In the winter of 1917 Lewis Fry Richardson was driving an ambulance for a French infantry division on the Western Front. At the same time, in spare moments behind the lines, he was completing a vast project of mathematical calculations. He had set out to compute the weather, to predict the temperature, the winds and the barometric pressure from first principles of physics. It would have been an ambitious undertaking in the best of circumstances—in a quiet study with a long oak table for spreading out the paperwork. Richardson did it in the middle of a war that was, among its other distinctions, the muddiest in human history. "My office," he wrote, "was a heap of hay in a cold rest billet."
Richardson's goal was to follow the development of the weather for a six-hour interval in a small area of central Europe. Even a forecast of this limited scope called for a calculation of daunting complexity. Needless to say, he had no electronic computer to do the arithmetic for him. He worked with pencil and paper, as well as a slide rule and a table of logarithms. To guard against careless mistakes he did everything twice.
The outcome was not a conspicuous success. The actual weather changed little over the period of the forecast, but Richardson's equations had the barometer rising fast enough to make your ears pop. The calculation was a brilliant piece of work nonetheless. Although he failed to predict the weather, he predicted the future of weather prediction. The forecast you now see on the six o'clock news is based on simulations remarkably like Richardson's (including the occasional error).
What went wrong with that first-ever numerical forecast? Many commentators have suggested explanations, starting with Richardson himself. A new analysis takes a direct approach to understanding the source of the problem. Peter Lynch of Met Éireann, the Irish Meteorological Service, has carefully reconstructed and replicated the entire calculation—everything but the heap of hay in the rest billet. He has been able to reproduce Richardson's error, and even to correct it.