Modern meteorology was born 60 years ago | Hacker News Reader

Modern meteorology was born 60 years ago

arstechnica.com

30 pointsTomte8 comments

The introductory chapters of James Gleick's Chaos are about meteorology. When I read it I was blown away by the fact that these mathematical systems had sensitive dependence on initial conditions to such a degree that there was fractal complexity.

Very cool! I had this belief that if I get another decimal point in in the observations I would get closer to the truth, but if 1.25 and 1.2 could yield diametrically opposing results that doesn't work. Threw me for a loop, honestly.

As Fry Richardson wrote:

"Big whorls have little whorls/ Which feed on their velocity,/ And little whorls have lesser whorls/ And so on to viscosity."

Lewis Fry Richardson was the person who first attempted numerical modelling of the weather. He spent his free time on the front line of WWI as an ambulance driver trying to complete a single 6 hour forecast. Unfortunately his source data was noisy and ruined the result, but it was his efforts that enabled computer generated weather forecasts - he wasn't a bit player.

I believe the problem was his numerical method was unstable, and blew up.

Somewhere in this talk by Philip Roe https://youtube.com/watch?v=uaH91P665PI

I understand that his method was rerun by the Met Office and when they excluded the bad source data they got the correct answer.

https://www.metoffice.gov.uk/about-us/who/our-history/celebr...

  limitations with the data
  method was eventually... proved correct

Does sound like they're blaming the data; though his "method" could well refer to his general method, not to every exact detail.

Philip Roe is himself a famous pioneer in CFD, so I lean towards trusting his version more... he said the forecast was of extraordinary 100mph winds, pretty specific. The bad data would have to be wildly bad to yield that! OTOH it inevitably happens with instability.

However, I really don't know. Maybe it's become a folktale, to illustrate instability, since it's a big deal in CFD!

I can't remember exactly which episode of the "Omega Tau" podcast it was, very likely "Weather Forecasting at the ECMWF", but in it it was stated that we're adding one day of reliable forecasting for every decade of research on the topic.

https://omegataupodcast.net/326-weather-forecasting-at-the-e...

I'm skeptical about the claims that Tiros-1 was useful, compared to ground station and radiosonde data for forecasting.

It would be useful for hurricane tracking, though.

Source: worked on Crays for a national weather lab.

Reply on news.ycombinator.com