Gfs vs predictwind, oh and windgusts

[Snowpetrel]

My folks just did a coastal passage. Since they had limited internet I watched the weather closely via all sources. They all pretty well hovered around 25 kn SW. Pwc had the higher winds, nearer 27 knots GFS gave around 24 knots and pwg somewhere between the two. The local weather forcast was for SW 20. In reality they all were right. I guess the award goes to PWC as they reckon they had 30kn for a while, and they aren't all that prone to over exaggeration. But in all honesty its a pretty good result for modern weather info.

I do think the more we as sailers can learn about the different models and the modern weather computers the better.

I am still facinated by the windgust varible and if its any use. I usually predict it like Barnicle does, looking at the synoptic, and the satpicand the surounding topography to guess at it.

At the moment it seems like a pretty deep rabithole of info you could get lost in if you aren't careful. It was so much easier in the old days were a newpaper weatherchart and a voice forcast was about all you had!

Then gribs came along and changed things. But now we have 4 or 5 base worldwide models, and a fair few modified versions of the base models to wade through. One varible I still feel we are missing is an uncertainty varible. It would be nice if the gribs gave a confidence rating, rather than having to watch them for a few days before, and then compare a bunch of different models to get a feel for the levels of error that the conputer is working under.

[barnakiel]

Originally Posted by Snowpetrel

(...)

One varible I still feel we are missing is an uncertainty varible. It would be nice if the gribs gave a confidence rating, rather than having to watch them for a few days before, and then compare a bunch of different models to get a feel for the levels of error that the conputer is working under.

Yes. As for many models watch which one is open water and which one allows for land masses. This is the first step as open water models 'forget' about the islands, coasts and lakes...

Some local agencies seem to twist the gribbed data for that. I believe the AUS BOM and NR YR.NO do just this. Also Polish ICM EDU seems to do better than the gribbed source average. I would guess Victoria Uni NZ does this too. Alas, Spanish AEMET seems to do nothing but use US data and plot it graphically for the masses. Etc.

Now why confidence intervals are not given I believe may be due to a plain banana methodology there. Maybe if we just ask NOAA in an email, they could show us how to calculate this, But I believe this will be a simple variation / bell curve thing. If we have the distribution of the model and the var/st dev, then we know the confidence/probability.

But I never asked and rather use a simple 75/50/25 rule for the next days. Not that I am not interested, but my sailing today does not ask for more. I bet I would love to get the data if sailing a long coastal or offshore race.

b.

[Snowpetrel]

Yes Australia's BOM Meteye program"twists" the wind data to better reflect the coastal conditions, land shadows and heat effects. It seems to work quite well, even often getting seabreezes right. Still no prediction for gusts. You need to work that one out yourself. Also local storms/squalls arent modeled by meteye. But they give warmings on the voice forcast. Fronts ocassionally caych them out.

The interesting thing with aus BOM is that they use a variety of base models, selecting those that seem best, so you dont know if its based on GFS or some other model.

Each state has their own system, and you see some weird discontinuities at the state boundarys at times, thats always a warning to me that their is uncertainties somewhere.

Anyway its a worldclass system that is hard to beat yourself. BOM refines meteye first, the automatically generates the voice forcasts from it.

My recent trip to NZ made me realise how lucky we are to have BOM in aussie. The NZ metservice had nowhere near as detailed information.

[barnakiel]

Originally Posted by Snowpetrel

(...)

My recent trip to NZ made me realise how lucky we are to have BOM in aussie. The NZ metservice had nowhere near as detailed information.

Well, you know; look at AUS size and NZ size, and look at respective GDP.

By all standards the Kiwis are doing plenty off outstanding work on the wx front. E.g. here:

New Zealand Weather Forecasts

Cheers,
b.

[Snowpetrel]

True enough about the size difference between the countries. But I really missed the observation pages with the listed 30 minute obs data from all the coastal stations.
Funny thing I found most usefull about predictwind was the really good realtime weather observation graphs. I was unable to find the info anywhere else online. Maybe PW buys access from the NZ Metservice? Anyway most useful, see the pic, though this function is crap for aussie on PW. Go figure.

All this just made me really appreciate what we have over here for free on one website.

I remember when metvuw first came on the scene. That was before we had easy access to gribs and it was awesome. But isnt it just repackaged GFS data? Or is there more to it than this?

[Emerald Sea]

Yes, Ive recorded daily observations for months during our crossings of actual wx conditions vs PredictWinds four models and the PWC (Canadian weather model modified by PW) came out as the more accurate model. Not all the time but most often. This is general and does not include any isolated or local events ie squalls, etc. As expected, once in coastal waters the accuracy diminishes and we typically go over to, say Windguru or another local forecast site.

I find the sat images are very handy but they are only available for that day and not an outlook. I understand PW may be changing this. But by observing the previous days with todays you can form a prognosis of whats expected in your area. This was quite useful in the ITCZ where it assisted me in choosing a route with minimal disturbance.

[Snowpetrel]

Interesting info Emerald, thanks for that. I must say I've never really used the canadian models but its great to hear they are viable competition to the GFS model. Obvously PW must modify the at sea forcasts as well. I just assumed they added a local land effect modification? Anyway good info. I had also just thought windguru was another repackaged version of GFS or something, but it seems like its more technical than that http://www.windguru.cz/int/help_index.php?sec=models

http://www.windguru.cz/int/help_index.php?sec=windmod

One way I like to use the satpics and cloud info is to use it to predict changes in wind strength. Ie if the gribs show a band of cloud clearing with the front, then the SW building after the skies clear I know that a clearing sky will likely bring the sw change and rising wind. This plus the baro trends helps me relate the gribs to the observed weather as it unfolds. When I sailed with a skyeye satpix receiver this was brilliant as I could overlay the gribs onto the satpic and adjust the gribs position and timing to suit the satpics.

Often you could see the lows on the satpic, and the fronts. Even wind direction from the lee waves on the mountains . This gave a good reality check on the gribs.

[Emerald Sea]

Jon at PredictWind is quite approachable and can be contacted (on his web email) where he can explain how they modify the PW forecasts. Id rather not re-explain as its a complicated process and Im likely to get it wrong!
Steve

[Paul L]

Originally Posted by Emerald Sea

Jon at PredictWind is quite approachable and can be contacted (on his web email) where he can explain how they modify the PW forecasts. Id rather not re-explain as its a complicated process and Im likely to get it wrong!
Steve

Too bad the description isn't on their website.

[Paul L]

Originally Posted by Emerald Sea

Yes, Ive recorded daily observations for months during our crossings of actual wx conditions vs PredictWinds four models and the PWC (Canadian weather model modified by PW) came out as the more accurate model. Not all the time but most often. ......

Is this over a particular area or in many areas/oceans.
It is so much easier to interpret GRIBs, WeFaxs and weather routers after you've already dealt with the weather in the same season. That's the hard part about cruising to a new area. You don't have that local knowledge/experience to appropriately weigh the different weather tools available.