Data assimilation techniques are a vital part of forecasting. ECMWF has pioneered work on assimilation methods such as 4D-Var.
Observations and short-range forecasts are combined by calculating a weighted average where the weights depend on the respective characteristic errors.
Error characterisation is thus an essential part of research and development work in data assimilation. ECMWF is also working on developments to improve the scalability/efficiency of the assimilation system as the forecasting model moves to higher vertical and horizontal resolution.