Hindcasting the Georges Bank circulation, Part II: wind-band inversion

• Published in: Continental shelf research Vol. 22; no. 15; pp. 2191 - 2224
• Main Authors: Lynch, Daniel R, Naimie, Christopher E
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.10.2002
• Subjects: Adjoint method; Data assimilation; Finite element; Adjoint method; Data assimilation; Finite element
• ISSN: 0278-4343; 1873-6955
• DOI: 10.1016/S0278-4343(02)00072-9

Nonlinear inversion of velocity observations to produce limited-area circulation hindcasts is examined in the context of the USGLOBEC-Georges Bank surveys. Two linear inverses representing tide- and wind-band responses are used serially in an iteration with a nonlinear forward model. Both inverses are linearized about the nonlinear model. The objective is generalized least-squares fit; the controls are sea level elevation on the boundary of a limited-area mesh. The iteration converges rapidly provided a good prior estimate is available. Observational system simulation experiments address prediction skill for an idealized shelf regime and for an isolated wind event on Georges Bank. The idealized shelf inversion shows the limitations of extrapolative skill. For the wind event, skill is sensitive to the space-time locations of the observations. A real Georges Bank cruise, ENDEAVOR 261, is studied and the hindcast based on shipboard ADCP and moored data is assessed. The prior estimate based on climatology plus observed wind captured much of the tidal variability, most of the time-mean circulation, and little of the wind-band variability. The inversion improved the tidal fit, captured the wind-band variability, and did not degrade the mean.