Hindcasting the Georges Bank circulation. part I: detiding

Detiding of ADCP observations on Georges Bank is addressed using 3D circulation models and an iterative inversion strategy. The forward model is a nonlinear, time-domain finite element model with advanced turbulence closure. The inverse model is its linearized frequency domain counterpart. The inversion is focused on the deduction of boundary conditions for a limited-area model. fferent inversion strategies are described and tested. Robust smoothing parameters are selected based on the variance of boundary condition estimates. Test cases using model-generated output as “data” show the possibility of near-perfect inversion given realistic sampling of the bank; portions of the boundary which are poorly constrained by the sampling; and various sources of imperfection. The mismatched physics between forward and inverse models is shown to be diminished by iteration such that a nonlinear inverse is constructed. ion of real data – a 12-day cruise in April 1995 – is demonstrated. Prior error generated by climatology-based model solutions is 15 cm/s RMS.The inversion reduces this error to 8 cm/s, close to the minimum achievable based on tidal analysis of moored time series. The hindcast circulation shows important departures from the climatology, including a meander in the Northern Flank jet; a northward, off-bank transport from the Central Bank and Northeast Peak; and an on-bank intrusion along the Southern Flank.