Winds aloft injection -- just an idea ...

[Hardy Heinlin]

Carl, I fear you still don't get it. Unfortunately, all my weather balloons have exploded and so I can't provide real aloft weather to you. And NOAA doesn't provide real aloft weather either.

So the real weather will be simulated by applying random variations to the received forecast data.

Only ATIS/METAR is real weather.

You can keep typing your wishes in big capitals; it won't change the world.

End of transmission. Roger over.

[cavaricooper]

HH-

Got it.

Unfortunately, my hope chest hasn't exploded, and perhaps one day there will be a way for PSX in much the same way as JP and Mark/Gary have done for the P3D/PSX combo (with AS).

In the meantime, Enrique's thoughts accompanied by your variability, do sound promising.

Roger Dodger, got my Vector Victor, over

C

[Hardy Heinlin]

No, you didn't. And you just can't stop repeating the same nonsense.

[cavaricooper]

Apologies, stopped.

C

[Jeroen Hoppenbrouwers]

So, just to get my facts straight. If this below is all correct, just reply "OK" or something.

1. METAR is near-actual, but always past, data on the ground. It can be extrapolated a bit up and in between airports but it is only relevant for, say, below 18,000 ft.

2. There is no actual winds/temps aloft data available anywhere, as the collected data points by weather balloon and aircraft ACARS are far between and hours old. This won't change any time soon. Nobody has this data, not the NOAA, not the KGB, not even Facebook.

3. So all available aloft data sources are forecasts, using advanced methods and as many real data points as they could get. But forecasts. Hazy crystal balls. If you inject 'live winds aloft' from these sources, you inject a guesstimation, not an observation.

4. Flight plans are built using these forecasts, and thus nearly always agree with the forecasts, as they are coming from the same hazy crystal balls.

5. Wind uploads come from the same forecasts.

6. Since neither forecast not flight plan nor FMC are locked to reality, surprise is around the corner in real-life and should be in simulated flight. The level of surprise may be selectable at the sim pilot's desire. Real-world pilots do not have such a surprise level selector, sadly.

7. It is as effective to inject the flight plan forecast into the PSX aloft surprise generator, as the NOAA aloft forecast, as they are the same. They are both slightly wrong. The real planet does what it wants and does not tell us.

8. Only when the world invents a truly realtime global winds/temps aloft detector/reporter, we need to revisit the PSX surpriser.


Hoppie

[torrence]

8. Only when the world invents a truly realtime global winds/temps aloft detector/reporter, we need to revisit the PSX surpriser.


Hoppie

Having tried to tackle the winds aloft issue from various angles since PS1 days, I agree with the analysis, Hoppie. 

I'd only add that even if you had 8 (in nerd speak - a global n-dimensional state vector for the global atmosphere) it wouldn't really solve the problem because of the 'butterfly' effect.  Weather on the scale of an aircraft is intrinsically chaotic - if you had the magic state vector and flew two aircraft through the same volume of atmosphere just minutes apart you'd still be likely to hear something like this on ATC: "Center, this is Airprobe 1 at FL320, did you get any turbulence reports from that Virgin flight ahead of us?  We just got some real jolts - have you got a smoother altitude for us?"

Cheers
Torrence

Added note:  The amazing thing actually is the forecasts are good enough that, if winds are the only issue, flights can usually arrive within minutes of scheduled time.

[Will]

To torrence's last point, the forecasts are usually good enough because the winds aloft are almost always coming from the west at 80 kts +/- 1-50 kts, and the temperature is almost always -55, +/- 1-10 degrees.

And this is the reason that the native PSX aloft environment often does so well when compared against performance planned by PFPX or other sources of real-world forecast data.

What really makes for significant changes is the jet stream, but the jet stream isn't everywhere. I've said this before, but if you just drag the jet stream sliders in PSX so that you get something that looks even a little like a current upper level prog chart, then a flight flown in the PSX atmosphere comes acceptably close to performance data predicted by PFPX. And has just the right amount of surprise.

I can see the utility, though, in having the forecast winds incarnated into the PSX environment, and that would be to test a PSX flight under completely controlled conditions. This could be useful and in some cases even fun, but that "surprise" factor that Jeroen eloquently describes would be missing, and so it would depart from reality. However, a simulator itself departs from reality, so there's not an a priori quibble there.

Unrelated thought: Making a weather randomizer can be simple, or complex.

Simple randomizer: every weather value varies randomly by somewhere between -25% and +25%.

Complex randomizer: the predicted front probably arrives, but it arrives randomly early or late, and is randomly stronger or weaker than forecast.

The complex randomizer approach probably needs supercomputer resources, as it is essentially a weather simulator.

[Dirk Schepmann]

So, just to get my facts straight. If this below is all correct, just reply "OK" or something.

2. There is no actual winds/temps aloft data available anywhere, as the collected data points by weather balloon and aircraft ACARS are far between and hours old. This won't change any time soon. Nobody has this data, not the NOAA, not the KGB, not even Facebook.

I think that this is the most important point - there are no real time upper wind data available, not even to the airlines (except of data transmitted with POS REPORTS). The weather baloons which collect real wind data in the upper atmosphere are launched only 2 times per day (at 00:00z and 12:00z) at approx. 800 representative positions around the globe.

The data collected by the balloons are then fed into the high performance computers which use these and other data for their forecast models (e.g. GFS, EZMF, UKMO, GEM to name a few). I assume that NOAA uses the GFS forecast data for their GRIB output. GFS creates a new model output every 6 hrs (00:00z, 6:00z, 12:00z and 18:00z). And these outputs are the basis of the flight planning.

So basically, the available upper wind data are never in real time and are always based on computer calculations. For short term prognosis the data are quite accurate. But if you compare the outputs of different models (e.g. GFS vs. EZMF) you can easily spot differences. Sometimes GFS is more accurate, sometimes EZMF. The reality is often in between.

I'm curious to see what Hardy comes up with. :-)

Best regards,
Dirk

[cagarini]

So, just to get my facts straight. If this below is all correct, just reply "OK" or something.

2. There is no actual winds/temps aloft data available anywhere, as the collected data points by weather balloon and aircraft ACARS are far between and hours old. This won't change any time soon. Nobody has this data, not the NOAA, not the KGB, not even Facebook.

I think that this is the most important point - there are no real time upper wind data available, not even to the airlines (except of data transmitted with POS REPORTS). The weather baloons which collect real wind data in the upper atmosphere are launched only 2 times per day (at 00:00z and 12:00z) at approx. 800 representative positions around the globe.

The data collected by the balloons are then fed into the high performance computers which use these and other data for their forecast models (e.g. GFS, EZMF, UKMO, GEM to name a few). I assume that NOAA uses the GFS forecast data for their GRIB output. GFS creates a new model output every 6 hrs (00:00z, 6:00z, 12:00z and 18:00z). And these outputs are the basis of the flight planning.

So basically, the available upper wind data are never in real time and are always based on computer calculations. For short term prognosis the data are quite accurate. But if you compare the outputs of different models (e.g. GFS vs. EZMF) you can easily spot differences. Sometimes GFS is more accurate, sometimes EZMF. The reality is often in between.

I'm curious to see what Hardy comes up with. :-)

Best regards,
Dirk

Don't know EZMF ?  We use the UKMO and ECMWF, both Global, the only two GLOBAL. Not that much of a difference for higher level winds and temps between the two.

Tropopause height is another important piece of data, but I don't know if HH is planning to use it for instance to also adapt PSX's Jetstream model accordingly "on-the-fly".

Regarding the approach followed by OFP generators, they should naturally adapt the data they use to the expected times at the successive waypoints of a given route. This is why, for instance, Active Sky (they have it from GRIB to the whole World) asks for a route to better set the winds and temps aloft because that way they can better adapt it to the route, and should the flight be a long one, interim downloads will occur with new fresh data.

I actually find Hardy's approach rather pragmatic, effective and easy to use for those who use PSX as it is supposed to be used, to simulate RL operations of such an airliner - not like I use it as a sim I load from time to time to play a few approaches under bad weather :-), but then again, METAR will be more than sufficient for my kind of playing ....

[Dirk Schepmann]

Hi jcomm,

EZMW is the German word for ECMWF.

All models are quite accurate for predictions in the range of 12-48 hrs. After 72 hrs it get's interesting but that's  not relevant for aviation.

However, I personally like to study the weather outputs for Germany to predict the weather. Right now we have very high temperatures and a draught in many parts of Europe. The forecasts of GFS and ECMWF are quite different for next week, but my bet is that ECMWF wins. We'll see next week.

But that's off-topic. Sorry for that.

[simonijs]

For the interested reader  , but you may want to wait for a rainy day:

http://www.wmo.int/pages/prog/www/GOS/ABO/AMDAR/resources/AMDAR_Programme_Development.html
https://pdfs.semanticscholar.org/2370/877464b5a608dac16f97488365d4fb52d01e.pdf

All about increasing forecast accuracy.

Regards,
Simon

[Markus Vitzethum]

Yet another approach to near-real-time wind data, as reported in this weeks edition of the german weekly newspaper DIE ZEIT: on August 21st 2018, ESA will launch the Aeolus spacecraft, which will be capable of measuring upper wind using a UV-laser powered LIDAR instrument (Aladin).

https://www.zeit.de/2018/31/satellit-aeolus-wetter-daten-messung-weltall-geschwindigkeit (in German)

According to the article it will measure up to 24 layers and will have max. 3h from data download to availabilty "on a public website" (whatever that means) via the weather center in Reading.

More from Airbus:
https://www.airbus.com/newsroom/press-releases/en/2018/06/Airbus-built-Aeolus-wind-sensor-satellite-ready-for-shipment.html

[United744]

I thought Hardy had a pretty good system with the jet stream model.

If you set the jet streams as accurately as you can based on current real-world data, then you take a chart showing upper level winds, PSX isn't too far off the mark (close enough that it doesn't matter).

Even if we were to get "real upper level winds" into the sim, would it still actually represent reality any more accurately than it already does? I doubt it.

I know many pilots use a "rule of thumb" for winds anyway, across the route. If they're doing trans-Atlantic stuff they'll use an average 50 kts wind (either head or tail depending on direction of flight) then check this against the flight plan. If they compute more fuel, they add the difference.

If you get "unusual" situations, such as a hurricane, or they're flying around the ITCZ (Inter-Tropical Convergence Zone), then again they'll do their own "back of the envelope" check and add the difference.

Once you're half-way into your 14 hour flight, not much you can do if the weather changes unexpectedly.

[Hardy Heinlin]

Hi, here's a new question.


Background:

An adjustable randomizer will generate a difference between the flight plan data and the actual data.
E.g. when 080°/50 kt, 10°C are set on the documents, the weather model may randomize this to,
for example, 071°/59 kt, 13°C.

A "randomize" slider will have a scale from "Off" (full left) to "Extreme" (full right).


Question:

When the slider is set to "Extreme", what should be the maximum allowable differences?


Suggestion:

Wind difference: Up to +/- 040°
Speed difference: Up to +/- 20 kt
OAT difference: Up to +/- 10°C


Regards,

|-|ardy

[G-CIVA]

Having looked at lots of real world OFPs & compared this data in real time to the out put data created by WX engines PSX users are using to inject WX into scenery generators I see the difference in the two is actually quite small - indicating that the designers of these WX engines are now getting their WX data from decent sources.

This is also true for the other WX data that forms the circle & which you also intend to utilise to create the upper wind/temp WX situation for PSX - from PFPX (or Garys flight planning software)

So what I am saying is that having studied this topic very closely now since the inception of PFPX the error margins between the real upper wind data - as in outside your front door, the simulated & injected wind data - as created by well known 3rd party WX engines & the planned WX data - as created by flight planning software tools like PFPX is now all very close to matching in detail.

Can I suggest the following:

Wind difference: Up to +/- 020°

Speed difference: Up to +/- 20 kt

OAT difference: Up to +/- 5°C - I have never seen a temp deviation greater than this across the various areas of data I have surveyed - Real World OFPs, WX engines, PFPX or the real WX on the planet

Any greater deviation has huge implications for flight planning & would be spotted during the planning phase.  To simply have a randomly generated number in excess of this would literally play havoc with the flight planning process.  Its one of the reasons why there are discreet Flight Planning Performance Cruise Tables for each aircraft at varying temperatures of deviation away from the ISA temperature at cruise flight levels

I think the slider is a great idea ... functional & simple & giving a really measured control over the random nature of the probabilities.

[Will]

Simple answer: Steve's suggestions are good.


More complicated answer: you can get much bigger deviations in wind speed and direction if the jet stream is farther north or farther south than predicted. But modeling this is more complicated, because it wouldn't be realistic to see a 50 kt deviation in one sector, followed by 0 kt deviation in the next sector, followed by a 45 kt deviation in the next, followed by a 7 kt deviation in the next.


If the deviations are strictly random, you might end up with a series of deviations that would be completely unrealistic. But if you could move the jet stream randomly +/- 150 miles farther south than predicted... then that might be both random and realistic at the same time. But that's obviously quite complicated.

[Jeroen Hoppenbrouwers]

I suppose that especially the jet stream would be detected quite quickly if it manages to escape forecast. The first few aircraft will yell bloody murder and then the rest can reroute?

Hoppie

[Hardy Heinlin]

I need to add that my question rather refers to real life operations. What are the most extreme differences re actual wind & OAT compared to flight plan data that a real crew experiences on typical flights every year?

I understand, if a PSX user works with real NOAA data on the plan and FSX weather add-ons in the model, the difference will be very low, and this technical difference by itself may already serve as a sufficient random effect. In this case, the user simply sets the slider to OFF.

The feature I want to implement is not restricted to the usage with NOAA forecast and FSX weather add-ons. The feature will be able to take any input from any source, and that may lead to near-zero differences. Even though such near-zero differences are unlikely, I want to assume that they are or will be possible, and hence I want to give the random slider more headroom for such rare cases. That's the reason for the slider's existence. Otherwise I could also replace the slider by an on/off checkbox for a fixed random value.

Please note that the randomizer will not set just one constant difference in all of the hundreds of flight plan variables. Each variable will be individually randomized. The slider just sets the maximum allowable difference.


Regards,

|-|ardy


P.S.: For the wind speed, the random value may be a percentage instead of a value in knots. That is, areas with higher wind speeds will be subject to higher differences (in knots). E.g. if 10% is set, a 20 kt speed may vary by 2 kt, and a 200 kt speed may vary by 20 kt. Therefore, jet stream areas will have a greater uncertainty knot-wise. The percentage cannot be applied to wind direction and OAT as those have to be linear; i.e. 300° is not more variable than 003°, and -50°C is not more variable than -10°C.

[Britjet]

No specific figures, I'm afraid, but perhaps to say that it would be unusual for the flight planned destination fuel predictions and the actual en-route predictions to be outside a tonne throughout the flight. It's normally very accurate.
Sorry I can't be more helpful.
Peter.

[G-CIVA]

perhaps to say that it would be unusual for the flight planned destination fuel predictions and the actual en-route predictions to be outside a tonne throughout the flight. It's normally very accurate.

Quoting Peters very real experience is invaluable here.

Using the metrics I have described in my earlier post & planning PSX flights with my PPFPX files I am able to achieve an accuracy close this tolerance with alarming regularity.