SOUE News 2014

Weather and Climate Prediction: Science and Applications Across Timescales

Lecture by Dr Andy Brown, Director of Science at the Met Office

Dr Brown explained that his intention was, first to explain the science behind weather forecasting, and then to look at some of the applications.

To forecast the weather in the UK for more than a few days ahead it is necessary to forecast it for the whole globe, starting from observations of the current state. A lot is known about the fundamental physics of the atmosphere, for instance that air tends to flow from high-pressure to low-pressure areas, that hot air rises, how clouds form, etc. So the whole global atmosphere is encapsulated in a large computer model, involving 25 km squares covering the Earth's surface, each square divided into 70 different levels vertically. This amounts to about 55 million blocks of air. The computer solves the differential equations relating the variables involved, and steps forward the model every 10 minutes. This requires very large computers doing trillions of calculations per second.

The observations fed into the model come from a variety of sources, and they all complement each other. At the simplest it is someone, in Africa say, reading the instruments in a surface weather station. Many such surface weather stations are read automatically. Measurements at sea are taken by ships, and by floating buoys. These surface measurements are distributed very unevenly of course - lots in Europe, but much thinner on the ground in many parts of the world and they do just measure conditions at the surface.

Since around 1980 a huge amount of data has been available from weather satellites. These give very good spatial coverage over the whole planet, but there are challenges in interpreting the data - for example, high how is that cloud? Aircraft, weather balloons and radar can tell us what is happening at various heights, but aircraft measurements come mainly, as one would expect, from the main aviation corridors.

Some interesting comparisons have been made between the challenges of weather forecasting and those of economic forecasting. Hard as it is, three fundamental differences between the two fields do appear to give an advantage to the weather forecasters: 1) The weather forecasters have much better world-wide information about what is actually happening now; 2) The laws of economics are much less well-defined than the laws of physics; and 3) forecasting the weather has no feedback effect on what actually happens!

A good illustration of the advances in forecasting accuracy is that of predicting pressure distribution over the Atlantic. The three-day-ahead forecast now is as good as the one-day-ahead of 20 years ago, and the ±3 mbar errors of 20 years ago are now down to ±1 mbar. These dramatic advances have come from a combination of better observations, better science in the models and more powerful computers allowing more detailed models (smaller grid squares) to be run.

Weather forecasting for the airlines is done by the US for the American continents, and by the UK for the rest of the world. The speed and direction of high-level winds have a large impact on the fuel an aircraft needs to carry to complete its journey. Reliable forecasting means less fuel needs to be carried, which reduces the weight of the aircraft and hence the amount of fuel it actually uses. The savings can be very substantial.

Among other economic and humanitarian applications are the forecasting of sandstorms in parts of Africa. These are not only very unpleasant for those exposed to them, but are also believed to be related to the spread of meningitis.

To forecast the weather in the UK, the 25 x 25 km squares are really too big. For instance a thunderstorm will cover a much smaller area. One revolution of the last few years has been to adopt a much smaller grid, of squares 1.5 x 1.5 km. For example floods in Carlisle in 2005 resulted from very heavy rainfall in the Eden catchment. The average rainfall over the 25 km square was nothing very extraordinary, but a study (after the event in this case, but the capability is now used routinely) showed that by using smaller squares, the maximum observed rainfall accumulations of around 200 mm (8 inches) could be reproduced. Predicting floods given the rainfall is a task for the Environment Agency rather than the Met Office, but close collaboration between the two is hugely improving services. In general, the need is not only to produce better forecasts, but crucially to also ensure they are used by people to take action..

Another example relates to December 2011 in Central Scotland. A forecast on Monday 5 December suggested there were going to be strong winds on Thursday the 8th, so a "yellow" warning was issued. By Wednesday the 7th, gusts of order 100 knots were forecast for that area the following day, as indeed happened, e.g. blowing over a school bus. So a "red" warning was issued by the Met Office (there are only two to three of these per year for the whole UK) and the Scottish Government decided to close all schools in the region for the day, so thankfully no one was on that bus.

Admiral Robert Fitzroy, who had captained the Beagle on Charles Darwin's famous voyage, and later founded the Met Office, was careful to point out that forecasts were not expected to be exact, merely to express probabilities, and would therefore sometimes be wrong. The Met Office still takes this view, and tries to estimate uncertainties by "nudging" the computer model 20 or more times in different directions, and looking at how this affects the results. If they still all say the same thing, then there is a high probability they are right and action can be taken with confidence. However even a forecast that there is, say, a 20% chance that something will occur, can in certain circumstances be very useful e.g. if it gives the civil contingencies community a 'heads-up' to start preparations for possible severe weather.

In 2012 there were serious floods in the Aberystwyth area in West Wales, e.g. washing caravans away. Running the "nudged" models suggested that 32 mm or more of rain was likely over most of West Wales, but there was a very high likelihood of 100 mm in the Plynlimon area, from where it would flow very fast down the rivers to the west coast - which is indeed what happened.

Air quality can also be forecast, as it was for the Olympics. Typically winds from the Atlantic come clean, while those from Europe tend to be dirty. If there is little wind, then locally generated pollution is more important. Re volcanic ash, which the President mentioned in his introduction, the Met Office can predict where the winds will carry it, but estimating how much material is being emitted requires expertise from vulcanologists, and the effect on aircraft engines is a question for the experts on those. Another example of partnership working is in the area of animal health. Strong winds from the South can bring clouds of midges from across the Channel. They can bring the blue-tongue virus with them, and in one case forecasts of this led to the pre-emptive vaccination of cattle in one area.

To produce forecasts for the Olympic sailing events in Weymouth Bay, they had to use even smaller grid squares, 300 x 300 m to allow for the shielding effect of Portland Bill.

Another area where forecasting needs the collaboration of another discipline is that of coastal floods like those in East Anglia in 1953, when a combination of North Winds and low pressure over the North Sea led to a large rise in sea level, with East coast flood defences being overtopped, with much loss of life. By combining the model of the atmosphere with one of the ocean, the risk of such events can be estimated, and warnings sent out. Another application of great interest to local authorities is the prediction of road icing. Gritting roads can cost hundreds of thousands of pounds a night, and reliable advice, based on forecasts of temperature and cloud cover, on which roads should be gritted, and how much is of huge value in keeping motorists safe without wasting large amounts of money through unnecessary gritting.

Forecasting climate change

For this, forecasts have to be extended into the future, months, years, decades or centuries ahead. And it needs models of the oceans, the sea ice and atmospheric chemistry as well as of the atmosphere and land, to be able to assess the probabilities of different scenarios. In principle the modelling is done in the same way, though the boxes have to be bigger. The UK is a world leader in this. It has the advantage that both weather and climate are forecast by the same organisation, the Met Office. Other countries tend to separate them.

There is a huge demand for forecasts of what the weather will be like a month or a season ahead. The speaker showed a graph of how monthly temperatures had departed from their long-term means over the last century. Some trends are just about visible, but the noise level is much larger than the signal. It would be useful, to the general public and to insurance companies, to be able to predict how harsh the next UK winter is likely to be. For many parts of the world this is fairly straightforward, but not here. The UK winter depends primarily on whether the prevailing wind is going to be from the West (mild) or from the North or East (harsh), and existing models, when applied to historic data, have not been very good at predicting this. A new model has been used on data from the last 20 winters. The old model tried to predict variations in the Gulf Stream temperature by dividing the Atlantic into 1° squares (of longitude or latitude). The new model uses ¼° squares. Its prediction of the winter weather is still not perfect, but the skill is much increased. The correlation between forecast and actual is 0.62, which if perhaps still not very useful to an individual planning where to spend the winter, could potentially be of huge use in some sectors such as insurance or retail.

Concluding remarks

The speaker concluded by claiming that the rapid advances of the recent years made this an exciting field to be working in - and encouraged the younger members of the audience to consider it! He then re-iterated his earlier point, that we need not only to make the forecast better, but also to engage with the user communities to make sure it is used for practical purposes.

The lecture was followed by a full question-and-answer session.

  Contents Next article   >>