A data-constrained large ensemble analysis of Antarctic evolution since the Eemian
R. D. Briggs, D. Pollard, and L. Tarasov
Quaternary Science Reviews (1 November 2014)
DOI: 10.1016/j.quascirev.2014.09.003
Abstract Reconstructing the historical behaviour of the Antarctic Ice Sheet is essential for understanding and predicting climatic interactions, ice sheet sensitivities, and sea level contributions. Furthermore, to interpret such reconstructions with any degree of confidence, meaningful uncertainty estimates are required. Working toward this goal, this article presents results from a large-ensemble data-constrained study of Antarctic evolution over the Last Glacial cycle. Ice sheet chronologies have been generated using a 3-D glacial systems model that includes sheet/stream/shelf flow; a parametrized basal drag coefficient (accounting for sub-grid topographic roughness, sediment likelihood, and systematic model-to-observation thickness misfits); a sub-grid grounding-line flux parametrization; a visco-elastic bedrock response component; parametrized climate forcing; separate shelf and tidewater calving treatments; and a physically-motivated, empirically derived, sub-shelf melt component. There are 31 ensemble parameters to capture uncertainties in the glacial cycle climate, mass-balance processes, and ice dynamics. Once generated, the ensemble of model runs is constrained using a suite of observational data (including constraints on relative sea level, past ice thickness and grounding line retreat data-points, as well as present-day ice volumes and configuration) and an evaluation methodology that produces a misfit-to-observations score for each run. Assuming variants of a Gaussian error model, the scores are used to generate probability distributions for the past evolution of the ice sheet. In our ensemble-based analysis the Last Glacial Maximum (LGM) of Antarctic ice volume occurs at ∼25–24 ka. The LGM ice volume excess relative to present-day is likely between 5.6 and 14.3 m equivalent sea level (mESL), and with less confidence >10 mESL. There is little change in the grounded ice volume from 24 ka until 17–16 ka at which time widespread deglaciation commences. The Ross and Weddell Sea sectors are the largest sources of ice mass loss. The major period of grounding line retreat in the Weddell Sea sector, resulting in the present-day Ronne–Filchner shelf system, occurs after 10 ka; the grounding line in the Ross Sea sector start its major retreat phase after 12 ka. There is no significant contribution (<1.8 mESL) to meltwater pulse 1a. During the Eemian, the modelled Antarctic Ice Sheet has a minimal ice volume at 114 ka contributing ∼2 mESL relative to present-day. The mass loss is principally due to a major retreat of the grounding line in the Siple Coast sector of the Ross Sea. We present standard deviation plots of the ensemble results for present day and at the LGM. These plots highlight geographic areas that would most benefit from additional constraint and therefore offer priorities and guidance for future field campaigns.
keywords: Model evaluation; Observational data; Ice sheet model; Antarctica; Deglaciation