ERCIM 7th Workshop of the Working Group

Matrix Computations and Statistics

Limassol, October 31, 2005

Two sessions were organized during the CSDA'05 conference :

Data assimilation and its application Contents of the session and slides
QR and other factorizations Contents of the session and slides

Data Assimilation and its Application

Observation data are becoming both more reliable and more easily available. It is evident that the best insight into a complex system will be achieved when observational and theoretical knowledge are combined by objective methods as it is done through application of advanced data assimilation techniques. Ongoing research and development in this field has shown that the quality of the output of comprehensive models is considerably increased through advanced data assimilation. Improvements can be achieved in several different ways:

One can increase the quality of the initial data used in the treatment of the models.
One can improve the quality of input data used continuously during the run of the model under consideration (as, for example, emission data).
One can use data assimilation to improve the chemical and physical mechanisms used in the mathematical formulation of the models.

Such tasks lead to very time-consuming runs. Some of these tasks cannot yet be comprehensively treated on now available computers. Special strategies have to be developed and applied to overcome such problems while waiting for further growth of computer performance.

The selection of representative and reliable measurement data (and measurement stations) is another difficulty when data assimilation techniques are to be implemented in a large-scale model. The results obtained by employing data assimilation techniques essentially depend on the availability of measurement data describing the treated system with sufficient complexity. Therefore, methods have to be developed which allow less accurate or even proxy-data to be implemented in data assimilation schemes. In particular, 4-dimensional variational data assimilation has the potential for such an approach.

Chair

Zahari Zlatev
National Environmental Research Institute
Frederiksborgvej 399, P. O. Box 358
DK-4000 Roskilde, Denmark
Tel: +45 4630 1149
Fax: +45 4630 1214
E-mail: zz@dmu.dk

Communications

Combining observations and ensemble air-quality forecasts
		Vivien Mallet, Bruno Sportisse slides

	Implementation issues related to data assimilation using Kalman filtering
		Gabriel Dimitriu slides

	Variational data assimilation: examination of results obtained by different combinations of numerical algorithms and splitting procedures
		Zahari Zlatev slides

	Error and Uncertainties in Data Assimilation
		Francois-Xavier Le Dimet slides

	Determination of the material properties of a solid elastic medium in contact with a fluid medium with MCMC and SPSA
		Samih Zein, Jocelyne Erhel, Nabil Nassif, Edouard Canot slides

QR and other factorizations

The Track is devoted to the analysis of direct methods which determine the rank of matrices and solve linear least squares problems (standard LS, generalized LS or Total LS). The two cases of dense or sparse matrices are considered.

The Track also focuses on the solution of linear least squares problems arising in diverse applications such as in image processing.

Chair

Bernard Philippe
INRIA/IRISA
Campus de Beaulieu,
35042 Rennes cedex,
France
Tel: +33 2 99 84 73 38
Fax: +33 2 99 84 71 71
E-mail: philippe@irisa.fr

Communications

Gaussian elimination and orthogonal factorization of sparse matrices
		Laura Grigori slides

	From sparse QR to sparse RRQR
		Frédéric Guyomarc'h, Dani Mezher, Bernard Philippe slides

	A Package of Parallel Algebraic Two-level Preconditioners Based on PSBLAS
		Daniela di Serafino, Pasqua D'Ambra, Salvatore Filippone slides

	Structured Total Least Squares for Color Image Restoration
		Jesse Barlow, Haoying Fu, Michael Ng slides