Variability of sea ice simulations assessed with RGPS kinematics

Sea ice drift and deformation from coupled ice-ocean models are compared with high-resolution ice motion from the RADARSAT Geophysical Processor System (RGPS). In contrast to buoy drift, the density and extent of the RGPS coverage allows a more extensive assessment and understanding of model simulations at spatial scales from similar to 10 km to near basin scales and from days to seasonal timescales. This work illustrates the strengths of the RGPS data set as a basis for examining model ice drift and its gradients. As it is not our intent to assess relative performance, we have selected four models with a range of attributes and grid resolution. Model fields are examined in terms of ice drift, export, deformation, deformation-related ice production, and spatial deformation patterns. Even though the models are capable of reproducing large-scale drift patterns, variability among model behavior is high. When compared to the RGPS kinematics, the characteristics shared by the models are (1) ice drift along coastal Alaska and Siberia is slower, (2) the skill in explaining the time series of regional divergence of the ice cover is poor, and (3) the deformation-related volume production is consistently lower. Attribution of some of these features to specific causes is beyond our current scope because of the complex interplay between model processes, parameters, and forcing. The present work suggests that high-resolution ice drift observations, like those from the RGPS, would be essential for future model developments, improvements, intercomparisons, and especially for evaluation of the small-scale behavior of models with finer grid spacing.