Cruisers Forum
 


Reply
  This discussion is proudly sponsored by:
Please support our sponsors and let them know you heard about their products on Cruisers Forums. Advertise Here
 
Thread Tools Search this Thread Rate Thread Display Modes
Old 13-06-2024, 11:50   #1
Registered User
 
Indolent58's Avatar

Join Date: Jun 2021
Boat: Beneteau 46
Posts: 61
Weather models vs actual weather

We have been using PredictWind and Windy for passage planning for the last year and a half as well as Chris Parker’s service for longer bluewater passages and have been surprised at how the automated models can often be consistently wrong over a matter of days. Predictwind, for example says they update their model data every 12hrs, but if they have wind direction in a given area off by 90 degrees (which matters a bit) for an entire day wouldn’t you expect them to sync up to or at least get closer to real observations by the next day? How about three days? Nope and nope.

I confess that I know little about how weather models work but don’t they take actual observations as inputs? If predicted vs actual wind and waves are way off over a matter of days or a week I would think that would trigger some response from the models but it doesn’t appear to. Not to pick on PredictWind but they addresse the issue of inaccurate forecasts like this in their FAQ.

“Q: What happens if the forecast is wrong yesterday, should I still trust today's forecast?
A: Yes. The process has no historical memory. Each model run starts with a new input file, and so you are best to put your trust in the PredictWind forecast.”

I would have to disagree. I don’t know what data they put in the new input file but in my experience having the model being wrong on day 1 is a pretty good predictor for it being wrong on day 2. We do continue to use both PredictWind and Windy and they are most definitely better than nothing, but I think they have been oversold to meteorologically impaired folks like me.
Indolent58 is offline   Reply With Quote
Old 13-06-2024, 12:03   #2
Registered User

Join Date: Oct 2011
Location: Northern Puget Sound
Boat: Pacific Seacraft 44
Posts: 75
Re: Weather models vs actual weather

I’m not educated enough to give the OP solid answers, but I’ll share a couple of thoughts.
- I was taught that weather forecasting is more reliable about conditions than it is about timing of those conditions. I.e., a forecast wind shift that does not occur at the forecast time may (likely) still occur at different time. (Of course, your three day issue speaks against this theory).
- Local diurnal conditions may not be reflected in the larger forecast area.
A la carte is offline   Reply With Quote
Old 13-06-2024, 12:14   #3
Registered User

Join Date: Jun 2019
Posts: 2,884
Re: Weather models vs actual weather

Though there are other influences, the single biggest indicator of ocean weather in terms of wind and waves is pressure gradiant between high and low pressure systems. PredictWind doesn't have a great display of isobars, Windy is okay but not great. The US Weather Service surface charts are pretty good readable - the charts at Passage weather.com are very similar and I find helpful.. Point being that just tracking the H/L systems is likely the next level of accuracy for your onboard weather decisions.

Back in the weather fax days, the surface and 500mb charts would scroll in a few times a day. The NWS forecaster signed his/her name to each forecast. It was not unusual to see the forecast shift significantly between two different meteorologists. I recently listened to a presentation on hurricane forecasting by one of the NWS forecasters - he said their cone-of-certainty was designed to be one-sigma of accuracy, meaning about one third of the actual paths were outside the predicted path!

For PredictWind, I find their "gust" wind models more accurate than their regular "wind" model, at least for wind strength.
mvweebles is offline   Reply With Quote
Old 13-06-2024, 12:23   #4
Registered User

Join Date: May 2008
Posts: 3,639
Re: Weather models vs actual weather

I’m assuming your complaint is about wind speed. This is very difficult for the models to forecast accurately. They do better with wind direction and precipitation. I consider a wind speed forecast more than 72 hours out to be of little value.

For same day and next day I find the new HRRR short term model to be significantly better for wind speed and direction. It updates hourly. I read that it uses radar and satellites to get real time wind speed. It’s available in both Windy and PredictWind
CarlF is offline   Reply With Quote
Old 13-06-2024, 15:28   #5
Registered User
 
Kettlewell's Avatar

Join Date: Jul 2006
Location: Massachusetts
Boat: Finnsailer 38
Posts: 5,702
Re: Weather models vs actual weather

It probably depends on where you are in the world too, and which met office covers your area. Personally, I like to use the National Weather Service forecasts whenever possible. I find them to be the most accurate for coastal weather and regions they cover offshore. A good way to supplement those forecasts is to check out the actual reports from towers and buoys at the National Buoy data center. But, forecasters often get things wrong, so expecting perfection is bound to disappoint. I remember one night snug in harbor with a howling northerly blow listening to the weather service forecast all night calling for light southerlys.
__________________
JJKettlewell
"Go small, Go simple, Go now"
Kettlewell is offline   Reply With Quote
Old 13-06-2024, 15:40   #6
Registered User

Join Date: Feb 2017
Location: Occasionally in Colorado. Generally live-aboard. Eastern Caribbean last winter. Nova Scotia and Newfoundland hopefully this summer.
Boat: Antares 44i
Posts: 789
Re: Weather models vs actual weather

Most people don't understand what information the weather models are providing, and become dissatisfied because they are interpreting the model output as something other than what it is. The wind models are not forecasting what the wind will be. What? Yes, that's right.

The models are forecasting what the gradient wind will be. Gradient wind is the wind caused by geographic variations in air pressure and "Coriolis force". Gradient wind is only one component of the wind, and unless you are far from land and there is no significant convection nearby you shouldn't expect the wind to match the model most of the time. If you are on an offshore passage and more than 100 miles from any significant body of land, and there is no recent convection within say 50 miles, then the wind you experience will be predominantly the gradient wind. Otherwise other wind components will change the wind you experience.

Understanding and being able to predict the other components of the wind is the value that a human forecaster, like Chris Parker, adds. You can learn to do this too, though becoming as good as a professional requires a lot of effort. However, with just some effort you can learn to account for land and sea breezes. Depending on the weather these can have a huge effect on the actual (total) wind. You can also learn to predict convective inflows and outflows. Sometimes that is easy but sometimes the closest clouds block your view of what is going on a bit further away and flows travel longer or shorter distances than you expect. There are other wind components, but understanding those two goes a long way towards knowing what to expect beyond the model output.

There are lots of good books on marine weather forecasting. I just reread Reeds Maritime Meteorology by Cornish and Ives. It is dated, but the fundamentals of weather don't change. A good reference book like that can be very helpful.
dougweibel is offline   Reply With Quote
Old 13-06-2024, 16:26   #7
Registered User
 
Thomas1985's Avatar

Join Date: Sep 2012
Boat: Downeaster 38
Posts: 381
Re: Weather models vs actual weather

I went out to sail Saturday in the forecast sunny 10-15

Sat there in dead calm for 4 hours then got pasted by driving rain and 35 kts for 3 hours
Thomas1985 is offline   Reply With Quote
Old 13-06-2024, 16:52   #8
Registered User

Join Date: Mar 2016
Location: San Francisco
Boat: Morgan 382
Posts: 3,127
Re: Weather models vs actual weather

Most people seem to get frustrated that the weather models are wrong because they are trying to sail a route that some software gave them. That will NEVER work. Always sail to the conditions and don't try to follow the routing.

The thing is, if the wind direction is off by only a couple degrees(which would be outstanding), and your polars are off by only a couple percent(again, that would be outstanding), and the effect of the waves isn't accounted for, now all of the sudden you are on the wrong tack, 50 miles from what would actually be "ideal." So, use it as a tool, but also use your experience and what you actually see to make your course decisions.

As for the quality of the prediction itself. Also remember that it is an average of a fairly large area, over many hours. Exactly how large and long depends on the model. But there could be a square that is 100 square miles, and you take the average wind speed and direction over that whole area for 4 hours, and that is the the plotted wind speed. In your software, when you move one hour at a time, or when you zoom in closer, you are looking at a calculated gradient from what the model actually produced between to low resolution points. Because wind can change quite a lot in 4 hours, and because it can be very different only a mile or 2 away, of course it will often be wrong at a very specific spot and time. Again, it is a tool in your arsenal, not a prediction to follow.

What you should be looking at is "is this a good time to leave?", "is it safe?" , "do the models agree, or are they all totally different?" , "is hell going to break loose in a few days?" (the exact time isn't relevant if it is, don't go.)

One last thing, that makes a larger difference than you might expect. Models give wind speed and direction as ground wind, not true wind, and not apparent wind. Whatever speed and direction your instrument measures is apparent. It isn't perfect. Then, using what is usually not a very good STW from a paddle wheel, it calculates true wind, which is actually relative to the movement of the water. So then you would need to apply whatever current there is to calculate the ground wind predicated by the model. There is a LOT of room for error there. I would hope not the 90 degrees you suggest, but 30-45 degrees would not surprise me, that what you are measuring for wind speed and direction is really quite an inaccurate measurement. And given the previous about how the model works and averages, a 90 degree error all the sudden seems like it might be within the total margin of error.
__________________
-Warren
wholybee is offline   Reply With Quote
Old 13-06-2024, 17:24   #9
registered user
 
HankOnthewater's Avatar

Join Date: Dec 2012
Location: back in West Australia
Boat: plastic production boat, suitable for deep blue water ;)
Posts: 1,138
Re: Weather models vs actual weather

Like most sailors, and even armchair sailors, I follow the weather forecasts. And Indolent58 is right, often things to not match: forecast and actual.

Apart for my own sailing I also do forecast for others, sailors and for various events, like film productions. I never use one source only.
I use all the common sources, and subscribed to a few. And then.... like A-la-carte, use isobar charts to check.
Indeed, with a 2-3 day forecast, for instance, we know there is a front coming our way, that is undeniable, just the timing might be out by a few hours or even half a day, or the front might go a little higher or lower (in latitude), but the front is coming!
With 6-8 days out things are a lot more unpredictable, and even the route planning of PredictWind will get it it wrong with so many days out.

What I do is:
- plan for the 10 days out, jot down to important details
- at 8 days out do the same
- then 6 days out
- and then daily
By then I have a fairly good idea, if the front is delayed, or weakened, or if a new path of such front is developing.

I believe weather models should have historical memory. Is that not how AI is learning? And all human do, learn from our mistakes?
I wish I had the time and resources to compare forecasts of the various models or weather bureaus, and really see how accurate (or not) things are. Hmmm, and that would be only valid for one local (my) area.

Of course, none of the computer models take local knowledge into account.
__________________
Wishing you all sunny skies above, clear water below, gentle winds behind and a safe port ahead,
and when coming this way check https://www.cruiserswiki.org/wiki/Albany,_Australia
HankOnthewater is offline   Reply With Quote
Old 13-06-2024, 20:12   #10
Registered User

Join Date: Feb 2013
Location: Oahu, Hawaiian Islands
Boat: Nauticat 43
Posts: 412
Re: Weather models vs actual weather

As wholybee said, sail to the conditions and not to the routing. We have Starlink which can usually enable daily updates of weather. We use Predict Wind, Windy, DWD ICON Global, and GFS for wind, and GFS and Copernicus for waves. Depending upon where you are, you will develop an affinity for some model. But, none of the models will excel. As mvweebles said, you need to track the development and dissipation of highs and lows, too. These chaotic systems wreak havoc on predictions.

In our experience, the predicted wind speed, wind direction, wave heights, and wave directions are garbage after 24 hours. This is true whether you are coastal cruising, island hopping, or bluewater sailing. I've heard some favorable testimonials about weather routing from professionals. Our experience is not so good.

If you are in a flotilla, it will be a great help to get periodic updates from boats in front of you on a similar course. The best way for this to happen is either with direct communication using SSB radio or else with WhatsApp messaging (using Starlink). Otherwise, you should keep a log when underway, and some items to record (notwithstanding wholybee's word of caution) are true wind speed and true wind direction. Besides helping you decide how to tailor your routing, you can adjust your polars to try to get a better match of your boat's modeled performance to reality for conditions you experience.
__________________
"If you don't know where you're going, you might wind up somewhere else." Yogi Berra
Ded reckoner is offline   Reply With Quote
Old 13-06-2024, 21:37   #11
Registered User

Join Date: Mar 2007
Location: Santa Cruz
Boat: SAnta Cruz 27
Posts: 6,888
Re: Weather models vs actual weather

You can get useful real time wind data with the SCAT systems. They are not predictions.


https://manati.star.nesdis.noaa.gov/.../OSCATData.php

Predict wind forecasts are based on the US and European models. I find that the ECMWF and GFS models under predict winds by about 5 knots, but the ECMWF is more accurate. The PWE and PWG models are tweaks of the above models, but not more accurate offshore.

Closer to the US shores, the NAM and HRRR are shorter term models which include land effects and do a better job. However none of the models are good at predicting small eddys.

If I am looking at a coastal delivery, I start watching the GFS and ECMF models 10 days out. They are usually not in agreement, but if they are in agreement on a good weather window 5 days out I have enough confidence to schedule crew to the boat. Further confidence comes if these models are in agreement with the NAM and HRRR 2 days out. These forecasts do not apply for cyclones/hurricanes.

The models are dependent on quality data, and it is hard to come by offshore. Timing of fronts, or crossing a ridge near the center of a big high, or a multiple high situation requires looking out the window.

Weather predictions are solutions to non linear differential equations, and are applied chaos theory. Sometimes you can get a good prediction of what is going to happen, and sometimes you can't.
donradcliffe is offline   Reply With Quote
Old 13-06-2024, 22:38   #12
Registered User
 
Searles's Avatar

Join Date: Feb 2020
Location: Port adelaide south australia
Boat: Cheoy lee perry 48
Posts: 758
Re: Weather models vs actual weather

The weather is wot you have at the time you are there , any thing else Iis someone else’s weather ,learn to love it .⛵️⚓️ .
Searles is offline   Reply With Quote
Old 14-06-2024, 03:26   #13
Registered User

Join Date: Apr 2024
Posts: 13
Re: Weather models vs actual weather

Quote:
Originally Posted by Searles View Post
The weather is wot you have at the time you are there , any thing else Iis someone else’s weather ,learn to love it .⛵️⚓️ .
It's tough to learn to love the wheater that makes you walk on lit charcoal
Henr1y is offline   Reply With Quote
Old 14-06-2024, 03:28   #14
Registered User

Join Date: Jun 2019
Posts: 2,884
Re: Weather models vs actual weather

Quote:
Originally Posted by Indolent58 View Post
“Q: What happens if the forecast is wrong yesterday, should I still trust today's forecast?
A: Yes. The process has no historical memory. Each model run starts with a new input file, and so you are best to put your trust in the PredictWind forecast.”

......I don’t know what data they put in the new input file but in my experience having the model being wrong on day 1 is a pretty good predictor for it being wrong on day 2. We do continue to use both PredictWind and Windy and they are most definitely better than nothing, but I think they have been oversold to meteorologically impaired folks like me.
The OP didn't ask what model is better, but rather why should he believe today's forecast when yesterday's forecast was wrong. I think PredictWind's response is pretty good overall - they have their algorithm that starts fresh each time it's run. It's not iterative.

I wish I had a dime for each time I've heard a sailor say the Euro Model (ECMWF) is more accurate the US Model (GFS). I read an interview of a US NWS forecaster who was asked if this was true. He responded "Do you really think if there was a better model available we wouldn't use it?" For the most part, the folks Ive met who 'model shop' haven't bothered to peek under the covers of weather forecasting to understand what's going on but somehow think that clicking one button over another makes them sound smart about weather.

But to the OP's question - sometimes the weather forces are predictable, sometimes they are not. Example of where the forecast is not reliable right now. Attached are two pics of the Central American isthmus which is where my boat is right now - border of Mexico and Guatemala (both from PassageWeather.com as it was easy to grab). Obviously, each model interprets current data differently let alone the 24-hour forecast. Quite a variance for modern forecasting as there is a significant difference between the ECMWF forecast and GFS. Net result: if one sailor uses one model and the other uses the other forecast model, at least one of them is going to be disappointed. But eventually, the conditions will normalize and the forecasts will be reliable - might be tomorrow which is the underlying point made by PredictWind's FAQ response.

1. Current Surface Pressure Analysis. This is the most current weather analysis available. It's the base data though how the isobars are interpolated between GFS and ECMWF models may differ somewhat. The bigger difference comes in how each algorithm predicts the pressure gradiants will move. The NWS has decent summary wx charts https://www.weather.gov/marine/sepacificbrief. There's a learning curve in digesting them, but suffice to say they see a relatively strong low pressure system over Central America. I'd venture a guess the ECMWF model sees that shifting southward over the next 48-hours and thereby moving the area of stronger winds south by quite a bit. Point being - base data is the same, but how it evolves over time can vary quite a bit.

2. Surface Analysis - Current vs 24-hour forecast; GFS vs ECMWF. The surface conditions will be heavily influenced by the movement of the pressure systems (the NWS link above identifies a "Monsoon Trof" that sweeps across the northern stretch of the Central American isthmus). If you bore down into the underlying data (which I haven't done - decent walk of NWS forecast products for the region HERE), pretty sure the GFS model shows this Trof stays a bit northward of the ECMWF model. Which is right? Time will tell. But clearly, there is significant divergence between the two models so I would not be surprised to see the forecasts vary widely each time they are re-issued.

Bottom line is sometimes the data is clear and the predicted progression solid. Sometimes it can vary and swing. Weather forecasting was one of the major use-cases for super computers.....even led IBM to purchase Weather.com several years ago. In 1991, Hurricane Andrew's devastation of south Florida spurred a massive investment in NWS forecasting and reliability of forecast products to provide better early warning. Computing power improvements have underpinned forecast accuracy though clearly not perfect which the NWS is quick to point out as a warning to mariners that they need to be very careful. I believe it was Hurricane Maria a few years ago that, by the NHC's own admission, the path was wrong by almost 1000 nms. As quantum computing hardware becomes a reality, it will only get better.

Good luck -
Attached Thumbnails
Click image for larger version

Name:	Surface Wx Data.jpg
Views:	12
Size:	60.2 KB
ID:	290923   Click image for larger version

Name:	WX Model Compare.jpg
Views:	15
Size:	479.9 KB
ID:	290924  

__________________
_______________________________________
Cruising our 36-foot trawler from California to Florida
Join our Instagram page @MVWeebles to follow along
mvweebles is offline   Reply With Quote
Old 15-06-2024, 05:50   #15
Registered User
 
Indolent58's Avatar

Join Date: Jun 2021
Boat: Beneteau 46
Posts: 61
Re: Weather models vs actual weather

Thanks for all the responses. They generally confirm my experience that weather model services such as Windy and PredictWind can be useful but are wildly oversold to the average mariner. PredictWind has been heavily marketing their DataHub device which creates a custom polar for your boat based on actual performance which in turn will supposedly lead to better routing. But if model data is really only useful as an input among many to local weather prediction the accuracy of your boat polar seems mostly beside the point when the computer offers up suggested routes. And PredictWind really is suggesting in their marketing that you can take their route planning seriously. We have learned not to and have found Chris Parker’s predictions and routing recommendations to be considerably more useful on average. Caveat emptor.
Indolent58 is offline   Reply With Quote
Reply

Tags
model, weather

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes Rate This Thread
Rate This Thread:

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off


Similar Threads
Thread Thread Starter Forum Replies Last Post
Yachtworld Asking Prices VS Actual Sale Price - or What is the Markup ? talus Dollars & Cents 111 01-05-2019 05:41
Weather models / Gribs for eastern US Coasts mdelatte Weather | Gear, Reports and Resources 6 10-02-2017 06:13
Towing Warps - Actual Test Data? thinwater Seamanship & Boat Handling 60 20-02-2012 18:51
An Actual Shower stall 2divers Monohull Sailboats 62 09-06-2011 22:31

Advertise Here


All times are GMT -7. The time now is 21:57.


Google+
Powered by vBulletin® Version 3.8.8 Beta 1
Copyright ©2000 - 2024, vBulletin Solutions, Inc.
Social Knowledge Networks
Powered by vBulletin® Version 3.8.8 Beta 1
Copyright ©2000 - 2024, vBulletin Solutions, Inc.

ShowCase vBulletin Plugins by Drive Thru Online, Inc.