Choosing a weather forecast

sighmoon

sighmoon

Well-Known Member
Joined
6 Feb 2006
Messages
4,114
Location
West Coast
Visit site
Looking at windguru and the met office shipping forecast for the next 6 hours, the forecasts are very different. For the Forth, the met office say NW becoming variable 3 to 4, whereas windguru say NW 1 becoming W 2.

Are they using entirely different models (i.e. one of them is closer to the truth), or is the difference because windguru is a shore forecast, and so takes local effects into account? i.e. would we encounter conditions like the met office forecast offshore, whereas on the beach, we'd get windguru's?
 
I was at a RYA event once where they had a forecaster to tell us how it all worked. They said that they do many plots starting at slightly different points based on the natural variability and error margins. They are run through various models and essentially end up as a bunch of lines piloted on a graph. At the start of the graph all the plots are pretty close as there is just the previously mentioned variation between them. They then slowly diverge as they look ahead. 95% of the weather falls between the upper and lower plots. If they diverge widely then the accuracy is poor, if they are tight then it is good. It would be useful if they made reference to this so that we could know if a forecast is likely to be right or not.
 
pmagowan said:
... If they diverge widely then the accuracy is poor, if they are tight then it is good. It would be useful if they made reference to this so that we could know if a forecast is likely to be right or not.
Click to expand...

i.e. assign a level of confidence to the forecast, which would inevitably reduce as the forecast looks further ahead in time.

Perhaps forecasters don't credit their public with sufficient intelligence to understand this sort of subtlety.

Perhaps however they also consider that the multitude of alternative 'predictions' would over-confuse the forecast.

The weather will never be perfectly forecastable, so if we as sailors are going to rely on their methods, then we also have to rely on our own experience and skepticism, and make our own judgements.
 
All I know is that I have cancelled/ cut short several of trips recently because of strong wind warnings from the Met Office. Windguru and other web-based forecasts gave much lower wind predictions but I played it safe.

Guess what - they wre right and I sat at home fuming about wasted opportunities
 
For me the key factor is knowing where the source data for the forecast has come from. As has been mentioned, a great many of the free (and even some of the paid-for) web-based forecasts rely ultimately on the GFS output, their forecasts can thus be expected to agree with one another quite well and it would be dangerous to assume that because two or more of these types of forecast agree that you can have confidence in them. I always use the GMDSS forecasts as well as the GFS (via zygrib).
 
sighmoon said:
Are they using entirely different models (i.e. one of them is closer to the truth), or is the difference because windguru is a shore forecast, and so takes local effects into account? i.e. would we encounter conditions like the met office forecast offshore, whereas on the beach, we'd get windguru's?
Click to expand...

Sea Area "Forth" is a whole lot bigger than the Firth of Forth, so you are probably looking at a much more generalised forecast taking in a big chunk of North Sea, whereas wind guru will be specific for a location.
 
And Sighmoon - What is the wind actually doing in the Firth at the moment?

Both your weather forecast quotes show NW although variable in strength.

http://www.forthweather.co.uk/ Shows ESE 8mph as current conditions.

This incidentally matches what Meteo Consult (Android App) predicts for Fife Ness - North Berwick at the moment. ESE F2 gusting F3.
As a bonus Meteo Consult also shows a reliabilty percentage and happens to show sea states (due to wind and swell) as well as tide times.
 
Listen or read carefully forecasts broadcast under the GMDSS. These are written with safety very much in mind.

Get the GFS, best in GRIB form via email (Saildocs or MailASail) m by FTP (zyGrib or Ugrib), or any number of Tablet Apps. These are mostly all the same data. They are also the same info that you will see on Windguru, Magic seaweed, XCWeather etc. Same data different appearance.

The GRIB file services are more useful because you can get data you want for an area and times of your choice. The data are stored automatically for later reference when offline.

Distrust any forecasts that is automatically generated and has no human input. Any service that offers detailed and highly localised information for many locations are almost certain to come into this category.

Use the GRIBs for planning up to 5 or 6 days ahead. Read local GMDSS forecasts in the light of the GFS. For more info wait until early 2014 for http://weather.mailasail.com/w/uploads/Franks-Weather/bookcover.pdf.
 
nigelmercier said:
The irony is that they do this in the US.
Click to expand...



This is a perennial problem that has exercised the Met office over the years. There are many questions such as what does a confidence of X% really mean. No forecast, apart from trivial ones (eg there will not be a frost on July 1st in London) is ever 100% certain.

Forecasts use words such as isolated showers or frequent showers. These could, in principle be quantified as, say 86% or 25&. Does that give a better idea to the user?

Forecasters use model ensembles to decide what confidence they have in computer forecast output. That enables them to say that they are sure or unsure about the next few days. I do not think that they can reasonably put numbers to the probabilities.

I do not know how the US forecasters quote percentages or what they really mean. I must say that I am not at all sanguine about the British public. Just cast your mind back to monthly and seasonal forecasts that were always expressed in a probabilistic sense. Such predictions were always probabilistic but there are still scars after the barbecue summer “forecast.”

I fear that it would be a big stick for the more irresponsible organs of the press to use to beat the Met Office. To say nothing of some of the more ludicrous posts that appear on these forums, maube in jest, maybe semi-serious, maybe serious. I can well understand the reluctance of the met Office management to introduce the idea widely. .
 
Last edited:
sighmoon said:
Looking at windguru and the met office shipping forecast for the next 6 hours, the forecasts are very different. For the Forth, the met office say NW becoming variable 3 to 4, whereas windguru say NW 1 becoming W 2.

Are they using entirely different models (i.e. one of them is closer to the truth), or is the difference because windguru is a shore forecast, and so takes local effects into account? i.e. would we encounter conditions like the met office forecast offshore, whereas on the beach, we'd get windguru's?
Click to expand...

If I'm going 'offshore', eg UK/Gib etc, I quite like these
http://www.ecmwf.int/products/forecasts/d/charts/medium/deterministic/msl_uv850_z500/
 
The Met Office sea areas are very large and subject to significant variations across them. Last week-end Solent Coastguard broadcast a gale warning for sea area Wight. I was in the Solent and felt it unlikely that we would be affected so:

  • I checked my little barometric pressure recorder - steady.
  • Looked at met office website. Sure enough - shipping forecast area Wight - shown red. But inshore waters forecast showed Selsey Bill to Lyme Regis as the only one without a strong wind warning.
  • Looked at http://passageweather.com/ and it confirmed the forecast - gales in Cherbourg and no more than 5 this side of the Channel.

So Met office spot on but I am sure there were many complaining.

http://passageweather.com/ will give you GFS, COAMPS and WRF forecasts and as a third system of display you can also use http://www.xcweather.co.uk/
They all tend to be pretty accurate at a local level.
 
Last edited:
pugwash94 said:
The Met Office sea areas are very large and subject to significant variations across them. Last week-end Solent Coastguard broadcast a gale warning for sea area Wight. I was in the Solent and felt it unlikely that we would be affected so:

  • I checked my little barometric pressure recorder - steady.
  • Looked at met office website. Sure enough - shipping forecast area Wight - shown red. But inshore waters forecast showed Selsey Bill to Lyme Regis as the only one without a strong wind warning.
  • Looked at http://passageweather.com/ and it confirmed the forecast - gales in Cherbourg and no more than 5 this side of the Channel.

So Met office spot on but I am sure there were many complaining.

http://passageweather.com/ will give you GFS, COAMPS and WRF forecasts and as a third system of display you can also use http://www.xcweather.co.uk/
They all tend to be pretty accurate at a local level.
Click to expand...


The Met Office MUST issue a gale warning if there is a gale expected somewhere within a sea area. That is a requirement placed upon it by the users represented by the Working Group for UK Safety of Navigation. (WGUKSON)

In the case you describe, it seems to me that it was all pretty clear. The inshore forecast had no strong winds but there were gales in Wight. Wight is not a large area and the word constraint may have made it difficult to differentiate between north and south of the area.

I was on duty many years ago when the C/O RNAS Lee-on-Solent complained that there was a southerly gale cone flying outside his window but the wind sock showed a light northerly, There was a cold front across mid-Channel. South of it winds were WSW F9 – hence the S cone. North of it there were light N winds.
 
alant said:
If I'm going 'offshore', eg UK/Gib etc, I quite like these
http://www.ecmwf.int/products/forecasts/d/charts/medium/deterministic/msl_uv850_z500/
Click to expand...

ECMWF Should be the best for extended range ie up to 10 days at present. Their model uses the shortest grid length of all global models. They can put great effort into data analysis because they do not have operational pressures such as aviation or meeting a TV schedule.

However, on any particular occasion, they may or may not be the best. Whichever you use, keep comparing each run with the ones before. Look for consistency. That is a message in Reeds Weather Handbook due out early 2014.
 
You must log in or register to reply here.

Share this page

Top