(D102) <<< Previous | Back to Abstracts Page | Next >>> (D104)
Variational assimilation of Lagrangian data in a Primitive Equations model
Maelle Nodet
Laboratoire J.A. Dieudonne, Universite de Nice, CNRS
nodet@unice.fr(Abstract received 04/26/2005 for session D)
ABSTRACT
We investigate variational assimilation of Lagrangian data. We begin our study by doing twin experiments using an idealized configuration of the North-Atlantic Ocean and simulated positions of drifting floats. We are using the four dimensional variational technique and the adjoint method: we aim at minimizing a cost function which represents the root mean square error between observed positions of drifting floats and positions generated by the model. This cost function is minimized with respect to the control vector, which is the initial velocity field of the ocean. Observed variables, namely the positions of the floats, are expressed as a function of the control vector thanks to a non linear observation operator. The minimization of the cost function requires the tangent linear observation operator and its adjoint, whose implementation is quite difficult because of the non linear aspect of the operator. This method has been implemented in the OPA Primitive Equations model in the incremental 4D-Var approach. It has the ability to reconstruct the main patterns of the oceanic circulation. Moreover it is very robust with respect to increase of time-sampling period of observations. We have run many twin experiments in order to analyze the sensitivity of our method to the number of floats, the time-sampling period and the vertical drift level. We compare also the performance of the Lagrangian method to that of the classical Eulerian one.
(D102) <<< Previous | Back to Abstracts Page | Next >>> (D104)
2005 LAPCOD Meeting, Lerici, Italy, June 13-17, 2005