Ice shelves in the Amundsen Sea Embayment have thinned, accelerating the seaward flow of ice sheets upstream over recent decades. This imbalance is caused by an increase in the ocean‐driven melting of the ice shelves. Observations and models show that the ocean heat content reaching the ice shelves is sensitive to the depth of thermocline, which separates the cool, fresh surface waters from warm, salty waters. Yet the processes controlling the variability of thermocline depth remain poorly constrained. Here we quantify the oceanic conditions and ocean‐driven melting of Cosgrove, Pine Island Glacier (PIG), Thwaites, Crosson, and Dotson ice shelves in the Amundsen Sea Embayment from 1991 to 2014 using a general circulation model. Ice‐shelf melting is coupled to variability in the wind field and the sea‐ice motions over the continental shelf break and associated onshore advection of warm waters in deep troughs. The layer of warm, salty waters at the calving front of PIG and Thwaites is thicker in austral spring (June–October) than in austral summer (December–March), whereas the seasonal cycle at the calving front of Dotson is reversed. Furthermore, the ocean‐driven melting in PIG is enhanced by an asymmetric response to changes in ocean heat transport anomalies at the continental shelf break: melting responds more rapidly to increases in ocean heat transport than to decreases. This asymmetry is caused by the inland deepening of bathymetry and the glacial meltwater circulation around the ice shelf.

Assessment of ocean‐forced ice sheet loss requires that ocean models be able to represent sub‐ice shelf melt rates. However, spatial accuracy of modeled melt is not well investigated, and neither is the level of accuracy required to assess ice sheet loss. Focusing on a fast‐thinning region of West Antarctica, we calculate spatially resolved ice‐shelf melt from satellite altimetry and compare against results from an ocean model with varying representations of cavity geometry and ocean physics. Then, we use an ice‐flow model to assess the impact of the results on grounded ice. We find that a number of factors influence model‐data agreement of melt rates, with bathymetry being the leading factor; but this agreement is only important in isolated regions under the ice shelves, such as shear margins and grounding lines. To improve ice sheet forecasts, both modeling and observations of ice‐ocean interactions must be improved in these critical regions.

Ice shelves play a vital role in regulating loss of grounded ice and in supplying freshwater to coastal seas. However, melt variability within ice shelves is poorly constrained and may be instrumental in driving ice shelf imbalance and collapse. High‐resolution altimetry measurements from 2010 to 2016 show that Dotson Ice Shelf (DIS), West Antarctica, thins in response to basal melting focused along a single 5 km‐wide and 60 km‐long channel extending from the ice shelf's grounding zone to its calving front. If focused thinning continues at present rates, the channel will melt through, and the ice shelf collapse, within 40–50 years, almost two centuries before collapse is projected from the average thinning rate. Our findings provide evidence of basal melt‐driven sub‐ice shelf channel formation and its potential for accelerating the weakening of ice shelves.