Adnan KhanAdnanProduct & AI
San Francisco Bay Area
Get in touch

Source note3 min read

AIFS-SUBS: Extending Data-Driven Forecasting to Sub-Seasonal Timescales

Source

arXiv

arxiv.org/abs/2607.05100v1

Read the source

Source

  • Source name: arXiv
  • Source kind: paper
  • Authors / organization: Jakob Schloer, Steffen Tietsche, Christopher D. Roberts, Lorenzo Zampieri, Simon Lang, Gert Mertes, Gareth Jones, Matthew Chantry, Frederic Vitart
  • Venue: arXiv: physics.ao-ph
  • Published: 2026-07-06
  • Collected: 2026-07-07
  • URL: http://arxiv.org/abs/2607.05100v1
  • PDF: https://arxiv.org/pdf/2607.05100v1
  • DOI: UNKNOWN

Plain-Language Brief

Researchers have developed a new data-driven forecasting model, called AIFS-SUBS, to extend the capabilities of existing models beyond medium range to sub-seasonal timescales. This new model shows improved skill in probabilistic forecasts for weeks 2-6, reducing systematic biases compared to the operational Integrated Forecasting System (IFS). The model also improves the prediction of various weather phenomena, such as the Madden--Julian Oscillation (MJO), tropical cyclone activity, and sudden stratospheric warming (SSW). AIFS-SUBS uses significantly less energy during inference than the IFS, allowing for larger real-time ensembles.

What the Document Says

  • Researchers developed a model called AIFS-SUBS based on ECMWF's AIFS-CRPS medium-range model.
  • AIFS-SUBS adopts a 24h autoregressive time step, adds stratospheric levels and top-of-atmosphere thermal radiation as predictors, and uses 2007--2011 as an independent verification window.
  • AIFS-SUBS was evaluated in two configurations: AIFS-SUBS fine-tuned on operational analyses and AIFS-SUBS-ERA5 trained on ERA5 alone.
  • Across weeks 2--6, AIFS-SUBS matches the operational Integrated Forecasting System (IFS) in probabilistic skill while reducing systematic biases.
  • For the convective (OLR) component of the Madden--Julian Oscillation (MJO), AIFS-SUBS extends skilful forecasts by eight days relative to the IFS.
  • AIFS-SUBS reproduces the observed MJO modulation of tropical cyclone activity comparably.
  • Stratospheric skill is particularly strong with AIFS-SUBS reproducing sudden stratospheric warming (SSW) frequency and surface impact.
  • In the AI Weather Quest, AIFS-SUBS-ERA5 attains a variable-averaged ranked probability skill score slightly ahead of the IFS at weeks 3 and 4.
  • At inference, AIFS-SUBS uses about 200 times less energy than the IFS.

Key Concepts Mentioned

  • AIFS-SUBS
  • sub-seasonal forecasting
  • probalistic skill
  • systematic biases
  • Madden--Julian Oscillation (MJO)
  • tropical cyclone activity
  • sudden stratospheric warming (SSW)
  • AI Weather Quest

Claims Made by the Document

  • AIFS-SUBS matches the operational Integrated Forecasting System (IFS) in probabilistic skill while reducing systematic biases across weeks 2--6.
  • AIFS-SUBS extends skilful forecasts by eight days relative to the IFS for the convective component of the Madden--Julian Oscillation (MJO).
  • AIFS-SUBS reproduces the observed MJO modulation of tropical cyclone activity comparably.

Evidence Used

  • Across weeks 2--6, AIFS-SUBS matches the operational Integrated Forecasting System (IFS) in probabilistic skill while reducing systematic biases.
  • For the convective (OLR) component of the Madden--Julian Oscillation (MJO), AIFS-SUBS extends skilful forecasts by eight days relative to the IFS.
  • AIFS-SUBS reproduces the observed MJO modulation of tropical cyclone activity comparably.

Methods / Approach

UNKNOWN

Limitations Stated by the Authors

UNKNOWN

Limitations Noted by the Agent

Agent interpretation — not from the source.

UNKNOWN

Open Questions

UNKNOWN

Notes for Future Q&A

May include interpretation. Labeled where it does.

UNKNOWN

Raw Abstract / Excerpt

Data-driven models now rival numerical weather prediction in the medium range, but extending them to sub-seasonal lead times raises challenges absent at shorter horizons. Errors accumulate over long autoregressive rollouts, systematic biases grow with lead time, and several years of data must be held out for independent verification, even though machine-learning models otherwise benefit from longer training records. To address these challenges, we adapt ECMWF's AIFS-CRPS medium-range model. AIFS-SUBS adopts a 24h autoregressive time step to reduce error accumulation, adds stratospheric levels and top-of-atmosphere thermal radiation as predictors, and reserves 2007--2011 as an independent verification window. We evaluate two config-durations: AIFS-SUBS, fine-tuned on operational analyses, and AIFS-SUBS-ERA5, trained on ERA5 alone. Across weeks 2--6, AIFS-SUBS matches the operational Integrated Forecasting System (IFS) in probabilistic skill while reducing systematic biases. For the convective (OLR) component of the Madden--Julian Oscillation (MJO), AIFS-SUBS extends skilful forecasts (correlation > 0.5) by eight days relative to the IFS, while matching or exceeding the IFS for the full multivariate RMM index. AIFS-SUBS also reproduces the observed MJO modulation of tropical cyclone activity comparably. Stratospheric skill is particularly strong with AIFS-SUBS reproducing sudden stratospheric warming (SSW) frequency and surface impact. In the AI Weather Quest, AIFS-SUBS-ERA5 attains a variable-averaged ranked probability skill score slightly ahead of the IFS at weeks 3 and 4. At inference, AIFS-SUBS uses about 200 times less energy than the IFS, opening the door to much larger real-time ensembles. AIFS-SUBS is ECMWF's first machine-learning model targeted at sub-seasonal time-scales.