The Met Office recieves approximately half a million observations of temperature; pressure; wind speed and direction; humidity, and many other atmospheric variables every day. This data is collected (by the instruments listed below) and is then sent to a central database for storage.
The next step is to calculate how the current atmosphere will change over time.To do this, the supercomputer uses a number of complex equations which model the atmosphere and oceans. These calculations are repeated many times. Each repetition the forecast is stepped a few minutes further into the future, and this enables us to produce forecasts from just a few hours ahead, to a climate prediction for the coming 100 years. This process is called Numerical Weather Prediction.
However, as we go further into the future, the accuracy of our predictions decreases. This is because the partial differential equations that govern the atmosphere are chaotic in nature.It is impossible to solve these equations exactly, and small errors grow with time (doubling about every five days). Present understanding is that this chaotic behavior limits accurate forecasts to about 14 days even with accurate input data and a flawless model.