Pith. sign in

REVIEW 2 major objections 5 minor 71 references

Transferable inference of turbulence models for urban flows with the Parameter-Regularised Ensemble Kalman Filter

T0 review · 2 major / 5 minor · reviewed 2026-07-12 · grok-4.5

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read A parameter-regularised ensemble Kalman filter yields transferable RANS turbulence coefficients for urban flows, cutting reconstruction error by up to 50% and improving predictions on unseen city-scale cases.

desk verdict Clean MAP derivation of a literature-regularised EnKF that actually improves SST transfer from CEDVAL to Shinjuku; the σ=0.2 prior is hand-tuned but the transfer evidence is real and the unregularised EnKF fails where it should. read the letter →

arxiv 2607.03571 v1 pith:O62O5QEW submitted 2026-07-03 physics.flu-dyn physics.comp-phphysics.data-an

Emanuele Bombardi , Andrea Nóvoa , Luca Magri , Alessandro Parente This is my paper
classification physics.flu-dynphysics.comp-phphysics.data-an
keywords urbanflowRANSSSTk-ωensembleKalmanfilterparameterregularisationtransferabilitydataassimilationturbulencemodelling
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

Urban wind and pollutant predictions often use RANS models whose closure coefficients are tuned on simple benchmarks and then applied to real cities—where they frequently fail. This paper derives a Parameter-Regularised Ensemble Kalman Filter (PR-EnKF) that adds a literature-consistent prior term to the standard EnKF cost function, so Bayesian updates stay inside physically meaningful ranges. Coefficients of the SST k–ω model are inferred on an isolated-building wind-tunnel case and then transferred, unchanged, to a high-rise, a building array, and the Shinjuku district. Relative to both a default RANS baseline and an unregularised EnKF, the regularised parameters converge faster, shrink uncertainty by roughly an order of magnitude, reduce reconstruction RMSE by up to 50% on the training geometry, and continue to improve predictions on the more complex, previously unseen configurations. The practical claim is that one can therefore calibrate once on a cheap laboratory case and still obtain more accurate large-scale urban RANS without re-optimising millions of cells.

What carries the argument

The Parameter-Regularised Ensemble Kalman Filter (PR-EnKF): the closed-form Kalman update that minimises a cost containing forecast prior, data misfit, and an extra term that pulls parameters toward literature values with adaptive strength set by the ensemble covariance.

What would settle it

Transfer the same PR-EnKF coefficients to another full-scale urban wind-tunnel or field campaign (different packing density or roughness) and check whether velocity RMSE still falls below the default SST baseline; if it does not, the claimed transferability fails.

Watch

Extended reading notes

Core claim

The analytical MAP solution of a three-term cost that regularises the standard EnKF against literature-consistent turbulence coefficients produces SST k–ω parameters that remain physically bounded, converge with low ensemble spread, and transfer from an isolated building to multi-building and full-district urban flows, cutting reconstruction error by up to 50% and improving baseline predictions where the unregularised EnKF degrades them.

Load-bearing premise

That a single relative uncertainty of 20% on every SST coefficient, plus coefficients learned on one small-scale isolated building, remain good enough for full-scale urban geometries whose Reynolds numbers, roughness and multi-building interactions are quite different.

Editorial extensions

If this is right

  • Urban RANS studies can calibrate SST coefficients once on a cheap laboratory building and reuse them for district-scale meshes without re-optimisation.
  • Regularisation selectively freezes well-constrained coefficients and only updates those that the data actually inform, reducing the risk of geometry-specific overfitting.
  • The same three-term cost can be applied to other two-equation closures or multi-field observations (velocity + TKE) with only modest extra cost.
  • Computational overhead of model optimisation for city-scale air-quality and ventilation studies is substantially lowered.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The same regularisation idea could be used for other empirical closures (e.g., atmospheric boundary-layer wall functions) whose coefficients also drift across geometries.
  • If the literature prior is replaced by a hierarchical hyper-prior on σ itself, the method might adapt the regularisation strength automatically to different data densities.
  • The observed selective activation of TKE-related coefficients suggests the filter could guide which sensors are most informative for future urban measurement campaigns.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

2 major / 5 minor

Summary. The paper derives a Parameter-Regularised Ensemble Kalman Filter (PR-EnKF) as the MAP solution of a three-term cost that adds a literature-consistent Gaussian prior on the SST k-ω coefficients to the standard EnKF objective. Parameters are inferred on CEDVAL A1-1 (velocity-only and multi-field) and then frozen and transferred, without re-optimisation, to AIJ Case B (high-rise), Case C (building array) and Case F (Shinjuku). On the assimilation case the PR-EnKF converges with ensemble spreads an order of magnitude smaller than the unregularised EnKF and reduces velocity RMSE by ~50 % relative to baseline CFD; on the transfer cases it improves baseline RMSE (13.8 % on the array, 16.2 % on Shinjuku) while the unregularised EnKF either degrades or improves less consistently. The authors attribute the transferability gain to the regularisation term that selectively updates only the most data-sensitive coefficients.

Significance. If the transferability claim holds under broader priors and geometries, the work supplies a practical, Bayesian route to calibrate RANS closures once on a cheap laboratory case and deploy them on city-scale meshes, which is of direct value for urban planning and air-quality assessment. The analytical MAP derivation (Eqs. 12–17) is clean, recovers the ordinary EnKF as C_pp → ∞, and is a useful addition to the regularised-EnKF literature. The multi-geometry assimilation-transfer protocol and the explicit comparison against unregularised EnKF are strengths that make the central claim falsifiable.

major comments (2)
  1. The transferability advantage is demonstrated for a single hand-chosen prior strength (σ = 0.2, diagonal C_pp) inferred only on CEDVAL A1-1. Appendix B varies σ on the same assimilation case but never re-infers under a different prior (or on a second geometry) and then re-tests transfer to AIJ C/F. Consequently it remains open whether the reported gains on the array and Shinjuku are a general property of the regularised filter or an artefact of that particular prior. A minimal additional experiment—e.g. transfer of the σ = 0.05 and warm-start posteriors already computed in Appendix B, or a leave-one-geometry-out re-inference—would substantially strengthen the central claim.
  2. Section 3.3.1 and Table 1 treat all eleven SST coefficients as free parameters under a uniform relative uncertainty. Several of these coefficients are linked by the original Menter blending construction (e.g. the pairs (α_k1,α_k2), (γ1,γ2), eta*). The paper does not discuss whether unconstrained independent updates preserve the intended near-wall / free-stream blending or the realisability properties of the SST model. A short check that the optimised coefficients still satisfy the original algebraic relations (or an explicit statement that those relations are deliberately relaxed) is needed for physical interpretability of the transferred parameters.
minor comments (5)
  1. Eq. (15) states ~M ∈ R^{N_φ imes (N_q+N_α)}; the first dimension should be N_q+N_α (or the transpose convention should be clarified).
  2. Table 2 reports EnKF ensemble spreads of O(100 %) and mean deviations of several hundred percent; a brief remark on whether those members remain numerically stable inside the RANS solver would help the reader assess the comparison.
  3. Figure 4b and the accompanying text use both “iterations” and “assimilation cycles”; a consistent terminology (and a reminder that Δn_a = 500) would improve readability.
  4. The multi-field normalisation (min-max to [0,1]) is described only briefly in §3.3.2; stating whether the same bounds are used for all ensemble members and all cycles would aid reproducibility.
  5. A few typographical issues: “N´ ovoa” accents, “K´ arm´ an”, and the arXiv date line “3 Jul 2026” should be corrected before final production.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: PR-EnKF is an independent MAP derivation with literature prior; parameters inferred on one case are frozen and tested on held-out geometries.

full rationale

The paper derives the PR-EnKF update (Eqs. 16–17) by maximising the three-term posterior (12)–(13) that adds a Gaussian pseudo-observation term on literature values p (Menter 1994 defaults) to the standard EnKF cost; the algebra is self-contained and recovers the ordinary EnKF when C_pp o∞. Parameters are inferred solely on CEDVAL A1-1 velocity (and optionally TKE) data, then frozen and transferred without re-fitting to three independent AIJ geometries whose observations never enter the filter. The scalar relative uncertainty σ=0.2 that builds the diagonal C_pp is a user-chosen regularisation strength (sensitivity shown in Appendix B on the same assimilation case only); it is not fitted to the transfer metrics that constitute the main claim. Self-citations (Nóvoa & Magri, Magri & Doan) supply background on ensemble methods and bias-aware filters but are not invoked as uniqueness theorems or load-bearing premises that force the transferability result. Inlet-profile fitting to the same public databases used for validation is ordinary CFD practice and does not make the posterior parameters or the reported RMSE reductions tautological. Consequently the derivation chain and the empirical transfer tests stand independently of their inputs.

Assumptions & free parameters 3 free parameters · 4 assumptions · 1 invented entities

The central claim rests on standard Bayesian/Gaussian EnKF assumptions, the SST k-ω model equations, literature default coefficients used as soft priors, and a handful of hand-chosen numerical hyperparameters (σ, ensemble size, inflation, assimilation interval). No new physical entities are postulated; the free parameters are the eleven SST coefficients themselves plus the regularisation strength.

free parameters (3)
  • σ (relative uncertainty on literature parameters) = 0.2 (default)
    User-defined scalar that sets the diagonal of C_pp; chosen as 0.2 after limited sensitivity tests (σ=0.05, 0.1, 0.2). Directly controls regularisation strength and therefore the transferred coefficients.
  • Eleven SST k-ω coefficients (a1, b1, c1, β*, αk1, αk2, αω1, αω2, γ1, γ2, β2) = see Table 2 (e.g. a1≈0.294, β*≈0.066 for σ=0.2 velocity-only)
    Inferred by the filter; their posterior means are the transferable result. Initialised from Menter (1994) defaults and updated under the regularised cost.
  • Ensemble size Ne, inflation λ, assimilation interval Δna = 70, 1.1, 500
    Numerical hyperparameters fixed at Ne=70, λ=1.1, Δna=500 after informal tests; they affect covariance rank and update frequency.
assumptions (4)
  • standard math Gaussian prior, likelihood and parameter pseudo-observation errors; linear Kalman update remains optimal for the MAP of the three-term quadratic cost.
    Standard EnKF modelling assumptions used to derive Eqs. 12–17.
  • domain assumption Steady RANS with the SST k-ω closure and ABL-consistent wall functions adequately represent the mean urban flow for the purpose of parameter transfer.
    Underlying CFD model throughout Sections 3–5; if the model form is inadequate, parameter transfer cannot compensate.
  • domain assumption Literature default SST coefficients (Menter 1994) constitute a physically meaningful soft prior that should not be abandoned without strong data evidence.
    Encoded as Term 3 of the cost (Eq. 13) and as the mean of the initial ensemble.
  • ad hoc to paper A single relative uncertainty σ applied uniformly to all eleven parameters is an adequate regularisation prior.
    Chosen for simplicity in §3.3.1; no parameter-specific or geometry-dependent uncertainty model is derived.
invented entities (1)
  • Parameter-Regularised Ensemble Kalman Filter (PR-EnKF)
    purpose: Provides the sequential Bayesian update that jointly assimilates flow observations and literature parameter pseudo-observations.
    New algorithmic object defined by the three-term cost and the resulting gains Kφ,p and Kα,p; independent evidence is the derivation itself and the numerical experiments, not an external physical discovery.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Transferable inference of turbulence models for urban flows with the Parameter-Regularised Ensemble Kalman Filter." pith.science (2026). https://pith.science/paper/O62O5QEW

@misc{pith2026260703571,
  author       = {Pith},
  title        = {Pith review of: Transferable inference of turbulence models for urban flows with the Parameter-Regularised Ensemble Kalman Filter},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/O62O5QEW}},
  note         = {Machine review of arXiv:2607.03571}
}
read the original abstract

The accurate simulation of urban flow is key to designing building ventilation, understanding cities' micrometeorology, and predicting pollutant dispersion. Reynolds-Averaged Navier-Stokes (RANS) simulations are a common modelling approach for simulating urban flow, but their accuracy depends on the closure model and its parameters. These parameters are inferred from benchmark cases, but they are not necessarily suitable for realistic urban environments, which involve different physical mechanisms. This is referred to as the transferability problem of RANS urban modelling. The objective of this work is to propose a robust Bayesian method to {sequentially} infer RANS parameters for urban flow modelling. Key to the approach is the mathematical derivation of the parameter-regularised ensemble Kalman filter (PR-EnKF), which is the analytical solution of the data assimilation problem for the sequential parameter estimation. The cost functional is regularised using the prior knowledge on the turbulence parameters, thereby ensuring that the Bayesian updates remain within physical ranges. The parameters are first inferred on an isolated building, and then transferred to three cases of increasing complexity: (i) a high-rise building, (ii) a multi-building array, and (iii) the Shinjuku district urban environment. Results show that the PR-EnKF achieves faster convergence, reducing parameter uncertainty by an order of magnitude and reconstruction errors by up to 50%. Because of the regularisation, the PR-EnKF selectively updates the most important parameters. This work enables robust large-scale urban flow simulation whilst reducing the computational overhead of model optimisation for urban planning and air quality assessment.

Figures

Figures reproduced from arXiv: 2607.03571 by the authors.

Figure 1
Figure 1. Schematic of the assimilation-transferability methodology. The PR-EnKF infers optimal [PITH_FULL_IMAGE:figures/full_fig_p009_1.png] view at source ↗
Figure 2
Figure 2. Schematic representation of the CEDVAL A1-1 configuration showing [Leitl and Schatz [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. Schematic representation of the three transferability cases from [AIJ, 2007]. (a) AIJ Case [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figures from the paper (10 more)
Figure 4
Figure 4. Figure 4: Parameter convergence with the standard EnKF, PR-EnKF with [PITH_FULL_IMAGE:figures/full_fig_p016_4.png]
Figure 5
Figure 5. Figure 5: Contribution from the observations and parameter regularisation to the mean analysis for [PITH_FULL_IMAGE:figures/full_fig_p017_5.png]
Figure 6
Figure 6. Figure 6: Inferred k-ω SST parameters from their literature-consistent values. Turbulent kinetic energy assimilation (orange, hatched) activates parameters associated with turbulence transport and dissipation, producing larger deviations than velocity-only optimisation (green, d…
Figure 7
Figure 7. Figure 7: Profiles of (a) streamwise velocity and (b) turbulent kinetic energy at six measurement [PITH_FULL_IMAGE:figures/full_fig_p018_7.png]
Figure 8
Figure 8. Figure 8: Transferability to high-rise building. Streamwise velocity profiles at six measurement [PITH_FULL_IMAGE:figures/full_fig_p019_8.png]
Figure 9
Figure 9. Figure 9: Transferability to building array. Comparison of the predictions from the baseline CFD [PITH_FULL_IMAGE:figures/full_fig_p020_9.png]
Figure 10
Figure 10. Figure 10: Transferability to building array. Parity plots of normalised velocity [PITH_FULL_IMAGE:figures/full_fig_p022_10.png]
Figure 11
Figure 11. Figure 11: Transferability to a real urban district (Shinjuku). Comparison of the predictions from [PITH_FULL_IMAGE:figures/full_fig_p023_11.png]
Figure 12
Figure 12. Figure 12: Flow field comparisons in the Shinjuku urban environment obtained using the PR [PITH_FULL_IMAGE:figures/full_fig_p024_12.png]
Figure 13
Figure 13. Figure 13: Profiles of (a) streamwise velocity and (b) turbulent kinetic energy at six measurement [PITH_FULL_IMAGE:figures/full_fig_p027_13.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

71 extracted references · 11 canonical work pages

  1. [1]

    , year =

    Anderson, J.L. , year =. An Ensemble adjustment

  2. [2]

    The ensemble

    Evensen, Geir , year =. The ensemble. Control Systems, IEEE , doi =

  3. [3]

    Physics-informed machine learning , journal =

    Karniadakis, George and Kevrekidis, Yannis and Lu, Lu and Perdikaris, Paris and Wang, Sifan and Yang, Liu , year =. Physics-informed machine learning , journal =

  4. [4]

    Application of a comprehensive

    Raghunathan Srikumar, Sampath Kumar and Cotteleer, Leo and Mosca, Gabriele and Gambale, Alessandro and Parente, Alessandro , year =. Application of a comprehensive. Building and Environment , doi =

  5. [5]

    A review of

    Bombardi, Emanuele and Gambale, Alessandro and Parente, Alessandro , year =. A review of. Building and Environment , doi =

  6. [6]

    Generalizability evaluation of k- models calibrated by using ensemble

    Zhao, Runmin and Liu, Sumei and Liu, Junjie and Jiang, Nan and Chen, Qingyan , year =. Generalizability evaluation of k- models calibrated by using ensemble. Building and Environment , doi =

  7. [7]

    Computers and Fluids , volume =

    Augmented state estimation of urban settings using on-the-fly sequential. Computers and Fluids , volume =. 2024 , issn =. doi:10.1016/j.compfluid.2023.106118 , author =

    work page doi:10.1016/j.compfluid.2023.106118 2024

  8. [8]

    Building and Environment , doi =

    Blocken, Bert , year =. Building and Environment , doi =

Show all 71 references
  1. [9]

    1991 , address =

    Daley, Roger , title =. 1991 , address =

    1991

  2. [10]

    Kalman, R. E. , title =. Journal of Basic Engineering , volume =. 1960 , month =

    1960

  3. [11]

    and Strijhak, Sergei , year =

    Romanova, Daria and Ivanov, Oleg and Trifonov, Vladimir and Ginzburg, Nika and Korovina, Daria and Ginzburg, Boris and Koltunov, Nikita and Eglit, M. and Strijhak, Sergei , year =. Calibration of the k -. Fluids , doi =

  4. [12]

    and Carneiro, F.O

    Rocha, Paulo and Rocha, H.H. and Carneiro, F.O. and Silva, M.E. and Bueno, Andre , year =. K-. Energy , doi =

  5. [13]

    Calibration of Turbulent Model Constants Based on Experimental Data Assimilation: Numerical Prediction of Subsonic Jet Flow Characteristics , volume =

    He, Xin and Yuan, Changjiang and Gao, Haoran and Chen, Yaqing and Zhao, Rui , year =. Calibration of Turbulent Model Constants Based on Experimental Data Assimilation: Numerical Prediction of Subsonic Jet Flow Characteristics , volume =. Sustainability , doi =

  6. [14]

    Calibration of k - turbulence model for thermal–hydraulic analyses in rib-roughened narrow rectangular channels using genetic algorithm , volume =

    Khan, Abid and Md Shafiqul, Islam and Sazzad, Istiak , year =. Calibration of k - turbulence model for thermal–hydraulic analyses in rib-roughened narrow rectangular channels using genetic algorithm , volume =. SN Applied Sciences , doi =

  7. [15]

    and Wang, J.-X

    Xiao, Heng and Wu, J.-L. and Wang, J.-X. and Sun, R. and Roy, C. J. , title =. Journal of Computational Physics , year =

  8. [16]

    and Moser, Robert D

    Oliver, Todd A. and Moser, Robert D. , title =. Journal of Physics: Conference Series , year =

  9. [17]

    , title =

    Schillings, Christian and Stuart, Andrew M. , title =. arXiv preprint arXiv:1602.02020 , year =

    arXiv

  10. [18]

    2020 , issn =

    Journal of Building Engineering , volume =. 2020 , issn =. doi:10.1016/j.jobe.2020.101756 , author =

    2020 doi

  11. [19]

    Stroud and Matthias Katzfuss and Christopher K

    Jonathan R. Stroud and Matthias Katzfuss and Christopher K. Wikle. A Bayesian Adaptive Ensemble Kalman Filter for Sequential State and Parameter Estimation. Monthly Weather Review. 2018. doi:10.1175/MWR-D-16-0427.1

    2018 doi

  12. [20]

    Journal of Wind Engineering and Industrial Aerodynamics , volume =

    Quantifying inflow and. Journal of Wind Engineering and Industrial Aerodynamics , volume =. 2015 , note =. doi:10.1016/j.jweia.2015.03.025 , author =

    2015 doi

  13. [21]

    Menter , title =

    Florian R. Menter , title =. AIAA Journal , year =

  14. [22]

    1974 , issn =

    The numerical computation of turbulent flows , journal =. 1974 , issn =. doi:10.1016/0045-7825(74)90029-2 , author =

    1974 doi

  15. [23]

    Wilcox , title =

    David C. Wilcox , title =. AIAA Journal , volume =. 1988 , doi =

    1988

  16. [24]

    A Comprehensive Modelling Approach for the Neutral Atmospheric Boundary Layer: Consistent Inflow Conditions, Wall Function and Turbulence Model , volume =

    Parente, Alessandro and Gorlé, Catherine and Beeck, Jeroen and Benocci, Carlo , year =. A Comprehensive Modelling Approach for the Neutral Atmospheric Boundary Layer: Consistent Inflow Conditions, Wall Function and Turbulence Model , volume =. Boundary-Layer Meteorology , doi =

  17. [25]

    An Extended

    Bellegoni, Marco and Cotteleer, Leo and Raghunathan Srikumar, Sampath Kumar and Mosca, Gabriele and Gambale, Alessandro and Tognotti, Leonardo and Galletti, Chiara and Parente, Alessandro , year =. An Extended. SSRN Electronic Journal , doi =

  18. [26]

    Improved k- model and wall function formulation for the

    Parente, Alessandro and Gorlé, Catherine and Beeck, Jeroen and Benocci, Carlo , year =. Improved k- model and wall function formulation for the. Journal of Wind Engineering and Industrial Aerodynamics , doi =

  19. [27]

    Advanced turbulence models and boundary conditions for flows around different configurations of ground-mounted buildings , volume =

    Longo, Riccardo and Ferrarotti, Marco and Garcia-Sanchez, Clara and Derudi, Marco and Parente, Alessandro , year =. Advanced turbulence models and boundary conditions for flows around different configurations of ground-mounted buildings , volume =. Journal of Wind Engineering ...

  20. [28]

    1998 , howpublished =

    Bernd Leitl and Michael Schatzmann , title =. 1998 , howpublished =

    1998

  21. [29]

    2008 , note =

    Journal of Wind Engineering and Industrial Aerodynamics , volume =. 2008 , note =. doi:10.1016/j.jweia.2008.02.058 , author =

    2008 doi

  22. [30]

    Cross Comparisons of

    Tominaga, Yoshihide and Mochida, Akashi and Shirasawa, Taichi and Yoshie, Ryuichiro and Kataoka, Hiroto and Harimoto, Kazuyoshi and Nozu, Tsuyoshi , year =. Cross Comparisons of. Journal of Asian Architecture and Building Engineering , volume =. doi:10.3130/jaabe.3.63 , issn =

    doi

  23. [31]

    Pedestrian Wind Comfort:

  24. [32]

    2024 , issn =

    Physics-informed machine learning: A comprehensive review on applications in anomaly detection and condition monitoring , journal =. 2024 , issn =. doi:10.1016/j.eswa.2024.124678 , author =

    2024 doi

  25. [33]

    2021 , issn =

    Journal of Computational Physics , volume =. 2021 , issn =. doi:10.1016/j.jcp.2020.109913 , author =

    2021 doi

  26. [34]

    Entropy , VOLUME =

    Mohammad-Djafari, Ali , TITLE =. Entropy , VOLUME =. 2021 , NUMBER =

    2021

  27. [35]

    Journal of Computational Physics , volume =

    Regularized ensemble. Journal of Computational Physics , volume =. 2020 , issn =. doi:10.1016/j.jcp.2020.109517 , author =

    2020 doi

  28. [36]

    Computer Methods in Applied Mechanics and Engineering , volume =

    Inferring unknown unknowns: Regularized bias-aware ensemble. Computer Methods in Applied Mechanics and Engineering , volume =. 2024 , issn =. doi:10.1016/j.cma.2023.116502 , author =

    2024 doi

  29. [37]

    A regularizing iterative ensemble

    Iglesias, Marco A , year=. A regularizing iterative ensemble. Inverse Problems , publisher=. doi:10.1088/0266-5611/32/2/025002 , number=

    doi

  30. [38]

    Computer Methods in Applied Mechanics and Engineering , volume =

    Parameter identification in explicit structural dynamics: performance of the extended. Computer Methods in Applied Mechanics and Engineering , volume =. 2004 , issn =. doi:10.1016/j.cma.2004.02.003 , author =

    2004 doi

  31. [39]

    , title =

    Franke, J. , title =. The Fourth International Symposium on Computational Wind Engineering , address =

  32. [40]

    2007 , note =

    Guidebook for. 2007 , note =

    2007

  33. [41]

    and Magri, L

    Nóvoa, A. and Magri, L. , year=. Real-time thermoacoustic data assimilation , volume=. doi:10.1017/jfm.2022.653 , journal=

    2022 doi

  34. [42]

    Anderson

    Jeffrey L. Anderson. An Ensemble Adjustment Kalman Filter for Data Assimilation. Monthly Weather Review. 2001. doi:10.1175/1520-0493(2001)129<2884:AEAKFF>2.0.CO;2

    2001 doi

  35. [43]

    2018 , journal =

    Bert Blocken , title =. 2018 , journal =

    2018

  36. [44]

    Numerical Simulation of Dispersion around an Isolated Cubic Building : Comparison of Various Types of k- Models , volume =

    Tominaga, Yoshihide and Stathopoulos, Ted , year =. Numerical Simulation of Dispersion around an Isolated Cubic Building : Comparison of Various Types of k- Models , volume =. Atmospheric Environment , doi =

  37. [45]

    Pope, Stephen B. , year=. Turbulent Flows , publisher=

  38. [46]

    P. G. Duynkerke. Application of the E – Turbulence Closure Model to the Neutral and Stable Atmospheric Boundary Layer. Journal of Atmospheric Sciences. 1988. doi:10.1175/1520-0469(1988)045<0865:AOTTCM>2.0.CO;2

    1988 doi

  39. [47]

    Statistical calibration of

    Glover, Nina and Guillas, Serge and Malki-Epshtein, Liora , year =. Statistical calibration of

  40. [48]

    Physics-Informed Data-Driven Prediction of Turbulent Reacting Flows with Lyapunov Analysis and Sequential Data Assimilation

    Magri, Luca and Doan, Nguyen Anh Khoa. Physics-Informed Data-Driven Prediction of Turbulent Reacting Flows with Lyapunov Analysis and Sequential Data Assimilation. Data Analysis for Direct Numerical Simulations of Turbulent Combustion: From Equation-Based Analysis to Machine L...

    2020 doi

  41. [49]

    , title =

    Zaki, Tamer A. , title =. Annual Review of Fluid Mechanics , year =

  42. [50]

    Wang, Mengze and Zaki, Tamer A. , year=. Variational data assimilation in wall turbulence: from outer observations to wall stress and pressure , volume=. doi:10.1017/jfm.2025.132 , journal=

    2025 doi

  43. [51]

    Journal of Computational Physics , volume =

    A reduced order model based on. Journal of Computational Physics , volume =. 2017 , issn =. doi:10.1016/j.jcp.2017.06.042 , author =

    2017 doi

  44. [52]

    and Hou, Wei and Eldredge, Jeff , year =

    Le Provost, M. and Hou, Wei and Eldredge, Jeff , year =. Deep learning and data assimilation approaches to sensor reduction in estimation of disturbed separated flows , doi =

  45. [53]

    , title =

    Vinuesa, Ricardo and Brunton, Steven L. , title =. Nature Computational Science , year =

  46. [54]

    Progress in Aerospace Sciences , volume =

    Quantification of model uncertainty in. Progress in Aerospace Sciences , volume =. 2019 , issn =. doi:10.1016/j.paerosci.2018.10.001 , author =

    2019 doi

  47. [55]

    Energies , VOLUME =

    Jiang, Chao and Mi, Junyi and Laima, Shujin and Li, Hui , TITLE =. Energies , VOLUME =. 2020 , NUMBER =

    2020

  48. [56]

    Journal of the Royal Statistical Society: Series B (Statistical Methodology) , volume=

    Bayesian calibration of computer models , author=. Journal of the Royal Statistical Society: Series B (Statistical Methodology) , volume=. 2001 , publisher=

    2001

  49. [57]

    Journal of Process Control , volume=

    Constrained ensemble Kalman filter based on Kullback--Leibler divergence , author=. Journal of Process Control , volume=. 2019 , publisher=

    2019

  50. [58]

    Journal of Computational Physics , volume=

    Space-dependent aggregation of stochastic data-driven turbulence models , author=. Journal of Computational Physics , volume=. 2025 , publisher=

    2025

  51. [59]

    Monthly weather review , volume=

    Analysis scheme in the ensemble Kalman filter , author=. Monthly weather review , volume=

  52. [60]

    Journal of Fluid Mechanics , year =

    Liu, Hao-Chen and Yin, Zifei and Zhang, Xin-Lei and He, Guowei , title =. Journal of Fluid Mechanics , year =

  53. [61]

    Journal of Fluid Mechanics , year =

    Zhang, Xin-Lei and Zhang, Fengshun and Li, Zhaobin and Yang, Xiaolei and He, Guowei , title =. Journal of Fluid Mechanics , year =

  54. [62]

    and Lehnasch, G

    Moldovan, G. and Lehnasch, G. and Cordier, L. and Meldi, M. , title =. Journal of Computational Physics , year =

  55. [63]

    Computers & Fluids , volume =

    Zhang, Xin-Lei and Xiao, Heng and He, Guo-Wei and Wang, Shi-Zhao , title =. Computers & Fluids , volume =. 2021 , doi =

    2021

  56. [64]

    1977 , address =

    Tikhonov, Andrei Nikolaevich and Arsenin, Vasiliy Yakovlevich , title =. 1977 , address =

    1977

  57. [65]

    and Mei, S

    Hu, J. and Mei, S. and Xu, L. and Hang, J. , title =. Urban Climate , volume =. 2026 , doi =

    2026

  58. [66]

    Computers & Mathematics with Applications , volume =

    Latt, Jonas and Malaspinas, Orestis and Kontaxakis, Dimitrios and others , title =. Computers & Mathematics with Applications , volume =. 2021 , doi =

    2021

  59. [67]

    , title =

    Woodbury, Max A. , title =. Memorandum Report , volume =

  60. [68]

    Computers & Geosciences , volume=

    Ensemble smoother with multiple data assimilation , author=. Computers & Geosciences , volume=. 2013 , publisher=

    2013

  61. [69]

    and Reynolds, Albert C

    Emerick, Alexandre A. and Reynolds, Albert C. , title =. Computational Geosciences , volume =. 2012 , doi =

    2012

  62. [70]

    and Ghili, Saman , title =

    Iaccarino, Gianluca and Mishra, Aashwin A. and Ghili, Saman , title =. Physical Review Fluids , volume =

  63. [71]

    Edeling, W. N. and Cinnella, P. and Dwight, R. P. and Bijl, H. , title =. Journal of Computational Physics , volume =

Pith tools

Reviewed July 12, 2026 · model on record in the stance chip above.