The Antarctic Ice Sheet (AIS) is the largest source of uncertainty in sea-level rise (SLR) projections, posing substantial risks to coastal communities worldwide. AIS projections are subject to cascading uncertainties propagating through emission scenarios, atmosphere-ocean general circulation models (AOGCMs), ice sheet dynamics, and sea-level physics. In numerical ice sheet model intercomparison projects, these uncertainties—typically attributed to intermodel differences—stem from modeling choices that may bias ensembles toward commonly adopted approaches regardless of observational consistency. Here we show that, unlike previous ensemble projections, observation-constrained projections indicate it is very likely (P$\ge$0.92) that historical warming has committed the AIS to continued mass loss through the 21st century even under aggressive emission reduction scenarios. Our results arise from a machine learning emulator that systematically quantifies how each physical assumption cascades into spatiotemporally variable SLR uncertainties, complemented by Bayesian calibration to reduce projection bias. Higher emissions drive greater mass loss within this century (P$\ge$0.89)—affecting both near-term risks and long-term millennial commitments. Notably, under very high emission scenarios, we reveal cascading physical mechanisms that could produce up to 25.4 cm of SLR by 2100 (95th percentile; median = 15.7 cm) while remaining consistent with recent satellite observations. Effective management of these substantial risks requires rapid emission reductions and improved constraints on AOGCM selection, sliding laws, and ice shelf melt parameterizations, the primary uncertainty drivers identified by our analysis.

暂无