The chart above was borrowed from Wikipedia. Sailors know that the winds and seas in the southern latitudes from 40 to 60 south are not for the faint of heart. Those high winds drive the seas that creates the circumpolar current. With the higher average surface wind velocity and little land mass to deflect the current, huge amount of energy are exchanged between the cold ocean waters and the atmosphere. This region is the thermal regulator of Earth's climate.
Update: The Flow of the Antarctic Circumpolar Current through the Drake Passage is estimated at 95 to more than 134 Sverdrup which is 10^6 cubic meters per second. The Gulf Stream current off Florida is estimated at 35 Sverdrup. Somebody asked so there is the best answer I could find.
Southern South American temperature reconstruction by Neukom et al. 2010. My apologies for misspelling the name on the chart, the link should soothe any perceived slight. With this "splice", the instrumental data fits well with the temperature anomaly of the paleo reconstruction. I have not include any error bars because at this time the degree of uncertainty is not easily determined. The centered 5 year smoothing applied to each data series is intended to not overly smooth any information that they may contain, just eliminate some of the noise in each. The object is to determine an appropriate baseline to start rebuilding a better picture of past climate with other regional reconstructions.
Cook et al 2000 Tasmania with the Neukom et al. Southern South America and the GISS LOTI 44-64S instrumental. Using the same p.b.l with 5 year centered smoothing there is a good deal of noise. The mean value lines for each series is included showing that the range of means is from about -0.3 to -0.5C. Despite the noise, that is a remarkably close range of mean for 4000 years of climate. Of course the reconstructions may have issues. By increasing the smoothing period, there would be less noise reducing the peak values. Smoothing them enough, we would have a hockey stick with current temperatures about 0.4 to 0.5C higher than the past mean, but that is already shown. Ideally, any more smoothing would match the natural smoothing of averaging the surface temperature instrumental.
Expanding the Time Frame:
Extending past climate beyond 900 AD is a bit of a challenge. Since the Ice Ages would have a much more pronounced impact near the poles and at higher elevations, tropical reconstruction would give a better indication of ocean temperatures but not global temperatures. The southern high latitudes may have been frozen to some point. Antarctic sea ice advance could have shut down or greatly reduced the Antarctic Circumpolar Current. So this next step is a bit of a guess.
The Tierney et al. 2008 Lake Tanganyika surface temperature reconstruction is added with the darker green full reconstruction period mean value and the 10000BC to 695AD section in the lighter green with its mean value. By subtracting 0.6C from the mean of the overlap period with the Tasmania reconstruction, we get the orientation shown. It could be higher or lower, this is just a rough fit.
During the modern era, the instrument data provides hints of the different oscillations and dampening constants of the hemispheres.
By selecting different smoothing and comparing regions, like the Tropics and Extra Tropics above, you can get a reasonable picture of the heat transfer between the regions of the oceans. The satellite series started with a small volcanic event that impacted the northern hemisphere. By 1991, the Southern na Northern Extratropics appeared to have been equalized only to that the Pinatubo eruption in 1991 drive northern hemisphere temperatures down again. The extra tropics reached the same capacity again in 1996 setting the stage for the temperature equivalent of a rogue wave in the 1998 Super El Nino. Since then, the temperatures a falling in a dampened manner with various harmonic synchronization generating smaller El Nino and La Nina events. Much longer term oscillations are likely which are probably generating the "noise" in the paleo climate reconstructions.
Once Antarctic sea ice extent increases to the South American peninsular, the efficiency of the ocean/atmosphere heat transfer most likely to released energy to space would be reduced. This would likely increase the heat loss at the northern pole increasing high latitude precipitation where the mass snow and ice could accumulate at high elevations, not only in the Northern Hemisphere, but in all higher elevations of the Earth. Oddly, the accumulation of snow on the more permanent Antarctic sea ice would likely be a significant driver of reduced global sea level where the sea ice was "fixed" or landed. The building and breaking of these Antarctic fixed ice accumulations could produce some interesting "Red Herrings" in the iconic Antarctic ice core history of global climate. I'll have to search for a few of those anomalies.
Anywho, to be continued.
I wasn't up to the Caribbean yet, but since I got into a discussion on the Milankovic Cycles that don't quite match the ice ages because ice ages are not all the same, I just threw this in for the moment. I hope it does give away the ending :) Schmidt, M.W. et al. 2006 is on the NOAA paleo site if you are interested.
The Schmidt et al. Western Caribbean is one of many longer term reconstructions not often mention in climate science. By combining the Western Caribbean with southern oceans reconstruction and the Lake Tanganyike lake surface temperature reconstruction you can see why. There are longer term internal ocean oscillations which tend to confuse most folks. The southern oceans are the main heat sink for the planet, but they don't catch up with global heat capacity changes quickly in all cases. That produces the internal oscillations on all time scales.