The Empire Written in Ice

AIce is an indifferent archivist, which is exactly what makes it useful. Snow falling on an Alpine summit traps whatever the wind is carrying, and once buried it is not easily disturbed, so a vertical column of ice reads as a calendar of the air above it. Among the substances it holds is lead, released whenever galena, the commonest ore of the metal, is roasted to free the silver locked inside it. The Roman world wanted that silver for coin, and the process that yielded it, known as cupellation, sent roughly three hundred kilograms of lead into the atmosphere for every kilogram of silver recovered. A core drilled at Colle Gnifetti, on the Swiss-Italian border, carries a lead signal that climbs through the last two centuries BCE, stays high for some two hundred years and then collapses in the second century CE. What that curve actually measures has been argued over for the better part of thirty years.

BIlse Marchetti, who directs the palaeo-atmospheric group at the Institute for Alpine Ice Archives in Bern, treats the curve as an economic index. Her team measured lead in 1,140 samples spanning the period from 1100 BCE to 800 CE and, more usefully, measured in each the ratio between two isotopes of the metal. Ores differ in that ratio according to the geological age of the deposit, and the ratio recorded in the Roman-era ice matches that of the sulphide bodies of south-western Iberia rather than those of Britain or the Balkans. By 40 CE the rate of deposition stood at 3.6 times the level of the ninth century BCE, before organised mining began. It then fell by 78 per cent over the twelve years following 165 CE, the year an epidemic reached the Italian peninsula. Marchetti sets that fall beside independent evidence of a contraction in minting, and reads the two records as one story told twice.

CThe reading makes a prediction capable of failing. If the ice tracks output, cores drilled far apart ought to record the same turning points, since a change in production is a change at the source. Marchetti's group compared the Alpine record with a Greenland one and found the major inflections agreeing to within four years across the interval from 50 BCE to 200 CE. That agreement, she is careful to add, is confined to the interval named: before 50 BCE the two series wander independently, which she puts down to the smaller quantities then involved rather than to any defect in the method. She accepts a second limit as well. Once the metal is airborne the ice cannot tell one smelting district from another, so what the curve indexes is the total quantity burnt, never the question of where or by whom it was burnt.

DCormac Delaney, an archaeometallurgist at the Danube Institute for Material Studies in Vienna, replies that a curve read this way confuses delivery with source. Lead reaches an Alpine summit only if the wind brings it, and the circulation carrying air northward from Iberia has not been constant. Delaney's reanalysis of three cores dated by counting annual layers puts the uncertainty in that counting at eleven years either side of 100 CE, wide enough to swallow the correspondence Marchetti draws with the coinage. His second objection comes from the ground: the slag heaps at Rio Tinto, some 6.6 million tonnes of them, imply a scale of working that the ice, taken as a measure of output across the empire, would understate, because furnace design altered over time and later furnaces lost less of their charge to the air. A falling curve, on this account, may record better furnaces rather than fewer of them.

EA second dispute runs alongside the first and has proved harder to keep indoors. Skeletons from a cemetery outside Ostia, 84 of them dated to the first and second centuries CE, carry bone lead concentrations averaging eight times those of Iron Age burials from the same stretch of coast, and that figure is not in doubt. Its cause is. The city's pipes were lead, yet hard water lays down a carbonate film inside a pipe within months and the film acts as a barrier, so the boiling of grape must in lead vessels to make the sweetener called defrutum has been proposed as the heavier route. Delaney, whose scepticism about the ice does not extend to this evidence, holds that the two routes cannot be told apart by bone measurements alone, since bone integrates exposure across decades and keeps no record of where that exposure came from. Marchetti answers that the question is secondary to the atmospheric one, since smelting smoke reached populations who never went near a pipe.

FA settlement of sorts is emerging in which the two readings are separated by scale rather than ranked against each other. At the scale of a single mining district the ice performs well: when Iberian production is known from other evidence to have halted, the isotope ratio in the ice moves away from Iberian values within a few years. At the scale of an empire it is a blunter instrument, since transport, furnace efficiency and dating error all push in the same direction as the quantity being measured. The stakes are no longer purely historical. Regulators define the natural concentration of lead in air and soil against pre-industrial values, and several of those values were taken from ice-core work of just this kind, so a Roman signal that is a poor guide to how much metal was burnt passes its error on to the baseline drawn from it. What neither researcher will accept is the older assumption that one number from one column of ice can stand for the industry of a continent.