Truace
Subscribe
52/100

Moderate evidence

Cite
More
Export CSVExport JSONEmailLinkedInX

Europe's AI weather model gets 4%–6% more accurate; rain forecasts gain about a day

Study title: AIFS Single 1.1.0: an update to ECMWF's machine-learned weather forecast model AIFS

Evidence type
Benchmark or lab evaluation
Sources
1 sourcePeer-reviewed
Published
2 min read
Truvace record
TRV-2026-2622cite or check it

Study title: AIFS Single 1.1.0: an update to ECMWF's machine-learned weather forecast model AIFS

Researchers at ECMWF describe version 1.1.0 of AIFS Single, its machine-learned weather forecast model that has been operational since 25 February 2025. The update raises overall skill by 4%–6% in the upper air and near-surface variables, with no loss of spatial variability.

Read the study abstract

We present version 1.1.0 of ECMWF's Artificial Intelligence Forecasting System (AIFS Single), operational since 25 February 2025. The revised system introduces a bounding-layer framework that enforces physical constraints, such as non-negativity and internal consistency within precipitation and cloud cover variables, alongside expanded training data, revised loss weighting, and an extended set of surface and atmospheric variables. Overall skill improves by 4 %–6 % in the upper air and near-surface variables with…

The gain

In a model evaluation, ECMWF's updated AI forecast system AIFS Single 1.1.0 improved upper-air and near-surface skill by 4%–6% and precipitation skill by up to 12%, gaining about 1 day on a categorical measure.

In brief

The biggest jump is in precipitation, where forecasts improved by up to 12% and gained roughly one day of lead time on a categorical skill measure. The authors trace this to a bounding-layer framework that enforces physical limits such as non-negative precipitation.

A controlled comparison found that adding more training data was the main source of upper-air gains, which the authors say shows why frequent model updates matter.

Main points

  1. AIFS Single has been operational at ECMWF since 25 February 2025; this paper describes version 1.1.0.
  2. The update adds a bounding-layer framework that enforces physical constraints, including non-negativity and internal consistency within precipitation and cloud cover variables.
  3. Other changes include expanded training data, revised loss weighting and an extended set of surface and atmospheric variables.
  4. Skill improves 4%–6% in upper-air and near-surface variables without degrading spatial variability.
  5. Precipitation improves by up to 12%, with an approximately 1 day advantage on a categorical skill measure.
  6. A controlled comparison identifies training data expansion as the dominant source of upper-air skill gains.
  7. Enforcing precipitation non-negativity resolves a gradient ambiguity at the zero-precipitation boundary under MSE training, which explains the reduced drizzle bias.

The full story

The paper documents an upgrade to a forecasting system already in operational use. Version 1.1.0 changes both the training setup and the model's output handling, adding a bounding-layer framework that keeps predicted fields physically plausible.

The authors separated the effects of the changes with a controlled comparison. Expanded training data drove most of the upper-air gains, while the bounding framework produced the largest precipitation improvements, up to 12%.

They also offer a technical explanation for the better rain forecasts. Under MSE training, the zero-precipitation boundary creates a gradient ambiguity; enforcing non-negativity resolves it, reducing drizzle bias.

Sources

1 on file
  1. Geoscientific Model Developmentdoi:10.5194/gmd-19-4703-2026
    Peer-reviewed2026-06-01

Talk it over

Four questions to start onWho gets this benefit, and who was left out of the study or story?

Start with a question

Europe's AI weather model gets 4%–6% more accurate; rain forecasts gain about a dayhttps://truvace.com/item/aifs-single-1-1-0-an-update-to-ecmwfs-machine-learned-weather-forecast-model-aif

  1. Who gets this benefit, and who was left out of the study or story?

    The gain In a model evaluation, ECMWF's updated AI forecast system AIFS Single 1.1.0 improved upper-air and near-surface skill by 4%–6% and precipitation skill by up to 12%, gaining about 1 day on a categorical measure.

    Answer in the thread

  2. How do we know? What can this kind of evidence show, and what can’t it?

    The evidence This rests on a benchmark or lab test, from Geoscientific Model Development.

    Answer in the thread

  3. What does it leave open? What else would you want to know before acting on it?

    Answer in the thread

  4. Have you seen AI used this way where you study, work or live? Did it go the same way?

    Answer in the thread

The questions are the same on every record. The details under them are taken from this record as written.

What readers say

More from Climate