A state-space model for ocean drifter motions dominated by inertial oscillations

A discrete-process model based on finite differences is used to regress ocean accelerations on ocean velocity estimates but fails because errors in the discrete approximations for the ocean velocities are biased and accumulate over time.

Model identification is achieved by fitting a stochastic differential equation model based on classical upper ocean physics to the drifter data via the Kalman filter. Ocean parameters are shown to be nonidentifiable in a direct application to the Labrador Sea data when the known Coriolis parameter is not identified by the model. To address this, the ocean parameters are estimated in an empirical sequence.