%0 Journal Article
%9 ACL : Articles dans des revues avec comité de lecture répertoriées par l'AERES
%A Koenig, G.
%A Aldebert, C.
%A Chevalier, Cristèle
%A Devenon, J. L.
%T Identifying lateral boundary conditions for the M2 tide in a coastal model using a stochastic gradient descent algorithm
%D 2020
%L fdi:010080411
%G ENG
%J Ocean Modelling
%@ 1463-5003
%K Parameters identification ; Tidal modeling ; Stochastic algorithms ; Data assimilation
%K PACIFIQUE ; NOUVELLE CALEDONIE
%K OUANO LAGON
%M ISI:000593758200005
%P 101709 [14 ]
%U https://www.documentation.ird.fr/hor/fdi:010080411
%> https://www.documentation.ird.fr/intranet/publi/2020/12/010080411.pdf
%V 156
%W Horizon (IRD)
%X While lateral boundary conditions are crucial for the physical modeling of ocean dynamics, their estimation may lack accuracy in coastal regions. Data-assimilation has long been used to improve accuracy, but most of the widely-used methods are difficult to implement. We tried a new and an easy-to-implement method to estimate boundary conditions. This method uses data assimilation with a stochastic gradient descent and successive approximations of the boundary conditions. We tested it with twin experiments and a more realistic setting on a tidal model in the lagoon of Ouano, in New-Caledonia. The method proved successful and provided good estimation of the boundary conditions with various settings of subsampling and noise for the pseudo-data in the twin experiments, but there were important oscillations in the experiments with more realistic settings. Here we present those results and discuss the use of our new and easy-to-implement method.
%$ 032 ; 020