The hidden science of weather forecasting

Guest post by Vincent Kelly

In one hilarious episode of “Curb Your Enthusiasm”, the principal grump – Larry David – becomes convinced that an acquaintance of his, a meteorologist, is a fraud. This acquaintance, he comes to suspect, forecasts rain only on days he is due to play golf, thus ensuring an empty course as everyone else stays away. He confronts his friend with his theory, who predictably reacts with righteous indignation. But Larry is having none of it, and is clearly convinced that his friend is a charlatan and, by extension, weather forecasting is a sham science. The final scene sees Larry, having convinced his other friends that the previous day’s forecast of rain was nothing but a ruse, stand on the golf course in a torrential downpour, looking thoroughly miserable. They have not an umbrella between them.

I’ve always had a bit of Larry in me on this subject. As a committed cyclist I am more finely attuned than most when the weather forecast gets it spectacularly wrong.  Probably, along with Larry, my cynicism of the science of weather forecasting was driven by my complete ignorance of how they come up with their predictions. So I sought to rectify this, to see if there was scientific basis for what is predicted, or if it is indeed – as Larry would contend – a charade.

Since the dawn of civilization, humans have attempted to forecast the weather. However, these forecasts were generally based on pattern recognition and usually were of the form of pithy aphorisms – some of wisdom  (“Red sky at night, shepherd’s delight; red sky in the morning, shepherd’s warning”, “No weather be ill, if the wind be still”) and some of more dubious merit (“Seagull, seagull, sit on the sand; it’s never good weather when you’re on land”, “Onion skins very thin, mild winter coming in; onion skins thick and tough, coming winter cold and rough”). Bill Murray, in the film “Groundhog Day”, plays a weatherman sent to cover the annual weather forecast based on the behaviour of a groundhog.  This is an old piece of weather lore in the US – if the designated groundhog sees its shadow, then winter will last another month – otherwise Spring will imminently be sprung. Good film; dubious concept.

All modern forecasting uses the Numerical Weather Prediction system (NWP). This is a mathematical model which aims to predict the evolution of weather systems based on the current weather status. The inputs to the model are numerous – inputs based on surface observations (barometric readings, temperature, wind direction, wind speed, precipitation), coastal weather stations, weather ships, buoys, radiosondes (helium filled balloons with transmitters, released into the atmosphere), commercial ships and planes, meteorological satellites (taking pictures of cloud formation) and weather radars. The various weather agencies, national or otherwise, pool their information so that sufficient data is available for the area, and its surrounds, of which a forecast is being made. This can mean thousands of miles either side of the longitudes.

This dazzling array of data is fed into the computer system, which then attempts to “roll forward” the current weather status into the future, using equations derived from studies of the physical processes at work in the atmosphere.

For example, suppose the range of inputs tell us that a cyclonic weather system is sitting off the west coast of Ireland. Wind speed and direction, and weather theory’s prediction of how it will interact with other systems, will tell us if and when we can expect it to sweep across Ireland. Barometric measurements and studies of cyclonic systems can tell us what weather we can expect it to bring. And bingo, we have forecast of another wet Monday in Ireland.

Weather theory is so advanced that forecasts are, by and large, very accurate over short time periods. However, despite the sophistication of modern weather techniques, the accuracy of forecasts falls dramatically as the time period is extended from 1 to 5 days. Further out than 5 days and any forecast, must be taken with a large grain of salt. The reason for this is the inherent chaos in any complex system. This was described by Edward Lorenz memorably as the “Butterfly Effect” – that small changes in the initial conditions can have large, unintended effects.

Modern weather forecasting addresses the Butterfly effect by using “ensemble forecasting”. The model may be run thousands of times, relying on databases of 50 years of time series – if in 60% of these scenarios, it is raining in Dublin in three days time – then the model is estimating the probability of rain in Dublin in three days time at 60%. Thus modern forecasts are probabilistic best estimates.

The weight of science in weather forecasting, the multitude of disciplines that it combines (meteorology, climatology, hydrometeorology, physics, fluid dynamics, hydrology, astronomy, telecommunications, linear algebra, partial differential calculus, chaos theory, aerology, nephology to name but a few) and the sheer range of uses to which it is put (agriculture – when the harvests can proceed; maritime – whether that boat should sail; air – whether that flight should take off, when it will land; supermarkets and electricity companies – try to estimate future demand based on forecast weather; extreme weather warnings – tornadoes and floods) must rank the science of weather forecasting as one of the most significant achievements in the history of mankind. We can now do what hundreds of generations of our ancestors dreamed – accurately predict whether we should wear the waterproof trousers or not.

Thus, following a bit of research, I have to say I now disagree with Larry David – although I do, as always, sympathize with his plight.

Vincent Kelly works as an actuary in Dublin, and is interested in science and mathematics, particularly applied mathematics. He has a BSc in Financial and Actuarial Mathematics from DCU.