Abstract
Flow laws for ice predict rates of deformation (strain) and are fundamental to modelling glacier and ice-sheet dynamics. Here we apply Bayesian inference to laboratory measurements accumulated over 70 years to constrain flow laws for ice-sheet modelling. At low strains, commonly used flow laws—derived from individual experimental datasets with narrow stress, temperature and grain-size ranges—fail to capture the full complexity of ice behaviour. We show that a multicomponent flow law that sums strain rates from different deformation mechanisms is needed to capture grain-size and temperature sensitivities observed in the full set of experiments. This multicomponent flow law is applicable to natural scenarios where the anisotropy of ice is weak or where the deformation kinematics differ from those that formed the crystallographic preferred orientation, making the ice more viscous. Low-strain flow laws, including this multicomponent flow law, have limited validity at high strain, where viscosity evolves and anisotropy develops, making ice less viscous. A one-component, grain-size insensitive flow law gives a reasonable fit to high-strain experimental data and is better suited to modelling the large-scale flow behaviour of ice sheets.