 |
| Titel |
Covariance localization with diffusion-based correlation model |
| VerfasserIn |
M. Yaremchuk, D. Nechaev |
| Konferenz |
EGU General Assembly 2012
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 14 (2012) |
| Datensatznummer |
250061202
|
|
|
|
|
|
| Zusammenfassung |
| Flow-dependent localization of the ensemble-generated covariances is essential for improving
the performance of the data assimilation algorithms. Using numerical experiments simulating
inhomogeneous Gaussian-shaped covariances B, two methods of retrieval the diffusion tensor
field from the simulated ensembles are tested. The diffusion-based approach is compared
with the adaptive ensemble covariance localization (AECL) technique based on the
modulated ensembles. Results of the experiments suggest that the diffusion-based correlation
model is capable of effectively approximating B with an accuracy of the AECL method when
the typical decorrelation scale gradients are less than unity. Comparison of the methods in the
strongly inhomogeneous situations demonstrates a substantial growth of the approximation
errors. |
|
|
|
|
|
|