Meanwhile the Met Office is switching from Cray to Microsoft https://www.metoffice.gov.uk/about-us/what/technology/supercomputer
Hewlett Packard Enterprise has bagged a contract to build a supercomputer for the United Weather Centres – West, the remit being to help improve the accuracy of forecasts in Northern Europe. UWC-West is a tie-up between the Danish Meteorological Institute, Icelandic Met Office, Ireland Met Éireann national weather service, and …
This post has been deleted by its author
Friday 12th November 2021 12:54 GMT Anonymous Coward
@Edward Ashford - Incorrect
Note, the withdrawn post was me, as I forgot to go Anon, whoopsie!
Whilst Microsoft will be acting as the provider of the supercomputer, the data storage and close-coupled post processing; the actual supercomputer will be a Cray. The provisioning of the actual tin has been subcontracted to HPE.
Anon as that is dangerously close to insider information
Friday 12th November 2021 14:34 GMT Pascal Monett
"help improve the accuracy of forecasts in Northern Europe"
Okay, first of all, they dreadfully need to improve accuracy. They're barely capable of forecasting tomorrow's weather properly, unless it's clear skies all over Europe in which case they just might get it right for 5 days in a row.
Second : if you have to change your forecast every hour, you're not forecasting anything. You're looking out the window and writing down what you see.
I can do that just as well.
Friday 12th November 2021 23:55 GMT the small snake
Re: "help improve the accuracy of forecasts in Northern Europe"
Not clear if you are trolling or stupid, but. (1) Is easy to check improvements in weather forecasts over time: think for a long time has been about 1 day/decade. So forecasts for 5 days out are now about as good as those for a day out made in 1980. Unsurprising since forecasts have SDIC but computing power has increased about exponentially over that time and also data on initial conditions has vastly improved due satellites. (2) if you are forecasting weather 24 hours out (quite interesting for many) then naturally forecast 23 hours out will be better and almost certainly more than linearly better. So you want to update forecast for ~1 day out at least hourly and probably much more often than that if you can. If you are forecasting 3 hours out (very interesting for, say, sports betting etc) then you probably want to be updating every 10 minutes if you can. And if you live somewhere where you can reliably tell if it will be raining in 3 hours, well, you are lucky.