Forecasting the weather has been undertaken informally for millennia with records showing that the Babylonians in 650 BC attempted weather prediction from cloud formations as well as astrology. Weather forecasting continued to rely on observing patterns of events (pattern recognition) locally and it was not until the arrival of the electric telegraph (1835), with its ability to allow weather conditions across a wide region to be received almost instantaneously, that the modern era of weather forecasting arrived.

The birth of forecasting as a science is credited to both Francis Beaufort (Beaufort scale) and Robert Fitzroy (Fitzroy barometer) and their approach is considered to form the basis of all of today’s weather forecasting knowledge. Numerous developments and refinements have since taken place, notably the first pictorial cloud atlas (1890), the first international cloud atlas (1896), the first proposed forecasting by differential equations (in a pre-computer era of 1922) by Lewis Fry Richardson entitled ‘Weather Prediction by Numerical Process’ and the practical implementation of numerical forecasting in 1955 due to the development of early programmable computers.

Weather forecasting is now a global, multi-billion dollar industry that operates across all forms of media and employs tens of thousands of people who utilize the latest technology, super computers and satellites.

In general, forecast accuracy is dependent on:

  • Length of forecast. Longer range forecasts are largely based on patterns and trends derived from weather data recorded over many decades for a particular locality i.e. long-term average weather for a particular date. Superimposed on this are recent and current weather patterns, and forecasting is made using computer mathematical weather models that produce a percentage likelihood for a particular variable e.g. temperature over a given period in the future.
  • Location. Variables such as latitude, position relative to a large land or water mass, trade winds, ocean currents etc. can affect the weather stability of any given location. In some cases they combine to produce stable weather generally, which makes prediction and accuracy rates higher compared to unstable regions, whereby an unexpected change in just one variable will lead to lower accuracy.
  • Season. Some locations have very stable and predictable weather during certain periods of the year, making forecasts during these times very accurate.

Therefore we can define a weather forecast as:

“An extrapolation from present conditions, based on a weather model. The mathematical model uses atmospheric-physics theory to predict the weather. Past weather patterns contribute to the picture, as do current observations”.

Given that real weather mixes chaos with pattern, such that presently and for the foreseeable future, no computer model can include or account for the myriad of possibilities inherent in chaos, how accurate can we predict the weather at present?

Surprisingly, actual data regarding accuracy rates for both general forecasts and for specific factors e.g. precipitation, are not readily available and those which are, tend to be highly variable for different regions and localities even within a given country. In general, forecast accuracy for the next day is usually accurate, up to 3 days reasonable, 3-5 days very variable,  5-7 days are unreliable and beyond 7 days no better than long-term averages. Therefore, at present, the ‘golden’ period where forecasts are of reasonable accuracy is up to a  maximum of 3 days.

However, on a positive note, progress is being made such that on average, a current 5 day weather forecast is as reliable as a 2 day weather forecast was approximately 20 years ago.