The origin of numerical weather forecast

As early as the early 20th century, the British scientist L.F. Richardson first tried numerical weather forecasting. 1922, in his book Numerical Methods of Weather Forecast, he discussed the principle and possibility of numerical forecast, and applied the complete primitive equation to forecast the ground pressure field in Europe for 6 hours. However, the result is not ideal: he predicted that the air pressure in this area would change to 154 millibar (hectopascal) within 6 hours, but the actual air pressure hardly changed. At that time, Richardson blamed the failure on the inaccurate initial value, and his failure once made people doubt the practical possibility of numerical weather forecast. Until the end of World War II, due to the appearance of computers and the development of meteorological observation networks, especially high-altitude observation, meteorological data have been greatly improved, and numerical weather forecasting has been paid attention to. In particular, people realize that Richardson's failure mainly lies in solving the equations he used, including not only slow processes such as long waves, but also high-speed sound waves and gravity waves. The actual amplitude of these high-speed waves is very small, but they are often amplified in the calculation process to cover up the meaningful meteorological disturbance. 1948, J.G. Qiu Chen put forward the filtering theory based on the work of C.-G. Rosby and others, and proved that the approximation of static balance and geostrophic balance can eliminate gravity waves and sound waves (see the equilibrium state of atmospheric motion). The simplified equation thus established avoids the influence of acoustic waves and gravity waves. 1950, Qiu Chen, R. F. Yotov and J. von Neumann successfully predicted the pressure field of 500 hectopascals in North America for the first time by using the quasi-geostrophic barotropic model.