Chris Roberts

We have taken the ensemble version of the Artificial Intelligence Forecasting System (#AIFS) into operations!

It will run side by side with the traditional physics-based Integrated Forecasting System (IFS).

Find out more here
➡️ www.ecmwf.int/en/about/med...

Final published version of “Unbiased calculation, evaluation, and calibration of ensemble forecast anomalies” now available online in QJRMS:

rmets.onlinelibrary.wiley.com/doi/10.1002/...

Original thread on the preprint from the other place:

x.com/ClimatePrima...

Last Tuesday, I was set to give a talk on 'climate change in the Northeast' at a retirement home but had to cancel due to hourly job threats.

After nearly two weeks of overwhelming uncertainty, today it happened. I was fired from my dream of working at NOAA. I'm so sorry to everyone also affected.

Hi @raspstephan.bsky.social - interesting graph! However, I am not sure it makes sense to plot the ensemble and deterministic comparisons on the same axes. I guess the AIFS-CRPS results are from Fig9, which compares against the Tco1279 ENS rather than HRES…?

Sorry - meridional wind, not zonal.

Fair scores for ensemble forecasts The notion of fair scores for ensemble forecasts was introduced recently to reward ensembles with members that behave as though they and the verifying observation are sampled from the same distributi...

More details on fair ensemble scores:

rmets.onlinelibrary.wiley.com/doi/abs/10.1...

rmets.onlinelibrary.wiley.com/doi/full/10....

This is because optimising a fair ensemble score means there is no tension between minimising error and maintaining realistic levels of variability, which is unavoidable for deterministic MSE training. fCRPS loss is minimised when the forecast is drawn from the same distribution as observations.

Hmm - maybe Bluesky doesn’t support self-uploaded GIFs yet? Sorry Simon!

And here is an example 10 day forecast of 850 hPa Zonal Wind from AIFS-CRPS provided by Simon Lang. This animation illustrates that probabilistic CRPS-based training does not suffer from smoothing and/or reduced variance for longer rollouts that is inherent to deterministic MSE-based training.

… Peter Deuben , Sara Hahner, Pedro Maciel, Ana Prieto-Nemesio, Cathal O'Brien, Florian Pinault, Jan Polster, Baudouin Raoult, Steffen Tietsche, Martin Leutbecher!

This work represents a huge team effort from @ECMWF colleagues Simon Lang, Mihai Alexe, Mariana Clare, Rilwan Adewoyin, Zied Ben Bouallègue, Matthew Chantry, Jesper Dramsch…

For subseasonal forecasts, AIFS-CRPS outperforms the IFS ensemble before calibration and is competitive with the IFS ensemble when forecasts are evaluated as anomalies to remove the influence of model biases.

For medium-range forecasts, AIFS-CRPS outperforms the physics-based Integrated Forecasting System (IFS) ensemble for the majority of variables and lead times.

The trained model is stochastic and can generate as many exchangeable members as desired and is computationally feasible in inference.

Specifically, it introduces the “almost fair CRPS”, which approximately removes the bias in the score due to finite ensemble size yet avoids a degeneracy of the fair CRPS.

AIFS-CRPS is an ensemble variant of the Artificial Intelligence Forecasting System (AIFS) developed at ECMWF. The training protocol utilises a probabilistic loss function based on the Continuous Ranked Probability Score (CRPS).

This paper presents a novel approach to obtain a weather forecast model for ensemble forecasting with machine-learning.

AIFS-CRPS: Ensemble forecasting using a model trained with a loss function based on the Continuous Ranked Probability Score Over the last three decades, ensemble forecasts have become an integral part of forecasting the weather. They provide users with more complete information than single forecasts as they permit to estim...

🚨New preprint🚨from @ecmwf.bsky.social introducing AIFS-CRPS, a new data-driven ensemble system for medium-range and subseasonal forecasting. arxiv.org/abs/2412.15832

AIFS-CRPS is a variant of the Artificial Intelligence Forecasting System (AIFS) developed at
ECMWF. Its loss function is based on a proper score, the Continuous Ranked Probability Score (CRPS).

This paper presents a novel approach to obtain a weather forecast model for ensemble forecasting with machine-learning.

Our news highlights from 2024 include a major upgrade of the #IntegratedForecastingSystem; initiatives to drive forward weather science, including #MachineLearning; and news from the EU-funded services implemented by ECMWF.

For details ➡️ www.ecmwf.int/en/about/med...

Background image of world map with graph used in graph neural network processors in Anemoi. Overlaid text: “ECMWF AIFS blog. A year in ML for weather forecasting. Matthew Chantry”

New #AIFS blog: #MachineLearning (ML) for #weather and climate has continued to evolve at breathtaking pace.
Matthew Chantry takes us on a tour of the year past and gives some previews of the year to come.
➡️ ecmwf.int/en/about/med...

The Predictability of Weather and Climate group at the Physics Department of the University of Oxford invites applications for an open-ended senior research position: my.corehr.com/pls/uoxrecru...

Director-General:Reading, UK

Are you up for a challenge? We are now looking for our next Director General: jobs.ecmwf.int/Job/JobDetai...

Join us at #EGU25 to hear and discuss the latest advances in subseasonal-to-seasonal #S2S prediction and applications. Please submit your abstracts to AS1.7 meetingorganizer.copernicus.org/EGU25/sessio... by 15 January 2025 13:00 CET.

EGU 2025 S2S session.

The #EGU25 call for abstracts is open! Join our session AS1.7 to discuss recent advances in #S2S prediction, including process understanding and applications. We look forward to your contribution!

I will be at the CMStatistics conference in London from Saturday to Monday, let me know if you'd like to meet, e.g. to talk about the available postdoc position (bwsyncandshare.kit.edu/s/HXJrexofCa...)

A reminder that liking a post on bsky doesn't make it appear on your followers feed in the same way that it did on X.

If you think your followers may be interested in a post, then please do re-post it don't just like it.

I was informed by editor of our newsletter 😭

Forecasting :)