Meteorologists Make the First Computerized Weather Prediction

Date April, 1950

Using one of the first electronic digital computers, John von Neumann and his colleagues attempted to prove that future weather could be predicted mathematically by quantitatively analyzing atmospheric conditions.

Also known as “Numerical Integration of the Barytropic Vorticity Equations”

Locale Princeton, New Jersey

Key Figures

• John von Neumann (1903-1957), Hungarian-born American mathematician and physicist
• Jule Gregory Charney (1917-1981), American mathematician and meteorologist
• Edward Lorenz (b. 1917), American meteorologist
• Lewis Fry Richardson (1881-1953), British physicist and mathematician
• Vilhelm Bjerknes (1862-1951), Norwegian meteorologist

Summary of Event

In April, 1950, in one of the first applications of electronic computers to civilian use, mathematicianJohn von Neumann and two collaborators attempted to predict the weather over the North American continent over a twenty-four-hour period. Electronic digital computers had been developed during World War II, in order to perform mathematical calculations of military interest much more quickly than could be done by pencil-and-paper calculation. At the close of the war, scientists were eager to identify civilian applications for the new computing machines.

Von Neumann had been one of the many eminent intellectuals to emigrate from Germany as the Nazi Party came to power. He had come to the United States in 1933, settling in Princeton, New Jersey, where he was given a position at the Institute for Advanced Study, a research institute on the Princeton University campus at which leading scientists, mathematicians, and other scholars could devote their full energies to basic research, freed of the teaching responsibilities of most university faculty members. Within a decade, von Neumann had become a consultant to numerous government projects, including the atomic bomb project, and had assisted in the development of the first large-scale electronic computer, placed into service at the University of Pennsylvania in 1943. At the end of the war, he returned to the Institute for Advanced Study while retaining many of his consulting positions. He was active in trying to persuade the military services to fund the construction of computers for civilian purposes, including weather prediction.

The first attempt to predict the weather by mathematically treating the motion of air, moisture, and heat had been made by the Englishman Lewis Fry Richardson, who published his results in 1922 in a book, Weather Prediction by Numerical Process. Richardson’s methods provided the basic framework within which almost all subsequent work has been done. The basic equations were much too complicated for any sort of exact solution, and the only possibility of solving them was to find approximate numerical values of pressure, wind speed, and temperature as a function of time, given a set of numerical values for them at some initial time.

Before the invention of digital computers, this sort of approximation had to be achieved by human computation, painstakingly performed and checked by arithmetical calculations for long hours. Richardson’s predictions, for a six-hour period on May 20, 1910, were far from satisfactory. Von Neumann and his group later pointed out that the specific combination of approximations used by Richardson was prone to magnify the effects of small errors very quickly. Had more been known about approximate mathematics at the time, Richardson might have had more encouraging results.

Von Neumann’s results were published in the second volume of a new scientific journal, Tellus, being introduced by the Swedish Geophysical Union. The authors of the paper, “Numerical Integration of the Barytropic Vorticity Equations,” were Jule Gregory Charney, a young scientist with a short-term postdoctoral appointment at the Institute for Advanced Study, Ragnar Fjörtoft, a meteorologist on leave from the Norwegian Meteorological Institute, and von Neumann. The title refers to a set of approximations, basically a neglect of the exchange of air between different layers of the atmosphere, that had to be made to simplify the equations for the computers of the time, which were very slow compared to more modern machines. The published paper expresses thanks to the U.S. Army for allowing the computation to be done on one of the military computers.

The calculation was an example of computer simulation, a technique in which a computer program is given information about the state of a physical system at one instant of time, in this case information about pressures and temperatures over some region of the earth’s surface. The equations governing the behavior of the system are used to calculate values for the same quantities at a slightly later instant, the calculated values then being used to calculate values for a later time, and so on. The calculation attempted to predict the weather over North America and the surrounding ocean regions for twenty-four-hour periods beginning on four different dates in 1949. The predictions varied in their ability to describe the overall movements of high- and low-pressure areas, and much of the paper is devoted to describing the possible sources of inaccuracy. Using the computers available at the time, predicting the weather over a twenty-four-hour period required nearly twenty-four hours of actual computation time.

Charney went on to a distinguished career as professor of meteorology at the Massachusetts Institute of Technology. As computers became faster and less expensive and meteorological data became more readily available, numerous other scientists became involved in attempts to simulate weather on the computer. Another researcher at the Massachusetts Institute of Technology, Edward Lorenz, discovered that some of the basic equations exhibited the kind of behavior that mathematicians have come to refer to as “chaotic,” meaning that the behavior of the system demonstrated extreme sensitivity to its initial conditions. In other words, very slight differences in the values of one or more of the quantities described by the equations would result in vastly different outcomes at later times.

Because weather is a chaotic system, it becomes increasingly difficult to predict accurately over time, as smaller and smaller initial events come to have greater and greater impacts. This effect is sometimes referred to as the “butterfly effect” for the notion that a single butterfly, flapping its wings somewhere in Asia, might be responsible for a blizzard taking place in North America. The chaotic nature of the system means that it may be impossible in principle to predict the weather with precision over a period of months or even weeks, although within those limits increasingly accurate weather predictions can be made by using more detailed input information and calculating results for smaller increments of time. Thus, unlike astronomy, in which the positions of planets could be computed centuries in advance, meteorology may forever remain primarily an observational science.

Significance

Throughout history, the inability to predict the weather has added major elements of uncertainty to human planning. Being able to predict the occurrence of rainstorms has always been important to farmers, sailors, and the commanders of troops. With the coming of industrialization, and the burning of coal for heat and power, urban areas became dependent on the circulation of the atmosphere to remove smoke particles and the products of combustion. With the development of motorized traffic and a diverse chemical industry, the distribution of carbon monoxide and ozone through the lower atmosphere became a matter of interest. The subsequent development of aviation, and later, of vast chemical and nuclear power plants, added vital importance to predictions of atmospheric movement. When an industrial accident results in the release of toxic or radioactive materials to the environment, knowledge of wind and precipitation patterns may be essential to identifying the areas of greatest danger to the population, so that evacuation or other protective measures might be taken.

Although scientists had learned to associate different types of clouds with different types of weather and to measure atmospheric pressure as it changed from day to day, accurate prediction of the weather was not feasible until the invention of the telegraph in 1844 made it possible to collect weather reports from a wide range of locations in a very short time. By the end of the nineteenth century, many nations had established weather bureaus, and meteorologists began to develop predictions based on similarities between present weather conditions and past weather conditions as recorded on detailed weather maps. The first meteorological predictions were based solely on past observations. By deciding which previous weather maps most closely resembled current conditions, the most likely future developments were predicted based on what had happened in the past.

At the beginning of the twentieth century, a few farsighted weather scientists began to propose that future weather could be predicted from the equations that physicists had developed to describe the motion of air, humidity, and heat. The first serious efforts in this direction took place around the year 1900 at the Norwegian Geophysical Institute under the leadership of Vilhelm Bjerknes. The extreme complexity of the equations, and the detailed information which had to be input into the calculation, did not allow much progress to be made, despite the efforts of such determined men as Lewis Richardson, until the invention of the digital computer.

Even with the advent of the computer, accurate predictions could not be made without far more information about the state of the atmosphere. It turned out that von Neumann’s success was in part a result of luck, as subsequent simulations using the same general approach resulted in far less accurate predictions. The missing information was supplied by weather balloons and earth satellites. Investigation of the upper atmosphere had been one of the major motivations for the development of the hot-air balloon. By the mid-nineteenth century, serious mapping of the lower atmosphere using piloted balloons had begun. The first earth satellite to be dedicated to weather prediction was the Television Infrared Observation Satellite (TIROS 1), launched in 1960. The rapid evolution of weather satellite technology since then has led to a great increase in the accuracy of short-term weather forecasts.

As scientific understanding of the weather increased, computer prediction of the weather, like the field of meteorology itself, eventually divided into several specialized areas. Attempts to predict the weather over large distances accurately, covering a continent or more, continued to be conducted out of general scientific interest. Attempts to simulate the weather over regions of a smaller scale took on importance as the potential for inadvertent release of airborne pollutants from industrial processes or power generation became appreciated. Such regional models can also be used for urban and industrial planning, by allowing the prediction in advance of the highest levels of ozone, carbon monoxide, or other pollutants that would accumulate as the result of the siting of new roads or industrial plants.

Bibliography

Charney, Jule G., Ragnar Fjörtoft, and John von Neumann. “Numerical Integration of the Barytropic Vorticity Equation.” Tellus 2 (November, 1950): 237-254. The original published report of von Neumann’s work is available in many university libraries. The introductory section, describing the motivation for the calculations, and the conclusion, summarizing the results, are accessible to the educated layperson.

Fiedler, Franz. “Development of Meteorological Computer Models.” Interdisciplinary Science Reviews 18 (September, 1993): 192-198. A nontechnical article describing the use of computer simulation to characterize the dispersal of pollutants over a region of modest size, taking into account the terrain and distribution of vegetation.

Gleick, James. Chaos: The Making of a New Science. New York: Viking Penguin, 1987. This book, which did much to popularize the ideas of chaos theory among general readers, begins with a detailed discussion of the butterfly effect discovered in computer simulations of the weather by Edward Lorenz.

Hargittai, István. The Martians of Science: Five Physicists Who Changed the Twentieth Century. New York: Oxford University Press, 2006. Tells the stories of von Neumann and four other brilliant scientists born in Budapest around the same time, all of whom emigrated to the United States and remained friends as they developed world-changing theories and technologies. Bibliographic references and index.

Heims, Steve J. John Von Neumann and Norbert Weiner: From Mathematics to the Technologies of Life and Death. Cambridge, Mass.: MIT Press, 1980. The development of computer weather prediction could not have occurred without the development of computer technology for military applications and a redefinition of the relationship between scientists and government. This process is described through the personal stories of von Neumann and other great mathematicians.

Macrae, Norman. John von Neumann: The Scientific Genius Who Pioneered the Modern Computer, Game Theory, Nuclear Deterrence, and Much More. Providence, R.I.: American Mathematical Society, 1999. Detailed examination of von Neumann’s contributions to science and culture. Bibliographic references and index.

Richardson, Lewis F. Weather Prediction by Numerical Process. New York: Dover, 1965. This is a reprint of the 1922 edition published by Cambridge University Press, to which a contemporary introduction has been added by Sidney Chapman. Although the bulk of this book requires familiarity with advanced mathematics, the layperson will find the introductory sections quite understandable.

Thompson, Philip D., and Robert O’Brien. Weather. New York: Time-Life Books, 1965. A general introduction to the subject of weather that gives a good overview of the introduction of computers into weather forecasting, including some historically significant photographs.