By comparing the top and middle panels of Figure 1, we can see that the increase in salinity off Brazil is a reflection of the fact that there is a sharper salinity front, presumably associated with the ACC, at year 1800.
The freshening in the North Atlantic is associated with an interesting "climate shift" about year 1500. Figure 2 shows the evolution of the salinity at 700 meters depth at model point (120,55) for the run. This point is a bit SW of iceland, at about 30W, 60N.
There is a distinct drop in salinity beginning around year 1500. This is associated with a coincident change in the structure of the convection in the North Atlantic. Figure 3 shows this evolution; it displays the Depth of Surface Convection across the breadth North Atlantic. Longitude is on the X axis and time is on the Y axis.
The band of high values between 0 and 10 W shows convection in the GIN sea. Before approximately year 1480, there is moderate convection between 20 and 35 West; at year 1480, there is a sudden burst of convection off the tip of Greenland (~50 W). After that, convection in the North Atlantic quiets down, and proceeds mainly in the GIN sea. The changing location of convection in the North Atlantic is also shown in Figure 4, which presents the depth of surface convection at years 1460, 1486, and 1700 (selected to show typical years before, during, and after transition in convective pattern, respectively).
The burst of convection off Greenland is shown in Figure 5, which shows the depth of surface convection at model point (110,54).
The change in convection is what is apparently driving the shift in salinities in the North Atlantic; what I believe is happening is that the decrease of convection at the more southerly latitudes in the North Atlantic (which can be seen in Figure 4) removes a salty contribution to the intermediate water; the water mass freshens as a result.
So why does the water column suddenly overturn off Greenland? To shed some light on this, I've plotted the time evolution of the water column at point 110,54 in Figure 6.
The upper panel shows the salinity, and the lower panel shows the temperature. Note that the vertical axis is model level, not physical depth, and that only the first 100 years is shown.
When I look at Figure 6, I notice two things in particular. 1) A salinifying event in the upper layers around year 1460. 2) A consistent cooling trend at model levels 3 and 4. My suspicion is that the model suddenly starts overturning off Greenland because of the consistant cooling happening just below the surface layer. The decline in temperature, and how it is synchronized to the sudden increase in convection, is also shown in Figure 7.
There is a good .4 C drop in temperature from year 1410 to year 1470.
There is, in fact, some distinct variability on the decade-to-century timescale which can be seen.
Where, geographically, is this variability coming from? This can be seen in Figure 9, which shows the time evolution of the depth of surface convection around the Antarctic continent (out to 60 degrees S).
The strongest burst of convection seen occurs about on the prime meridian at model year 1580. There is a somewhat more regular repeating pattern of convection at about 90 West, in the Amundsen sector (not in the Weddell). Apparently, both of these are contributing to the Drake Passage transport; this can be seen more directly in Figure 10, which shows the depth of convection at 90 West (Amundsen) and along the prime meridian (Enderby).
The peak in Drake Passage transport near year 1620, for example, is clearly associated with convection in the Enderby, but not the Amundsen; on the other hand, the increase transport after year 1810 is associated with activity in the Amundsen, but not the Enderby. So they each are having an effect.
One of the motivating questions for this study is: does the OPYC model show thermohaline variability on the decade to century time scale; and, if so, is it via the same physical mechanism as produces it in the LSG model? The data shown above certainly shows that there is some decadal variability in the Southern Ocean; but how similar is the physical mechanism to that in the LSG model?
The physical mechanism of variability in the LSG model is well characterized as a non-linear flip-flop oscillator. Some of its relevant attributes: an approximate (but oscillating) balance of stabilizing fresh water flux, and destabilizing subsurface heat flux; the accumulation of heat below the surface as fresh water accumulates on the surface; and the requirement of non-linearities in the equation of state for the oscillation to occur.
Figure 11 shows the evolution of the water column in the Amundsen region (124 to 72 West, 90 to 60 S).
Note that model years 1600 to 1700 are shown. Figure 10 shows that during this time, convection in the region dropped to a low value around year 1620, reached a peak around 1650, then dropped to another low at 1670. The structure of the water column in the Amundsen clearly shows the accumulation of subsurface heat when the convection drops, along with the accumulation of surface freshwater. So this is consistent with the idea that the non-linear flip-flop oscillator is causing the variability in this region. I think that more evidence would be required to prove that this is the mechanism. For instance, with the LSG model we demonstrated several additional points:
What about the buoyancy fluxes? Well, the average heat flux across the ACC during years 1600 to 1700 is about .72 PW. I haven't yet reprocessed the heat fluxes as per Josef's message, so I don't know exactly how much of this goes into the region of interest (the Amundsen). But let's just say that it is equal to the fractional area which the region of interest covers (.17 of the total area south of 60 S) times the total heat flux. Then, the heat flux would be .12 PW; this gives a total (area integrated) destabilizing buoyancy flux due to heat of 2.4e4 m**4/s**3. If we say that .5 m/year of net P-E accumulates on average in the region, then the stabilizing buoyancy flux due to rainfall is 1.5e4 m**4/s**3. These numbers are within a factor of two of each other, which (given the crudity of the estimation process) implies that they might be close enough to support the oscillations. A more detailed budget, using the corrected heat and fresh water fluxes, should show more.
Some general points: