CHRISTIANSEN et al. (2026) The winter mean NAO: white noise and predictability

Identification

• Journal: npj Climate and Atmospheric Science
• Year: 2026
• Date: 2026-01-16
• Authors: BO CHRISTIANSEN, Shuting Yang
• DOI: 10.1038/s41612-026-01326-7

Research Groups

• National Center for Climate Research, Danish Meteorological Institute, Copenhagen, Denmark

Short Summary

This study thoroughly investigates whether the winter mean North Atlantic Oscillation (NAO) can be distinguished from white noise in both observations and a large CMIP6 multi-model ensemble. The authors find little evidence to reject the null-hypothesis of white noise for either, suggesting that the NAO has limited predictability on inter-annual to decadal timescales.

Objective

• Is the winter mean North Atlantic Oscillation (NAO) distinguishable from white noise?
• Does the temporal structure of the NAO differ between observations and historical experiments with contemporary climate models (CMIP6)?

Study Configuration

• Spatial Scale: Northern Hemisphere, specifically the North Atlantic region, with NAO defined by pressure differences between the Azores (20–28°W, 36–40°N) and Iceland (16–25°W, 63–70°N).
• Temporal Scale: Inter-annual to decadal timescales, focusing on winter mean (December–February) NAO. Observational record from 1865 to 2023, and CMIP6 historical experiments from 1850 to 2014.

Methodology and Data

• Models used: CMIP6 historical experiments (215 ensemble members from 45 different models).
• Data sources:
  • Observations: Hurrell’s monthly station-based NAO index.
  • Statistical Tools: Power-spectra, wavelet-spectra (Morlet wavelet of order 6), autoregressive models (Yule-Walker equations, Whittle estimation, Akaike information criterion (AIC), Bayesian information criterion (BIC)), and various time series test statistics (Brock, Dechert, and Scheinkman (BDS) statistic, Turning Point statistics, Permutation Entropy, Box-Pierce statistic). Surrogate data tests were used for statistical significance, comparing results to white Gaussian noise series of the same length. A one-way ANOVA test was used for model inter-comparison.

Main Results

• Overall, there is little evidence to reject the null-hypothesis that the winter mean NAO is white Gaussian noise for both observations and CMIP6 models.
• For observations, a peak in the power spectrum near 8 years is found to be significant in the period after 1950 but not before, indicating non-stationarity. However, when considering the complete spectrum, this peak is not statistically significant compared to white noise.
• The large CMIP6 multi-model ensemble is statistically very similar to an ensemble of white noise of comparable size, with ensemble-averaged power spectra and wavelet power spectra appearing flat.
• Autoregressive coefficients and various time series statistics (BDS, Turning Point, Permutation Entropy, Box-Pierce) do not show significant deviations from white noise for either observations or the model ensemble.
• The in-model variance (internal variability) is considerably larger than the between-model variance (differences in model physics) for decadal power (ratio of approximately 3.37), suggesting internal variability dominates.
• The forced response in models is very weak, with an amplitude more than 10 times smaller than observations, and does not exhibit significant excess power at decadal timescales.
• These findings collectively suggest limited decadal predictability of the NAO.

Contributions

• Provides a comprehensive statistical investigation of the winter mean NAO's temporal and spectral structure, rigorously testing the white noise hypothesis using a wide array of statistical tools.
• Offers a thorough comparison of observed NAO characteristics with a large CMIP6 multi-model historical ensemble and white noise surrogates, enhancing the robustness of conclusions.
• Challenges previous reports of significant decadal peaks in observed NAO power spectra by demonstrating they are likely chance occurrences when considering the full spectrum and non-stationarity.
• Quantifies the relative importance of internal variability versus model physics differences for decadal power in CMIP6 models, showing internal variability is the dominant factor.
• Reinforces the conclusion of limited decadal predictability for the NAO, providing strong evidence against the presence of a consistent, predictable signal on these timescales in both observations and current climate models.

Funding

• Danish state-funded National Centre for Climate Research (NCFK)
• Impetus4Change project (grant agreement No 101000607)
• OptimESM project (grant agreement No 101081193)
• European Union’s Horizon 2020 research and innovation programme

Citation

@article{CHRISTIANSEN2026winter,
  author = {CHRISTIANSEN, BO and Yang, Shuting},
  title = {The winter mean NAO: white noise and predictability},
  journal = {npj Climate and Atmospheric Science},
  year = {2026},
  doi = {10.1038/s41612-026-01326-7},
  url = {https://doi.org/10.1038/s41612-026-01326-7}
}