@ClimatePrimate profile picture
Chris Roberts @ClimatePrimate

Senior scientist @ECMWF working on sub-seasonal (s2s) prediction.

Joined June 2009
  • Tweets 650
  • Following 319
  • Followers 335
25 Photos and videos
Pinned Tweet
Chris Roberts@ClimatePrimate
9 Dec 2024
🚨 New preprint 🚨 - "Ensemble reliability and the signal-to-noise paradox in large-ensemble subseasonal forecasts" arxiv.org/abs/2411.17694

Ensemble reliability and the signal-to-noise paradox in...

Ensemble forecasts can exhibit counterintuitive statistical properties such that the correlation between ensemble means and observations ($r_{mo}$) exceeds the correlation between ensemble means...

arxiv.org
1
1
3
148
Final published version of “Unbiased calculation, evaluation, and calibration of ensemble forecast anomalies” now available online in the Quarterly Journal of the Royal Meteorological Society: rmets.onlinelibrary.wiley.co…

Unbiased calculation, evaluation, and calibration of ensemble forecast anomalies

Standard methods for calculating ensemble forecast anomalies result in statistical inconsistencies between forecast and verification anomalies, even if the underlying forecasts are perfectly reliab...

rmets.onlinelibrary.wiley.com
Chris Roberts@ClimatePrimate
11 Oct 2024
Do you calculate anomalies from ensemble forecasts? If yes, then maybe you should be doing it differently! 🚨 New paper!🚨 "Unbiased evaluation and calibration of ensemble forecast anomalies" arxiv.org/abs/2410.06162
5
125
And here is an example 10 day forecast of 850 hPa Meridional Wind from AIFS-CRPS provided by Simon Lang. This animation shows that CRPS-based training does not suffer from smoothing and/or reduced variance for longer rollouts that is inherent to deterministic MSE-based training.
Chris Roberts@ClimatePrimate
23 Dec 2024
🚨New preprint🚨from @ECMWF introducing AIFS-CRPS, a new data-driven ensemble system for medium-range and subseasonal forecasting. arxiv.org/abs/2412.15832
1
7
258
This is because optimising a fair ensemble score means there is no trade-off between minimising error and maintaining realistic levels of variability, which is unavoidable for deterministic training. fCRPS loss is minimised when the fc is drawn from the same distribution as obs.
1
2
50
More details on fair ensemble scores: rmets.onlinelibrary.wiley.co… rmets.onlinelibrary.wiley.co…

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 distribu...

rmets.onlinelibrary.wiley.com
47
🚨New preprint🚨from @ECMWF introducing AIFS-CRPS, a new data-driven ensemble system for medium-range and subseasonal forecasting. arxiv.org/abs/2412.15832

AIFS-CRPS: Ensemble forecasting using a model trained with a loss...

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...

arxiv.org
1
7
22
6,111
This paper presents a novel approach to obtain a weather forecast model for ensemble forecasting with machine-learning.
2
2
174
Thread continued:
Chris Roberts@ClimatePrimate
23 Dec 2024
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).
1
134
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).
1
1
6
399
more replies
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…
1
4
96
… Peter Deuben , Sara Hahner, Pedro Maciel, Ana Prieto-Nemesio, Cathal O'Brien, Florian Pinault, Jan Polster, Baudouin Raoult, Steffen Tietsche, Martin Leutbecher!
1
149
Chris Roberts retweeted
Sol Osman@marisol_osman
17 Dec 2024
🚨 Don’t miss out! 🚨 Submit your abstract before the Christmas break and join us at #EGU25 to discuss recent progress in #S2S prediction @Dr_Chris_White @Domeisen_D @ClimatePrimate
Sol Osman@marisol_osman
25 Oct 2024
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! @Dr_Chris_White @ClimatePrimate @Domeisen_D
1
4
186
🚨 New preprint 🚨 - "Ensemble reliability and the signal-to-noise paradox in large-ensemble subseasonal forecasts" arxiv.org/abs/2411.17694

Ensemble reliability and the signal-to-noise paradox in...

Ensemble forecasts can exhibit counterintuitive statistical properties such that the correlation between ensemble means and observations ($r_{mo}$) exceeds the correlation between ensemble means...

arxiv.org
1
1
3
148
more replies
Crucially, there is no inconsistency between the objectives of eliminating an apparent signal-to-noise paradox and traditional approaches to ensemble forecast development guided by unbiased evaluation of forecast reliability and optimization of fair ensemble scores
1
1
42
And if you made it this far, come follow me on Bluesky! I am not sure how much longer I will be active on Twitter... bsky.app/profile/cdroberts.b…
34
Load more