Projecting Antarctic ice discharge using response functions from SeaRISE ice-sheet models

Abstract. The largest uncertainty in projections of future sea-level change still results from the potentially changing dynamical ice discharge from Antarctica. While ice discharge can alter through a number of processes, basal ice-shelf melting induced by a warming ocean has been identified as a major if not the major cause for possible additional ice flow across the grounding line. Here we derive dynamic ice-sheet response functions for basal ice-shelf melting using experiments carried out within the Sea-level Response to Ice Sheet Evolution (SeaRISE) intercomparison project with five different Antarctic ice-sheet models. As used here these response functions provide separate contributions for four different Antarctic drainage regions. Under the assumptions of linear-response theory we project future ice-discharge for each model, each region and each of the four Representative Concentration Pathways (RCP) using oceanic temperatures from 19 comprehensive climate models of the Coupled Model Intercomparison Project, CMIP-5, and two ocean models from the EU-project Ice2Sea. Uncertainty in the climatic forcing, the oceanic response and the ice-model differences is combined into an uncertainty range of future Antarctic ice-discharge induced from basal ice-shelf melt. The additional ice-loss (Table 6) is clearly scenario-dependent and results in a median of 0.07 m (66%-range: 0.04–0.10 m; 90%-range: −0.01–0.26 m) of global sea-level equivalent for the low-emission RCP-2.6 scenario and yields 0.1 m (66%-range: 0.06–0.14 m; 90%-range: −0.01–0.45 m) for the strongest RCP-8.5. If only models with an explicit representation of ice-shelves are taken into account the scenario dependence remains and the values change to: 0.05 m (66%-range: 0.03–0.08 m) for RCP-2.6 and 0.07 m (66%-range: 0.04–0.11 m) for RCP-8.5. These results were obtained using a time delay between the surface warming signal and the subsurface oceanic warming as observed in the CMIP-5 models. Without this time delay the ranges for all ice-models changes to 0.10 m (66%-range: 0.07–0.12 m; 90%-range: 0.01–0.28 m) for RCP-2.6 and 0.15 m (66%-range: 0.10–0.21 m; 90%-range: 0.02–0.53 m) for RCP-8.5. All probability distributions as provided in Fig. 10 are highly skewed towards high values.