Picture an immense concert hall filled with 64,000 computers 鈥� not machines but people who compute figures. This entire hall functions as a map: 鈥淭he ceiling represents the north polar regions. England is in the gallery, the tropics in the upper circle, Australia in the dress circle and the Antarctic in the pit.鈥� As weather data pours in from around the world, coloured stage lights cue the human computers to work in unison. Towering above them in a pulpit is 鈥渁 conductor of an orchestra in which the instruments are slide rules and calculating machines. But instead of waving a baton he turns a beam of rosy light upon any region that is running ahead of the rest and a beam of blue light upon those who are behindhand.鈥� Lewis Fry Richardson鈥檚fanciful proposalfora 鈥渟lide-rule orchestra鈥� was the culminatingvision of one of the most ambitious attempts ever made at weatherforecasting.
FOR a man who liked to be alone, Lewis Fry Richardson had come up with an idea guaranteed to make him feel uncomfortable. In 1913, he was superintendent of the Eskdalemuir Observatory, a remote weather station in Scotland. It was the perfect job for someone who counted solitude among his hobbies. Three years earlier, however, an event had taken place that had him yearning for a vast crowd of people 鈥� so long as everyone in it was handy with a slide rule.
On 20 May 1910, under the direction of the Norwegian meteorologist Vilhelm Bjerknes, weather stations across Europe simultaneously collected measurements from weather balloons at six different altitudes over a 6-hour stretch. The idea of all that data would make most heads reel. To Richardson it offered the prospect of creating the grandest of weather forecasting schemes. If he worked backwards from two sets of balloon data gathered across Europe that day, it might be possible to come up with differential equations 鈥渉indcasting鈥� the weather patterns that led to those readings. These could then be applied to current weather readings to forecast future weather.
Advertisement
The notion had also occurred to Bjerknes in 1904, but he had dismissed it as a Herculean labour. Richardson obligingly became meteorology鈥檚 Hercules. Laying a grid over a map of Europe and dividing it into 25 cells, each with five layers of altitude, Richardson faced an immense number of variables and a staggeringly complicated task for calculation. Worse, he had undertaken it in the most hideous environment imaginable. Prompted by his pacifist Quaker conscience, Richardson gave up his beloved observatory job to volunteer for ambulance duty on the battlefields of France in 1916.
He worked in harrowing conditions, perfecting his equations between shifts spent transporting shattered soldiers from the front; a heap of hay served as his desk. He had the particular misfortune to serve in the third battle of Champagne in April 1917. As the French army retreated and chaos ensued, Richardson had to abandon his precious forecasting manuscript. Five years later, he recounted that 鈥渢he working copy was sent to the rear, where it became lost, to be later rediscovered under a heap of coal鈥�.
Further trials were to come. Once Richardson had his final equations from two sets of readings, he checked their accuracy by forecasting the weather recorded by the third set of readings from 20 May 1910. If his forecast matched the data, he would know the equations were accurate. It did not quite work out that way. His forecast for Munich, for example, showed a huge drop in barometric pressure. In fact, the weather there had been rather calm, so clearly something was amiss. Richardson suspected bad data: sloppy figures from just one or two weather stations would throw a spanner in the works. Even worse, the magnitude of the errors would multiply if he tried forecasting even further ahead, and soon his model would be projecting apocalyptic weather.
鈥淟ewis Fry Richardson鈥檚 ideas proved startlingly prescient of computerised meteorology鈥�
鈥淗is methodology was unimpeachable,鈥� says Peter Lynch of the Irish Meteorological Service, 鈥渂ut his results were disastrously wrong.鈥� An analysis by Lynch in 1999 revealed that data was indeed the culprit, but not necessarily because the measurements were bad. The problem was what was measured. Massive but fleeting shifts in barometric pressure known as gravity wave oscillations, which bear little relation to the true overall shifts in pressure, created data 鈥渘oise鈥�; left unfiltered, these readings hopelessly skewed Richardson鈥檚 results.
It is a mark of Richardson鈥檚 confidence in his methods that he pressed ahead with his failed experiment anyway, publishing the results in 1922 as Weather Prediction by the Numerical Process. Even if his equations and data were wrong, he believed the fundamental methodology constituted an important advance in meteorology. But despite its exhilarating vision of a world where weather prediction would be performed by a massive slide-rule orchestra, initial reaction to Richardson鈥檚 book was muted. The Cambridge University Press only published the book after a partial subsidy from Richardson, and the scepticism proved justified. Even with a modest run of 750 copies, 30 years later a number remained unsold.
It hardly helped that the book was dauntingly and unapologetically difficult. 鈥淭he scheme is complicated,鈥� Richardson explained, 鈥渂ecause the atmosphere itself is complicated.鈥� He was unimpressed by the widespread perception in the 1920s that Einstein and other physicists were stripping physical laws down to an elegant simplicity: 鈥淚f they would condescend to attend to meteorology the subject might be greatly enriched. But I suspect that they would have to abandon the idea that truth is really simple.鈥�
However, one group was intrigued by the complexities of his work, much to Richardson鈥檚 distress. As a dedicated pacifist, Richardson was appalled to find that when the war ended his new job at the British Meteorological Office was to come under the control of the Air Ministry. And when he discovered that those with the keenest interest in his atmospheric research were scientists working on poison gases, he destroyed all his work that had not yet been published and left. His departure from the Met Office was one of the great losses to meteorology of the era.
He would not be moved from his fundamental belief, though, which he still expressed nearly 40 years later: 鈥淭he moral evil in war outweighs the moral good, although the latter is conspicuous.鈥� Richardson threw himself into the statistical study of the causes and prevention of warfare, a problem that he had been contemplating ever since his days on the battlefields of France. War, like weather, was a seemingly chaotic phenomenon that people accepted fatalistically but which Richardson believed amenable to quantitative analysis. The immense task Richardson set himself was never quite completed, even though it occupied him until his death in 1953.
But then, his reach always did exceed his grasp. Even the whimsical idea of his slide-rule orchestra was bigger than Richardson realised. As Lynch points out, Richardson鈥檚 staggering figure of 64,000 human computers was also short of the mark 鈥� something like 204,800 would have been needed for a global forecast. 鈥淓ven this vast multitude could compute the weather only as fast as it was evolving,鈥� Lynch says. 鈥淭o obtain useful and timely predictions, the calculations would need to go several times faster than the atmosphere. Allowing for a speed-up factor of five, the establishment of a practical forecast-factory would have reduced the ranks of the unemployed by over a million.鈥�
Richardson鈥檚 slide-rule orchestra, it seems, hit some wrong notes. But even in Weather Prediction, the author admitted his methodology outstripped the means to implement it: 鈥淧erhaps some day in the dim future it will be possible to advance the computations faster than the weather advances and at a cost less than the saving to mankind due to information gained. But that is a dream.鈥�
Yet that dream came true 鈥� not in rooms filled with 64,000 humans, but through a very different sort of 64K. Richardson鈥檚 ideas proved startlingly prescient of computerised meteorology, and as computing power increased there was renewed interest in his work 鈥� enough to reprint Weather Prediction By the Numerical Process in 1965. This time, it sold thousands of copies. Lewis Fry Richardson became an eccentric hero of meteorology, and his work is part of the foundation of modern forecasting algorithms. Not bad for failed calculations scratched out atop a heap of hay.