![Hier klicken, um den Treffer aus der Auswahl zu entfernen](images/unchecked.gif) |
Titel |
An accelerated data assimilation approach for volcanic ash forecast |
VerfasserIn |
Guangliang Fu, Haixiang Lin, Arnold Heemink, Arjo Segers, Sha Lu |
Konferenz |
EGU General Assembly 2016
|
Medientyp |
Artikel
|
Sprache |
en
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 18 (2016) |
Datensatznummer |
250130149
|
Publikation (Nr.) |
EGU/EGU2016-10360.pdf |
|
|
|
Zusammenfassung |
The 2010 Eyjafjallajökull volcano eruption had serious consequences to civil aviation. This
has initiated a lot of research on volcanic ash forecast in recent years. Ensemble-based data
assimilation uses the observation data to improve the parameter and state estimation and
subsequently the volcanic ash forecast accuracy. Due to the computational complexity of
ensemble-based algorithms and the large scale of real-life applications, application of these
methods usually introduces a large computational cost, particularly in the analysis step of
assimilation processes. Because the other time-consuming part in the single CPU case, the
forecast step, can be efficiently and easily parallelized.
In this study, we focus on speeding up of the analysis step. For volcanic ash assimilation
of aircraft-based measurements, the most time-consuming part in the analysis step
has been shown to be the computation of the Kalman gain matrix. After a careful
study on the characteristics of ensemble ash states, we propose a model-reduced
Kalman gain (MR-Gain) approach which transforms the ensemble state matrix into a
low-rank matrix by a multiplication with an index matrix which recorded the sparsity
information of the ensemble state matrix, and thus the computational cost of all the
ensemble-related matrix multiplications are reduced. After the computation of Kalman gain,
using the previously recorded state index, the full analyzed ensemble states are
reconstructed.
The result shows the MR-Gain approach is exact, which can be used to replace the
original full matrix with a much low computation cost. Computer experiments show that the
computing time for the analysis step with the new approach is a factor of ten times faster
than the conventional analysis step. The result also shows that with the accelerated
analysis step in volcanic ash assimilation system, the total amount of computing
time for volcanic ash forecast can be significantly reduced by up to a factor of 5. |
|
|
|
|
|