Le previsioni meteorologiche del Regno Unito si prefiggono di diventare “più accurate” come Met Office svela il supercomputer che “simula il futuro”

is “simulates the future” not what weather forecasting has always been?

I can Predict the weather without a supercomputer, if it’s a bank holiday or summer holidays it will be raining, if you’re forced to be indoors there will be beautiful sunshine outside, and on the day you repair your fence there will be a storm to knock it down again.

Quite an interesting history behind this. They put it out for tender and chose Microsoft but a rival bidder, Atos threw their toys out of the pram.

Atos did a bid to win the tender rather than provide the functionality – their test environment ran on cheaper, less powerful hardware making it pretty much junk. They would have known that, but because the tender wasn’t 100% clear the challenged it and ended up getting [$29m](https://www.theregister.com/2023/01/09/uk_gov_atos_settlement/) for not supplying a system.

Highlights a real issue with government procurement; anyone with any knowledge of the system and its function would know the test environment needs to be architecturally the same as production – otherwise it won’t be valid. But Atos used that to their advantage to either win the bid at a lower price or – I suspect – not win and get money for challenging it.

That’s no fault of the met office, the people there have done a great job and this new system should be really beneficial.

*You call this a super computer?*

[*https://frinkiac.com/img/S07E17/703803.jpg*](https://frinkiac.com/img/S07E17/703803.jpg)

TLDR;

They are able to run the ensemble models in parallel instead of sequentially.

i.e. They change the current conditions a little bit, this way and that, and see what falls out the back

*however, they are talking absolute bullshit about the 14 day having similar accuracy to the 7-8-9 day window – unless they are just talking about the MET’s models which were previously so shit (vs the ECMWF/GFS ensembles) that there is no real accuracy beyond day 7 anyway. e.g. a childs guess of the 7th day weather is probably of similar accuracy to his 14th day forecast.*

Yeah heard the same thing around 1999 when they sent millions on a weather simulator. Turned out it was rubbish.

lol so to tell me when its either going to be blazingly hawt or wet.

I don’t need a supercomputer to tell me it’ll be red hot whilst I’m at work then pissing down at the weekend.

You mean they’ve just built a small machine learning algorithm

If I recall correctly, isn’t this also the meaning of the word “forecast”?

>make 14-day forecasts nearly as accurate as seven-day forecasts

They can barely get a 2 day forecast accurate and we are talking about 14 days?

What about euro millions numbers ? I’ll pay £million for correct prediction.