Gfs And Different Input

[johnholmes]

hi all

Sometime ago I posted, well a very long time ago, and since then people like Steve Muir and others have posted about the same.

Namely the different data, over and above the basic observational data in each run, that goes into

00z

06z

12z

18z

For the life of me I cannot find it anywhere in the reams of files I have on the weather nor can I find it on the forum, although I know it was mentioned only a few days ago.

If anyone has access to it, I would be very grateful if someone could, either, post the information, or point me in the right direction.

many thanks

John

[Shiny_Bottom_1]

Is THIS what you were looking for John?

[MAF]

john thought it might be this :

00z - Weather buoy, satellite data, shipping data, country data, NOAA data

06Z - Weather buoy, satellite data, shipping data

12Z - Shipping data, Satellite data ONLY

18Z - Weather buoy, satellite data, shipping data, country data, NOAA data

argghhh sorry SR didn't realise you had posted whilst i was searching

Edited December 5, 2005 by sesnow1

[johnholmes]

thanks both of you that is it, I'll now try and put it in the Guides so its always available.

thanks again

John

[Thundersquall]

john thought it might be this :

00z - Weather buoy, satellite data, shipping data, country data, NOAA data

06Z - Weather buoy, satellite data, shipping data

12Z - Shipping data, Satellite data ONLY

18Z - Weather buoy, satellite data, shipping data, country data, NOAA data

argghhh sorry SR didn't realise you had posted whilst i was searching

The 00z and 18z - "too many data inputs spoil the broth"?

[Paul]

Not at all, you can't feed too much data into a numerical model..

[The PIT]

Interesting that the 06z unless my memory is fading picked up the Easterly very well.

Pity they all can't have the same data input.

[johnholmes]

Interesting that the 06z unless my memory is fading picked up the Easterly very well.

Pity they all can't have the same data input.

so did the 12Z

[AtlanticFlamethrower]

Not at all, you can't feed too much data into a numerical model..

You make an interesting statement but I'd have to agree to disagree with you. With more data you've potentially stronger agreement at longer range but conversely if your methodology (the way the data is collected and put into the model) is not watertight more errors creep in to the model and make that extra data a hindrance rather than a help. You can overload a shopping trolly if you see what I mean.

This might be why the 12z is best regarded on this site - satellite data is probably the most accurate source of weather info we have, the major sources of error on thie 12z run presumably confined to shipping data which involves human readings and presumably ever changing positions of the ships. You can't data cut indefinately without losing details and long term accuracy but maybe if shipping data was removed from the 12z the trends would not be much less easier to spot.

[Paul]

To be honest the list above is much simplified when compared to what actually goes into each run - I'll try and get something put together to show you..

[Paul]

As promised - a more comprehensive look at what data goes into the GFS. All of these types of data from around the entire globe are input into the model on each run. There are obviously different availabilities of some types of data (like for instance flight data will be more available during daylight hours as flights occur more often at that time).

But bear in mind that as the data is worldwide - there will be more data available from different parts of the globe at different times. The net result is probably that each model run has a similar amount of raw data fed into it.

As you can see there is a fair bit of raw data used...

Surface data – Land

Synoptic - restricted (WMO Resolution 40) manual and automatic

Synoptic - fixed manual and automatic

Synoptic - mobile manual and automatic

Aviation – METAR

Vertical soundings – other than Satellite

Rawinsonde - fixed land

Rawinsonde - mobile land

Rawinsonde – ship

Dropwinsonde

Pibal

NEXRAD Vertical Azimuth Display (VAD) winds

Single Level Upper-air data (other than satellite)

Flight level reconnaissance aircraft data

Automated MDCRS ACARS aircraft data (from ARINC)

Automated MDCRS ACARS aircraft data (from ARINC) via AFWA

Manual AIREP format aircraft data

Manual PIREP format aircraft data

Automated AMDAR format ASDAR/ACARS aircraft data

European ASDAR/ACARS Data Aquisition System (E-ADAS) aircraft data (originally in BUFR)

NOAA Profiler Network (NPN) winds

Profiler winds originating from PILOT (Pibal) bulletins

Japanese Meteorological Agency (JMA) profiler winds

Cooperative Agency Profiler (CAP) winds

RASS temperatures from NOAA Profiler Network (NPN) and Cooperative Agency Profilers (CAP)

Vertical soundings – satellite

GOES/NESDIS-processed 5x5 field-of-view soundings/brightness temperatures

GOES/NESDIS-processed 1x1 field-of-view soundings/brightness temperatures

Surface data – satellite

QuikSCAT products

DMSP/SSM-I - operational products derived at FNMOC

DMSP/SSM-I - Neural Net-3 products derived at NCEP

DMSP/SSM-I - processed brightness temperatures

NASA/Tropical Rainfall Measuring Mission (TRMM)/TMI)

Single level upper air data – satellite

GOES/NESDIS infrared derived cloud motion (from NESDIS server, originally in BUFR)

GOES/NESDIS water vapor imager derived cloud motion (from NESDIS server, originally in BUFR)

GOES/NESDIS visible derived cloud motion (from NESDIS server, originally in BUFR)

INSAT/India infrared derived cloud motion

INSAT/India visible derived cloud motion

INSAT/India water vapor derived cloud motion

GMS/JMA infrared derived cloud motion - low density (originally in SATOB format)

GMS/JMA visible derived cloud motion - low density (originally in SATOB format)

GMS/JMA water vapor imager derived cloud motion - low density (originally in SATOB format)

METEOSAT/EUMETSAT infrared derived cloud motion (originally in BUFR)

AQUA/TERRA MODIS infrared derived cloud motion

AQUA/TERRA MODIS water vapor imager derived cloud motion

Surface data – sea

Sea level pressure bogus

Ship - manual and automatic

Buoys in FM-18 format (moored or drifting)

Buoys in FM-13 format (moored)

Land-based CMAN station

Tide gauge reports in CREX format

Radiances – satellite measured

AQUA/AIRS, AMSU-A, HSB processed brightness temperatures

GOES NESDIS-processed 11x17 field-of-view imager (clear sky) brightness temperatures

NOAA-14/HIRS-2 (High resolution InfraRed Sounder-2) NCEP-processed brightness temperatures

NOAA-14/MSU (Microwave Sounding Unit) NCEP-processed brightness temperatures

NOAA-15,-16,-17,-18/AMSU-A (Advanced Microwave Sounding Unit-A) NCEP-processed brightness temperatures

NOAA-15,-16,-17/AMSU-B (Advanced Microwave Sounding Unit- B) NCEP-processed brightness temperatures

NOAA-15,-16,-17/HIRS-3 (High resolution InfraRed Sounder-3) NCEP-processed brightness temperatures

Physical/chemical constituents

POES/NESDIS Solar backscatter ultraviolet radiances

Edited January 2, 2006 by Paul

[Stoxs]

with all this data why is it ever wrong .

I know some one asked this quetion a few days ago but does it teach it`s self with possible trends or is it based on what it was programed will happen?

[AtlanticFlamethrower]

Thanks for that, Paul. Peak behind the green curtain always interesting.

Do you have any more details on each run? Or can you confirm it's as sesnow1 listed above.

[Paul]

with all this data why is it ever wrong .

I know some one asked this quetion a few days ago but does it teach it`s self with possible trends or is it based on what it was programed will happen?

I'm not sure how the model itself works - it's extremely complex!

Thanks for that, Paul. Peak behind the green curtain always interesting.

Do you have any more details on each run? Or can you confirm it's as sesnow1 listed above.

As I understand it all that data goes into each run, some runs will have more from one data source than others due to the availability of the data but I'm not sure about the list above..

[ChartViewer]

Hi AFT,

This was part of an email I had in response to my questions re the GFS. To be honest

I suspect as Paul quite rightly states above most of the data would try to be incorporated in each run.

The NOAA email makes it even more confusing though...sorry! <_<

'Model runs are generally made four times a day at 00Z (7:00 pm EST),

06Z (1:00 am EST), 12Z (7:00 am EST) and 18Z (1:00 pm EST). Since

radiosonde observation balloons are released throughout the world at

00Z and 12Z, the models are "initialized" at these times with real

data for subsequent runs. When there is little or no data, for example

for the "off hour" 06Z and 18Z runs, the models are initialized from

previous model run results. '

PS - It was New Years Eve when I asked the question though, so maybe he thought I was

strange asking that question around 5pm, so he could have been waiting to go home for a drink! :lol:

Kind regards

CV

Edited January 2, 2006 by ChartViewer

[Paul]

As I understand, basically how it works is that all of the data sources and types I listed above go into each 'data dump' before the model is run. If for example there is no new radiosonde data available for a run then old data will be used. Within each data type though there can be a varying amount of data available - so for instance there is probably more flight data from the atlantic within the 12z and 18z runs as there are more flights across the atlantic at those times.

Edited January 2, 2006 by Paul

[ChartViewer]

As I understand, basically how it works is that all of the data sources and types I listed above go into each 'data dump' before the model is run. If for example there is no new radiosonde data available for a run then old data will be used. Within each data type though there can be a varying amount of data available - so for instance there is probably more flight data from the atlantic within the 12z and 18z runs as there are more flights across the atlantic at those times.

Hi Paul,

I think that has pretty much nailed it <_<

The interesting thing to arise from this discussion is IF we find all runs are using the same data

where possible, it would be correct to compare the 12z to the 18z, etc yes?

Kind regards

Rob

Edited January 2, 2006 by ChartViewer

[Paul]

To be honest, I never really thought there was an issue with comparing runs - ok there will always some differences in the data, but I don't think that it will ever be different enough to say that comparing run for run won't work. I've never seen anything from the NCEP to suggest that each run cannot be followed in turn either <_<

[ChartViewer]

To be honest, I never really thought there was an issue with comparing runs - ok there will always some differences in the data, but I don't think that it will ever be different enough to say that comparing run for run won't work. I've never seen anything from the NCEP to suggest that each run cannot be followed in turn either <_<

Thanks Paul,

Seemed to be a big issue when looking at the Models as most users were under the Impression you should only

compare an 18z with its predecessor.

I think this has answered most of the questions raging in the Model runs.

I for one shall now start to use a run from the previous Model shown....i.e. 12z to 18z

Kind regards

Rob

[Methuselah]

with all this data why is it ever wrong .

I know some one asked this quetion a few days ago but does it teach it`s self with possible trends or is it based on what it was programed will happen?

It being only a MODEL, I can imagine a few inherent flaws:

It would need infinite computing capacity, and infinite AND continuous data-input for it to be a perfect replica of the real thing...Like all models, it's a highly simplified version of reality

<_< :lol:

[ChartViewer]

It being only a MODEL, I can imagine a few inherent flaws:

It would need infinite computing capacity, and infinite AND continuous data-input for it to be a perfect replica of the real thing...Like all models, it's a highly simplified version of reality

<_< :lol:

Yes and computer Models are always likely to receive corrupt data at some point Peter. Which means

unless they are updating in real time they are always open to interpretation and flaws.

Cant beat the real thing

Cheers

Rob

[Paul]

Totally agree Pete - even with the huge supercomputers and incredible scientists and modellers working on the software, a model can (at this point in time) never be able to match nature!

[johnholmes]

Thanks Paul,

Seemed to be a big issue when looking at the Models as most users were under the Impression you should only

compare an 18z with its predecessor.

I think this has answered most of the questions raging in the Model runs.

I for one shall now start to use a run from the previous Model shown....i.e. 12z to 18z

Kind regards

Rob

I think you will get extremely frustrated doing that mate but its up to you.

regards

John

To be honest, I never really thought there was an issue with comparing runs - ok there will always some differences in the data, but I don't think that it will ever be different enough to say that comparing run for run won't work. I've never seen anything from the NCEP to suggest that each run cannot be followed in turn either <_<

hi

I am no expert in this business but experience with GFS over 14-15 months does, for whatever reason, suggest that comparing successive runs with one another is pretty fruitless, at least beyond a' near' time scale, say 72-96 hours possibly 120 hours.

Like I say I have no real knowledge but 12-15 months with GFS has quite often shown this up.

regards

John

[ChartViewer]

I think you will get extremely frustrated doing that mate but its up to you.

regards

John

Hi John,

I am frustrated with the data full stop. You know that more than most <_<

I think I will give up and just watch events unfold :lol:

You are right though on a more technical basis running a 5 day 12z v 12z trend must be more

precise than just a 5 run trend.

Great discussion though

Kind regards

Rob

[johnholmes]

hi

It would help if NOAA had a dedicated line for amateurs like ourselves to get relatively simple answers from them. Either they ignore e mails or give such a complex answer that its almost as bad as no answer.

I do know, in the past at the Met Office, that each run had the same basic surface synoptic data fed into it, the upper air when that was most often available, and other(satellites, aircraft, buoys etc) when available.

regards

John