ECMWF vs GFS - over the top gust predictions?

webcraft

Well-known member
Joined
8 Jul 2001
Messages
39,942
Location
Cyberspace
www.bluemoment.com
.
Currently waiting on a weather window for a 3-day passage across Biscay and poring over the GRIBs, mostly via Windy.

The European model regularly shows 5 knots or more higher wind speeds, but it is the gust forecasts that I find most alarming. ECMWF is regularly showing gust speeds ten or fifteen knots higher than GFS, making all the difference between setting off and not setting off. 25 knot gusts yes, 35 knot gusts no thank you very much.

Is this a regular feature of the models, and which one should I believe? Is the truth likely to be somewhere between the two?

Currently looking at Friday afternoon just off the coast here . . . GFS is 9kts gusting 11, ECMWF is 19kts gusting 27 and this discrepancy is fairly standard for the period and area I am looking at. It makes a nonsense of passage planning, and frankly I wish I had never discovered there were two models.

- W
 

LittleSister

Well-known member
Joined
12 Nov 2007
Messages
17,547
Location
Me Norfolk/Suffolk border - Boat Deben & Southwold
Visit site
Hopefully Frank Singleton or some other expert will be along to explain the differences between the models, but as a more general thing I think you will usually find that as the time gets nearer the models will converge.

Nothing for it but to stock up ready, and keep an eye on the evolution of the forecasts,
 

Buck Turgidson

Well-known member
Joined
10 Apr 2012
Messages
3,125
Location
Zürich
Visit site
My own experience is that GFS under forecasts but is generally more accurate than ECM. Use GFS gusts and add 5kts. But to be honest it's only a forecast and the weather doesn't often check what it's supposed to be doing.
 

geem

Well-known member
Joined
27 Apr 2006
Messages
7,236
Location
Caribbean
Visit site
In the Caribbean we always found the ECMWF spot on. I also dont worry too much about gusts especially if sailing with wind behind the beam. Gusts tend not to determine sea state so not a big issue as long as you reef accordingly
 

Birdseye

Well-known member
Joined
9 Mar 2003
Messages
28,092
Location
s e wales
Visit site
.
Currently waiting on a weather window for a 3-day passage across Biscay and poring over the GRIBs, mostly via Windy.

The European model regularly shows 5 knots or more higher wind speeds, but it is the gust forecasts that I find most alarming. ECMWF is regularly showing gust speeds ten or fifteen knots higher than GFS, making all the difference between setting off and not setting off. 25 knot gusts yes, 35 knot gusts no thank you very much.

Is this a regular feature of the models, and which one should I believe? Is the truth likely to be somewhere between the two?

Currently looking at Friday afternoon just off the coast here . . . GFS is 9kts gusting 11, ECMWF is 19kts gusting 27 and this discrepancy is fairly standard for the period and area I am looking at. It makes a nonsense of passage planning, and frankly I wish I had never discovered there were two models.

- W
There are more than 2 models, and the prediction of weather patterns is such a complex job that its hardly surprising that there is minor variation between them particualrly over a 3 day period. You have two choices, well three if you just decide to chance it. First is to wait for a period of more stable weather when the forecasts will all come together. The second is to change the crossing - The Gironde to Bilbao is barely 36 hours and forcests that far ahead are likely to be way more reliable than 72 hours.

Had a look at the models and they are saying at the moment that there is likely to be a more stable period in 5 or 6 days time
 
Last edited by a moderator:

Carib

Active member
Joined
30 Mar 2011
Messages
352
Location
Southampton
www.sailinginlimbo.blogspot.com
I agree the different models are confusing. The way these sites display information suggests a high degree of accuracy, but what's the point if you then click on another model and it shows something entirely different? The main issue is probably that the sites don't indicate the uncertainty of the prediction.

Is there some established guidance on which is more accurate in different scenarios?

By way of example, supposing I want to be sailing near the Channel Islands on Thursday at 2100 (just 36 hours away). Below is the GFS forecast from passage weather alongside the COAMPS model. Just a tiny difference in conditions there (and these are averages, not gusts).

Of course there will be different models, but how do we know which to pay more attention to - alongside the more established forecasts, of course?

Screenshot_20200826-084443.pngScreenshot_20200826-084509.png
 
Last edited:

LittleSister

Well-known member
Joined
12 Nov 2007
Messages
17,547
Location
Me Norfolk/Suffolk border - Boat Deben & Southwold
Visit site
Of course there will be different models but how is the average punter to know which to trust?

None, until close to the time. Even then take them with a pinch of salt.

Forecasts do an amazing job of giving a good idea of what is likely to happen, give or take, but expecting an accurate, precision account of what the future will hold in something as complex and dynamic as our atmosphere is unrealistic.
 

Carib

Active member
Joined
30 Mar 2011
Messages
352
Location
Southampton
www.sailinginlimbo.blogspot.com
None, until close to the time. Even then take them with a pinch of salt.

Forecasts do an amazing job of giving a good idea of what is likely to happen, give or take, but expecting an accurate, precision account of what the future will hold in something as complex and dynamic as our atmosphere is unrealistic.
Ha - I edited the word 'trust' to anticipate this objection :) But 36 hours IS close to the time.

My point, though, is that all these apps and sites purport to do exactly that (give an accurate, precision account), and indeed they can often be highly accurate. Ignoring them seems a bit drastic, particularly when the Inshore forecast is giving (as it did on many days in July) a helpful 'Force 3 to 5'.

Looking just at my randomly chosen example above, it's the difference between expecting a dead calm or sailing in a near-gale. Hardly a minor variation.

So, if we don't ignore them entirely, I would still like to know essentially 'which is best'? If no-one can tell me, aren't we all left waiting in harbour just in case the 40 knot gusts predicted by one model turn out to be true..?
 

LittleSister

Well-known member
Joined
12 Nov 2007
Messages
17,547
Location
Me Norfolk/Suffolk border - Boat Deben & Southwold
Visit site
If no-one can tell me, aren't we all left waiting in harbour just in case the 40 knot gusts predicted by one model turn out to be true..?

I would be! The idea of sailing around the Channel Islands in 40 knots, even just gusts, doesn't appeal.

The divergence between the models suggests very dynamic, difficult to predict conditions, so I would expect greater certainty only over much shorter periods.

Personally, I'd err on the side of caution, and not set out until the models converge, or the situation becomes more settled.
 
Last edited:

GHA

Well-known member
Joined
26 Jun 2013
Messages
12,104
Location
Hopefully somewhere warm
Visit site
Unless you do a load of work finding a reliable 10m source of wind speed away from local effects and graph actual against forecast "my forecast is better than yours..." is just guesswork...........
Imho :)
This is met office data against actual in the Scillies
Scilly Wind – My First website

Webby - the compare view on windy is good for, well, comparing the models ;) Makes it easier to see what they're up to, discrepancies often are just with one model showing a front coming through a bit earlier >

CfOWobI.png
 

RobbieW

Well-known member
Joined
24 Jun 2007
Messages
4,596
Location
On land for now
Visit site
In principle the ECMWF 'should' be more accurate that the GFS. Forecasting is a function of grid length, GFS has a grid of 28km and ECMWF 9km. So ECMWF should be providing a finer grained picture than GFS. Frank has this to say about GFS...
In particular, it is important to remember that any model can, at best represent weather on a scale of about 5 x the grid length. For the GFS, this is about 100 to 150 NM. As a result, winds speeds over the open sea will be under-forecast by about one Beaufort force or 20%.
 

Carib

Active member
Joined
30 Mar 2011
Messages
352
Location
Southampton
www.sailinginlimbo.blogspot.com
That's interesting thanks. But I don't follow that quote - why does the large resolution area mean speeds will be under-forecast at sea? And if this is known, why do the apps not adjust for it (or do they)?

It's not looking good for my entirely imaginary Channel Isles sail.. now it's just become F8 not F7!
 

RobbieW

Well-known member
Joined
24 Jun 2007
Messages
4,596
Location
On land for now
Visit site
That's interesting thanks. But I don't follow that quote - why does the large resolution area mean speeds will be under-forecast at sea? And if this is known, why do the apps not adjust for it (or do they)?

It's not looking good for my entirely imaginary Channel Isles sail.. now it's just become F8 not F7!
Having looked a little wider, I suspect Franks site is a bit out of date. The GFS now give thier grid length at 22km so representing a scale of around 100km, or c. 65 miles, in length

I dont know why the under reporting happens, if I had to guess it'd be that averaging over the larger scale reduces the accuracy for specific points along the line.
 

GHA

Well-known member
Joined
26 Jun 2013
Messages
12,104
Location
Hopefully somewhere warm
Visit site
That's interesting thanks. But I don't follow that quote - why does the large resolution area mean speeds will be under-forecast at sea? And if this is known, why do the apps not adjust for it (or do they)?

It's not looking good for my entirely imaginary Channel Isles sail.. now it's just become F8 not F7!
Windy is a great site to check, lots of live reporting inc ships - it's usually obvious which are dodgy data but generally GFS wind speeds are close to reported data. Click on the station for live/forecast comparison for the different models. No gust data unfortunately.

1YRsum1.png
 

Koeketiene

Well-known member
Joined
24 Sep 2003
Messages
17,743
Location
Finistère
www.sailblogs.com
.Currently looking at Friday afternoon just off the coast here . . . GFS is 9kts gusting 11, ECMWF is 19kts gusting 27 and this discrepancy is fairly standard for the period and area I am looking at. It makes a nonsense of passage planning, and frankly I wish I had never discovered there were two models.

A man with a watch knows the time.
A man with two watches is never sure.

Pick a source you know from experience is trustworthy and don't look anywhere else.

In my experience, I have found www.passageweather.com to be spot on over 90% of the time.

When I started sailing, the weather was what was posted in the HM window and you planned accordingly.
 

Birdseye

Well-known member
Joined
9 Mar 2003
Messages
28,092
Location
s e wales
Visit site
Ha - I edited the word 'trust' to anticipate this objection :) But 36 hours IS close to the time.

My point, though, is that all these apps and sites purport to do exactly that (give an accurate, precision account), and indeed they can often be highly accurate. Ignoring them seems a bit drastic, particularly when the Inshore forecast is giving (as it did on many days in July) a helpful 'Force 3 to 5'.

Looking just at my randomly chosen example above, it's the difference between expecting a dead calm or sailing in a near-gale. Hardly a minor variation.

So, if we don't ignore them entirely, I would still like to know essentially 'which is best'? If no-one can tell me, aren't we all left waiting in harbour just in case the 40 knot gusts predicted by one model turn out to be true..?

None of them purport to give " an accurate precison account" - the movement and state of the atmosphere is beyond that sort of calculation, and thats without taking into account local issues. Your expectations are unrealistic. One site that lucks out on an accurate forecast one day can be the worst forecast the next day.

What you need to do for a three day passage is to study the pettern of weather systems coming through. Usually you find that they all forecast similar weather systems and the difference between them is as much timing as anything else. The forecast gale arrives 12 m hours early or 12 hours lkate but it still arrives.

The Met Office forecasts are the most helpful not only because they arent simple model print outs as many web sites are but becauuse they give an indication of uncertainty . W 3 to 5 is maybe better read as W 4 +/- 1.
 

webcraft

Well-known member
Joined
8 Jul 2001
Messages
39,942
Location
Cyberspace
www.bluemoment.com
.
Friends went from here to L'Orient last night. GFS gave 21kts gusting 29, with ECWMS a couple of knots more. They got sustained 28 gusting 39.

I am drawing no conclusions as a result, just pleased we stayed put. At the moment things are looking a little more stable next week. Two named storms in August is not good.

- W
 

Carib

Active member
Joined
30 Mar 2011
Messages
352
Location
Southampton
www.sailinginlimbo.blogspot.com
None of them purport to give " an accurate precison account" - the movement and state of the atmosphere is beyond that sort of calculation, and thats without taking into account local issues. Your expectations are unrealistic. One site that lucks out on an accurate forecast one day can be the worst forecast the next day.
I'm not expecting them to be more accurate than any other forecast; I really meant that the way the sites present information in a very precise, finely grained way - almost hour to hour - easily leads the user into expecting a higher degree of accuracy than will ever be the case.
 
Top