Hacker News Reader

Comment by bufferoverflow | Hacker News Reader

The big problem with precise weather simulations is that we don't have the sensors in 3D. They are all pretty much sitting on the surface, apart from some weather balloons. So we don't have precise data for wind, temperature, pressure apart from the surface. And all of these factors change A LOT with altitude.

View on HN

Are you stating this because you are a weather researcher or because you think that?

Because we do have weather balloons and weather satellites and weather stations in variant heights and situations.

But at a ridiculously sparse density. At anything above 10km, it’s not even 1 data source per hundred thousand square km of area at any given time.

Countries the size of Poland go without for at least a decent chunk of the 8760 hours in a year.

I see this pretty often here unfortunately. Very confidently wrong statements from people who aren’t actual experts. For historical reasons there’s a sort of guise of expert on HN that can be abused.

This is how the tech nerds operate. Read a wikipedia paragraph, and become expert.

Satellites only look at the surface (and maybe clouds).

Weather balloons are extremely sparce.

The Aeolus mission: https://earth.esa.int/eogateway/missions/aeolus

> The Aeolus mission objectives were to provide accurate global measurements of winds from the surface up to 30 km

We have weather satellites observing everything from above though. Combined with the ground based sensors we have a pretty decent clue.

Yeah, this seems like the perfect problem for AI to solve. Combine different data and find hidden relations.

Wrong. Satellites only look at the surface and maybe clouds when available. We have almost no data from above the surface. Very few weather balloons.

No, over a million meteorological observations are collected each day by commercial airplanes, and these are a principle source of data for weather models. Weather balloons and other sounding instruments (satellites are capable of this) also make significant contributions, of course.

It's been proven for decades, the big problem with weather simulation is that the dynamics are chaotic. They continue to be, even as we collect more data from the surface and above. Our forecasts improve, slowly (by a day per decade or so, historically), but simulation cannot be accurate and precise in the same sense.

A good introduction is "Weather Prediction: What everyone needs to know"

Reply on news.ycombinator.com