About this page
How weather forecasting, one of the oldest "sciences" if not the most precise, has developed over the past 30 or so years. What does the future hold?
- Using Weather Forecasts ,
- Weather Forecasts for the Sailor ,
- Calculating the Weather
- Single Observer Forecasting (Weather Lore)'''
On this page -
- Where it started - Weather Lore
- Before the computer Synoptic Meteorology
- Numerical Weather Prediction
- Why it goes wrong
- Where we are now
- What will the future bring? A long range forecast!
Setting the scene
Having worked out a passage plan, checked all the relevant pilotage information such as rise of tide at your destination and ETA, all then depends on the weather. For many yachtsmen that is when the uncertainties really creep in. Why do they use words like "perhaps" and "locally". What does West becoming North really mean? When will it become North? Even as a now long-retired UK Met Office senior forecaster I still find the uncertainties and apparent prevarication in weather forecasts to be frustrating. Why is it like this and is it really getting better?
Man has tried to predict the weather ever since he began to depend on agriculture and fishing for his livelihood. It may seem surprising that a science as old as astronomy and medicine can be so uncertain. I will try to show how meteorology has developed and why there is still much to be done. Weather forecasting has progressed through three phases - single observer forecasting, the synoptic approach and the current application of computers. The three phases overlap and each still has its place in the overall scheme.
Single observer forecasting, sometimes called Weather Lore, is limited by the amount of information used. We only know what is happening at one location and weather systems can travel great distances - 1500 miles or more in 24 hours is not unknown. Experience coupled to observation is very useful but still just one, small tool in the thinking yachtsman's armoury. For thousands of years it was the only possible technique and, as such, has gathered some considerable mystique.
Read more about single Observer Forecasting here.
Synoptic meteorology began in the mid 1850s. By then scientists had realized that weather had to be studied as a complete system and not at isolated locations without reference to what was happening elsewhere. The electric telegraph had become operational allowing the collection of data over large areas in real time. A study of a storm in the Black Sea during the Crimean war had shown that an area of strong winds around a low pressure centre had moved in a fairly steady manner. This showed that weather prediction was possible, at least in principle.
For the next 100 years advances in synoptic meteorology followed successive technological developments. Radio telegraphy and the Morse code enabled the collection of data over the oceans as well as from remote land areas. Radio then led to cheap, use-once-and-throw-away instruments that could be attached to free flying balloons. These enabled pressure, temperature and humidity to be measured above ground level and so give a three dimensional picture of the atmosphere.
Radar provided the means to track these radio-sonde balloons and so measure winds throughout the depth of the atmosphere. Improvements in aviation navigational techniques allowed commercial aircraft to measure winds at flying height en route.
Radar was developed to show rainfall extent and give a rough but useful measure of intensity. Satellites led, first, to pictures of overall cloud cover and development. Satellites are now being used increasingly to make measurements. Contrary to what many people think, in themselves, satellites do not provide forecasts; they are simply observing and communication platforms.
As more information became available so a better theoretical basis for prediction evolved. However, forecasting was still a very uncertain business. There were some significant successes, but also some serious failures. The D-Day forecast fell into the former category with a brief weather window, well forecast. But, without denigrating their efforts in any way, the Allied forecasters, led by the UK Met Office, did enjoy a little luck. The German meteorologists, no slouches themselves, had advised the Luftwaffe that conditions were too bad for an invasion and many of their fighter pilots were taking a break.
Synoptic forecasting methods fail, ultimately, because quite small differences in weather pattern can lead to quite different outcomes, especially for periods more than 24 hours ahead. The use of pattern matching techniques, tried particularly for long range prediction, fails for short periods as well.
The idea that weather developments could be calculated dates back to the 1904 when a Norwegian scientist said that it should be possible to use mathematical equations describing the physics of the atmosphere. This was at a time when computing was by mechanical calculators and log tables (does anyone remember those?) During WW1, a British scientist, L F Richardson, tried to do the sums but failed due to lack of computer power, observations and the knowledge required to do the calculations..
We all know that pressure gradients lead to wind. But air pressure is related to temperature. Temperature depends on how the air is heated by the earth, itself heated by the sun. Dry, sandy soils and man made large conurbation's heat up quickly and warm the air above them. Such areas also cool quickly at night. Oceans and forests heat up more slowly and, so, heat the air less. But they retain their temperature at night and still heat the air, but more slowly than the warm land by day. South facing slopes heat up while North facing ones stay cold.
Winds move areas of warm and cold air around causing changes to the pressure patterns. As air rises it cools. This leads to condensation and cloud formation. Water in the cloud may then freeze. During condensation and freezing, latent heat is released giving a heating effect while the cloud reduces both incoming solar radiation and radiation loss from the earth. Snow falling from cloud may melt, rain falling from cloud will evaporate, at least partially. These effects require heat and thus cause cooling of the air.
Topography acts on all scales. At the smallest end, houses and trees cause gusts. Higher up the scale we have sea breezes and winds around headlands foehn effect cause very warm often dry weather on the lee side of hills and mountains. Eastern Scotland and Northeast England come readily to mind. The Rockies have a downstream effect over Western Europe. At the largest scale, continents, or the lack of them, lead to the Roaring Forties.
All these effects change the pressure patterns that we are used to seeing on weather charts as isobars. Look at a series of weather charts and see how weather systems both move while changing shape and intensity. Richardson saw how these processes could be calculated in principle but did not have the tools necessary to make other than a very crude and ineffective attempt at even the simplest. prediction. The idea was shelved as being theoretically interesting but impossible in practice.
After the second World War, computers, developed and were seized upon as the means by which computing the weather might be realized. Since then meteorologists have been specifying needs for bigger and faster computers while improving the calculation of all the various processes that occur in the atmosphere. Perhaps, it should be emphasised here that statistical techniques are not used. The computers do not use past experience, they start with the current state of the atmosphere and calculate how the atmosphere will evolve. The physicists call this an initial value problem.
Why do forecasts still go wrong? What improvements have there been over the past 50 years? Forecast errors occur for three main reasons.
- First, even with many land observing stations, many ships, buoys and aircraft providing data over the oceans and with satellites continuously providing information from space, there are still uncertainties in knowing sufficiently accurately the present conditions. Improvements are taking place but the problems are great and very costly to solve.
- Secondly, even the biggest computers available cannot handle all the physical processes sufficiently accurately and never will. In particular, there are small, local variations in the weather that are currently too small for the computers to replicate.
- Thirdly, and the real killer is predictability. This comes down to chaos in the atmosphere. Small disturbance may grow into large ones - the so-called butterfly effect, although I doubt whether a butterfly flapping its wings in Tokyo can really result in a storm over New York or anywhere else. The problem, partly, is knowing which will and which will not develop coupled with the observational problem of detecting such initially small weather features.
Te improvements that have occurred most evident in the ability to provide useful, although by no means always accurate, information out to five or six days ahead. Improvements in local forecasts have been achieved over land for rainfall. These are due to a combination of the input from weather radar coupled with NWP output. There are no comparable wind data. Studies into modelling on a very small local scale for short periods have shown that it is possible to predict fog occurring in low lying areas. Again, the use of such models for wind is more problematic since, for example, small variations in wind direction can give a very different local effect.
National weather services, such as the Met Office, rely heavily on the numerical weather models but still use their synoptic experience. Every 6 hours, the Met Office produces 24 hour forecast charts by taking the computer output and modifying it using subjective assessments of experienced forecasters. Every 12 hours they produce forecast charts for 48 and 72 hours ahead in the same general manner. The time may not be far off that all these forecast charts are issued straight from the computer.
The Met Office issues, once a day, charts for 96 and 120 hours ahead. based on output not only from the Met Office computer output but also the results of other forecasts centres such as the European Centre at Reading, the US Weather Bureau and major European services. These 4 and 5 day predictions are, thus, a consensus view. Time will tell whether or not the results are statistically better than just taking the results of one computer model.
Producing forecast weather charts is one matter. Getting the weather detail right is another question altogether. Much effort is being directed at using computer output to generate, directly, forecasts in word or graphical form.
Developments in computer technology will undoubtedly lead to further improvements in our ability to forecast for longer periods ahead. Just how far ahead it will be possible to produce forecasts of interest to sailors is uncertain. For the kind of decisions that we want to make, my best guess is about 10 days. Currently, March 2017, my experience is is that I can often plan 6 or 7 days ahead.
Short period - less than 12 hours - forecasts are still a problem area for several reasons.
- Small weather details have short lifetimes
- Theleast of our worries, very great computing power is needed6 to generate forecasts of the scale of the Tor Bay sea breeze (for example) or thunderstorm scale features.
- The practical problem of how sailors would receive the enormous amount of information that would be available.
- Another practical problem is how such detailed information would be used or even be useful to a leisure sailor.
Yet to be developed meaningfully is the concept of probability in weather prediction. What would a sailor do if told that there was a 30% chance of winds reaching gale force? Go or no go?
Return to Home Page