Recently, a professor of meteorology at the Massachusetts Institute of Technology passed away. Dr. Edward N. Lorenz was 90 years old, and had a long and distinguished career at MIT, including serving as the head of the department of meteorology. Typically the passing of a long-standing faculty member will receive some notice at his institution and in his hometown, and that would be the end of it. However, Dr. Lorenz was far from typical, which is why his obituary appeared in the New York Times, the Washington Post, the Los Angeles Times, the London Times, the Toronto Globe and Mail and many other newspapers in the U.S. and worldwide. His passing also received notice from all U.S. broadcast television networks and cable news outlets like CNN and Fox News.
So why all the fuss about a meteorology professor? It rests with an inadvertent discovery made by Dr. Lorenz while doing research on numerical weather prediction in the 1960s.
Numerical weather prediction is carried out by defining the current state of the atmosphere by the values of several parameters (pressure, temperature, humidity, etc.) and by representing the processes that go on in the atmosphere with a system of equations. The numbers representing the current state of the atmosphere are inserted into the equations, which are then used to predict the future state of the atmosphere, thus generating a forecast. Because the calculations are so numerous and so complex, they are carried out by high-speed computers, and so when you hear a television weather forecaster cite “the computer models” while presenting a forecast, the forecaster is referring to numerical weather prediction.
While Dr. Lorenz was working with one of these computer models, he ran the same set of data through the same computer model twice. At least he thought it was the same set of data. When the results were wildly different, he looked a bit more carefully at his initial data sets, and found that one of the input values was slightly different. The difference arose because a number had been truncated — that is, in one of the data sets the digits in the fourth or fifth decimal place had been left off. Think about a piece of teak for a marine project, and having specs which require that teak to be cut 12.34567 inches long, but when the piece is actually cut, it is only 12.345 inches long. A tiny, almost immeasurable difference, but in the case of the computer model of the atmosphere, it turned out to be very significant. Dr. Lorenz did more experimentation that led to his development of what has come to be known as the chaos theory. After more research, he pub lished a paper titled, Predictability: Does the Flap of a Butterfly’s Wings in Brazil Set Off a Tornado in Texas?
His premise with chaos theory was that there was an upper limit to the predictability of the atmosphere, and by extension, other natural systems. He suggested that very small perturbations in the data, perhaps even too small to be measured, could eventually lead to much larger scale disturbances. The fact that these small disturbances are very difficult to measure, let alone predict, led him to conclude that no matter how precisely we can measure the atmosphere, or how sophisticated the mathematical models become, a perfect prediction was not possible.
So how does this impact today’s mariners? Computer models have become more powerful and more sophisticated in the decades since Lorenz’s discovery, and while these models have certainly improved the predictability of the atmosphere, they have not been able to achieve perfection, even in the short term. The computer model data has become more widely available in recent years thanks to advances in communications, and for mariners, this has manifested itself in the widespread use of grib data, which is computer model output of parameters of interest to sailors, like wind speed and direction, and sea state. This grib data is fairly reliable, but users of the data would do well to remember what Ed Lorenz discovered, namely that the output from the computer models cannot be perfect, and in some cases, can be significantly in error. The systems that display this data on yachts are very slick looking, presenting a professional-looking product, and this makes it very easy to assume that they will be 100 percent correct. But relying totally on this data puts the mariner at the mercy of a potential butterfly effect, and thus should be avoided. Mariners should routinely be consulting other forecast information, either government-issued forecasts, or forecasts from private meteorologists, both of which factor in many parameters in addition to the computer model output. Hopefully by considering all of these sources, mariners will avoid situations that could be chaotic for their vessel, crew, and passengers.
About the writer
Ken McKinley earned a bachelor’s degree in atmospheric science from Cornell University in 1980, and attended graduate school in meteorology at the Massachusetts Institute of Technology. After working as a meteorologist for nearly 10 years for a large private consulting firm in Massachusetts, he founded his own meteorological consulting firm, Locus Weather, in Camden, Maine in 1991. A large portion of his business at Locus Weather involves providing custom weather forecast services for oceangoing yachts, both racers and cruisers. Ken serves as an instructor for the Ocean Navigator School of Seamanship, and also as an adjunct instructor at the STAR Centers for Professional Maritime Officers in Dania, Fla., and Toledo, Ohio, and for MITAGS in Baltimore, Md. He has also taught meteorology at Maine Maritime Academy. He resides in Rockport, Maine with his wife and two sons.
Question for Ken? editors@oceannavigator.com