Pith. sign in

REVIEW 4 major objections 6 minor 98 references

RL-DAUNCE: Reinforcement Learning-Driven Data Assimilation with Uncertainty-Aware Constrained Ensembles

T0 review · 4 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

Pith's one-line read RL-DAUNCE matches constrained EnKF at 20x speed

desk verdict A potentially useful constrained-DA emulator that currently overclaims accuracy and mislabels imitation as RL; worth major revision before acceptance. read the letter →

arxiv 2505.05452 v1 pith:5QSOVUJL submitted 2025-05-08 cs.LG math-phmath.MP

classification cs.LGmath-phmath.MP
keywords dataassimilationreinforcementlearningensembleKalmanfilterphysicalconstraintsuncertaintyquantificationMadden-JulianOscillationprimal-dualoptimizationextremeevents
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

RL-DAUNCE claims that data assimilation can be run as a team of reinforcement-learning agents, one per ensemble member, trained to imitate a physically constrained ensemble Kalman filter. The paper tests this on the Madden-Julian Oscillation, where standard EnKF blows up because the estimated convective activity goes negative or energy drifts. RL-DAUNCE's ensemble of agents learns the constrained filter's one-step updates while keeping total energy inside a prescribed band and convective activity strictly positive. The result is assimilation accuracy and uncertainty spread close to constrained EnKF, with each step costing roughly 1.1 seconds instead of 22.96 seconds, a twentyfold speed-up. A sympathetic reader cares because it suggests learned filters can replace expensive constrained optimization while preserving both physics and uncertainty quantification.

What carries the argument

The central object is the ensemble-as-agents construction: $N$ policy networks $\pi_{\theta^{(i)}}$ that mirror EnKF ensemble members and are trained on constrained EnKF analyses. The carrying mechanism is the constraint-augmented Bellman operator with primal-dual reward $R_{\mathrm{PD}}(s,a)=R(s,a)-\lambda(s)\tilde{\zeta}(a)$, where $\tilde{\zeta}(a)=1/\delta E(a)-1/\epsilon$; dual ascent on $\lambda(s)$ makes the energy constraint active in expectation, and constraining the action space to $A_c=\{a:\ a_{\min}\le a_i\le a_{\max}\}$ enforces positivity by construction. This is what lets the learned filter keep energy in the tolerance band and keep $A+\bar{A}>0$ at every step.

What would settle it

Run RL-DAUNCE over a long assimilation window that includes MJO extreme events lying outside the training period, and compare time-averaged RMSE, ensemble energy, and blow-up incidents against constrained EnKF; if energy leaves the prescribed band or RMSE diverges while constrained EnKF stays stable, the central claim fails.

Watch

Extended reading notes

Core claim

The paper's central claim is that physical constraints can be built into a learned data-assimilation filter rather than imposed as post-hoc corrections. Each ensemble member becomes an independent policy network $\pi_{\theta^{(i)}}$ that maps the current state variables, their finite differences, and the time-space coordinates to the next filtered state, trained by regressing onto the constrained EnKF analysis. A primal-dual Lagrangian $L = R(s,a) - \lambda(s)(1/\delta E(a) - 1/\epsilon)$ penalizes energy deviations, with Lagrange multiplier $\lambda(s)$ updated per observation so the constraint is enforced in expectation, while positivity of convective activity $A+\bar{A}>0$ is enforced by bounding the action space. The paper reports that on the stochastic skeleton MJO model, RL-DAUNCE reproduces constrained EnKF's mean states, ensemble spread, and recovered extreme events, keeps energy inside the interval $[0.015,0.08]$, and completes each assimilation step in 1.1 seconds versus 22.96 seconds for constrained EnKF.

Load-bearing premise

The claim rests on the assumption that a policy trained to reproduce constrained EnKF's one-step analysis remains accurate when applied autoregressively over long assimilation cycles, without drift or unquantified error growth.

Editorial extensions

If this is right

  • Assimilation of the MJO from convective-activity observations alone can be performed at about 1.1 seconds per assimilation step, making real-time or large-ensemble applications feasible.
  • Physical consistency is learned rather than clipped: energy conservation and positivity hold at inference without post-hoc projection or re-optimization.
  • Uncertainty quantification is available from the ensemble spread, because inference samples from the ensemble of policies rather than producing a single mean-state estimate.
  • The method transfers to the warm-pool heating profile, recovering the same intermittent extreme events and MJO propagation, suggesting it is not tied to the spatially homogeneous setup.
  • A learned constrained filter can replace constrained EnKF in settings where the teacher's per-step optimization is too costly, provided the training data are available.

Reading between the lines

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

  • The reported 20x speed-up is per assimilation step at inference; the one-time training cost is not quantified, so the full computational advantage depends on amortization over a long enough deployment.
  • Because RL-DAUNCE imitates constrained EnKF, its performance is capped by the teacher; testing on a regime where constrained EnKF is itself degraded would reveal whether the learned policy inherits that bias.
  • The recipe should transfer to other constrained data-assimilation problems by choosing hard bounds for state validity, a conserved functional for the soft constraint, and any teacher filter; the open question is whether the inverse-violation dual update remains well behaved when the deviation functional can take negative values.
  • Replacing the constrained-EnKF teacher with a cheaper or weaker teacher in training would isolate how much of RL-DAUNCE's accuracy comes from the architecture versus from the quality of the distillation target.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes RL-DAUNCE, a reinforcement-learning-based data assimilation method in which each ensemble member is an independent policy network. The policies are trained to reproduce the analysis states produced by a constrained ensemble Kalman filter, while a primal-dual scheme with dynamically adjusted Lagrange multipliers enforces soft physical constraints (energy conservation) and a constrained action space enforces hard bounds (positivity of convective activity). The method is applied to the MJO skeleton model, where the authors report that RL-DAUNCE matches the accuracy and uncertainty quantification of constrained EnKF at a fraction of the computational cost, and that it remains physically consistent over long simulations. The paper also presents a comparison with an unconstrained RL variant to show the role of constraint enforcement.

Significance. If the reported results are valid, RL-DAUNCE could be a useful learned surrogate for constrained ensemble data assimilation, combining the speed of a neural-network filter with physical consistency and ensemble-based uncertainty quantification. The ensemble-as-agents design and the use of action-space constraints are sensible ideas, and the demonstration that energy conservation emerges from the primal-dual training (Figure 5) is a valuable experimental point. However, the paper's central empirical claim currently rests on an evaluation that may be in-sample, and the theoretical presentation contains sign inconsistencies. These issues must be resolved before the contribution can be assessed reliably. The paper does not provide code or machine-checked artifacts, so the empirical claims are not independently verifiable from the manuscript alone.

major comments (4)
  1. [Section 3.4, Eqs. (3.4) to (3.8)] The primal-dual formulation is internally inconsistent. Equation (3.4) defines a Lagrangian for minimizing the MSE, while Eq. (3.5) rewrites it as a reward minus a constraint penalty and the text then states that the agent solves 'max_θ min_λ L.' If L is a reward-based Lagrangian, the policy update should be gradient ascent, but Eq. (3.6) writes θ ← θ − α_θ ∇_θ L, which is gradient descent. Moreover, the dual update in Eq. (3.7) is then replaced by Eq. (3.8), which rewrites the constraint δE ≤ ε as 1/δE > 1/ε. These two constraints are not equivalent for positive δE: δE ≤ ε is equivalent to 1/δE ≥ 1/ε, so the inequality direction in Eq. (3.8) is wrong. This affects the derivation of the KKT conditions and the claimed convergence of the dual step. The authors should present a single, consistent saddle-point formulation with correct signs and update rules.
  2. [Sections 4.3 and 4.5] No train/test split is reported. Section 4.3 states that the RL agent is trained on a dataset generated by constrained EnKF 'at each time step,' and Section 4.5 evaluates RMSE and correlation at days 700, 800, 950, and 1100 without stating whether these days are within the training window. Since the training objective (3.2) regresses directly onto the constrained-EnKF analysis at the next time step, in-sample evaluation can make the policy appear nearly perfect by construction. The authors need to specify the exact training period and separately evaluate on held-out time intervals (or on the warm-pool case as an explicit out-of-sample test, with details). Without this, the headline claim that RL-DAUNCE 'matches constrained EnKF' is not established.
  3. [Section 4.5, Table 2] The evaluation is limited to four selected time instants. There is no time-averaged RMSE or correlation, no error-growth curve, and no long-horizon autoregressive assessment. Because the policy predicts a single next step and is then applied recursively, compounding errors could make the single-step fit misleading. Please add time-averaged skill scores over the full evaluation period and a plot of RMSE versus time (or at least a tabulated average) to substantiate the accuracy claim.
  4. [Sections 3.3 and 4.3] Reproducibility details are missing. The paper discusses PPO in Section 3.2 but never specifies the RL algorithm actually used, the network architecture, the reward function R(s,a) entering Eq. (3.5), the number of training steps, the discount factor, the ensemble size N, or the hyperparameters beyond α_θ and α_λ. These details are needed to reproduce the experiments and to interpret the reported speedups.
minor comments (6)
  1. [Abstract and Introduction] The text says RL-DAUNCE 'outperforms the standard ensemble Kalman filter,' but the experiments show that unconstrained EnKF becomes unstable and diverges; the wording should clarify that the improvement is in maintaining physical consistency and stability rather than in raw RMSE over a stable baseline.
  2. [Section 4.2, Eq. (4.8)] The symbol S in the convective energy term is not defined in the text; please define it or replace it with the source-term notation used elsewhere.
  3. [Section 4.3] The statement that input state variables are normalized to remain in [−1,1] is unclear for variables that are inherently unbounded, such as A before positivity clipping; please explain how the normalization is computed.
  4. [Figure 5] The legend includes 'Unconstrained RL,' but this variant is not described in the experimental setup; please add a sentence explaining how it was trained and how it differs from RL-DAUNCE.
  5. [Section 4.4, Eq. (4.10)] The notation mixes MJO_truth/MJO_est with u_truth/u_est in the same equation; use consistent variable names.
  6. [Throughout] There are several typographical issues, including 'T able 1' and 'T able 2' in table captions, 'we discrete the spatial domain' instead of 'we discretize', and text in Section 2.1 after Eq. (2.8) that is missing spaces. A careful proofreading pass is needed.

Circularity Check

1 steps flagged · score 5.0 of 10

The headline match between RL-DAUNCE and constrained EnKF is the training objective itself (Eq. 3.2), so the central empirical claim is partly a fit; truth-based scores and the speed comparison retain independent content.

  1. fitted input called prediction [Section 3.3, Eq. (3.2); results in Section 4.5, Table 2 and Fig. 3]
    "The learning objective is to minimize the mean squared error (MSE) between the predicted action a(i) (i.e., the estimated state of ensemble i at the next time step) and the reference solution at the next time step, provided by constrained EnKF."

    The policy is trained, by Eq. (3.2), to output the constrained EnKF analysis a* at the next time step. The paper's headline claim that 'RL-DAUNCE matches the performance of constrained EnKF' and the Figure 3 statement that RL-DAUNCE trajectories 'closely follow the constrained EnKF in both the mean state and uncertainty' are reports of exactly this fitted objective. No held-out period is specified in Section 4, so the match at days 700, 800, 950, and 1100 can be a verification of the training fit rather than an independent prediction. The RMSE/Corr values are computed against the true MJO state, which is not used in training, so the comparison to truth provides some independent content; this makes the circularity partial rather than total.

full rationale

The paper is transparent that constrained EnKF generates the training data, and the policy is explicitly an emulator of constrained EnKF. The central circular step is that the reported 'match' between RL-DAUNCE and constrained EnKF is the objective function minimized in Eq. (3.2): minimizing the MSE to the constrained EnKF analysis and then showing that RL-DAUNCE closely follows the constrained EnKF is a goodness-of-fit check, not an independent discovery. The lack of any stated train/test split in Section 4 strengthens this concern, because the evaluation times in Table 2 lie in the same simulation window from which training data are generated. However, the truth-based RMSE and correlation scores, the long-horizon energy plot, and the wall-clock speed comparison are independent of the fitted objective, so the paper is not entirely circular. The self-citations to the authors' prior work (e.g., [19], [69]) are used as method references and supporting statements rather than load-bearing uniqueness theorems, so they do not independently raise the circularity score.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The central claim rests on hand-chosen constraint tolerances, unreported hyperparameters, the teacher-student assumption, and the unverified premise that single-step imitation composes stably. No new physical entities are introduced.

free parameters (3)
  • Energy tolerance interval epsilon = [0.015, 0.08]
    Hand-chosen acceptable range for energy deviations; directly controls the constraint penalty strength and is matched to the training data's observed energy fluctuations.
  • Learning rates alpha_theta and alpha_lambda = not reported
    Primal and dual learning rates in Eqs. (3.6) and (3.8) are never specified; they are free choices that affect convergence and final performance.
  • Ensemble size N = not reported
    The number of policy networks and ensemble members is central to the uncertainty quantification claim, but no value is given in the paper.
assumptions (5)
  • domain assumption The stochastic skeleton model (4.3)-(4.6) is an adequate proxy for the true MJO, and the perfect-model setup means the same model generates both truth and observations.
    All conclusions are demonstrated only in this simulated perfect-model setting; real-world MJO filtering is not addressed.
  • domain assumption The constrained EnKF analysis used to generate training targets is a reliable teacher.
    RL-DAUNCE is trained by regression onto constrained EnKF outputs (Eq. 3.2), so any teacher bias or failure is inherited by the learned filter.
  • ad hoc to paper The energy function (4.8) and the chosen tolerance interval capture the physically relevant conserved quantity despite stochastic forcing.
    The tolerance [0.015, 0.08] is stated rather than derived, and the dual update uses an inverted constraint (1/deltaE) that presumes positive constraint violations.
  • ad hoc to paper Single-step minimization of MSE to the teacher generalizes over long horizons without compounding error.
    The policy is evaluated autoregressively at inference, but the paper reports no long-horizon error accumulation analysis or formal train/test split for extreme events.
  • standard math Standard constrained RL and Bellman contraction results apply to the proposed updates.
    Section 3.5 asserts gamma-contraction of T_lambda and KKT conditions, relying on textbook results [82, 84] without addressing the sign inconsistency in Eqs. (3.5)-(3.6).

how reviews work

0 comments
Cite this review

Pith. "Pith review of RL-DAUNCE: Reinforcement Learning-Driven Data Assimilation with Uncertainty-Aware Constrained Ensembles." pith.science (2026). https://pith.science/paper/5QSOVUJL

@misc{pith2026250505452,
  author       = {Pith},
  title        = {Pith review of: RL-DAUNCE: Reinforcement Learning-Driven Data Assimilation with Uncertainty-Aware Constrained Ensembles},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5QSOVUJL}},
  note         = {Machine review of arXiv:2505.05452}
}
read the original abstract

Machine learning has become a powerful tool for enhancing data assimilation. While supervised learning remains the standard method, reinforcement learning (RL) offers unique advantages through its sequential decision-making framework, which naturally fits the iterative nature of data assimilation by dynamically balancing model forecasts with observations. We develop RL-DAUNCE, a new RL-based method that enhances data assimilation with physical constraints through three key aspects. First, RL-DAUNCE inherits the computational efficiency of machine learning while it uniquely structures its agents to mirror ensemble members in conventional data assimilation methods. Second, RL-DAUNCE emphasizes uncertainty quantification by advancing multiple ensemble members, moving beyond simple mean-state optimization. Third, RL-DAUNCE's ensemble-as-agents design facilitates the enforcement of physical constraints during the assimilation process, which is crucial to improving the state estimation and subsequent forecasting. A primal-dual optimization strategy is developed to enforce constraints, which dynamically penalizes the reward function to ensure constraint satisfaction throughout the learning process. Also, state variable bounds are respected by constraining the RL action space. Together, these features ensure physical consistency without sacrificing efficiency. RL-DAUNCE is applied to the Madden-Julian Oscillation, an intermittent atmospheric phenomenon characterized by strongly non-Gaussian features and multiple physical constraints. RL-DAUNCE outperforms the standard ensemble Kalman filter (EnKF), which fails catastrophically due to the violation of physical constraints. Notably, RL-DAUNCE matches the performance of constrained EnKF, particularly in recovering intermittent signals, capturing extreme events, and quantifying uncertainties, while requiring substantially less computational effort.

Figures

Figures reproduced from arXiv: 2505.05452 by the authors.

Figure 1
Figure 1. Overview of the proposed RL framework for constrained data assimilation. The RL agent ensemble learns to propose physically consistent actions based on EnKF-generated data, with constraints applied via primal-dual optimization during training. The system evolves sequentially in time with learned actions subject to positivity, conservation, and other constraints. Ac =  a ∈ R n [PITH_FULL_IMAGE:figures/full_fig_p009… view at source ↗
Figure 2
Figure 2. Temporal evaluation of variable a and its corresponding observation at a fixed grid point. In generating the EnKF solution, we use an EAKF solver [73] and apply the Gaspari-Cohn localization matrix, which filters out weak correlations beyond a certain distance. Given that our system consists of 64 spatial points, we element-wise multiply this localization matrix with the variables at their respective spatial points.… view at source ↗
Figure 3
Figure 3. Temporal trajectories of the state variables K, R, Z, A, and MJO at a fixed spatial location. Each subplot compares the ground truth (black), the mean and uncertainty ( pm2 standard deviation) from the constrained EnKF (blue), and those from the RL-DAUNCE framework (red). The RL-DAUNCE predictions closely follow the constrained EnKF in both the mean state and uncertainty, demonstrating the RL-DAUNCE’s ability to rep… view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Hovmöller diagrams of the state variables K, R, Z, A, and MJO in the space-time domain. 15 [PITH_FULL_IMAGE:figures/full_fig_p015_4.png]
Figure 5
Figure 5. Figure 5: Comparison of the total energy across ensembles for different methods. RL-DAUNCE successfully conserves total energy through the deployment of our constraint enforcement algorithm. However, without applying the constraint enforcement, even though RL was trained using c…
Figure 6
Figure 6. Figure 6: Results for the warm pool case. Left: Time series of the state variables A and MJO at a fixed spatial point, comparing RL-DAUNCE (red) with the ground truth (black). The shaded regions represent the uncertainty (±2 standard deviation) estimated by RL-DAUNCE. Right: Hov…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

98 extracted references · 71 canonical work pages

  1. [1]

    K. Law, A. Stuart, K. Zygalakis, Data assimilation, Cham, Switzerland: Springer 214 (2015) 52

  2. [2]

    A. J. Majda, J. Harlim, Filtering complex turbulent systems, Cambridge University Press, 2012

  3. [3]

    M. Asch, M. Bocquet, M. Nodet, Data assimilation: methods, algorithms, and applications, SIAM, 2016

  4. [4]

    Evensen, F

    G. Evensen, F. C. Vossepoel, P. J. Van Leeuwen, Data assimilation fundamentals: A unified formulation of the state and parameter estimation problem, Springer Nature, 2022

  5. [5]

    M. A. Freitag, R. W. E. Potthast, Synergy of inverse problems and data assimilation techniques, DE GRUYTER, 2013, pp. 1–54. doi:10.1515/9783110282269.1

  6. [6]

    Chen, Stochastic Methods for Modeling and Predicting Complex Dynamical Systems, Springer, 2023

    N. Chen, Stochastic Methods for Modeling and Predicting Complex Dynamical Systems, Springer, 2023

  7. [7]

    B. Wang, X. Zou, J. Zhu, Data assimilation and its applications, Proceedings of the National Academy of Sciences 97 (2000) 11143–11144

  8. [8]

    Zhong, S

    C. Zhong, S. Cheng, M. Kasoar, R. Arcucci, Reduced-order digital twin and latent data assimilation for global wildfire prediction, Natural hazards and earth system sciences 23 (2023) 1755–1768. 17

Show all 98 references
  1. [9]

    Donato, C

    L. Donato, C. Galletti, A. Parente, Self-updating digital twin of a hydrogen-powered furnace using data assimilation, Applied Thermal Engineering 236 (2024) 121431

  2. [10]

    C. Chen, N. Chen, Y. Zhang, J.-L. Wu, CGKN: A deep learning framework for modeling complex dynamical systems and efficient data assimilation, Journal of Computational Physics 532 (2025) 113950

  3. [11]

    E. Bach, M. Ghil, A multi-model ensemble Kalman filter for data assimilation and forecasting, Journal of Advances in Modeling Earth Systems 15 (2023) e2022MS003123

  4. [12]

    Y. Chen, S. N. Stechmann, Multi-model communication and data assimilation for mitigating model error and improving forecasts, Chinese Annals of Mathematics, Series B 40 (2019) 689–720

  5. [13]

    Stephenson, C

    D. Stephenson, C. Coelho, F. Doblas-Reyes, M. Balmaseda, Forecast assimilation: A unified framework for the combination of multi-model weather and climate predictions, Tellus A: Dynamic Meteorology and Oceanography 57 (2005) 253–264

  6. [14]

    R. E. Kalman, A new approach to linear filtering and prediction problems (1960)

  7. [15]

    M. I. Ribeiro, Kalman and extended Kalman filters: Concept, derivation and properties, Institute for Systems and Robotics 43 (2004) 3736–3741

  8. [16]

    C. L. Keppenne, M. M. Rienecker, Initial testing of a massively parallel ensemble kalman filter with the poseidon isopycnal ocean general circulation model, Monthly weather review 130 (2002) 2951–2965

  9. [17]

    Carrassi, M

    A. Carrassi, M. Ghil, A. Trevisan, F. Uboldi, Data assimilation as a nonlinear dynamical systems problem: Stability and convergence of the prediction-assimilation system, Chaos: An Interdisciplinary Journal of Nonlinear Science 18 (2008). doi:10.1063/1.2909862

  10. [18]

    G. Evensen, Sequential data assimilation with a nonlinear quasi-geostrophic model using monte carlo methods to forecast error statistics, Journal of Geophysical Research: Oceans 99 (1994) 10143–10162

  11. [19]

    N. Chen, A. J. Majda, Filtering the stochastic skeleton model for the madden–julian oscillation, Monthly Weather Review 144 (2016) 501–527

  12. [20]

    R. v. d. Merwe, E. A. Wan, Sigma-point Kalman filters for integrated navigation, in: Proceedings of the 60th annual meeting of the institute of navigation (2004), 2004, pp. 641–654

  13. [21]

    Evensen, The ensemble kalman filter: Theoretical formulation and practical implementation, Ocean dynamics 53 (2003) 343–367

    G. Evensen, The ensemble kalman filter: Theoretical formulation and practical implementation, Ocean dynamics 53 (2003) 343–367

  14. [22]

    P. L. Houtekamer, H. L. Mitchell, Ensemble Kalman filtering, Quarterly Journal of the Royal Meteoro- logical Society: A journal of the atmospheric sciences, applied meteorology and physical oceanography 131 (2005) 3269–3289

  15. [23]

    D. L. T. Anderson, J. Sheinbaum, K. Haines, Data assimilation in ocean models, Reports on Progress in Physics 59 (1996) 1209–1266. doi:10.1088/0034-4885/59/10/001

  16. [24]

    Kalnay, Atmospheric modeling, data assimilation and predictability, Cambridge university press, 2003

    E. Kalnay, Atmospheric modeling, data assimilation and predictability, Cambridge university press, 2003

  17. [25]

    Lievens, G

    H. Lievens, G. De Lannoy, A. Al Bitar, M. Drusch, G. Dumedah, H.-J. H. Franssen, Y. Kerr, S. K. Tomer, B. Martens, O. Merlin, et al., Assimilation of smos soil moisture and brightness temperature products into a land surface model, Remote sensing of environment 180 (2016) 292–304

  18. [26]

    G. J. De Lannoy, R. H. Reichle, K. R. Arsenault, P. R. Houser, S. Kumar, N. E. Verhoest, V. R. Pauwels, Multiscale assimilation of advanced microwave scanning radiometer–eos snow water equivalent and moderate resolution imaging spectroradiometer snow cover fraction observation...

  19. [27]

    X. Li, F. Liu, C. Ma, J. Hou, D. Zheng, H. Ma, Y. Bai, X. Han, H. Vereecken, K. Yang, Q. Duan, C. Huang, Land Data Assimilation: Harmonizing Theory and Data in Land Surface Process Studies, Reviews of Geophysics 62 (2024). doi:10.1029/2022rg000801

  20. [28]

    J. A. Vrugt, H. V. Gupta, W. Bouten, S. Sorooshian, A shuffled complex evolution metropolis algorithm for optimization and uncertainty assessment of hydrologic model parameters, Water resources research 39 (2003)

  21. [29]

    W. A. Lahoz, P. Schneider, Data assimilation: making sense of earth observation, Frontiers in Envi- ronmental Science 2 (2014) 16

  22. [30]

    X. Li, F. Liu, M. Fang, Harmonizing models and observations: Data assimilation in earth system science, Science China Earth Sciences 63 (2020) 1059–1068

  23. [31]

    R. Dong, H. Leng, C. Zhao, J. Song, J. Zhao, X. Cao, A hybrid data assimilation system based on machine learning, Frontiers in Earth Science 10 (2023). doi:10.3389/feart.2022.1012165

  24. [32]

    J. L. Anderson, Localization and sampling error correction in ensemble Kalman filter data assimilation, Monthly Weather Review 140 (2012) 2359–2371

  25. [33]

    J. L. Anderson, Exploring the need for localization in ensemble data assimilation using a hierarchical ensemble filter, Physica D: Nonlinear Phenomena 230 (2007) 99–111

  26. [34]

    Buehner, M

    M. Buehner, M. Charron, Spectral and spatial localization of background-error correlations for data assimilation, Quarterly Journal of the Royal Meteorological Society: A journal of the atmospheric sciences, applied meteorology and physical oceanography 133 (2007) 615–630

  27. [35]

    J. S. Whitaker, T. M. Hamill, Evaluating methods to account for system errors in ensemble data assimilation, Monthly Weather Review 140 (2012) 3078–3089

  28. [36]

    Cheng, C

    S. Cheng, C. Quilodrán-Casas, S. Ouala, A. Farchi, C. Liu, P. Tandeo, R. Fablet, D. Lucor, B. Iooss, J. Brajard, et al., Machine learning with data assimilation and uncertainty quantification for dynamical systems: a review, IEEE/CAA Journal of Automatica Sinica 10 (2023) 1361–1387

  29. [37]

    Revach, N

    G. Revach, N. Shlezinger, X. Ni, A. L. Escoriza, R. J. Van Sloun, Y. C. Eldar, Kalmannet: Neural network aided Kalman filtering for partially known dynamics, IEEE Transactions on Signal Processing 70 (2022) 1532–1547

  30. [38]

    Boudier, A

    P. Boudier, A. Fillion, S. Gratton, S. Gürol, Dan–an optimal data assimilation framework based on machine learning recurrent networks, arXiv preprint arXiv:2010.09694 (2020)

  31. [39]

    Ouala, R

    S. Ouala, R. Fablet, C. Herzet, B. Chapron, A. Pascual, F. Collard, L. Gaultier, Neural network based Kalman filters for the spatio-temporal interpolation of satellite-derived sea surface temperature, Remote Sensing 10 (2018) 1864

  32. [40]

    G. E. Manucharyan, L. Siegelman, P. Klein, A deep learning approach to spatiotemporal sea sur- face height interpolation and estimation of deep currents in geostrophic ocean turbulence, Journal of Advances in Modeling Earth Systems 13 (2021) e2019MS001965

  33. [41]

    C.Mou, L.M.Smith, N.Chen, Combiningstochasticparameterizedreduced-ordermodelswithmachine learning for data assimilation and uncertainty quantification with partial observations, Journal of Advances in Modeling Earth Systems 15 (2023) e2022MS003597

  34. [42]

    Boudier, A

    P. Boudier, A. Fillion, S. Gratton, S. Gürol, S. Zhang, Data assimilation networks, Journal of Advances in Modeling Earth Systems 15 (2023) e2022MS003353. 19

  35. [43]

    G. A. Gottwald, S. Reich, Combining machine learning and data assimilation to forecast dynamical systems from noisy partial observations, Chaos: An Interdisciplinary Journal of Nonlinear Science 31 (2021) 101103

  36. [44]

    Chattopadhyay, E

    A. Chattopadhyay, E. Nabizadeh, E. Bach, P. Hassanzadeh, Deep learning-enhanced ensemble-based data assimilation for high-dimensional nonlinear dynamical systems, Journal of Computational Physics 477 (2023) 111918

  37. [45]

    R. C. Gilbert, M. B. Richman, T. B. Trafalis, L. M. Leslie, Machine learning methods for data assimilation, Computational Intelligence in Architecturing Complex Engineering Systems (2010) 105– 112

  38. [46]

    Y. Wang, H. Yao, S. Zhao, Auto-encoder based dimensionality reduction, Neurocomputing 184 (2016) 232–242

  39. [47]

    S. E. Otto, C. W. Rowley, Linearly recurrent autoencoder networks for learning dynamics, SIAM Journal on Applied Dynamical Systems 18 (2019) 558–593

  40. [48]

    Takeishi, Y

    N. Takeishi, Y. Kawahara, T. Yairi, Learning Koopman invariant subspaces for dynamic mode decom- position, Advances in Neural Information Processing Systems 30 (2017)

  41. [49]

    Brajard, A

    J. Brajard, A. Carrassi, M. Bocquet, L. Bertino, Combining data assimilation and machine learning to emulate a dynamical model from sparse and noisy observations: A case study with the lorenz 96 model, Journal of Computational Science 44 (2020) 101171

  42. [50]

    G. A. Gottwald, S. Reich, Supervised learning from noisy observations: Combining machine-learning techniques with data assimilation, Physica D: Nonlinear Phenomena 423 (2021) 132911

  43. [51]

    Buizza, C

    C. Buizza, C. Q. Casas, P. Nadler, J. Mack, S. Marrone, Z. Titus, C. Le Cornec, E. Heylen, T. Dur, L. B. Ruiz, et al., Data learning: Integrating data assimilation and machine learning, Journal of Computational Science 58 (2022) 101525

  44. [52]

    Farchi, P

    A. Farchi, P. Laloyaux, M. Bonavita, M. Bocquet, Using machine learning to correct model error in data assimilation and forecast applications, Quarterly Journal of the Royal Meteorological Society 147 (2021) 3067–3084

  45. [53]

    Bocquet, J

    M. Bocquet, J. Brajard, A. Carrassi, L. Bertino, Bayesian inference of chaotic dynamics by merging data assimilation, machine learning and expectation-maximization, arXiv preprint arXiv:2001.06270 (2020)

  46. [54]

    TSUYUKI, R

    T. TSUYUKI, R. TAMURA, Nonlinear data assimilation by deep learning embedded in an ensemble Kalman filter, Journal of the Meteorological Society of Japan. Ser. II (2022)

  47. [55]

    Maulik, V

    R. Maulik, V. Rao, J. Wang, G. Mengaldo, E. Constantinescu, B. Lusch, P. Balaprakash, I. Foster, R. Kotamarthi, AIEADA 1.0: Efficient high-dimensional variational data assimilation with machine- learned reduced-order models, Geoscientific Model Development Discussions 2022 (2022) 1–20

  48. [56]

    S. G. Penny, T. A. Smith, T.-C. Chen, J. A. Platt, H.-Y. Lin, M. Goodliff, H. D. Abarbanel, Integrating recurrent neural networks with data assimilation for scalable data-driven state estimation, Journal of Advances in Modeling Earth Systems 14 (2022) e2021MS002843

  49. [57]

    Pawar, S

    S. Pawar, S. E. Ahmed, O. San, A. Rasheed, I. M. Navon, Long short-term memory embedded nudging schemes for nonlinear data assimilation of geophysical flows, Physics of Fluids 32 (2020) 076606

  50. [58]

    M. Tang, Y. Liu, L. J. Durlofsky, A deep-learning-based surrogate model for data assimilation in dynamic subsurface flow problems, Journal of Computational Physics 413 (2020) 109456. 20

  51. [59]

    Janjić, D

    T. Janjić, D. McLaughlin, S. E. Cohn, M. Verlaan, Conservation of mass and preservation of positivity with ensemble-type kalman filter algorithms, Monthly Weather Review 142 (2014) 755–773

  52. [60]

    Chattopadhyay, M

    A. Chattopadhyay, M. Mustafa, P. Hassanzadeh, E. Bach, K. Kashinath, Towards physically consis- tent data-driven weather forecasting: Integrating data assimilation with equivariance-preserving spatial transformers in a case study with ERA5, Geoscientific Model Development Disc...

  53. [61]

    A. J. Majda, N. Chen, Model error, information barriers, state estimation and prediction in complex multiscale systems, Entropy 20 (2018) 644

  54. [62]

    L. P. Kaelbling, M. L. Littman, A. W. Moore, Reinforcement learning: A survey, Journal of artificial intelligence research 4 (1996) 237–285

  55. [63]

    M. A. Wiering, M. Van Otterlo, Reinforcement learning, Adaptation, learning, and optimization 12 (2012) 729

  56. [64]

    Raffin, A

    A. Raffin, A. Hill, A. Gleave, A. Kanervisto, M. Ernestus, N. Dormann, Stable-baselines3: Reliable reinforcement learning implementations, Journal of machine learning research 22 (2021) 1–8

  57. [65]

    Y.Feng, F.S.Schranner, J.Winter, N.A.Adams, Adeepreinforcementlearningframeworkfordynamic optimization of numerical schemes for compressible flow simulations, Journal of Computational Physics 493 (2023) 112436

  58. [66]

    M. A. E. R. Hammoud, N. Raboudi, E. S. Titi, O. Knio, I. Hoteit, Data assimilation in chaotic systems using deep reinforcement learning, Journal of Advances in Modeling Earth Systems 16 (2024) e2023MS004178

  59. [67]

    Janjić, Y

    T. Janjić, Y. Zeng, Weakly constrained LETKF for estimation of hydrometeor variables in convective- scale data assimilation, Geophysical Research Letters 48 (2021) e2021GL094962

  60. [68]

    Ruckstuhl, T

    Y. Ruckstuhl, T. Janjić, S. Rasp, Training a convolutional neural network to conserve mass in data assimilation, Nonlinear Processes in Geophysics 28 (2021) 111–119

  61. [69]

    Gleiter, T

    T. Gleiter, T. Janjić, N. Chen, Ensemble kalman filter based data assimilation for tropical waves in the mjo skeleton model, Quarterly Journal of the Royal Meteorological Society 148 (2022) 1035–1056

  62. [70]

    Burgers, P

    G. Burgers, P. Jan van Leeuwen, G. Evensen, Analysis scheme in the ensemble kalman filter, Monthly weather review 126 (1998) 1719–1724

  63. [71]

    C. H. Bishop, B. J. Etherton, S. J. Majumdar, Adaptive sampling with the ensemble transform kalman filter. part i: Theoretical aspects, Monthly weather review 129 (2001) 420–436

  64. [72]

    B. R. Hunt, E. J. Kostelich, I. Szunyogh, Efficient data assimilation for spatiotemporal chaos: A local ensemble transform Kalman filter, Physica D: Nonlinear Phenomena 230 (2007) 112–126

  65. [73]

    J. L. Anderson, An ensemble adjustment Kalman filter for data assimilation, Monthly weather review 129 (2001) 2884–2903

  66. [74]

    J. S. Whitaker, T. M. Hamill, Ensemble data assimilation without perturbed observations, Monthly weather review 130 (2002) 1913–1924

  67. [75]

    S. J. Thomas, J. Hacker, J. Anderson, A robust formulation of the ensemble Kalman filter, Quarterly Journal of the Royal Meteorological Society: A journal of the atmospheric sciences, applied meteorology and physical oceanography 135 (2009) 507–521

  68. [76]

    E. J. Fertig, B. R. Hunt, E. Ott, I. Szunyogh, Assimilating non-local observations with a local ensemble Kalman filter, Tellus A: Dynamic Meteorology and Oceanography 59 (2007) 719–730. 21

  69. [77]

    W. F. Campbell, C. H. Bishop, D. Hodyss, Vertical covariance localization for satellite radiances in ensemble Kalman filters, Monthly Weather Review 138 (2010) 282–290

  70. [78]

    P. L. Houtekamer, F. Zhang, Review of the ensemble Kalman filter for atmospheric data assimilation, Monthly Weather Review 144 (2016) 4489–4532

  71. [79]

    Ladosz, L

    P. Ladosz, L. Weng, M. Kim, H. Oh, Exploration in deep reinforcement learning: A survey, Information Fusion 85 (2022) 1–22

  72. [80]

    E. C.-O. DAD, Introduction to reinforcement learning (2019)

  73. [81]

    Paternain, M

    S. Paternain, M. Calvo-Fullana, L. F. Chamon, A. Ribeiro, Safe policies for reinforcement learning via primal-dual methods, IEEE Transactions on Automatic Control 68 (2022) 1321–1336

  74. [82]

    S. P. Boyd, L. Vandenberghe, Convex optimization, Cambridge university press, 2004

  75. [83]

    S. S. Mousavi, M. Schukat, E. Howley, Deep reinforcement learning: an overview, in: Proceedings of SAI Intelligent Systems Conference (IntelliSys) 2016: Volume 2, Springer, 2018, pp. 426–440

  76. [84]

    M. G. Bellemare, G. Ostrovski, A. Guez, P. Thomas, R. Munos, Increasing the action gap: New operators for reinforcement learning, in: Proceedings of the AAAI conference on artificial intelligence, volume 30, 2016

  77. [85]

    Paternain, M

    S. Paternain, M. Calvo-Fullana, L. F. Chamon, A. Ribeiro, Learning safe policies via primal-dual methods, in: 2019 IEEE 58th Conference on Decision and Control (CDC), IEEE, 2019, pp. 6491–6497

  78. [86]

    J. Hu, X. Liu, Z.-W. Wen, Y.-X. Yuan, A brief introduction to manifold optimization, Journal of the Operations Research Society of China 8 (2020) 199–248

  79. [87]

    Zhang, Madden-Julian oscillation, Reviews of Geophysics 43 (2005)

    C. Zhang, Madden-Julian oscillation, Reviews of Geophysics 43 (2005)

  80. [88]

    Zhang, Á

    C. Zhang, Á. Adames, B. Khouider, B. Wang, D. Yang, Four theories of the Madden-Julian oscillation, Reviews of Geophysics 58 (2020) e2019RG000685

  81. [89]

    Thual, A

    S. Thual, A. J. Majda, S. N. Stechmann, A stochastic skeleton model for the MJO, Journal of the Atmospheric Sciences 71 (2014) 697–715

  82. [90]

    A. J. Majda, S. N. Stechmann, The skeleton of tropical intraseasonal oscillations, Proceedings of the National Academy of Sciences 106 (2009) 8417–8422

  83. [91]

    A. J. Majda, S. N. Stechmann, Nonlinear dynamics and regional variations in the MJO skeleton, Journal of the Atmospheric Sciences 68 (2011) 3053–3071

  84. [92]

    J. P. Stachnik, D. E. Waliser, A. J. Majda, S. N. Stechmann, S. Thual, Evaluating MJO event ini- tiation and decay in the skeleton model using an RMM-like index, Journal of Geophysical Research: Atmospheres 120 (2015) 11–486

  85. [93]

    H. R. Ogrosky, S. N. Stechmann, The MJO skeleton model with observation-based background state and forcing, Quarterly Journal of the Royal Meteorological Society 141 (2015) 2654–2669

  86. [94]

    S. N. Stechmann, A. J. Majda, Identifying the skeleton of the madden–julian oscillation in observational data, Monthly Weather Review 143 (2015) 395–416

  87. [95]

    D. M. Frierson, A. J. Majda, O. M. Pauluis, Large scale dynamics of precipitation fronts in the tropical atmosphere: A novel relaxation limit, Comm. Math. Sci. 2 (2004) 591626

  88. [96]

    S. Chen, S. N. Stechmann, Nonlinear traveling waves for the skeleton of the madden-julian oscillation, arXiv preprint arXiv:1510.00880 (2015). 22

  89. [97]

    A. J. Majda, S. N. Stechmann, S. Chen, H. R. Ogrosky, S. Thual, Tropical Intraseasonal Variability and the Stochastic Skeleton Method, Springer, 2019

  90. [98]

    Székely, D

    E. Székely, D. Giannakis, A. J. Majda, Extraction and predictability of coherent intraseasonal signals in infrared brightness temperature data, Climate Dynamics 46 (2016) 1473–1502. Appendix A. Dual Optimization and Gradient-Based Updates It is important to emphasize that whil...

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.