AUTO METARs LHR

[gaffer]

Any one using the AUTO METARs from EGLL ? I'd be interested in their views as to their accuracy.

[Cough]

Fine for most stuff, but the cloud entries are gibberish.

The system can't seem integrate CB's and TCU's into a proper METAR, so you end up with SCT012 BKN024 //////CB or the like.

I've also seen it declare //////CB on a sky clear day. I've also (more worryingly) seen no TCU or CB in the METAR, when there were.

All the rest, fine. Cloud types, rubbish.

[T250]

Not great.

Poor vis - at least 1-2km out on each report
Poor cloud - CBs reported when there aren't any, no TCUs when there are
Poor weather - moderate or heavy precipitation reported, when light at best

Overall either exaggerates or underestimates conditions.

Amongst my airline we tend to make our own observations ourselves, and even our layman unaccredited met skills indicate the extent of the errors!

Seems a dangerous precedent to set, will it be going back to human method again in future? Far more reliable and trusted by our crews.

[BigDaddyBoxMeal]

One assumes the system is as accurate(!) as the same system used at a host of UK airfields... Edinburgh, Glasgow, Bristol, Cardiff, Manchester, Luton, Stansted.... There's a common factor

[eastern wiseguy]

Being able to report cloud types? Is that a new piece of software.

Certainly the kit I was using at EGAA three years ago could only report coverage and not type.

[250 kts]

Seems a dangerous precedent to set, will it be going back to human method again in future? Far more reliable and trusted by our crews.

I think not. I would urge any crews at any airfield who are noticing a reduction in the accuracy of the auto met reports to MOR it immediately. This is a clear example of a large UK ANSP trying to cut costs at the possible expense of safety. I know it's paperwork on your part but we need evidence from crews to back up our case that the system is not fit for purpose.

[HEATHROW DIRECTOR]

The rot started many years ago when they closed met offices at larger airfields and gave the observing task to ATC staff!

[ZOOKER]

When I was at EGNX, the SATCO was intent of transferring the Wx observing task from the ATCOs to the ATCAs, (mainly from a safety point of view), which he did after I left in 1976.
What is also interesting is that all the airport-based met offices were synoptic observing stations, and provided comprehensive observations of many other parameters, as well as those included in the METARS.
It's odd that with the present uncertainties over changing Wx and climate, the scientific community no longer have this kind of detailed data to work with. The cloud genera were quite useful when they were in the METARS too, they gave a rough guide to the physical processes taking place aloft. 'SCT, BKN' etc doesn't tell you much.
Just out of interest, how much met' is included in the NATS ATCO training course these days?

[good egg]

Originally Posted by T250

Not great.

Poor vis - at least 1-2km out on each report
Poor cloud - CBs reported when there aren't any, no TCUs when there are
Poor weather - moderate or heavy precipitation reported, when light at best

Overall either exaggerates or underestimates conditions.

Not that I want to decry your comments but it certainly doesn't surprise me that vis is "1-2km out on each report". Any "observation" is an assessment at a particular point in time. These are done at 30min intervals and only updated at intermediate times if there is a significant change, e.g. if the vis changes from one specific range to another. In this respect auto "observations" use the same parameters as human obs. In other words a METAR could be 29mins "out-of-date". Be it 1 min or 29mins the weather will always be changing but can't necessarily be updated using the reporting parameters.
I can't count the number of times I've had a different opinion from a colleague (both equally trained) on the weather obs. Nor could I count the number of missed SPECI reports by human method.
At least the auto method uses hard data from a specific measurement method (assuming all the sensors are correct! ) and so isn't subject to the vagaries of humans.
I'm guessing the argument is that at the moment it is sufficiently close to a human obs and is therefore suitable for the task. As technology improves and algorithms are refined then auto is likely to provide a more consistently accurate obs than a human.

[T250]

Not that I want to decry your comments but it certainly doesn't surprise me that vis is "1-2km out on each report". Any "observation" is an assessment at a particular point in time. These are done at 30min intervals and only updated at intermediate times if there is a significant change, e.g. if the vis changes from one specific range to another. In this respect auto "observations" use the same parameters as human obs. In other words a METAR could be 29mins "out-of-date". Be it 1 min or 29mins the weather will always be changing but can't necessarily be updated using the reporting parameters.
I can't count the number of times I've had a different opinion from a colleague (both equally trained) on the weather obs. Nor could I count the number of missed SPECI reports by human method.
At least the auto method uses hard data from a specific measurement method (assuming all the sensors are correct! ) and so isn't subject to the vagaries of humans.
I'm guessing the argument is that at the moment it is sufficiently close to a human obs and is therefore suitable for the task. As technology improves and algorithms are refined then auto is likely to provide a more consistently accurate obs than a human.

I understand your point, however my examples were based upon a day where there were constant CBs and TCUs throughout an afternoon and these were never promulgated on the METAR. This is at best an error and at worst a case to simply save money at expense of quality and safe service to your customers.

Not all parts of a METAR can (or in cause of this auto rubbish, should) be automatically formulated, despite your comment that auto method uses specific measurements. How does a computer or other equipment try to justify if there is a CB or TCU? What exact measurement can be used for that, only the naked human eye can really decipher and then identify such phenomena.

[good egg]

I'll cede that point to you as I think that has a greater impact safety-wise than vis (esp. as the airports we're talking about have IRVR). Fortunately I'm at a unit where human input is still required

Don't know how a TS would be sensed either.
If I'm unsure whether it's rain or a shower of rain (judging by the cloud) I tend to look at the TAF to see if I should input RA or SHRA. If there's no mention of rain on the TAF it's more likely to be a shower (as I understand it steady rain from stratus cloud is relatively easy to predict whereas whether it will rain from cumulus is a tad trickier). I can't imagine that's too tricky to build into the algorithm!
CBs & TCUs a tad trickier I guess but surely in this day & age it should be possible to determine by sensors...

[Cough]

Good egg.

I'll cede the point that human observations are prone to 'variation'.

However, on a blue sky day, you are unlikely to send a METAR with /////CB out on it.

But thats what we see! When they are there, they aren't (according to METAR)

Problem is, CB's really mess with the LHR landing rate and when you can't trust whats happening on the information flow, then the decision making process is flawed...

On balance, I'll take 'variation'. Thats because I trust you...

[good egg]

Cough

Yes, even my observations are not that inept!

Originally Posted by 250 kts

This is a clear example of a large UK ANSP trying to cut costs at the possible expense of safety.

The above quote is an interesting point and, incidentally, is not restricted to just the UK.
Where do you think the pressure to cut ANSP costs comes from?
In the UK there is an open, competitive market for ATS provision at airports and hence it'd be simple to assume that the ANSP looks to reduce costs as far as practicable in order to maintain/gain a contract - This is true regardless of the ANSP.
The pressure to reduce costs doesn't come solely from the need to maintain/gain contracts though. Airport Authorities want low costs. Airlines want low airport charges. Travellers want low ticket prices. This is true regardless of whether it is an open or closed market.

There is an inevitable march towards automatic reporting. On a plus note, having had a quick google (other search engines available), there appears to be a lot of research into improving the quality/accuracy of auto METAR. It seems though that in this case it is perhaps happening, according to the inaccuracies mentioned by others on here, after the horse has bolted?

[fisbangwollop]

In Auto reports when cloud is detected, three strokes (///) are placed at the end of each cloud layer group, to indicate that the Auto Weather Station could not identify TCU or CB. i.e. OVC019/// However, CB or TCU will be included if a forecaster at the met office is able to observe CB or TCU by reference to the additional observational data available to them i.e. weather radar etc. In that instance you would see OVC019/// //////TCU This would indicate that Cloud is Overcast at 1900ft cloud type not determined but Towering Cumulus has been detected in the area.

[good egg]

fisbangwollop

That's good to know. I was a little confused as the only time I've ever input seen "//" (on a TAF admittedly) is when appended to "VV" to indicate vertical visibility not determined (Sky obscured).

That's actually quite a handy feature of auto reporting as there are plenty of occasions when from the ground it's difficult to determine if CB or TCU is present because the tops of the clouds are not visible. (By referring to the forecast - which is based on the additional data available to the met office, weather radar etc, the auto system is then doing a similar check to what I do as a human observer.)

That, combined with on-board weather radar, maybe indicates that the horse hasn't bolted after all!

[TelsBoy]

All semi- or fully-automated Met Systems have the same problem, and that is that they rely on what the Wx sensors can actually see. Smaller airfields have only one Met Garden with one set of sensors - cloudbase recorder, visiometer, Temp/RH sensors, present Wx sensor - larger airfields have two or even three. However a CBR only points straight up and if the laser hits a cloud straight overhead at 11000' it won't register the stratus sitting over the other end of the runway at 100'. Similarly if the Visiometer works out the vis based on what it can see at the Met Garden as 10K+ it won't see the bank of fog on the other side of the airfield.

Larger airports with multiple Met Gardens have several sources of info making this less of a problem, but the problem is still there.

Someone made a point of ANSPs cutting costs and impact of safety - sadly ANSPs (especially one in particular in the UK) are under severe pressure to do so by profit-hungry low-cost airlines and the politicians they have complained to, yet the customers still demand top-notch ATS and ATE infrastructure which does not come at a cheap price.

[ShyTorque]

Hopefully all this will be sorted out just in time for fully automated, pilotless aircraft. Then there will be no need for METARS at all because nothing can go wrong.....go wrong...go wrong.

[T250]

...Until the automated pilotless systems get hacked into