Pith. sign in

REVIEW 4 major objections 5 minor 63 references

Climate network and complexity approach predict neutral ENSO event for 2025

T0 review · 4 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read The paper forecasts a neutral ENSO year for 2025/26 with 69.6% probability, based on a climate-network signal and a complexity-entropy signal that both rule out an El Niño onset in 2025.

desk verdict Useful annual ENSO outlook from an established pipeline, but the 69.6% neutral probability depends on a logistic regression that contradicts the paper's own empirical claim about neutral-to-La Niña transitions. read the letter →

arxiv 2502.00643 v1 pith:RC3RFIAG submitted 2025-01-19 physics.ao-ph

J. Ludescher , J. Meng , J. Fan , A. Bunde , H. J. Schellnhuber This is my paper · ORCID
classification physics.ao-ph
keywords ENSOforecastingElNiñoLaNiñaclimatenetworkSystemSampleEntropyOceanicIndexspringpredictabilitybarrierglobalmeantemperature
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 predicts that the 2025/26 ENSO season will be neutral: neither an El Niño nor a La Niña. It applies two forecasting methods developed in earlier work — a climate-network measure of Pacific cooperativity and a System Sample Entropy measure of disorder in the Niño 3.4 region — and both point to the absence of an El Niño, at 91.2% and 91.7% probability. Combining these with a logistic regression on the Oceanic Niño Index gives a 69.6% probability of neutral conditions, 21.8% for La Niña, and 8.6% for El Niño. The authors argue that if this is right, 2025 mean global temperature is likely to fall somewhat below the 2024 level, because strong El Niños are a major driver of record global warmth.

What carries the argument

The climate network consists of 14 grid points in the central and eastern equatorial Pacific linked to 193 points outside; the mean link strength is computed from cross-correlations of surface air temperature anomalies and compared with a threshold learned in 1950-1980. The System Sample Entropy is the negative log of the conditional probability that similar temperature-anomaly subsequences in the Niño 3.4 region stay similar for longer windows, and a linear regression maps its previous-year value to a forecast El Niño magnitude, with values below 1.31 forecasting absence. A logistic regression on the ONI values of non-El Niño years followed by non-El Niño years, plus Laplace's rule of succession, converts the no-El-Niño forecast into neutral-versus-La-Niña probabilities.

What would settle it

If the observed Oceanic Niño Index for November 2025 through January 2026 or the surrounding season reaches at least +0.5°C for five consecutive months, the central forecast of no El Niño in 2025 is false. A season recorded as La Niña, with ONI at or below -0.5°C for five months, would also contradict the paper's 69.6% neutral estimate, though less directly.

Watch

Extended reading notes

Core claim

The central claim is that the cooperative mode in the Pacific climate network and the low complexity of Niño 3.4 temperature anomalies in 2024 both signal that no El Niño will start in 2025. In the network approach the mean link strength stayed below the decision threshold through 2024, and in the entropy approach the 2024 SysSampEn value of 0.79 lies below the 1.31 threshold. Placing the current OND ONI value of -0.4°C in the historical OND-to-NDJ relationship then favors a neutral year over a La Niña. The authors present these as probabilistic forecasts, not deterministic statements.

Load-bearing premise

The forecast probabilities assume that the coming year behaves like the small set of past non-El Niño years used to calibrate them — 34 network cases, 12 entropy cases, and 8 neutral-to-neutral transitions — so if 2025 is not exchangeable with those samples, the quoted percentages lose their support.

Editorial extensions

If this is right

  • If correct, the 2025/26 season is far more likely to be neutral (69.6%) than La Niña (21.8%) or El Niño (8.6%).
  • The forecast implies 2025 global mean temperature will likely decline from the 2024 record, though other forcings could offset part of the drop.
  • Both methods would again demonstrate skill at lead times beyond the spring predictability barrier, reinforcing their value as early-warning tools.
  • The network algorithm's version (ii) would log its second consecutive correct no-El-Niño forecast after the 2023/24 El Niño call.

Reading between the lines

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

  • An extension the paper does not draw: the 69.6% neutral estimate inherits considerable uncertainty from tiny samples; adding one more neutral-to-neutral transition to the eight in the record would move the Laplace estimate from 90% to 91%, and a single La Niña following a neutral year would materially change the rough split.
  • A natural stress test is to run the same two predictors on the 2026 target season in January 2026: stable probabilities would support the claim that these methods genuinely bypass the spring barrier, while large swings would suggest overfitting.
  • The same ONI-based logistic regression could be extended to predict La Niña onset specifically, where historical support is even thinner; such an extension would sharpen the actionable part of the forecast for agriculture and water management.
  • Because the forecast depends only on data available in January, it is falsifiable within the calendar year, unlike typical decadal projections.
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

4 major / 5 minor

Summary. The paper applies two previously developed forecasting approaches to the 2025 ENSO state: a climate-network method based on mean link strength S(t) and a complexity method based on System Sample Entropy (SysSampEn). Both methods forecast the absence of an El Niño in 2025, with claimed probabilities of 91.2% and 91.7%. The authors then use a logistic regression on the Oceanic Niño Index (ONI) to split the no-El Niño outcome into neutral (69.6%) and La Niña (21.8%) probabilities for NDJ 2025/26, and compare these with the NOAA CPC forecast.

Significance. If the forecast is correct, the paper provides a useful early-warning demonstration for a key climate mode, and the network method has a genuine real-time track record since 2011 (12/13 correct forecasts in its original version, including the successful 2023/24 El Niño forecast). The SysSampEn method also has a published methodology with out-of-sample application. However, the quantitative probability claims are weakened by small sample sizes, in-sample parameter selection, and an internal inconsistency in the logistic-regression step. The central qualitative conclusion (no El Niño in 2025, with neutral more likely than La Niña) is plausible, but the specific numbers 91.2%, 91.7%, and 69.6% are not supported as stated.

major comments (4)
  1. [Section 4, Figures 6 and 7] The text states that 'neutral events were not followed by a La Niña phase' and reports 8 neutral-to-neutral transitions with zero neutral-to-La Niña transitions since 1950. Yet the logistic regression is fitted to all non-El Niño years followed by non-El Niño years, which includes La Niña years with OND values below -0.5°C. At the current OND = -0.4°C, the historical conditional frequency of La Niña is 0/8 (or 1/10 under Laplace's rule), but the fitted curve gives 23.8%. The final 69.6% neutral probability is therefore an artifact of pooling non-exchangeable data and extrapolating across the -0.5°C boundary. This is load-bearing for the headline probabilities and must be replaced with an analysis that conditions on the observed neutral OND state.
  2. [Section 3.2, Eq. (1)] The 91.7% probability is the in-sample hit rate of a threshold and parameter combination (m=30, p=30, γ=8, l_eff=360) selected for best hindcast skill in reference [4]. The same 12 cases are used to select the parameters and to evaluate the rule, so the reported probability is a fitted quantity rather than an out-of-sample predictive probability. No confidence interval or cross-validation is provided. The central no-El Niño conclusion may survive, but the 91.7% number is not a valid probability estimate as presented.
  3. [Section 2.2] The 91.2% no-El Niño probability for version (ii) is based on 31/34 absence forecasts over 1981-2024, but version (ii) was introduced in 2022 and applied retrospectively. Only a subset of those 34 cases are genuine out-of-sample forecasts; the explicitly reported real-time record is 12/13 correct for 2012-2024 using version (i). The paper should separate prospective out-of-sample performance from hindcast performance and report binomial confidence intervals for the proportions.
  4. [Section 4, combination of forecasts] The 'combined probability' of 91.4% is the arithmetic mean of 91.2% and 91.7%, not a combination of independent forecast probabilities. If the two methods were treated as independent, the no-El Niño probability would be roughly 1 - (1-0.912)(1-0.917) ≈ 0.993. The subsequent multiplication of the logistic-regression probabilities by 0.914 also treats the no-El Niño probability as a single number and double-counts it. The combination rule and any assumption about dependence between the two methods must be stated explicitly.
minor comments (5)
  1. [Section 3.1] The sentence 'For a brief description of the approach we follow [5]' appears twice in the same subsection; the duplicate should be removed.
  2. [Figure 7 caption] The caption contains a typo: '0 for a neural event' should read '0 for a neutral event'.
  3. [Section 4, Figure 6] The phrase 'in Figure 6, 2 such cases are on top of each other' is unclear; please explain explicitly how overlapping points are counted in the total of 8 cases.
  4. [Section 2.1] The description of version (ii) of the network algorithm says alarms are considered only when 'the ONI remains below 0.5°C for the rest of the calendar year'; the timing of this condition relative to the alarm date needs clarification.
  5. [Sections 3 and 4] All claimed probabilities should be accompanied by confidence intervals or Bayesian credibility intervals, given the small numbers of events on which they are based.

Circularity Check

2 steps flagged · score 6.0 of 10

Headline absence probabilities are partly in-sample hit rates: SysSampEn parameters and network version (ii) are chosen for best past skill and then counted as forecast probabilities.

  1. fitted input called prediction [Section 3.2, SysSampEn forecast for 2025 (parameter choice and 12-case count)]
    "We use the parameter values for the SysSampEn that lead to the best El Niño forecasting skill when applied to past events, as described in [4], m = 30, p = 30, γ = 8 and lef f = 360. ... There were 12 occurrences of a low SysSampEn accompanied by a lower than 0.5 °C ONI in December. In 11 out of these 12 cases, the hindcast was correct. ... Thus the method predicts with 91.7% probability the absence of an El Niño in 2025."

    The parameter set (m=30, p=30, γ=8, l_eff=360) is explicitly selected to maximize hindcast skill on past events, and the '12 occurrences' are those past events. Counting the rule's hits on the data used to choose the threshold and parameters gives the in-sample training accuracy, not an independent predictive probability. The 91.7% is therefore a fitted in-sample rate relabeled as a forecast probability, so the headline absence probability for 2025 is partly a direct restatement of the fit criterion.

  2. fitted input called prediction [Section 2.2, 'El Niño forecasts since 2011' (version (ii) rule and 31/34 absence accuracy)]
    "In a more restrictive version (ii) [56, 57], the algorithm considers only those alarms where the ONI remains below 0.5°C for the rest of the calendar year. ... The more restrictive version (ii) of the algorithm [56, 57] gave 9 El Niño alarms, all of which were correct. In this version, the forecasts for the absence of an El Niño onset are correct with 31/34≈ 91.2% probability."

    Version (ii) was introduced in the authors' own prior work [56, 57] to suppress known false alarms of the original version, including the incorrect alarms in 1994, 2004 and 2019. The 31/34 absence accuracy is then scored over the same 1981-2024 historical record that motivated this post hoc rule change. The reported 91.2% probability is therefore partly the in-sample performance of a rule adjusted after seeing past failures, not a purely out-of-sample forecast probability.

full rationale

The paper's central numerical forecasts inherit a partial circularity. The SysSampEn forecast states that parameter values were chosen for the best El Niño forecasting skill on past events and then reports 11 correct out of 12 past cases as a 91.7% probability; this is training accuracy, not an independent prediction probability. The climate-network 91.2% figure is computed under version (ii), a restriction introduced in the authors' own [56,57] that removes previously known false alarms, so scoring it on the same record is partly in-sample. The final 69.6% neutral and 21.8% La Niña probabilities are obtained by multiplying the logistic-regression split by the combined 91.4% absence probability, so they inherit these fitted rates. The Section 4 logistic regression itself is a standard statistical fit to historical transitions and is not circular by construction, but it is internally inconsistent with the paper's own statement that neutral events were never followed by La Niña: the text reports 0/8 such transitions yet the fitted curve gives 23.8% La Niña at OND = -0.4, which undermines the specific 69.6% number. The paper does contain independent content: the network method's real-time record since 2011 (12/13 correct), the random-guess p-value benchmark, and the qualitative conclusion that a neutral event is more likely than La Niña or El Niño. Thus the circularity is partial rather than total, supporting a score of 6.

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

The forecast depends on fitted thresholds and regression coefficients from prior work, plus small-sample historical transition data. No new physical entities are introduced.

free parameters (5)
  • Network decision threshold Θ = 2.82 (range 2.815-2.834)
    Fixed in the 1950-1980 learning phase; determines whether an alarm is issued each year and therefore the 91.2% absence probability.
  • SysSampEn parameters m, p, γ, leff = m=30, p=30, γ=8, leff=360
    Chosen in ref [4] for best El Niño forecasting skill on past events, not derived from first principles.
  • SysSampEn magnitude threshold = 1.31
    Threshold below which the forecasted magnitude is below 0.5 °C and absence of El Niño is predicted; determined from hindcasts.
  • Linear regression coefficients for SysSampEn vs El Niño magnitude = not reported
    Regression model trained on 'all correctly hindcasted El Niño events before 2024' (Section 3.2).
  • Logistic regression coefficients for neutral vs La Niña = not reported
    Fitted to non-El Niño years followed by non-El Niño years since 1950 (Section 4).
assumptions (5)
  • domain assumption ENSO phase is operationally defined by ONI ≥ 0.5 °C or ≤ -0.5 °C for at least five consecutive months.
    Used throughout to classify events and to define 'absence' forecasts; this is the standard NOAA definition adopted without re-derivation.
  • domain assumption The climate-network cooperative mode before El Niño, as measured by SATA cross-correlations, is a stable precursor.
    Adopted from refs [1,2]; the paper does not re-derive or independently test this mechanism.
  • domain assumption Previous-year SysSampEn is linearly related to next-year El Niño magnitude with r ≈ 0.90.
    Adopted from ref [4]; the paper uses this regression to convert 2024 SysSampEn into a 2025 magnitude forecast.
  • ad hoc to paper The 8 observed neutral-to-neutral transitions since 1950 are exchangeable with the 2024/25 state, supporting Laplace's rule of succession.
    Section 4 uses (8+1)/(8+2)=0.9; this small-sample exchangeability is not justified.
  • domain assumption Neutral events since 1950 have not been followed by a La Niña phase.
    Section 4 relies on this empirical regularity to assign a 90% prior probability to neutral, while the logistic regression then admits a 23.8% La Niña probability.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Climate network and complexity approach predict neutral ENSO event for 2025." pith.science (2026). https://pith.science/paper/RC3RFIAG

@misc{pith2026250200643,
  author       = {Pith},
  title        = {Pith review of: Climate network and complexity approach predict neutral ENSO event for 2025},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RC3RFIAG}},
  note         = {Machine review of arXiv:2502.00643}
}
read the original abstract

The El Ni\~no Southern Oscillation (ENSO) is the strongest driver of interannual global climate variability and can lead to extreme weather events like droughts and flooding. Additionally, ENSO influences the mean global temperature with strong El Ni\~no events often leading, in a warming climate, to new record highs. Recently, we have developed two approaches for the early forecasting of El Ni\~no. The climate network-based approach allows forecasting the onset of an El Ni\~no event about 1 year ahead. The complexity-based approach allows additionally to forecast the magnitude of an upcoming El Ni\~no event in the calendar year before. These methods successfully forecasted the onset of an Eastern Pacific El Ni\~no for 2023/24 and the subsequent record-breaking warming of 2024. Here, we apply these methods to forecast the ENSO state in 2025. Both methods forecast the absence of an El Ni\~no in 2025, with 91.2% and 91.7% probability, respectively. Combining these forecasts with a logistic regression based on the Oceanic Ni\~no Index (ONI) leads to a 69.6% probability that 2025/26 will be a neutral ENSO event. We estimate the probability of a La Ni\~na at 21.8%. This makes it likely that the mean global temperature in 2025 will decrease somewhat compared to the 2024 level.

Figures

Figures reproduced from arXiv: 2502.00643 by the authors.

Figure 1
Figure 1. The nodes of the climate network. The network consists of 14 grid points in the central and eastern equatorial Pacific (red dots) and 193 grid points outside this area (blue dots). The green rectangle shows the Ni˜no3.4 area. The grid points represent the nodes of the climate network that we use here to forecast the onset or absence of an El Ni˜no event. Each red node is linked to each blue node. The nodes are chara… view at source ↗
Figure 2
Figure 2. The network-based forecasting scheme. We compare the average link strength S(t) in the climate network (red curve) with a decision threshold Θ (horizontal line, here Θ = 2.82), (left scale), and the standard Ni˜no3.4 index (ONI), (right scale), between January 1950 and December 2024. When the link strength crosses the threshold from below, and the last available ONI is below 0.5 ◦C, we give an alarm and predict that… view at source ↗
Figure 3
Figure 3. The climate network-based real-time forecasting phase. Same as [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: The Ni˜no3.4 area and the SysSampEn input data. The red circles indicate the 22 nodes covering the Ni˜no 3.4 region at a spatial resolution of 5◦×5 ◦ . The curves are examples of the temperature anomaly time series for 3 nodes in the Ni˜no 3.4 region for one specific y…
Figure 5
Figure 5. Figure 5: Forecasted and observed El Ni˜no magnitudes. The magnitude forecast is shown as the height of rectangles in the year when the forecast is made, i.e., one year ahead of a potential El Ni˜no. The forecast is obtained by inserting the regarded calendar year’s SysSampEn va…
Figure 6
Figure 6. Figure 6: Interannual ONI relationship. The November-December-January (NDJ) ONI value of year y+1 vs. the OND value of the previous calendar year y (blue circles). Our forecast predicts the absence of an El Ni˜no with a high probability (91.2% and 91.7%). Thus, this outcome is e…
Figure 7
Figure 7. Figure 7: The probability of a La Ni˜na. We regard all years that are non-El Ni˜no years and are also followed by a non-El Ni˜no year, i.e., the events shown in the white area in [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: Summary of our forecast. The height of the bars shows the probability of El Ni˜no, neutral event and La Ni˜na for NDJ 2025/26 based on the climate network, the SysSampEn and on a logistic regression based on the events that correspond to the current state. We obtain 8.…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

63 extracted references · 61 canonical work pages

  1. [4]

    Proc Natl Acad Sci USA , 117:177-183; idid

    Meng J, Fan J, Ludescher J, Ankit A, Chen X, Bunde A, Kurths J, Schellnhuber HJ (2019) Complexity based approach for El Ni˜ no magnitude forecasting before the spring predictability barrier. Proc Natl Acad Sci USA , 117:177-183; idid. doi:10.1073/pnas.1917007117

    work page doi:10.1073/pnas.1917007117 2019

  2. [1]

    Proc Natl Acad Sci USA 110:11742- 11745; ibid

    Ludescher J, Gozolchiani A, Bogachev MI, Bunde A, Havlin S, Schellnhuber HJ (2013) Im- proved El Ni˜ no forecasting by cooperativity detection. Proc Natl Acad Sci USA 110:11742- 11745; ibid. doi:10.1073/pnas.1317354110

    work page doi:10.1073/pnas.1317354110 2013

  3. [2]

    Proc Natl Acad Sci USA 111:2064-2066; ibid

    Ludescher J, Gozolchiani A, Bogachev MI, Bunde A, Havlin S, Schellnhuber HJ (2014) Very early warning of next El Ni˜ no. Proc Natl Acad Sci USA 111:2064-2066; ibid. doi: 10.1073/pnas.1323058111

    work page doi:10.1073/pnas.1323058111 2014

  4. [3]

    https://doi.org/10.1007/s00704-024-05035-0

    Bunde A, Ludescher J, Schellnhuber HJ (2024) Evaluation of the real-time El Ni˜ no forecasts by the climate network approach between 2011 and present.Theor Appl Climatol 155, 6727–6736. https://doi.org/10.1007/s00704-024-05035-0

    work page doi:10.1007/s00704-024-05035-0 2024

  5. [5]

    Very early warning of a moderate-to-strong El Ni\~no in 2023

    Ludescher J, Meng J, Fan J, Bunde A, Schellnhuber HJ (2023) Very early warning of a moderate-to-strong El Ni˜ no in 2023. arXiv:2301.10763

    work page Pith review arXiv 2023

  6. [6]

    Dijkstra HA (2005) Nonlinear Physical Oceanography: A Dynamical Systems Approach to the Large-Scale Ocean Circulation and El Ni˜ no(Springer, New York)

    work page 2005

  7. [7]

    Clarke AJ (2008) An Introduction to the Dynamics of El Ni˜ no and the Southern Oscillation (Elsevier Academic Press, London)

    work page 2008

  8. [8]

    Sarachik ES, Cane MA (2010) The El Ni˜ no-Southern Oscillation Phenomenon (Cambridge University Press, Cambridge)

    work page 2010

Show all 63 references
  1. [9]

    (2017) El Ni˜ no and Southern Oscillation (ENSO): A review, in: Coral Reefs of the Eastern Tropical Pacific , eds Glymn PW, Manzello D, Enochs IC (Springer, Berlin)

    Wang C, et al. (2017) El Ni˜ no and Southern Oscillation (ENSO): A review, in: Coral Reefs of the Eastern Tropical Pacific , eds Glymn PW, Manzello D, Enochs IC (Springer, Berlin)

    2017

  2. [10]

    (2018) El Ni˜ no-Southern Oscillation complexity

    Timmermann A, et al. (2018) El Ni˜ no-Southern Oscillation complexity. Nature 559:535-545

    2018

  3. [11]

    McPhaden MJ, Santoso A, Cai W (Eds.) (2020) El Ni˜ no Southern Oscillation in a Changing Climate (John Wiley & Sons, Hoboken)

    2020

  4. [12]

    https://origin.cpc.ncep.noaa.gov/products/analysis monitoring/ensostuff/ONI v5.php

    National Oceanic and Atmospheric Administration, Climate Prediction Center. https://origin.cpc.ncep.noaa.gov/products/analysis monitoring/ensostuff/ONI v5.php

  5. [13]

    Davis M (2001) Late Victorian Holocaust: El Ni˜ no Famines and the Making of the Third World (Verso, London & New York)

    2001

  6. [14]

    Chin J Athmos Sci 26:359-376

    Wen C (2002) Impacts of El Ni˜ no and La Ni˜ na on the cycle of the East Asian winter and summer monsoon. Chin J Athmos Sci 26:359-376

    2002

  7. [15]

    Lancet 362:1481–1489

    Kovats RS, Bouma MJ, Hajat S, Worrall E, Haines A (2003) El Ni˜ no and health. Lancet 362:1481–1489

    2003

  8. [16]

    Donnelly JP, Woodruff JD (2007) Intense hurricane activity over the past 5,000 years con- trolled by El Ni˜ no and the West African monsoon.Nature 447:465-468

    2007

  9. [17]

    Corral A, Oss´ o A, Llebot JE (2010) Scaling of tropical-cyclone dissipation.Nature Phys 6:693- 696

    2010

  10. [18]

    Cane MA, Zebiak SE, Dolan SC (1986) Experimental forecasts of El Ni˜ no.Nature 321:827-832. 9

    1986

  11. [19]

    J Clim 8:1999-2024

    Penland C und Sardeshmukh PD (1995) The optimal growth of tropical sea surface tempera- ture anomalies. J Clim 8:1999-2024

    1995

  12. [20]

    Tziperman E, Scher H, Zebiak SE, Cane MA (1997) Controlling Spatiotemporal Chaos in a Realistic El Ni˜ no Prediction Model.Phys Rev Lett 79:1034-1037

    1997

  13. [21]

    Fedorov A V, Harper SL, Philander SG, Winter B, Wittenberg A (2003) How Predictable is El Ni˜ no?Bull Amer Meteor Soc 84:911-919

    2003

  14. [22]

    Mon Weather Rev 131:2748-2764

    Galanti E, Tziperman E, Rosati A, Sirkes Z (2003) A Study of ENSO Prediction Using a Hybrid Coupled Model and the Adjoint Method for Data Assimilation. Mon Weather Rev 131:2748-2764

    2003

  15. [23]

    Monthly Weather Review 131:2324-2341

    Kirtman BP (2003) The COLA anomaly coupled model: Ensemble ENSO prediction. Monthly Weather Review 131:2324-2341

    2003

  16. [24]

    Nature 428:733-736

    Chen D, Cane MA, Kaplan A, Zebiak SE, Huang D (2004) Predictability of El Ni˜ no over the past 148 years. Nature 428:733-736

    2004

  17. [25]

    (2004) Development of a european multimodel ensemble system for seasonal- to-interannual prediction (demeter)

    Palmer TN et al. (2004) Development of a european multimodel ensemble system for seasonal- to-interannual prediction (demeter). Bull. Am. Meteorol. Soc. 85, 853-872

    2004

  18. [26]

    J Comput Phys 227:3625–3640

    Chen D, Cane MA (2008) El Ni˜ no prediction and predictability. J Comput Phys 227:3625–3640

    2008

  19. [27]

    J Clim 21(1):84-93

    Luo JJ, Masson S, Behera SK, Yamagata T (2008) Extended ENSO predictions using a fully coupled ocean-atmosphere model. J Clim 21(1):84-93

    2008

  20. [28]

    Predicting stochastic systems by noise sampling, and application to the El Ni˜ no-Southern Oscillation

    Chekroun MD, Kondrashov D, Ghil M (2011). Predicting stochastic systems by noise sampling, and application to the El Ni˜ no-Southern Oscillation. Proc Nat Acad Sci USA 108(29):11766–11771

    2011

  21. [29]

    J Clim 28:8511-8520

    Chapman D, Cane MA, Henderson N, Lee DE, Chen C (2015) A Vector Autoregressive ENSO Prediction Model. J Clim 28:8511-8520

    2015

  22. [30]

    Meng J, Fan J, Ashkenazy Y, Bunde A, Havlin S (2018) Forecasting the magnitude and onset of El Ni˜ no based on climate networkNew J Phys 20:043036

    2018

  23. [31]

    Noteboom PD, Feng QY, Lopez C, Hern´ andez-Garc ´ ıa, Dijkstra HA (2018) Using network theory and machine learning to predict El Ni˜ no.Earth Syst Dynam 9:969-983

    2018

  24. [32]

    Nature, 573, 568-572

    Ham YG, Kim JH,Luo, JJ (2019) Deep learning for multi-year ENSO forecasts. Nature, 573, 568-572

    2019

  25. [33]

    (2014) The NCEP climate forecast system version 2

    Saha S et al. (2014) The NCEP climate forecast system version 2. Journal of climate 27(6), 2185-2208

    2014

  26. [34]

    (2016) ClimateLearn : A machine-learning approach for climate prediction using network measures

    Feng QY et al. (2016) ClimateLearn : A machine-learning approach for climate prediction using network measures. Geosci. Model Dev. 10.5194/gmd-2015-273

    2016 doi

  27. [35]

    Lu Z, Yuan N, Fu Z (2016) Percolation Phase Transition of Surface Air Temperature Networks under Attacks of El Ni˜ no/La Ni˜ na.Sci. Rep. 6, 26779

    2016

  28. [36]

    Rodriguez-Mendez V, Eguiluz VM, Hernandez-Garcia E, Ramasco JJ (2016) Percolation- based precursors of transitions in extended systems. Sci. Rep. 6, 29552

    2016

  29. [37]

    IEEE Access 8, 55711- 55723

    De Castro Santos MA, Vega-Oliveros DA, Zhao L, Berton L (2020) Classifying El Ni˜ no- Southern Oscillation combining network science and machine learning. IEEE Access 8, 55711- 55723

    2020

  30. [38]

    Petersik PJ, Dijkstra HA (2020) Probabilistic forecasting of El Ni˜ no using neural network models. Geophys. Res. Lett. 47, e2019GL086423. 10

    2020

  31. [39]

    Environmental Research: Climate 1(1), 011002

    Hassanibesheli F, Kurths J, Boers N (2022) Long-term ENSO prediction with echo-state networks. Environmental Research: Climate 1(1), 011002

    2022

  32. [40]

    (2024) Explainable El Ni˜ no predictability from climate mode interactions

    Zhao S, et al. (2024) Explainable El Ni˜ no predictability from climate mode interactions. Nature, 630(8018), 891-898

    2024

  33. [41]

    IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 17, 17066-17074, doi: 10.1109/JSTARS.2024.3447356

    Zhao A, Qin M, Wu S, Liu R, Du Z (2024) ENSO Forecasts With Spatiotemporal Fusion Transformer Network. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 17, 17066-17074, doi: 10.1109/JSTARS.2024.3447356

    2024

  34. [42]

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

    Schl¨ or J, Newman M, Thuemmel J, Capotondi A, Goswami B (2024). A Hybrid Deep- Learning Model for El Ni˜ no Southern Oscillation in the Low-Data Regime. arXiv preprint arXiv:2412.03743

    2024 arXiv

  35. [43]

    Meteorol Atmos Phys 56:33-55

    Webster PJ (1995) The annual cycle and the predictability of the tropical coupled ocean- athomosphere system. Meteorol Atmos Phys 56:33-55

    1995

  36. [44]

    (2001) Current approaches to seasonal to interannual climate predictions

    Goddard L et al. (2001) Current approaches to seasonal to interannual climate predictions. Int J Clim 21:1111-1152

    2001

  37. [45]

    Barnston AG, Tippett MK, L’Heureux ML, Li S, DeWitt DG (2012) Skill of real-time sea- sonal ENSO model predictions during 2002-11: Is our capability increasing? Bulletin of the American Meteorological Society 93(5), 631-651

    2012

  38. [46]

    npj Climate and Atmospheric Science, 7(1), 301

    Ehsan MA, L’Heureux ML, Tippett MK, Robertson A W, Turmelle J (2024) Real-time ENSO forecast skill evaluated over the last two decades, with focus on the onset of ENSO events. npj Climate and Atmospheric Science, 7(1), 301

    2024

  39. [47]

    arXiv:2102.02192

    Ludescher J, Meng J, Fan J (2021) Climate network and complexity based El Ni˜ no forecast for 2021. arXiv:2102.02192

    2021 arXiv

  40. [48]

    Tsonis AA, Swanson KL, Roebber PJ (2006) What do networks have to do with climate? Bull Amer Meteor Soc 87:585-595

    2006

  41. [49]

    Yamasaki K, Gozolchiani A, Havlin S (2008) Climate networks around the globe are signifi- cantly affected by El Ni˜ no.Phys Rev Lett 100:228501

    2008

  42. [50]

    EPL (Europhysics Letters) 87:48007

    Donges JF, Zou Y, Marwan N, Kurths, J (2009) The backbone of the climate network. EPL (Europhysics Letters) 87:48007

    2009

  43. [51]

    Phys Rev Lett 107:148501

    Gozolchiani A, Havlin S, Yamasaki K (2011) Emergence of El Ni˜ no as an autonomous com- ponent in the climate network. Phys Rev Lett 107:148501

    2011

  44. [52]

    Dijkstra HA, Hern´ andez-Garc ´ ıa E, Masoller C, Barreiro M (2019)Networks in Climate (Cam- bridge Univ Press, Cambridge, UK)

    2019

  45. [53]

    Physics Reports 896:1-84

    Fan J, Meng J, Ludescher J, Chen X, Ashkenazy Y, Kurths J, Havlin S, Schellnhuber HJ (2020) Statistical physics approaches to the complex Earth system. Physics Reports 896:1-84

    2020

  46. [54]

    Proc Natl Acad Sci USA 118(47) e1922872118, doi: 10.1073/pnas.1922872118

    Ludescher J, Martin M, Boers N, Bunde A, Ciemer C, Fan J, Havlin S, Kretschmer M, Kurths J, Runge J, Stolbova V, Surovyatkina E, Schellnhuber HJ (2021) Network-based forecasting of climate phenomena. Proc Natl Acad Sci USA 118(47) e1922872118, doi: 10.1073/pnas.1922872118

    2021 doi

  47. [55]

    Journal of Climate , 35(3), 1009-1020

    Fan J, Meng J, Ludescher J, Li Z, Surovyatkina E, Chen X, Kurths J, Schellnhuber HJ (2022) Network-based approach and climate change benefits for forecasting the amount of indian monsoon rainfall. Journal of Climate , 35(3), 1009-1020

    2022

  48. [56]

    arXiv preprint arXiv:2212.14025

    Ludescher J, Bunde A, Schellnhuber HJ (2022) Forecasting the El Ni˜ no type well before the spring predictability barrier. arXiv preprint arXiv:2212.14025. 11

    2022 arXiv

  49. [57]

    npj Clim Atmos Sci 6, 196

    Ludescher J, Bunde A, Schellnhuber HJ (2023) Forecasting the El Ni˜ no type well before the spring predictability barrier. npj Clim Atmos Sci 6, 196. https://doi.org/10.1038/s41612-023- 00519-8

    2023 doi

  50. [58]

    (1996) The NCEP/NCAR 40-year reanalysis project

    Kalnay et al. (1996) The NCEP/NCAR 40-year reanalysis project. Bull Am Meteorol Soc 77:437-471

    1996

  51. [59]

    http.//www.esrl.noaa.gov/psd/data/gridded/data.ncep/reanalyis.html

    National Oceanic and Atmospheric Administration, Earth System Research Laboratory. http.//www.esrl.noaa.gov/psd/data/gridded/data.ncep/reanalyis.html

  52. [60]

    Am J Physiol Heart Circ Physiol 278, H2039–2049

    Richman JS, Moorman JR (2000) Physiological time-series analysis using approximate entropy and sample entropy. Am J Physiol Heart Circ Physiol 278, H2039–2049

    2000

  53. [61]

    https://climate.copernicus.eu/climate-reanalysis?q=products/ climate-reanalysis

    ERA5 Climate reanalyis. https://climate.copernicus.eu/climate-reanalysis?q=products/ climate-reanalysis

  54. [62]

    Cambridge university press

    Jaynes ET, Probability theory: The logic of science. Cambridge university press. (2003)

    2003

  55. [63]

    National Oceanic and Atmospheric Administration (NOAA), Climate Prediction Center (CPC), Official Probabilistic ENSO Forecast: https://iri.columbia.edu/our- expertise/climate/forecasts/enso/current/?enso tab=enso-cpc plume 12

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.