Terry,
very interesting thoughts by your side that made me re-find that old Madsen paper, which I found quite instructive once upon a time:
ftp://io.usp.br/lado_20130322/papers/madsen_1977.pdfThat also shows that lovely struggle between Ekman's 45 degrees, and more "realistic" surface turning angles against the wind stress that are far smaller. If you combine that with the wind turning angle, you see again that the surface behaves different from your intuition (i.e. the very surface should then flow into a hurricane). Not a problem though on the large scale, because the climatology is made by the mean transport direction, which is at a large angle to the surface current. Similarly, you make the pumping worse for the hurricane, because it is far more effective if you dredge up around the thermocline and then mix towards the surface by turbulence.
That paper also shows a bit the overall misery with Ekman transport, since that all depends on the vertical friction/shear terms, which are quite a mess (turbulent friction, viscosity, etc.).
I would point to that your view on the pumping poses a couple of problems: The horizontal scale is far larger than the vertical scale, so even if you have some stratification, you cannot just swap the driving agents: Ekman pumping happens because the mean surface transport has a divergence, not the other way round. Your surface waters will be forced down even at minimal differences in the wind stress. Not saying that it doesn't play a role, though.
You have a point about the flowing in deeper water: This is essentially the issue with the thicker ice that will have a larger turning angle as linked in the previous papers. However, I would like to be a bit picky and again stress that this effect is not caused _by_ the Ekman pumping, but it is the cause _of_ the Ekman pumping.
Also, concerning the ice: You have both a boundary towards the water (which exerts the stress on the water) and a boundary towards the air, which is why the previously linked paper introduced two turning angles at a surface. Never forget that the friction with the air might be lower, but that you have >10 times the velocity difference, which makes up for that... (or also think about that all motion of your surface and below must come from that friction, so your argument violates the cause-effect principle a bit).
So, sorry for the lengthy words from my side.
Plinius
The math, particularly in your second linked paper, is too much for me.
I have a much simpler model in mind.
As cyclonic winds react with <100% sea ice covered waters the water moves vertically upwards in the center and displace surface water, as well as anything floating in it, away from the center. This effect increases with wind speed, but is present at all velocities.
Sea ice, floating 90% submerged, is far more sensitive to water velocity and direction than to air and diffusion takes place more or less symmetrically.
Ignoring the ice for a moment, can you point to any literature showing an inward flow of water in a cyclonic situation at any wind speed? If not I think it reasonable to assume that anything floating heavily in this water will closely follow the outflowing currents generated by Ekman pumping/suction.
Terry