pith. sign in

arxiv: 2601.06304 · v2 · pith:2W2LW3PQnew · submitted 2026-01-09 · ⚛️ physics.flu-dyn

Localization of sources in weakly nonlinear fluid systems using linear and quadratic sensitivity analysis

Qi Wang , Zejian You This is my paper

Pith reviewed 2026-05-21 15:14 UTC · model grok-4.3

classification ⚛️ physics.flu-dyn
keywords source localizationadjoint sensitivityquadratic embeddingsnonlinear PDEsinverse problemsfluid dynamicsprincipal angle minimizationweakly nonlinear systems
0
0 comments X

The pith

Quadratic positional embeddings from second-order sensitivity analysis improve source localization in weakly nonlinear fluid systems beyond linear adjoint methods.

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

The paper develops a method to locate sources in dynamical systems governed by nonlinear partial differential equations. It begins with the standard linear adjoint relation between measurements and sources, then adds a quadratic correction represented as a symmetric bilinear operator. This correction is approximated through a truncated eigen-expansion using Krylov subspace iterations, producing quadratic positional embeddings. These embeddings augment the linear field so that measurement data can be projected onto a higher-dimensional hyperplane. A search algorithm then minimizes the principal angle between that hyperplane and the observation vector to identify source positions.

Core claim

We introduce quadratic positional embeddings that augment the linear adjoint field, enabling measurement data to be projected onto a higher-dimensional hyperplane spanned by the linear and quadratic embeddings. A source search algorithm is formulated based on principal angle minimization between this hyperplane and the observation vector, providing a natural probabilistic interpretation of source location. The method operates in a one-shot fashion without iterative updates of candidate source positions, and it can be readily extended to scenarios involving multiple sources. Demonstrations on perturbation-source identification in the viscous Burgers equation and heat-source detection in a two

What carries the argument

Quadratic positional embeddings obtained from a symmetric bilinear operator approximated by truncated eigen-expansion with Krylov subspace iterations, which together with the linear adjoint field span a hyperplane for measurement projection and principal-angle minimization.

Load-bearing premise

The nonlinearity stays weak enough that a quadratic correction via a symmetric bilinear operator is sufficient to capture the effects without introducing large approximation errors.

What would settle it

Perform source localization on the Burgers equation benchmark in a region where the linear adjoint sensitivity is near zero; if adding the quadratic embeddings produces no reduction in localization error, the claimed improvement from the hyperplane projection is false.

Figures

Figures reproduced from arXiv: 2601.06304 by Qi Wang, Zejian You.

Figure 1
Figure 1. Figure 1: FIG. 1. Schematic of the linear and quadratic positional embedding for source localization. [PITH_FULL_IMAGE:figures/full_fig_p010_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. (a) The surface plot of [PITH_FULL_IMAGE:figures/full_fig_p013_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. Taylor test for the accuracy of linear and quadratic embeddings, showing (a) the relative [PITH_FULL_IMAGE:figures/full_fig_p014_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4. (a.i) Predicted probability distribution [PITH_FULL_IMAGE:figures/full_fig_p015_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5. (a-c) Forward temperature field [PITH_FULL_IMAGE:figures/full_fig_p018_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: FIG. 6. (a) The linear embedding fields ˜s [PITH_FULL_IMAGE:figures/full_fig_p019_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: FIG. 7. (a) Forward field [PITH_FULL_IMAGE:figures/full_fig_p020_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: FIG. 8. Probability distribution of reconstructed source location using linear (top) embedding [PITH_FULL_IMAGE:figures/full_fig_p021_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: FIG. 9. Error in the location prediction using linear (top) and quadratic (bottom) position embed [PITH_FULL_IMAGE:figures/full_fig_p022_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: FIG. 10. Reconstruction of two sources using (a) linear positional embedding and (b) quadratic [PITH_FULL_IMAGE:figures/full_fig_p024_10.png] view at source ↗
read the original abstract

We develop a framework for localized source detection in dynamical systems governed by nonlinear partial differential equations based on first and second-order sensitivity analysis. Building on the standard adjoint formulation, which relates multiple measurements to external sources through a linear duality relation, we first introduce a linear positional embedding that identifies the source location by aligning the measurement vector with the embedding. To capture weakly nonlinear effects that arise when the source intensity is finite, we then incorporate a quadratic correction represented as a symmetric bilinear operator and approximated via a truncated eigen-expansion obtained with Krylov subspace iterations. This yields quadratic positional embeddings that augment the linear adjoint field, enabling measurement data to be projected onto a higher-dimensional hyperplane, spanned by the linear and quadratic embeddings. A source search algorithm is formulated based on principal angle minimization between this hyperplane and the observation vector, providing a natural probabilistic interpretation of source location. The method operates in a one-shot fashion without iterative updates of candidate source positions, and it can be readily extended to scenarios involving multiple sources. Demonstrations on benchmark inverse problems include perturbation-source identification in the viscous Burgers equation and heat-source detection in a two-dimensional laminar stratified channel. The results with quadratic embeddings show significant improvements in localization accuracy compared with linear adjoint-based sensitivity methods, especially in the region where linear adjoint sensitivity vanishes.

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 develops a framework for localizing sources in weakly nonlinear fluid systems governed by nonlinear PDEs. It extends the standard adjoint formulation with a linear positional embedding and a quadratic correction represented as a symmetric bilinear operator, which is approximated via a truncated eigen-expansion using Krylov subspace iterations. This produces quadratic positional embeddings that augment the linear adjoint field, allowing measurement data to be projected onto a higher-dimensional hyperplane. Source location is identified via principal-angle minimization between the hyperplane and the observation vector, yielding a one-shot algorithm with a probabilistic interpretation. The approach is demonstrated on perturbation-source identification in the viscous Burgers equation and heat-source detection in a two-dimensional laminar stratified channel flow, with claimed improvements over linear adjoint methods especially where linear sensitivity vanishes.

Significance. If the quadratic correction proves sufficient and the truncation errors are demonstrably small, the method could provide an efficient, non-iterative extension of adjoint sensitivity analysis to weakly nonlinear regimes where linear methods fail. The geometric interpretation via hyperplane projection and principal angles offers a natural probabilistic framing and potential extensibility to multiple sources. The focus on benchmark problems in fluid dynamics is appropriate, and the avoidance of iterative source-position updates is a practical strength. However, the current lack of a-priori bounds on Taylor remainders and eigen-truncation errors prevents a firm assessment of robustness.

major comments (2)
  1. [Methods (quadratic correction and approximation)] The central claim that quadratic embeddings yield significant localization gains when linear adjoint sensitivity vanishes rests on the assumption that the quadratic term is the leading correction and that the truncated eigen-expansion introduces negligible distortion. The manuscript provides no explicit a-priori bounds on the remainder of the Taylor expansion or on the truncation error of the symmetric bilinear operator (see the description of the quadratic approximation and Krylov iterations). Without these, the hyperplane projection and principal-angle minimization cannot be guaranteed to remain accurate for finite source intensities.
  2. [Numerical demonstrations and results] The results section claims 'significant improvements in localization accuracy' for the quadratic embeddings relative to linear adjoint methods, particularly in regions of vanishing linear sensitivity. However, the demonstrations on the Burgers equation and stratified channel flow lack reported quantitative metrics (e.g., localization error norms, success rates, or comparison tables with effect sizes), numerical error analysis, or verification that the observed gains are not artifacts of the specific truncation level.
minor comments (2)
  1. [Abstract] The abstract introduces 'principal angle minimization' without a brief definition or reference to its geometric meaning in the context of the hyperplane; adding one sentence would improve accessibility for readers unfamiliar with the concept.
  2. [Formulation] Notation for the symmetric bilinear operator and the eigen-expansion truncation level should be introduced consistently in the first appearance to avoid ambiguity when the method is extended to multiple sources.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their thorough review and constructive feedback on our manuscript. We appreciate the recognition of the potential of the quadratic sensitivity approach for source localization in weakly nonlinear systems. Below, we provide point-by-point responses to the major comments and outline the revisions we plan to implement.

read point-by-point responses

  1. Referee: The central claim that quadratic embeddings yield significant localization gains when linear adjoint sensitivity vanishes rests on the assumption that the quadratic term is the leading correction and that the truncated eigen-expansion introduces negligible distortion. The manuscript provides no explicit a-priori bounds on the remainder of the Taylor expansion or on the truncation error of the symmetric bilinear operator (see the description of the quadratic approximation and Krylov iterations). Without these, the hyperplane projection and principal-angle minimization cannot be guaranteed to remain accurate for finite source intensities.

    Authors: We agree that providing bounds or estimates on the approximation errors would strengthen the theoretical foundation. While deriving general a priori bounds for arbitrary nonlinear PDEs is nontrivial and beyond the scope of the current work, we will revise the manuscript to include a dedicated subsection on error analysis. This will feature: (i) a discussion of the conditions under which the quadratic term dominates the Taylor remainder for weakly nonlinear regimes, (ii) numerical convergence studies showing the decay of truncation errors with increasing Krylov subspace dimension for the benchmark problems, and (iii) empirical estimates of the Taylor remainder for the source intensities used in the demonstrations. These additions will support the robustness of the hyperplane projection method. revision: yes

  2. Referee: The results section claims 'significant improvements in localization accuracy' for the quadratic embeddings relative to linear adjoint methods, particularly in regions of vanishing linear sensitivity. However, the demonstrations on the Burgers equation and stratified channel flow lack reported quantitative metrics (e.g., localization error norms, success rates, or comparison tables with effect sizes), numerical error analysis, or verification that the observed gains are not artifacts of the specific truncation level.

    Authors: We acknowledge that the current presentation relies primarily on visual comparisons and qualitative descriptions of improved localization. To address this, we will augment the results section with quantitative metrics, including tables of average localization errors (e.g., Euclidean distance between true and estimated source positions) over ensembles of test cases, success rates defined as the fraction of trials where the estimated location falls within a specified tolerance, and direct comparisons of linear vs. quadratic methods with varying numbers of retained eigenmodes. Additionally, we will include error bars or statistical analysis to demonstrate that the observed improvements are statistically significant and not sensitive to the truncation level within the range explored. revision: yes

Circularity Check

0 steps flagged

Derivation chain is self-contained; no reductions to inputs by construction

full rationale

The framework extends the standard adjoint method with a quadratic bilinear operator approximated by truncated eigen-expansion and Krylov iterations, then defines positional embeddings and a principal-angle search algorithm directly from these operators. No equation in the abstract or described chain equates a claimed prediction or result to a fitted parameter or prior self-citation by construction; the quadratic correction is introduced as an explicit extension rather than a redefinition of the linear adjoint field. Demonstrations on Burgers and stratified flow serve as external checks rather than tautological verifications. The approach therefore remains non-circular under the enumerated patterns.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 1 invented entities

Ledger entries are inferred from the abstract description; full text would permit more complete enumeration of parameters and assumptions.

free parameters (1)
  • eigen-expansion truncation level
    Number of terms retained in the truncated eigen-expansion used to approximate the symmetric bilinear operator for the quadratic correction.
axioms (1)
  • domain assumption The standard adjoint formulation relates multiple measurements to external sources through a linear duality relation.
    Explicitly invoked as the foundation before introducing the quadratic extension.
invented entities (1)
  • quadratic positional embedding no independent evidence
    purpose: To represent the quadratic correction as a symmetric bilinear operator and augment the linear adjoint field for hyperplane projection.
    Newly introduced construct approximated via Krylov subspace iterations.

pith-pipeline@v0.9.0 · 5755 in / 1215 out tokens · 103886 ms · 2026-05-21T15:14:58.183647+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

45 extracted references · 45 canonical work pages

  1. [1]

    blockage

    Sensor sensitivity analysis We first examine the evolution of sensor sensitivity in time. Figure 5 illustrates the scalar field, adjoint sensitivity field, and the leading eigenmode of the Hessian for measurement timesT={1,4,8}, and for source and sensor located at the same heighty s =y m = 0.33. The forward scalar field exhibits characteristic lee-wave p...

    work page

  2. [2]

    The forward field and source reconstruction results for this setup are summarized in figure 7

    Source inference results We consider a sample case where the intensity of the source isI s = 0.05, the mea- surement timeT= 4, and a 5-sensor array is arranged at the same height as the source (ys =y m = 0.33), as shown in figure 7. The forward field and source reconstruction results for this setup are summarized in figure 7. Panel (a) shows the normalize...

    work page

  3. [3]

    T. A. Zaki, Turbulence from an observer perspective, Annual Review of Fluid Mechanics57 (2024)

    work page 2024

  4. [4]

    Wang and J.-H

    Q. Wang and J.-H. Gao, The drag-adjoint field of a circular cylinder wake at reynolds numbers 20, 100 and 500, Journal of Fluid Mechanics730, 145 (2013)

    work page 2013

  5. [5]

    M. C. Hall and D. G. Cacuci, Physical interpretation of the adjoint functions for sensitivity analysis of atmospheric models, Journal of Atmospheric Sciences40, 2537 (1983)

    work page 1983

  6. [6]

    Dom´ ınguez-V´ azquez, N

    D. Dom´ ınguez-V´ azquez, N. Escobar-Castaneda, Q. Wang, and G. B. Jacobs, Inference of inertial particle dynamics from limited measurements, Physics of Fluids37(2025)

    work page 2025

  7. [7]

    Wang and T

    Q. Wang and T. A. Zaki, Domain of dependence for wall-pressure measurements in high-speed boundary layers, Journal of Fluid Mechanics1009, A67 (2025)

    work page 2025

  8. [8]

    Q. Wang, M. Wang, and T. A. Zaki, What is observable from wall data in turbulent channel flow?, Journal of Fluid Mechanics941, A48 (2022)

    work page 2022

  9. [9]

    Q. Wang, Y. Hasegawa, and T. A. Zaki, Spatial reconstruction of steady scalar sources from remote measurements in turbulent flow, Journal of Fluid Mechanics870, 316 (2019)

    work page 2019

  10. [10]

    Tanogami, Scale locality of information flow in turbulence, arXiv preprint arXiv:2407.20572 (2024)

    T. Tanogami, Scale locality of information flow in turbulence, arXiv preprint arXiv:2407.20572 (2024)

    work page arXiv 2024

  11. [11]

    J. P. Antenucci and J. Imberger, Energetics of long internal gravity waves in large lakes, Limnology and oceanography46, 1760 (2001)

    work page 2001

  12. [12]

    B. R. Sutherland,Internal gravity waves(Cambridge university press, 2010)

    work page 2010

  13. [13]

    Staquet and J

    C. Staquet and J. Sommeria, Internal gravity waves: from instabilities to turbulence, Annual Review of Fluid Mechanics34, 559 (2002)

    work page 2002

  14. [14]

    Hasegawa and N

    Y. Hasegawa and N. Kasagi, Low-pass filtering effects of viscous sublayer on high schmidt number mass transfer close to a solid wall, International Journal of Heat and Fluid Flow30, 525 (2009)

    work page 2009

  15. [15]

    Liu and Z

    X. Liu and Z. Zhai, Inverse modeling methods for indoor airborne pollutant tracking: literature 26 review and fundamentals, Indoor air17, 419 (2007)

    work page 2007

  16. [16]

    S. M. Gorelick, B. Evans, and I. Remson, Identifying sources of groundwater pollution: an optimization approach, Water Resour. Res19, 779 (1983)

    work page 1983

  17. [17]

    Alapati, Z

    S. Alapati, Z. Kabala,et al., Recovering the release history of a groundwater contaminant using a non-linear least-squares method, Hydrological processes14, 1003 (2000)

    work page 2000

  18. [18]

    T. H. Skaggs and Z. J. Kabala, Recovering the release history of a groundwater contaminant, Water Resources Research30, 71 (1994)

    work page 1994

  19. [19]

    P. S. Mahar and B. Datta, Optimal monitoring network and ground-water-pollution source identification, Journal of water resources planning and management123, 199 (1997)

    work page 1997

  20. [20]

    P. S. Mahar and B. Datta, Identification of pollution sources in transient groundwater systems, Water Resources Management14, 209 (2000)

    work page 2000

  21. [21]

    P. S. Mahar and B. Datta, Optimal identification of ground-water pollution sources and pa- rameter estimation, Journal of Water Resources Planning and Management127, 20 (2001)

    work page 2001

  22. [22]

    Misaka, S

    T. Misaka, S. Obayashi, and E. Endo, Measurement-integrated simulation of clear air turbu- lence using a four-dimensional variational method, Journal of aircraft45, 1217 (2008)

    work page 2008

  23. [23]

    E. P. Olaguer, Adjoint model enhanced plume reconstruction from tomographic remote sensing measurements, Atmospheric environment45, 6980 (2011)

    work page 2011

  24. [24]

    Courtier and F

    P. Courtier and F. Rabier, The use of adjoint equations in numerical weather prediction, Atmosphere-Ocean35, 303 (1997)

    work page 1997

  25. [25]

    Pires and P

    C. Pires and P. Miranda, Sensitivity of the adjoint method in the inversion of tsunami source parameters, Natural Hazards and Earth System Sciences3, 341 (2003)

    work page 2003

  26. [26]

    X. Li, C. Wang, W. Fan, and X. Lv, Optimization of the spatiotemporal parameters in a dynamical marine ecosystem model based on the adjoint assimilation, Mathematical Problems in Engineering2013(2013)

    work page 2013

  27. [27]

    M. Kim, H. M. Kim, J. Kim, S.-M. Kim, C. Velden, and B. Hoover, Effect of enhanced satellite-derived atmospheric motion vectors on numerical weather prediction in east asia using an adjoint-based observation impact method, Weather and Forecasting32, 579 (2017)

    work page 2017

  28. [28]

    Cerizza, W

    D. Cerizza, W. Sekiguchi, T. Tsukahara, T. Zaki, and Y. Hasegawa, Reconstruction of scalar source intensity based on sensor signal in turbulent channel flow, Flow, Turbulence and Com- bustion97, 1211 (2016)

    work page 2016

  29. [29]

    Q. Wang, Y. Hasegawa, and T. A. Zaki, Spatial reconstruction of steady scalar sources from 27 remote measurements in turbulent flow, Journal of Fluid Mechanics870, 316–352 (2019)

    work page 2019

  30. [30]

    M. A. Abassi, Q. Wang, and X. Liu, Adjoint-based data assimilation in a subdomain using omnidirectional-integration-enabled pressure dirichlet boundary conditions, Physics of Fluids 37(2025)

    work page 2025

  31. [31]

    M. Wang, Q. Wang, and T. A. Zaki, Discrete adjoint of fractional-step incompressible navier- stokes solver in curvilinear coordinates and application to data assimilation, Journal of Com- putational Physics396, 427 (2019)

    work page 2019

  32. [32]

    Keats, E

    A. Keats, E. Yee, and F.-S. Lien, Bayesian inference for source determination with applications to a complex urban environment, Atmospheric environment41, 465 (2007)

    work page 2007

  33. [33]

    J. D. Eldredge and M. Le Provost, Bayesian inference of vorticity in unbounded flow from limited pressure measurements, Journal of Fluid Mechanics986, A18 (2024)

    work page 2024

  34. [34]

    T. A. Zaki, Turbulence from an observer perspective, Annual Review of Fluid Mechanics57, 311 (2025)

    work page 2025

  35. [35]

    T. A. Zaki and M. Wang, From limited observations to the state of turbulence: Fundamental difficulties of flow reconstruction, Phys. Rev. Fluids6, 100501 (2021)

    work page 2021

  36. [36]

    Wang and T

    M. Wang and T. A. Zaki, State estimation in turbulent channel flow from limited observations, Journal of Fluid Mechanics917, A9 (2021)

    work page 2021

  37. [37]

    Z. Wang, I. M. Navon, F.-X. Le Dimet, and X. Zou, The second order adjoint analysis: theory and applications, Meteorology and atmospheric physics50, 3 (1992)

    work page 1992

  38. [38]

    Cioaca, M

    A. Cioaca, M. Alexe, and A. Sandu, Second-order adjoints for solving pde-constrained opti- mization problems, Optimization methods and software27, 625 (2012)

    work page 2012

  39. [39]

    Le Dimet and V

    F.-X. Le Dimet and V. Shutyaev, Second-order methods in variational data assimilation, in Data Assimilation for Atmospheric, Oceanic and Hydrologic Applications (Vol. IV)(Springer,

    work page

  40. [40]

    D. G. Cacuci, The second-order adjoint sensitivity analysis methodology for nonlinear sys- tems—i: Theory, Nuclear Science and Engineering184, 16 (2016)

    work page 2016

  41. [41]

    Le Dimet, I

    F.-X. Le Dimet, I. M. Navon, and D. N. Daescu, Second-order information in data assimilation, Monthly Weather Review130, 629 (2002)

    work page 2002

  42. [42]

    Boujo, Second-order adjoint-based sensitivity for hydrodynamic stability and control, Jour- nal of Fluid Mechanics920, A12 (2021)

    E. Boujo, Second-order adjoint-based sensitivity for hydrodynamic stability and control, Jour- nal of Fluid Mechanics920, A12 (2021)

    work page 2021

  43. [43]

    B. N. Parlett,The symmetric eigenvalue problem(SIAM, 1998). 28

    work page 1998

  44. [44]

    R. B. Lehoucq, D. C. Sorensen, and C. Yang,ARPACK users’ guide: solution of large-scale eigenvalue problems with implicitly restarted Arnoldi methods(SIAM, 1998)

    work page 1998

  45. [45]

    M. Wang, Q. Wang, and T. A. Zaki, Discrete adjoint of fractional-step incompressible navier- stokes solver in curvilinear coordinates and application to data assimilation, Journal of Com- putational Physics396, 427 (2019). 29

    work page 2019