Recent developments in optimization methods for data assimilation in oceanography

  • Patrick Laloyaux

    Student thesis: Doc typesDoctor of Sciences

    Abstract

    This work is concerned with the development and the study of novel optimization algorithms for solving data assimilation problems with application in oceanography. Motivated by the theoretical equivalence between the SEEK filter and a reduced variant of the 4D-Var problem, two innovative minimization methods for the variational approach are proposed. The first one derives an appropriate starting point and a limited-memory preconditioner to accelerate the conjugate gradient-like algorithm used in the incremental method. The second one proposes a derivative-free approach which minimizes the 4D-Var function by constructing and exploring a sequence of appropriate low dimensional subspaces. Both methods have been numerically analyzed on academical shallow water models before being assessed on the GYRE configuration of the NEMO framework.
    Date of Award28 Sept 2012
    Original languageFrench
    Awarding Institution
    • University of Namur
    SupervisorAnnick Sartenaer (Supervisor), Timoteo Carletti (Jury), Anne Lemaitre (President), Serge Gratton (Jury), Luis VICENTE (Jury) & Arthur VIDARD (Jury)

    Cite this

    '