News:

Precision Simulator update 10.173 (24 February 2024) is now available.
Navburo update 13 (23 November 2022) is now available.
NG FMC and More is released.

Main Menu

SIGMET / AIRMET

Started by Hardy Heinlin, Mon, 10 Dec 2018 09:47

Jeroen Hoppenbrouwers

Abuse the typos. If your parser breaks on a SIGMET, that SIGMET becomes volcanic ash. Fun!

Hardy Heinlin

And the poor PSX users must suffer? :-)

No, erroneous SIGMETs are just skipped.



But, seriously, is it vodka or what when someone is mixing at random upper case "O" letters with zeros "0" in a sequence of lat/lon points? I mean, one even need to hold the SHIFT key to produce this nonsense!

Hardy Heinlin

Now with download system:

(Click on pic to enlarge)






|-|

cagarini

#23
I feel your pain Hardy...

Mine were written, long ago, using Lex & Yak + Good-old "C" on Vax-VMS, then latter Ultrix... :-)

It's a true PITA... Don't know how our Swedish colleagues work, but sometimes our aviation meteorologists are really in stress, during the whole 12hr shift they do, sometimes alone! It's not that difficult to do typos, I guess...

It's looking Astonishingly Beautiful!!!

Hardy Heinlin

I now understand the reason for that typo better: The guys probably always have the Caps Lock on, so no extra shift key press; and the O key is directly under the 0 key :-)

I implemented a filter that converts every O in an assumed lat/lon word to a 0.

A more serious problem is to assure whether an arc "minute" value actually indicates
minutes or decimal fractions of 1.
E.g. W07733 could mean 33 minutes or 0.33 degrees. In almost all cases it's minutes. Almost.

Jeroen Hoppenbrouwers

Quote from: Hardy Heinlin on Fri, 11 Jan 2019 00:11
But, seriously, is it vodka or what when someone is mixing at random upper case "O" letters with zeros "0" in a sequence of lat/lon points? I mean, one even need to hold the SHIFT key to produce this nonsense!
It's my wild, wild guess. You know I always first go to stupidity before I fall towards malicious intent, when I see something silly.

I bet this is an OCR job of the stupid kind. The capital O occurrences are next to the letter E. The system just didn't know for sure what to make of the circular shape and errored towards letters.

Why o why OCR is in the loop -- don't ask me.

Quote
I now understand the reason for that typo better: The guys probably always have the Caps Lock on, so no extra shift key press; and the O key is directly under the 0 key :-)

This is a lot more plausible in any case. Still -- I bet they don't even know how much trouble it will cause when doing this.


Hoppie

Hardy Heinlin

#26
Quote from: Jeroen Hoppenbrouwers on Sat, 12 Jan 2019 00:10
I bet this is an OCR job of the stupid kind. The capital O occurrences are next to the letter E. The system just didn't know for sure what to make of the circular shape and errored towards letters.

That reminds me of the Xerox scanner problems some years ago; the software used an image compression algorithm which sometimes replaced certain similar characters by wrong reference characters. It was used in scientific works, financial papers etc. Big effects, and nobody noticed it for years. It only happend in the highest compression mode, but this was the default mode that Xerox recommended.

http://www.dkriesel.com/en/blog/2013/0802_xerox-workcentres_are_switching_written_numbers_when_scanning

https://youtu.be/c0O6UXrOZJo?t=484

cagarini

I checked yesterday, and indeed the zones are hand written in the meteorological telecommunications system by my colleagues. They get the coordinates from the various model outputs, use the mouse to read the locations and then hand type them in the MTS... which is of course, just as with AIRMETs, GAMETs and TAFs prone to errors...

Filters are then used. Mine were written in C long ago, but after the various code changes and since I was already away from that area, the new filters are coded in Fortran 99 :-)

I'll try to check how our filters cope with examples like the one you posted above Hardy..  Maybe you can also start selling MTS filters in Java !!!!! :-)

Your work is progressing amazingly !!! Looking fwd into it.

Hardy Heinlin

Here's an example of a manually created, fictitious SIGMET that generates a small thunderstorm area. The area is defined by some lat/lon points, and the fuzzy details within the area are randomized.

(Click on pic to enlarge)








|-|

Gary Oliver

Gary approves of this :-)

G-CIVA

Steve Bell
aka The CC

Hardy Heinlin

By the way, when such smaller thunderstorm areas occur in the real world, and Active Sky downloads SIGMETs from the Internet, you can be pretty sure that the thunderstorm locations in PSX (which can download SIGMETs on its own) will be very close to the locations in Active Sky. The randomization of the cells within an area depends on the area size. Some areas are as small as Luxemburg, some as big as France ...


|-|

cagarini

#32
Unless it's a multi-cell system, it'll always be very random, no matter how small the SIGMET polygon is...

Yet, one of the  small scale models we use at IPMA can sometimes forecast the almost exact position of some thunderstorms. The IBM P8 also gets in a true thunderstorm when running it :-)

AS uses the very same sources PSX will use for SIGMET, AIRMET, GAMET, TAF and METAR. It complements it with some large scale forecast models, and, I believe, also with satellite image processing to gt better results regarding cloud coverage and types...

EDIT: One aspect I really like in the way PSX models TS systems is their dynamics... I sometimes seat there, either using "M" to freeze the aircraft in place, or by flying in circles and taking shots of the returns in the ND to observe as the cell changes position and shape with time.  ( although I believe using "M" doesn't have the very same effect ? )

Hardy Heinlin

Motion freeze just freezes the motion of the aircraft.

CB shapes on the radar also change by radar gain, tilt angle and scan altitude. Most PSX users probably always use the radar's AUTO function and don't notice what differences the tilt and gain can make.

CBs generated by the "local zone" features will move as usual. SIGMET generated CBs will not move. They will only move or disappear by SIGMET updates. PSX ignores the forecast parameters in the SIGMETs.


|-|

cagarini

Quote from: Hardy Heinlin on Sat, 19 Jan 2019 10:28
Motion freeze just freezes the motion of the aircraft.

CB shapes on the radar also change by radar gain, tilt angle and scan altitude. Most PSX users probably always use the radar's AUTO function and don't notice what differences the tilt and gain can make.

CBs generated by the "local zone" features will move as usual. SIGMET generated CBs will not move. They will only move or disappear by SIGMET updates. PSX ignores the forecast parameters in the SIGMETs.


|-|

Ok, looks acceptable to me ( regarding the SIGMET-induced TS ).

Yes I do play with tilt & gain, depending on weather conditions, and on if I'm climbing or descending...

Nice / easy read here: http://code7700.com/radar_gain.htm

Will

Will /Chicago /USA

Holger Wende

Quote from: jcomm on Sat, 19 Jan 2019 09:14
...One aspect I really like in the way PSX models TS systems is their dynamics... I sometimes seat there, either using "M" to freeze the aircraft in place, or by flying in circles ...

...but sometime I hate this dynamics: Quite often, when I approach an airport and there is a likelyhood of TS I have the suspicion, that the PSX dynamics "knows" where I want to go and puts a TS cell exactly such, that is moves right into my approach path. Evil  :-[
Then I fly holdings with either inbound or outbound track towards my preferred runway, such that I can monitor the weather cell, waiting to move away from the field and hopefully not moving towards my holding position.

Regards, Holger

Hardy Heinlin

Speaking of OCR again ...

Some SIGMETs are even so bizarre that they randomly pick some lat/lon points within a sequence of boundary points, and swap these points within this sequence so that the boundary shape becomes a random zigzag shape.

How can this be? How can this be that an OCR picks single words of a sequence of words and swaps them? How be this can? Be how can this? This how be can?

Even more bizarre: Skyvector displays the published incorrect text, yet displays the boundary shape correctly on the Skyvector map. Be it could, it be could, could it be that there's a human at Skyvector that checks the parsed SIGMET geometry manually from time to time?


|-|ardy

Dirk Schepmann

Not really. I found the following statement on the Skyvector forums:

QuoteWe attempt to parse the shape from the Sigmet text. It is very prone to errors. Often, our parser gives up and simply uses the FIR shape as the sigmet shape. If you see a parse error, please bring it to our attention and we'll see if we can make it better.

Maybe this is the secret? 😊

Best regards,
Dirk

Hardy Heinlin

I didn't know this quote, but, indeed, I noticed on their maps that they just use the entire FIR if the SIGMET contains pure nonsense.

However, their zigzag corrections refer to the points in the SIGMET texts; in that case they don't draw the FIR. There must be a huge correction algorithm that checks for the most reasonable sequence. It's relatively easy to detect a self-crossing shape. But it's not easy to find out which point is the wrong one, especially if there's no self-crossing at all in the zizag line. Some areas intentionally include angles greater than 90°.


Cheers,

|-|ardy