Please make METAR optional till optimized

that makes sense. when ideed it is just a different data set from a single data provider, it indeed might not be possible. But then at least the data should be from the same time…

I for some reason assumed that data is from tow differen sources - which would explain that the timestamps are not the same - and then it shouldn’t be to hard to turn one data source on or off with a switch.

But yeah. I didnt report it as bug. I didnt even dare to put it on the wishlist.I just wanted to throw my thoughts into discussion, that it would be nice to have a switch like fpr DirectX until the new system is improved.

You’re correct. Prior to SU7, I believe the METAR data was from a different source. Someone on the forum traced the network connections on their PC and confirmed the data stream.

But I think METAR data is now coming from Meteoblue also, which leads me to believe whatever data set they get must be different than what they were getting prior to SU7.

A point that might sound pedantic but I think isn’t, at least sometimes: a METAR is a direct observation of current conditions at an airport, and model data is a forecast, not a direct observation.

If I understand correctly, the Meteoblue data is a model. If so, that means it’s a global computer simulation of what the computer thinks the weather should be like at any given time, based on synoptic-scale features such as pressure systems, fronts, jetstreams/streaks etc.

The problem with this sort of model is that it often can’t make adjustments for local or mesoscale features / effects like marine or terrain influences. So your local weather forecaster will examine model data (several different models in fact), but will then apply his/her local knowledge and judgement to create a forecast that is typically more correct than the model.

TL;DR - any forecast can be in error, a pure model forecast even more so. This is why the transition from model-predicted conditions to METAR reported conditions can be so jarring - it’s not the METAR that’s wrong, it’s the Meteoblue model. This is also why it’s so important to be using METAR data near airports; the model data is often entirely inaccurate, especially when dealing with localized effects like fog or marine layer.

They definitely need to smooth out the transitions to / from the METAR data areas, but even with the jarring transitions this is still an improvement over the old purely-model-based ceilings, vis and precip, which were often entirely at odds with reality.

yeah, i think the same.
But I see the advantage of the model as well.
A METAR is a static observation and only updated every 30, 60, sometimes only every 180 minutes or even less.
Without a model ( based on real world observtions like METARs) the weather will change from one second to another when METARs get updated and transition can be very hardfrom one to the next.
So I really like the idea of MeteoBlue Model, but it is only as good as the underlying data.
And mixing a dynamic model with static METARs is bringing new challanges and we can observe these in the sim right now.
Especialy at mornings and evenings weather can change significantly within an hour depending on region and season. MeteoBlue is dynamicaly adapting to these changes, while METARs “jump” from one static observation to another.
Getting this right will take some time, if possible at all.

  • A METAR is a tiny fragment of historical weather. It’s also wrong in the sense that it’s old, incomplete, and sometimes even the observation itself can be bad.
  • Meteoblue’s model is a cartoon guess at the current, whole atmosphere.

Pick your poison. And Asobo has chosen both, along with the side effects and consequences that come with both.

I think ultimately Asobo is on the right path. “Live Weather” doesn’t exist, and needs to be created from disparate pieces. Adding data to fill the gaps in the forecast model is the way forward, but it needs to be done carefully and with proper testing. Otherwise we get SU7, a concoction of an inaccurate picture of the whole weather scene, polluted with bits of stale, unrealistic weather observation.

I just wanted to point out that this system will always be flawed and quirky, and the fix is not to regress to previous versions or rely more heavily on a single set of flawed data like METARs.

1 Like

METAR and dynamic weather don’t really fit together like I mentioned before:

Has anyone watched this video? The meteoblue model sounds quite complex already. Was it not working for them to decide to use METAR instead?

of course MeteoBlue Model is complex and working very well.
THey are offering their services to companies, especialy aviation on a professional basis.
When I start a flight, it is not as important as what the METAR is at destination now, but what the weather will be when I arrive.
And of course as an airline I also need to know what the weather is enroute, where no METAR coverage is givien.
This is where the MeteoBlue Model kicks in and is among the best on the market.
The question is, what dataset MSFS is getting and how noch of it can be rendered and simulated in the sim.

METAR is only gving airport weather observation, it does not deliver any information on weather in high altitudes or outside the Airport Area. It does not include Forecasts and are updated in various intervalls.
Major Airports update them every hour ( at least during operation) but some METARs can bei updated only every 3 hours and not at night ( when there is no traffic for that airport).
Given this limitations fo METAR , a good model is not only better but a must-have in modern aviation.
Bringing this to the SIm and blending nicely with the METARs that are most important for departures and arrivals is the tricky part Asobo is currently trying to achieve.

METARs were introduced on community request. It is very important for those flying online on a network like VATSIM and IVAO. There curent METARs are used and without that in the Sim, weather can differ quite a lot even ending with a tailwind landing because the METAR for the AIrport is not the same for Sim and online Controller.
So yeah, for those people METAR is very important and I am happy for them they got it.
To them the not so nice transitions is probably a small price they are willing to pay.

For those not flying on online network, that dont care if the weather is 100% accurate or just realistic, the current implementation of METAR is a degradation of their expirience.
Therefore I suggested, to make the METAR priority over MeteoBlue Model optional, so everyone can have a good experiece.
If this is technically possible, is a different question, though…

2 Likes

Very very good question. The answer, if there is one, is loaded with a lot of history. My summary that follows is based on my engagement in the forums over the past year and a half, so if I’m missing key parts of what has transpired, please feel free to correct me.

Shortly after the sim was released, people started noticing that live weather was very inaccurate. It wasn’t happening to everyone, but a good portion of users were definitely having issues. Now, I’m not talking about live weather being completely down (that’s a separate issue). What I’m referring to is users loading into a flight and seeing wildly different weather than what was forecasted and observed for that area.

Asobo had decided to completely close off the weather system to any third parties and had provided very few details on how the weather system actually works. Yes we had the feature discovery videos and Q&As so we knew some things, but without granular details, many of us were left guessing as to how the system gathers and interpolates the weather data, how it’s delivered to us, and most importantly…when it’s delivered to us.

Due to that lack of knowledge and not knowing what to expect as output from live weather, the forums became rampant with discussions of using METAR instead of Meteoblue’s weather model. Many users began comparing METARs to what was depicted in the sim at any given time.

After some time though, a bunch of forum users started really observing and began to identify trends. Eventually, it was discovered that for about half the day, every day, the live weather system was incorrectly rendering the wrong forecast hour, pushing it ahead by almost 12 hours. This was the actual cause of a lot of users seeing incorrect weather. Not a problem necessarily with the weather modeling itself or lack of METAR data, but a problem with delivery.

During that time however, Asobo began to inject temp, pressure, and winds from METAR into the simulation as a supplement to Meteoblue’s weather model, and that’s the way it had been up until the recent changes introduced with SU7. I found that this worked pretty well as it allowed for the correct assignment of runways for ATC users and provided an accurate baseline to determine aircraft performance on landing and takeoff. The visual weather wasn’t always a 1:1 match with METAR, but it was reliable enough to determine whether you prepare for VFR or possible IFR conditions.

After the forecast hour issue was fixed, the forum discussion of live weather inaccuracies dropped dramatically. Yes there were still minor issues like missing visibility, temperature spikes, overdone lightning, etc., but for the most part things settled down and weather was mostly accurate for a lot of people. However, by that time Asobo already began developing plans to overhaul live weather.

My point with all this is that it feels that SU7 is the result of Asobo attempting to fix live weather inaccuracies based on assumptions everyone was making earlier this year even though many of those problems were rectified by an entirely different solution. It also seems to be to cater to those simmers who for some reason expect to see a 1:1 representation of what METAR is reporting. Those legacy simmers for decades became used to METAR based weather systems in flight sim, especially those who fly in online networks like VATSIM.

Personally, all I think they really needed to do was find a way to add in visibility, improve the transitions of METAR (temp, pressure, and winds) and improve the accuracy of preciptation. I think they’d be better off sourcing visual weather data (clouds, vis, precip) from non-METAR sources. METAR has inherent limitations to the data it reports (ie. 10SM visibility limit, cloud layers only up to a certain height, etc).

But unfortunately, we are where we are now. I assume that the data sets provided by Meteoblue have changed with SU7, and the idea of making METAR optional may not be possible without Meteoblue providing two different data sets, one for SU7 weather and one for pre-SU7 weather, if indeed that is how it’s working.

3 Likes

METAR data isn’t “flawed”, “wrong”, or “old”… it is updated as necessary (not just once an hour) based on meaningful changes in conditions, via SPECI observations. I’ve seen this happen, with a corresponding change in weather depiction, in the sim post-SU7.

Nor is it some new frontier of boundary-pushing magic to blend observation and model data in the sim; ActiveSky has been doing it for years.

It sure can be. The weather can change quite a bit with no special report being issued. I’ve watched stations update hourly with 90 degree swings in the wind direction. If the wind speed isn’t high enough or the direction not changing fast enough, they might not issue an update. Other reported criteria may not be considered significant enough to warrant an update like changes in sky conditions, ceiling or visibility to levels that aren’t critical. The numerical model might actually be more accurate in such instances, even if it’s a forecast generated hours ago.

Stations may also have known biases and error due to instrumentation issues and limitations. What’s cool is many numerical models actually take this into account when ingesting data so that the forecast isn’t corrupted as a result. Stations where the temp is known to be off by 10 degrees get filtered, otherwise you see weird spikes and blips in the model output.

1 Like

Anything can be wrong. But setting aside outlier cases like that, what’s more likely to be accurate at a given time: a direct observation that’s issued at least once per hour and more frequently when conditions dictate, or a computer’s best guess that’s updated every 12 hours (Meteoblue is only updated twice a day)?

The whole reason they incorporated METARs in the sim in the first place was that the Meteoblue weather depictions were often severely inaccurate, right?

You’re right that the METARs, the few bits of weather they’re tracking, are more accurate in general than numerical models. But it’s not just outlier error cases. It really depends on the situation. When changing weather is occurring, which is pretty much everywhere, you will always see noticeable lag in the METAR, even when special reports are issued. The METAR is old as soon as it’s released because the weather has already moved on. And after an hour the weather can be markedly different. I stare at screens for my job where METARs are overlaid on top of radar and visible satellite, and you can tell when the METAR is old because the station observations no longer match what you’re seeing on the other feeds. It’s most noticeable when there’s a front moving near the station, where the model is trying to get the timing right on that front, but the METAR is reporting conditions from an hour ago. If the model is right on that timing, blending in METAR information will actually give you a less accurate picture. We wind up using the METARs not to get real time reports, but more for verification purposes for what the models said the weather should be doing, awhile ago, and still have to interpolate or use forecasts forward from there to get currency

I think a lot of people are hung up on the idea the METAR is the real weather. It was the real weather. It was a tiny bit of real weather.

If the numerical models were severely inaccurate, they wouldn’t be useful to Meteoblue’s customers. There’s uncertainty and error of course, but they still have valuable skill in prediction. Instead, the model has gaps where it’s incapable of showing highly localized or specific types of conditions. The goal of the METAR injection was to try to fill some of those holes, and hopefully also improve on some of the error.

But you’re still mixing old data (METARs) with wrong data (numerical forecasts). Both are increasingly diverging from reality. It’s a lot harder of a problem to solve now than it used to be on decades old solutions that were blending much more coarse and basic weather information such that it really didn’t matter if it was using hour old fragments.

2 Likes

It seems like you’re trying to debate the point, but you end up agreeing with me in your last paragraph, by pointing out that a METAR is a direct observation but the model data is likely wrong ;).

Sure the METAR can lag a fast weather change until a SPECI is issued, but it’s still the “official” weather at that airport, and thus as long as a sim (any sim) is using METAR- based weather depiction, we’re all seeing the same weather on VATSIM etc. There’s no other way to achieve that across multiple sim platforms. (It also makes real-world flight planning and EFB tools like Foreflight usable with the sim, where purely using model data would not.) And the large majority of the time, that weather will closely match actual conditions because rapidly changing conditions aren’t the norm. Besides, a model that was last run 9 hours ago isn’t going to be any better at nailing the timing of a rapid change, if it accurately forecasts the change at all.

The value of the model (in real life and the sim) is that it provides a large-scale forecast, while METARs are specific to small areas but are at least direct observations (though there’s no reason not to incorporate TAFs too). But these areas (airports) are the ones that matter most for any sort of organized flight simming, like on the networks.

The right approach seems to be what other weather injection systems like Active Sky have done for years now - use model data for general conditions, refined near airports by METAR data. Asobo just needs to smooth out the transitions and limit the popping, but this should be doable (since systems like Active Sky have been doing it).

As you say, METAR are most important for airports and for organised flying on online networks.

Unfortunately in its current implementation the transition between airport weather and the weather model covering the rest of the world needs improvement and as not everyone is flying on online networks this is a throwback for the vast majority of users to give a better compatibility for a small frsction of the community.

Noone is asking to pull the METAR implementation. It is great it is here,it is fantastic that online networks should now be more enjoyable. But we should not forget the other simmers either and the title of my post says it quite clear: please make it optional till optimized

So it is not about taking it away from you, it is about leaving it for you but giving me the option to deactivate it for myself.

Of course this depends on the way it is technically done and if a separate on/off switch for the newly added METAR injection can be done.
This can only be answered by the devs, though, as the Wearher API is still not available.

1 Like

Solid points and fair enough.

you have that backwards… the forecast model is run by the supercomputers twice a day, so it can be up to 12 hours old. metars are usually less than an hour old.

(and im going by what one of the weather geeks here has told me, 12 hours – on the meteoblue site it says every 24 hours)

True, but the model is including a forecast to dynamically fill the gap between the updates and therefore gives a smoother transition than a static METAR that remains the same and then changes abruptly when updated.

but the resolution of the model is only 1 hour as well. so its still just smoothing between 1 hour points.

you can smooth metars as well.

Of course you can. That’s what you call a Model.
You Interpolate real World data and calculated the missing values in between.
Thats what I mean by “optimization”.

Right now, the sim does not smooth it properly

It doesn’t even load more than 1 METAR right now.
Best observed in area with high density of airports, lile eg. Los Angeles.
Fly over the City and observed the weather Co stolz changing instantly while flying through the METARs of different airports.

I appreciate the implementation of METARs and their will to improve it.
But for now, I would prefer to go without METAR for visibilty and clouds.
I dont mind temperature, pressure or Winds changinging spontanously, it is the visual weather poppig I dislike…