Weather proving disaster for boat trips

[aqua_sax]

we are sailing in French waters for the summer

Lucky you! If you are south of Brest you are probably also enjoying better weather!

I know what you mean about the importance of defining terms. Woolly definitions = woolly thought. When talking about measurements, in medical research, we tend to use the terms reliability and validity, in other fields the terms precision and accuracy are used to mean the same things.

Reliability/precision: repeated measurements of the same thing yields the same result. You measure the length a nail ten times and each it comes out at exactly 5 inches. The measurement is reliable/precise. Unfortunately, someone stretched your plastic tape measure without you realising it, and it under-reads. Your nail is actually a 6 inch nail.

Validity/accuracy: you get the right, correct result. When you measure a 6 inch nail, the result you get is 6 inches. The measurement is valid/accurate.

It then follows that to assess validity/accuracy, you need to know what the correct measurement is. In medicine, we use the concept of the gold standard test, a test that we believe produces the most valid/accurate result (which may itself be more or less valid/accurate, but we believe it is the best we have) and then test other tests against it.

In assessing weather forecasting accuracy, the gold standard is the actual, real weather, But it is hard to measure and record in a satisfactory manner, because of the spatial and temporal variation we have already mentioned. A common workaround is to use modelling, but by definition (sic), it is just a model, not the real thing. We don't really have valid, reliable gold standard data.

One answer might be to use a summary which is good enough for our purposes. This can certainly be done visually, which I did as part of my attempt to determine whether the wind patterns have changed over the years (a separate but related effort definitely related to this thread's title, though these plots aren't going to be right for testing forecast accuracy, but they might set others thinking). Here is an example for the Isle of Portland. This is a useful weather observation station because (a) it has a long reasonably complete (MIDAS) data set (though they are questions about how the data was collected eg manual vs automatic) and (b) the station is located pretty much in the Channel (see google maps), meaning less but not no land effects. Hourly wind speed is actually the mean wind speed in the period 20 minutes to 10 minutes to the hour (ie mean for HH-20 to HH-10).



These so-called violin plots give a visual impression of the range (top to bottom) and dispersion (width of the 'violin' at each point shows the number of readings at that point) of the wind in July at the Isle of Portland over the last four decades. The first four years (1980 to 1983) do seem to me more sailing friendly, mostly moderate winds (bulges mostly in the 10-15 knot range) but for the rest of the series I can't really say I see a visual trend one way or the other. There may be more Julys with often not enough wind (bulges low down) but the peak (mean not gust) wind speed looks pretty random to me, certainly no obvious trend.

Frank, no need for an immediate response, more important you enjoy the sailing while you can! I'm just filing it here for possible future discussion, if you are so inclined, no obligation.

 

[franksingleton]

That is my assessment of the current state of play as well. My hunch is that will continue to be the state of play for at least the foreseeable future, perhaps even forever, mostly for the reasons you give: the spatial (and temporal) variations in actual wind happen at a scale well below the forecast resolution. It is like trying to spot microscopic changes with a magnifying glass, it can't be done.

I have been able to get some family free time sooner than I expected and have been able to read your post fairly slowly.

We seem to agree there about the difficulty/impossibility of having any verification scheme that would help lay users of forecasts. Onlective assessments used by modellers are unlikely to help most users.

However, they may be indirect methods. I have already mentioned the concept of forecast stability: if forecasts are accurate, then they should not change very much if at all in the short term. If they do, then at least one if not both of the forecasts must be wrong. Since the IWF includes both a 24 hour forecast and an outlook for the following 24 hours, we can compare today's 24 hour forecast with yesterdays 24 hour outlook, and see how much they agree or disagree.

That is what I recommend. In practice, there is rarely much real difference between different models for the first 2 or more days. Over those short time periods differences between models probably reflect the uncertainty in the weather itself.

Two more general points:

(1) I personally am very wary of modelling. To my mind, it is a fancy name for whatiffery - what happens if this happens? In very simple systems it might work, but the moment you go outside the parameter limits of the model (ie extrapolate) you risk entering a fantasy world. A simple example (not perfect but it gives the idea): you model a boat's behaviour at displacement speeds. What does that model tell you about what happens at planning speeds? Or consider the modelling used during the pandemic, again no more than random whatiffery, that time with very damaging consequences. The modern belief that anything can be modelled given enough data and a big enough computer is just another example of human hubris. Take human behaviour for example: try developing a model for that. Maybe the weather isn't far behind, because both human behaviour and the weather at profoundly complex, such that the numbers of ones and zeros you need to model their behaviours are infinite, and there is no way that I can see of managing let alone understanding an infinite number of ones and zeros. Perhaps in the future mankind will come up with something cleverer than manipulating ones and zeros, but it hasn't happened yet.

I would say, categorically, that you are wrong as far as weather modelling is concerned. These apply the laws of classical physics, starting with Newton’s third law. You are probably correct for statistical models. Of course, there are uncertainties in model formulation, data and data analysis. Over the past (nearly 80 years) the great effort that has gone into understanding these effects has resulted in continue increase in short term forecasting and in the period ahead for which forecast models give useful guidance. I saw this in the mid-70s, when the Met Office model was, basically hemispheric. As a then forecaster, I could see increasing skill beyond 3 days.
AI forecasting is, in a sense, going backwards in that it is, essentially, a pattern matching technique. Whether, when fully implemented, it will improve on physical models when properly implemented has yet to be seen. I do not know enough about weather AI models to know how the models are being development. Essentially, the AI model will look at historical outcomes to see how the current situation might evolve over the next few hours. I have no feel for Howe long a time step is used.

(2) what exactly do I want from the forecast? Do I want to know there is a 40% chance of a light shower at 1400 this afternoon? I don't think so. If, for example, I am considering whether to put some varnish on the boat this afternoon, all I need to know is whether it is going to rain or not. Likewise, I use the shipping forecasts to decide whether to go to sea, or whether to stay snug in harbour. A 60% probability of 25 knot gusts at 1200 is too much detail, all I really need to know is whether there is going to be not enough wind, a useful breeze, or too much wind, plus general direction, because what might be too much wind from one quarter might be a fair wind from the opposite quarter. Perhaps modern forecasting has strayed down a path where it ends up being too clever by half. Perhaps more is less, and if so, less might be more.

Unless you are going to pay for a forecast tailored to your specific needs, then, inevitably it is going to be a matter of interpretation of a general forecast. However, there are some tools that can help. For example, for rain sensitive activities, the Met Office app is usually good. But, even that can be a little misleading even for what is happening right now. I remember a cartoon many years ago showing a weather radar spinning round but with a cumulus type cloud sitting right above the radar and grinning. There are situations where some very light rain might be so low down that it is below radar beams and not show up on the app. In any case, drizzle droops are too small to be seen by weather radars.

In summary, you seem to be falling into the all too common trap of not appreciating that weather models are physical constructs, not statistical, data driven ones.

My frustration about anonymity is that those who hide behind a pseudonym can make outlandish statements with impunity and no evidence about their knowledge base.

 

[franksingleton]

Lucky you! If you are south of Brest you are probably also enjoying better weather!

I know what you mean about the importance of defining terms. Woolly definitions = woolly thought. When talking about measurements, in medical research, we tend to use the terms reliability and validity, in other fields the terms precision and accuracy are used to mean the same things.

Reliability/precision: repeated measurements of the same thing yields the same result. You measure the length a nail ten times and each it comes out at exactly 5 inches. The measurement is reliable/precise. Unfortunately, someone stretched your plastic tape measure without you realising it, and it under-reads. Your nail is actually a 6 inch nail.

Validity/accuracy: you get the right, correct result. When you measure a 6 inch nail, the result you get is 6 inches. The measurement is valid/accurate.

It then follows that to assess validity/accuracy, you need to know what the correct measurement is. In medicine, we use the concept of the gold standard test, a test that we believe produces the most valid/accurate result (which may itself be more or less valid/accurate, but we believe it is the best we have) and then test other tests against it.

In assessing weather forecasting accuracy, the gold standard is the actual, real weather, But it is hard to measure and record in a satisfactory manner, because of the spatial and temporal variation we have already mentioned. A common workaround is to use modelling, but by definition (sic), it is just a model, not the real thing. We don't really have valid, reliable gold standard data.

One answer might be to use a summary which is good enough for our purposes. This can certainly be done visually, which I did as part of my attempt to determine whether the wind patterns have changed over the years (a separate but related effort definitely related to this thread's title, though these plots aren't going to be right for testing forecast accuracy, but they might set others thinking). Here is an example for the Isle of Portland. This is a useful weather observation station because (a) it has a long reasonably complete (MIDAS) data set (though they are questions about how the data was collected eg manual vs automatic) and (b) the station is located pretty much in the Channel (see google maps), meaning less but not no land effects. Hourly wind speed is actually the mean wind speed in the period 20 minutes to 10 minutes to the hour (ie mean for HH-20 to HH-10).

View attachment 197309

These so-called violin plots give a visual impression of the range (top to bottom) and dispersion (width of the 'violin' at each point shows the number of readings at that point) of the wind in July at the Isle of Portland over the last four decades. The first four years (1980 to 1983) do seem to me more sailing friendly, mostly moderate winds (bulges mostly in the 10-15 knot range) but for the rest of the series I can't really say I see a visual trend one way or the other. There may be more Julys with often not enough wind (bulges low down) but the peak (mean not gust) wind speed looks pretty random to me, certainly no obvious trend.

Frank, no need for an immediate response, more important you enjoy the sailing while you can! I'm just filing it here for possible future discussion, if you are so inclined, no obligation.

We crossed in the post. My reply above still stands.

 

[Refueler]

Most of the Shipping Co's I worked in / with - contracted Oceanroutes for route advisorys ....... and too many I observed were rubbish !

Problem being that ships voyages are not short and OR would set a route that looked good - but could not foresee the next stages ...

But the worst case of Route planning I ever experienced ....

Gulf Mexico - the Seismic Ship I was working on ... was sold to a UK outifit in Galveston. The UK outfit wanted the ship ... 900ton converted Stern Trawler to move to Falmouth UK for dry-docking.
They could not get a full crew - so we were asked if we would stay on and assist the skeleton crew they would send out. Myself and 2 engineers remained on for the trip.
The Master they sent out had never captained such a vessel and was a liability.
The CH.Eng was dragged out of retirement and couldn't give a monkey's ****

Anyway - I suggested a route taking into account seasonal weather etc ... new owners sent us Oceanroutes suggested - which was to be honest - laughable ! But they were the owners and they insisted we follow it.
Being a Seismic vessel - we had satcoms with telex / fax / voice. Once we exited Gulf of Mexico and skirted south of islands to enter Atlantic ... the weather which I had feared hit us ... myself and master were on Satcom to owners in London ... advising that weather was atrocious and we would be delayed etc.
You cannot beloieve the reply we had ... Owner said - I'm looking at weather map and you should be in good weather .. so hang in there ... weather is shown to be good .....
We both asked what map he was looking at as we had downloaded the forecast via the satcom and it showed worsening for significant period ...

Owner replied ..... the waether map in the corner of his newspaper !!!

I was then guilty of voicing my opinion .....

By time we reached Falmouth - we had water in the holds where the Computer rooms were installed .... our cabins were slopping water around the floors ... our bedding / clothes / stores were damp / wet ....
The engineers had been maintaining regular pumping routines during the trip ....

Needless to say - I was quick to vacate the 'old girl' in Falmouth .....

Next I heard of her - new owners had sent her up around Shetland for work and she suffered a serious fire ... was burnt out.

 

[aqua_sax]

I would say, categorically, that you are wrong as far as weather modelling is concerned. These apply the laws of classical physics, starting with Newton’s third law. You are probably correct for statistical models.

In summary, you seem to be falling into the all too common trap of not appreciating that weather models are physical constructs, not statistical, data driven ones.

That's an interesting point, because I always thought, mistakenly, that the heart of modern forecasting was, to put it very simply, data collection, and then looking for matching patterns, with bigger and better computers making this much easier. But, if I have understood you correctly, which I think I have, because I have re-read the relevant section in your recent CA article, the current models start with the fundamental laws of physics, and then apply them to weather. Fundamentals to specifics rather than specifics to specifics. Incorporating AI does certainly seem to be a retrograde step, but then there is nothing unusual in that. Take google's 'AI overviews' which can be very wrong, or ebay's AI generated item descriptions, like this one for for a NWP super-computer: "The Met Office inspired NWP super-computer embodies the spirit of numerical weather prediction. With it's trillions of calculations per nanosecond, this supercomputer is perfect for weather forecast enthusiasts and aficionados. Housed in a sleek brushed aluminium casing, it will be sure to put the shine on your forecasts. Get one today and up your forecasting game!"

My frustration about anonymity is that those who hide behind a pseudonym can make outlandish statements with impunity and no evidence about their knowledge base.

Agreed, but I generally avoid making outlandish statements (though some might say that I sometimes have a somewhat flamboyant style). I may get something wrong, but that is not the same thing as outlandish, and I am very happy to be corrected if someone points out I am wrong, or argue my case should I happen to believe I am right, and the other party is mistaken. The anonymity is really a hangover/habit from a time and a place where anonymity was both the norm and a sensible precaution.

Needless to say - I was quick to vacate the 'old girl' in Falmouth

Very wise!

 

[franksingleton]

That's an interesting point, because I always thought, mistakenly, that the heart of modern forecasting was, to put it very simply, data collection, and then looking for matching patterns, with bigger and better computers making this much easier. But, if I have understood you correctly, which I think I have, because I have re-read the relevant section in your recent CA article, the current models start with the fundamental laws of physics, and then apply them to weather. Fundamentals to specifics rather than specifics to specifics. Incorporating AI does certainly seem to be a retrograde step, but then there is nothing unusual in that. Take google's 'AI overviews' which can be very wrong, or ebay's AI generated item descriptions, like this one for for a NWP super-computer: "The Met Office inspired NWP super-computer embodies the spirit of numerical weather prediction. With it's trillions of calculations per nanosecond, this supercomputer is perfect for weather forecast enthusiasts and aficionados. Housed in a sleek brushed aluminium casing, it will be sure to put the shine on your forecasts. Get one today and up your forecasting game!"



Agreed, but I generally avoid making outlandish statements (though some might say that I sometimes have a somewhat flamboyant style). I may get something wrong, but that is not the same thing as outlandish, and I am very happy to be corrected if someone points out I am wrong, or argue my case should I happen to believe I am right, and the other party is mistaken. The anonymity is really a hangover/habit from a time and a place where anonymity was both the norm and a sensible precaution.



Very wise!

Perhaps you did not read Calculating the Weather - Franks-Weather - The Weather Window
NWP was first suggested as a theoretical possibility in 1906. An attempt at a manual calculation was made in 1918. A subsequent book in the early 1920s foresaw parallel processing. First electronic computer models were trialled in the 1950s. The first, very crude Met Office model was introduced in the mid 1960s. It used an English Electronic KDF9.

 

[aqua_sax]

Perhaps you did not read Calculating the Weather - Franks-Weather - The Weather Window

I have now, thank you for the link.

One thing that has been bothering me has now crystallised after reading the above (which for the first time describes as well as numbers the law). I didn't mention it before because I wasn't sure of my ground, thought it might be a typo or whatever, and even now it could be seen as more of a semantic thing. It is which Newton's law of motion is in use. In everything you have written that I have read, you say it is the third law, but I don't think it is, it is the second law, as described in the above, but named as the third law.

In one English version of the Principia, the law of motion are given as follows:



Extracting the key bits:

Law II: The alteration of motion is ever proportional to the motive force impressed

Law III: To every action there is always opposed an equal reaction

In your article linked to above, you state:

"At the centre of these equations is one that says, in effect, that applying a net force to an object, including a parcel of air, results in change – acceleration or deceleration. This is Newton’s third law." But it seems to me this is the second law. The alteration of motion (ie your 'change – acceleration or deceleration') is ever proportional to the motive force impressed (ie your 'applying a net force to an object') with the subsequent elaboration under the second law making it clear it is that it is net force ie "compounded from the determination of both". The other important point in the Principia but not in your description is that the alteration in motion is proportional, a double force will generate double the motion. This leads in turn to the formulaic expression of the second law, acceleration = net force / mass, which tells us that the acceleration (or deceleration) is (a) proportional to the net force applied and (b) inversely proportional to the mass of the object being accelerated (or decelerated).

Maybe I have got this all wrong, but I don't think I have. I have after all gone back to the original source.

Does this matter? I don't know. Perhaps it is just a typo that having become fixed in the author's mind has been perpetuated ever since. Does it even matter what number we give to the law, given what really counts is the content and meaning of the law? Or perhaps it does matter, because when we just refer to using Newton's third law, without defining the law, it amounts at best to woolliness and, perhaps of more concern, it is a source of perplexion and confusion to mere mortals attempting to understand the complexities of numerical weather prediction.

 

[franksingleton]

I have now, thank you for the link.

One thing that has been bothering me has now crystallised after reading the above (which for the first time describes as well as numbers the law). I didn't mention it before because I wasn't sure of my ground, thought it might be a typo or whatever, and even now it could be seen as more of a semantic thing. It is which Newton's law of motion is in use. In everything you have written that I have read, you say it is the third law, but I don't think it is, it is the second law, as described in the above, but named as the third law.

In one English version of the Principia, the law of motion are given as follows:

View attachment 197370

Extracting the key bits:

Law II: The alteration of motion is ever proportional to the motive force impressed

Law III: To every action there is always opposed an equal reaction

In your article linked to above, you state:

"At the centre of these equations is one that says, in effect, that applying a net force to an object, including a parcel of air, results in change – acceleration or deceleration. This is Newton’s third law." But it seems to me this is the second law. The alteration of motion (ie your 'change – acceleration or deceleration') is ever proportional to the motive force impressed (ie your 'applying a net force to an object') with the subsequent elaboration under the second law making it clear it is that it is net force ie "compounded from the determination of both". The other important point in the Principia but not in your description is that the alteration in motion is proportional, a double force will generate double the motion. This leads in turn to the formulaic expression of the second law, acceleration = net force / mass, which tells us that the acceleration (or deceleration) is (a) proportional to the net force applied and (b) inversely proportional to the mass of the object being accelerated (or decelerated).

Maybe I have got this all wrong, but I don't think I have. I have after all gone back to the original source.

Does this matter? I don't know. Perhaps it is just a typo that having become fixed in the author's mind has been perpetuated ever since. Does it even matter what number we give to the law, given what really counts is the content and meaning of the law? Or perhaps it does matter, because when we just refer to using Newton's third law, without defining the law, it amounts at best to woolliness and, perhaps of more concern, it is a source of perplexion and confusion to mere mortals attempting to understand the complexities of numerical weather prediction.

See #22

 

[aqua_sax]

See #22

But that also says Newton's third law, as does your recent CA article. #22 does not describe the law, it just refers to the third law, whereas the CA article gives a brief summary ("a net force applied to a mass creates a change") which is the point at which I started wondering because that is not normally how the third law is described, but didn't initially bring it up. Seeing in effect the same thing on your 'Calculating the weather' web page is what made me think I needed to clarify the position.

 

[franksingleton]

But that also says Newton's third law, as does your recent CA article. #22 does not describe the law, it just refers to the third law, whereas the CA article gives a brief summary ("a net force applied to a mass creates a change") which is the point at which I started wondering because that is not normally how the third law is described, but didn't initially bring it up. Seeing in effect the same thing on your 'Calculating the weather' web page is what made me think I needed to clarify the position.

Sorry. I rather assumed that every sailor would have a basic understanding of Newton’s laws. You must surely know that the Third Law says, basically, apply a force to a mass and you create an acceleration. Every air particle is subject to gravity, pressure gradient forces, the Coriolis ettect, friction and viscous effects.
In principle, all these ever changing forces can be calculated. In practice, of course, they cannot and NWP models have to make many approximations and estimates. That it works at all is a tribute to the many scientists in many countries.

 

[aqua_sax]

You must surely know that the Third Law says, basically, apply a force to a mass and you create an acceleration.

See #27 and please read what is says in the image and text. What you call the third law is the second law.

 

[Dockhead]

When I came to Latvia in 1993 ... summers were glorious for boating .... winters were glorious for Postcard Winter scenes.

Over time - summers have continued in the 25 - 30C mark - but high humidity if not serious rain and winds.... winters are more like wet UK crap ...

I want to enjoy my boats more - but weather is just proving so darned awful ... and forecasts are becoming unreliable for more than 24hrs ...o

Its not only impacting use of the boats - its affecting maintenance ......

Well, the climate, she is a-changin . . .

Summer in the Northern and Eastern Baltic is always a crap shoot and I suspect always has been. My first summer in Finland, I was told that midsummer (21 June) was one of the biggest holidays, and that everyone goes to the islands. So I dutifully sailed to one of the islands, anchored out, dinghied to shore -- and found no one there. That night, it snowed, and I found an inch of snow on my deck. That was 2014.

I spent a couple years in Riga and kept my boat in the old yacht club over the summer. Those summers were both glorious.

I've only recently moved the boat from Denmark back to the Solent. Last winter I had a lot of damage from freezing. I had been keeping the boat in Denmark during winters specifically to avoid freezing. But there was a freakish cold snap this year, which took out a water pressure pump, some taps, and other plumbing items.

I'd still take all that over Florida or the Med, any day of the year. Just need a well-found boat, good central heating, a good anchor, and patience. There's some joy in every kind of weather. Especially in this part of the world.

And sometimes the unusual weather even works to your advantage. We sailed from Denmark, near Copenhagen, to the Solent during June, and instead of the usual headwinds, we had Easterlies and Nor'easterlies the entire way, making for some gorgeous sailing and a pleasant North Sea crossing. An anti-cyclone caused by a high pressure dome over the North Sea.

 

[franksingleton]

See #27 and please read what is says in the image and text. What you call the third law is the second law.

Forget the semantics as you read them. NWP begins with Force = Mass x Acceleration applied to a fluid in what are known as the Navier-Stokes equations. Just pause, think about basic physics at school level.

 

[aqua_sax]

Another pair of adjacent short term forecasts that are significantly different, 3 to 5 has become 4 to 6 occasionally 7):

Selsey Bill to Lyme Regis Inshore Waters Forecast:

Yesterday's 0600 Outlook for the following 24 hours (ie today): East or southeast 3 to 5.

Today's 0600 24 Hour forecast (ie today): East or northeast, veering southeast for a time at first, 2 to 4, increasing 4 to 6, occasionally 7 later.

These are short term forecasts ie 24-48 then 0-24 hours ahead. At least one must be wrong, or possibly both.

 

[franksingleton]

Another pair of adjacent short term forecasts that are significantly different, 3 to 5 has become 4 to 6 occasionally 7):

Selsey Bill to Lyme Regis Inshore Waters Forecast:

Yesterday's 0600 Outlook for the following 24 hours (ie today): East or southeast 3 to 5.

Today's 0600 24 Hour forecast (ie today): East or northeast, veering southeast for a time at first, 2 to 4, increasing 4 to 6, occasionally 7 later.

These are short term forecasts ie 24-48 then 0-24 hours ahead. At least one must be wrong, or possibly both.

So? As a sailor you will know all too well just,how variable is the weather. I will give you a challenge. Sail along that bit of coast up and down for a 24 hour period. Then try to describe what you have experienced in as few words as the forecasters used. Even just do one passage and describe what you have seen.

 

[Refueler]

So? As a sailor you will know all too well just,how variable is the weather. I will give you a challenge. Sail along that bit of coast up and down for a 24 hour period. Then try to describe what you have experienced in as few words as the forecasters used. Even just do one passage and describe what you have seen.

Much as we'd like to accept better forecast - the prudent mariner takes the worst as the expected.

 

[franksingleton]

Much as we'd like to accept better forecast - the prudent mariner takes the worst as the expected.

Quite! The prudent sailor has an understanding of the realities of a briefly worded forecast.