ECMWF | Reading | 26 January 2018
One of ECMWF’s biggest-ever projects was the development of the variational data assimilation system. The project culminated with the operational implementation of 4D-Var on 25 November 1997, less than two years after the introduction of 3D-Var.
4D-Var is a four-dimensional variational data assimilation technique. It performs a statistical interpolation in space and time between a distribution of meteorological observations and an estimate of the model state (the background). This is done in such a way that account is taken of the dynamics and physics of the forecast model to ensure the observations are used in a meteorologically consistent way.
4D-Var development continued after initial implementation with a longer window (12 hour instead of 6 hour, in September 2001), improved physics and multiple outer loops (e.g. in June 2007), and an ensemble of lower-resolution 4D-Var analyses to provide background uncertainty information (in May 2011). This development path continues with 4D-Var now being extended to include atmosphere-ocean coupling in the outer loop and a hybrid extended control vector. A notable success of 4D-Var has been the all-sky assimilation system, which has allowed dynamical information to be extracted from rain- and cloud-affected radiances through the 4D-Var tracer mechanism.
In parallel, ECMWF is working on a major project to provide an Object Oriented Programming System. This system is being developed and interfaced to the Integrated Forecasting System (IFS) to allow a more flexible calling of the data assimilation algorithm, allowing alternative solvers and time parallel weak-constraint approaches such as the saddle point algorithm to be tested.
This symposium is being organised to mark 20 years of 4D-Var at ECMWF.