Quantifying Uncertainty in Inferred Viscosity and Basal Shear Stress Over Ice Streams

Abstract:

Basal friction and ice viscosity are both essential controls on glacier motion that cannot be measured by remote sensing. In order to initialize models, it is common practice to use inverse methods to determine the basal shear stress over grounded ice and the viscosity of floating ice. It is difficult to quantify the uncertainty in the inferred parameters due to the computational expense of the procedure, the choice of regularization parameter, and the errors in the various measurements used as input, as well as differences in inversion method. Various methods can be used to perform the inversion, and these differing procedures cause discrepancies in the inferred properties of the ice streams. Additionally, the inferred parameters depend on the sophistication of the approximation for ice flow that is used, e.g. full-Stokes or the shallow-shelf approximation.

We analyze the impact the choices of modeling procedure and inversion method have on inferred ice properties. To do this we perform a number of inversions for basal shear stress and for ice shelf viscosity over Smith, Pope, and Kohler Glaciers in West Antarctica and assess the sensitivity to modelers' choices. We use both a three dimensional full-Stokes model and a two dimensional shallow-shelf model, with both Robin and adjoint type inversion procedures, to infer basal shear stress and ice viscosity. We compare the results of these different methods and evaluate their implication on uncertainty in the unknown parameters.