Accuracy of Grib weather data

[macd]

I thought in the UK they were all the same forecast now. NOAA dish it out free from an ftp address every 6 hours?

No. Data comes from heaps of sources, but is fed into different predictive models. See http://passageweather.com/ for a brief explanation. For several areas covered by Passageweather, you have a choice of up to three forecast models. Often they will substantially agree with each other, but sometimes not. Old varnish advocated using gribs over a period of time to become familiar with their foibles; the same is true of using these different models.

 

[charles_reed]

GRIBs are not forecasts in the true sense of the word and anyone relying entirely on them is being naive.
They take no account of local topographic effects, and are frequently well out on timings of changes.
Having said that I use them in preference to any forecasting sites as the raw material to develop my own forecasts.

Forecasting local weather in Atlantic waters is pretty easy, but when you get to enclosed "lakes" like the Med it's an entirely different matter.

 

[ytd]

I used grib for a 1100nm trip down the east coast of Australia last november and they were consistently better on windspeed and direction than the official forecasts. Only problem is that you need and internet connection and when out of mobile phone range its back to the official forecasts.

 

[macd]

GRIBs are not forecasts in the true sense of the word and anyone relying entirely on them is being naive.
They take no account of local topographic effects, and are frequently well out on timings of changes.
Having said that I use them in preference to any forecasting sites as the raw material to develop my own forecasts.

Forecasting local weather in Atlantic waters is pretty easy, but when you get to enclosed "lakes" like the Med it's an entirely different matter.

I'd dispute your definition. A forecast is a forecast whether by a meteorologist, computer...or 'reading' a pine cone or granny's lumbago, come to that.
But of course you're right that anyone expecting accuracy on a micro level is in for a disappointment, since the resolution is typically around 60 miles. That said, certain macro effects are modelled in: the Straights of Gib and Bonifacio, for instance (which isn't the same as saying they always get it right).

In the open Atlantic I've had gribs prove spot-on for a full seven days. Good when it happens, if not to be relied on.

 

[Talbot]

I used grib for a 1100nm trip down the east coast of Australia last november and they were consistently better on windspeed and direction than the official forecasts. Only problem is that you need and internet connection and when out of mobile phone range its back to the official forecasts.

Perhaps you should have bought an SSB receiver for weather fax.

CHARLEVILLE, AUSTRALIA
CALL SIGNS FREQUENCIES TIMES EMISSION POWER
VMC 2628 kHz 0900-1900 F3C 1 KW
VMC 5100 kHz All Broadcast Times F3C 1 KW
VMC 11030 kHz All Broadcast Times F3C 1 KW
VMC 13920 kHz All Broadcast Times F3C 1 KW
VMC 20469 kHz 1900-0900 F3C 1 KW

WILUNA, AUSTRALIA
CALL SIGN FREQUENCIES TIMES EMISSION POWER
VMW 5755 kHz 1100-2100 F3C 1 KW
VMW 7535 kHz All Broadcast Times F3C 1 KW
VMW 10555 kHz All Broadcast Times F3C 1 KW
VMW 15615 kHz All Broadcast Times F3C 1 KW
VMW 18060 kHz 2100-1100 F3C 1 KW

TIME CONTENTS OF TRANSMISSION RPM/IOC VALID MAP
TIME AREA
------/1200 Australian MSLP Prog (H+36) 120/576 1200 AUST
0015/1215 VMC/VMW Schedule Page 1 of 2 120/576
0030/1230 VMC/VMW Schedule Page 2 of 2 120/576
0045/------- VMC/VMW Information Notice 120/576
0100/------- IPS Recommended Frequencies for VMC (Charleville)) 120/576
0130/------- IPS RECOMMENDED FREQUENCIES FOR VMW 120/576
------/1245 Indian Ocean MSLP Prog (H+36) 120/576 1200 IO
------/1315 South Pacific Ocean Total Waves (H+48) 120/576 0000 SWP
------/1330 Indian Ocean Total Waves (H+48) 120/576 0000 IO
------/1345 Pacific Ocean Sea Surface Temps (Weekly) 120/576 LATEST SWP
------/1400 Indian Ocean Sea Surface Temps (Weekly) 120/576 LATEST IO
0200/------- Australian MSLP Prog (H+36) 120/576 0000 AUST
------/1415 Casey Eastern and Western High Seas (H+48) 120/576 0000
0245/1430 Australian MSLP Anal (Manual) 120/576 00/12 AUST
0300/1500 Australian 500 hPa Anal 120/576 00/12 AUST
0315/------ Voice Broadcast Information for VMW (Wiluna) 120/576
------/1515 Australian MSLP Prog (H+36) 120/576 1200 AUST
0400/------ Australian 500 hPa (H+24) Prog 120/576 0000 AUST
0430/1530 Australian MSLP 4-day forecast, Days 1 and 2 120/576
0445/1545 Australian MSLP 4-day forecast, Days 3 and 4 120/576
------/1600 Australian 500 hPa (H+24) Prog 120/576 1200 AUST
------/1630 IPS Recommended Frequencies for VMC (Charleville) 120/576
------/1700 IPS Recommended Frequencies for VMW (Wiluna) 120/576
0600/1800 Asian (Part A) Gradient Level Wind Anal (Manual) 120/576 00/12 A
0623/1823 Asian (Part B) Gradient Level Wind Anal (Manual) 120/576 00/12 B
0645/------- Asian MSLP Anal (Manual) 120/576 0000 C
0730/1915 Indian Ocean MSLP Anal (Manual) 120/576 00/12 IO
0745/1930 Australian Wind Waves Ht(m) Prog 120/576 00/12 AUST
0800/1945 Australian Swell Waves Ht(m) Prog (H+24) 120/576 00/12 AUST
0830/------- South Pacific Ocean MSLP Anal 120/576 0000 SWP
0845/------- Australian MSLP Anal (Manual) 120/576 0600 AUST
0900/------- Australian MSLP Prog (H+36) (Repeat) 120/576 0000 AUST
0915/------- Australian MSLP 4-day forecast, Days 1 and 2 (Repeat) 120/576
0930/------- Australian MSLP 4-day forecast, Days 3 and 4 (Repeat) 120/576
------/2000 South Pacific Ocean MSLP Anal (Manual) 120/576 1200 SWP
------/2015 Casey Eastern and Western High Seas (H+24) 120/576 1200
------/2030 Australian MSLP Anal (Manual) 120/576 1800 AUST
1015/------ Casey Eastern and Western High Seas (H+24) 120/576 0000
------/2215 Casey Eastern and Western High Seas (H+36) 120/576 1200
1030/2230 S.H. 500 hPa Prog (H+48) 120/576 00/12 SH
1045/2245 S.H. MSLP Prog (H+48) 120/576 00/12 SH
1100/------ Casey Eastern and Western High Seas (H+36) 120/576 0000
1115/2300 S.H. 500 hPa Anal 120/576 00/12 SH
------/2315 Casey Eastern and Western High Seas (H+48) 120/576 1200
1130/------- Asian Sea Surface Temp Anal (Weekly) 120/576 LATEST E
------/2330 Australian MSLP Prog (H+36) 120/576 0000 AUST
------/2345 Indian Ocean MSLP Prog (H+48) 120/576 1200 IO
1145/------- VMC/VMW Information Notice 120/576

 

[Nico]

I think I understand the resolution (and other) limitations of GFS, but I don't see why the error would always be an under-estimate or why it would be worse with a stronger wind. It seems generally agreed to be the case, but I can't see why. Also I don't see why if it's accepted that GFS is 1 force under over 15 knots they don't just add a 1 force offset under those circumstances.

I think it is because the GFS is a logical model based on pressure differential. So having crunched all the numbers and decided that isobar x will here and isobar y will there physics tells us that the wind speed will be z knots. They may be aware that for coastal areas in northern Europe it will be 10 knots livelier but they don't want to muck up the purity and universality of a relatively simple model that shows the overall pattern very well. I'm guessing all this, but it makes sense.

If you look at Windguru, which is a GFS-based web site designed for windsurfers they have added some factors in, because windsurfers tend to frequent places where local effects increase the wind significantly, but this is a bit piecemeal and isn't what the GFS is about. It isn't a forecast - it is a model from which you are welcome to derive a forecast. The Met Office will always consult this and other models (including their own) before making a forecast.

 

[estarzinger]

just to add a few facts to the discussion.

If you look at the following graph, you will see two things #1 that the grib model skill has increased dramatically the past 20 years, and #2 the european model is (slightly) better than the gfs (american) model.

View attachment 22318

This graph, shown using a bit more intuitive metric, the great increase in accuracy in storm track prediction. But you will note that even today the mean error at 72 hrs is 100nm, and that can easily be the difference between hurricane force winds and only a gale.

View attachment 22319

Finally, here is a page from a 2006 NOAA presentation where they suggest their consider their 24hr marine wind forecast to be 55% 'accurate', and wave heights to be 70% 'accurate'.

 

[westhinder]

The thought of you drooling into the mouthpiece of yon valve powered radio and electrocuting yersel .. what a way to go

LOL

 

[RobbieW]

I think I understand the resolution (and other) limitations of GFS, but I don't see why the error would always be an under-estimate or why it would be worse with a stronger wind. It seems generally agreed to be the case, but I can't see why....

I understand this to be because the model prediction is an average over about 900 sq miles (0.5x0.5 degrees). There can be quite a difference, + or -, in actuals across the linear 30 miles or so within each model 'box'.

 

[ytd]

Perhaps you should have bought an SSB receiver for weather fax.

I wouldn't presume to be better at interpreting a synoptic chart than the bureau of meteorology and you can always get their forecasts on VHF. But their ocean forecast districts are quite large and their inshore forecasts are designed for inshore fishermen so includes those local effects. If you are transiting via the offshore islands you get the true ocean winds for which grib is better. Especially for passage planning the grib layout of wind arrows in each cell for each 3 hours is very useful.

 

[charles_reed]

I'd dispute your definition. A forecast is a forecast whether by a meteorologist, computer...or 'reading' a pine cone or granny's lumbago, come to that.
But of course you're right that anyone expecting accuracy on a micro level is in for a disappointment, since the resolution is typically around 60 miles. That said, certain macro effects are modelled in: the Straights of Gib and Bonifacio, for instance (which isn't the same as saying they always get it right).

In the open Atlantic I've had gribs prove spot-on for a full seven days. Good when it happens, if not to be relied on.

By all means dispute my definition but you run the risk of obscuring the point under discussion.
GRIB is an extrapolation of existing data, based upon a computer model, a forecast is an attempt to foretell what is going to happen, either about financial ratios, (relatively simple though banks have been singularly incompetent) or weather (far more difficult).
If you wish to indulge in semantics, be my guest, but don't expect me to agree.

 

[macd]

GRIB is an extrapolation of existing data, based upon a computer model, a forecast is an attempt to foretell what is going to happen, either about financial ratios, (relatively simple though banks have been singularly incompetent) or weather (far more difficult).
If you wish to indulge in semantics, be my guest, but don't expect me to agree.

What is a forecast, as you define it, if it is not also "an extrapolation of existing data", albeit by people? It was you who first took the semantic route in declaring that gribs were not forecasts when the words "grib" and "forecast" go together like "Charles" and "Reed". And I forecast that you will not agree, but so what?

 

[Tomahawk]

Frank will correct me if I am wrong..

I thought that a grib file is simply a standard method of transmitting a forecast in text form... Grided Binary which allows transmission of large data dets over limited bandwith... You then use a resident program on your own PC to turn binary file set into a visual set of images?

I was also under the impression from "weatheronline" that all the major weather forecasting centres use the same grib format to output their owin forecasts.. It is interesting to compare GFS, UKMO, ECMWF, CMC, NOGAPS.

 

[Nico]

Frank will correct me if I am wrong..

I thought that a grib file is simply a standard method of transmitting a forecast in text form... Grided Binary which allows transmission of large data dets over limited bandwith... You then use a resident program on your own PC to turn binary file set into a visual set of images?

I was also under the impression from "weatheronline" that all the major weather forecasting centres use the same grib format to output their owin forecasts.. It is interesting to compare GFS, UKMO, ECMWF, CMC, NOGAPS.

Seeing as we are all being really pedantic I should say that gridded binary (grib) is, by definition, not text, but otherwise I think you are right.

To say that grib forecasts are inaccurate is like saying that Excel accounts are inaccurate. Don't shoot the format.

 

[KellysEye]

GRIB file data points are 25 miles apart so we found them accurate for long distance sailing forecasts but hopeless for coastal sailing. UK inshore forcasts are accurate but, for example, Greek forecasts don't take acccount of inshore Katabatic winds.

 

[ParaHandy]

Your remark about looking out of the window was obviously a joke but as I'm sure you know there is no human intervention in the US model, unlike the Met Office forecast.

Not quite! You see, to generate the Grib data you must have input and Sandettie is part of NOAA buoy data, therefor it seems reasonable that data will be collected from Sandettie and included, in some way, to the Grib forecast at the appropriate time? These Grib files appear at least 5 hours 'late', plenty of time to collect data to refine the GFS model. Clearly this isn't happening and I'm wondering why not? Maintaining the light ship isn't cheap. I'm beginning to wonder if Sandettie is there for the convenience of the Dover ferries.

 

[RobbieW]

Could it be that the forecast time is that of the input data? The model then needs to run against that data; computing, collating and distributing the output. Cant find anything easily that gives a computer run time for the GFS model but given the amount of data and complexity of the model it'd be hours rather than minutes - meteorology has long been one of the bigger consumers of compute power.

EDIT:
Possible insights here and here

 

[franksingleton]

I see several of the usual misunderstandings in this thread but do not have time to go through them all. One rather odd comment was that “GRIBs are note forecasts.” Nonsense. I agree that they should not be used in isolation if anything else is available. Similarly, I would not use a shipping forecast in isolation if anything else was available.

The best way to check accuracy of GRIBs is to repeat the exercise that I have done on several occasions for talks. See http://weather.mailasail.com/Franks-Weather/Grib-Forecast-Examples.

Compare the 1 day forecast with the T=) for that day, Compare the 2 day forecast with the T=) for that day. Do this for the 8 days of the forecast.

An Alternative, but equally revealing test is to look at an 8 day forecast on day 1. On day 2, look at the 7 day forecast. On day 3, look at the 6 day forecast. Etc.. These will all verify on the same day. Two things outcomes happen. Either there first three or four will be consistent. If that is so, then the chances are that the 5 day forecast will be pretty good and you can plan ahead with come confidence.

If the First few are inconsistent, then you will not be able to plan ahead more than a day or two at a time.

Rather than arguing on the basis of subjective impressions, get some numeracy into the argument.

Always remember that the GFS use a 27 km grid for its calculations. That ,means that it can only resolve weather on a scale of about 130 km. It only provides the data on a 50 km grid. There can be no detail.

 

[franksingleton]

Frank will correct me if I am wrong..

I thought that a grib file is simply a standard method of transmitting a forecast in text form... Grided Binary which allows transmission of large data dets over limited bandwith... You then use a resident program on your own PC to turn binary file set into a visual set of images?

I was also under the impression from "weatheronline" that all the major weather forecasting centres use the same grib format to output their owin forecasts.. It is interesting to compare GFS, UKMO, ECMWF, CMC, NOGAPS.

GRIB is a WMO code designed to store and exchange charted information rather than use the old, slow facsimile. It is absolutely standard and used by all national Met services, ECMWF, USN, HIRLAM, etc.

All Met services make their data available to all other Met services. Unfortunately, politicians get in the way. Only the US and Canada, as far as I know, make GRIB files free to all. My understanding is that weatheronline has to pay for ECMWF, JMA, UK info. If anyone knows differently, then we have a stick to beat the Met Office and demand free access.

Can I make a plea. I do not at my laptop or tablet all day monitoring YBW. If anyone really wants me to add a bit of real knowledgee it would help if the sent me an email. It can be either using the YBW or my home email. "frank at franksweather.co.uk."

Or, is there a way of being automatically aleretd if certain topics arise?

 

[franksingleton]

PS. Many of the queries about GRIBs can be found on my site.

You will find info on why they often underestimate winds, detail possible, whether meso-scale GRIBs are any real use etc.

I am very happy to take part in discussions but find it frustrating to see the same queries raised again and again. Also the same misconceptions.