Some examples of the improvement

This is from the early days of XP12

This is from the sim about a month ago after the release of 12.3

You can see the improvement in that time. 

This was P3DV5 with True Sky

 

Excellent comparison @coastaldriver👍👍👍👍

9 hours ago, jcomm said:

merge not only grib data but also satellite and Earth cam data to better paint the skies

Blending this together would certainly be a good task for ai/ml. Pergunta: does satellite provide only top-visible layers? Or are there any other wavelengths/spectra that might show mid layer clouds? Seems like that would be difficult to parse out of the noise given that it's all water.

Where meteoblue likely hit snags was that it sounded like they were using ai/ml to do the thermodynamic predictions. XP is now simply saying "Oh this spot is in the metar turf? Make it so!" and blending that with the grib, instead of letting it be set by a as-yet-fully-trained ai/ml statistical model.

4 hours ago, blingthinger said:

Blending this together would certainly be a good task for ai/ml. Pergunta: does satellite provide only top-visible layers? Or are there any other wavelengths/spectra that might show mid layer clouds? Seems like that would be difficult to parse out of the noise given that it's all water.

Where meteoblue likely hit snags was that it sounded like they were using ai/ml to do the thermodynamic predictions. XP is now simply saying "Oh this spot is in the metar turf? Make it so!" and blending that with the grib, instead of letting it be set by a as-yet-fully-trained ai/ml statistical model.

Where this all goes with the sim depiction is in my view hitting limits. To do all this data crunching on the fly means your CPU would do nothing else. Does LR want that level of burden on its current servers vis a vis real time weather in the sim. Again the more you draw in terms of cloud or wind or precipitation depiction means more CPU GPU time and load. It is only cloud and weather at the moment that produces a hit on my FPS - not drastically by the way but notable with the FPS counter running. 

Satellite data for the current met satellites is limited to high level cloud or cloud so if there are layers you only see from the satellite image the high level cloud. Look there are heaps of satellites doing other atmospheric measurements or ocean investigation. This is not a unified collected system of data that is represented to aviation as we see it , met reports, grid charts area forecasts and other operational data. 

I am not looking for a completely realistic depiction of the sky in the sim but a reasonable representation of major cloud, weather and winds. That they do ok now. Are their some unexpected anomalies or errors? Yes of course but guess what you get the same thing happen in the real world of aviation as well. The Metar said this but out the window you can see that is not at all and either much better for flight or worse! There is also a time lag in reporting 30 mins for Metars, 4, 8 and 12 hours for the various forecasts sure a Speci can be issue for rapid changes which they do but your always looking a picture of the past with this data. 

Edited January 18, 20251 yr by coastaldriver

Its never going through be about accuracy,its always going to be about what is pleasing to the eye. There is simply not enough data to provide "accurate" weather, i dont understand stand how people dont get it. If a billion dollar super computer cant do it what makes you think your cheap pc in comparison could. Madness.

5 hours ago, blingthinger said:

Blending this together would certainly be a good task for ai/ml. Pergunta: does satellite provide only top-visible layers? Or are there any other wavelengths/spectra that might show mid layer clouds? Seems like that would be difficult to parse out of the noise given that it's all water.

Yes, some satellites can indeed "see" layers bellow the tops, using microwave sensors.

26 minutes ago, coastaldriver said:

To do all this data crunching on the fly means your CPU would do nothing else.

One of the benefits of the ai/ml concept is a significant reduction in processing requirements and yes, it would be LR's servers doing that blending/crunching prior to sending off to the sim clients. Unless of course, they open up their weather API and a specialized plugin does it instead. Another benefit is the ability to merge significant numbers of inputs of varying type and resolution into one simplified model. These models are built to handle precisely these situations and use cases if you give 'em enough training data. And as far as weather goes, there's tons and tons of public data available.

 

31 minutes ago, coastaldriver said:

That they do ok now.

I agree. Hard-wiring to the METAR is pretty good enough for the most part. It would be very interesting to see a web cam feed blended into the process to spit out a cloud formation type. Maybe do that with a single frame off a camera at KJFK every half hr or so and then send the resulting type code out to the sim clients along with the METAR. Pick a few major webcams at major airports around the world and that's really not a terrible amount of processing in the end. Nerds gonna dream nerdy dreams.

I'm just curious what else those satellites offer beyond the common info fed out to an aviation forecast station. If you look at the current GOES sensor capabilities for example, you'll see over a dozen different energy wavelengths and a variety of different sensors. I just wondered if any of those could be used to parse out mid-level clouds.

 

1 minute ago, jcomm said:

using microwave sensors

Aha! I wonder what the spatial resolution comes out to there though... I guess you could at least pick off the altitude of a cloud shelf.

1 hour ago, blingthinger said:

I just wondered if any of those could be used to parse out mid-level clouds.

 

Yea that's the grib portion. Completely unrelated to my question.

2 hours ago, blingthinger said:

Yea that's the grib portion. Completely unrelated to my question.

grib is just the data format? knowing you have a grib says nothing about the source of the data? - it all comes as gribs.

On 1/17/2025 at 6:52 AM, jcomm said:

Could you be more specific please?

What exactly is failing ? Low visibility settings ? Precipitation effects on visibility ? ...

Thx!

P.S.: I use xEnviro 1.34 and it does it's job better than default weather in some aspects, although I find default weather has progressed quite promissingly in the latest updates.

Two years later, XP 12.1.3's cloud bases are still broken with real weather download on.

In this sim, I can pick any random airport with a real world current METAR showing under 1000 foot ceiling.. in the sim the ground can still be seen when well above 1000 ft AGL.  

In XP 11, same conditions... I can't see the ground when above the reported cloud base on the METAR. 

 

On 1/17/2025 at 7:30 AM, Franz007 said:

Sure XP11...with all clouds disappearing all 10 minutes and being drawn again...one of the worst weather depictions ever.

Yes this happens briefly sometimes on my end as well... but the clouds come back right away. 

4 hours ago, drix9 said:

In this sim, I can pick any random airport with a real world current METAR showing under 1000 foot ceiling.. in the sim the ground can still be seen when well above 1000 ft AGL.

This has not been my experience at all. And it is the thing I am always looking very closely to and I have been amazed about how close it is to what is being reported by the metar. It can happen in some rate occasions that a reported overcast-layer is depicted as broken but most of the time because it was in transition in real as well. A metar is already outdated after it has been published. In real you won’t get the exact same conditions as the metar during the whole 30- or 60min cycle. Often we see transitions from overcast to broken. If you see some situations where it doesn’t match at all, you absolutely should report it. But it is for sure not true in general.

4 hours ago, drix9 said:

Yes this happens briefly sometimes on my end as well... but the clouds come back right away. 

It doesn’t happen just « sometimes » but always because that’s the way XP11 draw its clouds. They are just static and will delete everything and redraw the new situation, meaning it will go from full of clouds to scattered within 2 seconds. I cannot think of a worst way; it’s completely immersion-breaking and highly unrealistic. In XP12 the clouds are dynamic. But well, if you prefer the XP-clouds in XP11 it’s your opinion but I think you are part of a very small minority, because the weather in XP12 is probably the thing that has improved the most compared to XP11.

5 hours ago, drix9 said:

in the sim the ground can still be seen when well above 1000 ft AGL.  

with the exception of the night clouds bug, really not seen this

around 1000 feet I believe it switches from cloud coverage to visibility (because the clouds do no accurately capture the vis, and vis is the important number for landing)

I snapped this at the "relief" of seeing the runway lights just breaking out of the clouds

for the prior 1000 feet descent before that there were just "hints" at the tops of buildings

We've talked a lot about scud running before

having solid vis done properly has always been a strong point of X-Plane - and that hasn't changed.

 

And the amazing thing is how close xplane looks to that video 😉

 

 thanks for sharing

20 hours ago, mSparks said:

the source of the data?

The gribs in this case are from forecast simulations: a CFD run. Mid level cloud coverage (any cloud coverage for that matter) is an output because it's a 3D domain.