Latos et al. (2026) Subseasonal forecasts of the MJO and convectively coupled equatorial waves in ACCESS‐S2 and GraphCast: A multi‐year analysis

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Quarterly Journal of the Royal Meteorological Society
• Year: 2026
• Date: 2026-08-17
• Authors: Beata Latos, Hanh Nguyen, Matthew C. Wheeler, Chen Li, Catherine de Burgh-Day, Muhammad E. E. Hassim, Sandeep Sahany, Aurel Moise
• DOI: 10.1002/qj.70293

Research Groups

• Australian Centre for Climate and Environmental Modelling (ACCEM)
• GraphCast Development Team

Short Summary

This study evaluates the predictive skill of two contrasting approaches, ACCESS-S2 and GraphCast, in forecasting Madden–Julian Oscillation (MJO) and convectively coupled equatorial waves at subseasonal-to-seasonal time-scales. The results reveal distinct complementary capabilities between the two models.

Objective

• Investigate the predictive skill of dynamical and machine-learning-based approaches for MJO and equatorial wave forecasting

Study Configuration

• Spatial Scale: Global tropics, with a focus on Maritime Continent and Indian Ocean regions
• Temporal Scale: Subseasonal-to-seasonal (S2S) time-scales, with 42-day lead times

Methodology and Data

• Models used:
  • Australian Community Climate and Earth-System Simulator–Seasonal version 2 (ACCESS-S2)
  • Machine-learning-based GraphCast
• Data sources: Multi-Source Weighted-Ensemble Precipitation (MSWEP) estimates, with 38 years of hindcast data (1981-2018)

Main Results

• ACCESS-S2 demonstrates better skill for longer-time-scale phenomena, maintaining meaningful correlations up to 21 days for MJO forecasts over the Maritime Continent and Indian Ocean.
• GraphCast shows better performance for shorter-time-scale wave prediction, achieving 6-9 days of skill for Kelvin waves and 4-8 days for MRG waves.
• Both models systematically underestimate precipitation amplitude across all wave types.

Contributions

This study highlights complementary capabilities between dynamical and machine-learning-based approaches for S2S equatorial wave and tropical precipitation predictions, informing hybrid ensemble forecasting strategies for improved accuracy.

Funding

• Australian Research Council (ARC) projects: DP180100966, DP190102111
• GraphCast Development Team funding: not specified

Citation

@article{Latos2026Subseasonal,
  author = {Latos, Beata and Nguyen, Hanh and Wheeler, Matthew C. and Li, Chen and Burgh-Day, Catherine de and Hassim, Muhammad E. E. and Sahany, Sandeep and Moise, Aurel},
  title = {Subseasonal forecasts of the MJO and convectively coupled equatorial waves in ACCESS‐S2 and GraphCast: A multi‐year analysis},
  journal = {Quarterly Journal of the Royal Meteorological Society},
  year = {2026},
  doi = {10.1002/qj.70293},
  url = {https://doi.org/10.1002/qj.70293}
}