The Greenland Ice Sheet (GIS) is losing mass at a high rate. Given the short-term nature of the observational record, it is difficult to assess the historical importance of this mass-loss trend. Unlike records of greenhouse gas concentrations and global temperature, in which observations have been merged with palaeoclimate datasets, there are no comparably long records for rates of GIS mass change. Here we reveal unprecedented mass loss from the GIS this century, by placing contemporary and future rates of GIS mass loss within the context of the natural variability over the past 12,000 years. We force a high-resolution ice-sheet model with an ensemble of climate histories constrained by ice-core data. Our simulation domain covers southwestern Greenland, the mass change of which is dominated by surface mass balance. The results agree favourably with an independent chronology of the history of the GIS margin. The largest pre-industrial rates of mass loss (up to 6,000 billion tonnes per century) occurred in the early Holocene, and were similar to the contemporary (ad 2000-2018) rate of around 6,100 billion tonnes per century. Simulations of future mass loss from southwestern GIS, based on Representative Concentration Pathway (RCP) scenarios corresponding to low (RCP2.6) and high (RCP8.5) greenhouse gas concentration trajectories, predict mass loss of between 8,800 and 35,900 billion tonnes over the twenty-first century. These rates of GIS mass loss exceed the maximum rates over the past 12,000 years. Because rates of mass loss from the southwestern GIS scale linearly with the GIS as a whole, our results indicate, with high confidence, that the rate of mass loss from the GIS will exceed Holocene rates this century.

Establishing the timing of maximum Holocene warmth in the Arctic is critical for understanding global climate system response to external forcing. In Greenland, challenges in obtaining climate records that span the full Holocene have hampered efforts to robustly identify when the Holocene Thermal Maximum occurred. Reconstructing land-based ice sheet history can fill this gap because these ice sheet regions respond sensitively to summer temperature. We synthesize new and published 10Be and 14C ages from southwest Greenland to map Greenland ice sheet margin positions from 12 to 7 ka and calculate retreat rates from 12 to 0 ka. We found that maximum Greenland ice sheet recession occurred between ~10.4 and 9.1 ka. Our reconstruction suggests that summer air temperatures in southwest Greenland were highest from ~10.4 to 9.1 ka, providing support for an early regional Holocene Thermal Maximum. These results can serve as benchmarks for comparison with ice sheet and climate model simulations.