franksingleton
Well-Known Member
From time to time, I take a look at YBW forums to see what is being said about weather, weather forecasts, warnings etc. Inevitably, I see the same old misunderstandings of the realities and practicalities of weather forecasting. Rather than get involved in any particular aspect, I thought that I would try to set down my thoughts on the whole business. I do this as a cruising sailor with a background as a Met Office senior forecaster – in the distant past.
It was all much simpler 40 years ago when all we had was the shipping forecast, 4 times daily each with actuals, covering the next 24 hours. It was only available on BBC LW (Radio 2 in those far off days). People like me wrote books and articles on how to draw a mini-synoptic chart to put flesh on the bare bones of the forecast and even deduce an outlook. Very hairy; difficult to teach, more difficult to learn and to maintain the skill. However, it did make you think about the weather.
Now we have many forecasts, from several sources (eg National Met services of UK, USA, France, Germany, ECMWF as well as various private sector services that use output from the national centres), via various routes. The mass of information, coupled with some ridiculous claims about “accuracy” and “precision” have, I believe, generated a black box syndrome not unlike that that I see affecting navigation. My perception is that some sailors are taking forecasts verbatim and not using their own experience and nous.
Why are forecasts not precise? Why do they go wrong?
First, computer speeds are still far less than is needed. Although speeds have increased by about 10 to the power of 12 ( 1 and 12 zeros) over the past 50 years, it has been estimated that to replicate the atmosphere requires a further speed increase of 10 to the power 36 (1 and 36 zeros). At best, that is 150 years away. Realistically, it may never happen.
Secondly, data are simply not good enough to define the atmosphere as accurately as we sailors observe it. There are always small errors in the analysis, despite massive efforts by the major weather services involving some very heavy mathematics. Running a numerical weather forecast is the easy bit; initialising the models is the difficult part. Like medicine a good prognosis requires a good diagnosis. Again, like medicine, a good diagnosis will not guarantee a good prognosis.
The third problem is the atmosphere itself; this is because of chaos. I doubt that any thinking meteorologist really believes that a butterfly flapping its wings in Tokyo can create a storm in New York. However, hurricanes start from a small group of thunderstorms in the right place, just at the right time. Many of our lows start from a small wave on a front to the east of the USA. In neither case can anyone predict where and when the next one will form. Throughout the lifetime of such weather features, chaos is such that skill in the forecast will decrease the further ahead is a prediction made. I am talking here about major features and not the detail within them.
Large scale prediction requires global modelling partly because weather can travel great distances in 24 hours and partly because of tele-connections eg the El Nino. Global models currently work on a grid of around 20 NM and so can only define weather or topography on a scale of about 100 NM.
My experience, as a user, is that a 24 hour synoptic chart or GRIB forecast will be pretty good, but never be precisely correct in all detail. A 48 hour forecast will have more errors. At 5 days, there will be appreciable skill but significant errors in places. By 7 days, skill will be too small for our use. By 15 days there will be no skill whatsoever.
Detailed (meso-scale) forecasts use global input but then calculate over a smaller area on a finer grid with better representation of the physics ie latent heat, radiation, topographic effects and so on. The Met Office uses a 6 NM grid over much of the eastern USA/N Atlantic/European area and 2 NM over the British Isles. At best the latter can define weather on a 10 NM scale.
Meso-scale forecasts can and, like all forecasts, do go wrong.
First, they are critically dependent on the analysis. A small area of cloud can affect how a sea breeze forms; a small change in the gradient wind can have a big effect on how the wind comes round Portland Bill, for example.
Secondly, they depend on the global model. A meso-scale model can improve on a good global forecast but cannot repair a poor one.
Thirdly, weather can develop in situ eg thunderstorms. The model might say that they will probably occur, but, again, they are chaotic and, currently, unpredictable in detail.
Fourth, allied to the last is the lifetime of small weather features. A gust last seconds; a small cumulus cloud lasts about 30 minutes; a thunderstorm has a life span of about 6 hours; a group of storms perhaps 36 to 48 hours; a frontal depression can have a life span of a few days. These facts determine how long ahead it is worthwhile using a meso-scale forecast. Anything up to 36 hours is my suggestion and no more.
All that is the reality.
The practicality comes in when someone has to write a forecast in a word length that can be read out by HM coastguard. Sail for 30 miles and then describe the wind, visibility and weather as though it were a forecast. Then imagine doing that for a whole coastal area bearing in mind the effects of topography and the uncertainties of the forecast outlined above.
There are two lessons that I have learned as a sailor tempered by my meteorological background. First, that forecasts of general weather patterns are pretty good to about 4 or 5 days ahead. They are excellent as planning tools and help avoid being caught in the wrong place at the wrong time. This starts from looking at the next 5 days, each and every day when we are sailing. I can assure you that it pays dividends.
Secondly, detailed local forecasting is nigh impossible at present. Whether you use the Inshore Waters forecast or a Meso-scale GRIB forecast, these can only give broad guidance. The sailor must be prepared to put experience and commonsense to use. If you sail in Weymouth Bay, you will know how the wind comes around the Bill; if you sail in Torbay, you will know how the sea breeze occurs there. In neither case will you always get it right, but you will improve upon the forecast. Ignoring either the forecast or your own knowledge is fatal.
I apologise for the length of this but I saw that my name cropped up from time to time. I am not a lackey of the Met Office nor an apologist for any Met service. I do understand the background to forecasts better than most sailors. To see more about the limitations of forecast, the value of single observer forecasting etc go to my website pages starting at http://weather.mailasail.com/Franks-Weather/Understanding-Using-Marine-Weather-Forecasts. There, I try to lay it on the line.
It was all much simpler 40 years ago when all we had was the shipping forecast, 4 times daily each with actuals, covering the next 24 hours. It was only available on BBC LW (Radio 2 in those far off days). People like me wrote books and articles on how to draw a mini-synoptic chart to put flesh on the bare bones of the forecast and even deduce an outlook. Very hairy; difficult to teach, more difficult to learn and to maintain the skill. However, it did make you think about the weather.
Now we have many forecasts, from several sources (eg National Met services of UK, USA, France, Germany, ECMWF as well as various private sector services that use output from the national centres), via various routes. The mass of information, coupled with some ridiculous claims about “accuracy” and “precision” have, I believe, generated a black box syndrome not unlike that that I see affecting navigation. My perception is that some sailors are taking forecasts verbatim and not using their own experience and nous.
Why are forecasts not precise? Why do they go wrong?
First, computer speeds are still far less than is needed. Although speeds have increased by about 10 to the power of 12 ( 1 and 12 zeros) over the past 50 years, it has been estimated that to replicate the atmosphere requires a further speed increase of 10 to the power 36 (1 and 36 zeros). At best, that is 150 years away. Realistically, it may never happen.
Secondly, data are simply not good enough to define the atmosphere as accurately as we sailors observe it. There are always small errors in the analysis, despite massive efforts by the major weather services involving some very heavy mathematics. Running a numerical weather forecast is the easy bit; initialising the models is the difficult part. Like medicine a good prognosis requires a good diagnosis. Again, like medicine, a good diagnosis will not guarantee a good prognosis.
The third problem is the atmosphere itself; this is because of chaos. I doubt that any thinking meteorologist really believes that a butterfly flapping its wings in Tokyo can create a storm in New York. However, hurricanes start from a small group of thunderstorms in the right place, just at the right time. Many of our lows start from a small wave on a front to the east of the USA. In neither case can anyone predict where and when the next one will form. Throughout the lifetime of such weather features, chaos is such that skill in the forecast will decrease the further ahead is a prediction made. I am talking here about major features and not the detail within them.
Large scale prediction requires global modelling partly because weather can travel great distances in 24 hours and partly because of tele-connections eg the El Nino. Global models currently work on a grid of around 20 NM and so can only define weather or topography on a scale of about 100 NM.
My experience, as a user, is that a 24 hour synoptic chart or GRIB forecast will be pretty good, but never be precisely correct in all detail. A 48 hour forecast will have more errors. At 5 days, there will be appreciable skill but significant errors in places. By 7 days, skill will be too small for our use. By 15 days there will be no skill whatsoever.
Detailed (meso-scale) forecasts use global input but then calculate over a smaller area on a finer grid with better representation of the physics ie latent heat, radiation, topographic effects and so on. The Met Office uses a 6 NM grid over much of the eastern USA/N Atlantic/European area and 2 NM over the British Isles. At best the latter can define weather on a 10 NM scale.
Meso-scale forecasts can and, like all forecasts, do go wrong.
First, they are critically dependent on the analysis. A small area of cloud can affect how a sea breeze forms; a small change in the gradient wind can have a big effect on how the wind comes round Portland Bill, for example.
Secondly, they depend on the global model. A meso-scale model can improve on a good global forecast but cannot repair a poor one.
Thirdly, weather can develop in situ eg thunderstorms. The model might say that they will probably occur, but, again, they are chaotic and, currently, unpredictable in detail.
Fourth, allied to the last is the lifetime of small weather features. A gust last seconds; a small cumulus cloud lasts about 30 minutes; a thunderstorm has a life span of about 6 hours; a group of storms perhaps 36 to 48 hours; a frontal depression can have a life span of a few days. These facts determine how long ahead it is worthwhile using a meso-scale forecast. Anything up to 36 hours is my suggestion and no more.
All that is the reality.
The practicality comes in when someone has to write a forecast in a word length that can be read out by HM coastguard. Sail for 30 miles and then describe the wind, visibility and weather as though it were a forecast. Then imagine doing that for a whole coastal area bearing in mind the effects of topography and the uncertainties of the forecast outlined above.
There are two lessons that I have learned as a sailor tempered by my meteorological background. First, that forecasts of general weather patterns are pretty good to about 4 or 5 days ahead. They are excellent as planning tools and help avoid being caught in the wrong place at the wrong time. This starts from looking at the next 5 days, each and every day when we are sailing. I can assure you that it pays dividends.
Secondly, detailed local forecasting is nigh impossible at present. Whether you use the Inshore Waters forecast or a Meso-scale GRIB forecast, these can only give broad guidance. The sailor must be prepared to put experience and commonsense to use. If you sail in Weymouth Bay, you will know how the wind comes around the Bill; if you sail in Torbay, you will know how the sea breeze occurs there. In neither case will you always get it right, but you will improve upon the forecast. Ignoring either the forecast or your own knowledge is fatal.
I apologise for the length of this but I saw that my name cropped up from time to time. I am not a lackey of the Met Office nor an apologist for any Met service. I do understand the background to forecasts better than most sailors. To see more about the limitations of forecast, the value of single observer forecasting etc go to my website pages starting at http://weather.mailasail.com/Franks-Weather/Understanding-Using-Marine-Weather-Forecasts. There, I try to lay it on the line.
