Data assimilation algorithms

Data assimilation is a mathematical process by which a numerical evolution model is constrained to stay close to a partially observed reality. Here, it is a set of techniques to prepare the initial forecast states of ARPEGE, ALADIN and AROME using observations of the atmosphere. Most activities concentrate on understanding sequential variational assimilation techniques, called 3D-Var and 4D-Var, and tuning their most important ingredient, the covariances of estimation errors of the background states.

This activity has many quick and useful applications to improving the numerical weather forecasts. It is supported by research grants such as INSU/LEFE (from CNRS) and ANR.


References