 , publisher=

    2018

  10. [10]

    Advances in water resources , volume=

    Ground-water models cannot be validated , author=. Advances in water resources , volume=. 1992 , publisher=

    1992

  11. [11]

    Water Resources Research , volume=

    Efficient posterior exploration of a high-dimensional groundwater model from two-stage Markov chain Monte Carlo simulation and polynomial chaos expansion , author=. Water Resources Research , volume=. 2013 , publisher=

    2013

  12. [12]

    Mathematical Programming , pages=

    McCormick envelopes in mixed-integer PDE-constrained optimization , author=. Mathematical Programming , pages=. 2025 , publisher=

    2025

  13. [13]

    Advances in Water Resources , volume=

    On uncertainty quantification in hydrogeology and hydrogeophysics , author=. Advances in Water Resources , volume=. 2017 , publisher=

    2017

  14. [14]

    Mathematical programming , volume=

    Computability of global solutions to factorable nonconvex programs: Part I—Convex underestimating problems , author=. Mathematical programming , volume=. 1976 , publisher=

    1976

  15. [15]

    Water Resources Research , volume=

    A reassessment of the groundwater inverse problem , author=. Water Resources Research , volume=. 1996 , publisher=

    1996

  16. [16]

    Waterlines report series , number=

    Australian groundwater modelling guidelines , author=. Waterlines report series , number=

  17. [17]

    Summary report of the national groundwater modelling uncertainty workshop , volume=

    Groundwater Modelling Uncertainty—Implications for Decision Making , author=. Summary report of the national groundwater modelling uncertainty workshop , volume=

  18. [18]

    2009 , publisher=

    Introduction to interval analysis , author=. 2009 , publisher=

    2009

  19. [19]

    PyMC: A Modern, and Comprehensive Probabilistic Programming Framework in Python

    Oriol Abril-Pla and Virgile Andreani and Colin Carroll and Larry Dong and Christopher J. Fonnesbeck and Maxim Kochurov and Ravin Kumar and Junpeng Lao and Christian C. Luhmann and Osvaldo A. Martin and Michael Osthege and Ricardo Vieira and Thomas Wiecki and Robert Zinkov , journal =. doi:10.7717/peerj-cs.1516 , year =

    work page doi:10.7717/peerj-cs.1516

  20. [20]

    Water Resources Research , volume=

    Non-Gaussian parameter inference for hydrogeological models using Stein variational gradient descent , author=. Water Resources Research , volume=. 2021 , publisher=

    2021

  21. [21]

    Groundwater , volume =

    Renard, Philippe , title =. Groundwater , volume =. doi:https://doi.org/10.1111/j.1745-6584.2007.00340.x , url =. https://ngwa.onlinelibrary.wiley.com/doi/pdf/10.1111/j.1745-6584.2007.00340.x , abstract =

    work page doi:10.1111/j.1745-6584.2007.00340.x 2007

  22. [22]

    Reviews of Geophysics , volume=

    Beyond classical observations in hydrogeology: The advantages of including exchange flux, temperature, tracer concentration, residence time, and soil moisture observations in groundwater model calibration , author=. Reviews of Geophysics , volume=. 2019 , publisher=

    2019

  23. [23]

    Journal of the ACM (JACM) , volume=

    Smoothed analysis of algorithms: Why the simplex algorithm usually takes polynomial time , author=. Journal of the ACM (JACM) , volume=. 2004 , publisher=

    2004

  24. [24]

    Statistics and Computing , volume=

    Differential evolution Markov chain with snooker updater and fewer chains , author=. Statistics and Computing , volume=. 2008 , publisher=

    2008

  25. [25]

    Water Resources Research , volume=

    Calibration-constrained Monte Carlo analysis of highly parameterized models using subspace techniques , author=. Water Resources Research , volume=. 2009 , publisher=

    2009

  26. [26]

    Nature methods , volume=

    SciPy 1.0: fundamental algorithms for scientific computing in Python , author=. Nature methods , volume=. 2020 , publisher=

    2020

  27. [27]

    Mathematical programming , volume=

    On the implementation of an interior-point filter line-search algorithm for large-scale nonlinear programming , author=. Mathematical programming , volume=. 2006 , publisher=

    2006