Kim et al. (2026) ENSO phase transition enables prediction of winter North Atlantic Oscillation one year ahead

Identification

• Journal: Nature Communications
• Year: 2026
• Date: 2026-03-25
• Authors: YK Kim, Myong-In Lee, Adam A. Scaife, Doug Smith
• DOI: 10.1038/s41467-026-70646-2

Research Groups

• Department of Civil, Urban, Earth and Environmental Engineering, Ulsan National Institute of Science and Technology, Ulsan, South Korea
• Met Office Hadley Centre, Exeter, United Kingdom
• Department of Mathematics and Statistics, University of Exeter, Exeter, United Kingdom

Short Summary

This study demonstrates that the predictability of the winter North Atlantic Oscillation (NAO) one year ahead significantly improves during El Niño–Southern Oscillation (ENSO) phase transition years, a phenomenon linked to the northward propagation of atmospheric angular momentum anomalies.

Objective

• To investigate if El Niño–Southern Oscillation (ENSO) phase transitions enhance the predictability of the winter North Atlantic Oscillation (NAO) one year ahead.

Study Configuration

• Spatial Scale: Northern Hemisphere, North Atlantic (specifically Iceland and Azores for NAO), equatorial Pacific (15°S–15°N, 140°E–80°W for ENSO), 20°–80°N, 60°W–30°E for pattern correlation coefficients, and zonal-mean atmospheric angular momentum (AAM) in the Northern Hemisphere.
• Temporal Scale: Multi-year hindcasts covering 1962–2019 (DCPP models) and 1972–2019 (CESM2). Focus on one-year ahead predictions (Lead Year 1, LY1), corresponding to January–February (JF) mean winter (14–15 months lead time). ENSO index averaged over November–January (NDJ). AAM averaged over January–June (JFMAMJ).

Methodology and Data

• Models used:
  • Multi-Model Ensemble (MME) comprising 70 ensemble members from:
    • Decadal Climate Prediction Project (DCPP) models: HadGEM3-GC3.1-DePreSys4, CMCC-CM2-SR5, EC-Earth3, MPI-ESM1-2-HR.
    • Community Earth System Model 2-Seasonal to Multiyear Large Ensemble (CESM2-SMYLE).
  • Coupled Model Intercomparison Project Phase 6 (CMIP6) historical simulations (32 models).
• Data sources:
  • Reanalysis: European Centre for Medium-Range Weather Forecasts Reanalysis version 5 (ERA5) for atmospheric variables.
  • Observation: Hadley Centre Sea Level Pressure 2 (HadSLP2) for sea level pressure (SLP), Hadley Centre Sea Ice and SST dataset (HadISST1) for sea surface temperature (SST).

Main Results

• Winter NAO prediction skill one year ahead (LY1) significantly improves during ENSO phase transition years (Anomaly Correlation Coefficient (ACC) = 0.60, p = 0.012; Ratio of Predictable Components (RPC) = 2.86), compared to ENSO persistence years which show little to no skill.
• During ENSO transition years, observed NAO variability is significantly larger (standard deviation = 10.53 hPa) than during persistence years (standard deviation = 6.34 hPa).
• A signal-to-noise paradox (SNP) emerges during ENSO transition years, where real-world predictability (rmo) exceeds model-estimated skill (PCm) when the ensemble size is greater than approximately 10 members.
• The enhanced NAO predictability is dynamically linked to the northward propagation of anomalous atmospheric angular momentum (AAM), which is robustly observed in ERA5 reanalysis and reproduced by models during ENSO transition years.
• The MME accurately reproduces the observed NAO dipole structure in sea level pressure anomalies during transition years (Pattern Correlation Coefficient (PCC) = 0.89), while showing no resemblance during persistence years (PCC = 0.05).
• CMIP6 historical simulations generally underestimate both the ENSO–AAM and AAM–NAO correlations compared to observations, indicating limitations in current climate models to fully capture this delayed teleconnection.
• The MME exhibits statistically significant long-range ENSO prediction skill through the LY1 winter, confirming the practical relevance of conditioning on ENSO phase evolution.

Contributions

• Demonstrates for the first time, using a multi-model ensemble, that ENSO phase transitions substantially enhance the one-year-ahead predictability of the winter North Atlantic Oscillation.
• Identifies and confirms the poleward propagation of atmospheric angular momentum (AAM) anomalies as the key dynamical mechanism linking ENSO transitions to delayed NAO variability.
• Reveals the emergence of a signal-to-noise paradox for NAO prediction at a one-year lead during ENSO transition years, suggesting that real-world predictability is higher than models imply.
• Highlights the critical role of combining ENSO phase transition signals with large-ensemble strategies for improving subseasonal-to-interannual (S2I) and seasonal-to-decadal (S2D) NAO forecasts.
• Emphasizes the importance of accurately representing the ENSO-AAM-NAO pathway in future climate models for enhanced long-range predictability.

Funding

• Korea Meteorological Administration Research and Development Program “Operating and Developing Global Seasonal Forecast System” (KMA2018-00322)
• Korea Meteorological Administration Research and Development Program “Grant RS-2025-02313090”
• Met Office Hadley Centre Climate Programme (HCCP) funded by the UK Department for Science, Innovation and Technology (DSIT)
• Met Office Hadley Centre Climate Programme (HCCP) funded by the UK Public Weather Service

Citation

@article{Kim2026ENSO,
  author = {Kim, YK and Lee, Myong-In and Scaife, Adam A. and Smith, Doug},
  title = {ENSO phase transition enables prediction of winter North Atlantic Oscillation one year ahead},
  journal = {Nature Communications},
  year = {2026},
  doi = {10.1038/s41467-026-70646-2},
  url = {https://doi.org/10.1038/s41467-026-70646-2}
}