Poor Performance of a Common Crevasse Model at Marine-Terminating Glaciers

Crevasses are both affected by and effect stresses and surface mass balance of glaciers, potentially exerting important controls on meltwater routing, glacier viscosity, and iceberg calving, yet there are few direct observations of crevasse depth. Here we assess one of the most common models for crevasse formation, in which crevasse depths depend on the local stress state, through analysis of 52644 crevasse depth observations from 19 Greenland glaciers. We find that modeled depths are uncorrelated with observed depths and are generally too deep. Model performance can be improved with glacier-by-glacier tuning of viscosity and water depth parameters, but spatial variations in tuning parameters are unlikely to have a physical basis, and the model still fails to capture smaller-scale variations in crevassing that may control calving. Thus, numerical ice flow models drawing on this parameterization are likely to yield inaccurate projections of glacier mass change or crevasse depth-driven terminus position changes.