Yellow Brick

[jesterchallenger]

GRIB files

GRIB files: I have found that downloading GRIB files is quick (and therefore inexpensive) and easy. The only failing I have experienced with them is that the wind speeds are consistently understated, which I understand is a generic problem. Other than that, I found them to be consistently accurate in giving a big picture of what was going to happen and therefore a useful tool in deciding where to point the boat! Bearing in mind that my old boat only did 6 knots, anything other than a big picture was all a bit academic. Bearing in mind the number of times the local shipping forecast does not exactly coincide with what's actually happening (come on, we've all moaned about it!), I think you have to appreciate that a mid ocean forecast is just that - a forecast. I now have a significantly faster boat, so it will be interesting to see if that makes any difference to my interpretation of GRIB files.
Mailsail: During my dealings with Ed Wildgoose and his team at Mailasail, I have always found them most helpful and efficient, and very understanding of computer duffers like myself. They offer good service and I would not hesitate to recommend them. I use a cheap netbook with Windows XP, on the basis that when it gets drowned it won't cost a fortune to replace. Needless to say it's never missed a beat.

 

[franksingleton]

GRIB speeds

The speed problem with any NWP, the GFS included, is because they can only calculate their forecasts on a 25 km grid, although the GFS GRIBs are only available on a 50 km grid (~ 1/2 degree.) This has the effect of smoothing out detail. Stan Honey (the Volvo RTW winner in ????) said that he adds 25% to winds in the southern ocean and 20% elsewhere.

Used in the way that I described earlier in this thread, they can give good planning/forward looks up to 7 or 8 days at best, 3 to 5 days at worst. Any worded High Seas, Offshore or Inshore forecast is only ever going to be a brief summary, almost a headline. GRIBs can help to amplify them.

 

[franksingleton]

GRIB speeds

The speed problem with any NWP, the GFS included, is because they can only calculate their forecasts on a 25 km grid, although the GFS GRIBs are only available on a 50 km grid (well. 1/2 degree.) This has the effect of smoothing out detail. Stan Honey (the Volvo RTW winner in ????) said that he adds 25% to winds in the southern ocean and 20% elsewhere.

Used in the way that I described earlier in this thread, they can give good planning/forward looks up to 7 or 8 days at best, 3 to 5 days at worst. Any worded High Seas, Offshore or Inshore forecast is only ever going to be a brief summary, almost a headline. GRIBs can help to amplify them.

 

[MailASail]

Additionally remember that wind speeds are calculated for various "heights" above sea level. Friction of the wind against the land is considerable and often you can see great differences between "sea level", "10m", "25m" and "50m" heights (you can go all the way up to upper atmosphere)

MailASail supply "10m wind speed" when you request wind speed. This may vary from other sources and may deviate from the height of your mast...

I think the NOAA website has a discussion of predicted speeds versus observed speeds. As well as the reasoning above, remember that they produce something similar to an "average" figure, whereas as a sailor we tend to notice more the peak of the gust and don't mentally average down for the wind not being constantly at that peak value. Therefore there are various rules of thumb (such as from Frank) to give a number that is closer to what you see on the meter

Cheers

Ed W

 

[MailASail]

I am talking about day after day of GRIB files saying the wind would be Southerly but in fact we had constant Easterlies. The GRIB files also ignored a F8 depression that was with us for 3 days. I don't consider those small details.

Just to clear up a potential misunderstanding:

"GRIB" is a file format for transferring meteorological data around. It says nothing about the source of such data, nor is a viewer for such data provided.

There are some *sources* for grib data that is extremely "free" in both copyright and cost, and that is the output of several different computer models from NOAA (ie the USA weather office). Now the most common of those several computer models (but not the only) is the GFS model, and 95% of the time when someone says "this grib", they actually mean "the output from the NOAA GFS model"

This on the Yachting Monthly Scuttlebutt forum is illustrative of how bad GRIB files are:

And unless ParaHandy is talking about a different Sandette Light Vessel, I put it in the the Dover Straight which is somewhere you would expect forecasts to be correct.

Again, GRIB files can be used by you and I to send all kinds of numbers between us. GRIB is a file format and not definition of something which predicts weather. So you would more correctly say "...illustrative of how bad the *NOAA* GFS data was during this time period..."


Furthermore if you are looking at *GFS* data then you should be wary of making claims such as "Dover straight which is somewhere you would expect forecasts to be correct" without backing that up? Why? Because the GFS is a model which is intended largely to be correct on a macro scale and has only limited ability to handle fine details where local landmass, etc might influence the forecast (I'm sure Frank can entertain you for hours on the subtle details, but if you will forgive a broad brush explanation)

So you would expect little more from the GFS model than an attempt to predict where various lows and highs will move over the next 14 days. Please do not be surprised that inshore waters are poorly predicted, or that features which are obvious to sailors such as wind shifts in the morning/evening are badly predicted.

If you want the same macro forecast as GFS, but with better inshore details (meseo prediction) then I believe you can get data from Theyr relatively cheaply and my (limited) understanding is that they take the GFS model, but recrunch it at higher resolution to get a better model where local features (land) will influence the forecast, but unfortunately you have the same basic (and possibly wrong) prediction of the global/macro forecast


Now, not to defend NOAA, but they do several things right in my opinion:
- They argue the tax payer paid for the model and then *give away* the data for free, this is very generous
- They spend a ton of money on collecting data from bouys, balloons, satellites and weather stations (the raw satellite data can actually often be intercepted for free if you are interested - they are pretty open). Additionally SSB users will be familiar that individual yachts can also contribute observations of local weather
- They then use this mass of data to compute a backtest and see how well the forecast did versus the prediction
- Finally, and in my mind the most important: they *publish* the results of how well they do versus predictions and that data is available every day.
- (They also re-run the model lots of times with slightly perturbed inputs and publish this as an ensemble model - it takes longer to appear, but gives some information on how well the computer is likely to predict reality)

Currently NOAA ranks around 3rd for predictive performance in the Atlantic region. UK Met office (something like £20K/year for data) ranks second, and ECMWF (up to 2K euros a month) ranks first. You should consider ponying up for ECMWF data if you want the best (but note that they are only slightly better for quite a lot more dosh...)

I think some are forgetting that computers have limited ability to predict the weather and computers do better some days than others...


Finally, please read and heed the large warnings on the NOAA data. The data is generated completely automatically, and placed on the internet with NO checking whatsoever, nor is there any warranty... They do not pull or alter the data if it's completely different to observations - buyer beware (and given it would take another 6 hours to repredict, what would you realistically expect them to do anyway?)


"GRIB" is not at fault. GRIB is just a file format. Please be careful to define *where* you get your data from and critique the source, not the file format. The corollary is to beware that MailASail, Saildocs, GMN, Moving Weather all simply feed you data from NOAA, so please try and look for the *source* of the data, don't quote the name of the feed. (If you do this you will find that there is only actually a small source of raw weather data available to you...)

Good luck

Ed W

 

[MailASail]

I think you might be talking to the wrong sailors. I was really interested to hear the accumulated experience of the Jester Challengers and the Jester Azores Challengers indicated that GRIB files were generally inaccurate, which has also been my experience. Albeit my GRIB file experience has only been in the Pacific not the North Atlantic.

I should add that 'Triple Venture' mentioned that Weatherfax which produces synoptic charts and are aided by human interpretation were quite reliable. Tony has a dedicated Weatherfax machine and someone else mentioned they had tried to download Weatherfax to their computer but unfortunately found it incredibly slow as it scanned each line.

You need to watch that the *source* of the GRIB files you mention is probably NOAA, and likely the *source* of the synoptic charts is also NOAA, and that they generate the synoptic charts approximately by printing out the GRIB files and scribbling on them...

I very commonly see people using the *same* data either at two different times, or in slightly different presentations and yet holding the opinion that one source of data is better than the other (when we can see that the sources are sometimes the same). It's easy to be misled - be careful

Synoptic charts, although possibly deriving from the same source as your GRIB files do arguably have value because the man "scribbling" on them has access to a lot more information than just the grib data, eg he can see whether the predictions have been good recently, he has radar views of the systems, etc. Therefore arguably he can augment the printout in useful ways that *might* make them more valuable than the raw grib data

It's a complicated subject, I would recommend reading Franks web pages thoroughly which gives you a good understanding of the underlying problems. If you find yourself believing that A is better than B then just make sure that you understand the underlying reasons why this should be, and it's not just empirical, otherwise I think there is a large risk of it being simply coincidence?

Computers fundamentally don't (yet) predict weather to the level that many people would desire. But in fact if you relax your definition of "correct" then over the last few years predictions have got quite a lot better (just beware that good macro predictions don't always lead to knowing the wind spot on in some inshore area - which is what many sailors would argue is "correct")

Have fun!

Ed W

 

[franksingleton]

Additionally remember that wind speeds are calculated for various "heights" above sea level. Friction of the wind against the land is considerable and often you can see great differences between "sea level", "10m", "25m" and "50m" heights (you can go all the way up to upper atmosphere)
.............

Just a little amplification. The bottom layer of the model is at a pressure of 0.998 of the ]b]surface[/b] pressure. In other words. it is a terrain following model. That is a height of around 20 metres above sea level. The 10 m wind is derived using an algorithm that takes into account surface roughnwess amd the stability of the air. For a given pressure gradient, it should show a stronger wind in an unstable NW air flow than in a stable SW one.

If you use some GRIB sources - zyGrib is one, you can get gust speeds as shown by the model.

 

[PacketRat]

This proved to be quite an informative thread. Thanks to all concerned.

My personal observations on weather are generally when beating, it's blowing a hooly, and when running, it's flat calm. And when the sun shines, it's time to shake out a reef.