I wanted to estimate the probabilities for each quantiles using all the members of our model and pre-calculated quantiles in AI_WQ_package but need the guidance. Note that Climatological quintile boundaries are only available for every Monday .
Suppose I have a forecasts issued at 12June,2025. So, to estimate week3 and week4 probabilities, I need to get the quantiles for the date 30 June,20205 and 7July, 2025, right?
Thanks, Nachi
Hi Nachi,
Thanks for the message.
I am in the process of making climatological quintile boundaries available at a daily resolution. These will be accessible via the AI Weather Quest python package by the end of the day. You will need to update the python package to download non-Monday climatological boundaries.
python3 -m pip install --upgrade AI-WQ-package
Climatological boundaries should be available up to present day plus 8 months.
Regarding your example of 12th June 2025, the 30th June and 7th July would be the appropriate quintile boundaries.
Thanks,
Josh
1 Like
Dear Josh, Thanks so much for the confirmation of using 30th June and 7th July and also thanks for doing daily climatology boundaries. So along with weekly boundaries, we can now have daily, this is awesome! I have another request, if possible can we have the real value of climatology along with the quantile? It would be very useful to calculate the anomaly. It is too expensive to run the reforcasts for our AI foundation model, so it will be easy to use ERA5’s climatology. One quick question, for “TP”, should we do the weekly average (day 19-25 and day 26 to 32) or aggregation? Thanks, Nachi
Hi Nachi,
Weekly-aggregated ERA5 data is available to download at Training Data — AI_Weather_Quest 1.0 documentation . From this, you should be able to calculate the climatology of weekly-aggregated fields. Is this what you were looking for?
For precipitation, weekly-sum please.
Thanks,
Josh
Dear Josh, Thanks for the replies. I will use that data to estimate climatology. Thanks! Also, thanks for the confirmation on the precipitation. I have submitted the forecasts based on 20250626 for tas and mslp. Thanks for making such easy to use package. Much appreciated.
You’re very welcome. And I pleased to see a forecast submitted from SAIS2S.
Josh