The degree to which a system's future states can be accurately forecast.Predictability is a measure of how well a system's future behavior can be known in advance. In physics, it is not an all-or-nothing property but a matter of degree: some systems, like a swinging pendulum, can be forecast far into the future with high accuracy, while others, like the weather, lose reliability after a few days. The concept is central to distinguishing between systems that are deterministic—where the same starting conditions always lead to the same outcome—and those that are chaotic, where tiny differences in initial conditions grow exponentially, making long-term forecasts impossible in practice.
The modern understanding of predictability emerged in the late 19th century with Henri Poincaré's work on the three-body problem, where he showed that even simple gravitational systems can be unpredictable. In the 1960s, Edward Lorenz rediscovered this in meteorology, coining the term 'butterfly effect'.
Predictability depends on two factors: the precision of initial measurements and the system's sensitivity to those measurements. In a stable system, small errors in measurement remain small over time, so forecasts stay accurate. In a chaotic system, errors amplify rapidly, so even minuscule uncertainties—like the flap of a butterfly's wings—soon dominate the outcome. This is quantified by the Lyapunov exponent, which measures how quickly nearby trajectories diverge. When the exponent is positive, the horizon of predictability is finite and often short.
Predictability limits are used in weather forecasting to set confidence intervals, in climate modeling to distinguish signal from noise, and in engineering to design control systems that remain stable despite uncertainties.
The atmosphere's predictability horizon is roughly two weeks, not because of any fundamental law, but because of the chaotic amplification of measurement errors.
