Preconditioning Representer-based Variational Data Assimilation Systems: Application to NAVDAS-AR

Abstract

Assimilation of observations into numerical models has emerged as an essential modeling component in geosciences. This procedure requires the solution of large systems of linear equations. Solving these systems in ``real-time'' or ``near-real-time'' in a timely manner is still a computational challenge. This paper shows how new methods in computational linear algebra are used to ``speed-up'' the representer-based algorithm in a variety of assimilation problems, with particular application to the Naval Research Laboratory (NRL) Atmospheric Variational Data Assimilation System-Accelerated Representer (NAVDAS-AR) system.