NY Times, Science News) are covering some results from measuring the amount of lead in an ice core from Greenland, which the study's authors say traces out the history of lead production in the ancient world. They think they can see in this data several major historical events, such as the civil wars of the late Roman Republic and the crisis of AD 250-270. This is pretty cool science; I love the basic idea of studying ancient industrial production by measuring the residue of pollution. But there is something wrong with this study.
First, look how spiky their graph is. A certain amount of random variation is inevitable in any series like this, but why would samples from some years have ten times as much lead as the years before or after? Industrial production in the Mediterranean cannot explain this. So there is another variable at work.
Supplementary Material. On these three graphs the simplified ice core data is shown in red. The black line is data from three studies of lead in bog sediments. Notice that all three show the Classical expansion, but only the one at the top shows dramatic growth after 650 AD. That sample is from Flanders Moss in northern Scotland.
Aha! So the missing factor here is wind patterns. The graph we see from the ice core is a combination of at least two variables: how much lead the Romans were pumping into the air, and whether the wind was blowing toward Greenland. Lead concentrations rise in the Carolingian period not because of increased production but because the wind shifted and blew more often toward the North Atlantic. Of course there might be yet more variables, such as changes in refining techniques, or the seasonality of smelting work, or the location of active mines. But anyway there are at least two.
Notice that the bottom bog sample has much higher lead levels than the other two; as you might guess, that bog is in Spain, the epicenter of Rome's metal industries. And it traces out a graph that more closely matches my idea of lead production in the Mediterranean than the ice core graph does. Probably cost about 1% as much, too.
I don't want to seem too negative; the ice cores are a great resource and we should push them as hard as we can. But they don't reveal everything, and sometimes they can be misleading. I understand that a couple of well attested volcanic eruptions do not appear in the ice core data, presumably because weird wind patterns kept any detectable amount of their ash from reaching Greenland. In the case of lead from Classical and Medieval times ice core data clearly should not be used alone, but only in combination with other information. And people who publish about it should be a lot more circumspect and do less grandiose self-promotion; their data is nice but not as good as data from bogs produced by people nobody has ever heard of, whose results completely failed to make it into the news.