METAR keeps disrupting the weather/ bugged weather/Cumulus/CB clouds only/no medium to high cloud coverage

Bug Reporting Hub Weather & Live Weather
, , , ,
1330

That’s an assumption, and you know what they say about those…

I’m looking for data and valid comparisons, rather than “I think”, and other blind assumptions. And, lo:

That’s how it’s done. I’ve done the same a few times - measured in-game weather against real-world radar, in game radar, surrounding areas, live weather cams etc (California, with their extensive fire/weather camera network is a great place to do this, btw). Anyway, great work bringing evidence! In order to really bring it home, the last bit of evidence we need is a live-cam or hi-res visible satellite shot (understandably hard to do at night). Because that will resolve conjecture like this:

”Broken at 9,000” does not preclude cumulus-type clouds. However, chances are, if they were cumulus, they may be altocumulus or stratocumulus versus the cumulus congestus we see in the sim. What kind of cloud do you think it should have been?

That aside, observations don’t really differentiate cloud types except for CB and TCU (ok, maybe ACSL and other lenticular clouds at some locations). A satellite shot around the same time (I know, it’s night time) would be the final bit of evidence needed to prove that it’s wrong.

But we’re stuck decrying two variables here - METAR and cumulus, and I’ve not seen any evidence that the two are linked (other than the former produces the cloud base of the latter). Meaning theres no evidence the sim is drawing cumulus because of METARs, it simply seems to be that at some point they’ve chosen to draw fairly hefty volumetric clouds throughout the sim.

Regardless, I am not sure how they’re planning on generating different, correct cloud types. Cloud tops and types are rarely reported, outside of PIREPs. You can infer them from radar (precipitation tops) and visible or IR satellite data, but otherwise, you basically need a complete dataset of upper-air observations to understand what types of clouds you’re going to get at each layer. Unfortunately, those observations are even more widely spaced, both in geography and time, than other observations.

I don’t have an answer, I have questions as to the assumptions and conclusions I see drawn here, and the understanding that the road to accuracy is not going to be easy. I also have plenty of experience that it wasn’t great prior, either. It was also a mess. At least here, it’s closer to what’s really going on, cumulus observations (and assumptions) aside.

However, if you’re wanting whether to look correct, why not use a preset? “Live weather” imparts the presumption of matching the spatial/temporal accuracy of the real-world. I’m just not sure how it’s also, ever, going to be the correct cloud types and animate the shift between as the weather changes. Herculean task.

1 Like
1331

The accuracy limits the sim to have the weather change smoothly over time. If making a weather that feels like weather limits the weather to be accurate.

Example 1: if the weather didn’t need to match the real world weather at all it would never need hard transitions.

Example 2: If the weather needs to match weather that has already occured 100% it needs transition between those known states.

Simple fact.

I preffer the first example even if it’s not accurate. I bet you preffer the second right? I would rather observe a weather in the sim that look like weather when i observe it rather than have the observed weather instantly injected without care if it behave correct.

I think both of those examples is valid because some want weather that behaves and looks like weather when flying and some just wants it accurate but i think these examples can’t work together. Add example 1 into example 2 makes those like example 1 dissapointed it doesn’t behave like weather anymore and adding example 2 into example 1 will make those want it accurate complain.

5 Likes
1332

What I seemed to see was a mixture of weather depending on how far away from it I was. When I was further away, all I saw was the popcorn clouds. As I used slew to move closer I saw a less defined fog like cloud bank appear, but you could still make out the popcorn like clouds below it or through it. It was like the MB weather was over the top of the METAR weather. Very odd.

1 Like
1333

With the 60 levels of cloud information in the Meteoblue model, from anywhere in the vicinity of the observation, they should be able to estimate the vertical extent, and subsequently the most likely cloud species that satisfies the observation of BKN090. I think this is the big missing component in the present implementation - not using the Meteoblue model for what it can provide, basically they are throwing the baby away with the water.

1 Like
1334

I don’t disagree with this, but…

This conclusion makes several assumptions. Are they throwing away the Meteoblue data? If so, what about all the area between reporting stations? From where is that weather being generated?

It would need to have transitions, and they’d be as faked as anything else because you have to inject updated weather somewhere. Our current problem is at which time resolution to do that. Some folks are comfortable with it being several hours off and jumping back or forward when the revised forecast doesn’t meet the reality portrayed in the sim (based on the old forecast). Some are not comfortable with that.

That’s an assumption. I prefer it to be as accurate as possible and useable with forecast tools and real-time observations. Weather can’t look good if it’s not accurate because those who know, know. They see through the gaps in realism and no matter how well it’s presented, it looks bad.

I understand the data with which to do this correctly is limited. I don’t know why (almost) real-time radar data aren’t used more in areas in which they are available, especially with regard to thunderstorm development. FWIW, I agree here with everybody who wants more layers and realistic-looking clouds (and thunderstorms). 100%. And I don’t think you’re going to get it from METAR, never have I made that claim - I agree with you all there.

METAR is best for wind speed, cloud height, temperature, and barometric pressure at an aerodrome. I personally wouldn’t use it to generate cloud coverage or type as those are some of its major limitations, especially over a wider area. Even visibility is usually a function of precipitation (or fog/haze/mist/smoke) and I wouldn’t use it for that, either. Maybe the answer is to only inject METAR ceilings and visibility when the weather is MVFR or lower (or just let it deal with smaller “bubbles” less than 5SM vis and 3000’ ceilings. But that still doesn’t rectify the aforementioned forecast limitations. I don’t know what the answer is there.

But in the end, I’m actually okay with it being a little off, as that’s realistic enough. However, I’m not okay with it being several hours or a hundred or more miles off, which it has proved to be many times.

There are a lot of assumptions being made in the thread and it all boils down to one thing - not one of us here knows how it’s actually working behind the scenes. We’re still at the “collect observations and data to compare how incorrect it is to the naked eye or reported weather” stage. And “I feel like it should look like this” or “it looked great before su7” are personal anecdotes that don’t always match the reality observed by others.

Anybody miss the puffy clouds forming from the surface of flat land? Or suddenly swapping between regions of weather that were completely incongruent to anything that was forecast? Or wildly inaccurate altimeter setting or wind readings? I don’t.

Do I like to be able to use aviation weather tools to be able to plan a route and make in-flight decisions based on real-world observations? Absolutely. That’s the closest simulation to real-world you’ll get. Anything that can’t use those tools (or generate a complete set of their own) reliably is bogus for “simming,” because it’s outright missing a huge part of general aviation - planning and ADM based on forecast and observation.

1335

that’s why the weather is mostly the same everywhere, the bug generated from the METAR implementation since su7 is that probably the weather generated from METAR is extended at the whole world, and only the layer reported from metar, hence the one layer of clouds (cumulus only), because METAR does not report what type of clouds they are, so MFS draws the standard cumulus clouds. You might call it an assumption as you like to say to us when we are trying ro work out what’s going on, but the fact that this topic is logged as a bug means the developers agree with what’s reported here since a year and half from sim update 7!

2 Likes
1336

ALL: Just a reminder that all bug topics are to report a bug and to supply further information for the developers regarding the bug. Discussions should be kept out of bug topics. If you wish to have a discussion, please create a new topic in #gd-interests:general-discussion or take it into a private message.

Slow mode on bug topics are to help prevent discussions.

Thank you.

1 Like
1337

In the interest of finding common ground with which to resolve this bug (or even discover the realities of it): There is definitely weather between METAR reporting stations. METARs seem to modify it, not the other way around.

For an example, take a look look at northern CA right now. There is a storm front rolling through the Central Valley and the METAR are only changing things on the periphery where rain does or doesn’t really exist at that moment. But the whole storm system is definitely not coming from METAR alone - it’s a much longer, congruent front (not entirely accurate in position as of 90 minutes ago when I looked into it), which would be impossible to re-create with METAR alone.

So how do we go forward in squashing this bug? I think it’s important to understand what the sim is actually doing compared to different forecast sets, current observations, and observed reality.

2 Likes
1338

The thing is it’s not a bug. The data of clouds we have now in the sim coming from Meteoblue. The fog layer is on the client side though (I know because when the live-weather has been down i still get the fog from METAR) It’s different data than what we had pre su7. Back then we had the simulation model without postprocessed data. After su7 we get the postprocessed data that includes Messured and observation data. It’s not Asobo that decides where the data of clouds should be shown. If Meteoblue postprocess the data to be in a circle around airports we will see it as in a circle in the sim.

I preffered the data without postprocessing done.

Asobo also explained pre su7 that they changed the global clouds to be more realistic they said back then. In my opinion it were the oposite to more realistic.

Here they explain the system before su7 were released.

Meteoblue injects METAR data before they sends it to Asobo. It’s a different model in use. It’s intended. But not all agrees thats better (me included). Thats because this topic exists i bet.

2 Likes
1339

This illustrates exactly I’m saying. We can’t even come to terms as to whether this is a bug or not. We’ve been warned several times that it is one and to not have side discussions, but you say it isn’t a bug.

Then there’s the claim that it’s post-processed, and that may be the case on the boundaries where weather is moving in unforecast ways, but why does that preclude drawing the rest of it correctly? And if it is precluding that meterologically-correct drawing, then is that the bug?

Again, it seems to have nothing to do with METARs, per se, rather it is coincidental with when MB decided to post-process, if that’s indeed the case. The METARs, as far as clouds go, seem to be only affecting the edge cases where things are moving differently than forecast. And I will say that my observations have been that these discrepancies have been a lot closer to reality than it was several updates ago. To the point where it’s almost not noticeable in the windscreen - I haven’t seen a full-on METAR “bubble” in the visual depiction in a long time and now if I happen to come across one, to me it’s evidence that the forecast was extremely off (or the surface reporting station is broken). Instead everything is moving and flowing and pretty convincingly. It’s just the layers/variation of clouds that’s missing.

It will be interesting to observe these forecast discrepancies during the upcoming Midwest thunderstorm season. Storms may be forecast at a certain time of day, but actually only kick off when diurnal heating and a combination of other conditions break the convective-inhibiting “cap.” Those initial, developing stages can go from relatively clear skies, to an agitated AC/CU field, to CB really quickly.

But to bring it back on-topic to the bug: What is it, then? Why are we not discussing it as one and why do we keep doing it in this thread if it’s not a bug?

1 Like
1340

It’s good if they found a bug and i bet there is bugs in weather but nobody here can report bugs of weather because nobody knows how the system works besides Asobo. We can only tell what we observe. And they mentioned in su7 they added more data from METAR that is injected in the meteoblue model. And in this title it shows what we have observed since they implemented the METAR. Then it’s Asobo that needs to find the cause if it’s a bug. We can’t help them with that more than give feedback of what we see. Maybe use the weather debug tool but that doesn’t show much how the weather model/engine works.

1341

https://forums.flightsimulator.com/t/beta-update-1-31-19-0-february-27th-2023/578445

1342

Correct. Even with myself being in the Beta, I really have no way of knowing or verifying any of those changes. Someone else who has kept a close eye on this stuff who is either in the Beta or when SU12 comes out can probably give a better before and after comparison. Only weather change I can tell has changed is the issue with the weather not injecting over 5 hours.

1343

Its not just high level cirrus clouds… its thunderstorm clouds and or supercells that expand up to FL600 that needs to happen.

6 Likes
1344

Stratus and so on and so forth, a lot has been lost on the way…

8 Likes
1345

Ues indeed

1346

Apparently the weather over Leeds UK right now…

image
image1920×1027 130 KB

5 Likes
1347

No way that is Leeds in the evening, it’s the Inter-Tropical Convergence Zone at noon.

Leeds has been approached by a warmfront, with a low ceiling and light rain.
But as MSFS is not capable of displaying a high-reaching, multilayered cloudscape of BKN SC/AC/AS/CS, as it would be typical for a warmfront, it displays a single layer of high reaching cauliflower puffs.

Pre-SU5, with a superior lighting, cloud rendering and convincing layering, it would probably have looked more like this:

Multilayeres clouds with AS
Multilayeres clouds with AS1920×810 36.3 KB

17 Likes
1348

I miss it so much :frowning:

14 Likes
1349

Well, as much as i want the 2020 weather system back. Actually it’s my number 1 wish for this sim but I think the post su7 weather is what we will see gets improved on. Too much work to get the old system back again i bet.

Accuracy is the dominant thing in flight simulators. The look, feel and behaviour of weather will allways come second. I hope some day it will be the opposite and we get observations of the weather in the sim instead of using real observations to create static/generic weather with and let the weather be a fluid as it is IRL.

2 Likes