Weather Forecast Models

[tpbrady]

While at anchor in a small nook in Gambier Bay waiting out winds in Stephens passage, I started comparing the weather forecast models prediction for tomorrow. The range of winds at 8 am at a point I need to go around before I can head north, is from 5 (GFS) to 21 (HRRR), with other models clustered between 12 and 15. When I look at the forecast winds at locations where I have current observation sites, all the models understated the winds by 3-5 knots in the south and overestimated the winds in the north by the same.

All the models converged toward the observation in their forecast. Based on this, it appears the NWS forecast is more or less based on the scoring system used in ice skating, throw out the low and high score and average the rest.

Anyone have any insight into how the NWS marine forecasts are developed?

Tom

PS. Thanks for having Starlink, otherwise, I could not see the current observations and with the Cape Fanshaw weather radio down, I would be in the same condition as last year. The only way to get weather was to pull the anchor and go out and look.

 

[O C Diver]

Anyone have any insight into how the NWS marine forecasts are developed?

Ouija Board.

Ted

 

[Don L]

The only time it seems the forecast is right is when it predicts bad weather. After that adding 5 to the wind forecast normally seems to work.

I crossed from the Bahamas back to Florida in April and all 4 predictions plus Chris Parker were wrong and it SUCKED!

 

[sledge]

Tom,
I know you're on the opposite coast but yes its a struggle to pick the best model. I can tell you in the SE US coast in the past year, we seem to have the closest match by using the GFS model, but no guarantees its still kind of a crap shoot....

 

[Bacchus]

My rule of thumb and recommendation to other boaters is to look at and consider the various models but NEVER try to puck one and rely on it.
Don't take a chance on a single or multi midels... consider that any are possible so consider worst case NOT best case.
I've taught the USPS Wx course for 25 yrs.
I encourage boaters to ask themselves if they are comfortable living when CONSIDERING ALL the possibilities?

 

[OldDan1943]

Models are just that, best guess.
I have yet to to meet a ‘weather man’ willing to hang his hat on a prediction made more than 1 day and then it is too late to move your boat any meaningful distance.

 

[mvweebles]

I don't know the OPs area, but sounds like a very small area where local conditions could greatly influence variations to the macro forecast. There is still a place for local knowledge.

I realize it's fashionable to reference various models and highlight when they're wrong. But more often than not these days, folks are actually referencing an app that has an algorithm that autonomously interprets the weather data from a source such as GFS.

I realize NOAA surface charts and synoptic charts can be a bit boring and perhaps intimidating. But they have a ton of information and give an accurate view of the weather forecast - its a relatively macro view that covers a lot of large expanse. Asking for micro-interpolation by an algorithm is a nice idea, but please don't complain when it changes.

What I really like about the NOAA surface charts is that each one is digitally signed by an experienced meteorologist.

Sure forecasts change. But they change for a reason. If you're not tracking pressure systems, you are at risk of being informed too late that a forecast has changed.

Peter

 

[LovetoBoat]

So two things that have worked for me.
1) I have the buoys from Florida to Nova Scotia on my PC and I watch them as I approach the areas - wind speed and direction - wave heights and direction for the last 6 hours to see if they are growing or shrinking.
2) I then look at the forecast and ask if the forecast is off by 10 knots in the wrong direction would I want to be out there?

If the answer is no - I stay where I am and wait...
Stay Safe.

 

[fryedaze]

Ouija Board.



Ted

[emoji106][emoji106][emoji106][emoji3060][emoji106][emoji106]
I have been burnt so many times by the weather magicians. I just did the Albemarle sound with a 15 mph north predicted wind. It was 20-24. My 14 ton 42' boat was half out of the water several times until I powered up and got the nose and speed up. We should dock these guys pay.

 

[Hippocampus]

Once again agree with peter. Synoptics are definitely worth looking at and would take one step further and look at the 500mb if doing anything near or off shore of any significance. Have always liked Chris if dealing with US east coast and eastern Caribbean but commanders for elsewhere. Totally agree with looking at all the models. Plan for the worst outlook among the models. Overall my impression is the European model is the most reliable in general.
Did the Lee Chesneau course and afterwards realized how little I know but still think it’s your boat and your life so it’s worthwhile to independently review as many sources as you can. To judge a particular source at a particular time having a baseline of knowledge even if like me it’s limited is worthwhile. You then can sense those times when predictions will be less reliable. It’s rarely a mistake to stay put if you’re nervous and remember schedules kill. Still use a recording barometer (meteoman) and still don’t trust GRIB files. They’re ok if looking at a large enough field of view but as a boater all weather is local so they can be very misleading. For US coastal find NWS surprisingly good.
Don’t understand why people pay for GRIB files when they’re available or free. And don’t understand why people only look at them.

 

[Retriever]

I’ve found the “gust” predictions far more accurate. If I’m happy with the gust forecast on the worst model, I’ll probably go.

I have much more faith in the models when they are in agreement about timing and trends.

I’m writing this on passage between Hokkaido, Japan and Attu at the end of the Aleutians. This is one of the few passages I’ve been on where the GFS gust model has often predicted stronger wind than the ECMWF model.

In both cases the “gust” forecast has, so far, overestimated wind speed and the “wind” model has underestimated wind speed. They’ve been spot on with the wind and sea trends, but not very good with current velocity or direction predictions. It will be interesting to see how they do with local effects once we get up to the Aleutians.

 

[tpbrady]

Sam,

Have a good trip to Attu. I’ve been there once and Shemya numerous times. I look forward to seeing details of your trip.

I happen to be rereading 1000 Mile War, a history of WWII in the Aleutians. If you haven’t read it, get it for Kindle because it will put context to places you are going to see and hear their impressions of weather. The only inhabited islands I am aware of are Shemya (USAF, don’t go there), Atka, and Adak. There might be a few people at Nikolski.

Tom

 

[HeadedToTexas]

Ouija Board.

Ted

That’s my read as well.

 

[sledge]

For the East and the gulf states at least, NASA and science continues to attempt to improve prediction tools. I found this news item interesting, a new set of satellites launched for this problem and are apparently operational just in time for the 2023 Hurricane season...
Includes an incredibly long acronym so it must be good, Ha!

TROPICS (Time-Resolved Observations of Precipitation structure and storm Intensity with a Constellation of Smallsats)

https://www.nasa.gov/press-release/nasa-rocket-lab-complete-launch-of-tropics-cubesat-constellation

 

[OldDan1943]

Ah yes, we look at the weather prediction models and then finally, look out the window. HAHA

I will say, seldom does a major weather event sneak up on us.

 

[Bacchus]

Frequently, ( not always) the difference between models is either timing or location to location variation. Models cover a fairly wide ( it varies) area and different areas within the larger may see local conditions different that other areas. All that leads individuals to think fcsts are never correct... when they are correct God a different nearby area or at a different time than anticipated.
Local terrain/conditions can have a significant affect in coastal & inland areas... not so much off shore.

 

[DDW]

Often there is a forecast discussion page associated with the forecast, though it may be hard to find. In it is a narrative description of how the particular forecaster arrived at what he/she wrote. We have this for NOAA aviation forecasts for example. It can be quite illuminating. They may say they looked at this and that, discounted the other, but feel that what they wrote is most likely because.... One thing about weather is you need to embrace uncertainty. If I look at 5 models and they all say the same thing, then I'm pretty confident in what I will find. If they all say something different, they you need to be ready for anything.

I'm in somewhat the same position, departing south from Craig intending to go around Cape Chacon and run the Dixon west to east in the next few days. Several models seem pretty consistent for that period, but 5 or 6' waves don't sound appealing.

 

[murofrelwart]

Often there is a forecast discussion page associated with the forecast, though it may be hard to find. In it is a narrative description of how the particular forecaster arrived at what he/she wrote. We have this for NOAA aviation forecasts for example. It can be quite illuminating. They may say they looked at this and that, discounted the other, but feel that what they wrote is most likely because.... One thing about weather is you need to embrace uncertainty. If I look at 5 models and they all say the same thing, then I'm pretty confident in what I will find. If they all say something different, they you need to be ready for anything.

I'm in somewhat the same position, departing south from Craig intending to go around Cape Chacon and run the Dixon west to east in the next few days. Several models seem pretty consistent for that period, but 5 or 6' waves don't sound appealing.

Link to forecast discussion for SE Alaska. Today's discussion doesn't go into as much depth of discussion of the various models as I've seen in the past.



https://www.weather.gov/aawu/afd

 

[Mike H]

You have to remember your location. Having previously lived in Alaska with the numerous mountains, valleys, and terrain that wanders this way and that, weather can change rapidly simply by a small shift in wind direction. I was once in Dutch Harbor/Unalaska waiting to give a presentation on wind energy to the City Council. Sitting in my hotel room, I could see the ripples on Margaret's Bay traveling one way. A few seconds later, the ripples are going the complete opposite direction. The overall wind did not shift that much, but a few degrees difference can mean going left or right along a valley sitting cross to the wind. Hard to put up a wind turbine there.

After moving to Maine, it was pure pleasure actually being able to plan a few days in advance.

 

[Jc180]

Have spent considerable time over the past number of years looking at the weather model forecasts for the SE US, Florida, and the Bahamas. The two prominent models are the ECMWF (European model) and GFS (model used by US NWS). Agree with others who have said best when the models agree. Also agree with others who have said actual wind speed and direction best described in gust forecast. Obviously have no idea about the actual algorithms in each model, but have observed that the ECMWF more accurate more often. Like the Windy App. You can easily select the forecast for each model, compare, and then develop cruising strategy.

 

[Jc180]

Should have also mentioned that local affects often overide the model forecast. The obvious onshore afternoon sea breeze is captured in the models, but other unique geographical features impacting wind are more sophisticated than the models.

 

[mvweebles]

Have spent considerable time over the past number of years looking at the weather model forecasts for the SE US, Florida, and the Bahamas. The two prominent models are the ECMWF (European model) and GFS (model used by US NWS). Agree with others who have said best when the models agree. Also agree with others who have said actual wind speed and direction best described in gust forecast. Obviously have no idea about the actual algorithms in each model, but have observed that the ECMWF more accurate more often. Like the Windy App. You can easily select the forecast for each model, compare, and then develop cruising strategy.

Here's the part I just don't understand. Can someone point me to the base-data for the European Model? Here are the synoptic charts for GFS (NOAA/NWS) station New Orleans for the GoM/Atlantic. Where is the base data for ECMWF forecasts, and does anyone have any idea why it would be more accurate? Is the European model more accurate; or does their algorithm better accept the European data? Or maybe Windy/PredictWind accepts raw GFS data before live-human (such as Lee Chesneau - RIP) intervention to tweak the forecasts? NOAA sends up weather balloons and takes-in real-time weather readings from hundreds of buoys and thousands of ships at sea. Where does the data from Euro models come from? How is the data crunched?

I know it's sort of sexy to talk about GFS vs Euro, but what exactly does that mean, and why would one be more accurate than the other? I use Windy a lot but for longer passage planning, it just doesn't cut it. I want to know the base data. Too much at risk.

Thoughts?

Peter

 

[Jc180]

Here's the part I just don't understand. Can someone point me to the base-data for the European Model? Here are the synoptic charts for GFS (NOAA/NWS) station New Orleans for the GoM/Atlantic. Where are is the base data for ECMWF forecasts, and does anyone have any idea why it would be more accurate?

My sense is it's popular to talk about "GFS" and "European" models because there are two different buttons to push on Windy, PredictWind, etc. Saying one button is more accurate than another is not that useful - it's a lagging indicator . Weather forecasting is most useful as a leading indicator. To give yourself the most time to make meaningful decisions, best to understand the base data. Looking to understand where that is for European Model, not just the outcomes or some algorithmic interpretation. I find the GFS information very accurate as long as I can see the synoptic charts with 'adult supervision' by NWS forecasters. Does similar Euro data exist? Is the issue the Euro model is more accurate, or is it that the GFS is more accurate but the Windy algorithms screw it up more often?

GFS base data and human intervention is easily accessible. Where do I find similar for European?

Thoughts?

Peter

Peter, All good questions. First, do not believe Windy App influences or modifies model results. Second, I have no idea of the basis of either the GFS or ECMWF models. Just reporting my observations over many years. Having said that, don’t want to mislead, GFS is a good model, and can be more accurate than ECMWF on any given day or time frame.

 

[mvweebles]

Peter, All good questions. First, do not believe Windy App influences or modifies model results. Second, I have no idea of the basis of either the GFS or ECMWF models. Just reporting my observations over many years. Having said that, don’t want to mislead, GFS is a good model, and can be more accurate than ECMWF on any given day or time frame.

I chose your quote because its a very common sentiment. I don't think you're right or wrong, just I hear it a lot. I've said it myself. But I have to question it.

The NWS/NOAA has the budget to derive a LOT of data. From ships at sea; from hundreds of buoys. Yet it's popular to say the forecast model isn't as good as the Euro model. Why would that be? Maybe it doesn't make a difference - if one is more accurate, who cares why? But you have to ask yourself - is the Euro model always more accurate? Or is it just the Hurricane Spaghetti Model thats more accurate? Or is it close-coastal that's more accurate? What exactly is more accurate and why? For me, heading 100's of miles, I want to know what is influencing the weather.

Each NWS/NOAA synoptic chart has the digital "John Hancock" of a career meteoroligist on it. There are less than ten of them so it's pretty easy to follow each. When one goes off-shift and the next comes on, the forecast changes, sometimes a lot. You could say that's not a good thing for a forecast and I agree, but my take is it's variable. If it changes a lot, that's meaningful. If the forecast between shifts doesn't change much that's meaningful too.

Just trying to probe a bit deeper. I was in a Rip Van Winkle boating slumber for 10+ years - access to weather products was a primary reason I joined TF. It's independence.

Peter

 

[Jc180]

So this thread has exposed the dilemma those of us who cruise face. It is all about the weather. But the weather forecasts are based on models that disagree and have inherent inaccuracies. So what to do? Think the best we can do is to understand the track record of the models, but more importantly, the limitations. And then be patient in waiting for favorable weather before setting out in open water.

 

[Simi 60]

Models are just that, best guess.
I have yet to to meet a ‘weather man’ willing to hang his hat on a prediction made more than 1 day and then it is too late to move your boat any meaningful distance.

Back in the day (90's)there used to be a weather report on one section of QLD coastline put out by Rocky Met who was the Rockhampton office of the Australian Bureau of Meteorology.(BOM)

They started out by reading the official BOM weather and then did their own version and at the end of report asked for observations from boats out there and....many of us did
From that collection of "live" data, they did actually get to be very accurate and we'd always be glad entering their area knowing we could pretty accurately plan several days out.

Best weather service ever but, like all good things it came to an end, govt probably needed to buy some out of date phone system or weaponry system from the US

They were to my knowledge the only office to do so, everything else came from BOM in the capitals and was read out verbatim by volunteer coast guard up and down the coast.

 

[Jc180]

I chose your quote because its a very common sentiment. I don't think you're right or wrong, just I hear it a lot. I've said it myself. But I have to question it.

The NWS/NOAA has the budget to derive a LOT of data. From ships at sea; from hundreds of buoys. Yet it's popular to say the forecast model isn't as good as the Euro model. Why would that be? Maybe it doesn't make a difference - if one is more accurate, who cares why? But you have to ask yourself - is the Euro model always more accurate? Or is it just the Hurricane Spaghetti Model thats more accurate? Or is it close-coastal that's more accurate? What exactly is more accurate and why? For me, heading 100's of miles, I want to know what is influencing the weather.

Each NWS/NOAA synoptic chart has the digital "John Hancock" of a career meteoroligist on it. There are less than ten of them so it's pretty easy to follow each. When one goes off-shift and the next comes on, the forecast changes, sometimes a lot. You could say that's not a good thing for a forecast and I agree, but my take is it's variable. If it changes a lot, that's meaningful. If the forecast between shifts doesn't change much that's meaningful too.

Just trying to probe a bit deeper. I was in a Rip Van Winkle boating slumber for 10+ years - access to weather products was a primary reason I joined TF. It's independence.

Peter

Peter, not sure that I share your endearment with the NWS/NOAA meteorologists. Go back and take a look if you can find the data at Hurricane Ian. NWS meteorologists were for many days forecasting land fall from the Florida panhandle/big bend to Tampa Bay and only at the last minute so to speak resolved to Ft Myers area. ECMWF was showing Ft Myers all along with eye passing over New Smyrna Beach as it exited Florida to the east back into the Atlantic which is exactly what happened. I vividly recall because my motorsailer is in New Smyrna Beach.

 

[mvweebles]

Peter, not sure that I share your endearment with the NWS/NOAA meteorologists. Go back and take a look if you can find the data at Hurricane Ian. NWS meteorologists were for many days forecasting land fall from the Florida panhandle/big bend to Tampa Bay and only at the last minute so to speak resolved to Ft Myers area. ECMWF was showing Ft Myers all along with eye passing over New Smyrna Beach as it exited Florida to the east back into the Atlantic which is exactly what happened. I vividly recall because my motorsailer is in New Smyrna Beach.

I live in Madeira Beach which had a bullseye on it for Ian. Was the reason I dashed from Ensenada to Florida.

The models were all over the map, fairly normal for hurricane forecasting though Ian was more unpredictable due to stagnation of pressure systems.

My point being is that if you don't know why a model is calling for a certain forecast, it's a crap shoot. To my read, GFS in Ian provided excellent base data including explanation of why the Cone of Uncertainty was wide and shifted. Perhaps in hindsight one model was more accurate than the other. But as they say in investments, past performance does not guarantee future performance.

Until someone can tell me why one forecast model is better than another, I will continue to urge cruisers to get base data and interpret it for themselves. The data for rhe European model may be out there but I can't find it, though I haven't looked too hard.

But it's a fair question - is the euro model actually better, or is it that the various forecast algorithms cannot assimilate the best data from NOAA and NWS, likely the world's best funded weather forecasters?

Peter

 

[Jc180]

I live in Madeira Beach which had a bullseye on it for Ian. Was the reason I dashed from Ensenada to Florida.

The models were all over the map, fairly normal for hurricane forecasting though Ian was more unpredictable due to stagnation of pressure systems.

My point being is that if you don't know why a model is calling for a certain forecast, it's a crap shoot. To my read, GFS in Ian provided excellent base data including explanation of why the Cone of Uncertainty was wide and shifted. Perhaps in hindsight one model was more accurate than the other. But as they say in investments, past performance does not guarantee future performance.

Until someone can tell me why one forecast model is better than another, I will continue to urge cruisers to get base data and interpret it for themselves. The data for rhe European model may be out there but I can't find it, though I haven't looked too hard.

But it's a fair question - is the euro model actually better, or is it that the various forecast algorithms cannot assimilate the best data from NOAA and NWS, likely the world's best funded weather forecasters?

Peter

But the ECMWF model never wavered in forecasting Ian. Always Ft Myers. Never Madeira Beach. And the NWS meteorologists never told us except in the “cone of uncertainty” discussion. But I do agree, who knows which model is better on any given day. And your point about understanding why or how each model works is unknowable to us laymen. My point is not to disregard the ECMWF model because the US NWS favors GFS. Furthermore my own observations say do so at your own peril. We should look at both.

 

[Jc180]

One more point about Ian. Did the cone of uncertainty really shift? Again ECMWF never changed. Why did GFS?