<jats:title>Abstract</jats:title><jats:p>Ice flow from the ice sheets to the ocean contains the maximum potential contributing to future eustatic sea-level rise. In Antarctica most mass fluxes occur via the extended ice-shelf regions covering more than half the Antarctic coastline. The most extended ice shelves are the Filchner–Ronne and Ross Ice Shelves, which contribute ~30% to the total mass loss caused by basal melting. Basal melt rates here show small to moderate average amplitudes of &lt;0.5ma<jats:sup>–1</jats:sup>. By comparison, the smaller but most vulnerable ice shelves in the Amundsen and Bellinghausen Seas show much higher melt rates (up to 30 ma<jats:sup>–1</jats:sup>), but overall basal mass loss is comparably small due to the small size of the ice shelves. The pivotal question for both characteristic ice-shelf regions, however, is the impact of ocean melting, and, coevally, change in ice-shelf thickness, on the flow dynamics of the hinterland ice masses. In theory, ice-shelf back-pressure acts to stabilize the ice sheet, and thus the ice volume stored above sea level. We use the three-dimensional (3-D) thermomechanical ice-flow model RIMBAY to investigate the ice flow in a regularly shaped model domain, including ice-sheet, ice-shelf and open-ocean regions. By using melting scenarios for perturbation studies, we find a hysteresis-like behaviour. The experiments show that the system regains its initial state when perturbations are switched off. Average basal melt rates of up to 2 ma<jats:sup>–1</jats:sup> as well as spatially variable melting calculated by our 3-D ocean model ROMBAX act as basal boundary conditions in time-dependent model studies. Changes in ice volume and grounding-line position are monitored after 1000 years of modelling and reveal mass losses of up to 40 Gt a<jats:sup>–1</jats:sup>.</jats:p>