The climate of East Africa became drier between about 5 million and 2.5 million years ago, and that may have been the catalyst that forced our ancestors to adapt to a savannah environment as the forests dwindled1, 2. At about the same time, the Earth entered a climate mode dominated by the waxing and waning of large continental ice sheets. The coincidental timing of global cooling, African aridity and human evolution invites speculation about a common link3. For that, we must look to the oceans — in redistributing heat and influencing greenhouse-gas concentrations globally, they are the main component in determining climate change. Marine records tell us that the transition to large-scale glacial cycles took at least a million years; and plate-tectonic motions that opened or closed ocean gateways are thought to have triggered these events4.
On page 157 of this issue, Cane and Molnar5 present an analysis of changes in surface-ocean circulation that they believe occurred as an oceanic gateway — the Indonesian seaway — narrowed over the past 5 million years. This gateway acts as a valve for water flowing from the Pacific into the Indian Ocean. Plate tectonics in the Indonesian region is complicated, but Cane and Molnar show that the passages regulating water flow from the Pacific to the Indian Ocean 5 million years ago were wider and deeper, and were located further to the south, than they are today. Surface water in the South Pacific is warmer and saltier than that in the North Pacific. Cane and Molnar argue that the more southerly position for the Indo-Pacific connection meant that the warmer South Pacific flowed into the Indian Ocean. The result was warmer sea surface temperatures in the Indian Ocean and high levels of evaporation and precipitation — and wet East African climates.
Over the past 5 million years, the constriction and northern movement of the Indonesian seaway have progressively shut off the South Pacific source of water, while increasing the influence of the colder North Pacific. These changes should have cooled the tropical Indian Ocean and reduced the precipitation, leading to a gradual drying of East Africa. Again, then, the idea is that the Indonesian seaway, controlled by the northern movement of New Guinea and smaller islands, has acted like a valve, regulating the relative amount of warm and cool water entering the Indian Ocean.
A more speculative aspect of Cane and Molnar's paper deals with the possible effects on global climate of this narrowing of the Indonesian seaway. The authors argue that, when the seaway was farther south, conditions in the tropical Pacific would have been more like those observed during modern El Niños (that is, both east–west and vertical thermal gradients in the ocean would have been weaker). This configuration would have promoted greater heat transport to the high northern latitudes than at present; and that higher heat flux would have inhibited the growth of large ice sheets in the Northern Hemisphere. In support of Cane and Molnar's speculation, palaeoceanographic reconstruction of the tropical Pacific between 5 million and 3 million years ago matches the prediction of smaller east–west and vertical temperature differences6.
This speculation is especially provocative because it requires a new principle for understanding the glacial cycles that developed 3.5–2.5 million years ago. Existing models for large-scale Northern Hemisphere glaciation focus on increased circulation of the North Atlantic 'conveyor', which includes the Gulf Stream, as the cause of ice-sheet development4, 7, 8, 9. The finer points in these models vary, but most of them begin with increased precipitation at high northern latitudes as a result of a more vigorous Gulf Stream. And going back one step, the closure of the Panamanian isthmus is seen as the trigger for initiating the more powerful Gulf Stream.
Recently, however, a fly in the ointment for these hypotheses has appeared10. An evaluation of North Atlantic circulation for between 5 million and 2 million years ago contradicts the enhanced Gulf Stream mechanism for Northern Hemisphere glaciation. The new data10 indicate that the North Atlantic conveyor became considerably weaker, not stronger, 3.5–2.5 million years ago.
Cane and Molnar5 propose a shift in thinking away from the North Atlantic to the Indonesian gateway as a factor governing global climate change. There are good reasons to take this idea seriously, speculative though it may be, because the Indonesian region is undoubtedly important in the redistribution of heat received at the Earth's surface and in moisture fluxes to the atmosphere. Moreover, Cane and Molnar make several predictions that can be tested.
The most obvious prediction is that sea surface temperatures in the Indian Ocean cooled between 5 million and 2 million years ago. This may be difficult to test, however. The change in temperature was probably 2–3 °C. Oxygen-isotope analysis of planktonic foraminifera, a useful tool for estimating past sea surface temperatures, can resolve this change. But because of evaporation, the South Pacific is saltier, as well as warmer, than the North Pacific. The evaporation that increases the salinity also increases the oxygen-isotope value of the surface water, offsetting the temperature effect. Advances in measuring the Mg/Ca ratios in foraminifera, and then using those ratios to estimate temperature, may solve the problem.
My own view is that Cane and Molnar are correct in their view that African aridity is linked to sea surface temperatures in the Indian Ocean, and that the most likely cause of cooling there was a narrowing of the Indonesian seaway. I am less confident that these changes had much to do with glaciation of the Northern Hemisphere, for one simple reason: from 10 million to 5.6 million years ago, cyclic glaciation was highly active in the Northern hemisphere and glaciation was suppressed between 5.5 million and 3.5 million years ago. Moreover, changes in the North Atlantic conveyor circulation cannot be ruled out in driving these glaciations. The conveyor delivers a substantial amount of heat to the high northern latitudes; the link with glaciation might have been through reduced heat fluxes as conveyor circulation decreased11, rather than through precipitation. Nonetheless, Cane and Molnar have armed the palaeoclimate community with the theory and predictions that will allow us to examine events in the Pacific, Indian and Atlantic oceans — especially the valves controlling water flow into and out of them — as drivers of Earth's climate.
