I disagree

19 hours ago, blingthinger said:

We can accurately predict a weather pattern many days

<24 hours when accuracy actually matters (such as in aviation)

else TAFs (which are still often wrong) would go out further than 24 hours. 

see also

https://wiki.ivao.aero/en/home/training/documentation/TAF_explanation

See again hurricane predictions, multiple day's warning about the recent polar vortex deviation into the southern US. What was that storm that recently hit the UK? Bet there were a few days warning there. All still "black magic" and improving further with the next gen of CFD-trained ai/ml models.

1 hour ago, blingthinger said:

See again hurricane predictions, multiple day's warning about the recent polar vortex deviation into the southern US. What was that storm that recently hit the UK? Bet there were a few days warning there. All still "black magic" and improving further with the next gen of CFD-trained ai/ml models.

They finally managed to predict it would be very windy during a hurricane and last for several days.

"slow clap"

I mean, sure, it is a vast improvement over

but to call what they forecast now accurate is still a stretch at best.

enough pilots and their passengers still die on a regular basis getting caught out by incorrect weather forecasts that putting that idea to one side is definitely for the best.

even being out by a few minutes on the weather now can end up being lethal

 

Edited January 28, 20251 yr by mSparks

Hey, don't let me stop you from actively downplaying and minimizing humanity's scientific achievements. I guess this is all par for your course. And bonus, I'm sure the LR employees who do peruse these forums love seeing how well you interpret the technical world around you.

18 minutes ago, blingthinger said:

minimizing humanity's scientific achievements

show me real science - you know, stuff that meets the definition of science 

"the systematic study of the structure and behaviour of the physical and natural world through observation, experimentation, and the testing of theories against the evidence obtained."

and Ill be convinced.

But when most climate "science" is usually in the form of "most climate scientists agree"

the name for that is

https://en.m.wikipedia.org/wiki/Argumentum_ad_populum

not just "not science", but pure scientific fallacy.

mix that in with corruption of the peer review process, refusal to accept large volumes of actual evidence that undermines their theories, and a huge corporatocratic machine vested in making people believe things that almost certainly are not true so they can buy themselves luxury yachts in monaco. 

And I'm gonna stick with what I know has been proven to work with a clear conscience. 

btw

LOC = line of code

still waiting for the links to any CFD stuff they are actually running to generate these forecasts.

Yeah yeah. Keep going. This is good stuff.

Large Eddy Simulation (LES): LES is a CFD technique used to simulate turbulent flows by resolving the larger eddies directly and modeling the smaller scales using subgrid-scale models. This approach balances detailed simulations with manageable computational demands, making it ideal for capturing transient turbulent structures.

Meso-NH: The Meso-NH model is a non-hydrostatic mesoscale atmospheric model that incorporates advanced numerical techniques, including CFD methods. It is designed to handle a wide range of scales, from synoptic to turbulent, and includes advanced physical parameterizations for representing clouds, precipitation, and other atmospheric phenomena.

A preliminary study of assimilating numerical weather prediction data into computational fluid dynamics models for wind prediction - ScienceDirect

Both models leverage CFD techniques to provide accurate and detailed simulations of fluid dynamics in their respective applications.

and... the FUTURE Now!

Anemoi: a new framework for weather forecasting based on machine learning | ECMWF

Machine learning and physics in weather forecasting: a discussion between Alan Thorpe and Florian Pappenberger | ECMWF

Edited January 28, 20251 yr by jcomm

Earth weather is an interesting application for ai/ml models. The geometry doesn't change, the gas state is relatively consistent, and there's tons and tons and tons of data and accurate CFD training inputs available now.

The rub is in that discussion link:

"...many users probably do not always recognise or care about dynamical and physical consistency."

As is obvious in this thread, it's very easy to fall into the trap of thinking that a stats model is obeying the laws of physics just because you see pretty colors on the screen. 

Now... CFD with some ai/ml mixed into those physics processes that are clearly statistical: e.g. LES subgrid models... that would be interesting to watch. 

 

1 hour ago, jcomm said:

A preliminary study of assimilating numerical weather prediction data into computational fluid dynamics models for wind prediction - ScienceDirect

This link very clearly states that they used the data as source terms in the CFD. A step beyond initial condition but I would think for a steady-state simulation only (1 instant in time). Is this a common configuration? I wonder if that could possibly be another point to inject some time-history statistical trends into the model.

50 minutes ago, blingthinger said:

This link very clearly states that they used the data as source terms in the CFD. A step beyond initial condition but I would think for a steady-state simulation only (1 instant in time). Is this a common configuration? I wonder if that could possibly be another point to inject some time-history statistical trends into the model.

Yep! It's called reanalysis 🙂

Climate reanalysis | ECMWF

OTOH, also worth reading:

Fact sheet: Ensemble weather forecasting | ECMWF (watch the vid at the end...)

Edited January 28, 20251 yr by jcomm

1 hour ago, blingthinger said:

stats model

just to be sure, you dont think dtmc is a "stats" model do you?

because its sounding very much like you think it is.

And if not, what "stats model" are you referring to?

Edited January 28, 20251 yr by mSparks

I wonder how all this "bible of theories" has any objective correspondence with the functioning of my XP12 for PC.
In yesterday's flight the real weather behaved terribly, both in terms of physics and graphics.
So please excuse my pessimism, but it seems to me that this "bible of theories" concerns more science fiction films than the real (and current) behavior of XP12. 🤔

8 minutes ago, efis007 said:

I wonder how all this "bible of theories" has any objective correspondence with the functioning of my XP12 for PC.
In yesterday's flight the real weather behaved terribly, both in terms of physics and graphics.
So please excuse my pessimism, but it seems to me that this "bible of theories" concerns more science fiction films than the real (and current) behavior of XP12. 🤔

You weren't really expecting real weather where. If you are your going to be disapointed. Be close in places with the corcect data, others where there is littke be a best guess. Which it is anyway.

1 hour ago, efis007 said:

I wonder how all this "bible of theories" has any objective correspondence with the functioning of my XP12 for PC.
In yesterday's flight the real weather behaved terribly, both in terms of physics and graphics.
So please excuse my pessimism, but it seems to me that this "bible of theories" concerns more science fiction films than the real (and current) behavior of XP12. 🤔

It's just us with bla-bla-bla... X-Plane 12, or any other desktop platform, are far from reaching that level of detail, but they can still manage to do an acceptable work, provided they're fine tuned.

The runs of the forecast models take hours in our HPC, fully dedicated to that task... Just imagine trying to do the same in a desktop flightsim...

But AI might change this, actually a mix of AI and streaming ... 

"physics" needs clarification. Physics of the turbulence bouncy? In that case yes, I somewhat agree 😐 but it's much better than 12 at release a couple yrs ago.

Physics of the atmosphere? What aspect? XP is downloading these +3 hr predictions and interpolating between the 2 time stamps. We'll see pretty good winds aloft (jetstream), temperatures, pressures, humidity and some broad strokes of cloud cover. If it's clear/sunny IRL for km in all directions, it will be sunny in sim. All that alone is a huge jump from what it was in the past and regarding OP, a bit better than a consumer ai/ml model.

I have yet to see XP totally barf on the real weather enough to cry tears. Every situation I see in a METAR area is broadly pretty well-replicated in sim.