Why predictions are difficult

In 1972 the American meteorologist Edward Lorenz wrote a paper with an arresting title: “Predictability: Does the Flap of a Butterfly’s Wings in Brazil Set Off a Tornado in Texas?” A decade earlier, Lorenz had discovered by accident that tiny data entry variations in computer simulations of weather patterns-like replacing 0.506127 with 0.506-could produce dramatically different longterm forecasts. It was an insight that would inspire “chaos theory”: in nonlinear systems like the atmosphere, even small changes in initial conditions can mushroom to enormous proportions. So, in principle, a lone butterfly in Brazil could flap its wings and set off a tornado in Texas even though swarms of other Brazilian butterflies could flap frantically their whole lives and never cause a noticeable gust a few miles away. Of course Lorenz didn’t mean that the butterfly* “causes” the tornado in the same sense that I cause a wineglass to break when I hit it with a hammer.