Inverse model for limited-area hindcasts on the continental shelf

The authors have constructed an inverse model for interpreting subtidal vel ocity observations in a limited-area hindcasting context. The intended use is following removal of a best prior circulation estimate that accounts for tide, local baroclinicity, and the direct response to local wind forcing. The remaining, subtidal velocity signal is inverted to provide far-field su btidal pressure forcing. The forward portion of the model is a linearization of a full-physics 3D si mulator (QUODDY). Its inversion is achieved by gradient descent, using an e xact algebraic adjoint and strong dynamical constraints. The cost function is a weighted least squares blend of velocity mismatch plus boundary condit ion size, slope, and tendency. The control parameters are barotropic open-w ater boundary conditions. Solution is achieved in the time domain, as a com plement to the frequency-domain "detiding'' inversion presented earlier. A simple test case is introduced to demonstrate features of the inversion: precision, accuracy, interpolation, and extrapolation. A representative app lication on realistic topography (Georges Bank) is given.