Before 1955: Numerical Models

and the Prehistory of AGCMs

In the early 20th century, the Norwegian Vilhelm Bjerknes argued that atmospheric physics had advanced sufficiently to allow weather to be forecast using calculations. He developed a set of seven equations whose solution would, in principle, predict large-scale atmospheric motions.

Bjerknes proposed a "graphical calculus," based on weather maps, for solving the equations. Although his methods continued to be used and developed until the 1950s, both the lack of faster calculating methods and the dearth of accurate observational data limited their success as forecasting techniques.

Richardson's "Forecast Factory"

In 1922, Lewis Fry Richardson developed the first numerical weather prediction (NWP) system. His calculating techniques -- division of space into grid cells; finite difference solutions of differential equations -- were the same ones employed by the first generations of AGCM builders. Richardson's method, based on simplified versions of Bjerknes's "primitive equations" of motion and state (and adding an eighth variable, for atmospheric dust) reduced the calculations required to a level where manual solution could be contemplated. Still, this task remained so large that Richardson did not imagine it as a weather forecast technique. His own attempt to calculate weather for a single eight-hour period took six weeks and ended in failure.

His model's enormous calculation requirements led Richardson to propose a fanciful solution he called the "forecast-factory." The "factory" -- really more like a vast orchestral performance -- would have filled a vast stadium with 64,000 people. Each one, armed with a mechanical calculator, would perform part of the calculation. A leader in the center, using colored signal lights and telegraph communication, would coordinate the forecast.

Yet even with this fanciful apparatus, Richardson thought he would probably be able to calculate weather only about as fast as it actually happens. Only in the 1940s, when digital computers made possible automatic calculation on an unprecedented scale, did Richardson's technique become practical.

Computers, Weather, and War in the 1940s

The Princeton mathematician John von Neumann was among the earliest computer pioneers. Engaged in computer simulations of nuclear weapons explosions, he immediately saw parallels to weather prediction. (Both are non-linear problems of fluid dynamics.) In 1946, soon after the ENIAC became operational, von Neumann began to advocate the application of computers to weather prediction.[3] As a committed opponent of Communism and a key member of the WWII-era national security establishment, von Neumann hoped that weather modeling might lead to weather control, which might be used as a weapon of war. Soviet harvests, for example, might be ruined by a US-induced drought.[4]

Under grants from the Weather Bureau, the Navy, and the Air Force, he assembled a group of theoretical meteorologists at Princeton's Institute for Advanced Study (IAS). If regional weather prediction proved feasible, von Neumann planned to move on to the extremely ambitious problem of simulating the entire atmosphere. This, in turn, would allow the modeling of climate. Jule Charney, an energetic and visionary meteorologist who had worked with Carl-Gustaf Rossby at the University of Chicago and with Arnt Eliassen at the University of Oslo, was invited to head the new Meteorology Group.

The Meteorology Project ran its first computerized weather forecast on the ENIAC in 1950. The group's model, like Richardson's, divided the atmosphere into a set of grid cells and employed finite difference methods to solve differential equations numerically. The 1950 forecasts, covering North America, used a two-dimensional grid with 270 points about 700 km apart. The time step was three hours. Results, while far from perfect, justified further work.

The Swedish Institute of Meteorology

The Royal Swedish Air Force Weather Service in Stockholm was first in the world to begin routine real-time numerical weather forecasting (i.e., with broadcast of forecasts in advance of weather). The Institute of Meteorology at the University of Stockholm, associated with the eminent meteorologist Carl-Gustaf Rossby, developed the model. Forecasts for the North Atlantic region were made three times a week on the Swedish BESK computer using a barotropic model, starting in December, 1954.[6]

The Joint Numerical Weather Prediction Unit

About 1952, Von Neumann, Charney, and others convinced the Weather Bureau and several research and forecasting agencies of the Air Force and Navy to establish a Joint Numerical Weather Prediction (JNWP) Unit. The JNWP Unit opened in Suitland, Maryland in 1954, under George Cressman. It began routine real-time weather forecasting in May, 1955.[7] Yet it was well over a decade before numerical methods began to outstrip in accuracy the "subjective method" employed by human forecasters.

Initially, the computer models used for NWP employed simplifying assumptions. Only in the 1960s did models based on the Bjerknes/Richardson primitive equations replace barotropic and baroclinic models.

Top of page | Back to Introduction | Forward to 1955-65 | View AGCM Family Tree


[1] F. Nebeker, Calculating the Weather: Meteorology in the 20th Century (New York: Academic Press, 1995).

[2] L.F. Richardson, Weather Prediction by Numerical Process (Cambridge: Cambridge University Press, 1922).

[3] W. Aspray, John von Neumann and the Origins of Modern Computing (Cambridge, MA: MIT Press, 1990).

[4] C. Kwa, "The Rise and Fall of Weather Modification," in Changing the Atmosphere: Expert Knowledge and Global Environmental Governance, eds. C.A. Miller and P.N. Edwards and (Cambridge, MA: MIT Press, forthcoming, 2000).

C. Kwa, "Modelling Technologies of Control," Science as Culture 4, no. 20 (1994): 363-391.

[5] J.G. Charney, R. Fjörtoft, and J.v. Neumann, "Numerical integration of the barotropic vorticity equation," Tellus 2 (1950): 237-254.

G.W. Platzman, "The ENIAC computations of 1950 -- gateway to numerical weather prediction," Bulletin of the American Meteorological Society 60 (1979): 302-312.

[6] Staff Members of the Institute of Meteorology, University of Stockholm, "Results of Forecasting with the Barotropic Model on an Electronic Computer (BESK)," Tellus 6 (1954): 139-149.

P. Bergthorsson et al., "Routine Forecasting with the Barotropic Model," Tellus 7 (1955): 272-274.

[7] F. Nebeker, Calculating the Weather: Meteorology in the 20th Century (New York: Academic Press, 1995).

Top of page | Back to Introduction | Forward to 1955-65 | View AGCM Family Tree