Gust and wind speeds on Windy

[desodave]

I've just started to use Windy and am confused by the huge difference between wind speeds and gust speeds for the area I'm looking at - specifically Johnstone Strait in BC - and also in other parts of the Salish Sea. I'm looking at the Fanny Island weather station area and Windy shows wind speeds not exceeding 6 knots over the next 5 days, yet forecast gusting as high as 27 knots at the same time. Doesn't seem realistic to me. And a Canadian marine forecast of a gale warning tomorrow - 30-40 knots - at the same time.


I don't understand how the predicted wind measurements can be so low and the gusts so high. I'm used to seeing wind speeds of 20 associated with gusts of 27 in this PNW area.


I've compared measurements at other stations in the area and have found Windy's gust speeds to be more representative of the actual wind speeds recorded by the stations than Windy wind speeds.


Can anyone explain to me why their predicted wind speeds appear so low?

[JDWS]

Dave, I’ve noticed the same. A week or two ago near Lummi Island Windy showed 6 kt all afternoon gusting to 15. We spent the afternoon reefed, steady at 16, gusting to 22. I wrote it off as forecasting error and switched to Wind Alert (and had a blast however, Windy didn’t really ever update for actual conditions

[gamayun]

I sail San Francisco Bay and have never found Windy to be accurate to that micro scale where topography, day time heating of land, etc. are more likely to influence sea surface wind flow patterns.

[Paul Elliott]

It's not a Windy.com issue so much, but a model issue. None of the available models do local microclimates very well, but some have more resolution and possibly better accuracy closer to land. The GFS22km model and the NAM5km model are often worlds apart in the Salish Sea area. The few times I've checked, the NAM5km model was a better match to the Cape Flattery buoy wind speed report, but that could have been just luck.

[desodave]

Thanks everybody for your replies. Likely expecting too much detail on a micro level from a macro model.

[mikedefieslife]

Presumably ECMWF is the one to go for. The resolution is finer than GFS at 9km, and the model is nonhydrostatic so does take into account topography (in terms of altitude) rather than merely basing itself on pressure.

[sailorboy1]

NOAA doesn't get wind predictions right, why would I expect anyone else to?

[GregW1309]

As previously mentioned, almost all the prediction models ignore local topography, they are ok offshore but not much use for coastal & inland waters. Also, weather reports consistently are biased to the worste case. To understand wind in the Salish Sea get this book, it is amazing:
Taken By the Wind: The Northwest Coast: A Guide to Sailing the Coasts of British Columbia and Southeast Alaska
Johnson, Marilyn



Two others are:

Marine weather hazards manual: A guide to local forecasts and conditions
Environment Canad

The Wind Came All Ways: A Quest to Understand the Winds, Waves, and Weather in the Georgia Basin
Lange, Owen S

Books are available at Amazon.


The only weather forecastes that do take into account the local topography are available from the Northwest Modeling Consortium. It helps if you are a meterologist...as there site is, to say the least, Geeky.
Here are some links:
https://atmos.washington.edu/mm5rt/mm5naminfo.html
Example:
https://atmos.washington.edu/~ovens/...d3_wa_slp+///3

Listing of all the models results:

https://atmos.washington.edu/wrfrt/gfsinit.html

[eyschulman]

I live and boat in PNW and do look at windy and other sources of weather information. I find no one source is fully appropriate and have come to rely on a rough average of predicted conditions. I want some rough idea if I will find quiet or windy conditions direction of wind and time of day or night of expected conditions. Depending on accuracy beyond that has been disappointing throughout most of PNW.

[CAPAZ]

we just sailed from Seattle to LA , using predict wind vs windy when we could get it. What I observed is when we were 50-60 mi offshore they were both good. Near shore it fell apart quite often. I have heard that somehow the data progs don't really know that just off the water is land (as if?)

anyway just my recent experience with both, would still use in the right way, I would use either/both again

rocco

[FabioC]

I sail in the Puget Sound and observed the same as the OP many times.

I think that the macro/micro climate explanation leading to inaccuracy in the model, while true, should be easily solved as part of the model. A "good" model should have validations in it to highlight forecasts that are clearly inaccurate, like in the scenario of this thread.

Even if the model does not have those validations, I think it is not unreasonable to expect that a service like Windy (especially the paid version) would adjust the forecast "a posteriori", rather than simply reporting blindly out of the model.

A related aspect that I find baffling with these models is the apparent lack of feedback in the forecast from current conditions.

For example, the forecast for 2:00pm, 5:00pm, 8:00pm, etc. shows, say, 5 knots from NE, gusting to 7. At 2:00pm, the real conditions are totally different, say, it blows 20+ from SW (it happened like this only a few days ago). The 5:00pm, 8:00pm, and later forecasts continue to show the 5 knots from NE.

Clearly, the inputs to the model never (or only rarely) get updated... Wouldn't a frequent feedback loop with input from "real" conditions (they have the data) be an obvious, easy way to improve short term forecast?

Finally, I do not understand why these models appear to be "stubborn". For example, say that the actual general weather conditions are relatively constant for 10 days. Say, every day is sunny, with calm winds in the early afternoon, and 10-15 knots coming up at 6:00pm, until 8:00pm, then back to calm until midnight. If the model misses this pattern on the first day, it will miss it in the same way every day for the next 10 days... Why do these models never learn? We are in the era of Artificial Intelligence, and models that we build for other things are able to learn pretty fast in very convoluted use cases, but not these...

[Paul Elliott]

Originally Posted by FabioC

We are in the era of Artificial Intelligence, and models that we build for other things are able to learn pretty fast in very convoluted use cases, but not these...

I think you underestimate just how tough this problem is. GFS and the other models use the largest and fastest computers out there, use very complex math, and are updated every six hours using all available real-time data. However, they are attempting to model and predict a *huge* global system, from the surface up to the troposphere (?), and the spatial resolution for GFS is 13km -- not exactly good enough for local microclimates.

Windy and other similar services go a great job of post-processing the model outputs, but they aren't going to be implementing their own supercomputer modeling system.

[westwon]

We race on English Bay every 2 weeks, and tried to use apps like Windy, but realized it's just not accurate for anything more than rough wind direction while sailing in enclosed waters. Fun to play with though!


Steve

[Kenomac]

It happens all the time in Croatia before or following a Bora. Wind speed 5-8 knots with gusts up to 30 knots and it’s very real.

[surazo]

please remember that all the models ,NOAA ,european and others are based in a extrapolation matrix based in some true points ,consequently the accuracy depends of the mathematical model used ,the number of data and the area predicted .In general ,according my experience running different model is that normally work well as a general tendency ,in particular when they have some degree of coincidence, but for local conditions ,ie 2-3 miles around may be a good guess .Barometric pressure is a good local indicator