On Friday, 16 February 2018 03:18:30 UTC+13, krasw wrote:
torstai 15. helmikuuta 2018 14.01.12 UTC+2 kirjoitti:

We are actually experimenting with ECMWF as an initialisation source. If it provides measurable improvement we will utilise it for regions where it does so.


Wow, I thought ECMWF boundaries would cost arm and leg to use commercially. I bet it provides measurable improvement, it is the best model currently available and on different level than GFS. Now if someone would run ensemble of RASP using ECMWF ENS members as boundaries, we would get usable thermal forecasts for next week instead of 2 days...

We run the WRF model with EC data. It is noticeably better - there are many more levels and the horizontal resolution we get is at 0.1 degrees vs the 0.25 for the GFS/GDAS. In addition, the EC analyses and forecasts are generally a lot better.