Revised estimates of Greenland ice sheet thinning histories based on ice-core records

Ice core records were recently used to infer elevation changes of the Greenland ice sheet throughout the Holocene. The inferred elevation changes show a significantly greater elevation reduction than those output from numerical models, bringing into question the accuracy of the model-based reconstructions and, to some extent, the estimated elevation histories. A key component of the ice core analysis involved removing the influence of vertical surface motion on the δ18O signal measured from the Agassiz and Renland ice caps. We re-visit the original analysis with the intent to determine if the use of more accurate land uplift curves can account for some of the above noted discrepancy. To improve on the original analysis, we apply a geophysical model of glacial isostatic adjustment calibrated to sea-level records from the Queen Elizabeth Islands and Greenland to calculate the influence of land height changes on the δ18O signal from the two ice cores. This procedure is complicated by the fact that δ18O contained in Agassiz ice is influenced by land height changes distant from the ice cap and so selecting a single location at which to compute the land height signal is not possible. Uncertainty in this selection is further complicated by the possible influence of Innuitian ice during the early Holocene (12–8 ka BP). Our results indicate that a more accurate treatment of the uplift correction leads to elevation histories that are, in general, shifted down relative to the original curves at GRIP, NGRIP, DYE-3 and Camp Century. In addition, compared to the original analysis, the 1-σ uncertainty is considerably larger at GRIP and NGRIP. These changes reduce the data-model discrepancy reported by Vinther et al. (2009) at GRIP, NGRIP, DYE-3 and Camp Century. A more accurate treatment of isostasy and surface loading also acts to improve the data-model fits such that the residuals at all four sites for the period 8 ka BP to present are significantly reduced compared to the original analysis. Prior to 8 ka BP, the possible influence of Innuitian ice on the inferred elevation histories prevents a meaningful comparison. ► We revisited the analysis of Vinther et al. (2009). ► Using GIA modelling we produce more accurate land uplift corrections. ► Results in a series of new ice thinning histories at Greenland ice core sites. ► With a decrease in amplitude and a larger 1-σ uncertainty at GRIP and NGRIP. ► This decreases the misfit between numerical ice models and the thinnin