Passageweather.com for the Mediterranean

[Ric]

Hello,

Passageweather.com uses three different models for wind predictions in the Mediterranean - often they differ substantially.

What is collective wisdom of forumites as to the most accurate of the three?

 

[billreilly]

All 3 of the models (GFS, COAMPS and SWS) have their strong points, but neither of them is going to be more accurate all the time.

For long-term forecasts, and for more open stretches of water, I'd say the GFS model would be your best bet.

In areas where the local terrain/topography is a big influence on the wind speed and direction, such as around islands or in the Strait of Gibraltar and the Strait of Bonifacio, the SWS model is by far the best.

It's best to have a look at both the GFS and the higher resolution (SWS or COAMPS) forecasts to get a better idea of what to expect... It's always good to have a second opinion.

Bill

PS - For other questions regarding PassageWeather.com, check out our new Support Forum at http://www.passageweather.net

 

[jordanbasset]

Thanks for the info and details of the support site, very useful

 

[franksingleton]

All 3 of the models (GFS, COAMPS and SWS) have their strong points, but neither of them is going to be more accurate all the time.

For long-term forecasts, and for more open stretches of water, I'd say the GFS model would be your best bet.

In areas where the local terrain/topography is a big influence on the wind speed and direction, such as around islands or in the Strait of Gibraltar and the Strait of Bonifacio, the SWS model is by far the best...............l]

Perhaps I can amplify what Bill has said.

The GFS, like the UK, calculates on a 25 km grid worldwide. Any model can only describe weather and topography on a scale of about 4 to 5 grid lengths. For the GFS, this means about 100 km at best. Like the UK and other global models ithe GFS will handle large scale weather well up to about 5 days. See http://weather.mailasail.com/Franks-Weather/Grib-Forecast-Examples.

Finer scale (meso-scale) models can describe smaller scale features. COAMPS uses ~ 10 km grid and, so, can define weather and topography down to about 50 km. I do not know what grid SWS use for their free service ie as available on Passageweather. I doubt that it is much less than 10 km.

How far ahead, a model can reasonably be used will depend upon the typical lifetimes of the weather that it can describe. In the GFS case, small errors (and there always are errors) in the analysis are likely to take up to 5+ das to grow sufficiently to create serious errors.

With a model using a 10 km grid, assuming that all available data have been incorporated into the analysis, the time taken for small errors to manifest themselves will be that much shorter. The UK only runs its 10 km meso-scale model (the NAE) up to 48 hours and their forecasters probably do not use the information beyond 36 hours.

The COAMPS produces output to 72 hours. I think that this is probably too far ahead. The SWS model runs to 120 hours which is far too far ahead, especially if they use a grid length less than my 10 km guess.

Bill has probably studied the behaviour of these models in the Med more diligently than I have so that I cannot query what he says about the SWS being better than COANPS in certain areas. This leaves me in something of a quandary because few of these models run by private firms use all the observational data available. See http://weather.mailasail.com/Franks-Weather/Weather-Observations-Nwp. Put simply, just handling that amount of data, applying quality checks, experimenting to see how the greatly differing kinds of data have to be merged is beyond their capability. This is as opposed to the big boys of NOAA, The UK, French etc Met services who have the data sources, computing power and the scientific ability needed.

As a result, I always have to question how these firms can produce well founded detailed forecasts. COAMPS is run by USN, and, I assume, benefit from NOAA experience and expertise. On the other hand, the smaller the grid length, the better is the ability to model small scale topography. This leaves me in something of a quandary.

My reading of the situation, or how I square the circle, is to guess that these fine scale models do well for short periods ahead in fairly settled weather when the main drivers are topography rather than large scale, synoptic, effects. If my guess about the SWS grid length is correct, then this does not explain why Bill finds that the SWS out-perform COAMPS in the Med.

After well over 50 years in meteorology, there is still much to learn.

 

[billreilly]

All of what Frank points out is correct, in theory, and I'm in no position to argue with any of his points... He has much more experience than I do and I respect him for that...

I look at it all from a more practical side... Between regattas and yacht deliveries, I sail several thousand miles a year in and around the Mediterranean, and I always compare the 3 models (GFS, COAMPS and SWS)...

Just last week I delivered a race boat from Cascais, Portugal to Valencia, Spain... Going down the Portuguese coast I would say all 3 models were pretty much the same, with minor differences between them, neither of them better than the others... Turning the corner and all the way to the Strait of Gibraltar the SWS model was way better than the others... no comparison... There was a fairly brisk easterly (30-35 knots) blowing in the Strait and the SWS model pretty much nailed it, in both speed and direction... The COAMPS was better direction-wise than the GFS, but both underestimated the wind speed by 10-15 knots... Once in the Med and away from the strait they were all relatively similar, with the SWS probably a bit better at first, but later on it all evened out...

I've observed very similar results on many deliveries over the past couple of years... In the Strait of Bonifacio the SWS model almost always wins... In the Gulf of Lion I give the GFS more confidence than the others, and everywhere else it's usually too close to call...

The bottom line: What I always do is look at all 3 models, see where they coincide and where they disagree, and then make my own decision based on the current (real) conditions... then, after a few hours, I look at which model has performed better up until then and I start to give that one a bit more weight... It usually works well for me...

Bill

 

[haydude]

Hello,

Passageweather.com uses three different models for wind predictions in the Mediterranean - often they differ substantially.

What is collective wisdom of forumites as to the most accurate of the three?

I did not know there was a choice therefore I always used the default I am presented with and I found it very, very accurate, up to four days in advance.

 

[franksingleton]

All of what Frank points out is correct, in theory, and I'm in no position to argue with any of his points... He has much more experience than I do and I respect him for that................/QUOTE]

Thanks, Bill. I guess that I usually add one BF to the GFS so find it generally OK. Certainly, when it comes to Mistrals and those lows that come up from N Africa it does very well. Even back in 2005/6 it was giving an indication of Boras. I have no doubt that it will do better now than 12 months ago.

How far ahead do you find that the SWS out performs the GFS in areas such as the Gib Strait?