News for hackers. Stuff that matters.

A comment on U.S. has its wettest 12 months on record again. Parent comment

Re:U.S. has its wettest 12 months on record again by pazimzadeh on Tuesday July 09, 2019 @06:33PM (#20397399)
I'm not sure what the use is of comparing human processing to "the type of operations required to simulate something like climate" since predicting the weather is not something our computers are particularly good at. Why is a 3D model even necessary?

Personal anecdote: my grandmother is capable of predicting rain 1-2 days ahead of time based on smell and just "a feeling." Granted, we probably don't remember all of the times that she was wrong.

[ Parent | on HN ]
Re:U.S. has its wettest 12 months on record again by dlivingston on Tuesday July 09, 2019 @06:42PM (#20397472)
> Why is a 3D model even necessary?

See [0]. PDEs can be solved discreetly in three dimensions on a three-dimensional grid (3D model, sort of).

[0]: https://en.wikipedia.org/wiki/Finite_volume_method

[ Parent | on HN ]
Re:U.S. has its wettest 12 months on record again by lutorm on Tuesday July 09, 2019 @08:44PM (#20398222)
That doesn't really explain why you need a 3d model, though. If there was no important vertical process, you could get away with not resolving it. However, vertical heat transport is the entire game since the lower atmosphere is opaque to heat.
[ Parent | on HN ]
Re:U.S. has its wettest 12 months on record again by dlivingston on Wednesday July 10, 2019 @01:37PM (#20404303)
I mean, it does though. The atmosphere is a 3D volume. Parameters that you might be interested in solving for, like advection and convection, are 3D processes. Consequently, you will need a 3D grid to solve the PDEs describing these parameters.

EDIT: Ah, I see what you're saying. Yeah, I believe we fundamentally agree and our posts describe the same reasoning in two different ways.

[ Parent | on HN ]