Skilful global seasonal predictions from a machine learning weather model trained on reanalysis data

Chris Kent, Adam A. Scaife, Nick J. Dunstone, Doug Smith, Steven C. Hardiman, Tom Dunstan, Oliver Watt-Meyer

npj Climate and Atmospheric Science · 31 March 2025 · Allen Institute for AI (Ai2, ACE), Met Office

Why it matters. Shows ACE2 emulator has skilful seasonal NAO prediction, hinting ML weather models can serve seasonal forecasting; notable signal-to-noise finding.

In one sentence. ACE2 produces skilful winter seasonal forecasts from SST-persisted initialisation, reaching NAO correlation 0.47 over 1993–2016, comparable to a physics-based model.

Abstract

Machine learning weather models trained on observed atmospheric conditions can outperform conventional physics-based models at short- to medium-range (1-14 day) forecast timescales. Here we take the machine learning weather model ACE2, trained to predict 6-hourly steps in atmospheric evolution and which can remain stable over long forecast periods, and assess it from a seasonal forecasting perspective. Applying persisted sea surface temperature (SST) and sea-ice anomalies centred on 1st November each year, we initialise a lagged ensemble of winter predictions covering 1993/1994 to 2015/2016. Over this 23-year period there is remarkable similarity in the patterns of predictability with a leading physics-based model. The ACE2 model exhibits skilful predictions of the North Atlantic Oscillation (NAO) with a correlation score of 0.47 (p=0.02), as well as a realistic global distribution of skill and ensemble spread. Surprisingly, ACE2 is found to exhibit a signal-to-noise error as seen in physics-based models, in which it is better at predicting the real world than itself. Examining predictions of winter 2009/2010 indicates potential limitations of ACE2 in capturing extreme seasonal conditions that extend outside the training data. Nevertheless, this study reveals that machine learning weather models can produce skilful global seasonal predictions and heralds a new era of increased understanding, development and generation of near-term climate predictions.

Labels

Labels and summary are generated by Claude and reviewed weekly. Wrong? Edit this entry.