Design and Implementation of a Parallel Multivariate Ensemble Kalman Filter for the Poseidon Ocean General Circulation Model
- ISBN 13:
- 9781289272913
- format:
- Paperback
- publisher:
- Bibliogov
- language:
- English
- Publication Year:
- 2013
- Genre:
- Government Textbook
- Condition:
- New
- Availability:
- Item usually sent within 7 working days
Description
A detailed analysis of a parallel multivariate ensemble Kalman filter for ocean circulation models has been presented in this technical report. The implementation was tested using a Pacific Basin model configuration, with approximately two million prognostic state-vector variables. The approach used regionalization to achieve parallelism in the data assimilation step, and employed a Hadamard product to compactly support background-error covariances. This methodology is discussed in detail, along with the MvEnKF configuration. Results show that this method has negligible impact on analysis quality, but is efficient for large numbers of observations. However, it does not scale well beyond 100 processing elements at current model resolution, where memory rather than speed becomes a limiting factor.