An inverse method for tracking ice motion in the marginal ice zone using sequential satellite images

A new method for tracking ice motion and estimating ocean surface currents from sequential satellite images is presented. It is particularly suited for the marginal ice zone. A simple ice advection model, driven by wind and surface currents, is formulated with several important model parameters, the ''controls,'' treated as unknown. The goal of the method is to estimate the optimal values of the controls using sequential satellite images. Following the approach of variational data assimilation, a cost function is formulated that includes the mean squared difference between the images after advection to a common time by the ice-ocean model. The controls are then adjusted to minimize the cost function. Consequently, the method gives estimates of the ocean currents and wind-driven motion that form the optimal ice motion. The approach is applied to a pair of Advanced Very High Resolution Radiometer (AVHRR) images from the Labrador Shelf separated by about 7.5 h. Preliminary results are promising. The resulting ice advection model successfully explains 78% of the variance present in the original image pair. Also, the ocean currents agree well with the observed mean circulation in the region. The proposed approach provides a framework for ice tracking and forecasting using an appropriate dynamic model to assimilate all possible data collected at different times by different types of sensors.