Error analysis and assimilation of remotely sensed ice motion within an Arctic sea ice model

New sea ice motion fields available from remotely sensed data are potentially useful for assessing and improving models of the polar ice pack. Here we investigate the error characteristics of the observed ice motions relative to drifting buoys and a dynamic-thermodynamic ice model. A data assimilation approach is then used to assess the potential of the motion data for reducing model biases, as well as the potential of the model to serve as an interpolation tool to generate improved ice motion data sets. Special Sensor Microwave/Imager (SSM/I) derived and model simulated ice motions for the years 1988 through 1993 are compared with ice displacement observations from drifting buoys. Variability and biases are summarized for seasonal and regional means. SSM/I motions are assimilated into the model using an optimal interpolation method that accounts for the modeled and SSM/I motion error variances and the number and distribution of the SSM/I motions. Modeled and SSM/I-derived motions are found to have comparable mean errors, With some notable regional and seasonal differences. Assimilation substantially reduces the error standard deviation and improves the correlation of the simulated motions relative to the buoy observations, but some biases remain. In the model framework used here, assimilation of the SSM/I data substantially alters average ice thickness in some regions of the Arctic and affects ice mass outflow through the Fram Strait but has a small effect on mean ice concentration. The assimilation yields an increase in the spatial and temporal variability in ice deformation. The observations are particularly suited for improving the simulation of specific synoptic events, where substantial differences can occur between simulated and observed ice transport.