Atlas Mathematical Conference Abstracts ||
Conferences |
Abstracts |
for Organizers |
About AMCA
AD 2000 - From Simulation to Optimization
June 19-23, 2000
INRIA Sophia Antipolis
Sophia Antipolis, France |
|
Organizers George Corliss, Christele Faure, Andre Galligo, Andreas Griewank, Laurent Hascoet, Uwe Naumann
View Abstracts
Conference Homepage |
Second Order Methods for Geophysical Flows
by
Francois-Xavier Le Dimet
Université Joseph Fourier, Grenoble, France
Coauthors: I.M. Navon (Florida State University, USA), P. Ngnepieba (LMC and Université de Yaondé, Cameroon)
A geophysical episode is a unique situation. To retrieve it is necessary to
gather all the available information ( models, data, statistics). The information is linked
through a variational principle involving the derivative of the model with respect to its
variables. Most of the time the variational principle uses a so-called adjoint model which is
obtained after a derivation and a transposition of th direct model.
The information o the second derivative is useful for several purposes:
- to estimate a preconditioning of the problem.
- to carry out efficient numerical methods ( Newton and related methods).
- to compute the sensitivity with respect to some inputs of the model, especially the
observations.
- to evaluate statistical quantities such as correlation matrices.
We will show how to compute these second order quantities through a so-called second order
adjoint which permits to compute the product of the Hessian matrix by a vector and how this
property is used for prctical purposes.
We will discuss about the automatic derivation of the second order adjoint.
An application will be shown to a problem of identification for underground water. A
comparison of hand written adjoint and adjoint generated by ODYSSEE will be carried out.
Date received: February 10, 2000
Copyright © 2000 by the author(s).
The author(s) of this document and the organizers of the conference
have granted their consent to include this abstract in
Atlas Mathematical Conference Abstracts.
Document # cads-59.