Pith. sign in

REVIEW 3 major objections 5 minor 2 cited by

A Hybrid Deep-Learning Model for El Ni\~no Southern Oscillation in the Low-Data Regime

T0 review · 3 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read A hybrid linear-plus-neural model beats full deep learning for El Niño forecasts on century-long records.

desk verdict Solid hybrid ENSO model paper whose low-data headline needs resampling error bars before it fully lands. read the letter →

arxiv 2412.03743 v2 pith:F47B7VAG submitted 2024-12-04 cs.LG physics.ao-ph

Jakob Schloer , Matthew Newman , Jannik Thuemmel , Antonietta Capotondi , Bedartha Goswami This is my paper · ORCID
classification cs.LGphysics.ao-ph
keywords ENSOforecastinghybridmodelLinearInverseLSTMlow-dataregimeasymmetryprobabilisticCESM2
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

This paper asks whether El Niño forecasts can remain skillful when only about a century of training data is available, the regime of real observations rather than thousand-year climate simulations. The authors' answer is a hybrid: start with a cyclostationary Linear Inverse Model (CS-LIM), a linear stochastic model that already captures the predictable linear dynamics from short records, and train an LSTM to learn the CS-LIM's forecast residuals. In tests on a 2000-year CESM2 pre-industrial simulation, the hybrid is more accurate than the CS-LIM alone and also beats a full LSTM trained on the same data when training sets are 50 to 300 years long. The improvement is concentrated at leads beyond about six months and in the western tropical Pacific, and the hybrid reproduces the asymmetry between warm and cold ENSO events that the linear model cannot. If this transfers to observations, hybrid linear-plus-neural models could be a practical path for seasonal prediction in data-scarce settings.

What carries the argument

The load-bearing object is the residual-corrected cyclostationary LIM. The CS-LIM is estimated from the leading principal components of tropical Pacific SSTA and SSHA, with a separate linear operator and noise covariance for each calendar month; it supplies the linear forecast and its theoretical predictability. The LSTM takes the CS-LIM's 16-member ensemble forecast sequence, applies a learned affine month conditioning to its latent state, and outputs a correction added to the LIM forecast, trained end-to-end by minimizing the continuous ranked probability score (CRPS). The LIM's optimal initial condition, defined as the singular vector of the forecast propagator with the largest singular value, provides the paper's tool for showing that predictable growth in the linear model also controls the hybrid's skill.

What would settle it

Train the same LIM-LSTM architecture on data generated by a purely linear Gaussian process with known dynamics: if the hybrid still beats the CS-LIM, the residual network is correcting linear misspecification rather than capturing nonlinear physics; a complementary check is whether the western-Pacific skill gain survives on observed reanalysis data rather than CESM2 alone.

Watch

Extended reading notes

Core claim

The central discovery is that the predictable part of ENSO's nonlinearity, especially its warm-cold asymmetry, can be extracted and forecast by a residual LSTM placed on top of a cyclostationary Linear Inverse Model. With roughly a century of training data, the LIM-LSTM has higher deterministic and probabilistic skill (RMSE and CRPS) than both the CS-LIM and a full LSTM, and it reaches the CS-LIM's skill with 50 to 100 years of training while the full LSTM needs about 300 years. The hybrid also shows that the LIM's optimal initial conditions still identify which forecast states are most predictable, so the linear model's predictability theory carries over to the nonlinear correction. Forecasts of warm and cold events from the hybrid show the observed zonal dipole in asymmetry, indicating that the network is adding nonlinear dynamics rather than just more linear skill.

Load-bearing premise

The load-bearing premise is that the LSTM's improvement over the CS-LIM represents genuine predictable nonlinear ocean dynamics, not the linear model's unfinished business or the climate model's biased variability.

Editorial extensions

If this is right

  • With 50 to 100 years of monthly training data, the hybrid's 12-month Niño4 forecast anomaly correlation exceeds 0.5, while a full LSTM stays below 0.4 at the same data size.
  • The full LSTM needs roughly 300 years of training to match the CS-LIM's skill, so the hybrid is the data-efficient choice for observational records.
  • Skill gains over the CS-LIM are significant at leads above 6 months, strongest in the western tropical Pacific at 9 to 18 month leads.
  • The hybrid reproduces the warm-minus-cold zonal dipole in ENSO events, meaning the asymmetry is predictable and not just a statistical property of the climate.
  • Because LIM-based optimal initial conditions predict the hybrid's skill, forecasts can carry an a priori estimate of their own reliability.

Reading between the lines

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

  • If the same design is trained on multiple climate models or directly on reanalyses, the western-Pacific improvement could be separated from CESM2's known overestimate of SSTA variability there.
  • The residual-correction idea should transfer to other subseasonal-to-seasonal targets with short observational records, wherever a linear stochastic core already captures most of the skill.
  • A direct decomposition of the LSTM's learned correction could test whether it corresponds to known nonlinear ENSO mechanisms such as state-dependent growth or skewness of extremes.
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

3 major / 5 minor

Summary. The paper proposes a hybrid LIM-LSTM model for probabilistic ENSO forecasting in the tropical Pacific, where an LSTM learns a residual correction to a cyclostationary Linear Inverse Model (CS-LIM) applied to SSTA and SSHA principal components. Using a 2000-year CESM2 pre-industrial control simulation, the models are trained on random subsets of 50-1500 years and evaluated on a 200-year test period with RMSE and CRPS skill scores. The headline results are that the hybrid exceeds the CS-LIM and a full LSTM at O(100 yr) training lengths, that the hybrid matches or exceeds fully deep-learning baselines at 1500 years with fewer parameters, and that the nonlinear models capture warm-cold ENSO asymmetry that the linear model misses.

Significance. If the results hold, the paper makes a useful contribution: it demonstrates a data-efficient hybrid that combines a physics-based linear stochastic model with a neural network residual correction, and it evaluates the approach with probabilistic CRPS, 16-member ensembles, multiple training repetitions, and bootstrap significance tests. The optimal-initial-condition (OIC) predictability analysis is a novel and valuable link between linear predictability theory and neural-network forecasts. The paper also releases code, which supports reproducibility. The main caveats are the robustness of the low-data comparison and the interpretation of the learned residuals as nonlinear dynamics.

major comments (3)
  1. [Section 2.2, Figs. 1 and 5c-d] The central claim that the hybrid exceeds a full LSTM at O(100 yr) training data is not robust because the random subsets of the 1500-year training set are not resampled. The error bars in Fig. 5 reflect repeated training runs with varied weight initialization and data shuffling, not redrawing which 50-100 years are used. Since the CS-LIM operator and ENSO event diversity are strongly sample-dependent at these record lengths, a favorable subset could drive the apparent advantage. Please redraw the subsets (e.g., 10-20 independent draws per training length), report the subset-to-subset variability, and state the number of draws used.
  2. [Section 2.1, Eq. (9)] The attribution of skill improvement to 'predictable nonlinearities' is not supported without a control. The LSTM residual may be correcting linear mis-specification caused by EOF truncation (20 SSTA plus 10 SSHA PCs) or by a CS-LIM that omits some linear dynamics; the paper's own statement in Section 2.1 ('to the extent that we have tried to ensure...') acknowledges this. A concrete test would be to train a linear residual model (e.g., ridge regression on the same PC inputs) and show that the LSTM adds nonlinear skill beyond it, or to demonstrate that the learned corrections are not linearly predictable from the input PCs.
  3. [Section 2.4, Fig. 7] The asymmetry claim is based on composites of April-initiated states with top-10% absolute optimal initial growth, but the number of warm and cold events is not reported, and the bootstrap significance test is applied grid-point-wise without accounting for multiple comparisons or event-level autocorrelation. With a single 200-year test segment, the event sample is small; please report the event counts and add an event-based bootstrap (resample events rather than grid points) to quantify the uncertainty of the warm-minus-cold dipole.
minor comments (5)
  1. [Section 4.7, Eqs. (12)-(13)] There is a duplicated equation number and an empty (13); the loss function should be presented in a single numbered equation.
  2. [Figure 3 caption] The caption text 'differences in RMSE skill scores relative to the Hybrid model' conflicts with the description of red as improvement in the LIM-LSTM model; it should say 'relative to the CS-LIM'.
  3. [Section 4.1] The training/validation/test split is described as years 1-1500, 1500-1800, and 1800-2000, which double-counts year 1500; clarify the intended boundaries (e.g., 1-1500, 1501-1800, 1801-2000).
  4. [Sections 4.5 and 4.6] The cross-references to 'sec. 4.24.3' are mangled; they should refer to Sections 4.3 and 4.4.
  5. [Section 4.2, Eq. (1)] The recurrence notation in Eq. (1), especially the nested f_{m(t-delta)}, is confusing; please define the history horizon t_hist and the autoregressive structure in words.

Circularity Check

1 steps flagged · score 3.0 of 10

Main skill comparison is not circular, but the mechanistic attribution of LIM-LSTM skill gains to 'predictable nonlinearities' is a definitional relabeling of the fitted residual term.

  1. self definitional [Section 2.1, Section 4.5, Eq. (9)]
    "While the LIM captures the predictable linear dynamics, the LSTM learns the residuals between the LIM predictions and the actual data, thus the nonlinear dynamics. ... Improved forecast skill of the LIM-LSTM, relative to the LIM itself, can thus be attributed to its ability to capture predictable nonlinearities of the tropical Pacific Ocean dynamics. ... We anticipate that the skill improvements can be largely attributed to predictable nonlinearities to the extent that we have tried to ensure that all known linear dynamics are captured by the CS-LIM."

    The final hybrid forecast is defined as z(t+τ) = z_LIM(t+τ) + z_res(t+τ) (Eq. 9), and z_res is trained to minimize the CRPS between the LIM forecast plus correction and the target data. Any skill gain of the hybrid over the LIM is therefore, by construction, exactly the contribution of the fitted residual term. Calling that contribution 'predictable nonlinearity' is valid only if the CS-LIM already captures all linear dynamics; the paper does not establish this independently and explicitly hedges ('to the extent that we have tried to ensure'). The mechanistic conclusion that the improvement is due to nonlinearity is thus a definitional relabeling of the fitted residual rather than an independently demonstrated nonlinear effect.

full rationale

The headline forecast-skill comparisons are evaluated on an independent 200-year test period, with models trained on 50-1500 year subsets, so the central claim that the hybrid exceeds both the CS-LIM and the full LSTM is not circular in itself. The missing subset-resampling variability is a robustness concern, not a circularity. The paper's self-citations to Shin et al. and other LIM literature are method citations and are not load-bearing in a way that forces the results. The one genuinely circular element is the interpretation of the LSTM residual as 'predictable nonlinearity': because the LSTM is fit to whatever the LIM misses, the skill improvement over the LIM is definitionally due to the residual term, and labeling that term 'nonlinear' presupposes the unverified assumption that the CS-LIM is a complete linear model. This affects the mechanistic interpretation and the asymmetry attribution, but not the independent skill evaluation, so the overall circularity is moderate rather than pervasive.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central claims depend on a few manually chosen hyperparameters (gamma, EOF truncation, ensemble size) and on the representativeness of CESM2. No new physical entities are introduced. The key assumptions are the completeness of the CS-LIM and the transferability of CESM2 results to the real ocean.

free parameters (3)
  • CRPS lead-time decay weight (gamma) = 0.65
    Set empirically in Section 4.7 to down-weight longer lead times in the loss function. The paper does not report sensitivity to this choice.
  • Number of EOF PCs (SSTA, SSHA) = 20 SSTA, 10 SSHA
    Chosen for the CS-LIM and LSTM inputs. Section 4.3 notes higher-order PCs do not affect results, but no supporting experiment is shown.
  • Ensemble size = 16
    Number of ensemble members generated by LIM and used by LSTM. Fixed without a sensitivity study.
assumptions (4)
  • standard math The LIM assumptions of statistical stationarity and exponential decay of autocorrelation hold for the detrended CESM2 anomalies.
    Invoked in Section 4.4 to estimate the linear operator L and noise covariance Q via the fluctuation-dissipation relationship (Eq. 5-6).
  • ad hoc to paper The residual between the CS-LIM forecast and the target is dominated by predictable nonlinear dynamics that an LSTM can learn.
    Section 2.1: the improvement is attributed to 'predictable nonlinearities' based on the completeness of the CS-LIM. This is an interpretive assumption, not directly proven.
  • domain assumption The CESM2 pre-industrial control simulation faithfully represents the real ENSO dynamics, including warm-cold asymmetry and the behavior in low-data regimes.
    All experiments use CESM2 data. The paper acknowledges this limitation in the Discussion and does not test on observations.
  • domain assumption The LSTM nonlinear correction generalizes to the held-out test period.
    Standard supervised learning assumption; the test set is from the same model and same statistical distribution as training, so generalization within the simulation is plausible but not proven across models or real climate.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Hybrid Deep-Learning Model for El Ni\~no Southern Oscillation in the Low-Data Regime." pith.science (2026). https://pith.science/paper/F47B7VAG

@misc{pith2026241203743,
  author       = {Pith},
  title        = {Pith review of: A Hybrid Deep-Learning Model for El Ni\~no Southern Oscillation in the Low-Data Regime},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/F47B7VAG}},
  note         = {Machine review of arXiv:2412.03743}
}
read the original abstract

While deep-learning models have demonstrated skillful El Ni\~no Southern Oscillation (ENSO) forecasts up to one year in advance, they are predominantly trained on climate model simulations that provide thousands of years of training data at the expense of introducing climate model biases. Simpler Linear Inverse Models (LIMs) trained on the much shorter observational record also make skillful ENSO predictions but do not capture predictable nonlinear processes. This motivates a hybrid approach, combining the LIMs modest data needs with a deep-learning non-Markovian correction of the LIM. For O(100 yr) datasets, our resulting Hybrid model is more skillful than the LIM while also exceeding the skill of a full deep-learning model. Additionally, while the most predictable ENSO events are still identified in advance by the LIM, they are better predicted by the Hybrid model, especially in the western tropical Pacific for leads beyond about 9 months, by capturing the subsequent asymmetric (warm versus cold phases) evolution of ENSO.

Figures

Figures reproduced from arXiv: 2412.03743 by the authors.

Figure 1
Figure 1. Variations in forecast skill over training data length. The forecast skill of the CS-LIM, LIM-LSTM hybrid model, and LSTM changes with number of years in the training data. Models are trained on random subsets, ranging from 50 - 1500 years, of the training set. The anomaly correlation coefficient (ACC) of the Niño4-index is computed for a forecast lead time of 12-months over the 200-year test period (see Methods Sec… view at source ↗
Figure 2
Figure 2. RMSE and CRPS skill scores of the LIM versions and our LIM-LSTM model. Skill scores for RMSE (a) and CRPS (b) across various LIM versions and the LIM-LSTM model are evaluated over forecast lead time (τ) using the average SSTA in the Niño4 region on the test set. The progression in LIM versions from the stationary (ST)-LIM, which uses SSTA data and does not include seasonally-varying operators, to the more advanced c… view at source ↗
Figure 3
Figure 3. Spatial distribution of skill improvement. RMSE skill score of SSTA and SSHA for the τ = 12 month forecast of CS-LIM (a, c) and the differences in RMSE skill scores relative to the Hybrid model (b, d). Red colors indicate an improvement in skill in the LIM-LSTM model, while blue colors indicate a decrease in skill. Using a two-sided t-test, we evaluate the significance of the difference between the 1000 randomly boo… view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Example El Niño forecast: Example of a forecast initialized 12 months prior to an El Niño event exemplar. The Niño4 mean (solid line) and spread (shading) of the 16 ensemble members of the CS-LIM and LIM-LSTM forecast are shown in a. The dashed line in a indicates the …
Figure 5
Figure 5. Figure 5: Skill of deep learning baselines. Same as [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Predictability is captured by linear optimals. The optimal initial condition (OIC) (a) of the CS-LIM for a 12-month lead forecast initialized in April evolves into an El Niño-like pattern after 12 months (b). Forecasts are initialized from states in the test set whose …
Figure 7
Figure 7. Figure 7: Nonlinear models capture ENSO asymmetries. The 12-month evolutions of states initialized in April with the absolute largest optimal initial growth (>90%) show warm and cold patterns. The average target state of warm-cold patterns (a) and warm+cold patterns (b) are depi…
Figure 8
Figure 8. Figure 8: Schematic Representation of the LIM-LSTM model. First, the initial state at time t is projected onto the PCs. This is followed by an ensemble forecast using the CS-LIM which is conditioned on the forecast months. Subsequently, the LSTM adjusts each ensemble member of t…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. AIFS-SUBS: Extending Data-Driven Forecasting to Sub-Seasonal Timescales

    physics.ao-ph 2026-07 conditional novelty 6.0 of 10

    AIFS-SUBS matches IFS probabilistic skill at weeks 2–6 with reduced biases, extends skilful MJO OLR forecasts by eight days, and runs at ~200× lower energy cost.

  2. Climate network and complexity approach predict neutral ENSO event for 2025

    physics.ao-ph 2025-01 conditional novelty 3.0 of 10

    The paper forecasts a neutral ENSO event for 2025/26, with 69.6% probability, based on climate-network and complexity methods plus an ONI-based logistic regression.

Reference graph

Works this paper leans on

72 extracted references · 42 canonical work pages · cited by 2 Pith papers

  1. [1]

    J. R. Albers and M. Newman. A Priori Identification of Skillful Extratropical Subseasonal Forecasts . Geophysical Research Letters, 46 0 (21): 0 12527--12536, 2019. ISSN 1944-8007. doi:10.1029/2019GL085270

    work page doi:10.1029/2019gl085270 2019

  2. [2]

    J. R. Albers and M. Newman. Subseasonal predictability of the North Atlantic Oscillation . Environ. Res. Lett., 16 0 (4): 0 044024, Mar. 2021. ISSN 1748-9326. doi:10.1088/1748-9326/abe781

    work page doi:10.1088/1748-9326/abe781 2021

  3. [3]

    Bauer, A

    P. Bauer, A. Thorpe, and G. Brunet. The quiet revolution of numerical weather prediction. Nature, 525 0 (7567): 0 47--55, Sept. 2015. ISSN 1476-4687. doi:10.1038/nature14956

    doi:10.1038/nature14956 2015

  4. [4]

    Ben-Bouallegue , M

    Z. Ben-Bouallegue , M. C. A. Clare, L. Magnusson, E. Gascon, M. Maier-Gerber , M. Janousek, M. Rodwell, F. Pinault, J. S. Dramsch, S. T. K. Lang, B. Raoult, F. Rabier, M. Chevallier, I. Sandu, P. Dueben, M. Chantry, and F. Pappenberger. The rise of data-driven weather forecasting, July 2023

    work page 2023

  5. [5]

    J. D. Beverley, M. Newman, and A. Hoell. Rapid Development of Systematic ENSO-Related Seasonal Forecast Errors . Geophysical Research Letters, 50 0 (10): 0 e2022GL102249, 2023. ISSN 1944-8007. doi:10.1029/2022GL102249

    work page doi:10.1029/2022gl102249 2023

  6. [6]

    K. Bi, L. Xie, H. Zhang, X. Chen, X. Gu, and Q. Tian. Accurate medium-range global weather forecasting with 3D neural networks. Nature, 619 0 (7970): 0 533--538, July 2023. ISSN 1476-4687. doi:10.1038/s41586-023-06185-3

    doi:10.1038/s41586-023-06185-3 2023

  7. [7]

    S. R. Cachay, E. Erickson, A. F. C. Bucker, E. Pokropek, W. Potosnak, S. Bire, S. Osei, and B. L \"u tjens. The World as a Graph : Improving El Ni no Forecasts with Graph Neural Networks , May 2021. URL http://arxiv.org/abs/2104.05089

    arXiv 2021

  8. [8]

    Callahan and J

    C. Callahan and J. S. Mankin. Persistent effect of El Ni \ n o on global economic growth Science , 2023. URL https://www.science.org/doi/10.1126/science.adf2983

    doi:10.1126/science.adf2983 2023

Show all 72 references
  1. [9]

    Capotondi

    A. Capotondi. ENSO diversity in the NCAR CCSM4 climate model: Enso Diversity in the NCAR CCSM4 . J. Geophys. Res. Oceans, 118 0 (10): 0 4755--4770, Oct. 2013. ISSN 21699275. doi:10.1002/jgrc.20335

    2013 doi

  2. [10]

    Capotondi and P

    A. Capotondi and P. D. Sardeshmukh. Optimal precursors of different types of ENSO events. Geophysical Research Letters, 42 0 (22): 0 9952--9960, 2015. ISSN 1944-8007. doi:10.1002/2015GL066171

    2015 doi

  3. [11]

    Capotondi, A

    A. Capotondi, A. T. Wittenberg, M. Newman, E. D. Lorenzo, J. Y. Yu, P. Braconnot, J. Cole, B. Dewitte, B. Giese, E. Guilyardi, F. F. Jin, K. Karnauskas, B. Kirtman, T. Lee, N. Schneider, Y. Xue, and S. W. Yeh. Understanding enso diversity. Bulletin of the American Meteorologic...

    2015 doi

  4. [12]

    Capotondi, C

    A. Capotondi, C. Deser, A. S. Phillips, Y. Okumura, and S. M. Larson. ENSO and Pacific Decadal Variability in the Community Earth System Model Version 2. Journal of Advances in Modeling Earth Systems, 12 0 (12): 0 e2019MS002022, 2020. ISSN 1942-2466. doi:10.1029/2019MS002022

    2020 doi

  5. [13]

    C. Chen, M. A. Cane, N. Henderson, D. E. Lee, D. Chapman, D. Kondrashov, and M. D. Chekroun. Diversity, Nonlinearity , Seasonality , and Memory Effect in ENSO Simulation and Prediction Using Empirical Model Reduction . Journal of Climate, 29 0 (5): 0 1809--1830, Mar. 2016. ISS...

    2016 doi

  6. [14]

    Chen, Fei-Fei-Jin , S

    H.-C. Chen, Fei-Fei-Jin , S. Zhao, A. T. Wittenberg, and S. Xie. ENSO Dynamics in the E3SM-1-0 , CESM2 , and GFDL-CM4 Climate Models . Journal of Climate, 34 0 (23): 0 9365--9384, Dec. 2021. ISSN 0894-8755, 1520-0442. doi:10.1175/JCLI-D-21-0355.1

    2021 doi

  7. [15]

    Danabasoglu, J.-F

    G. Danabasoglu, J.-F. Lamarque, J. Bacmeister, D. A. Bailey, A. K. DuVivier, J. Edwards, L. K. Emmons, J. Fasullo, R. Garcia, A. Gettelman, C. Hannay, M. M. Holland, W. G. Large, P. H. Lauritzen, D. M. Lawrence, J. T. M. Lenaerts, K. Lindsay, W. H. Lipscomb, M. J. Mills, R. Ne...

    2020

  8. [16]

    Dommenget, T

    D. Dommenget, T. Bayr, and C. Frauen. Analysis of the non-linearity in the pattern and time evolution of El Ni \ n o southern oscillation. Clim Dyn, 40 0 (11): 0 2825--2847, June 2013. ISSN 1432-0894. doi:10.1007/s00382-012-1475-0

    2013 doi

  9. [17]

    Frankignoul and K

    C. Frankignoul and K. Hasselmann. Stochastic climate models, Part II Application to sea-surface temperature anomalies and thermocline variability. Tellus, 29 0 (4): 0 289--305, 1977. ISSN 2153-3490. doi:10.1111/j.2153-3490.1977.tb00740.x

    1977

  10. [18]

    Geng and F.-F

    L. Geng and F.-F. Jin. ENSO Diversity Simulated in a Revised Cane-Zebiak Model . Frontiers in Earth Science, 10, 2022. ISSN 2296-6463. URL https://www.frontiersin.org/articles/10.3389/feart.2022.899323

    2022

  11. [19]

    Gneiting and A

    T. Gneiting and A. E. Raftery. Strictly Proper Scoring Rules , Prediction , and Estimation . Journal of the American Statistical Association, 102 0 (477): 0 359--378, Mar. 2007. ISSN 0162-1459. doi:10.1198/016214506000001437

    2007 doi

  12. [20]

    Gneiting, A

    T. Gneiting, A. E. Raftery, A. H. Westveld, and T. Goldman. Calibrated Probabilistic Forecasting Using Ensemble Model Output Statistics and Minimum CRPS Estimation . Monthly Weather Review, 133 0 (5): 0 1098--1118, May 2005. ISSN 1520-0493, 0027-0644. doi:10.1175/MWR2904.1

    2005 doi

  13. [21]

    H. Goel, I. Melnyk, and A. Banerjee. R2N2 : Residual Recurrent Neural Networks for Multivariate Time Series Forecasting , Sept. 2017

    2017

  14. [22]

    Ham, J.-H

    Y.-G. Ham, J.-H. Kim, and J.-J. Luo. Deep learning for multi-year ENSO forecasts. Nature, 573 0 (7775): 0 568--572, Sept. 2019. ISSN 0028-0836, 1476-4687. doi:10.1038/s41586-019-1559-7

    2019 doi

  15. [23]

    Hasselmann

    K. Hasselmann. Stochastic climate models Part I . Theory . Tellus, 28 0 (6): 0 473--485, 1976. ISSN 2153-3490. doi:10.1111/j.2153-3490.1976.tb00696.x

    1976

  16. [24]

    Hayashi, F.-F

    M. Hayashi, F.-F. Jin, and M. F. Stuecker. Dynamics for El Ni \ n o-La Ni \ n a asymmetry constrain equatorial- Pacific warming pattern. Nat Commun, 11 0 (1): 0 4230, Aug. 2020. ISSN 2041-1723. doi:10.1038/s41467-020-17983-y

    2020 doi

  17. [25]

    Hersbach

    H. Hersbach. Decomposition of the Continuous Ranked Probability Score for Ensemble Prediction Systems . Weather and Forecasting, 15 0 (5): 0 559--570, Oct. 2000. ISSN 1520-0434, 0882-8156. doi:10.1175/1520-0434(2000)015<0559:DOTCRP>2.0.CO;2

    2000 doi

  18. [26]

    Hochreiter and J

    S. Hochreiter and J. Schmidhuber. Long Short-Term Memory . Neural Computation, 9 0 (8): 0 1735--1780, Nov. 1997. ISSN 0899-7667. doi:10.1162/neco.1997.9.8.1735

    1997 doi

  19. [27]

    Irrgang, N

    C. Irrgang, N. Boers, M. Sonnewald, E. A. Barnes, C. Kadow, J. Staneva, and J. Saynisch-Wagner . Towards neural Earth system modelling by integrating artificial intelligence in Earth system science. Nat Mach Intell, 3 0 (8): 0 667--674, Aug. 2021. ISSN 2522-5839. doi:10.1038/s...

    2021 doi

  20. [28]

    G. C. Johnson. The Pacific Ocean Subtropical cell surface limb. Geophysical Research Letters, 28 0 (9): 0 1771--1774, 2001. ISSN 1944-8007. doi:10.1029/2000GL012723

    2001 doi

  21. [29]

    u ben, M. Kl \

    D. Kochkov, J. Yuval, I. Langmore, P. Norgaard, J. Smith, G. Mooers, J. Lottes, S. Rasp, P. D \"u ben, M. Kl \"o wer, S. Hatfield, P. Battaglia, A. Sanchez-Gonzalez , M. Willson, M. P. Brenner, and S. Hoyer. Neural General Circulation Models , Nov. 2023

    2023

  22. [30]

    R. Lam, A. Sanchez-Gonzalez , M. Willson, P. Wirnsberger, M. Fortunato, F. Alet, S. Ravuri, T. Ewalds, Z. Eaton-Rosen , W. Hu, A. Merose, S. Hoyer, G. Holland, O. Vinyals, J. Stott, A. Pritzel, S. Mohamed, and P. Battaglia. GraphCast : Learning skillful medium-range global wea...

    2023

  23. [31]

    S. Lang, M. Alexe, M. Chantry, J. Dramsch, F. Pinault, B. Raoult, M. C. A. Clare, C. Lessig, M. Maier-Gerber , L. Magnusson, Z. B. Bouall \`e gue, A. P. Nemesio, P. D. Dueben, A. Brown, F. Pappenberger, and F. Rabier. AIFS - ECMWF 's data-driven forecasting system, June 2024

    2024

  24. [32]

    Lessig, I

    C. Lessig, I. Luise, B. Gong, M. Langguth, S. Stadler, and M. Schultz. AtmoRep : A stochastic model of atmosphere dynamics using large scale representation learning, Aug. 2023. URL http://arxiv.org/abs/2308.13280

    2023 arXiv

  25. [33]

    M. L. L'Heureux, A. F. Z. Levine, M. Newman, C. Ganter, J.-J. Luo, M. K. Tippett, and T. N. Stockdale. ENSO Prediction . In El Ni \ n o Southern Oscillation in a Changing Climate , chapter 10, pages 227--246. American Geophysical Union (AGU), 2020. ISBN 978-1-119-54816-4. doi:...

    2020 doi

  26. [34]

    Z. Liu, Y. Lin, Y. Cao, H. Hu, Y. Wei, Z. Zhang, S. Lin, and B. Guo. Swin Transformer : Hierarchical Vision Transformer using Shifted Windows , Aug. 2021

    2021

  27. [35]

    Z. Liu, H. Mao, C.-Y. Wu, C. Feichtenhofer, T. Darrell, and S. Xie. A ConvNet for the 2020s, Mar. 2022

    2022

  28. [36]

    Loshchilov and F

    I. Loshchilov and F. Hutter. SGDR : Stochastic Gradient Descent with Warm Restarts , May 2017

    2017

  29. [37]

    Loshchilov and F

    I. Loshchilov and F. Hutter. Decoupled Weight Decay Regularization , Jan. 2019

    2019

  30. [38]

    Martinez-Villalobos , D

    C. Martinez-Villalobos , D. J. Vimont, C. Penland, M. Newman, and J. D. Neelin. Calculating State-Dependent Noise in a Linear Inverse Model Framework . Journal of the Atmospheric Sciences, 75 0 (2): 0 479--496, Feb. 2018. ISSN 0022-4928, 1520-0469. doi:10.1175/JAS-D-17-0235.1

    2018 doi

  31. [39]

    Martinez-Villalobos , M

    C. Martinez-Villalobos , M. Newman, D. J. Vimont, C. Penland, and J. David Neelin. Observed El Ni \ n o-La Ni \ n a Asymmetry in a Linear Model . Geophysical Research Letters, 46 0 (16): 0 9909--9919, 2019. ISSN 1944-8007. doi:10.1029/2019GL082922

    2019 doi

  32. [40]

    Newman and P

    M. Newman and P. D. Sardeshmukh. Are we near the predictability limit of tropical Indo - Pacific sea surface temperatures? Geophysical Research Letters, 44 0 (16): 0 8520--8529, Aug. 2017. ISSN 0094-8276, 1944-8007. doi:10.1002/2017GL074088

    2017 doi

  33. [41]

    Newman, P

    M. Newman, P. D. Sardeshmukh, C. R. Winkler, and J. S. Whitaker. A Study of Subseasonal Predictability . Monthly Weather Review, 131 0 (8): 0 1715--1732, Aug. 2003. ISSN 1520-0493, 0027-0644. doi:10.1175//2558.1

    2003 doi

  34. [42]

    Newman, M

    M. Newman, M. A. Alexander, and J. D. Scott. An empirical model of tropical ocean dynamics. Clim Dyn, 37 0 (9-10): 0 1823--1841, Nov. 2011. ISSN 0930-7575, 1432-0894. doi:10.1007/s00382-011-1034-0

    2011 doi

  35. [43]

    Y. M. Okumura. ENSO Diversity from an Atmospheric Perspective . Curr Clim Change Rep, 5 0 (3): 0 245--257, Sept. 2019. ISSN 2198-6061. doi:10.1007/s40641-019-00138-7

    2019 doi

  36. [44]

    Pathak, S

    J. Pathak, S. Subramanian, P. Harrington, S. Raja, A. Chattopadhyay, M. Mardani, T. Kurth, D. Hall, Z. Li, K. Azizzadenesheli, P. Hassanzadeh, K. Kashinath, and A. Anandkumar. FourCastNet : A Global Data-driven High-resolution Weather Model using Adaptive Fourier Neural Operat...

    2022

  37. [45]

    C. Penland. A stochastic model of IndoPacific sea surface temperature anomalies. Physica D: Nonlinear Phenomena, 98 0 (2): 0 534--558, Nov. 1996. ISSN 0167-2789. doi:10.1016/0167-2789(96)00124-8

    1996 doi

  38. [46]

    Penland and P

    C. Penland and P. D. Sardeshmukh. The Optimal Growth of Tropical Sea Surface Temperature Anomalies . Journal of Climate, 8 0 (8): 0 1999--2024, Aug. 1995. ISSN 0894-8755, 1520-0442. doi:10.1175/1520-0442(1995)008<1999:TOGOTS>2.0.CO;2

    1995 doi

  39. [47]

    Perez, F

    E. Perez, F. Strub, H. de Vries , V. Dumoulin, and A. Courville. FiLM : Visual Reasoning with a General Conditioning Layer , Dec. 2017

    2017

  40. [48]

    P. J. Petersik and H. A. Dijkstra. Probabilistic Forecasting of El Ni \ n o Using Neural Network Models . Geophys. Res. Lett., 47 0 (6), Mar. 2020. ISSN 0094-8276, 1944-8007. doi:10.1029/2019GL086423

    2020 doi

  41. [49]

    E. M. Rasmusson and T. H. Carpenter. Variations in Tropical Sea Surface Temperature and Surface Wind Fields Associated with the Southern Oscillation / El Ni \ n o . Monthly Weather Review, 110 0 (5): 0 354--384, May 1982. ISSN 1520-0493, 0027-0644. doi:10.1175/1520-0493(1982)1...

    1982 doi

  42. [50]

    Rasp and S

    S. Rasp and S. Lerch. Neural Networks for Postprocessing Ensemble Weather Forecasts . Monthly Weather Review, 146 0 (11): 0 3885--3900, Nov. 2018. ISSN 1520-0493, 0027-0644. doi:10.1175/MWR-D-18-0187.1

    2018 doi

  43. [51]

    Rodrigues, B

    E. Rodrigues, B. Zadrozny, C. Watson, and D. Gold. Decadal Forecasts with ResDMD : A Residual DMD Neural Network , June 2021. URL http://arxiv.org/abs/2106.11111

    2021 arXiv

  44. [52]

    P. D. Sardeshmukh and C. Penland. Understanding the distinctively skewed and heavy tailed character of atmospheric and oceanic probability distributions. Chaos: An Interdisciplinary Journal of Nonlinear Science, 25 0 (3): 0 036410, Mar. 2015. ISSN 1054-1500. doi:10.1063/1.4914169

    2015 doi

  45. [53]

    P. D. Sardeshmukh, G. P. Compo, and C. Penland. Need for Caution in Interpreting Extreme Weather Statistics . Journal of Climate, 28 0 (23): 0 9166--9187, Dec. 2015. ISSN 0894-8755, 1520-0442. doi:10.1175/JCLI-D-15-0020.1

    2015 doi

  46. [54]

    X. Shi, Z. Chen, H. Wang, D.-Y. Yeung, W.-k. Wong, and W.-c. Woo. Convolutional LSTM Network : A Machine Learning Approach for Precipitation Nowcasting , Sept. 2015

    2015

  47. [55]

    S.-I. Shin, P. D. Sardeshmukh, M. Newman, C. Penland, and M. A. Alexander. Impact of Annual Cycle on ENSO Variability and Predictability . Journal of Climate, 34 0 (1): 0 171--193, Jan. 2021. ISSN 0894-8755, 1520-0442. doi:10.1175/JCLI-D-20-0291.1

    2021 doi

  48. [56]

    o r, C. Fr \

    F. M. Strnad, J. Schl \"o r, C. Fr \"o hlich, and B. Goswami. Teleconnection Patterns of Different El Ni \ n o Types Revealed by Climate Network Curvature . Geophysical Research Letters, 49 0 (17): 0 e2022GL098571, 2022. ISSN 1944-8007. doi:10.1029/2022GL098571

    2022 doi

  49. [57]

    Sutskever, O

    I. Sutskever, O. Vinyals, and Q. V. Le. Sequence to Sequence Learning with Neural Networks , Dec. 2014

    2014

  50. [58]

    Takahashi and B

    K. Takahashi and B. Dewitte. Strong and moderate nonlinear El Ni \ n o regimes. Climate Dynamics, 46 0 (5-6): 0 1627--1645, Mar. 2016. ISSN 14320894. doi:10.1007/s00382-015-2665-3

    2016 doi

  51. [59]

    Takahashi, A

    K. Takahashi, A. Montecinos, K. Goubanova, and B. Dewitte. ENSO regimes: Reinterpreting the canonical and Modoki El Ni \ n o . Geophysical Research Letters, 38 0 (10): 0 L10704, May 2011. ISSN 00948276. doi:10.1029/2011GL047364

    2011 doi

  52. [60]

    Takahashi, C

    K. Takahashi, C. Karamperidou, and B. Dewitte. A theoretical model of strong and moderate El Ni \ n o regimes. Clim Dyn, 52 0 (12): 0 7477--7493, June 2019. ISSN 1432-0894. doi:10.1007/s00382-018-4100-z

    2019 doi

  53. [61]

    A. S. Taschetto, C. C. Ummenhofer, M. F. Stuecker, D. Dommenget, K. Ashok, R. R. Rodrigues, and S.-W. Yeh. ENSO Atmospheric Teleconnections . In El Ni \ n o Southern Oscillation in a Changing Climate , chapter 14, pages 309--335. American Geophysical Union (AGU), 2020. ISBN 97...

    2020 doi

  54. [62]

    Thual and B

    S. Thual and B. Dewitte. ENSO complexity controlled by zonal shifts in the Walker circulation. Nat. Geosci., 16 0 (4): 0 328--332, Apr. 2023. ISSN 1752-0908. doi:10.1038/s41561-023-01154-x

    2023 doi

  55. [63]

    Timmermann, S.-I

    A. Timmermann, S.-I. An, J.-S. Kug, F.-F. Jin, W. Cai, A. Capotondi, K. M. Cobb, M. Lengaigne, M. J. McPhaden, M. F. Stuecker, K. Stein, A. T. Wittenberg, K.-S. Yun, T. Bayr, H.-C. Chen, Y. Chikamoto, B. Dewitte, D. Dommenget, P. Grothe, E. Guilyardi, Y.-G. Ham, M. Hayashi, S....

    2018

  56. [64]

    D. J. Vimont, M. A. Alexander, and M. Newman. Optimal growth of Central and East Pacific ENSO events. Geophys. Res. Lett., 41 0 (11): 0 4027--4034, June 2014. ISSN 00948276. doi:10.1002/2014GL059997

    2014 doi

  57. [65]

    D. J. Vimont, M. Newman, D. S. Battisti, and S.-I. Shin. The Role of Seasonality and the ENSO Mode in Central and East Pacific ENSO Growth and Evolution . Journal of Climate, 35 0 (11): 0 3195--3209, June 2022. ISSN 0894-8755, 1520-0442. doi:10.1175/JCLI-D-21-0599.1

    2022 doi

  58. [66]

    S. Wang, L. Mu, and D. Liu. A hybrid approach for El Ni \ n o prediction based on Empirical Mode Decomposition and convolutional LSTM Encoder-Decoder . Computers & Geosciences, 149: 0 104695, Apr. 2021. ISSN 0098-3004. doi:10.1016/j.cageo.2021.104695

    2021

  59. [67]

    Watt-Meyer , N

    O. Watt-Meyer , N. D. Brenowitz, S. K. Clark, B. Henn, A. Kwa, J. McGibbon, W. A. Perkins, and C. S. Bretherton. Correcting Weather and Climate Models by Machine Learning Nudged Historical Simulations . Geophysical Research Letters, 48 0 (15): 0 e2021GL092555, 2021. ISSN 1944-...

    2021 doi

  60. [68]

    A. T. Wittenberg. Are historical records sufficient to constrain ENSO simulations? Geophysical Research Letters, 36 0 (12): 0 L12702, 2009. ISSN 1944-8007. doi:10.1029/2009GL038710

    2009 doi

  61. [69]

    Zhou and R.-H

    L. Zhou and R.-H. Zhang. A Hybrid Neural Network Model for ENSO Prediction in Combination with Principal Oscillation Pattern Analyses . Adv. Atmos. Sci., 39 0 (6): 0 889--902, June 2022. ISSN 1861-9533. doi:10.1007/s00376-021-1368-4

    2022 doi

  62. [70]

    Zhou and R.-H

    L. Zhou and R.-H. Zhang. A self-attention--based neural network for three-dimensional multivariate modeling and its skillful ENSO predictions. Sci. Adv., 9 0 (10): 0 eadf2827, Mar. 2023. ISSN 2375-2548. doi:10.1126/sciadv.adf2827

    2023 doi

  63. [71]

    , " * write output.state after.block = add.period write newline

    ENTRY address archive author booktitle chapter edition editor eprint howpublished institution journal key month note number organization pages publisher school series title type url doi volume year label INTEGERS output.state before.all mid.sentence after.sentence after.block ...

  64. [72]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.