ABSTRACT
Our goal is regional assimilation of Lagrangian data into a stratified primitive equation model. Such models are ill-posed as initial-boundary value problems; however, if the boundary moves with the flow, the problem is well-posed. Lagrangian coordinates are convenient for expressing a `comoving' boundary, and lead to linear measurement functionals for float data. We have compared the Lagrangian form of the reduced gravity model to the Eulerian form and determined that solutions are acceptably close to each other when low order, high resolution numerics are used. We have also assimilated synthetic data into the Lagrangian form for a few hours, and determined that an efficient algorithm is feasible.