Climate and weather: The predictability of weather
You need to have Flash Player 10 or above installed to view this video
Weather is a series of short-term events that can change dramatically from one day to the next. Small fluctuations in atmospheric conditions can build up into large weather effects over time, a phenomenon known as ‘chaos’. The chaotic nature of weather makes it impossible to predict more than a few weeks in advance, even with the best possible computer model.
A tiny imprecision in the measurements of weather conditions used as the starting point of a weather forecast can skew the whole prediction. This effect is referred to as ‘chaos’. Many physical and biological systems are chaotic, including our short-term weather. Very slight disturbances in the atmosphere or oceans can escalate into significant events over time. Day-to-day weather is dominated by these chaotic effects, so much so that scientists would need the ability to measure the atmosphere and oceans with infinite precision in order to forecast the long-term weather with high accuracy. However, chaotic effects diminish over longer-term averages, so chaos no longer dominates on multi-decade climate timescales.
To forecast the weather, scientists use computer models that simulate the Earth’s atmosphere and oceans in great detail. These models have good track records in predicting the weather over a few days. But accuracy diminishes when forecasting more than about a week into the future, because even a perfect computer model relies on measurements of the initial weather conditions. Day-to-day weather is chaotic, so a tiny imprecision in these initial conditions can render the whole forecast inaccurate. The atmosphere has a greater tendency towards chaotic effects at some times than at others, and some regions tend to have more stable conditions than others. So forecast accuracy depends partly on location and atmospheric conditions.
The maximum weather forecasting range depends both on the accuracy of the computer model used and the precision of the measurements of initial weather conditions. Improvements are continually being made to models, but owing to the chaotic nature of day-to-day weather, scientists estimate that even a perfect model would only improve the range of useful forecasts from one week to about two weeks. Scientists are also working towards better measurements of weather conditions, including ground-based, airborne and satellite observation systems. However, it’s likely that even near-perfect weather data, combined with a perfect computer model, would only enable a useful day-to-day weather forecast up to three weeks into the future.
Scientists estimate that even with near-perfect computer models and measurements, they could only forecast the day-to-day weather two to three weeks into the future. So how can they attempt to forecast the weather weeks or months in advance? While day-to-day forecasting on these timescales is impossible, average weekly weather is less chaotic. Monthly average conditions are even more stable. Scientists may never be able to forecast a particularly hot or cold day months in advance, but a hot or cold week, month or year may be possible to forecast because chaotic effects become less pronounced over longer-term averages. However, seasonal forecasting is an area of intensive research and the reliability of such forecasts remains uncertain.