This data allowed Richardson to model a mathematical weather forecast. Of course, he already knew the weather for the day in question (he had Bjerknes’s record to hand, after all); the challenge was to generate from this record a numerical model which he could then apply to the future. And so he drew up a grid over Europe, each cell incorporating Bjerknes’s weather data, including locational variables such as the extent of open water affecting evaporation, and five vertical divisions in the upper air.
Richardson claimed that it took him six weeks to calculate a six-hour forecast for a single location. Critics have wondered whether even six weeks was enough time. In any case, the first numerical forecast was woefully out of sync with what actually happened. Not only did Richardson’s forecast take longer to calculate than the weather it was calculating took to happen, but it was also a prediction after the fact that remained manifestly wrong.
Yet scientific failures of this magnitude often have important consequences, not least in this case because Richardson’s mathematical approach to weather forecasting was largely vindicated in the 1940s with the invention of the first digital computers, or “probability machines”. These are still the basis for much weather forecasting today. His experiment also contributed to the development of an international field of scientific meteorology.