Reliability of Wind Predictions

[viben]

I guess each weather forecaster has their strength and weaknesses, as shown in this screen grab from PredictWind. The 4 models displayed in PW indicate that we can expect anywhere between 15 and 49 knots gusts after midnight tonight. So, we are hoping for the best while preparing for the worst.

Any rational opinions, based on actual use and experience, about the various wind forecasts, models and apps available out there?

[Suijin]

The accuracy of wind predictions are pretty much related to the bigger picture of the weather systems in the forecast. If you have closely spaces high and lows, and pretty closely space pressure gradients, the prediction is not worth a hill of beans. If you have more settled weather with slower moving systems, then the accuracy goes up.

That's one reason why I still look at the NOAA surface chart forecasts. Not only do you get a visual snapshot of what's going on but you get a sense from experience of the forecaster's confidence in their forecasts.

That said, I pretty much never head out unprepared for the forecast to be absolutely wrong, based on the experience of it having been absolutely wrong a lot.

[conachair]

Originally Posted by viben

Any rational opinions, based on actual use and experience, about the various wind forecasts, models and apps available out there?

Yes, no one here knows much at all about the models' accuracy, IMHO.

To get even a hint of accuracy is a load of work and also us mere mortals don't really have the data available (or the time/skills? ) to get some kind of handle on the different models accuracy wherever you happen to be.

I've just signed up for another year with predictwind, though don't rate their own 2 models much, but the display option are excellent so it's easy to see just how much the different models agree.

Many will argue that "My seaguru site is better than windweed" without even realising most free sites are looking at the same numbers.

Personally GFS is favoured onboard, but a few days out anything can happen.

[robert sailor]

We were in the West End in Roatan Honduras and the European grib files were calling for strong N to NW winds, we prepared to leave prior to that as the anchorage is open to wind and seas from those directions but a friend gave me a call and said they were staying because Predictwind was calling for winds to remain easterly. I had been watching the weather for a couple of days and 2 days before the low was expected I noticed a halo around the sun which was always a prediction of an incoming low but my friend checked his fancy Predictwind and still easterly winds. We advised them a day before that we were leaving and surprise surprise he said they were leaving as well because just a day before Predictwind changed their forecast. All of this simply means that forecasts are just that and you can't bet the farm on them especially in the longer term. Generally good info but be prepared for changes.

[Captain Bill]

That's an interesting forecast. I checked several other sources and none of them are predicting that much wind in New Bern. Some are predicting gusts into the 40's off the coast. That being said Towndock.net in Oriental is reporting gusts to 35 already and no one is predicting that at this hour. I guess you'll have to let us know how it turned out tomorrow. Plan for the worst, hope for the best. If you're in the Trent river in New Bern you are fairly well sheltered, especially if you're in one of the marinas between the bridges.

[El Pinguino]

Predictwinds product is still just gift wrapped gribs.

The warning which comes with gribs via Sailmail is still relevant... in part...

'Also remember that grib data is not reviewed by forecasters before being made available. You are getting a small part of the raw model data that the forecasters themselves use when writing a forecast, and it is your responsibility to make sure that the data is consistent with your local conditions and with the professionally-generated forecasts (e.g. text bulletins and weather-fax charts).
Grib data also has limitations along shore, where local effects often dominate and may not be adequately modeled. In addition these models cannot provide adequate prediction for tropical systems, frontal activity or convergence zones. '

[Paul L]

If you are comparing models then comparing winds at the same hour isn't a great comparison. While the models will have different wind speeds, they more importantly have the front passings occuring at different times. If you look at the wind peaks of the front passage and compare those peaks the difference is a lot less - although that 49 really seems like an outlier.

[viben]

Originally Posted by Captain Bill

That's an interesting forecast. I checked several other sources and none of them are predicting that much wind in New Bern. Some are predicting gusts into the 40's off the coast. That being said Towndock.net in Oriental is reporting gusts to 35 already and no one is predicting that at this hour. I guess you'll have to let us know how it turned out tomorrow. Plan for the worst, hope for the best. If you're in the Trent river in New Bern you are fairly well sheltered, especially if you're in one of the marinas between the bridges.

Reality at 10:45pm: Hardly any significant wind. The gusts make the wet flags move leasurely. Nevertheless, PredictWind​ still forecasts gusts between 17 (GFS) and (45 PWG) for 11:00pm. Still before the storm?

Yet, the afternoon and early evening predictions were more accurate. We did experience gusts above 30 knots.

[jkleins]

I have been watching this all winter along the East coast of US since we are a little big to do the ICW comfortably we go out a lot. It seems that PredictWind will almost always over call the wind. They probably work on a a "worst case" algorithm or something but we would hardly ever travel if we just had the Predict Wind forecast.
The GFS seems to be be the closest to reality but it always feels nice when all the models agree.
While I hear people talk about how the GRIB's are not as good as human forecasts I haven't noticed that much difference between the GFS model forecast and the NOAA Marine forecasts which I have always assumed had more human interaction. Maybe they are better for this area?

Jim

[barnakiel]

They are just mathematical models taking care of data. They are getting slowly better at that too. You can easily calculate predictability from such models and if you lived long enough you would find the predictability figure to be correct.

Some models seem to be better in some LOCATIONS but no model seems to be better overall.

Watch result sets over a year or two in any specific area and you will develop a sense of who 'knows' and who 'thinks he knows'.

Surprisingly, local offices (e.g. Spanish AEMET in Canary Islands) are NOT the best. I use yr.no (they are Norwegian guys) here with much better hit / miss score.

Beware some providers provide nothing but further splitting or manipulation of otherwise publicly funded and publicly available data. These guys tend to do much fuss while there is in fact no added value in what they sell.

Look for long-term repetitively accurate guesses. They exist.

b.

[fallingeggs]

Originally Posted by robert sailor

We were in the West End in Roatan Honduras and the European grib files were calling for strong N to NW winds [...]

For what it is worth, PredictWind now (as of late 2016, at least) includes a European model... though not sure if it "the" European model. PredictWind calls it the "European Centre for Medium-Range Weather Forecasts." I am curious if it uses the same data as the grib you refer to?

[viben]

The wind definitely increased again after the calm period an hour or two before and after midnight, with gusts maybe reaching low 30s, judged by sounds and movements of the boat, as observed from my comfortable v-berth.

Forecasting is not an exact science... That meteorologists may come up with different predictions based on the same data is understandable, but it never ceases to astonish me that they can be diametrically different, and also completely wrong, even within a timeframe of only a couple of hours. Longer term forecasting is a completely different story....

As long as the experts can't get it right, I don't expect that I will be any better at interpreting the data. Hence, maybe a better idea would be to increase the number of forecast sources (PW presents 4) and simply look for differences and commonalities? If they all align, the odds are that it is reliable. If they are all over the place, anything could happen...

Anyone having compiled an extensive list of marine weather forecast apps and sites?

[roland stockham]

When comparing accuracy for grib files you need to be sure that they both display the same cell size and location. if one is based on say a 1deg cell and another on 5deg you will see different results. Also remember that the predicted wind is the strongest fore cast in that cell so can be 50miles offshore or only or only at the local cape. Grib files are by nature less accurate than weather forecasts but can be useful. Main issue I have is they don't show fronts. Offshore Wfax is way more accurate and useful and more reliable to get.

[conachair]

Lots of discussion but no one, it seems, has even put a metaphorical pen to paper to log any forecast / actual data.

Possibly a thread as much about human bias and inability to be objective as much as anything else

[Tar34]

I have had excellent results with Sail Flow
https://dl.dropboxusercontent.com/u/...2041%2018.jpeg