A new reduced rank square root Kalman filter for data assimilation in mathematical models

• Authors: Dimitri Treebushny;Henrik Madsen
• Venue: ICCS'03 Proceedings of the 1st international conference on Computational science: PartI
• Year: 2003

The reduced rank square root filter is a special formulation of the Kalman filter for assimilation of data in large scale models that represent simple linear or complex nonlinear systems. In this formulation, the covariance matrix of the model state is expressed in a limited number of modes. In the classical implementation [15] some sort of normalization of the square-root matrix is required when variables of different scales are considered in the model. A new approach is formulated that avoids the normalization step. In addition, it provides a more cost-efficient scheme and includes a precision coefficient that can be tuned for specific applications depending on the trade-off between precision and computational load.