data assimilation

The first process carried out in the execution of a numerical forecasting model, consisting of determining the initial fields of the various meteorological variables making up the model, such that the best starting analysis is obtained. Through variational methods requiring a large amount of mathematical calculation, the model's initial state (data assigned to each grid point) is established from available observations (of different types and irregularly distributed across the Earth) and the model's predictions from the previous run. Data assimilation is the process on which the powerful computers responsible for running general circulation models of the atmosphere spend the most calculation time.