Overview: Data Assimilation and Model Reduction

Abstract

For the very large systems that arise in the environmental sciences, the available data are not sufficient to initiate a complex computational forecasting model. The technique of data assimilation enables measured observations (over time) to be combined with model predictions to generate accurate estimates of the system states – both current and future. An overview of data assimilation is presented here and the major types of assimilation method for treating very large nonlinear systems are described. Novel numerical assimilation techniques based on model reduction methods from control theory are then presented. The benefits of these techniques are demonstrated in a practical application from atmosphere and ocean modelling. (© 2008 WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim)