University of East Anglia, UK

An oceanographer describes a missing piece of the climate puzzle.

Most school students know that increasing atmospheric carbon dioxide raises global temperatures. But I've always been fascinated by the other half of the climate–CO2 connection: why, in the past, have increasing temperatures driven up atmospheric CO2?

That CO2 and temperature are locked in a powerful, positive-feedback embrace is obvious from ice-core evidence. But if we add up all the mechanisms that we know about, we fall short of explaining the rise in CO2 levels seen at the end of glaciations.

Upwelling in the Southern Ocean may be the missing piece. Today, this process brings deep CO2-rich water rapidly to the surface, where it vents carbon to the atmosphere. If upwelling was shut down during glaciations, we could fit the data better.

It has been suggested that sea ice might have blocked the air–sea transfer of CO2 during times of glacial maxima (B. B Stephens & R. F. Keeling Nature 404, 171–174; 2000).

There is good evidence that sea ice was extensive in the region, but any upwelling would have melted that ice, because the rising water has a temperature above freezing point. So the sea ice is evidence that the upwelling itself was absent. What stopped it?

One recent paper (J. R. Toggweiler et al. Paleoceanography 21, PA2005; 2006) argues that the critical factor was a shift to the north of the westerly wind belts that drive the upwelling. I and a colleague propose a subtler connection, a change in the balance of surface heat flux, that would also reduce the upwelling to near-zero (A. J. Watson & A. C. N. Garabato Tellus B 58, 73–87; 2006).

The theories are convergent in many respects, but make distinct predictions that we can test against new proxy evidence. This problem will be solved pretty soon, I think.