Accuracy of weather forecasts

franksingleton

franksingleton

Well-Known Member
Joined
27 Oct 2002
Messages
3,964
Location
UK when not sailing
weather.mailasail.com
Verification of forecast winds from GRIB and similar output

There are various ways, at no cost, of getting GRIB forecasts that are all output from the US GFS model. There are also a number of sources, mainly on repayment, of more detailed predictions using meso-scale models. To the best of my knowledge there has been no systematic study of how good these forecasts are and whether the meso-scale forecasts are significantly better than the global (GFS) forecasts and, if so for what times ahead is there an improvement. Thwy may be better at 24 hours but are they at 48 hours or 72 hours?

Is there anyone out there with the requisite computer skills to set up such a study? This is beyond my capabilities although I could make some suggestions regarding what forecasts could be looked and how to analyse the data.

You can contact me direct at frank (at) franksweather.co.uk
 
This is just a general feeling with no scientific data to back it up whatsoever, but I get the impression that the Met Office inshore forecast always seems to forecast stronger winds than what actually transpires.

I sail out of Peterhead, so I'm close to Rattray Head, the border with 2 areas, both of which usually have higher forecast winds.

I find Xcweather to be more accurate most of the time.

Any comments from others in different areas ? ? ?
 
Hi Frank,

of course, most internet sites use the GFS 1 deg model.

There are some verification stats published by NCEP here http://www.emc.ncep.noaa.gov/gmb/STATS/STATS.html and these make interesting reading to compare various models.

I do see MOS performance data from the model that Weatheronline use, and this is updated daily for their purposes. MOS data generalluy performs better than GFS for locations greater than 1 deg, although of course we both nknow that much of this can be put down to the smoothing effects of the model.

I have GFS GRIB datat converted to text here for the area 0 to 90N and 30E to 60W in a text format if someopne would like to run a verification through it? I guess we could try by looking at the T+24,48,72,96 and 120 time steps and compare the analysis with each of those forecasts?

Simon
 
Simon


I have looked at the US verification stats. Not really very helpful. I am more interested in a comparison of forecasts using the GFS versus actuals at a number of locations over the sea and for periods up to 5 days out. It would be useful to know whether meso-scale forecasts were really any better. I am a little sceptical here. Also whether there is any merit in using the WW3 winds (as do GlobalMarineNet) which are GFS but using a different algorithm to get to the 10 metre wind from that at a pressure height of 0.998 of the surface pressure (the bottom layer of the GFS model). Also, perhaps, how good is NOGAPS?


I see and hear so much hearsay, andI I have my own, no doubt, biased opinions. We need some hard facts to help put some of what I suspect are outrageous claims put into perspective.


I did not get a notification email of your comments, hence the delay in replying.
 
Rowena

I did a reply by email but it seems to have got lost. I will repeat here and expand a little further.

Ref you comment about XCWeather vs the Met Office Inshore forecast. Do you mean Inshore or "Coastal Forecasts" as on the BBC website?

Whichever, you could well be correct. XCWeather, like Movingweather, UGrib, Saildocs, MailASail, the free WindGuru, Windfinder (global version), in fact any GFS output are all the same. Differences can only be in the presentation. The wind data are calculated on a 20 NM grid (approx). The data can thus only refer, effectively, to an area of around 100 NM. XCWeather (like Movingweather and WindGuru) can only interpolate between the grid points. The values can be no better than if you were to use UGrib or any other and do your own interpolation.

The Met Office "Coastal Forecast" on the BBC SHOULD be better because it is calculated on a 6 NM grid and the values should refer to a 30 NM area. They also start with a very good analysis. Likewise, the Met Office Coastal Forecast SHOULD be better than the meso-scale forecasts produced by the various commercial firms none of which, as far as I know, start from a good meso-scale analysis. Like a medical prognosis, a good diagnosis is a prerequisite for a good prognosis, however good is the model being used for the Numerical Weather Prediction. Go to my site and look up Calculating the weather.

I use the word SHOULD because I would like to see an objective and truly independent study of the results. It is only in that way that gut feelings, such as yours, can be verified. I have looked at a limited number of examples comparing the US GFS against actual observations taken from the US NDBC pages. (Go to my site; navigate to the Home page, Forecasts on the Net, Weather actuals to find various links to actual observations.) Up to 48 hours over the open sea, eg Channel LV and the K2 buoy, the forecasts were so good that any improvement would not be possible. Beyond 48 hours, general patterns were good but actual values less so. At inshore locations, eg Sandettie LV, Irish Buoy M2, the forecasts up to 48 hours were good in general terms but actual values less good than those more over the open sea.

The Inshore Waters forecasts, as broadcast by HMCG, NAVTEX and on the net, is a human interpretation of the UK Numerical Weather Prediction model. The forecaster has to cover a 24 plus 24 hour period whereas GRIB output lets you be rather more time specific. I can well see that the GFS would look better than the GMDSS Inshore forecast. Remember that GMDSS forecasts are, essentially, of a warning nature. As a worded forecast you cannot do much better.

Mind you, absolute accuracy is not synonymous with usefulness. One of my gut feeling is that the GFS forecasts are sufficiently good for most of us that there is no reason to pay for meso-scale forecasts. But, it is only a gut feeling. Hence my call for an objective study.

www.franksweather.co.uk
 
Frank,

I never received your email.

I was refering to the Met Office inshore forecast, which I just have a gut feeling gives higher forecast winds than what actually transpires.

When I'm at home, and thinking of going for a sail, I use various websites to try and build up a picture of what is likely to happen. Here's some links to a few that I use -


http://meteonet.nl/aktueel/brackall.htm - the charts on here fit my screen.


http://meteocentre.com/cgi-bin/get_sao_stn?STN=EGPD&DELT=24 - Aberdeen airport.

http://www.wunderground.com/weatherstation/WXDailyHistory.asp?ID=IABERDEE12 - This site is 5 miles from my house. Scrol down to see the graphs.

http://www.metoffice.gov.uk/weather/uk/radar/index.html - Scroll through the times to see if I'm likely to need my oilies!

http://xcweather.co.uk/ - Very useful for wind predictions, I find.


When I'm on the boat, all I've got is the Shipping/Inshore forecasts from the BBC & coastguard. I've got a radio cassete recorder that I set to record the early morning BBC broadcast, then get the same forecast from HMCG on the VHF. I get to see the synoptic charts from the met office which are sent to the ASCO base in Peterhead for use by the North Sea Supply boats, and a copy is pinned up on the marina noticeboard.

As I said, I've no evidence that the Met Office forecasts higher wind speeds than what we actually get. It's just a feeling from being aboard and going for a sail. I would be most interested in any comparison of forecast/actual winds that you get.

Cheers
Jim
 
Thanks Rowena. I had not seen http://meteonet.nl/aktueel/brackall.htm before. It looks a more useful link for several charts than http://www.weather.org.uk/charts/thumbs.html which uses thumbnail compression. For single charts, there are many links, of course.

I guess that it is really all a question of what you use forecasts for. My main use is as planning tools, by looking at forecasts each day it is rarely a surprise on the day. My main decisions then become not so much "Can I go?" but "Is it sensible to go?" or "Must I go?". In other words, two or three days later, will I regret going or not going, as the case might be.

I have really come round to GRIBs as my main source together with whatever GMDSS forecasts I can get.
 
You must log in or register to reply here.

Share this page

Top