A clarification has been posted on the website:
Participants may use any data with a timestamp strictly prior or equal to Thursday 00 UTC to initialise their models, regardless of when the data becomes available . The four-day submission window allows participants to process initial data with relatively long latency (e.g. dynamical sub-seasonal forecast data which has a two-day delay) and provides additional time for those with limited computational resource to submit their forecasts.
I’d like to suggest modifying this to earlier or equal to 00 UTC. 00 UTC is an extremely important data collection time, rivaled by only 12 UTC. It would surely improve the forecasts. Indeed, when I asked a question about data in the March 17th meeting, the answer then was that 00 UTC data could be used.
Hello Dale,
Thanks for the comment! Indeed, that was what was meant and it has now been precised on the website (Submitting Forecasts • AI Weather Quest): Participants may use any data with a timestamp strictly prior or equal to Thursday 00 UTC to initialise their models, regardless of when the data becomes available .
Dear Joshua Talib
If I use the data in ECMWF | S2S, ECMWF, Realtime, Instantaneous and Accumulated
In the first competition week, I can use the realtime forecast in 0814 or 0813 ?
Hi @Peng_Lu ,
Good question.
On the website we state the following:
Participants may use any data with a timestamp strictly prior or equal to Thursday 00 UTC to initialise their models, regardless of when the data becomes available . The four-day submission window allows participants to process initial data with relatively long latency (e.g. dynamical sub-seasonal forecast data which has a two-day delay) and provides additional time for those with limited computational resource to submit their forecasts.
Given that dynamical sub-seasonal forecast data is initialised at 00:00 on Thursday 14th Aug, this can be used for the AI WQ. You should NOT use forecasts initialised beyond 00:00 on the 14th August.
Good luck,
Josh
Thank you very much for your response. I’m not entirely clear about the initialization time in S2S forecasting. For example, if I select forecasts initialized on August 14th, when choosing step 0, does the corresponding prediction result represent August 14th at 00 UTC or August 15th at 00 UTC?
Hi @Peng_Lu ,
The step dimension typically corresponds to the number of hours into the forecast.
When selecting sub-seasonal forecasts from 14th August, you are selecting forecasts that have been initialised at 00:00 14th August. At step 0 output is representative of 14th Aug 00:00. At step 6, for example, output is representative of 14th Aug 06:00 (six hours later).
Hope that makes sense,
Josh
Thank you for your detailed and clear explanation