Could a hall full of human calculators have forecast the weather?

In principle yes, because the hall would have run the method used today, but by hand it was hopelessly slow: Richardson's own six-hour forecast took at least six weeks. The English scientist Lewis Fry Richardson proposed forecasting by solving differential equations, and published Weather Prediction by Numerical Process in 1922, when suitable fast computing was unavailable. The core idea is to treat the atmosphere as a fluid, sample its state at one moment, and use the equations of fluid dynamics and thermodynamics to estimate its state later.

To show how it might work, Richardson allowed himself a fantasy. Imagine a large hall like a theatre, with walls painted as a map of the globe. A myriad human computers are at work, each on the weather of the part of the map where they sit, each attending only to one equation or part of an equation. A man in a pulpit keeps all regions at a uniform speed, like the conductor of an orchestra whose instruments are slide-rules and calculating machines.

His real attempt was brutal. Using procedures developed by Vilhelm Bjerknes, Richardson produced by hand a six-hour forecast for two points in central Europe, and it took at least six weeks. The result looked like a failure: his forecast predicted a huge 145 hectopascal pressure rise over six hours, when the pressure was actually more or less static. But meteorologist Peter Lynch later showed that the problem was a failure to smooth the data to remove unphysical surges in pressure. With smoothing, Richardson's forecast is essentially accurate, a remarkable achievement considering the calculations were done by hand while he was serving with the Quaker ambulance unit in France.

Machines caught up. The ENIAC made the first computer weather forecasts in 1950. Richardson called it an enormous scientific advance, although the first 24-hour forecast took the ENIAC nearly 24 hours to produce.