I understand your position and I respect your ability to create a fantastic product. I have been a fan of your work for FSX and P3D for ages. Please understand that when I say I’m not attacking you or your product. I am only inclined to disagree with your statement that the ONLY ground truth is the METAR report, especially because of their inability to report transient weather, and the fact that this is the only input to local terminal weather is the primary reason I will not buy REX Weather Force in its current state of development if that’s the engine’s purpose.
In the screenshot below, although Frankfort is close to the center of that area of precip, the report from half an hour prior doesn’t report any precipitation, and this kind of thing has caused Active Sky in FSX to generate, you guessed it, Ceiling Overcast 10,000, no precipitation:
Similarly, while a METAR reports a snapshot of weather, it can be incredibly outdated by the time the aircraft is actually there. While an ATIS reports a certain altimeter setting, a controller may provide an altimeter setting up to 0.04" different (a personal experience flying GA aircraft in Southern Ontario on a particularly windy day when pressure was dropping rapidly). Or when winds calm down as the sun sets, the wind check on final is usually far calmer than ATIS reports (they update with a SPEC METAR when wind increases, but why would they bother doing it when wind decreases?).
Some convective thunderstorms around Minneapolis, which may affect routing in the vicinity of the terminal area. Once again, METARs do not capture weather to this level of fidelity:
Some weather I found in Northern Australia: once again, no METAR nearby to capture the presence of these isolated thunderstorm cells (which may otherwise provide an interesting challenge enroute):
Then there’s obvious issues:
- METARs can’t report cirrus clouds, you need weather forecasts to be able to predict that
- METARs can’t report haze at altitudes, you need weather forecasts to be able to predict that (hell I’m looking out my window at home and seeing a hazy and dull afternoon with the sun barely shining through, yet my local airport METAR is reporting clear skies)
- METARs can’t give an idea of the type of cloud in a certain layer, approximate size, vertical development, or convection intensity (you need forecast data for that).
- In certain countries, METARs don’t even report clouds above a certain altitude that’s absurdly low (something like 7000’).
- METARs can’t provide an idea of an overcast cloud shelf: the type to exist from coastal weather (typically seen when arriving into the Netherlands from over the North Sea) or from a moving front, or even the horizontal movement of stratus clouds I’ve seen so often when flying in Southern Ontario in the fall and spring.
- Automated METARs can generate erroneous readings if equipment were to fail (happens more often than you’d think), and injecting the weather blindly into the sim creates a false result (example below, find the station with a faulty automatic visibility reporting system):
METARs simply do not provide spatial or temporal fidelity, and if there is no METAR present there’s no weather to provide. However, I do agree that using forecast to inject weather is wrong as well (as we see from Asobo’s implementation as it stands today). A next gen weather engine would require a combination method to:
- Use forecast data to determine local weather phenomena and expected weather reports to generate a fluid and continuous picture of global weather
- Use real-time weather data (such as METARs, weather stations, radar where available, and satellite where available) in order to adjust forecast data as needed and correct local deviations
- Be able to calculate expected temporal change to account for the fact that weather reports are indeed outdated, and are outdated the second they’re issued. This is a matter of fact whether one likes it or not
As for computing power, although it would require a lot of power (and that I agree with), I disagree with it being impossible: These calculations are already being done numerous times a day by supercomputers across the world (which is the most computationally intensive part of determining weather), Meteoblue simply streams the calculated data to local MSFS clients, which then insert the weather accordingly. To then take the forecast model’s output and correct it to match real world data is an afterthought in comparison, and adjusting for satellite and radar input should be a no brainer in comparison.
Man, that was a lot of typing…