When the Royal Meteorological Society was founded, weather forecasting was a sorry business: methods were dubious and results unreliable. Worse still, prospects for improvement looked bleak. Around 1900, Vilhelm Bjerknes realised that the development of thermodynamics had completed the jig-saw of atmospheric laws, and that these laws provided a rational basis for precise scientific forecasting. But his was a method without a means: he saw no possibility to put his ideas to practical use. Lewis Fry Richardson was bolder, attempting a direct assault on the equations of motion. His methodology was unimpeachable but his results disastrously wrong. His glorious failure discouraged further investigation along these lines.
Developments in meteorology and numerical analysis in the ensuing decades, and the invention of the radiosonde and digital computer, led to the possibility of a practical implementation of Bjerknes' and Richardson's ideas. In 1950 the first computer forecast was made using a greatly simplified mathematical model. It was so successful that within ten years computer predictions were used in operational forecasting. Numerical weather prediction has advanced rapidly, with the development of efficient algorithms and refinement of the treatment of physical processes. By Trojan efforts, assimilation techniques have been devised to squeeze benefits from quantitative satellite data. Forecasts with some skill up to a week ahead are now routine.
Progress since 1950 has been spectacular. Advances have been based on sound theoretical principles so that the woolly art of weather forecasting has been ennobled to a precise scientific discipline.