Clarification on models included in the multimodel_mean

Workshops, competitions and events AI Weather Quest
1

Hi, I have a quick question about the S2S models used in the AI Weather Quest.
I noticed that the model overview page lists configuration details for seven​ dynamical models (ECMWF, ECCC, NOAA, KMA, JMA, CMA, HMCR). However, the leader board page seems to only show results for five​ of these models (ECMWF, ECCC, HMCR, JMA, NOAA).
Could you please clarify: is the multimodel_mean result calculated from all seven​ available models, or only from the subset of five​ that appear on the leader board?
Thanks for your help

2

Hello @Peng_Lu

Thanks for your message.

Unfortunately the inclusion of dynamical model scores is dependent on forecast data stored on ECMWF | S2S, ECMWF, Realtime, Daily averaged . For certain modelling centres, this dataset can be temperamental and miss out certain weeks.

For every week I attempt to download all seven models, compute a model average from all models downloaded, and then evaluate the respected fields for the leaderboards. I have tried to keep information regarding missing forecasts updated on Sub-seasonal forecast models - AI Weather Quest - ECMWF Confluence Wiki, but please note I sometimes forget to update this page. An example of an issue includes CMA where no forecast data is present post 20250925.

Thanks,

Josh

3

Thank You for Your Swift and Clear Response

4

Pleasure. I have also updated the confluence page with information regarding the latest missed forecasts.

5

Super! Thank you very much.

1 Like