We reconstruct human–landscape interactions in the context of ancient bloomery smelting on Elba Island, Tyrrhenian Sea, based on sedimentological evidence. Elba was, together with Populonia, the center of iron production in antiquity (mid‐6th century B.C.E. to 1st century C.E.). The sediment sequences obtained by vibracoring in the Campo plain reveal that the Holocene transgression triggered aggradation during the Early–Middle Holocene (>5.4 ka cal BP). Nevertheless, wetland conditions prevailed in low‐lying areas during Roman times. Correspondingly, ancient smelting sites are found only on elevated areas along the edge of the plain. The palaeolandscape thus plays a key role in understanding location factors for smelting sites in antiquity. Our sediment sequences indicate that during Etrusco–Roman times (2.4–2.0 ka cal BP) morphodynamics on the island increased, as is evident from the alteration of slope deposits and slack water deposits in upstream positions, and the change from lacustrine to alluvial deposition in downstream areas. The charcoal record shows that fire dynamics increased simultaneously. The sediments that are synsedimentary to the ancient smelting period show significantly increased arsenic contents and high sediment magnetic susceptibility, whereas overall concentrations of heavy metals remain in the range of the geogenic background signal.

No abstract

Scholars frequently cite fuel scarcity after deforestation as a reason for the abandonment of most of the Roman iron smelting sites on Elba Island (Tuscan Archipelago, Italy) in the 1st century bce. Whereas the archaeological record clearly indicates the decrease in smelting activities, evidence confirming the ‘deforestation narrative’ is ambiguous. Therefore, we employed a stochastic, spatio-temporal model of the wood required and consumed for iron smelting on Elba Island in order to assess the availability of fuelwood on the island. We used Monte Carlo simulations to cope with the limited knowledge available on the past conditions on Elba Island and the related uncertainties in the input parameters. The model includes both, wood required for the furnaces and to supply the workforce employed in smelting. Although subject to high uncertainties, the outcomes of our model clearly indicate that it is unlikely that all woodlands on the island were cleared in the 1st century bce. A lack of fuel seems only likely if a relatively ineffective production process is assumed. Therefore, we propose taking a closer look at other reasons for the abandonment of smelting sites, e.g. the occupation of new Roman provinces with important iron ore deposits; or a resource-saving strategy in Italia. Additionally, we propose to read the development of the ‘deforestation narrative’ originating from the 18th/19th century in its historical context.