On the Sea
The Navy uses computers too. At the David Taylor Test Basin in Maryland, a UNIVAC LARC is busy doing design work on ship hulls. Other computers mounted in completed Navy vessels perform navigation and gun-ranging functions. At New London, Connecticut, a Minneapolis-Honeywell computer simulates full scale naval battles. Radar and sonar screens in mock submarine command posts show the maneuvering of many ships in realistic simulations. Polaris submarines depend on special computers to launch their missiles, and the missiles themselves mount tiny computers that navigate Polaris to its target. Another computer task was the “sea testing” of the nuclear submarine “Sea Wolf” before it was launched!
Computers are being used by the Navy in a project that has tremendous applications not only for military application but for civilian use as well. Mark Twain to the contrary, a lot of people have tried to do something about the weather, among them an Englishman named Richardson. Back in 1922 he came up with the idea of predicting the weather for a good-sized chunk of England. Basically his ambitious scheme was sound. Drawing on weather stations for the data, he determined to produce a 24-hour forecast.
Unfortunately for Mr. Richardson, the English, and the world in general, the mathematics required was so complicated that he labored for three months on that first prediction. By then it had lost much of its value—and it was also wrong! The only solution that Richardson could think of was to enlist the aid of about 60,000 helpers who would be packed into a huge stadium. Each of these people would be given data upon which to perform some mathematical operation, and then pass on to the next person in line. Pages would transfer results from one section of the stadium to another, and a “conductor,” armed with a megaphone undoubtedly along with his baton, would “direct” the weather symphony, or perhaps cacaphony. As he lifted his baton, the helpers were to calculate like crazy, when he lowered it they would pass the result along. What Richardson had invented, of course, was the first large-scale computer, a serial computer with human components. For a number of reasons, this colossal machine was never completed. It was obviously much easier to simply damn the weatherman.
Actually, Richardson had stumbled onto something big. He had brought into being the idea of “numerical weather prediction.” It is known that weather is caused by the movement of air and variations in its pressure. Basically it is simple, knowing pressure conditions yesterday and today, to project a line or extrapolate the conditions for tomorrow. If we know the conditions tomorrow, we can then predict or forecast the temperature, precipitation, and winds.
There was even the mathematics to make this possible in Richardson’s day: the so-called “primitive equations” of the pioneer mathematician Euler. These are six partial differential equations involving velocity, pressure, density, temperature, and so on. But though the principle is simple, the practical application is hopelessly involved—unless you have a stadium filled with 60,000 willing mathematicians or a fast computer of some other type.
In 1950 the stage was finally set for the implementation of numerical weather prediction. First, electronic computers were available. Second, and importantly, mathematician C. G. Rossby had worked some magic with the original primitive equations and reduced them to a single neat equation with only four terms. The new tool is called the Rossby equation. Meteorologists and mathematicians at Princeton’s Institute for Advanced Study decided to combine the Rossby equation, the MANIAC computer, and some money available from the Office of Naval Research. The result was JNWPU, Joint Numerical Weather Prediction Unit, later to become NANWEP, for Navy Numerical Weather Problems Group. It is too bad that pioneer Richardson did not live to see the exploitation of his dream.
What NANWEP does is to take the meteorological data from some 3,000 reporting stations, compare them with those existing yesterday, and print out a weather map for the Northern Hemisphere for tomorrow. Because there are so many more stations reporting than the handful that Richardson used, the number of computations has risen to the astronomical total of about 300,000,000. Despite this, a Control Data Corporation 1604 digital computer does the job in a good bit less time than the three months it took Richardson. NANWEP prints out its weather maps 40 minutes from the time all data are in.
Teletype reports come in from the thousands of weather stations; these are punched on tape and fed to the 1604. Since the information includes geographical position in addition to meteorological data, the computer prints out numbers that form a map of weather coming up. Although the meteorologist adds some clarifying lines by connecting points of equal pressure, the “raw” map with its distinctive shaded areas is meaningful even to the layman.
Further refinements are in the offing. As many as 10,000 weather stations may eventually report to the central computer, which may also learn to accept the teletype information directly with no need for the intermediate step of punching a tape. Although it will be a long time before a positive forecast, exact in every detail, is possible, NANWEP already has lifted weather prediction from the educated guesswork of the older meteorologists to truly scientific forecasting.
It turns out that numerical weather prediction brings with it some bonuses. NANWEP can predict the action of ocean waves three days in advance, in addition to its regular wind, temperature, and precipitation information. So it is now being put to work preparing optimum routes for ships. Here’s the way it would work. A ship sailing from California to Japan requests the best routes for the voyage. Initially the computer is given the ship’s characteristics and told how it will perform in various sea conditions. It then integrates this information with the predicted sea conditions for the first day’s leg, and plots several different courses. Distances the ship would travel on each of these courses are plotted, and a curve is drawn to connect them. Now the computer repeats the process for the next day, so that each of the tentative courses branches out with its own alternates. The process is repeated for each of the five days of the voyage. Then the computer works backward, picking the best route for the entire voyage, and gives the course to be followed for optimum time. If that isn’t sufficiently informative for the captain, he can request and receive not one but three courses: one for the fastest trip regardless of sea condition, another for the fastest trip with waves of only a certain height, and finally a course for the fastest trip through calm water! The advantages of such a service are immediately obvious and give a hint at many other applications of the technique to air travel, truck transport, and so on.
NANWEP is ground-based, of course. There are also airborne weather computers like those of the U.S. Weather Bureau’s National Severe Storm Research Aircraft Project. The Weather Bureau has jumped its computer budget from $1.5 to $2.5 million to extend this and other projects. The compact airborne computers ride along in DC-6 and B-57 aircraft to monitor hurricanes off Florida and tornadoes in the Great Plains area. The computers gather forty different kinds of information and convert it to digital form at thousands of characters a second. Such monitoring of violent weather by means of computers suggests an intriguing use of the machine. Man has long considered the prospect of going the step beyond weather recording and prediction to actually changing or even creating his own weather. He has done a few rather startling things of this kind, admittedly on a small scale but with tremendous implications. Cloud-seeding experiments are samples, as attempts both to induce precipitation and to create or destroy storms. These experiments, though inconclusive, have led to results—including precipitation of lawsuits and ill feeling. Meteorologists attempted to divert a hurricane along the Atlantic coast line once, apparently with results. But the storm swerved too far and the weathermen incurred the justifiable wrath of those living in the area affected. Why not simulate such an experiment in the computer? Besides being safer, it is also far cheaper. In the long run, we may do something about the weather at that.