back to article Weather gets granular with GPUs

Everyone complains about the weather, but no one is doing anything about it. The folks at the National Oceanic and Atmospheric Administration (NOAA) aren't doing anything about the weather. They're too busy trying to figure out what it's going to do tomorrow and next week. I sat in on a very interesting presentation from NOAA …


This topic is closed for new posts.
Silver badge

Just a thought

When our national weather service had not yet "upgraded" its system, the weather reports were, as I remember them, accurate enough to actually plan upon. In other words, if they said this week-end was going to see rain, you didn't plan a trip to the beach and, more often than not, they were right and you had no regrets. Well, sometimes you did, but more often you didn't.

Then they upgraded their system, and spanned their prediction time to 5 days, inserting a "confidence ratio" while they did it. It soon became obvious that, although they had more than doubled the computing capacity, they had also more than doubled the error margin.

Now, unless there are clear skies over half of Europe, we have trouble knowing whether or not it will rain tomorrow. Forget about a 5-day forecast - they haven't made one that actually panned out.

So what I'm asking is this : given the dismal failure France has already recorded in improving on its computing capacity, what steps are being taken to ensure that all this improved granularity isn't going to end up going the same way ?

Bronze badge

NOAA's problems run deeper than that

About 80% of all US weather stations fail to meet NOAA's minimum requirements to provide accurate data according to On a scale of 1 to 5, with 1 being the best, over half are ranked a 4. To be considered accurate, a weather station is supposed to be 150 feet away from any structure and be 5 foot off the ground. Many sites sit near pavement, buildings, and AC units. The result is, of course, higher temperatures than what is really exists, especially at night.

When you feed garbage into the model, garbage comes out. If you can't get the here and now right, how can you get the future right? NOAA needs to focus on getting the basics right first. Until then, is my source.


Why do you think they do QA ?

Strangely enough, there is a QA step where observations are blacklisted before being fed into the forecast. Many / most observations don't make it into the operational forecast, and a lot of the work at NOAA and national weather services goes into QA.

Why do you think NOAA checks the standards of the weather stations if not to do this ?

And as for, where do you think they get the data from, and the computer model ? They run a statistical analysis on top of the GFS output and data from NOAA / NCAR ...


Great British pastime

Why on earth would anybody want to improve these predictions?

What us Brits do best is talk about the weather, it brings generations together with a common talking point.

I have enjoyed nothing more than moaning to my fellow man about the lack of predictability of the weather.

Like cups of tea, Heinz Soup, and Cadbury (is that off the list now?) and crumpets, there is something so British when talking about the weather.

This topic is closed for new posts.