Data Assimilation Meetings at Reading

Date Meeting type Speakers
21 November 2012 DARC seminar Africa Perianez (UoR and DWD)
Adaptive Localization for Ensemble Methods in Data Assimilation
In ensemble Kalman filter techniques it is very common to use space localization in order to reduce the effect of spurious long range correlations. It is the goal of our analysis to understand the basic theoretical properties of localization in the ensemble Kalman filter scheme. A decomposition study of the error sources is performed in order to determine its effect in the computation of the optimal localization size. The different sources of error are studied separately: the ensemble space, the localization and the observation measurements. We explore the performance of the analysis in the limits either when the ensemble size is increased or when the localization radius tends to zero. This approach is analyzed from a mathematical perspective addressed with numerical experimental results.

Page navigation