How well do we know Ice Ages?
A bit of the backstory to the wonderful exchange between the Congressional Committee on Science, Space and Technology and the President’s Science Advisor. A question posed by Rep. Stockman was basically, why don’t we include the cause of ice ages in our climate models? It turns out this arose from a conversation he had with NASA Earth Science Director Piers Sellers (who recently wrote a column in the New York Times about it). The earlier exchange had Stockman asking how we could predict climate change when we couldn’t say how ice ages were caused; Sellers responded by saying, basically, we “understood very well” and it is because of Milankovitch cycles, which reflect changes in the Earth’s orbit over thousands of years.
This makes GG somewhat nervous; it might be a good answer to give a Congressman but it leaves out some important stuff.
You see, the statement Sellers made is a gross simplification in two ways: there are mechanisms that modify Milankovitch cycles and there is the question of why northern hemisphere glaciations started about 2.75 million years ago. The diagram below (borrowed from Ruddiman’s very nice textbook) makes the first point nicely:
The curve on the left is basically the output from Milankovitch cycles; that on the right is compiled from various geologic proxies. You can see that ice ages (when the right hand curve is bluer and more to the left) are not simply when strength of sunlight at high latitudes is weakest (when curve on left is farthest left). The reason is simple in concept and more challenging in detail: there are feedbacks in the climate system that will make the amount of ice present differ from a simple multiple of the amount of summer sunshine. The most basic element is that sunshine affects the rate of glacial growth, so you might expect the largest glaciers just as sunshine had passed its minimum and was getting stronger (basically the same reason we say winter is from late December to mid-March instead of early November to early February, which is when the sun is lowest). Other feedbacks are less obvious: if there is a big ice cap, a few warm years won’t melt it. Cooling the earth will cause the ocean to absorb more carbon dioxide, further cooling the earth and amplifying the signal. Changes in sea level and vegetation will play roles. Toss in all these effects and the ice signal is bound to look different than the starting Milankovitch insolation variations. (It is sort of like rocking a car to get it out of mud or snow; pushing continuously doesn’t get you anywhere, and pushing at a fast rate won’t do the work, but a rhythmic push at the right frequency will get the job done. So too will the climate system respond more for certain frequencies of forcing).
There are models that take the lefthand curve and predict the righthand one, and by and large they do pretty well. But there are different versions that have somewhat different feedbacks, and there are probably some feedbacks we don’t know about yet. So at the basic level of “Milankovitch + feedbacks = ice ages” our understanding is pretty solid (we don’t need anything else), but at the level of identifying exactly the strength of all the feedbacks, we are still trying to get all the pieces lined up.
How about the start of northern hemisphere glaciation (the beginning of ice ages as we usually think of them)? This is utterly different and Milankovitch plays no role. Milankovitch cycles go back probably to the earth’s origin (though with changing frequencies), so something changed about 2.75 million years ago. The safest statement here is that we don’t know the cause. It probably requires some significant changes to oceanic circulation and increasingly it looks like this change was in the Pacific and not the Atlantic as many texts still claim. One suggestion, for instance, is that the world used to be in a near-perpetual “El Niño” state because of the connection of the western Pacific to the Indian Ocean and the much lower amounts of rainfall owing to the much smaller area of land in the Indonesian seaway. (During modern El Niños temperatures are higher in the Canadian shield and Arctic, reducing the ability to grow ice sheets). As Australia and New Guinea moved north, land area increased, precipitation in the west Pacific increased and ocean currents were changed. Others point to possible issues in the Bering Sea (and probably there are still advocates for changes in the Atlantic). Although the thing we want to know about is a climate change, the causes in this case are probably tectonic.
So the next time Rep. Stockman chats with the Science Advisor, he can ask why plate tectonics isn’t being included in climate models addressing global warming…and we can all roll our eyes at that question.