In part 1 (Andreas and Claffey, this issue) we observed some characteristics of the neutral stability air-ice drag coefficient at a reference height of 10 m (CDN10) that had not been documented before. Our main conclusion was that wind-driven snow continually alters the sea ice surface; the resulting snowdrifts determine how large CDN10 is. In particular, part 1 reported three observations that I would like to explain. (1) CDN10 is near 1.5¿10-3 when the wind is well aligned with the drifted snow. (2) CDN10 is near 2.5¿10-3 the when wind makes a large angle with the dominant orientation of the snowdrifts. (3) CDN10 can increase by 20% if, after being well aligned with the drift patterns, the mean wind direction shifts by as little as 20¿. To investigate this behavior of CDN10, here I adapt a model developed by Raupach (1992) that partitions the total surface stress into contributions from form drag and skin friction. An essential part of this development was extending Raupach's model to the more complex geometry of sastrugi-like roughness elements. Assuming that 10-cm high sastrugi cover 15% of the surface, this physically based model reproduces the three main observations listed above. Thus the model seems to include the basic physics of air--ice momentum exchange. The main conclusion from this modeling is that 10-cm, sastrugilike snowdrifts, rather than pressure ridges, sustain most of the form drag over compact sea ice in the western Weddell Sea. Secondly, the modeling suggests that skin friction accounts for about 60% of the surface stress when the wind is well aligned with the sastrugi; but when the wind is not well aligned, form drag accounts for about 80% of the stress. The sastrugi are thus quite effective in streamlining the surface.