Seminar: On the marriage between data assimilation and ensemble prediction by dr. Nils Gustafsson

Torek, 11. februar 2014, ob 13:15
Predavalnica F2, 1. nadstropje, FMF, Jadranska 19, Ljubljana

Na Katedri za meteorologijo FMF trenutno gostimo dr. Nilsa Gustafssona iz Švedskega meteorološkega in hidrološkega inštituta (SMHI). Predavanje bo v angleškem jeziku. Seminar je odprt za vse obiskovalce.


Traditionally meteorological data assimilation (DA) systems and meteorological ensemble prediction systems (EPS) have been progressing along different lines of development. This is no longer the case, the reason is that the core problem in DA and EPS is a joint problem, the estimation of uncertainty. For example, the ECMWF EPS utilizes ensemble member initial data based on randomly perturbed observations (to simulate observation uncertainty) and the short range EPS forecasts are applied to estimate forecast background uncertainty in DA. Eventually this first coupling between DA and EPS will grow into a unifying marriage. This has already happened in the weather service of Canada, where the EPS system is based on an Ensemble Kalman Filtering DA system.

The recent progress in developing a joint DA/EPS system within the HIRLAM community will be described and discussed during the seminar. This development includes Hybrid DA techniques, forming a synthesis between variational and ensemble DA techniques and also 4D-En-Var (4-dimensional ensemble variational DA) that can be considered to be a “4D-Var (4-dimensional variational DA) without tears” since the tangent linear and adjoint forecast models are no longer needed, beeing replaced by ensembles of model states in 4 dimensions. A further development includes improved correction of spatial phase and timing errors by injection of structures from the "best" ensemble member, as determined by comparison with satellite images, for example.