Techniques for kinematic analysis and dynamic, “free-drift” ice modeling are described and applied to interpretation of ice-drift data from recent marginal ice zone (MIZ) experiments. Kinematic description is based on a complex demodulation algorithm that separates inertial and tidal components from lower frequency, “synoptic” drift. Complex demodulation produces the time series of phasors (complex numbers describing phase and amplitude of the oscillating components), useful for separating the physical processes active in the upper ocean/ice system. Free-drift ice motion modeling utilizes a similarity theory for planetary-boundary-layer dynamics that includes the effect of buoyancy, both from rapid melting at the ice/ocean interface, and/or from a pre-existing density gradient (pycnocline) within the boundary layer. Two examples are considered: one in which a band of ice in the Bering Sea drifted rapidly away from the rest of the pack when it encountered warm water at the ice edge; and a second in which drift in the Greenland Sea was apparently affected by both a shallow pycnocline and a period of rapid melt.