Prof Dr Martin Ehrendorfer
Institut für Meteorologie und Geophysik der
Universität Innsbruck
Innrain 52
A-6020 Innsbruck
Austria
martin.ehrendorfer@uibk.ac.at
Project description
Singular vectors (SVs) computed using analysis-error covariance information
provide a square-root decomposition of the anlaysis error covariance
matrix P^a. This SV-decomposition evolves into the eigendecomposition
of the forecast error covariance matrix. As such, the SV-decomposition
of P^a is a primary candidate for generating - from a multivariately
standard-normal random variable - a set of M initial-time perturbations
fully consistent with P^a as described through the leading N SVs. Within
this Special Project this SV-based multinormal sampling technique will
be tested with the aim of assessing its potential for operational implementation
(especially in the case M > N). Ensembles using SVs based on the
"total energy norm", as well as "Hessian" SVs will
be generated and investigated with this sampling technique. Another
important part of this Special Project is formed by data assimilation
and ensemble prediction experiments with an exact quasigeostrophic Kalman
Filter which will allow to assess explicitly the implications of limiting
initial-time covariance information to N SVs on the time-evolving covariance
structures.
Final report
Additional information
Project period 2000 - 2003.
Special Projects