In the Arctic sea ice grows thermodynamically until an equilibrium thickness of 2-3m is reached. This formation of level ice is disturbed by deformation processes, which evolve under convergent and shear conditions of sea ice motion. Rafting and ridging of the ice cover occur. This builds thicker ice, accounting for about two-thirds of the Arctic sea ice volume. It is important to describe these dynamical deformation processes in numerical models, which are used to investigate climate change in Arctic and Subarctic regions.Efforts have been made during the last thirty years to implement these deformation processes to numerical models based on very different theories. In this study three different approaches to model a deformed sea ice cover are compared. These approaches include (1) an additional prognostic equation for ice roughness from which ridge parameters are diagnostically derived, applicable to single ice class models, (2) a redistribution model with two ice categories, level and ridged ice, including a statistical derivation of ridge parameters, and (3) a set of prognostic equations for ridge parameters, i.e. ridge density and sail height, from which a ridged ice class is finally derived.Basically all three models were able to reproduce the governing spatial distribution of Arctic sea ice thickness. However, there are differences in structure and absolute values. For example, the algorithm listed first tends to give a less discrete distribution of deformed ice while the third shows very distinct areas of heavy ridging. Results of a model experiment with simplified grid and forcing are presented and differences between the ridging schemes are discussed. The results of a first application to realistic Arctic conditions are compared to data from a control run of the sea ice model without ridging in order to express the effect of the ridging on the model behaviour.