Accuracy of Grib weather data
- Thread starter ParaHandy
- Start date
franksingleton
Well-Known Member
Do not take statements by me, Simon or anyone else at face value. Do your own checks – systematically. As a cruising sailor, I suggest the following.
First, get a GRIB for 8 days for whatever area takes your fancy. Compare the forecasts for H+24, H+48m etc to H+192 with the H+0 for that day.
Secondly, get 8 day forecasts on successive days. Compare the H+192 on day 1 with the H+168 on day 2, the H+144 on day 3 etc.
Do these exercises a few times then make up your own mind on whether you think GRIBs are useful or not. I think that you will find this far more helpful than discussions on this or any other forum. My personal experience is they make a significant contribution to my cruising decisions for up to a week ahead. You may find differently.
macd
Well-Known Member
Couldn't agree more, Frank (and many thanks for all your patient contributions).
Only the reckless would trust themselves to a forecast for a week hence, but gribs are great for putting yourself in the right place for a passage should successive forecasts evolve as predicted. And on the whole I find them remarkably accurate (notwithstanding local effects, about which, 'nuff said)...even here the Med.
franksingleton
Well-Known Member
We were first in the Med in 2010 when GRIBs were just becoming available. At that time, I used to use the DWD 5 day spot wind forecasts as well. Both were giving reasonably reliable predictions of mistrals to 4 days ahead. When we left the Med in 2009, the GRIBs were giving 6 day warnings of mistrals and I had stopped using DWD.
ParaHandy
Well-Known Member
For what it is worth, Sandettie reported an average over 30 days of 21kn and the forecast T+ 12 average was 12kn.
T+12 for a Grib means the Grib issued with the first Grib at 0000 and the forecast read off starting from twelve hours later at 1200, 1300 etc to 2300; then followed by the Grib file with the first Grib at 1200 and the forecasts read off starting from 0000 etc.
The result was the same at all the short ranges I used.
The statistical treatment is almost worthless but it was never intended to be treated in this way. The actual wind on Sandettie was used to calculate the output from a Siemens 3.6MW turbine and compare that with the forecast output using the most accurate, near range, forecast. The actual was 100% greater than forecast.
T+12 for a Grib means the Grib issued with the first Grib at 0000 and the forecast read off starting from twelve hours later at 1200, 1300 etc to 2300; then followed by the Grib file with the first Grib at 1200 and the forecasts read off starting from 0000 etc.
The result was the same at all the short ranges I used.
The statistical treatment is almost worthless but it was never intended to be treated in this way. The actual wind on Sandettie was used to calculate the output from a Siemens 3.6MW turbine and compare that with the forecast output using the most accurate, near range, forecast. The actual was 100% greater than forecast.
franksingleton
Well-Known Member
Sandettie is in the Dover Strait, more or less. With funnelling through the Strait, there will often be a stronger wind than generally in the area. So, looking at this one location, you have two effects; first, the smoothing resulting from the use of a 25 km grid coupled with the fact that the GFS only gives winds at 1/2 degree spacing. Secondly, the Dover strait effect. I did such an exercise some while ago looking at Chanel Light. The GRIBs were surprisingly good.
You have to remember that GRIBs will only give a general pattern. They cannot be expected to get detail right whether it is due to topography, as here, or due to some small feature of the weather pattern. I have said on many occasions that any grid can only resolve features on a scale of around 5 grid lengths. That is over 100 km; the Dover Strait is about 50 km avross. The grid cannot resolve the effects of the Strait. Please see my IOM diagrams at http://weather.mailasail.com/Franks-Weather/Grid-Length-Resolution.
With any forecast, whether it is an inshore forecast or a GRIB you, the user, have to put in your own experience.
Last edited:
Bilgediver
Well-Known Member
Not really out of range if you have a Winlink or Airmail station in range.
Winlink software has the software to select and request grib files using SSB
these files once received can be viewed using an Airmail reader Viewfax.
http://www.winlink.org/
There are Winlink stations in Oz
http://www.winlink.org/RMSHFPositions
You need a ham licence for Winlink however this might not be needed for airmail
though you will need a pactor modem.
Last edited:
franksingleton
Well-Known Member
Or use conventional shore radio stations; Sailmail is the user funded service or Globalmarinenet a commercial one.
ParaHandy
Well-Known Member
I'm not completely unaware of conditions in the channel ...
Big dollops of wind from the North didn't improve accuracy. There seems a permanent error.
franksingleton
Well-Known Member
I have not sailed in UK waters for a long tome now. However, I would expect the human forecaster to correct computer output for the, as you say, well known effects.
What I am saying is that the models simply cannot cope with such detailed topographic effects - as my page at http://weather.mailasail.com/Franks-Weather/Grid-Length-Resolution makes quite clear. Have you looked at my Isle of man example?