(FORTUNE Magazine) – How do climate scientists know the greenhouse effect will bring about the woes that they predict? They don't know to a total certainty. What they do know is based on half a dozen high-powered computer simulation programs, called general circulation models, in North America and Europe. Researchers feed in equations based on the laws of physics, along with assumptions about clouds, sea ice, ocean currents, soil moisture, atmospheric convection, and emission of heat from the ground. More complicated things happen in the heavens and on earth, however, than are dreamt of in the equations of scientists. Even using the best supercomputers, none of the models is so good that it can start with known weather conditions at a given point in the past and reproduce precisely what has happened since. To make the calculations manageable even by computers, most of the models suppose either that the oceans are a shallow, motionless swamp or that they don't exist at all. Despite that oversimplification, an especially sophisticated computer model at the National Center for Atmospheric Research (NCAR) in Boulder, Colorado, requires 1.5 trillion calculations to advance its predictions a single day. One basic problem is called grid resolution. Climatologists divide the world into a grid. Most use a grid with squares the size of France. The grid defines France as a single set of numbers, failing to distinguish the cool, rainy north from the sunny, drier south. In his 1987 book Chaos, James Gleick, a New York Times reporter, imagined a world covered with a vast jungle gym of sensors spaced a foot apart and rising 35 miles to the top of the atmosphere. Each sensor measures with great precision temperature, pressure, humidity, and every other meteorological variable. An infinitely powerful computer processes all the data. This seemingly perfect monitoring system still could not predict exactly the weather next month in Atlanta. The reason: The computer would not detect microfluctuations that took place in between the sensors. Errors multiply so quickly that within hours the reality of weather diverges from its predicted course. In effect, you can never have enough grid squares to forecast weather accurately. Tiny variations matter. The Butterfly Effect, known technically as ''sensitive dependence on initial conditions,'' gets its name from the thought that a butterfly flapping its wings today in Nagasaki could conceivably influence storms next month in New York. While most scientists agree that the greenhouse effect is coming, there aren't enough data yet to say with absolute conviction what its consequences will be. Certainties in science are a long time in the making. In a profession where tentative conclusions require decades' worth of data, one swallow does not make a summer. As recently as the 1970s some climatologists were worrying about global cooling, because world temperatures had peaked in the 1940s and then declined into the 1970s. Air pollutants such as volcanic and man-made dust may have blocked enough sunlight to lower global temperatures.